7 # bug number for skipped test:
8 ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT"
9 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
11 # bug number for skipped test:
12 # a tool to create lustre filesystem images
13 ALWAYS_EXCEPT="32newtarball $ALWAYS_EXCEPT"
15 # bug number for skipped tests: LU-9795 (all below)
16 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 0 31 32a 32d 35a"
17 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 53a 53b 54b 76a 76b"
18 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 76c 76d 78 103"
22 PATH=$PWD/$SRCDIR:$SRCDIR:$SRCDIR/../utils:$PATH
24 PTLDEBUG=${PTLDEBUG:--1}
26 LUSTRE=${LUSTRE:-$(dirname $0)/..}
27 RLUSTRE=${RLUSTRE:-$LUSTRE}
28 export MULTIOP=${MULTIOP:-multiop}
30 . $LUSTRE/tests/test-framework.sh
32 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
34 # use small MDS + OST size to speed formatting time
35 # do not use too small MDSSIZE/OSTSIZE, which affect the default journal size
36 # STORED_MDSSIZE is used in test_18
37 STORED_MDSSIZE=$MDSSIZE
38 STORED_OSTSIZE=$OSTSIZE
40 [ $(facet_fstype $SINGLEMDS) = "zfs" ] && MDSSIZE=400000
42 [ $(facet_fstype ost1) = "zfs" ] && OSTSIZE=400000
52 if ! combined_mgs_mds; then
53 # bug number for skipped test: LU-9860 LU-9860 LU-9860
54 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 43b 53b 54b"
55 # bug number for skipped test: LU-9875 LU-9879 LU-9879 LU-9879 LU-9879
56 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 70e 80 84 87 100"
57 # bug number for skipped test: LU-8110 LU-9879 LU-9879 LU-9879
58 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 102 104 105 107"
61 # pass "-E lazy_itable_init" to mke2fs to speed up the formatting time
62 if [[ "$LDISKFS_MKFS_OPTS" != *lazy_itable_init* ]]; then
63 LDISKFS_MKFS_OPTS=$(csa_add "$LDISKFS_MKFS_OPTS" -E lazy_itable_init)
66 [ $(facet_fstype $SINGLEMDS) = "zfs" ] &&
67 # bug number for skipped test:
68 ALWAYS_EXCEPT="$ALWAYS_EXCEPT"
69 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
74 require_dsh_mds || exit 0
75 require_dsh_ost || exit 0
78 [ "$SLOW" = "no" ] && EXCEPT_SLOW="45 69 106 111"
83 # The MGS must be started before the OSTs for a new fs, so start
84 # and stop to generate the startup logs.
87 wait_osc_import_state mds ost FULL
92 reformat_and_config() {
94 if ! combined_mgs_mds ; then
100 writeconf_or_reformat() {
101 # There are at most 2 OSTs for write_conf test
102 # who knows if/where $TUNEFS is installed?
103 # Better reformat if it fails...
104 writeconf_all $MDSCOUNT 2 ||
105 { echo "tunefs failed, reformatting instead" &&
106 reformat_and_config && return 0; }
115 echo "start mgs service on $(facet_active_host mgs)"
116 start mgs $(mgsdevname) $MGS_MOUNT_OPTS $@
122 local dev=$(mdsdevname $num)
125 echo "start mds service on `facet_active_host $facet`"
126 start $facet ${dev} $MDS_MOUNT_OPTS $@ || return 94
132 local dev=$(mdsdevname $num)
135 echo "stop mds service on `facet_active_host $facet`"
136 # These tests all use non-failover stop
137 stop $facet -f || return 97
143 for num in $(seq $MDSCOUNT); do
144 start_mdt $num $@ || return 94
149 if ! combined_mgs_mds ; then
157 for num in $(seq $MDSCOUNT); do
158 stop_mdt $num || return 97
163 echo "stop mgs service on `facet_active_host mgs`"
164 # These tests all use non-failover stop
165 stop mgs -f || return 97
169 echo "start ost1 service on `facet_active_host ost1`"
170 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS $@ || return 95
174 echo "stop ost1 service on `facet_active_host ost1`"
175 # These tests all use non-failover stop
176 stop ost1 -f || return 98
180 echo "start ost2 service on `facet_active_host ost2`"
181 start ost2 $(ostdevname 2) $OST_MOUNT_OPTS $@ || return 92
185 echo "stop ost2 service on `facet_active_host ost2`"
186 # These tests all use non-failover stop
187 stop ost2 -f || return 93
192 echo "mount $FSNAME on ${MOUNTPATH}....."
193 zconf_mount $(hostname) $MOUNTPATH || return 96
197 local mountopt="remount,$1"
199 echo "remount '$1' lustre on ${MOUNTPATH}....."
200 zconf_mount $(hostname) $MOUNTPATH "$mountopt" || return 96
206 echo "umount lustre on $mountpath....."
207 zconf_umount $HOSTNAME $mountpath $@ || return 97
210 manual_umount_client(){
213 echo "manual umount lustre on ${MOUNT}...."
214 do_facet client "umount ${FORCE} $MOUNT"
220 start_mds || error "MDT start failed"
221 start_ost || error "Unable to start OST1"
222 mount_client $MOUNT || error "client start failed"
223 client_up || error "client_up failed"
232 unload_modules_conf () {
233 if combined_mgs_mds || ! local_mode; then
234 unload_modules || return 1
239 stop_ost || return 202
240 stop_mds || return 201
241 unload_modules_conf || return 203
246 [ "x$1" != "x" ] && force='-f'
247 umount_client $MOUNT $force|| return 200
248 cleanup_nocli || return $?
253 echo "umount $MOUNT2 ..."
254 umount $MOUNT2 || true
255 echo "stopping fs2mds ..."
256 stop fs2mds -f || true
257 echo "stopping fs2ost ..."
258 stop fs2ost -f || true
262 do_facet client "cp /etc/passwd $DIR/a" || return 71
263 do_facet client "rm $DIR/a" || return 72
264 # make sure lustre is actually mounted (touch will block,
265 # but grep won't, so do it after)
266 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
267 echo "setup single mount lustre success"
271 do_facet client "touch $DIR/a" || return 71
272 do_facet client "rm $DIR/a" || return 72
273 do_facet client "touch $DIR2/a" || return 73
274 do_facet client "rm $DIR2/a" || return 74
275 echo "setup double mount lustre success"
279 cat /dev/urandom | tr -dc 'a-zA-Z0-9' | fold -w $1 | head -n 1
284 if [ "$ONLY" == "setup" ]; then
289 if [ "$ONLY" == "cleanup" ]; then
296 #create single point mountpoint
302 check_mount || error "check_mount failed"
303 cleanup || error "cleanup failed with $?"
305 run_test 0 "single mount setup"
308 start_mds || error "MDS start failed"
309 start_ost || error "unable to start OST"
310 echo "start ost second time..."
311 start_ost && error "2nd OST start should fail"
312 mount_client $MOUNT || error "client start failed"
313 check_mount || error "check_mount failed"
314 cleanup || error "cleanup failed with $?"
316 run_test 1 "start up ost twice (should return errors)"
319 start_mds || error "MDT start failed"
320 echo "start mds second time.."
321 start_mds && error "2nd MDT start should fail"
322 start_ost || error "OST start failed"
323 mount_client $MOUNT || error "mount_client failed to start client"
324 check_mount || error "check_mount failed"
325 cleanup || error "cleanup failed with $?"
327 run_test 2 "start up mds twice (should return err)"
331 #mount.lustre returns an error if already in mtab
332 mount_client $MOUNT && error "2nd client mount should fail"
333 check_mount || error "check_mount failed"
334 cleanup || error "cleanup failed with $?"
336 run_test 3 "mount client twice (should return err)"
340 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
341 stop_ost || error "Unable to stop OST1"
342 umount_client $MOUNT -f || error “unmount $MOUNT failed”
345 # ok for ost to fail shutdown
346 if [ 202 -ne $eno ] && [ 0 -ne $eno ]; then
347 error "cleanup failed with $?"
350 run_test 4 "force cleanup ost, then cleanup"
352 test_5a() { # was test_5
354 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
355 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
357 stop_mds || error "Unable to stop MDS"
359 # cleanup may return an error from the failed
360 # disconnects; for now I'll consider this successful
361 # if all the modules have unloaded.
365 echo "killing umount"
366 kill -TERM $UMOUNT_PID
367 echo "waiting for umount to finish"
369 if grep " $MOUNT " /proc/mounts; then
370 echo "test 5: /proc/mounts after failed umount"
374 echo "killing umount"
375 kill -TERM $UMOUNT_PID
376 echo "waiting for umount to finish"
378 grep " $MOUNT " /proc/mounts &&
379 error "/proc/mounts after second umount"
382 # manual_mount_client may fail due to umount succeeding above
384 # stop_mds is a no-op here, and should not fail
385 cleanup_nocli || error "cleanup_nocli failed with $?"
386 # df may have lingering entry
388 # mtab may have lingering entry
392 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
394 grep -q $MOUNT" " /etc/mtab || break
395 echo "Waiting /etc/mtab updated ... "
396 WAIT=$(( WAIT + sleep))
398 [ "$WAIT" -eq "$MAX_WAIT" ] &&
399 error "/etc/mtab is not updated in $WAIT secs"
400 echo "/etc/mtab updated in $WAIT secs"
402 run_test 5a "force cleanup mds, then cleanup"
410 grep " $MOUNT " /etc/mtab &&
411 error false "unexpected entry in mtab before mount" && return 10
413 start_ost || error "OST start failed"
414 if ! combined_mgs_mds ; then
415 trap cleanup_5b EXIT ERR
416 start_mds || error "MDS start failed"
420 mount_client $MOUNT && error "mount_client $MOUNT should fail"
421 grep " $MOUNT " /etc/mtab &&
422 error "$MOUNT entry in mtab after failed mount"
424 # stop_mds is a no-op here, and should not fail
425 cleanup_nocli || error "cleanup_nocli failed with $?"
426 if ! combined_mgs_mds ; then
430 run_test 5b "Try to start a client with no MGS (should return errs)"
433 grep " $MOUNT " /etc/mtab &&
434 error false "unexpected entry in mtab before mount" && return 10
436 start_mds || error "MDS start failed"
437 start_ost || error "OST start failed"
438 local oldfs="${FSNAME}"
439 FSNAME="wrong.${FSNAME}"
440 mount_client $MOUNT || :
442 grep " $MOUNT " /etc/mtab &&
443 error "$MOUNT entry in mtab after failed mount"
445 cleanup_nocli || error "cleanup_nocli failed with $?"
447 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
450 grep " $MOUNT " /etc/mtab &&
451 error "unexpected entry in mtab before mount"
453 start_ost || error "OST start failed"
454 start_mds || error "MDS start failed"
455 stop_ost -f || error "Unable to stop OST1"
456 mount_client $MOUNT || error "mount_client $MOUNT failed"
457 umount_client $MOUNT -f || error "umount_client $MOUNT failed"
458 cleanup_nocli || error "cleanup_nocli failed with $?"
459 ! grep " $MOUNT " /etc/mtab ||
460 error "$MOUNT entry in mtab after unmount"
462 run_test 5d "mount with ost down"
465 grep " $MOUNT " /etc/mtab &&
466 error false "unexpected entry in mtab before mount" && return 10
468 start_mds || error "MDS start failed"
469 start_ost || error "OST start failed"
471 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
472 do_facet client "$LCTL set_param fail_loc=0x80000506"
473 mount_client $MOUNT || echo "mount failed (not fatal)"
474 cleanup || error "cleanup failed with $?"
475 grep " $MOUNT " /etc/mtab &&
476 error "$MOUNT entry in mtab after unmount"
479 run_test 5e "delayed connect, don't crash (bug 10268)"
482 if combined_mgs_mds ; then
483 skip "needs separate mgs and mds"
487 grep " $MOUNT " /etc/mtab &&
488 error false "unexpected entry in mtab before mount" && return 10
491 start_ost || error "OST start failed"
492 mount_client $MOUNT &
494 echo client_mount pid is $pid
498 if ! ps -f -p $pid >/dev/null; then
501 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
502 error "mount returns $rc, expected to hang"
504 cleanup || error "cleanup failed with $?"
509 start_mds || error "start MDS failed"
511 # mount should succeed after start mds
513 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
514 cleanup || error "final call to cleanup failed with rc $?"
516 run_test 5f "mds down, cleanup after failed mount (bug 2712)"
520 [ $(lustre_version_code client) -lt $(version_code 2.9.53) ] &&
521 { skip "automount of debugfs missing before 2.9.53" && return 0; }
522 umount /sys/kernel/debug
523 $LCTL get_param -n devices | egrep -v "error" && \
524 error "lctl can't access debugfs data"
525 grep " debugfs " /etc/mtab || error "debugfs failed to remount"
527 run_test 5g "handle missing debugfs"
532 mount_client $MOUNT || error "mount_client $MOUNT failed"
533 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
534 cleanup || error "cleanup failed with rc $?"
536 run_test 6 "manual umount, then mount again"
541 cleanup_nocli || error "cleanup_nocli failed with $?"
543 run_test 7 "manual umount, then cleanup"
547 mount_client $MOUNT2 || error "mount_client $MOUNT2 failed"
548 check_mount2 || error "check_mount2 failed"
549 umount_client $MOUNT2 || error "umount_client $MOUNT2 failed"
550 cleanup || error "cleanup failed with rc $?"
552 run_test 8 "double mount setup"
555 start_ost || error "OST start failed"
557 do_facet ost1 $LCTL set_param debug=\'inode trace\' ||
558 error "do_facet ost1 set_param inode trace failed."
559 do_facet ost1 $LCTL set_param subsystem_debug=\'mds ost\' ||
560 error "do_facet ost1 set_param debug mds ost failed."
562 CHECK_PTLDEBUG="`do_facet ost1 $LCTL get_param -n debug`"
563 if [ "$CHECK_PTLDEBUG" ] && { \
564 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
565 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
566 echo "lnet.debug success"
568 error "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
570 CHECK_SUBSYS="`do_facet ost1 $LCTL get_param -n subsystem_debug`"
571 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
572 echo "lnet.subsystem_debug success"
574 error "lnet.subsystem_debug: want 'mds ost' got '$CHECK_SUBSYS'"
576 stop_ost || error "Unable to stop OST1"
578 run_test 9 "test ptldebug and subsystem for mkfs"
581 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
582 # But with new MDS stack we don't care about the mode of local objects
583 # anymore, so this test is removed. See bug 22944 for more details.
587 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
588 skip "ldiskfs only test"
593 check_mount || error "check_mount failed"
594 cleanup || error "cleanup failed with rc $?"
596 echo "Remove mds config log"
597 if ! combined_mgs_mds ; then
601 do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' \
602 $(mgsdevname) || return \$?" ||
603 error "do_facet mgs failed with $?"
605 if ! combined_mgs_mds ; then
609 start_ost || error "OST start failed"
610 start_mds && error "MDS start succeeded, but should fail"
613 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
616 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
617 skip "ldiskfs only test"
621 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
626 # check if current MDSSIZE is large enough
627 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE &&
628 log "use MDSSIZE=$MDSSIZE"
630 # check if the global config has a large enough MDSSIZE
631 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] &&
632 [ $STORED_MDSSIZE -ge $MIN ] &&
633 OK=1 && myMDSSIZE=$STORED_MDSSIZE &&
634 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
636 # check if the block device is large enough
637 is_blkdev $SINGLEMDS $MDSDEV $MIN
638 local large_enough=$?
639 if [ -n "$OK" ]; then
640 [ $large_enough -ne 0 ] && OK=""
642 [ $large_enough -eq 0 ] && OK=1 && myMDSSIZE=$MIN &&
643 log "use device $MDSDEV with MIN=$MIN"
646 # check if a loopback device has enough space for fs metadata (5%)
648 if [ -z "$OK" ]; then
649 local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! \
650 -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
651 awk '($1 != "Filesystem") { print $4 }')
652 ! [ -z "$SPACE" ] && [ $SPACE -gt $((MIN / 20)) ] &&
653 OK=1 && myMDSSIZE=$MIN &&
654 log "use file $MDSDEV with MIN=$MIN"
657 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS" && return
659 echo "mount mds with large journal..."
661 local OLD_MDSSIZE=$MDSSIZE
665 echo "mount lustre system..."
667 check_mount || error "check_mount failed"
669 echo "check journal size..."
670 local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
671 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
672 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
674 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
677 cleanup || error "cleanup failed with rc $?"
682 run_test 18 "check mkfs creates large journals"
685 start_mds || error "MDS start failed"
686 stop_mds || error "Unable to stop MDS"
688 run_test 19a "start/stop MDS without OSTs"
691 start_ost || error "Unable to start OST1"
692 stop_ost -f || error "Unable to stop OST1"
694 run_test 19b "start/stop OSTs without MDS"
697 # first format the ost/mdt
698 start_mds || error "MDS start failed"
699 start_ost || error "Unable to start OST1"
700 mount_client $MOUNT || error "mount_client $MOUNT failed"
701 check_mount || error "check_mount failed"
702 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed."
703 remount_client ro $MOUNT || error "remount_client with ro failed"
704 touch $DIR/$tfile && error "$DIR/$tfile created incorrectly"
705 [ -e $DIR/$tfile ] && error "$DIR/$tfile exists incorrectly"
706 remount_client rw $MOUNT || error "remount_client with rw failed"
707 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
708 MCNT=$(grep -c $MOUNT' ' /etc/mtab)
709 [ "$MCNT" -ne 1 ] && error "$MOUNT in /etc/mtab $MCNT times"
711 stop_mds || error "Unable to stop MDS"
712 stop_ost || error "Unable to stop OST1"
714 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
717 start_mds || error "MDS start failed"
718 start_ost || error "unable to start OST1"
719 wait_osc_import_state mds ost FULL
720 stop_ost || error "unable to stop OST1"
721 stop_mds || error "unable to stop MDS"
723 run_test 21a "start mds before ost, stop ost first"
726 start_ost || error "unable to start OST1"
727 start_mds || error "MDS start failed"
728 wait_osc_import_state mds ost FULL
729 stop_mds || error "unable to stop MDS"
730 stop_ost || error "unable to stop OST1"
732 run_test 21b "start ost before mds, stop mds first"
735 start_ost || error "Unable to start OST1"
736 start_mds || error "MDS start failed"
737 start_ost2 || error "Unable to start OST2"
738 wait_osc_import_state mds ost2 FULL
739 stop_ost || error "Unable to stop OST1"
740 stop_ost2 || error "Unable to stop OST2"
741 stop_mds || error "Unable to stop MDS"
742 #writeconf to remove all ost2 traces for subsequent tests
743 writeconf_or_reformat
745 run_test 21c "start mds between two osts, stop mds last"
748 if combined_mgs_mds ; then
749 skip "need separate mgs device" && return 0
755 start_mgs || error "unable to start MGS"
756 start_ost || error "unable to start OST1"
757 start_ost2 || error "unable to start OST2"
758 start_mds || error "MDS start failed"
759 wait_osc_import_state mds ost2 FULL
761 stop_ost || error "Unable to stop OST1"
762 stop_ost2 || error "Unable to stop OST2"
763 stop_mds || error "Unable to stop MDS"
765 #writeconf to remove all ost2 traces for subsequent tests
766 writeconf_or_reformat
767 start_mgs || error "unable to start MGS"
769 run_test 21d "start mgs then ost and then mds"
772 MGSNID="$saved_mgsnid"
774 echo "stopping fs2mgs ..."
775 stop $fs2mgs -f || true
778 test_21e() { # LU-5863
779 if [[ -z "$fs3ost_DEV" || -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
780 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
781 skip_env "mixed loopback and real device not working" && return
784 local fs2mdsdev=$(mdsdevname 1_2)
785 local fs2ostdev=$(ostdevname 1_2)
786 local fs3ostdev=$(ostdevname 2_2)
788 local fs2mdsvdev=$(mdsvdevname 1_2)
789 local fs2ostvdev=$(ostvdevname 1_2)
790 local fs3ostvdev=$(ostvdevname 2_2)
792 # temporarily use fs3ost as fs2mgs
794 local fs2mgsdev=$fs3ostdev
795 local fs2mgsvdev=$fs3ostvdev
797 local fsname=test1234
799 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
800 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
801 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS && trap cleanup_21e EXIT INT ||
802 error "start fs2mgs failed"
804 local saved_mgsnid="$MGSNID"
805 MGSNID=$(do_facet $fs2mgs $LCTL list_nids | xargs | tr ' ' ,)
807 add fs2mds $(mkfs_opts mds1 $fs2mdsdev $fsname) \
808 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
809 add fs2ost $(mkfs_opts ost1 $fs2ostdev $fsname) \
810 --reformat $fs2ostdev $fs2ostvdev || error "add fs2ost failed"
812 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
813 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS || error "start fs2mds failed"
815 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
816 $MOUNT_CMD $MGSNID:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
817 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
821 run_test 21e "separate MGS and MDS"
824 start_mds || error "MDS start failed"
826 echo "Client mount with ost in logs, but none running"
827 start_ost || error "unable to start OST1"
828 # wait until mds connected to ost and open client connection
829 wait_osc_import_state mds ost FULL
830 stop_ost || error "unable to stop OST1"
831 mount_client $MOUNT || error "mount_client $MOUNT failed"
832 # check_mount will block trying to contact ost
833 mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed"
834 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed"
835 umount_client $MOUNT -f
838 echo "Client mount with a running ost"
839 start_ost || error "unable to start OST1"
841 # if gss enabled, wait full time to let connection from
842 # mds to ost be established, due to the mismatch between
843 # initial connect timeout and gss context negotiation timeout.
844 # This perhaps could be remove after AT landed.
845 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
846 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
848 mount_client $MOUNT || error "mount_client $MOUNT failed"
849 wait_osc_import_state mds ost FULL
850 wait_osc_import_ready client ost
851 check_mount || error "check_mount failed"
854 cleanup || error "cleanup failed with rc $?"
856 run_test 22 "start a client before osts (should return errs)"
858 test_23a() { # was test_23
861 stop $SINGLEMDS || error "failed to stop $SINGLEMDS"
862 # force down client so that recovering mds waits for reconnect
863 local running=$(grep -c $MOUNT /proc/mounts) || true
864 if [ $running -ne 0 ]; then
865 echo "Stopping client $MOUNT (opts: -f)"
869 # enter recovery on failed mds
870 local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/})
871 start $SINGLEMDS $MDT_DEV $MDS_MOUNT_OPTS || error "MDS start failed"
872 # try to start a new client
873 mount_client $MOUNT &
875 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
876 MOUNT_LUSTRE_PID=$(ps -ef | grep mount.lustre |
877 grep -v grep | awk '{print $2}')
878 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
880 ps --ppid $MOUNT_LUSTRE_PID
881 echo "waiting for mount to finish"
883 # "ctrl-c" sends SIGINT but it usually (in script) does not work on child process
884 # SIGTERM works but it does not spread to offspring processses
885 kill -s TERM $MOUNT_PID
886 kill -s TERM $MOUNT_LUSTRE_PID
887 # we can not wait $MOUNT_PID because it is not a child of this shell
893 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
895 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
896 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
899 [ -z "$PID1" -a -z "$PID2" ] && break
900 echo "waiting for mount to finish ... "
901 WAIT=$(( WAIT + sleep))
903 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
904 error "MOUNT_PID $MOUNT_PID and "\
905 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
908 cleanup || error "cleanup failed with rc $?"
910 run_test 23a "interrupt client during recovery mount delay"
912 test_23b() { # was test_23
913 start_mds || error "MDS start failed"
914 start_ost || error "Unable to start OST1"
915 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
916 $LCTL set_param fail_loc=0x80000313
918 cleanup || error "cleanup failed with rc $?"
920 run_test 23b "Simulate -EINTR during mount"
923 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
925 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
926 is_blkdev $SINGLEMDS $MDSDEV &&
927 skip_env "mixed loopback and real device not working" && return
930 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
932 local fs2mdsdev=$(mdsdevname 1_2)
933 local fs2ostdev=$(ostdevname 1_2)
934 local fs2mdsvdev=$(mdsvdevname 1_2)
935 local fs2ostvdev=$(ostvdevname 1_2)
938 # LU-9733 test fsname started with numbers as well
939 local FSNAME2=969362ae
941 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --nomgs --mgsnode=$MGSNID \
942 --fsname=${FSNAME2} --reformat $fs2mdsdev $fs2mdsvdev || exit 10
944 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --fsname=${FSNAME2} \
945 --reformat $fs2ostdev $fs2ostvdev || exit 10
948 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
949 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
950 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
951 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
953 # LU-9733 test fsname started with numbers
954 cl_user=$(do_facet $SINGLEMDS lctl --device $FSNAME2-MDT0000 \
955 changelog_register -n) ||
956 error "register changelog failed"
958 do_facet $SINGLEMDS lctl --device $FSNAME2-MDT0000 \
959 changelog_deregister $cl_user ||
960 error "deregister changelog failed"
962 check_mount || error "check_mount failed"
963 # files written on 1 should not show up on 2
964 cp /etc/passwd $DIR/$tfile
966 [ -e $MOUNT2/$tfile ] && error "File bleed"
969 cp /etc/passwd $MOUNT2/$tfile ||
970 error "cp /etc/passwd $MOUNT2/$tfile failed"
971 rm $MOUNT2/$tfile || error "remove $MOUNT2/$tfile failed"
972 # 2 is actually mounted
973 grep $MOUNT2' ' /proc/mounts > /dev/null || error "$MOUNT2 not mounted"
975 facet_failover fs2mds
976 facet_failover fs2ost
979 # the MDS must remain up until last MDT
981 MDS=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
982 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
983 [ -z "$MDS" ] && error "No MDT"
985 cleanup_nocli || error "cleanup_nocli failed with rc $?"
987 run_test 24a "Multiple MDTs on a single node"
990 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
992 if [ -z "$fs2mds_DEV" ]; then
993 local dev=${SINGLEMDS}_dev
995 is_blkdev $SINGLEMDS $MDSDEV &&
996 skip_env "mixed loopback and real device not working" && return
999 local fs2mdsdev=$(mdsdevname 1_2)
1000 local fs2mdsvdev=$(mdsvdevname 1_2)
1002 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --mgs --fsname=${FSNAME}2 \
1003 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
1005 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS &&
1006 error "start MDS should fail"
1008 cleanup || error "cleanup failed with rc $?"
1010 run_test 24b "Multiple MGSs on a single node (should return err)"
1014 check_mount || error "check_mount failed"
1015 local MODULES=$($LCTL modules | awk '{ print $2 }')
1016 rmmod $MODULES 2>/dev/null || true
1017 cleanup || error "cleanup failed with $?"
1019 run_test 25 "Verify modules are referenced"
1023 # we need modules before mount for sysctl, so make sure...
1024 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
1025 #define OBD_FAIL_MDS_FS_SETUP 0x135
1026 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000135"
1027 start_mds && error "MDS started but should not have started"
1028 $LCTL get_param -n devices
1029 DEVS=$($LCTL get_param -n devices | egrep -v MG | wc -l)
1030 [ $DEVS -gt 0 ] && error "number of devices is $DEVS, should be zero"
1031 # start mds to drop writeconf setting
1032 start_mds || error "Unable to start MDS"
1033 stop_mds || error "Unable to stop MDS"
1034 unload_modules_conf || error "unload_modules_conf failed with $?"
1036 run_test 26 "MDT startup failure cleans LOV (should return errs)"
1039 start_ost || error "Unable to start OST1"
1040 start_mds || error "Unable to start MDS"
1041 echo "Requeue thread should have started: "
1042 ps -e | grep ll_cfg_requeue
1043 set_persistent_param_and_check ost1 \
1044 "obdfilter.$FSNAME-OST0000.client_cache_seconds" \
1045 "$FSNAME-OST0000.ost.client_cache_seconds"
1046 cleanup_nocli || error "cleanup_nocli failed with rc $?"
1048 run_test 27a "Reacquire MGS lock if OST started first"
1053 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
1054 awk '($3 ~ "mdt" && $4 ~ "MDT0000") { print $4 }')
1056 facet_failover $SINGLEMDS
1057 set_persistent_param_and_check $SINGLEMDS \
1058 "mdt.$device.identity_acquire_expire" \
1059 "$device.mdt.identity_acquire_expire"
1060 set_persistent_param_and_check client \
1061 "mdc.$device-mdc-*.max_rpcs_in_flight" \
1062 "$device.mdc.max_rpcs_in_flight"
1064 cleanup || error "cleanup failed with $?"
1066 run_test 27b "Reacquire MGS lock after failover"
1068 test_28A() { # was test_28
1070 TEST="llite.$FSNAME-*.max_read_ahead_whole_mb"
1071 PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
1072 ORIG=$($LCTL get_param -n $TEST)
1073 FINAL=$(($ORIG + 1))
1074 set_persistent_param_and_check client "$TEST" "$PARAM" $FINAL
1075 FINAL=$(($FINAL + 1))
1076 set_persistent_param_and_check client "$TEST" "$PARAM" $FINAL
1077 umount_client $MOUNT || error "umount_client $MOUNT failed"
1078 mount_client $MOUNT || error "mount_client $MOUNT failed"
1079 RESULT=$($LCTL get_param -n $TEST)
1080 if [ $RESULT -ne $FINAL ]; then
1081 error "New config not seen: wanted $FINAL got $RESULT"
1083 echo "New config success: got $RESULT"
1085 set_persistent_param_and_check client "$TEST" "$PARAM" $ORIG
1086 cleanup || error "cleanup failed with rc $?"
1088 run_test 28A "permanent parameter setting"
1090 test_28a() { # LU-4221
1091 [[ $(lustre_version_code ost1) -ge $(version_code 2.5.52) ]] ||
1092 { skip "Need OST version at least 2.5.52" && return 0; }
1093 [ "$(facet_fstype ost1)" = "zfs" ] &&
1094 skip "LU-4221: no such proc params for ZFS OSTs" && return
1101 local device="$FSNAME-OST0000"
1105 # In this test we will set three kinds of proc parameters with
1106 # lctl set_param -P or lctl conf_param:
1107 # 1. non-symlink ones in the OFD
1108 # 2. non-symlink ones in the OSD
1111 # prepare a non-symlink parameter in the OFD
1112 name="client_cache_seconds"
1113 param="$device.ost.$name"
1114 cmd="obdfilter.$device.$name"
1116 # permanently setting the parameter in the OFD
1117 old=$(do_facet ost1 $LCTL get_param -n $cmd)
1119 set_persistent_param_and_check ost1 "$cmd" "$param" $new
1120 set_persistent_param_and_check ost1 "$cmd" "$param" $old
1123 # prepare a non-symlink parameter in the OSD
1125 param="$device.osd.$name"
1126 cmd="osd-*.$device.$name"
1128 # conf_param the parameter in the OSD
1129 old=$(do_facet ost1 $LCTL get_param -n $cmd)
1130 new=$(((old + 1) % 2))
1131 set_persistent_param_and_check ost1 "$cmd" "$param" $new
1132 set_persistent_param_and_check ost1 "$cmd" "$param" $old
1134 cleanup || error "cleanup failed with $?"
1136 run_test 28a "set symlink parameters permanently with lctl"
1139 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >= 2 OSTs" && return
1140 setup > /dev/null 2>&1
1141 start_ost2 || error "Unable to start OST2"
1144 local PARAM="$FSNAME-OST0001.osc.active"
1145 # With lctl set_param -P the value $PROC_ACT will be sent to
1146 # all nodes. The [!M] filter out the ability to set active
1147 # on the MDS servers which is tested with wait_osp_* below.
