7 # bug number for skipped test: LU-8972
8 ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT 101"
9 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
13 if [ -r /etc/SuSE-release ]
15 local vers=$(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 false "unexpected entry in mtab before mount" && return 10
460 start_ost || error "OST start failed"
461 start_mds || error "MDS start failed"
462 stop_ost || 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"
470 run_test 5d "mount with ost down"
473 grep " $MOUNT " /etc/mtab &&
474 error false "unexpected entry in mtab before mount" && return 10
476 start_mds || error "MDS start failed"
477 start_ost || error "OST start failed"
479 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
480 do_facet client "$LCTL set_param fail_loc=0x80000506"
481 mount_client $MOUNT || echo "mount failed (not fatal)"
482 cleanup || error "cleanup failed with $?"
483 grep " $MOUNT " /etc/mtab &&
484 error "$MOUNT entry in mtab after unmount"
487 run_test 5e "delayed connect, don't crash (bug 10268)"
490 if combined_mgs_mds ; then
491 skip "combined mgs and mds"
495 grep " $MOUNT " /etc/mtab &&
496 error false "unexpected entry in mtab before mount" && return 10
499 start_ost || error "OST start failed"
500 mount_client $MOUNT &
502 echo client_mount pid is $pid
506 if ! ps -f -p $pid >/dev/null; then
509 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
510 error "mount returns $rc, expected to hang"
512 cleanup || error "cleanup failed with $?"
517 start_mds || error "start MDS failed"
519 # mount should succeed after start mds
521 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
522 cleanup || error "final call to cleanup failed with rc $?"
524 run_test 5f "mds down, cleanup after failed mount (bug 2712)"
529 mount_client $MOUNT || error "mount_client $MOUNT failed"
530 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
531 cleanup || error "cleanup failed with rc $?"
533 run_test 6 "manual umount, then mount again"
538 cleanup_nocli || error "cleanup_nocli failed with $?"
540 run_test 7 "manual umount, then cleanup"
544 mount_client $MOUNT2 || error "mount_client $MOUNT2 failed"
545 check_mount2 || error "check_mount2 failed"
546 umount_client $MOUNT2 || error "umount_client $MOUNT2 failed"
547 cleanup || error "cleanup failed with rc $?"
549 run_test 8 "double mount setup"
552 start_ost || error "OST start failed"
554 do_facet ost1 $LCTL set_param debug=\'inode trace\' ||
555 error "do_facet ost1 set_param inode trace failed."
556 do_facet ost1 $LCTL set_param subsystem_debug=\'mds ost\' ||
557 error "do_facet ost1 set_param debug mds ost failed."
559 CHECK_PTLDEBUG="`do_facet ost1 $LCTL get_param -n debug`"
560 if [ "$CHECK_PTLDEBUG" ] && { \
561 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
562 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
563 echo "lnet.debug success"
565 error "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
567 CHECK_SUBSYS="`do_facet ost1 $LCTL get_param -n subsystem_debug`"
568 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
569 echo "lnet.subsystem_debug success"
571 error "lnet.subsystem_debug: want 'mds ost' got '$CHECK_SUBSYS'"
573 stop_ost || error "Unable to stop OST1"
575 run_test 9 "test ptldebug and subsystem for mkfs"
583 do_facet $facet "test -b $dev" || rc=1
584 if [[ "$size" ]]; then
585 local in=$(do_facet $facet "dd if=$dev of=/dev/null bs=1k \
586 count=1 skip=$size 2>&1" |
587 awk '($3 == "in") { print $1 }')
588 [[ $in = "1+0" ]] || rc=1
594 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
595 # But with new MDS stack we don't care about the mode of local objects
596 # anymore, so this test is removed. See bug 22944 for more details.
600 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
601 skip "Only applicable to ldiskfs-based MDTs"
606 check_mount || error "check_mount failed"
607 cleanup || error "cleanup failed with rc $?"
609 echo "Remove mds config log"
610 if ! combined_mgs_mds ; then
614 do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' \
615 $(mgsdevname) || return \$?" ||
616 error "do_facet mgs failed with $?"
618 if ! combined_mgs_mds ; then
622 start_ost || error "OST start failed"
623 start_mds && error "MDS start succeeded, but should fail"
626 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
629 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
630 skip "Only applicable to ldiskfs-based MDTs"
634 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
639 # check if current MDSSIZE is large enough
640 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE &&
641 log "use MDSSIZE=$MDSSIZE"
643 # check if the global config has a large enough MDSSIZE
644 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] &&
645 [ $STORED_MDSSIZE -ge $MIN ] &&
646 OK=1 && myMDSSIZE=$STORED_MDSSIZE &&
647 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
649 # check if the block device is large enough
650 is_blkdev $SINGLEMDS $MDSDEV $MIN
651 local large_enough=$?
652 if [ -n "$OK" ]; then
653 [ $large_enough -ne 0 ] && OK=""
655 [ $large_enough -eq 0 ] && OK=1 && myMDSSIZE=$MIN &&
656 log "use device $MDSDEV with MIN=$MIN"
659 # check if a loopback device has enough space for fs metadata (5%)
661 if [ -z "$OK" ]; then
662 local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! \
663 -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
664 awk '($1 != "Filesystem") { print $4 }')
665 ! [ -z "$SPACE" ] && [ $SPACE -gt $((MIN / 20)) ] &&
666 OK=1 && myMDSSIZE=$MIN &&
667 log "use file $MDSDEV with MIN=$MIN"
670 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS" && return
672 echo "mount mds with large journal..."
674 local OLD_MDSSIZE=$MDSSIZE
678 echo "mount lustre system..."
680 check_mount || error "check_mount failed"
682 echo "check journal size..."
683 local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
684 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
685 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
687 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
690 cleanup || error "cleanup failed with rc $?"
695 run_test 18 "check mkfs creates large journals"
698 start_mds || error "MDS start failed"
699 stop_mds || error "Unable to stop MDS"
701 run_test 19a "start/stop MDS without OSTs"
704 start_ost || error "Unable to start OST1"
705 stop_ost || error "Unable to stop OST1"
707 run_test 19b "start/stop OSTs without MDS"
710 # first format the ost/mdt
711 start_mds || error "MDS start failed"
712 start_ost || error "Unable to start OST1"
713 mount_client $MOUNT || error "mount_client $MOUNT failed"
714 check_mount || error "check_mount failed"
715 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed."
716 remount_client ro $MOUNT || error "remount_client with ro failed"
717 touch $DIR/$tfile && error "$DIR/$tfile created incorrectly"
718 [ -e $DIR/$tfile ] && error "$DIR/$tfile exists incorrectly"
719 remount_client rw $MOUNT || error "remount_client with rw failed"
720 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
721 MCNT=$(grep -c $MOUNT /etc/mtab)
722 [ "$MCNT" -ne 1 ] && error "$MOUNT in /etc/mtab $MCNT times"
724 stop_mds || error "Unable to stop MDS"
725 stop_ost || error "Unable to stop OST1"
727 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
730 start_mds || error "MDS start failed"
731 start_ost || error "unable to start OST1"
732 wait_osc_import_state mds ost FULL
733 stop_ost || error "unable to stop OST1"
734 stop_mds || error "unable to stop MDS"
736 run_test 21a "start mds before ost, stop ost first"
739 start_ost || error "unable to start OST1"
740 start_mds || error "MDS start failed"
741 wait_osc_import_state mds ost FULL
742 stop_mds || error "unable to stop MDS"
743 stop_ost || error "unable to stop OST1"
745 run_test 21b "start ost before mds, stop mds first"
748 start_ost || error "Unable to start OST1"
749 start_mds || error "MDS start failed"
750 start_ost2 || error "Unable to start OST2"
751 wait_osc_import_state mds ost2 FULL
752 stop_ost || error "Unable to stop OST1"
753 stop_ost2 || error "Unable to stop OST2"
754 stop_mds || error "Unable to stop MDS"
755 #writeconf to remove all ost2 traces for subsequent tests
756 writeconf_or_reformat
758 run_test 21c "start mds between two osts, stop mds last"
761 if combined_mgs_mds ; then
762 skip "need separate mgs device" && return 0
768 start_mgs || error "unable to start MGS"
769 start_ost || error "unable to start OST1"
770 start_ost2 || error "unable to start OST2"
771 start_mds || error "MDS start failed"
772 wait_osc_import_state mds ost2 FULL
774 stop_ost || error "Unable to stop OST1"
775 stop_ost2 || error "Unable to stop OST2"
776 stop_mds || error "Unable to stop MDS"
778 #writeconf to remove all ost2 traces for subsequent tests
779 writeconf_or_reformat
780 start_mgs || error "unable to start MGS"
782 run_test 21d "start mgs then ost and then mds"
785 MGSNID="$saved_mgsnid"
787 echo "stopping fs2mgs ..."
788 stop $fs2mgs -f || true
791 test_21e() { # LU-5863
792 if [[ -z "$fs3ost_DEV" || -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
793 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
794 skip_env "mixed loopback and real device not working" && return
797 local fs2mdsdev=$(mdsdevname 1_2)
798 local fs2ostdev=$(ostdevname 1_2)
799 local fs3ostdev=$(ostdevname 2_2)
801 local fs2mdsvdev=$(mdsvdevname 1_2)
802 local fs2ostvdev=$(ostvdevname 1_2)
803 local fs3ostvdev=$(ostvdevname 2_2)
805 # temporarily use fs3ost as fs2mgs
807 local fs2mgsdev=$fs3ostdev
808 local fs2mgsvdev=$fs3ostvdev
810 local fsname=test1234
812 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
813 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
814 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS && trap cleanup_21e EXIT INT ||
815 error "start fs2mgs failed"
817 local saved_mgsnid="$MGSNID"
818 MGSNID=$(do_facet $fs2mgs $LCTL list_nids | xargs | tr ' ' ,)
820 add fs2mds $(mkfs_opts mds1 $fs2mdsdev $fsname) \
821 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
822 add fs2ost $(mkfs_opts ost1 $fs2ostdev $fsname) \
823 --reformat $fs2ostdev $fs2ostvdev || error "add fs2ost failed"
825 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
826 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS || error "start fs2mds failed"
828 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
829 $MOUNT_CMD $MGSNID:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
830 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
834 run_test 21e "separate MGS and MDS"
837 start_mds || error "MDS start failed"
839 echo "Client mount with ost in logs, but none running"
840 start_ost || error "unable to start OST1"
841 # wait until mds connected to ost and open client connection
842 wait_osc_import_state mds ost FULL
843 stop_ost || error "unable to stop OST1"
844 mount_client $MOUNT || error "mount_client $MOUNT failed"
845 # check_mount will block trying to contact ost
846 mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed"
847 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed"
848 umount_client $MOUNT -f
851 echo "Client mount with a running ost"
852 start_ost || error "unable to start OST1"
854 # if gss enabled, wait full time to let connection from
855 # mds to ost be established, due to the mismatch between
856 # initial connect timeout and gss context negotiation timeout.
857 # This perhaps could be remove after AT landed.
858 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
859 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
861 mount_client $MOUNT || error "mount_client $MOUNT failed"
862 wait_osc_import_state mds ost FULL
863 wait_osc_import_state client ost FULL
864 check_mount || error "check_mount failed"
867 cleanup || error "cleanup failed with rc $?"
869 run_test 22 "start a client before osts (should return errs)"
871 test_23a() { # was test_23
874 stop $SINGLEMDS || error "failed to stop $SINGLEMDS"
875 # force down client so that recovering mds waits for reconnect
876 local running=$(grep -c $MOUNT /proc/mounts) || true
877 if [ $running -ne 0 ]; then
878 echo "Stopping client $MOUNT (opts: -f)"
882 # enter recovery on failed mds
883 local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/})
884 start $SINGLEMDS $MDT_DEV $MDS_MOUNT_OPTS || error "MDS start failed"
885 # try to start a new client
886 mount_client $MOUNT &
888 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
889 MOUNT_LUSTRE_PID=$(ps -ef | grep mount.lustre |
890 grep -v grep | awk '{print $2}')
891 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
893 ps --ppid $MOUNT_LUSTRE_PID
894 echo "waiting for mount to finish"
896 # "ctrl-c" sends SIGINT but it usually (in script) does not work on child process
897 # SIGTERM works but it does not spread to offspring processses
898 kill -s TERM $MOUNT_PID
899 kill -s TERM $MOUNT_LUSTRE_PID
900 # we can not wait $MOUNT_PID because it is not a child of this shell
906 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
908 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
909 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
912 [ -z "$PID1" -a -z "$PID2" ] && break
913 echo "waiting for mount to finish ... "
914 WAIT=$(( WAIT + sleep))
916 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
917 error "MOUNT_PID $MOUNT_PID and "\
918 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
921 stop_mds || error "stopping MDSes failed"
922 stop_ost || error "stopping OSSes failed"
924 run_test 23a "interrupt client during recovery mount delay"
929 test_23b() { # was test_23
930 start_mds || error "MDS start failed"
931 start_ost || error "Unable to start OST1"
932 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
933 $LCTL set_param fail_loc=0x80000313
935 cleanup || error "cleanup failed with rc $?"
937 run_test 23b "Simulate -EINTR during mount"
940 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
942 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
943 is_blkdev $SINGLEMDS $MDSDEV &&
944 skip_env "mixed loopback and real device not working" && return
947 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
949 local fs2mdsdev=$(mdsdevname 1_2)
950 local fs2ostdev=$(ostdevname 1_2)
951 local fs2mdsvdev=$(mdsvdevname 1_2)
952 local fs2ostvdev=$(ostvdevname 1_2)
954 # test 8-char fsname as well
955 local FSNAME2=test1234
957 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --nomgs --mgsnode=$MGSNID \
958 --fsname=${FSNAME2} --reformat $fs2mdsdev $fs2mdsvdev || exit 10
960 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --fsname=${FSNAME2} \
961 --reformat $fs2ostdev $fs2ostvdev || exit 10
964 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
965 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
966 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
967 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
969 check_mount || error "check_mount failed"
970 # files written on 1 should not show up on 2
971 cp /etc/passwd $DIR/$tfile
973 [ -e $MOUNT2/$tfile ] && error "File bleed"
976 cp /etc/passwd $MOUNT2/$tfile ||
977 error "cp /etc/passwd $MOUNT2/$tfile failed"
978 rm $MOUNT2/$tfile || error "remove $MOUNT2/$tfile failed"
979 # 2 is actually mounted
980 grep $MOUNT2' ' /proc/mounts > /dev/null || error "$MOUNT2 not mounted"
982 facet_failover fs2mds
983 facet_failover fs2ost
986 # the MDS must remain up until last MDT
988 MDS=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
989 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
990 [ -z "$MDS" ] && error "No MDT"
992 cleanup_nocli || error "cleanup_nocli failed with rc $?"
994 run_test 24a "Multiple MDTs on a single node"
997 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
999 if [ -z "$fs2mds_DEV" ]; then
1000 local dev=${SINGLEMDS}_dev
1001 local MDSDEV=${!dev}
1002 is_blkdev $SINGLEMDS $MDSDEV &&
1003 skip_env "mixed loopback and real device not working" && return
1006 local fs2mdsdev=$(mdsdevname 1_2)
1007 local fs2mdsvdev=$(mdsvdevname 1_2)
1009 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --mgs --fsname=${FSNAME}2 \
1010 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
1012 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS &&
1013 error "start MDS should fail"
1015 cleanup || error "cleanup failed with rc $?"
1017 run_test 24b "Multiple MGSs on a single node (should return err)"
1021 check_mount || error "check_mount failed"
1022 local MODULES=$($LCTL modules | awk '{ print $2 }')
1023 rmmod $MODULES 2>/dev/null || true
1024 cleanup || error "cleanup failed with $?"
1026 run_test 25 "Verify modules are referenced"
1030 # we need modules before mount for sysctl, so make sure...
1031 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
1032 #define OBD_FAIL_MDS_FS_SETUP 0x135
1033 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000135"
1034 start_mds && error "MDS started but should not have started"
1035 $LCTL get_param -n devices
1036 DEVS=$($LCTL get_param -n devices | egrep -v MG | wc -l)
1037 [ $DEVS -gt 0 ] && error "number of devices is $DEVS, should be zero"
1038 # start mds to drop writeconf setting
1039 start_mds || error "Unable to start MDS"
1040 stop_mds || error "Unable to stop MDS"
1041 unload_modules_conf || error "unload_modules_conf failed with $?"
1043 run_test 26 "MDT startup failure cleans LOV (should return errs)"
1046 start_ost || error "Unable to start OST1"
1047 start_mds || error "Unable to start MDS"
1048 echo "Requeue thread should have started: "
1049 ps -e | grep ll_cfg_requeue
1050 set_conf_param_and_check ost1 \
1051 "$LCTL get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" \
1052 "$FSNAME-OST0000.ost.client_cache_seconds" ||
1053 error "set_conf_param_and_check ost1 failed"
1054 cleanup_nocli || error "cleanup_nocli failed with rc $?"
1056 run_test 27a "Reacquire MGS lock if OST started first"
1061 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
1062 awk '($3 ~ "mdt" && $4 ~ "MDT0000") { print $4 }')
1064 facet_failover $SINGLEMDS
1065 set_conf_param_and_check $SINGLEMDS \
1066 "$LCTL get_param -n mdt.$device.identity_acquire_expire" \
1067 "$device.mdt.identity_acquire_expire" ||
1068 error "set_conf_param_and_check $SINGLEMDS failed"
1069 set_conf_param_and_check client \
1070 "$LCTL get_param -n mdc.$device-mdc-*.max_rpcs_in_flight"\
1071 "$device.mdc.max_rpcs_in_flight" ||
1072 error "set_conf_param_and_check client failed"
1074 cleanup || error "cleanup failed with $?"
1076 run_test 27b "Reacquire MGS lock after failover"
1080 TEST="$LCTL get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1081 PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
1083 FINAL=$(($ORIG + 1))
1084 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL ||
1085 error "first set_conf_param_and_check client failed"
1086 FINAL=$(($FINAL + 1))
1087 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL ||
1088 error "second set_conf_param_and_check client failed"
1089 umount_client $MOUNT || error "umount_client $MOUNT failed"
1090 mount_client $MOUNT || error "mount_client $MOUNT failed"
1092 if [ $RESULT -ne $FINAL ]; then
1093 error "New config not seen: wanted $FINAL got $RESULT"
1095 echo "New config success: got $RESULT"
1097 set_conf_param_and_check client "$TEST" "$PARAM" $ORIG ||
1098 error "third set_conf_param_and_check client failed"
1099 cleanup || error "cleanup failed with rc $?"
1101 run_test 28 "permanent parameter setting"
1103 test_28a() { # LU-4221
1104 [[ $(lustre_version_code ost1) -ge $(version_code 2.5.52) ]] ||
1105 { skip "Need OST version at least 2.5.52" && return 0; }
1106 [ "$(facet_fstype ost1)" = "zfs" ] &&
1107 skip "LU-4221: no such proc params for ZFS OSTs" && return
1114 local device="$FSNAME-OST0000"
1118 # In this test we will set three kinds of proc parameters with
1120 # 1. the ones moved from the OFD to the OSD, and only their
1121 # symlinks kept in obdfilter
1122 # 2. non-symlink ones in the OFD
1123 # 3. non-symlink ones in the OSD
1126 # prepare a symlink parameter in the OFD
1127 name="writethrough_cache_enable"
1128 param="$device.ost.$name"
1129 cmd="$LCTL get_param -n obdfilter.$device.$name"
1131 # conf_param the symlink parameter in the OFD
1132 old=$(do_facet ost1 $cmd)
1133 new=$(((old + 1) % 2))
1134 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1135 error "lctl conf_param $device.ost.$param=$new failed"
1137 # conf_param the target parameter in the OSD
1138 param="$device.osd.$name"
1139 cmd="$LCTL get_param -n osd-*.$device.$name"
1140 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1141 error "lctl conf_param $device.osd.$param=$old failed"
1144 # prepare a non-symlink parameter in the OFD
1145 name="client_cache_seconds"
1146 param="$device.ost.$name"
1147 cmd="$LCTL get_param -n obdfilter.$device.$name"
1149 # conf_param the parameter in the OFD
1150 old=$(do_facet ost1 $cmd)
1152 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1153 error "lctl conf_param $device.ost.$param=$new failed"
1154 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1155 error "lctl conf_param $device.ost.$param=$old failed"
1158 # prepare a non-symlink parameter in the OSD
1160 param="$device.osd.$name"
1161 cmd="$LCTL get_param -n osd-*.$device.$name"
1163 # conf_param the parameter in the OSD
1164 old=$(do_facet ost1 $cmd)
1165 new=$(((old + 1) % 2))
1166 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1167 error "lctl conf_param $device.osd.$param=$new failed"
1168 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1169 error "lctl conf_param $device.osd.$param=$old failed"
1171 cleanup || error "cleanup failed with $?"
1173 run_test 28a "set symlink parameters permanently with conf_param"
1176 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
1177 setup > /dev/null 2>&1
1178 start_ost2 || error "Unable to start OST2"
1181 local PARAM="$FSNAME-OST0001.osc.active"
1182 local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
1183 local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
1185 ACTV=$($LCTL get_param -n $PROC_ACT)
1187 set_conf_param_and_check client \
1188 "$LCTL get_param -n $PROC_ACT" "$PARAM" $DEAC ||
1189 error "set_conf_param_and_check client failed"
1190 # also check ost_server_uuid status
1191 RESULT=$($LCTL get_param -n $PROC_UUID | grep DEACTIV)
1192 if [ -z "$RESULT" ]; then
1193 error "Client not deactivated: $($LCTL get_param \
1196 echo "Live client success: got $RESULT"
1200 for num in $(seq $MDSCOUNT); do
1201 local mdtosc=$(get_mdtosc_proc_path mds${num} $FSNAME-OST0001)
1202 local MPROC="osc.$mdtosc.active"
1207 RESULT=$(do_facet mds${num} "$LCTL get_param -n $MPROC")
1208 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
1209 if [ $RESULT -eq $DEAC ]; then
1210 echo -n "MDT deactivated also after"
1211 echo "$WAIT sec (got $RESULT)"
1215 if [ $WAIT -eq $MAX ]; then
1216 error "MDT active: wanted $DEAC got $RESULT"
1218 echo "Waiting $(($MAX - $WAIT))secs for MDT deactivated"
1221 # test new client starts deactivated
1222 umount_client $MOUNT || error "umount_client $MOUNT failed"
1223 mount_client $MOUNT || error "mount_client $MOUNT failed"
1224 RESULT=$($LCTL get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
1225 if [ -z "$RESULT" ]; then
1226 error "New client start active: $(lctl get_param -n $PROC_UUID)"
1228 echo "New client success: got $RESULT"
1231 # make sure it reactivates
1232 set_conf_param_and_check client \
1233 "$LCTL get_param -n $PROC_ACT" "$PARAM" $ACTV ||
1234 error "lctl get_param $PROC_ACT $PARAM $ACTV failed"
1236 umount_client $MOUNT
1237 stop_ost2 || error "Unable to stop OST2"
1238 cleanup_nocli || error "cleanup_nocli failed with $?"
