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/$NAME.sh}
35 # use small MDS + OST size to speed formatting time
36 # do not use too small MDSSIZE/OSTSIZE, which affect the default journal size
37 # STORED_MDSSIZE is used in test_18
38 STORED_MDSSIZE=$MDSSIZE
39 STORED_OSTSIZE=$OSTSIZE
41 [ $(facet_fstype $SINGLEMDS) = "zfs" ] && MDSSIZE=400000
43 [ $(facet_fstype ost1) = "zfs" ] && OSTSIZE=400000
53 # bug number for skipped test: LU-11915
54 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 110"
56 if ! combined_mgs_mds; then
57 # bug number for skipped test: LU-11991 LU-11990
58 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 32a 32b 32c 32d 32e 66"
59 # bug number for skipped test: LU-9897 LU-12032
60 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 84 123F"
63 # pass "-E lazy_itable_init" to mke2fs to speed up the formatting time
64 if [[ "$LDISKFS_MKFS_OPTS" != *lazy_itable_init* ]]; then
65 LDISKFS_MKFS_OPTS=$(csa_add "$LDISKFS_MKFS_OPTS" -E lazy_itable_init)
71 require_dsh_mds || exit 0
72 require_dsh_ost || exit 0
75 [ "$SLOW" = "no" ] && EXCEPT_SLOW="45 69"
80 # The MGS must be started before the OSTs for a new fs, so start
81 # and stop to generate the startup logs.
84 wait_osc_import_state mds ost FULL
89 reformat_and_config() {
91 if ! combined_mgs_mds ; then
97 writeconf_or_reformat() {
98 # There are at most 2 OSTs for write_conf test
99 # who knows if/where $TUNEFS is installed?
100 # Better reformat if it fails...
101 writeconf_all $MDSCOUNT 2 ||
102 { echo "tunefs failed, reformatting instead" &&
103 reformat_and_config && return 0; }
112 echo "start mgs service on $(facet_active_host mgs)"
113 start mgs $(mgsdevname) $MGS_MOUNT_OPTS $@
119 local dev=$(mdsdevname $num)
122 echo "start mds service on `facet_active_host $facet`"
123 start $facet ${dev} $MDS_MOUNT_OPTS $@ || return 94
129 local dev=$(mdsdevname $num)
132 echo "stop mds service on `facet_active_host $facet`"
133 # These tests all use non-failover stop
134 stop $facet -f || return 97
140 for num in $(seq $MDSCOUNT); do
141 start_mdt $num $@ || return 94
146 if ! combined_mgs_mds ; then
154 for num in $(seq $MDSCOUNT); do
155 stop_mdt $num || return 97
160 echo "stop mgs service on `facet_active_host mgs`"
161 # These tests all use non-failover stop
162 stop mgs -f || return 97
166 echo "start ost1 service on `facet_active_host ost1`"
167 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS $@ || return 95
171 echo "stop ost1 service on `facet_active_host ost1`"
172 # These tests all use non-failover stop
173 stop ost1 -f || return 98
177 echo "start ost2 service on `facet_active_host ost2`"
178 start ost2 $(ostdevname 2) $OST_MOUNT_OPTS $@ || return 92
182 echo "stop ost2 service on `facet_active_host ost2`"
183 # These tests all use non-failover stop
184 stop ost2 -f || return 93
189 echo "mount $FSNAME on ${MOUNTPATH}....."
190 zconf_mount $(hostname) $MOUNTPATH || return 96
194 local mountopt="remount,$1"
196 echo "remount '$1' lustre on ${MOUNTPATH}....."
197 zconf_mount $(hostname) $MOUNTPATH "$mountopt" || return 96
203 echo "umount lustre on $mountpath....."
204 zconf_umount $HOSTNAME $mountpath $@ || return 97
207 manual_umount_client(){
210 echo "manual umount lustre on ${MOUNT}...."
211 do_facet client "umount ${FORCE} $MOUNT"
217 start_mds || error "MDT start failed"
218 start_ost || error "Unable to start OST1"
219 mount_client $MOUNT || error "client start failed"
220 client_up || error "client_up failed"
229 unload_modules_conf () {
230 if combined_mgs_mds || ! local_mode; then
231 unload_modules || return 1
236 stop_ost || return 202
237 stop_mds || return 201
238 unload_modules_conf || return 203
243 [ "x$1" != "x" ] && force='-f'
244 umount_client $MOUNT $force|| return 200
245 cleanup_nocli || return $?
250 echo "umount $MOUNT2 ..."
251 umount $MOUNT2 || true
252 echo "stopping fs2mds ..."
253 stop fs2mds -f || true
254 echo "stopping fs2ost ..."
255 stop fs2ost -f || true
259 do_facet client "cp /etc/passwd $DIR/a" || return 71
260 do_facet client "rm $DIR/a" || return 72
261 # make sure lustre is actually mounted (touch will block,
262 # but grep won't, so do it after)
263 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
264 echo "setup single mount lustre success"
268 do_facet client "touch $DIR/a" || return 71
269 do_facet client "rm $DIR/a" || return 72
270 do_facet client "touch $DIR2/a" || return 73
271 do_facet client "rm $DIR2/a" || return 74
272 echo "setup double mount lustre success"
277 if [ "$ONLY" == "setup" ]; then
282 if [ "$ONLY" == "cleanup" ]; then
289 #create single point mountpoint
295 check_mount || error "check_mount failed"
296 cleanup || error "cleanup failed with $?"
298 run_test 0 "single mount setup"
301 start_mds || error "MDS start failed"
302 start_ost || error "unable to start OST"
303 echo "start ost second time..."
304 start_ost && error "2nd OST start should fail"
305 mount_client $MOUNT || error "client start failed"
306 check_mount || error "check_mount failed"
307 cleanup || error "cleanup failed with $?"
309 run_test 1 "start up ost twice (should return errors)"
312 start_mds || error "MDT start failed"
313 echo "start mds second time.."
314 start_mds && error "2nd MDT start should fail"
315 start_ost || error "OST start failed"
316 mount_client $MOUNT || error "mount_client failed to start client"
317 check_mount || error "check_mount failed"
318 cleanup || error "cleanup failed with $?"
320 run_test 2 "start up mds twice (should return err)"
324 #mount.lustre returns an error if already in mtab
325 mount_client $MOUNT && error "2nd client mount should fail"
326 check_mount || error "check_mount failed"
327 cleanup || error "cleanup failed with $?"
329 run_test 3 "mount client twice (should return err)"
333 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
334 stop_ost || error "Unable to stop OST1"
335 umount_client $MOUNT -f || error “unmount $MOUNT failed”
338 # ok for ost to fail shutdown
339 if [ 202 -ne $eno ] && [ 0 -ne $eno ]; then
340 error "cleanup failed with $?"
343 run_test 4 "force cleanup ost, then cleanup"
345 test_5a() { # was test_5
347 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
348 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
350 stop_mds || error "Unable to stop MDS"
352 # cleanup may return an error from the failed
353 # disconnects; for now I'll consider this successful
354 # if all the modules have unloaded.
358 echo "killing umount"
359 kill -TERM $UMOUNT_PID
360 echo "waiting for umount to finish"
362 if grep " $MOUNT " /proc/mounts; then
363 echo "test 5: /proc/mounts after failed umount"
367 echo "killing umount"
368 kill -TERM $UMOUNT_PID
369 echo "waiting for umount to finish"
371 grep " $MOUNT " /proc/mounts &&
372 error "/proc/mounts after second umount"
375 # manual_mount_client may fail due to umount succeeding above
377 # stop_mds is a no-op here, and should not fail
378 cleanup_nocli || error "cleanup_nocli failed with $?"
379 # df may have lingering entry
381 # mtab may have lingering entry
385 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
387 grep -q $MOUNT" " /etc/mtab || break
388 echo "Waiting /etc/mtab updated ... "
389 WAIT=$(( WAIT + sleep))
391 [ "$WAIT" -eq "$MAX_WAIT" ] &&
392 error "/etc/mtab is not updated in $WAIT secs"
393 echo "/etc/mtab updated in $WAIT secs"
395 run_test 5a "force cleanup mds, then cleanup"
403 grep " $MOUNT " /etc/mtab &&
404 error false "unexpected entry in mtab before mount" && return 10
406 start_ost || error "OST start failed"
407 if ! combined_mgs_mds ; then
408 trap cleanup_5b EXIT ERR
409 start_mds || error "MDS start failed"
413 mount_client $MOUNT && error "mount_client $MOUNT should fail"
414 grep " $MOUNT " /etc/mtab &&
415 error "$MOUNT entry in mtab after failed mount"
417 # stop_mds is a no-op here, and should not fail
418 cleanup_nocli || error "cleanup_nocli failed with $?"
419 if ! combined_mgs_mds ; then
423 run_test 5b "Try to start a client with no MGS (should return errs)"
426 grep " $MOUNT " /etc/mtab &&
427 error false "unexpected entry in mtab before mount" && return 10
429 start_mds || error "MDS start failed"
430 start_ost || error "OST start failed"
431 local oldfs="${FSNAME}"
432 FSNAME="wrong.${FSNAME}"
433 mount_client $MOUNT || :
435 grep " $MOUNT " /etc/mtab &&
436 error "$MOUNT entry in mtab after failed mount"
438 cleanup_nocli || error "cleanup_nocli failed with $?"
440 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
443 grep " $MOUNT " /etc/mtab &&
444 error "unexpected entry in mtab before mount"
446 start_ost || error "OST start failed"
447 start_mds || error "MDS start failed"
448 stop_ost -f || error "Unable to stop OST1"
449 mount_client $MOUNT || error "mount_client $MOUNT failed"
450 umount_client $MOUNT -f || error "umount_client $MOUNT failed"
451 cleanup_nocli || error "cleanup_nocli failed with $?"
452 ! grep " $MOUNT " /etc/mtab ||
453 error "$MOUNT entry in mtab after unmount"
455 run_test 5d "mount with ost down"
458 grep " $MOUNT " /etc/mtab &&
459 error false "unexpected entry in mtab before mount" && return 10
461 start_mds || error "MDS start failed"
462 start_ost || error "OST start failed"
464 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
465 do_facet client "$LCTL set_param fail_loc=0x80000506"
466 mount_client $MOUNT || echo "mount failed (not fatal)"
467 cleanup || error "cleanup failed with $?"
468 grep " $MOUNT " /etc/mtab &&
469 error "$MOUNT entry in mtab after unmount"
472 run_test 5e "delayed connect, don't crash (bug 10268)"
475 combined_mgs_mds && skip "needs separate mgs and mds"
477 grep " $MOUNT " /etc/mtab &&
478 error false "unexpected entry in mtab before mount" && return 10
481 start_ost || error "OST start failed"
482 mount_client $MOUNT &
484 echo client_mount pid is $pid
488 if ! ps -f -p $pid >/dev/null; then
491 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
492 error "mount returns $rc, expected to hang"
494 cleanup || error "cleanup failed with $?"
499 start_mds || error "start MDS failed"
501 # mount should succeed after start mds
503 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
504 cleanup || error "final call to cleanup failed with rc $?"
506 run_test 5f "mds down, cleanup after failed mount (bug 2712)"
510 [ $(lustre_version_code client) -lt $(version_code 2.9.53) ] &&
511 skip "automount of debugfs missing before 2.9.53"
512 umount /sys/kernel/debug
513 $LCTL get_param -n devices | egrep -v "error" && \
514 error "lctl can't access debugfs data"
515 grep " debugfs " /etc/mtab || error "debugfs failed to remount"
517 run_test 5g "handle missing debugfs"
522 mount_client $MOUNT || error "mount_client $MOUNT failed"
523 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
524 cleanup || error "cleanup failed with rc $?"