1148 # For ost_server_uuid that only exist on client so filtering
1150 local PROC_ACT="osc.$FSNAME-OST0001-osc-*.active"
1151 local PROC_UUID="osc.$FSNAME-OST0001-osc-[!M]*.ost_server_uuid"
1153 ACTV=$($LCTL get_param -n $PROC_ACT)
1155 set_persistent_param_and_check client $PROC_ACT $PARAM $DEAC
1156 # also check ost_server_uuid status
1157 RESULT=$($LCTL get_param -n $PROC_UUID | grep DEACTIV)
1158 if [ -z "$RESULT" ]; then
1159 error "Client not deactivated: $($LCTL get_param \
1162 echo "Live client success: got $RESULT"
1166 wait_osp_active ost ${FSNAME}-OST0001 1 0
1168 # test new client starts deactivated
1169 umount_client $MOUNT || error "umount_client $MOUNT failed"
1170 mount_client $MOUNT || error "mount_client $MOUNT failed"
1172 # the 2nd and 3rd field of ost_server_uuid do not update at the same
1173 # time when using lctl set_param -P
1174 wait_update_facet client \
1175 "$LCTL get_param -n $PROC_UUID | awk '{print \\\$3 }'" \
1177 error "New client start active: $($LCTL get_param -n $PROC_UUID)"
1179 echo "New client success: got '$($LCTL get_param -n $PROC_UUID)'"
1181 # make sure it reactivates
1182 set_persistent_param_and_check client $PROC_ACT $PARAM $ACTV
1184 umount_client $MOUNT
1185 stop_ost2 || error "Unable to stop OST2"
1186 cleanup_nocli || error "cleanup_nocli failed with $?"
1187 #writeconf to remove all ost2 traces for subsequent tests
1188 writeconf_or_reformat
1190 run_test 29 "permanently remove an OST"
1195 echo Big config llog
1196 TEST="llite.$FSNAME-*.max_read_ahead_whole_mb"
1197 ORIG=$($LCTL get_param -n $TEST)
1198 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1199 for i in ${LIST[@]}; do
1200 set_persistent_param_and_check client "$TEST" \
1201 "$FSNAME.llite.max_read_ahead_whole_mb" $i
1203 # make sure client restart still works
1204 umount_client $MOUNT
1205 mount_client $MOUNT || error "mount_client $MOUNT failed"
1206 [ "$($LCTL get_param -n $TEST)" -ne "$i" ] &&
1207 error "Param didn't stick across restart $($TEST) != $i"
1210 echo Erase parameter setting
1211 if [[ $PERM_CMD == *"set_param -P"* ]]; then
1212 do_facet mgs "$PERM_CMD -d $TEST" ||
1213 error "Erase param $TEST failed"
1215 do_facet mgs "$PERM_CMD \
1216 -d $FSNAME.llite.max_read_ahead_whole_mb" ||
1217 error "Erase param $FSNAME.llite.max_read_ahead_whole_mb failed"
1219 umount_client $MOUNT
1220 mount_client $MOUNT || error "mount_client $MOUNT failed"
1221 FINAL=$($LCTL get_param -n $TEST)
1222 echo "deleted (default) value=$FINAL, orig=$ORIG"
1223 # assumes this parameter started at the default value
1224 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
1226 cleanup || error "cleanup failed with rc $?"
1228 run_test 30a "Big config llog and permanent parameter deletion"
1233 local orignids=$($LCTL get_param -n \
1234 osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1236 local orignidcount=$(echo "$orignids" | wc -w)
1238 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1239 # numerical part of it. Hopefully that's not already a failover address
1241 local OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | \
1243 local ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1244 local NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1245 local NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1246 echo "Using fake nid $NEW"
1248 local TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1249 grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1250 if [[ $PERM_CMD == *"set_param -P"* ]]; then
1251 PARAM="osc.$FSNAME-OST0000-osc-[^M]*.import"
1252 echo "Setting $PARAM from $TEST to $NEW"
1253 do_facet mgs "$PERM_CMD $PARAM='connection=$NEW'" ||
1254 error "$PERM_CMD $PARAM failed"
1256 PARAM="$FSNAME-OST0000.failover.node"
1257 echo "Setting $PARAM from $TEST to $NEW"
1258 do_facet mgs "$PERM_CMD $PARAM='$NEW'" ||
1259 error "$PARAM $PARAM failed"
1261 wait_update_facet client "$TEST" "$NEW" ||
1262 error "check $PARAM failed!"
1264 local NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1266 local NIDCOUNT=$(echo "$NIDS" | wc -w)
1267 echo "should have $((orignidcount + 1)) entries \
1268 in failover nids string, have $NIDCOUNT"
1269 [ $NIDCOUNT -eq $((orignidcount + 1)) ] ||
1270 error "Failover nid not added"
1272 if [[ $PERM_CMD == *"set_param -P"* ]]; then
1273 do_facet mgs "$PERM_CMD -d osc.$FSNAME-OST0000-osc-*.import"
1275 do_facet mgs "$PERM_CMD -d $FSNAME-OST0000.failover.node" ||
1276 error "$PERM_CMD delete failed"
1278 umount_client $MOUNT
1279 mount_client $MOUNT || error "mount_client $MOUNT failed"
1281 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1283 NIDCOUNT=$(echo "$NIDS" | wc -w)
1284 echo "only $orignidcount final entries should remain \
1285 in failover nids string, have $NIDCOUNT"
1286 [ $NIDCOUNT -eq $orignidcount ] || error "Failover nids not removed"
1288 cleanup || error "cleanup failed with rc $?"
1290 run_test 30b "Remove failover nids"
1292 test_31() { # bug 10734
1293 # ipaddr must not exist
1294 $MOUNT_CMD 4.3.2.1@tcp:/lustre $MOUNT || true
1295 cleanup || error "cleanup failed with rc $?"
1297 run_test 31 "Connect to non-existent node (shouldn't crash)"
1301 T32_BLIMIT=40960 # Kbytes
1305 # This is not really a test but a tool to create new disk
1306 # image tarballs for the upgrade tests.
1308 # Disk image tarballs should be created on single-node
1309 # clusters by running this test with default configurations
1310 # plus a few mandatory environment settings that are verified
1311 # at the beginning of the test.
1313 test_32newtarball() {
1317 local tmp=$TMP/t32_image_create
1318 local server_version=$(lustre_version_code $SINGLEMDS)
1323 if [ $FSNAME != t32fs -o \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o \
1324 $OSTCOUNT -ne 1 -o -z "$OSTDEV1" ]; then
1325 error "Needs FSNAME=t32fs MDSCOUNT=2 " \
1326 "MDSDEV1=<nonexistent_file>" \
1327 "MDSDEV2=<nonexistent_file>" \
1328 "(or MDSDEV, in the case of b1_8)" \
1329 "OSTCOUNT=1 OSTDEV1=<nonexistent_file>"
1333 echo "Found stale $tmp"
1337 mkdir $tmp/src || return 1
1338 tar cf - -C $src . | tar xf - -C $tmp/src
1339 dd if=/dev/zero of=$tmp/src/t32_qf_old bs=1M \
1340 count=$(($T32_BLIMIT / 1024 / 4))
1341 chown $T32_QID.$T32_QID $tmp/src/t32_qf_old
1343 # format ost with comma-separated NIDs to verify LU-4460
1344 local failnid="$(h2nettype 1.2.3.4),$(h2nettype 4.3.2.1)"
1345 MGSNID="$MGSNID,$MGSNID" OSTOPT="--failnode=$failnid" formatall
1349 [[ $server_version -ge $(version_code 2.3.50) ]] ||
1350 $LFS quotacheck -ug /mnt/$FSNAME
1351 $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
1354 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1356 if [[ $MDSCOUNT -ge 2 ]]; then
1357 remote_dir=/mnt/$FSNAME/remote_dir
1358 $LFS mkdir -i 1 $remote_dir
1359 tar cf - -C $tmp/src . | tar xf - -C $remote_dir
1361 if [[ $server_version -ge $(version_code 2.7.0) ]]; then
1362 striped_dir=/mnt/$FSNAME/striped_dir_old
1363 $LFS mkdir -i 1 -c 2 $striped_dir
1364 tar cf - -C $tmp/src . | tar xf - -C $striped_dir
1370 mkdir $tmp/img || return 1
1374 pushd_dir=/mnt/$FSNAME
1375 if [[ $MDSCOUNT -ge 2 ]]; then
1376 pushd_dir=$remote_dir
1377 if [[ $server_version -ge $(version_code 2.7.0) ]]; then
1379 ls -Rni --time-style=+%s >$tmp/img/list2
1385 ls -Rni --time-style=+%s >$tmp/img/list
1386 find ! -name .lustre -type f -exec sha1sum {} \; |
1387 sort -k 2 >$tmp/img/sha1sums
1389 $LCTL get_param -n version | head -n 1 |
1390 sed -e 's/^lustre: *//' >$tmp/img/commit
1392 [[ $server_version -ge $(version_code 2.3.50) ]] ||
1393 $LFS quotaon -ug /mnt/$FSNAME
1394 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1395 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1396 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1397 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1398 | tr -d "*" > $tmp/img/bspace
1399 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1400 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1401 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1402 | tr -d "*" > $tmp/img/ispace
1403 echo $T32_BLIMIT > $tmp/img/blimit
1404 echo $T32_ILIMIT > $tmp/img/ilimit
1409 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1412 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1413 echo "Data verification failed"
1416 uname -r >$tmp/img/kernel
1417 uname -m >$tmp/img/arch
1419 mv ${MDSDEV1:-$MDSDEV} $tmp/img
1420 for num in $(seq 2 $MDSCOUNT); do
1421 local devname=$(mdsdevname $num)
1423 [[ $(facet_fstype $facet) != zfs ]] ||
1424 devname=$(mdsvdevname $num)
1425 mv $devname $tmp/img
1427 mv $OSTDEV1 $tmp/img
1429 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1430 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1433 tar cjvf $dst/disk$version-$(facet_fstype $SINGLEMDS).tar.bz2 -S *
1438 #run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1441 # The list of applicable tarballs is returned via the caller's
1442 # variable "tarballs".
1445 local node=$(facet_active_host $SINGLEMDS)
1446 local r="do_node $node"
1448 if [ "$CLIENTONLY" ]; then
1449 skip "Client-only testing"
1453 if ! $r which $TUNEFS; then
1454 skip_env "tunefs.lustre required on $node"
1458 local IMGTYPE=$(facet_fstype $SINGLEMDS)
1460 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 \
1461 -name \'disk*-$IMGTYPE.tar.bz2\')
1463 if [ -z "$tarballs" ]; then
1464 skip "No applicable tarballs found"
1469 t32_test_cleanup() {
1471 local facet=$SINGLEMDS
1472 local fstype=$(facet_fstype $facet)
1475 if $shall_cleanup_lustre; then
1476 umount $tmp/mnt/lustre || rc=$?
1478 if $shall_cleanup_mdt; then
1479 $r $UMOUNT $tmp/mnt/mdt || rc=$?
1481 if $shall_cleanup_mdt1; then
1482 $r $UMOUNT $tmp/mnt/mdt1 || rc=$?
1484 if $shall_cleanup_ost; then
1485 $r $UMOUNT $tmp/mnt/ost || rc=$?
1490 if [[ $fstype == zfs ]]; then
1492 local poolname_list="t32fs-mdt1 t32fs-ost1"
1494 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1496 for poolname in $poolname_list; do
1497 destroy_zpool $facet $poolname
1500 combined_mgs_mds || start_mgs || rc=$?
1504 t32_bits_per_long() {
1506 # Yes, this is not meant to be perfect.
1516 t32_reload_modules() {
1518 local all_removed=false
1520 local fstype=$(facet_fstype $SINGLEMDS)
1522 [ $fstype == "zfs" ] && do_rpc_nodes $node "service zed stop"
1524 while ((i < 20)); do
1525 echo "Unloading modules on $node: Attempt $i"
1526 do_rpc_nodes $node $LUSTRE_RMMOD $fstype &&
1528 do_rpc_nodes $node check_mem_leak || return 1
1529 if $all_removed; then
1530 do_rpc_nodes $node load_modules
1533 if [ $fstype == "zfs" ]; then
1534 do_rpc_nodes $node "$ZPOOL status -v"
1539 echo "Unloading modules on $node: Given up"
1543 t32_wait_til_devices_gone() {
1549 echo wait for devices to go
1550 while ((i < 20)); do
1551 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1552 loops=$(do_rpc_nodes $node losetup -a | grep -c t32)
1553 ((devices == 0 && loops == 0)) && return 0
1557 echo "waiting for dev on $node: dev $devices loop $loops given up"
1558 do_rpc_nodes $node "losetup -a"
1559 do_rpc_nodes $node "$LCTL device_list"
1563 t32_verify_quota() {
1567 local fstype=$(facet_fstype $SINGLEMDS)
1571 # LU-2435: if the underlying zfs doesn't support userobj_accounting,
1572 # lustre will estimate the object count usage. This fails quota
1573 # verification in 32b. The object quota usage should be accurate after
1574 # zfs-0.7.0 is released.
1575 [ $fstype == "zfs" ] && {
1576 local zfs_version=$(do_facet $facet cat /sys/module/zfs/version)
1578 [ $(version_code $zfs_version) -lt $(version_code 0.7.0) ] && {
1579 echo "Skip quota verify for zfs: $zfs_version"
1584 $LFS quota -u $T32_QID -v $mnt
1586 qval=$($LFS quota -v -u $T32_QID $mnt |
1587 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1588 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1590 [ $qval -eq $img_bspace ] || {
1591 echo "bspace, act:$qval, exp:$img_bspace"
1595 qval=$($LFS quota -v -u $T32_QID $mnt |
1596 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1597 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1599 [ $qval -eq $img_ispace ] || {
1600 echo "ispace, act:$qval, exp:$img_ispace"
1604 qval=$($LFS quota -v -u $T32_QID $mnt |
1605 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1606 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1608 [ $qval -eq $img_blimit ] || {
1609 echo "blimit, act:$qval, exp:$img_blimit"
1613 qval=$($LFS quota -v -u $T32_QID $mnt |
1614 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1615 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1617 [ $qval -eq $img_ilimit ] || {
1618 echo "ilimit, act:$qval, exp:$img_ilimit"
1622 set_persistent_param_and_check $facet \
1623 "osd-$fstype.$fsname-MDT0000.quota_slave.enabled" \
1624 "$fsname.quota.mdt" ug
1626 set_persistent_param_and_check $facet \
1627 "osd-$fstype.$fsname-OST0000.quota_slave.enabled" \
1628 "$fsname.quota.ost" ug
1631 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1632 bs=1M count=$((img_blimit / 1024)) oflag=sync && {
1633 echo "Write succeed, but expect -EDQUOT"
1636 rm -f $mnt/t32_qf_new
1638 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1640 echo "Create succeed, but expect -EDQUOT"
1643 unlinkmany $mnt/t32_qf_ $img_ilimit
1651 local dne_upgrade=${dne_upgrade:-"no"}
1652 local dom_upgrade=${dom_upgrade:-"no"}
1653 local ff_convert=${ff_convert:-"no"}
1654 local shall_cleanup_mdt=false
1655 local shall_cleanup_mdt1=false
1656 local shall_cleanup_ost=false
1657 local shall_cleanup_lustre=false
1658 local mdt2_is_available=false
1659 local node=$(facet_active_host $SINGLEMDS)
1660 local r="do_node $node"
1676 local fstype=$(facet_fstype $SINGLEMDS)
1677 local mdt_dev=$tmp/mdt
1678 local mdt2_dev=$tmp/mdt2
1679 local ost_dev=$tmp/ost
1684 combined_mgs_mds || stop_mgs || error "Unable to stop MGS"
1685 trap 'trap - RETURN; t32_test_cleanup' RETURN
1688 nid=$($r $LCTL list_nids | head -1)
1690 mkdir -p $tmp/mnt/lustre || error "mkdir $tmp/mnt/lustre failed"
1691 $r mkdir -p $tmp/mnt/{mdt,mdt1,ost}
1692 $r tar xjvf $tarball -S -C $tmp || {
1693 error_noexit "Unpacking the disk image tarball"
1696 img_commit=$($r cat $tmp/commit)
1697 img_kernel=$($r cat $tmp/kernel)
1698 img_arch=$($r cat $tmp/arch)
1699 img_bspace=$($r cat $tmp/bspace)
1700 img_ispace=$($r cat $tmp/ispace)
1702 # older images did not have "blimit" and "ilimit" files
1703 # use old values for T32_BLIMIT and T32_ILIMIT
1704 $r test -f $tmp/blimit && img_blimit=$($r cat $tmp/blimit) ||
1706 $r test -f $tmp/ilimit && img_ilimit=$($r cat $tmp/ilimit) ||
1709 echo "Upgrading from $(basename $tarball), created with:"
1710 echo " Commit: $img_commit"
1711 echo " Kernel: $img_kernel"
1712 echo " Arch: $img_arch"
1713 echo "OST version: $(lustre_build_version ost1)"
1715 # The conversion can be made only when both of the following
1716 # conditions are satisfied:
1717 # - ost device img version < 2.3.64
1718 # - ost server version >= 2.5
1719 [ $(version_code $img_commit) -ge $(version_code 2.3.64) -o \
1720 $(lustre_version_code ost1) -lt $(version_code 2.5.0) ] &&
1723 ! $r test -f $mdt2_dev || mdt2_is_available=true
1725 if [[ $fstype == zfs ]]; then
1728 local poolname_list="t32fs-mdt1 t32fs-ost1"
1730 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1732 for poolname in $poolname_list; do
1734 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
1735 $ZPOOL import -f -d $tmp $poolname"
1738 # upgrade zpool to latest supported features, including
1739 # dnode quota accounting in 0.7.0
1740 $r "$ZPOOL upgrade -a"
1742 mdt_dev=t32fs-mdt1/mdt1
1743 ost_dev=t32fs-ost1/ost1
1744 ! $mdt2_is_available || mdt2_dev=t32fs-mdt2/mdt2
1745 wait_update_facet $SINGLEMDS "$ZPOOL list |
1746 awk '/^t32fs-mdt1/ { print \\\$1 }'" "t32fs-mdt1" || {
1747 error_noexit "import zfs pool failed"
1752 $r $LCTL set_param debug="$PTLDEBUG"
1754 $r $TUNEFS --dryrun $mdt_dev || {
1756 error_noexit "tunefs.lustre before mounting the MDT"
1760 if $mdt2_is_available; then
1761 $r $TUNEFS --dryrun $mdt2_dev || {
1763 error_noexit "tunefs.lustre before mounting the MDT"
1768 if [ "$writeconf" ]; then
1770 if [ $fstype == "ldiskfs" ]; then
1772 $r $TUNEFS --quota $mdt_dev || {
1774 error_noexit "Enable mdt quota feature"
1777 if $mdt2_is_available; then
1778 $r $TUNEFS --quota $mdt2_dev || {
1780 error_noexit "Enable mdt quota feature"
1786 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
1787 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
1788 { skip "LU-2200: Cannot run over Inifiniband w/o lctl replace_nids "
1789 "(Need MGS version at least 2.3.59)"; return 0; }
1791 local osthost=$(facet_active_host ost1)
1792 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
1795 if [ $fstype == "ldiskfs" ]; then
1798 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt
1799 $r $LCTL replace_nids $fsname-OST0000 $ostnid
1800 $r $LCTL replace_nids $fsname-MDT0000 $nid
1801 $r $UMOUNT $tmp/mnt/mdt
1804 mopts=exclude=$fsname-OST0000
1805 if [ $fstype == "ldiskfs" ]; then
1810 t32_wait_til_devices_gone $node
1812 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
1814 error_noexit "Mounting the MDT"
1817 shall_cleanup_mdt=true
1819 if $mdt2_is_available; then
1820 mopts=mgsnode=$nid,$mopts
1821 $r $MOUNT_CMD -o $mopts $mdt2_dev $tmp/mnt/mdt1 || {
1823 error_noexit "Mounting the MDT"
1827 echo "mount new MDT....$mdt2_dev"
1828 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1829 error_noexit "enable remote dir create failed"
1831 shall_cleanup_mdt1=true
1832 elif [ "$dne_upgrade" != "no" ]; then
1833 local fs2mdsdev=$(mdsdevname 1_2)
1834 local fs2mdsvdev=$(mdsvdevname 1_2)
1836 echo "mkfs new MDT on ${fs2mdsdev}...."
1837 if [ $(facet_fstype mds1) == ldiskfs ]; then
1838 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
1841 add $SINGLEMDS $(mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
1842 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
1843 error_noexit "Mkfs new MDT failed"
1847 [[ $(facet_fstype mds1) != zfs ]] || import_zpool mds1
1849 $r $TUNEFS --dryrun $fs2mdsdev || {
1850 error_noexit "tunefs.lustre before mounting the MDT"
1854 echo "mount new MDT....$fs2mdsdev"
1855 $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
1856 error_noexit "mount mdt1 failed"
1860 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1861 error_noexit "enable remote dir create failed"
1863 shall_cleanup_mdt1=true
1866 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
1867 error_noexit "Getting MDT UUID"
1870 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
1871 error_noexit "Unexpected MDT UUID: \"$uuid\""
1875 $r $TUNEFS --dryrun $ost_dev || {
1876 error_noexit "tunefs.lustre before mounting the OST"
1879 if [ "$writeconf" ]; then
1880 mopts=mgsnode=$nid,$writeconf
1881 if [ $fstype == "ldiskfs" ]; then
1883 $r $TUNEFS --quota $ost_dev || {
1885 error_noexit "Enable ost quota feature"
1891 if [ $fstype == "ldiskfs" ]; then
1896 $r $MOUNT_CMD -onomgs -o$mopts $ost_dev $tmp/mnt/ost || {
1897 error_noexit "Mounting the OST"
1900 shall_cleanup_ost=true
1902 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
1903 error_noexit "Getting OST UUID"
1906 if [ "$uuid" != $fsname-OST0000_UUID ]; then
1907 error_noexit "Unexpected OST UUID: \"$uuid\""
1911 if [[ $PERM_CMD == *"set_param -P"* ]]; then
1912 $r $PERM_CMD osc.$fsname-OST0000*.import=connection=$nid || {
1913 error_noexit "Setting OST \"failover.node\""
1916 $r $PERM_CMD mdc.$fsname-MDT0000*.import=connection=$nid || {
1917 error_noexit "Setting MDT \"failover.node\""
1920 $r $PERM_CMD osc.$fsname-OST0000-*.max_dirty_mb=15 || {
1921 error_noexit "Setting \"max_dirty_mb\""
1924 $r $PERM_CMD mdc.$fsname-MDT0000-*.max_rpcs_in_flight=9 || {
1925 error_noexit "Setting \"max_rpcs_in_flight\""
1928 $r $PERM_CMD lov.$fsname-MDT0000-*.stripesize=4M || {
1929 error_noexit "Setting \"lov.stripesize\""
1932 $r $PERM_CMD mdd.$fsname-MDT0000-*.atime_diff=70 || {
1933 error_noexit "Setting \"mdd.atime_diff\""
1937 $r $PERM_CMD $fsname-OST0000.failover.node=$nid || {
1938 error_noexit "Setting OST \"failover.node\""
1942 $r $PERM_CMD $fsname-MDT0000.failover.node=$nid || {
1943 error_noexit "Setting MDT \"failover.node\""
1947 $r $PERM_CMD $fsname-OST0000.osc.max_dirty_mb=15 || {
1948 error_noexit "Setting \"max_dirty_mb\""
1951 $r $PERM_CMD $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
1952 error_noexit "Setting \"max_rpcs_in_flight\""
1955 $r $PERM_CMD $fsname-MDT0000.lov.stripesize=4M || {
1956 error_noexit "Setting \"lov.stripesize\""
1959 $r $PERM_CMD $fsname-MDT0000.mdd.atime_diff=70 || {
1960 error_noexit "Setting \"mdd.atime_diff\""
1965 $r $LCTL pool_new $fsname.interop || {
1966 error_noexit "Setting \"interop\""
1970 if [ "$ff_convert" != "no" -a $(facet_fstype ost1) == "ldiskfs" ]; then
1971 $r $LCTL lfsck_start -M $fsname-OST0000 || {
1972 error_noexit "Start OI scrub on OST0"
1976 # The oi_scrub should be on ost1, but for test_32(),
1977 # all on the SINGLEMDS.
1978 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
1979 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1980 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
1981 error_noexit "Failed to get the expected 'completed'"
1985 local UPDATED=$($r $LCTL get_param -n \
1986 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1987 awk '/^updated/ { print $2 }')
1988 [ $UPDATED -ge 1 ] || {
1989 error_noexit "Only $UPDATED objects have been converted"
1994 if [ "$dne_upgrade" != "no" ]; then
1995 if [[ $PERM_CMD == *"set_param -P"* ]]; then
1996 $r $PERM_CMD mdc.$fsname-MDT0001*.import=connection=$nid || {
1997 error_noexit "Setting MDT1 \"failover.node\""
2001 $r $PERM_CMD mdc.$fsname-MDT0001-*.max_rpcs_in_flight=9 || {
2002 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
2005 $r $PERM_CMD lov.$fsname-MDT0001-*.stripesize=4M || {
2006 error_noexit "Setting MDT1 \"lov.stripesize\""
2010 $r $PERM_CMD $fsname-MDT0001.failover.node=$nid || {
2011 error_noexit "Setting MDT1 \"failover.node\""
2014 $r $PERM_CMD $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
2015 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
2018 $r $PERM_CMD $fsname-MDT0001.lov.stripesize=4M || {
2019 error_noexit "Setting MDT1 \"lov.stripesize\""
2025 if [ "$writeconf" ]; then
2026 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
2027 error_noexit "Mounting the client"
2031 shall_cleanup_lustre=true
2032 $r $LCTL set_param debug="$PTLDEBUG"
2034 # Leave re-enabling this to a separate patch for LU-11558
2035 # t32_verify_quota $SINGLEMDS $fsname $tmp/mnt/lustre || {
2036 # error_noexit "verify quota failed"
2040 if $r test -f $tmp/list; then
2042 # There is not a Test Framework API to copy files to or
2043 # from a remote node.
2045 # LU-2393 - do both sorts on same node to ensure locale
2047 local list_file=$tmp/list
2049 if $mdt2_is_available; then
2050 if [[ -d $tmp/mnt/lustre/striped_dir_old ]] &&
2051 $r test -f $tmp/list2; then
2052 list_file=$tmp/list2
2053 pushd $tmp/mnt/lustre/striped_dir_old
2055 pushd $tmp/mnt/lustre/remote_dir
2058 pushd $tmp/mnt/lustre
2060 $r cat $list_file | sort -k 6 >$tmp/list.orig
2061 ls -Rni --time-style=+%s | sort -k 6 |
2062 sed 's/\. / /' >$tmp/list || {
2068 # 32-bit and 64-bit clients use different algorithms to
2069 # convert FIDs into inode numbers. Hence, remove the
2070 # inode numbers from the lists, if the original list was
2071 # created on an architecture with different number of
2074 if [ $(t32_bits_per_long $(uname -m)) != \
2075 $(t32_bits_per_long $img_arch) ]; then
2076 echo "Different number of bits per \"long\"" \
2077 "from the disk image"
2078 for list in list.orig list; do
2079 sed -i -e 's/^[0-9]\+[ \t]\+//' \
2083 if ! diff -ub $tmp/list.orig $tmp/list; then
2084 error_noexit "list verification failed"
2088 echo "list verification skipped"
2091 if [ "$dom_upgrade" != "no" ]; then
2092 echo "Check DoM file can be created"
2093 $LFS setstripe -E 1M -L mdt -E EOF $tmp/mnt/lustre/dom || {
2094 error_noexit "Verify DoM creation"
2097 [ $($LFS getstripe -L $tmp/mnt/lustre/dom) == "mdt" ] || {
2098 error_noexit "Verify a DoM file"
2101 dd if=/dev/urandom of=$tmp/mnt/lustre/dom bs=4096 \
2102 count=1 conv=fsync || {
2103 error_noexit "Cannot write to DoM file"
2106 [ $(stat -c%s $tmp/mnt/lustre/dom) == 4096 ] || {
2107 error_noexit "DoM: bad size after write"
2110 rm $tmp/mnt/lustre/dom
2112 set_persistent_param_and_check mds \
2113 "lod.*MDT0000*.dom_stripesize" \
2114 "$fsname-MDT0000.lod.dom_stripesize" 0 || {
2115 error_noexit "Changing \"dom_stripesize\""
2120 if [ "$dne_upgrade" != "no" ]; then
2121 $LFS mkdir -i 1 -c2 $tmp/mnt/lustre/striped_dir || {
2122 error_noexit "set striped dir failed"
2126 $LFS setdirstripe -D -c2 $tmp/mnt/lustre/striped_dir
2128 pushd $tmp/mnt/lustre
2129 tar -cf - . --exclude=./striped_dir \
2130 --exclude=./striped_dir_old \
2131 --exclude=./remote_dir |
2132 tar -xvf - -C striped_dir 1>/dev/null || {
2133 error_noexit "cp to striped dir failed"
2139 # If it is upgrade from DNE (2.5), then rename the remote dir,
2140 # which is created in 2.5 to striped dir.
2141 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]]; then
2142 stripe_index=$($LFS getdirstripe -i \
2143 $tmp/mnt/lustre/remote_dir)
2145 [[ $stripe_index -eq 1 ]] || {
2146 error_noexit "get index \"$stripe_index\"" \
2147 "from remote dir failed"
2150 mv $tmp/mnt/lustre/remote_dir \
2151 $tmp/mnt/lustre/striped_dir/ || {
2152 error_noexit "mv remote dir failed"
2157 # If it is upgraded from DNE (2.7), then move the striped dir
2158 # which was created in 2.7 to the new striped dir.