1239 #writeconf to remove all ost2 traces for subsequent tests
1240 writeconf_or_reformat
1242 run_test 29 "permanently remove an OST"
1247 echo Big config llog
1248 TEST="$LCTL get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1250 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1251 for i in ${LIST[@]}; do
1252 set_conf_param_and_check client "$TEST" \
1253 "$FSNAME.llite.max_read_ahead_whole_mb" $i ||
1254 error "Set $FSNAME.llite.max_read_ahead_whole_mb failed"
1256 # make sure client restart still works
1257 umount_client $MOUNT
1258 mount_client $MOUNT || error "mount_client $MOUNT failed"
1259 [ "$($TEST)" -ne "$i" ] &&
1260 error "Param didn't stick across restart $($TEST) != $i"
1263 echo Erase parameter setting
1264 do_facet mgs "$LCTL conf_param \
1265 -d $FSNAME.llite.max_read_ahead_whole_mb" ||
1266 error "Erase param $FSNAME.llite.max_read_ahead_whole_mb failed"
1267 umount_client $MOUNT
1268 mount_client $MOUNT || error "mount_client $MOUNT failed"
1270 echo "deleted (default) value=$FINAL, orig=$ORIG"
1271 # assumes this parameter started at the default value
1272 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
1274 cleanup || error "cleanup failed with rc $?"
1276 run_test 30a "Big config llog and conf_param deletion"
1281 local orignids=$($LCTL get_param -n \
1282 osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1284 local orignidcount=$(echo "$orignids" | wc -w)
1286 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1287 # numerical part of it. Hopefully that's not already a failover address
1289 local OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | \
1291 local ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1292 local NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1293 local NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1294 echo "Using fake nid $NEW"
1296 local TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1297 grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1298 set_conf_param_and_check client "$TEST" \
1299 "$FSNAME-OST0000.failover.node" $NEW ||
1300 error "didn't add failover nid $NEW"
1301 local NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1304 local NIDCOUNT=$(echo "$NIDS" | wc -w)
1305 echo "should have $((orignidcount + 1)) entries \
1306 in failover nids string, have $NIDCOUNT"
1307 [ $NIDCOUNT -eq $((orignidcount + 1)) ] ||
1308 error "Failover nid not added"
1310 do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" ||
1311 error "conf_param delete failed"
1312 umount_client $MOUNT
1313 mount_client $MOUNT || error "mount_client $MOUNT failed"
1315 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1318 NIDCOUNT=$(echo "$NIDS" | wc -w)
1319 echo "only $orignidcount final entries should remain \
1320 in failover nids string, have $NIDCOUNT"
1321 [ $NIDCOUNT -eq $orignidcount ] || error "Failover nids not removed"
1323 cleanup || error "cleanup failed with rc $?"
1325 run_test 30b "Remove failover nids"
1327 test_31() { # bug 10734
1328 # ipaddr must not exist
1329 $MOUNT_CMD 4.3.2.1@tcp:/lustre $MOUNT || true
1330 cleanup || error "cleanup failed with rc $?"
1332 run_test 31 "Connect to non-existent node (shouldn't crash)"
1336 T32_BLIMIT=40960 # Kbytes
1340 # This is not really a test but a tool to create new disk
1341 # image tarballs for the upgrade tests.
1343 # Disk image tarballs should be created on single-node
1344 # clusters by running this test with default configurations
1345 # plus a few mandatory environment settings that are verified
1346 # at the beginning of the test.
1348 test_32newtarball() {
1352 local tmp=$TMP/t32_image_create
1353 local server_version=$(lustre_version_code $SINGLEMDS)
1358 if [ $FSNAME != t32fs -o \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o \
1359 $OSTCOUNT -ne 1 -o -z "$OSTDEV1" ]; then
1360 error "Needs FSNAME=t32fs MDSCOUNT=2 " \
1361 "MDSDEV1=<nonexistent_file>" \
1362 "MDSDEV2=<nonexistent_file>" \
1363 "(or MDSDEV, in the case of b1_8)" \
1364 "OSTCOUNT=1 OSTDEV1=<nonexistent_file>"
1368 echo "Found stale $tmp"
1372 mkdir $tmp/src || return 1
1373 tar cf - -C $src . | tar xf - -C $tmp/src
1374 dd if=/dev/zero of=$tmp/src/t32_qf_old bs=1M \
1375 count=$(($T32_BLIMIT / 1024 / 4))
1376 chown $T32_QID.$T32_QID $tmp/src/t32_qf_old
1378 # format ost with comma-separated NIDs to verify LU-4460
1379 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
1380 MGSNID="$MGSNID,$MGSNID" OSTOPT="--failnode=$failnid" formatall
1384 [[ $server_version -ge $(version_code 2.3.50) ]] ||
1385 $LFS quotacheck -ug /mnt/$FSNAME
1386 $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
1389 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1391 if [[ $MDSCOUNT -ge 2 ]]; then
1392 remote_dir=/mnt/$FSNAME/remote_dir
1393 $LFS mkdir -i 1 $remote_dir
1394 tar cf - -C $tmp/src . | tar xf - -C $remote_dir
1396 if [[ $server_version -ge $(version_code 2.7.0) ]]; then
1397 striped_dir=/mnt/$FSNAME/striped_dir_old
1398 $LFS mkdir -i 1 -c 2 $striped_dir
1399 tar cf - -C $tmp/src . | tar xf - -C $striped_dir
1405 mkdir $tmp/img || return 1
1409 pushd_dir=/mnt/$FSNAME
1410 if [[ $MDSCOUNT -ge 2 ]]; then
1411 pushd_dir=$remote_dir
1412 if [[ $server_version -ge $(version_code 2.7.0) ]]; then
1414 ls -Rni --time-style=+%s >$tmp/img/list2
1420 ls -Rni --time-style=+%s >$tmp/img/list
1421 find ! -name .lustre -type f -exec sha1sum {} \; |
1422 sort -k 2 >$tmp/img/sha1sums
1424 $LCTL get_param -n version | head -n 1 |
1425 sed -e 's/^lustre: *//' >$tmp/img/commit
1427 [[ $server_version -ge $(version_code 2.3.50) ]] ||
1428 $LFS quotaon -ug /mnt/$FSNAME
1429 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1430 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1431 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1432 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1433 | tr -d "*" > $tmp/img/bspace
1434 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1435 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1436 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1437 | tr -d "*" > $tmp/img/ispace
1438 echo $T32_BLIMIT > $tmp/img/blimit
1439 echo $T32_ILIMIT > $tmp/img/ilimit
1444 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1447 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1448 echo "Data verification failed"
1451 uname -r >$tmp/img/kernel
1452 uname -m >$tmp/img/arch
1454 mv ${MDSDEV1:-$MDSDEV} $tmp/img
1455 for num in $(seq 2 $MDSCOUNT); do
1456 local devname=$(mdsdevname $num)
1458 [[ $(facet_fstype $facet) != zfs ]] ||
1459 devname=$(mdsvdevname $num)
1460 mv $devname $tmp/img
1462 mv $OSTDEV1 $tmp/img
1464 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1465 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1468 tar cjvf $dst/disk$version-$(facet_fstype $SINGLEMDS).tar.bz2 -S *
1473 #run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1476 # The list of applicable tarballs is returned via the caller's
1477 # variable "tarballs".
1480 local node=$(facet_active_host $SINGLEMDS)
1481 local r="do_node $node"
1483 if [ "$CLIENTONLY" ]; then
1484 skip "Client-only testing"
1488 if ! $r which $TUNEFS; then
1489 skip_env "tunefs.lustre required on $node"
1493 local IMGTYPE=$(facet_fstype $SINGLEMDS)
1495 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 -name \'disk*-$IMGTYPE.tar.bz2\')
1497 if [ -z "$tarballs" ]; then
1498 skip "No applicable tarballs found"
1503 t32_test_cleanup() {
1505 local facet=$SINGLEMDS
1506 local fstype=$(facet_fstype $facet)
1509 if $shall_cleanup_lustre; then
1510 umount $tmp/mnt/lustre || rc=$?
1512 if $shall_cleanup_mdt; then
1513 $r $UMOUNT $tmp/mnt/mdt || rc=$?
1515 if $shall_cleanup_mdt1; then
1516 $r $UMOUNT $tmp/mnt/mdt1 || rc=$?
1518 if $shall_cleanup_ost; then
1519 $r $UMOUNT $tmp/mnt/ost || rc=$?
1524 if [[ $fstype == zfs ]]; then
1526 local poolname_list="t32fs-mdt1 t32fs-ost1"
1528 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1530 for poolname in $poolname_list; do
1531 destroy_zpool $facet $poolname
1537 t32_bits_per_long() {
1539 # Yes, this is not meant to be perfect.
1549 t32_reload_modules() {
1551 local all_removed=false
1554 while ((i < 20)); do
1555 echo "Unloading modules on $node: Attempt $i"
1556 do_rpc_nodes $node $LUSTRE_RMMOD $(facet_fstype $SINGLEMDS) &&
1558 do_rpc_nodes $node check_mem_leak || return 1
1559 if $all_removed; then
1560 do_rpc_nodes $node load_modules
1566 echo "Unloading modules on $node: Given up"
1570 t32_wait_til_devices_gone() {
1576 echo wait for devices to go
1577 while ((i < 20)); do
1578 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1579 loops=$(do_rpc_nodes $node losetup -a | grep -c t32)
1580 ((devices == 0 && loops == 0)) && return 0
1584 echo "waiting for dev on $node: dev $devices loop $loops given up"
1585 do_rpc_nodes $node "losetup -a"
1586 do_rpc_nodes $node "$LCTL device_list"
1590 t32_verify_quota() {
1594 local fstype=$(facet_fstype $SINGLEMDS)
1598 $LFS quota -u $T32_QID -v $mnt
1600 qval=$($LFS quota -v -u $T32_QID $mnt |
1601 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1602 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1604 [ $qval -eq $img_bspace ] || {
1605 echo "bspace, act:$qval, exp:$img_bspace"
1609 qval=$($LFS quota -v -u $T32_QID $mnt |
1610 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1611 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1613 [ $qval -eq $img_ispace ] || {
1614 echo "ispace, act:$qval, exp:$img_ispace"
1618 qval=$($LFS quota -v -u $T32_QID $mnt |
1619 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1620 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1622 [ $qval -eq $img_blimit ] || {
1623 echo "blimit, act:$qval, exp:$img_blimit"
1627 qval=$($LFS quota -v -u $T32_QID $mnt |
1628 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1629 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1631 [ $qval -eq $img_ilimit ] || {
1632 echo "ilimit, act:$qval, exp:$img_ilimit"
1636 do_node $node $LCTL conf_param $fsname.quota.mdt=ug
1637 cmd="$LCTL get_param -n osd-$fstype.$fsname-MDT0000"
1638 cmd=$cmd.quota_slave.enabled
1639 wait_update $node "$cmd" "ug" || {
1640 echo "Enable mdt quota failed"
1644 do_node $node $LCTL conf_param $fsname.quota.ost=ug
1645 cmd="$LCTL get_param -n osd-$fstype.$fsname-OST0000"
1646 cmd=$cmd.quota_slave.enabled
1647 wait_update $node "$cmd" "ug" || {
1648 echo "Enable ost quota failed"
1653 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1654 bs=1M count=$((img_blimit / 1024)) oflag=sync && {
1655 echo "Write succeed, but expect -EDQUOT"
1658 rm -f $mnt/t32_qf_new
1660 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1662 echo "Create succeed, but expect -EDQUOT"
1665 unlinkmany $mnt/t32_qf_ $img_ilimit
1673 local dne_upgrade=${dne_upgrade:-"no"}
1674 local ff_convert=${ff_convert:-"no"}
1675 local shall_cleanup_mdt=false
1676 local shall_cleanup_mdt1=false
1677 local shall_cleanup_ost=false
1678 local shall_cleanup_lustre=false
1679 local mdt2_is_available=false
1680 local node=$(facet_active_host $SINGLEMDS)
1681 local r="do_node $node"
1682 local node2=$(facet_active_host mds2)
1692 local nid=$($r $LCTL list_nids | head -1)
1698 local fstype=$(facet_fstype $SINGLEMDS)
1699 local mdt_dev=$tmp/mdt
1700 local mdt2_dev=$tmp/mdt2
1701 local ost_dev=$tmp/ost
1706 trap 'trap - RETURN; t32_test_cleanup' RETURN
1709 mkdir -p $tmp/mnt/lustre || error "mkdir $tmp/mnt/lustre failed"
1710 $r mkdir -p $tmp/mnt/{mdt,mdt1,ost}
1711 $r tar xjvf $tarball -S -C $tmp || {
1712 error_noexit "Unpacking the disk image tarball"
1715 img_commit=$($r cat $tmp/commit)
1716 img_kernel=$($r cat $tmp/kernel)
1717 img_arch=$($r cat $tmp/arch)
1718 img_bspace=$($r cat $tmp/bspace)
1719 img_ispace=$($r cat $tmp/ispace)
1721 # older images did not have "blimit" and "ilimit" files
1722 # use old values for T32_BLIMIT and T32_ILIMIT
1723 $r test -f $tmp/blimit && img_blimit=$($r cat $tmp/blimit) ||
1725 $r test -f $tmp/ilimit && img_ilimit=$($r cat $tmp/ilimit) ||
1728 echo "Upgrading from $(basename $tarball), created with:"
1729 echo " Commit: $img_commit"
1730 echo " Kernel: $img_kernel"
1731 echo " Arch: $img_arch"
1732 echo "OST version: $(lustre_build_version ost1)"
1734 # The conversion can be made only when both of the following
1735 # conditions are satisfied:
1736 # - ost device img version < 2.3.64
1737 # - ost server version >= 2.5
1738 [ $(version_code $img_commit) -ge $(version_code 2.3.64) -o \
1739 $(lustre_version_code ost1) -lt $(version_code 2.5.0) ] &&
1742 ! $r test -f $mdt2_dev || mdt2_is_available=true
1744 if [[ $fstype == zfs ]]; then
1747 local poolname_list="t32fs-mdt1 t32fs-ost1"
1749 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1751 for poolname in $poolname_list; do
1753 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
1754 $ZPOOL import -f -d $tmp $poolname"
1757 mdt_dev=t32fs-mdt1/mdt1
1758 ost_dev=t32fs-ost1/ost1
1759 ! $mdt2_is_available || mdt2_dev=t32fs-mdt2/mdt2
1760 wait_update_facet $SINGLEMDS "$ZPOOL list |
1761 awk '/^t32fs-mdt1/ { print \\\$1 }'" "t32fs-mdt1" || {
1762 error_noexit "import zfs pool failed"
1767 $r $LCTL set_param debug="$PTLDEBUG"
1769 $r $TUNEFS --dryrun $mdt_dev || {
1771 error_noexit "tunefs.lustre before mounting the MDT"
1775 if $mdt2_is_available; then
1776 $r $TUNEFS --dryrun $mdt2_dev || {
1778 error_noexit "tunefs.lustre before mounting the MDT"
1783 if [ "$writeconf" ]; then
1785 if [ $fstype == "ldiskfs" ]; then
1787 $r $TUNEFS --quota $mdt_dev || {
1789 error_noexit "Enable mdt quota feature"
1792 if $mdt2_is_available; then
1793 $r $TUNEFS --quota $mdt2_dev || {
1795 error_noexit "Enable mdt quota feature"
1801 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
1802 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
1803 { skip "LU-2200: Cannot run over Inifiniband w/o lctl replace_nids "
1804 "(Need MGS version at least 2.3.59)"; return 0; }
1806 local osthost=$(facet_active_host ost1)
1807 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
1810 if [ $fstype == "ldiskfs" ]; then
1813 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt
1814 $r $LCTL replace_nids $fsname-OST0000 $ostnid
1815 $r $LCTL replace_nids $fsname-MDT0000 $nid
1816 $r $UMOUNT $tmp/mnt/mdt
1819 mopts=exclude=$fsname-OST0000
1820 if [ $fstype == "ldiskfs" ]; then
1825 t32_wait_til_devices_gone $node
1827 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
1829 error_noexit "Mounting the MDT"
1832 shall_cleanup_mdt=true
1834 if $mdt2_is_available; then
1835 mopts=mgsnode=$nid,$mopts
1836 $r $MOUNT_CMD -o $mopts $mdt2_dev $tmp/mnt/mdt1 || {
1838 error_noexit "Mounting the MDT"
1842 echo "mount new MDT....$mdt2_dev"
1843 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1844 error_noexit "enable remote dir create failed"
1846 shall_cleanup_mdt1=true
1847 elif [ "$dne_upgrade" != "no" ]; then
1848 local fs2mdsdev=$(mdsdevname 1_2)
1849 local fs2mdsvdev=$(mdsvdevname 1_2)
1851 echo "mkfs new MDT on ${fs2mdsdev}...."
1852 if [ $(facet_fstype mds1) == ldiskfs ]; then
1853 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
1856 add fs2mds $(mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
1857 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
1858 error_noexit "Mkfs new MDT failed"
1862 [[ $(facet_fstype mds1) != zfs ]] || import_zpool fs2mds
1864 $r $TUNEFS --dryrun $fs2mdsdev || {
1865 error_noexit "tunefs.lustre before mounting the MDT"
1869 echo "mount new MDT....$fs2mdsdev"
1870 $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
1871 error_noexit "mount mdt1 failed"
1875 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1876 error_noexit "enable remote dir create failed"
1878 shall_cleanup_mdt1=true
1881 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
1882 error_noexit "Getting MDT UUID"
1885 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
1886 error_noexit "Unexpected MDT UUID: \"$uuid\""
1890 $r $TUNEFS --dryrun $ost_dev || {
1891 error_noexit "tunefs.lustre before mounting the OST"
1894 if [ "$writeconf" ]; then
1895 mopts=mgsnode=$nid,$writeconf
1896 if [ $fstype == "ldiskfs" ]; then
1898 $r $TUNEFS --quota $ost_dev || {
1900 error_noexit "Enable ost quota feature"
1906 if [ $fstype == "ldiskfs" ]; then
1910 $r $MOUNT_CMD -o $mopts $ost_dev $tmp/mnt/ost || {
1911 error_noexit "Mounting the OST"
1914 shall_cleanup_ost=true
1916 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
1917 error_noexit "Getting OST UUID"
1920 if [ "$uuid" != $fsname-OST0000_UUID ]; then
1921 error_noexit "Unexpected OST UUID: \"$uuid\""
1925 $r $LCTL conf_param $fsname-OST0000.osc.max_dirty_mb=15 || {
1926 error_noexit "Setting \"max_dirty_mb\""
1929 $r $LCTL conf_param $fsname-OST0000.failover.node=$nid || {
1930 error_noexit "Setting OST \"failover.node\""
1933 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
1934 error_noexit "Setting \"max_rpcs_in_flight\""
1937 $r $LCTL conf_param $fsname-MDT0000.failover.node=$nid || {
1938 error_noexit "Setting MDT \"failover.node\""
1941 $r $LCTL pool_new $fsname.interop || {
1942 error_noexit "Setting \"interop\""
1945 $r $LCTL conf_param $fsname-MDT0000.lov.stripesize=4M || {
1946 error_noexit "Setting \"lov.stripesize\""
1949 $r $LCTL conf_param $fsname-MDT0000.mdd.atime_diff=70 || {
1950 error_noexit "Setting \"mdd.atime_diff\""
1954 if [ "$ff_convert" != "no" -a $(facet_fstype ost1) == "ldiskfs" ]; then
1955 $r $LCTL lfsck_start -M $fsname-OST0000 || {
1956 error_noexit "Start OI scrub on OST0"
1960 # The oi_scrub should be on ost1, but for test_32(),
1961 # all on the SINGLEMDS.
1962 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
1963 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1964 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
1965 error_noexit "Failed to get the expected 'completed'"
1969 local UPDATED=$($r $LCTL get_param -n \
1970 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1971 awk '/^updated/ { print $2 }')
1972 [ $UPDATED -ge 1 ] || {
1973 error_noexit "Only $UPDATED objects have been converted"
1978 if [ "$dne_upgrade" != "no" ]; then
1979 $r $LCTL conf_param \
1980 $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
1981 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
1984 $r $LCTL conf_param $fsname-MDT0001.failover.node=$nid || {
1985 error_noexit "Setting MDT1 \"failover.node\""
1988 $r $LCTL conf_param $fsname-MDT0001.lov.stripesize=4M || {
1989 error_noexit "Setting MDT1 \"lov.stripesize\""
1994 if [ "$writeconf" ]; then
1995 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
1996 error_noexit "Mounting the client"
2000 shall_cleanup_lustre=true
2001 $r $LCTL set_param debug="$PTLDEBUG"
2003 t32_verify_quota $node $fsname $tmp/mnt/lustre || {
2004 error_noexit "verify quota failed"
2008 if $r test -f $tmp/list; then
2010 # There is not a Test Framework API to copy files to or
2011 # from a remote node.
2013 # LU-2393 - do both sorts on same node to ensure locale
2015 local list_file=$tmp/list
2017 if $mdt2_is_available; then
2018 if [[ -d $tmp/mnt/lustre/striped_dir_old ]] &&
2019 $r test -f $tmp/list2; then
2020 list_file=$tmp/list2
2021 pushd $tmp/mnt/lustre/striped_dir_old
2023 pushd $tmp/mnt/lustre/remote_dir
2026 pushd $tmp/mnt/lustre
2028 $r cat $list_file | sort -k 6 >$tmp/list.orig
2029 ls -Rni --time-style=+%s | sort -k 6 |
2030 sed 's/\. / /' >$tmp/list || {
2036 # 32-bit and 64-bit clients use different algorithms to
2037 # convert FIDs into inode numbers. Hence, remove the
2038 # inode numbers from the lists, if the original list was
2039 # created on an architecture with different number of
2042 if [ $(t32_bits_per_long $(uname -m)) != \
2043 $(t32_bits_per_long $img_arch) ]; then
2044 echo "Different number of bits per \"long\"" \
2045 "from the disk image"
2046 for list in list.orig list; do
2047 sed -i -e 's/^[0-9]\+[ \t]\+//' \
2051 if ! diff -ub $tmp/list.orig $tmp/list; then
2052 error_noexit "list verification failed"
2056 echo "list verification skipped"
2059 if [ "$dne_upgrade" != "no" ]; then
2060 $LFS mkdir -i 1 -c2 $tmp/mnt/lustre/striped_dir || {
2061 error_noexit "set striped dir failed"
2065 $LFS setdirstripe -D -c2 $tmp/mnt/lustre/striped_dir
2067 pushd $tmp/mnt/lustre
2068 tar -cf - . --exclude=./striped_dir \
2069 --exclude=./striped_dir_old \
2070 --exclude=./remote_dir |
2071 tar -xvf - -C striped_dir 1>/dev/null || {
2072 error_noexit "cp to striped dir failed"
2078 # If it is upgrade from DNE (2.5), then rename the remote dir,
2079 # which is created in 2.5 to striped dir.
2080 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]]; then
2081 stripe_index=$($LFS getdirstripe -i \
2082 $tmp/mnt/lustre/remote_dir)
2084 [[ $stripe_index -eq 1 ]] || {
2085 error_noexit "get index \"$stripe_index\"" \
2086 "from remote dir failed"
2089 mv $tmp/mnt/lustre/remote_dir \
2090 $tmp/mnt/lustre/striped_dir/ || {
2091 error_noexit "mv remote dir failed"
2096 # If it is upgraded from DNE (2.7), then move the striped dir
2097 # which was created in 2.7 to the new striped dir.
