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
3708 [ $(facet_fstype mds2) == zfs ] && import_zpool mds2
3709 do_facet mds2 "$TUNEFS --param mdc.active=0 $(mdsdevname 2)" ||
3710 error "tunefs MDT2 failed"
3711 start_mds || error "Unable to start MDT"
3712 start_ost || error "Unable to start OST1"
3713 start_ost2 || error "Unable to start OST2"
3714 mount_client $MOUNT || error "client start failed"
3716 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3718 $LCTL conf_param ${FSNAME}-MDT0000.mdc.active=0 &&
3719 error "deactive MDC0 succeeds"
3720 # activate MDC for MDT2
3721 local TEST="$LCTL get_param -n mdc.${FSNAME}-MDT0001-mdc-[!M]*.active"
3722 set_conf_param_and_check client \
3723 "$TEST" "${FSNAME}-MDT0001.mdc.active" 1 ||
3724 error "Unable to activate MDT2"
3726 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
3727 if [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.60) ]
3729 wait_dne_interconnect
3731 $LFS mkdir -i1 $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3733 createmany -o $DIR/$tdir/2/$tfile-%d 1 || error "create files failed"
3735 rm -rf $DIR/$tdir/2 || error "unlink dir failed"
3737 # deactivate MDC for MDT2
3738 local TEST="$LCTL get_param -n mdc.${FSNAME}-MDT0001-mdc-[!M]*.active"
3739 set_conf_param_and_check client \
3740 "$TEST" "${FSNAME}-MDT0001.mdc.active" 0 ||
3741 error "Unable to deactivate MDT2"
3743 $LFS mkdir -i1 $DIR/$tdir/2 &&
3744 error "mkdir $DIR/$tdir/2 succeeds after deactive MDT"
3747 umount_client $MOUNT || error "Unable to umount client"
3752 run_test 50i "activate deactivated MDT"
3755 local LOCAL_TIMEOUT=20
3759 check_mount || error "check_mount failed"
3761 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
3762 $SETSTRIPE -c -1 $MOUNT/$tdir ||
3763 error "$SETSTRIPE -c -1 $MOUNT/$tdir failed"
3764 #define OBD_FAIL_MDS_REINT_DELAY 0x142
3765 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x142"
3766 touch $MOUNT/$tdir/$tfile &
3769 start_ost2 || error "Unable to start OST1"
3771 stop_ost2 || error "Unable to stop OST1"
3772 umount_client $MOUNT -f || error “unmount $MOUNT failed”
3773 cleanup_nocli || error “stop server failed”
3774 #writeconf to remove all ost2 traces for subsequent tests
3775 writeconf_or_reformat
3777 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
3786 do_node $node mkdir -p $dest
3787 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3789 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
3790 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3791 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
3793 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
3794 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
3804 local backup2=${TMP}/backup2
3806 do_node $node mkdir -p $backup2
3807 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3809 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
3810 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3811 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
3813 do_node $node "diff -rq $backup $backup2"
3814 [ $? -eq 0 ] || { error "contents differ"; return 3; }
3816 local xattrs2=${TMP}/xattrs2
3817 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
3818 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
3820 do_node $node "diff $xattrs $xattrs2"
3821 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
3823 do_node $node "rm -rf $backup2 $xattrs2"
3824 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
3828 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3829 skip "Only applicable to ldiskfs-based MDTs"
3833 start_mds || error "Unable to start MDS"
3834 start_ost || error "Unable to start OST1"
3835 mount_client $MOUNT || error "Unable to mount client"
3838 local ost1mnt=$(facet_mntpt ost1)
3839 local ost1node=$(facet_active_host ost1)
3840 local ost1tmp=$TMP/conf52
3843 mkdir $DIR/$tdir || error "Unable to create $DIR/$tdir"
3844 touch $TMP/modified_first || error "Unable to create temporary file"
3845 local mtime=$(stat -c %Y $TMP/modified_first)
3846 do_node $ost1node "mkdir -p $ost1tmp &&
3847 touch -m -d @$mtime $ost1tmp/modified_first" ||
3848 error "Unable to create temporary file"
3851 $SETSTRIPE -c -1 -S 1M $DIR/$tdir || error "$SETSTRIPE failed"
3853 for (( i=0; i < nrfiles; i++ )); do
3854 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c ||
3855 error "multiop failed"
3861 echo backup files to $TMP/$tdir
3862 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
3863 copy_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
3864 error "Unable to copy files"
3866 umount_client $MOUNT || error "Unable to umount client"
3867 stop_ost || error "Unable to stop ost1"
3869 echo mount ost1 as ldiskfs
3870 do_node $ost1node mkdir -p $ost1mnt || error "Unable to create $ost1mnt"
3871 if ! do_node $ost1node test -b $ost1_dev; then
3874 do_node $ost1node mount -t $(facet_fstype ost1) $loop $ost1_dev \
3876 error "Unable to mount ost1 as ldiskfs"
3879 echo backup objects to $ost1tmp/objects
3880 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
3881 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
3882 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
3884 error "Unable to copy objects"
3886 # move objects to lost+found
3887 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
3888 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
3890 do_node $ost1node "umount $ost1mnt" ||
3891 error "Unable to umount ost1 as ldiskfs"
3893 start_ost || error "Unable to start OST1"
3894 mount_client $MOUNT || error "Unable to mount client"
3896 local REPAIRED=$(do_node $ost1node "$LCTL get_param \
3897 -n osd-ldiskfs.$FSNAME-OST0000.oi_scrub" |
3898 awk '/^lf_repa[ri]*ed/ { print $2 }')
3899 [ $REPAIRED -gt 0 ] ||
3900 error "Some entry under /lost+found should be repaired"
3903 diff_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
3904 error "Unable to diff files"
3906 rm -rf $TMP/$tdir $TMP/file_xattrs ||
3907 error "Unable to delete temporary files"
3908 do_node $ost1node "rm -rf $ost1tmp" ||
3909 error "Unable to delete temporary files"
3910 cleanup || error "cleanup failed with $?"
3912 run_test 52 "check recovering objects from lost+found"
3914 # Checks threads_min/max/started for some service
3916 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
3917 # parameter pattern prefix like 'ost.*.ost'.
3930 local msg="Insane $modname thread counts"
3931 local ncpts=$(check_cpt_number $facet)
3935 check_mount || return 41
3937 # We need to expand $parampat, but it may match multiple parameters, so
3938 # we'll pick the first one
3939 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
3940 error "Couldn't expand ${parampat}.threads_min parameter name"
3944 # Remove the .threads_min part
3945 paramp=${paramp%.threads_min}
3947 # Check for sanity in defaults
3948 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3950 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3952 tstarted=$(do_facet $facet "$LCTL get_param \
3953 -n ${paramp}.threads_started" || echo 0)
3954 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' ||
3956 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
3958 nthrs=$(expr $tmax - $tmin)
3959 if [ $nthrs -lt $ncpts ]; then
3965 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
3966 skip_env "module parameter forced $facet thread count" &&
3967 tmin=3 && tmax=$((3 * tmax))
3969 # Check that we can change min/max
3970 do_facet $facet "$LCTL set_param \
3971 ${paramp}.threads_min=$((tmin + nthrs))"
3972 do_facet $facet "$LCTL set_param \
3973 ${paramp}.threads_max=$((tmax - nthrs))"
3974 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3976 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3978 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) &&
3979 $tmax2 == ($tmax - $nthrs)))' || return $?
3981 # Check that we can set min/max to the same value
3982 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3984 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$tmin"
3985 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3987 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3989 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
3991 # Check that we can't set max < min
3992 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$((tmin - 1))"
3993 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3995 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3997 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
3999 # We need to ensure that we get the module options desired; to do this
4000 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
4001 LOAD_MODULES_REMOTE=true
4004 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
4005 setmodopts -a $modname "$newvalue" oldvalue
4008 check_mount || return 41
4010 # Restore previous setting of MODOPTS_*
4011 setmodopts $modname "$oldvalue"
4013 # Check that $opts took
4014 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min")
4015 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max")
4016 tstarted=$(do_facet $facet \
4017 "$LCTL get_param -n ${paramp}.threads_started")
4018 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4027 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
4028 cleanup || error "cleanup failed with rc $?"
4030 run_test 53a "check OSS thread count params"
4034 local mds=$(do_facet $SINGLEMDS "$LCTL get_param \
4035 -N mds.*.*.threads_max 2>/dev/null")
4036 if [ -z "$mds" ]; then
4037 #running this on an old MDT
4038 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
4040 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
4042 cleanup || error "cleanup failed with $?"
4044 run_test 53b "check MDS thread count params"
4047 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4048 skip "Only applicable to ldiskfs-based MDTs"
4052 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p ||
4053 error "llverdev failed with rc=$?"
4056 run_test 54a "test llverdev and partial verify of device"
4059 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4060 skip "Only applicable to ldiskfs-based MDTs"
4065 run_llverfs $MOUNT -p || error "llverfs failed with rc=$?"
4066 cleanup || error "cleanup failed with rc=$?"
4068 run_test 54b "test llverfs and partial verify of filesystem"
4072 local max_ost_index=$1
4073 echo -n $(((max_ost_index + 1) * 8))
4077 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4078 skip "Only applicable to ldiskfs-based MDTs"
4082 local mdsdev=$(mdsdevname 1)
4083 local mdsvdev=$(mdsvdevname 1)
4087 if ! combined_mgs_mds; then
4088 stop_mgs || error "stopping MGS service failed"
4089 format_mgs || error "formatting MGT failed"
4091 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
4093 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
4094 --reformat $(ostdevname 1) $(ostvdevname 1)
4100 echo checking size of lov_objid for ost index $i
4101 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
4102 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
4103 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
4105 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
4112 run_test 55 "check lov_objid size"
4115 local server_version=$(lustre_version_code $SINGLEMDS)
4116 local mds_journal_size_orig=$MDSJOURNALSIZE
4121 for num in $(seq 1 $MDSCOUNT); do
4124 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
4125 $(ostdevname 1) $(ostvdevname 1)
4126 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
4127 $(ostdevname 2) $(ostvdevname 2)
4130 start_ost || error "Unable to start first ost (idx 10000)"
4131 start_ost2 || error "Unable to start second ost (idx 1000)"
4132 mount_client $MOUNT || error "Unable to mount client"
4136 if [[ $server_version -ge $(version_code 2.6.54) ]] ||
4137 [[ $server_version -ge $(version_code 2.5.4) &&
4138 $server_version -lt $(version_code 2.5.11) ]]; then
4139 wait_osc_import_state mds ost1 FULL
4140 wait_osc_import_state mds ost2 FULL
4141 $SETSTRIPE --stripe-count=-1 $DIR/$tfile ||
4142 error "Unable to setstripe $DIR/$tfile"
4143 n=$($LFS getstripe --stripe-count $DIR/$tfile)
4144 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
4149 MDSJOURNALSIZE=$mds_journal_size_orig
4152 run_test 56a "check big OST indexes and out-of-index-order start"
4157 umount_client $MOUNT -f || error "unmount client failed"
4166 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs" && return
4168 trap cleanup_56b EXIT RETURN ERR
4171 if ! combined_mgs_mds ; then
4176 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) --index=0 --reformat \
4177 $(mdsdevname 1) $(mdsvdevname 1)
4178 add mds2 $(mkfs_opts mds2 $(mdsdevname 2)) --index=1 --reformat \
4179 $(mdsdevname 2) $(mdsvdevname 2)
4180 add mds3 $(mkfs_opts mds3 $(mdsdevname 3)) --index=1000 --reformat \
4181 $(mdsdevname 3) $(mdsvdevname 3)
4185 start_mdt 1 || error "MDT 1 (idx 0) start failed"
4186 start_mdt 2 || error "MDT 2 (idx 1) start failed"
4187 start_mdt 3 || error "MDT 3 (idx 1000) start failed"
4188 start_ost || error "Unable to start first ost"
4189 start_ost2 || error "Unable to start second ost"
4191 do_nodes $(comma_list $(mdts_nodes)) \
4192 "$LCTL set_param mdt.*.enable_remote_dir=1 \
4193 mdt.*.enable_remote_dir_gid=-1"
4195 mount_client $MOUNT || error "Unable to mount client"
4197 $LFS mkdir -c3 $MOUNT/$tdir || error "failed to make testdir"
4199 echo "This is test file 1!" > $MOUNT/$tdir/$tfile.1 ||
4200 error "failed to make test file 1"
4201 echo "This is test file 2!" > $MOUNT/$tdir/$tfile.2 ||
4202 error "failed to make test file 2"
4203 echo "This is test file 1000!" > $MOUNT/$tdir/$tfile.1000 ||
4204 error "failed to make test file 1000"
4206 rm -rf $MOUNT/$tdir || error "failed to remove testdir"
4208 $LFS mkdir -i1000 $MOUNT/$tdir.1000 ||
4209 error "create remote dir at idx 1000 failed"
4212 echo "=== START lfs df OUTPUT ==="
4214 echo "==== END lfs df OUTPUT ===="
4216 mdtcnt=$(echo -e "$output" | grep $FSNAME-MDT | wc -l)
4217 ostcnt=$(echo -e "$output" | grep $FSNAME-OST | wc -l)
4219 echo "lfs df returned mdt count $mdtcnt and ost count $ostcnt"
4220 [ $mdtcnt -eq 3 ] || error "lfs df returned wrong mdt count"
4221 [ $ostcnt -eq 2 ] || error "lfs df returned wrong ost count"
4223 echo "This is test file 1!" > $MOUNT/$tdir.1000/$tfile.1 ||
4224 error "failed to make test file 1"
4225 echo "This is test file 2!" > $MOUNT/$tdir.1000/$tfile.2 ||
4226 error "failed to make test file 2"
4227 echo "This is test file 1000!" > $MOUNT/$tdir.1000/$tfile.1000 ||
4228 error "failed to make test file 1000"
4229 rm -rf $MOUNT/$tdir.1000 || error "failed to remove remote_dir"
4232 echo "=== START lfs mdts OUTPUT ==="
4234 echo "==== END lfs mdts OUTPUT ===="
4236 echo -e "$output" | grep -v "MDTS:" | awk '{print $1}' |
4237 sed 's/://g' > $TMP/mdts-actual.txt
4238 sort $TMP/mdts-actual.txt -o $TMP/mdts-actual.txt
4240 echo -e "0\n1\n1000" > $TMP/mdts-expected.txt
4242 diff $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4245 rm $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4247 [ $result -eq 0 ] || error "target_obd proc file is incorrect!"
