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 # LU-2435: if the underlying zfs doesn't support userobj_accounting,
1610 # lustre will estimate the object count usage. This fails quota
1611 # verification in 32b. The object quota usage should be accurate after
1612 # zfs-0.7.0 is released.
1613 [ $fstype == "zfs" ] && {
1614 local zfs_version=$(do_node $node cat /sys/module/zfs/version)
1616 [ $(version_code $zfs_version) -lt $(version_code 0.7.0) ] && {
1617 echo "Skip quota verify for zfs: $zfs_version"
1622 $LFS quota -u $T32_QID -v $mnt
1624 qval=$($LFS quota -v -u $T32_QID $mnt |
1625 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1626 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1628 [ $qval -eq $img_bspace ] || {
1629 echo "bspace, act:$qval, exp:$img_bspace"
1633 qval=$($LFS quota -v -u $T32_QID $mnt |
1634 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1635 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1637 [ $qval -eq $img_ispace ] || {
1638 echo "ispace, act:$qval, exp:$img_ispace"
1642 qval=$($LFS quota -v -u $T32_QID $mnt |
1643 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1644 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1646 [ $qval -eq $img_blimit ] || {
1647 echo "blimit, act:$qval, exp:$img_blimit"
1651 qval=$($LFS quota -v -u $T32_QID $mnt |
1652 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1653 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1655 [ $qval -eq $img_ilimit ] || {
1656 echo "ilimit, act:$qval, exp:$img_ilimit"
1660 do_node $node $LCTL conf_param $fsname.quota.mdt=ug
1661 cmd="$LCTL get_param -n osd-$fstype.$fsname-MDT0000"
1662 cmd=$cmd.quota_slave.enabled
1663 wait_update $node "$cmd" "ug" || {
1664 echo "Enable mdt quota failed"
1668 do_node $node $LCTL conf_param $fsname.quota.ost=ug
1669 cmd="$LCTL get_param -n osd-$fstype.$fsname-OST0000"
1670 cmd=$cmd.quota_slave.enabled
1671 wait_update $node "$cmd" "ug" || {
1672 echo "Enable ost quota failed"
1677 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1678 bs=1M count=$((img_blimit / 1024)) oflag=sync && {
1679 echo "Write succeed, but expect -EDQUOT"
1682 rm -f $mnt/t32_qf_new
1684 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1686 echo "Create succeed, but expect -EDQUOT"
1689 unlinkmany $mnt/t32_qf_ $img_ilimit
1697 local dne_upgrade=${dne_upgrade:-"no"}
1698 local ff_convert=${ff_convert:-"no"}
1699 local shall_cleanup_mdt=false
1700 local shall_cleanup_mdt1=false
1701 local shall_cleanup_ost=false
1702 local shall_cleanup_lustre=false
1703 local mdt2_is_available=false
1704 local node=$(facet_active_host $SINGLEMDS)
1705 local r="do_node $node"
1721 local fstype=$(facet_fstype $SINGLEMDS)
1722 local mdt_dev=$tmp/mdt
1723 local mdt2_dev=$tmp/mdt2
1724 local ost_dev=$tmp/ost
1729 combined_mgs_mds || stop_mgs || error "Unable to stop MGS"
1730 trap 'trap - RETURN; t32_test_cleanup' RETURN
1733 nid=$($r $LCTL list_nids | head -1)
1735 mkdir -p $tmp/mnt/lustre || error "mkdir $tmp/mnt/lustre failed"
1736 $r mkdir -p $tmp/mnt/{mdt,mdt1,ost}
1737 $r tar xjvf $tarball -S -C $tmp || {
1738 error_noexit "Unpacking the disk image tarball"
1741 img_commit=$($r cat $tmp/commit)
1742 img_kernel=$($r cat $tmp/kernel)
1743 img_arch=$($r cat $tmp/arch)
1744 img_bspace=$($r cat $tmp/bspace)
1745 img_ispace=$($r cat $tmp/ispace)
1747 # older images did not have "blimit" and "ilimit" files
1748 # use old values for T32_BLIMIT and T32_ILIMIT
1749 $r test -f $tmp/blimit && img_blimit=$($r cat $tmp/blimit) ||
1751 $r test -f $tmp/ilimit && img_ilimit=$($r cat $tmp/ilimit) ||
1754 echo "Upgrading from $(basename $tarball), created with:"
1755 echo " Commit: $img_commit"
1756 echo " Kernel: $img_kernel"
1757 echo " Arch: $img_arch"
1758 echo "OST version: $(lustre_build_version ost1)"
1760 # The conversion can be made only when both of the following
1761 # conditions are satisfied:
1762 # - ost device img version < 2.3.64
1763 # - ost server version >= 2.5
1764 [ $(version_code $img_commit) -ge $(version_code 2.3.64) -o \
1765 $(lustre_version_code ost1) -lt $(version_code 2.5.0) ] &&
1768 ! $r test -f $mdt2_dev || mdt2_is_available=true
1770 if [[ $fstype == zfs ]]; then
1773 local poolname_list="t32fs-mdt1 t32fs-ost1"
1775 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1777 for poolname in $poolname_list; do
1779 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
1780 $ZPOOL import -f -d $tmp $poolname"
1783 # upgrade zpool to latest supported features, including
1784 # dnode quota accounting in 0.7.0
1785 $r "$ZPOOL upgrade -a"
1787 mdt_dev=t32fs-mdt1/mdt1
1788 ost_dev=t32fs-ost1/ost1
1789 ! $mdt2_is_available || mdt2_dev=t32fs-mdt2/mdt2
1790 wait_update_facet $SINGLEMDS "$ZPOOL list |
1791 awk '/^t32fs-mdt1/ { print \\\$1 }'" "t32fs-mdt1" || {
1792 error_noexit "import zfs pool failed"
1797 $r $LCTL set_param debug="$PTLDEBUG"
1799 $r $TUNEFS --dryrun $mdt_dev || {
1801 error_noexit "tunefs.lustre before mounting the MDT"
1805 if $mdt2_is_available; then
1806 $r $TUNEFS --dryrun $mdt2_dev || {
1808 error_noexit "tunefs.lustre before mounting the MDT"
1813 if [ "$writeconf" ]; then
1815 if [ $fstype == "ldiskfs" ]; then
1817 $r $TUNEFS --quota $mdt_dev || {
1819 error_noexit "Enable mdt quota feature"
1822 if $mdt2_is_available; then
1823 $r $TUNEFS --quota $mdt2_dev || {
1825 error_noexit "Enable mdt quota feature"
1831 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
1832 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
1833 { skip "LU-2200: Cannot run over Inifiniband w/o lctl replace_nids "
1834 "(Need MGS version at least 2.3.59)"; return 0; }
1836 local osthost=$(facet_active_host ost1)
1837 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
1840 if [ $fstype == "ldiskfs" ]; then
1843 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt
1844 $r $LCTL replace_nids $fsname-OST0000 $ostnid
1845 $r $LCTL replace_nids $fsname-MDT0000 $nid
1846 $r $UMOUNT $tmp/mnt/mdt
1849 mopts=exclude=$fsname-OST0000
1850 if [ $fstype == "ldiskfs" ]; then
1855 t32_wait_til_devices_gone $node
1857 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
1859 error_noexit "Mounting the MDT"
1862 shall_cleanup_mdt=true
1864 if $mdt2_is_available; then
1865 mopts=mgsnode=$nid,$mopts
1866 $r $MOUNT_CMD -o $mopts $mdt2_dev $tmp/mnt/mdt1 || {
1868 error_noexit "Mounting the MDT"
1872 echo "mount new MDT....$mdt2_dev"
1873 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1874 error_noexit "enable remote dir create failed"
1876 shall_cleanup_mdt1=true
1877 elif [ "$dne_upgrade" != "no" ]; then
1878 local fs2mdsdev=$(mdsdevname 1_2)
1879 local fs2mdsvdev=$(mdsvdevname 1_2)
1881 echo "mkfs new MDT on ${fs2mdsdev}...."
1882 if [ $(facet_fstype mds1) == ldiskfs ]; then
1883 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
1886 add $SINGLEMDS $(mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
1887 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
1888 error_noexit "Mkfs new MDT failed"
1892 [[ $(facet_fstype mds1) != zfs ]] || import_zpool mds1
1894 $r $TUNEFS --dryrun $fs2mdsdev || {
1895 error_noexit "tunefs.lustre before mounting the MDT"
1899 echo "mount new MDT....$fs2mdsdev"
1900 $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
1901 error_noexit "mount mdt1 failed"
1905 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1906 error_noexit "enable remote dir create failed"
1908 shall_cleanup_mdt1=true
1911 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
1912 error_noexit "Getting MDT UUID"
1915 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
1916 error_noexit "Unexpected MDT UUID: \"$uuid\""
1920 $r $TUNEFS --dryrun $ost_dev || {
1921 error_noexit "tunefs.lustre before mounting the OST"
1924 if [ "$writeconf" ]; then
1925 mopts=mgsnode=$nid,$writeconf
1926 if [ $fstype == "ldiskfs" ]; then
1928 $r $TUNEFS --quota $ost_dev || {
1930 error_noexit "Enable ost quota feature"
1936 if [ $fstype == "ldiskfs" ]; then
1941 $r $MOUNT_CMD -onomgs -o$mopts $ost_dev $tmp/mnt/ost || {
1942 error_noexit "Mounting the OST"
1945 shall_cleanup_ost=true
1947 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
1948 error_noexit "Getting OST UUID"
1951 if [ "$uuid" != $fsname-OST0000_UUID ]; then
1952 error_noexit "Unexpected OST UUID: \"$uuid\""
1956 $r $LCTL conf_param $fsname-OST0000.osc.max_dirty_mb=15 || {
1957 error_noexit "Setting \"max_dirty_mb\""
1960 $r $LCTL conf_param $fsname-OST0000.failover.node=$nid || {
1961 error_noexit "Setting OST \"failover.node\""
1964 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
1965 error_noexit "Setting \"max_rpcs_in_flight\""
1968 $r $LCTL conf_param $fsname-MDT0000.failover.node=$nid || {
1969 error_noexit "Setting MDT \"failover.node\""
1972 $r $LCTL pool_new $fsname.interop || {
1973 error_noexit "Setting \"interop\""
1976 $r $LCTL conf_param $fsname-MDT0000.lov.stripesize=4M || {
1977 error_noexit "Setting \"lov.stripesize\""
1980 $r $LCTL conf_param $fsname-MDT0000.mdd.atime_diff=70 || {
1981 error_noexit "Setting \"mdd.atime_diff\""
1985 if [ "$ff_convert" != "no" -a $(facet_fstype ost1) == "ldiskfs" ]; then
1986 $r $LCTL lfsck_start -M $fsname-OST0000 || {
1987 error_noexit "Start OI scrub on OST0"
1991 # The oi_scrub should be on ost1, but for test_32(),
1992 # all on the SINGLEMDS.
1993 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
1994 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1995 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
1996 error_noexit "Failed to get the expected 'completed'"
2000 local UPDATED=$($r $LCTL get_param -n \
2001 osd-ldiskfs.$fsname-OST0000.oi_scrub |
2002 awk '/^updated/ { print $2 }')
2003 [ $UPDATED -ge 1 ] || {
2004 error_noexit "Only $UPDATED objects have been converted"
2009 if [ "$dne_upgrade" != "no" ]; then
2010 $r $LCTL conf_param \
2011 $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
2012 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
2015 $r $LCTL conf_param $fsname-MDT0001.failover.node=$nid || {
2016 error_noexit "Setting MDT1 \"failover.node\""
2019 $r $LCTL conf_param $fsname-MDT0001.lov.stripesize=4M || {
2020 error_noexit "Setting MDT1 \"lov.stripesize\""
2025 if [ "$writeconf" ]; then
2026 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
2027 error_noexit "Mounting the client"
2031 shall_cleanup_lustre=true
2032 $r $LCTL set_param debug="$PTLDEBUG"
2034 t32_verify_quota $node $fsname $tmp/mnt/lustre || {
2035 error_noexit "verify quota failed"
2039 if $r test -f $tmp/list; then
2041 # There is not a Test Framework API to copy files to or
2042 # from a remote node.
2044 # LU-2393 - do both sorts on same node to ensure locale
2046 local list_file=$tmp/list
2048 if $mdt2_is_available; then
2049 if [[ -d $tmp/mnt/lustre/striped_dir_old ]] &&
2050 $r test -f $tmp/list2; then
2051 list_file=$tmp/list2
2052 pushd $tmp/mnt/lustre/striped_dir_old
2054 pushd $tmp/mnt/lustre/remote_dir
2057 pushd $tmp/mnt/lustre
2059 $r cat $list_file | sort -k 6 >$tmp/list.orig
2060 ls -Rni --time-style=+%s | sort -k 6 |
2061 sed 's/\. / /' >$tmp/list || {
2067 # 32-bit and 64-bit clients use different algorithms to
2068 # convert FIDs into inode numbers. Hence, remove the
2069 # inode numbers from the lists, if the original list was
2070 # created on an architecture with different number of
2073 if [ $(t32_bits_per_long $(uname -m)) != \
2074 $(t32_bits_per_long $img_arch) ]; then
2075 echo "Different number of bits per \"long\"" \
2076 "from the disk image"
2077 for list in list.orig list; do
2078 sed -i -e 's/^[0-9]\+[ \t]\+//' \
2082 if ! diff -ub $tmp/list.orig $tmp/list; then
2083 error_noexit "list verification failed"
2087 echo "list verification skipped"
2090 if [ "$dne_upgrade" != "no" ]; then
2091 $LFS mkdir -i 1 -c2 $tmp/mnt/lustre/striped_dir || {
2092 error_noexit "set striped dir failed"
2096 $LFS setdirstripe -D -c2 $tmp/mnt/lustre/striped_dir
2098 pushd $tmp/mnt/lustre
2099 tar -cf - . --exclude=./striped_dir \
2100 --exclude=./striped_dir_old \
2101 --exclude=./remote_dir |
2102 tar -xvf - -C striped_dir 1>/dev/null || {
2103 error_noexit "cp to striped dir failed"
2109 # If it is upgrade from DNE (2.5), then rename the remote dir,
2110 # which is created in 2.5 to striped dir.
2111 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]]; then
2112 stripe_index=$($LFS getdirstripe -i \
2113 $tmp/mnt/lustre/remote_dir)
2115 [[ $stripe_index -eq 1 ]] || {
2116 error_noexit "get index \"$stripe_index\"" \
2117 "from remote dir failed"
2120 mv $tmp/mnt/lustre/remote_dir \
2121 $tmp/mnt/lustre/striped_dir/ || {
2122 error_noexit "mv remote dir failed"
2127 # If it is upgraded from DNE (2.7), then move the striped dir
2128 # which was created in 2.7 to the new striped dir.
2129 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]] &&
2130 [[ -d $tmp/mnt/lustre/striped_dir_old ]]; then
2131 stripe_count=$($LFS getdirstripe -c \
2132 $tmp/mnt/lustre/striped_dir_old)
2133 [[ $stripe_count -eq 2 ]] || {
2134 error_noexit "get count $stripe_count" \
2135 "from striped dir failed"
2138 mv $tmp/mnt/lustre/striped_dir_old \
2139 $tmp/mnt/lustre/striped_dir/ || {
2140 error_noexit "mv striped dir failed"
2146 $r $LCTL set_param -n osd*.*.force_sync=1
2147 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
2148 error_noexit "dd failed"
2151 rm -rf $tmp/mnt/lustre/tmp_file || {
2152 error_noexit "rm failed"
2156 if $r test -f $tmp/sha1sums; then
2157 # LU-2393 - do both sorts on same node to ensure locale
2159 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
2160 if [ "$dne_upgrade" != "no" ]; then
2161 pushd $tmp/mnt/lustre/striped_dir
2163 pushd $tmp/mnt/lustre
2166 find ! -path "*remote_dir*" ! -path "*striped_dir*" \
2167 ! -name .lustre -type f -exec sha1sum {} \; |
2168 sort -k 2 >$tmp/sha1sums || {
2170 error_noexit "sha1sum"
2174 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
2175 error_noexit "sha1sum verification failed"
2179 # if upgrade from DNE(2.5), then check remote directory
2180 # if upgrade from DNE(2.7), then check striped directory
2181 if $mdt2_is_available &&
2182 [[ "$dne_upgrade" != "no" ]]; then
2183 local new_dir="$tmp/mnt/lustre/striped_dir"
2184 local striped_dir_old="$new_dir/striped_dir_old"
2186 local dir_list="$new_dir/remote_dir"
2187 [[ ! -d $triped_dir_old ]] ||
2188 dir_list+=" $striped_dir_old"
2190 for dir in $dir_list; do
2192 find ! -name .lustre -type f \
2193 -exec sha1sum {} \; |
2194 sort -k 2 >$tmp/sha1sums || {
2196 error_noexit "sha1sum"
2200 if ! diff -ub $tmp/sha1sums.orig \
2202 error_noexit "sha1sum $dir" \
2209 echo "sha1sum verification skipped"
2212 if [ "$dne_upgrade" != "no" ]; then
2213 rm -rf $tmp/mnt/lustre/striped_dir || {
2214 error_noexit "remove remote dir failed"
2219 # migrate files/dirs to remote MDT, then move them back
2220 if [ $(lustre_version_code mds1) -ge $(version_code 2.7.50) -a \
2221 $dne_upgrade != "no" ]; then
2222 $r $LCTL set_param -n \
2223 mdt.${fsname}*.enable_remote_dir=1 2>/dev/null
2225 echo "test migration"
2226 pushd $tmp/mnt/lustre
2227 for dir in $(find ! -name .lustre ! -name . -type d); do
2228 mdt_index=$($LFS getdirstripe -i $dir)
2229 stripe_cnt=$($LFS getdirstripe -c $dir)
2230 if [ $mdt_index = 0 -a $stripe_cnt -le 1 ]; then
2231 $LFS mv -M 1 $dir || {
2233 error_noexit "migrate MDT1 failed"
2239 for dir in $(find ! -name . ! -name .lustre -type d); do
2240 mdt_index=$($LFS getdirstripe -i $dir)
2241 stripe_cnt=$($LFS getdirstripe -c $dir)
2242 if [ $mdt_index = 1 -a $stripe_cnt -le 1 ]; then
2243 $LFS mv -M 0 $dir || {
2245 error_noexit "migrate MDT0 failed"
2254 # When adding new data verification tests, please check for
2255 # the presence of the required reference files first, like
2256 # the "sha1sums" and "list" tests above, to avoid the need to
2257 # regenerate every image for each test addition.
2260 nrpcs_orig=$($LCTL get_param \
2261 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
2262 error_noexit "Getting \"max_rpcs_in_flight\""
2265 nrpcs=$((nrpcs_orig + 5))
2266 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=$nrpcs || {
2267 error_noexit "Changing \"max_rpcs_in_flight\""
2270 wait_update $HOSTNAME "$LCTL get_param \
2271 -n mdc.*MDT0000*.max_rpcs_in_flight" $nrpcs || {
2272 error_noexit "Verifying \"max_rpcs_in_flight\""
2276 umount $tmp/mnt/lustre || {
2277 error_noexit "Unmounting the client"
2280 shall_cleanup_lustre=false
2282 if [[ "$dne_upgrade" != "no" ]] || $mdt2_is_available; then
2283 $r $UMOUNT $tmp/mnt/mdt1 || {
2284 error_noexit "Unmounting the MDT2"
2287 shall_cleanup_mdt1=false
2290 $r $UMOUNT $tmp/mnt/mdt || {
2291 error_noexit "Unmounting the MDT"
2294 shall_cleanup_mdt=false
2296 $r $UMOUNT $tmp/mnt/ost || {
2297 error_noexit "Unmounting the OST"
2300 shall_cleanup_ost=false
2302 t32_reload_modules $node || {
2303 error_noexit "Reloading modules"
2307 if [[ $fstype == zfs ]]; then
2308 local poolname=t32fs-mdt1
2310 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
2311 $ZPOOL import -f -d $tmp $poolname"
2313 # upgrade zpool to latest supported features,
2314 # including dnode quota accounting in 0.7.0
2315 $r "$ZPOOL upgrade $poolname"
2318 # mount a second time to make sure we didnt leave upgrade flag on
2319 $r $TUNEFS --dryrun $mdt_dev || {
2321 error_noexit "tunefs.lustre before remounting the MDT"
2325 mopts=exclude=$fsname-OST0000
2326 if [ $fstype == "ldiskfs" ]; then
2329 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2330 error_noexit "Remounting the MDT"
2333 shall_cleanup_mdt=true
2343 for tarball in $tarballs; do
2344 t32_test $tarball || let "rc += $?"