526 run_test 6 "manual umount, then mount again"
531 cleanup_nocli || error "cleanup_nocli failed with $?"
533 run_test 7 "manual umount, then cleanup"
537 mount_client $MOUNT2 || error "mount_client $MOUNT2 failed"
538 check_mount2 || error "check_mount2 failed"
539 umount_client $MOUNT2 || error "umount_client $MOUNT2 failed"
540 cleanup || error "cleanup failed with rc $?"
542 run_test 8 "double mount setup"
545 start_ost || error "OST start failed"
547 do_facet ost1 $LCTL set_param debug=\'inode trace\' ||
548 error "do_facet ost1 set_param inode trace failed."
549 do_facet ost1 $LCTL set_param subsystem_debug=\'mds ost\' ||
550 error "do_facet ost1 set_param debug mds ost failed."
552 CHECK_PTLDEBUG="`do_facet ost1 $LCTL get_param -n debug`"
553 if [ "$CHECK_PTLDEBUG" ] && { \
554 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
555 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
556 echo "lnet.debug success"
558 error "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
560 CHECK_SUBSYS="`do_facet ost1 $LCTL get_param -n subsystem_debug`"
561 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
562 echo "lnet.subsystem_debug success"
564 error "lnet.subsystem_debug: want 'mds ost' got '$CHECK_SUBSYS'"
566 stop_ost || error "Unable to stop OST1"
568 run_test 9 "test ptldebug and subsystem for mkfs"
571 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
572 # But with new MDS stack we don't care about the mode of local objects
573 # anymore, so this test is removed. See bug 22944 for more details.
577 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
578 skip "ldiskfs only test"
582 check_mount || error "check_mount failed"
583 cleanup || error "cleanup failed with rc $?"
585 echo "Remove mds config log"
586 if ! combined_mgs_mds ; then
590 do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' \
591 $(mgsdevname) || return \$?" ||
592 error "do_facet mgs failed with $?"
594 if ! combined_mgs_mds ; then
598 start_ost || error "OST start failed"
599 start_mds && error "MDS start succeeded, but should fail"
602 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
605 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
606 skip "ldiskfs only test"
609 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
614 # check if current MDSSIZE is large enough
615 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE &&
616 log "use MDSSIZE=$MDSSIZE"
618 # check if the global config has a large enough MDSSIZE
619 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] &&
620 [ $STORED_MDSSIZE -ge $MIN ] &&
621 OK=1 && myMDSSIZE=$STORED_MDSSIZE &&
622 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
624 # check if the block device is large enough
625 is_blkdev $SINGLEMDS $MDSDEV $MIN
626 local large_enough=$?
627 if [ -n "$OK" ]; then
628 [ $large_enough -ne 0 ] && OK=""
630 [ $large_enough -eq 0 ] && OK=1 && myMDSSIZE=$MIN &&
631 log "use device $MDSDEV with MIN=$MIN"
634 # check if a loopback device has enough space for fs metadata (5%)
636 if [ -z "$OK" ]; then
637 local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! \
638 -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
639 awk '($1 != "Filesystem") { print $4 }')
640 ! [ -z "$SPACE" ] && [ $SPACE -gt $((MIN / 20)) ] &&
641 OK=1 && myMDSSIZE=$MIN &&
642 log "use file $MDSDEV with MIN=$MIN"
645 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS"
647 echo "mount mds with large journal..."
649 local OLD_MDSSIZE=$MDSSIZE
653 echo "mount lustre system..."
655 check_mount || error "check_mount failed"
657 echo "check journal size..."
658 local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
659 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
660 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
662 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
665 cleanup || error "cleanup failed with rc $?"
670 run_test 18 "check mkfs creates large journals"
673 start_mds || error "MDS start failed"
674 stop_mds || error "Unable to stop MDS"
676 run_test 19a "start/stop MDS without OSTs"
679 start_ost || error "Unable to start OST1"
680 stop_ost -f || error "Unable to stop OST1"
682 run_test 19b "start/stop OSTs without MDS"
685 # first format the ost/mdt
686 start_mds || error "MDS start failed"
687 start_ost || error "Unable to start OST1"
688 mount_client $MOUNT || error "mount_client $MOUNT failed"
689 check_mount || error "check_mount failed"
690 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed."
691 remount_client ro $MOUNT || error "remount_client with ro failed"
692 touch $DIR/$tfile && error "$DIR/$tfile created incorrectly"
693 [ -e $DIR/$tfile ] && error "$DIR/$tfile exists incorrectly"
694 remount_client rw $MOUNT || error "remount_client with rw failed"
695 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
696 MCNT=$(grep -c $MOUNT' ' /etc/mtab)
697 [ "$MCNT" -ne 1 ] && error "$MOUNT in /etc/mtab $MCNT times"
699 stop_mds || error "Unable to stop MDS"
700 stop_ost || error "Unable to stop OST1"
702 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
705 start_mds || error "MDS start failed"
706 start_ost || error "unable to start OST1"
707 wait_osc_import_state mds ost FULL
708 stop_ost || error "unable to stop OST1"
709 stop_mds || error "unable to stop MDS"
711 run_test 21a "start mds before ost, stop ost first"
714 start_ost || error "unable to start OST1"
715 start_mds || error "MDS start failed"
716 wait_osc_import_state mds ost FULL
717 stop_mds || error "unable to stop MDS"
718 stop_ost || error "unable to stop OST1"
720 run_test 21b "start ost before mds, stop mds first"
723 start_ost || error "Unable to start OST1"
724 start_mds || error "MDS start failed"
725 start_ost2 || error "Unable to start OST2"
726 wait_osc_import_state mds ost2 FULL
727 stop_ost || error "Unable to stop OST1"
728 stop_ost2 || error "Unable to stop OST2"
729 stop_mds || error "Unable to stop MDS"
730 #writeconf to remove all ost2 traces for subsequent tests
731 writeconf_or_reformat
733 run_test 21c "start mds between two osts, stop mds last"
736 combined_mgs_mds && skip "need separate mgs device"
741 start_mgs || error "unable to start MGS"
742 start_ost || error "unable to start OST1"
743 start_ost2 || error "unable to start OST2"
744 start_mds || error "MDS start failed"
745 wait_osc_import_state mds ost2 FULL
747 stop_ost || error "Unable to stop OST1"
748 stop_ost2 || error "Unable to stop OST2"
749 stop_mds || error "Unable to stop MDS"
751 #writeconf to remove all ost2 traces for subsequent tests
752 writeconf_or_reformat
753 start_mgs || error "unable to start MGS"
755 run_test 21d "start mgs then ost and then mds"
758 MGSNID="$saved_mgsnid"
760 echo "stopping fs2mgs ..."
761 stop $fs2mgs -f || true
764 test_21e() { # LU-5863
765 if [[ -z "$fs3ost_DEV" || -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
766 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
767 skip_env "mixed loopback and real device not working"
770 local fs2mdsdev=$(mdsdevname 1_2)
771 local fs2ostdev=$(ostdevname 1_2)
772 local fs3ostdev=$(ostdevname 2_2)
774 local fs2mdsvdev=$(mdsvdevname 1_2)
775 local fs2ostvdev=$(ostvdevname 1_2)
776 local fs3ostvdev=$(ostvdevname 2_2)
778 # temporarily use fs3ost as fs2mgs
780 local fs2mgsdev=$fs3ostdev
781 local fs2mgsvdev=$fs3ostvdev
783 local fsname=test1234
785 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
786 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
787 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS && trap cleanup_21e EXIT INT ||
788 error "start fs2mgs failed"
790 local saved_mgsnid="$MGSNID"
791 MGSNID=$(do_facet $fs2mgs $LCTL list_nids | xargs | tr ' ' ,)
793 add fs2mds $(mkfs_opts mds1 $fs2mdsdev $fsname) \
794 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
795 add fs2ost $(mkfs_opts ost1 $fs2ostdev $fsname) \
796 --reformat $fs2ostdev $fs2ostvdev || error "add fs2ost failed"
798 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
799 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS || error "start fs2mds failed"
801 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
802 $MOUNT_CMD $MGSNID:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
803 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
807 run_test 21e "separate MGS and MDS"
810 start_mds || error "MDS start failed"
812 echo "Client mount with ost in logs, but none running"
813 start_ost || error "unable to start OST1"
814 # wait until mds connected to ost and open client connection
815 wait_osc_import_state mds ost FULL
816 stop_ost || error "unable to stop OST1"
817 mount_client $MOUNT || error "mount_client $MOUNT failed"
818 # check_mount will block trying to contact ost
819 mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed"
820 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed"
821 umount_client $MOUNT -f
824 echo "Client mount with a running ost"
825 start_ost || error "unable to start OST1"
827 # if gss enabled, wait full time to let connection from
828 # mds to ost be established, due to the mismatch between
829 # initial connect timeout and gss context negotiation timeout.
830 # This perhaps could be remove after AT landed.
831 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
832 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
834 mount_client $MOUNT || error "mount_client $MOUNT failed"
835 wait_osc_import_state mds ost FULL
836 wait_osc_import_ready client ost
837 check_mount || error "check_mount failed"
840 cleanup || error "cleanup failed with rc $?"
842 run_test 22 "start a client before osts (should return errs)"
844 test_23a() { # was test_23
847 stop $SINGLEMDS || error "failed to stop $SINGLEMDS"
848 # force down client so that recovering mds waits for reconnect
849 local running=$(grep -c $MOUNT /proc/mounts) || true
850 if [ $running -ne 0 ]; then
851 echo "Stopping client $MOUNT (opts: -f)"
855 # enter recovery on failed mds
856 local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/})
857 start $SINGLEMDS $MDT_DEV $MDS_MOUNT_OPTS || error "MDS start failed"
858 # try to start a new client
859 mount_client $MOUNT &
861 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
862 MOUNT_LUSTRE_PID=$(ps -ef | grep mount.lustre |
863 grep -v grep | awk '{print $2}')
864 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
866 ps --ppid $MOUNT_LUSTRE_PID
867 echo "waiting for mount to finish"
869 # "ctrl-c" sends SIGINT but it usually (in script) does not work on child process
870 # SIGTERM works but it does not spread to offspring processses
871 kill -s TERM $MOUNT_PID
872 kill -s TERM $MOUNT_LUSTRE_PID
873 # we can not wait $MOUNT_PID because it is not a child of this shell
879 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
881 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
882 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
885 [ -z "$PID1" -a -z "$PID2" ] && break
886 echo "waiting for mount to finish ... "
887 WAIT=$(( WAIT + sleep))
889 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
890 error "MOUNT_PID $MOUNT_PID and "\
891 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
894 cleanup || error "cleanup failed with rc $?"
896 run_test 23a "interrupt client during recovery mount delay"
898 test_23b() { # was test_23
899 start_mds || error "MDS start failed"
900 start_ost || error "Unable to start OST1"
901 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
902 $LCTL set_param fail_loc=0x80000313
904 cleanup || error "cleanup failed with rc $?"