4249 run_test 56b "test target_obd correctness with nonconsecutive MDTs"
4251 test_57a() { # bug 22656
4252 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4253 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4254 tail -1 | awk '{print $1}')
4255 writeconf_or_reformat
4256 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4257 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" ||
4258 error "tunefs failed"
4260 start_ost && error "OST registration from failnode should fail"
4263 run_test 57a "initial registration from failnode should fail (should return errs)"
4266 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4267 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4268 tail -1 | awk '{print $1}')
4269 writeconf_or_reformat
4270 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4271 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" ||
4272 error "tunefs failed"
4274 start_ost || error "OST registration from servicenode should not fail"
4277 run_test 57b "initial registration from servicenode should not fail"
4280 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
4283 test_58() { # bug 22658
4284 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
4286 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4287 createmany -o $DIR/$tdir/$tfile-%d 100
4288 # make sure that OSTs do not cancel llog cookies before we unmount the MDS
4289 #define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601
4290 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x601"
4291 unlinkmany $DIR/$tdir/$tfile-%d 100
4292 stop_mds || error "Unable to stop MDS"
4294 local MNTDIR=$(facet_mntpt $SINGLEMDS)
4295 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4297 # remove all files from the OBJECTS dir
4298 mount_fstype $SINGLEMDS
4300 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
4302 unmount_fstype $SINGLEMDS
4303 # restart MDS with missing llog files
4304 start_mds || error "unable to start MDS"
4305 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0"
4308 run_test 58 "missing llog files must not prevent MDT from mounting"
4311 start_mgsmds >> /dev/null
4312 local C1=$(count_osts)
4313 if [ $C1 -eq 0 ]; then
4314 start_ost >> /dev/null
4318 echo "original ost count: $C1 (expect > 0)"
4319 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
4320 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
4321 local C2=$(count_osts)
4322 echo "after mdt writeconf count: $C2 (expect 0)"
4323 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
4324 echo "OST start without writeconf should fail:"
4325 start_ost >> /dev/null &&
4326 error "OST start without writeconf didn't fail"
4327 echo "OST start with writeconf should succeed:"
4328 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
4329 local C3=$(count_osts)
4330 echo "after ost writeconf count: $C3 (expect 1)"
4331 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
4332 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
4333 local C4=$(count_osts)
4334 echo "after ost2 writeconf count: $C4 (expect 2)"
4335 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
4336 stop_ost2 >> /dev/null
4337 cleanup_nocli >> /dev/null
4338 #writeconf to remove all ost2 traces for subsequent tests
4339 writeconf_or_reformat
4341 run_test 59 "writeconf mount option"
4343 test_60() { # LU-471
4346 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4347 skip "Only applicable to ldiskfs-based MDTs"
4351 for num in $(seq $MDSCOUNT); do
4352 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
4353 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
4354 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4358 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
4359 [ ${PIPESTATUS[0]} -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
4361 # MDT default has dirdata feature
4362 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
4363 # we disable uninit_bg feature
4364 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
4365 # we set stride extended options
4366 echo $dump | grep stride > /dev/null || error "stride is not set"
4369 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
4374 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] ||
4375 { skip "Need MDS version at least 2.1.53"; return 0; }
4377 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
4378 ! large_xattr_enabled; then
4381 for num in $(seq $MDSCOUNT); do
4382 do_facet mds${num} $TUNE2FS -O large_xattr \
4383 $(mdsdevname $num) ||
4384 error "tune2fs on mds $num failed"
4388 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
4389 setup_noconfig || error "setting up the filesystem failed"
4390 client_up || error "starting client failed"
4392 local file=$DIR/$tfile
4393 touch $file || error "touch $file failed"
4395 local large_value="$(generate_string $(max_xattr_size))"
4396 local small_value="bar"
4398 local name="trusted.big"
4399 log "save large xattr $name on $file"
4400 setfattr -n $name -v $large_value $file ||
4401 error "saving $name on $file failed"
4403 local new_value=$(get_xattr_value $name $file)
4404 [[ "$new_value" != "$large_value" ]] &&
4405 error "$name different after saving"
4407 log "shrink value of $name on $file"
4408 setfattr -n $name -v $small_value $file ||
4409 error "shrinking value of $name on $file failed"
4411 new_value=$(get_xattr_value $name $file)
4412 [[ "$new_value" != "$small_value" ]] &&
4413 error "$name different after shrinking"
4415 log "grow value of $name on $file"
4416 setfattr -n $name -v $large_value $file ||
4417 error "growing value of $name on $file failed"
4419 new_value=$(get_xattr_value $name $file)
4420 [[ "$new_value" != "$large_value" ]] &&
4421 error "$name different after growing"
4423 log "check value of $name on $file after remounting MDS"
4425 new_value=$(get_xattr_value $name $file)
4426 [[ "$new_value" != "$large_value" ]] &&
4427 error "$name different after remounting MDS"
4429 log "remove large xattr $name from $file"
4430 setfattr -x $name $file || error "removing $name from $file failed"
4433 stopall || error "stopping for e2fsck run"
4434 for num in $(seq $MDSCOUNT); do
4435 run_e2fsck $(facet_active_host mds$num) \
4436 $(mdsdevname $num) "-y" ||
4437 error "e2fsck MDT$num failed"
4439 setup_noconfig || error "remounting the filesystem failed"
4442 # need to delete this file to avoid problems in other tests
4444 stopall || error "stopping systems to turn off large_xattr"
4446 for num in $(seq $MDSCOUNT); do
4447 do_facet mds${num} $TUNE2FS -O ^large_xattr \
4448 $(mdsdevname $num) ||
4449 error "tune2fs on mds $num failed"
4453 run_test 61 "large xattr"
4456 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4457 skip "Only applicable to ldiskfs-based MDTs"
4462 local mdsdev=$(mdsdevname 1)
4463 local ostdev=$(ostdevname 1)
4465 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] ||
4466 { skip "Need MDS version at least 2.2.51"; return 0; }
4468 echo "disable journal for mds"
4469 do_facet mds1 $TUNE2FS -O ^has_journal $mdsdev || error "tune2fs failed"
4470 start_mds && error "MDT start should fail"
4471 echo "disable journal for ost"
4472 do_facet ost1 $TUNE2FS -O ^has_journal $ostdev || error "tune2fs failed"
4473 start_ost && error "OST start should fail"
4474 cleanup || error "cleanup failed with rc $?"
4477 run_test 62 "start with disabled journal"
4480 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4481 skip "Only applicable to ldiskfs-based MDTs"
4485 local inode_slab=$(do_facet $SINGLEMDS \
4486 "awk '/ldiskfs_inode_cache/ { print \\\$5 }' /proc/slabinfo")
4487 if [ -z "$inode_slab" ]; then
4488 skip "ldiskfs module has not been loaded"
4492 echo "$inode_slab ldisk inodes per page"
4493 if [ "$inode_slab" -ge "3" ] ; then
4494 # If kmalloc-128 is also 1 per page - this is a debug kernel
4495 # and so this is not an error.
4496 local kmalloc128=$(do_facet $SINGLEMDS \
4497 "awk '/^(kmalloc|size)-128 / { print \\\$5 }' /proc/slabinfo")
4498 # 32 128-byte chunks in 4k
4499 [ "$kmalloc128" -eq "32" ] ||
4500 error "ldisk inode size is too big, $inode_slab objs per page"
4505 run_test 63 "Verify each page can at least hold 3 ldisk inodes"
4508 start_mds || error "unable to start MDS"
4509 start_ost || error "Unable to start OST1"
4510 start_ost2 || error "Unable to start second ost"
4511 mount_client $MOUNT || error "Unable to mount client"
4512 stop_ost2 || error "Unable to stop second ost"
4515 umount_client $MOUNT -f || error “unmount $MOUNT failed”
4516 cleanup_nocli || error "cleanup_nocli failed with $?"
4517 #writeconf to remove all ost2 traces for subsequent tests
4518 writeconf_or_reformat
4520 run_test 64 "check lfs df --lazy "
4522 test_65() { # LU-2237
4523 # Currently, the test is only valid for ldiskfs backend
4524 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4525 skip "non-ldiskfs backend" && return
4527 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4528 local brpt=$(facet_mntpt brpt)
4531 if ! do_facet $SINGLEMDS "test -b $devname"; then
4535 stop_mds || error "Unable to stop MDS"
4536 local obj=$(do_facet $SINGLEMDS \
4537 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
4539 if [ -z "$obj" ]; then
4540 # The MDT may be just re-formatted, mount the MDT for the
4541 # first time to guarantee the "last_rcvd" file is there.
4542 start_mds || error "fail to mount the MDS for the first time"
4543 stop_mds || error "Unable to stop MDS"
4546 # remove the "last_rcvd" file
4547 do_facet $SINGLEMDS "mkdir -p $brpt"
4548 do_facet $SINGLEMDS \
4549 "mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
4550 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
4551 do_facet $SINGLEMDS "$UMOUNT $brpt"
4553 # restart MDS, the "last_rcvd" file should be recreated.
4554 start_mds || error "fail to restart the MDS"
4555 stop_mds || error "Unable to stop MDS"
4556 obj=$(do_facet $SINGLEMDS \
4557 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
4558 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
4560 run_test 65 "re-create the lost last_rcvd file when server mount"
4563 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
4564 { skip "Need MGS version at least 2.3.59"; return 0; }
4567 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
4568 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
4570 echo "replace_nids should fail if MDS, OSTs and clients are UP"
4571 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4572 error "replace_nids fail"
4574 umount_client $MOUNT || error "unmounting client failed"
4575 echo "replace_nids should fail if MDS and OSTs are UP"
4576 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4577 error "replace_nids fail"
4579 stop_ost || error "Unable to stop OST1"
4580 echo "replace_nids should fail if MDS is UP"
4581 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4582 error "replace_nids fail"
4584 stop_mds || error "stopping mds failed"
4586 if combined_mgs_mds; then
4587 start_mdt 1 "-o nosvc" ||
4588 error "starting mds with nosvc option failed"
4591 echo "command should accept two parameters"
4592 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
4593 error "command should accept two params"
4595 echo "correct device name should be passed"
4596 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
4597 error "wrong devname"
4599 echo "wrong nids list should not destroy the system"
4600 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
4603 echo "replace OST nid"
4604 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
4605 error "replace nids failed"
4607 echo "command should accept two parameters"
4608 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
4609 error "command should accept two params"
4611 echo "wrong nids list should not destroy the system"
4612 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
4615 echo "replace MDS nid"
4616 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
4617 error "replace nids failed"
4619 if ! combined_mgs_mds ; then
4622 stop_mds || error "Unable to stop MDS"
4626 check_mount || error "error after nid replace"
4627 cleanup || error "cleanup failed"
4630 run_test 66 "replace nids"
4632 test_67() { #LU-2950
4633 local legacy="$TMP/legacy_lnet_config"
4634 local new="$TMP/new_routes_test"
4635 local out="$TMP/config_out_file"
4636 local verify="$TMP/conv_verify"
4637 local verify_conf="$TMP/conf_verify"
4639 # Create the legacy file that will be run through the
4640 # lustre_routes_conversion script
4641 cat <<- LEGACY_LNET_CONFIG > $legacy
4642 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
4643 tcp2 54 10.1.3.2@tcp;
4644 tcp3 10.3.4.3@tcp:3;
4648 # Create the verification file to verify the output of
4649 # lustre_routes_conversion script against.