2348 run_test 32a "Upgrade (not live)"
2356 for tarball in $tarballs; do
2357 t32_test $tarball writeconf || let "rc += $?"
2361 run_test 32b "Upgrade with writeconf"
2368 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
2370 for tarball in $tarballs; do
2371 # Do not support 1_8 and 2_1 direct upgrade to DNE2 anymore */
2372 echo $tarball | grep "1_8" && continue
2373 echo $tarball | grep "2_1" && continue
2375 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
2379 run_test 32c "dne upgrade test"
2387 for tarball in $tarballs; do
2388 ff_convert=yes t32_test $tarball || rc=$?
2392 run_test 32d "convert ff test"
2394 test_33a() { # bug 12333, was test_33
2395 local FSNAME2=test-123
2396 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2399 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
2401 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
2402 local dev=${SINGLEMDS}_dev
2403 local MDSDEV=${!dev}
2404 is_blkdev $SINGLEMDS $MDSDEV &&
2405 skip_env "mixed loopback and real device not working" &&
2409 local fs2mdsdev=$(mdsdevname 1_2)
2410 local fs2ostdev=$(ostdevname 1_2)
2411 local fs2mdsvdev=$(mdsvdevname 1_2)
2412 local fs2ostvdev=$(ostvdevname 1_2)
2414 if [ $(facet_fstype mds1) == ldiskfs ]; then
2415 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
2418 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2419 --reformat $mkfsoptions $fs2mdsdev $fs2mdsvdev || exit 10
2420 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2421 --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \
2422 $fs2ostvdev || exit 10
2424 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
2425 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2426 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" ||
2427 error "$LCTL conf_param $FSNAME2.sys.timeout=200 failed"
2428 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2429 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2432 cp /etc/hosts $MOUNT2/ || error "copy /etc/hosts $MOUNT2/ failed"
2433 $GETSTRIPE $MOUNT2/hosts || error "$GETSTRIPE $MOUNT2/hosts failed"
2438 cleanup_nocli || error "cleanup_nocli failed with $?"
2440 run_test 33a "Mount ost with a large index number"
2442 test_33b() { # was test_34
2445 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
2446 # Drop lock cancelation reply during umount
2447 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
2448 do_facet client $LCTL set_param fail_loc=0x80000304
2449 #lctl set_param debug=-1
2450 umount_client $MOUNT
2451 cleanup || error "cleanup failed with $?"
2453 run_test 33b "Drop cancel during umount"
2457 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
2458 manual_umount_client
2460 do_facet client killall -USR1 multiop
2461 if [ $rc -eq 0 ]; then
2462 error "umount not fail!"
2465 cleanup || error "cleanup failed with rc $?"
2467 run_test 34a "umount with opened file should be fail"
2471 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2472 stop_mds || error "Unable to stop MDS"
2474 manual_umount_client --force || error "mtab after failed umount with $?"
2476 cleanup || error "cleanup failed with $?"
2478 run_test 34b "force umount with failed mds should be normal"
2482 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2483 stop_ost || error "Unable to stop OST1"
2485 manual_umount_client --force || error "mtab after failed umount with $?"
2487 cleanup || error "cleanup failed with $?"
2489 run_test 34c "force umount with failed ost should be normal"
2491 test_35a() { # bug 12459
2494 DBG_SAVE="`$LCTL get_param -n debug`"
2495 $LCTL set_param debug="ha"
2497 log "Set up a fake failnode for the MDS"
2499 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2500 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2501 do_facet mgs "$LCTL conf_param \
2502 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" ||
2503 error "Setting ${device}.failover.node=\
2504 $(h2$NETTYPE $FAKENID) failed."
2506 log "Wait for RECONNECT_INTERVAL seconds (10s)"
2509 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
2512 log "Stopping the MDT: $device"
2513 stop_mdt 1 || error "MDT0 stop fail"
2515 df $MOUNT > /dev/null 2>&1 &
2517 log "Restarting the MDT: $device"
2518 start_mdt 1 || error "MDT0 start fail"
2519 log "Wait for df ($DFPID) ... "
2522 $LCTL set_param debug="$DBG_SAVE"
2524 # retrieve from the log the first server that the client tried to
2525 # contact after the connection loss
2526 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2527 NEXTCONN=`awk "/${MSG}/ {start = 1;}
2528 /import_select_connection.*$device-mdc.* using connection/ {
2530 if (\\\$NF ~ /$FAKENID/)
2536 }" $TMP/lustre-log-$TESTNAME.log`
2537 [ "$NEXTCONN" != "0" ] &&
2538 error "Tried to connect to ${NEXTCONN} not last active server"
2539 cleanup || error "cleanup failed with $?"
2540 # remove nid settings
2541 writeconf_or_reformat
2543 run_test 35a "Reconnect to the last active server first"
2545 test_35b() { # bug 18674
2546 remote_mds || { skip "local MDS" && return 0; }
2550 $LCTL set_param debug="ha"
2552 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
2555 log "Set up a fake failnode for the MDS"
2557 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2558 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2559 do_facet mgs "$LCTL conf_param \
2560 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" ||
2561 error "Set ${device}.failover.node=\
2562 $(h2$NETTYPE $FAKENID) failed"
2564 local at_max_saved=0
2565 # adaptive timeouts may prevent seeing the issue
2566 if at_is_enabled; then
2567 at_max_saved=$(at_max_get mds)
2568 at_max_set 0 mds client
2571 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
2573 log "Injecting EBUSY on MDS"
2574 # Setting OBD_FAIL_MDS_RESEND=0x136
2575 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" ||
2576 error "unable to set param fail_loc=0x80000136"
2578 $LCTL set_param mdc.${FSNAME}*.stats=clear
2580 log "Creating a test file and stat it"
2581 touch $MOUNT/$tdir/$tfile || error "touch $MOUNT/$tdir/$tfile failed"
2582 stat $MOUNT/$tdir/$tfile
2584 log "Stop injecting EBUSY on MDS"
2585 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" ||
2586 error "unable to set param fail_loc=0"
2587 rm -f $MOUNT/$tdir/$tfile || error "remove $MOUNT/$tdir/$tfile failed"
2590 # restore adaptive timeout
2591 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
2593 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2595 CONNCNT=$($LCTL get_param mdc.${FSNAME}*.stats |
2596 awk '/mds_connect/{print $2}')
2598 # retrieve from the log if the client has ever tried to
2599 # contact the fake server after the loss of connection
2600 FAILCONN=`awk "BEGIN {ret = 0;}
2601 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
2603 if (\\\$NF ~ /$FAKENID/) {
2608 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
2610 [ "$FAILCONN" == "0" ] &&
2611 error "The client reconnection has not been triggered"
2612 [ "$FAILCONN" == "2" ] &&
2613 error "Primary server busy, client reconnect to failover failed"
2616 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
2617 # Reconnects are supposed to be rate limited to one every 5s
2618 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] &&
2619 error "Too many reconnects $CONNCNT"
2621 cleanup || error "cleanup failed with $?"
2622 # remove nid settings
2623 writeconf_or_reformat
2625 run_test 35b "Continue reconnection retries, if the active server is busy"
2628 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
2630 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] ||
2631 { skip "remote OST" && return 0; }
2634 local FSNAME2=test1234
2635 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2637 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
2639 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
2640 is_blkdev $SINGLEMDS $MDSDEV &&
2641 skip_env "mixed loopback and real device not working" && return
2644 local fs2mdsdev=$(mdsdevname 1_2)
2645 local fs2ostdev=$(ostdevname 1_2)
2646 local fs3ostdev=$(ostdevname 2_2)
2647 local fs2mdsvdev=$(mdsvdevname 1_2)
2648 local fs2ostvdev=$(ostvdevname 1_2)
2649 local fs3ostvdev=$(ostvdevname 2_2)
2651 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2652 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
2653 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
2654 # different one than the default value here.
2655 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2656 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
2657 add fs3ost $(mkfs_opts ost2 ${fs3ostdev}) --mgsnode=$MGSNID \
2658 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
2660 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
2661 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2662 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
2663 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2664 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2666 sleep 5 # until 11778 fixed
2668 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || error "dd failed"
2670 BKTOTAL=$($LCTL get_param -n obdfilter.*.kbytestotal |
2671 awk 'BEGIN{total=0}; {total+=$1}; END{print total}')
2672 BKFREE=$($LCTL get_param -n obdfilter.*.kbytesfree |
2673 awk 'BEGIN{free=0}; {free+=$1}; END{print free}')
2674 BKAVAIL=$($LCTL get_param -n obdfilter.*.kbytesavail |
2675 awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}')
2676 STRING=$(df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}')
2677 DFTOTAL=$(echo $STRING | cut -d, -f1)
2678 DFUSED=$(echo $STRING | cut -d, -f2)
2679 DFAVAIL=$(echo $STRING | cut -d, -f3)
2680 DFFREE=$(($DFTOTAL - $DFUSED))
2682 ALLOWANCE=$((64 * $OSTCOUNT))
2684 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
2685 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
2686 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
2689 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
2690 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
2691 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
2694 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
2695 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
2696 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
2701 stop fs3ost -f || error "unable to stop OST3"
2702 stop fs2ost -f || error "unable to stop OST2"
2703 stop fs2mds -f || error "unable to stop second MDS"
2704 unload_modules_conf || error "unable unload modules"
2707 run_test 36 "df report consistency on OSTs with different block size"
2710 local mntpt=$(facet_mntpt $SINGLEMDS)
2711 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
2712 local mdsdev_sym="$TMP/sym_mdt.img"
2713 local opts=$MDS_MOUNT_OPTS
2716 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2717 skip "Currently only applicable to ldiskfs-based MDTs"
2721 echo "MDS : $mdsdev"
2722 echo "SYMLINK : $mdsdev_sym"
2723 do_facet $SINGLEMDS rm -f $mdsdev_sym
2725 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
2727 echo "mount symlink device - $mdsdev_sym"
2729 if ! do_facet $SINGLEMDS test -b $mdsdev; then
2730 opts=$(csa_add "$opts" -o loop)
2732 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
2733 $mdsdev_sym $mntpt 2>&1)
2736 echo mount_op=$mount_op
2738 do_facet $SINGLEMDS "$UMOUNT $mntpt && rm -f $mdsdev_sym"
2740 if $(echo $mount_op | grep -q "unable to set tunable"); then
2741 error "set tunables failed for symlink device"
2744 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
2746 run_test 37 "verify set tunables works for symlink device"
2748 test_38() { # bug 14222
2749 local fstype=$(facet_fstype $SINGLEMDS)
2750 local mntpt=$(facet_mntpt $SINGLEMDS)
2755 local SRC="/etc /bin"
2756 local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT)
2757 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
2758 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2759 tar cf - $FILES | tar xf - -C $DIR/$tdir ||
2760 error "copying $SRC to $DIR/$tdir"
2762 umount_client $MOUNT || error "umount_client $MOUNT failed"
2763 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2764 stop_mds || error "Unable to stop MDS"
2765 log "delete lov_objid file on MDS"
2767 mount_fstype $SINGLEMDS || error "mount MDS failed (1)"
2769 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid; rm $mntpt/lov_objid"
2771 unmount_fstype $SINGLEMDS || error "umount failed (1)"
2773 # check create in mds_lov_connect
2774 start_mds || error "unable to start MDS"
2775 mount_client $MOUNT || error "mount_client $MOUNT failed"
2777 [ $V ] && log "verifying $DIR/$tdir/$f"
2778 diff -q $f $DIR/$tdir/$f || ERROR=y
2780 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2781 if [ "$ERROR" = "y" ]; then
2782 # check it's updates in sync
2783 umount_client $MOUNT
2785 mount_fstype $SIGNLEMDS
2786 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2787 unmount_fstype $SINGLEMDS
2788 error "old and new files are different after connect" || true
2790 touch $DIR/$tdir/f2 || error "f2 file create failed"
2792 # check it's updates in sync
2793 umount_client $MOUNT || error "second umount_client $MOUNT failed"
2796 mount_fstype $SINGLEMDS || error "mount MDS failed (3)"
2798 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2799 do_facet $SINGLEMDS dd if=/dev/zero of=$mntpt/lov_objid.clear count=8
2801 unmount_fstype $SINGLEMDS || error "umount failed (3)"
2803 start_mds || error "unable to start MDS"
2804 mount_client $MOUNT || error "mount_client $MOUNT failed"
2806 [ $V ] && log "verifying $DIR/$tdir/$f"
2807 diff -q $f $DIR/$tdir/$f || ERROR=y
2809 touch $DIR/$tdir/f3 || error "f3 file create failed"
2810 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2811 umount_client $MOUNT || error "third umount_client $MOUNT failed"
2813 mount_fstype $SINGLEMDS || error "mount MDS failed (4)"
2814 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2815 unmount_fstype $SINGLEMDS || error "umount failed (4)"
2817 [ "$ERROR" = "y" ] &&
2818 error "old and new files are different after sync" || true
2820 log "files compared the same"
2821 cleanup || error "cleanup failed with $?"
2823 run_test 38 "MDS recreates missing lov_objid file from OST data"
2828 cleanup || error "cleanup failed with $?"
2829 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
2830 error "memory leak detected" || true
2832 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
2834 test_40() { # bug 15759
2835 start_ost || error "Unable to start OST1"
2836 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
2837 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
2839 cleanup || error "cleanup failed with rc $?"
2841 run_test 40 "race during service thread startup"
2843 test_41a() { #bug 14134
2844 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2845 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2846 skip "Loop devices does not work with nosvc option"
2851 { skip "needs combined MGT and MDT device" && return 0; }
2853 start_mdt 1 -o nosvc -n
2854 if [ $MDSCOUNT -ge 2 ]; then
2855 for num in $(seq 2 $MDSCOUNT); do
2856 start_mdt $num || return
2859 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
2860 start_mdt 1 -o nomgs,force
2861 mount_client $MOUNT || error "mount_client $MOUNT failed"
2864 echo "blah blah" > $MOUNT/$tfile
2867 umount_client $MOUNT || error "umount_client $MOUNT failed"
2868 stop ost1 -f || error "unable to stop OST1"
2869 stop_mds || error "Unable to stop MDS"
2870 stop_mds || error "Unable to stop MDS on second try"
2872 run_test 41a "mount mds with --nosvc and --nomgs"
2875 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2876 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2877 skip "Loop devices does not work with nosvc option"
2881 ! combined_mgs_mds && skip "needs combined mgs device" && return 0
2885 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2887 start_mdt 1 -o nosvc -n
2888 if [ $MDSCOUNT -ge 2 ]; then
2889 for num in $(seq 2 $MDSCOUNT); do
2890 start_mdt $num || return
2893 start_ost || error "Unable to start OST1"
2894 start_mdt 1 -o nomgs,force
2895 mount_client $MOUNT || error "mount_client $MOUNT failed"
2898 echo "blah blah" > $MOUNT/$tfile
2899 cat $MOUNT/$tfile || error "cat $MOUNT/$tfile failed"
2901 umount_client $MOUNT || error "umount_client $MOUNT failed"
2902 stop_ost || error "Unable to stop OST1"
2903 stop_mds || error "Unable to stop MDS"
2904 stop_mds || error "Unable to stop MDS on second try"
2906 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
2909 local server_version=$(lustre_version_code $SINGLEMDS)
2910 local oss_list=$(comma_list $(osts_nodes))
2912 [[ $server_version -ge $(version_code 2.6.52) ]] ||
2913 [[ $server_version -ge $(version_code 2.5.26) &&
2914 $server_version -lt $(version_code 2.5.50) ]] ||
2915 [[ $server_version -ge $(version_code 2.5.4) &&
2916 $server_version -lt $(version_code 2.5.11) ]] ||
2917 { skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"; return; }
2919 # ensure mds1 ost1 have been created even if running sub-test standalone
2922 cleanup || error "cleanup failed"
2924 # using directly mount command instead of start() function to avoid
2925 # any side effect of // with others/externals tools/features
2926 # ("zpool import", ...)
2928 # MDT concurrent start
2930 LOAD_MODULES_REMOTE=true load_modules
2931 do_facet $SINGLEMDS "lsmod | grep -q libcfs" ||
2932 error "MDT concurrent start: libcfs module not loaded"
2934 local mds1dev=$(mdsdevname 1)
2935 local mds1mnt=$(facet_mntpt mds1)
2936 local mds1fstype=$(facet_fstype mds1)
2937 local mds1opts=$MDS_MOUNT_OPTS
2939 if [ $mds1fstype == ldiskfs ] &&
2940 ! do_facet mds1 test -b $mds1dev; then
2941 mds1opts=$(csa_add "$mds1opts" -o loop)
2943 if [[ $mds1fstype == zfs ]]; then
2944 import_zpool mds1 || return ${PIPESTATUS[0]}
2947 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
2948 do_facet mds1 "$LCTL set_param fail_loc=0x80000716"
2950 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts &
2953 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts
2957 do_facet mds1 "$LCTL set_param fail_loc=0x0"
2958 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
2959 echo "1st MDT start succeed"
2960 echo "2nd MDT start failed with $rc2"
2961 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
2962 echo "1st MDT start failed with $rc"
2963 echo "2nd MDT start succeed"
2966 error "unexpected concurrent MDT mounts result, rc=$rc rc2=$rc2"
2969 if [ $MDSCOUNT -ge 2 ]; then
2970 for num in $(seq 2 $MDSCOUNT); do
2971 start_mdt $num || return
2975 # OST concurrent start
2977 do_rpc_nodes $oss_list "lsmod | grep -q libcfs" ||
2978 error "OST concurrent start: libcfs module not loaded"
2980 local ost1dev=$(ostdevname 1)
2981 local ost1mnt=$(facet_mntpt ost1)
2982 local ost1fstype=$(facet_fstype ost1)
2983 local ost1opts=$OST_MOUNT_OPTS
2985 if [ $ost1fstype == ldiskfs ] &&
2986 ! do_facet ost1 test -b $ost1dev; then
2987 ost1opts=$(csa_add "$ost1opts" -o loop)
2989 if [[ $ost1fstype == zfs ]]; then
2990 import_zpool ost1 || return ${PIPESTATUS[0]}
2993 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
2994 do_facet ost1 "$LCTL set_param fail_loc=0x80000716"
2996 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts &
2999 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts
3003 do_facet ost1 "$LCTL set_param fail_loc=0x0"
3004 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3005 echo "1st OST start succeed"
3006 echo "2nd OST start failed with $rc2"
3007 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3008 echo "1st OST start failed with $rc"
3009 echo "2nd OST start succeed"
3013 error "unexpected concurrent OST mounts result, rc=$rc rc2=$rc2"
3019 # verify everything ok
3024 error "MDT(s) start failed"
3032 error "OST(s) start failed"
3040 error "client start failed"
3047 error "client mount failed"
3051 run_test 41c "concurrent mounts of MDT/OST should all fail but one"
3053 test_42() { #bug 14693
3055 check_mount || error "client was not mounted"
3057 do_facet mgs $LCTL conf_param $FSNAME.llite.some_wrong_param=10
3058 umount_client $MOUNT ||
3059 error "unmounting client failed with invalid llite param"
3060 mount_client $MOUNT ||
3061 error "mounting client failed with invalid llite param"
3063 do_facet mgs $LCTL conf_param $FSNAME.sys.some_wrong_param=20
3064 cleanup || error "stopping $FSNAME failed with invalid sys param"
3066 check_mount || error "client was not mounted with invalid sys param"
3067 cleanup || error "stopping $FSNAME failed with invalid sys param"
3069 run_test 42 "allow client/server mount/unmount with invalid config param"
3072 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.58) ]] ||
3073 { skip "Need MDS version at least 2.5.58" && return 0; }
3074 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
3077 USER1=$(getent passwd | grep :$ID1:$ID1: | cut -d: -f1)
3078 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1" &&
3082 chmod ugo+x $DIR || error "chmod 0 failed"
3083 set_conf_param_and_check mds1 \
3084 "$LCTL get_param -n mdt.$FSNAME-MDT0000.root_squash" \
3085 "$FSNAME.mdt.root_squash" \
3087 wait_update $HOSTNAME \
3088 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3090 error "check llite root_squash failed!"
3091 set_conf_param_and_check mds1 \
3092 "$LCTL get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
3093 "$FSNAME.mdt.nosquash_nids" \
3095 wait_update $HOSTNAME \
3096 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3098 error "check llite nosquash_nids failed!"