906 run_test 23b "Simulate -EINTR during mount"
909 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
911 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
912 is_blkdev $SINGLEMDS $MDSDEV &&
913 skip_env "mixed loopback and real device not working"
916 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
918 local fs2mdsdev=$(mdsdevname 1_2)
919 local fs2ostdev=$(ostdevname 1_2)
920 local fs2mdsvdev=$(mdsvdevname 1_2)
921 local fs2ostvdev=$(ostvdevname 1_2)
924 # LU-9733 test fsname started with numbers as well
925 local FSNAME2=969362ae
927 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --nomgs --mgsnode=$MGSNID \
928 --fsname=${FSNAME2} --reformat $fs2mdsdev $fs2mdsvdev || exit 10
930 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --fsname=${FSNAME2} \
931 --reformat $fs2ostdev $fs2ostvdev || exit 10
934 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
935 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
936 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
937 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
939 # LU-9733 test fsname started with numbers
940 cl_user=$(do_facet $SINGLEMDS lctl --device $FSNAME2-MDT0000 \
941 changelog_register -n) ||
942 error "register changelog failed"
944 do_facet $SINGLEMDS lctl --device $FSNAME2-MDT0000 \
945 changelog_deregister $cl_user ||
946 error "deregister changelog failed"
948 check_mount || error "check_mount failed"
949 # files written on 1 should not show up on 2
950 cp /etc/passwd $DIR/$tfile
952 [ -e $MOUNT2/$tfile ] && error "File bleed"
955 cp /etc/passwd $MOUNT2/$tfile ||
956 error "cp /etc/passwd $MOUNT2/$tfile failed"
957 rm $MOUNT2/$tfile || error "remove $MOUNT2/$tfile failed"
958 # 2 is actually mounted
959 grep $MOUNT2' ' /proc/mounts > /dev/null || error "$MOUNT2 not mounted"
961 facet_failover fs2mds
962 facet_failover fs2ost
965 # the MDS must remain up until last MDT
967 MDS=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
968 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
969 [ -z "$MDS" ] && error "No MDT"
971 cleanup_nocli || error "cleanup_nocli failed with rc $?"
973 run_test 24a "Multiple MDTs on a single node"
976 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
978 skip "needs combined MGT and MDT device"
980 if [ -z "$fs2mds_DEV" ]; then
981 local dev=${SINGLEMDS}_dev
983 is_blkdev $SINGLEMDS $MDSDEV &&
984 skip_env "mixed loopback and real device not working"
987 local fs2mdsdev=$(mdsdevname 1_2)
988 local fs2mdsvdev=$(mdsvdevname 1_2)
990 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --mgs --fsname=${FSNAME}2 \
991 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
993 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && {
995 error "start MDS should fail"
998 cleanup || error "cleanup failed with rc $?"
1000 run_test 24b "Multiple MGSs on a single node (should return err)"
1004 check_mount || error "check_mount failed"
1005 local MODULES=$($LCTL modules | awk '{ print $2 }')
1006 rmmod $MODULES 2>/dev/null || true
1007 cleanup || error "cleanup failed with $?"
1009 run_test 25 "Verify modules are referenced"
1013 # we need modules before mount for sysctl, so make sure...
1014 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
1015 #define OBD_FAIL_MDS_FS_SETUP 0x135
1016 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000135"
1017 start_mds && error "MDS started but should not have started"
1018 $LCTL get_param -n devices
1019 DEVS=$($LCTL get_param -n devices | egrep -v MG | wc -l)
1020 [ $DEVS -gt 0 ] && error "number of devices is $DEVS, should be zero"
1021 # start mds to drop writeconf setting
1022 start_mds || error "Unable to start MDS"
1023 stop_mds || error "Unable to stop MDS"
1024 unload_modules_conf || error "unload_modules_conf failed with $?"
1026 run_test 26 "MDT startup failure cleans LOV (should return errs)"
1030 start_ost || error "Unable to start OST1"
1031 start_mds || error "Unable to start MDS"
1032 echo "Requeue thread should have started: "
1033 ps -e | grep ll_cfg_requeue
1034 set_persistent_param_and_check ost1 \
1035 "obdfilter.$FSNAME-OST0000.client_cache_seconds" \
1036 "$FSNAME-OST0000.ost.client_cache_seconds"
1037 cleanup_nocli || error "cleanup_nocli failed with rc $?"
1039 run_test 27a "Reacquire MGS lock if OST started first"
1044 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
1045 awk '($3 ~ "mdt" && $4 ~ "MDT0000") { print $4 }')
1047 facet_failover $SINGLEMDS
1048 set_persistent_param_and_check $SINGLEMDS \
1049 "mdt.$device.identity_acquire_expire" \
1050 "$device.mdt.identity_acquire_expire"
1051 set_persistent_param_and_check client \
1052 "mdc.$device-mdc-*.max_rpcs_in_flight" \
1053 "$device.mdc.max_rpcs_in_flight"
1055 cleanup || error "cleanup failed with $?"
1057 run_test 27b "Reacquire MGS lock after failover"
1059 test_28A() { # was test_28
1062 local TEST="llite.$FSNAME-*.max_read_ahead_whole_mb"
1063 local PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
1064 local orig=$($LCTL get_param -n $TEST)
1065 local max=$($LCTL get_param -n \
1066 llite.$FSNAME-*.max_read_ahead_per_file_mb)
1068 orig=${orig%%.[0-9]*}
1070 echo "ORIG:$orig MAX:$max"
1071 [[ $max -le $orig ]] && orig=$((max - 3))
1072 echo "ORIG:$orig MAX:$max"
1074 local final=$((orig + 1))
1076 set_persistent_param_and_check client "$TEST" "$PARAM" $final
1077 final=$((final + 1))
1078 set_persistent_param_and_check client "$TEST" "$PARAM" $final
1079 umount_client $MOUNT || error "umount_client $MOUNT failed"
1080 mount_client $MOUNT || error "mount_client $MOUNT failed"
1082 local result=$($LCTL get_param -n $TEST)
1084 if [ $result -ne $final ]; then
1085 error "New config not seen: wanted $final got $result"
1087 echo "New config success: got $result"
1089 set_persistent_param_and_check client "$TEST" "$PARAM" $orig
1090 cleanup || error "cleanup failed with rc $?"
1092 run_test 28A "permanent parameter setting"
1094 test_28a() { # LU-4221
1095 [[ $(lustre_version_code ost1) -ge $(version_code 2.5.52) ]] ||
1096 skip "Need OST version at least 2.5.52"
1097 [ "$(facet_fstype ost1)" = "zfs" ] &&
1098 skip "LU-4221: no such proc params for ZFS OSTs"
1105 local device="$FSNAME-OST0000"
1109 # In this test we will set three kinds of proc parameters with
1110 # lctl set_param -P or lctl conf_param:
1111 # 1. non-symlink ones in the OFD
1112 # 2. non-symlink ones in the OSD
1115 # prepare a non-symlink parameter in the OFD
1116 name="client_cache_seconds"
1117 param="$device.ost.$name"
1118 cmd="obdfilter.$device.$name"
1120 # permanently setting the parameter in the OFD
1121 old=$(do_facet ost1 $LCTL get_param -n $cmd)
1123 set_persistent_param_and_check ost1 "$cmd" "$param" $new
1124 set_persistent_param_and_check ost1 "$cmd" "$param" $old
1127 # prepare a non-symlink parameter in the OSD
1129 param="$device.osd.$name"
1130 cmd="osd-*.$device.$name"
1132 # conf_param the parameter in the OSD
1133 old=$(do_facet ost1 $LCTL get_param -n $cmd)
1134 new=$(((old + 1) % 2))
1135 set_persistent_param_and_check ost1 "$cmd" "$param" $new
1136 set_persistent_param_and_check ost1 "$cmd" "$param" $old
1138 cleanup || error "cleanup failed with $?"
1140 run_test 28a "set symlink parameters permanently with lctl"
1143 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >= 2 OSTs"
1144 setup_noconfig > /dev/null 2>&1
1145 start_ost2 || error "Unable to start OST2"
1148 local PARAM="$FSNAME-OST0001.osc.active"
1149 # With lctl set_param -P the value $PROC_ACT will be sent to
1150 # all nodes. The [^M] filter out the ability to set active
1151 # on the MDS servers which is tested with wait_osp_* below.
1152 # For ost_server_uuid that only exist on client so filtering
1154 local PROC_ACT="os[cp].$FSNAME-OST0001-osc-[^M]*.active"
1155 local PROC_UUID="os[cp].$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
1157 ACTV=$($LCTL get_param -n $PROC_ACT)
1159 set_persistent_param_and_check client $PROC_ACT $PARAM $DEAC
1160 # also check ost_server_uuid status
1161 RESULT=$($LCTL get_param -n $PROC_UUID | grep DEACTIV)
1162 if [ -z "$RESULT" ]; then
1163 error "Client not deactivated: $($LCTL get_param \
1166 echo "Live client success: got $RESULT"
1170 wait_osp_active ost ${FSNAME}-OST0001 1 0
1172 # test new client starts deactivated
1173 umount_client $MOUNT || error "umount_client $MOUNT failed"
1174 mount_client $MOUNT || error "mount_client $MOUNT failed"
1175 RESULT=$($LCTL get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
1176 if [ -z "$RESULT" ]; then
1177 error "New client start active: $(lctl get_param -n $PROC_UUID)"
1179 echo "New client success: got $RESULT"
1182 # make sure it reactivates
1183 set_persistent_param_and_check client $PROC_ACT $PARAM $ACTV
1185 umount_client $MOUNT
1186 stop_ost2 || error "Unable to stop OST2"
1187 cleanup_nocli || error "cleanup_nocli failed with $?"
1188 #writeconf to remove all ost2 traces for subsequent tests
1189 writeconf_or_reformat
1191 run_test 29 "permanently remove an OST"
1196 echo Big config llog
1197 TEST="llite.$FSNAME-*.max_read_ahead_whole_mb"
1198 ORIG=$($LCTL get_param -n $TEST)
1199 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1200 for i in ${LIST[@]}; do
1201 set_persistent_param_and_check client "$TEST" \
1202 "$FSNAME.llite.max_read_ahead_whole_mb" $i
1204 # make sure client restart still works
1205 umount_client $MOUNT
1206 mount_client $MOUNT || error "mount_client $MOUNT failed"
1207 [ "$($LCTL get_param -n $TEST)" -ne "$i" ] &&
1208 error "Param didn't stick across restart $($TEST) != $i"
1211 echo Erase parameter setting
1212 if [[ $PERM_CMD = *"set_param -P"* ]]; then
1213 do_facet mgs "$PERM_CMD -d $TEST" ||
1214 error "Erase param $TEST failed"
1216 do_facet mgs "$PERM_CMD \
1217 -d $FSNAME.llite.max_read_ahead_whole_mb" ||
1218 error "Erase param $FSNAME.llite.max_read_ahead_whole_mb failed"
1220 umount_client $MOUNT
1221 mount_client $MOUNT || error "mount_client $MOUNT failed"
1222 FINAL=$($LCTL get_param -n $TEST)
1223 echo "deleted (default) value=$FINAL, orig=$ORIG"
1224 ORIG=${ORIG%%.[0-9]*}
1225 FINAL=${FINAL%%.[0-9]*}
1226 # assumes this parameter started at the default value
1227 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
1229 cleanup || error "cleanup failed with rc $?"
1231 run_test 30a "Big config llog and permanent parameter deletion"
1236 local orignids=$($LCTL get_param -n \
1237 osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1239 local orignidcount=$(echo "$orignids" | wc -w)
1241 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1242 # numerical part of it. Hopefully that's not already a failover address
1244 local OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | \
1246 local ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1247 local NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1248 local NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1249 echo "Using fake nid $NEW"
1251 local TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1252 grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1253 if [[ $PERM_CMD = *"set_param -P"* ]]; then
1254 PARAM="osc.$FSNAME-OST0000-osc-[^M]*.import"
1255 echo "Setting $PARAM from $TEST to $NEW"
1256 do_facet mgs "$PERM_CMD $PARAM='connection=$NEW'" ||
1257 error "$PERM_CMD $PARAM failed"
1259 PARAM="$FSNAME-OST0000.failover.node"
1260 echo "Setting $PARAM from $TEST to $NEW"
1261 do_facet mgs "$PERM_CMD $PARAM='$NEW'" ||
1262 error "$PARAM $PARAM failed"
1264 wait_update_facet client "$TEST" "$NEW" ||
1265 error "check $PARAM failed!"