4650 cat <<- VERIFY_LNET_CONFIG > $verify
4651 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
4652 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
4653 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
4654 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
4655 tcp4: { gateway: 10.3.3.4@tcp }
4658 # Create the verification file to verify the output of
4659 # lustre_routes_config script against
4660 cat <<- VERIFY_LNET_CONFIG > $verify_conf
4661 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
4662 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
4663 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
4664 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
4665 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
4668 lustre_routes_conversion $legacy $new > /dev/null
4669 if [ -f $new ]; then
4670 # verify the conversion output
4671 cmp -s $new $verify > /dev/null
4672 if [ $? -eq 1 ]; then
4673 error "routes conversion failed"
4676 lustre_routes_config --dry-run --verbose $new > $out
4677 # check that the script succeeded
4678 cmp -s $out $verify_conf > /dev/null
4679 if [ $? -eq 1 ]; then
4680 error "routes config failed"
4683 error "routes conversion test failed"
4685 # remove generated files
4686 rm -f $new $legacy $verify $verify_conf $out
4688 run_test 67 "test routes conversion and configuration"
4696 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.4.53) ] ||
4697 { skip "Need MDS version at least 2.4.53"; return 0; }
4699 umount_client $MOUNT || error "umount client failed"
4701 start_mdt 1 || error "MDT start failed"
4702 start_ost || error "Unable to start OST1"
4704 # START-END - the sequences we'll be reserving
4705 START=$(do_facet $SINGLEMDS \
4706 $LCTL get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
4707 END=$((START + (1 << 30)))
4708 do_facet $SINGLEMDS \
4709 $LCTL set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
4711 # reset the sequences MDT0000 has already assigned
4712 do_facet $SINGLEMDS \
4713 $LCTL set_param seq.srv*MDT0000.space=clear
4715 # remount to let the client allocate new sequence
4716 mount_client $MOUNT || error "mount client failed"
4718 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4719 do_facet $SINGLEMDS \
4720 $LCTL get_param seq.srv*MDT0000.space
4721 $LFS path2fid $DIR/$tfile
4723 local old_ifs="$IFS"
4725 fid=($($LFS path2fid $DIR/$tfile))
4729 if [[ $seq < $END ]]; then
4730 error "used reserved sequence $seq?"
4732 cleanup || error "cleanup failed with $?"
4734 run_test 68 "be able to reserve specific sequences in FLDB"
4736 # Test 69: is about the total number of objects ever created on an OST.
4737 # so that when it is reformatted the normal MDS->OST orphan recovery won't
4738 # just "precreate" the missing objects. In the past it might try to recreate
4739 # millions of objects after an OST was reformatted
4741 local server_version=$(lustre_version_code $SINGLEMDS)
4743 [[ $server_version -lt $(version_code 2.4.2) ]] &&
4744 skip "Need MDS version at least 2.4.2" && return
4746 [[ $server_version -ge $(version_code 2.4.50) ]] &&
4747 [[ $server_version -lt $(version_code 2.5.0) ]] &&
4748 skip "Need MDS version at least 2.5.0" && return
4751 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4753 # use OST0000 since it probably has the most creations
4754 local OSTNAME=$(ostname_from_index 0)
4755 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
4756 local last_id=$(do_facet mds1 $LCTL get_param -n \
4757 osc.$mdtosc_proc1.prealloc_last_id)
4759 # Want to have OST LAST_ID over 5 * OST_MAX_PRECREATE to
4760 # verify that the LAST_ID recovery is working properly. If
4761 # not, then the OST will refuse to allow the MDS connect
4762 # because the LAST_ID value is too different from the MDS
4763 #define OST_MAX_PRECREATE=20000
4764 local ost_max_pre=20000
4765 local num_create=$(( ost_max_pre * 5 + 1 - last_id))
4767 # If the LAST_ID is already over 5 * OST_MAX_PRECREATE, we don't
4768 # need to create any files. So, skip this section.
4769 if [ $num_create -gt 0 ]; then
4770 # Check the number of inodes available on OST0
4772 local ifree=$($LFS df -i $MOUNT | awk '/OST0000/ { print $4 }')
4773 log "On OST0, $ifree inodes available. Want $num_create."
4775 $SETSTRIPE -i 0 $DIR/$tdir ||
4776 error "$SETSTRIPE -i 0 $DIR/$tdir failed"
4777 if [ $ifree -lt 10000 ]; then
4778 files=$(( ifree - 50 ))
4783 local j=$((num_create / files + 1))
4784 for i in $(seq 1 $j); do
4785 createmany -o $DIR/$tdir/$tfile-$i- $files ||
4786 error "createmany fail create $files files: $?"
4787 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
4788 error "unlinkmany failed unlink $files files"
4792 # delete all of the files with objects on OST0 so the
4793 # filesystem is not inconsistent later on
4794 $LFS find $MOUNT --ost 0 -print0 | xargs -0 rm
4796 umount_client $MOUNT || error "umount client failed"
4797 stop_ost || error "OST0 stop failure"
4798 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
4799 $(ostdevname 1) $(ostvdevname 1) ||
4800 error "reformat and replace $ostdev failed"
4801 start_ost || error "OST0 restart failure"
4802 wait_osc_import_state mds ost FULL
4804 mount_client $MOUNT || error "mount client failed"
4805 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
4806 local idx=$($GETSTRIPE -i $DIR/$tdir/$tfile-last)
4807 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
4809 local iused=$($LFS df -i $MOUNT | awk '/OST0000/ { print $3 }')
4810 log "On OST0, $iused used inodes"
4811 [ $iused -ge $((ost_max_pre/2 + 1000)) ] &&
4812 error "OST replacement created too many inodes; $iused"
4813 cleanup || error "cleanup failed with $?"
4815 run_test 69 "replace an OST with the same index"
4818 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4821 cleanup || error "cleanup failed with $?"
4823 start_mdt 1 || error "MDT0 start fail"
4825 start_ost || error "OST0 start fail"
4826 for num in $(seq 2 $MDSCOUNT); do
4827 start_mdt $num || return
4830 mount_client $MOUNT || error "mount client fails"
4832 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
4834 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4835 error "create remote dir fail"
4837 rm -rf $DIR/$tdir || error "delete dir fail"
4838 cleanup || error "cleanup failed with $?"
4840 run_test 70a "start MDT0, then OST, then MDT1"
4843 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4846 start_ost || error "OST0 start fail"
4848 start_mds || error "MDS start fail"
4850 mount_client $MOUNT || error "mount client fails"
4852 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
4854 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4855 error "create remote dir fail"
4857 rm -rf $DIR/$tdir || error "delete dir fail"
4859 cleanup || error "cleanup failed with $?"
4861 run_test 70b "start OST, MDT1, MDT0"
4864 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4867 start_mds || error "MDS start fail"
4868 start_ost || error "OST0 start fail"
4870 mount_client $MOUNT || error "mount client fails"
4871 stop_mdt 1 || error "MDT1 start fail"
4873 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
4874 echo "deactivate $mdc_for_mdt1"
4875 $LCTL --device $mdc_for_mdt1 deactivate ||
4876 error "set $mdc_for_mdt1 deactivate failed"
4878 mkdir $DIR/$tdir && error "mkdir succeed"
4880 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4881 error "create remote dir succeed"
4883 cleanup || error "cleanup failed with $?"
4885 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
4888 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4891 start_mds || error "MDS start fail"
4892 start_ost || error "OST0 start fail"
4894 mount_client $MOUNT || error "mount client fails"
4896 stop_mdt 2 || error "MDT1 start fail"
4898 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
4900 echo "deactivate $mdc_for_mdt2"
4901 $LCTL --device $mdc_for_mdt2 deactivate ||
4902 error "set $mdc_for_mdt2 deactivate failed"
4904 mkdir $DIR/$tdir || error "mkdir fail"
4905 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4906 error "create remote dir succeed"
4908 rm -rf $DIR/$tdir || error "delete dir fail"
4910 cleanup || error "cleanup failed with $?"
4912 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
4915 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4917 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.62) ] ||
4918 { skip "Need MDS version at least 2.7.62"; return 0; }
4920 cleanup || error "cleanup failed with $?"
4922 local mdsdev=$(mdsdevname 1)
4923 local ostdev=$(ostdevname 1)
4924 local mdsvdev=$(mdsvdevname 1)
4925 local ostvdev=$(ostvdevname 1)
4926 local opts_mds="$(mkfs_opts mds1 $mdsdev) --reformat $mdsdev $mdsvdev"
4927 local opts_ost="$(mkfs_opts ost1 $ostdev) --reformat $ostdev $ostvdev"
4929 add mds1 $opts_mds || error "add mds1 failed"
4930 start_mdt 1 || error "start mdt1 failed"
4931 add ost1 $opts_ost || error "add ost1 failed"
4932 start_ost || error "start ost failed"
4933 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
4935 local soc=$(do_facet mds1 "$LCTL get_param -n \
4936 mdt.*MDT0000.sync_lock_cancel")
4937 [ $soc == "never" ] || error "SoC enabled on single MDS"
4939 for i in $(seq 2 $MDSCOUNT); do
4940 mdsdev=$(mdsdevname $i)
4941 mdsvdev=$(mdsvdevname $i)
4942 opts_mds="$(mkfs_opts mds$i $mdsdev) --reformat $mdsdev \
4944 add mds$i $opts_mds || error "add mds$i failed"
4945 start_mdt $i || error "start mdt$i fail"
4948 wait_dne_interconnect
4950 for i in $(seq $MDSCOUNT); do
4951 soc=$(do_facet mds$i "$LCTL get_param -n \
4952 mdt.*MDT000$((i - 1)).sync_lock_cancel")
4953 [ $soc == "blocking" ] || error "SoC not enabled on DNE"
4956 for i in $(seq 2 $MDSCOUNT); do
4957 stop_mdt $i || error "stop mdt$i fail"
4959 soc=$(do_facet mds1 "$LCTL get_param -n \
4960 mdt.*MDT0000.sync_lock_cancel")
4961 [ $soc == "never" ] || error "SoC enabled on single MDS"
4963 cleanup || error "cleanup failed with $?"