2098 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]] &&
2099 [[ -d $tmp/mnt/lustre/striped_dir_old ]]; then
2100 stripe_count=$($LFS getdirstripe -c \
2101 $tmp/mnt/lustre/striped_dir_old)
2102 [[ $stripe_count -eq 2 ]] || {
2103 error_noexit "get count $stripe_count" \
2104 "from striped dir failed"
2107 mv $tmp/mnt/lustre/striped_dir_old \
2108 $tmp/mnt/lustre/striped_dir/ || {
2109 error_noexit "mv striped dir failed"
2115 $r $LCTL set_param -n osd*.*.force_sync=1
2116 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
2117 error_noexit "dd failed"
2120 rm -rf $tmp/mnt/lustre/tmp_file || {
2121 error_noexit "rm failed"
2125 if $r test -f $tmp/sha1sums; then
2126 # LU-2393 - do both sorts on same node to ensure locale
2128 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
2129 if [ "$dne_upgrade" != "no" ]; then
2130 pushd $tmp/mnt/lustre/striped_dir
2132 pushd $tmp/mnt/lustre
2135 find ! -path "*remote_dir*" ! -path "*striped_dir*" \
2136 ! -name .lustre -type f -exec sha1sum {} \; |
2137 sort -k 2 >$tmp/sha1sums || {
2139 error_noexit "sha1sum"
2143 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
2144 error_noexit "sha1sum verification failed"
2148 # if upgrade from DNE(2.5), then check remote directory
2149 # if upgrade from DNE(2.7), then check striped directory
2150 if $mdt2_is_available &&
2151 [[ "$dne_upgrade" != "no" ]]; then
2152 local new_dir="$tmp/mnt/lustre/striped_dir"
2153 local striped_dir_old="$new_dir/striped_dir_old"
2155 local dir_list="$new_dir/remote_dir"
2156 [[ ! -d $triped_dir_old ]] ||
2157 dir_list+=" $striped_dir_old"
2159 for dir in $dir_list; do
2161 find ! -name .lustre -type f \
2162 -exec sha1sum {} \; |
2163 sort -k 2 >$tmp/sha1sums || {
2165 error_noexit "sha1sum"
2169 if ! diff -ub $tmp/sha1sums.orig \
2171 error_noexit "sha1sum $dir" \
2178 echo "sha1sum verification skipped"
2181 if [ "$dne_upgrade" != "no" ]; then
2182 rm -rf $tmp/mnt/lustre/striped_dir || {
2183 error_noexit "remove remote dir failed"
2188 # migrate files/dirs to remote MDT, then move them back
2189 if [ $(lustre_version_code mds1) -ge $(version_code 2.7.50) -a \
2190 $dne_upgrade != "no" ]; then
2191 $r $LCTL set_param -n \
2192 mdt.${fsname}*.enable_remote_dir=1 2>/dev/null
2194 echo "test migration"
2195 pushd $tmp/mnt/lustre
2196 for dir in $(find ! -name .lustre ! -name . -type d); do
2197 mdt_index=$($LFS getdirstripe -i $dir)
2198 stripe_cnt=$($LFS getdirstripe -c $dir)
2199 if [ $mdt_index = 0 -a $stripe_cnt -le 1 ]; then
2200 $LFS mv -M 1 $dir || {
2202 error_noexit "migrate MDT1 failed"
2208 for dir in $(find ! -name . ! -name .lustre -type d); do
2209 mdt_index=$($LFS getdirstripe -i $dir)
2210 stripe_cnt=$($LFS getdirstripe -c $dir)
2211 if [ $mdt_index = 1 -a $stripe_cnt -le 1 ]; then
2212 $LFS mv -M 0 $dir || {
2214 error_noexit "migrate MDT0 failed"
2223 # When adding new data verification tests, please check for
2224 # the presence of the required reference files first, like
2225 # the "sha1sums" and "list" tests above, to avoid the need to
2226 # regenerate every image for each test addition.
2229 nrpcs_orig=$($LCTL get_param \
2230 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
2231 error_noexit "Getting \"max_rpcs_in_flight\""
2234 nrpcs=$((nrpcs_orig + 5))
2235 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=$nrpcs || {
2236 error_noexit "Changing \"max_rpcs_in_flight\""
2239 wait_update $HOSTNAME "$LCTL get_param \
2240 -n mdc.*MDT0000*.max_rpcs_in_flight" $nrpcs || {
2241 error_noexit "Verifying \"max_rpcs_in_flight\""
2245 umount $tmp/mnt/lustre || {
2246 error_noexit "Unmounting the client"
2249 shall_cleanup_lustre=false
2251 if [[ "$dne_upgrade" != "no" ]] || $mdt2_is_available; then
2252 $r $UMOUNT $tmp/mnt/mdt1 || {
2253 error_noexit "Unmounting the MDT2"
2256 shall_cleanup_mdt1=false
2259 $r $UMOUNT $tmp/mnt/mdt || {
2260 error_noexit "Unmounting the MDT"
2263 shall_cleanup_mdt=false
2265 $r $UMOUNT $tmp/mnt/ost || {
2266 error_noexit "Unmounting the OST"
2269 shall_cleanup_ost=false
2271 t32_reload_modules $node || {
2272 error_noexit "Reloading modules"
2276 if [[ $fstype == zfs ]]; then
2277 local poolname=t32fs-mdt1
2279 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
2280 $ZPOOL import -f -d $tmp $poolname"
2283 # mount a second time to make sure we didnt leave upgrade flag on
2284 $r $TUNEFS --dryrun $mdt_dev || {
2286 error_noexit "tunefs.lustre before remounting the MDT"
2290 mopts=exclude=$fsname-OST0000
2291 if [ $fstype == "ldiskfs" ]; then
2294 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2295 error_noexit "Remounting the MDT"
2298 shall_cleanup_mdt=true
2308 for tarball in $tarballs; do
2309 t32_test $tarball || let "rc += $?"
2313 run_test 32a "Upgrade (not live)"
2321 for tarball in $tarballs; do
2322 t32_test $tarball writeconf || let "rc += $?"
2326 run_test 32b "Upgrade with writeconf"
2333 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
2335 for tarball in $tarballs; do
2336 # Do not support 1_8 and 2_1 direct upgrade to DNE2 anymore */
2337 echo $tarball | grep "1_8" && continue
2338 echo $tarball | grep "2_1" && continue
2340 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
2344 run_test 32c "dne upgrade test"
2352 for tarball in $tarballs; do
2353 ff_convert=yes t32_test $tarball || rc=$?
2357 run_test 32d "convert ff test"
2359 test_33a() { # bug 12333, was test_33
2360 local FSNAME2=test-123
2361 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2364 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
2366 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
2367 local dev=${SINGLEMDS}_dev
2368 local MDSDEV=${!dev}
2369 is_blkdev $SINGLEMDS $MDSDEV &&
2370 skip_env "mixed loopback and real device not working" &&
2374 local fs2mdsdev=$(mdsdevname 1_2)
2375 local fs2ostdev=$(ostdevname 1_2)
2376 local fs2mdsvdev=$(mdsvdevname 1_2)
2377 local fs2ostvdev=$(ostvdevname 1_2)
2379 if [ $(facet_fstype mds1) == ldiskfs ]; then
2380 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
2383 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2384 --reformat $mkfsoptions $fs2mdsdev $fs2mdsvdev || exit 10
2385 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2386 --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \
2387 $fs2ostvdev || exit 10
2389 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
2390 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2391 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" ||
2392 error "$LCTL conf_param $FSNAME2.sys.timeout=200 failed"
2393 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2394 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2397 cp /etc/hosts $MOUNT2/ || error "copy /etc/hosts $MOUNT2/ failed"
2398 $GETSTRIPE $MOUNT2/hosts || error "$GETSTRIPE $MOUNT2/hosts failed"
2403 cleanup_nocli || error "cleanup_nocli failed with $?"
2405 run_test 33a "Mount ost with a large index number"
2407 test_33b() { # was test_34
2410 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
2411 # Drop lock cancelation reply during umount
2412 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
2413 do_facet client $LCTL set_param fail_loc=0x80000304
2414 #lctl set_param debug=-1
2415 umount_client $MOUNT
2416 cleanup || error "cleanup failed with $?"
2418 run_test 33b "Drop cancel during umount"
2422 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
2423 manual_umount_client
2425 do_facet client killall -USR1 multiop
2426 if [ $rc -eq 0 ]; then
2427 error "umount not fail!"
2430 cleanup || error "cleanup failed with rc $?"
2432 run_test 34a "umount with opened file should be fail"
2436 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2437 stop_mds || error "Unable to stop MDS"
2439 manual_umount_client --force || error "mtab after failed umount with $?"
2441 cleanup || error "cleanup failed with $?"
2443 run_test 34b "force umount with failed mds should be normal"
2447 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2448 stop_ost || error "Unable to stop OST1"
2450 manual_umount_client --force || error "mtab after failed umount with $?"
2452 cleanup || error "cleanup failed with $?"
2454 run_test 34c "force umount with failed ost should be normal"
2456 test_35a() { # bug 12459
2459 DBG_SAVE="`$LCTL get_param -n debug`"
2460 $LCTL set_param debug="ha"
2462 log "Set up a fake failnode for the MDS"
2464 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2465 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2466 do_facet mgs "$LCTL conf_param \
2467 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" ||
2468 error "Setting ${device}.failover.node=\
2469 $(h2$NETTYPE $FAKENID) failed."
2471 log "Wait for RECONNECT_INTERVAL seconds (10s)"
2474 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
2477 log "Stopping the MDT: $device"
2478 stop_mdt 1 || error "MDT0 stop fail"
2480 df $MOUNT > /dev/null 2>&1 &
2482 log "Restarting the MDT: $device"
2483 start_mdt 1 || error "MDT0 start fail"
2484 log "Wait for df ($DFPID) ... "
2487 $LCTL set_param debug="$DBG_SAVE"
2489 # retrieve from the log the first server that the client tried to
2490 # contact after the connection loss
2491 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2492 NEXTCONN=`awk "/${MSG}/ {start = 1;}
2493 /import_select_connection.*$device-mdc.* using connection/ {
2495 if (\\\$NF ~ /$FAKENID/)
2501 }" $TMP/lustre-log-$TESTNAME.log`
2502 [ "$NEXTCONN" != "0" ] &&
2503 error "Tried to connect to ${NEXTCONN} not last active server"
2504 cleanup || error "cleanup failed with $?"
2505 # remove nid settings
2506 writeconf_or_reformat
2508 run_test 35a "Reconnect to the last active server first"
2510 test_35b() { # bug 18674
2511 remote_mds || { skip "local MDS" && return 0; }
2515 $LCTL set_param debug="ha"
2517 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
2520 log "Set up a fake failnode for the MDS"
2522 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2523 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2524 do_facet mgs "$LCTL conf_param \
2525 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" ||
2526 error "Set ${device}.failover.node=\
2527 $(h2$NETTYPE $FAKENID) failed"
2529 local at_max_saved=0
2530 # adaptive timeouts may prevent seeing the issue
2531 if at_is_enabled; then
2532 at_max_saved=$(at_max_get mds)
2533 at_max_set 0 mds client
2536 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
2538 log "Injecting EBUSY on MDS"
2539 # Setting OBD_FAIL_MDS_RESEND=0x136
2540 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" ||
2541 error "unable to set param fail_loc=0x80000136"
2543 $LCTL set_param mdc.${FSNAME}*.stats=clear
2545 log "Creating a test file and stat it"
2546 touch $MOUNT/$tdir/$tfile || error "touch $MOUNT/$tdir/$tfile failed"
2547 stat $MOUNT/$tdir/$tfile
2549 log "Stop injecting EBUSY on MDS"
2550 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" ||
2551 error "unable to set param fail_loc=0"
2552 rm -f $MOUNT/$tdir/$tfile || error "remove $MOUNT/$tdir/$tfile failed"
2555 # restore adaptive timeout
2556 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
2558 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2560 CONNCNT=$($LCTL get_param mdc.${FSNAME}*.stats |
2561 awk '/mds_connect/{print $2}')
2563 # retrieve from the log if the client has ever tried to
2564 # contact the fake server after the loss of connection
2565 FAILCONN=`awk "BEGIN {ret = 0;}
2566 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
2568 if (\\\$NF ~ /$FAKENID/) {
2573 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
2575 [ "$FAILCONN" == "0" ] &&
2576 error "The client reconnection has not been triggered"
2577 [ "$FAILCONN" == "2" ] &&
2578 error "Primary server busy, client reconnect to failover failed"
2581 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
2582 # Reconnects are supposed to be rate limited to one every 5s
2583 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] &&
2584 error "Too many reconnects $CONNCNT"
2586 cleanup || error "cleanup failed with $?"
2587 # remove nid settings
2588 writeconf_or_reformat
2590 run_test 35b "Continue reconnection retries, if the active server is busy"
2593 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
2595 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] ||
2596 { skip "remote OST" && return 0; }
2599 local FSNAME2=test1234
2600 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2602 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
2604 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
2605 is_blkdev $SINGLEMDS $MDSDEV &&
2606 skip_env "mixed loopback and real device not working" && return
2609 local fs2mdsdev=$(mdsdevname 1_2)
2610 local fs2ostdev=$(ostdevname 1_2)
2611 local fs3ostdev=$(ostdevname 2_2)
2612 local fs2mdsvdev=$(mdsvdevname 1_2)
2613 local fs2ostvdev=$(ostvdevname 1_2)
2614 local fs3ostvdev=$(ostvdevname 2_2)
2616 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2617 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
2618 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
2619 # different one than the default value here.
2620 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2621 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
2622 add fs3ost $(mkfs_opts ost2 ${fs3ostdev}) --mgsnode=$MGSNID \
2623 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
2625 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
2626 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2627 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
2628 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2629 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2631 sleep 5 # until 11778 fixed
2633 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || error "dd failed"
2635 BKTOTAL=$($LCTL get_param -n obdfilter.*.kbytestotal |
2636 awk 'BEGIN{total=0}; {total+=$1}; END{print total}')
2637 BKFREE=$($LCTL get_param -n obdfilter.*.kbytesfree |
2638 awk 'BEGIN{free=0}; {free+=$1}; END{print free}')
2639 BKAVAIL=$($LCTL get_param -n obdfilter.*.kbytesavail |
2640 awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}')
2641 STRING=$(df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}')
2642 DFTOTAL=$(echo $STRING | cut -d, -f1)
2643 DFUSED=$(echo $STRING | cut -d, -f2)
2644 DFAVAIL=$(echo $STRING | cut -d, -f3)
2645 DFFREE=$(($DFTOTAL - $DFUSED))
2647 ALLOWANCE=$((64 * $OSTCOUNT))
2649 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
2650 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
2651 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
2654 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
2655 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
2656 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
2659 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
2660 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
2661 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
2666 stop fs3ost -f || error "unable to stop OST3"
2667 stop fs2ost -f || error "unable to stop OST2"
2668 stop fs2mds -f || error "unable to stop second MDS"
2669 unload_modules_conf || error "unable unload modules"
2672 run_test 36 "df report consistency on OSTs with different block size"
2675 local mntpt=$(facet_mntpt $SINGLEMDS)
2676 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
2677 local mdsdev_sym="$TMP/sym_mdt.img"
2678 local opts=$MDS_MOUNT_OPTS
2681 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2682 skip "Currently only applicable to ldiskfs-based MDTs"
2686 echo "MDS : $mdsdev"
2687 echo "SYMLINK : $mdsdev_sym"
2688 do_facet $SINGLEMDS rm -f $mdsdev_sym
2690 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
2692 echo "mount symlink device - $mdsdev_sym"
2694 if ! do_facet $SINGLEMDS test -b $mdsdev; then
2695 opts=$(csa_add "$opts" -o loop)
2697 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
2698 $mdsdev_sym $mntpt 2>&1)
2701 echo mount_op=$mount_op
2703 do_facet $SINGLEMDS "$UMOUNT $mntpt && rm -f $mdsdev_sym"
2705 if $(echo $mount_op | grep -q "unable to set tunable"); then
2706 error "set tunables failed for symlink device"
2709 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
2711 run_test 37 "verify set tunables works for symlink device"
2713 test_38() { # bug 14222
2714 local fstype=$(facet_fstype $SINGLEMDS)
2715 local mntpt=$(facet_mntpt $SINGLEMDS)
2720 local SRC="/etc /bin"
2721 local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT)
2722 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
2723 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2724 tar cf - $FILES | tar xf - -C $DIR/$tdir ||
2725 error "copying $SRC to $DIR/$tdir"
2727 umount_client $MOUNT || error "umount_client $MOUNT failed"
2728 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2729 stop_mds || error "Unable to stop MDS"
2730 log "delete lov_objid file on MDS"
2732 mount_fstype $SINGLEMDS || error "mount MDS failed (1)"
2734 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid; rm $mntpt/lov_objid"
2736 unmount_fstype $SINGLEMDS || error "umount failed (1)"
2738 # check create in mds_lov_connect
2739 start_mds || error "unable to start MDS"
2740 mount_client $MOUNT || error "mount_client $MOUNT failed"
2742 [ $V ] && log "verifying $DIR/$tdir/$f"
2743 diff -q $f $DIR/$tdir/$f || ERROR=y
2745 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2746 if [ "$ERROR" = "y" ]; then
2747 # check it's updates in sync
2748 umount_client $MOUNT
2750 mount_fstype $SIGNLEMDS
2751 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2752 unmount_fstype $SINGLEMDS
2753 error "old and new files are different after connect" || true
2755 touch $DIR/$tdir/f2 || error "f2 file create failed"
2757 # check it's updates in sync
2758 umount_client $MOUNT || error "second umount_client $MOUNT failed"
2761 mount_fstype $SINGLEMDS || error "mount MDS failed (3)"
2763 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2764 do_facet $SINGLEMDS dd if=/dev/zero of=$mntpt/lov_objid.clear count=8
2766 unmount_fstype $SINGLEMDS || error "umount failed (3)"
2768 start_mds || error "unable to start MDS"
2769 mount_client $MOUNT || error "mount_client $MOUNT failed"
2771 [ $V ] && log "verifying $DIR/$tdir/$f"
2772 diff -q $f $DIR/$tdir/$f || ERROR=y
2774 touch $DIR/$tdir/f3 || error "f3 file create failed"
2775 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2776 umount_client $MOUNT || error "third umount_client $MOUNT failed"
2778 mount_fstype $SINGLEMDS || error "mount MDS failed (4)"
2779 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2780 unmount_fstype $SINGLEMDS || error "umount failed (4)"
2782 [ "$ERROR" = "y" ] &&
2783 error "old and new files are different after sync" || true
2785 log "files compared the same"
2786 cleanup || error "cleanup failed with $?"
2788 run_test 38 "MDS recreates missing lov_objid file from OST data"
2793 cleanup || error "cleanup failed with $?"
2794 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
2795 error "memory leak detected" || true
2797 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
2799 test_40() { # bug 15759
2800 start_ost || error "Unable to start OST1"
2801 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
2802 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
2804 cleanup || error "cleanup failed with rc $?"
2806 run_test 40 "race during service thread startup"
2808 test_41a() { #bug 14134
2809 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2810 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2811 skip "Loop devices does not work with nosvc option"
2816 { skip "needs combined MGT and MDT device" && return 0; }
2818 start_mdt 1 -o nosvc -n
2819 if [ $MDSCOUNT -ge 2 ]; then
2820 for num in $(seq 2 $MDSCOUNT); do
2821 start_mdt $num || return
2824 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
2825 start_mdt 1 -o nomgs,force
2826 mount_client $MOUNT || error "mount_client $MOUNT failed"
2829 echo "blah blah" > $MOUNT/$tfile
2832 umount_client $MOUNT || error "umount_client $MOUNT failed"
2833 stop ost1 -f || error "unable to stop OST1"
2834 stop_mds || error "Unable to stop MDS"
2835 stop_mds || error "Unable to stop MDS on second try"
2837 run_test 41a "mount mds with --nosvc and --nomgs"
2840 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2841 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2842 skip "Loop devices does not work with nosvc option"
2846 ! combined_mgs_mds && skip "needs combined mgs device" && return 0
2850 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2852 start_mdt 1 -o nosvc -n
2853 if [ $MDSCOUNT -ge 2 ]; then
2854 for num in $(seq 2 $MDSCOUNT); do
2855 start_mdt $num || return
2858 start_ost || error "Unable to start OST1"
2859 start_mdt 1 -o nomgs,force
2860 mount_client $MOUNT || error "mount_client $MOUNT failed"
2863 echo "blah blah" > $MOUNT/$tfile
2864 cat $MOUNT/$tfile || error "cat $MOUNT/$tfile failed"
2866 umount_client $MOUNT || error "umount_client $MOUNT failed"
2867 stop_ost || error "Unable to stop OST1"
2868 stop_mds || error "Unable to stop MDS"
2869 stop_mds || error "Unable to stop MDS on second try"
2871 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
2874 local server_version=$(lustre_version_code $SINGLEMDS)
2875 local oss_list=$(comma_list $(osts_nodes))
2877 [[ $server_version -ge $(version_code 2.6.52) ]] ||
2878 [[ $server_version -ge $(version_code 2.5.26) &&
2879 $server_version -lt $(version_code 2.5.50) ]] ||
2880 [[ $server_version -ge $(version_code 2.5.4) &&
2881 $server_version -lt $(version_code 2.5.11) ]] ||
2882 { skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"; return; }
2884 # ensure mds1 ost1 have been created even if running sub-test standalone
2887 cleanup || error "cleanup failed"
2889 # using directly mount command instead of start() function to avoid
2890 # any side effect of // with others/externals tools/features
2891 # ("zpool import", ...)
2893 # MDT concurrent start
2895 LOAD_MODULES_REMOTE=true load_modules
2896 do_facet $SINGLEMDS "lsmod | grep -q libcfs" ||
2897 error "MDT concurrent start: libcfs module not loaded"
2899 local mds1dev=$(mdsdevname 1)
2900 local mds1mnt=$(facet_mntpt mds1)
2901 local mds1fstype=$(facet_fstype mds1)
2902 local mds1opts=$MDS_MOUNT_OPTS
2904 if [ $mds1fstype == ldiskfs ] &&
2905 ! do_facet mds1 test -b $mds1dev; then
2906 mds1opts=$(csa_add "$mds1opts" -o loop)
2908 if [[ $mds1fstype == zfs ]]; then
2909 import_zpool mds1 || return ${PIPESTATUS[0]}
2912 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
2913 do_facet mds1 "$LCTL set_param fail_loc=0x80000716"
2915 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts &
2918 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts
2922 do_facet mds1 "$LCTL set_param fail_loc=0x0"
2923 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
2924 echo "1st MDT start succeed"
2925 echo "2nd MDT start failed with $rc2"
2926 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
2927 echo "1st MDT start failed with $rc"
2928 echo "2nd MDT start succeed"
2931 error "unexpected concurrent MDT mounts result, rc=$rc rc2=$rc2"
2934 if [ $MDSCOUNT -ge 2 ]; then
2935 for num in $(seq 2 $MDSCOUNT); do
2936 start_mdt $num || return
2940 # OST concurrent start
2942 do_rpc_nodes $oss_list "lsmod | grep -q libcfs" ||
2943 error "OST concurrent start: libcfs module not loaded"
2945 local ost1dev=$(ostdevname 1)
2946 local ost1mnt=$(facet_mntpt ost1)
2947 local ost1fstype=$(facet_fstype ost1)
2948 local ost1opts=$OST_MOUNT_OPTS
2950 if [ $ost1fstype == ldiskfs ] &&
2951 ! do_facet ost1 test -b $ost1dev; then
2952 ost1opts=$(csa_add "$ost1opts" -o loop)
2954 if [[ $ost1fstype == zfs ]]; then
2955 import_zpool ost1 || return ${PIPESTATUS[0]}
2958 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
2959 do_facet ost1 "$LCTL set_param fail_loc=0x80000716"
2961 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts &
2964 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts
2968 do_facet ost1 "$LCTL set_param fail_loc=0x0"
2969 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
2970 echo "1st OST start succeed"
2971 echo "2nd OST start failed with $rc2"
2972 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
2973 echo "1st OST start failed with $rc"
2974 echo "2nd OST start succeed"
2978 error "unexpected concurrent OST mounts result, rc=$rc rc2=$rc2"
2984 # verify everything ok
2989 error "MDT(s) start failed"
2997 error "OST(s) start failed"
3005 error "client start failed"
3012 error "client mount failed"
3016 run_test 41c "concurrent mounts of MDT/OST should all fail but one"
3018 test_42() { #bug 14693
3020 check_mount || error "client was not mounted"
3022 do_facet mgs $LCTL conf_param $FSNAME.llite.some_wrong_param=10
3023 umount_client $MOUNT ||
3024 error "unmounting client failed with invalid llite param"
3025 mount_client $MOUNT ||
3026 error "mounting client failed with invalid llite param"
3028 do_facet mgs $LCTL conf_param $FSNAME.sys.some_wrong_param=20
3029 cleanup || error "stopping $FSNAME failed with invalid sys param"
3031 check_mount || error "client was not mounted with invalid sys param"
3032 cleanup || error "stopping $FSNAME failed with invalid sys param"
3034 run_test 42 "allow client/server mount/unmount with invalid config param"
3037 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.58) ]] ||
3038 { skip "Need MDS version at least 2.5.58" && return 0; }
3039 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
3042 USER1=$(getent passwd | grep :$ID1:$ID1: | cut -d: -f1)
3043 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1" &&
3047 chmod ugo+x $DIR || error "chmod 0 failed"
3048 set_conf_param_and_check mds1 \
3049 "$LCTL get_param -n mdt.$FSNAME-MDT0000.root_squash" \
3050 "$FSNAME.mdt.root_squash" \
3052 wait_update $HOSTNAME \
3053 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3055 error "check llite root_squash failed!"