1267 local NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1269 local NIDCOUNT=$(echo "$NIDS" | wc -w)
1270 echo "should have $((orignidcount + 1)) entries \
1271 in failover nids string, have $NIDCOUNT"
1272 [ $NIDCOUNT -eq $((orignidcount + 1)) ] ||
1273 error "Failover nid not added"
1275 if [[ $PERM_CMD = *"set_param -P"* ]]; then
1276 do_facet mgs "$PERM_CMD -d osc.$FSNAME-OST0000-osc-*.import"
1278 do_facet mgs "$PERM_CMD -d $FSNAME-OST0000.failover.node" ||
1279 error "$PERM_CMD delete failed"
1281 umount_client $MOUNT
1282 mount_client $MOUNT || error "mount_client $MOUNT failed"
1284 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1286 NIDCOUNT=$(echo "$NIDS" | wc -w)
1287 echo "only $orignidcount final entries should remain \
1288 in failover nids string, have $NIDCOUNT"
1289 [ $NIDCOUNT -eq $orignidcount ] || error "Failover nids not removed"
1291 cleanup || error "cleanup failed with rc $?"
1293 run_test 30b "Remove failover nids"
1295 test_31() { # bug 10734
1296 # ipaddr must not exist
1297 $MOUNT_CMD 4.3.2.1@tcp:/lustre $MOUNT || true
1298 cleanup || error "cleanup failed with rc $?"
1300 run_test 31 "Connect to non-existent node (shouldn't crash)"
1304 T32_BLIMIT=40960 # Kbytes
1308 # This is not really a test but a tool to create new disk
1309 # image tarballs for the upgrade tests.
1311 # Disk image tarballs should be created on single-node
1312 # clusters by running this test with default configurations
1313 # plus a few mandatory environment settings that are verified
1314 # at the beginning of the test.
1316 test_32newtarball() {
1320 local tmp=$TMP/t32_image_create
1321 local server_version=$(lustre_version_code $SINGLEMDS)
1326 if [ $FSNAME != t32fs -o \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o \
1327 $OSTCOUNT -ne 1 -o -z "$OSTDEV1" ]; then
1328 error "Needs FSNAME=t32fs MDSCOUNT=2 " \
1329 "MDSDEV1=<nonexistent_file>" \
1330 "MDSDEV2=<nonexistent_file>" \
1331 "(or MDSDEV, in the case of b1_8)" \
1332 "OSTCOUNT=1 OSTDEV1=<nonexistent_file>"
1336 echo "Found stale $tmp"
1340 mkdir $tmp/src || return 1
1341 tar cf - -C $src . | tar xf - -C $tmp/src
1342 dd if=/dev/zero of=$tmp/src/t32_qf_old bs=1M \
1343 count=$(($T32_BLIMIT / 1024 / 4))
1344 chown $T32_QID.$T32_QID $tmp/src/t32_qf_old
1346 # format ost with comma-separated NIDs to verify LU-4460
1347 local failnid="$(h2nettype 1.2.3.4),$(h2nettype 4.3.2.1)"
1348 MGSNID="$MGSNID,$MGSNID" OSTOPT="--failnode=$failnid" formatall
1352 [[ $server_version -ge $(version_code 2.3.50) ]] ||
1353 $LFS quotacheck -ug /mnt/$FSNAME
1354 $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
1357 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1359 if [[ $MDSCOUNT -ge 2 ]]; then
1360 remote_dir=/mnt/$FSNAME/remote_dir
1361 $LFS mkdir -i 1 $remote_dir
1362 tar cf - -C $tmp/src . | tar xf - -C $remote_dir
1364 if [[ $server_version -ge $(version_code 2.7.0) ]]; then
1365 striped_dir=/mnt/$FSNAME/striped_dir_old
1366 $LFS mkdir -i 1 -c 2 $striped_dir
1367 tar cf - -C $tmp/src . | tar xf - -C $striped_dir
1373 mkdir $tmp/img || return 1
1377 pushd_dir=/mnt/$FSNAME
1378 if [[ $MDSCOUNT -ge 2 ]]; then
1379 pushd_dir=$remote_dir
1380 if [[ $server_version -ge $(version_code 2.7.0) ]]; then
1382 ls -Rni --time-style=+%s >$tmp/img/list2
1388 ls -Rni --time-style=+%s >$tmp/img/list
1389 find ! -name .lustre -type f -exec sha1sum {} \; |
1390 sort -k 2 >$tmp/img/sha1sums
1392 $LCTL get_param -n version | head -n 1 |
1393 sed -e 's/^lustre: *//' >$tmp/img/commit
1395 [[ $server_version -ge $(version_code 2.3.50) ]] ||
1396 $LFS quotaon -ug /mnt/$FSNAME
1397 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1398 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1399 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1400 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1401 | tr -d "*" > $tmp/img/bspace
1402 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1403 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1404 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1405 | tr -d "*" > $tmp/img/ispace
1406 echo $T32_BLIMIT > $tmp/img/blimit
1407 echo $T32_ILIMIT > $tmp/img/ilimit
1412 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1415 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1416 echo "Data verification failed"
1419 uname -r >$tmp/img/kernel
1420 uname -m >$tmp/img/arch
1422 mv ${MDSDEV1:-$MDSDEV} $tmp/img
1423 for num in $(seq 2 $MDSCOUNT); do
1424 local devname=$(mdsdevname $num)
1426 [[ $(facet_fstype $facet) != zfs ]] ||
1427 devname=$(mdsvdevname $num)
1428 mv $devname $tmp/img
1430 mv $OSTDEV1 $tmp/img
1432 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1433 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1436 tar cjvf $dst/disk$version-$(facet_fstype $SINGLEMDS).tar.bz2 -S *
1441 #run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1444 # The list of applicable tarballs is returned via the caller's
1445 # variable "tarballs".
1448 [ "$CLIENTONLY" ] && skip "Client-only testing"
1450 local node=$(facet_active_host $SINGLEMDS)
1451 local r="do_node $node"
1453 ! $r which "$TUNEFS" && skip_env "tunefs.lustre required on $node"
1455 local IMGTYPE=$(facet_fstype $SINGLEMDS)
1457 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 \
1458 -name \'disk*-$IMGTYPE.tar.bz2\')
1460 [ -z "$tarballs" ] && skip "No applicable tarballs found"
1463 t32_test_cleanup() {
1465 local facet=$SINGLEMDS
1466 local fstype=$(facet_fstype $facet)
1469 if $shall_cleanup_lustre; then
1470 umount $tmp/mnt/lustre || rc=$?
1472 if $shall_cleanup_mdt; then
1473 $r $UMOUNT $tmp/mnt/mdt || rc=$?
1475 if $shall_cleanup_mdt1; then
1476 $r $UMOUNT $tmp/mnt/mdt1 || rc=$?
1478 if $shall_cleanup_ost; then
1479 $r $UMOUNT $tmp/mnt/ost || rc=$?
1484 if [[ $fstype == zfs ]]; then
1486 local poolname_list="t32fs-mdt1 t32fs-ost1"
1488 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1490 for poolname in $poolname_list; do
1491 destroy_zpool $facet $poolname
1494 combined_mgs_mds || start_mgs || rc=$?
1498 t32_bits_per_long() {
1500 # Yes, this is not meant to be perfect.
1510 t32_reload_modules() {
1512 local all_removed=false
1514 local fstype=$(facet_fstype $SINGLEMDS)
1516 [ $fstype == "zfs" ] && do_rpc_nodes $node "service zed stop"
1518 while ((i < 20)); do
1519 echo "Unloading modules on $node: Attempt $i"
1520 do_rpc_nodes $node $LUSTRE_RMMOD $fstype &&
1522 do_rpc_nodes $node check_mem_leak || return 1
1523 if $all_removed; then
1524 do_rpc_nodes $node load_modules
1527 if [ $fstype == "zfs" ]; then
1528 do_rpc_nodes $node "$ZPOOL status -v"
1533 echo "Unloading modules on $node: Given up"
1537 t32_wait_til_devices_gone() {
1543 echo wait for devices to go
1544 while ((i < 20)); do
1545 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1546 loops=$(do_rpc_nodes $node losetup -a | grep -c t32)
1547 ((devices == 0 && loops == 0)) && return 0
1551 echo "waiting for dev on $node: dev $devices loop $loops given up"
1552 do_rpc_nodes $node "losetup -a"
1553 do_rpc_nodes $node "$LCTL device_list"
1557 t32_verify_quota() {
1561 local fstype=$(facet_fstype $SINGLEMDS)
1565 # LU-2435: if the underlying zfs doesn't support userobj_accounting,
1566 # lustre will estimate the object count usage. This fails quota
1567 # verification in 32b. The object quota usage should be accurate after
1568 # zfs-0.7.0 is released.