3101 # create set of test files
3103 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
3104 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
3105 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
3107 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
3108 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
3110 mkdir $DIR/$tdir-rootdir || error "mkdir failed"
3111 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
3112 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
3114 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
3115 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
3116 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
3119 # check root_squash:
3120 # set root squash UID:GID to RUNAS_ID
3121 # root should be able to access only files owned by RUNAS_ID
3123 set_conf_param_and_check mds1 \
3124 "$LCTL get_param -n mdt.$FSNAME-MDT0000.root_squash" \
3125 "$FSNAME.mdt.root_squash" \
3126 "$RUNAS_ID:$RUNAS_ID"
3127 wait_update $HOSTNAME \
3128 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3129 "$RUNAS_ID:$RUNAS_ID" ||
3130 error "check llite root_squash failed!"
3132 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
3133 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3134 error "$ST: root read permission is denied"
3135 echo "$ST: root read permission is granted - ok"
3138 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3139 error "$ST: root write permission is denied"
3140 echo "$ST: root write permission is granted - ok"
3142 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3143 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3144 error "$ST: root read permission is granted"
3145 echo "$ST: root read permission is denied - ok"
3148 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3149 error "$ST: root write permission is granted"
3150 echo "$ST: root write permission is denied - ok"
3152 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3153 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null &&
3154 error "$ST: root unlink permission is granted"
3155 echo "$ST: root unlink permission is denied - ok"
3157 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null &&
3158 error "$ST: root create permission is granted"
3159 echo "$ST: root create permission is denied - ok"
3162 # check root_squash is enforced independently
3163 # of client cache content
3165 # access file by USER1, keep access open
3166 # root should be denied access to user file
3168 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
3172 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
3173 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3174 { kill $pid; error "$ST: root read permission is granted"; }
3175 echo "$ST: root read permission is denied - ok"
3178 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3179 { kill $pid; error "$ST: root write permission is granted"; }
3180 echo "$ST: root write permission is denied - ok"
3186 # check nosquash_nids:
3187 # put client's NID into nosquash_nids list,
3188 # root should be able to access root file after that
3190 local NIDLIST=$($LCTL list_nids all | tr '\n' ' ')
3191 NIDLIST="2@gni $NIDLIST 192.168.0.[2,10]@tcp"
3192 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
3193 set_conf_param_and_check mds1 \
3194 "$LCTL get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
3195 "$FSNAME-MDTall.mdt.nosquash_nids" \
3197 wait_update $HOSTNAME \
3198 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3200 error "check llite nosquash_nids failed!"
3202 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3203 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3204 error "$ST: root read permission is denied"
3205 echo "$ST: root read permission is granted - ok"
3208 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3209 error "$ST: root write permission is denied"
3210 echo "$ST: root write permission is granted - ok"
3212 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3213 rm $DIR/$tdir-rootdir/tfile-1 ||
3214 error "$ST: root unlink permission is denied"
3215 echo "$ST: root unlink permission is granted - ok"
3216 touch $DIR/$tdir-rootdir/tfile-2 ||
3217 error "$ST: root create permission is denied"
3218 echo "$ST: root create permission is granted - ok"
3219 cleanup || error "cleanup failed with $?"
3221 run_test 43a "check root_squash and nosquash_nids"
3223 test_43b() { # LU-5690
3224 [[ $(lustre_version_code mgs) -ge $(version_code 2.7.62) ]] ||
3225 { skip "Need MGS version 2.7.62+"; return; }
3227 if [[ -z "$fs2mds_DEV" ]]; then
3228 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
3229 skip_env "mixed loopback and real device not working" && return
3232 local fs2mdsdev=$(mdsdevname 1_2)
3233 local fs2mdsvdev=$(mdsvdevname 1_2)
3235 # temporarily use fs2mds as fs2mgs
3237 local fs2mgsdev=$fs2mdsdev
3238 local fs2mgsvdev=$fs2mdsvdev
3240 local fsname=test1234
3242 load_module llite/lustre
3243 local client_ip=$(host_nids_address $HOSTNAME $NETTYPE)
3244 local host=${client_ip//*./}
3245 local net=${client_ip/%$host/}
3246 local nosquash_nids=$(h2$NETTYPE $net[$host,$host,$host])
3248 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
3249 --param mdt.root_squash=$RUNAS_ID:$RUNAS_ID \
3250 --param mdt.nosquash_nids=$nosquash_nids \
3251 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
3252 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS || error "start fs2mgs failed"
3253 stop $fs2mgs -f || error "stop fs2mgs failed"
3255 run_test 43b "parse nosquash_nids with commas in expr_list"
3257 umount_client $MOUNT
3262 check_mount || error "check_mount"
3263 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
3265 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
3266 for VAL in $UUIDS; do
3267 NID=$(echo $VAL | cut -d= -f1)
3268 CLUUID=$(echo $VAL | cut -d= -f2)
3269 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
3271 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
3272 cleanup || error "cleanup failed with $?"
3274 run_test 44 "mounted client proc entry exists"
3278 check_mount || error "check_mount"
3279 stop_mds || error "Unable to stop MDS"
3283 #define OBD_FAIL_PTLRPC_LONG_REPL_UNLINK 0x50f
3284 do_facet client "$LCTL set_param fail_loc=0x8000050f"
3287 manual_umount_client --force || error "manual_umount_client failed"
3288 do_facet client "$LCTL set_param fail_loc=0x0"
3289 start_mds || error "unable to start MDS"
3290 mount_client $MOUNT || error "mount_client $MOUNT failed"
3291 cleanup || error "cleanup failed with $?"
3293 run_test 45 "long unlink handling in ptlrpcd"
3300 umount_client $MOUNT2 || rc=$?
3301 umount_client $MOUNT || rc=$?
3302 while [ $count -gt 0 ]; do
3303 stop ost${count} -f || rc=$?
3307 cleanup_nocli || rc=$?
3308 #writeconf to remove all ost2 traces for subsequent tests
3309 writeconf_or_reformat
3314 echo "Testing with $OSTCOUNT OSTs"
3316 start_mds || error "unable to start MDS"
3317 #first client should see only one ost
3318 start_ost || error "Unable to start OST1"
3319 wait_osc_import_state mds ost FULL
3321 mount_client $MOUNT || error "mount_client $MOUNT failed"
3322 trap "cleanup_46a $OSTCOUNT" EXIT ERR
3325 for (( i=2; i<=$OSTCOUNT; i++ )); do
3326 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
3327 error "start_ost$i $(ostdevname $i) failed"
3330 # wait until osts in sync
3331 for (( i=2; i<=$OSTCOUNT; i++ )); do
3332 wait_osc_import_state mds ost$i FULL
3333 wait_osc_import_state client ost$i FULL
3336 #second client see all ost's
3338 mount_client $MOUNT2 || error "mount_client failed"
3339 $SETSTRIPE -c -1 $MOUNT2 || error "$SETSTRIPE -c -1 $MOUNT2 failed"
3340 $GETSTRIPE $MOUNT2 || error "$GETSTRIPE $MOUNT2 failed"
3342 echo "ok" > $MOUNT2/widestripe
3343 $GETSTRIPE $MOUNT2/widestripe ||
3344 error "$GETSTRIPE $MOUNT2/widestripe failed"
3345 # fill acl buffer for avoid expand lsm to them
3346 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd |
3348 setfacl -m $acl $MOUNT2/widestripe
3352 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3354 cleanup_46a $OSTCOUNT || error "cleanup_46a failed"
3356 run_test 46a "handle ost additional - wide striped file"
3361 check_mount || error "check_mount failed"
3362 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
3366 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3367 if echo $ns | grep "MDT[[:digit:]]*"; then
3370 lrs=$(echo $ns | sed 's/.*lru_size=//')
3371 lru_size[count]=$lrs
3376 facet_failover $SINGLEMDS
3377 client_up || error "client_up failed"
3380 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3381 if echo $ns | grep "MDT[[:digit:]]*"; then
3384 lrs=$(echo $ns | sed 's/.*lru_size=//')
3385 if ! test "$lrs" -eq "${lru_size[count]}"; then
3386 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
3387 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
3392 cleanup || error "cleanup failed with $?"
3394 run_test 47 "server restart does not make client loss lru_resize settings"
3399 # reformat after this test is needed - if the test fails,
3400 # we will have unkillable file at FS
3404 test_48() { # bug 17636
3407 check_mount || error "check_mount failed"
3409 $SETSTRIPE -c -1 $MOUNT || error "$SETSTRIPE -c -1 $MOUNT failed"
3410 $GETSTRIPE $MOUNT || error "$GETSTRIPE $MOUNT failed"
3412 echo "ok" > $MOUNT/widestripe
3413 $GETSTRIPE $MOUNT/widestripe ||
3414 error "$GETSTRIPE $MOUNT/widestripe failed"
3416 trap cleanup_48 EXIT ERR
3418 # fill acl buffer for avoid expand lsm to them
3419 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
3420 setfacl -m $acl $MOUNT/widestripe
3423 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3427 run_test 48 "too many acls on file"
3429 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
3430 test_49a() { # bug 17710
3431 local timeout_orig=$TIMEOUT
3432 local ldlm_timeout_orig=$LDLM_TIMEOUT
3433 local LOCAL_TIMEOUT=20
3435 LDLM_TIMEOUT=$LOCAL_TIMEOUT
3436 TIMEOUT=$LOCAL_TIMEOUT
3440 check_mount || error "client mount failed"
3442 echo "check ldlm_timout..."
3443 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3444 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3445 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3447 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3448 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3451 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
3452 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
3455 umount_client $MOUNT || error "umount_client $MOUNT failed"
3456 stop_ost || error "problem stopping OSS"
3457 stop_mds || error "problem stopping MDS"
3459 LDLM_TIMEOUT=$ldlm_timeout_orig
3460 TIMEOUT=$timeout_orig
3462 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3464 test_49b() { # bug 17710
3465 local timeout_orig=$TIMEOUT
3466 local ldlm_timeout_orig=$LDLM_TIMEOUT
3467 local LOCAL_TIMEOUT=20
3469 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
3470 TIMEOUT=$LOCAL_TIMEOUT
3474 check_mount || error "client mount failed"
3476 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3477 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3478 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3480 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3481 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3484 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
3485 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
3488 cleanup || error "cleanup failed"
3490 LDLM_TIMEOUT=$ldlm_timeout_orig
3491 TIMEOUT=$timeout_orig
3493 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3496 # Test both statfs and lfs df and fail if either one fails
3497 multiop_bg_pause $1 f_
3500 killall -USR1 multiop
3501 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
3502 wait $PID || { RC1=$?; log "multiop return error "; }
3509 if [ $RC2 -eq 0 ]; then
3511 log "lazystatfs df failed"
3515 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
3521 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3522 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3524 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
3526 cleanup || error "cleanup failed with rc $?"
3528 run_test 50a "lazystatfs all servers available"
3532 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3533 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3535 # Wait for client to detect down OST
3536 stop_ost || error "Unable to stop OST1"
3537 wait_osc_import_state mds ost DISCONN
3539 lazystatfs $MOUNT || error "lazystatfs should not return EIO"
3541 umount_client $MOUNT || error "Unable to unmount client"
3542 stop_mds || error "Unable to stop MDS"
3544 run_test 50b "lazystatfs all servers down"
3547 start_mds || error "Unable to start MDS"
3548 start_ost || error "Unable to start OST1"
3549 start_ost2 || error "Unable to start OST2"
3550 mount_client $MOUNT || error "Unable to mount client"
3551 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3552 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3554 # Wait for client to detect down OST
3555 stop_ost || error "Unable to stop OST1"
3556 wait_osc_import_state mds ost DISCONN
3557 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3559 umount_client $MOUNT || error "Unable to unmount client"
3560 stop_ost2 || error "Unable to stop OST2"
3561 stop_mds || error "Unable to stop MDS"
3562 #writeconf to remove all ost2 traces for subsequent tests
3563 writeconf_or_reformat
3565 run_test 50c "lazystatfs one server down"
3568 start_mds || error "Unable to start MDS"
3569 start_ost || error "Unable to start OST1"
3570 start_ost2 || error "Unable to start OST2"
3571 mount_client $MOUNT || error "Unable to mount client"
3572 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3573 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3575 # Issue the statfs during the window where the client still
3576 # belives the OST to be available but it is in fact down.
3577 # No failure just a statfs which hangs for a timeout interval.
3578 stop_ost || error "Unable to stop OST1"
3579 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3581 umount_client $MOUNT || error "Unable to unmount client"
3582 stop_ost2 || error "Unable to stop OST2"
3583 stop_mds || error "Unable to stop MDS"
3584 #writeconf to remove all ost2 traces for subsequent tests
3585 writeconf_or_reformat
3587 run_test 50d "lazystatfs client/server conn race"
3594 start_mds || error "Unable to start MDS"
3595 #first client should see only one ost
3596 start_ost || error "Unable to start OST1"
3597 wait_osc_import_state mds ost FULL
3599 # Wait for client to detect down OST
3600 stop_ost || error "Unable to stop OST1"
3601 wait_osc_import_state mds ost DISCONN
3603 mount_client $MOUNT || error "Unable to mount client"
3604 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3606 multiop_bg_pause $MOUNT _f
3610 if [ $RC1 -ne 0 ]; then
3611 log "multiop failed $RC1"
3614 sleep $(( $TIMEOUT+1 ))
3616 [ $? -ne 0 ] && error "process isn't sleep"
3617 start_ost || error "Unable to start OST1"
3618 wait $pid || error "statfs failed"
3621 umount_client $MOUNT || error "Unable to unmount client"
3622 stop_ost || error "Unable to stop OST1"
3623 stop_mds || error "Unable to stop MDS"
3625 run_test 50e "normal statfs all servers down"
3630 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
3632 start_mds || error "Unable to start mds"
3633 #first client should see only one ost
3634 start_ost || error "Unable to start OST1"
3635 wait_osc_import_state mds ost FULL
3637 start_ost2 || error "Unable to start OST2"
3638 wait_osc_import_state mds ost2 FULL
3640 # Wait for client to detect down OST
3641 stop_ost2 || error "Unable to stop OST2"
3643 wait_osc_import_state mds ost2 DISCONN
3644 mount_client $MOUNT || error "Unable to mount client"
3645 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3647 multiop_bg_pause $MOUNT _f
3651 if [ $RC1 -ne 0 ]; then
3652 log "lazystatfs multiop failed $RC1"
3655 sleep $(( $TIMEOUT+1 ))
3657 [ $? -ne 0 ] && error "process isn't sleep"
3658 start_ost2 || error "Unable to start OST2"
3659 wait $pid || error "statfs failed"
3660 stop_ost2 || error "Unable to stop OST2"
3663 umount_client $MOUNT -f || error "Unable to unmount client"
3664 stop_ost || error "Unable to stop OST1"
3665 stop_mds || error "Unable to stop MDS"
3666 #writeconf to remove all ost2 traces for subsequent tests
3667 writeconf_or_reformat
3669 run_test 50f "normal statfs one server in down"
3672 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3674 start_ost2 || error "Unable to start OST2"
3675 wait_osc_import_state mds ost2 FULL
3676 wait_osc_import_state client ost2 FULL
3678 local PARAM="${FSNAME}-OST0001.osc.active"
3680 $SETSTRIPE -c -1 $DIR/$tfile || error "$SETSTRIPE failed"
3681 do_facet mgs $LCTL conf_param $PARAM=0 ||
3682 error "Unable to deactivate OST"
3684 umount_client $MOUNT || error "Unable to unmount client"
3685 mount_client $MOUNT || error "Unable to mount client"
3686 # This df should not cause a panic
3689 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
3690 rm -f $DIR/$tfile || error "unable to remove file $DIR/$tfile"
3691 umount_client $MOUNT || error "Unable to unmount client"
3692 stop_ost2 || error "Unable to stop OST2"
3693 stop_ost || error "Unable to stop OST1"
3694 stop_mds || error "Unable to stop MDS"
3695 #writeconf to remove all ost2 traces for subsequent tests
3696 writeconf_or_reformat
3698 run_test 50g "deactivated OST should not cause panic"
3702 # prepare MDT/OST, make OSC inactive for OST1
3703 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3705 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3706 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
3707 error "tunefs OST1 failed"
3708 start_mds || error "Unable to start MDT"
3709 start_ost || error "Unable to start OST1"
3710 start_ost2 || error "Unable to start OST2"
3711 mount_client $MOUNT || error "client start failed"
3713 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3715 # activatate OSC for OST1
3716 local TEST="$LCTL get_param -n osc.${FSNAME}-OST0000-osc-[!M]*.active"
3717 set_conf_param_and_check client \
3718 "$TEST" "${FSNAME}-OST0000.osc.active" 1 ||
3719 error "Unable to activate OST1"
3721 mkdir $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3722 $SETSTRIPE -c -1 -i 0 $DIR/$tdir/2 ||
3723 error "$SETSTRIPE $DIR/$tdir/2 failed"
3724 sleep 1 && echo "create a file after OST1 is activated"
3726 createmany -o $DIR/$tdir/2/$tfile-%d 1
3728 # check OSC import is working
3729 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
3730 error "some OSC imports are still not connected"
3733 umount_client $MOUNT || error "Unable to umount client"
3734 stop_ost2 || error "Unable to stop OST2"
3735 cleanup_nocli || error "cleanup_nocli failed with $?"
3737 run_test 50h "LU-642: activate deactivated OST"
3740 # prepare MDT/OST, make OSC inactive for OST1
3741 [ "$MDSCOUNT" -lt "2" ] && skip_env "$MDSCOUNT < 2, skipping" && return
3743 [ $(facet_fstype mds2) == zfs ] && import_zpool mds2
3744 do_facet mds2 "$TUNEFS --param mdc.active=0 $(mdsdevname 2)" ||
3745 error "tunefs MDT2 failed"
3746 start_mds || error "Unable to start MDT"
3747 start_ost || error "Unable to start OST1"
3748 start_ost2 || error "Unable to start OST2"
3749 mount_client $MOUNT || error "client start failed"
3751 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3753 $LCTL conf_param ${FSNAME}-MDT0000.mdc.active=0 &&
3754 error "deactive MDC0 succeeds"
3755 # activate MDC for MDT2
3756 local TEST="$LCTL get_param -n mdc.${FSNAME}-MDT0001-mdc-[!M]*.active"
3757 set_conf_param_and_check client \
3758 "$TEST" "${FSNAME}-MDT0001.mdc.active" 1 ||
3759 error "Unable to activate MDT2"
3761 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
3762 if [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.60) ]
3764 wait_dne_interconnect
3766 $LFS mkdir -i1 $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3768 createmany -o $DIR/$tdir/2/$tfile-%d 1 || error "create files failed"
3770 rm -rf $DIR/$tdir/2 || error "unlink dir failed"
3772 # deactivate MDC for MDT2
3773 local TEST="$LCTL get_param -n mdc.${FSNAME}-MDT0001-mdc-[!M]*.active"
3774 set_conf_param_and_check client \
3775 "$TEST" "${FSNAME}-MDT0001.mdc.active" 0 ||
3776 error "Unable to deactivate MDT2"
3778 $LFS mkdir -i1 $DIR/$tdir/2 &&
3779 error "mkdir $DIR/$tdir/2 succeeds after deactive MDT"
3782 umount_client $MOUNT || error "Unable to umount client"
3787 run_test 50i "activate deactivated MDT"
3790 local LOCAL_TIMEOUT=20
3794 check_mount || error "check_mount failed"
3796 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
3797 $SETSTRIPE -c -1 $MOUNT/$tdir ||
3798 error "$SETSTRIPE -c -1 $MOUNT/$tdir failed"
3799 #define OBD_FAIL_MDS_REINT_DELAY 0x142
3800 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x142"
3801 touch $MOUNT/$tdir/$tfile &
3804 start_ost2 || error "Unable to start OST1"
3806 stop_ost2 || error "Unable to stop OST1"
3807 umount_client $MOUNT -f || error “unmount $MOUNT failed”
3808 cleanup_nocli || error “stop server failed”
3809 #writeconf to remove all ost2 traces for subsequent tests
3810 writeconf_or_reformat
3812 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
3821 do_node $node mkdir -p $dest
3822 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3824 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
3825 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3826 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
3828 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
3829 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
3839 local backup2=${TMP}/backup2
3841 do_node $node mkdir -p $backup2
3842 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3844 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
3845 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3846 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
3848 do_node $node "diff -rq $backup $backup2"
3849 [ $? -eq 0 ] || { error "contents differ"; return 3; }
3851 local xattrs2=${TMP}/xattrs2
3852 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
3853 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
3855 do_node $node "diff $xattrs $xattrs2"
3856 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
3858 do_node $node "rm -rf $backup2 $xattrs2"
3859 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
3863 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3864 skip "Only applicable to ldiskfs-based MDTs"
3868 start_mds || error "Unable to start MDS"
3869 start_ost || error "Unable to start OST1"
3870 mount_client $MOUNT || error "Unable to mount client"
3873 local ost1mnt=$(facet_mntpt ost1)
3874 local ost1node=$(facet_active_host ost1)
3875 local ost1tmp=$TMP/conf52
3878 mkdir $DIR/$tdir || error "Unable to create $DIR/$tdir"
3879 touch $TMP/modified_first || error "Unable to create temporary file"
3880 local mtime=$(stat -c %Y $TMP/modified_first)
3881 do_node $ost1node "mkdir -p $ost1tmp &&
3882 touch -m -d @$mtime $ost1tmp/modified_first" ||
3883 error "Unable to create temporary file"
3886 $SETSTRIPE -c -1 -S 1M $DIR/$tdir || error "$SETSTRIPE failed"
3888 for (( i=0; i < nrfiles; i++ )); do
3889 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c ||
3890 error "multiop failed"
3896 echo backup files to $TMP/$tdir
3897 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
3898 copy_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
3899 error "Unable to copy files"
3901 umount_client $MOUNT || error "Unable to umount client"
3902 stop_ost || error "Unable to stop ost1"
3904 echo mount ost1 as ldiskfs
3905 do_node $ost1node mkdir -p $ost1mnt || error "Unable to create $ost1mnt"
3906 if ! do_node $ost1node test -b $ost1_dev; then
3909 do_node $ost1node mount -t $(facet_fstype ost1) $loop $ost1_dev \
3911 error "Unable to mount ost1 as ldiskfs"
3914 echo backup objects to $ost1tmp/objects
3915 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
3916 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
3917 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
3919 error "Unable to copy objects"
3921 # move objects to lost+found
3922 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
3923 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
3925 do_node $ost1node "umount $ost1mnt" ||
3926 error "Unable to umount ost1 as ldiskfs"
3928 start_ost || error "Unable to start OST1"
3929 mount_client $MOUNT || error "Unable to mount client"
3931 local REPAIRED=$(do_node $ost1node "$LCTL get_param \
3932 -n osd-ldiskfs.$FSNAME-OST0000.oi_scrub" |
3933 awk '/^lf_repa[ri]*ed/ { print $2 }')
3934 [ $REPAIRED -gt 0 ] ||
3935 error "Some entry under /lost+found should be repaired"
3938 diff_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
3939 error "Unable to diff files"
3941 rm -rf $TMP/$tdir $TMP/file_xattrs ||
3942 error "Unable to delete temporary files"
3943 do_node $ost1node "rm -rf $ost1tmp" ||
3944 error "Unable to delete temporary files"
3945 cleanup || error "cleanup failed with $?"