4965 run_test 70e "Sync-on-Cancel will be enabled by default on DNE"
4968 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4969 if combined_mgs_mds; then
4970 skip "needs separate MGS/MDT" && return
4974 start_mdt 1 || error "MDT0 start fail"
4975 start_ost || error "OST0 start fail"
4976 for num in $(seq 2 $MDSCOUNT); do
4977 start_mdt $num || return
4980 start_ost2 || error "OST1 start fail"
4982 mount_client $MOUNT || error "mount client fails"
4984 mkdir $DIR/$tdir || error "mkdir fail"
4985 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4986 error "create remote dir succeed"
4988 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4989 rm -rf $DIR/$tdir || error "delete dir fail"
4991 umount_client $MOUNT || error "umount_client failed"
4992 stop_mds || error "MDS stop fail"
4993 stop_ost || error "OST0 stop fail"
4994 stop_ost2 || error "OST1 stop fail"
4996 run_test 71a "start MDT0 OST0, MDT1, OST1"
4999 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5000 if combined_mgs_mds; then
5001 skip "needs separate MGS/MDT" && return
5005 for num in $(seq 2 $MDSCOUNT); do
5006 start_mdt $num || return
5008 start_ost || error "OST0 start fail"
5009 start_mdt 1 || error "MDT0 start fail"
5010 start_ost2 || error "OST1 start fail"
5012 mount_client $MOUNT || error "mount client fails"
5014 mkdir $DIR/$tdir || error "mkdir fail"
5015 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5016 error "create remote dir succeed"
5018 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5019 rm -rf $DIR/$tdir || error "delete dir fail"
5021 umount_client $MOUNT || error "umount_client failed"
5022 stop_mds || error "MDT0 stop fail"
5023 stop_ost || error "OST0 stop fail"
5024 stop_ost2 || error "OST1 stop fail"
5026 run_test 71b "start MDT1, OST0, MDT0, OST1"
5029 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5030 if combined_mgs_mds; then
5031 skip "needs separate MGS/MDT" && return
5035 start_ost || error "OST0 start fail"
5036 start_ost2 || error "OST1 start fail"
5037 for num in $(seq 2 $MDSCOUNT); do
5038 start_mdt $num || return
5040 start_mdt 1 || error "MDT0 start fail"
5042 mount_client $MOUNT || error "mount client fails"
5044 mkdir $DIR/$tdir || error "mkdir fail"
5045 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5046 error "create remote dir succeed"
5048 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5049 rm -rf $DIR/$tdir || error "delete dir fail"
5051 umount_client $MOUNT || error "umount_client failed"
5052 stop_mds || error "MDS stop fail"
5053 stop_ost || error "OST0 stop fail"
5054 stop_ost2 || error "OST1 stop fail"
5057 run_test 71c "start OST0, OST1, MDT1, MDT0"
5060 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5061 if combined_mgs_mds; then
5062 skip "needs separate MGS/MDT" && return
5066 start_ost || error "OST0 start fail"
5067 for num in $(seq 2 $MDSCOUNT); do
5068 start_mdt $num || return
5070 start_mdt 1 || error "MDT0 start fail"
5071 start_ost2 || error "OST1 start fail"
5073 mount_client $MOUNT || error "mount client fails"
5075 mkdir $DIR/$tdir || error "mkdir fail"
5076 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5077 error "create remote dir succeed"
5079 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5080 rm -rf $DIR/$tdir || error "delete dir fail"
5082 umount_client $MOUNT || error "umount_client failed"
5083 stop_mds || error "MDS stop fail"
5084 stop_ost || error "OST0 stop fail"
5085 stop_ost2 || error "OST1 stop fail"
5088 run_test 71d "start OST0, MDT1, MDT0, OST1"
5091 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5092 if combined_mgs_mds; then
5093 skip "needs separate MGS/MDT" && return
5097 start_ost || error "OST0 start fail"
5098 for num in $(seq 2 $MDSCOUNT); do
5099 start_mdt $num || return
5101 start_ost2 || error "OST1 start fail"
5102 start_mdt 1 || error "MDT0 start fail"
5104 mount_client $MOUNT || error "mount client fails"
5106 mkdir $DIR/$tdir || error "mkdir fail"
5107 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5108 error "create remote dir succeed"
5110 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5111 rm -rf $DIR/$tdir || error "delete dir fail"
5113 umount_client $MOUNT || error "umount_client failed"
5114 stop_mds || error "MDS stop fail"
5115 stop_ost || error "OST0 stop fail"
5116 stop_ost2 || error "OST1 stop fail"
5119 run_test 71e "start OST0, MDT1, OST1, MDT0"
5121 test_72() { #LU-2634
5122 local mdsdev=$(mdsdevname 1)
5123 local ostdev=$(ostdevname 1)
5124 local cmd="$E2FSCK -fnvd $mdsdev"
5127 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
5128 skip "ldiskfs only test" && return
5130 #tune MDT with "-O extents"
5132 for num in $(seq $MDSCOUNT); do
5133 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
5134 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
5135 error "add mds $num failed"
5136 do_facet mds${num} "$TUNE2FS -O extents $(mdsdevname $num)" ||
5137 error "$TUNE2FS failed on mds${num}"
5140 add ost1 $(mkfs_opts ost1 $ostdev) --reformat $ostdev ||
5141 error "add $ostdev failed"
5142 start_mgsmds || error "start mds failed"
5143 start_ost || error "start ost failed"
5144 mount_client $MOUNT || error "mount client failed"
5146 #create some short symlinks
5147 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5148 createmany -o $DIR/$tdir/$tfile-%d $fn
5149 echo "create $fn short symlinks"
5150 for i in $(seq -w 1 $fn); do
5151 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
5156 umount_client $MOUNT || error "umount client failed"
5157 stop_mds || error "stop mds failed"
5158 stop_ost || error "stop ost failed"
5161 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
5163 run_test 72 "test fast symlink with extents flag enabled"
5165 test_73() { #LU-3006
5166 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
5167 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
5168 error "1st tunefs failed"
5169 start_mgsmds || error "start mds failed"
5170 start_ost || error "start ost failed"
5171 mount_client $MOUNT || error "mount client failed"
5172 $LCTL get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
5173 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
5174 umount_client $MOUNT || error "umount client failed"
5178 run_test 73 "failnode to update from mountdata properly"
5180 test_75() { # LU-2374
5181 [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] &&
5182 skip "Need MDS version at least 2.4.1" && return
5185 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
5186 --reformat $(mdsdevname 1) $(mdsvdevname 1)"
5187 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
5188 --reformat $(ostdevname 1) $(ostvdevname 1)"
5190 #check with default parameters
5191 add mds1 $opts_mds || error "add mds1 failed for default params"
5192 add ost1 $opts_ost || error "add ost1 failed for default params"
5194 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
5195 opts_mds=$(echo $opts_mds |
5196 sed -e "s/--index=$index/--index=$index --mdt/")
5197 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
5198 opts_ost=$(echo $opts_ost |
5199 sed -e "s/--index=$index/--index=$index --ost/")
5201 add mds1 $opts_mds || error "add mds1 failed for new params"
5202 add ost1 $opts_ost || error "add ost1 failed for new params"
5205 run_test 75 "The order of --index should be irrelevant"
5208 [[ $(lustre_version_code mgs) -ge $(version_code 2.4.52) ]] ||
5209 { skip "Need MDS version at least 2.4.52" && return 0; }
5211 local MDMB_PARAM="osc.*.max_dirty_mb"
5212 echo "Change MGS params"
5213 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
5215 echo "max_dirty_mb: $MAX_DIRTY_MB"
5216 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB + MAX_DIRTY_MB))
5217 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
5218 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
5219 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5220 head -1" $NEW_MAX_DIRTY_MB
5221 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5222 echo "$MAX_DIRTY_MB"
5223 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5224 error "error while apply max_dirty_mb"
5226 echo "Check the value is stored after remount"
5229 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5230 head -1" $NEW_MAX_DIRTY_MB
5231 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5232 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5233 error "max_dirty_mb is not saved after remount"
5235 echo "Change OST params"
5236 CLIENT_PARAM="obdfilter.*.client_cache_count"
5237 local CLIENT_CACHE_COUNT
5238 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5240 echo "client_cache_count: $CLIENT_CACHE_COUNT"
5241 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
5242 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
5243 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
5244 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5245 head -1" $NEW_CLIENT_CACHE_COUNT
5246 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5248 echo "$CLIENT_CACHE_COUNT"
5249 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5250 error "error while apply client_cache_count"
5252 echo "Check the value is stored after remount"
5255 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5256 head -1" $NEW_CLIENT_CACHE_COUNT
5257 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5259 echo "$CLIENT_CACHE_COUNT"
5260 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5261 error "client_cache_count is not saved after remount"
5264 run_test 76a "set permanent params set_param -P"
5266 test_76b() { # LU-4783
5267 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.57) ]] ||
5268 { skip "Need MGS version at least 2.5.57" && return 0; }
5271 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
5272 error "start params log failed"
5275 run_test 76b "verify params log setup correctly"
5278 [[ $(lustre_version_code mgs) -ge $(version_code 2.8.54) ]] ||
5279 { skip "Need MDS version at least 2.4.52" && return 0; }
5281 local MASK_PARAM="mdd.*.changelog_mask"
5282 echo "Change changelog_mask"
5283 do_facet mgs $LCTL set_param -P $MASK_PARAM=-CLOSE ||
5284 error "Can't change changlog_mask"
5285 wait_update $(facet_host mds) "$LCTL get_param -n $MASK_PARAM |
5288 echo "Check the value is stored after mds remount"
5289 stop_mds || error "Failed to stop MDS"
5290 start_mds || error "Failed to start MDS"
5291 local CHANGELOG_MASK=$(do_facet mgs $LCTL get_param -n $MASK_PARAM)
5292 echo $CHANGELOG_MASK | grep CLOSE > /dev/null &&
5293 error "changelog_mask is not changed"
5297 run_test 76c "verify changelog_mask is applied with set_param -P"
5299 test_77() { # LU-3445
5300 local server_version=$(lustre_version_code $SINGLEMDS)
5301 [[ $server_version -ge $(version_code 2.8.55) ]] ||
5302 { skip "Need MDS version 2.8.55+ "; return; }
5304 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
5305 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
5306 skip_env "mixed loopback and real device not working" && return
5309 local fs2mdsdev=$(mdsdevname 1_2)
5310 local fs2ostdev=$(ostdevname 1_2)
5311 local fs2mdsvdev=$(mdsvdevname 1_2)
5312 local fs2ostvdev=$(ostvdevname 1_2)
5313 local fsname=test1234
5315 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
5317 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
5318 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
5319 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
5320 error "start fs2mds failed"
5322 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
5323 mgsnid="$mgsnid,$mgsnid:$mgsnid"
5325 add fs2ost $(mkfs_opts ost1 $fs2ostdev) --mgsnode=$mgsnid \
5326 --failnode=$failnid --fsname=$fsname \
5327 --reformat $fs2ostdev $fs2ostvdev ||
5328 error "add fs2ost failed"
5329 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
5331 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
5332 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
5333 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
5336 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
5339 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ||
5340 $(facet_fstype ost1) != ldiskfs ]] &&
5341 skip "only applicable to ldiskfs-based MDTs and OSTs" && return
5343 # reformat the Lustre filesystem with a smaller size
5344 local saved_MDSCOUNT=$MDSCOUNT
5345 local saved_MDSSIZE=$MDSSIZE
5346 local saved_OSTCOUNT=$OSTCOUNT
5347 local saved_OSTSIZE=$OSTSIZE
5350 MDSSIZE=$((MDSSIZE - 20000))
5351 OSTSIZE=$((OSTSIZE - 20000))
5352 reformat || error "(1) reformat Lustre filesystem failed"
5353 MDSSIZE=$saved_MDSSIZE
5354 OSTSIZE=$saved_OSTSIZE
5356 # mount the Lustre filesystem
5357 setup_noconfig || error "(2) setup Lustre filesystem failed"
5360 log "create test files"
5365 mkdir $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
5367 for i in $(seq $num_files); do
5368 file=$MOUNT/$tdir/$tfile-$i
5369 dd if=/dev/urandom of=$file count=1 bs=1M || {
5370 $LCTL get_param osc.*.cur*grant*
5371 $LFS df; $LFS df -i;
5372 # stop creating files if there is no more space
5373 if [ ! -e $file ]; then
5374 num_files=$((i - 1))
5378 $LFS getstripe -v $file
5379 local ost_idx=$(LFS getstripe -i $file)