1569 [ $fstype == "zfs" ] && {
1570 local zfs_version=$(do_facet $facet cat /sys/module/zfs/version)
1572 [ $(version_code $zfs_version) -lt $(version_code 0.7.0) ] && {
1573 echo "Skip quota verify for zfs: $zfs_version"
1578 $LFS quota -u $T32_QID -v $mnt
1580 qval=$($LFS quota -v -u $T32_QID $mnt |
1581 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1582 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1584 [ $qval -eq $img_bspace ] || {
1585 echo "bspace, act:$qval, exp:$img_bspace"
1589 qval=$($LFS quota -v -u $T32_QID $mnt |
1590 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1591 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1593 [ $qval -eq $img_ispace ] || {
1594 echo "ispace, act:$qval, exp:$img_ispace"
1598 qval=$($LFS quota -v -u $T32_QID $mnt |
1599 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1600 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1602 [ $qval -eq $img_blimit ] || {
1603 echo "blimit, act:$qval, exp:$img_blimit"
1607 qval=$($LFS quota -v -u $T32_QID $mnt |
1608 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1609 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1611 [ $qval -eq $img_ilimit ] || {
1612 echo "ilimit, act:$qval, exp:$img_ilimit"
1616 set_persistent_param_and_check $facet \
1617 "osd-$fstype.$fsname-MDT0000.quota_slave.enabled" \
1618 "$fsname.quota.mdt" ug
1620 set_persistent_param_and_check $facet \
1621 "osd-$fstype.$fsname-OST0000.quota_slave.enabled" \
1622 "$fsname.quota.ost" ug
1625 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1626 bs=1M count=$((img_blimit / 1024)) oflag=sync && {
1627 echo "Write succeed, but expect -EDQUOT"
1630 rm -f $mnt/t32_qf_new
1632 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1634 echo "Create succeed, but expect -EDQUOT"
1637 unlinkmany $mnt/t32_qf_ $img_ilimit
1645 local dne_upgrade=${dne_upgrade:-"no"}
1646 local dom_upgrade=${dom_upgrade:-"no"}
1647 local ff_convert=${ff_convert:-"no"}
1648 local shall_cleanup_mdt=false
1649 local shall_cleanup_mdt1=false
1650 local shall_cleanup_ost=false
1651 local shall_cleanup_lustre=false
1652 local mdt2_is_available=false
1653 local node=$(facet_active_host $SINGLEMDS)
1654 local r="do_node $node"
1670 local fstype=$(facet_fstype $SINGLEMDS)
1671 local mdt_dev=$tmp/mdt
1672 local mdt2_dev=$tmp/mdt2
1673 local ost_dev=$tmp/ost
1678 combined_mgs_mds || stop_mgs || error "Unable to stop MGS"
1679 trap 'trap - RETURN; t32_test_cleanup' RETURN
1682 nid=$($r $LCTL list_nids | head -1)
1684 mkdir -p $tmp/mnt/lustre || error "mkdir $tmp/mnt/lustre failed"
1685 $r mkdir -p $tmp/mnt/{mdt,mdt1,ost}
1686 $r tar xjvf $tarball -S -C $tmp || {
1687 error_noexit "Unpacking the disk image tarball"
1690 img_commit=$($r cat $tmp/commit)
1691 img_kernel=$($r cat $tmp/kernel)
1692 img_arch=$($r cat $tmp/arch)
1693 img_bspace=$($r cat $tmp/bspace)
1694 img_ispace=$($r cat $tmp/ispace)
1696 # older images did not have "blimit" and "ilimit" files
1697 # use old values for T32_BLIMIT and T32_ILIMIT
1698 $r test -f $tmp/blimit && img_blimit=$($r cat $tmp/blimit) ||
1700 $r test -f $tmp/ilimit && img_ilimit=$($r cat $tmp/ilimit) ||
1703 echo "Upgrading from $(basename $tarball), created with:"
1704 echo " Commit: $img_commit"
1705 echo " Kernel: $img_kernel"
1706 echo " Arch: $img_arch"
1707 echo "OST version: $(lustre_build_version ost1)"
1709 # The conversion can be made only when both of the following
1710 # conditions are satisfied:
1711 # - ost device img version < 2.3.64
1712 # - ost server version >= 2.5
1713 [ $(version_code $img_commit) -ge $(version_code 2.3.64) -o \
1714 $(lustre_version_code ost1) -lt $(version_code 2.5.0) ] &&
1717 ! $r test -f $mdt2_dev || mdt2_is_available=true
1719 if [[ $fstype == zfs ]]; then
1722 local poolname_list="t32fs-mdt1 t32fs-ost1"
1724 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1726 for poolname in $poolname_list; do
1728 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
1729 $ZPOOL import -f -d $tmp $poolname"
1732 # upgrade zpool to latest supported features, including
1733 # dnode quota accounting in 0.7.0
1734 $r "$ZPOOL upgrade -a"
1736 mdt_dev=t32fs-mdt1/mdt1
1737 ost_dev=t32fs-ost1/ost1
1738 ! $mdt2_is_available || mdt2_dev=t32fs-mdt2/mdt2
1739 wait_update_facet $SINGLEMDS "$ZPOOL list |
1740 awk '/^t32fs-mdt1/ { print \\\$1 }'" "t32fs-mdt1" || {
1741 error_noexit "import zfs pool failed"
1746 $r $LCTL set_param debug="$PTLDEBUG"
1748 $r $TUNEFS --dryrun $mdt_dev || {
1750 error_noexit "tunefs.lustre before mounting the MDT"
1754 if $mdt2_is_available; then
1755 $r $TUNEFS --dryrun $mdt2_dev || {
1757 error_noexit "tunefs.lustre before mounting the MDT"
1762 if [ "$writeconf" ]; then
1764 if [ $fstype == "ldiskfs" ]; then
1766 $r $TUNEFS --quota $mdt_dev || {
1768 error_noexit "Enable mdt quota feature"
1771 if $mdt2_is_available; then
1772 $r $TUNEFS --quota $mdt2_dev || {
1774 error_noexit "Enable mdt quota feature"
1780 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
1781 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
1782 skip "LU-2200: Cannot run over IB w/o lctl replace_nids "
1783 "(Need MGS version at least 2.3.59)"
1785 local osthost=$(facet_active_host ost1)
1786 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
1789 if [ $fstype == "ldiskfs" ]; then
1792 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt
1793 $r $LCTL replace_nids $fsname-OST0000 $ostnid
1794 $r $LCTL replace_nids $fsname-MDT0000 $nid
1795 $r $UMOUNT $tmp/mnt/mdt
1798 mopts=exclude=$fsname-OST0000
1799 if [ $fstype == "ldiskfs" ]; then
1804 t32_wait_til_devices_gone $node
1806 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
1808 error_noexit "Mounting the MDT"
1811 shall_cleanup_mdt=true
1813 if $mdt2_is_available; then
1814 mopts=mgsnode=$nid,$mopts
1815 $r $MOUNT_CMD -o $mopts $mdt2_dev $tmp/mnt/mdt1 || {
1817 error_noexit "Mounting the MDT"
1821 echo "mount new MDT....$mdt2_dev"
1822 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1823 error_noexit "enable remote dir create failed"
1825 shall_cleanup_mdt1=true
1826 elif [ "$dne_upgrade" != "no" ]; then
1827 local fs2mdsdev=$(mdsdevname 1_2)
1828 local fs2mdsvdev=$(mdsvdevname 1_2)
1830 echo "mkfs new MDT on ${fs2mdsdev}...."
1831 if [ $(facet_fstype mds1) == ldiskfs ]; then
1832 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
1835 add $SINGLEMDS $(mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
1836 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
1837 error_noexit "Mkfs new MDT failed"
1841 [[ $(facet_fstype mds1) != zfs ]] || import_zpool mds1
1843 $r $TUNEFS --dryrun $fs2mdsdev || {
1844 error_noexit "tunefs.lustre before mounting the MDT"
1848 echo "mount new MDT....$fs2mdsdev"
1849 $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
1850 error_noexit "mount mdt1 failed"
1854 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1855 error_noexit "enable remote dir create failed"
1857 shall_cleanup_mdt1=true
1860 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
1861 error_noexit "Getting MDT UUID"
1864 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
1865 error_noexit "Unexpected MDT UUID: \"$uuid\""
1869 $r $TUNEFS --dryrun $ost_dev || {
1870 error_noexit "tunefs.lustre before mounting the OST"
1873 if [ "$writeconf" ]; then
1874 mopts=mgsnode=$nid,$writeconf
1875 if [ $fstype == "ldiskfs" ]; then
1877 $r $TUNEFS --quota $ost_dev || {
1879 error_noexit "Enable ost quota feature"
1885 if [ $fstype == "ldiskfs" ]; then
1890 $r $MOUNT_CMD -onomgs -o$mopts $ost_dev $tmp/mnt/ost || {
1891 error_noexit "Mounting the OST"
1894 shall_cleanup_ost=true
1896 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
1897 error_noexit "Getting OST UUID"
1900 if [ "$uuid" != $fsname-OST0000_UUID ]; then
1901 error_noexit "Unexpected OST UUID: \"$uuid\""
1905 if [[ $PERM_CMD = *"set_param -P"* ]]; then
1906 $r $PERM_CMD osc.$fsname-OST0000*.import=connection=$nid || {
1907 error_noexit "Setting OST \"failover.node\""
1910 $r $PERM_CMD mdc.$fsname-MDT0000*.import=connection=$nid || {
1911 error_noexit "Setting MDT \"failover.node\""
1914 $r $PERM_CMD osc.$fsname-OST0000-*.max_dirty_mb=15 || {
1915 error_noexit "Setting \"max_dirty_mb\""
1918 $r $PERM_CMD mdc.$fsname-MDT0000-*.max_rpcs_in_flight=9 || {
1919 error_noexit "Setting \"max_rpcs_in_flight\""
1922 $r $PERM_CMD lov.$fsname-MDT0000-*.stripesize=4M || {
1923 error_noexit "Setting \"lov.stripesize\""
1926 $r $PERM_CMD mdd.$fsname-MDT0000-*.atime_diff=70 || {
1927 error_noexit "Setting \"mdd.atime_diff\""
1931 $r $PERM_CMD $fsname-OST0000.failover.node=$nid || {
1932 error_noexit "Setting OST \"failover.node\""
1936 $r $PERM_CMD $fsname-MDT0000.failover.node=$nid || {
1937 error_noexit "Setting MDT \"failover.node\""
1941 $r $PERM_CMD $fsname-OST0000.osc.max_dirty_mb=15 || {
1942 error_noexit "Setting \"max_dirty_mb\""
1945 $r $PERM_CMD $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
1946 error_noexit "Setting \"max_rpcs_in_flight\""
1949 $r $PERM_CMD $fsname-MDT0000.lov.stripesize=4M || {
1950 error_noexit "Setting \"lov.stripesize\""
1953 $r $PERM_CMD $fsname-MDT0000.mdd.atime_diff=70 || {
1954 error_noexit "Setting \"mdd.atime_diff\""
1959 $r $LCTL pool_new $fsname.interop || {
1960 error_noexit "Setting \"interop\""
1964 if [ "$ff_convert" != "no" -a $(facet_fstype ost1) == "ldiskfs" ]; then
1965 $r $LCTL lfsck_start -M $fsname-OST0000 || {
1966 error_noexit "Start OI scrub on OST0"
1970 # The oi_scrub should be on ost1, but for test_32(),
1971 # all on the SINGLEMDS.
1972 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
1973 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1974 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
1975 error_noexit "Failed to get the expected 'completed'"
1979 local UPDATED=$($r $LCTL get_param -n \
1980 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1981 awk '/^updated/ { print $2 }')
1982 [ $UPDATED -ge 1 ] || {
1983 error_noexit "Only $UPDATED objects have been converted"
1988 if [ "$dne_upgrade" != "no" ]; then
1989 if [[ $PERM_CMD = *"set_param -P"* ]]; then
1990 $r $PERM_CMD mdc.$fsname-MDT0001*.import=connection=$nid || {
1991 error_noexit "Setting MDT1 \"failover.node\""
1995 $r $PERM_CMD mdc.$fsname-MDT0001-*.max_rpcs_in_flight=9 || {
1996 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
1999 $r $PERM_CMD lov.$fsname-MDT0001-*.stripesize=4M || {
2000 error_noexit "Setting MDT1 \"lov.stripesize\""
2004 $r $PERM_CMD $fsname-MDT0001.failover.node=$nid || {
2005 error_noexit "Setting MDT1 \"failover.node\""
2008 $r $PERM_CMD $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
2009 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
2012 $r $PERM_CMD $fsname-MDT0001.lov.stripesize=4M || {
2013 error_noexit "Setting MDT1 \"lov.stripesize\""
2019 if [ "$writeconf" ]; then
2020 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
2021 error_noexit "Mounting the client"
2025 shall_cleanup_lustre=true
2026 $r $LCTL set_param debug="$PTLDEBUG"
2028 # Leave re-enabling this to a separate patch for LU-11558
2029 # t32_verify_quota $SINGLEMDS $fsname $tmp/mnt/lustre || {
2030 # error_noexit "verify quota failed"
2034 if $r test -f $tmp/list; then
2036 # There is not a Test Framework API to copy files to or
2037 # from a remote node.