3947 run_test 52 "check recovering objects from lost+found"
3949 # Checks threads_min/max/started for some service
3951 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
3952 # parameter pattern prefix like 'ost.*.ost'.
3965 local msg="Insane $modname thread counts"
3966 local ncpts=$(check_cpt_number $facet)
3970 check_mount || return 41
3972 # We need to expand $parampat, but it may match multiple parameters, so
3973 # we'll pick the first one
3974 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
3975 error "Couldn't expand ${parampat}.threads_min parameter name"
3979 # Remove the .threads_min part
3980 paramp=${paramp%.threads_min}
3982 # Check for sanity in defaults
3983 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3985 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3987 tstarted=$(do_facet $facet "$LCTL get_param \
3988 -n ${paramp}.threads_started" || echo 0)
3989 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' ||
3991 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
3993 nthrs=$(expr $tmax - $tmin)
3994 if [ $nthrs -lt $ncpts ]; then
4000 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
4001 skip_env "module parameter forced $facet thread count" &&
4002 tmin=3 && tmax=$((3 * tmax))
4004 # Check that we can change min/max
4005 do_facet $facet "$LCTL set_param \
4006 ${paramp}.threads_min=$((tmin + nthrs))"
4007 do_facet $facet "$LCTL set_param \
4008 ${paramp}.threads_max=$((tmax - nthrs))"
4009 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4011 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4013 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) &&
4014 $tmax2 == ($tmax - $nthrs)))' || return $?
4016 # Check that we can set min/max to the same value
4017 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4019 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$tmin"
4020 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4022 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4024 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
4026 # Check that we can't set max < min
4027 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$((tmin - 1))"
4028 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4030 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4032 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
4034 # We need to ensure that we get the module options desired; to do this
4035 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
4036 LOAD_MODULES_REMOTE=true
4039 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
4040 setmodopts -a $modname "$newvalue" oldvalue
4043 check_mount || return 41
4045 # Restore previous setting of MODOPTS_*
4046 setmodopts $modname "$oldvalue"
4048 # Check that $opts took
4049 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min")
4050 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max")
4051 tstarted=$(do_facet $facet \
4052 "$LCTL get_param -n ${paramp}.threads_started")
4053 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4062 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
4063 cleanup || error "cleanup failed with rc $?"
4065 run_test 53a "check OSS thread count params"
4069 local mds=$(do_facet $SINGLEMDS "$LCTL get_param \
4070 -N mds.*.*.threads_max 2>/dev/null")
4071 if [ -z "$mds" ]; then
4072 #running this on an old MDT
4073 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
4075 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
4077 cleanup || error "cleanup failed with $?"
4079 run_test 53b "check MDS thread count params"
4082 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4083 skip "Only applicable to ldiskfs-based MDTs"
4087 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p ||
4088 error "llverdev failed with rc=$?"
4091 run_test 54a "test llverdev and partial verify of device"
4094 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4095 skip "Only applicable to ldiskfs-based MDTs"
4100 run_llverfs $MOUNT -p || error "llverfs failed with rc=$?"
4101 cleanup || error "cleanup failed with rc=$?"
4103 run_test 54b "test llverfs and partial verify of filesystem"
4107 local max_ost_index=$1
4108 echo -n $(((max_ost_index + 1) * 8))
4112 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4113 skip "Only applicable to ldiskfs-based MDTs"
4117 local mdsdev=$(mdsdevname 1)
4118 local mdsvdev=$(mdsvdevname 1)
4122 if ! combined_mgs_mds; then
4123 stop_mgs || error "stopping MGS service failed"
4124 format_mgs || error "formatting MGT failed"
4126 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
4128 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
4129 --reformat $(ostdevname 1) $(ostvdevname 1)
4135 echo checking size of lov_objid for ost index $i
4136 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" |
4137 grep ^User | awk -F 'Size: ' '{print $2}')
4138 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
4139 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
4141 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
4148 run_test 55 "check lov_objid size"
4151 local server_version=$(lustre_version_code $SINGLEMDS)
4152 local mds_journal_size_orig=$MDSJOURNALSIZE
4157 for num in $(seq 1 $MDSCOUNT); do
4160 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
4161 $(ostdevname 1) $(ostvdevname 1)
4162 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
4163 $(ostdevname 2) $(ostvdevname 2)
4166 start_ost || error "Unable to start first ost (idx 10000)"
4167 start_ost2 || error "Unable to start second ost (idx 1000)"
4168 mount_client $MOUNT || error "Unable to mount client"
4172 if [[ $server_version -ge $(version_code 2.6.54) ]] ||
4173 [[ $server_version -ge $(version_code 2.5.4) &&
4174 $server_version -lt $(version_code 2.5.11) ]]; then
4175 wait_osc_import_state mds ost1 FULL
4176 wait_osc_import_state mds ost2 FULL
4177 $SETSTRIPE --stripe-count=-1 $DIR/$tfile ||
4178 error "Unable to setstripe $DIR/$tfile"
4179 n=$($LFS getstripe --stripe-count $DIR/$tfile)
4180 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
4185 MDSJOURNALSIZE=$mds_journal_size_orig
4188 run_test 56a "check big OST indexes and out-of-index-order start"
4193 umount_client $MOUNT -f || error "unmount client failed"
4202 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs" && return
4204 trap cleanup_56b EXIT RETURN ERR
4207 if ! combined_mgs_mds ; then
4212 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) --index=0 --reformat \
4213 $(mdsdevname 1) $(mdsvdevname 1)
4214 add mds2 $(mkfs_opts mds2 $(mdsdevname 2)) --index=1 --reformat \
4215 $(mdsdevname 2) $(mdsvdevname 2)
4216 add mds3 $(mkfs_opts mds3 $(mdsdevname 3)) --index=1000 --reformat \
4217 $(mdsdevname 3) $(mdsvdevname 3)
4221 start_mdt 1 || error "MDT 1 (idx 0) start failed"
4222 start_mdt 2 || error "MDT 2 (idx 1) start failed"
4223 start_mdt 3 || error "MDT 3 (idx 1000) start failed"
4224 start_ost || error "Unable to start first ost"
4225 start_ost2 || error "Unable to start second ost"
4227 do_nodes $(comma_list $(mdts_nodes)) \
4228 "$LCTL set_param mdt.*.enable_remote_dir=1 \
4229 mdt.*.enable_remote_dir_gid=-1"
4231 mount_client $MOUNT || error "Unable to mount client"
4233 $LFS mkdir -c3 $MOUNT/$tdir || error "failed to make testdir"
4235 echo "This is test file 1!" > $MOUNT/$tdir/$tfile.1 ||
4236 error "failed to make test file 1"
4237 echo "This is test file 2!" > $MOUNT/$tdir/$tfile.2 ||
4238 error "failed to make test file 2"
4239 echo "This is test file 1000!" > $MOUNT/$tdir/$tfile.1000 ||
4240 error "failed to make test file 1000"
4242 rm -rf $MOUNT/$tdir || error "failed to remove testdir"
4244 $LFS mkdir -i1000 $MOUNT/$tdir.1000 ||
4245 error "create remote dir at idx 1000 failed"
4248 echo "=== START lfs df OUTPUT ==="
4250 echo "==== END lfs df OUTPUT ===="
4252 mdtcnt=$(echo -e "$output" | grep $FSNAME-MDT | wc -l)
4253 ostcnt=$(echo -e "$output" | grep $FSNAME-OST | wc -l)
4255 echo "lfs df returned mdt count $mdtcnt and ost count $ostcnt"
4256 [ $mdtcnt -eq 3 ] || error "lfs df returned wrong mdt count"
4257 [ $ostcnt -eq 2 ] || error "lfs df returned wrong ost count"
4259 echo "This is test file 1!" > $MOUNT/$tdir.1000/$tfile.1 ||
4260 error "failed to make test file 1"
4261 echo "This is test file 2!" > $MOUNT/$tdir.1000/$tfile.2 ||
4262 error "failed to make test file 2"
4263 echo "This is test file 1000!" > $MOUNT/$tdir.1000/$tfile.1000 ||
4264 error "failed to make test file 1000"
4265 rm -rf $MOUNT/$tdir.1000 || error "failed to remove remote_dir"
4268 echo "=== START lfs mdts OUTPUT ==="
4270 echo "==== END lfs mdts OUTPUT ===="
4272 echo -e "$output" | grep -v "MDTS:" | awk '{print $1}' |
4273 sed 's/://g' > $TMP/mdts-actual.txt
4274 sort $TMP/mdts-actual.txt -o $TMP/mdts-actual.txt
4276 echo -e "0\n1\n1000" > $TMP/mdts-expected.txt
4278 diff $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4281 rm $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4283 [ $result -eq 0 ] || error "target_obd proc file is incorrect!"
4285 run_test 56b "test target_obd correctness with nonconsecutive MDTs"
4287 test_57a() { # bug 22656
4288 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4289 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4290 tail -1 | awk '{print $1}')
4291 writeconf_or_reformat
4292 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4293 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" ||
4294 error "tunefs failed"
4296 start_ost && error "OST registration from failnode should fail"
4299 run_test 57a "initial registration from failnode should fail (should return errs)"
4302 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4303 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4304 tail -1 | awk '{print $1}')
4305 writeconf_or_reformat
4306 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4307 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" ||
4308 error "tunefs failed"
4310 start_ost || error "OST registration from servicenode should not fail"
4313 run_test 57b "initial registration from servicenode should not fail"
4316 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
4319 test_58() { # bug 22658
4320 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
4322 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4323 createmany -o $DIR/$tdir/$tfile-%d 100
4324 # make sure that OSTs do not cancel llog cookies before we unmount the MDS
4325 #define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601
4326 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x601"
4327 unlinkmany $DIR/$tdir/$tfile-%d 100
4328 stop_mds || error "Unable to stop MDS"
4330 local MNTDIR=$(facet_mntpt $SINGLEMDS)
4331 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4333 # remove all files from the OBJECTS dir
4334 mount_fstype $SINGLEMDS
4336 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
4338 unmount_fstype $SINGLEMDS
4339 # restart MDS with missing llog files
4340 start_mds || error "unable to start MDS"
4341 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0"
4344 run_test 58 "missing llog files must not prevent MDT from mounting"
4347 start_mgsmds >> /dev/null
4348 local C1=$(count_osts)
4349 if [ $C1 -eq 0 ]; then
4350 start_ost >> /dev/null
4354 echo "original ost count: $C1 (expect > 0)"
4355 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
4356 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
4357 local C2=$(count_osts)
4358 echo "after mdt writeconf count: $C2 (expect 0)"
4359 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
4360 echo "OST start without writeconf should fail:"
4361 start_ost >> /dev/null &&
4362 error "OST start without writeconf didn't fail"
4363 echo "OST start with writeconf should succeed:"
4364 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
4365 local C3=$(count_osts)
4366 echo "after ost writeconf count: $C3 (expect 1)"
4367 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
4368 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
4369 local C4=$(count_osts)
4370 echo "after ost2 writeconf count: $C4 (expect 2)"
4371 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
4372 stop_ost2 >> /dev/null
4373 cleanup_nocli >> /dev/null
4374 #writeconf to remove all ost2 traces for subsequent tests
4375 writeconf_or_reformat
4377 run_test 59 "writeconf mount option"
4379 test_60() { # LU-471
4382 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4383 skip "Only applicable to ldiskfs-based MDTs"
4387 for num in $(seq $MDSCOUNT); do
4388 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
4389 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
4390 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4394 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
4395 [ ${PIPESTATUS[0]} -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
4397 # MDT default has dirdata feature
4398 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
4399 # we disable uninit_bg feature
4400 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
4401 # we set stride extended options
4402 echo $dump | grep stride > /dev/null || error "stride is not set"
4405 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
4410 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] ||
4411 { skip "Need MDS version at least 2.1.53"; return 0; }
4413 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
4414 ! large_xattr_enabled; then
4417 for num in $(seq $MDSCOUNT); do
4418 do_facet mds${num} $TUNE2FS -O large_xattr \
4419 $(mdsdevname $num) ||
4420 error "tune2fs on mds $num failed"
4424 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
4425 setup_noconfig || error "setting up the filesystem failed"
4426 client_up || error "starting client failed"
4428 local file=$DIR/$tfile
4429 touch $file || error "touch $file failed"
4431 local large_value="$(generate_string $(max_xattr_size))"
4432 local small_value="bar"
4434 local name="trusted.big"
4435 log "save large xattr $name on $file"
4436 setfattr -n $name -v $large_value $file ||
4437 error "saving $name on $file failed"
4439 local new_value=$(get_xattr_value $name $file)
4440 [[ "$new_value" != "$large_value" ]] &&
4441 error "$name different after saving"
4443 log "shrink value of $name on $file"
4444 setfattr -n $name -v $small_value $file ||
4445 error "shrinking value of $name on $file failed"
4447 new_value=$(get_xattr_value $name $file)
4448 [[ "$new_value" != "$small_value" ]] &&
4449 error "$name different after shrinking"
4451 log "grow value of $name on $file"
4452 setfattr -n $name -v $large_value $file ||
4453 error "growing value of $name on $file failed"
4455 new_value=$(get_xattr_value $name $file)
4456 [[ "$new_value" != "$large_value" ]] &&
4457 error "$name different after growing"
4459 log "check value of $name on $file after remounting MDS"
4461 new_value=$(get_xattr_value $name $file)
4462 [[ "$new_value" != "$large_value" ]] &&
4463 error "$name different after remounting MDS"
4465 log "remove large xattr $name from $file"
4466 setfattr -x $name $file || error "removing $name from $file failed"
4469 stopall || error "stopping for e2fsck run"
4470 for num in $(seq $MDSCOUNT); do
4471 run_e2fsck $(facet_active_host mds$num) \
4472 $(mdsdevname $num) "-y" ||
4473 error "e2fsck MDT$num failed"
4475 setup_noconfig || error "remounting the filesystem failed"
4478 # need to delete this file to avoid problems in other tests
4480 stopall || error "stopping systems to turn off large_xattr"
4482 for num in $(seq $MDSCOUNT); do
4483 do_facet mds${num} $TUNE2FS -O ^large_xattr \
4484 $(mdsdevname $num) ||
4485 error "tune2fs on mds $num failed"
4489 run_test 61 "large xattr"
4492 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4493 skip "Only applicable to ldiskfs-based MDTs"
4498 local mdsdev=$(mdsdevname 1)
4499 local ostdev=$(ostdevname 1)
4501 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] ||
4502 { skip "Need MDS version at least 2.2.51"; return 0; }
4504 echo "disable journal for mds"
4505 do_facet mds1 $TUNE2FS -O ^has_journal $mdsdev || error "tune2fs failed"
4506 start_mds && error "MDT start should fail"
4507 echo "disable journal for ost"
4508 do_facet ost1 $TUNE2FS -O ^has_journal $ostdev || error "tune2fs failed"
4509 start_ost && error "OST start should fail"
4510 cleanup || error "cleanup failed with rc $?"
4513 run_test 62 "start with disabled journal"
4516 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4517 skip "Only applicable to ldiskfs-based MDTs"
4521 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ldiskfs
4522 local inode_slab=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
4523 awk '/ldiskfs_inode_cache/ { print $5 / $6 }')
4524 if [ -z "$inode_slab" ]; then
4525 skip "ldiskfs module has not been loaded"
4529 echo "$inode_slab ldiskfs inodes per page"
4530 [ "${inode_slab%.*}" -ge "3" ] && return 0
4532 # If kmalloc-128 is also 1 per page - this is a debug kernel
4533 # and so this is not an error.
4534 local kmalloc128=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
4535 awk '/^(kmalloc|size)-128 / { print $5 / $6 }')
4536 # 32 128-byte chunks in 4k
4537 [ "${kmalloc128%.*}" -lt "32" ] ||
4538 error "ldiskfs inode too big, only $inode_slab objs/page, " \
4539 "kmalloc128 = $kmalloc128 objs/page"
4541 run_test 63 "Verify each page can at least hold 3 ldiskfs inodes"
4544 start_mds || error "unable to start MDS"
4545 start_ost || error "Unable to start OST1"
4546 start_ost2 || error "Unable to start second ost"
4547 mount_client $MOUNT || error "Unable to mount client"
4548 stop_ost2 || error "Unable to stop second ost"
4551 umount_client $MOUNT -f || error “unmount $MOUNT failed”
4552 cleanup_nocli || error "cleanup_nocli failed with $?"
4553 #writeconf to remove all ost2 traces for subsequent tests
4554 writeconf_or_reformat
4556 run_test 64 "check lfs df --lazy "
4558 test_65() { # LU-2237
4559 # Currently, the test is only valid for ldiskfs backend
4560 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4561 skip "non-ldiskfs backend" && return
4563 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4564 local brpt=$(facet_mntpt brpt)
4567 if ! do_facet $SINGLEMDS "test -b $devname"; then
4571 stop_mds || error "Unable to stop MDS"
4572 local obj=$(do_facet $SINGLEMDS \
4573 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
4575 if [ -z "$obj" ]; then
4576 # The MDT may be just re-formatted, mount the MDT for the
4577 # first time to guarantee the "last_rcvd" file is there.
4578 start_mds || error "fail to mount the MDS for the first time"
4579 stop_mds || error "Unable to stop MDS"
4582 # remove the "last_rcvd" file
4583 do_facet $SINGLEMDS "mkdir -p $brpt"
4584 do_facet $SINGLEMDS \
4585 "mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
4586 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
4587 do_facet $SINGLEMDS "$UMOUNT $brpt"
4589 # restart MDS, the "last_rcvd" file should be recreated.