5380 do_facet ost$((ost_idx + 1)) \
5381 $LCTL get_param obdfilter.*.*grant*
5382 error "(4) create $file failed"
5386 # unmount the Lustre filesystem
5387 cleanup || error "(5) cleanup Lustre filesystem failed"
5389 # run e2fsck on the MDT and OST devices
5390 local mds_host=$(facet_active_host $SINGLEMDS)
5391 local ost_host=$(facet_active_host ost1)
5392 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
5393 local ost_dev=$(ostdevname 1)
5395 run_e2fsck $mds_host $mds_dev "-y"
5396 run_e2fsck $ost_host $ost_dev "-y"
5398 # get the original block count of the MDT and OST filesystems
5399 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
5400 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
5402 # expand the MDT and OST filesystems to the device size
5403 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
5404 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
5406 # run e2fsck on the MDT and OST devices again
5407 run_e2fsck $mds_host $mds_dev "-y"
5408 run_e2fsck $ost_host $ost_dev "-y"
5410 # mount the Lustre filesystem
5414 log "check files after expanding the MDT and OST filesystems"
5415 for i in $(seq $num_files); do
5416 file=$MOUNT/$tdir/$tfile-$i
5417 $CHECKSTAT -t file -s 1048576 $file ||
5418 error "(6) checkstat $file failed"
5422 log "create more files after expanding the MDT and OST filesystems"
5423 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
5424 file=$MOUNT/$tdir/$tfile-$i
5425 dd if=/dev/urandom of=$file count=1 bs=1M ||
5426 error "(7) create $file failed"
5429 # unmount the Lustre filesystem
5430 cleanup || error "(8) cleanup Lustre filesystem failed"
5432 # run e2fsck on the MDT and OST devices
5433 run_e2fsck $mds_host $mds_dev "-y"
5434 run_e2fsck $ost_host $ost_dev "-y"
5436 # get the maximum block count of the MDT and OST filesystems
5437 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
5438 local ost_max_blks=$(get_block_count ost1 $ost_dev)
5440 # get the minimum block count of the MDT and OST filesystems
5441 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
5442 grep minimum | sed -e 's/^.*filesystem: //g')
5443 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
5444 grep minimum | sed -e 's/^.*filesystem: //g')
5446 # shrink the MDT and OST filesystems to a smaller size
5450 if [[ $mds_max_blks -gt $mds_min_blks &&
5451 $mds_max_blks -gt $mds_orig_blks ]]; then
5452 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
5453 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
5454 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
5455 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
5456 error "shrink $SINGLEMDS to $new_blks failed"
5460 if [[ $ost_max_blks -gt $ost_min_blks &&
5461 $ost_max_blks -gt $ost_orig_blks ]]; then
5462 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
5463 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
5464 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
5465 run_resize2fs ost1 $ost_dev $new_blks ||
5466 error "shrink ost1 to $new_blks failed"
5470 # check whether the MDT or OST filesystem was shrunk or not
5472 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
5473 reformat || error "(10) reformat Lustre filesystem failed"
5477 # run e2fsck on the MDT and OST devices again
5478 run_e2fsck $mds_host $mds_dev "-y"
5479 run_e2fsck $ost_host $ost_dev "-y"
5481 # mount the Lustre filesystem again
5485 log "check files after shrinking the MDT and OST filesystems"
5486 for i in $(seq $((num_files + 10))); do
5487 file=$MOUNT/$tdir/$tfile-$i
5488 $CHECKSTAT -t file -s 1048576 $file ||
5489 error "(11) checkstat $file failed"
5492 # unmount and reformat the Lustre filesystem
5493 cleanup || error "(12) cleanup Lustre filesystem failed"
5494 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
5496 MDSCOUNT=$saved_MDSCOUNT
5497 OSTCOUNT=$saved_OSTCOUNT
5498 reformat || error "(14) reformat Lustre filesystem failed"
5500 run_test 78 "run resize2fs on MDT and OST filesystems"
5502 test_79() { # LU-4227
5503 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.5.59) ]] ||
5504 { skip "Need MDS version at least 2.5.59"; return 0; }
5506 local mdsdev1=$(mdsdevname 1)
5507 local mdsvdev1=$(mdsvdevname 1)
5508 local mdsdev2=$(mdsdevname 2)
5509 local mdsvdev2=$(mdsvdevname 2)
5510 local ostdev1=$(ostdevname 1)
5511 local ostvdev1=$(ostvdevname 1)
5512 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
5513 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
5514 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
5517 # remove --mgs/--mgsnode from mkfs.lustre options
5518 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
5520 mgsnode_opt=$(echo $opts_mds2 |
5521 awk '{ for ( i = 1; i < NF; i++ )
5522 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5523 [ -n $mgsnode_opt ] &&
5524 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
5526 mgsnode_opt=$(echo $opts_ost1 |
5527 awk '{ for ( i = 1; i < NF; i++ )
5528 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5529 [ -n $mgsnode_opt ] &&
5530 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
5532 # -MGS, format a mdt without --mgs option
5533 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
5534 error "Must specify --mgs when formatting mdt combined with mgs"
5536 # +MGS, format a mdt/ost without --mgsnode option
5537 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
5538 > /dev/null || error "start mds1 failed"
5539 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
5540 error "Must specify --mgsnode when formatting a mdt"
5541 add ost1 $opts_ost1 $ostdev1 $ostvdev2 &&
5542 error "Must specify --mgsnode when formatting an ost"
5546 run_test 79 "format MDT/OST without mgs option (should return errors)"
5549 start_mds || error "Failed to start MDT"
5550 start_ost || error "Failed to start OST1"
5551 uuid=$(do_facet ost1 $LCTL get_param -n mgc.*.uuid)
5552 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
5553 do_facet ost1 "$LCTL set_param fail_val=10 fail_loc=0x906"
5554 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x906"
5555 do_facet mgs "$LCTL set_param -n mgs/MGS/evict_client $uuid"
5557 start_ost2 || error "Failed to start OST2"
5559 do_facet ost1 "$LCTL set_param fail_loc=0"
5562 run_test 80 "mgc import reconnect race"
5564 #Save the original values of $OSTCOUNT and $OSTINDEX$i.
5566 local new_ostcount=$1
5567 saved_ostcount=$OSTCOUNT
5568 OSTCOUNT=$new_ostcount
5572 for i in $(seq $OSTCOUNT); do
5574 eval saved_ostindex$i=${!index}
5579 # Restore the original values of $OSTCOUNT and $OSTINDEX$i.
5580 restore_ostindex() {
5585 for i in $(seq $OSTCOUNT); do
5586 index=saved_ostindex$i
5587 eval OSTINDEX$i=${!index}
5589 OSTCOUNT=$saved_ostcount
5594 # The main purpose of this test is to ensure the OST_INDEX_LIST functions as
5595 # expected. This test uses OST_INDEX_LIST to format OSTs with a randomly
5596 # assigned index and ensures we can mount such a formatted file system
5597 test_81() { # LU-4665
5598 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5599 { skip "Need MDS version at least 2.6.54" && return; }
5600 [[ $OSTCOUNT -ge 3 ]] || { skip_env "Need at least 3 OSTs" && return; }
5604 # Each time RANDOM is referenced, a random integer between 0 and 32767
5607 local saved_ostindex1=$OSTINDEX1
5608 for i in 65535 $((RANDOM + 65536)); do
5609 echo -e "\nFormat ost1 with --index=$i, should fail"
5611 if add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat \
5612 $(ostdevname 1) $(ostvdevname 1); then
5613 OSTINDEX1=$saved_ostindex1
5614 error "format ost1 with --index=$i should fail"
5617 OSTINDEX1=$saved_ostindex1
5621 # Format OSTs with random sparse indices.
5622 trap "restore_ostindex" EXIT
5623 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices"
5624 OST_INDEX_LIST=[0,$((RANDOM * 2 % 65533 + 1)),65534] formatall
5626 # Setup and check Lustre filesystem.
5627 start_mgsmds || error "start_mgsmds failed"
5628 for i in $(seq $OSTCOUNT); do
5629 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5630 error "start ost$i failed"
5633 mount_client $MOUNT || error "mount client $MOUNT failed"
5634 check_mount || error "check client $MOUNT failed"
5637 local max_easize=$($LCTL get_param -n llite.*.max_easize)
5638 [[ $max_easize -eq 128 ]] ||
5639 error "max_easize is $max_easize, should be 128 bytes"
5643 run_test 81 "sparse OST indexing"
5645 # Wait OSTs to be active on both client and MDT side.
5647 local cmd="$LCTL get_param -n lov.$FSNAME-clilov-*.target_obd |
5648 awk 'BEGIN {c = 0} /ACTIVE/{c += 1} END {printf \\\"%d\\\", c}'"
5649 wait_update $HOSTNAME "eval $cmd" $OSTCOUNT ||
5650 error "wait_update OSTs up on client failed"
5652 cmd="$LCTL get_param -n lod.$FSNAME-MDT*-*.target_obd | sort -u |
5653 awk 'BEGIN {c = 0} /ACTIVE/{c += 1} END {printf \\\"%d\\\", c}'"
5654 wait_update_facet $SINGLEMDS "eval $cmd" $OSTCOUNT ||
5655 error "wait_update OSTs up on MDT failed"
5658 # Here we exercise the stripe placement functionality on a file system that
5659 # has formatted the OST with a random index. With the file system the following
5660 # functionality is tested:
5662 # 1. Creating a new file with a specific stripe layout.
5664 # 2. Modifiy a existing empty file with a specific stripe layout.
5666 # 3. Ensure we fail to set the stripe layout of a file that already has one.
5668 # 4. If ost-index is defined we need to ensure it is the first entry in the
5669 # ost index list returned by lfs getstripe.
5671 # 5. Lastly ensure this functionality fails with directories.
5672 test_82a() { # LU-4665
5673 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5674 { skip "Need MDS version at least 2.6.54" && return; }
5675 [[ $OSTCOUNT -ge 3 ]] || { skip_env "Need at least 3 OSTs" && return; }
5681 # Format OSTs with random sparse indices.
5685 local LOV_V1_INSANE_STRIPE_COUNT=65532
5686 for i in $(seq $OSTCOUNT); do
5687 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
5688 ost_indices+=" $index"
5690 ost_indices=$(comma_list $ost_indices)
5692 trap "restore_ostindex" EXIT
5693 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5694 OST_INDEX_LIST=[$ost_indices] formatall
5696 # Setup Lustre filesystem.
5697 start_mgsmds || error "start_mgsmds failed"
5698 for i in $(seq $OSTCOUNT); do
5699 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5700 error "start ost$i failed"
5703 mount_client $MOUNT || error "mount client $MOUNT failed"
5706 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5707 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5709 # 1. If the file does not exist, new file will be created
5710 # with specified OSTs.
5711 local file=$DIR/$tdir/$tfile-1
5712 local cmd="$SETSTRIPE -o $ost_indices $file"
5714 eval $cmd || error "$cmd failed"
5715 check_stripe_count $file $OSTCOUNT
5716 check_obdidx $file $ost_indices
5717 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5718 error "write $file failed"
5720 # 2. If the file already exists and is an empty file, the file
5721 # will be attached with specified layout.
5722 file=$DIR/$tdir/$tfile-2
5723 mcreate $file || error "mcreate $file failed"
5724 cmd="$SETSTRIPE -o $ost_indices $file"
5726 eval $cmd || error "$cmd failed"
5727 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5728 error "write $file failed"
5729 check_stripe_count $file $OSTCOUNT
5730 check_obdidx $file $ost_indices
5732 # 3. If the file already has a valid layout attached, the command
5733 # should fail with EBUSY.
5735 eval $cmd && error "stripe is already set on $file, $cmd should fail"
5737 # 4. If [--stripe-index|-i <start_ost_idx>] is used, the index must
5738 # be in the OST indices list.
5739 local start_ost_idx=${ost_indices##*,}
5740 file=$DIR/$tdir/$tfile-3
5741 cmd="$SETSTRIPE -o $ost_indices -i $start_ost_idx $file"
5743 eval $cmd || error "$cmd failed"
5744 check_stripe_count $file $OSTCOUNT
5745 check_obdidx $file $ost_indices
5746 check_start_ost_idx $file $start_ost_idx
5748 file=$DIR/$tdir/$tfile-4
5750 cmd+=" -o $(exclude_items_from_list $ost_indices $start_ost_idx)"
5751 cmd+=" -i $start_ost_idx $file"
5753 eval $cmd && error "index $start_ost_idx should be in $ost_indices"
5755 # 5. Specifying OST indices for directory should fail with ENOSUPP.
5756 local dir=$DIR/$tdir/$tdir
5757 mkdir $dir || error "mkdir $dir failed"
5758 cmd="$SETSTRIPE -o $ost_indices $dir"
5760 eval $cmd && error "$cmd should fail, specifying OST indices" \
5761 "for directory is not supported"
5765 run_test 82a "specify OSTs for file (succeed) or directory (fail)"
5770 # Remove OSTs from a pool and destroy the pool.
5771 destroy_pool $ost_pool || true
5776 # Test 82b is run to ensure that if the user supplies a pool with a specific
5777 # stripe layout that it behaves proprerly. It should fail in the case that
5778 # the supplied OST index list points to OSTs not contained in the user
5780 test_82b() { # LU-4665
5781 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5782 { skip "Need MDS version at least 2.6.54" && return; }
5783 [[ $OSTCOUNT -ge 4 ]] || { skip_env "Need at least 4 OSTs" && return; }
5789 # Format OSTs with random sparse indices.