2039 # LU-2393 - do both sorts on same node to ensure locale
2041 local list_file=$tmp/list
2043 if $mdt2_is_available; then
2044 if [[ -d $tmp/mnt/lustre/striped_dir_old ]] &&
2045 $r test -f $tmp/list2; then
2046 list_file=$tmp/list2
2047 pushd $tmp/mnt/lustre/striped_dir_old
2049 pushd $tmp/mnt/lustre/remote_dir
2052 pushd $tmp/mnt/lustre
2054 $r cat $list_file | sort -k 6 >$tmp/list.orig
2055 ls -Rni --time-style=+%s | sort -k 6 |
2056 sed 's/\. / /' >$tmp/list || {
2062 # 32-bit and 64-bit clients use different algorithms to
2063 # convert FIDs into inode numbers. Hence, remove the
2064 # inode numbers from the lists, if the original list was
2065 # created on an architecture with different number of
2068 if [ $(t32_bits_per_long $(uname -m)) != \
2069 $(t32_bits_per_long $img_arch) ]; then
2070 echo "Different number of bits per \"long\"" \
2071 "from the disk image"
2072 for list in list.orig list; do
2073 sed -i -e 's/^[0-9]\+[ \t]\+//' \
2077 if ! diff -ub $tmp/list.orig $tmp/list; then
2078 error_noexit "list verification failed"
2082 echo "list verification skipped"
2085 if [ "$dom_upgrade" != "no" ]; then
2086 echo "Check DoM file can be created"
2087 $LFS setstripe -E 1M -L mdt -E EOF $tmp/mnt/lustre/dom || {
2088 error_noexit "Verify DoM creation"
2091 [ $($LFS getstripe -L $tmp/mnt/lustre/dom) == "mdt" ] || {
2092 error_noexit "Verify a DoM file"
2095 dd if=/dev/urandom of=$tmp/mnt/lustre/dom bs=4096 \
2096 count=1 conv=fsync || {
2097 error_noexit "Cannot write to DoM file"
2100 [ $(stat -c%s $tmp/mnt/lustre/dom) == 4096 ] || {
2101 error_noexit "DoM: bad size after write"
2104 rm $tmp/mnt/lustre/dom
2106 set_persistent_param_and_check mds \
2107 "lod.*MDT0000*.dom_stripesize" \
2108 "$fsname-MDT0000.lod.dom_stripesize" 0 || {
2109 error_noexit "Changing \"dom_stripesize\""
2114 if [ "$dne_upgrade" != "no" ]; then
2115 $LFS mkdir -i 1 -c2 $tmp/mnt/lustre/striped_dir || {
2116 error_noexit "set striped dir failed"
2120 $LFS setdirstripe -D -c2 $tmp/mnt/lustre/striped_dir
2122 pushd $tmp/mnt/lustre
2123 tar -cf - . --exclude=./striped_dir \
2124 --exclude=./striped_dir_old \
2125 --exclude=./remote_dir |
2126 tar -xvf - -C striped_dir 1>/dev/null || {
2127 error_noexit "cp to striped dir failed"
2133 # If it is upgrade from DNE (2.5), then rename the remote dir,
2134 # which is created in 2.5 to striped dir.
2135 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]]; then
2136 stripe_index=$($LFS getdirstripe -i \
2137 $tmp/mnt/lustre/remote_dir)
2139 [[ $stripe_index -eq 1 ]] || {
2140 error_noexit "get index \"$stripe_index\"" \
2141 "from remote dir failed"
2144 mv $tmp/mnt/lustre/remote_dir \
2145 $tmp/mnt/lustre/striped_dir/ || {
2146 error_noexit "mv remote dir failed"
2151 # If it is upgraded from DNE (2.7), then move the striped dir
2152 # which was created in 2.7 to the new striped dir.
2153 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]] &&
2154 [[ -d $tmp/mnt/lustre/striped_dir_old ]]; then
2155 stripe_count=$($LFS getdirstripe -c \
2156 $tmp/mnt/lustre/striped_dir_old)
2157 [[ $stripe_count -eq 2 ]] || {
2158 error_noexit "get count $stripe_count" \
2159 "from striped dir failed"
2162 mv $tmp/mnt/lustre/striped_dir_old \
2163 $tmp/mnt/lustre/striped_dir/ || {
2164 error_noexit "mv striped dir failed"
2170 $r $LCTL set_param -n osd*.*.force_sync=1
2171 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
2172 error_noexit "dd failed"
2175 rm -rf $tmp/mnt/lustre/tmp_file || {
2176 error_noexit "rm failed"
2180 if $r test -f $tmp/sha1sums; then
2181 # LU-2393 - do both sorts on same node to ensure locale
2183 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
2184 if [ "$dne_upgrade" != "no" ]; then
2185 pushd $tmp/mnt/lustre/striped_dir
2187 pushd $tmp/mnt/lustre
2190 find ! -path "*remote_dir*" ! -path "*striped_dir*" \
2191 ! -name .lustre -type f -exec sha1sum {} \; |
2192 sort -k 2 >$tmp/sha1sums || {
2194 error_noexit "sha1sum"
2198 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
2199 error_noexit "sha1sum verification failed"
2203 # if upgrade from DNE(2.5), then check remote directory
2204 # if upgrade from DNE(2.7), then check striped directory
2205 if $mdt2_is_available &&
2206 [[ "$dne_upgrade" != "no" ]]; then
2207 local new_dir="$tmp/mnt/lustre/striped_dir"
2208 local striped_dir_old="$new_dir/striped_dir_old"
2210 local dir_list="$new_dir/remote_dir"
2211 [[ ! -d $triped_dir_old ]] ||
2212 dir_list+=" $striped_dir_old"
2214 for dir in $dir_list; do
2216 find ! -name .lustre -type f \
2217 -exec sha1sum {} \; |
2218 sort -k 2 >$tmp/sha1sums || {
2220 error_noexit "sha1sum"
2224 if ! diff -ub $tmp/sha1sums.orig \
2226 error_noexit "sha1sum $dir" \
2233 echo "sha1sum verification skipped"
2236 if [ "$dne_upgrade" != "no" ]; then
2237 rm -rf $tmp/mnt/lustre/striped_dir || {
2238 error_noexit "remove remote dir failed"
2243 # migrate files/dirs to remote MDT, then move them back
2244 if [ $(lustre_version_code mds1) -ge $(version_code 2.7.50) -a \
2245 $dne_upgrade != "no" ]; then
2246 $r $LCTL set_param -n \
2247 mdt.${fsname}*.enable_remote_dir=1 2>/dev/null
2249 echo "test migration"
2250 pushd $tmp/mnt/lustre
2251 for dir in $(find ! -name .lustre ! -name . -type d); do
2252 mdt_index=$($LFS getdirstripe -i $dir)
2253 stripe_cnt=$($LFS getdirstripe -c $dir)
2254 if [ $mdt_index = 0 -a $stripe_cnt -le 1 ]; then
2255 $LFS migrate -m 1 $dir || {
2257 error_noexit "migrate MDT1 failed"
2263 for dir in $(find ! -name . ! -name .lustre -type d); do
2264 mdt_index=$($LFS getdirstripe -i $dir)
2265 stripe_cnt=$($LFS getdirstripe -c $dir)
2266 if [ $mdt_index = 1 -a $stripe_cnt -le 1 ]; then
2267 $LFS migrate -m 0 $dir || {
2269 error_noexit "migrate MDT0 failed"
2278 # When adding new data verification tests, please check for
2279 # the presence of the required reference files first, like
2280 # the "sha1sums" and "list" tests above, to avoid the need to
2281 # regenerate every image for each test addition.
2284 nrpcs_orig=$($LCTL get_param \
2285 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
2286 error_noexit "Getting \"max_rpcs_in_flight\""
2289 nrpcs=$((nrpcs_orig + 5))
2291 set_persistent_param_and_check client \
2292 "mdc.$fsname-MDT0000*.max_rpcs_in_flight" \
2293 "$fsname-MDT0000.mdc.max_rpcs_in_flight" $nrpcs || {
2294 error_noexit "Changing \"max_rpcs_in_flight\""
2298 umount $tmp/mnt/lustre || {
2299 error_noexit "Unmounting the client"
2302 shall_cleanup_lustre=false
2304 if [[ "$dne_upgrade" != "no" ]] || $mdt2_is_available; then
2305 $r $UMOUNT $tmp/mnt/mdt1 || {
2306 error_noexit "Unmounting the MDT2"
2309 if [[ $fstype == zfs ]]; then
2310 $r "$ZPOOL export t32fs-mdt2"
2312 shall_cleanup_mdt1=false
2315 $r $UMOUNT $tmp/mnt/mdt || {
2316 error_noexit "Unmounting the MDT"
2319 if [[ $fstype == zfs ]]; then
2320 $r "$ZPOOL export t32fs-mdt1"
2322 shall_cleanup_mdt=false
2324 $r $UMOUNT $tmp/mnt/ost || {
2325 error_noexit "Unmounting the OST"
2328 if [[ $fstype == zfs ]]; then
2329 $r "$ZPOOL export t32fs-ost1"
2331 shall_cleanup_ost=false
2333 t32_reload_modules $node || {
2334 error_noexit "Reloading modules"
2338 if [[ $fstype == zfs ]]; then
2339 local poolname=t32fs-mdt1
2341 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
2342 $ZPOOL import -f -d $tmp $poolname"
2344 # upgrade zpool to latest supported features,
2345 # including dnode quota accounting in 0.7.0
2346 $r "$ZPOOL upgrade $poolname"
2349 # mount a second time to make sure we didnt leave upgrade flag on
2350 $r $TUNEFS --dryrun $mdt_dev || {
2352 error_noexit "tunefs.lustre before remounting the MDT"
2356 mopts=exclude=$fsname-OST0000
2357 if [ $fstype == "ldiskfs" ]; then
2360 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2361 error_noexit "Remounting the MDT"
2364 shall_cleanup_mdt=true
2374 for tarball in $tarballs; do
2375 t32_test $tarball || let "rc += $?"
2379 run_test 32a "Upgrade (not live)"
2387 for tarball in $tarballs; do
2388 t32_test $tarball writeconf || let "rc += $?"
2392 run_test 32b "Upgrade with writeconf"
2399 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
2401 for tarball in $tarballs; do
2402 # Do not support 1_8 and 2_1 direct upgrade to DNE2 anymore */
2403 echo $tarball | grep "1_8" && continue
2404 echo $tarball | grep "2_1" && continue
2406 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
2410 run_test 32c "dne upgrade test"
2418 for tarball in $tarballs; do
2419 ff_convert=yes t32_test $tarball || rc=$?
2423 run_test 32d "convert ff test"
2426 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.10.56) ]] ||
2427 skip "Need MDS version at least 2.10.56"
2434 for tarball in $tarballs; do
2435 echo $tarball | grep "2_9" || continue
2437 dom_upgrade=yes t32_test $tarball writeconf || let "rc += $?"
2441 run_test 32e "dom upgrade test"
2443 test_33a() { # bug 12333, was test_33
2444 local FSNAME2=test-123
2445 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2448 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
2450 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
2451 local dev=${SINGLEMDS}_dev
2452 local MDSDEV=${!dev}
2453 is_blkdev $SINGLEMDS $MDSDEV &&
2454 skip_env "mixed loopback and real device not working"
2457 local fs2mdsdev=$(mdsdevname 1_2)
2458 local fs2ostdev=$(ostdevname 1_2)
2459 local fs2mdsvdev=$(mdsvdevname 1_2)
2460 local fs2ostvdev=$(ostvdevname 1_2)
2462 if [ $(facet_fstype mds1) == ldiskfs ]; then
2463 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
2466 if combined_mgs_mds; then
2467 local mgs_flag="--mgs"
2470 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --fsname=${FSNAME2} \
2471 --reformat $mgs_flag $mkfsoptions $fs2mdsdev $fs2mdsvdev ||
2473 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2474 --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \
2475 $fs2ostvdev || exit 10
2477 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
2478 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2480 if [[ $PERM_CMD = *"set_param -P"* ]]; then
2481 do_facet mgs "$PERM_CMD timeout=200" ||
2482 error "$PERM_CMD timeout=200 failed"
2484 do_facet mgs "$PERM_CMD $FSNAME2.sys.timeout=200" ||
2485 error "$PERM_CMD $FSNAME2.sys.timeout=200 failed"
2487 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2488 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2491 cp /etc/hosts $MOUNT2/ || error "copy /etc/hosts $MOUNT2/ failed"
2492 $GETSTRIPE $MOUNT2/hosts || error "$GETSTRIPE $MOUNT2/hosts failed"
2497 cleanup_nocli || error "cleanup_nocli failed with $?"