3056 set_conf_param_and_check mds1 \
3057 "$LCTL get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
3058 "$FSNAME.mdt.nosquash_nids" \
3060 wait_update $HOSTNAME \
3061 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3063 error "check llite nosquash_nids failed!"
3066 # create set of test files
3068 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
3069 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
3070 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
3072 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
3073 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
3075 mkdir $DIR/$tdir-rootdir || error "mkdir failed"
3076 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
3077 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
3079 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
3080 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
3081 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
3084 # check root_squash:
3085 # set root squash UID:GID to RUNAS_ID
3086 # root should be able to access only files owned by RUNAS_ID
3088 set_conf_param_and_check mds1 \
3089 "$LCTL get_param -n mdt.$FSNAME-MDT0000.root_squash" \
3090 "$FSNAME.mdt.root_squash" \
3091 "$RUNAS_ID:$RUNAS_ID"
3092 wait_update $HOSTNAME \
3093 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3094 "$RUNAS_ID:$RUNAS_ID" ||
3095 error "check llite root_squash failed!"
3097 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
3098 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3099 error "$ST: root read permission is denied"
3100 echo "$ST: root read permission is granted - ok"
3103 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3104 error "$ST: root write permission is denied"
3105 echo "$ST: root write permission is granted - ok"
3107 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3108 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3109 error "$ST: root read permission is granted"
3110 echo "$ST: root read permission is denied - ok"
3113 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3114 error "$ST: root write permission is granted"
3115 echo "$ST: root write permission is denied - ok"
3117 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3118 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null &&
3119 error "$ST: root unlink permission is granted"
3120 echo "$ST: root unlink permission is denied - ok"
3122 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null &&
3123 error "$ST: root create permission is granted"
3124 echo "$ST: root create permission is denied - ok"
3127 # check root_squash is enforced independently
3128 # of client cache content
3130 # access file by USER1, keep access open
3131 # root should be denied access to user file
3133 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
3137 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
3138 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3139 { kill $pid; error "$ST: root read permission is granted"; }
3140 echo "$ST: root read permission is denied - ok"
3143 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3144 { kill $pid; error "$ST: root write permission is granted"; }
3145 echo "$ST: root write permission is denied - ok"
3151 # check nosquash_nids:
3152 # put client's NID into nosquash_nids list,
3153 # root should be able to access root file after that
3155 local NIDLIST=$($LCTL list_nids all | tr '\n' ' ')
3156 NIDLIST="2@gni $NIDLIST 192.168.0.[2,10]@tcp"
3157 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
3158 set_conf_param_and_check mds1 \
3159 "$LCTL get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
3160 "$FSNAME-MDTall.mdt.nosquash_nids" \
3162 wait_update $HOSTNAME \
3163 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3165 error "check llite nosquash_nids failed!"
3167 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3168 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3169 error "$ST: root read permission is denied"
3170 echo "$ST: root read permission is granted - ok"
3173 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3174 error "$ST: root write permission is denied"
3175 echo "$ST: root write permission is granted - ok"
3177 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3178 rm $DIR/$tdir-rootdir/tfile-1 ||
3179 error "$ST: root unlink permission is denied"
3180 echo "$ST: root unlink permission is granted - ok"
3181 touch $DIR/$tdir-rootdir/tfile-2 ||
3182 error "$ST: root create permission is denied"
3183 echo "$ST: root create permission is granted - ok"
3184 cleanup || error "cleanup failed with $?"
3186 run_test 43a "check root_squash and nosquash_nids"
3188 test_43b() { # LU-5690
3189 [[ $(lustre_version_code mgs) -ge $(version_code 2.7.62) ]] ||
3190 { skip "Need MGS version 2.7.62+"; return; }
3192 if [[ -z "$fs2mds_DEV" ]]; then
3193 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
3194 skip_env "mixed loopback and real device not working" && return
3197 local fs2mdsdev=$(mdsdevname 1_2)
3198 local fs2mdsvdev=$(mdsvdevname 1_2)
3200 # temporarily use fs2mds as fs2mgs
3202 local fs2mgsdev=$fs2mdsdev
3203 local fs2mgsvdev=$fs2mdsvdev
3205 local fsname=test1234
3207 load_module llite/lustre
3208 local client_ip=$(host_nids_address $HOSTNAME $NETTYPE)
3209 local host=${client_ip//*./}
3210 local net=${client_ip/%$host/}
3211 local nosquash_nids=$(h2$NETTYPE $net[$host,$host,$host])
3213 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
3214 --param mdt.root_squash=$RUNAS_ID:$RUNAS_ID \
3215 --param mdt.nosquash_nids=$nosquash_nids \
3216 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
3217 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS || error "start fs2mgs failed"
3218 stop $fs2mgs -f || error "stop fs2mgs failed"
3220 run_test 43b "parse nosquash_nids with commas in expr_list"
3222 umount_client $MOUNT
3227 check_mount || error "check_mount"
3228 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
3230 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
3231 for VAL in $UUIDS; do
3232 NID=$(echo $VAL | cut -d= -f1)
3233 CLUUID=$(echo $VAL | cut -d= -f2)
3234 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
3236 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
3237 cleanup || error "cleanup failed with $?"
3239 run_test 44 "mounted client proc entry exists"
3243 check_mount || error "check_mount"
3244 stop_mds || error "Unable to stop MDS"
3248 #define OBD_FAIL_PTLRPC_LONG_REPL_UNLINK 0x50f
3249 do_facet client "$LCTL set_param fail_loc=0x8000050f"
3252 manual_umount_client --force || error "manual_umount_client failed"
3253 do_facet client "$LCTL set_param fail_loc=0x0"
3254 start_mds || error "unable to start MDS"
3255 mount_client $MOUNT || error "mount_client $MOUNT failed"
3256 cleanup || error "cleanup failed with $?"
3258 run_test 45 "long unlink handling in ptlrpcd"
3265 umount_client $MOUNT2 || rc=$?
3266 umount_client $MOUNT || rc=$?
3267 while [ $count -gt 0 ]; do
3268 stop ost${count} -f || rc=$?
3272 cleanup_nocli || rc=$?
3273 #writeconf to remove all ost2 traces for subsequent tests
3274 writeconf_or_reformat
3279 echo "Testing with $OSTCOUNT OSTs"
3281 start_mds || error "unable to start MDS"
3282 #first client should see only one ost
3283 start_ost || error "Unable to start OST1"
3284 wait_osc_import_state mds ost FULL
3286 mount_client $MOUNT || error "mount_client $MOUNT failed"
3287 trap "cleanup_46a $OSTCOUNT" EXIT ERR
3290 for (( i=2; i<=$OSTCOUNT; i++ )); do
3291 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
3292 error "start_ost$i $(ostdevname $i) failed"
3295 # wait until osts in sync
3296 for (( i=2; i<=$OSTCOUNT; i++ )); do
3297 wait_osc_import_state mds ost$i FULL
3298 wait_osc_import_state client ost$i FULL
3301 #second client see all ost's
3303 mount_client $MOUNT2 || error "mount_client failed"
3304 $SETSTRIPE -c -1 $MOUNT2 || error "$SETSTRIPE -c -1 $MOUNT2 failed"
3305 $GETSTRIPE $MOUNT2 || error "$GETSTRIPE $MOUNT2 failed"
3307 echo "ok" > $MOUNT2/widestripe
3308 $GETSTRIPE $MOUNT2/widestripe ||
3309 error "$GETSTRIPE $MOUNT2/widestripe failed"
3310 # fill acl buffer for avoid expand lsm to them
3311 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd |
3313 setfacl -m $acl $MOUNT2/widestripe
3317 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3319 cleanup_46a $OSTCOUNT || error "cleanup_46a failed"
3321 run_test 46a "handle ost additional - wide striped file"
3326 check_mount || error "check_mount failed"
3327 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
3331 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3332 if echo $ns | grep "MDT[[:digit:]]*"; then
3335 lrs=$(echo $ns | sed 's/.*lru_size=//')
3336 lru_size[count]=$lrs
3341 facet_failover $SINGLEMDS
3342 client_up || error "client_up failed"
3345 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3346 if echo $ns | grep "MDT[[:digit:]]*"; then
3349 lrs=$(echo $ns | sed 's/.*lru_size=//')
3350 if ! test "$lrs" -eq "${lru_size[count]}"; then
3351 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
3352 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
3357 cleanup || error "cleanup failed with $?"
3359 run_test 47 "server restart does not make client loss lru_resize settings"
3364 # reformat after this test is needed - if the test fails,
3365 # we will have unkillable file at FS
3369 test_48() { # bug 17636
3372 check_mount || error "check_mount failed"
3374 $SETSTRIPE -c -1 $MOUNT || error "$SETSTRIPE -c -1 $MOUNT failed"
3375 $GETSTRIPE $MOUNT || error "$GETSTRIPE $MOUNT failed"
3377 echo "ok" > $MOUNT/widestripe
3378 $GETSTRIPE $MOUNT/widestripe ||
3379 error "$GETSTRIPE $MOUNT/widestripe failed"
3381 trap cleanup_48 EXIT ERR
3383 # fill acl buffer for avoid expand lsm to them
3384 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
3385 setfacl -m $acl $MOUNT/widestripe
3388 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3392 run_test 48 "too many acls on file"
3394 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
3395 test_49a() { # bug 17710
3396 local timeout_orig=$TIMEOUT
3397 local ldlm_timeout_orig=$LDLM_TIMEOUT
3398 local LOCAL_TIMEOUT=20
3400 LDLM_TIMEOUT=$LOCAL_TIMEOUT
3401 TIMEOUT=$LOCAL_TIMEOUT
3405 check_mount || error "client mount failed"
3407 echo "check ldlm_timout..."
3408 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3409 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3410 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3412 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3413 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3416 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
3417 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
3420 umount_client $MOUNT || error "umount_client $MOUNT failed"
3421 stop_ost || error "problem stopping OSS"
3422 stop_mds || error "problem stopping MDS"
3424 LDLM_TIMEOUT=$ldlm_timeout_orig
3425 TIMEOUT=$timeout_orig
3427 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3429 test_49b() { # bug 17710
3430 local timeout_orig=$TIMEOUT
3431 local ldlm_timeout_orig=$LDLM_TIMEOUT
3432 local LOCAL_TIMEOUT=20
3434 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
3435 TIMEOUT=$LOCAL_TIMEOUT
3439 check_mount || error "client mount failed"
3441 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3442 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3443 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3445 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3446 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3449 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
3450 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
3453 cleanup || error "cleanup failed"
3455 LDLM_TIMEOUT=$ldlm_timeout_orig
3456 TIMEOUT=$timeout_orig
3458 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3461 # Test both statfs and lfs df and fail if either one fails
3462 multiop_bg_pause $1 f_
3465 killall -USR1 multiop
3466 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
3467 wait $PID || { RC1=$?; log "multiop return error "; }
3474 if [ $RC2 -eq 0 ]; then
3476 log "lazystatfs df failed"
3480 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
3486 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3487 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3489 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
3491 cleanup || error "cleanup failed with rc $?"
3493 run_test 50a "lazystatfs all servers available"
3497 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3498 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3500 # Wait for client to detect down OST
3501 stop_ost || error "Unable to stop OST1"
3502 wait_osc_import_state mds ost DISCONN
3504 lazystatfs $MOUNT || error "lazystatfs should not return EIO"
3506 umount_client $MOUNT || error "Unable to unmount client"
3507 stop_mds || error "Unable to stop MDS"
3509 run_test 50b "lazystatfs all servers down"
3512 start_mds || error "Unable to start MDS"
3513 start_ost || error "Unable to start OST1"
3514 start_ost2 || error "Unable to start OST2"
3515 mount_client $MOUNT || error "Unable to mount client"
3516 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3517 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3519 # Wait for client to detect down OST
3520 stop_ost || error "Unable to stop OST1"
3521 wait_osc_import_state mds ost DISCONN
3522 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3524 umount_client $MOUNT || error "Unable to unmount client"
3525 stop_ost2 || error "Unable to stop OST2"
3526 stop_mds || error "Unable to stop MDS"
3527 #writeconf to remove all ost2 traces for subsequent tests
3528 writeconf_or_reformat
3530 run_test 50c "lazystatfs one server down"
3533 start_mds || error "Unable to start MDS"
3534 start_ost || error "Unable to start OST1"
3535 start_ost2 || error "Unable to start OST2"
3536 mount_client $MOUNT || error "Unable to mount client"
3537 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3538 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3540 # Issue the statfs during the window where the client still
3541 # belives the OST to be available but it is in fact down.
3542 # No failure just a statfs which hangs for a timeout interval.
3543 stop_ost || error "Unable to stop OST1"
3544 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3546 umount_client $MOUNT || error "Unable to unmount client"
3547 stop_ost2 || error "Unable to stop OST2"
3548 stop_mds || error "Unable to stop MDS"
3549 #writeconf to remove all ost2 traces for subsequent tests
3550 writeconf_or_reformat
3552 run_test 50d "lazystatfs client/server conn race"
3559 start_mds || error "Unable to start MDS"
3560 #first client should see only one ost
3561 start_ost || error "Unable to start OST1"
3562 wait_osc_import_state mds ost FULL
3564 # Wait for client to detect down OST
3565 stop_ost || error "Unable to stop OST1"
3566 wait_osc_import_state mds ost DISCONN
3568 mount_client $MOUNT || error "Unable to mount client"
3569 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3571 multiop_bg_pause $MOUNT _f
3575 if [ $RC1 -ne 0 ]; then
3576 log "multiop failed $RC1"
3579 sleep $(( $TIMEOUT+1 ))
3581 [ $? -ne 0 ] && error "process isn't sleep"
3582 start_ost || error "Unable to start OST1"
3583 wait $pid || error "statfs failed"
3586 umount_client $MOUNT || error "Unable to unmount client"
3587 stop_ost || error "Unable to stop OST1"
3588 stop_mds || error "Unable to stop MDS"
3590 run_test 50e "normal statfs all servers down"
3595 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
3597 start_mds || error "Unable to start mds"
3598 #first client should see only one ost
3599 start_ost || error "Unable to start OST1"
3600 wait_osc_import_state mds ost FULL
3602 start_ost2 || error "Unable to start OST2"
3603 wait_osc_import_state mds ost2 FULL
3605 # Wait for client to detect down OST
3606 stop_ost2 || error "Unable to stop OST2"
3608 wait_osc_import_state mds ost2 DISCONN
3609 mount_client $MOUNT || error "Unable to mount client"
3610 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3612 multiop_bg_pause $MOUNT _f
3616 if [ $RC1 -ne 0 ]; then
3617 log "lazystatfs multiop failed $RC1"
3620 sleep $(( $TIMEOUT+1 ))
3622 [ $? -ne 0 ] && error "process isn't sleep"
3623 start_ost2 || error "Unable to start OST2"
3624 wait $pid || error "statfs failed"
3625 stop_ost2 || error "Unable to stop OST2"
3628 umount_client $MOUNT -f || error "Unable to unmount client"
3629 stop_ost || error "Unable to stop OST1"
3630 stop_mds || error "Unable to stop MDS"
3631 #writeconf to remove all ost2 traces for subsequent tests
3632 writeconf_or_reformat
3634 run_test 50f "normal statfs one server in down"
3637 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3639 start_ost2 || error "Unable to start OST2"
3640 wait_osc_import_state mds ost2 FULL
3641 wait_osc_import_state client ost2 FULL
3643 local PARAM="${FSNAME}-OST0001.osc.active"
3645 $SETSTRIPE -c -1 $DIR/$tfile || error "$SETSTRIPE failed"
3646 do_facet mgs $LCTL conf_param $PARAM=0 ||
3647 error "Unable to deactivate OST"
3649 umount_client $MOUNT || error "Unable to unmount client"
3650 mount_client $MOUNT || error "Unable to mount client"
3651 # This df should not cause a panic
3654 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
3655 rm -f $DIR/$tfile || error "unable to remove file $DIR/$tfile"
3656 umount_client $MOUNT || error "Unable to unmount client"
3657 stop_ost2 || error "Unable to stop OST2"
3658 stop_ost || error "Unable to stop OST1"
3659 stop_mds || error "Unable to stop MDS"
3660 #writeconf to remove all ost2 traces for subsequent tests
3661 writeconf_or_reformat
3663 run_test 50g "deactivated OST should not cause panic"
3667 # prepare MDT/OST, make OSC inactive for OST1
3668 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3670 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3671 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
3672 error "tunefs OST1 failed"
3673 start_mds || error "Unable to start MDT"
3674 start_ost || error "Unable to start OST1"
3675 start_ost2 || error "Unable to start OST2"
3676 mount_client $MOUNT || error "client start failed"
3678 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3680 # activatate OSC for OST1
3681 local TEST="$LCTL get_param -n osc.${FSNAME}-OST0000-osc-[!M]*.active"
3682 set_conf_param_and_check client \
3683 "$TEST" "${FSNAME}-OST0000.osc.active" 1 ||
3684 error "Unable to activate OST1"
3686 mkdir $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3687 $SETSTRIPE -c -1 -i 0 $DIR/$tdir/2 ||
3688 error "$SETSTRIPE $DIR/$tdir/2 failed"
3689 sleep 1 && echo "create a file after OST1 is activated"
3691 createmany -o $DIR/$tdir/2/$tfile-%d 1
3693 # check OSC import is working
3694 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
3695 error "some OSC imports are still not connected"
3698 umount_client $MOUNT || error "Unable to umount client"
3699 stop_ost2 || error "Unable to stop OST2"
3700 cleanup_nocli || error "cleanup_nocli failed with $?"
3702 run_test 50h "LU-642: activate deactivated OST"
3705 # prepare MDT/OST, make OSC inactive for OST1
3706 [ "$MDSCOUNT" -lt "2" ] && skip_env "$MDSCOUNT < 2, skipping" && return
3709 [ $(facet_fstype mds2) == zfs ] && import_zpool mds2
3710 do_facet mds2 "$TUNEFS --param mdc.active=0 $(mdsdevname 2)" ||
3711 error "tunefs MDT2 failed"
3712 start_mds || error "Unable to start MDT"
3713 start_ost || error "Unable to start OST1"
3714 start_ost2 || error "Unable to start OST2"
3715 mount_client $MOUNT || error "client start failed"
3717 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3719 $LCTL conf_param ${FSNAME}-MDT0000.mdc.active=0 &&
3720 error "deactive MDC0 succeeds"
3721 # activate MDC for MDT2
3722 local TEST="$LCTL get_param -n mdc.${FSNAME}-MDT0001-mdc-[!M]*.active"
3723 set_conf_param_and_check client \
3724 "$TEST" "${FSNAME}-MDT0001.mdc.active" 1 ||
3725 error "Unable to activate MDT2"
3727 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
3728 if [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.60) ]
3730 wait_dne_interconnect
3732 $LFS mkdir -i1 $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3734 createmany -o $DIR/$tdir/2/$tfile-%d 1 || error "create files failed"
3736 rm -rf $DIR/$tdir/2 || error "unlink dir failed"
3738 # deactivate MDC for MDT2
3739 TEST="$LCTL get_param -n mdc.${FSNAME}-MDT0001-mdc-[!M]*.active"
3740 set_conf_param_and_check client \
3741 "$TEST" "${FSNAME}-MDT0001.mdc.active" 0 ||
3742 error "Unable to deactivate MDT2"
3744 $LFS mkdir -i1 $DIR/$tdir/2 &&
3745 error "mkdir $DIR/$tdir/2 succeeds after deactive MDT"
3747 $LFS mkdir -i0 -c$MDSCOUNT $DIR/$tdir/striped_dir ||
3748 error "mkdir $DIR/$tdir/striped_dir fails after deactive MDT2"
3750 local stripe_count=$($LFS getdirstripe -c $DIR/$tdir/striped_dir)
3751 [ $stripe_count -eq $((MDSCOUNT - 1)) ] ||
3752 error "wrong $stripe_count != $((MDSCOUNT -1)) for striped_dir"
3755 umount_client $MOUNT || error "Unable to umount client"
3760 run_test 50i "activate deactivated MDT"
3763 local LOCAL_TIMEOUT=20
3767 check_mount || error "check_mount failed"
3769 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
3770 $SETSTRIPE -c -1 $MOUNT/$tdir ||
3771 error "$SETSTRIPE -c -1 $MOUNT/$tdir failed"
3772 #define OBD_FAIL_MDS_REINT_DELAY 0x142
3773 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x142"
3774 touch $MOUNT/$tdir/$tfile &
3777 start_ost2 || error "Unable to start OST1"
3779 stop_ost2 || error "Unable to stop OST1"
3780 umount_client $MOUNT -f || error “unmount $MOUNT failed”
3781 cleanup_nocli || error “stop server failed”
3782 #writeconf to remove all ost2 traces for subsequent tests
3783 writeconf_or_reformat
3785 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
3794 do_node $node mkdir -p $dest
3795 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3797 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
3798 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3799 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
3801 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
3802 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
3812 local backup2=${TMP}/backup2
3814 do_node $node mkdir -p $backup2
3815 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3817 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
3818 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3819 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
3821 do_node $node "diff -rq $backup $backup2"
3822 [ $? -eq 0 ] || { error "contents differ"; return 3; }
3824 local xattrs2=${TMP}/xattrs2
3825 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
3826 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
3828 do_node $node "diff $xattrs $xattrs2"
3829 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
3831 do_node $node "rm -rf $backup2 $xattrs2"
3832 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
3836 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3837 skip "Only applicable to ldiskfs-based MDTs"
3841 start_mds || error "Unable to start MDS"
3842 start_ost || error "Unable to start OST1"
3843 mount_client $MOUNT || error "Unable to mount client"
3846 local ost1mnt=$(facet_mntpt ost1)
3847 local ost1node=$(facet_active_host ost1)
3848 local ost1tmp=$TMP/conf52
3851 mkdir $DIR/$tdir || error "Unable to create $DIR/$tdir"
3852 touch $TMP/modified_first || error "Unable to create temporary file"
3853 local mtime=$(stat -c %Y $TMP/modified_first)
3854 do_node $ost1node "mkdir -p $ost1tmp &&
3855 touch -m -d @$mtime $ost1tmp/modified_first" ||
3856 error "Unable to create temporary file"
3859 $SETSTRIPE -c -1 -S 1M $DIR/$tdir || error "$SETSTRIPE failed"
3861 for (( i=0; i < nrfiles; i++ )); do
3862 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c ||
3863 error "multiop failed"
3869 echo backup files to $TMP/$tdir
3870 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
3871 copy_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
3872 error "Unable to copy files"
3874 umount_client $MOUNT || error "Unable to umount client"
3875 stop_ost || error "Unable to stop ost1"
3877 echo mount ost1 as ldiskfs
3878 do_node $ost1node mkdir -p $ost1mnt || error "Unable to create $ost1mnt"
3879 if ! do_node $ost1node test -b $ost1_dev; then
3882 do_node $ost1node mount -t $(facet_fstype ost1) $loop $ost1_dev \
3884 error "Unable to mount ost1 as ldiskfs"
3887 echo backup objects to $ost1tmp/objects
3888 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
3889 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
3890 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
3892 error "Unable to copy objects"
3894 # move objects to lost+found
3895 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
3896 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
3898 do_node $ost1node "umount $ost1mnt" ||
3899 error "Unable to umount ost1 as ldiskfs"
3901 start_ost || error "Unable to start OST1"
3902 mount_client $MOUNT || error "Unable to mount client"
3904 local REPAIRED=$(do_node $ost1node "$LCTL get_param \
3905 -n osd-ldiskfs.$FSNAME-OST0000.oi_scrub" |
3906 awk '/^lf_repa[ri]*ed/ { print $2 }')
3907 [ $REPAIRED -gt 0 ] ||
3908 error "Some entry under /lost+found should be repaired"
3911 diff_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
3912 error "Unable to diff files"
3914 rm -rf $TMP/$tdir $TMP/file_xattrs ||
3915 error "Unable to delete temporary files"
3916 do_node $ost1node "rm -rf $ost1tmp" ||
3917 error "Unable to delete temporary files"
3918 cleanup || error "cleanup failed with $?"