2159 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]] &&
2160 [[ -d $tmp/mnt/lustre/striped_dir_old ]]; then
2161 stripe_count=$($LFS getdirstripe -c \
2162 $tmp/mnt/lustre/striped_dir_old)
2163 [[ $stripe_count -eq 2 ]] || {
2164 error_noexit "get count $stripe_count" \
2165 "from striped dir failed"
2168 mv $tmp/mnt/lustre/striped_dir_old \
2169 $tmp/mnt/lustre/striped_dir/ || {
2170 error_noexit "mv striped dir failed"
2176 $r $LCTL set_param -n osd*.*.force_sync=1
2177 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
2178 error_noexit "dd failed"
2181 rm -rf $tmp/mnt/lustre/tmp_file || {
2182 error_noexit "rm failed"
2186 if $r test -f $tmp/sha1sums; then
2187 # LU-2393 - do both sorts on same node to ensure locale
2189 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
2190 if [ "$dne_upgrade" != "no" ]; then
2191 pushd $tmp/mnt/lustre/striped_dir
2193 pushd $tmp/mnt/lustre
2196 find ! -path "*remote_dir*" ! -path "*striped_dir*" \
2197 ! -name .lustre -type f -exec sha1sum {} \; |
2198 sort -k 2 >$tmp/sha1sums || {
2200 error_noexit "sha1sum"
2204 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
2205 error_noexit "sha1sum verification failed"
2209 # if upgrade from DNE(2.5), then check remote directory
2210 # if upgrade from DNE(2.7), then check striped directory
2211 if $mdt2_is_available &&
2212 [[ "$dne_upgrade" != "no" ]]; then
2213 local new_dir="$tmp/mnt/lustre/striped_dir"
2214 local striped_dir_old="$new_dir/striped_dir_old"
2216 local dir_list="$new_dir/remote_dir"
2217 [[ ! -d $triped_dir_old ]] ||
2218 dir_list+=" $striped_dir_old"
2220 for dir in $dir_list; do
2222 find ! -name .lustre -type f \
2223 -exec sha1sum {} \; |
2224 sort -k 2 >$tmp/sha1sums || {
2226 error_noexit "sha1sum"
2230 if ! diff -ub $tmp/sha1sums.orig \
2232 error_noexit "sha1sum $dir" \
2239 echo "sha1sum verification skipped"
2242 if [ "$dne_upgrade" != "no" ]; then
2243 rm -rf $tmp/mnt/lustre/striped_dir || {
2244 error_noexit "remove remote dir failed"
2249 # migrate files/dirs to remote MDT, then move them back
2250 if [ $(lustre_version_code mds1) -ge $(version_code 2.7.50) -a \
2251 $dne_upgrade != "no" ]; then
2252 $r $LCTL set_param -n \
2253 mdt.${fsname}*.enable_remote_dir=1 2>/dev/null
2255 echo "test migration"
2256 pushd $tmp/mnt/lustre
2257 for dir in $(find ! -name .lustre ! -name . -type d); do
2258 mdt_index=$($LFS getdirstripe -i $dir)
2259 stripe_cnt=$($LFS getdirstripe -c $dir)
2260 if [ $mdt_index = 0 -a $stripe_cnt -le 1 ]; then
2261 $LFS migrate -m 1 $dir || {
2263 error_noexit "migrate MDT1 failed"
2269 for dir in $(find ! -name . ! -name .lustre -type d); do
2270 mdt_index=$($LFS getdirstripe -i $dir)
2271 stripe_cnt=$($LFS getdirstripe -c $dir)
2272 if [ $mdt_index = 1 -a $stripe_cnt -le 1 ]; then
2273 $LFS migrate -m 0 $dir || {
2275 error_noexit "migrate MDT0 failed"
2284 # When adding new data verification tests, please check for
2285 # the presence of the required reference files first, like
2286 # the "sha1sums" and "list" tests above, to avoid the need to
2287 # regenerate every image for each test addition.
2290 nrpcs_orig=$($LCTL get_param \
2291 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
2292 error_noexit "Getting \"max_rpcs_in_flight\""
2295 nrpcs=$((nrpcs_orig + 5))
2297 set_persistent_param_and_check client \
2298 "mdc.$fsname-MDT0000*.max_rpcs_in_flight" \
2299 "$fsname-MDT0000.mdc.max_rpcs_in_flight" $nrpcs || {
2300 error_noexit "Changing \"max_rpcs_in_flight\""
2304 umount $tmp/mnt/lustre || {
2305 error_noexit "Unmounting the client"
2308 shall_cleanup_lustre=false
2310 if [[ "$dne_upgrade" != "no" ]] || $mdt2_is_available; then
2311 $r $UMOUNT $tmp/mnt/mdt1 || {
2312 error_noexit "Unmounting the MDT2"
2315 if [[ $fstype == zfs ]]; then
2316 $r "$ZPOOL export t32fs-mdt2"
2318 shall_cleanup_mdt1=false
2321 $r $UMOUNT $tmp/mnt/mdt || {
2322 error_noexit "Unmounting the MDT"
2325 if [[ $fstype == zfs ]]; then
2326 $r "$ZPOOL export t32fs-mdt1"
2328 shall_cleanup_mdt=false
2330 $r $UMOUNT $tmp/mnt/ost || {
2331 error_noexit "Unmounting the OST"
2334 if [[ $fstype == zfs ]]; then
2335 $r "$ZPOOL export t32fs-ost1"
2337 shall_cleanup_ost=false
2339 t32_reload_modules $node || {
2340 error_noexit "Reloading modules"
2344 if [[ $fstype == zfs ]]; then
2345 local poolname=t32fs-mdt1
2347 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
2348 $ZPOOL import -f -d $tmp $poolname"
2350 # upgrade zpool to latest supported features,
2351 # including dnode quota accounting in 0.7.0
2352 $r "$ZPOOL upgrade $poolname"
2355 # mount a second time to make sure we didnt leave upgrade flag on
2356 $r $TUNEFS --dryrun $mdt_dev || {
2358 error_noexit "tunefs.lustre before remounting the MDT"
2362 mopts=exclude=$fsname-OST0000
2363 if [ $fstype == "ldiskfs" ]; then
2366 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2367 error_noexit "Remounting the MDT"
2370 shall_cleanup_mdt=true
2380 for tarball in $tarballs; do
2381 t32_test $tarball || let "rc += $?"
2385 run_test 32a "Upgrade (not live)"
2393 for tarball in $tarballs; do
2394 t32_test $tarball writeconf || let "rc += $?"
2398 run_test 32b "Upgrade with writeconf"
2405 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
2407 for tarball in $tarballs; do
2408 # Do not support 1_8 and 2_1 direct upgrade to DNE2 anymore */
2409 echo $tarball | grep "1_8" && continue
2410 echo $tarball | grep "2_1" && continue
2412 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
2416 run_test 32c "dne upgrade test"
2424 for tarball in $tarballs; do
2425 ff_convert=yes t32_test $tarball || rc=$?
2429 run_test 32d "convert ff test"
2432 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.10.56) ]] ||
2433 { skip "Need MDS version at least 2.10.56"; return 0; }
2440 for tarball in $tarballs; do
2441 echo $tarball | grep "2_9" || continue
2443 dom_upgrade=yes t32_test $tarball writeconf || let "rc += $?"
2447 run_test 32e "dom upgrade test"
2449 test_33a() { # bug 12333, was test_33
2450 local FSNAME2=test-123
2451 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2454 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
2456 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
2457 local dev=${SINGLEMDS}_dev
2458 local MDSDEV=${!dev}
2459 is_blkdev $SINGLEMDS $MDSDEV &&
2460 skip_env "mixed loopback and real device not working" &&
2464 local fs2mdsdev=$(mdsdevname 1_2)
2465 local fs2ostdev=$(ostdevname 1_2)
2466 local fs2mdsvdev=$(mdsvdevname 1_2)
2467 local fs2ostvdev=$(ostvdevname 1_2)
2469 if [ $(facet_fstype mds1) == ldiskfs ]; then
2470 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
2473 if combined_mgs_mds; then
2474 local mgs_flag="--mgs"
2477 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --fsname=${FSNAME2} \
2478 --reformat $mgs_flag $mkfsoptions $fs2mdsdev $fs2mdsvdev ||
2480 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2481 --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \
2482 $fs2ostvdev || exit 10
2484 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
2485 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2487 if [[ $PERM_CMD == *"set_param -P"* ]]; then
2488 do_facet mgs "$PERM_CMD timeout=200" ||
2489 error "$PERM_CMD timeout=200 failed"
2491 do_facet mgs "$PERM_CMD $FSNAME2.sys.timeout=200" ||
2492 error "$PERM_CMD $FSNAME2.sys.timeout=200 failed"
2494 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2495 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2498 cp /etc/hosts $MOUNT2/ || error "copy /etc/hosts $MOUNT2/ failed"
2499 $GETSTRIPE $MOUNT2/hosts || error "$GETSTRIPE $MOUNT2/hosts failed"
2504 cleanup_nocli || error "cleanup_nocli failed with $?"
2506 run_test 33a "Mount ost with a large index number"
2508 test_33b() { # was test_34
2511 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
2512 # Drop lock cancelation reply during umount
2513 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
2514 do_facet client $LCTL set_param fail_loc=0x80000304
2515 #lctl set_param debug=-1
2516 umount_client $MOUNT
2517 cleanup || error "cleanup failed with $?"
2519 run_test 33b "Drop cancel during umount"
2523 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
2524 manual_umount_client
2526 do_facet client killall -USR1 multiop
2527 if [ $rc -eq 0 ]; then
2528 error "umount not fail!"
2531 cleanup || error "cleanup failed with rc $?"
2533 run_test 34a "umount with opened file should be fail"
2537 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2538 stop_mds || error "Unable to stop MDS"
2540 manual_umount_client --force || error "mtab after failed umount with $?"
2542 cleanup || error "cleanup failed with $?"
2544 run_test 34b "force umount with failed mds should be normal"
2548 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2549 stop_ost || error "Unable to stop OST1"
2551 manual_umount_client --force || error "mtab after failed umount with $?"
2553 cleanup || error "cleanup failed with $?"
2555 run_test 34c "force umount with failed ost should be normal"
2557 test_35a() { # bug 12459
2560 DBG_SAVE="`$LCTL get_param -n debug`"
2561 $LCTL set_param debug="ha"
2563 log "Set up a fake failnode for the MDS"
2565 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2566 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2568 if [[ $PERM_CMD == *"set_param -P"* ]]; then
2569 do_facet mgs "$PERM_CMD \
2570 mdc.*${device}*.import=connection=$(h2nettype $FAKENID)" ||
2571 error "Setting mdc.*${device}*.import=connection=\
2572 $(h2nettype $FAKENID) failed."
2574 do_facet mgs "$PERM_CMD \
2575 ${device}.failover.node=$(h2nettype $FAKENID)" ||
2576 error "Setting ${device}.failover.node=\
2577 $(h2nettype $FAKENID) failed."
2579 log "Wait for RECONNECT_INTERVAL seconds (10s)"
2582 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
2585 log "Stopping the MDT: $device"
2586 stop_mdt 1 || error "MDT0 stop fail"
2588 df $MOUNT > /dev/null 2>&1 &
2590 log "Restarting the MDT: $device"
2591 start_mdt 1 || error "MDT0 start fail"
2592 log "Wait for df ($DFPID) ... "
2595 $LCTL set_param debug="$DBG_SAVE"
2597 # retrieve from the log the first server that the client tried to
2598 # contact after the connection loss
2599 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2600 NEXTCONN=`awk "/${MSG}/ {start = 1;}
2601 /import_select_connection.*$device-mdc.* using connection/ {
2603 if (\\\$NF ~ /$FAKENID/)
2609 }" $TMP/lustre-log-$TESTNAME.log`
2610 [ "$NEXTCONN" != "0" ] &&
2611 error "Tried to connect to ${NEXTCONN} not last active server"
2612 cleanup || error "cleanup failed with $?"
2613 # remove nid settings
2614 writeconf_or_reformat
2616 run_test 35a "Reconnect to the last active server first"
2618 test_35b() { # bug 18674
2619 remote_mds || { skip "local MDS" && return 0; }
2623 $LCTL set_param debug="ha"
2625 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
2628 log "Set up a fake failnode for the MDS"
2630 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2631 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2633 if [[ $PERM_CMD == *"set_param -P"* ]]; then
2634 do_facet mgs "$PERM_CMD \
2635 mdc.*${device}*.import=connection=$(h2nettype $FAKENID)" ||
2636 error "Set mdc.*${device}*.import=connection=\
2637 $(h2nettype $FAKENID) failed"
2639 do_facet mgs "$PERM_CMD \
2640 ${device}.failover.node=$(h2nettype $FAKENID)" ||
2641 error "Set ${device}.failover.node=\
2642 $(h2nettype $FAKENID) failed"
2645 local at_max_saved=0
2646 # adaptive timeouts may prevent seeing the issue
2647 if at_is_enabled; then
2648 at_max_saved=$(at_max_get mds)
2649 at_max_set 0 mds client
2652 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
2654 log "Injecting EBUSY on MDS"
2655 # Setting OBD_FAIL_MDS_RESEND=0x136
2656 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" ||
2657 error "unable to set param fail_loc=0x80000136"
2659 $LCTL set_param mdc.${FSNAME}*.stats=clear
2661 log "Creating a test file and stat it"
2662 touch $MOUNT/$tdir/$tfile || error "touch $MOUNT/$tdir/$tfile failed"
2663 stat $MOUNT/$tdir/$tfile
2665 log "Stop injecting EBUSY on MDS"
2666 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" ||
2667 error "unable to set param fail_loc=0"
2668 rm -f $MOUNT/$tdir/$tfile || error "remove $MOUNT/$tdir/$tfile failed"
2671 # restore adaptive timeout
2672 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
2674 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2676 CONNCNT=$($LCTL get_param mdc.${FSNAME}*.stats |
2677 awk '/mds_connect/{print $2}')
2679 # retrieve from the log if the client has ever tried to
2680 # contact the fake server after the loss of connection
2681 FAILCONN=`awk "BEGIN {ret = 0;}
2682 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
2684 if (\\\$NF ~ /$FAKENID/) {
2689 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
2691 [ "$FAILCONN" == "0" ] &&
2692 error "The client reconnection has not been triggered"
2693 [ "$FAILCONN" == "2" ] &&
2694 error "Primary server busy, client reconnect to failover failed"
2697 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
2698 # Reconnects are supposed to be rate limited to one every 5s
2699 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] &&
2700 error "Too many reconnects $CONNCNT"
2702 cleanup || error "cleanup failed with $?"
2703 # remove nid settings
2704 writeconf_or_reformat
2706 run_test 35b "Continue reconnection retries, if the active server is busy"
2709 [ $OSTCOUNT -lt 2 ] && skip_env "needs >= 2 OSTs" && return
2711 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] ||
2712 { skip "remote OST" && return 0; }
2715 local FSNAME2=test1234
2716 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2718 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
2720 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
2721 is_blkdev $SINGLEMDS $MDSDEV &&
2722 skip_env "mixed loopback and real device not working" && return
2725 local fs2mdsdev=$(mdsdevname 1_2)
2726 local fs2ostdev=$(ostdevname 1_2)
2727 local fs3ostdev=$(ostdevname 2_2)
2728 local fs2mdsvdev=$(mdsvdevname 1_2)
2729 local fs2ostvdev=$(ostvdevname 1_2)
2730 local fs3ostvdev=$(ostvdevname 2_2)
2732 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2733 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
2734 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
2735 # different one than the default value here.
2736 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2737 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
2738 add fs3ost $(mkfs_opts ost2 ${fs3ostdev}) --mgsnode=$MGSNID \
2739 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
2741 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
2742 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2743 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
2744 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2745 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2747 sleep 5 # until 11778 fixed
2749 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || error "dd failed"
2751 BKTOTAL=$($LCTL get_param -n obdfilter.*.kbytestotal |
2752 awk 'BEGIN{total=0}; {total+=$1}; END{print total}')
2753 BKFREE=$($LCTL get_param -n obdfilter.*.kbytesfree |
2754 awk 'BEGIN{free=0}; {free+=$1}; END{print free}')
2755 BKAVAIL=$($LCTL get_param -n obdfilter.*.kbytesavail |
2756 awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}')
2757 STRING=$(df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}')
2758 DFTOTAL=$(echo $STRING | cut -d, -f1)
2759 DFUSED=$(echo $STRING | cut -d, -f2)
2760 DFAVAIL=$(echo $STRING | cut -d, -f3)
2761 DFFREE=$(($DFTOTAL - $DFUSED))
2763 ALLOWANCE=$((64 * $OSTCOUNT))
2765 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
2766 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
2767 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
2770 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
2771 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
2772 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
2775 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
2776 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
2777 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
2782 stop fs3ost -f || error "unable to stop OST3"
2783 stop fs2ost -f || error "unable to stop OST2"
2784 stop fs2mds -f || error "unable to stop second MDS"
2785 unload_modules_conf || error "unable unload modules"
2788 run_test 36 "df report consistency on OSTs with different block size"
2791 local mntpt=$(facet_mntpt $SINGLEMDS)
2792 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
2793 local mdsdev_sym="$TMP/sym_mdt.img"
2794 local opts=$MDS_MOUNT_OPTS
2797 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2798 skip "ldiskfs only test"
2802 echo "MDS : $mdsdev"
2803 echo "SYMLINK : $mdsdev_sym"
2804 do_facet $SINGLEMDS rm -f $mdsdev_sym
2806 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
2808 echo "mount symlink device - $mdsdev_sym"
2810 if ! do_facet $SINGLEMDS test -b $mdsdev; then
2811 opts=$(csa_add "$opts" -o loop)
2813 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
2814 $mdsdev_sym $mntpt 2>&1)
2817 echo mount_op=$mount_op
2819 do_facet $SINGLEMDS "$UMOUNT $mntpt && rm -f $mdsdev_sym"
2821 if $(echo $mount_op | grep -q "unable to set tunable"); then
2822 error "set tunables failed for symlink device"
2825 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
2827 run_test 37 "verify set tunables works for symlink device"
2829 test_38() { # bug 14222
2830 local fstype=$(facet_fstype $SINGLEMDS)
2831 local mntpt=$(facet_mntpt $SINGLEMDS)
2836 local SRC="/etc /bin"
2837 local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT)
2838 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
2839 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2840 tar cf - $FILES | tar xf - -C $DIR/$tdir ||
2841 error "copying $SRC to $DIR/$tdir"
2843 umount_client $MOUNT || error "umount_client $MOUNT failed"
2844 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2845 stop_mds || error "Unable to stop MDS"
2846 log "delete lov_objid file on MDS"
2848 mount_fstype $SINGLEMDS || error "mount MDS failed (1)"
2850 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid; rm $mntpt/lov_objid"
2852 unmount_fstype $SINGLEMDS || error "umount failed (1)"
2854 # check create in mds_lov_connect
2855 start_mds || error "unable to start MDS"
2856 mount_client $MOUNT || error "mount_client $MOUNT failed"
2858 [ $V ] && log "verifying $DIR/$tdir/$f"
2859 diff -q $f $DIR/$tdir/$f || ERROR=y
2861 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2862 if [ "$ERROR" = "y" ]; then
2863 # check it's updates in sync
2864 umount_client $MOUNT
2866 mount_fstype $SIGNLEMDS
2867 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2868 unmount_fstype $SINGLEMDS
2869 error "old and new files are different after connect" || true
2871 touch $DIR/$tdir/f2 || error "f2 file create failed"
2873 # check it's updates in sync
2874 umount_client $MOUNT || error "second umount_client $MOUNT failed"
2877 mount_fstype $SINGLEMDS || error "mount MDS failed (3)"
2879 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2880 do_facet $SINGLEMDS dd if=/dev/zero of=$mntpt/lov_objid.clear count=8
2882 unmount_fstype $SINGLEMDS || error "umount failed (3)"
2884 start_mds || error "unable to start MDS"
2885 mount_client $MOUNT || error "mount_client $MOUNT failed"
2887 [ $V ] && log "verifying $DIR/$tdir/$f"
2888 diff -q $f $DIR/$tdir/$f || ERROR=y
2890 touch $DIR/$tdir/f3 || error "f3 file create failed"
2891 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2892 umount_client $MOUNT || error "third umount_client $MOUNT failed"
2894 mount_fstype $SINGLEMDS || error "mount MDS failed (4)"
2895 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2896 unmount_fstype $SINGLEMDS || error "umount failed (4)"
2898 [ "$ERROR" = "y" ] &&
2899 error "old and new files are different after sync" || true
2901 log "files compared the same"
2902 cleanup || error "cleanup failed with $?"
2904 run_test 38 "MDS recreates missing lov_objid file from OST data"
2909 cleanup || error "cleanup failed with $?"
2910 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
2911 error "memory leak detected" || true
2913 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
2915 test_40() { # bug 15759
2916 start_ost || error "Unable to start OST1"
2917 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
2918 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
2920 cleanup || error "cleanup failed with rc $?"
2922 run_test 40 "race during service thread startup"
2924 test_41a() { #bug 14134
2925 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2926 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2927 skip "Loop devices does not work with nosvc option"
2932 { skip "needs combined MGT and MDT device" && return 0; }
2934 start_mdt 1 -o nosvc -n
2935 if [ $MDSCOUNT -ge 2 ]; then
2936 for num in $(seq 2 $MDSCOUNT); do
2937 start_mdt $num || return
2940 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
2941 start_mdt 1 -o nomgs,force
2942 mount_client $MOUNT || error "mount_client $MOUNT failed"
2945 echo "blah blah" > $MOUNT/$tfile
2948 umount_client $MOUNT || error "umount_client $MOUNT failed"
2949 stop ost1 -f || error "unable to stop OST1"
2950 stop_mds || error "Unable to stop MDS"
2951 stop_mds || error "Unable to stop MDS on second try"
2953 run_test 41a "mount mds with --nosvc and --nomgs"
2956 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2957 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2958 skip "Loop devices does not work with nosvc option"
2962 ! combined_mgs_mds && skip "needs combined mgs device" && return 0
2966 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2968 start_mdt 1 -o nosvc -n
2969 if [ $MDSCOUNT -ge 2 ]; then
2970 for num in $(seq 2 $MDSCOUNT); do
2971 start_mdt $num || return
2974 start_ost || error "Unable to start OST1"
2975 start_mdt 1 -o nomgs,force
2976 mount_client $MOUNT || error "mount_client $MOUNT failed"
2979 echo "blah blah" > $MOUNT/$tfile
2980 cat $MOUNT/$tfile || error "cat $MOUNT/$tfile failed"
2982 umount_client $MOUNT -f || error "umount_client $MOUNT failed"
2983 stop_ost || error "Unable to stop OST1"
2984 stop_mds || error "Unable to stop MDS"
2985 stop_mds || error "Unable to stop MDS on second try"
2987 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
2990 local server_version=$(lustre_version_code $SINGLEMDS)
2991 local oss_list=$(comma_list $(osts_nodes))
2993 [[ $server_version -ge $(version_code 2.6.52) ]] ||
2994 [[ $server_version -ge $(version_code 2.5.26) &&
2995 $server_version -lt $(version_code 2.5.50) ]] ||
2996 [[ $server_version -ge $(version_code 2.5.4) &&
2997 $server_version -lt $(version_code 2.5.11) ]] ||
2998 { skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"; return; }
3000 # ensure mds1 ost1 have been created even if running sub-test standalone
3003 cleanup || error "cleanup failed"
3005 # using directly mount command instead of start() function to avoid
3006 # any side effect of // with others/externals tools/features
3007 # ("zpool import", ...)
3009 # MDT concurrent start
3011 LOAD_MODULES_REMOTE=true load_modules
3012 do_facet $SINGLEMDS "lsmod | grep -q libcfs" ||
3013 error "MDT concurrent start: libcfs module not loaded"
3015 local mds1dev=$(mdsdevname 1)
3016 local mds1mnt=$(facet_mntpt mds1)
3017 local mds1fstype=$(facet_fstype mds1)
3018 local mds1opts=$MDS_MOUNT_OPTS
3020 if [ $mds1fstype == ldiskfs ] &&
3021 ! do_facet mds1 test -b $mds1dev; then
3022 mds1opts=$(csa_add "$mds1opts" -o loop)
3024 if [[ $mds1fstype == zfs ]]; then
3025 import_zpool mds1 || return ${PIPESTATUS[0]}
3028 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3029 do_facet mds1 "$LCTL set_param fail_loc=0x80000716"
3031 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts &
3034 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts
3038 do_facet mds1 "$LCTL set_param fail_loc=0x0"
3039 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3040 echo "1st MDT start succeed"
3041 echo "2nd MDT start failed with $rc2"
3042 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3043 echo "1st MDT start failed with $rc"
3044 echo "2nd MDT start succeed"
3047 error "unexpected concurrent MDT mounts result, rc=$rc rc2=$rc2"
3050 if [ $MDSCOUNT -ge 2 ]; then
3051 for num in $(seq 2 $MDSCOUNT); do
3052 start_mdt $num || return
3056 # OST concurrent start
3058 do_rpc_nodes $oss_list "lsmod | grep -q libcfs" ||
3059 error "OST concurrent start: libcfs module not loaded"
3061 local ost1dev=$(ostdevname 1)
3062 local ost1mnt=$(facet_mntpt ost1)
3063 local ost1fstype=$(facet_fstype ost1)
3064 local ost1opts=$OST_MOUNT_OPTS
3066 if [ $ost1fstype == ldiskfs ] &&
3067 ! do_facet ost1 test -b $ost1dev; then
3068 ost1opts=$(csa_add "$ost1opts" -o loop)
3070 if [[ $ost1fstype == zfs ]]; then
3071 import_zpool ost1 || return ${PIPESTATUS[0]}
3074 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3075 do_facet ost1 "$LCTL set_param fail_loc=0x80000716"
3077 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts &
3080 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts
3084 do_facet ost1 "$LCTL set_param fail_loc=0x0"
3085 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3086 echo "1st OST start succeed"
3087 echo "2nd OST start failed with $rc2"
3088 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3089 echo "1st OST start failed with $rc"
3090 echo "2nd OST start succeed"
3094 error "unexpected concurrent OST mounts result, rc=$rc rc2=$rc2"
3100 # verify everything ok
3105 error "MDT(s) start failed"
3113 error "OST(s) start failed"
3121 error "client start failed"
3128 error "client mount failed"
3132 run_test 41c "concurrent mounts of MDT/OST should all fail but one"
3134 test_42() { #bug 14693
3138 check_mount || error "client was not mounted"
3140 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3141 PARAM="llite.$FSNAME-*.some_wrong_param"
3143 PARAM="$FSNAME.llite.some_wrong_param"
3146 do_facet mgs $PERM_CMD $PARAM=10
3147 umount_client $MOUNT ||
3148 error "unmounting client failed with invalid llite param"
3149 mount_client $MOUNT ||
3150 error "mounting client failed with invalid llite param"
3152 do_facet mgs $PERM_CMD $PARAM=20
3153 cleanup || error "stopping $FSNAME failed with invalid sys param"
3155 check_mount || error "client was not mounted with invalid sys param"
3156 cleanup || error "stopping $FSNAME failed with invalid sys param"
3158 run_test 42 "allow client/server mount/unmount with invalid config param"
3161 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.58) ]] ||
3162 { skip "Need MDS version at least 2.5.58" && return 0; }
3163 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
3166 USER1=$(getent passwd | grep :$ID1:$ID1: | cut -d: -f1)
3167 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1" &&
3171 chmod ugo+x $DIR || error "chmod 0 failed"
3172 set_persistent_param_and_check mds1 \
3173 "mdt.$FSNAME-MDT0000.root_squash" \
3174 "$FSNAME.mdt.root_squash" \
3176 wait_update $HOSTNAME \
3177 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3179 error "check llite root_squash failed!"
3180 set_persistent_param_and_check mds1 \
3181 "mdt.$FSNAME-MDT0000.nosquash_nids" \
3182 "$FSNAME.mdt.nosquash_nids" \
3184 wait_update $HOSTNAME \
3185 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3187 error "check llite nosquash_nids failed!"
3190 # create set of test files
3192 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
3193 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
3194 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
3196 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
3197 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
3199 mkdir $DIR/$tdir-rootdir || error "mkdir failed"
3200 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
3201 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
3203 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
3204 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
3205 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
3208 # check root_squash:
3209 # set root squash UID:GID to RUNAS_ID
3210 # root should be able to access only files owned by RUNAS_ID
3212 set_persistent_param_and_check mds1 \
3213 "mdt.$FSNAME-MDT0000.root_squash" \
3214 "$FSNAME.mdt.root_squash" \
3215 "$RUNAS_ID:$RUNAS_ID"
3216 wait_update $HOSTNAME \
3217 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3218 "$RUNAS_ID:$RUNAS_ID" ||
3219 error "check llite root_squash failed!"
3221 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
3222 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3223 error "$ST: root read permission is denied"
3224 echo "$ST: root read permission is granted - ok"
3227 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3228 error "$ST: root write permission is denied"
3229 echo "$ST: root write permission is granted - ok"
3231 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3232 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3233 error "$ST: root read permission is granted"
3234 echo "$ST: root read permission is denied - ok"
3237 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3238 error "$ST: root write permission is granted"
3239 echo "$ST: root write permission is denied - ok"
3241 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3242 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null &&
3243 error "$ST: root unlink permission is granted"
3244 echo "$ST: root unlink permission is denied - ok"
3246 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null &&
3247 error "$ST: root create permission is granted"
3248 echo "$ST: root create permission is denied - ok"
3251 # check root_squash is enforced independently
3252 # of client cache content
3254 # access file by USER1, keep access open
3255 # root should be denied access to user file
3257 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
3261 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
3262 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3263 { kill $pid; error "$ST: root read permission is granted"; }
3264 echo "$ST: root read permission is denied - ok"
3267 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3268 { kill $pid; error "$ST: root write permission is granted"; }
3269 echo "$ST: root write permission is denied - ok"
3275 # check nosquash_nids:
3276 # put client's NID into nosquash_nids list,
3277 # root should be able to access root file after that
3279 local NIDLIST=$($LCTL list_nids all | tr '\n' ' ')
3280 NIDLIST="2@gni $NIDLIST 192.168.0.[2,10]@tcp"
3281 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
3282 set_persistent_param_and_check mds1 \
3283 "mdt.$FSNAME-MDT0000.nosquash_nids" \
3284 "$FSNAME-MDTall.mdt.nosquash_nids" \
3286 wait_update $HOSTNAME \
3287 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3289 error "check llite nosquash_nids failed!"
3291 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3292 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3293 error "$ST: root read permission is denied"
3294 echo "$ST: root read permission is granted - ok"
3297 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3298 error "$ST: root write permission is denied"
3299 echo "$ST: root write permission is granted - ok"
3301 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3302 rm $DIR/$tdir-rootdir/tfile-1 ||
3303 error "$ST: root unlink permission is denied"
3304 echo "$ST: root unlink permission is granted - ok"
3305 touch $DIR/$tdir-rootdir/tfile-2 ||
3306 error "$ST: root create permission is denied"
3307 echo "$ST: root create permission is granted - ok"
3308 cleanup || error "cleanup failed with $?"
3310 run_test 43a "check root_squash and nosquash_nids"
3312 test_43b() { # LU-5690
3313 [[ $(lustre_version_code mgs) -ge $(version_code 2.7.62) ]] ||
3314 { skip "Need MGS version 2.7.62+"; return; }
3316 if [[ -z "$fs2mds_DEV" ]]; then
3317 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
3318 skip_env "mixed loopback and real device not working" && return
3321 local fs2mdsdev=$(mdsdevname 1_2)
3322 local fs2mdsvdev=$(mdsvdevname 1_2)
3324 # temporarily use fs2mds as fs2mgs
3326 local fs2mgsdev=$fs2mdsdev
3327 local fs2mgsvdev=$fs2mdsvdev
3329 local fsname=test1234
3331 load_module llite/lustre
3332 local client_ip=$(host_nids_address $HOSTNAME $NETTYPE)
3333 local host=${client_ip//*./}
3334 local net=${client_ip/%$host/}
3335 local nosquash_nids=$(h2nettype $net[$host,$host,$host])
3337 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
3338 --param mdt.root_squash=$RUNAS_ID:$RUNAS_ID \
3339 --param mdt.nosquash_nids=$nosquash_nids \
3340 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
3341 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS || error "start fs2mgs failed"
3342 stop $fs2mgs -f || error "stop fs2mgs failed"
3343 cleanup || error "cleanup failed with $?"
3345 run_test 43b "parse nosquash_nids with commas in expr_list"
3349 check_mount || error "check_mount"
3350 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
3352 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
3353 for VAL in $UUIDS; do
3354 NID=$(echo $VAL | cut -d= -f1)
3355 CLUUID=$(echo $VAL | cut -d= -f2)
3356 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
3358 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
3359 cleanup || error "cleanup failed with $?"
3361 run_test 44 "mounted client proc entry exists"
3365 check_mount || error "check_mount"
3366 stop_mds || error "Unable to stop MDS"
3370 #define OBD_FAIL_PTLRPC_LONG_REPL_UNLINK 0x50f
3371 do_facet client "$LCTL set_param fail_loc=0x8000050f"
3374 manual_umount_client --force || error "manual_umount_client failed"
3375 do_facet client "$LCTL set_param fail_loc=0x0"
3376 start_mds || error "unable to start MDS"
3377 mount_client $MOUNT || error "mount_client $MOUNT failed"
3378 cleanup || error "cleanup failed with $?"