2499 run_test 33a "Mount ost with a large index number"
2501 test_33b() { # was test_34
2504 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
2505 # Drop lock cancelation reply during umount
2506 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
2507 do_facet client $LCTL set_param fail_loc=0x80000304
2508 #lctl set_param debug=-1
2509 umount_client $MOUNT
2510 cleanup || error "cleanup failed with $?"
2512 run_test 33b "Drop cancel during umount"
2516 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
2517 manual_umount_client
2519 do_facet client killall -USR1 multiop
2520 if [ $rc -eq 0 ]; then
2521 error "umount not fail!"
2524 cleanup || error "cleanup failed with rc $?"
2526 run_test 34a "umount with opened file should be fail"
2530 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2531 stop_mds || error "Unable to stop MDS"
2533 manual_umount_client --force || error "mtab after failed umount with $?"
2535 cleanup || error "cleanup failed with $?"
2537 run_test 34b "force umount with failed mds should be normal"
2541 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2542 stop_ost || error "Unable to stop OST1"
2544 manual_umount_client --force || error "mtab after failed umount with $?"
2546 cleanup || error "cleanup failed with $?"
2548 run_test 34c "force umount with failed ost should be normal"
2550 test_35a() { # bug 12459
2553 DBG_SAVE="`$LCTL get_param -n debug`"
2554 $LCTL set_param debug="ha"
2556 log "Set up a fake failnode for the MDS"
2558 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2559 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2561 if [[ $PERM_CMD = *"set_param -P"* ]]; then
2562 do_facet mgs "$PERM_CMD \
2563 mdc.*${device}*.import=connection=$(h2nettype $FAKENID)" ||
2564 error "Setting mdc.*${device}*.import=connection=\
2565 $(h2nettype $FAKENID) failed."
2567 do_facet mgs "$PERM_CMD \
2568 ${device}.failover.node=$(h2nettype $FAKENID)" ||
2569 error "Setting ${device}.failover.node=\
2570 $(h2nettype $FAKENID) failed."
2572 log "Wait for RECONNECT_INTERVAL seconds (10s)"
2575 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
2578 log "Stopping the MDT: $device"
2579 stop_mdt 1 || error "MDT0 stop fail"
2581 df $MOUNT > /dev/null 2>&1 &
2583 log "Restarting the MDT: $device"
2584 start_mdt 1 || error "MDT0 start fail"
2585 log "Wait for df ($DFPID) ... "
2588 $LCTL set_param debug="$DBG_SAVE"
2590 # retrieve from the log the first server that the client tried to
2591 # contact after the connection loss
2592 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2593 NEXTCONN=`awk "/${MSG}/ {start = 1;}
2594 /import_select_connection.*$device-mdc.* using connection/ {
2596 if (\\\$NF ~ /$FAKENID/)
2602 }" $TMP/lustre-log-$TESTNAME.log`
2603 [ "$NEXTCONN" != "0" ] &&
2604 error "Tried to connect to ${NEXTCONN} not last active server"
2605 cleanup || error "cleanup failed with $?"
2606 # remove nid settings
2607 writeconf_or_reformat
2609 run_test 35a "Reconnect to the last active server first"
2611 test_35b() { # bug 18674
2612 remote_mds || skip "local MDS"
2616 $LCTL set_param debug="ha"
2618 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
2621 log "Set up a fake failnode for the MDS"
2623 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2624 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2626 if [[ $PERM_CMD = *"set_param -P"* ]]; then
2627 do_facet mgs "$PERM_CMD \
2628 mdc.*${device}*.import=connection=$(h2nettype $FAKENID)" ||
2629 error "Set mdc.*${device}*.import=connection=\
2630 $(h2nettype $FAKENID) failed"
2632 do_facet mgs "$PERM_CMD \
2633 ${device}.failover.node=$(h2nettype $FAKENID)" ||
2634 error "Set ${device}.failover.node=\
2635 $(h2nettype $FAKENID) failed"
2638 local at_max_saved=0
2639 # adaptive timeouts may prevent seeing the issue
2640 if at_is_enabled; then
2641 at_max_saved=$(at_max_get mds)
2642 at_max_set 0 mds client
2645 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
2647 log "Injecting EBUSY on MDS"
2648 # Setting OBD_FAIL_MDS_RESEND=0x136
2649 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" ||
2650 error "unable to set param fail_loc=0x80000136"
2652 $LCTL set_param mdc.${FSNAME}*.stats=clear
2654 log "Creating a test file and stat it"
2655 touch $MOUNT/$tdir/$tfile || error "touch $MOUNT/$tdir/$tfile failed"
2656 stat $MOUNT/$tdir/$tfile
2658 log "Stop injecting EBUSY on MDS"
2659 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" ||
2660 error "unable to set param fail_loc=0"
2661 rm -f $MOUNT/$tdir/$tfile || error "remove $MOUNT/$tdir/$tfile failed"
2664 # restore adaptive timeout
2665 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
2667 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2669 CONNCNT=$($LCTL get_param mdc.${FSNAME}*.stats |
2670 awk '/mds_connect/{print $2}')
2672 # retrieve from the log if the client has ever tried to
2673 # contact the fake server after the loss of connection
2674 FAILCONN=`awk "BEGIN {ret = 0;}
2675 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
2677 if (\\\$NF ~ /$FAKENID/) {
2682 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
2684 [ "$FAILCONN" == "0" ] &&
2685 error "The client reconnection has not been triggered"
2686 [ "$FAILCONN" == "2" ] &&
2687 error "Primary server busy, client reconnect to failover failed"
2690 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
2691 # Reconnects are supposed to be rate limited to one every 5s
2692 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] &&
2693 error "Too many reconnects $CONNCNT"
2695 cleanup || error "cleanup failed with $?"
2696 # remove nid settings
2697 writeconf_or_reformat
2699 run_test 35b "Continue reconnection retries, if the active server is busy"
2702 [ $OSTCOUNT -lt 2 ] && skip_env "needs >= 2 OSTs"
2704 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] ||
2708 local FSNAME2=test1234
2709 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2711 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
2713 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
2714 is_blkdev $SINGLEMDS $MDSDEV &&
2715 skip_env "mixed loopback and real device not working"
2718 local fs2mdsdev=$(mdsdevname 1_2)
2719 local fs2ostdev=$(ostdevname 1_2)
2720 local fs3ostdev=$(ostdevname 2_2)
2721 local fs2mdsvdev=$(mdsvdevname 1_2)
2722 local fs2ostvdev=$(ostvdevname 1_2)
2723 local fs3ostvdev=$(ostvdevname 2_2)
2726 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2727 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
2728 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
2729 # different one than the default value here.
2730 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2731 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
2732 add fs3ost $(mkfs_opts ost2 ${fs3ostdev}) --mgsnode=$MGSNID \
2733 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
2735 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
2736 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2737 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
2738 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2739 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2741 sleep 5 # until 11778 fixed
2743 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || error "dd failed"
2745 BKTOTAL=$($LCTL get_param -n obdfilter.*.kbytestotal |
2746 awk 'BEGIN{total=0}; {total+=$1}; END{print total}')
2747 BKFREE=$($LCTL get_param -n obdfilter.*.kbytesfree |
2748 awk 'BEGIN{free=0}; {free+=$1}; END{print free}')
2749 BKAVAIL=$($LCTL get_param -n obdfilter.*.kbytesavail |
2750 awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}')
2751 STRING=$(df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}')
2752 DFTOTAL=$(echo $STRING | cut -d, -f1)
2753 DFUSED=$(echo $STRING | cut -d, -f2)
2754 DFAVAIL=$(echo $STRING | cut -d, -f3)
2755 DFFREE=$(($DFTOTAL - $DFUSED))
2757 ALLOWANCE=$((64 * $OSTCOUNT))
2759 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
2760 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
2761 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
2764 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
2765 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
2766 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
2769 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
2770 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
2771 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
2776 stop fs3ost -f || error "unable to stop OST3"
2777 stop fs2ost -f || error "unable to stop OST2"
2778 stop fs2mds -f || error "unable to stop second MDS"
2779 unload_modules_conf || error "unable unload modules"
2782 run_test 36 "df report consistency on OSTs with different block size"
2785 local mntpt=$(facet_mntpt $SINGLEMDS)
2786 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
2787 local mdsdev_sym="$TMP/sym_mdt.img"
2788 local opts=$MDS_MOUNT_OPTS
2791 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2792 skip "ldiskfs only test"
2795 echo "MDS : $mdsdev"
2796 echo "SYMLINK : $mdsdev_sym"
2797 do_facet $SINGLEMDS rm -f $mdsdev_sym
2799 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
2801 echo "mount symlink device - $mdsdev_sym"
2803 if ! do_facet $SINGLEMDS test -b $mdsdev; then
2804 opts=$(csa_add "$opts" -o loop)
2808 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
2809 $mdsdev_sym $mntpt 2>&1)
2812 echo mount_op=$mount_op
2814 do_facet $SINGLEMDS "$UMOUNT $mntpt && rm -f $mdsdev_sym"
2816 if $(echo $mount_op | grep -q "unable to set tunable"); then
2817 error "set tunables failed for symlink device"
2820 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
2822 run_test 37 "verify set tunables works for symlink device"
2824 test_38() { # bug 14222
2825 local fstype=$(facet_fstype $SINGLEMDS)
2826 local mntpt=$(facet_mntpt $SINGLEMDS)
2831 local SRC="/etc /bin"
2832 local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT)
2833 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
2834 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2835 tar cf - $FILES | tar xf - -C $DIR/$tdir ||
2836 error "copying $SRC to $DIR/$tdir"
2838 umount_client $MOUNT || error "umount_client $MOUNT failed"
2839 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2840 stop_mds || error "Unable to stop MDS"
2841 log "delete lov_objid file on MDS"
2843 mount_fstype $SINGLEMDS || error "mount MDS failed (1)"
2845 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid; rm $mntpt/lov_objid"
2847 unmount_fstype $SINGLEMDS || error "umount failed (1)"
2849 # check create in mds_lov_connect
2850 start_mds || error "unable to start MDS"
2851 mount_client $MOUNT || error "mount_client $MOUNT failed"
2853 [ $V ] && log "verifying $DIR/$tdir/$f"
2854 diff -q $f $DIR/$tdir/$f || ERROR=y
2856 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2857 if [ "$ERROR" = "y" ]; then
2858 # check it's updates in sync
2859 umount_client $MOUNT
2861 mount_fstype $SIGNLEMDS
2862 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2863 unmount_fstype $SINGLEMDS
2864 error "old and new files are different after connect" || true
2866 touch $DIR/$tdir/f2 || error "f2 file create failed"
2868 # check it's updates in sync
2869 umount_client $MOUNT || error "second umount_client $MOUNT failed"
2872 mount_fstype $SINGLEMDS || error "mount MDS failed (3)"
2874 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2875 do_facet $SINGLEMDS dd if=/dev/zero of=$mntpt/lov_objid.clear count=8
2877 unmount_fstype $SINGLEMDS || error "umount failed (3)"
2879 start_mds || error "unable to start MDS"
2880 mount_client $MOUNT || error "mount_client $MOUNT failed"
2882 [ $V ] && log "verifying $DIR/$tdir/$f"
2883 diff -q $f $DIR/$tdir/$f || ERROR=y
2885 touch $DIR/$tdir/f3 || error "f3 file create failed"
2886 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2887 umount_client $MOUNT || error "third umount_client $MOUNT failed"
2889 mount_fstype $SINGLEMDS || error "mount MDS failed (4)"
2890 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2891 unmount_fstype $SINGLEMDS || error "umount failed (4)"
2893 [ "$ERROR" = "y" ] &&
2894 error "old and new files are different after sync" || true
2896 log "files compared the same"
2897 cleanup || error "cleanup failed with $?"