3920 run_test 52 "check recovering objects from lost+found"
3922 # Checks threads_min/max/started for some service
3924 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
3925 # parameter pattern prefix like 'ost.*.ost'.
3938 local msg="Insane $modname thread counts"
3939 local ncpts=$(check_cpt_number $facet)
3943 check_mount || return 41
3945 # We need to expand $parampat, but it may match multiple parameters, so
3946 # we'll pick the first one
3947 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
3948 error "Couldn't expand ${parampat}.threads_min parameter name"
3952 # Remove the .threads_min part
3953 paramp=${paramp%.threads_min}
3955 # Check for sanity in defaults
3956 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3958 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3960 tstarted=$(do_facet $facet "$LCTL get_param \
3961 -n ${paramp}.threads_started" || echo 0)
3962 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' ||
3964 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
3966 nthrs=$(expr $tmax - $tmin)
3967 if [ $nthrs -lt $ncpts ]; then
3973 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
3974 skip_env "module parameter forced $facet thread count" &&
3975 tmin=3 && tmax=$((3 * tmax))
3977 # Check that we can change min/max
3978 do_facet $facet "$LCTL set_param \
3979 ${paramp}.threads_min=$((tmin + nthrs))"
3980 do_facet $facet "$LCTL set_param \
3981 ${paramp}.threads_max=$((tmax - nthrs))"
3982 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3984 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3986 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) &&
3987 $tmax2 == ($tmax - $nthrs)))' || return $?
3989 # Check that we can set min/max to the same value
3990 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3992 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$tmin"
3993 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3995 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3997 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
3999 # Check that we can't set max < min
4000 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$((tmin - 1))"
4001 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4003 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4005 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
4007 # We need to ensure that we get the module options desired; to do this
4008 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
4009 LOAD_MODULES_REMOTE=true
4012 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
4013 setmodopts -a $modname "$newvalue" oldvalue
4016 check_mount || return 41
4018 # Restore previous setting of MODOPTS_*
4019 setmodopts $modname "$oldvalue"
4021 # Check that $opts took
4022 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min")
4023 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max")
4024 tstarted=$(do_facet $facet \
4025 "$LCTL get_param -n ${paramp}.threads_started")
4026 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4035 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
4036 cleanup || error "cleanup failed with rc $?"
4038 run_test 53a "check OSS thread count params"
4042 local mds=$(do_facet $SINGLEMDS "$LCTL get_param \
4043 -N mds.*.*.threads_max 2>/dev/null")
4044 if [ -z "$mds" ]; then
4045 #running this on an old MDT
4046 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
4048 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
4050 cleanup || error "cleanup failed with $?"
4052 run_test 53b "check MDS thread count params"
4055 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4056 skip "Only applicable to ldiskfs-based MDTs"
4060 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p ||
4061 error "llverdev failed with rc=$?"
4064 run_test 54a "test llverdev and partial verify of device"
4067 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4068 skip "Only applicable to ldiskfs-based MDTs"
4073 run_llverfs $MOUNT -p || error "llverfs failed with rc=$?"
4074 cleanup || error "cleanup failed with rc=$?"
4076 run_test 54b "test llverfs and partial verify of filesystem"
4080 local max_ost_index=$1
4081 echo -n $(((max_ost_index + 1) * 8))
4085 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4086 skip "Only applicable to ldiskfs-based MDTs"
4090 local mdsdev=$(mdsdevname 1)
4091 local mdsvdev=$(mdsvdevname 1)
4095 if ! combined_mgs_mds; then
4096 stop_mgs || error "stopping MGS service failed"
4097 format_mgs || error "formatting MGT failed"
4099 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
4101 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
4102 --reformat $(ostdevname 1) $(ostvdevname 1)
4108 echo checking size of lov_objid for ost index $i
4109 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
4110 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
4111 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
4113 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
4120 run_test 55 "check lov_objid size"
4123 local server_version=$(lustre_version_code $SINGLEMDS)
4124 local mds_journal_size_orig=$MDSJOURNALSIZE
4129 for num in $(seq 1 $MDSCOUNT); do
4132 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
4133 $(ostdevname 1) $(ostvdevname 1)
4134 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
4135 $(ostdevname 2) $(ostvdevname 2)
4138 start_ost || error "Unable to start first ost (idx 10000)"
4139 start_ost2 || error "Unable to start second ost (idx 1000)"
4140 mount_client $MOUNT || error "Unable to mount client"
4144 if [[ $server_version -ge $(version_code 2.6.54) ]] ||
4145 [[ $server_version -ge $(version_code 2.5.4) &&
4146 $server_version -lt $(version_code 2.5.11) ]]; then
4147 wait_osc_import_state mds ost1 FULL
4148 wait_osc_import_state mds ost2 FULL
4149 $SETSTRIPE --stripe-count=-1 $DIR/$tfile ||
4150 error "Unable to setstripe $DIR/$tfile"
4151 n=$($LFS getstripe --stripe-count $DIR/$tfile)
4152 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
4157 MDSJOURNALSIZE=$mds_journal_size_orig
4160 run_test 56a "check big OST indexes and out-of-index-order start"
4165 umount_client $MOUNT -f || error "unmount client failed"
4174 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs" && return
4176 trap cleanup_56b EXIT RETURN ERR
4179 if ! combined_mgs_mds ; then
4184 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) --index=0 --reformat \
4185 $(mdsdevname 1) $(mdsvdevname 1)
4186 add mds2 $(mkfs_opts mds2 $(mdsdevname 2)) --index=1 --reformat \
4187 $(mdsdevname 2) $(mdsvdevname 2)
4188 add mds3 $(mkfs_opts mds3 $(mdsdevname 3)) --index=1000 --reformat \
4189 $(mdsdevname 3) $(mdsvdevname 3)
4193 start_mdt 1 || error "MDT 1 (idx 0) start failed"
4194 start_mdt 2 || error "MDT 2 (idx 1) start failed"
4195 start_mdt 3 || error "MDT 3 (idx 1000) start failed"
4196 start_ost || error "Unable to start first ost"
4197 start_ost2 || error "Unable to start second ost"
4199 do_nodes $(comma_list $(mdts_nodes)) \
4200 "$LCTL set_param mdt.*.enable_remote_dir=1 \
4201 mdt.*.enable_remote_dir_gid=-1"
4203 mount_client $MOUNT || error "Unable to mount client"
4205 $LFS mkdir -c3 $MOUNT/$tdir || error "failed to make testdir"
4207 echo "This is test file 1!" > $MOUNT/$tdir/$tfile.1 ||
4208 error "failed to make test file 1"
4209 echo "This is test file 2!" > $MOUNT/$tdir/$tfile.2 ||
4210 error "failed to make test file 2"
4211 echo "This is test file 1000!" > $MOUNT/$tdir/$tfile.1000 ||
4212 error "failed to make test file 1000"
4214 rm -rf $MOUNT/$tdir || error "failed to remove testdir"
4216 $LFS mkdir -i1000 $MOUNT/$tdir.1000 ||
4217 error "create remote dir at idx 1000 failed"
4220 echo "=== START lfs df OUTPUT ==="
4222 echo "==== END lfs df OUTPUT ===="
4224 mdtcnt=$(echo -e "$output" | grep $FSNAME-MDT | wc -l)
4225 ostcnt=$(echo -e "$output" | grep $FSNAME-OST | wc -l)
4227 echo "lfs df returned mdt count $mdtcnt and ost count $ostcnt"
4228 [ $mdtcnt -eq 3 ] || error "lfs df returned wrong mdt count"
4229 [ $ostcnt -eq 2 ] || error "lfs df returned wrong ost count"
4231 echo "This is test file 1!" > $MOUNT/$tdir.1000/$tfile.1 ||
4232 error "failed to make test file 1"
4233 echo "This is test file 2!" > $MOUNT/$tdir.1000/$tfile.2 ||
4234 error "failed to make test file 2"
4235 echo "This is test file 1000!" > $MOUNT/$tdir.1000/$tfile.1000 ||
4236 error "failed to make test file 1000"
4237 rm -rf $MOUNT/$tdir.1000 || error "failed to remove remote_dir"
4240 echo "=== START lfs mdts OUTPUT ==="
4242 echo "==== END lfs mdts OUTPUT ===="
4244 echo -e "$output" | grep -v "MDTS:" | awk '{print $1}' |
4245 sed 's/://g' > $TMP/mdts-actual.txt
4246 sort $TMP/mdts-actual.txt -o $TMP/mdts-actual.txt
4248 echo -e "0\n1\n1000" > $TMP/mdts-expected.txt
4250 diff $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4253 rm $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4255 [ $result -eq 0 ] || error "target_obd proc file is incorrect!"
4257 run_test 56b "test target_obd correctness with nonconsecutive MDTs"
4259 test_57a() { # bug 22656
4260 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4261 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4262 tail -1 | awk '{print $1}')
4263 writeconf_or_reformat
4264 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4265 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" ||
4266 error "tunefs failed"
4268 start_ost && error "OST registration from failnode should fail"
4271 run_test 57a "initial registration from failnode should fail (should return errs)"
4274 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4275 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4276 tail -1 | awk '{print $1}')
4277 writeconf_or_reformat
4278 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4279 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" ||
4280 error "tunefs failed"
4282 start_ost || error "OST registration from servicenode should not fail"
4285 run_test 57b "initial registration from servicenode should not fail"
4288 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
4291 test_58() { # bug 22658
4292 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
4294 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4295 createmany -o $DIR/$tdir/$tfile-%d 100
4296 # make sure that OSTs do not cancel llog cookies before we unmount the MDS
4297 #define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601
4298 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x601"
4299 unlinkmany $DIR/$tdir/$tfile-%d 100
4300 stop_mds || error "Unable to stop MDS"
4302 local MNTDIR=$(facet_mntpt $SINGLEMDS)
4303 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4305 # remove all files from the OBJECTS dir
4306 mount_fstype $SINGLEMDS
4308 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
4310 unmount_fstype $SINGLEMDS
4311 # restart MDS with missing llog files
4312 start_mds || error "unable to start MDS"
4313 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0"
4316 run_test 58 "missing llog files must not prevent MDT from mounting"
4319 start_mgsmds >> /dev/null
4320 local C1=$(count_osts)
4321 if [ $C1 -eq 0 ]; then
4322 start_ost >> /dev/null
4326 echo "original ost count: $C1 (expect > 0)"
4327 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
4328 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
4329 local C2=$(count_osts)
4330 echo "after mdt writeconf count: $C2 (expect 0)"
4331 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
4332 echo "OST start without writeconf should fail:"
4333 start_ost >> /dev/null &&
4334 error "OST start without writeconf didn't fail"
4335 echo "OST start with writeconf should succeed:"
4336 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
4337 local C3=$(count_osts)
4338 echo "after ost writeconf count: $C3 (expect 1)"
4339 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
4340 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
4341 local C4=$(count_osts)
4342 echo "after ost2 writeconf count: $C4 (expect 2)"
4343 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
4344 stop_ost2 >> /dev/null
4345 cleanup_nocli >> /dev/null
4346 #writeconf to remove all ost2 traces for subsequent tests
4347 writeconf_or_reformat
4349 run_test 59 "writeconf mount option"
4351 test_60() { # LU-471
4354 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4355 skip "Only applicable to ldiskfs-based MDTs"
4359 for num in $(seq $MDSCOUNT); do
4360 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
4361 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
4362 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4366 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
4367 [ ${PIPESTATUS[0]} -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
4369 # MDT default has dirdata feature
4370 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
4371 # we disable uninit_bg feature
4372 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
4373 # we set stride extended options
4374 echo $dump | grep stride > /dev/null || error "stride is not set"
4377 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
4382 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] ||
4383 { skip "Need MDS version at least 2.1.53"; return 0; }
4385 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
4386 ! large_xattr_enabled; then
4389 for num in $(seq $MDSCOUNT); do
4390 do_facet mds${num} $TUNE2FS -O large_xattr \
4391 $(mdsdevname $num) ||
4392 error "tune2fs on mds $num failed"
4396 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
4397 setup_noconfig || error "setting up the filesystem failed"
4398 client_up || error "starting client failed"
4400 local file=$DIR/$tfile
4401 touch $file || error "touch $file failed"
4403 local large_value="$(generate_string $(max_xattr_size))"
4404 local small_value="bar"
4406 local name="trusted.big"
4407 log "save large xattr $name on $file"
4408 setfattr -n $name -v $large_value $file ||
4409 error "saving $name on $file failed"
4411 local new_value=$(get_xattr_value $name $file)
4412 [[ "$new_value" != "$large_value" ]] &&
4413 error "$name different after saving"
4415 log "shrink value of $name on $file"
4416 setfattr -n $name -v $small_value $file ||
4417 error "shrinking value of $name on $file failed"
4419 new_value=$(get_xattr_value $name $file)
4420 [[ "$new_value" != "$small_value" ]] &&
4421 error "$name different after shrinking"
4423 log "grow value of $name on $file"
4424 setfattr -n $name -v $large_value $file ||
4425 error "growing value of $name on $file failed"
4427 new_value=$(get_xattr_value $name $file)
4428 [[ "$new_value" != "$large_value" ]] &&
4429 error "$name different after growing"
4431 log "check value of $name on $file after remounting MDS"
4433 new_value=$(get_xattr_value $name $file)
4434 [[ "$new_value" != "$large_value" ]] &&
4435 error "$name different after remounting MDS"
4437 log "remove large xattr $name from $file"
4438 setfattr -x $name $file || error "removing $name from $file failed"
4441 stopall || error "stopping for e2fsck run"
4442 for num in $(seq $MDSCOUNT); do
4443 run_e2fsck $(facet_active_host mds$num) \
4444 $(mdsdevname $num) "-y" ||
4445 error "e2fsck MDT$num failed"
4447 setup_noconfig || error "remounting the filesystem failed"
4450 # need to delete this file to avoid problems in other tests
4452 stopall || error "stopping systems to turn off large_xattr"
4454 for num in $(seq $MDSCOUNT); do
4455 do_facet mds${num} $TUNE2FS -O ^large_xattr \
4456 $(mdsdevname $num) ||
4457 error "tune2fs on mds $num failed"
4461 run_test 61 "large xattr"
4464 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4465 skip "Only applicable to ldiskfs-based MDTs"
4470 local mdsdev=$(mdsdevname 1)
4471 local ostdev=$(ostdevname 1)
4473 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] ||
4474 { skip "Need MDS version at least 2.2.51"; return 0; }
4476 echo "disable journal for mds"
4477 do_facet mds1 $TUNE2FS -O ^has_journal $mdsdev || error "tune2fs failed"
4478 start_mds && error "MDT start should fail"
4479 echo "disable journal for ost"
4480 do_facet ost1 $TUNE2FS -O ^has_journal $ostdev || error "tune2fs failed"
4481 start_ost && error "OST start should fail"
4482 cleanup || error "cleanup failed with rc $?"
4485 run_test 62 "start with disabled journal"
4488 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4489 skip "Only applicable to ldiskfs-based MDTs"
4493 local inode_slab=$(do_facet $SINGLEMDS \
4494 "awk '/ldiskfs_inode_cache/ { print \\\$5 }' /proc/slabinfo")
4495 if [ -z "$inode_slab" ]; then
4496 skip "ldiskfs module has not been loaded"
4500 echo "$inode_slab ldisk inodes per page"
4501 if [ "$inode_slab" -ge "3" ] ; then
4502 # If kmalloc-128 is also 1 per page - this is a debug kernel
4503 # and so this is not an error.
4504 local kmalloc128=$(do_facet $SINGLEMDS \
4505 "awk '/^(kmalloc|size)-128 / { print \\\$5 }' /proc/slabinfo")
4506 # 32 128-byte chunks in 4k
4507 [ "$kmalloc128" -eq "32" ] ||
4508 error "ldisk inode size is too big, $inode_slab objs per page"
4513 run_test 63 "Verify each page can at least hold 3 ldisk inodes"
4516 start_mds || error "unable to start MDS"
4517 start_ost || error "Unable to start OST1"
4518 start_ost2 || error "Unable to start second ost"
4519 mount_client $MOUNT || error "Unable to mount client"
4520 stop_ost2 || error "Unable to stop second ost"
4523 umount_client $MOUNT -f || error “unmount $MOUNT failed”
4524 cleanup_nocli || error "cleanup_nocli failed with $?"
4525 #writeconf to remove all ost2 traces for subsequent tests
4526 writeconf_or_reformat
4528 run_test 64 "check lfs df --lazy "
4530 test_65() { # LU-2237
4531 # Currently, the test is only valid for ldiskfs backend
4532 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4533 skip "non-ldiskfs backend" && return
4535 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4536 local brpt=$(facet_mntpt brpt)
4539 if ! do_facet $SINGLEMDS "test -b $devname"; then
4543 stop_mds || error "Unable to stop MDS"
4544 local obj=$(do_facet $SINGLEMDS \
4545 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
4547 if [ -z "$obj" ]; then
4548 # The MDT may be just re-formatted, mount the MDT for the
4549 # first time to guarantee the "last_rcvd" file is there.
4550 start_mds || error "fail to mount the MDS for the first time"
4551 stop_mds || error "Unable to stop MDS"
4554 # remove the "last_rcvd" file
4555 do_facet $SINGLEMDS "mkdir -p $brpt"
4556 do_facet $SINGLEMDS \
4557 "mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
4558 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
4559 do_facet $SINGLEMDS "$UMOUNT $brpt"
4561 # restart MDS, the "last_rcvd" file should be recreated.
4562 start_mds || error "fail to restart the MDS"
4563 stop_mds || error "Unable to stop MDS"
4564 obj=$(do_facet $SINGLEMDS \
4565 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
4566 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
4568 run_test 65 "re-create the lost last_rcvd file when server mount"
4571 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
4572 { skip "Need MGS version at least 2.3.59"; return 0; }
4575 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
4576 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
4578 echo "replace_nids should fail if MDS, OSTs and clients are UP"
4579 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4580 error "replace_nids fail"
4582 umount_client $MOUNT || error "unmounting client failed"
4583 echo "replace_nids should fail if MDS and OSTs are UP"
4584 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4585 error "replace_nids fail"
4587 stop_ost || error "Unable to stop OST1"
4588 echo "replace_nids should fail if MDS is UP"
4589 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4590 error "replace_nids fail"
4592 stop_mds || error "stopping mds failed"
4594 if combined_mgs_mds; then
4595 start_mdt 1 "-o nosvc" ||
4596 error "starting mds with nosvc option failed"
4599 echo "command should accept two parameters"
4600 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
4601 error "command should accept two params"
4603 echo "correct device name should be passed"
4604 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
4605 error "wrong devname"
4607 echo "wrong nids list should not destroy the system"
4608 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
4611 echo "replace OST nid"
4612 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
4613 error "replace nids failed"
4615 echo "command should accept two parameters"
4616 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
4617 error "command should accept two params"
4619 echo "wrong nids list should not destroy the system"
4620 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
4623 echo "replace MDS nid"
4624 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
4625 error "replace nids failed"
4627 if ! combined_mgs_mds ; then
4630 stop_mds || error "Unable to stop MDS"
4634 check_mount || error "error after nid replace"
4635 cleanup || error "cleanup failed"
4638 run_test 66 "replace nids"
4640 test_67() { #LU-2950
4641 local legacy="$TMP/legacy_lnet_config"
4642 local new="$TMP/new_routes_test"
4643 local out="$TMP/config_out_file"
4644 local verify="$TMP/conv_verify"
4645 local verify_conf="$TMP/conf_verify"
4647 # Create the legacy file that will be run through the
4648 # lustre_routes_conversion script
4649 cat <<- LEGACY_LNET_CONFIG > $legacy
4650 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
4651 tcp2 54 10.1.3.2@tcp;
4652 tcp3 10.3.4.3@tcp:3;
4656 # Create the verification file to verify the output of
4657 # lustre_routes_conversion script against.
4658 cat <<- VERIFY_LNET_CONFIG > $verify
4659 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
4660 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
4661 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
4662 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
4663 tcp4: { gateway: 10.3.3.4@tcp }
4666 # Create the verification file to verify the output of
4667 # lustre_routes_config script against
4668 cat <<- VERIFY_LNET_CONFIG > $verify_conf
4669 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
4670 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
4671 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
4672 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
4673 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
4676 lustre_routes_conversion $legacy $new > /dev/null
4677 if [ -f $new ]; then
4678 # verify the conversion output
4679 cmp -s $new $verify > /dev/null
4680 if [ $? -eq 1 ]; then
4681 error "routes conversion failed"
4684 lustre_routes_config --dry-run --verbose $new > $out
4685 # check that the script succeeded
4686 cmp -s $out $verify_conf > /dev/null
4687 if [ $? -eq 1 ]; then
4688 error "routes config failed"
4691 error "routes conversion test failed"
4693 # remove generated files
4694 rm -f $new $legacy $verify $verify_conf $out
4696 run_test 67 "test routes conversion and configuration"
4704 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.4.53) ] ||
4705 { skip "Need MDS version at least 2.4.53"; return 0; }
4707 umount_client $MOUNT || error "umount client failed"
4709 start_mdt 1 || error "MDT start failed"
4710 start_ost || error "Unable to start OST1"
4712 # START-END - the sequences we'll be reserving
4713 START=$(do_facet $SINGLEMDS \
4714 $LCTL get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
4715 END=$((START + (1 << 30)))
4716 do_facet $SINGLEMDS \
4717 $LCTL set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
4719 # reset the sequences MDT0000 has already assigned
4720 do_facet $SINGLEMDS \
4721 $LCTL set_param seq.srv*MDT0000.space=clear
4723 # remount to let the client allocate new sequence
4724 mount_client $MOUNT || error "mount client failed"
4726 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4727 do_facet $SINGLEMDS \
4728 $LCTL get_param seq.srv*MDT0000.space
4729 $LFS path2fid $DIR/$tfile
4731 local old_ifs="$IFS"
4733 fid=($($LFS path2fid $DIR/$tfile))
4737 if [[ $seq < $END ]]; then
4738 error "used reserved sequence $seq?"