3380 run_test 45 "long unlink handling in ptlrpcd"
3387 umount_client $MOUNT2 || rc=$?
3388 umount_client $MOUNT || rc=$?
3389 while [ $count -gt 0 ]; do
3390 stop ost${count} -f || rc=$?
3394 cleanup_nocli || rc=$?
3395 #writeconf to remove all ost2 traces for subsequent tests
3396 writeconf_or_reformat
3401 echo "Testing with $OSTCOUNT OSTs"
3403 start_mds || error "unable to start MDS"
3404 #first client should see only one ost
3405 start_ost || error "Unable to start OST1"
3406 wait_osc_import_state mds ost FULL
3408 mount_client $MOUNT || error "mount_client $MOUNT failed"
3409 trap "cleanup_46a $OSTCOUNT" EXIT ERR
3412 for (( i=2; i<=$OSTCOUNT; i++ )); do
3413 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
3414 error "start_ost$i $(ostdevname $i) failed"
3417 # wait until osts in sync
3418 for (( i=2; i<=$OSTCOUNT; i++ )); do
3419 wait_osc_import_state mds ost$i FULL
3420 wait_osc_import_ready client ost$i
3423 #second client see all ost's
3425 mount_client $MOUNT2 || error "mount_client failed"
3426 $SETSTRIPE -c -1 $MOUNT2 || error "$SETSTRIPE -c -1 $MOUNT2 failed"
3427 $GETSTRIPE $MOUNT2 || error "$GETSTRIPE $MOUNT2 failed"
3429 echo "ok" > $MOUNT2/widestripe
3430 $GETSTRIPE $MOUNT2/widestripe ||
3431 error "$GETSTRIPE $MOUNT2/widestripe failed"
3432 # fill acl buffer for avoid expand lsm to them
3433 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd |
3435 setfacl -m $acl $MOUNT2/widestripe
3439 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3441 cleanup_46a $OSTCOUNT || error "cleanup_46a failed"
3443 run_test 46a "handle ost additional - wide striped file"
3448 check_mount || error "check_mount failed"
3449 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
3453 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3454 if echo $ns | grep "MDT[[:digit:]]*"; then
3457 lrs=$(echo $ns | sed 's/.*lru_size=//')
3458 lru_size[count]=$lrs
3463 facet_failover $SINGLEMDS
3464 client_up || error "client_up failed"
3467 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3468 if echo $ns | grep "MDT[[:digit:]]*"; then
3471 lrs=$(echo $ns | sed 's/.*lru_size=//')
3472 if ! test "$lrs" -eq "${lru_size[count]}"; then
3473 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
3474 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
3479 cleanup || error "cleanup failed with $?"
3481 run_test 47 "server restart does not make client loss lru_resize settings"
3486 # reformat after this test is needed - if the test fails,
3487 # we will have unkillable file at FS
3491 test_48() { # bz-17636 LU-7473
3495 check_mount || error "check_mount failed"
3497 $SETSTRIPE -c -1 $MOUNT || error "$SETSTRIPE -c -1 $MOUNT failed"
3498 $GETSTRIPE $MOUNT || error "$GETSTRIPE $MOUNT failed"
3500 echo "ok" > $MOUNT/widestripe
3501 $GETSTRIPE $MOUNT/widestripe ||
3502 error "$GETSTRIPE $MOUNT/widestripe failed"
3504 # In the future, we may introduce more EAs, such as selinux, enlarged
3505 # LOV EA, and so on. These EA will use some EA space that is shared by
3506 # ACL entries. So here we only check some reasonable ACL entries count,
3507 # instead of the max number that is calculated from the max_ea_size.
3508 if [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.8.57) ];
3510 count=28 # hard coded of RPC protocol
3511 elif [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3512 count=4000 # max_num 4091 max_ea_size = 32768
3513 elif ! large_xattr_enabled; then
3514 count=450 # max_num 497 max_ea_size = 4012
3516 count=4500 # max_num 8187 max_ea_size = 1048492
3517 # not create too much (>5000) to save test time
3520 echo "It is expected to hold at least $count ACL entries"
3521 trap cleanup_48 EXIT ERR
3522 for ((i = 0; i < $count; i++)) do
3523 setfacl -m u:$((i + 100)):rw $MOUNT/widestripe ||
3524 error "Fail to setfacl for $MOUNT/widestripe at $i"
3527 cancel_lru_locks mdc
3528 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3529 local r_count=$(getfacl $MOUNT/widestripe | grep "user:" | wc -l)
3530 count=$((count + 1)) # for the entry "user::rw-"
3532 [ $count -eq $r_count ] ||
3533 error "Expected ACL entries $count, but got $r_count"
3537 run_test 48 "too many acls on file"
3539 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
3540 test_49a() { # bug 17710
3541 local timeout_orig=$TIMEOUT
3542 local ldlm_timeout_orig=$LDLM_TIMEOUT
3543 local LOCAL_TIMEOUT=20
3545 LDLM_TIMEOUT=$LOCAL_TIMEOUT
3546 TIMEOUT=$LOCAL_TIMEOUT
3550 check_mount || error "client mount failed"
3552 echo "check ldlm_timout..."
3553 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3554 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3555 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3557 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3558 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3561 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
3562 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
3565 umount_client $MOUNT || error "umount_client $MOUNT failed"
3566 stop_ost || error "problem stopping OSS"
3567 stop_mds || error "problem stopping MDS"
3569 LDLM_TIMEOUT=$ldlm_timeout_orig
3570 TIMEOUT=$timeout_orig
3572 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3574 test_49b() { # bug 17710
3575 local timeout_orig=$TIMEOUT
3576 local ldlm_timeout_orig=$LDLM_TIMEOUT
3577 local LOCAL_TIMEOUT=20
3579 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
3580 TIMEOUT=$LOCAL_TIMEOUT
3584 check_mount || error "client mount failed"
3586 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3587 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3588 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3590 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3591 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3594 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
3595 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
3598 cleanup || error "cleanup failed"
3600 LDLM_TIMEOUT=$ldlm_timeout_orig
3601 TIMEOUT=$timeout_orig
3603 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3606 # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1
3608 # Test both statfs and lfs df and fail if either one fails
3609 multiop_bg_pause $1 f_
3612 killall -USR1 multiop
3613 [ $RC -ne 0 ] && log "lazystatfs multiop failed"
3614 wait $PID || { RC=$?; log "multiop return error "; }
3616 # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1
3621 if kill -s 0 $PID; then
3624 log "lazystatfs lfs df failed to complete in 5s"
3632 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3633 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3635 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
3637 cleanup || error "cleanup failed with rc $?"
3639 run_test 50a "lazystatfs all servers available"
3643 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3644 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3646 # Wait for client to detect down OST
3647 stop_ost || error "Unable to stop OST1"
3648 wait_osc_import_state client ost DISCONN
3650 log "OSCs should all be DISCONN"
3652 lazystatfs $MOUNT || error "lazystatfs should not return EIO"
3654 umount_client $MOUNT || error "Unable to unmount client"
3655 stop_mds || error "Unable to stop MDS"
3657 run_test 50b "lazystatfs all servers down"
3660 start_mds || error "Unable to start MDS"
3661 start_ost || error "Unable to start OST1"
3662 start_ost2 || error "Unable to start OST2"
3663 mount_client $MOUNT || error "Unable to mount client"
3664 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3665 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3667 # Wait for client to detect down OST
3668 stop_ost || error "Unable to stop OST1"
3669 wait_osc_import_state mds ost DISCONN
3670 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3672 umount_client $MOUNT || error "Unable to unmount client"
3673 stop_ost2 || error "Unable to stop OST2"
3674 stop_mds || error "Unable to stop MDS"
3675 #writeconf to remove all ost2 traces for subsequent tests
3676 writeconf_or_reformat
3678 run_test 50c "lazystatfs one server down"
3681 start_mds || error "Unable to start MDS"
3682 start_ost || error "Unable to start OST1"
3683 start_ost2 || error "Unable to start OST2"
3684 mount_client $MOUNT || error "Unable to mount client"
3685 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3686 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3688 # Issue the statfs during the window where the client still
3689 # belives the OST to be available but it is in fact down.
3690 # No failure just a statfs which hangs for a timeout interval.
3691 stop_ost || error "Unable to stop OST1"
3692 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3694 umount_client $MOUNT || error "Unable to unmount client"
3695 stop_ost2 || error "Unable to stop OST2"
3696 stop_mds || error "Unable to stop MDS"
3697 #writeconf to remove all ost2 traces for subsequent tests
3698 writeconf_or_reformat
3700 run_test 50d "lazystatfs client/server conn race"
3707 start_mds || error "Unable to start MDS"
3708 #first client should see only one ost
3709 start_ost || error "Unable to start OST1"
3710 wait_osc_import_state mds ost FULL
3712 # Wait for client to detect down OST
3713 stop_ost || error "Unable to stop OST1"
3714 wait_osc_import_state mds ost DISCONN
3716 mount_client $MOUNT || error "Unable to mount client"
3717 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3719 multiop_bg_pause $MOUNT _f
3723 if [ $RC1 -ne 0 ]; then
3724 log "multiop failed $RC1"
3727 sleep $(( $TIMEOUT+1 ))
3729 [ $? -ne 0 ] && error "process isn't sleep"
3730 start_ost || error "Unable to start OST1"
3731 wait $pid || error "statfs failed"
3734 umount_client $MOUNT || error "Unable to unmount client"
3735 stop_ost || error "Unable to stop OST1"
3736 stop_mds || error "Unable to stop MDS"
3738 run_test 50e "normal statfs all servers down"
3743 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
3745 start_mds || error "Unable to start mds"
3746 #first client should see only one ost
3747 start_ost || error "Unable to start OST1"
3748 wait_osc_import_state mds ost FULL
3750 start_ost2 || error "Unable to start OST2"
3751 wait_osc_import_state mds ost2 FULL
3753 # Wait for client to detect down OST
3754 stop_ost2 || error "Unable to stop OST2"
3756 wait_osc_import_state mds ost2 DISCONN
3757 mount_client $MOUNT || error "Unable to mount client"
3758 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3760 multiop_bg_pause $MOUNT _f
3764 if [ $RC1 -ne 0 ]; then
3765 log "lazystatfs multiop failed $RC1"
3768 sleep $(( $TIMEOUT+1 ))
3770 [ $? -ne 0 ] && error "process isn't sleep"
3771 start_ost2 || error "Unable to start OST2"
3772 wait $pid || error "statfs failed"
3773 stop_ost2 || error "Unable to stop OST2"
3776 umount_client $MOUNT -f || error "Unable to unmount client"
3777 stop_ost || error "Unable to stop OST1"
3778 stop_mds || error "Unable to stop MDS"
3779 #writeconf to remove all ost2 traces for subsequent tests
3780 writeconf_or_reformat
3782 run_test 50f "normal statfs one server in down"
3785 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs" && return
3787 start_ost2 || error "Unable to start OST2"
3788 wait_osc_import_state mds ost2 FULL
3789 wait_osc_import_ready client ost2
3791 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3792 local PARAM="osc.${FSNAME}-OST0001*.active"
3794 local PARAM="${FSNAME}-OST0001.osc.active"
3797 $SETSTRIPE -c -1 $DIR/$tfile || error "$SETSTRIPE failed"
3798 do_facet mgs $PERM_CMD $PARAM=0 || error "Unable to deactivate OST"
3800 umount_client $MOUNT || error "Unable to unmount client"
3801 mount_client $MOUNT || error "Unable to mount client"
3802 # This df should not cause a panic
3805 do_facet mgs $PERM_CMD $PARAM=1 || error "Unable to activate OST"
3806 rm -f $DIR/$tfile || error "unable to remove file $DIR/$tfile"
3807 umount_client $MOUNT || error "Unable to unmount client"
3808 stop_ost2 || error "Unable to stop OST2"
3809 stop_ost || error "Unable to stop OST1"
3810 stop_mds || error "Unable to stop MDS"
3811 #writeconf to remove all ost2 traces for subsequent tests
3812 writeconf_or_reformat
3814 run_test 50g "deactivated OST should not cause panic"
3818 # prepare MDT/OST, make OSC inactive for OST1
3819 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs" && return
3821 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3822 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
3823 error "tunefs OST1 failed"
3824 start_mds || error "Unable to start MDT"
3825 start_ost || error "Unable to start OST1"
3826 start_ost2 || error "Unable to start OST2"
3827 mount_client $MOUNT || error "client start failed"
3829 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3831 # activatate OSC for OST1
3832 set_persistent_param_and_check client \
3833 "osc.${FSNAME}-OST0000-osc-[!M]*.active" \
3834 "${FSNAME}-OST0000.osc.active" 1
3836 mkdir $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3837 $SETSTRIPE -c -1 -i 0 $DIR/$tdir/2 ||
3838 error "$SETSTRIPE $DIR/$tdir/2 failed"
3839 sleep 1 && echo "create a file after OST1 is activated"
3840 # doing some io, shouldn't crash
3841 dd if=/dev/zero of=$DIR/$tdir/2/$tfile-io bs=1M count=10
3843 # check OSC import is working
3844 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
3845 error "some OSC imports are still not connected"
3849 umount_client $MOUNT || error "Unable to umount client"
3850 stop_ost2 || error "Unable to stop OST2"
3851 cleanup_nocli || error "cleanup_nocli failed with $?"
3853 run_test 50h "LU-642: activate deactivated OST"
3856 # prepare MDT/OST, make OSC inactive for OST1
3857 [ "$MDSCOUNT" -lt "2" ] && skip_env "needs >= 2 MDTs" && return
3860 [ $(facet_fstype mds2) == zfs ] && import_zpool mds2
3861 do_facet mds2 "$TUNEFS --param mdc.active=0 $(mdsdevname 2)" ||
3862 error "tunefs MDT2 failed"
3863 start_mds || error "Unable to start MDT"
3864 start_ost || error "Unable to start OST1"
3865 start_ost2 || error "Unable to start OST2"
3866 mount_client $MOUNT || error "client start failed"
3868 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3870 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3871 $PERM_CMD mdc.${FSNAME}-MDT0001-mdc-*.active=0 &&
3872 error "deactive MDC0 succeeds"
3874 $PERM_CMD ${FSNAME}-MDT0000.mdc.active=0 &&
3875 error "deactive MDC0 succeeds"
3878 # activate MDC for MDT2
3879 set_persistent_param_and_check client \
3880 "mdc.${FSNAME}-MDT0001-mdc-*.active" \
3881 "${FSNAME}-MDT0001.mdc.active" 1
3883 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
3884 if [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.60) ]
3886 wait_dne_interconnect
3888 $LFS mkdir -i1 $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3890 createmany -o $DIR/$tdir/2/$tfile-%d 1 || error "create files failed"
3892 rm -rf $DIR/$tdir/2 || error "unlink dir failed"
3894 # deactivate MDC for MDT2
3895 set_persistent_param_and_check client \
3896 "mdc.${FSNAME}-MDT0001-mdc-*.active" \
3897 "${FSNAME}-MDT0001.mdc.active" 0
3899 wait_osp_active mds ${FSNAME}-MDT0001 1 0
3901 $LFS mkdir -i1 $DIR/$tdir/2 &&
3902 error "mkdir $DIR/$tdir/2 succeeds after deactive MDT"
3904 $LFS mkdir -i0 -c$MDSCOUNT $DIR/$tdir/striped_dir ||
3905 error "mkdir $DIR/$tdir/striped_dir fails after deactive MDT2"
3907 local stripe_count=$($LFS getdirstripe -c $DIR/$tdir/striped_dir)
3908 [ $stripe_count -eq $((MDSCOUNT - 1)) ] ||
3909 error "wrong $stripe_count != $((MDSCOUNT -1)) for striped_dir"
3912 umount_client $MOUNT || error "Unable to umount client"
3917 run_test 50i "activate deactivated MDT"
3920 local LOCAL_TIMEOUT=20
3924 check_mount || error "check_mount failed"
3926 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
3927 $SETSTRIPE -c -1 $MOUNT/$tdir ||
3928 error "$SETSTRIPE -c -1 $MOUNT/$tdir failed"
3929 #define OBD_FAIL_MDS_REINT_DELAY 0x142
3930 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x142"
3931 touch $MOUNT/$tdir/$tfile &
3934 start_ost2 || error "Unable to start OST1"
3936 stop_ost2 || error "Unable to stop OST1"
3937 umount_client $MOUNT -f || error “unmount $MOUNT failed”
3938 cleanup_nocli || error “stop server failed”
3939 #writeconf to remove all ost2 traces for subsequent tests
3940 writeconf_or_reformat
3942 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
3951 do_node $node mkdir -p $dest
3952 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3954 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
3955 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3956 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
3958 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
3959 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
3969 local backup2=${TMP}/backup2
3971 do_node $node mkdir -p $backup2
3972 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3974 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
3975 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3976 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
3978 do_node $node "diff -rq $backup $backup2"
3979 [ $? -eq 0 ] || { error "contents differ"; return 3; }
3981 local xattrs2=${TMP}/xattrs2
3982 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
3983 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
3985 do_node $node "diff $xattrs $xattrs2"
3986 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
3988 do_node $node "rm -rf $backup2 $xattrs2"
3989 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
3993 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3994 skip "ldiskfs only test"
3998 start_mds || error "Unable to start MDS"
3999 start_ost || error "Unable to start OST1"
4000 mount_client $MOUNT || error "Unable to mount client"
4003 local ost1mnt=$(facet_mntpt ost1)
4004 local ost1node=$(facet_active_host ost1)
4005 local ost1tmp=$TMP/conf52
4008 mkdir $DIR/$tdir || error "Unable to create $DIR/$tdir"
4009 touch $TMP/modified_first || error "Unable to create temporary file"
4010 local mtime=$(stat -c %Y $TMP/modified_first)
4011 do_node $ost1node "mkdir -p $ost1tmp &&
4012 touch -m -d @$mtime $ost1tmp/modified_first" ||
4013 error "Unable to create temporary file"
4016 $SETSTRIPE -c -1 -S 1M $DIR/$tdir || error "$SETSTRIPE failed"
4018 for (( i=0; i < nrfiles; i++ )); do
4019 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c ||
4020 error "multiop failed"
4025 # sync all the data and make sure no pending data on the client,
4026 # thus the SOM xattr would not be changed any more.
4027 cancel_lru_locks osc
4030 echo backup files to $TMP/$tdir
4031 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
4032 copy_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
4033 error "Unable to copy files"
4035 umount_client $MOUNT || error "Unable to umount client"
4036 stop_ost || error "Unable to stop ost1"
4038 echo mount ost1 as ldiskfs
4039 do_node $ost1node mkdir -p $ost1mnt || error "Unable to create $ost1mnt"
4040 if ! do_node $ost1node test -b $ost1_dev; then
4043 do_node $ost1node mount -t $(facet_fstype ost1) $loop $ost1_dev \
4045 error "Unable to mount ost1 as ldiskfs"
4048 echo backup objects to $ost1tmp/objects
4049 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
4050 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
4051 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
4053 error "Unable to copy objects"
4055 # move objects to lost+found
4056 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
4057 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
4059 do_node $ost1node "umount $ost1mnt" ||
4060 error "Unable to umount ost1 as ldiskfs"
4062 start_ost || error "Unable to start OST1"
4063 mount_client $MOUNT || error "Unable to mount client"
4065 local REPAIRED=$(do_node $ost1node "$LCTL get_param \
4066 -n osd-ldiskfs.$FSNAME-OST0000.oi_scrub" |
4067 awk '/^lf_repa[ri]*ed/ { print $2 }')
4068 [ $REPAIRED -gt 0 ] ||
4069 error "Some entry under /lost+found should be repaired"
4072 diff_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
4073 error "Unable to diff files"
4075 rm -rf $TMP/$tdir $TMP/file_xattrs ||
4076 error "Unable to delete temporary files"
4077 do_node $ost1node "rm -rf $ost1tmp" ||
4078 error "Unable to delete temporary files"
4079 cleanup || error "cleanup failed with $?"
4081 run_test 52 "check recovering objects from lost+found"
4083 # Checks threads_min/max/started for some service
4085 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
4086 # parameter pattern prefix like 'ost.*.ost'.
4099 local msg="Insane $modname thread counts"
4100 local ncpts=$(check_cpt_number $facet)
4104 check_mount || return 41
4106 # We need to expand $parampat, but it may match multiple parameters, so
4107 # we'll pick the first one
4108 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
4109 error "Couldn't expand ${parampat}.threads_min parameter name"
4113 # Remove the .threads_min part
4114 paramp=${paramp%.threads_min}
4116 # Check for sanity in defaults
4117 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4119 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4121 tstarted=$(do_facet $facet "$LCTL get_param \
4122 -n ${paramp}.threads_started" || echo 0)
4123 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' ||
4125 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4127 nthrs=$(expr $tmax - $tmin)
4128 if [ $nthrs -lt $ncpts ]; then
4134 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
4135 skip_env "module parameter forced $facet thread count" &&
4136 tmin=3 && tmax=$((3 * tmax))
4138 # Check that we can change min/max
4139 do_facet $facet "$LCTL set_param \
4140 ${paramp}.threads_min=$((tmin + nthrs))"
4141 do_facet $facet "$LCTL set_param \
4142 ${paramp}.threads_max=$((tmax - nthrs))"
4143 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4145 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4147 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) &&
4148 $tmax2 == ($tmax - $nthrs)))' || return $?
4150 # Check that we can set min/max to the same value
4151 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4153 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$tmin"
4154 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4156 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4158 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
4160 # Check that we can't set max < min
4161 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$((tmin - 1))"
4162 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4164 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4166 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
4168 # We need to ensure that we get the module options desired; to do this
4169 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
4170 LOAD_MODULES_REMOTE=true
4173 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
4174 setmodopts -a $modname "$newvalue" oldvalue
4177 check_mount || return 41
4179 # Restore previous setting of MODOPTS_*
4180 setmodopts $modname "$oldvalue"
4182 # Check that $opts took
4183 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4185 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4187 tstarted=$(do_facet $facet \
4188 "$LCTL get_param -n ${paramp}.threads_started" || echo 0)
4189 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4198 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
4199 cleanup || error "cleanup failed with rc $?"
4201 run_test 53a "check OSS thread count params"
4205 local mds=$(do_facet $SINGLEMDS "$LCTL get_param \
4206 -N mds.*.*.threads_max 2>/dev/null")
4207 if [ -z "$mds" ]; then
4208 #running this on an old MDT
4209 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
4211 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
4213 cleanup || error "cleanup failed with $?"
4215 run_test 53b "check MDS thread count params"
4218 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4219 skip "ldiskfs only test"
4223 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p ||
4224 error "llverdev failed with rc=$?"
4227 run_test 54a "test llverdev and partial verify of device"
4230 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4231 skip "ldiskfs only test"
4236 run_llverfs $MOUNT -p || error "llverfs failed with rc=$?"
4237 cleanup || error "cleanup failed with rc=$?"
4239 run_test 54b "test llverfs and partial verify of filesystem"
4243 local max_ost_index=$1
4244 echo -n $(((max_ost_index + 1) * 8))
4248 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4249 skip "ldiskfs only test"
4253 local mdsdev=$(mdsdevname 1)
4254 local mdsvdev=$(mdsvdevname 1)
4258 if ! combined_mgs_mds; then
4259 stop_mgs || error "stopping MGS service failed"
4260 format_mgs || error "formatting MGT failed"
4262 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
4264 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
4265 --reformat $(ostdevname 1) $(ostvdevname 1)
4271 echo checking size of lov_objid for ost index $i
4272 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" |
4273 grep ^User | awk -F 'Size: ' '{print $2}')
4274 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
4275 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
4277 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
4284 run_test 55 "check lov_objid size"
4287 local server_version=$(lustre_version_code $SINGLEMDS)
4288 local mds_journal_size_orig=$MDSJOURNALSIZE
4294 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
4295 $(ostdevname 1) $(ostvdevname 1)
4296 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
4297 $(ostdevname 2) $(ostvdevname 2)
4300 start_ost || error "Unable to start first ost (idx 10000)"
4301 start_ost2 || error "Unable to start second ost (idx 1000)"
4302 mount_client $MOUNT || error "Unable to mount client"
4306 if [[ $server_version -ge $(version_code 2.6.54) ]] ||
4307 [[ $server_version -ge $(version_code 2.5.4) &&
4308 $server_version -lt $(version_code 2.5.11) ]]; then
4309 wait_osc_import_state mds ost1 FULL
4310 wait_osc_import_state mds ost2 FULL
4311 $SETSTRIPE --stripe-count=-1 $DIR/$tfile ||
4312 error "Unable to setstripe $DIR/$tfile"
4313 n=$($LFS getstripe --stripe-count $DIR/$tfile)
4314 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
4319 MDSJOURNALSIZE=$mds_journal_size_orig
4322 run_test 56a "check big OST indexes and out-of-index-order start"
4327 umount_client $MOUNT -f || error "unmount client failed"
4336 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs" && return
4338 trap cleanup_56b EXIT RETURN ERR
4341 if ! combined_mgs_mds ; then
4346 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) --index=0 --reformat \
4347 $(mdsdevname 1) $(mdsvdevname 1)
4348 add mds2 $(mkfs_opts mds2 $(mdsdevname 2)) --index=1 --reformat \
4349 $(mdsdevname 2) $(mdsvdevname 2)
4350 add mds3 $(mkfs_opts mds3 $(mdsdevname 3)) --index=1000 --reformat \
4351 $(mdsdevname 3) $(mdsvdevname 3)
4355 start_mdt 1 || error "MDT 1 (idx 0) start failed"
4356 start_mdt 2 || error "MDT 2 (idx 1) start failed"
4357 start_mdt 3 || error "MDT 3 (idx 1000) start failed"
4358 start_ost || error "Unable to start first ost"
4359 start_ost2 || error "Unable to start second ost"
4361 do_nodes $(comma_list $(mdts_nodes)) \
4362 "$LCTL set_param mdt.*.enable_remote_dir=1 \
4363 mdt.*.enable_remote_dir_gid=-1"
4365 mount_client $MOUNT || error "Unable to mount client"
4367 $LFS mkdir -c3 $MOUNT/$tdir || error "failed to make testdir"
4369 echo "This is test file 1!" > $MOUNT/$tdir/$tfile.1 ||
4370 error "failed to make test file 1"
4371 echo "This is test file 2!" > $MOUNT/$tdir/$tfile.2 ||
4372 error "failed to make test file 2"
4373 echo "This is test file 1000!" > $MOUNT/$tdir/$tfile.1000 ||
4374 error "failed to make test file 1000"
4376 rm -rf $MOUNT/$tdir || error "failed to remove testdir"
4378 $LFS mkdir -i1000 $MOUNT/$tdir.1000 ||
4379 error "create remote dir at idx 1000 failed"
4382 echo "=== START lfs df OUTPUT ==="
4384 echo "==== END lfs df OUTPUT ===="
4386 mdtcnt=$(echo -e "$output" | grep $FSNAME-MDT | wc -l)
4387 ostcnt=$(echo -e "$output" | grep $FSNAME-OST | wc -l)
4389 echo "lfs df returned mdt count $mdtcnt and ost count $ostcnt"
4390 [ $mdtcnt -eq 3 ] || error "lfs df returned wrong mdt count"
4391 [ $ostcnt -eq 2 ] || error "lfs df returned wrong ost count"
4393 echo "This is test file 1!" > $MOUNT/$tdir.1000/$tfile.1 ||
4394 error "failed to make test file 1"
4395 echo "This is test file 2!" > $MOUNT/$tdir.1000/$tfile.2 ||
4396 error "failed to make test file 2"
4397 echo "This is test file 1000!" > $MOUNT/$tdir.1000/$tfile.1000 ||
4398 error "failed to make test file 1000"
4399 rm -rf $MOUNT/$tdir.1000 || error "failed to remove remote_dir"
4402 echo "=== START lfs mdts OUTPUT ==="
4404 echo "==== END lfs mdts OUTPUT ===="
4406 echo -e "$output" | grep -v "MDTS:" | awk '{print $1}' |
4407 sed 's/://g' > $TMP/mdts-actual.txt
4408 sort $TMP/mdts-actual.txt -o $TMP/mdts-actual.txt
4410 echo -e "0\n1\n1000" > $TMP/mdts-expected.txt
4412 diff $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4415 rm $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4417 [ $result -eq 0 ] || error "target_obd proc file is incorrect!"
4419 run_test 56b "test target_obd correctness with nonconsecutive MDTs"
4421 test_57a() { # bug 22656
4422 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4423 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4424 tail -1 | awk '{print $1}')
4425 writeconf_or_reformat
4426 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4427 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" ||
4428 error "tunefs failed"
4430 start_ost && error "OST registration from failnode should fail"
4433 run_test 57a "initial registration from failnode should fail (should return errs)"
4436 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4437 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4438 tail -1 | awk '{print $1}')
4439 writeconf_or_reformat
4440 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4441 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" ||
4442 error "tunefs failed"
4444 start_ost || error "OST registration from servicenode should not fail"
4447 run_test 57b "initial registration from servicenode should not fail"
4450 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
4453 test_58() { # bug 22658
4454 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
4456 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4457 createmany -o $DIR/$tdir/$tfile-%d 100
4458 unlinkmany $DIR/$tdir/$tfile-%d 100
4459 stop_mds || error "Unable to stop MDS"
4461 local MNTDIR=$(facet_mntpt $SINGLEMDS)
4462 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4464 # remove all files from the OBJECTS dir
4465 mount_fstype $SINGLEMDS
4467 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
4469 unmount_fstype $SINGLEMDS
4470 # restart MDS with missing llog files
4471 start_mds || error "unable to start MDS"
4472 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0"
4475 run_test 58 "missing llog files must not prevent MDT from mounting"
4478 start_mgsmds >> /dev/null
4479 local C1=$(count_osts)
4480 if [ $C1 -eq 0 ]; then
4481 start_ost >> /dev/null
4485 echo "original ost count: $C1 (expect > 0)"
4486 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
4487 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
4488 local C2=$(count_osts)
4489 echo "after mdt writeconf count: $C2 (expect 0)"
4490 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
4491 echo "OST start without writeconf should fail:"
4492 start_ost >> /dev/null &&
4493 error "OST start without writeconf didn't fail"
4494 echo "OST start with writeconf should succeed:"
4495 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
4496 local C3=$(count_osts)
4497 echo "after ost writeconf count: $C3 (expect 1)"
4498 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
4499 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
4500 local C4=$(count_osts)
4501 echo "after ost2 writeconf count: $C4 (expect 2)"
4502 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
4503 stop_ost2 >> /dev/null
4504 cleanup_nocli >> /dev/null
4505 #writeconf to remove all ost2 traces for subsequent tests
4506 writeconf_or_reformat
4508 run_test 59 "writeconf mount option"
4510 test_60() { # LU-471
4513 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4514 skip "ldiskfs only test"
4518 for num in $(seq $MDSCOUNT); do
4519 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
4520 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
4521 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4525 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
4526 [ ${PIPESTATUS[0]} -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
4528 # MDT default has dirdata feature
4529 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
4530 # we disable uninit_bg feature
4531 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
4532 # we set stride extended options
4533 echo $dump | grep stride > /dev/null || error "stride is not set"
4536 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
4541 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] ||
4542 { skip "Need MDS version at least 2.1.53"; return 0; }
4544 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
4545 ! large_xattr_enabled; then
4548 for num in $(seq $MDSCOUNT); do
4549 do_facet mds${num} $TUNE2FS -O ea_inode \
4550 $(mdsdevname $num) ||
4551 error "tune2fs on mds $num failed"
4555 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
4556 setup_noconfig || error "setting up the filesystem failed"
4557 client_up || error "starting client failed"
4559 local file=$DIR/$tfile
4560 touch $file || error "touch $file failed"
4562 local large_value="$(generate_string $(max_xattr_size))"
4563 local small_value="bar"
4565 local name="trusted.big"
4566 log "save large xattr $name on $file"
4567 setfattr -n $name -v $large_value $file ||
4568 error "saving $name on $file failed"
4570 local new_value=$(get_xattr_value $name $file)
4571 [[ "$new_value" != "$large_value" ]] &&
4572 error "$name different after saving"
4574 log "shrink value of $name on $file"
4575 setfattr -n $name -v $small_value $file ||
4576 error "shrinking value of $name on $file failed"
4578 new_value=$(get_xattr_value $name $file)
4579 [[ "$new_value" != "$small_value" ]] &&
4580 error "$name different after shrinking"
4582 log "grow value of $name on $file"
4583 setfattr -n $name -v $large_value $file ||
4584 error "growing value of $name on $file failed"
4586 new_value=$(get_xattr_value $name $file)
4587 [[ "$new_value" != "$large_value" ]] &&
4588 error "$name different after growing"
4590 log "check value of $name on $file after remounting MDS"
4592 new_value=$(get_xattr_value $name $file)
4593 [[ "$new_value" != "$large_value" ]] &&
4594 error "$name different after remounting MDS"
4596 log "remove large xattr $name from $file"
4597 setfattr -x $name $file || error "removing $name from $file failed"
4600 stopall || error "stopping for e2fsck run"
4601 for num in $(seq $MDSCOUNT); do
4602 run_e2fsck $(facet_active_host mds$num) \
4603 $(mdsdevname $num) "-y" ||
4604 error "e2fsck MDT$num failed"
4606 setup_noconfig || error "remounting the filesystem failed"
4609 # need to delete this file to avoid problems in other tests
4611 stopall || error "stopping systems failed"
4613 run_test 61 "large xattr"
4616 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4617 skip "ldiskfs only test"
4622 local mdsdev=$(mdsdevname 1)
4623 local ostdev=$(ostdevname 1)
4625 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] ||
4626 { skip "Need MDS version at least 2.2.51"; return 0; }
4628 echo "disable journal for mds"
4629 do_facet mds1 $TUNE2FS -O ^has_journal $mdsdev || error "tune2fs failed"
4630 start_mds && error "MDT start should fail"
4631 echo "disable journal for ost"
4632 do_facet ost1 $TUNE2FS -O ^has_journal $ostdev || error "tune2fs failed"
4633 start_ost && error "OST start should fail"
4634 cleanup || error "cleanup failed with rc $?"