4590 start_mds || error "fail to restart the MDS"
4591 stop_mds || error "Unable to stop MDS"
4592 obj=$(do_facet $SINGLEMDS \
4593 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
4594 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
4596 run_test 65 "re-create the lost last_rcvd file when server mount"
4599 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
4600 { skip "Need MGS version at least 2.3.59"; return 0; }
4603 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
4604 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
4606 echo "replace_nids should fail if MDS, OSTs and clients are UP"
4607 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4608 error "replace_nids fail"
4610 umount_client $MOUNT || error "unmounting client failed"
4611 echo "replace_nids should fail if MDS and OSTs are UP"
4612 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4613 error "replace_nids fail"
4615 stop_ost || error "Unable to stop OST1"
4616 echo "replace_nids should fail if MDS is UP"
4617 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4618 error "replace_nids fail"
4620 stop_mds || error "stopping mds failed"
4622 if combined_mgs_mds; then
4623 start_mdt 1 "-o nosvc" ||
4624 error "starting mds with nosvc option failed"
4627 echo "command should accept two parameters"
4628 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
4629 error "command should accept two params"
4631 echo "correct device name should be passed"
4632 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
4633 error "wrong devname"
4635 echo "wrong nids list should not destroy the system"
4636 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
4639 echo "replace OST nid"
4640 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
4641 error "replace nids failed"
4643 echo "command should accept two parameters"
4644 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
4645 error "command should accept two params"
4647 echo "wrong nids list should not destroy the system"
4648 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
4651 echo "replace MDS nid"
4652 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
4653 error "replace nids failed"
4655 if ! combined_mgs_mds ; then
4658 stop_mds || error "Unable to stop MDS"
4662 check_mount || error "error after nid replace"
4663 cleanup || error "cleanup failed"
4666 run_test 66 "replace nids"
4668 test_67() { #LU-2950
4669 local legacy="$TMP/legacy_lnet_config"
4670 local new="$TMP/new_routes_test"
4671 local out="$TMP/config_out_file"
4672 local verify="$TMP/conv_verify"
4673 local verify_conf="$TMP/conf_verify"
4675 # Create the legacy file that will be run through the
4676 # lustre_routes_conversion script
4677 cat <<- LEGACY_LNET_CONFIG > $legacy
4678 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
4679 tcp2 54 10.1.3.2@tcp;
4680 tcp3 10.3.4.3@tcp:3;
4684 # Create the verification file to verify the output of
4685 # lustre_routes_conversion script against.
4686 cat <<- VERIFY_LNET_CONFIG > $verify
4687 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
4688 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
4689 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
4690 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
4691 tcp4: { gateway: 10.3.3.4@tcp }
4694 # Create the verification file to verify the output of
4695 # lustre_routes_config script against
4696 cat <<- VERIFY_LNET_CONFIG > $verify_conf
4697 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
4698 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
4699 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
4700 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
4701 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
4704 lustre_routes_conversion $legacy $new > /dev/null
4705 if [ -f $new ]; then
4706 # verify the conversion output
4707 cmp -s $new $verify > /dev/null
4708 if [ $? -eq 1 ]; then
4709 error "routes conversion failed"
4712 lustre_routes_config --dry-run --verbose $new > $out
4713 # check that the script succeeded
4714 cmp -s $out $verify_conf > /dev/null
4715 if [ $? -eq 1 ]; then
4716 error "routes config failed"
4719 error "routes conversion test failed"
4721 # remove generated files
4722 rm -f $new $legacy $verify $verify_conf $out
4724 run_test 67 "test routes conversion and configuration"
4732 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.4.53) ] ||
4733 { skip "Need MDS version at least 2.4.53"; return 0; }
4735 umount_client $MOUNT || error "umount client failed"
4737 if ! combined_mgs_mds; then
4738 start_mgs || error "start mgs failed"
4741 start_mdt 1 || error "MDT start failed"
4742 start_ost || error "Unable to start OST1"
4744 # START-END - the sequences we'll be reserving
4745 START=$(do_facet $SINGLEMDS \
4746 $LCTL get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
4747 END=$((START + (1 << 30)))
4748 do_facet $SINGLEMDS \
4749 $LCTL set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
4751 # reset the sequences MDT0000 has already assigned
4752 do_facet $SINGLEMDS \
4753 $LCTL set_param seq.srv*MDT0000.space=clear
4755 # remount to let the client allocate new sequence
4756 mount_client $MOUNT || error "mount client failed"
4758 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4759 do_facet $SINGLEMDS \
4760 $LCTL get_param seq.srv*MDT0000.space
4761 $LFS path2fid $DIR/$tfile
4763 local old_ifs="$IFS"
4765 fid=($($LFS path2fid $DIR/$tfile))
4769 if [[ $seq < $END ]]; then
4770 error "used reserved sequence $seq?"
4772 cleanup || error "cleanup failed with $?"
4774 run_test 68 "be able to reserve specific sequences in FLDB"
4776 # Test 69: is about the total number of objects ever created on an OST.
4777 # so that when it is reformatted the normal MDS->OST orphan recovery won't
4778 # just "precreate" the missing objects. In the past it might try to recreate
4779 # millions of objects after an OST was reformatted
4781 local server_version=$(lustre_version_code $SINGLEMDS)
4783 [[ $server_version -lt $(version_code 2.4.2) ]] &&
4784 skip "Need MDS version at least 2.4.2" && return
4786 [[ $server_version -ge $(version_code 2.4.50) ]] &&
4787 [[ $server_version -lt $(version_code 2.5.0) ]] &&
4788 skip "Need MDS version at least 2.5.0" && return
4791 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4793 # use OST0000 since it probably has the most creations
4794 local OSTNAME=$(ostname_from_index 0)
4795 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
4796 local last_id=$(do_facet mds1 $LCTL get_param -n \
4797 osc.$mdtosc_proc1.prealloc_last_id)
4799 # Want to have OST LAST_ID over 5 * OST_MAX_PRECREATE to
4800 # verify that the LAST_ID recovery is working properly. If
4801 # not, then the OST will refuse to allow the MDS connect
4802 # because the LAST_ID value is too different from the MDS
4803 #define OST_MAX_PRECREATE=20000
4804 local ost_max_pre=20000
4805 local num_create=$(( ost_max_pre * 5 + 1 - last_id))
4807 # If the LAST_ID is already over 5 * OST_MAX_PRECREATE, we don't
4808 # need to create any files. So, skip this section.
4809 if [ $num_create -gt 0 ]; then
4810 # Check the number of inodes available on OST0
4812 local ifree=$($LFS df -i $MOUNT | awk '/OST0000/ { print $4 }')
4813 log "On OST0, $ifree inodes available. Want $num_create."
4815 $SETSTRIPE -i 0 $DIR/$tdir ||
4816 error "$SETSTRIPE -i 0 $DIR/$tdir failed"
4817 if [ $ifree -lt 10000 ]; then
4818 files=$(( ifree - 50 ))
4823 local j=$((num_create / files + 1))
4824 for i in $(seq 1 $j); do
4825 createmany -o $DIR/$tdir/$tfile-$i- $files ||
4826 error "createmany fail create $files files: $?"
4827 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
4828 error "unlinkmany failed unlink $files files"
4832 # delete all of the files with objects on OST0 so the
4833 # filesystem is not inconsistent later on
4834 $LFS find $MOUNT --ost 0 -print0 | xargs -0 rm
4836 umount_client $MOUNT || error "umount client failed"
4837 stop_ost || error "OST0 stop failure"
4838 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
4839 $(ostdevname 1) $(ostvdevname 1) ||
4840 error "reformat and replace $ostdev failed"
4841 start_ost || error "OST0 restart failure"
4842 wait_osc_import_state mds ost FULL
4844 mount_client $MOUNT || error "mount client failed"
4845 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
4846 local idx=$($GETSTRIPE -i $DIR/$tdir/$tfile-last)
4847 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
4849 local iused=$($LFS df -i $MOUNT | awk '/OST0000/ { print $3 }')
4850 log "On OST0, $iused used inodes"
4851 [ $iused -ge $((ost_max_pre/2 + 1000)) ] &&
4852 error "OST replacement created too many inodes; $iused"
4853 cleanup || error "cleanup failed with $?"
4855 run_test 69 "replace an OST with the same index"
4858 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4861 cleanup || error "cleanup failed with $?"
4863 start_mdt 1 || error "MDT0 start fail"
4865 start_ost || error "OST0 start fail"
4866 for num in $(seq 2 $MDSCOUNT); do
4867 start_mdt $num || return
4870 mount_client $MOUNT || error "mount client fails"
4872 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
4874 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4875 error "create remote dir fail"
4877 rm -rf $DIR/$tdir || error "delete dir fail"
4878 cleanup || error "cleanup failed with $?"
4880 run_test 70a "start MDT0, then OST, then MDT1"
4883 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4886 start_ost || error "OST0 start fail"
4888 start_mds || error "MDS start fail"
4890 mount_client $MOUNT || error "mount client fails"
4892 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
4894 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4895 error "create remote dir fail"
4897 rm -rf $DIR/$tdir || error "delete dir fail"
4899 cleanup || error "cleanup failed with $?"
4901 run_test 70b "start OST, MDT1, MDT0"
4904 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4907 start_mds || error "MDS start fail"
4908 start_ost || error "OST0 start fail"
4910 mount_client $MOUNT || error "mount client fails"
4911 stop_mdt 1 || error "MDT1 start fail"
4913 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
4914 echo "deactivate $mdc_for_mdt1"
4915 $LCTL --device $mdc_for_mdt1 deactivate ||
4916 error "set $mdc_for_mdt1 deactivate failed"
4918 mkdir $DIR/$tdir && error "mkdir succeed"
4920 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4921 error "create remote dir succeed"
4923 cleanup || error "cleanup failed with $?"
4925 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
4928 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4931 start_mds || error "MDS start fail"
4932 start_ost || error "OST0 start fail"
4934 mount_client $MOUNT || error "mount client fails"
4936 stop_mdt 2 || error "MDT1 start fail"
4938 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
4940 echo "deactivate $mdc_for_mdt2"
4941 $LCTL --device $mdc_for_mdt2 deactivate ||
4942 error "set $mdc_for_mdt2 deactivate failed"
4944 mkdir $DIR/$tdir || error "mkdir fail"
4945 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4946 error "create remote dir succeed"
4948 rm -rf $DIR/$tdir || error "delete dir fail"
4950 cleanup || error "cleanup failed with $?"
4952 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
4955 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4957 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.62) ] ||
4958 { skip "Need MDS version at least 2.7.62"; return 0; }
4960 cleanup || error "cleanup failed with $?"
4962 local mdsdev=$(mdsdevname 1)
4963 local ostdev=$(ostdevname 1)
4964 local mdsvdev=$(mdsvdevname 1)
4965 local ostvdev=$(ostvdevname 1)
4966 local opts_mds="$(mkfs_opts mds1 $mdsdev) --reformat $mdsdev $mdsvdev"
4967 local opts_ost="$(mkfs_opts ost1 $ostdev) --reformat $ostdev $ostvdev"
4969 add mds1 $opts_mds || error "add mds1 failed"
4970 start_mdt 1 || error "start mdt1 failed"
4971 add ost1 $opts_ost || error "add ost1 failed"
4972 start_ost || error "start ost failed"
4973 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
4975 local soc=$(do_facet mds1 "$LCTL get_param -n \
4976 mdt.*MDT0000.sync_lock_cancel")
4977 [ $soc == "never" ] || error "SoC enabled on single MDS"
4979 for i in $(seq 2 $MDSCOUNT); do
4980 mdsdev=$(mdsdevname $i)
4981 mdsvdev=$(mdsvdevname $i)
4982 opts_mds="$(mkfs_opts mds$i $mdsdev) --reformat $mdsdev \
4984 add mds$i $opts_mds || error "add mds$i failed"
4985 start_mdt $i || error "start mdt$i fail"
4988 wait_dne_interconnect
4990 for i in $(seq $MDSCOUNT); do
4991 soc=$(do_facet mds$i "$LCTL get_param -n \
4992 mdt.*MDT000$((i - 1)).sync_lock_cancel")
4993 [ $soc == "blocking" ] || error "SoC not enabled on DNE"
4996 for i in $(seq 2 $MDSCOUNT); do
4997 stop_mdt $i || error "stop mdt$i fail"
4999 soc=$(do_facet mds1 "$LCTL get_param -n \
5000 mdt.*MDT0000.sync_lock_cancel")
5001 [ $soc == "never" ] || error "SoC enabled on single MDS"
5003 cleanup || error "cleanup failed with $?"
5005 run_test 70e "Sync-on-Cancel will be enabled by default on DNE"
5008 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5009 if combined_mgs_mds; then
5010 skip "needs separate MGS/MDT" && return
5014 start_mdt 1 || error "MDT0 start fail"
5015 start_ost || error "OST0 start fail"
5016 for num in $(seq 2 $MDSCOUNT); do
5017 start_mdt $num || return
5020 start_ost2 || error "OST1 start fail"
5022 mount_client $MOUNT || error "mount client fails"
5024 mkdir $DIR/$tdir || error "mkdir fail"
5025 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5026 error "create remote dir succeed"
5028 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5029 rm -rf $DIR/$tdir || error "delete dir fail"
5031 umount_client $MOUNT || error "umount_client failed"
5032 stop_mds || error "MDS stop fail"
5033 stop_ost || error "OST0 stop fail"
5034 stop_ost2 || error "OST1 stop fail"
5036 run_test 71a "start MDT0 OST0, MDT1, OST1"
5039 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5040 if combined_mgs_mds; then
5041 skip "needs separate MGS/MDT" && return
5045 for num in $(seq 2 $MDSCOUNT); do
5046 start_mdt $num || return
5048 start_ost || error "OST0 start fail"
5049 start_mdt 1 || error "MDT0 start fail"
5050 start_ost2 || error "OST1 start fail"
5052 mount_client $MOUNT || error "mount client fails"
5054 mkdir $DIR/$tdir || error "mkdir fail"
5055 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5056 error "create remote dir succeed"
5058 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5059 rm -rf $DIR/$tdir || error "delete dir fail"
5061 umount_client $MOUNT || error "umount_client failed"
5062 stop_mds || error "MDT0 stop fail"
5063 stop_ost || error "OST0 stop fail"
5064 stop_ost2 || error "OST1 stop fail"
5066 run_test 71b "start MDT1, OST0, MDT0, OST1"
5069 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5070 if combined_mgs_mds; then
5071 skip "needs separate MGS/MDT" && return
5075 start_ost || error "OST0 start fail"
5076 start_ost2 || error "OST1 start fail"
5077 for num in $(seq 2 $MDSCOUNT); do
5078 start_mdt $num || return
5080 start_mdt 1 || error "MDT0 start fail"
5082 mount_client $MOUNT || error "mount client fails"
5084 mkdir $DIR/$tdir || error "mkdir fail"
5085 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5086 error "create remote dir succeed"
5088 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5089 rm -rf $DIR/$tdir || error "delete dir fail"
5091 umount_client $MOUNT || error "umount_client failed"
5092 stop_mds || error "MDS stop fail"
5093 stop_ost || error "OST0 stop fail"
5094 stop_ost2 || error "OST1 stop fail"
5097 run_test 71c "start OST0, OST1, MDT1, MDT0"
5100 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5101 if combined_mgs_mds; then
5102 skip "needs separate MGS/MDT" && return
5106 start_ost || error "OST0 start fail"
5107 for num in $(seq 2 $MDSCOUNT); do
5108 start_mdt $num || return
5110 start_mdt 1 || error "MDT0 start fail"
5111 start_ost2 || error "OST1 start fail"
5113 mount_client $MOUNT || error "mount client fails"
5115 mkdir $DIR/$tdir || error "mkdir fail"
5116 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5117 error "create remote dir succeed"
5119 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5120 rm -rf $DIR/$tdir || error "delete dir fail"
5122 umount_client $MOUNT || error "umount_client failed"
5123 stop_mds || error "MDS stop fail"
5124 stop_ost || error "OST0 stop fail"
5125 stop_ost2 || error "OST1 stop fail"
5128 run_test 71d "start OST0, MDT1, MDT0, OST1"
5131 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5132 if combined_mgs_mds; then
5133 skip "needs separate MGS/MDT" && return
5137 start_ost || error "OST0 start fail"
5138 for num in $(seq 2 $MDSCOUNT); do
5139 start_mdt $num || return
5141 start_ost2 || error "OST1 start fail"
5142 start_mdt 1 || error "MDT0 start fail"
5144 mount_client $MOUNT || error "mount client fails"
5146 mkdir $DIR/$tdir || error "mkdir fail"
5147 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5148 error "create remote dir succeed"
5150 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5151 rm -rf $DIR/$tdir || error "delete dir fail"
5153 umount_client $MOUNT || error "umount_client failed"
5154 stop_mds || error "MDS stop fail"
5155 stop_ost || error "OST0 stop fail"
5156 stop_ost2 || error "OST1 stop fail"
5159 run_test 71e "start OST0, MDT1, OST1, MDT0"
5161 test_72() { #LU-2634
5162 local mdsdev=$(mdsdevname 1)
5163 local ostdev=$(ostdevname 1)
5164 local cmd="$E2FSCK -fnvd $mdsdev"
5168 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
5169 skip "ldiskfs only test" && return
5171 if combined_mgs_mds; then
5172 add_options='--reformat'
5174 add_options='--reformat --replace'
5177 #tune MDT with "-O extents"
5179 for num in $(seq $MDSCOUNT); do
5180 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
5181 $add_options $(mdsdevname $num) $(mdsvdevname $num) ||
5182 error "add mds $num failed"
5183 do_facet mds${num} "$TUNE2FS -O extents $(mdsdevname $num)" ||
5184 error "$TUNE2FS failed on mds${num}"
5187 add ost1 $(mkfs_opts ost1 $ostdev) $add_options $ostdev ||
5188 error "add $ostdev failed"
5189 start_mds || error "start mds failed"
5190 start_ost || error "start ost failed"
5191 mount_client $MOUNT || error "mount client failed"
5193 #create some short symlinks
5194 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5195 createmany -o $DIR/$tdir/$tfile-%d $fn
5196 echo "create $fn short symlinks"
5197 for i in $(seq -w 1 $fn); do
5198 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
5203 umount_client $MOUNT || error "umount client failed"
5204 stop_mds || error "stop mds failed"
5205 stop_ost || error "stop ost failed"
5208 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
5210 run_test 72 "test fast symlink with extents flag enabled"
5212 test_73() { #LU-3006
5213 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
5214 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
5215 error "1st tunefs failed"
5216 start_mgsmds || error "start mds failed"
5217 start_ost || error "start ost failed"
5218 mount_client $MOUNT || error "mount client failed"
5219 $LCTL get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
5220 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
5221 umount_client $MOUNT || error "umount client failed"
5225 run_test 73 "failnode to update from mountdata properly"
5227 test_75() { # LU-2374
5228 [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] &&
5229 skip "Need MDS version at least 2.4.1" && return
5232 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
5233 --reformat $(mdsdevname 1) $(mdsvdevname 1)"
5234 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
5235 --reformat $(ostdevname 1) $(ostvdevname 1)"
5237 #check with default parameters
5238 add mds1 $opts_mds || error "add mds1 failed for default params"
5239 add ost1 $opts_ost || error "add ost1 failed for default params"
5241 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
5242 opts_mds=$(echo $opts_mds |
5243 sed -e "s/--index=$index/--index=$index --mdt/")
5244 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
5245 opts_ost=$(echo $opts_ost |
5246 sed -e "s/--index=$index/--index=$index --ost/")
5248 add mds1 $opts_mds || error "add mds1 failed for new params"
5249 add ost1 $opts_ost || error "add ost1 failed for new params"
5250 if ! combined_mgs_mds; then
5251 stop_mgs || error "stop mgs failed"
5256 run_test 75 "The order of --index should be irrelevant"
5259 [[ $(lustre_version_code mgs) -ge $(version_code 2.4.52) ]] ||
5260 { skip "Need MDS version at least 2.4.52" && return 0; }
5262 if ! combined_mgs_mds; then
5263 start_mgs || error "start mgs failed"
5266 local MDMB_PARAM="osc.*.max_dirty_mb"
5267 echo "Change MGS params"
5268 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
5270 echo "max_dirty_mb: $MAX_DIRTY_MB"
5271 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB + MAX_DIRTY_MB))
5272 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
5273 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
5274 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5275 head -1" $NEW_MAX_DIRTY_MB