5793 local LOV_V1_INSANE_STRIPE_COUNT=65532
5794 for i in $(seq $OSTCOUNT); do
5795 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
5796 ost_indices+=" $index"
5798 ost_indices=$(comma_list $ost_indices)
5800 trap "restore_ostindex" EXIT
5801 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5802 OST_INDEX_LIST=[$ost_indices] formatall
5804 # Setup Lustre filesystem.
5805 start_mgsmds || error "start_mgsmds failed"
5806 for i in $(seq $OSTCOUNT); do
5807 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5808 error "start ost$i failed"
5811 mount_client $MOUNT || error "mount client $MOUNT failed"
5813 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5814 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5816 # Create a new pool and add OSTs into it.
5817 local ost_pool=$FSNAME.$TESTNAME
5818 create_pool $ost_pool || error "create OST pool $ost_pool failed"
5821 trap "cleanup_82b" EXIT
5823 local ost_idx_in_list=${ost_indices##*,}
5824 local ost_idx_in_pool=$(exclude_items_from_list $ost_indices \
5827 local ost_targets="$FSNAME-OST["
5828 for i in ${ost_idx_in_pool//,/ }; do
5829 ost_targets=$ost_targets$(printf "%04x," $i)
5831 ost_targets="${ost_targets%,}]"
5833 local ost_targets_uuid=$(for i in ${ost_idx_in_pool//,/ }; \
5834 do printf "$FSNAME-OST%04x_UUID\n" $i; done |
5835 sort -u | tr '\n' ' ')
5837 local cmd="$LCTL pool_add $ost_pool $ost_targets"
5838 do_facet mgs $cmd || error "$cmd failed"
5839 wait_update $HOSTNAME "$LCTL get_param -n lov.$FSNAME-*.pools.$TESTNAME|
5840 sort -u | tr '\n' ' ' " "$ost_targets_uuid" ||
5841 error "wait_update $ost_pool failed"
5842 pool_list $ost_pool || error "list OST pool $ost_pool failed"
5844 # If [--pool|-p <pool_name>] is set with [--ost-list|-o <ost_indices>],
5845 # then the OSTs must be the members of the pool.
5846 local file=$DIR/$tdir/$tfile
5847 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
5849 eval $cmd && error "OST with index $ost_idx_in_list should be" \
5850 "in OST pool $ost_pool"
5852 # Only select OST $ost_idx_in_list from $ost_pool for file.
5853 ost_idx_in_list=${ost_idx_in_pool#*,}
5854 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
5856 eval $cmd || error "$cmd failed"
5857 cmd="$GETSTRIPE $file"
5859 eval $cmd || error "$cmd failed"
5860 check_stripe_count $file 2
5861 check_obdidx $file $ost_idx_in_list
5862 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5863 error "write $file failed"
5867 run_test 82b "specify OSTs for file with --pool and --ost-list options"
5870 [[ $(lustre_version_code ost1) -ge $(version_code 2.6.91) ]] ||
5871 { skip "Need OST version at least 2.6.91" && return 0; }
5872 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
5873 skip "Only applicable to ldiskfs-based MDTs"
5883 ostmnt=$(facet_mntpt ost1)
5884 fstype=$(facet_fstype ost1)
5886 # Mount the OST as an ldiskfs filesystem.
5887 log "mount the OST $dev as a $fstype filesystem"
5888 add ost1 $(mkfs_opts ost1 $dev) $FSTYPE_OPT \
5889 --reformat $dev $dev > /dev/null ||
5890 error "format ost1 error"
5892 if ! test -b $dev; then
5893 mnt_opts=$(csa_add "$OST_MOUNT_OPTS" -o loop)
5895 echo "mnt_opts $mnt_opts"
5896 do_facet ost1 mount -t $fstype $dev \
5898 # Run llverfs on the mounted ldiskfs filesystem.
5899 # It is needed to get ENOSPACE.
5900 log "run llverfs in partial mode on the OST $fstype $ostmnt"
5901 do_rpc_nodes $(facet_host ost1) run_llverfs $ostmnt -vpl \
5902 "no" || error "run_llverfs error on $fstype"
5905 log "unmount the OST $dev"
5908 # Delete file IO_scrub. Later osd_scrub_setup will try to
5909 # create "IO_scrub" but will get ENOSPACE.
5911 echo "start ost1 service on `facet_active_host ost1`"
5912 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
5915 err=$(do_facet ost1 dmesg | grep "VFS: Busy inodes after unmount of")
5916 echo "string err $err"
5917 [ -z "$err" ] || error $err
5920 run_test 83 "ENOSPACE on OST doesn't cause message VFS: \
5921 Busy inodes after unmount ..."
5924 local facet=$SINGLEMDS
5925 local num=$(echo $facet | tr -d "mds")
5926 local dev=$(mdsdevname $num)
5927 local time_min=$(recovery_time_min)
5928 local recovery_duration
5929 local completed_clients
5930 local correct_clients
5933 echo "start mds service on $(facet_active_host $facet)"
5935 "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
5936 error "start MDS failed"
5938 start_ost || error "start OST0000 failed"
5939 start_ost2 || error "start OST0001 failed"
5941 echo "recovery_time=$time_min, timeout=$TIMEOUT, wrap_up=$wrap_up"
5943 mount_client $MOUNT1 || error "mount $MOUNT1 failed"
5944 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
5945 # make sure new superblock labels are sync'd before disabling writes
5949 replay_barrier $SINGLEMDS
5950 createmany -o $DIR1/$tfile-%d 1000
5952 # We need to catch the end of recovery window to extend it.
5953 # Skip 5 requests and add delay to request handling.
5954 #define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP
5955 do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5"
5957 facet_failover --fsck $SINGLEMDS || error "failover: $?"
5960 echo "recovery status"
5961 do_facet $SINGLEMDS \
5962 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status"
5964 recovery_duration=$(do_facet $SINGLEMDS \
5965 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
5966 awk '/recovery_duration/ { print $2 }')
5967 (( $recovery_duration > $time_min + $wrap_up )) &&
5968 error "recovery_duration > recovery_time_hard + wrap up"
5969 completed_clients=$(do_facet $SINGLEMDS \
5970 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
5971 awk '/completed_clients/ { print $2 }')
5973 correct_clients="$MDSCOUNT/$((MDSCOUNT+1))"
5974 [ "$completed_clients" = "${correct_clients}" ] ||
5975 error "$completed_clients != $correct_clients"
5977 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
5978 umount_client $MOUNT1
5979 umount_client $MOUNT2
5985 run_test 84 "check recovery_hard_time"
5988 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.55) ]] ||
5989 { skip "Need OST version at least 2.7.55" && return 0; }
5990 ##define OBD_FAIL_OSD_OST_EA_FID_SET 0x197
5991 do_facet ost1 "lctl set_param fail_loc=0x197"
5995 run_test 85 "osd_ost init: fail ea_fid_set"
6000 # ost1 has already registered to the MGS before the reformat.
6001 # So after reformatting it with option "-G", it could not be
6002 # mounted to the MGS. Cleanup the system for subsequent tests.
6007 local server_version=$(lustre_version_code $SINGLEMDS)
6008 [ "$(facet_fstype ost1)" = "zfs" ] &&
6009 skip "LU-6442: no such mkfs params for ZFS OSTs" && return
6010 [[ $server_version -ge $(version_code 2.7.56) ]] ||
6011 { skip "Need server version newer than 2.7.55"; return 0; }
6013 local OST_OPTS="$(mkfs_opts ost1 $(ostdevname 1)) \
6014 --reformat $(ostdevname 1) $(ostvdevname 1)"
6017 local OLDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6018 awk '/Flex block group size: / { print $NF; exit; }')
6021 if [[ ${!opts} != *mkfsoptions* ]]; then
6022 eval opts=\"${!opts} \
6023 --mkfsoptions='\\\"-O flex_bg -G $NEWSIZE\\\"'\"
6025 val=${!opts//--mkfsoptions=\\\"/ \
6026 --mkfsoptions=\\\"-O flex_bg -G $NEWSIZE }
6030 echo "params: $opts"
6032 trap cleanup_86 EXIT ERR
6035 add ost1 $opts || error "add ost1 failed with new params"
6037 local FOUNDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6038 awk '/Flex block group size: / { print $NF; exit; }')
6040 [[ $FOUNDSIZE == $NEWSIZE ]] ||
6041 error "Flex block group size: $FOUNDSIZE, expected: $NEWSIZE"
6045 run_test 86 "Replacing mkfs.lustre -G option"
6047 test_87() { #LU-6544
6048 [[ $(lustre_version_code $SINGLEMDS1) -ge $(version_code 2.7.56) ]] ||
6049 { skip "Need MDS version at least 2.7.56" && return; }
6050 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ]] &&
6051 { skip "Only applicable to ldiskfs-based MDTs" && return; }
6052 [[ $OSTCOUNT -gt 69 ]] &&
6053 { skip "Ignore wide striping situation" && return; }
6055 local mdsdev=$(mdsdevname 1)
6056 local mdsvdev=$(mdsvdevname 1)
6057 local file=$DIR/$tfile
6058 local mntpt=$(facet_mntpt $SINGLEMDS)
6059 local used_xattr_blk=0
6060 local inode_size=${1:-512}
6062 local xtest="trusted.test"
6067 #Please see LU-6544 for MDT inode size calculation
6068 if [ $OSTCOUNT -gt 26 ]; then
6070 elif [ $OSTCOUNT -gt 5 ]; then
6073 left_size=$(expr $inode_size - \
6076 32 - $OSTCOUNT \* 24 - 16 - 3 - \
6078 24 - 18 - $(expr length $tfile) - 16 - 4)
6079 if [ $left_size -le 0 ]; then
6080 echo "No space($left_size) is expected in inode."
6081 echo "Try 1-byte xattr instead to verify this."
6084 echo "Estimate: at most $left_size-byte space left in inode."
6090 add mds1 $(mkfs_opts mds1 ${mdsdev}) --stripe-count-hint=$OSTCOUNT \
6091 --reformat $mdsdev $mdsvdev || error "add mds1 failed"
6092 start_mdt 1 > /dev/null || error "start mdt1 failed"
6093 for i in $(seq $OSTCOUNT); do
6094 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS > /dev/null ||
6095 error "start ost$i failed"
6097 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
6098 check_mount || error "check client $MOUNT failed"
6101 $SETSTRIPE -c -1 $file || error "$SETSTRIPE -c -1 $file failed"
6102 $GETSTRIPE $file || error "$GETSTRIPE $file failed"
6103 i=$($GETSTRIPE -c $file)
6104 if [ $i -ne $OSTCOUNT ]; then
6105 left_size=$(expr $left_size + $(expr $OSTCOUNT - $i) \* 24)
6106 echo -n "Since only $i out $OSTCOUNT OSTs are used, "
6107 echo -n "the expected left space is changed to "
6108 echo "$left_size bytes at most."
6110 value=$(generate_string $left_size)
6111 setfattr -n $xtest -v $value $file
6112 orig=$(get_xattr_value $xtest $file)
6113 [[ "$orig" != "$value" ]] && error "$xtest changed"
6115 #Verify if inode has some expected space left
6116 umount $MOUNT > /dev/null || error "umount $MOUNT failed"
6117 stop_mdt 1 > /dev/null || error "stop mdt1 failed"
6118 mount_ldiskfs $SINGLEMDS || error "mount -t ldiskfs $SINGLEMDS failed"
6120 do_facet $SINGLEMDS ls -sal $mntpt/ROOT/$tfile
6121 used_xattr_blk=$(do_facet $SINGLEMDS ls -s $mntpt/ROOT/$tfile |
6123 [[ $used_xattr_blk -eq 0 ]] &&
6124 error "Please check MDS inode size calculation: \
6125 more than $left_size-byte space left in inode."
6126 echo "Verified: at most $left_size-byte space left in inode."