4637 run_test 62 "start with disabled journal"
4640 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4641 skip "ldiskfs only test"
4645 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ldiskfs
4646 local inode_slab=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
4647 awk '/ldiskfs_inode_cache/ { print $5 / $6 }')
4648 if [ -z "$inode_slab" ]; then
4649 skip "ldiskfs module has not been loaded"
4653 echo "$inode_slab ldiskfs inodes per page"
4654 [ "${inode_slab%.*}" -ge "3" ] && return 0
4656 # If kmalloc-128 is also 1 per page - this is a debug kernel
4657 # and so this is not an error.
4658 local kmalloc128=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
4659 awk '/^(kmalloc|size)-128 / { print $5 / $6 }')
4660 # 32 128-byte chunks in 4k
4661 [ "${kmalloc128%.*}" -lt "32" ] ||
4662 error "ldiskfs inode too big, only $inode_slab objs/page, " \
4663 "kmalloc128 = $kmalloc128 objs/page"
4665 run_test 63 "Verify each page can at least hold 3 ldiskfs inodes"
4668 start_mds || error "unable to start MDS"
4669 start_ost || error "Unable to start OST1"
4670 start_ost2 || error "Unable to start second ost"
4671 mount_client $MOUNT || error "Unable to mount client"
4672 stop_ost2 || error "Unable to stop second ost"
4675 umount_client $MOUNT -f || error “unmount $MOUNT failed”
4676 cleanup_nocli || error "cleanup_nocli failed with $?"
4677 #writeconf to remove all ost2 traces for subsequent tests
4678 writeconf_or_reformat
4680 run_test 64 "check lfs df --lazy "
4682 test_65() { # LU-2237
4683 # Currently, the test is only valid for ldiskfs backend
4684 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4685 skip "ldiskfs only test" && return
4687 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4688 local brpt=$(facet_mntpt brpt)
4691 if ! do_facet $SINGLEMDS "test -b $devname"; then
4695 stop_mds || error "Unable to stop MDS"
4696 local obj=$(do_facet $SINGLEMDS \
4697 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
4699 if [ -z "$obj" ]; then
4700 # The MDT may be just re-formatted, mount the MDT for the
4701 # first time to guarantee the "last_rcvd" file is there.
4702 start_mds || error "fail to mount the MDS for the first time"
4703 stop_mds || error "Unable to stop MDS"
4706 # remove the "last_rcvd" file
4707 do_facet $SINGLEMDS "mkdir -p $brpt"
4708 do_facet $SINGLEMDS \
4709 "mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
4710 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
4711 do_facet $SINGLEMDS "$UMOUNT $brpt"
4713 # restart MDS, the "last_rcvd" file should be recreated.
4714 start_mds || error "fail to restart the MDS"
4715 stop_mds || error "Unable to stop MDS"
4716 obj=$(do_facet $SINGLEMDS \
4717 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
4718 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
4720 run_test 65 "re-create the lost last_rcvd file when server mount"
4723 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
4724 { skip "Need MGS version at least 2.3.59"; return 0; }
4727 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
4728 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
4730 # add EXCLUDE records to config log, they are not to be
4731 # removed by lctl replace_nids
4732 set_conf_param_and_check mds \
4733 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
4734 "$FSNAME-OST0000.osc.active" \
4737 echo "replace_nids should fail if MDS, OSTs and clients are UP"
4738 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4739 error "replace_nids fail"
4741 umount_client $MOUNT || error "unmounting client failed"
4742 echo "replace_nids should fail if MDS and OSTs are UP"
4743 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4744 error "replace_nids fail"
4746 stop_ost || error "Unable to stop OST1"
4747 echo "replace_nids should fail if MDS is UP"
4748 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4749 error "replace_nids fail"
4751 stop_mds || error "stopping mds failed"
4753 if combined_mgs_mds; then
4754 start_mdt 1 "-o nosvc" ||
4755 error "starting mds with nosvc option failed"
4758 echo "command should accept two parameters"
4759 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
4760 error "command should accept two params"
4762 echo "correct device name should be passed"
4763 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
4764 error "wrong devname"
4766 echo "wrong nids list should not destroy the system"
4767 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
4769 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "asdfasdf, asdfadf" &&
4772 echo "replace OST nid"
4773 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
4774 error "replace nids failed"
4776 echo "command should accept two parameters"
4777 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
4778 error "command should accept two params"
4780 echo "wrong nids list should not destroy the system"
4781 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
4784 local FAKE_NIDS="192.168.0.112@tcp1,192.168.0.112@tcp2"
4785 local FAKE_FAILOVER="192.168.0.113@tcp1,192.168.0.113@tcp2"
4786 local NIDS_AND_FAILOVER="$MDS_NID,$FAKE_NIDS:$FAKE_FAILOVER"
4787 echo "set NIDs with failover"
4788 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $NIDS_AND_FAILOVER ||
4789 error "replace nids failed"
4792 echo "replace MDS nid"
4793 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
4794 error "replace nids failed"
4796 if ! combined_mgs_mds ; then
4799 stop_mds || error "Unable to stop MDS"
4802 start_mgsmds || error "start mgsmds failed"
4803 set_conf_param_and_check mds \
4804 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
4805 "$FSNAME-OST0000.osc.active" \
4807 start_ost || error "unable to start OST"
4808 mount_client $MOUNT || error "mount client failed"
4810 check_mount || error "error after nid replace"
4811 cleanup || error "cleanup failed"
4814 run_test 66 "replace nids"
4816 test_67() { #LU-2950
4817 local legacy="$TMP/legacy_lnet_config"
4818 local new="$TMP/new_routes_test"
4819 local out="$TMP/config_out_file"
4820 local verify="$TMP/conv_verify"
4821 local verify_conf="$TMP/conf_verify"
4823 # Create the legacy file that will be run through the
4824 # lustre_routes_conversion script
4825 cat <<- LEGACY_LNET_CONFIG > $legacy
4826 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
4827 tcp2 54 10.1.3.2@tcp;
4828 tcp3 10.3.4.3@tcp:3;
4832 # Create the verification file to verify the output of
4833 # lustre_routes_conversion script against.
4834 cat <<- VERIFY_LNET_CONFIG > $verify
4835 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
4836 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
4837 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
4838 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
4839 tcp4: { gateway: 10.3.3.4@tcp }
4842 # Create the verification file to verify the output of
4843 # lustre_routes_config script against
4844 cat <<- VERIFY_LNET_CONFIG > $verify_conf
4845 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
4846 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
4847 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
4848 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
4849 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
4852 $LUSTRE_ROUTES_CONVERSION $legacy $new > /dev/null
4853 if [ -f $new ]; then
4854 # verify the conversion output
4855 cmp -s $new $verify > /dev/null
4856 if [ $? -eq 1 ]; then
4857 error "routes conversion failed"
4860 lustre_routes_config --dry-run --verbose $new > $out
4861 # check that the script succeeded
4862 cmp -s $out $verify_conf > /dev/null
4863 if [ $? -eq 1 ]; then
4864 error "routes config failed"
4867 error "routes conversion test failed"
4869 # remove generated files
4870 rm -f $new $legacy $verify $verify_conf $out
4872 run_test 67 "test routes conversion and configuration"
4880 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.4.53) ] ||
4881 { skip "Need MDS version at least 2.4.53"; return 0; }
4883 umount_client $MOUNT || error "umount client failed"
4885 if ! combined_mgs_mds; then
4886 start_mgs || error "start mgs failed"
4889 start_mdt 1 || error "MDT start failed"
4890 start_ost || error "Unable to start OST1"
4892 # START-END - the sequences we'll be reserving
4893 START=$(do_facet $SINGLEMDS \
4894 $LCTL get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
4895 END=$((START + (1 << 30)))
4896 do_facet $SINGLEMDS \
4897 $LCTL set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
4899 # reset the sequences MDT0000 has already assigned
4900 do_facet $SINGLEMDS \
4901 $LCTL set_param seq.srv*MDT0000.space=clear
4903 # remount to let the client allocate new sequence
4904 mount_client $MOUNT || error "mount client failed"
4906 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4907 do_facet $SINGLEMDS \
4908 $LCTL get_param seq.srv*MDT0000.space
4909 $LFS path2fid $DIR/$tfile
4911 local old_ifs="$IFS"
4913 fid=($($LFS path2fid $DIR/$tfile))
4917 if [[ $seq < $END ]]; then
4918 error "used reserved sequence $seq?"
4920 cleanup || error "cleanup failed with $?"
4922 run_test 68 "be able to reserve specific sequences in FLDB"
4924 # Test 69: is about the total number of objects ever created on an OST.
4925 # so that when it is reformatted the normal MDS->OST orphan recovery won't
4926 # just "precreate" the missing objects. In the past it might try to recreate
4927 # millions of objects after an OST was reformatted
4929 local server_version=$(lustre_version_code $SINGLEMDS)
4931 [[ $server_version -lt $(version_code 2.4.2) ]] &&
4932 skip "Need MDS version at least 2.4.2" && return
4934 [[ $server_version -ge $(version_code 2.4.50) ]] &&
4935 [[ $server_version -lt $(version_code 2.5.0) ]] &&
4936 skip "Need MDS version at least 2.5.0" && return
4939 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4941 # use OST0000 since it probably has the most creations
4942 local OSTNAME=$(ostname_from_index 0)
4943 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
4944 local last_id=$(do_facet mds1 $LCTL get_param -n \
4945 osc.$mdtosc_proc1.prealloc_last_id)
4947 # Want to have OST LAST_ID over 5 * OST_MAX_PRECREATE to
4948 # verify that the LAST_ID recovery is working properly. If
4949 # not, then the OST will refuse to allow the MDS connect
4950 # because the LAST_ID value is too different from the MDS
4951 #define OST_MAX_PRECREATE=20000
4952 local ost_max_pre=20000
4953 local num_create=$(( ost_max_pre * 5 + 1 - last_id))
4955 # If the LAST_ID is already over 5 * OST_MAX_PRECREATE, we don't
4956 # need to create any files. So, skip this section.
4957 if [ $num_create -gt 0 ]; then
4958 # Check the number of inodes available on OST0
4960 local ifree=$($LFS df -i $MOUNT | awk '/OST0000/ { print $4 }')
4961 log "On OST0, $ifree inodes available. Want $num_create."
4963 $SETSTRIPE -i 0 $DIR/$tdir ||
4964 error "$SETSTRIPE -i 0 $DIR/$tdir failed"
4965 if [ $ifree -lt 10000 ]; then
4966 files=$(( ifree - 50 ))
4971 local j=$((num_create / files + 1))
4972 for i in $(seq 1 $j); do
4973 createmany -o $DIR/$tdir/$tfile-$i- $files ||
4974 error "createmany fail create $files files: $?"
4975 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
4976 error "unlinkmany failed unlink $files files"
4980 # delete all of the files with objects on OST0 so the
4981 # filesystem is not inconsistent later on
4982 $LFS find $MOUNT --ost 0 -print0 | xargs -0 rm
4984 umount_client $MOUNT || error "umount client failed"
4985 stop_ost || error "OST0 stop failure"
4986 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
4987 $(ostdevname 1) $(ostvdevname 1) ||
4988 error "reformat and replace $ostdev failed"
4989 start_ost || error "OST0 restart failure"
4990 wait_osc_import_state mds ost FULL
4992 mount_client $MOUNT || error "mount client failed"
4993 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
4994 local idx=$($GETSTRIPE -i $DIR/$tdir/$tfile-last)
4995 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
4997 local iused=$($LFS df -i $MOUNT | awk '/OST0000/ { print $3 }')
4998 log "On OST0, $iused used inodes"
4999 [ $iused -ge $((ost_max_pre/2 + 1000)) ] &&
5000 error "OST replacement created too many inodes; $iused"
5001 cleanup || error "cleanup failed with $?"
5003 run_test 69 "replace an OST with the same index"
5006 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5009 cleanup || error "cleanup failed with $?"
5011 start_mdt 1 || error "MDT0 start fail"
5013 start_ost || error "OST0 start fail"
5014 for num in $(seq 2 $MDSCOUNT); do
5015 start_mdt $num || return
5018 mount_client $MOUNT || error "mount client fails"
5020 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
5022 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5023 error "create remote dir fail"
5025 rm -rf $DIR/$tdir || error "delete dir fail"
5026 cleanup || error "cleanup failed with $?"
5028 run_test 70a "start MDT0, then OST, then MDT1"
5031 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5034 start_ost || error "OST0 start fail"
5036 start_mds || error "MDS start fail"
5038 mount_client $MOUNT || error "mount client fails"
5040 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
5042 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5043 error "create remote dir fail"
5045 rm -rf $DIR/$tdir || error "delete dir fail"
5047 cleanup || error "cleanup failed with $?"
5049 run_test 70b "start OST, MDT1, MDT0"
5052 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5055 start_mds || error "MDS start fail"
5056 start_ost || error "OST0 start fail"
5058 mount_client $MOUNT || error "mount client fails"
5059 stop_mdt 1 || error "MDT1 start fail"
5061 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
5062 echo "deactivate $mdc_for_mdt1"
5063 $LCTL --device $mdc_for_mdt1 deactivate ||
5064 error "set $mdc_for_mdt1 deactivate failed"
5066 mkdir $DIR/$tdir && error "mkdir succeed"
5068 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
5069 error "create remote dir succeed"
5071 cleanup || error "cleanup failed with $?"
5073 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
5076 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5079 start_mds || error "MDS start fail"
5080 start_ost || error "OST0 start fail"
5082 mount_client $MOUNT || error "mount client fails"
5084 stop_mdt 2 || error "MDT1 start fail"
5086 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
5088 echo "deactivate $mdc_for_mdt2"
5089 $LCTL --device $mdc_for_mdt2 deactivate ||
5090 error "set $mdc_for_mdt2 deactivate failed"
5092 mkdir $DIR/$tdir || error "mkdir fail"
5093 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
5094 error "create remote dir succeed"
5096 rm -rf $DIR/$tdir || error "delete dir fail"
5098 cleanup || error "cleanup failed with $?"
5100 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
5103 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5105 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.62) ] ||
5106 { skip "Need MDS version at least 2.7.62"; return 0; }
5108 cleanup || error "cleanup failed with $?"
5110 local mdsdev=$(mdsdevname 1)
5111 local ostdev=$(ostdevname 1)
5112 local mdsvdev=$(mdsvdevname 1)
5113 local ostvdev=$(ostvdevname 1)
5114 local opts_mds="$(mkfs_opts mds1 $mdsdev) --reformat $mdsdev $mdsvdev"
5115 local opts_ost="$(mkfs_opts ost1 $ostdev) --reformat $ostdev $ostvdev"
5117 add mds1 $opts_mds || error "add mds1 failed"
5118 start_mdt 1 || error "start mdt1 failed"
5119 add ost1 $opts_ost || error "add ost1 failed"
5120 start_ost || error "start ost failed"
5121 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
5123 local soc=$(do_facet mds1 "$LCTL get_param -n \
5124 mdt.*MDT0000.sync_lock_cancel")
5125 [ $soc == "never" ] || error "SoC enabled on single MDS"
5127 for i in $(seq 2 $MDSCOUNT); do
5128 mdsdev=$(mdsdevname $i)
5129 mdsvdev=$(mdsvdevname $i)
5130 opts_mds="$(mkfs_opts mds$i $mdsdev) --reformat $mdsdev \
5132 add mds$i $opts_mds || error "add mds$i failed"
5133 start_mdt $i || error "start mdt$i fail"
5136 wait_dne_interconnect
5138 for i in $(seq $MDSCOUNT); do
5139 soc=$(do_facet mds$i "$LCTL get_param -n \
5140 mdt.*MDT000$((i - 1)).sync_lock_cancel")
5141 [ $soc == "blocking" ] || error "SoC not enabled on DNE"
5144 for i in $(seq 2 $MDSCOUNT); do
5145 stop_mdt $i || error "stop mdt$i fail"
5147 soc=$(do_facet mds1 "$LCTL get_param -n \
5148 mdt.*MDT0000.sync_lock_cancel")
5149 [ $soc == "never" ] || error "SoC enabled on single MDS"
5150 umount_client $MOUNT -f > /dev/null
5152 cleanup || error "cleanup failed with $?"
5154 run_test 70e "Sync-on-Cancel will be enabled by default on DNE"
5157 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5158 if combined_mgs_mds; then
5159 skip "needs separate MGS/MDT" && return
5163 start_mdt 1 || error "MDT0 start fail"
5164 start_ost || error "OST0 start fail"
5165 for num in $(seq 2 $MDSCOUNT); do
5166 start_mdt $num || return
5169 start_ost2 || error "OST1 start fail"
5171 mount_client $MOUNT || error "mount client fails"
5173 mkdir $DIR/$tdir || error "mkdir fail"
5174 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5175 error "create remote dir succeed"
5177 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5178 rm -rf $DIR/$tdir || error "delete dir fail"
5180 umount_client $MOUNT || error "umount_client failed"
5181 stop_mds || error "MDS stop fail"
5182 stop_ost || error "OST0 stop fail"
5183 stop_ost2 || error "OST1 stop fail"
5185 run_test 71a "start MDT0 OST0, MDT1, OST1"
5188 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5189 if combined_mgs_mds; then
5190 skip "needs separate MGS/MDT" && return
5194 for num in $(seq 2 $MDSCOUNT); do
5195 start_mdt $num || return
5197 start_ost || error "OST0 start fail"
5198 start_mdt 1 || error "MDT0 start fail"
5199 start_ost2 || error "OST1 start fail"
5201 mount_client $MOUNT || error "mount client fails"
5203 mkdir $DIR/$tdir || error "mkdir fail"
5204 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5205 error "create remote dir succeed"
5207 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5208 rm -rf $DIR/$tdir || error "delete dir fail"
5210 umount_client $MOUNT || error "umount_client failed"
5211 stop_mds || error "MDT0 stop fail"
5212 stop_ost || error "OST0 stop fail"
5213 stop_ost2 || error "OST1 stop fail"
5215 run_test 71b "start MDT1, OST0, MDT0, OST1"
5218 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5219 if combined_mgs_mds; then
5220 skip "needs separate MGS/MDT" && return
5224 start_ost || error "OST0 start fail"
5225 start_ost2 || error "OST1 start fail"
5226 for num in $(seq 2 $MDSCOUNT); do
5227 start_mdt $num || return
5229 start_mdt 1 || error "MDT0 start fail"
5231 mount_client $MOUNT || error "mount client fails"
5233 mkdir $DIR/$tdir || error "mkdir fail"
5234 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5235 error "create remote dir succeed"
5237 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5238 rm -rf $DIR/$tdir || error "delete dir fail"
5240 umount_client $MOUNT || error "umount_client failed"
5241 stop_mds || error "MDS stop fail"
5242 stop_ost || error "OST0 stop fail"
5243 stop_ost2 || error "OST1 stop fail"
5246 run_test 71c "start OST0, OST1, MDT1, MDT0"
5249 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5250 if combined_mgs_mds; then
5251 skip "needs separate MGS/MDT" && return
5255 start_ost || error "OST0 start fail"
5256 for num in $(seq 2 $MDSCOUNT); do
5257 start_mdt $num || return
5259 start_mdt 1 || error "MDT0 start fail"
5260 start_ost2 || error "OST1 start fail"
5262 mount_client $MOUNT || error "mount client fails"
5264 mkdir $DIR/$tdir || error "mkdir fail"
5265 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5266 error "create remote dir succeed"
5268 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5269 rm -rf $DIR/$tdir || error "delete dir fail"
5271 umount_client $MOUNT || error "umount_client failed"
5272 stop_mds || error "MDS stop fail"
5273 stop_ost || error "OST0 stop fail"
5274 stop_ost2 || error "OST1 stop fail"
5277 run_test 71d "start OST0, MDT1, MDT0, OST1"
5280 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5281 if combined_mgs_mds; then
5282 skip "needs separate MGS/MDT" && return
5286 start_ost || error "OST0 start fail"
5287 for num in $(seq 2 $MDSCOUNT); do
5288 start_mdt $num || return
5290 start_ost2 || error "OST1 start fail"
5291 start_mdt 1 || error "MDT0 start fail"
5293 mount_client $MOUNT || error "mount client fails"
5295 mkdir $DIR/$tdir || error "mkdir fail"
5296 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5297 error "create remote dir succeed"
5299 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5300 rm -rf $DIR/$tdir || error "delete dir fail"
5302 umount_client $MOUNT || error "umount_client failed"
5303 stop_mds || error "MDS stop fail"
5304 stop_ost || error "OST0 stop fail"
5305 stop_ost2 || error "OST1 stop fail"
5308 run_test 71e "start OST0, MDT1, OST1, MDT0"
5310 test_72() { #LU-2634
5311 local mdsdev=$(mdsdevname 1)
5312 local ostdev=$(ostdevname 1)
5313 local cmd="$E2FSCK -fnvd $mdsdev"
5317 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
5318 skip "ldiskfs only test" && return
5320 if combined_mgs_mds; then
5321 add_options='--reformat'
5323 add_options='--reformat --replace'
5326 #tune MDT with "-O extents"
5328 for num in $(seq $MDSCOUNT); do
5329 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
5330 $add_options $(mdsdevname $num) $(mdsvdevname $num) ||
5331 error "add mds $num failed"
5332 do_facet mds${num} "$TUNE2FS -O extents $(mdsdevname $num)" ||
5333 error "$TUNE2FS failed on mds${num}"
5336 add ost1 $(mkfs_opts ost1 $ostdev) $add_options $ostdev ||
5337 error "add $ostdev failed"
5338 start_mds || error "start mds failed"
5339 start_ost || error "start ost failed"
5340 mount_client $MOUNT || error "mount client failed"
5342 #create some short symlinks
5343 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5344 createmany -o $DIR/$tdir/$tfile-%d $fn
5345 echo "create $fn short symlinks"
5346 for i in $(seq -w 1 $fn); do
5347 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
5352 umount_client $MOUNT || error "umount client failed"
5353 stop_mds || error "stop mds failed"
5354 stop_ost || error "stop ost failed"
5357 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
5359 run_test 72 "test fast symlink with extents flag enabled"
5361 test_73() { #LU-3006
5362 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
5363 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
5364 error "1st tunefs failed"
5365 start_mgsmds || error "start mds failed"
5366 start_ost || error "start ost failed"
5367 mount_client $MOUNT || error "mount client failed"
5368 $LCTL get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
5369 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
5370 umount_client $MOUNT || error "umount client failed"
5374 run_test 73 "failnode to update from mountdata properly"
5376 test_75() { # LU-2374
5377 [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] &&
5378 skip "Need MDS version at least 2.4.1" && return
5381 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
5382 --reformat $(mdsdevname 1) $(mdsvdevname 1)"
5383 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
5384 --reformat $(ostdevname 1) $(ostvdevname 1)"
5386 #check with default parameters
5387 add mds1 $opts_mds || error "add mds1 failed for default params"
5388 add ost1 $opts_ost || error "add ost1 failed for default params"
5390 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
5391 opts_mds=$(echo $opts_mds |
5392 sed -e "s/--index=$index/--index=$index --mdt/")
5393 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
5394 opts_ost=$(echo $opts_ost |
5395 sed -e "s/--index=$index/--index=$index --ost/")
5397 add mds1 $opts_mds || error "add mds1 failed for new params"
5398 add ost1 $opts_ost || error "add ost1 failed for new params"
5399 if ! combined_mgs_mds; then
5400 stop_mgs || error "stop mgs failed"
5405 run_test 75 "The order of --index should be irrelevant"
5408 [[ $(lustre_version_code mgs) -ge $(version_code 2.4.52) ]] ||
5409 { skip "Need MDS version at least 2.4.52" && return 0; }
5411 if ! combined_mgs_mds; then
5412 start_mgs || error "start mgs failed"
5415 local MDMB_PARAM="osc.*.max_dirty_mb"
5416 echo "Change MGS params"
5417 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
5419 echo "max_dirty_mb: $MAX_DIRTY_MB"
5420 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB - 10))
5421 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
5422 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
5423 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5424 head -1" $NEW_MAX_DIRTY_MB
5425 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5426 echo "$MAX_DIRTY_MB"
5427 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5428 error "error while apply max_dirty_mb"
5430 echo "Check the value is stored after remount"
5433 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5434 head -1" $NEW_MAX_DIRTY_MB
5435 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5436 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5437 error "max_dirty_mb is not saved after remount"
5439 echo "Change OST params"
5440 CLIENT_PARAM="obdfilter.*.client_cache_count"
5441 local CLIENT_CACHE_COUNT
5442 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5444 echo "client_cache_count: $CLIENT_CACHE_COUNT"
5445 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
5446 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
5447 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
5448 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5449 head -1" $NEW_CLIENT_CACHE_COUNT
5450 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5452 echo "$CLIENT_CACHE_COUNT"
5453 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5454 error "error while apply client_cache_count"
5456 echo "Check the value is stored after remount"
5459 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5460 head -1" $NEW_CLIENT_CACHE_COUNT
5461 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5463 echo "$CLIENT_CACHE_COUNT"
5464 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5465 error "client_cache_count is not saved after remount"
5468 run_test 76a "set permanent params with lctl across mounts"
5470 test_76b() { # LU-4783
5471 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.57) ]] ||
5472 { skip "Need MGS version at least 2.5.57" && return 0; }
5475 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
5476 error "start params log failed"
5479 run_test 76b "verify params log setup correctly"
5482 [[ $(lustre_version_code mgs) -ge $(version_code 2.8.54) ]] ||
5483 { skip "Need MDS version at least 2.4.52" && return 0; }
5485 local MASK_PARAM="mdd.*.changelog_mask"
5486 echo "Change changelog_mask"
5487 do_facet mgs $LCTL set_param -P $MASK_PARAM=-CLOSE ||
5488 error "Can't change changlog_mask"
5489 wait_update $(facet_host mds) "$LCTL get_param -n $MASK_PARAM |
5492 echo "Check the value is stored after mds remount"
5493 stop_mds || error "Failed to stop MDS"
5494 start_mds || error "Failed to start MDS"
5495 local CHANGELOG_MASK=$(do_facet mgs $LCTL get_param -n $MASK_PARAM)
5496 echo $CHANGELOG_MASK | grep CLOSE > /dev/null &&
5497 error "changelog_mask is not changed"
5501 run_test 76c "verify changelog_mask is applied with lctl set_param -P"
5503 test_76d() { #LU-9399
5506 local xattr_cache="llite.*.xattr_cache"
5507 local cmd="$LCTL get_param -n $xattr_cache | head -1"
5508 local new=$((($(eval $cmd) + 1) % 2))
5510 echo "lctl set_param -P llite.*.xattr_cache=$new"
5511 do_facet mgs $LCTL set_param -P $xattr_cache=$new ||
5512 error "Can't change xattr_cache"
5513 wait_update $HOSTNAME "$cmd" "$new"
5515 echo "Check $xattr_cache on client $MOUNT"
5516 umount_client $MOUNT || error "umount $MOUNT failed"
5517 mount_client $MOUNT || error "mount $MOUNT failed"
5518 [ $(eval $cmd) -eq $new ] ||
5519 error "$xattr_cache != $new on client $MOUNT"
5521 echo "Check $xattr_cache on the new client $MOUNT2"
5522 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
5523 [ $(eval $cmd) -eq $new ] ||
5524 error "$xattr_cache != $new on client $MOUNT2"
5525 umount_client $MOUNT2 || error "umount $MOUNT2 failed"
5529 run_test 76d "verify llite.*.xattr_cache can be set by 'lctl set_param -P' correctly"
5531 test_77() { # LU-3445
5532 local server_version=$(lustre_version_code $SINGLEMDS)
5533 [[ $server_version -ge $(version_code 2.8.55) ]] ||
5534 { skip "Need MDS version 2.8.55+ "; return; }
5536 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
5537 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
5538 skip_env "mixed loopback and real device not working" && return
5541 local fs2mdsdev=$(mdsdevname 1_2)
5542 local fs2ostdev=$(ostdevname 1_2)
5543 local fs2mdsvdev=$(mdsvdevname 1_2)
5544 local fs2ostvdev=$(ostvdevname 1_2)
5545 local fsname=test1234
5547 local failnid="$(h2nettype 1.2.3.4),$(h2nettype 4.3.2.1)"
5549 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
5551 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
5552 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
5553 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
5554 error "start fs2mds failed"
5556 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
5557 mgsnid="0.0.0.0@tcp,$mgsnid,$mgsnid:$mgsnid"
5559 add fs2ost --mgsnode=$mgsnid $(mkfs_opts ost1 $fs2ostdev) \
5560 --failnode=$failnid --fsname=$fsname \
5561 --reformat $fs2ostdev $fs2ostvdev ||
5562 error "add fs2ost failed"
5563 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
5565 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
5566 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
5567 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
5570 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
5573 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ||
5574 $(facet_fstype ost1) != ldiskfs ]] &&
5575 skip "ldiskfs only test" && return
5577 # reformat the Lustre filesystem with a smaller size
5578 local saved_MDSCOUNT=$MDSCOUNT
5579 local saved_MDSSIZE=$MDSSIZE
5580 local saved_OSTCOUNT=$OSTCOUNT
5581 local saved_OSTSIZE=$OSTSIZE
5584 MDSSIZE=$((MDSSIZE - 20000))
5585 OSTSIZE=$((OSTSIZE - 20000))
5586 reformat || error "(1) reformat Lustre filesystem failed"
5587 MDSSIZE=$saved_MDSSIZE
5588 OSTSIZE=$saved_OSTSIZE
5590 # mount the Lustre filesystem
5591 setup_noconfig || error "(2) setup Lustre filesystem failed"
5594 log "create test files"
5599 mkdir $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
5601 for i in $(seq $num_files); do
5602 file=$MOUNT/$tdir/$tfile-$i
5603 dd if=/dev/urandom of=$file count=1 bs=1M || {
5604 $LCTL get_param osc.*.cur*grant*
5605 $LFS df; $LFS df -i;
5606 # stop creating files if there is no more space
5607 if [ ! -e $file ]; then
5608 num_files=$((i - 1))
5612 $LFS getstripe -v $file
5613 local ost_idx=$(LFS getstripe -i $file)
5614 do_facet ost$((ost_idx + 1)) \
5615 $LCTL get_param obdfilter.*.*grant*
5616 error "(4) create $file failed"
5620 # unmount the Lustre filesystem
5621 cleanup || error "(5) cleanup Lustre filesystem failed"
5623 # run e2fsck on the MDT and OST devices
5624 local mds_host=$(facet_active_host $SINGLEMDS)
5625 local ost_host=$(facet_active_host ost1)
5626 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
5627 local ost_dev=$(ostdevname 1)
5629 run_e2fsck $mds_host $mds_dev "-y"
5630 run_e2fsck $ost_host $ost_dev "-y"
5632 # get the original block count of the MDT and OST filesystems
5633 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
5634 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
5636 # expand the MDT and OST filesystems to the device size
5637 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
5638 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
5640 # run e2fsck on the MDT and OST devices again
5641 run_e2fsck $mds_host $mds_dev "-y"
5642 run_e2fsck $ost_host $ost_dev "-y"
5644 # mount the Lustre filesystem
5648 log "check files after expanding the MDT and OST filesystems"
5649 for i in $(seq $num_files); do
5650 file=$MOUNT/$tdir/$tfile-$i
5651 $CHECKSTAT -t file -s 1048576 $file ||
5652 error "(6) checkstat $file failed"
5656 log "create more files after expanding the MDT and OST filesystems"
5657 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
5658 file=$MOUNT/$tdir/$tfile-$i
5659 dd if=/dev/urandom of=$file count=1 bs=1M ||
5660 error "(7) create $file failed"
5663 # unmount the Lustre filesystem
5664 cleanup || error "(8) cleanup Lustre filesystem failed"
5666 # run e2fsck on the MDT and OST devices
5667 run_e2fsck $mds_host $mds_dev "-y"
5668 run_e2fsck $ost_host $ost_dev "-y"
5670 # get the maximum block count of the MDT and OST filesystems
5671 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
5672 local ost_max_blks=$(get_block_count ost1 $ost_dev)
5674 # get the minimum block count of the MDT and OST filesystems
5675 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
5676 grep minimum | sed -e 's/^.*filesystem: //g')
5677 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
5678 grep minimum | sed -e 's/^.*filesystem: //g')
5680 # shrink the MDT and OST filesystems to a smaller size
5684 if [[ $mds_max_blks -gt $mds_min_blks &&
5685 $mds_max_blks -gt $mds_orig_blks ]]; then
5686 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
5687 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
5688 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
5689 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
5690 error "shrink $SINGLEMDS to $new_blks failed"
5694 if [[ $ost_max_blks -gt $ost_min_blks &&
5695 $ost_max_blks -gt $ost_orig_blks ]]; then
5696 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
5697 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
5698 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
5699 run_resize2fs ost1 $ost_dev $new_blks ||
5700 error "shrink ost1 to $new_blks failed"
5704 # check whether the MDT or OST filesystem was shrunk or not
5706 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
5707 reformat || error "(10) reformat Lustre filesystem failed"
5711 # run e2fsck on the MDT and OST devices again
5712 run_e2fsck $mds_host $mds_dev "-y"
5713 run_e2fsck $ost_host $ost_dev "-y"
5715 # mount the Lustre filesystem again
5719 log "check files after shrinking the MDT and OST filesystems"
5720 for i in $(seq $((num_files + 10))); do
5721 file=$MOUNT/$tdir/$tfile-$i
5722 $CHECKSTAT -t file -s 1048576 $file ||
5723 error "(11) checkstat $file failed"
5726 # unmount and reformat the Lustre filesystem
5727 cleanup || error "(12) cleanup Lustre filesystem failed"
5728 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
5730 MDSCOUNT=$saved_MDSCOUNT
5731 OSTCOUNT=$saved_OSTCOUNT
5732 reformat || error "(14) reformat Lustre filesystem failed"
5734 run_test 78 "run resize2fs on MDT and OST filesystems"
5736 test_79() { # LU-4227
5737 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.5.59) ]] ||
5738 { skip "Need MDS version at least 2.5.59"; return 0; }
5740 local mdsdev1=$(mdsdevname 1)
5741 local mdsvdev1=$(mdsvdevname 1)
5742 local mdsdev2=$(mdsdevname 2)
5743 local mdsvdev2=$(mdsvdevname 2)
5744 local ostdev1=$(ostdevname 1)
5745 local ostvdev1=$(ostvdevname 1)
5746 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
5747 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
5748 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
5751 # remove --mgs/--mgsnode from mkfs.lustre options
5752 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
5754 mgsnode_opt=$(echo $opts_mds2 |
5755 awk '{ for ( i = 1; i < NF; i++ )
5756 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5757 [ -n $mgsnode_opt ] &&
5758 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
5760 mgsnode_opt=$(echo $opts_ost1 |
5761 awk '{ for ( i = 1; i < NF; i++ )
5762 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5763 [ -n $mgsnode_opt ] &&
5764 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
5766 # -MGS, format a mdt without --mgs option
5767 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
5768 error "Must specify --mgs when formatting mdt combined with mgs"
5770 # +MGS, format a mdt/ost without --mgsnode option
5771 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
5772 > /dev/null || error "start mds1 failed"
5773 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
5774 error "Must specify --mgsnode when formatting a mdt"
5775 add ost1 $opts_ost1 $ostdev1 $ostvdev2 &&
5776 error "Must specify --mgsnode when formatting an ost"
5780 run_test 79 "format MDT/OST without mgs option (should return errors)"
5783 start_mds || error "Failed to start MDT"
5784 start_ost || error "Failed to start OST1"
5785 uuid=$(do_facet ost1 $LCTL get_param -n mgc.*.uuid)
5786 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
5787 do_facet ost1 "$LCTL set_param fail_val=10 fail_loc=0x906"
5788 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x906"
5789 do_facet mgs "$LCTL set_param -n mgs/MGS/evict_client $uuid"
5791 start_ost2 || error "Failed to start OST2"
5793 do_facet ost1 "$LCTL set_param fail_loc=0"
5796 run_test 80 "mgc import reconnect race"
5798 #Save the original values of $OSTCOUNT and $OSTINDEX$i.