5276 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5277 echo "$MAX_DIRTY_MB"
5278 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5279 error "error while apply max_dirty_mb"
5281 echo "Check the value is stored after remount"
5284 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5285 head -1" $NEW_MAX_DIRTY_MB
5286 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5287 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5288 error "max_dirty_mb is not saved after remount"
5290 echo "Change OST params"
5291 CLIENT_PARAM="obdfilter.*.client_cache_count"
5292 local CLIENT_CACHE_COUNT
5293 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5295 echo "client_cache_count: $CLIENT_CACHE_COUNT"
5296 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
5297 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
5298 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
5299 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5300 head -1" $NEW_CLIENT_CACHE_COUNT
5301 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5303 echo "$CLIENT_CACHE_COUNT"
5304 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5305 error "error while apply client_cache_count"
5307 echo "Check the value is stored after remount"
5310 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5311 head -1" $NEW_CLIENT_CACHE_COUNT
5312 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5314 echo "$CLIENT_CACHE_COUNT"
5315 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5316 error "client_cache_count is not saved after remount"
5319 run_test 76a "set permanent params set_param -P"
5321 test_76b() { # LU-4783
5322 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.57) ]] ||
5323 { skip "Need MGS version at least 2.5.57" && return 0; }
5326 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
5327 error "start params log failed"
5330 run_test 76b "verify params log setup correctly"
5333 [[ $(lustre_version_code mgs) -ge $(version_code 2.8.54) ]] ||
5334 { skip "Need MDS version at least 2.4.52" && return 0; }
5336 local MASK_PARAM="mdd.*.changelog_mask"
5337 echo "Change changelog_mask"
5338 do_facet mgs $LCTL set_param -P $MASK_PARAM=-CLOSE ||
5339 error "Can't change changlog_mask"
5340 wait_update $(facet_host mds) "$LCTL get_param -n $MASK_PARAM |
5343 echo "Check the value is stored after mds remount"
5344 stop_mds || error "Failed to stop MDS"
5345 start_mds || error "Failed to start MDS"
5346 local CHANGELOG_MASK=$(do_facet mgs $LCTL get_param -n $MASK_PARAM)
5347 echo $CHANGELOG_MASK | grep CLOSE > /dev/null &&
5348 error "changelog_mask is not changed"
5352 run_test 76c "verify changelog_mask is applied with set_param -P"
5354 test_77() { # LU-3445
5355 local server_version=$(lustre_version_code $SINGLEMDS)
5356 [[ $server_version -ge $(version_code 2.8.55) ]] ||
5357 { skip "Need MDS version 2.8.55+ "; return; }
5359 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
5360 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
5361 skip_env "mixed loopback and real device not working" && return
5364 local fs2mdsdev=$(mdsdevname 1_2)
5365 local fs2ostdev=$(ostdevname 1_2)
5366 local fs2mdsvdev=$(mdsvdevname 1_2)
5367 local fs2ostvdev=$(ostvdevname 1_2)
5368 local fsname=test1234
5370 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
5372 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
5373 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
5374 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
5375 error "start fs2mds failed"
5377 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
5378 mgsnid="$mgsnid,$mgsnid:$mgsnid"
5380 add fs2ost $(mkfs_opts ost1 $fs2ostdev) --mgsnode=$mgsnid \
5381 --failnode=$failnid --fsname=$fsname \
5382 --reformat $fs2ostdev $fs2ostvdev ||
5383 error "add fs2ost failed"
5384 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
5386 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
5387 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
5388 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
5391 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
5394 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ||
5395 $(facet_fstype ost1) != ldiskfs ]] &&
5396 skip "only applicable to ldiskfs-based MDTs and OSTs" && return
5398 # reformat the Lustre filesystem with a smaller size
5399 local saved_MDSCOUNT=$MDSCOUNT
5400 local saved_MDSSIZE=$MDSSIZE
5401 local saved_OSTCOUNT=$OSTCOUNT
5402 local saved_OSTSIZE=$OSTSIZE
5405 MDSSIZE=$((MDSSIZE - 20000))
5406 OSTSIZE=$((OSTSIZE - 20000))
5407 reformat || error "(1) reformat Lustre filesystem failed"
5408 MDSSIZE=$saved_MDSSIZE
5409 OSTSIZE=$saved_OSTSIZE
5411 # mount the Lustre filesystem
5412 setup_noconfig || error "(2) setup Lustre filesystem failed"
5415 log "create test files"
5420 mkdir $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
5422 for i in $(seq $num_files); do
5423 file=$MOUNT/$tdir/$tfile-$i
5424 dd if=/dev/urandom of=$file count=1 bs=1M || {
5425 $LCTL get_param osc.*.cur*grant*
5426 $LFS df; $LFS df -i;
5427 # stop creating files if there is no more space
5428 if [ ! -e $file ]; then
5429 num_files=$((i - 1))
5433 $LFS getstripe -v $file
5434 local ost_idx=$(LFS getstripe -i $file)
5435 do_facet ost$((ost_idx + 1)) \
5436 $LCTL get_param obdfilter.*.*grant*
5437 error "(4) create $file failed"
5441 # unmount the Lustre filesystem
5442 cleanup || error "(5) cleanup Lustre filesystem failed"
5444 # run e2fsck on the MDT and OST devices
5445 local mds_host=$(facet_active_host $SINGLEMDS)
5446 local ost_host=$(facet_active_host ost1)
5447 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
5448 local ost_dev=$(ostdevname 1)
5450 run_e2fsck $mds_host $mds_dev "-y"
5451 run_e2fsck $ost_host $ost_dev "-y"
5453 # get the original block count of the MDT and OST filesystems
5454 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
5455 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
5457 # expand the MDT and OST filesystems to the device size
5458 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
5459 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
5461 # run e2fsck on the MDT and OST devices again
5462 run_e2fsck $mds_host $mds_dev "-y"
5463 run_e2fsck $ost_host $ost_dev "-y"
5465 # mount the Lustre filesystem
5469 log "check files after expanding the MDT and OST filesystems"
5470 for i in $(seq $num_files); do
5471 file=$MOUNT/$tdir/$tfile-$i
5472 $CHECKSTAT -t file -s 1048576 $file ||
5473 error "(6) checkstat $file failed"
5477 log "create more files after expanding the MDT and OST filesystems"
5478 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
5479 file=$MOUNT/$tdir/$tfile-$i
5480 dd if=/dev/urandom of=$file count=1 bs=1M ||
5481 error "(7) create $file failed"
5484 # unmount the Lustre filesystem
5485 cleanup || error "(8) cleanup Lustre filesystem failed"
5487 # run e2fsck on the MDT and OST devices
5488 run_e2fsck $mds_host $mds_dev "-y"
5489 run_e2fsck $ost_host $ost_dev "-y"
5491 # get the maximum block count of the MDT and OST filesystems
5492 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
5493 local ost_max_blks=$(get_block_count ost1 $ost_dev)
5495 # get the minimum block count of the MDT and OST filesystems
5496 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
5497 grep minimum | sed -e 's/^.*filesystem: //g')
5498 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
5499 grep minimum | sed -e 's/^.*filesystem: //g')
5501 # shrink the MDT and OST filesystems to a smaller size
5505 if [[ $mds_max_blks -gt $mds_min_blks &&
5506 $mds_max_blks -gt $mds_orig_blks ]]; then
5507 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
5508 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
5509 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
5510 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
5511 error "shrink $SINGLEMDS to $new_blks failed"
5515 if [[ $ost_max_blks -gt $ost_min_blks &&
5516 $ost_max_blks -gt $ost_orig_blks ]]; then
5517 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
5518 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
5519 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
5520 run_resize2fs ost1 $ost_dev $new_blks ||
5521 error "shrink ost1 to $new_blks failed"
5525 # check whether the MDT or OST filesystem was shrunk or not
5527 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
5528 reformat || error "(10) reformat Lustre filesystem failed"
5532 # run e2fsck on the MDT and OST devices again
5533 run_e2fsck $mds_host $mds_dev "-y"
5534 run_e2fsck $ost_host $ost_dev "-y"
5536 # mount the Lustre filesystem again
5540 log "check files after shrinking the MDT and OST filesystems"
5541 for i in $(seq $((num_files + 10))); do
5542 file=$MOUNT/$tdir/$tfile-$i
5543 $CHECKSTAT -t file -s 1048576 $file ||
5544 error "(11) checkstat $file failed"
5547 # unmount and reformat the Lustre filesystem
5548 cleanup || error "(12) cleanup Lustre filesystem failed"
5549 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
5551 MDSCOUNT=$saved_MDSCOUNT
5552 OSTCOUNT=$saved_OSTCOUNT
5553 reformat || error "(14) reformat Lustre filesystem failed"
5555 run_test 78 "run resize2fs on MDT and OST filesystems"
5557 test_79() { # LU-4227
5558 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.5.59) ]] ||
5559 { skip "Need MDS version at least 2.5.59"; return 0; }
5561 local mdsdev1=$(mdsdevname 1)
5562 local mdsvdev1=$(mdsvdevname 1)
5563 local mdsdev2=$(mdsdevname 2)
5564 local mdsvdev2=$(mdsvdevname 2)
5565 local ostdev1=$(ostdevname 1)
5566 local ostvdev1=$(ostvdevname 1)
5567 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
5568 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
5569 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
5572 # remove --mgs/--mgsnode from mkfs.lustre options
5573 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
5575 mgsnode_opt=$(echo $opts_mds2 |
5576 awk '{ for ( i = 1; i < NF; i++ )
5577 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5578 [ -n $mgsnode_opt ] &&
5579 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
5581 mgsnode_opt=$(echo $opts_ost1 |
5582 awk '{ for ( i = 1; i < NF; i++ )
5583 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5584 [ -n $mgsnode_opt ] &&
5585 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
5587 # -MGS, format a mdt without --mgs option
5588 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
5589 error "Must specify --mgs when formatting mdt combined with mgs"
5591 # +MGS, format a mdt/ost without --mgsnode option
5592 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
5593 > /dev/null || error "start mds1 failed"
5594 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
5595 error "Must specify --mgsnode when formatting a mdt"
5596 add ost1 $opts_ost1 $ostdev1 $ostvdev2 &&
5597 error "Must specify --mgsnode when formatting an ost"
5601 run_test 79 "format MDT/OST without mgs option (should return errors)"
5604 start_mds || error "Failed to start MDT"
5605 start_ost || error "Failed to start OST1"
5606 uuid=$(do_facet ost1 $LCTL get_param -n mgc.*.uuid)
5607 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
5608 do_facet ost1 "$LCTL set_param fail_val=10 fail_loc=0x906"
5609 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x906"
5610 do_facet mgs "$LCTL set_param -n mgs/MGS/evict_client $uuid"
5612 start_ost2 || error "Failed to start OST2"
5614 do_facet ost1 "$LCTL set_param fail_loc=0"
5617 run_test 80 "mgc import reconnect race"
5619 #Save the original values of $OSTCOUNT and $OSTINDEX$i.
5621 local new_ostcount=$1
5622 saved_ostcount=$OSTCOUNT
5623 OSTCOUNT=$new_ostcount
5627 for i in $(seq $OSTCOUNT); do
5629 eval saved_ostindex$i=${!index}
5634 # Restore the original values of $OSTCOUNT and $OSTINDEX$i.
5635 restore_ostindex() {
5640 for i in $(seq $OSTCOUNT); do
5641 index=saved_ostindex$i
5642 eval OSTINDEX$i=${!index}
5644 OSTCOUNT=$saved_ostcount
5649 # The main purpose of this test is to ensure the OST_INDEX_LIST functions as
5650 # expected. This test uses OST_INDEX_LIST to format OSTs with a randomly
5651 # assigned index and ensures we can mount such a formatted file system
5652 test_81() { # LU-4665
5653 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5654 { skip "Need MDS version at least 2.6.54" && return; }
5655 [[ $OSTCOUNT -ge 3 ]] || { skip_env "Need at least 3 OSTs" && return; }
5659 # Each time RANDOM is referenced, a random integer between 0 and 32767
5662 local saved_ostindex1=$OSTINDEX1
5663 for i in 65535 $((RANDOM + 65536)); do
5664 echo -e "\nFormat ost1 with --index=$i, should fail"
5666 if add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat \
5667 $(ostdevname 1) $(ostvdevname 1); then
5668 OSTINDEX1=$saved_ostindex1
5669 error "format ost1 with --index=$i should fail"
5672 OSTINDEX1=$saved_ostindex1
5676 # Format OSTs with random sparse indices.
5677 trap "restore_ostindex" EXIT
5678 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices"
5679 OST_INDEX_LIST=[0,$((RANDOM * 2 % 65533 + 1)),65534] formatall
5681 # Setup and check Lustre filesystem.
5682 start_mgsmds || error "start_mgsmds failed"
5683 for i in $(seq $OSTCOUNT); do
5684 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5685 error "start ost$i failed"
5688 mount_client $MOUNT || error "mount client $MOUNT failed"
5689 check_mount || error "check client $MOUNT failed"
5692 local max_easize=$($LCTL get_param -n llite.*.max_easize)
5693 [[ $max_easize -eq 128 ]] ||
5694 error "max_easize is $max_easize, should be 128 bytes"
5698 run_test 81 "sparse OST indexing"
5700 # Here we exercise the stripe placement functionality on a file system that
5701 # has formatted the OST with a random index. With the file system the following
5702 # functionality is tested:
5704 # 1. Creating a new file with a specific stripe layout.
5706 # 2. Modifiy a existing empty file with a specific stripe layout.
5708 # 3. Ensure we fail to set the stripe layout of a file that already has one.
5710 # 4. If ost-index is defined we need to ensure it is the first entry in the
5711 # ost index list returned by lfs getstripe.
5713 # 5. Lastly ensure this functionality fails with directories.
5714 test_82a() { # LU-4665
5715 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5716 { skip "Need MDS version at least 2.6.54" && return; }
5717 [[ $OSTCOUNT -ge 3 ]] || { skip_env "Need at least 3 OSTs" && return; }
5723 # Format OSTs with random sparse indices.
5727 local LOV_V1_INSANE_STRIPE_COUNT=65532
5728 for i in $(seq $OSTCOUNT); do
5729 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
5730 ost_indices+=" $index"
5732 ost_indices=$(comma_list $ost_indices)
5734 trap "restore_ostindex" EXIT
5735 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5736 OST_INDEX_LIST=[$ost_indices] formatall
5738 # Setup Lustre filesystem.
5739 start_mgsmds || error "start_mgsmds failed"
5740 for i in $(seq $OSTCOUNT); do
5741 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5742 error "start ost$i failed"
5745 mount_client $MOUNT || error "mount client $MOUNT failed"
5748 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5749 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5751 # 1. If the file does not exist, new file will be created
5752 # with specified OSTs.
5753 local file=$DIR/$tdir/$tfile-1
5754 local cmd="$SETSTRIPE -o $ost_indices $file"
5756 eval $cmd || error "$cmd failed"
5757 check_stripe_count $file $OSTCOUNT
5758 check_obdidx $file $ost_indices
5759 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5760 error "write $file failed"
5762 # 2. If the file already exists and is an empty file, the file
5763 # will be attached with specified layout.
5764 file=$DIR/$tdir/$tfile-2
5765 mcreate $file || error "mcreate $file failed"
5766 cmd="$SETSTRIPE -o $ost_indices $file"
5768 eval $cmd || error "$cmd failed"
5769 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5770 error "write $file failed"
5771 check_stripe_count $file $OSTCOUNT
5772 check_obdidx $file $ost_indices
5774 # 3. If the file already has a valid layout attached, the command
5775 # should fail with EBUSY.
5777 eval $cmd && error "stripe is already set on $file, $cmd should fail"
5779 # 4. If [--stripe-index|-i <start_ost_idx>] is used, the index must
5780 # be in the OST indices list.
5781 local start_ost_idx=${ost_indices##*,}
5782 file=$DIR/$tdir/$tfile-3
5783 cmd="$SETSTRIPE -o $ost_indices -i $start_ost_idx $file"
5785 eval $cmd || error "$cmd failed"
5786 check_stripe_count $file $OSTCOUNT
5787 check_obdidx $file $ost_indices
5788 check_start_ost_idx $file $start_ost_idx
5790 file=$DIR/$tdir/$tfile-4
5792 cmd+=" -o $(exclude_items_from_list $ost_indices $start_ost_idx)"
5793 cmd+=" -i $start_ost_idx $file"
5795 eval $cmd && error "index $start_ost_idx should be in $ost_indices"
5797 # 5. Specifying OST indices for directory should fail with ENOSUPP.
5798 local dir=$DIR/$tdir/$tdir
5799 mkdir $dir || error "mkdir $dir failed"
5800 cmd="$SETSTRIPE -o $ost_indices $dir"
5802 eval $cmd && error "$cmd should fail, specifying OST indices" \
5803 "for directory is not supported"
5807 run_test 82a "specify OSTs for file (succeed) or directory (fail)"
5812 # Remove OSTs from a pool and destroy the pool.
5813 destroy_pool $ost_pool || true
5818 # Test 82b is run to ensure that if the user supplies a pool with a specific
5819 # stripe layout that it behaves proprerly. It should fail in the case that
5820 # the supplied OST index list points to OSTs not contained in the user
5822 test_82b() { # LU-4665
5823 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5824 { skip "Need MDS version at least 2.6.54" && return; }
5825 [[ $OSTCOUNT -ge 4 ]] || { skip_env "Need at least 4 OSTs" && return; }
5831 # Format OSTs with random sparse indices.
5835 local LOV_V1_INSANE_STRIPE_COUNT=65532
5836 for i in $(seq $OSTCOUNT); do
5837 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
5838 ost_indices+=" $index"
5840 ost_indices=$(comma_list $ost_indices)
5842 trap "restore_ostindex" EXIT
5843 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5844 OST_INDEX_LIST=[$ost_indices] formatall
5846 # Setup Lustre filesystem.
5847 start_mgsmds || error "start_mgsmds failed"
5848 for i in $(seq $OSTCOUNT); do
5849 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5850 error "start ost$i failed"
5853 mount_client $MOUNT || error "mount client $MOUNT failed"
5855 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5856 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5858 # Create a new pool and add OSTs into it.
5859 local ost_pool=$FSNAME.$TESTNAME
5860 create_pool $ost_pool || error "create OST pool $ost_pool failed"
5863 trap "cleanup_82b" EXIT
5865 local ost_idx_in_list=${ost_indices##*,}
5866 local ost_idx_in_pool=$(exclude_items_from_list $ost_indices \
5869 local ost_targets="$FSNAME-OST["
5870 for i in ${ost_idx_in_pool//,/ }; do
5871 ost_targets=$ost_targets$(printf "%04x," $i)
5873 ost_targets="${ost_targets%,}]"
5875 local ost_targets_uuid=$(for i in ${ost_idx_in_pool//,/ }; \
5876 do printf "$FSNAME-OST%04x_UUID\n" $i; done |
5877 sort -u | tr '\n' ' ')
5879 local cmd="$LCTL pool_add $ost_pool $ost_targets"
5880 do_facet mgs $cmd || error "$cmd failed"
5881 wait_update $HOSTNAME "$LCTL get_param -n lov.$FSNAME-*.pools.$TESTNAME|
5882 sort -u | tr '\n' ' ' " "$ost_targets_uuid" ||
5883 error "wait_update $ost_pool failed"
5884 wait_update_facet $SINGLEMDS "$LCTL pool_list $ost_pool | wc -l" 4 ||
5885 error "wait_update pool_list $ost_pool failed"
5887 # If [--pool|-p <pool_name>] is set with [--ost-list|-o <ost_indices>],
5888 # then the OSTs must be the members of the pool.
5889 local file=$DIR/$tdir/$tfile
5890 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
5892 eval $cmd && error "OST with index $ost_idx_in_list should be" \
5893 "in OST pool $ost_pool"
5895 # Only select OST $ost_idx_in_list from $ost_pool for file.
5896 ost_idx_in_list=${ost_idx_in_pool#*,}
5897 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
5899 eval $cmd || error "$cmd failed"
5900 cmd="$GETSTRIPE $file"
5902 eval $cmd || error "$cmd failed"
5903 check_stripe_count $file 2
5904 check_obdidx $file $ost_idx_in_list
5905 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5906 error "write $file failed"
5910 run_test 82b "specify OSTs for file with --pool and --ost-list options"
5913 [[ $(lustre_version_code ost1) -ge $(version_code 2.6.91) ]] ||
5914 { skip "Need OST version at least 2.6.91" && return 0; }
5915 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
5916 skip "Only applicable to ldiskfs-based MDTs"
5926 ostmnt=$(facet_mntpt ost1)
5927 fstype=$(facet_fstype ost1)
5929 # Mount the OST as an ldiskfs filesystem.
5930 log "mount the OST $dev as a $fstype filesystem"
5931 add ost1 $(mkfs_opts ost1 $dev) $FSTYPE_OPT \
5932 --reformat $dev $dev > /dev/null ||
5933 error "format ost1 error"
5935 if ! test -b $dev; then
5936 mnt_opts=$(csa_add "$OST_MOUNT_OPTS" -o loop)
5938 echo "mnt_opts $mnt_opts"
5939 do_facet ost1 mount -t $fstype $dev \
5941 # Run llverfs on the mounted ldiskfs filesystem.
5942 # It is needed to get ENOSPACE.