6130 run_test 87 "check if MDT inode can hold EAs with N stripes properly"
6133 [ "$(facet_fstype mds1)" == "zfs" ] &&
6134 skip "LU-6662: no implementation for ZFS" && return
6138 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
6139 --reformat $(mdsdevname 1) || error "add mds1 failed"
6141 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
6142 grep -e \".*opts:.*errors=remount-ro.*\"" ||
6143 error "default mount options is missing"
6145 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
6146 --mountfsoptions="user_xattr,errors=panic" \
6147 --reformat $(mdsdevname 1) || error "add mds1 failed"
6149 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
6150 grep -e \".*opts:.*errors=panic.*\"" ||
6151 error "user can't override default mount options"
6153 run_test 88 "check the default mount options can be overridden"
6156 # $2 (optional) value of max_mod_rpcs_in_flight to set
6157 check_max_mod_rpcs_in_flight() {
6165 idx=$(printf "%04x" $($LFS getdirstripe -i $dir))
6166 facet="mds$((0x$idx + 1))"
6168 if [ -z "$mmr" ]; then
6169 # get value of max_mod_rcps_in_flight
6170 mmr=$($LCTL get_param -n \
6171 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) ||
6172 error "Unable to get max_mod_rpcs_in_flight"
6173 echo "max_mod_rcps_in_flight is $mmr"
6175 # set value of max_mod_rpcs_in_flight
6177 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight=$mmr ||
6178 error "Unable to set max_mod_rpcs_in_flight to $mmr"
6179 echo "max_mod_rpcs_in_flight set to $mmr"
6182 # create mmr+1 files
6183 echo "creating $((mmr + 1)) files ..."
6185 for i in $(seq $((mmr + 1))); do
6191 # consumes mmr-1 modify RPC slots
6192 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6193 # drop requests on MDT so that RPC slots are consumed
6194 # during all the request resend interval
6195 do_facet $facet "$LCTL set_param fail_loc=0x159"
6196 echo "launch $((mmr - 1)) chmod in parallel ..."
6197 for i in $(seq $((mmr - 1))); do
6198 chmod 0600 $dir/file-$i &
6202 # send one additional modify RPC
6203 do_facet $facet "$LCTL set_param fail_loc=0"
6204 echo "launch 1 additional chmod in parallel ..."
6205 chmod 0600 $dir/file-$mmr &
6208 # check this additional modify RPC get a modify RPC slot
6209 # and succeed its operation
6210 checkstat -vp 0600 $dir/file-$mmr ||
6211 error "Unable to send $mmr modify RPCs in parallel"
6216 # consumes mmr modify RPC slots
6217 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6218 # drop requests on MDT so that RPC slots are consumed
6219 # during all the request resend interval
6220 do_facet $facet "$LCTL set_param fail_loc=0x159"
6221 echo "launch $mmr chmod in parallel ..."
6222 for i in $(seq $mmr); do
6223 chmod 0666 $dir/file-$i &
6227 # send one additional modify RPC
6228 do_facet $facet "$LCTL set_param fail_loc=0"
6229 echo "launch 1 additional chmod in parallel ..."
6230 chmod 0666 $dir/file-$((mmr + 1)) &
6233 # check this additional modify RPC blocked getting a modify RPC slot
6234 checkstat -vp 0644 $dir/file-$((mmr + 1)) ||
6235 error "Unexpectedly send $(($mmr + 1)) modify RPCs in parallel"
6241 if ! combined_mgs_mds ; then
6246 [[ $($LCTL get_param mdc.*.import |
6247 grep "connect_flags:.*multi_mod_rpc") ]] ||
6248 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6250 # check default value
6251 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6252 check_max_mod_rpcs_in_flight $DIR/$tdir
6256 run_test 90a "check max_mod_rpcs_in_flight is enforced"
6266 [[ $($LCTL get_param mdc.*.import |
6267 grep "connect_flags:.*multi_mod_rpc") ]] ||
6268 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6271 # update max_mod_rpcs_in_flight
6272 $LFS mkdir -c1 $DIR/${tdir}1 || error "mkdir $DIR/${tdir}1 failed"
6273 check_max_mod_rpcs_in_flight $DIR/${tdir}1 1
6276 # check client is able to send multiple modify RPCs in paralell
6277 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
6278 grep -c "multi_mod_rpcs")
6279 if [ "$tmp" -ne $MDSCOUNT ]; then
6280 echo "Client not able to send multiple modify RPCs in parallel"
6285 # update max_mod_rpcs_in_flight
6286 $LFS mkdir -c1 $DIR/${tdir}2 || error "mkdir $DIR/${tdir}2 failed"
6287 check_max_mod_rpcs_in_flight $DIR/${tdir}2 5
6290 $LFS mkdir -c1 $DIR/${tdir}3 || error "mkdir $DIR/${tdir}3 failed"
6291 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/${tdir}3))
6292 facet="mds$((0x$idx + 1))"
6294 # save MDT max_mod_rpcs_per_client
6295 mmrpc=$(do_facet $facet \
6296 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
6298 # update max_mod_rpcs_in_flight
6299 umount_client $MOUNT
6301 "echo 16 > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6303 $LCTL set_param mdc.$FSNAME-MDT$idx-mdc-*.max_rpcs_in_flight=17
6304 check_max_mod_rpcs_in_flight $DIR/${tdir}3 16
6306 # restore MDT max_mod_rpcs_per_client initial value
6308 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6310 rm -rf $DIR/${tdir}?
6313 run_test 90b "check max_mod_rpcs_in_flight is enforced after update"
6322 [[ $($LCTL get_param mdc.*.import |
6323 grep "connect_flags:.*multi_mod_rpc") ]] ||
6324 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6326 # check client is able to send multiple modify RPCs in paralell
6327 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
6328 grep -c "multi_mod_rpcs")
6329 if [ "$tmp" -ne $MDSCOUNT ]; then
6330 skip "Client not able to send multiple modify RPCs in parallel"
6335 # get max_rpcs_in_flight value
6336 mrif=$($LCTL get_param -n mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
6337 echo "max_rpcs_in_flight is $mrif"
6339 # get MDT max_mod_rpcs_per_client
6340 mmrpc=$(do_facet mds1 \
6341 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
6342 echo "max_mod_rpcs_per_client is $mmrpc"
6345 # attempt to set max_mod_rpcs_in_flight to max_rpcs_in_flight value
6346 # prerequisite: set max_mod_rpcs_per_client to max_rpcs_in_flight value
6347 umount_client $MOUNT
6349 "echo $mrif > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6353 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif &&
6354 error "set max_mod_rpcs_in_flight to $mrif should fail"
6356 umount_client $MOUNT
6358 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6362 # attempt to set max_mod_rpcs_in_flight to max_mod_rpcs_per_client+1
6363 # prerequisite: set max_rpcs_in_flight to max_mod_rpcs_per_client+2
6365 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$((mmrpc + 2))
6368 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$((mmrpc + 1)) &&
6369 error "set max_mod_rpcs_in_flight to $((mmrpc + 1)) should fail"
6373 run_test 90c "check max_mod_rpcs_in_flight update limits"
6384 [[ $($LCTL get_param mdc.*.import |
6385 grep "connect_flags:.*multi_mod_rpc") ]] ||
6386 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6388 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6389 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
6390 facet="mds$((0x$idx + 1))"
6392 # check client version supports multislots
6393 tmp=$($LCTL get_param -N \
6394 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
6395 if [ -z "$tmp" ]; then
6396 skip "Client does not support multiple modify RPCs in flight"
6401 # get current value of max_mod_rcps_in_flight
6402 mmr=$($LCTL get_param -n \
6403 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
6404 echo "max_mod_rcps_in_flight is $mmr"
6407 echo "creating $mmr files ..."
6409 for i in $(seq $mmr); do
6410 touch $DIR/$tdir/file-$i
6413 # prepare for close RPC
6414 multiop_bg_pause $DIR/$tdir/file-close O_c
6417 # consumes mmr modify RPC slots
6418 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6419 # drop requests on MDT so that RPC slots are consumed
6420 # during all the request resend interval
6421 do_facet $facet "$LCTL set_param fail_loc=0x159"
6422 echo "launch $mmr chmod in parallel ..."
6423 for i in $(seq $mmr); do
6424 chmod 0600 $DIR/$tdir/file-$i &
6427 # send one additional close RPC
6428 do_facet $facet "$LCTL set_param fail_loc=0"
6429 echo "launch 1 additional close in parallel ..."
6431 cancel_lru_locks mdc
6434 # check this additional close RPC get a modify RPC slot
6435 # and multiop process completed
6436 [ -d /proc/$pid ] &&
6437 error "Unable to send the additional close RPC in parallel"
6442 run_test 90d "check one close RPC is allowed above max_mod_rpcs_in_flight"
6444 check_uuid_on_ost() {
6446 do_facet ost1 "$LCTL get_param obdfilter.${FSNAME}*.exports.'$nid'.uuid"
6449 check_uuid_on_mdt() {
6451 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
6459 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.63) ]] ||
6460 { skip "Need OST version at least 2.7.63" && return 0; }
6461 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.63) ]] ||
6462 { skip "Need MDT version at least 2.7.63" && return 0; }
6464 start_mds || error "MDS start failed"
6465 start_ost || error "unable to start OST"
6466 mount_client $MOUNT || error "client start failed"
6467 check_mount || error "check_mount failed"
6470 nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g")
6474 uuid=$(get_client_uuid $MOUNT)
6476 echo "list nids on mdt:"
6477 do_facet $SINGLEMDS "$LCTL list_param mdt.${FSNAME}*.exports.*"
6478 echo "uuid from $nid:"
6479 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
6481 found=$(check_uuid_on_mdt $nid | grep $uuid)
6482 [ -z "$found" ] && error "can't find $uuid $nid on MDT"
6483 found=$(check_uuid_on_ost $nid | grep $uuid)
6484 [ -z "$found" ] && error "can't find $uuid $nid on OST"
6486 # umount the client so it won't reconnect
6487 manual_umount_client --force || error "failed to umount $?"
6488 # shouldn't disappear on MDS after forced umount
6489 found=$(check_uuid_on_mdt $nid | grep $uuid)
6490 [ -z "$found" ] && error "can't find $uuid $nid"
6493 do_facet $SINGLEMDS \
6494 "$LCTL set_param -n mdt.${mds1_svc}.evict_client nid:$nid"
6496 found=$(check_uuid_on_mdt $nid | grep $uuid)
6497 [ -n "$found" ] && error "found $uuid $nid on MDT"
6498 found=$(check_uuid_on_ost $nid | grep $uuid)
6499 [ -n "$found" ] && error "found $uuid $nid on OST"
6501 # check it didn't reconnect (being umounted)
6502 sleep $((TIMEOUT+1))
6503 found=$(check_uuid_on_mdt $nid | grep $uuid)
6504 [ -n "$found" ] && error "found $uuid $nid on MDT"
6505 found=$(check_uuid_on_ost $nid | grep $uuid)
6506 [ -n "$found" ] && error "found $uuid $nid on OST"
6510 run_test 91 "evict-by-nid support"
6512 generate_ldev_conf() {
6513 # generate an ldev.conf file
6514 local ldevconfpath=$1
6516 local fsldevformat=""
6519 fstype=$(facet_fstype mgs)
6520 if [ "$fstype" == "zfs" ]; then
6521 fsldevformat="$fstype:"
6526 printf "%s\t-\t%s-MGS0000\t%s%s\n" \
6530 $(mgsdevname) > $ldevconfpath
6532 local mdsfo_host=$mdsfailover_HOST;
6533 if [ -z "$mdsfo_host" ]; then
6537 for num in $(seq $MDSCOUNT); do
6538 fstype=$(facet_fstype mds$num)
6539 if [ "$fstype" == "zfs" ]; then
6540 fsldevformat="$fstype:"
6545 printf "%s\t%s\t%s-MDT%04d\t%s%s\n" \
6551 $(mdsdevname $num) >> $ldevconfpath
6554 local ostfo_host=$ostfailover_HOST;
6555 if [ -z "$ostfo_host" ]; then
6559 for num in $(seq $OSTCOUNT); do
6560 fstype=$(facet_fstype ost$num)
6561 if [ "$fstype" == "zfs" ]; then
6562 fsldevformat="$fstype:"
6567 printf "%s\t%s\t%s-OST%04d\t%s%s\n" \
6573 $(ostdevname $num) >> $ldevconfpath
6576 echo "----- $ldevconfpath -----"
6578 echo "--- END $ldevconfpath ---"
6583 # generate a nids file (mapping between hostname to nid)
6584 # looks like we only have the MGS nid available to us
6585 # so just echo that to a file
6587 echo -e "${mgs_HOST}\t${MGSNID}" > $nidspath
6589 echo "----- $nidspath -----"
6591 echo "--- END $nidspath ---"
6594 compare_ldev_output() {
6598 sort $expected_output -o $expected_output
6599 sort $ldev_output -o $ldev_output
6601 echo "-- START OF LDEV OUTPUT --"
6603 echo "--- END OF LDEV OUTPUT ---"
6605 echo "-- START OF EXPECTED OUTPUT --"
6606 cat $expected_output
6607 echo "--- END OF EXPECTED OUTPUT ---"
6609 diff $expected_output $ldev_output
6614 if [ -z "$LDEV" ]; then
6615 error "ldev is missing!"