5800 local new_ostcount=$1
5801 saved_ostcount=$OSTCOUNT
5802 OSTCOUNT=$new_ostcount
5806 for i in $(seq $OSTCOUNT); do
5808 eval saved_ostindex$i=${!index}
5813 # Restore the original values of $OSTCOUNT and $OSTINDEX$i.
5814 restore_ostindex() {
5819 for i in $(seq $OSTCOUNT); do
5820 index=saved_ostindex$i
5821 eval OSTINDEX$i=${!index}
5823 OSTCOUNT=$saved_ostcount
5828 # The main purpose of this test is to ensure the OST_INDEX_LIST functions as
5829 # expected. This test uses OST_INDEX_LIST to format OSTs with a randomly
5830 # assigned index and ensures we can mount such a formatted file system
5831 test_81() { # LU-4665
5832 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5833 { skip "Need MDS version at least 2.6.54" && return; }
5834 [[ $OSTCOUNT -ge 3 ]] || { skip_env "needs >= 3 OSTs" && return; }
5838 # Each time RANDOM is referenced, a random integer between 0 and 32767
5841 local saved_ostindex1=$OSTINDEX1
5842 for i in 65535 $((RANDOM + 65536)); do
5843 echo -e "\nFormat ost1 with --index=$i, should fail"
5845 if add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat \
5846 $(ostdevname 1) $(ostvdevname 1); then
5847 OSTINDEX1=$saved_ostindex1
5848 error "format ost1 with --index=$i should fail"
5851 OSTINDEX1=$saved_ostindex1
5855 # Format OSTs with random sparse indices.
5856 trap "restore_ostindex" EXIT
5857 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices"
5858 OST_INDEX_LIST=[0,$((RANDOM * 2 % 65533 + 1)),65534] formatall
5860 # Setup and check Lustre filesystem.
5861 start_mgsmds || error "start_mgsmds failed"
5862 for i in $(seq $OSTCOUNT); do
5863 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5864 error "start ost$i failed"
5867 mount_client $MOUNT || error "mount client $MOUNT failed"
5868 check_mount || error "check client $MOUNT failed"
5871 local max_easize=$($LCTL get_param -n llite.*.max_easize)
5872 [[ $max_easize -eq 128 ]] ||
5873 error "max_easize is $max_easize, should be 128 bytes"
5877 run_test 81 "sparse OST indexing"
5879 # Here we exercise the stripe placement functionality on a file system that
5880 # has formatted the OST with a random index. With the file system the following
5881 # functionality is tested:
5883 # 1. Creating a new file with a specific stripe layout.
5885 # 2. Modifiy a existing empty file with a specific stripe layout.
5887 # 3. Ensure we fail to set the stripe layout of a file that already has one.
5889 # 4. If ost-index is defined we need to ensure it is the first entry in the
5890 # ost index list returned by lfs getstripe.
5892 # 5. Lastly ensure this functionality fails with directories.
5893 test_82a() { # LU-4665
5894 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5895 { skip "Need MDS version at least 2.6.54" && return; }
5896 [[ $OSTCOUNT -ge 3 ]] || { skip_env "needs >= 3 OSTs" && return; }
5902 # Format OSTs with random sparse indices.
5906 local LOV_V1_INSANE_STRIPE_COUNT=65532
5907 for i in $(seq $OSTCOUNT); do
5908 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
5909 ost_indices+=" $index"
5911 ost_indices=$(comma_list $ost_indices)
5913 stack_trap "restore_ostindex" EXIT
5914 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5915 OST_INDEX_LIST=[$ost_indices] formatall
5917 # Setup Lustre filesystem.
5918 start_mgsmds || error "start_mgsmds failed"
5919 for i in $(seq $OSTCOUNT); do
5920 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5921 error "start ost$i failed"
5924 # Collect debug information - start of test
5925 do_nodes $(comma_list $(mdts_nodes)) \
5926 $LCTL get_param osc.*.prealloc_*_id
5928 mount_client $MOUNT || error "mount client $MOUNT failed"
5931 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5932 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5934 stack_trap "do_nodes $(comma_list $(mdts_nodes)) \
5935 $LCTL get_param osc.*.prealloc_*_id" EXIT
5937 # 1. If the file does not exist, new file will be created
5938 # with specified OSTs.
5939 local file=$DIR/$tdir/$tfile-1
5940 local cmd="$SETSTRIPE -o $ost_indices $file"
5942 eval $cmd || error "$cmd failed"
5943 check_stripe_count $file $OSTCOUNT
5944 check_obdidx $file $ost_indices
5945 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5946 error "write $file failed"
5948 # 2. If the file already exists and is an empty file, the file
5949 # will be attached with specified layout.
5950 file=$DIR/$tdir/$tfile-2
5951 mcreate $file || error "mcreate $file failed"
5952 cmd="$SETSTRIPE -o $ost_indices $file"
5954 eval $cmd || error "$cmd failed"
5955 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5956 error "write $file failed"
5957 check_stripe_count $file $OSTCOUNT
5958 check_obdidx $file $ost_indices
5960 # 3. If the file already has a valid layout attached, the command
5961 # should fail with EBUSY.
5963 eval $cmd && error "stripe is already set on $file, $cmd should fail"
5965 # 4. If [--stripe-index|-i <start_ost_idx>] is used, the index must
5966 # be in the OST indices list.
5967 local start_ost_idx=${ost_indices##*,}
5968 file=$DIR/$tdir/$tfile-3
5969 cmd="$SETSTRIPE -o $ost_indices -i $start_ost_idx $file"
5971 eval $cmd || error "$cmd failed"
5972 check_stripe_count $file $OSTCOUNT
5973 check_obdidx $file $ost_indices
5974 check_start_ost_idx $file $start_ost_idx
5976 file=$DIR/$tdir/$tfile-4
5978 cmd+=" -o $(exclude_items_from_list $ost_indices $start_ost_idx)"
5979 cmd+=" -i $start_ost_idx $file"
5981 eval $cmd && error "index $start_ost_idx should be in $ost_indices"
5983 # 5. Specifying OST indices for directory should succeed.
5984 local dir=$DIR/$tdir/$tdir
5985 mkdir $dir || error "mkdir $dir failed"
5986 cmd="$SETSTRIPE -o $ost_indices $dir"
5987 if [[ $(lustre_version_code $SINGLEMDS) -gt $(version_code 2.11.53) &&
5988 $(lustre_version_code client -gt $(version_code 2.11.53)) ]]; then
5990 eval $cmd || error "unable to specify OST indices on directory"
5992 echo "need MDS+client version at least 2.11.53"
5997 run_test 82a "specify OSTs for file (succeed) or directory (succeed)"
6002 # Remove OSTs from a pool and destroy the pool.
6003 destroy_pool $ost_pool || true
6005 if ! combined_mgs_mds ; then
6011 # Test 82b is run to ensure that if the user supplies a pool with a specific
6012 # stripe layout that it behaves proprerly. It should fail in the case that
6013 # the supplied OST index list points to OSTs not contained in the user
6015 test_82b() { # LU-4665
6016 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
6017 { skip "Need MDS version at least 2.6.54" && return; }
6018 [[ $OSTCOUNT -ge 4 ]] || { skip_env "needs >= 4 OSTs" && return; }
6024 # Format OSTs with random sparse indices.
6028 local LOV_V1_INSANE_STRIPE_COUNT=65532
6029 for i in $(seq $OSTCOUNT); do
6030 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
6031 ost_indices+=" $index"
6033 ost_indices=$(comma_list $ost_indices)
6035 trap "restore_ostindex" EXIT
6036 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
6037 OST_INDEX_LIST=[$ost_indices] formatall
6039 # Setup Lustre filesystem.
6040 start_mgsmds || error "start_mgsmds failed"
6041 for i in $(seq $OSTCOUNT); do
6042 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6043 error "start ost$i failed"
6046 mount_client $MOUNT || error "mount client $MOUNT failed"
6047 if ! combined_mgs_mds ; then
6052 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
6053 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6055 # Create a new pool and add OSTs into it.
6056 local ost_pool=$FSNAME.$TESTNAME
6057 create_pool $ost_pool || error "create OST pool $ost_pool failed"
6060 trap "cleanup_82b" EXIT
6062 local ost_idx_in_list=${ost_indices##*,}
6063 local ost_idx_in_pool=$(exclude_items_from_list $ost_indices \
6066 local ost_targets="$FSNAME-OST["
6067 for i in ${ost_idx_in_pool//,/ }; do
6068 ost_targets=$ost_targets$(printf "%04x," $i)
6070 ost_targets="${ost_targets%,}]"
6072 local ost_targets_uuid=$(for i in ${ost_idx_in_pool//,/ }; \
6073 do printf "$FSNAME-OST%04x_UUID\n" $i; done |
6074 sort -u | tr '\n' ' ')
6076 local cmd="$LCTL pool_add $ost_pool $ost_targets"
6077 do_facet mgs $cmd || error "$cmd failed"
6078 wait_update $HOSTNAME "$LCTL get_param -n lov.$FSNAME-*.pools.$TESTNAME|
6079 sort -u | tr '\n' ' ' " "$ost_targets_uuid" ||
6080 error "wait_update $ost_pool failed"
6081 wait_update_facet $SINGLEMDS "$LCTL pool_list $ost_pool | wc -l" 4 ||
6082 error "wait_update pool_list $ost_pool failed"
6084 # If [--pool|-p <pool_name>] is set with [--ost-list|-o <ost_indices>],
6085 # then the OSTs must be the members of the pool.
6086 local file=$DIR/$tdir/$tfile
6087 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
6089 eval $cmd && error "OST with index $ost_idx_in_list should be" \
6090 "in OST pool $ost_pool"
6092 # Only select OST $ost_idx_in_list from $ost_pool for file.
6093 ost_idx_in_list=${ost_idx_in_pool#*,}
6094 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
6096 eval $cmd || error "$cmd failed"
6097 cmd="$GETSTRIPE $file"
6099 eval $cmd || error "$cmd failed"
6100 check_stripe_count $file 2
6101 check_obdidx $file $ost_idx_in_list
6102 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6103 error "write $file failed"
6107 run_test 82b "specify OSTs for file with --pool and --ost-list options"
6110 [[ $(lustre_version_code ost1) -ge $(version_code 2.6.91) ]] ||
6111 { skip "Need OST version at least 2.6.91" && return 0; }
6112 if [ $(facet_fstype ost1) != ldiskfs ]; then
6113 skip "ldiskfs only test"
6123 ostmnt=$(facet_mntpt ost1)
6124 fstype=$(facet_fstype ost1)
6126 # Mount the OST as an ldiskfs filesystem.
6127 log "mount the OST $dev as a $fstype filesystem"
6128 add ost1 $(mkfs_opts ost1 $dev) $FSTYPE_OPT \
6129 --reformat $dev > /dev/null ||
6130 error "format ost1 error"
6132 if ! test -b $dev; then
6133 mnt_opts=$(csa_add "$OST_MOUNT_OPTS" -o loop)
6135 echo "mnt_opts $mnt_opts"
6136 do_facet ost1 mount -t $fstype $dev \
6138 # Run llverfs on the mounted ldiskfs filesystem.
6139 # It is needed to get ENOSPACE.
6140 log "run llverfs in partial mode on the OST $fstype $ostmnt"
6141 do_rpc_nodes $(facet_host ost1) run_llverfs $ostmnt -vpl \
6142 "no" || error "run_llverfs error on $fstype"
6145 log "unmount the OST $dev"
6148 # Delete file IO_scrub. Later osd_scrub_setup will try to
6149 # create "IO_scrub" but will get ENOSPACE.
6151 echo "start ost1 service on `facet_active_host ost1`"
6152 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
6155 err=$(do_facet ost1 dmesg | grep "VFS: Busy inodes after unmount of")
6156 echo "string err $err"
6157 [ -z "$err" ] || error $err
6160 run_test 83 "ENOSPACE on OST doesn't cause message VFS: \
6161 Busy inodes after unmount ..."
6164 local facet=$SINGLEMDS
6165 local num=$(echo $facet | tr -d "mds")
6166 local dev=$(mdsdevname $num)
6167 local time_min=$(recovery_time_min)
6168 local recovery_duration
6169 local completed_clients
6170 local correct_clients
6173 echo "start mds service on $(facet_active_host $facet)"
6175 "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
6176 error "start MDS failed"
6178 start_ost || error "start OST0000 failed"
6179 start_ost2 || error "start OST0001 failed"
6181 echo "recovery_time=$time_min, timeout=$TIMEOUT, wrap_up=$wrap_up"
6183 mount_client $MOUNT1 || error "mount $MOUNT1 failed"
6184 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
6185 # make sure new superblock labels are sync'd before disabling writes
6189 replay_barrier $SINGLEMDS
6190 createmany -o $DIR1/$tfile-%d 1000
6192 # We need to catch the end of recovery window to extend it.
6193 # Skip 5 requests and add delay to request handling.
6194 #define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP
6195 do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5"
6197 facet_failover --fsck $SINGLEMDS || error "failover: $?"
6200 echo "recovery status"
6201 do_facet $SINGLEMDS \
6202 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status"
6204 recovery_duration=$(do_facet $SINGLEMDS \
6205 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
6206 awk '/recovery_duration/ { print $2 }')
6207 (( $recovery_duration > $time_min + $wrap_up )) &&
6208 error "recovery_duration > recovery_time_hard + wrap up"
6209 completed_clients=$(do_facet $SINGLEMDS \
6210 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
6211 awk '/completed_clients/ { print $2 }')
6213 correct_clients="$MDSCOUNT/$((MDSCOUNT+1))"
6214 [ "$completed_clients" = "${correct_clients}" ] ||
6215 error "$completed_clients != $correct_clients"
6217 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
6218 umount_client $MOUNT1
6219 umount_client $MOUNT2
6225 run_test 84 "check recovery_hard_time"
6228 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.55) ]] ||
6229 { skip "Need OST version at least 2.7.55" && return 0; }
6230 ##define OBD_FAIL_OSD_OST_EA_FID_SET 0x197
6231 do_facet ost1 "lctl set_param fail_loc=0x197"
6235 run_test 85 "osd_ost init: fail ea_fid_set"
6240 # ost1 has already registered to the MGS before the reformat.
6241 # So after reformatting it with option "-G", it could not be
6242 # mounted to the MGS. Cleanup the system for subsequent tests.
6247 local server_version=$(lustre_version_code $SINGLEMDS)
6248 [ "$(facet_fstype ost1)" = "zfs" ] &&
6249 skip "LU-6442: no such mkfs params for ZFS OSTs" && return
6250 [[ $server_version -ge $(version_code 2.7.56) ]] ||
6251 { skip "Need server version newer than 2.7.55"; return 0; }
6253 local OST_OPTS="$(mkfs_opts ost1 $(ostdevname 1)) \
6254 --reformat $(ostdevname 1) $(ostvdevname 1)"
6257 local OLDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6258 awk '/Flex block group size: / { print $NF; exit; }')
6261 if [[ ${!opts} != *mkfsoptions* ]]; then
6262 eval opts=\"${!opts} \
6263 --mkfsoptions='\\\"-O flex_bg -G $NEWSIZE\\\"'\"
6265 val=${!opts//--mkfsoptions=\\\"/ \
6266 --mkfsoptions=\\\"-O flex_bg -G $NEWSIZE }
6270 echo "params: $opts"
6272 trap cleanup_86 EXIT ERR
6275 add ost1 $opts || error "add ost1 failed with new params"
6277 local FOUNDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6278 awk '/Flex block group size: / { print $NF; exit; }')
6280 [[ $FOUNDSIZE == $NEWSIZE ]] ||
6281 error "Flex block group size: $FOUNDSIZE, expected: $NEWSIZE"
6285 run_test 86 "Replacing mkfs.lustre -G option"
6287 test_87() { #LU-6544
6288 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.9.51) ]] ||
6289 { skip "Need MDS version at least 2.9.51" && return; }
6290 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ]] &&
6291 { skip "ldiskfs only test" && return; }
6292 [[ $OSTCOUNT -gt 59 ]] &&
6293 { skip "Ignore wide striping situation" && return; }
6295 local mdsdev=$(mdsdevname 1)
6296 local mdsvdev=$(mdsvdevname 1)
6297 local file=$DIR/$tfile
6298 local mntpt=$(facet_mntpt $SINGLEMDS)
6299 local used_xattr_blk=0
6300 local inode_size=${1:-1024}
6302 local xtest="trusted.test"
6306 local stripe_cnt=$(($OSTCOUNT + 2))
6308 #Please see ldiskfs_make_lustre() for MDT inode size calculation
6309 if [ $stripe_cnt -gt 16 ]; then
6312 left_size=$(expr $inode_size - \
6315 32 - 40 \* 3 - 32 \* 3 - $stripe_cnt \* 24 - 16 - 3 - \
6317 24 - 18 - $(expr length $tfile) - 16 - 4)
6318 if [ $left_size -le 0 ]; then
6319 echo "No space($left_size) is expected in inode."
6320 echo "Try 1-byte xattr instead to verify this."
6323 echo "Estimate: at most $left_size-byte space left in inode."
6329 add mds1 $(mkfs_opts mds1 ${mdsdev}) --stripe-count-hint=$stripe_cnt \
6330 --reformat $mdsdev $mdsvdev || error "add mds1 failed"
6331 start_mdt 1 > /dev/null || error "start mdt1 failed"
6332 for i in $(seq $OSTCOUNT); do
6333 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS > /dev/null ||
6334 error "start ost$i failed"
6336 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
6337 check_mount || error "check client $MOUNT failed"
6340 $SETSTRIPE -E 1M -S 1M -c 1 -E 64M -c 1 -E -1 -c -1 $file ||
6341 error "Create file with 3 components failed"
6342 $TRUNCATE $file $((1024*1024*64+1)) || error "truncate file failed"
6343 i=$($GETSTRIPE -I3 -c $file) || error "get 3rd stripe count failed"
6344 if [ $i -ne $OSTCOUNT ]; then
6345 left_size=$(expr $left_size + $(expr $OSTCOUNT - $i) \* 24)
6346 echo -n "Since only $i out $OSTCOUNT OSTs are used, "
6347 echo -n "the expected left space is changed to "
6348 echo "$left_size bytes at most."
6350 value=$(generate_string $left_size)
6351 setfattr -n $xtest -v $value $file
6352 orig=$(get_xattr_value $xtest $file)
6353 [[ "$orig" != "$value" ]] && error "$xtest changed"
6355 #Verify if inode has some expected space left
6356 umount $MOUNT > /dev/null || error "umount $MOUNT failed"
6357 stop_mdt 1 > /dev/null || error "stop mdt1 failed"
6358 mount_ldiskfs $SINGLEMDS || error "mount -t ldiskfs $SINGLEMDS failed"
6360 do_facet $SINGLEMDS ls -sal $mntpt/ROOT/$tfile
6361 used_xattr_blk=$(do_facet $SINGLEMDS ls -s $mntpt/ROOT/$tfile |
6363 [[ $used_xattr_blk -eq 0 ]] &&
6364 error "Please check MDS inode size calculation: \
6365 more than $left_size-byte space left in inode."
6366 echo "Verified: at most $left_size-byte space left in inode."
6370 run_test 87 "check if MDT inode can hold EAs with N stripes properly"
6373 [ "$(facet_fstype mds1)" == "zfs" ] &&
6374 skip "LU-6662: no implementation for ZFS" && return
6378 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
6379 --reformat $(mdsdevname 1) || error "add mds1 failed"
6381 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
6382 grep -e \".*opts:.*errors=remount-ro.*\"" ||
6383 error "default mount options is missing"
6385 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
6386 --mountfsoptions="user_xattr,errors=panic" \
6387 --reformat $(mdsdevname 1) || error "add mds1 failed"
6389 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
6390 grep -e \".*opts:.*errors=panic.*\"" ||
6391 error "user can't override default mount options"
6393 run_test 88 "check the default mount options can be overridden"
6395 test_89() { # LU-7131
6396 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.9.54) ]] ||
6397 { skip "Need MDT version at least 2.9.54" && return 0; }
6399 local key=failover.node
6400 local val1=192.0.2.254@tcp0 # Reserved IPs, see RFC 5735
6401 local val2=192.0.2.255@tcp0
6402 local mdsdev=$(mdsdevname 1)
6407 [ $(facet_fstype mds1) == zfs ] && import_zpool mds1
6408 # Check that parameters are added correctly
6409 echo "tunefs --param $key=$val1"
6410 do_facet mds "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
6411 error "tunefs --param $key=$val1 failed"
6412 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6413 error "tunefs --dryrun failed"
6414 params=${params##*Parameters:}
6415 params=${params%%exiting*}
6416 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val1) = "1" ] ||
6417 error "on-disk parameter not added correctly via tunefs"
6419 # Check that parameters replace existing instances when added
6420 echo "tunefs --param $key=$val2"
6421 do_facet mds "$TUNEFS --param $key=$val2 $mdsdev >/dev/null" ||
6422 error "tunefs --param $key=$val2 failed"
6423 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6424 error "tunefs --dryrun failed"
6425 params=${params##*Parameters:}
6426 params=${params%%exiting*}
6427 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "1" ] ||
6428 error "on-disk parameter not replaced via tunefs"
6429 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val2) = "1" ] ||
6430 error "on-disk parameter not replaced correctly via tunefs"
6432 # Check that a parameter is erased properly
6433 echo "tunefs --erase-param $key"
6434 do_facet mds "$TUNEFS --erase-param $key $mdsdev >/dev/null" ||
6435 error "tunefs --erase-param $key failed"
6436 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6437 error "tunefs --dryrun failed"
6438 params=${params##*Parameters:}
6439 params=${params%%exiting*}
6440 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "0" ] ||
6441 error "on-disk parameter not erased correctly via tunefs"
6443 # Check that all the parameters are erased
6444 echo "tunefs --erase-params"
6445 do_facet mds "$TUNEFS --erase-params $mdsdev >/dev/null" ||
6446 error "tunefs --erase-params failed"
6447 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6448 error "tunefs --dryrun failed"
6449 params=${params##*Parameters:}
6450 params=${params%%exiting*}
6452 error "all on-disk parameters not erased correctly via tunefs"
6454 # Check the order of options --erase-params and --param
6455 echo "tunefs --param $key=$val1 --erase-params"
6457 "$TUNEFS --param $key=$val1 --erase-params $mdsdev >/dev/null"||
6458 error "tunefs --param $key=$val1 --erase-params failed"
6459 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6460 error "tunefs --dryrun failed"
6461 params=${params##*Parameters:}
6462 params=${params%%exiting*}
6463 [ $(echo $params | tr ' ' '\n') == "$key=$val1" ] ||
6464 error "on-disk param not added correctly with --erase-params"
6468 run_test 89 "check tunefs --param and --erase-param{s} options"
6471 # $2 (optional) value of max_mod_rpcs_in_flight to set
6472 check_max_mod_rpcs_in_flight() {
6480 idx=$(printf "%04x" $($LFS getdirstripe -i $dir))
6481 facet="mds$((0x$idx + 1))"
6483 if [ -z "$mmr" ]; then
6484 # get value of max_mod_rcps_in_flight
6485 mmr=$($LCTL get_param -n \
6486 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) ||
6487 error "Unable to get max_mod_rpcs_in_flight"
6488 echo "max_mod_rcps_in_flight is $mmr"
6490 # set value of max_mod_rpcs_in_flight
6492 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight=$mmr ||
6493 error "Unable to set max_mod_rpcs_in_flight to $mmr"
6494 echo "max_mod_rpcs_in_flight set to $mmr"
6497 # create mmr+1 files
6498 echo "creating $((mmr + 1)) files ..."
6500 for i in $(seq $((mmr + 1))); do
6506 # consumes mmr-1 modify RPC slots
6507 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6508 # drop requests on MDT so that RPC slots are consumed
6509 # during all the request resend interval
6510 do_facet $facet "$LCTL set_param fail_loc=0x159"
6511 echo "launch $((mmr - 1)) chmod in parallel ..."
6512 for i in $(seq $((mmr - 1))); do
6513 chmod 0600 $dir/file-$i &
6517 # send one additional modify RPC
6518 do_facet $facet "$LCTL set_param fail_loc=0"
6519 echo "launch 1 additional chmod in parallel ..."