4740 cleanup || error "cleanup failed with $?"
4742 run_test 68 "be able to reserve specific sequences in FLDB"
4744 # Test 69: is about the total number of objects ever created on an OST.
4745 # so that when it is reformatted the normal MDS->OST orphan recovery won't
4746 # just "precreate" the missing objects. In the past it might try to recreate
4747 # millions of objects after an OST was reformatted
4749 local server_version=$(lustre_version_code $SINGLEMDS)
4751 [[ $server_version -lt $(version_code 2.4.2) ]] &&
4752 skip "Need MDS version at least 2.4.2" && return
4754 [[ $server_version -ge $(version_code 2.4.50) ]] &&
4755 [[ $server_version -lt $(version_code 2.5.0) ]] &&
4756 skip "Need MDS version at least 2.5.0" && return
4759 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4761 # use OST0000 since it probably has the most creations
4762 local OSTNAME=$(ostname_from_index 0)
4763 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
4764 local last_id=$(do_facet mds1 $LCTL get_param -n \
4765 osc.$mdtosc_proc1.prealloc_last_id)
4767 # Want to have OST LAST_ID over 5 * OST_MAX_PRECREATE to
4768 # verify that the LAST_ID recovery is working properly. If
4769 # not, then the OST will refuse to allow the MDS connect
4770 # because the LAST_ID value is too different from the MDS
4771 #define OST_MAX_PRECREATE=20000
4772 local ost_max_pre=20000
4773 local num_create=$(( ost_max_pre * 5 + 1 - last_id))
4775 # If the LAST_ID is already over 5 * OST_MAX_PRECREATE, we don't
4776 # need to create any files. So, skip this section.
4777 if [ $num_create -gt 0 ]; then
4778 # Check the number of inodes available on OST0
4780 local ifree=$($LFS df -i $MOUNT | awk '/OST0000/ { print $4 }')
4781 log "On OST0, $ifree inodes available. Want $num_create."
4783 $SETSTRIPE -i 0 $DIR/$tdir ||
4784 error "$SETSTRIPE -i 0 $DIR/$tdir failed"
4785 if [ $ifree -lt 10000 ]; then
4786 files=$(( ifree - 50 ))
4791 local j=$((num_create / files + 1))
4792 for i in $(seq 1 $j); do
4793 createmany -o $DIR/$tdir/$tfile-$i- $files ||
4794 error "createmany fail create $files files: $?"
4795 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
4796 error "unlinkmany failed unlink $files files"
4800 # delete all of the files with objects on OST0 so the
4801 # filesystem is not inconsistent later on
4802 $LFS find $MOUNT --ost 0 -print0 | xargs -0 rm
4804 umount_client $MOUNT || error "umount client failed"
4805 stop_ost || error "OST0 stop failure"
4806 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
4807 $(ostdevname 1) $(ostvdevname 1) ||
4808 error "reformat and replace $ostdev failed"
4809 start_ost || error "OST0 restart failure"
4810 wait_osc_import_state mds ost FULL
4812 mount_client $MOUNT || error "mount client failed"
4813 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
4814 local idx=$($GETSTRIPE -i $DIR/$tdir/$tfile-last)
4815 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
4817 local iused=$($LFS df -i $MOUNT | awk '/OST0000/ { print $3 }')
4818 log "On OST0, $iused used inodes"
4819 [ $iused -ge $((ost_max_pre/2 + 1000)) ] &&
4820 error "OST replacement created too many inodes; $iused"
4821 cleanup || error "cleanup failed with $?"
4823 run_test 69 "replace an OST with the same index"
4826 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4829 cleanup || error "cleanup failed with $?"
4831 start_mdt 1 || error "MDT0 start fail"
4833 start_ost || error "OST0 start fail"
4834 for num in $(seq 2 $MDSCOUNT); do
4835 start_mdt $num || return
4838 mount_client $MOUNT || error "mount client fails"
4840 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
4842 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4843 error "create remote dir fail"
4845 rm -rf $DIR/$tdir || error "delete dir fail"
4846 cleanup || error "cleanup failed with $?"
4848 run_test 70a "start MDT0, then OST, then MDT1"
4851 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4854 start_ost || error "OST0 start fail"
4856 start_mds || error "MDS start fail"
4858 mount_client $MOUNT || error "mount client fails"
4860 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
4862 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4863 error "create remote dir fail"
4865 rm -rf $DIR/$tdir || error "delete dir fail"
4867 cleanup || error "cleanup failed with $?"
4869 run_test 70b "start OST, MDT1, MDT0"
4872 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4875 start_mds || error "MDS start fail"
4876 start_ost || error "OST0 start fail"
4878 mount_client $MOUNT || error "mount client fails"
4879 stop_mdt 1 || error "MDT1 start fail"
4881 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
4882 echo "deactivate $mdc_for_mdt1"
4883 $LCTL --device $mdc_for_mdt1 deactivate ||
4884 error "set $mdc_for_mdt1 deactivate failed"
4886 mkdir $DIR/$tdir && error "mkdir succeed"
4888 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4889 error "create remote dir succeed"
4891 cleanup || error "cleanup failed with $?"
4893 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
4896 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4899 start_mds || error "MDS start fail"
4900 start_ost || error "OST0 start fail"
4902 mount_client $MOUNT || error "mount client fails"
4904 stop_mdt 2 || error "MDT1 start fail"
4906 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
4908 echo "deactivate $mdc_for_mdt2"
4909 $LCTL --device $mdc_for_mdt2 deactivate ||
4910 error "set $mdc_for_mdt2 deactivate failed"
4912 mkdir $DIR/$tdir || error "mkdir fail"
4913 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4914 error "create remote dir succeed"
4916 rm -rf $DIR/$tdir || error "delete dir fail"
4918 cleanup || error "cleanup failed with $?"
4920 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
4923 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4925 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.62) ] ||
4926 { skip "Need MDS version at least 2.7.62"; return 0; }
4928 cleanup || error "cleanup failed with $?"
4930 local mdsdev=$(mdsdevname 1)
4931 local ostdev=$(ostdevname 1)
4932 local mdsvdev=$(mdsvdevname 1)
4933 local ostvdev=$(ostvdevname 1)
4934 local opts_mds="$(mkfs_opts mds1 $mdsdev) --reformat $mdsdev $mdsvdev"
4935 local opts_ost="$(mkfs_opts ost1 $ostdev) --reformat $ostdev $ostvdev"
4937 add mds1 $opts_mds || error "add mds1 failed"
4938 start_mdt 1 || error "start mdt1 failed"
4939 add ost1 $opts_ost || error "add ost1 failed"
4940 start_ost || error "start ost failed"
4941 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
4943 local soc=$(do_facet mds1 "$LCTL get_param -n \
4944 mdt.*MDT0000.sync_lock_cancel")
4945 [ $soc == "never" ] || error "SoC enabled on single MDS"
4947 for i in $(seq 2 $MDSCOUNT); do
4948 mdsdev=$(mdsdevname $i)
4949 mdsvdev=$(mdsvdevname $i)
4950 opts_mds="$(mkfs_opts mds$i $mdsdev) --reformat $mdsdev \
4952 add mds$i $opts_mds || error "add mds$i failed"
4953 start_mdt $i || error "start mdt$i fail"
4956 wait_dne_interconnect
4958 for i in $(seq $MDSCOUNT); do
4959 soc=$(do_facet mds$i "$LCTL get_param -n \
4960 mdt.*MDT000$((i - 1)).sync_lock_cancel")
4961 [ $soc == "blocking" ] || error "SoC not enabled on DNE"
4964 for i in $(seq 2 $MDSCOUNT); do
4965 stop_mdt $i || error "stop mdt$i fail"
4967 soc=$(do_facet mds1 "$LCTL get_param -n \
4968 mdt.*MDT0000.sync_lock_cancel")
4969 [ $soc == "never" ] || error "SoC enabled on single MDS"
4971 cleanup || error "cleanup failed with $?"
4973 run_test 70e "Sync-on-Cancel will be enabled by default on DNE"
4976 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4977 if combined_mgs_mds; then
4978 skip "needs separate MGS/MDT" && return
4982 start_mdt 1 || error "MDT0 start fail"
4983 start_ost || error "OST0 start fail"
4984 for num in $(seq 2 $MDSCOUNT); do
4985 start_mdt $num || return
4988 start_ost2 || error "OST1 start fail"
4990 mount_client $MOUNT || error "mount client fails"
4992 mkdir $DIR/$tdir || error "mkdir fail"
4993 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4994 error "create remote dir succeed"
4996 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4997 rm -rf $DIR/$tdir || error "delete dir fail"
4999 umount_client $MOUNT || error "umount_client failed"
5000 stop_mds || error "MDS stop fail"
5001 stop_ost || error "OST0 stop fail"
5002 stop_ost2 || error "OST1 stop fail"
5004 run_test 71a "start MDT0 OST0, MDT1, OST1"
5007 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5008 if combined_mgs_mds; then
5009 skip "needs separate MGS/MDT" && return
5013 for num in $(seq 2 $MDSCOUNT); do
5014 start_mdt $num || return
5016 start_ost || error "OST0 start fail"
5017 start_mdt 1 || error "MDT0 start fail"
5018 start_ost2 || error "OST1 start fail"
5020 mount_client $MOUNT || error "mount client fails"
5022 mkdir $DIR/$tdir || error "mkdir fail"
5023 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5024 error "create remote dir succeed"
5026 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5027 rm -rf $DIR/$tdir || error "delete dir fail"
5029 umount_client $MOUNT || error "umount_client failed"
5030 stop_mds || error "MDT0 stop fail"
5031 stop_ost || error "OST0 stop fail"
5032 stop_ost2 || error "OST1 stop fail"
5034 run_test 71b "start MDT1, OST0, MDT0, OST1"
5037 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5038 if combined_mgs_mds; then
5039 skip "needs separate MGS/MDT" && return
5043 start_ost || error "OST0 start fail"
5044 start_ost2 || error "OST1 start fail"
5045 for num in $(seq 2 $MDSCOUNT); do
5046 start_mdt $num || return
5048 start_mdt 1 || error "MDT0 start fail"
5050 mount_client $MOUNT || error "mount client fails"
5052 mkdir $DIR/$tdir || error "mkdir fail"
5053 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5054 error "create remote dir succeed"
5056 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5057 rm -rf $DIR/$tdir || error "delete dir fail"
5059 umount_client $MOUNT || error "umount_client failed"
5060 stop_mds || error "MDS stop fail"
5061 stop_ost || error "OST0 stop fail"
5062 stop_ost2 || error "OST1 stop fail"
5065 run_test 71c "start OST0, OST1, MDT1, MDT0"
5068 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5069 if combined_mgs_mds; then
5070 skip "needs separate MGS/MDT" && return
5074 start_ost || error "OST0 start fail"
5075 for num in $(seq 2 $MDSCOUNT); do
5076 start_mdt $num || return
5078 start_mdt 1 || error "MDT0 start fail"
5079 start_ost2 || error "OST1 start fail"
5081 mount_client $MOUNT || error "mount client fails"
5083 mkdir $DIR/$tdir || error "mkdir fail"
5084 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5085 error "create remote dir succeed"
5087 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5088 rm -rf $DIR/$tdir || error "delete dir fail"
5090 umount_client $MOUNT || error "umount_client failed"
5091 stop_mds || error "MDS stop fail"
5092 stop_ost || error "OST0 stop fail"
5093 stop_ost2 || error "OST1 stop fail"
5096 run_test 71d "start OST0, MDT1, MDT0, OST1"
5099 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5100 if combined_mgs_mds; then
5101 skip "needs separate MGS/MDT" && return
5105 start_ost || error "OST0 start fail"
5106 for num in $(seq 2 $MDSCOUNT); do
5107 start_mdt $num || return
5109 start_ost2 || error "OST1 start fail"
5110 start_mdt 1 || error "MDT0 start fail"
5112 mount_client $MOUNT || error "mount client fails"
5114 mkdir $DIR/$tdir || error "mkdir fail"
5115 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5116 error "create remote dir succeed"
5118 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5119 rm -rf $DIR/$tdir || error "delete dir fail"
5121 umount_client $MOUNT || error "umount_client failed"
5122 stop_mds || error "MDS stop fail"
5123 stop_ost || error "OST0 stop fail"
5124 stop_ost2 || error "OST1 stop fail"
5127 run_test 71e "start OST0, MDT1, OST1, MDT0"
5129 test_72() { #LU-2634
5130 local mdsdev=$(mdsdevname 1)
5131 local ostdev=$(ostdevname 1)
5132 local cmd="$E2FSCK -fnvd $mdsdev"
5135 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
5136 skip "ldiskfs only test" && return
5138 #tune MDT with "-O extents"
5140 for num in $(seq $MDSCOUNT); do
5141 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
5142 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
5143 error "add mds $num failed"
5144 do_facet mds${num} "$TUNE2FS -O extents $(mdsdevname $num)" ||
5145 error "$TUNE2FS failed on mds${num}"
5148 add ost1 $(mkfs_opts ost1 $ostdev) --reformat $ostdev ||
5149 error "add $ostdev failed"
5150 start_mgsmds || error "start mds failed"
5151 start_ost || error "start ost failed"
5152 mount_client $MOUNT || error "mount client failed"
5154 #create some short symlinks
5155 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5156 createmany -o $DIR/$tdir/$tfile-%d $fn
5157 echo "create $fn short symlinks"
5158 for i in $(seq -w 1 $fn); do
5159 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
5164 umount_client $MOUNT || error "umount client failed"
5165 stop_mds || error "stop mds failed"
5166 stop_ost || error "stop ost failed"
5169 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
5171 run_test 72 "test fast symlink with extents flag enabled"
5173 test_73() { #LU-3006
5174 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
5175 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
5176 error "1st tunefs failed"
5177 start_mgsmds || error "start mds failed"
5178 start_ost || error "start ost failed"
5179 mount_client $MOUNT || error "mount client failed"
5180 $LCTL get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
5181 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
5182 umount_client $MOUNT || error "umount client failed"
5186 run_test 73 "failnode to update from mountdata properly"
5188 test_75() { # LU-2374
5189 [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] &&
5190 skip "Need MDS version at least 2.4.1" && return
5193 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
5194 --reformat $(mdsdevname 1) $(mdsvdevname 1)"
5195 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
5196 --reformat $(ostdevname 1) $(ostvdevname 1)"
5198 #check with default parameters
5199 add mds1 $opts_mds || error "add mds1 failed for default params"
5200 add ost1 $opts_ost || error "add ost1 failed for default params"
5202 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
5203 opts_mds=$(echo $opts_mds |
5204 sed -e "s/--index=$index/--index=$index --mdt/")
5205 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
5206 opts_ost=$(echo $opts_ost |
5207 sed -e "s/--index=$index/--index=$index --ost/")
5209 add mds1 $opts_mds || error "add mds1 failed for new params"
5210 add ost1 $opts_ost || error "add ost1 failed for new params"
5213 run_test 75 "The order of --index should be irrelevant"
5216 [[ $(lustre_version_code mgs) -ge $(version_code 2.4.52) ]] ||
5217 { skip "Need MDS version at least 2.4.52" && return 0; }
5219 local MDMB_PARAM="osc.*.max_dirty_mb"
5220 echo "Change MGS params"
5221 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
5223 echo "max_dirty_mb: $MAX_DIRTY_MB"
5224 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB + MAX_DIRTY_MB))
5225 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
5226 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
5227 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5228 head -1" $NEW_MAX_DIRTY_MB
5229 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5230 echo "$MAX_DIRTY_MB"
5231 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5232 error "error while apply max_dirty_mb"
5234 echo "Check the value is stored after remount"
5237 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5238 head -1" $NEW_MAX_DIRTY_MB
5239 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5240 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5241 error "max_dirty_mb is not saved after remount"
5243 echo "Change OST params"
5244 CLIENT_PARAM="obdfilter.*.client_cache_count"
5245 local CLIENT_CACHE_COUNT
5246 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5248 echo "client_cache_count: $CLIENT_CACHE_COUNT"
5249 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
5250 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
5251 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
5252 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5253 head -1" $NEW_CLIENT_CACHE_COUNT
5254 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5256 echo "$CLIENT_CACHE_COUNT"
5257 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5258 error "error while apply client_cache_count"
5260 echo "Check the value is stored after remount"
5263 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5264 head -1" $NEW_CLIENT_CACHE_COUNT
5265 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5267 echo "$CLIENT_CACHE_COUNT"
5268 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5269 error "client_cache_count is not saved after remount"
5272 run_test 76a "set permanent params set_param -P"
5274 test_76b() { # LU-4783
5275 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.57) ]] ||
5276 { skip "Need MGS version at least 2.5.57" && return 0; }
5279 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
5280 error "start params log failed"
5283 run_test 76b "verify params log setup correctly"
5286 [[ $(lustre_version_code mgs) -ge $(version_code 2.8.54) ]] ||
5287 { skip "Need MDS version at least 2.4.52" && return 0; }
5289 local MASK_PARAM="mdd.*.changelog_mask"
5290 echo "Change changelog_mask"
5291 do_facet mgs $LCTL set_param -P $MASK_PARAM=-CLOSE ||
5292 error "Can't change changlog_mask"
5293 wait_update $(facet_host mds) "$LCTL get_param -n $MASK_PARAM |
5296 echo "Check the value is stored after mds remount"
5297 stop_mds || error "Failed to stop MDS"
5298 start_mds || error "Failed to start MDS"
5299 local CHANGELOG_MASK=$(do_facet mgs $LCTL get_param -n $MASK_PARAM)
5300 echo $CHANGELOG_MASK | grep CLOSE > /dev/null &&
5301 error "changelog_mask is not changed"
5305 run_test 76c "verify changelog_mask is applied with set_param -P"
5307 test_77() { # LU-3445
5308 local server_version=$(lustre_version_code $SINGLEMDS)
5309 [[ $server_version -ge $(version_code 2.8.55) ]] ||
5310 { skip "Need MDS version 2.8.55+ "; return; }
5312 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
5313 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
5314 skip_env "mixed loopback and real device not working" && return
5317 local fs2mdsdev=$(mdsdevname 1_2)
5318 local fs2ostdev=$(ostdevname 1_2)
5319 local fs2mdsvdev=$(mdsvdevname 1_2)
5320 local fs2ostvdev=$(ostvdevname 1_2)
5321 local fsname=test1234
5323 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
5325 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
5326 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
5327 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
5328 error "start fs2mds failed"
5330 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
5331 mgsnid="$mgsnid,$mgsnid:$mgsnid"
5333 add fs2ost $(mkfs_opts ost1 $fs2ostdev) --mgsnode=$mgsnid \
5334 --failnode=$failnid --fsname=$fsname \
5335 --reformat $fs2ostdev $fs2ostvdev ||
5336 error "add fs2ost failed"
5337 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
5339 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
5340 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
5341 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
5344 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
5347 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ||
5348 $(facet_fstype ost1) != ldiskfs ]] &&
5349 skip "only applicable to ldiskfs-based MDTs and OSTs" && return
5351 # reformat the Lustre filesystem with a smaller size
5352 local saved_MDSCOUNT=$MDSCOUNT
5353 local saved_MDSSIZE=$MDSSIZE
5354 local saved_OSTCOUNT=$OSTCOUNT
5355 local saved_OSTSIZE=$OSTSIZE
5358 MDSSIZE=$((MDSSIZE - 20000))
5359 OSTSIZE=$((OSTSIZE - 20000))
5360 reformat || error "(1) reformat Lustre filesystem failed"
5361 MDSSIZE=$saved_MDSSIZE
5362 OSTSIZE=$saved_OSTSIZE
5364 # mount the Lustre filesystem
5365 setup_noconfig || error "(2) setup Lustre filesystem failed"
5368 log "create test files"
5373 mkdir $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
5375 for i in $(seq $num_files); do
5376 file=$MOUNT/$tdir/$tfile-$i
5377 dd if=/dev/urandom of=$file count=1 bs=1M || {
5378 $LCTL get_param osc.*.cur*grant*
5379 $LFS df; $LFS df -i;
5380 # stop creating files if there is no more space
5381 if [ ! -e $file ]; then
5382 num_files=$((i - 1))
5386 $LFS getstripe -v $file
5387 local ost_idx=$(LFS getstripe -i $file)
5388 do_facet ost$((ost_idx + 1)) \
5389 $LCTL get_param obdfilter.*.*grant*
5390 error "(4) create $file failed"
5394 # unmount the Lustre filesystem
5395 cleanup || error "(5) cleanup Lustre filesystem failed"
5397 # run e2fsck on the MDT and OST devices
5398 local mds_host=$(facet_active_host $SINGLEMDS)
5399 local ost_host=$(facet_active_host ost1)
5400 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
5401 local ost_dev=$(ostdevname 1)
5403 run_e2fsck $mds_host $mds_dev "-y"
5404 run_e2fsck $ost_host $ost_dev "-y"
5406 # get the original block count of the MDT and OST filesystems
5407 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
5408 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
5410 # expand the MDT and OST filesystems to the device size
5411 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
5412 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
5414 # run e2fsck on the MDT and OST devices again
5415 run_e2fsck $mds_host $mds_dev "-y"
5416 run_e2fsck $ost_host $ost_dev "-y"
5418 # mount the Lustre filesystem
5422 log "check files after expanding the MDT and OST filesystems"
5423 for i in $(seq $num_files); do
5424 file=$MOUNT/$tdir/$tfile-$i
5425 $CHECKSTAT -t file -s 1048576 $file ||
5426 error "(6) checkstat $file failed"
5430 log "create more files after expanding the MDT and OST filesystems"
5431 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
5432 file=$MOUNT/$tdir/$tfile-$i
5433 dd if=/dev/urandom of=$file count=1 bs=1M ||
5434 error "(7) create $file failed"
5437 # unmount the Lustre filesystem
5438 cleanup || error "(8) cleanup Lustre filesystem failed"
5440 # run e2fsck on the MDT and OST devices
5441 run_e2fsck $mds_host $mds_dev "-y"
5442 run_e2fsck $ost_host $ost_dev "-y"
5444 # get the maximum block count of the MDT and OST filesystems
5445 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
5446 local ost_max_blks=$(get_block_count ost1 $ost_dev)
5448 # get the minimum block count of the MDT and OST filesystems
5449 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
5450 grep minimum | sed -e 's/^.*filesystem: //g')
5451 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
5452 grep minimum | sed -e 's/^.*filesystem: //g')
5454 # shrink the MDT and OST filesystems to a smaller size
5458 if [[ $mds_max_blks -gt $mds_min_blks &&
5459 $mds_max_blks -gt $mds_orig_blks ]]; then
5460 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
5461 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
5462 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
5463 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
5464 error "shrink $SINGLEMDS to $new_blks failed"
5468 if [[ $ost_max_blks -gt $ost_min_blks &&
5469 $ost_max_blks -gt $ost_orig_blks ]]; then
5470 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
5471 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
5472 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
5473 run_resize2fs ost1 $ost_dev $new_blks ||
5474 error "shrink ost1 to $new_blks failed"
5478 # check whether the MDT or OST filesystem was shrunk or not
5480 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
5481 reformat || error "(10) reformat Lustre filesystem failed"
5485 # run e2fsck on the MDT and OST devices again
5486 run_e2fsck $mds_host $mds_dev "-y"
5487 run_e2fsck $ost_host $ost_dev "-y"
5489 # mount the Lustre filesystem again
5493 log "check files after shrinking the MDT and OST filesystems"
5494 for i in $(seq $((num_files + 10))); do
5495 file=$MOUNT/$tdir/$tfile-$i
5496 $CHECKSTAT -t file -s 1048576 $file ||
5497 error "(11) checkstat $file failed"
5500 # unmount and reformat the Lustre filesystem
5501 cleanup || error "(12) cleanup Lustre filesystem failed"
5502 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
5504 MDSCOUNT=$saved_MDSCOUNT
5505 OSTCOUNT=$saved_OSTCOUNT
5506 reformat || error "(14) reformat Lustre filesystem failed"
5508 run_test 78 "run resize2fs on MDT and OST filesystems"
5510 test_79() { # LU-4227
5511 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.5.59) ]] ||
5512 { skip "Need MDS version at least 2.5.59"; return 0; }
5514 local mdsdev1=$(mdsdevname 1)
5515 local mdsvdev1=$(mdsvdevname 1)
5516 local mdsdev2=$(mdsdevname 2)
5517 local mdsvdev2=$(mdsvdevname 2)
5518 local ostdev1=$(ostdevname 1)
5519 local ostvdev1=$(ostvdevname 1)
5520 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
5521 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
5522 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
5525 # remove --mgs/--mgsnode from mkfs.lustre options
5526 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
5528 mgsnode_opt=$(echo $opts_mds2 |
5529 awk '{ for ( i = 1; i < NF; i++ )
5530 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5531 [ -n $mgsnode_opt ] &&
5532 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
5534 mgsnode_opt=$(echo $opts_ost1 |
5535 awk '{ for ( i = 1; i < NF; i++ )
5536 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5537 [ -n $mgsnode_opt ] &&
5538 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
5540 # -MGS, format a mdt without --mgs option
5541 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
5542 error "Must specify --mgs when formatting mdt combined with mgs"
5544 # +MGS, format a mdt/ost without --mgsnode option
5545 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
5546 > /dev/null || error "start mds1 failed"
5547 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
5548 error "Must specify --mgsnode when formatting a mdt"
5549 add ost1 $opts_ost1 $ostdev1 $ostvdev2 &&
5550 error "Must specify --mgsnode when formatting an ost"
5554 run_test 79 "format MDT/OST without mgs option (should return errors)"
5557 start_mds || error "Failed to start MDT"
5558 start_ost || error "Failed to start OST1"
5559 uuid=$(do_facet ost1 $LCTL get_param -n mgc.*.uuid)
5560 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
5561 do_facet ost1 "$LCTL set_param fail_val=10 fail_loc=0x906"
5562 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x906"
5563 do_facet mgs "$LCTL set_param -n mgs/MGS/evict_client $uuid"
5565 start_ost2 || error "Failed to start OST2"
5567 do_facet ost1 "$LCTL set_param fail_loc=0"
5570 run_test 80 "mgc import reconnect race"
5572 #Save the original values of $OSTCOUNT and $OSTINDEX$i.