5943 log "run llverfs in partial mode on the OST $fstype $ostmnt"
5944 do_rpc_nodes $(facet_host ost1) run_llverfs $ostmnt -vpl \
5945 "no" || error "run_llverfs error on $fstype"
5948 log "unmount the OST $dev"
5951 # Delete file IO_scrub. Later osd_scrub_setup will try to
5952 # create "IO_scrub" but will get ENOSPACE.
5954 echo "start ost1 service on `facet_active_host ost1`"
5955 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
5958 err=$(do_facet ost1 dmesg | grep "VFS: Busy inodes after unmount of")
5959 echo "string err $err"
5960 [ -z "$err" ] || error $err
5963 run_test 83 "ENOSPACE on OST doesn't cause message VFS: \
5964 Busy inodes after unmount ..."
5967 local facet=$SINGLEMDS
5968 local num=$(echo $facet | tr -d "mds")
5969 local dev=$(mdsdevname $num)
5970 local time_min=$(recovery_time_min)
5971 local recovery_duration
5972 local completed_clients
5973 local correct_clients
5976 echo "start mds service on $(facet_active_host $facet)"
5978 "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
5979 error "start MDS failed"
5981 start_ost || error "start OST0000 failed"
5982 start_ost2 || error "start OST0001 failed"
5984 echo "recovery_time=$time_min, timeout=$TIMEOUT, wrap_up=$wrap_up"
5986 mount_client $MOUNT1 || error "mount $MOUNT1 failed"
5987 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
5988 # make sure new superblock labels are sync'd before disabling writes
5992 replay_barrier $SINGLEMDS
5993 createmany -o $DIR1/$tfile-%d 1000
5995 # We need to catch the end of recovery window to extend it.
5996 # Skip 5 requests and add delay to request handling.
5997 #define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP
5998 do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5"
6000 facet_failover --fsck $SINGLEMDS || error "failover: $?"
6003 echo "recovery status"
6004 do_facet $SINGLEMDS \
6005 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status"
6007 recovery_duration=$(do_facet $SINGLEMDS \
6008 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
6009 awk '/recovery_duration/ { print $2 }')
6010 (( $recovery_duration > $time_min + $wrap_up )) &&
6011 error "recovery_duration > recovery_time_hard + wrap up"
6012 completed_clients=$(do_facet $SINGLEMDS \
6013 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
6014 awk '/completed_clients/ { print $2 }')
6016 correct_clients="$MDSCOUNT/$((MDSCOUNT+1))"
6017 [ "$completed_clients" = "${correct_clients}" ] ||
6018 error "$completed_clients != $correct_clients"
6020 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
6021 umount_client $MOUNT1
6022 umount_client $MOUNT2
6028 run_test 84 "check recovery_hard_time"
6031 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.55) ]] ||
6032 { skip "Need OST version at least 2.7.55" && return 0; }
6033 ##define OBD_FAIL_OSD_OST_EA_FID_SET 0x197
6034 do_facet ost1 "lctl set_param fail_loc=0x197"
6038 run_test 85 "osd_ost init: fail ea_fid_set"
6043 # ost1 has already registered to the MGS before the reformat.
6044 # So after reformatting it with option "-G", it could not be
6045 # mounted to the MGS. Cleanup the system for subsequent tests.
6050 local server_version=$(lustre_version_code $SINGLEMDS)
6051 [ "$(facet_fstype ost1)" = "zfs" ] &&
6052 skip "LU-6442: no such mkfs params for ZFS OSTs" && return
6053 [[ $server_version -ge $(version_code 2.7.56) ]] ||
6054 { skip "Need server version newer than 2.7.55"; return 0; }
6056 local OST_OPTS="$(mkfs_opts ost1 $(ostdevname 1)) \
6057 --reformat $(ostdevname 1) $(ostvdevname 1)"
6060 local OLDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6061 awk '/Flex block group size: / { print $NF; exit; }')
6064 if [[ ${!opts} != *mkfsoptions* ]]; then
6065 eval opts=\"${!opts} \
6066 --mkfsoptions='\\\"-O flex_bg -G $NEWSIZE\\\"'\"
6068 val=${!opts//--mkfsoptions=\\\"/ \
6069 --mkfsoptions=\\\"-O flex_bg -G $NEWSIZE }
6073 echo "params: $opts"
6075 trap cleanup_86 EXIT ERR
6078 add ost1 $opts || error "add ost1 failed with new params"
6080 local FOUNDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6081 awk '/Flex block group size: / { print $NF; exit; }')
6083 [[ $FOUNDSIZE == $NEWSIZE ]] ||
6084 error "Flex block group size: $FOUNDSIZE, expected: $NEWSIZE"
6088 run_test 86 "Replacing mkfs.lustre -G option"
6090 test_87() { #LU-6544
6091 [[ $(lustre_version_code $SINGLEMDS1) -ge $(version_code 2.9.51) ]] ||
6092 { skip "Need MDS version at least 2.9.51" && return; }
6093 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ]] &&
6094 { skip "Only applicable to ldiskfs-based MDTs" && return; }
6095 [[ $OSTCOUNT -gt 59 ]] &&
6096 { skip "Ignore wide striping situation" && return; }
6098 local mdsdev=$(mdsdevname 1)
6099 local mdsvdev=$(mdsvdevname 1)
6100 local file=$DIR/$tfile
6101 local mntpt=$(facet_mntpt $SINGLEMDS)
6102 local used_xattr_blk=0
6103 local inode_size=${1:-1024}
6105 local xtest="trusted.test"
6109 local stripe_cnt=$(($OSTCOUNT + 2))
6111 #Please see ldiskfs_make_lustre() for MDT inode size calculation
6112 if [ $stripe_cnt -gt 16 ]; then
6115 left_size=$(expr $inode_size - \
6118 32 - 40 \* 3 - 32 \* 3 - $stripe_cnt \* 24 - 16 - 3 - \
6120 24 - 18 - $(expr length $tfile) - 16 - 4)
6121 if [ $left_size -le 0 ]; then
6122 echo "No space($left_size) is expected in inode."
6123 echo "Try 1-byte xattr instead to verify this."
6126 echo "Estimate: at most $left_size-byte space left in inode."
6132 add mds1 $(mkfs_opts mds1 ${mdsdev}) --stripe-count-hint=$stripe_cnt \
6133 --reformat $mdsdev $mdsvdev || error "add mds1 failed"
6134 start_mdt 1 > /dev/null || error "start mdt1 failed"
6135 for i in $(seq $OSTCOUNT); do
6136 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS > /dev/null ||
6137 error "start ost$i failed"
6139 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
6140 check_mount || error "check client $MOUNT failed"
6143 $SETSTRIPE -E 1M -c 1 -E 64M -c 1 -E -1 -c -1 $file ||
6144 error "Create file with 3 components failed"
6145 i=$($GETSTRIPE -I 3 -c $file)
6146 if [ $i -ne $OSTCOUNT ]; then
6147 left_size=$(expr $left_size + $(expr $OSTCOUNT - $i) \* 24)
6148 echo -n "Since only $i out $OSTCOUNT OSTs are used, "
6149 echo -n "the expected left space is changed to "
6150 echo "$left_size bytes at most."
6152 value=$(generate_string $left_size)
6153 setfattr -n $xtest -v $value $file
6154 orig=$(get_xattr_value $xtest $file)
6155 [[ "$orig" != "$value" ]] && error "$xtest changed"
6157 #Verify if inode has some expected space left
6158 umount $MOUNT > /dev/null || error "umount $MOUNT failed"
6159 stop_mdt 1 > /dev/null || error "stop mdt1 failed"
6160 mount_ldiskfs $SINGLEMDS || error "mount -t ldiskfs $SINGLEMDS failed"
6162 do_facet $SINGLEMDS ls -sal $mntpt/ROOT/$tfile
6163 used_xattr_blk=$(do_facet $SINGLEMDS ls -s $mntpt/ROOT/$tfile |
6165 [[ $used_xattr_blk -eq 0 ]] &&
6166 error "Please check MDS inode size calculation: \
6167 more than $left_size-byte space left in inode."
6168 echo "Verified: at most $left_size-byte space left in inode."
6172 run_test 87 "check if MDT inode can hold EAs with N stripes properly"
6175 [ "$(facet_fstype mds1)" == "zfs" ] &&
6176 skip "LU-6662: no implementation for ZFS" && return
6180 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
6181 --reformat $(mdsdevname 1) || error "add mds1 failed"
6183 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
6184 grep -e \".*opts:.*errors=remount-ro.*\"" ||
6185 error "default mount options is missing"
6187 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
6188 --mountfsoptions="user_xattr,errors=panic" \
6189 --reformat $(mdsdevname 1) || error "add mds1 failed"
6191 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
6192 grep -e \".*opts:.*errors=panic.*\"" ||
6193 error "user can't override default mount options"
6195 run_test 88 "check the default mount options can be overridden"
6197 test_89() { # LU-7131
6198 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.9.54) ]] ||
6199 { skip "Need MDT version at least 2.9.54" && return 0; }
6201 local key=failover.node
6202 local val1=192.0.2.254@tcp0 # Reserved IPs, see RFC 5735
6203 local val2=192.0.2.255@tcp0
6204 local mdsdev=$(mdsdevname 1)
6209 [ $(facet_fstype mds1) == zfs ] && import_zpool mds1
6210 # Check that parameters are added correctly
6211 echo "tunefs --param $key=$val1"
6212 do_facet mds "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
6213 error "tunefs --param $key=$val1 failed"
6214 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6215 error "tunefs --dryrun failed"
6216 params=${params##*Parameters:}
6217 params=${params%%exiting*}
6218 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val1) = "1" ] ||
6219 error "on-disk parameter not added correctly via tunefs"
6221 # Check that parameters replace existing instances when added
6222 echo "tunefs --param $key=$val2"
6223 do_facet mds "$TUNEFS --param $key=$val2 $mdsdev >/dev/null" ||
6224 error "tunefs --param $key=$val2 failed"
6225 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6226 error "tunefs --dryrun failed"
6227 params=${params##*Parameters:}
6228 params=${params%%exiting*}
6229 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "1" ] ||
6230 error "on-disk parameter not replaced via tunefs"
6231 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val2) = "1" ] ||
6232 error "on-disk parameter not replaced correctly via tunefs"
6234 # Check that a parameter is erased properly
6235 echo "tunefs --erase-param $key"
6236 do_facet mds "$TUNEFS --erase-param $key $mdsdev >/dev/null" ||
6237 error "tunefs --erase-param $key failed"
6238 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6239 error "tunefs --dryrun failed"
6240 params=${params##*Parameters:}
6241 params=${params%%exiting*}
6242 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "0" ] ||
6243 error "on-disk parameter not erased correctly via tunefs"
6245 # Check that all the parameters are erased
6246 echo "tunefs --erase-params"
6247 do_facet mds "$TUNEFS --erase-params $mdsdev >/dev/null" ||
6248 error "tunefs --erase-params failed"
6249 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6250 error "tunefs --dryrun failed"
6251 params=${params##*Parameters:}
6252 params=${params%%exiting*}
6254 error "all on-disk parameters not erased correctly via tunefs"
6256 # Check the order of options --erase-params and --param
6257 echo "tunefs --param $key=$val1 --erase-params"
6259 "$TUNEFS --param $key=$val1 --erase-params $mdsdev >/dev/null"||
6260 error "tunefs --param $key=$val1 --erase-params failed"
6261 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6262 error "tunefs --dryrun failed"
6263 params=${params##*Parameters:}
6264 params=${params%%exiting*}
6265 [ $(echo $params | tr ' ' '\n') == "$key=$val1" ] ||
6266 error "on-disk param not added correctly with --erase-params"
6270 run_test 89 "check tunefs --param and --erase-param{s} options"
6273 # $2 (optional) value of max_mod_rpcs_in_flight to set
6274 check_max_mod_rpcs_in_flight() {
6282 idx=$(printf "%04x" $($LFS getdirstripe -i $dir))
6283 facet="mds$((0x$idx + 1))"
6285 if [ -z "$mmr" ]; then
6286 # get value of max_mod_rcps_in_flight
6287 mmr=$($LCTL get_param -n \
6288 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) ||
6289 error "Unable to get max_mod_rpcs_in_flight"
6290 echo "max_mod_rcps_in_flight is $mmr"
6292 # set value of max_mod_rpcs_in_flight
6294 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight=$mmr ||
6295 error "Unable to set max_mod_rpcs_in_flight to $mmr"
6296 echo "max_mod_rpcs_in_flight set to $mmr"
6299 # create mmr+1 files
6300 echo "creating $((mmr + 1)) files ..."
6302 for i in $(seq $((mmr + 1))); do
6308 # consumes mmr-1 modify RPC slots
6309 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6310 # drop requests on MDT so that RPC slots are consumed
6311 # during all the request resend interval
6312 do_facet $facet "$LCTL set_param fail_loc=0x159"
6313 echo "launch $((mmr - 1)) chmod in parallel ..."
6314 for i in $(seq $((mmr - 1))); do
6315 chmod 0600 $dir/file-$i &
6319 # send one additional modify RPC
6320 do_facet $facet "$LCTL set_param fail_loc=0"
6321 echo "launch 1 additional chmod in parallel ..."
6322 chmod 0600 $dir/file-$mmr &
6325 # check this additional modify RPC get a modify RPC slot
6326 # and succeed its operation
6327 checkstat -vp 0600 $dir/file-$mmr ||
6328 error "Unable to send $mmr modify RPCs in parallel"
6333 # consumes mmr modify RPC slots
6334 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6335 # drop requests on MDT so that RPC slots are consumed
6336 # during all the request resend interval
6337 do_facet $facet "$LCTL set_param fail_loc=0x159"
6338 echo "launch $mmr chmod in parallel ..."
6339 for i in $(seq $mmr); do
6340 chmod 0666 $dir/file-$i &
6344 # send one additional modify RPC
6345 do_facet $facet "$LCTL set_param fail_loc=0"
6346 echo "launch 1 additional chmod in parallel ..."
6347 chmod 0666 $dir/file-$((mmr + 1)) &
6350 # check this additional modify RPC blocked getting a modify RPC slot
6351 checkstat -vp 0644 $dir/file-$((mmr + 1)) ||
6352 error "Unexpectedly send $(($mmr + 1)) modify RPCs in parallel"
6358 if ! combined_mgs_mds ; then
6363 [[ $($LCTL get_param mdc.*.import |
6364 grep "connect_flags:.*multi_mod_rpc") ]] ||
6365 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6367 # check default value
6368 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6369 check_max_mod_rpcs_in_flight $DIR/$tdir
6373 run_test 90a "check max_mod_rpcs_in_flight is enforced"
6383 [[ $($LCTL get_param mdc.*.import |
6384 grep "connect_flags:.*multi_mod_rpc") ]] ||
6385 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6388 # update max_mod_rpcs_in_flight
6389 $LFS mkdir -c1 $DIR/${tdir}1 || error "mkdir $DIR/${tdir}1 failed"
6390 check_max_mod_rpcs_in_flight $DIR/${tdir}1 1
6393 # check client is able to send multiple modify RPCs in paralell
6394 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
6395 grep -c "multi_mod_rpcs")
6396 if [ "$tmp" -ne $MDSCOUNT ]; then
6397 echo "Client not able to send multiple modify RPCs in parallel"
6402 # update max_mod_rpcs_in_flight
6403 $LFS mkdir -c1 $DIR/${tdir}2 || error "mkdir $DIR/${tdir}2 failed"
6404 check_max_mod_rpcs_in_flight $DIR/${tdir}2 5
6407 $LFS mkdir -c1 $DIR/${tdir}3 || error "mkdir $DIR/${tdir}3 failed"
6408 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/${tdir}3))
6409 facet="mds$((0x$idx + 1))"
6411 # save MDT max_mod_rpcs_per_client
6412 mmrpc=$(do_facet $facet \
6413 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
6415 # update max_mod_rpcs_in_flight
6416 umount_client $MOUNT
6418 "echo 16 > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6420 $LCTL set_param mdc.$FSNAME-MDT$idx-mdc-*.max_rpcs_in_flight=17
6421 check_max_mod_rpcs_in_flight $DIR/${tdir}3 16
6423 # restore MDT max_mod_rpcs_per_client initial value
6425 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6427 rm -rf $DIR/${tdir}?
6430 run_test 90b "check max_mod_rpcs_in_flight is enforced after update"
6439 [[ $($LCTL get_param mdc.*.import |
6440 grep "connect_flags:.*multi_mod_rpc") ]] ||
6441 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6443 # check client is able to send multiple modify RPCs in paralell
6444 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
6445 grep -c "multi_mod_rpcs")
6446 if [ "$tmp" -ne $MDSCOUNT ]; then
6447 skip "Client not able to send multiple modify RPCs in parallel"
6452 # get max_rpcs_in_flight value
6453 mrif=$($LCTL get_param -n mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
6454 echo "max_rpcs_in_flight is $mrif"
6456 # get MDT max_mod_rpcs_per_client
6457 mmrpc=$(do_facet mds1 \
6458 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
6459 echo "max_mod_rpcs_per_client is $mmrpc"
6462 # attempt to set max_mod_rpcs_in_flight to max_rpcs_in_flight value
6463 # prerequisite: set max_mod_rpcs_per_client to max_rpcs_in_flight value
6464 umount_client $MOUNT
6466 "echo $mrif > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6470 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif &&
6471 error "set max_mod_rpcs_in_flight to $mrif should fail"
6473 umount_client $MOUNT
6475 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6479 # attempt to set max_mod_rpcs_in_flight to max_mod_rpcs_per_client+1
6480 # prerequisite: set max_rpcs_in_flight to max_mod_rpcs_per_client+2
6482 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$((mmrpc + 2))
6485 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$((mmrpc + 1)) &&
6486 error "set max_mod_rpcs_in_flight to $((mmrpc + 1)) should fail"
6490 run_test 90c "check max_mod_rpcs_in_flight update limits"
6501 [[ $($LCTL get_param mdc.*.import |
6502 grep "connect_flags:.*multi_mod_rpc") ]] ||
6503 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6505 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6506 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
6507 facet="mds$((0x$idx + 1))"
6509 # check client version supports multislots
6510 tmp=$($LCTL get_param -N \
6511 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
6512 if [ -z "$tmp" ]; then
6513 skip "Client does not support multiple modify RPCs in flight"
6518 # get current value of max_mod_rcps_in_flight
6519 mmr=$($LCTL get_param -n \
6520 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
6521 echo "max_mod_rcps_in_flight is $mmr"
6524 echo "creating $mmr files ..."
6526 for i in $(seq $mmr); do
6527 touch $DIR/$tdir/file-$i
6530 # prepare for close RPC
6531 multiop_bg_pause $DIR/$tdir/file-close O_c
6534 # consumes mmr modify RPC slots
6535 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6536 # drop requests on MDT so that RPC slots are consumed
6537 # during all the request resend interval
6538 do_facet $facet "$LCTL set_param fail_loc=0x159"
6539 echo "launch $mmr chmod in parallel ..."
6540 for i in $(seq $mmr); do
6541 chmod 0600 $DIR/$tdir/file-$i &
6544 # send one additional close RPC
6545 do_facet $facet "$LCTL set_param fail_loc=0"
6546 echo "launch 1 additional close in parallel ..."
6548 cancel_lru_locks mdc
6551 # check this additional close RPC get a modify RPC slot
6552 # and multiop process completed
6553 [ -d /proc/$pid ] &&
6554 error "Unable to send the additional close RPC in parallel"
6559 run_test 90d "check one close RPC is allowed above max_mod_rpcs_in_flight"
6561 check_uuid_on_ost() {
6563 do_facet ost1 "$LCTL get_param obdfilter.${FSNAME}*.exports.'$nid'.uuid"
6566 check_uuid_on_mdt() {
6568 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
6576 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.63) ]] ||
6577 { skip "Need OST version at least 2.7.63" && return 0; }
6578 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.63) ]] ||
6579 { skip "Need MDT version at least 2.7.63" && return 0; }
6581 start_mds || error "MDS start failed"
6582 start_ost || error "unable to start OST"
6583 mount_client $MOUNT || error "client start failed"
6584 check_mount || error "check_mount failed"
6587 nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g")
6591 uuid=$(get_client_uuid $MOUNT)
6593 echo "list nids on mdt:"
6594 do_facet $SINGLEMDS "$LCTL list_param mdt.${FSNAME}*.exports.*"
6595 echo "uuid from $nid:"
6596 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
6598 found=$(check_uuid_on_mdt $nid | grep $uuid)
6599 [ -z "$found" ] && error "can't find $uuid $nid on MDT"
6600 found=$(check_uuid_on_ost $nid | grep $uuid)
6601 [ -z "$found" ] && error "can't find $uuid $nid on OST"
6603 # umount the client so it won't reconnect
6604 manual_umount_client --force || error "failed to umount $?"