6520 chmod 0600 $dir/file-$mmr &
6523 # check this additional modify RPC get a modify RPC slot
6524 # and succeed its operation
6525 checkstat -vp 0600 $dir/file-$mmr ||
6526 error "Unable to send $mmr modify RPCs in parallel"
6531 # consumes mmr modify RPC slots
6532 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6533 # drop requests on MDT so that RPC slots are consumed
6534 # during all the request resend interval
6535 do_facet $facet "$LCTL set_param fail_loc=0x159"
6536 echo "launch $mmr chmod in parallel ..."
6537 for i in $(seq $mmr); do
6538 chmod 0666 $dir/file-$i &
6542 # send one additional modify RPC
6543 do_facet $facet "$LCTL set_param fail_loc=0"
6544 echo "launch 1 additional chmod in parallel ..."
6545 chmod 0666 $dir/file-$((mmr + 1)) &
6548 # check this additional modify RPC blocked getting a modify RPC slot
6549 checkstat -vp 0644 $dir/file-$((mmr + 1)) ||
6550 error "Unexpectedly send $(($mmr + 1)) modify RPCs in parallel"
6556 if ! combined_mgs_mds ; then
6561 [[ $($LCTL get_param mdc.*.import |
6562 grep "connect_flags:.*multi_mod_rpc") ]] ||
6563 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6565 # check default value
6566 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6567 check_max_mod_rpcs_in_flight $DIR/$tdir
6571 run_test 90a "check max_mod_rpcs_in_flight is enforced"
6581 [[ $($LCTL get_param mdc.*.import |
6582 grep "connect_flags:.*multi_mod_rpc") ]] ||
6583 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6586 # update max_mod_rpcs_in_flight
6587 $LFS mkdir -c1 $DIR/${tdir}1 || error "mkdir $DIR/${tdir}1 failed"
6588 check_max_mod_rpcs_in_flight $DIR/${tdir}1 1
6591 # check client is able to send multiple modify RPCs in paralell
6592 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
6593 grep -c "multi_mod_rpcs")
6594 if [ "$tmp" -ne $MDSCOUNT ]; then
6595 echo "Client not able to send multiple modify RPCs in parallel"
6600 # update max_mod_rpcs_in_flight
6601 $LFS mkdir -c1 $DIR/${tdir}2 || error "mkdir $DIR/${tdir}2 failed"
6602 check_max_mod_rpcs_in_flight $DIR/${tdir}2 5
6605 $LFS mkdir -c1 $DIR/${tdir}3 || error "mkdir $DIR/${tdir}3 failed"
6606 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/${tdir}3))
6607 facet="mds$((0x$idx + 1))"
6609 # save MDT max_mod_rpcs_per_client
6610 mmrpc=$(do_facet $facet \
6611 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
6613 # update max_mod_rpcs_in_flight
6614 umount_client $MOUNT
6616 "echo 16 > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6618 $LCTL set_param mdc.$FSNAME-MDT$idx-mdc-*.max_rpcs_in_flight=17
6619 check_max_mod_rpcs_in_flight $DIR/${tdir}3 16
6621 # restore MDT max_mod_rpcs_per_client initial value
6623 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6625 rm -rf $DIR/${tdir}?
6628 run_test 90b "check max_mod_rpcs_in_flight is enforced after update"
6637 [[ $($LCTL get_param mdc.*.import |
6638 grep "connect_flags:.*multi_mod_rpc") ]] ||
6639 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6641 # check client is able to send multiple modify RPCs in paralell
6642 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
6643 grep -c "multi_mod_rpcs")
6644 if [ "$tmp" -ne $MDSCOUNT ]; then
6645 skip "Client not able to send multiple modify RPCs in parallel"
6650 # get max_rpcs_in_flight value
6651 mrif=$($LCTL get_param -n mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
6652 echo "max_rpcs_in_flight is $mrif"
6654 # get MDT max_mod_rpcs_per_client
6655 mmrpc=$(do_facet mds1 \
6656 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
6657 echo "max_mod_rpcs_per_client is $mmrpc"
6660 # attempt to set max_mod_rpcs_in_flight to max_rpcs_in_flight value
6661 # prerequisite: set max_mod_rpcs_per_client to max_rpcs_in_flight value
6662 umount_client $MOUNT
6664 "echo $mrif > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6668 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif &&
6669 error "set max_mod_rpcs_in_flight to $mrif should fail"
6671 umount_client $MOUNT
6673 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6677 # attempt to set max_mod_rpcs_in_flight to max_mod_rpcs_per_client+1
6678 # prerequisite: set max_rpcs_in_flight to max_mod_rpcs_per_client+2
6680 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$((mmrpc + 2))
6683 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$((mmrpc + 1)) &&
6684 error "set max_mod_rpcs_in_flight to $((mmrpc + 1)) should fail"
6688 run_test 90c "check max_mod_rpcs_in_flight update limits"
6699 [[ $($LCTL get_param mdc.*.import |
6700 grep "connect_flags:.*multi_mod_rpc") ]] ||
6701 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6703 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6704 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
6705 facet="mds$((0x$idx + 1))"
6707 # check client version supports multislots
6708 tmp=$($LCTL get_param -N \
6709 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
6710 if [ -z "$tmp" ]; then
6711 skip "Client does not support multiple modify RPCs in flight"
6716 # get current value of max_mod_rcps_in_flight
6717 mmr=$($LCTL get_param -n \
6718 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
6719 echo "max_mod_rcps_in_flight is $mmr"
6722 echo "creating $mmr files ..."
6724 for i in $(seq $mmr); do
6725 touch $DIR/$tdir/file-$i
6728 # prepare for close RPC
6729 multiop_bg_pause $DIR/$tdir/file-close O_c
6732 # consumes mmr modify RPC slots
6733 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6734 # drop requests on MDT so that RPC slots are consumed
6735 # during all the request resend interval
6736 do_facet $facet "$LCTL set_param fail_loc=0x159"
6737 echo "launch $mmr chmod in parallel ..."
6738 for i in $(seq $mmr); do
6739 chmod 0600 $DIR/$tdir/file-$i &
6742 # send one additional close RPC
6743 do_facet $facet "$LCTL set_param fail_loc=0"
6744 echo "launch 1 additional close in parallel ..."
6746 cancel_lru_locks mdc
6749 # check this additional close RPC get a modify RPC slot
6750 # and multiop process completed
6751 [ -d /proc/$pid ] &&
6752 error "Unable to send the additional close RPC in parallel"
6757 run_test 90d "check one close RPC is allowed above max_mod_rpcs_in_flight"
6759 check_uuid_on_ost() {
6761 do_facet ost1 "$LCTL get_param obdfilter.${FSNAME}*.exports.'$nid'.uuid"
6764 check_uuid_on_mdt() {
6766 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
6774 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.63) ]] ||
6775 { skip "Need OST version at least 2.7.63" && return 0; }
6776 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.63) ]] ||
6777 { skip "Need MDT version at least 2.7.63" && return 0; }
6779 start_mds || error "MDS start failed"
6780 start_ost || error "unable to start OST"
6781 mount_client $MOUNT || error "client start failed"
6782 check_mount || error "check_mount failed"
6785 nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g")
6789 uuid=$(get_client_uuid $MOUNT)
6791 echo "list nids on mdt:"
6792 do_facet $SINGLEMDS "$LCTL list_param mdt.${FSNAME}*.exports.*"
6793 echo "uuid from $nid:"
6794 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
6796 found=$(check_uuid_on_mdt $nid | grep $uuid)
6797 [ -z "$found" ] && error "can't find $uuid $nid on MDT"
6798 found=$(check_uuid_on_ost $nid | grep $uuid)
6799 [ -z "$found" ] && error "can't find $uuid $nid on OST"
6801 # umount the client so it won't reconnect
6802 manual_umount_client --force || error "failed to umount $?"
6803 # shouldn't disappear on MDS after forced umount
6804 found=$(check_uuid_on_mdt $nid | grep $uuid)
6805 [ -z "$found" ] && error "can't find $uuid $nid"
6808 do_facet $SINGLEMDS \
6809 "$LCTL set_param -n mdt.${mds1_svc}.evict_client nid:$nid"
6811 found=$(check_uuid_on_mdt $nid | grep $uuid)
6812 [ -n "$found" ] && error "found $uuid $nid on MDT"
6813 found=$(check_uuid_on_ost $nid | grep $uuid)
6814 [ -n "$found" ] && error "found $uuid $nid on OST"
6816 # check it didn't reconnect (being umounted)
6817 sleep $((TIMEOUT+1))
6818 found=$(check_uuid_on_mdt $nid | grep $uuid)
6819 [ -n "$found" ] && error "found $uuid $nid on MDT"
6820 found=$(check_uuid_on_ost $nid | grep $uuid)
6821 [ -n "$found" ] && error "found $uuid $nid on OST"
6825 run_test 91 "evict-by-nid support"
6827 generate_ldev_conf() {
6828 # generate an ldev.conf file
6829 local ldevconfpath=$1
6831 local fsldevformat=""
6834 fstype=$(facet_fstype mgs)
6835 if [ "$fstype" == "zfs" ]; then
6836 fsldevformat="$fstype:"
6841 printf "%s\t-\t%s-MGS0000\t%s%s\n" \
6845 $(mgsdevname) > $ldevconfpath
6847 local mdsfo_host=$mdsfailover_HOST;
6848 if [ -z "$mdsfo_host" ]; then
6852 for num in $(seq $MDSCOUNT); do
6853 fstype=$(facet_fstype mds$num)
6854 if [ "$fstype" == "zfs" ]; then
6855 fsldevformat="$fstype:"
6860 printf "%s\t%s\t%s-MDT%04d\t%s%s\n" \
6866 $(mdsdevname $num) >> $ldevconfpath
6869 local ostfo_host=$ostfailover_HOST;
6870 if [ -z "$ostfo_host" ]; then
6874 for num in $(seq $OSTCOUNT); do
6875 fstype=$(facet_fstype ost$num)
6876 if [ "$fstype" == "zfs" ]; then
6877 fsldevformat="$fstype:"
6882 printf "%s\t%s\t%s-OST%04d\t%s%s\n" \
6888 $(ostdevname $num) >> $ldevconfpath
6891 echo "----- $ldevconfpath -----"
6893 echo "--- END $ldevconfpath ---"
6898 # generate a nids file (mapping between hostname to nid)
6899 # looks like we only have the MGS nid available to us
6900 # so just echo that to a file
6902 echo -e "${mgs_HOST}\t${MGSNID}" > $nidspath
6904 echo "----- $nidspath -----"
6906 echo "--- END $nidspath ---"
6909 compare_ldev_output() {
6913 sort $expected_output -o $expected_output
6914 sort $ldev_output -o $ldev_output
6916 echo "-- START OF LDEV OUTPUT --"
6918 echo "--- END OF LDEV OUTPUT ---"
6920 echo "-- START OF EXPECTED OUTPUT --"
6921 cat $expected_output
6922 echo "--- END OF EXPECTED OUTPUT ---"
6924 diff $expected_output $ldev_output
6929 if [ -z "$LDEV" ]; then
6930 error "ldev is missing!"
6933 local LDEVCONFPATH=$TMP/ldev.conf
6934 local NIDSPATH=$TMP/nids
6936 echo "Host is $(hostname)"
6938 generate_ldev_conf $LDEVCONFPATH
6939 generate_nids $NIDSPATH
6941 # echo the mgs nid and compare it to environment variable MGSNID
6942 # also, ldev.conf and nids is a server side thing, use the OSS
6945 output=$($LDEV -c $LDEVCONFPATH -H $ost_HOST -n $NIDSPATH echo %m)
6947 echo "-- START OF LDEV OUTPUT --"
6949 echo "--- END OF LDEV OUTPUT ---"
6951 # ldev failed, error
6952 if [ $? -ne 0 ]; then
6953 rm $LDEVCONFPATH $NIDSPATH
6954 error "ldev failed to execute!"
6957 # need to process multiple lines because of combined MGS and MDS
6958 echo -e $output | awk '{ print $2 }' | while read -r line ; do
6959 if [ "$line" != "$MGSNID" ]; then
6960 rm $LDEVCONFPATH $NIDSPATH
6961 error "ldev failed mgs nid '$line', expected '$MGSNID'"
6965 rm $LDEVCONFPATH $NIDSPATH
6967 run_test 92 "ldev returns MGS NID correctly in command substitution"
6970 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs" && return
6973 #start mgs or mgs/mdt0
6974 if ! combined_mgs_mds ; then
6981 start_ost || error "OST0 start fail"
6983 #define OBD_FAIL_MGS_WRITE_TARGET_DELAY 0x90e
6984 do_facet mgs "$LCTL set_param fail_val = 10 fail_loc=0x8000090e"
6985 for num in $(seq 2 $MDSCOUNT); do
6989 mount_client $MOUNT || error "mount client fails"
6990 wait_osc_import_state mds ost FULL
6991 wait_osc_import_state client ost FULL
6992 check_mount || error "check_mount failed"
6994 cleanup || error "cleanup failed with $?"
6996 run_test 93 "register mulitple MDT at the same time"
6999 if [ -z "$LDEV" ]; then
7000 error "ldev is missing!"
7003 local LDEVCONFPATH=$TMP/ldev.conf
7004 local NIDSPATH=$TMP/nids
7006 generate_ldev_conf $LDEVCONFPATH
7007 generate_nids $NIDSPATH
7009 local LDEV_OUTPUT=$TMP/ldev-output.txt
7010 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME > $LDEV_OUTPUT
7012 # ldev failed, error
7013 if [ $? -ne 0 ]; then
7014 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7015 error "ldev failed to execute!"
7019 local EXPECTED_OUTPUT=$TMP/ldev-expected.txt
7021 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
7023 for num in $(seq $MDSCOUNT); do
7024 printf "%s-MDT%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
7027 for num in $(seq $OSTCOUNT); do
7028 printf "%s-OST%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
7031 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7033 if [ $? -ne 0 ]; then
7034 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7035 error "ldev failed to produce the correct hostlist!"
7038 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7040 run_test 94 "ldev outputs correct labels for file system name query"
7043 if [ -z "$LDEV" ]; then
7044 error "ldev is missing!"
7047 local LDEVCONFPATH=$TMP/ldev.conf
7048 local NIDSPATH=$TMP/nids
7050 generate_ldev_conf $LDEVCONFPATH
7051 generate_nids $NIDSPATH
7055 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME &>/dev/null
7056 if [ $? -ne 0 ]; then
7057 rm $LDEVCONFPATH $NIDSPATH
7058 error "ldev label filtering w/ -F failed!"
7062 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -l &>/dev/null
7063 if [ $? -ne 0 ]; then
7064 rm $LDEVCONFPATH $NIDSPATH
7065 error "ldev label filtering w/ -l failed!"
7069 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f &>/dev/null
7070 if [ $? -ne 0 ]; then
7071 rm $LDEVCONFPATH $NIDSPATH
7072 error "ldev label filtering w/ -f failed!"
7076 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a &>/dev/null
7077 if [ $? -ne 0 ]; then
7078 rm $LDEVCONFPATH $NIDSPATH
7079 error "ldev label filtering w/ -a failed!"
7084 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -F $FSNAME &>/dev/null
7085 if [ $? -eq 0 ]; then
7086 rm $LDEVCONFPATH $NIDSPATH
7087 error "ldev label filtering w/ -a and -F incorrectly succeeded"
7091 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -f &>/dev/null
7092 if [ $? -eq 0 ]; then
7093 rm $LDEVCONFPATH $NIDSPATH
7094 error "ldev label filtering w/ -a and -f incorrectly succeeded"
7098 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -l &>/dev/null
7099 if [ $? -eq 0 ]; then
7100 rm $LDEVCONFPATH $NIDSPATH
7101 error "ldev label filtering w/ -a and -l incorrectly succeeded"
7105 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f -l &>/dev/null
7106 if [ $? -eq 0 ]; then
7107 rm $LDEVCONFPATH $NIDSPATH
7108 error "ldev label filtering w/ -f and -l incorrectly succeeded"
7112 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -l &>/dev/null
7113 if [ $? -eq 0 ]; then
7114 rm $LDEVCONFPATH $NIDSPATH
7115 error "ldev label filtering w/ -F and -l incorrectly succeeded"
7118 # file sys & foreign
7119 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -f &>/dev/null
7120 if [ $? -eq 0 ]; then
7121 rm $LDEVCONFPATH $NIDSPATH
7122 error "ldev label filtering w/ -F and -f incorrectly succeeded"
7125 rm $LDEVCONFPATH $NIDSPATH
7127 run_test 95 "ldev should only allow one label filter"
7130 if [ -z "$LDEV" ]; then
7131 error "ldev is missing!"
7134 local LDEVCONFPATH=$TMP/ldev.conf
7135 local NIDSPATH=$TMP/nids
7137 generate_ldev_conf $LDEVCONFPATH
7138 generate_nids $NIDSPATH
7140 local LDEV_OUTPUT=$TMP/ldev-output.txt
7141 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -H $mgs_HOST \
7143 awk '{print $2}' > $LDEV_OUTPUT
7145 # ldev failed, error
7146 if [ $? -ne 0 ]; then
7147 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7148 error "ldev failed to execute!"
7152 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
7154 echo "$mgs_HOST-$(facet_fstype mgs)" > $EXPECTED_OUTPUT
7156 if [ "$mgs_HOST" == "$mds_HOST" ]; then
7157 for num in $(seq $MDSCOUNT); do
7158 echo "$mds_HOST-$(facet_fstype mds$num)" \
7163 if [ "$mgs_HOST" == "$ost_HOST" ]; then
7164 for num in $(seq $OSTCOUNT); do
7165 echo "$ost_HOST-$(facet_fstype ost$num)" \
7170 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7172 if [ $? -ne 0 ]; then
7173 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7174 error "ldev failed to produce the correct output!"
7177 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7179 run_test 96 "ldev returns hostname and backend fs correctly in command sub"
7182 if [ -z "$LDEV" ]; then
7183 error "ldev is missing!"
7186 local LDEVCONFPATH=$TMP/ldev.conf
7187 local NIDSPATH=$TMP/nids
7189 generate_ldev_conf $LDEVCONFPATH
7190 generate_nids $NIDSPATH
7192 local LDEV_OUTPUT=$TMP/ldev-output.txt
7193 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
7195 echo -e "\nMDT role"
7196 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mdt > $LDEV_OUTPUT
7198 if [ $? -ne 0 ]; then
7199 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7200 error "ldev failed to execute for mdt role!"
7203 for num in $(seq $MDSCOUNT); do
7204 printf "%s-MDT%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
7207 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7209 if [ $? -ne 0 ]; then
7210 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7211 error "ldev failed to produce the correct output for mdt role!"
7214 echo -e "\nOST role"
7215 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R ost > $LDEV_OUTPUT
7217 if [ $? -ne 0 ]; then
7218 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
7219 error "ldev failed to execute for ost role!"
7223 for num in $(seq $OSTCOUNT); do
7224 printf "%s-OST%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
7227 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7229 if [ $? -ne 0 ]; then
7230 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7231 error "ldev failed to produce the correct output for ost role!"
7234 echo -e "\nMGS role"
7235 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mgs > $LDEV_OUTPUT
7237 if [ $? -ne 0 ]; then
7238 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
7239 error "ldev failed to execute for mgs role!"
7242 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
7244 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7246 if [ $? -ne 0 ]; then
7247 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7248 error "ldev failed to produce the correct output for mgs role!"
7251 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7253 run_test 97 "ldev returns correct ouput when querying based on role"
7258 local temp=$MDS_MOUNT_OPTS
7261 check_mount || error "mount failed"
7262 mountopt="user_xattr"
7263 for ((x = 1; x <= 400; x++)); do
7264 mountopt="$mountopt,user_xattr"
7266 remount_client $mountopt $MOUNT 2>&1 | grep "too long" ||
7267 error "Buffer overflow check failed"
7268 cleanup || error "cleanup failed"
7270 run_test 98 "Buffer-overflow check while parsing mount_opts"
7274 [[ $(facet_fstype ost1) != ldiskfs ]] &&
7275 { skip "ldiskfs only test" && return; }
7276 [[ $(lustre_version_code ost1) -ge $(version_code 2.8.57) ]] ||
7277 { skip "Need OST version at least 2.8.57" && return 0; }
7279 local ost_opts="$(mkfs_opts ost1 $(ostdevname 1)) \
7280 --reformat $(ostdevname 1) $(ostvdevname 1)"
7281 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" &&
7282 skip "meta_bg already set" && return
7285 if [[ ${!opts} != *mkfsoptions* ]]; then
7286 eval opts=\"${!opts} \
7287 --mkfsoptions='\\\"-O ^resize_inode,meta_bg\\\"'\"
7289 local val=${!opts//--mkfsoptions=\\\"/ \
7290 --mkfsoptions=\\\"-O ^resize_inode,meta_bg }
7294 echo "params: $opts"
7296 add ost1 $opts || error "add ost1 failed with new params"
7298 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" ||
7299 error "meta_bg is not set"
7303 run_test 99 "Adding meta_bg option"
7307 start_mds || error "MDS start failed"
7308 start_ost || error "unable to start OST"
7309 mount_client $MOUNT || error "client start failed"
7310 check_mount || error "check_mount failed"
7319 do_facet mgs 'lshowmount -v' | awk 'BEGIN {NR == 0; rc=1} /MGS:/ {rc=0}
7320 END {exit rc}' || error "lshowmount have no output MGS"
7322 do_facet mds1 'lshowmount -v' | awk 'BEGIN {NR == 2; rc=1} /-MDT0000:/
7323 {rc=0} END {exit rc}' || error "lshowmount have no output MDT0"
7325 do_facet ost1 'lshowmount -v' | awk 'BEGIN {NR == 4; rc=1} /-OST0000:/
7326 {rc=0} END {exit rc}' || error "lshowmount have no output OST0"
7328 cleanup || error "cleanup failed with $?"
7330 run_test 100 "check lshowmount lists MGS, MDT, OST and 0@lo"
7333 local createmany_pid
7334 local dev=$FSNAME-OST0000-osc-MDT0000
7338 createmany -o $DIR1/$tdir/$tfile-%d 50000 &
7340 # MDT->OST reconnection causes MDT<->OST last_id synchornisation
7341 # via osp_precreate_cleanup_orphans.
7342 for ((i = 0; i < 100; i++)); do
7343 for ((k = 0; k < 10; k++)); do
7344 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;" \
7345 "$LCTL --device $dev activate"
7348 ls -asl $MOUNT | grep '???' &&
7349 { kill -9 $createmany_pid &>/dev/null;
7350 error "File has no object on OST"; }
7352 kill -s 0 $createmany_pid || break
7354 wait $createmany_pid
7356 unlinkmany $DIR1/$tdir/$tfile-%d 50000
7359 run_test 101 "Race MDT->OST reconnection with create"
7362 [[ $(lustre_version_code $SINGLEMDS) -gt $(version_code 2.9.53) ]] ||
7363 skip "Need server version greater than 2.9.53"
7364 cleanup || error "cleanup failed with $?"
7366 local mds1dev=$(mdsdevname 1)
7367 local mds1mnt=$(facet_mntpt mds1)
7368 local mds1fstype=$(facet_fstype mds1)
7369 local mds1opts=$MDS_MOUNT_OPTS
7371 if [ $mds1fstype == ldiskfs ] &&
7372 ! do_facet mds1 test -b $mds1dev; then
7373 mds1opts=$(csa_add "$mds1opts" -o loop)
7375 if [[ $mds1fstype == zfs ]]; then
7376 import_zpool mds1 || return ${PIPESTATUS[0]}
7379 # unload all and only load libcfs to allow fail_loc setting
7380 do_facet mds1 $LUSTRE_RMMOD || error "unable to unload modules"
7381 do_facet mds1 modprobe libcfs || error "libcfs not loaded"
7382 do_facet mds1 lsmod \| grep libcfs || error "libcfs not loaded"
7384 #define OBD_FAIL_OBDCLASS_MODULE_LOAD 0x60a
7385 do_facet mds1 "$LCTL set_param fail_loc=0x8000060a"
7387 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts &&
7388 error "mdt start must fail"
7389 do_facet mds1 lsmod \| grep obdclass && error "obdclass must not load"
7391 do_facet mds1 "$LCTL set_param fail_loc=0x0"
7393 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts ||
7394 error "mdt start must not fail"
7396 cleanup || error "cleanup failed with $?"
7398 run_test 102 "obdclass module cleanup upon error"
7403 echo "rename $FSNAME to $newname"
7405 if ! combined_mgs_mds ; then
7406 local facet=$(mgsdevname)
7409 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $facet"||
7410 error "(7) Fail to rename MGS"
7411 if [ "$(facet_fstype $facet)" = "zfs" ]; then
7412 reimport_zpool mgs $newname-mgs
7416 for num in $(seq $MDSCOUNT); do
7417 local facet=$(mdsdevname $num)
7419 do_facet mds${num} \
7420 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $facet"||
7421 error "(8) Fail to rename MDT $num"
7422 if [ "$(facet_fstype $facet)" = "zfs" ]; then
7423 reimport_zpool mds${num} $newname-mdt${num}
7427 for num in $(seq $OSTCOUNT); do
7428 local facet=$(ostdevname $num)
7430 do_facet ost${num} \
7431 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $facet"||
7432 error "(9) Fail to rename OST $num"
7433 if [ "$(facet_fstype $facet)" = "zfs" ]; then
7434 reimport_zpool ost${num} $newname-ost${num}
7439 test_103_set_pool() {
7443 do_facet mgs $LCTL pool_add $FSNAME.$pname ${FSNAME}-$ost_x ||
7444 error "Fail to add $ost_x to $FSNAME.$pname"
7445 wait_update $HOSTNAME \
7446 "lctl get_param -n lov.$FSNAME-clilov-*.pools.$pname |
7447 grep $ost_x" "$FSNAME-${ost_x}_UUID" ||
7448 error "$ost_x is NOT in pool $FSNAME.$pname"
7451 test_103_check_pool() {
7452 local save_fsname=$1
7455 stat $DIR/$tdir/test-framework.sh ||
7456 error "($errno) Fail to stat"
7457 do_facet mgs $LCTL pool_list $FSNAME.pool1 ||
7458 error "($errno) Fail to list $FSNAME.pool1"
7459 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname ||
7460 error "($errno) Fail to list $FSNAME.$save_fsname"
7461 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname |
7462 grep ${FSNAME}-OST0000 ||
7463 error "($errno) List $FSNAME.$save_fsname is invalid"
7465 local pname=$($LFS getstripe --pool $DIR/$tdir/d0)
7466 [ "$pname" = "$save_fsname" ] ||
7467 error "($errno) Unexpected pool name $pname"
7471 check_mount_and_prep
7473 mkdir $DIR/$tdir || error "(1) Fail to mkdir $DIR/$tdir"
7474 cp $LUSTRE/tests/test-framework.sh $DIR/$tdir ||
7475 error "(2) Fail to copy test-framework.sh"
7477 if ! combined_mgs_mds ; then
7480 do_facet mgs $LCTL pool_new $FSNAME.pool1 ||
7481 error "(3) Fail to create $FSNAME.pool1"
7482 # name the pool name as the fsname
7483 do_facet mgs $LCTL pool_new $FSNAME.$FSNAME ||
7484 error "(4) Fail to create $FSNAME.$FSNAME"
7486 test_103_set_pool $FSNAME OST0000
7488 $SETSTRIPE -p $FSNAME $DIR/$tdir/d0 ||
7489 error "(6) Fail to setstripe on $DIR/$tdir/d0"
7491 if ! combined_mgs_mds ; then
7497 test_renamefs mylustre
7499 local save_fsname=$FSNAME
7503 if ! combined_mgs_mds ; then
7506 test_103_check_pool $save_fsname 7
7508 if [ $OSTCOUNT -ge 2 ]; then
7509 test_103_set_pool $save_fsname OST0001
7512 $SETSTRIPE -p $save_fsname $DIR/$tdir/f0 ||
7513 error "(16) Fail to setstripe on $DIR/$tdir/f0"
7514 if ! combined_mgs_mds ; then
7525 if ! combined_mgs_mds ; then
7528 test_103_check_pool $save_fsname 17
7530 if ! combined_mgs_mds ; then
7535 test_renamefs $save_fsname
7541 run_test 103 "rename filesystem name"
7543 test_104() { # LU-6952
7544 local mds_mountopts=$MDS_MOUNT_OPTS
7545 local ost_mountopts=$OST_MOUNT_OPTS
7546 local mds_mountfsopts=$MDS_MOUNT_FS_OPTS
7547 local lctl_ver=$(do_facet $SINGLEMDS $LCTL --version |
7550 [[ $(version_code $lctl_ver) -lt $(version_code 2.9.55) ]] &&
7551 { skip "this test needs utils above 2.9.55" && return 0; }
7553 # specify "acl" in mount options used by mkfs.lustre
7554 if [ -z "$MDS_MOUNT_FS_OPTS" ]; then
7555 MDS_MOUNT_FS_OPTS="acl,user_xattr"
7558 MDS_MOUNT_FS_OPTS="${MDS_MOUNT_FS_OPTS},acl,user_xattr"
7561 echo "mountfsopt: $MDS_MOUNT_FS_OPTS"
7563 #reformat/remount the MDT to apply the MDT_MOUNT_FS_OPT options
7565 if [ -z "$MDS_MOUNT_OPTS" ]; then
7566 MDS_MOUNT_OPTS="-o noacl"
7568 MDS_MOUNT_OPTS="${MDS_MOUNT_OPTS},noacl"
7571 for num in $(seq $MDSCOUNT); do
7572 start mds$num $(mdsdevname $num) $MDS_MOUNT_OPTS ||
7573 error "Failed to start MDS"
7576 for num in $(seq $OSTCOUNT); do
7577 start ost$num $(ostdevname $num) $OST_MOUNT_OPTS ||
7578 error "Failed to start OST"
7582 setfacl -m "d:$RUNAS_ID:rwx" $MOUNT &&
7583 error "ACL is applied when FS is mounted with noacl."
7585 MDS_MOUNT_OPTS=$mds_mountopts
7586 OST_MOUNT_OPTS=$ost_mountopts
7587 MDS_MOUNT_FS_OPTS=$mds_mountfsopts
7592 run_test 104 "Make sure user defined options are reflected in mount"
7594 error_and_umount() {
7605 mount --bind $DIR $TMP/$tdir || error "mount bind mnt pt failed"
7606 rm -f $TMP/$tdir/$tfile
7607 rm -f $TMP/$tdir/${tfile}1
7609 # Files should not be created in ro bind mount point
7610 # remounting from rw to ro
7611 mount -o remount,ro $TMP/$tdir ||
7612 error_and_umount "readonly remount of bind mnt pt failed"
7613 touch $TMP/$tdir/$tfile &&
7614 error_and_umount "touch succeeds on ro bind mnt pt"
7615 [ -e $TMP/$tdir/$tfile ] &&
7616 error_and_umount "file created on ro bind mnt pt"
7618 # Files should be created in rw bind mount point
7619 # remounting from ro to rw
7620 mount -o remount,rw $TMP/$tdir ||
7621 error_and_umount "read-write remount of bind mnt pt failed"
7622 touch $TMP/$tdir/${tfile}1 ||
7623 error_and_umount "touch fails on rw bind mnt pt"
7624 [ -e $TMP/$tdir/${tfile}1 ] ||
7625 error_and_umount "file not created on rw bind mnt pt"
7626 umount $TMP/$tdir || error "umount of bind mnt pt failed"
7628 cleanup || error "cleanup failed with $?"