5574 local new_ostcount=$1
5575 saved_ostcount=$OSTCOUNT
5576 OSTCOUNT=$new_ostcount
5580 for i in $(seq $OSTCOUNT); do
5582 eval saved_ostindex$i=${!index}
5587 # Restore the original values of $OSTCOUNT and $OSTINDEX$i.
5588 restore_ostindex() {
5593 for i in $(seq $OSTCOUNT); do
5594 index=saved_ostindex$i
5595 eval OSTINDEX$i=${!index}
5597 OSTCOUNT=$saved_ostcount
5602 # The main purpose of this test is to ensure the OST_INDEX_LIST functions as
5603 # expected. This test uses OST_INDEX_LIST to format OSTs with a randomly
5604 # assigned index and ensures we can mount such a formatted file system
5605 test_81() { # LU-4665
5606 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5607 { skip "Need MDS version at least 2.6.54" && return; }
5608 [[ $OSTCOUNT -ge 3 ]] || { skip_env "Need at least 3 OSTs" && return; }
5612 # Each time RANDOM is referenced, a random integer between 0 and 32767
5615 local saved_ostindex1=$OSTINDEX1
5616 for i in 65535 $((RANDOM + 65536)); do
5617 echo -e "\nFormat ost1 with --index=$i, should fail"
5619 if add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat \
5620 $(ostdevname 1) $(ostvdevname 1); then
5621 OSTINDEX1=$saved_ostindex1
5622 error "format ost1 with --index=$i should fail"
5625 OSTINDEX1=$saved_ostindex1
5629 # Format OSTs with random sparse indices.
5630 trap "restore_ostindex" EXIT
5631 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices"
5632 OST_INDEX_LIST=[0,$((RANDOM * 2 % 65533 + 1)),65534] formatall
5634 # Setup and check Lustre filesystem.
5635 start_mgsmds || error "start_mgsmds failed"
5636 for i in $(seq $OSTCOUNT); do
5637 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5638 error "start ost$i failed"
5641 mount_client $MOUNT || error "mount client $MOUNT failed"
5642 check_mount || error "check client $MOUNT failed"
5645 local max_easize=$($LCTL get_param -n llite.*.max_easize)
5646 [[ $max_easize -eq 128 ]] ||
5647 error "max_easize is $max_easize, should be 128 bytes"
5651 run_test 81 "sparse OST indexing"
5653 # Wait OSTs to be active on both client and MDT side.
5655 local cmd="$LCTL get_param -n lov.$FSNAME-clilov-*.target_obd |
5656 awk 'BEGIN {c = 0} /ACTIVE/{c += 1} END {printf \\\"%d\\\", c}'"
5657 wait_update $HOSTNAME "eval $cmd" $OSTCOUNT ||
5658 error "wait_update OSTs up on client failed"
5660 cmd="$LCTL get_param -n lod.$FSNAME-MDT*-*.target_obd | sort -u |
5661 awk 'BEGIN {c = 0} /ACTIVE/{c += 1} END {printf \\\"%d\\\", c}'"
5662 wait_update_facet $SINGLEMDS "eval $cmd" $OSTCOUNT ||
5663 error "wait_update OSTs up on MDT failed"
5666 # Here we exercise the stripe placement functionality on a file system that
5667 # has formatted the OST with a random index. With the file system the following
5668 # functionality is tested:
5670 # 1. Creating a new file with a specific stripe layout.
5672 # 2. Modifiy a existing empty file with a specific stripe layout.
5674 # 3. Ensure we fail to set the stripe layout of a file that already has one.
5676 # 4. If ost-index is defined we need to ensure it is the first entry in the
5677 # ost index list returned by lfs getstripe.
5679 # 5. Lastly ensure this functionality fails with directories.
5680 test_82a() { # LU-4665
5681 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5682 { skip "Need MDS version at least 2.6.54" && return; }
5683 [[ $OSTCOUNT -ge 3 ]] || { skip_env "Need at least 3 OSTs" && return; }
5689 # Format OSTs with random sparse indices.
5693 local LOV_V1_INSANE_STRIPE_COUNT=65532
5694 for i in $(seq $OSTCOUNT); do
5695 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
5696 ost_indices+=" $index"
5698 ost_indices=$(comma_list $ost_indices)
5700 trap "restore_ostindex" EXIT
5701 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5702 OST_INDEX_LIST=[$ost_indices] formatall
5704 # Setup Lustre filesystem.
5705 start_mgsmds || error "start_mgsmds failed"
5706 for i in $(seq $OSTCOUNT); do
5707 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5708 error "start ost$i failed"
5711 mount_client $MOUNT || error "mount client $MOUNT failed"
5714 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5715 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5717 # 1. If the file does not exist, new file will be created
5718 # with specified OSTs.
5719 local file=$DIR/$tdir/$tfile-1
5720 local cmd="$SETSTRIPE -o $ost_indices $file"
5722 eval $cmd || error "$cmd failed"
5723 check_stripe_count $file $OSTCOUNT
5724 check_obdidx $file $ost_indices
5725 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5726 error "write $file failed"
5728 # 2. If the file already exists and is an empty file, the file
5729 # will be attached with specified layout.
5730 file=$DIR/$tdir/$tfile-2
5731 mcreate $file || error "mcreate $file failed"
5732 cmd="$SETSTRIPE -o $ost_indices $file"
5734 eval $cmd || error "$cmd failed"
5735 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5736 error "write $file failed"
5737 check_stripe_count $file $OSTCOUNT
5738 check_obdidx $file $ost_indices
5740 # 3. If the file already has a valid layout attached, the command
5741 # should fail with EBUSY.
5743 eval $cmd && error "stripe is already set on $file, $cmd should fail"
5745 # 4. If [--stripe-index|-i <start_ost_idx>] is used, the index must
5746 # be in the OST indices list.
5747 local start_ost_idx=${ost_indices##*,}
5748 file=$DIR/$tdir/$tfile-3
5749 cmd="$SETSTRIPE -o $ost_indices -i $start_ost_idx $file"
5751 eval $cmd || error "$cmd failed"
5752 check_stripe_count $file $OSTCOUNT
5753 check_obdidx $file $ost_indices
5754 check_start_ost_idx $file $start_ost_idx
5756 file=$DIR/$tdir/$tfile-4
5758 cmd+=" -o $(exclude_items_from_list $ost_indices $start_ost_idx)"
5759 cmd+=" -i $start_ost_idx $file"
5761 eval $cmd && error "index $start_ost_idx should be in $ost_indices"
5763 # 5. Specifying OST indices for directory should fail with ENOSUPP.
5764 local dir=$DIR/$tdir/$tdir
5765 mkdir $dir || error "mkdir $dir failed"
5766 cmd="$SETSTRIPE -o $ost_indices $dir"
5768 eval $cmd && error "$cmd should fail, specifying OST indices" \
5769 "for directory is not supported"
5773 run_test 82a "specify OSTs for file (succeed) or directory (fail)"
5778 # Remove OSTs from a pool and destroy the pool.
5779 destroy_pool $ost_pool || true
5784 # Test 82b is run to ensure that if the user supplies a pool with a specific
5785 # stripe layout that it behaves proprerly. It should fail in the case that
5786 # the supplied OST index list points to OSTs not contained in the user
5788 test_82b() { # LU-4665
5789 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5790 { skip "Need MDS version at least 2.6.54" && return; }
5791 [[ $OSTCOUNT -ge 4 ]] || { skip_env "Need at least 4 OSTs" && return; }
5797 # Format OSTs with random sparse indices.
5801 local LOV_V1_INSANE_STRIPE_COUNT=65532
5802 for i in $(seq $OSTCOUNT); do
5803 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
5804 ost_indices+=" $index"
5806 ost_indices=$(comma_list $ost_indices)
5808 trap "restore_ostindex" EXIT
5809 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5810 OST_INDEX_LIST=[$ost_indices] formatall
5812 # Setup Lustre filesystem.
5813 start_mgsmds || error "start_mgsmds failed"
5814 for i in $(seq $OSTCOUNT); do
5815 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5816 error "start ost$i failed"
5819 mount_client $MOUNT || error "mount client $MOUNT failed"
5821 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5822 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5824 # Create a new pool and add OSTs into it.
5825 local ost_pool=$FSNAME.$TESTNAME
5826 create_pool $ost_pool || error "create OST pool $ost_pool failed"
5829 trap "cleanup_82b" EXIT
5831 local ost_idx_in_list=${ost_indices##*,}
5832 local ost_idx_in_pool=$(exclude_items_from_list $ost_indices \
5835 local ost_targets="$FSNAME-OST["
5836 for i in ${ost_idx_in_pool//,/ }; do
5837 ost_targets=$ost_targets$(printf "%04x," $i)
5839 ost_targets="${ost_targets%,}]"
5841 local ost_targets_uuid=$(for i in ${ost_idx_in_pool//,/ }; \
5842 do printf "$FSNAME-OST%04x_UUID\n" $i; done |
5843 sort -u | tr '\n' ' ')
5845 local cmd="$LCTL pool_add $ost_pool $ost_targets"
5846 do_facet mgs $cmd || error "$cmd failed"
5847 wait_update $HOSTNAME "$LCTL get_param -n lov.$FSNAME-*.pools.$TESTNAME|
5848 sort -u | tr '\n' ' ' " "$ost_targets_uuid" ||
5849 error "wait_update $ost_pool failed"
5850 pool_list $ost_pool || error "list OST pool $ost_pool failed"
5852 # If [--pool|-p <pool_name>] is set with [--ost-list|-o <ost_indices>],
5853 # then the OSTs must be the members of the pool.
5854 local file=$DIR/$tdir/$tfile
5855 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
5857 eval $cmd && error "OST with index $ost_idx_in_list should be" \
5858 "in OST pool $ost_pool"
5860 # Only select OST $ost_idx_in_list from $ost_pool for file.
5861 ost_idx_in_list=${ost_idx_in_pool#*,}
5862 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
5864 eval $cmd || error "$cmd failed"
5865 cmd="$GETSTRIPE $file"
5867 eval $cmd || error "$cmd failed"
5868 check_stripe_count $file 2
5869 check_obdidx $file $ost_idx_in_list
5870 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5871 error "write $file failed"
5875 run_test 82b "specify OSTs for file with --pool and --ost-list options"
5878 [[ $(lustre_version_code ost1) -ge $(version_code 2.6.91) ]] ||
5879 { skip "Need OST version at least 2.6.91" && return 0; }
5880 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
5881 skip "Only applicable to ldiskfs-based MDTs"
5891 ostmnt=$(facet_mntpt ost1)
5892 fstype=$(facet_fstype ost1)
5894 # Mount the OST as an ldiskfs filesystem.
5895 log "mount the OST $dev as a $fstype filesystem"
5896 add ost1 $(mkfs_opts ost1 $dev) $FSTYPE_OPT \
5897 --reformat $dev $dev > /dev/null ||
5898 error "format ost1 error"
5900 if ! test -b $dev; then
5901 mnt_opts=$(csa_add "$OST_MOUNT_OPTS" -o loop)
5903 echo "mnt_opts $mnt_opts"
5904 do_facet ost1 mount -t $fstype $dev \
5906 # Run llverfs on the mounted ldiskfs filesystem.
5907 # It is needed to get ENOSPACE.
5908 log "run llverfs in partial mode on the OST $fstype $ostmnt"
5909 do_rpc_nodes $(facet_host ost1) run_llverfs $ostmnt -vpl \
5910 "no" || error "run_llverfs error on $fstype"
5913 log "unmount the OST $dev"
5916 # Delete file IO_scrub. Later osd_scrub_setup will try to
5917 # create "IO_scrub" but will get ENOSPACE.
5919 echo "start ost1 service on `facet_active_host ost1`"
5920 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
5923 err=$(do_facet ost1 dmesg | grep "VFS: Busy inodes after unmount of")
5924 echo "string err $err"
5925 [ -z "$err" ] || error $err
5928 run_test 83 "ENOSPACE on OST doesn't cause message VFS: \
5929 Busy inodes after unmount ..."
5932 local facet=$SINGLEMDS
5933 local num=$(echo $facet | tr -d "mds")
5934 local dev=$(mdsdevname $num)
5935 local time_min=$(recovery_time_min)
5936 local recovery_duration
5937 local completed_clients
5938 local correct_clients
5941 echo "start mds service on $(facet_active_host $facet)"
5943 "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
5944 error "start MDS failed"
5946 start_ost || error "start OST0000 failed"
5947 start_ost2 || error "start OST0001 failed"
5949 echo "recovery_time=$time_min, timeout=$TIMEOUT, wrap_up=$wrap_up"
5951 mount_client $MOUNT1 || error "mount $MOUNT1 failed"
5952 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
5953 # make sure new superblock labels are sync'd before disabling writes
5957 replay_barrier $SINGLEMDS
5958 createmany -o $DIR1/$tfile-%d 1000
5960 # We need to catch the end of recovery window to extend it.
5961 # Skip 5 requests and add delay to request handling.
5962 #define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP
5963 do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5"
5965 facet_failover --fsck $SINGLEMDS || error "failover: $?"
5968 echo "recovery status"
5969 do_facet $SINGLEMDS \
5970 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status"
5972 recovery_duration=$(do_facet $SINGLEMDS \
5973 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
5974 awk '/recovery_duration/ { print $2 }')
5975 (( $recovery_duration > $time_min + $wrap_up )) &&
5976 error "recovery_duration > recovery_time_hard + wrap up"
5977 completed_clients=$(do_facet $SINGLEMDS \
5978 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
5979 awk '/completed_clients/ { print $2 }')
5981 correct_clients="$MDSCOUNT/$((MDSCOUNT+1))"
5982 [ "$completed_clients" = "${correct_clients}" ] ||
5983 error "$completed_clients != $correct_clients"
5985 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
5986 umount_client $MOUNT1
5987 umount_client $MOUNT2
5993 run_test 84 "check recovery_hard_time"
5996 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.55) ]] ||
5997 { skip "Need OST version at least 2.7.55" && return 0; }
5998 ##define OBD_FAIL_OSD_OST_EA_FID_SET 0x197
5999 do_facet ost1 "lctl set_param fail_loc=0x197"
6003 run_test 85 "osd_ost init: fail ea_fid_set"
6008 # ost1 has already registered to the MGS before the reformat.
6009 # So after reformatting it with option "-G", it could not be
6010 # mounted to the MGS. Cleanup the system for subsequent tests.
6015 local server_version=$(lustre_version_code $SINGLEMDS)
6016 [ "$(facet_fstype ost1)" = "zfs" ] &&
6017 skip "LU-6442: no such mkfs params for ZFS OSTs" && return
6018 [[ $server_version -ge $(version_code 2.7.56) ]] ||
6019 { skip "Need server version newer than 2.7.55"; return 0; }
6021 local OST_OPTS="$(mkfs_opts ost1 $(ostdevname 1)) \
6022 --reformat $(ostdevname 1) $(ostvdevname 1)"
6025 local OLDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6026 awk '/Flex block group size: / { print $NF; exit; }')
6029 if [[ ${!opts} != *mkfsoptions* ]]; then
6030 eval opts=\"${!opts} \
6031 --mkfsoptions='\\\"-O flex_bg -G $NEWSIZE\\\"'\"
6033 val=${!opts//--mkfsoptions=\\\"/ \
6034 --mkfsoptions=\\\"-O flex_bg -G $NEWSIZE }
6038 echo "params: $opts"
6040 trap cleanup_86 EXIT ERR
6043 add ost1 $opts || error "add ost1 failed with new params"
6045 local FOUNDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6046 awk '/Flex block group size: / { print $NF; exit; }')
6048 [[ $FOUNDSIZE == $NEWSIZE ]] ||
6049 error "Flex block group size: $FOUNDSIZE, expected: $NEWSIZE"
6053 run_test 86 "Replacing mkfs.lustre -G option"
6055 test_87() { #LU-6544
6056 [[ $(lustre_version_code $SINGLEMDS1) -ge $(version_code 2.7.56) ]] ||
6057 { skip "Need MDS version at least 2.7.56" && return; }
6058 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ]] &&
6059 { skip "Only applicable to ldiskfs-based MDTs" && return; }
6060 [[ $OSTCOUNT -gt 69 ]] &&
6061 { skip "Ignore wide striping situation" && return; }
6063 local mdsdev=$(mdsdevname 1)
6064 local mdsvdev=$(mdsvdevname 1)
6065 local file=$DIR/$tfile
6066 local mntpt=$(facet_mntpt $SINGLEMDS)
6067 local used_xattr_blk=0
6068 local inode_size=${1:-512}
6070 local xtest="trusted.test"
6075 #Please see LU-6544 for MDT inode size calculation
6076 if [ $OSTCOUNT -gt 26 ]; then
6078 elif [ $OSTCOUNT -gt 5 ]; then
6081 left_size=$(expr $inode_size - \
6084 32 - $OSTCOUNT \* 24 - 16 - 3 - \
6086 24 - 18 - $(expr length $tfile) - 16 - 4)
6087 if [ $left_size -le 0 ]; then
6088 echo "No space($left_size) is expected in inode."
6089 echo "Try 1-byte xattr instead to verify this."
6092 echo "Estimate: at most $left_size-byte space left in inode."
6098 add mds1 $(mkfs_opts mds1 ${mdsdev}) --stripe-count-hint=$OSTCOUNT \
6099 --reformat $mdsdev $mdsvdev || error "add mds1 failed"
6100 start_mdt 1 > /dev/null || error "start mdt1 failed"
6101 for i in $(seq $OSTCOUNT); do
6102 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS > /dev/null ||
6103 error "start ost$i failed"
6105 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
6106 check_mount || error "check client $MOUNT failed"
6109 $SETSTRIPE -c -1 $file || error "$SETSTRIPE -c -1 $file failed"
6110 $GETSTRIPE $file || error "$GETSTRIPE $file failed"
6111 i=$($GETSTRIPE -c $file)
6112 if [ $i -ne $OSTCOUNT ]; then
6113 left_size=$(expr $left_size + $(expr $OSTCOUNT - $i) \* 24)
6114 echo -n "Since only $i out $OSTCOUNT OSTs are used, "
6115 echo -n "the expected left space is changed to "
6116 echo "$left_size bytes at most."
6118 value=$(generate_string $left_size)
6119 setfattr -n $xtest -v $value $file
6120 orig=$(get_xattr_value $xtest $file)
6121 [[ "$orig" != "$value" ]] && error "$xtest changed"
6123 #Verify if inode has some expected space left
6124 umount $MOUNT > /dev/null || error "umount $MOUNT failed"
6125 stop_mdt 1 > /dev/null || error "stop mdt1 failed"
6126 mount_ldiskfs $SINGLEMDS || error "mount -t ldiskfs $SINGLEMDS failed"
6128 do_facet $SINGLEMDS ls -sal $mntpt/ROOT/$tfile
6129 used_xattr_blk=$(do_facet $SINGLEMDS ls -s $mntpt/ROOT/$tfile |
6131 [[ $used_xattr_blk -eq 0 ]] &&
6132 error "Please check MDS inode size calculation: \
6133 more than $left_size-byte space left in inode."
6134 echo "Verified: at most $left_size-byte space left in inode."
6138 run_test 87 "check if MDT inode can hold EAs with N stripes properly"
6141 [ "$(facet_fstype mds1)" == "zfs" ] &&
6142 skip "LU-6662: no implementation for ZFS" && return
6146 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
6147 --reformat $(mdsdevname 1) || error "add mds1 failed"
6149 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
6150 grep -e \".*opts:.*errors=remount-ro.*\"" ||
6151 error "default mount options is missing"
6153 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
6154 --mountfsoptions="user_xattr,errors=panic" \
6155 --reformat $(mdsdevname 1) || error "add mds1 failed"
6157 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
6158 grep -e \".*opts:.*errors=panic.*\"" ||
6159 error "user can't override default mount options"
6161 run_test 88 "check the default mount options can be overridden"
6164 # $2 (optional) value of max_mod_rpcs_in_flight to set
6165 check_max_mod_rpcs_in_flight() {
6173 idx=$(printf "%04x" $($LFS getdirstripe -i $dir))
6174 facet="mds$((0x$idx + 1))"
6176 if [ -z "$mmr" ]; then
6177 # get value of max_mod_rcps_in_flight
6178 mmr=$($LCTL get_param -n \
6179 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) ||
6180 error "Unable to get max_mod_rpcs_in_flight"
6181 echo "max_mod_rcps_in_flight is $mmr"
6183 # set value of max_mod_rpcs_in_flight
6185 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight=$mmr ||
6186 error "Unable to set max_mod_rpcs_in_flight to $mmr"
6187 echo "max_mod_rpcs_in_flight set to $mmr"
6190 # create mmr+1 files
6191 echo "creating $((mmr + 1)) files ..."