6605 # shouldn't disappear on MDS after forced umount
6606 found=$(check_uuid_on_mdt $nid | grep $uuid)
6607 [ -z "$found" ] && error "can't find $uuid $nid"
6610 do_facet $SINGLEMDS \
6611 "$LCTL set_param -n mdt.${mds1_svc}.evict_client nid:$nid"
6613 found=$(check_uuid_on_mdt $nid | grep $uuid)
6614 [ -n "$found" ] && error "found $uuid $nid on MDT"
6615 found=$(check_uuid_on_ost $nid | grep $uuid)
6616 [ -n "$found" ] && error "found $uuid $nid on OST"
6618 # check it didn't reconnect (being umounted)
6619 sleep $((TIMEOUT+1))
6620 found=$(check_uuid_on_mdt $nid | grep $uuid)
6621 [ -n "$found" ] && error "found $uuid $nid on MDT"
6622 found=$(check_uuid_on_ost $nid | grep $uuid)
6623 [ -n "$found" ] && error "found $uuid $nid on OST"
6627 run_test 91 "evict-by-nid support"
6629 generate_ldev_conf() {
6630 # generate an ldev.conf file
6631 local ldevconfpath=$1
6633 local fsldevformat=""
6636 fstype=$(facet_fstype mgs)
6637 if [ "$fstype" == "zfs" ]; then
6638 fsldevformat="$fstype:"
6643 printf "%s\t-\t%s-MGS0000\t%s%s\n" \
6647 $(mgsdevname) > $ldevconfpath
6649 local mdsfo_host=$mdsfailover_HOST;
6650 if [ -z "$mdsfo_host" ]; then
6654 for num in $(seq $MDSCOUNT); do
6655 fstype=$(facet_fstype mds$num)
6656 if [ "$fstype" == "zfs" ]; then
6657 fsldevformat="$fstype:"
6662 printf "%s\t%s\t%s-MDT%04d\t%s%s\n" \
6668 $(mdsdevname $num) >> $ldevconfpath
6671 local ostfo_host=$ostfailover_HOST;
6672 if [ -z "$ostfo_host" ]; then
6676 for num in $(seq $OSTCOUNT); do
6677 fstype=$(facet_fstype ost$num)
6678 if [ "$fstype" == "zfs" ]; then
6679 fsldevformat="$fstype:"
6684 printf "%s\t%s\t%s-OST%04d\t%s%s\n" \
6690 $(ostdevname $num) >> $ldevconfpath
6693 echo "----- $ldevconfpath -----"
6695 echo "--- END $ldevconfpath ---"
6700 # generate a nids file (mapping between hostname to nid)
6701 # looks like we only have the MGS nid available to us
6702 # so just echo that to a file
6704 echo -e "${mgs_HOST}\t${MGSNID}" > $nidspath
6706 echo "----- $nidspath -----"
6708 echo "--- END $nidspath ---"
6711 compare_ldev_output() {
6715 sort $expected_output -o $expected_output
6716 sort $ldev_output -o $ldev_output
6718 echo "-- START OF LDEV OUTPUT --"
6720 echo "--- END OF LDEV OUTPUT ---"
6722 echo "-- START OF EXPECTED OUTPUT --"
6723 cat $expected_output
6724 echo "--- END OF EXPECTED OUTPUT ---"
6726 diff $expected_output $ldev_output
6731 if [ -z "$LDEV" ]; then
6732 error "ldev is missing!"
6735 local LDEVCONFPATH=$TMP/ldev.conf
6736 local NIDSPATH=$TMP/nids
6738 echo "Host is $(hostname)"
6740 generate_ldev_conf $LDEVCONFPATH
6741 generate_nids $NIDSPATH
6743 # echo the mgs nid and compare it to environment variable MGSNID
6744 # also, ldev.conf and nids is a server side thing, use the OSS
6747 output=$($LDEV -c $LDEVCONFPATH -H $ost_HOST -n $NIDSPATH echo %m)
6749 echo "-- START OF LDEV OUTPUT --"
6751 echo "--- END OF LDEV OUTPUT ---"
6753 # ldev failed, error
6754 if [ $? -ne 0 ]; then
6755 rm $LDEVCONFPATH $NIDSPATH
6756 error "ldev failed to execute!"
6759 # need to process multiple lines because of combined MGS and MDS
6760 echo -e $output | awk '{ print $2 }' | while read -r line ; do
6761 if [ "$line" != "$MGSNID" ]; then
6762 rm $LDEVCONFPATH $NIDSPATH
6763 error "ldev failed mgs nid '$line', expected '$MGSNID'"
6767 rm $LDEVCONFPATH $NIDSPATH
6769 run_test 92 "ldev returns MGS NID correctly in command substitution"
6772 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs" && return
6775 #start mgs or mgs/mdt0
6776 if ! combined_mgs_mds ; then
6783 start_ost || error "OST0 start fail"
6785 #define OBD_FAIL_MGS_WRITE_TARGET_DELAY 0x90e
6786 do_facet mgs "$LCTL set_param fail_val = 10 fail_loc=0x8000090e"
6787 for num in $(seq 2 $MDSCOUNT); do
6791 mount_client $MOUNT || error "mount client fails"
6792 wait_osc_import_state mds ost FULL
6793 wait_osc_import_state client ost FULL
6794 check_mount || error "check_mount failed"
6796 cleanup || error "cleanup failed with $?"
6798 run_test 93 "register mulitple MDT at the same time"
6801 if [ -z "$LDEV" ]; then
6802 error "ldev is missing!"
6805 local LDEVCONFPATH=$TMP/ldev.conf
6806 local NIDSPATH=$TMP/nids
6808 generate_ldev_conf $LDEVCONFPATH
6809 generate_nids $NIDSPATH
6811 local LDEV_OUTPUT=$TMP/ldev-output.txt
6812 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME > $LDEV_OUTPUT
6814 # ldev failed, error
6815 if [ $? -ne 0 ]; then
6816 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
6817 error "ldev failed to execute!"
6821 local EXPECTED_OUTPUT=$TMP/ldev-expected.txt
6823 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
6825 for num in $(seq $MDSCOUNT); do
6826 printf "%s-MDT%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
6829 for num in $(seq $OSTCOUNT); do
6830 printf "%s-OST%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
6833 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6835 if [ $? -ne 0 ]; then
6836 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6837 error "ldev failed to produce the correct hostlist!"
6840 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6842 run_test 94 "ldev outputs correct labels for file system name query"
6845 if [ -z "$LDEV" ]; then
6846 error "ldev is missing!"
6849 local LDEVCONFPATH=$TMP/ldev.conf
6850 local NIDSPATH=$TMP/nids
6852 generate_ldev_conf $LDEVCONFPATH
6853 generate_nids $NIDSPATH
6857 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME &>/dev/null
6858 if [ $? -ne 0 ]; then
6859 rm $LDEVCONFPATH $NIDSPATH
6860 error "ldev label filtering w/ -F failed!"
6864 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -l &>/dev/null
6865 if [ $? -ne 0 ]; then
6866 rm $LDEVCONFPATH $NIDSPATH
6867 error "ldev label filtering w/ -l failed!"
6871 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f &>/dev/null
6872 if [ $? -ne 0 ]; then
6873 rm $LDEVCONFPATH $NIDSPATH
6874 error "ldev label filtering w/ -f failed!"
6878 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a &>/dev/null
6879 if [ $? -ne 0 ]; then
6880 rm $LDEVCONFPATH $NIDSPATH
6881 error "ldev label filtering w/ -a failed!"
6886 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -F $FSNAME &>/dev/null
6887 if [ $? -eq 0 ]; then
6888 rm $LDEVCONFPATH $NIDSPATH
6889 error "ldev label filtering w/ -a and -F incorrectly succeeded"
6893 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -f &>/dev/null
6894 if [ $? -eq 0 ]; then
6895 rm $LDEVCONFPATH $NIDSPATH
6896 error "ldev label filtering w/ -a and -f incorrectly succeeded"
6900 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -l &>/dev/null
6901 if [ $? -eq 0 ]; then
6902 rm $LDEVCONFPATH $NIDSPATH
6903 error "ldev label filtering w/ -a and -l incorrectly succeeded"
6907 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f -l &>/dev/null
6908 if [ $? -eq 0 ]; then
6909 rm $LDEVCONFPATH $NIDSPATH
6910 error "ldev label filtering w/ -f and -l incorrectly succeeded"
6914 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -l &>/dev/null
6915 if [ $? -eq 0 ]; then
6916 rm $LDEVCONFPATH $NIDSPATH
6917 error "ldev label filtering w/ -F and -l incorrectly succeeded"
6920 # file sys & foreign
6921 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -f &>/dev/null
6922 if [ $? -eq 0 ]; then
6923 rm $LDEVCONFPATH $NIDSPATH
6924 error "ldev label filtering w/ -F and -f incorrectly succeeded"
6927 rm $LDEVCONFPATH $NIDSPATH
6929 run_test 95 "ldev should only allow one label filter"
6932 if [ -z "$LDEV" ]; then
6933 error "ldev is missing!"
6936 local LDEVCONFPATH=$TMP/ldev.conf
6937 local NIDSPATH=$TMP/nids
6939 generate_ldev_conf $LDEVCONFPATH
6940 generate_nids $NIDSPATH
6942 local LDEV_OUTPUT=$TMP/ldev-output.txt
6943 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -H $mgs_HOST \
6945 awk '{print $2}' > $LDEV_OUTPUT
6947 # ldev failed, error
6948 if [ $? -ne 0 ]; then
6949 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
6950 error "ldev failed to execute!"
6954 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
6956 echo "$mgs_HOST-$(facet_fstype mgs)" > $EXPECTED_OUTPUT
6958 if [ "$mgs_HOST" == "$mds_HOST" ]; then
6959 for num in $(seq $MDSCOUNT); do
6960 echo "$mds_HOST-$(facet_fstype mds$num)" \
6965 if [ "$mgs_HOST" == "$ost_HOST" ]; then
6966 for num in $(seq $OSTCOUNT); do
6967 echo "$ost_HOST-$(facet_fstype ost$num)" \
6972 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6974 if [ $? -ne 0 ]; then
6975 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6976 error "ldev failed to produce the correct output!"
6979 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6981 run_test 96 "ldev returns hostname and backend fs correctly in command sub"
6984 if [ -z "$LDEV" ]; then
6985 error "ldev is missing!"
6988 local LDEVCONFPATH=$TMP/ldev.conf
6989 local NIDSPATH=$TMP/nids
6991 generate_ldev_conf $LDEVCONFPATH
6992 generate_nids $NIDSPATH
6994 local LDEV_OUTPUT=$TMP/ldev-output.txt
6995 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
6997 echo -e "\nMDT role"
6998 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mdt > $LDEV_OUTPUT
7000 if [ $? -ne 0 ]; then
7001 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7002 error "ldev failed to execute for mdt role!"
7005 for num in $(seq $MDSCOUNT); do
7006 printf "%s-MDT%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
7009 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7011 if [ $? -ne 0 ]; then
7012 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7013 error "ldev failed to produce the correct output for mdt role!"
7016 echo -e "\nOST role"
7017 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R ost > $LDEV_OUTPUT
7019 if [ $? -ne 0 ]; then
7020 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
7021 error "ldev failed to execute for ost role!"
7025 for num in $(seq $OSTCOUNT); do
7026 printf "%s-OST%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
7029 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7031 if [ $? -ne 0 ]; then
7032 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7033 error "ldev failed to produce the correct output for ost role!"
7036 echo -e "\nMGS role"
7037 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mgs > $LDEV_OUTPUT
7039 if [ $? -ne 0 ]; then
7040 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
7041 error "ldev failed to execute for mgs role!"
7044 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
7046 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7048 if [ $? -ne 0 ]; then
7049 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7050 error "ldev failed to produce the correct output for mgs role!"
7053 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7055 run_test 97 "ldev returns correct ouput when querying based on role"
7060 local temp=$MDS_MOUNT_OPTS
7063 check_mount || error "mount failed"
7064 mountopt="user_xattr"
7065 for ((x = 1; x <= 400; x++)); do
7066 mountopt="$mountopt,user_xattr"
7068 remount_client $mountopt $MOUNT 2>&1 | grep "too long" ||
7069 error "Buffer overflow check failed"
7070 cleanup || error "cleanup failed"
7072 run_test 98 "Buffer-overflow check while parsing mount_opts"
7076 [[ $(facet_fstype ost1) != ldiskfs ]] &&
7077 { skip "Only applicable to ldiskfs-based OSTs" && return; }
7078 [[ $(lustre_version_code ost1) -ge $(version_code 2.8.57) ]] ||
7079 { skip "Need OST version at least 2.8.57" && return 0; }
7081 local ost_opts="$(mkfs_opts ost1 $(ostdevname 1)) \
7082 --reformat $(ostdevname 1) $(ostvdevname 1)"
7083 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" &&
7084 skip "meta_bg already set" && return
7087 if [[ ${!opts} != *mkfsoptions* ]]; then
7088 eval opts=\"${!opts} \
7089 --mkfsoptions='\\\"-O ^resize_inode,meta_bg\\\"'\"
7091 local val=${!opts//--mkfsoptions=\\\"/ \
7092 --mkfsoptions=\\\"-O ^resize_inode,meta_bg }
7096 echo "params: $opts"
7098 add ost1 $opts || error "add ost1 failed with new params"
7100 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" ||
7101 error "meta_bg is not set"
7105 run_test 99 "Adding meta_bg option"
7109 start_mds || error "MDS start failed"
7110 start_ost || error "unable to start OST"
7111 mount_client $MOUNT || error "client start failed"
7112 check_mount || error "check_mount failed"
7121 do_facet mgs 'lshowmount -v' | awk 'BEGIN {NR == 0; rc=1} /MGS:/ {rc=0}
7122 END {exit rc}' || error "lshowmount have no output MGS"
7124 do_facet mds1 'lshowmount -v' | awk 'BEGIN {NR == 2; rc=1} /-MDT0000:/
7125 {rc=0} END {exit rc}' || error "lshowmount have no output MDT0"
7127 do_facet ost1 'lshowmount -v' | awk 'BEGIN {NR == 4; rc=1} /-OST0000:/
7128 {rc=0} END {exit rc}' || error "lshowmount have no output OST0"
7130 cleanup || error "cleanup failed with $?"
7132 run_test 100 "check lshowmount lists MGS, MDT, OST and 0@lo"
7135 local createmany_oid
7136 local dev=$FSNAME-OST0000-osc-MDT0000
7139 createmany -o $DIR1/$tfile-%d 50000 &
7141 # MDT->OST reconnection causes MDT<->OST last_id synchornisation
7142 # via osp_precreate_cleanup_orphans.
7143 for ((i = 0; i < 100; i++)); do
7144 for ((k = 0; k < 10; k++)); do
7145 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;" \
7146 "$LCTL --device $dev activate"
7149 ls -asl $MOUNT | grep '???' &&
7150 (kill -9 $createmany_oid &>/dev/null; \
7151 error "File hasn't object on OST")
7153 kill -s 0 $createmany_oid || break
7155 wait $createmany_oid
7158 run_test 101 "Race MDT->OST reconnection with create"
7161 cleanup || error "cleanup failed with $?"
7163 local mds1dev=$(mdsdevname 1)
7164 local mds1mnt=$(facet_mntpt mds1)
7165 local mds1fstype=$(facet_fstype mds1)
7166 local mds1opts=$MDS_MOUNT_OPTS
7168 if [ $mds1fstype == ldiskfs ] &&
7169 ! do_facet mds1 test -b $mds1dev; then
7170 mds1opts=$(csa_add "$mds1opts" -o loop)
7172 if [[ $mds1fstype == zfs ]]; then
7173 import_zpool mds1 || return ${PIPESTATUS[0]}
7176 # unload all and only load libcfs to allow fail_loc setting
7177 do_facet mds1 lustre_rmmod || error "unable to unload modules"
7178 do_facet mds1 modprobe libcfs || error "libcfs not loaded"
7179 do_facet mds1 lsmod \| grep libcfs || error "libcfs not loaded"
7181 #define OBD_FAIL_OBDCLASS_MODULE_LOAD 0x60a
7182 do_facet mds1 "$LCTL set_param fail_loc=0x8000060a"
7184 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts &&
7185 error "mdt start must fail"
7186 do_facet mds1 lsmod \| grep obdclass && error "obdclass must not load"
7188 do_facet mds1 "$LCTL set_param fail_loc=0x0"
7190 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts ||
7191 error "mdt start must not fail"
7193 cleanup || error "cleanup failed with $?"
7195 run_test 102 "obdclass module cleanup upon error"
7200 echo "rename $FSNAME to $newname"
7202 if [ ! combined_mgs_mds ]; then
7203 local facet=$(mgsdevname)
7206 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $facet"||
7207 error "(7) Fail to rename MGS"
7208 if [ "$(facet_fstype $facet)" = "zfs" ]; then
7209 reimport_zpool mgs $newname-mgs
7213 for num in $(seq $MDSCOUNT); do
7214 local facet=$(mdsdevname $num)
7216 do_facet mds${num} \
7217 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $facet"||
7218 error "(8) Fail to rename MDT $num"
7219 if [ "$(facet_fstype $facet)" = "zfs" ]; then
7220 reimport_zpool mds${num} $newname-mdt${num}
7224 for num in $(seq $OSTCOUNT); do
7225 local facet=$(ostdevname $num)
7227 do_facet ost${num} \
7228 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $facet"||
7229 error "(9) Fail to rename OST $num"
7230 if [ "$(facet_fstype $facet)" = "zfs" ]; then
7231 reimport_zpool ost${num} $newname-ost${num}
7236 test_103_set_pool() {
7240 do_facet mgs $LCTL pool_add $FSNAME.$pname ${FSNAME}-$ost_x ||
7241 error "Fail to add $ost_x to $FSNAME.$pname"
7242 wait_update $HOSTNAME \
7243 "lctl get_param -n lov.$FSNAME-clilov-*.pools.$pname |
7244 grep $ost_x" "$FSNAME-${ost_x}_UUID" ||
7245 error "$ost_x is NOT in pool $FSNAME.$pname"
7248 test_103_check_pool() {
7249 local save_fsname=$1
7252 stat $DIR/$tdir/test-framework.sh ||
7253 error "($errno) Fail to stat"
7254 do_facet mgs $LCTL pool_list $FSNAME.pool1 ||
7255 error "($errno) Fail to list $FSNAME.pool1"
7256 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname ||
7257 error "($errno) Fail to list $FSNAME.$save_fsname"
7258 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname |
7259 grep ${FSNAME}-OST0000 ||
7260 error "($errno) List $FSNAME.$save_fsname is invalid"
7262 local pname=$($LFS getstripe --pool $DIR/$tdir/d0)
7263 [ "$pname" = "$save_fsname" ] ||
7264 error "($errno) Unexpected pool name $pname"
7268 check_mount_and_prep
7270 mkdir $DIR/$tdir || error "(1) Fail to mkdir $DIR/$tdir"
7271 cp $LUSTRE/tests/test-framework.sh $DIR/$tdir ||
7272 error "(2) Fail to copy test-framework.sh"
7274 do_facet mgs $LCTL pool_new $FSNAME.pool1 ||
7275 error "(3) Fail to create $FSNAME.pool1"
7276 # name the pool name as the fsname
7277 do_facet mgs $LCTL pool_new $FSNAME.$FSNAME ||
7278 error "(4) Fail to create $FSNAME.$FSNAME"
7280 test_103_set_pool $FSNAME OST0000
7282 $SETSTRIPE -p $FSNAME $DIR/$tdir/d0 ||
7283 error "(6) Fail to setstripe on $DIR/$tdir/d0"
7288 test_renamefs mylustre
7290 local save_fsname=$FSNAME
7294 test_103_check_pool $save_fsname 7
7296 if [ $OSTCOUNT -ge 2 ]; then
7297 test_103_set_pool $save_fsname OST0001
7300 $SETSTRIPE -p $save_fsname $DIR/$tdir/f0 ||
7301 error "(16) Fail to setstripe on $DIR/$tdir/f0"
7310 test_103_check_pool $save_fsname 17
7314 test_renamefs $save_fsname
7320 run_test 103 "rename filesystem name"
7322 if ! combined_mgs_mds ; then
7328 # restore the values of MDSSIZE and OSTSIZE
7329 MDSSIZE=$STORED_MDSSIZE
7330 OSTSIZE=$STORED_OSTSIZE