6618 local LDEVCONFPATH=$TMP/ldev.conf
6619 local NIDSPATH=$TMP/nids
6621 echo "Host is $(hostname)"
6623 generate_ldev_conf $LDEVCONFPATH
6624 generate_nids $NIDSPATH
6626 # echo the mgs nid and compare it to environment variable MGSNID
6627 # also, ldev.conf and nids is a server side thing, use the OSS
6630 output=$($LDEV -c $LDEVCONFPATH -H $ost_HOST -n $NIDSPATH echo %m)
6632 echo "-- START OF LDEV OUTPUT --"
6634 echo "--- END OF LDEV OUTPUT ---"
6636 # ldev failed, error
6637 if [ $? -ne 0 ]; then
6638 rm $LDEVCONFPATH $NIDSPATH
6639 error "ldev failed to execute!"
6642 # need to process multiple lines because of combined MGS and MDS
6643 echo -e $output | awk '{ print $2 }' | while read -r line ; do
6644 if [ "$line" != "$MGSNID" ]; then
6645 rm $LDEVCONFPATH $NIDSPATH
6646 error "ldev failed mgs nid '$line', expected '$MGSNID'"
6650 rm $LDEVCONFPATH $NIDSPATH
6652 run_test 92 "ldev returns MGS NID correctly in command substitution"
6655 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs" && return
6658 #start mgs or mgs/mdt0
6659 if ! combined_mgs_mds ; then
6666 start_ost || error "OST0 start fail"
6668 #define OBD_FAIL_MGS_WRITE_TARGET_DELAY 0x90e
6669 do_facet mgs "$LCTL set_param fail_val = 10 fail_loc=0x8000090e"
6670 for num in $(seq 2 $MDSCOUNT); do
6674 mount_client $MOUNT || error "mount client fails"
6675 wait_osc_import_state mds ost FULL
6676 wait_osc_import_state client ost FULL
6677 check_mount || error "check_mount failed"
6679 cleanup || error "cleanup failed with $?"
6681 run_test 93 "register mulitple MDT at the same time"
6684 if [ -z "$LDEV" ]; then
6685 error "ldev is missing!"
6688 local LDEVCONFPATH=$TMP/ldev.conf
6689 local NIDSPATH=$TMP/nids
6691 generate_ldev_conf $LDEVCONFPATH
6692 generate_nids $NIDSPATH
6694 local LDEV_OUTPUT=$TMP/ldev-output.txt
6695 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME > $LDEV_OUTPUT
6697 # ldev failed, error
6698 if [ $? -ne 0 ]; then
6699 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
6700 error "ldev failed to execute!"
6704 local EXPECTED_OUTPUT=$TMP/ldev-expected.txt
6706 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
6708 for num in $(seq $MDSCOUNT); do
6709 printf "%s-MDT%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
6712 for num in $(seq $OSTCOUNT); do
6713 printf "%s-OST%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
6716 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6718 if [ $? -ne 0 ]; then
6719 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6720 error "ldev failed to produce the correct hostlist!"
6723 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6725 run_test 94 "ldev outputs correct labels for file system name query"
6728 if [ -z "$LDEV" ]; then
6729 error "ldev is missing!"
6732 local LDEVCONFPATH=$TMP/ldev.conf
6733 local NIDSPATH=$TMP/nids
6735 generate_ldev_conf $LDEVCONFPATH
6736 generate_nids $NIDSPATH
6740 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME &>/dev/null
6741 if [ $? -ne 0 ]; then
6742 rm $LDEVCONFPATH $NIDSPATH
6743 error "ldev label filtering w/ -F failed!"
6747 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -l &>/dev/null
6748 if [ $? -ne 0 ]; then
6749 rm $LDEVCONFPATH $NIDSPATH
6750 error "ldev label filtering w/ -l failed!"
6754 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f &>/dev/null
6755 if [ $? -ne 0 ]; then
6756 rm $LDEVCONFPATH $NIDSPATH
6757 error "ldev label filtering w/ -f failed!"
6761 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a &>/dev/null
6762 if [ $? -ne 0 ]; then
6763 rm $LDEVCONFPATH $NIDSPATH
6764 error "ldev label filtering w/ -a failed!"
6769 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -F $FSNAME &>/dev/null
6770 if [ $? -eq 0 ]; then
6771 rm $LDEVCONFPATH $NIDSPATH
6772 error "ldev label filtering w/ -a and -F incorrectly succeeded"
6776 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -f &>/dev/null
6777 if [ $? -eq 0 ]; then
6778 rm $LDEVCONFPATH $NIDSPATH
6779 error "ldev label filtering w/ -a and -f incorrectly succeeded"
6783 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -l &>/dev/null
6784 if [ $? -eq 0 ]; then
6785 rm $LDEVCONFPATH $NIDSPATH
6786 error "ldev label filtering w/ -a and -l incorrectly succeeded"
6790 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f -l &>/dev/null
6791 if [ $? -eq 0 ]; then
6792 rm $LDEVCONFPATH $NIDSPATH
6793 error "ldev label filtering w/ -f and -l incorrectly succeeded"
6797 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -l &>/dev/null
6798 if [ $? -eq 0 ]; then
6799 rm $LDEVCONFPATH $NIDSPATH
6800 error "ldev label filtering w/ -F and -l incorrectly succeeded"
6803 # file sys & foreign
6804 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -f &>/dev/null
6805 if [ $? -eq 0 ]; then
6806 rm $LDEVCONFPATH $NIDSPATH
6807 error "ldev label filtering w/ -F and -f incorrectly succeeded"
6810 rm $LDEVCONFPATH $NIDSPATH
6812 run_test 95 "ldev should only allow one label filter"
6815 if [ -z "$LDEV" ]; then
6816 error "ldev is missing!"
6819 local LDEVCONFPATH=$TMP/ldev.conf
6820 local NIDSPATH=$TMP/nids
6822 generate_ldev_conf $LDEVCONFPATH
6823 generate_nids $NIDSPATH
6825 local LDEV_OUTPUT=$TMP/ldev-output.txt
6826 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -H $mgs_HOST \
6828 awk '{print $2}' > $LDEV_OUTPUT
6830 # ldev failed, error
6831 if [ $? -ne 0 ]; then
6832 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
6833 error "ldev failed to execute!"
6837 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
6839 echo "$mgs_HOST-$(facet_fstype mgs)" > $EXPECTED_OUTPUT
6841 if [ "$mgs_HOST" == "$mds_HOST" ]; then
6842 for num in $(seq $MDSCOUNT); do
6843 echo "$mds_HOST-$(facet_fstype mds$num)" \
6848 if [ "$mgs_HOST" == "$ost_HOST" ]; then
6849 for num in $(seq $OSTCOUNT); do
6850 echo "$ost_HOST-$(facet_fstype ost$num)" \
6855 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6857 if [ $? -ne 0 ]; then
6858 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6859 error "ldev failed to produce the correct output!"
6862 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6864 run_test 96 "ldev returns hostname and backend fs correctly in command sub"
6867 if [ -z "$LDEV" ]; then
6868 error "ldev is missing!"
6871 local LDEVCONFPATH=$TMP/ldev.conf
6872 local NIDSPATH=$TMP/nids
6874 generate_ldev_conf $LDEVCONFPATH
6875 generate_nids $NIDSPATH
6877 local LDEV_OUTPUT=$TMP/ldev-output.txt
6878 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
6880 echo -e "\nMDT role"
6881 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mdt > $LDEV_OUTPUT
6883 if [ $? -ne 0 ]; then
6884 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
6885 error "ldev failed to execute for mdt role!"
6888 for num in $(seq $MDSCOUNT); do
6889 printf "%s-MDT%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
6892 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6894 if [ $? -ne 0 ]; then
6895 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6896 error "ldev failed to produce the correct output for mdt role!"
6899 echo -e "\nOST role"
6900 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R ost > $LDEV_OUTPUT
6902 if [ $? -ne 0 ]; then
6903 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
6904 error "ldev failed to execute for ost role!"
6908 for num in $(seq $OSTCOUNT); do
6909 printf "%s-OST%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
6912 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6914 if [ $? -ne 0 ]; then
6915 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6916 error "ldev failed to produce the correct output for ost role!"
6919 echo -e "\nMGS role"
6920 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mgs > $LDEV_OUTPUT
6922 if [ $? -ne 0 ]; then
6923 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
6924 error "ldev failed to execute for mgs role!"
6927 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
6929 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6931 if [ $? -ne 0 ]; then
6932 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6933 error "ldev failed to produce the correct output for mgs role!"
6936 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6938 run_test 97 "ldev returns correct ouput when querying based on role"
6943 local temp=$MDS_MOUNT_OPTS
6946 check_mount || error "mount failed"
6947 mountopt="user_xattr"
6948 for ((x = 1; x <= 400; x++)); do
6949 mountopt="$mountopt,user_xattr"
6951 remount_client $mountopt $MOUNT 2>&1 | grep "too long" ||
6952 error "Buffer overflow check failed"
6953 cleanup || error "cleanup failed"
6955 run_test 98 "Buffer-overflow check while parsing mount_opts"
6959 [[ $(facet_fstype ost1) != ldiskfs ]] &&
6960 { skip "Only applicable to ldiskfs-based OSTs" && return; }
6961 [[ $(lustre_version_code ost1) -ge $(version_code 2.8.57) ]] ||
6962 { skip "Need OST version at least 2.8.57" && return 0; }
6964 local ost_opts="$(mkfs_opts ost1 $(ostdevname 1)) \
6965 --reformat $(ostdevname 1) $(ostvdevname 1)"
6966 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" &&
6967 skip "meta_bg already set" && return
6970 if [[ ${!opts} != *mkfsoptions* ]]; then
6971 eval opts=\"${!opts} \
6972 --mkfsoptions='\\\"-O ^resize_inode,meta_bg\\\"'\"
6974 local val=${!opts//--mkfsoptions=\\\"/ \
6975 --mkfsoptions=\\\"-O ^resize_inode,meta_bg }
6979 echo "params: $opts"
6981 add ost1 $opts || error "add ost1 failed with new params"
6983 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" ||
6984 error "meta_bg is not set"
6988 run_test 99 "Adding meta_bg option"
6992 start_mds || error "MDS start failed"
6993 start_ost || error "unable to start OST"
6994 mount_client $MOUNT || error "client start failed"
6995 check_mount || error "check_mount failed"
7004 do_facet mgs 'lshowmount -v' | awk 'BEGIN {NR == 0; rc=1} /MGS:/ {rc=0}
7005 END {exit rc}' || error "lshowmount have no output MGS"
7007 do_facet mds1 'lshowmount -v' | awk 'BEGIN {NR == 2; rc=1} /-MDT0000:/
7008 {rc=0} END {exit rc}' || error "lshowmount have no output MDT0"
7010 do_facet ost1 'lshowmount -v' | awk 'BEGIN {NR == 4; rc=1} /-OST0000:/
7011 {rc=0} END {exit rc}' || error "lshowmount have no output OST0"
7013 cleanup || error "cleanup failed with $?"
7015 run_test 100 "check lshowmount lists MGS, MDT, OST and 0@lo"
7018 local createmany_oid
7019 local dev=$FSNAME-OST0000-osc-MDT0000
7022 createmany -o $DIR1/$tfile-%d 50000 &
7024 # MDT->OST reconnection causes MDT<->OST last_id synchornisation
7025 # via osp_precreate_cleanup_orphans.
7026 for ((i = 0; i < 100; i++)); do
7027 for ((k = 0; k < 10; k++)); do
7028 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;" \
7029 "$LCTL --device $dev activate"
7032 ls -asl $MOUNT | grep '???' &&
7033 (kill -9 $createmany_oid &>/dev/null; \
7034 error "File hasn't object on OST")
7036 kill -s 0 $createmany_oid || break
7038 wait $createmany_oid
7041 run_test 101 "Race MDT->OST reconnection with create"
7043 if ! combined_mgs_mds ; then
7049 # restore the values of MDSSIZE and OSTSIZE
7050 MDSSIZE=$STORED_MDSSIZE
7051 OSTSIZE=$STORED_OSTSIZE