6193 for i in $(seq $((mmr + 1))); do
6199 # consumes mmr-1 modify RPC slots
6200 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6201 # drop requests on MDT so that RPC slots are consumed
6202 # during all the request resend interval
6203 do_facet $facet "$LCTL set_param fail_loc=0x159"
6204 echo "launch $((mmr - 1)) chmod in parallel ..."
6205 for i in $(seq $((mmr - 1))); do
6206 chmod 0600 $dir/file-$i &
6210 # send one additional modify RPC
6211 do_facet $facet "$LCTL set_param fail_loc=0"
6212 echo "launch 1 additional chmod in parallel ..."
6213 chmod 0600 $dir/file-$mmr &
6216 # check this additional modify RPC get a modify RPC slot
6217 # and succeed its operation
6218 checkstat -vp 0600 $dir/file-$mmr ||
6219 error "Unable to send $mmr modify RPCs in parallel"
6224 # consumes mmr modify RPC slots
6225 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6226 # drop requests on MDT so that RPC slots are consumed
6227 # during all the request resend interval
6228 do_facet $facet "$LCTL set_param fail_loc=0x159"
6229 echo "launch $mmr chmod in parallel ..."
6230 for i in $(seq $mmr); do
6231 chmod 0666 $dir/file-$i &
6235 # send one additional modify RPC
6236 do_facet $facet "$LCTL set_param fail_loc=0"
6237 echo "launch 1 additional chmod in parallel ..."
6238 chmod 0666 $dir/file-$((mmr + 1)) &
6241 # check this additional modify RPC blocked getting a modify RPC slot
6242 checkstat -vp 0644 $dir/file-$((mmr + 1)) ||
6243 error "Unexpectedly send $(($mmr + 1)) modify RPCs in parallel"
6249 if ! combined_mgs_mds ; then
6254 [[ $($LCTL get_param mdc.*.import |
6255 grep "connect_flags:.*multi_mod_rpc") ]] ||
6256 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6258 # check default value
6259 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6260 check_max_mod_rpcs_in_flight $DIR/$tdir
6264 run_test 90a "check max_mod_rpcs_in_flight is enforced"
6274 [[ $($LCTL get_param mdc.*.import |
6275 grep "connect_flags:.*multi_mod_rpc") ]] ||
6276 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6279 # update max_mod_rpcs_in_flight
6280 $LFS mkdir -c1 $DIR/${tdir}1 || error "mkdir $DIR/${tdir}1 failed"
6281 check_max_mod_rpcs_in_flight $DIR/${tdir}1 1
6284 # check client is able to send multiple modify RPCs in paralell
6285 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
6286 grep -c "multi_mod_rpcs")
6287 if [ "$tmp" -ne $MDSCOUNT ]; then
6288 echo "Client not able to send multiple modify RPCs in parallel"
6293 # update max_mod_rpcs_in_flight
6294 $LFS mkdir -c1 $DIR/${tdir}2 || error "mkdir $DIR/${tdir}2 failed"
6295 check_max_mod_rpcs_in_flight $DIR/${tdir}2 5
6298 $LFS mkdir -c1 $DIR/${tdir}3 || error "mkdir $DIR/${tdir}3 failed"
6299 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/${tdir}3))
6300 facet="mds$((0x$idx + 1))"
6302 # save MDT max_mod_rpcs_per_client
6303 mmrpc=$(do_facet $facet \
6304 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
6306 # update max_mod_rpcs_in_flight
6307 umount_client $MOUNT
6309 "echo 16 > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6311 $LCTL set_param mdc.$FSNAME-MDT$idx-mdc-*.max_rpcs_in_flight=17
6312 check_max_mod_rpcs_in_flight $DIR/${tdir}3 16
6314 # restore MDT max_mod_rpcs_per_client initial value
6316 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6318 rm -rf $DIR/${tdir}?
6321 run_test 90b "check max_mod_rpcs_in_flight is enforced after update"
6330 [[ $($LCTL get_param mdc.*.import |
6331 grep "connect_flags:.*multi_mod_rpc") ]] ||
6332 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6334 # check client is able to send multiple modify RPCs in paralell
6335 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
6336 grep -c "multi_mod_rpcs")
6337 if [ "$tmp" -ne $MDSCOUNT ]; then
6338 skip "Client not able to send multiple modify RPCs in parallel"
6343 # get max_rpcs_in_flight value
6344 mrif=$($LCTL get_param -n mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
6345 echo "max_rpcs_in_flight is $mrif"
6347 # get MDT max_mod_rpcs_per_client
6348 mmrpc=$(do_facet mds1 \
6349 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
6350 echo "max_mod_rpcs_per_client is $mmrpc"
6353 # attempt to set max_mod_rpcs_in_flight to max_rpcs_in_flight value
6354 # prerequisite: set max_mod_rpcs_per_client to max_rpcs_in_flight value
6355 umount_client $MOUNT
6357 "echo $mrif > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6361 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif &&
6362 error "set max_mod_rpcs_in_flight to $mrif should fail"
6364 umount_client $MOUNT
6366 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6370 # attempt to set max_mod_rpcs_in_flight to max_mod_rpcs_per_client+1
6371 # prerequisite: set max_rpcs_in_flight to max_mod_rpcs_per_client+2
6373 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$((mmrpc + 2))
6376 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$((mmrpc + 1)) &&
6377 error "set max_mod_rpcs_in_flight to $((mmrpc + 1)) should fail"
6381 run_test 90c "check max_mod_rpcs_in_flight update limits"
6392 [[ $($LCTL get_param mdc.*.import |
6393 grep "connect_flags:.*multi_mod_rpc") ]] ||
6394 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6396 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6397 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
6398 facet="mds$((0x$idx + 1))"
6400 # check client version supports multislots
6401 tmp=$($LCTL get_param -N \
6402 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
6403 if [ -z "$tmp" ]; then
6404 skip "Client does not support multiple modify RPCs in flight"
6409 # get current value of max_mod_rcps_in_flight
6410 mmr=$($LCTL get_param -n \
6411 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
6412 echo "max_mod_rcps_in_flight is $mmr"
6415 echo "creating $mmr files ..."
6417 for i in $(seq $mmr); do
6418 touch $DIR/$tdir/file-$i
6421 # prepare for close RPC
6422 multiop_bg_pause $DIR/$tdir/file-close O_c
6425 # consumes mmr modify RPC slots
6426 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6427 # drop requests on MDT so that RPC slots are consumed
6428 # during all the request resend interval
6429 do_facet $facet "$LCTL set_param fail_loc=0x159"
6430 echo "launch $mmr chmod in parallel ..."
6431 for i in $(seq $mmr); do
6432 chmod 0600 $DIR/$tdir/file-$i &
6435 # send one additional close RPC
6436 do_facet $facet "$LCTL set_param fail_loc=0"
6437 echo "launch 1 additional close in parallel ..."
6439 cancel_lru_locks mdc
6442 # check this additional close RPC get a modify RPC slot
6443 # and multiop process completed
6444 [ -d /proc/$pid ] &&
6445 error "Unable to send the additional close RPC in parallel"
6450 run_test 90d "check one close RPC is allowed above max_mod_rpcs_in_flight"
6452 check_uuid_on_ost() {
6454 do_facet ost1 "$LCTL get_param obdfilter.${FSNAME}*.exports.'$nid'.uuid"
6457 check_uuid_on_mdt() {
6459 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
6467 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.63) ]] ||
6468 { skip "Need OST version at least 2.7.63" && return 0; }
6469 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.63) ]] ||
6470 { skip "Need MDT version at least 2.7.63" && return 0; }
6472 start_mds || error "MDS start failed"
6473 start_ost || error "unable to start OST"
6474 mount_client $MOUNT || error "client start failed"
6475 check_mount || error "check_mount failed"
6478 nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g")
6482 uuid=$(get_client_uuid $MOUNT)
6484 echo "list nids on mdt:"
6485 do_facet $SINGLEMDS "$LCTL list_param mdt.${FSNAME}*.exports.*"
6486 echo "uuid from $nid:"
6487 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
6489 found=$(check_uuid_on_mdt $nid | grep $uuid)
6490 [ -z "$found" ] && error "can't find $uuid $nid on MDT"
6491 found=$(check_uuid_on_ost $nid | grep $uuid)
6492 [ -z "$found" ] && error "can't find $uuid $nid on OST"
6494 # umount the client so it won't reconnect
6495 manual_umount_client --force || error "failed to umount $?"
6496 # shouldn't disappear on MDS after forced umount
6497 found=$(check_uuid_on_mdt $nid | grep $uuid)
6498 [ -z "$found" ] && error "can't find $uuid $nid"
6501 do_facet $SINGLEMDS \
6502 "$LCTL set_param -n mdt.${mds1_svc}.evict_client nid:$nid"
6504 found=$(check_uuid_on_mdt $nid | grep $uuid)
6505 [ -n "$found" ] && error "found $uuid $nid on MDT"
6506 found=$(check_uuid_on_ost $nid | grep $uuid)
6507 [ -n "$found" ] && error "found $uuid $nid on OST"
6509 # check it didn't reconnect (being umounted)
6510 sleep $((TIMEOUT+1))
6511 found=$(check_uuid_on_mdt $nid | grep $uuid)
6512 [ -n "$found" ] && error "found $uuid $nid on MDT"
6513 found=$(check_uuid_on_ost $nid | grep $uuid)
6514 [ -n "$found" ] && error "found $uuid $nid on OST"
6518 run_test 91 "evict-by-nid support"
6520 generate_ldev_conf() {
6521 # generate an ldev.conf file
6522 local ldevconfpath=$1
6524 local fsldevformat=""
6527 fstype=$(facet_fstype mgs)
6528 if [ "$fstype" == "zfs" ]; then
6529 fsldevformat="$fstype:"
6534 printf "%s\t-\t%s-MGS0000\t%s%s\n" \
6538 $(mgsdevname) > $ldevconfpath
6540 local mdsfo_host=$mdsfailover_HOST;
6541 if [ -z "$mdsfo_host" ]; then
6545 for num in $(seq $MDSCOUNT); do
6546 fstype=$(facet_fstype mds$num)
6547 if [ "$fstype" == "zfs" ]; then
6548 fsldevformat="$fstype:"
6553 printf "%s\t%s\t%s-MDT%04d\t%s%s\n" \
6559 $(mdsdevname $num) >> $ldevconfpath
6562 local ostfo_host=$ostfailover_HOST;
6563 if [ -z "$ostfo_host" ]; then
6567 for num in $(seq $OSTCOUNT); do
6568 fstype=$(facet_fstype ost$num)
6569 if [ "$fstype" == "zfs" ]; then
6570 fsldevformat="$fstype:"
6575 printf "%s\t%s\t%s-OST%04d\t%s%s\n" \
6581 $(ostdevname $num) >> $ldevconfpath
6584 echo "----- $ldevconfpath -----"
6586 echo "--- END $ldevconfpath ---"
6591 # generate a nids file (mapping between hostname to nid)
6592 # looks like we only have the MGS nid available to us
6593 # so just echo that to a file
6595 echo -e "${mgs_HOST}\t${MGSNID}" > $nidspath
6597 echo "----- $nidspath -----"
6599 echo "--- END $nidspath ---"
6602 compare_ldev_output() {
6606 sort $expected_output -o $expected_output
6607 sort $ldev_output -o $ldev_output
6609 echo "-- START OF LDEV OUTPUT --"
6611 echo "--- END OF LDEV OUTPUT ---"
6613 echo "-- START OF EXPECTED OUTPUT --"
6614 cat $expected_output
6615 echo "--- END OF EXPECTED OUTPUT ---"
6617 diff $expected_output $ldev_output
6622 if [ -z "$LDEV" ]; then
6623 error "ldev is missing!"
6626 local LDEVCONFPATH=$TMP/ldev.conf
6627 local NIDSPATH=$TMP/nids
6629 echo "Host is $(hostname)"
6631 generate_ldev_conf $LDEVCONFPATH
6632 generate_nids $NIDSPATH
6634 # echo the mgs nid and compare it to environment variable MGSNID
6635 # also, ldev.conf and nids is a server side thing, use the OSS
6638 output=$($LDEV -c $LDEVCONFPATH -H $ost_HOST -n $NIDSPATH echo %m)
6640 echo "-- START OF LDEV OUTPUT --"
6642 echo "--- END OF LDEV OUTPUT ---"
6644 # ldev failed, error
6645 if [ $? -ne 0 ]; then
6646 rm $LDEVCONFPATH $NIDSPATH
6647 error "ldev failed to execute!"
6650 # need to process multiple lines because of combined MGS and MDS
6651 echo -e $output | awk '{ print $2 }' | while read -r line ; do
6652 if [ "$line" != "$MGSNID" ]; then
6653 rm $LDEVCONFPATH $NIDSPATH
6654 error "ldev failed mgs nid '$line', expected '$MGSNID'"
6658 rm $LDEVCONFPATH $NIDSPATH
6660 run_test 92 "ldev returns MGS NID correctly in command substitution"
6663 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs" && return
6666 #start mgs or mgs/mdt0
6667 if ! combined_mgs_mds ; then
6674 start_ost || error "OST0 start fail"
6676 #define OBD_FAIL_MGS_WRITE_TARGET_DELAY 0x90e
6677 do_facet mgs "$LCTL set_param fail_val = 10 fail_loc=0x8000090e"
6678 for num in $(seq 2 $MDSCOUNT); do
6682 mount_client $MOUNT || error "mount client fails"
6683 wait_osc_import_state mds ost FULL
6684 wait_osc_import_state client ost FULL
6685 check_mount || error "check_mount failed"
6687 cleanup || error "cleanup failed with $?"
6689 run_test 93 "register mulitple MDT at the same time"
6692 if [ -z "$LDEV" ]; then
6693 error "ldev is missing!"
6696 local LDEVCONFPATH=$TMP/ldev.conf
6697 local NIDSPATH=$TMP/nids
6699 generate_ldev_conf $LDEVCONFPATH
6700 generate_nids $NIDSPATH
6702 local LDEV_OUTPUT=$TMP/ldev-output.txt
6703 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME > $LDEV_OUTPUT
6705 # ldev failed, error
6706 if [ $? -ne 0 ]; then
6707 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
6708 error "ldev failed to execute!"
6712 local EXPECTED_OUTPUT=$TMP/ldev-expected.txt
6714 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
6716 for num in $(seq $MDSCOUNT); do
6717 printf "%s-MDT%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
6720 for num in $(seq $OSTCOUNT); do
6721 printf "%s-OST%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
6724 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6726 if [ $? -ne 0 ]; then
6727 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6728 error "ldev failed to produce the correct hostlist!"
6731 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6733 run_test 94 "ldev outputs correct labels for file system name query"
6736 if [ -z "$LDEV" ]; then
6737 error "ldev is missing!"
6740 local LDEVCONFPATH=$TMP/ldev.conf
6741 local NIDSPATH=$TMP/nids
6743 generate_ldev_conf $LDEVCONFPATH
6744 generate_nids $NIDSPATH
6748 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME &>/dev/null
6749 if [ $? -ne 0 ]; then
6750 rm $LDEVCONFPATH $NIDSPATH
6751 error "ldev label filtering w/ -F failed!"
6755 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -l &>/dev/null
6756 if [ $? -ne 0 ]; then
6757 rm $LDEVCONFPATH $NIDSPATH
6758 error "ldev label filtering w/ -l failed!"
6762 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f &>/dev/null
6763 if [ $? -ne 0 ]; then
6764 rm $LDEVCONFPATH $NIDSPATH
6765 error "ldev label filtering w/ -f failed!"
6769 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a &>/dev/null
6770 if [ $? -ne 0 ]; then
6771 rm $LDEVCONFPATH $NIDSPATH
6772 error "ldev label filtering w/ -a failed!"
6777 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -F $FSNAME &>/dev/null
6778 if [ $? -eq 0 ]; then
6779 rm $LDEVCONFPATH $NIDSPATH
6780 error "ldev label filtering w/ -a and -F incorrectly succeeded"
6784 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -f &>/dev/null
6785 if [ $? -eq 0 ]; then
6786 rm $LDEVCONFPATH $NIDSPATH
6787 error "ldev label filtering w/ -a and -f incorrectly succeeded"
6791 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -l &>/dev/null
6792 if [ $? -eq 0 ]; then
6793 rm $LDEVCONFPATH $NIDSPATH
6794 error "ldev label filtering w/ -a and -l incorrectly succeeded"
6798 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f -l &>/dev/null
6799 if [ $? -eq 0 ]; then
6800 rm $LDEVCONFPATH $NIDSPATH
6801 error "ldev label filtering w/ -f and -l incorrectly succeeded"
6805 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -l &>/dev/null
6806 if [ $? -eq 0 ]; then
6807 rm $LDEVCONFPATH $NIDSPATH
6808 error "ldev label filtering w/ -F and -l incorrectly succeeded"
6811 # file sys & foreign
6812 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -f &>/dev/null
6813 if [ $? -eq 0 ]; then
6814 rm $LDEVCONFPATH $NIDSPATH
6815 error "ldev label filtering w/ -F and -f incorrectly succeeded"
6818 rm $LDEVCONFPATH $NIDSPATH
6820 run_test 95 "ldev should only allow one label filter"
6823 if [ -z "$LDEV" ]; then
6824 error "ldev is missing!"
6827 local LDEVCONFPATH=$TMP/ldev.conf
6828 local NIDSPATH=$TMP/nids
6830 generate_ldev_conf $LDEVCONFPATH
6831 generate_nids $NIDSPATH
6833 local LDEV_OUTPUT=$TMP/ldev-output.txt
6834 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -H $mgs_HOST \
6836 awk '{print $2}' > $LDEV_OUTPUT
6838 # ldev failed, error
6839 if [ $? -ne 0 ]; then
6840 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
6841 error "ldev failed to execute!"
6845 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
6847 echo "$mgs_HOST-$(facet_fstype mgs)" > $EXPECTED_OUTPUT
6849 if [ "$mgs_HOST" == "$mds_HOST" ]; then
6850 for num in $(seq $MDSCOUNT); do
6851 echo "$mds_HOST-$(facet_fstype mds$num)" \
6856 if [ "$mgs_HOST" == "$ost_HOST" ]; then
6857 for num in $(seq $OSTCOUNT); do
6858 echo "$ost_HOST-$(facet_fstype ost$num)" \
6863 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6865 if [ $? -ne 0 ]; then
6866 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6867 error "ldev failed to produce the correct output!"
6870 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6872 run_test 96 "ldev returns hostname and backend fs correctly in command sub"
6875 if [ -z "$LDEV" ]; then
6876 error "ldev is missing!"
6879 local LDEVCONFPATH=$TMP/ldev.conf
6880 local NIDSPATH=$TMP/nids
6882 generate_ldev_conf $LDEVCONFPATH
6883 generate_nids $NIDSPATH
6885 local LDEV_OUTPUT=$TMP/ldev-output.txt
6886 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
6888 echo -e "\nMDT role"
6889 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mdt > $LDEV_OUTPUT
6891 if [ $? -ne 0 ]; then
6892 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
6893 error "ldev failed to execute for mdt role!"
6896 for num in $(seq $MDSCOUNT); do
6897 printf "%s-MDT%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
6900 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6902 if [ $? -ne 0 ]; then
6903 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6904 error "ldev failed to produce the correct output for mdt role!"
6907 echo -e "\nOST role"
6908 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R ost > $LDEV_OUTPUT
6910 if [ $? -ne 0 ]; then
6911 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
6912 error "ldev failed to execute for ost role!"
6916 for num in $(seq $OSTCOUNT); do
6917 printf "%s-OST%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
6920 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6922 if [ $? -ne 0 ]; then
6923 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6924 error "ldev failed to produce the correct output for ost role!"
6927 echo -e "\nMGS role"
6928 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mgs > $LDEV_OUTPUT
6930 if [ $? -ne 0 ]; then
6931 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
6932 error "ldev failed to execute for mgs role!"
6935 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
6937 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6939 if [ $? -ne 0 ]; then
6940 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6941 error "ldev failed to produce the correct output for mgs role!"
6944 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6946 run_test 97 "ldev returns correct ouput when querying based on role"
6951 local temp=$MDS_MOUNT_OPTS
6954 check_mount || error "mount failed"
6955 mountopt="user_xattr"
6956 for ((x = 1; x <= 400; x++)); do
6957 mountopt="$mountopt,user_xattr"
6959 remount_client $mountopt $MOUNT 2>&1 | grep "too long" ||
6960 error "Buffer overflow check failed"
6961 cleanup || error "cleanup failed"
6963 run_test 98 "Buffer-overflow check while parsing mount_opts"
6967 [[ $(facet_fstype ost1) != ldiskfs ]] &&
6968 { skip "Only applicable to ldiskfs-based OSTs" && return; }
6969 [[ $(lustre_version_code ost1) -ge $(version_code 2.8.57) ]] ||
6970 { skip "Need OST version at least 2.8.57" && return 0; }
6972 local ost_opts="$(mkfs_opts ost1 $(ostdevname 1)) \
6973 --reformat $(ostdevname 1) $(ostvdevname 1)"
6974 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" &&
6975 skip "meta_bg already set" && return
6978 if [[ ${!opts} != *mkfsoptions* ]]; then
6979 eval opts=\"${!opts} \
6980 --mkfsoptions='\\\"-O ^resize_inode,meta_bg\\\"'\"
6982 local val=${!opts//--mkfsoptions=\\\"/ \
6983 --mkfsoptions=\\\"-O ^resize_inode,meta_bg }
6987 echo "params: $opts"
6989 add ost1 $opts || error "add ost1 failed with new params"
6991 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" ||
6992 error "meta_bg is not set"
6996 run_test 99 "Adding meta_bg option"
7000 start_mds || error "MDS start failed"
7001 start_ost || error "unable to start OST"
7002 mount_client $MOUNT || error "client start failed"
7003 check_mount || error "check_mount failed"
7012 do_facet mgs 'lshowmount -v' | awk 'BEGIN {NR == 0; rc=1} /MGS:/ {rc=0}
7013 END {exit rc}' || error "lshowmount have no output MGS"
7015 do_facet mds1 'lshowmount -v' | awk 'BEGIN {NR == 2; rc=1} /-MDT0000:/
7016 {rc=0} END {exit rc}' || error "lshowmount have no output MDT0"
7018 do_facet ost1 'lshowmount -v' | awk 'BEGIN {NR == 4; rc=1} /-OST0000:/
7019 {rc=0} END {exit rc}' || error "lshowmount have no output OST0"
7021 cleanup || error "cleanup failed with $?"
7023 run_test 100 "check lshowmount lists MGS, MDT, OST and 0@lo"
7026 local createmany_oid
7027 local dev=$FSNAME-OST0000-osc-MDT0000
7030 createmany -o $DIR1/$tfile-%d 50000 &
7032 # MDT->OST reconnection causes MDT<->OST last_id synchornisation
7033 # via osp_precreate_cleanup_orphans.
7034 for ((i = 0; i < 100; i++)); do
7035 for ((k = 0; k < 10; k++)); do
7036 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;" \
7037 "$LCTL --device $dev activate"
7040 ls -asl $MOUNT | grep '???' &&
7041 (kill -9 $createmany_oid &>/dev/null; \
7042 error "File hasn't object on OST")
7044 kill -s 0 $createmany_oid || break
7046 wait $createmany_oid
7049 run_test 101 "Race MDT->OST reconnection with create"
7051 if ! combined_mgs_mds ; then
7057 # restore the values of MDSSIZE and OSTSIZE
7058 MDSSIZE=$STORED_MDSSIZE
7059 OSTSIZE=$STORED_OSTSIZE