7630 run_test 105 "check file creation for ro and rw bind mnt pt"
7637 mkdir -p $DIR/$tdir || error "create $tdir failed"
7638 lfs setstripe -c 1 -i 0 $DIR/$tdir
7639 #define OBD_FAIL_CAT_RECORDS 0x1312
7640 do_facet mds1 $LCTL set_param fail_loc=0x1312 fail_val=$repeat
7642 for ((i = 1; i <= $repeat; i++)); do
7644 #one full plain llog
7645 createmany -o $DIR/$tdir/f- 64768
7647 createmany -u $DIR/$tdir/f- 64768
7649 wait_delete_completed $((TIMEOUT * 7))
7650 #ASSERTION osp_sync_thread() ( thread->t_flags != SVC_RUNNING ) failed
7651 #shows that osp code is buggy
7652 do_facet mds1 $LCTL set_param fail_loc=0 fail_val=0
7656 run_test 106 "check osp llog processing when catalog is wrapped"
7659 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.10.50) ]] ||
7660 { skip "Need MDS version > 2.10.50"; return; }
7663 start_mgsmds || error "start_mgsmds failed"
7664 start_ost || error "unable to start OST"
7666 # add unknown configuration parameter.
7667 if [[ $PERM_CMD == *"set_param -P"* ]]; then
7668 cmd="$PERM_CMD ost.$FSNAME-OST0000*.unknown_param"
7670 cmd="$PERM_CMD $FSNAME-OST0000*.ost.unknown_param"
7672 do_facet mgs "$cmd=50"
7673 cleanup_nocli || error "cleanup_nocli failed with $?"
7676 # unknown param should be ignored while mounting.
7677 start_ost || error "unable to start OST after unknown param set"
7679 cleanup || error "cleanup failed with $?"
7681 run_test 107 "Unknown config param should not fail target mounting"
7686 $rcmd rm -rf $tmp > /dev/null 2>&1
7687 $rcmd mkdir -p $tmp/{mnt,images} || error "failed to mkdir remotely"
7689 for facet in $facets; do
7690 [ $(facet_fstype $SINGLEMDS) = "zfs" ] &&
7691 $rcmd $ZPOOL -f export lustre-$facet > /dev/null 2>&1
7692 $rcmd mkdir $tmp/mnt/$facet ||
7693 error "failed to mkdir $tmp/mnt/$facet"
7694 $rcmd dd if=/dev/zero of=$tmp/images/$facet \
7695 seek=199 bs=1M count=1 ||
7696 error "failed to create $tmp/images/$facet"
7705 local facet=${role}$((idx + 1))
7707 [ $# -eq 5 ] && pool=$5
7709 do_facet $SINGLEMDS $MKFS --fsname=lustre --$mgs \
7710 --$role --index=$idx --replace --backfstype=$bkfs \
7711 --device-size=200000 --reformat $pool $tmp/images/$facet ||
7712 error "failed to mkfs for $facet"
7716 echo "mounting client..."
7717 mount -t lustre ${nid}:/lustre $MOUNT ||
7718 error "failed to mount lustre"
7721 ls -l $MOUNT/local_dir || error "failed to list"
7723 echo "check truncate && write"
7724 echo "dummmmmmmmmmmmm" > $MOUNT/remote_dir/fsx.c ||
7725 error "failed to tuncate & write"
7728 touch $MOUNT/foooo ||
7729 error "failed to create"
7731 echo "check read && write && append"
7732 sha1sum $MOUNT/conf-sanity.sh |
7733 awk '{ print $1 }' > $MOUNT/checksum.new ||
7734 error "failed to read(1)"
7735 sha1sum $MOUNT/remote_dir/unlinkmany.c |
7736 awk '{ print $1 }' >> $MOUNT/checksum.new ||
7737 error "failed to read(2)"
7738 sha1sum $MOUNT/striped_dir/lockahead_test.o |
7739 awk '{ print $1 }' >> $MOUNT/checksum.new ||
7740 error "failed to read(3)"
7743 diff $MOUNT/checksum.new $MOUNT/checksum.src ||
7744 error "failed to verify data"
7754 umount -f $MOUNT || error "failed to umount client"
7755 for facet in $facets; do
7756 $rcmd umount -f $tmp/mnt/$facet ||
7757 error "failed to umount $facet"
7758 if [ $(facet_fstype $SINGLEMDS) = "zfs" ]; then
7759 $rcmd $ZPOOL export -f lustre-$facet ||
7760 error "failed to export lustre-$facet"
7764 $rcmd rm -rf $tmp || error "failed to rm the dir $tmp"
7768 [ "$CLIENTONLY" ] && skip "Client-only testing" && return
7770 [ $(facet_fstype $SINGLEMDS) != "zfs" ] &&
7771 skip "zfs only test" && return
7773 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.58) ] &&
7774 skip "Need server version at least 2.10.58" && return
7779 local tmp=$TMP/$tdir
7780 local rcmd="do_facet $SINGLEMDS"
7781 local facets="mdt1 mdt2 ost1 ost2"
7782 local nid=$($rcmd $LCTL list_nids | head -1)
7785 trap t_108_cleanup EXIT ERR
7788 t_108_mkfs mdt 0 zfs mgs lustre-mdt1/mdt1
7789 t_108_mkfs mdt 1 zfs mgsnode=$nid lustre-mdt2/mdt2
7790 t_108_mkfs ost 0 zfs mgsnode=$nid lustre-ost1/ost1
7791 t_108_mkfs ost 1 zfs mgsnode=$nid lustre-ost2/ost2
7793 for facet in $facets; do
7794 $rcmd zfs set mountpoint=$tmp/mnt/$facet canmount=on \
7795 lustre-$facet/$facet ||
7796 error "failed to zfs set for $facet (1)"
7797 $rcmd zfs mount lustre-$facet/$facet ||
7798 error "failed to local mount $facet"
7799 $rcmd tar jxf $LUSTRE/tests/ldiskfs_${facet}_2_11.tar.bz2 \
7800 --xattrs --xattrs-include="trusted.*" \
7801 -C $tmp/mnt/$facet/ > /dev/null 2>&1 ||
7802 error "failed to untar image for $facet"
7803 $rcmd "cd $tmp/mnt/$facet && rm -rf oi.* OI_* lfsck_* LFSCK" ||
7804 error "failed to cleanup for $facet"
7805 $rcmd zfs umount lustre-$facet/$facet ||
7806 error "failed to local umount $facet"
7807 $rcmd zfs set canmount=off lustre-$facet/$facet ||
7808 error "failed to zfs set $facet (2)"
7811 echo "changing server nid..."
7812 $rcmd mount -t lustre -o nosvc lustre-mdt1/mdt1 $tmp/mnt/mdt1
7813 $rcmd lctl replace_nids lustre-MDT0000 $nid
7814 $rcmd lctl replace_nids lustre-MDT0001 $nid
7815 $rcmd lctl replace_nids lustre-OST0000 $nid
7816 $rcmd lctl replace_nids lustre-OST0001 $nid
7817 $rcmd umount $tmp/mnt/mdt1
7819 for facet in $facets; do
7820 echo "mounting $facet from backup..."
7821 $rcmd mount -t lustre -o abort_recov lustre-$facet/$facet \
7822 $tmp/mnt/$facet || error "failed to mount $facet"
7825 # ZFS backend can detect migration and trigger OI scrub automatically
7826 # sleep 3 seconds for scrub done
7832 run_test 108a "migrate from ldiskfs to ZFS"
7835 [ "$CLIENTONLY" ] && skip "Client-only testing" && return
7837 [ $(facet_fstype $SINGLEMDS) != "ldiskfs" ] &&
7838 skip "ldiskfs only test" && return
7840 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.58) ] &&
7841 skip "Need server version at least 2.10.58" && return
7846 local tmp=$TMP/$tdir
7847 local rcmd="do_facet $SINGLEMDS"
7848 local facets="mdt1 mdt2 ost1 ost2"
7849 local scrub_list="MDT0000 MDT0001 OST0000 OST0001"
7850 local nid=$($rcmd $LCTL list_nids | head -1)
7853 trap t_108_cleanup EXIT ERR
7856 t_108_mkfs mdt 0 ldiskfs mgs
7857 t_108_mkfs mdt 1 ldiskfs mgsnode=$nid
7858 t_108_mkfs ost 0 ldiskfs mgsnode=$nid
7859 t_108_mkfs ost 1 ldiskfs mgsnode=$nid
7861 for facet in $facets; do
7862 $rcmd mount -t ldiskfs -o loop $tmp/images/$facet \
7864 error "failed to local mount $facet"
7866 $rcmd tar jxf $LUSTRE/tests/zfs_${facet}_2_11.tar.bz2 \
7867 --xattrs --xattrs-include="*.*" \
7868 -C $tmp/mnt/$facet/ > /dev/null 2>&1 ||
7869 error "failed to untar image for $facet"
7870 $rcmd "cd $tmp/mnt/$facet && rm -rf oi.* OI_* lfsck_* LFSCK" ||
7871 error "failed to cleanup for $facet"
7872 $rcmd umount $tmp/mnt/$facet ||
7873 error "failed to local umount $facet"
7876 echo "changing server nid..."
7877 $rcmd mount -t lustre -o nosvc,loop $tmp/images/mdt1 $tmp/mnt/mdt1
7878 $rcmd lctl replace_nids lustre-MDT0000 $nid
7879 $rcmd lctl replace_nids lustre-MDT0001 $nid
7880 $rcmd lctl replace_nids lustre-OST0000 $nid
7881 $rcmd lctl replace_nids lustre-OST0001 $nid
7882 $rcmd umount $tmp/mnt/mdt1
7884 for facet in $facets; do
7885 echo "mounting $facet from backup..."
7886 $rcmd mount -t lustre -o loop,abort_recov $tmp/images/$facet \
7887 $tmp/mnt/$facet || error "failed to mount $facet"
7890 for facet in $scrub_list; do
7891 $rcmd $LCTL lfsck_start -M lustre-$facet -t scrub ||
7892 error "failed to start OI scrub on $facet"
7895 # sleep 3 seconds for scrub done
7901 run_test 108b "migrate from ZFS to ldiskfs"
7905 # set number of permanent parameters
7907 test_109_set_params() {
7910 set_persistent_param_and_check mds \
7911 "mdd.$fsname-MDT0000.atime_diff" \
7912 "$fsname-MDT0000.mdd.atime_diff" \
7914 set_persistent_param_and_check mds \
7915 "mdd.$fsname-MDT0000.atime_diff" \
7916 "$fsname-MDT0000.mdd.atime_diff" \
7918 set_persistent_param_and_check client \
7919 "llite.$fsname*.max_read_ahead_mb" \
7920 "$fsname.llite.max_read_ahead_mb" \
7922 set_persistent_param_and_check client \
7923 "llite.$fsname*.max_read_ahead_mb" \
7924 "$fsname.llite.max_read_ahead_mb" \
7926 create_pool $fsname.pool1 || error "create pool failed"
7927 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
7928 error "pool_add failed"
7929 do_facet mgs $LCTL pool_remove $fsname.pool1 OST0000 ||
7930 error "pool_remove failed"
7931 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
7932 error "pool_add failed"
7936 # check permanent parameters
7938 test_109_test_params() {
7941 local atime_diff=$(do_facet mds $LCTL \
7942 get_param -n mdd.$fsname-MDT0000.atime_diff)
7943 [ $atime_diff == 63 ] || error "wrong mdd parameter after clear_conf"
7944 local max_read_ahead_mb=$(do_facet client $LCTL \
7945 get_param -n llite.$fsname*.max_read_ahead_mb)
7946 [ $max_read_ahead_mb == 64 ] ||
7947 error "wrong llite parameter after clear_conf"
7948 local ost_in_pool=$(do_facet mds $LCTL pool_list $fsname.pool1 |
7949 grep -v "^Pool:" | sed 's/_UUID//')
7950 [ $ost_in_pool = "$fsname-OST0000" ] ||
7951 error "wrong pool after clear_conf"
7955 # run lctl clear_conf, store CONFIGS before and after that
7957 test_109_clear_conf()
7959 local clear_conf_arg=$1
7962 if ! combined_mgs_mds ; then
7964 stop_mgs || error "stop_mgs failed"
7965 start_mgs "-o nosvc" || error "start_mgs nosvc failed"
7967 mgsdev=$(mdsdevname 1)
7968 start_mdt 1 "-o nosvc" || error "start_mdt 1 nosvc failed"
7971 do_facet mgs "rm -rf $TMP/${tdir}/conf1; mkdir -p $TMP/${tdir}/conf1;" \
7972 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf1\\\" \
7976 # the command being tested
7978 do_facet mgs $LCTL clear_conf $clear_conf_arg ||
7979 error "clear_conf failed"
7980 if ! combined_mgs_mds ; then
7981 stop_mgs || error "stop_mgs failed"
7983 stop_mdt 1 || error "stop_mdt 1 failed"
7986 do_facet mgs "rm -rf $TMP/${tdir}/conf2; mkdir -p $TMP/${tdir}/conf2;" \
7987 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf2\\\" \
7991 test_109_file_shortened() {
7993 local sizes=($(do_facet mgs "stat -c %s " \
7994 "$TMP/${tdir}/conf1/CONFIGS/$file" \
7995 "$TMP/${tdir}/conf2/CONFIGS/$file"))
7996 [ ${sizes[1]} -lt ${sizes[0]} ] && return 0
8002 [ "$(facet_fstype mgs)" == "zfs" ] &&
8003 skip "LU-8727: no implementation for ZFS" && return
8007 client_up || error "client_up failed"
8010 # set number of permanent parameters
8012 test_109_set_params $FSNAME
8014 umount_client $MOUNT || error "umount_client failed"
8015 stop_ost || error "stop_ost failed"
8016 stop_mds || error "stop_mds failed"
8018 test_109_clear_conf $FSNAME
8020 # make sure that all configs are cleared
8022 test_109_file_shortened $FSNAME-MDT0000 ||
8023 error "failed to clear MDT0000 config"
8024 test_109_file_shortened $FSNAME-client ||
8025 error "failed to clear client config"
8030 # check that configurations are intact
8032 test_109_test_params $FSNAME
8037 destroy_test_pools || error "destroy test pools failed"
8041 run_test 109a "test lctl clear_conf fsname"
8045 [ "$(facet_fstype mgs)" == "zfs" ] &&
8046 skip "LU-8727: no implementation for ZFS" && return
8050 client_up || error "client_up failed"
8053 # set number of permanent parameters
8055 test_109_set_params $FSNAME
8057 umount_client $MOUNT || error "umount_client failed"
8058 stop_ost || error "stop_ost failed"
8059 stop_mds || error "stop_mds failed"
8061 test_109_clear_conf $FSNAME-MDT0000
8063 # make sure that only one config is cleared
8065 test_109_file_shortened $FSNAME-MDT0000 ||
8066 error "failed to clear MDT0000 config"
8067 test_109_file_shortened $FSNAME-client &&
8068 error "failed to clear client config"
8073 # check that configurations are intact
8075 test_109_test_params $FSNAME
8080 destroy_test_pools || error "destroy test pools failed"
8084 run_test 109b "test lctl clear_conf one config"
8088 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ]] &&
8089 skip "Only applicable to ldiskfs-based MDTs"
8091 do_facet $SINGLEMDS $DEBUGFS -w -R supported_features |grep large_dir ||
8092 skip "large_dir option is not supported on MDS"
8093 do_facet ost1 $DEBUGFS -w -R supported_features | grep large_dir ||
8094 skip "large_dir option is not supported on OSS"
8096 stopall # stop all targets before modifying the target counts
8097 stack_trap "MDSCOUNT=$MDSCOUNT OSTCOUNT=$OSTCOUNT" EXIT
8101 # ext4_dir_entry_2 struct size:264
8102 # dx_root struct size:8
8103 # dx_node struct size:8
8104 # dx_entry struct size:8
8105 # For 1024 bytes block size.
8106 # First level directory entries: 126
8107 # Second level directory entries: 127
8108 # Entries in leaf: 3
8109 # For 2 levels limit: 48006
8110 # For 3 levels limit : 6096762
8111 # Create 80000 files to safely exceed 2-level htree limit.
8112 CONF_SANITY_110_LINKS=${CONF_SANITY_110_LINKS:-80000}
8114 # can fit at most 3 filenames per 1KB leaf block, but each
8115 # leaf/index block will only be 3/4 full before split at each level
8116 (( MDSSIZE < CONF_SANITY_110_LINKS / 3 * 4/3 * 4/3 )) &&
8117 CONF_SANITY_110_LINKS=$((MDSSIZE * 3 * 3/4 * 3/4))
8119 local opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
8120 --reformat $(mdsdevname 1) $(mdsvdevname 1)"
8121 if [[ $opts != *mkfsoptions* ]]; then
8122 opts+=" --mkfsoptions=\\\"-O large_dir -b 1024 -i 65536\\\""
8124 opts="${opts//--mkfsoptions=\\\"/ \
8125 --mkfsoptions=\\\"-O large_dir -b 1024 -i 65536 }"
8127 echo "MDT params: $opts"
8128 add mds1 $opts || error "add mds1 failed with new params"
8129 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
8131 opts="$(mkfs_opts ost1 $(ostdevname 1)) \
8132 --reformat $(ostdevname 1) $(ostvdevname 1)"
8134 if [[ $opts != *mkfsoptions* ]]; then
8135 opts+=" --mkfsoptions=\\\"-O large_dir\\\" "
8137 opts="${opts//--mkfsoptions=\\\"/ \
8138 --mkfsoptions=\\\"-O large_dir }"
8140 echo "OST params: $opts"
8141 add ost1 $opts || error "add ost1 failed with new params"
8142 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
8144 MOUNT_2=yes mountcli || error "mount clients failed"
8146 mkdir -v $DIR/$tdir || error "cannot create $DIR/$tdir"
8147 local pids count=0 group=0
8149 echo "creating $CONF_SANITY_110_LINKS in total"
8150 while (( count < CONF_SANITY_110_LINKS )); do
8151 local len=$((253 - $(wc -c <<<"$tfile-$group-40000-")))
8152 local dir=DIR$((group % 2 + 1))
8153 local target=${!dir}/$tdir/$tfile-$group
8154 local long=$target-$(generate_name $len)-
8155 local create=$((CONF_SANITY_110_LINKS - count))
8157 (( create > 40000 )) && create=40000
8158 touch $target || error "creating $target failed"
8159 echo "creating $create hard links to $target"
8160 createmany -l $target $long $create &
8163 count=$((count + create))
8164 group=$((group + 1))
8166 echo "waiting for PIDs$pids to complete"
8167 wait $pids || error "createmany failed after $group groups"
8171 run_e2fsck $(facet_active_host mds1) $(mdsdevname 1) -n
8173 run_test 110 "Adding large_dir with 3-level htree"
8176 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ]] &&
8177 skip "Only applicable to ldiskfs-based MDTs"
8179 is_dm_flakey_dev $SINGLEMDS $(mdsdevname 1) &&
8180 skip "This test can not be executed on flakey dev"
8182 do_facet $SINGLEMDS $DEBUGFS -w -R supported_features |grep large_dir ||
8183 skip "large_dir option is not supported on MDS"
8185 do_facet ost1 $DEBUGFS -w -R supported_features | grep large_dir ||
8186 skip "large_dir option is not supported on OSS"
8188 # cleanup before changing target counts
8190 stack_trap "MDSSIZE=$MDSSIZE MDSCOUNT=$MDSCOUNT OSTCOUNT=$OSTCOUNT" EXIT
8193 (( MDSSIZE < 2400000 )) && MDSSIZE=2400000 # need at least 2.4GB
8195 local mdsdev=$(mdsdevname 1)
8197 local opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
8198 --reformat $(mdsdevname 1) $(mdsvdevname 1)"
8199 if [[ $opts != *mkfsoptions* ]]; then
8200 opts+=" --mkfsoptions=\\\"-O large_dir -i 1048576 \\\" "
8202 opts="${opts//--mkfsoptions=\\\"/ \
8203 --mkfsoptions=\\\"-O large_dir -i 1048576 }"
8205 echo "MDT params: $opts"
8206 __touch_device mds 1
8207 add mds1 $opts || error "add mds1 failed with new params"
8208 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
8210 opts="$(mkfs_opts ost1 $(ostdevname 1)) \
8211 --reformat $(ostdevname 1) $(ostvdevname 1)"
8212 if [[ $opts != *mkfsoptions* ]]; then
8213 opts+=" --mkfsoptions=\\\"-O large_dir \\\""
8215 opts="${opts//--mkfsoptions=\\\"/ --mkfsoptions=\\\"-O large_dir }"
8217 echo "OST params: $opts"
8218 __touch_device ost 1
8219 add ost1 $opts || error "add ost1 failed with new params"
8220 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
8222 MOUNT_2=yes mountcli
8223 mkdir $DIR/$tdir || error "cannot create $DIR/$tdir"
8225 lfs df -i $DIR/$tdir
8229 local start=$SECONDS
8231 local dirmax=$((2 << 30))
8237 while (( !needskip & dirsize < dirmax )); do
8240 for cli in ${CLIENTS//,/ }; do
8241 local len=$((253 - $(wc -c <<<"$cli-$group-60000-")))
8242 local target=$cli-$group
8243 local long=$DIR/$tdir/$target-$(generate_name $len)-
8245 RPWD=$DIR/$tdir do_node $cli touch $target ||
8246 error "creating $target failed"
8247 echo "creating 60000 hardlinks to $target"
8248 RPWD=$DIR/$tdir do_node $cli createmany -l $target $long 60000 &
8251 group=$((group + 1))
8253 long=$DIR2/$tdir/$target-$(generate_name $len)-
8255 RPWD=$DIR2/$tdir do_node $cli touch $target ||
8256 error "creating $target failed"
8257 echo "creating 60000 hardlinks to $target"
8258 RPWD=$DIR2/$tdir do_node $cli createmany -l $target $long 60000 &
8261 group=$((group + 1))
8263 echo "waiting for PIDs$pids to complete"
8264 wait $pids || error "createmany failed after $group groups"
8265 dirsize=$(stat -c %s $DIR/$tdir)
8266 taken=$((SECONDS - start))
8267 rate=$((dirsize / taken))
8268 left=$(((dirmax - dirsize) / rate))
8269 num=$((group * 60000))
8270 echo "estimate ${left}s left after $num files / ${taken}s"
8271 # if the estimated time remaining is too large (it may change
8272 # over time as the create rate is not constant) then exit
8273 # without declaring a failure.
8274 (( left > 1200 )) && needskip=1
8279 (( $needskip )) && skip "ETA ${left}s after $num files / ${taken}s is too long"
8281 run_e2fsck $(facet_active_host mds1) $(mdsdevname 1) -n
8283 run_test 111 "Adding large_dir with over 2GB directory"
8290 rm -f $TMP/$tdir/lustre-mdt
8295 IMAGESIZE=$((3072 << 30)) # 3072 GiB
8297 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
8298 skip "Only applicable to ldiskfs-based MDTs"
8303 # We need MDT size 3072GB, because it is smallest
8304 # partition that can store 2B inodes
8305 do_facet $SINGLEMDS "mkdir -p $TMP/$tdir"
8306 local mdsimgname=$TMP/$tdir/lustre-mdt
8307 do_facet $SINGLEMDS "rm -f $mdsimgname"
8308 do_facet $SINGLEMDS "touch $mdsimgname"
8309 trap cleanup_115 RETURN EXIT
8310 do_facet $SINGLEMDS "$TRUNCATE $mdsimgname $IMAGESIZE" ||
8311 { skip "Backend FS doesn't support sparse files"; return 0; }
8312 local mdsdev=$(do_facet $SINGLEMDS "losetup -f")
8313 do_facet $SINGLEMDS "losetup $mdsdev $mdsimgname"
8315 local mds_opts="$(mkfs_opts mds1 ${mdsdev}) --device-size=$IMAGESIZE \
8316 --mkfsoptions='-O lazy_itable_init,ea_inode,^resize_inode,meta_bg \
8318 add mds1 $mds_opts --mgs --reformat $mdsdev ||
8319 { skip_env "format large MDT failed"; return 0; }
8320 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
8321 --reformat $(ostdevname 1) $(ostvdevname 1)
8323 start $SINGLEMDS ${mdsdev} $MDS_MOUNT_OPTS || error "start MDS failed"
8324 start_ost || error "start OSS failed"
8325 mount_client $MOUNT || error "mount client failed"
8327 mkdir -p $DIR/$tdir || error "mkdir $DIR/$tdir fail"
8328 for goal in $(do_facet $SINGLEMDS "ls /sys/fs/ldiskfs/*/inode_goal"); do
8329 do_facet $SINGLEMDS "echo 2147483947 >> $goal; grep . $goal"
8332 touch $DIR/$tdir/$tfile
8334 # Add > 5k bytes to xattr
8335 for i in {1..30}; do
8336 ln $DIR/$tdir/$tfile $DIR/$tdir/$(printf "link%0250d" $i) ||
8337 error "Can't make link"
8342 local inode_num=$(do_facet $SINGLEMDS \
8343 "$DEBUGFS -c -R 'stat ROOT/$tdir/$tfile' $mdsimgname" |
8344 awk '/link =/ { print $4 }' |
8345 sed -e 's/>//' -e 's/<//' -e 's/\"//')
8346 echo "inode num: $inode_num"
8347 [ $inode_num -ge 2147483947 ] || error "inode $inode_num too small"
8348 do_facet $SINGLEMDS "losetup -d $mdsdev"
8351 run_test 115 "Access large xattr with inodes number over 2TB"
8354 [ $(facet_fstype $SINGLEMDS) != "ldiskfs" ] &&
8355 skip "ldiskfs only test" && return
8357 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.59) ] &&
8358 skip "Need server version at least 2.10.59" && return
8360 do_facet $SINGLEMDS which mkfs.xfs || {
8361 skip_env "No mkfs.xfs installed"
8368 local tmpmnt=$TMP/$tdir
8369 local mdtimg=$tfile-mdt0
8371 do_facet $SINGLEMDS mkdir -p $tmpmnt
8372 stack_trap "do_facet $SINGLEMDS rmdir $tmpmnt" EXIT
8374 do_facet $SINGLEMDS touch $TMP/$mdtimg
8375 stack_trap "do_facet $SINGLEMDS rm -f $TMP/$mdtimg" EXIT
8376 do_facet $SINGLEMDS mkfs -t xfs -d file,size=1t,name=$TMP/$mdtimg ||
8377 error "mkfs temporary xfs image"
8379 do_facet $SINGLEMDS mount $TMP/$mdtimg $tmpmnt ||
8380 error "mount temporary xfs image"
8381 stack_trap "do_facet $SINGLEMDS umount $tmpmnt" EXIT
8382 local old_mdssize=$MDSSIZE
8383 local old_mdsisize=$MDSISIZE
8385 MDSSIZE=$((17 * 1024 * 1024 * 1024)) # 17T MDT
8386 MDSISIZE=$((16 << 20))
8387 local opts17t="$(mkfs_opts $SINGLEMDS)"
8389 MDSSIZE=$old_mdssize
8390 MDSISIZE=$old_mdsisize
8391 do_facet $SINGLEMDS $MKFS $opts17t $tmpmnt/$mdtimg ||
8392 error "failed to mkfs for $tmpmnt/$mdtimg"
8394 do_facet $SINGLEMDS $TUNE2FS -l $tmpmnt/$mdtimg |
8395 grep -qw 'features.*extent' || error "extent should be enabled"
8397 run_test 116 "big size MDT support"
8401 do_facet ost1 "$LCTL set_param ost.OSS.ost_io.nrs_policies=fifo"
8402 do_facet ost1 "$LCTL get_param -n ost.OSS.ost_io.nrs_tbf_rule" &&
8403 error "get_param should fail"
8404 cleanup || error "cleanup failed with rc $?"
8406 run_test 117 "lctl get_param return errors properly"
8408 test_120() { # LU-11130
8409 [ "$MDSCOUNT" -lt 2 ] && skip "mdt count < 2"
8410 [ $(facet_fstype $SINGLEMDS) != "ldiskfs" ] &&
8411 skip "ldiskfs only test"
8412 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.56) ] &&
8413 skip "Need DNE2 capable MD target with LU-11130 fix"
8417 local mds1host=$(facet_active_host mds1)
8418 local mds1dev=$(mdsdevname 1)
8420 $LFS mkdir -i 1 $DIR/$tdir
8421 $LFS mkdir -i 0 $DIR/$tdir/mds1dir
8423 ln -s foo $DIR/$tdir/bar
8424 mv $DIR/$tdir/bar $DIR/$tdir/mds1dir/bar2 ||
8425 error "cross-target rename failed"
8429 run_e2fsck $mds1host $mds1dev "-n"
8431 run_test 120 "cross-target rename should not create bad symlinks"
8434 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
8435 [[ $(lustre_version_code ost1) -ge $(version_code 2.11.53) ]] ||
8436 { skip "Need OST version at least 2.11.53" && return 0; }
8440 LOAD_MODULES_REMOTE=true load_modules
8441 #define OBD_FAIL_OFD_SET_OID 0x1e0
8442 do_facet ost1 $LCTL set_param fail_loc=0x00001e0
8445 $LFS mkdir -i1 -c1 $DIR/$tdir
8446 $LFS setstripe -i0 -c1 $DIR/$tdir
8447 do_facet ost1 $LCTL set_param fail_loc=0
8448 createmany -o $DIR/$tdir/file_ 1000 ||
8449 error "Fail to create a new sequence"
8453 run_test 122 "Check OST sequence update"
8457 local yaml_file="$TMP/$tfile.yaml"
8458 do_facet mgs rm "$yaml_file"
8459 local cfgfiles=$(do_facet mgs "lctl --device MGS llog_catlist |"\
8460 " sed 's/config_log://'")
8462 # set jobid_var to a different value for test
8463 local orig_val=$(do_facet mgs $LCTL get_param jobid_var)
8464 do_facet mgs $LCTL set_param -P jobid_var="testname"
8466 for i in $cfgfiles params; do
8467 do_facet mgs "lctl --device MGS llog_print ${i} >> $yaml_file"
8470 echo "Unmounting FS"
8480 # Reapply the config from before
8481 echo "Setting configuration parameters"
8482 do_facet mgs "lctl set_param -F $yaml_file"
8484 local set_val=$(do_facet mgs $LCTL get_param jobid_var)
8485 do_facet mgs $LCTL set_param -P $orig_val
8487 [ $set_val == "jobid_var=testname" ] ||
8488 error "$set_val is not testname"
8490 do_facet mgs rm "$yaml_file"
8492 run_test 123 "clear and reset all parameters using set_param -F"
8496 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
8497 [ -z $mds2failover_HOST ] && skip "needs MDT failover setup" && return
8503 if combined_mgs_mds; then
8504 start_mdt 1 "-o nosvc" ||
8505 error "starting mds with nosvc option failed"
8507 local nid=$(do_facet mds2 $LCTL list_nids | head -1)
8508 local failover_nid=$(do_node $mds2failover_HOST $LCTL list_nids | head -1)
8509 do_facet mgs $LCTL replace_nids $FSNAME-MDT0001 $nid:$failover_nid ||
8510 error "replace_nids execution error"
8512 if combined_mgs_mds; then
8518 echo "lfs setdirstripe"
8519 $LFS setdirstripe -i 1 $MOUNT/$tdir || error "setdirstirpe error"
8522 run_test 124 "check failover after replace_nids"
8524 if ! combined_mgs_mds ; then
8530 # restore the values of MDSSIZE and OSTSIZE
8531 MDSSIZE=$STORED_MDSSIZE
8532 OSTSIZE=$STORED_OSTSIZE
8536 check_and_cleanup_lustre