2899 run_test 38 "MDS recreates missing lov_objid file from OST data"
2904 cleanup || error "cleanup failed with $?"
2905 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
2906 error "memory leak detected" || true
2908 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
2910 test_40() { # bug 15759
2911 start_ost || error "Unable to start OST1"
2912 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
2913 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
2915 cleanup || error "cleanup failed with rc $?"
2917 run_test 40 "race during service thread startup"
2919 test_41a() { #bug 14134
2920 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2921 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2922 skip "Loop devices does not work with nosvc option"
2925 combined_mgs_mds || skip "needs combined MGT and MDT device"
2927 start_mdt 1 -o nosvc -n
2928 if [ $MDSCOUNT -ge 2 ]; then
2929 for num in $(seq 2 $MDSCOUNT); do
2930 start_mdt $num || return
2933 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
2934 start_mdt 1 -o nomgs,force
2935 mount_client $MOUNT || error "mount_client $MOUNT failed"
2938 echo "blah blah" > $MOUNT/$tfile
2941 umount_client $MOUNT || error "umount_client $MOUNT failed"
2942 stop ost1 -f || error "unable to stop OST1"
2943 stop_mds || error "Unable to stop MDS"
2944 stop_mds || error "Unable to stop MDS on second try"
2946 run_test 41a "mount mds with --nosvc and --nomgs"
2949 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2950 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2951 skip "Loop devices does not work with nosvc option"
2954 ! combined_mgs_mds && skip "needs combined mgs device"
2958 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2960 start_mdt 1 -o nosvc -n
2961 if [ $MDSCOUNT -ge 2 ]; then
2962 for num in $(seq 2 $MDSCOUNT); do
2963 start_mdt $num || return
2966 start_ost || error "Unable to start OST1"
2967 start_mdt 1 -o nomgs,force
2968 mount_client $MOUNT || error "mount_client $MOUNT failed"
2971 echo "blah blah" > $MOUNT/$tfile
2972 cat $MOUNT/$tfile || error "cat $MOUNT/$tfile failed"
2974 umount_client $MOUNT -f || error "umount_client $MOUNT failed"
2975 stop_ost || error "Unable to stop OST1"
2976 stop_mds || error "Unable to stop MDS"
2977 stop_mds || error "Unable to stop MDS on second try"
2979 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
2982 local server_version=$(lustre_version_code $SINGLEMDS)
2983 local oss_list=$(comma_list $(osts_nodes))
2985 [[ $server_version -ge $(version_code 2.6.52) ]] ||
2986 [[ $server_version -ge $(version_code 2.5.26) &&
2987 $server_version -lt $(version_code 2.5.50) ]] ||
2988 [[ $server_version -ge $(version_code 2.5.4) &&
2989 $server_version -lt $(version_code 2.5.11) ]] ||
2990 skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"
2992 # ensure mds1 ost1 have been created even if running sub-test standalone
2995 cleanup || error "cleanup failed"
2997 # using directly mount command instead of start() function to avoid
2998 # any side effect of // with others/externals tools/features
2999 # ("zpool import", ...)
3001 # MDT concurrent start
3003 LOAD_MODULES_REMOTE=true load_modules
3004 do_facet $SINGLEMDS "lsmod | grep -q libcfs" ||
3005 error "MDT concurrent start: libcfs module not loaded"
3007 local mds1dev=$(mdsdevname 1)
3008 local mds1mnt=$(facet_mntpt mds1)
3009 local mds1fstype=$(facet_fstype mds1)
3010 local mds1opts=$MDS_MOUNT_OPTS
3012 if [ $mds1fstype == ldiskfs ] &&
3013 ! do_facet mds1 test -b $mds1dev; then
3014 mds1opts=$(csa_add "$mds1opts" -o loop)
3016 if [[ $mds1fstype == zfs ]]; then
3017 import_zpool mds1 || return ${PIPESTATUS[0]}
3020 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3021 do_facet mds1 "$LCTL set_param fail_loc=0x80000716"
3023 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts &
3026 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts
3030 do_facet mds1 "$LCTL set_param fail_loc=0x0"
3031 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3032 echo "1st MDT start succeed"
3033 echo "2nd MDT start failed with $rc2"
3034 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3035 echo "1st MDT start failed with $rc"
3036 echo "2nd MDT start succeed"
3039 error "unexpected concurrent MDT mounts result, rc=$rc rc2=$rc2"
3042 if [ $MDSCOUNT -ge 2 ]; then
3043 for num in $(seq 2 $MDSCOUNT); do
3044 start_mdt $num || return
3048 # OST concurrent start
3050 do_rpc_nodes $oss_list "lsmod | grep -q libcfs" ||
3051 error "OST concurrent start: libcfs module not loaded"
3053 local ost1dev=$(ostdevname 1)
3054 local ost1mnt=$(facet_mntpt ost1)
3055 local ost1fstype=$(facet_fstype ost1)
3056 local ost1opts=$OST_MOUNT_OPTS
3058 if [ $ost1fstype == ldiskfs ] &&
3059 ! do_facet ost1 test -b $ost1dev; then
3060 ost1opts=$(csa_add "$ost1opts" -o loop)
3062 if [[ $ost1fstype == zfs ]]; then
3063 import_zpool ost1 || return ${PIPESTATUS[0]}
3066 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3067 do_facet ost1 "$LCTL set_param fail_loc=0x80000716"
3069 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts &
3072 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts
3076 do_facet ost1 "$LCTL set_param fail_loc=0x0"
3077 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3078 echo "1st OST start succeed"
3079 echo "2nd OST start failed with $rc2"
3080 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3081 echo "1st OST start failed with $rc"
3082 echo "2nd OST start succeed"
3086 error "unexpected concurrent OST mounts result, rc=$rc rc2=$rc2"
3092 # verify everything ok
3097 error "MDT(s) start failed"
3105 error "OST(s) start failed"
3113 error "client start failed"
3120 error "client mount failed"
3124 run_test 41c "concurrent mounts of MDT/OST should all fail but one"
3126 test_42() { #bug 14693
3130 check_mount || error "client was not mounted"
3132 if [[ $PERM_CMD = *"set_param -P"* ]]; then
3133 PARAM="llite.$FSNAME-*.some_wrong_param"
3135 PARAM="$FSNAME.llite.some_wrong_param"
3138 do_facet mgs $PERM_CMD $PARAM=10
3139 umount_client $MOUNT ||
3140 error "unmounting client failed with invalid llite param"
3141 mount_client $MOUNT ||
3142 error "mounting client failed with invalid llite param"
3144 do_facet mgs $PERM_CMD $PARAM=20
3145 cleanup || error "stopping $FSNAME failed with invalid sys param"
3147 check_mount || error "client was not mounted with invalid sys param"
3148 cleanup || error "stopping $FSNAME failed with invalid sys param"
3150 run_test 42 "allow client/server mount/unmount with invalid config param"
3153 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.58) ]] ||
3154 skip "Need MDS version at least 2.5.58"
3155 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
3158 USER1=$(getent passwd | grep :$ID1:$ID1: | cut -d: -f1)
3159 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1"
3162 chmod ugo+x $DIR || error "chmod 0 failed"
3163 set_persistent_param_and_check mds1 \
3164 "mdt.$FSNAME-MDT0000.root_squash" \
3165 "$FSNAME.mdt.root_squash" \
3167 wait_update $HOSTNAME \
3168 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3170 error "check llite root_squash failed!"
3171 set_persistent_param_and_check mds1 \
3172 "mdt.$FSNAME-MDT0000.nosquash_nids" \
3173 "$FSNAME.mdt.nosquash_nids" \
3175 wait_update $HOSTNAME \
3176 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3178 error "check llite nosquash_nids failed!"
3181 # create set of test files
3183 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
3184 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
3185 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
3187 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
3188 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
3190 mkdir $DIR/$tdir-rootdir || error "mkdir failed"
3191 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
3192 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
3194 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
3195 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
3196 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
3199 # check root_squash:
3200 # set root squash UID:GID to RUNAS_ID
3201 # root should be able to access only files owned by RUNAS_ID
3203 set_persistent_param_and_check mds1 \
3204 "mdt.$FSNAME-MDT0000.root_squash" \
3205 "$FSNAME.mdt.root_squash" \
3206 "$RUNAS_ID:$RUNAS_ID"
3207 wait_update $HOSTNAME \
3208 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3209 "$RUNAS_ID:$RUNAS_ID" ||
3210 error "check llite root_squash failed!"
3212 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
3213 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3214 error "$ST: root read permission is denied"
3215 echo "$ST: root read permission is granted - ok"
3218 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3219 error "$ST: root write permission is denied"
3220 echo "$ST: root write permission is granted - ok"
3222 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3223 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3224 error "$ST: root read permission is granted"
3225 echo "$ST: root read permission is denied - ok"
3228 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3229 error "$ST: root write permission is granted"
3230 echo "$ST: root write permission is denied - ok"
3232 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3233 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null &&
3234 error "$ST: root unlink permission is granted"
3235 echo "$ST: root unlink permission is denied - ok"
3237 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null &&
3238 error "$ST: root create permission is granted"
3239 echo "$ST: root create permission is denied - ok"
3242 # check root_squash is enforced independently
3243 # of client cache content
3245 # access file by USER1, keep access open
3246 # root should be denied access to user file
3248 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
3252 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
3253 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3254 { kill $pid; error "$ST: root read permission is granted"; }
3255 echo "$ST: root read permission is denied - ok"
3258 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3259 { kill $pid; error "$ST: root write permission is granted"; }
3260 echo "$ST: root write permission is denied - ok"
3266 # check nosquash_nids:
3267 # put client's NID into nosquash_nids list,
3268 # root should be able to access root file after that
3270 local NIDLIST=$($LCTL list_nids all | tr '\n' ' ')
3271 NIDLIST="2@gni $NIDLIST 192.168.0.[2,10]@tcp"
3272 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
3273 set_persistent_param_and_check mds1 \
3274 "mdt.$FSNAME-MDT0000.nosquash_nids" \
3275 "$FSNAME-MDTall.mdt.nosquash_nids" \
3277 wait_update $HOSTNAME \
3278 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3280 error "check llite nosquash_nids failed!"
3282 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3283 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3284 error "$ST: root read permission is denied"
3285 echo "$ST: root read permission is granted - ok"
3288 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3289 error "$ST: root write permission is denied"
3290 echo "$ST: root write permission is granted - ok"
3292 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3293 rm $DIR/$tdir-rootdir/tfile-1 ||
3294 error "$ST: root unlink permission is denied"
3295 echo "$ST: root unlink permission is granted - ok"
3296 touch $DIR/$tdir-rootdir/tfile-2 ||
3297 error "$ST: root create permission is denied"
3298 echo "$ST: root create permission is granted - ok"
3299 cleanup || error "cleanup failed with $?"
3301 run_test 43a "check root_squash and nosquash_nids"
3303 test_43b() { # LU-5690
3304 [[ $(lustre_version_code mgs) -ge $(version_code 2.7.62) ]] ||
3305 skip "Need MGS version 2.7.62+"
3307 if [[ -z "$fs2mds_DEV" ]]; then
3308 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
3309 skip_env "mixed loopback and real device not working"
3312 local fs2mdsdev=$(mdsdevname 1_2)
3313 local fs2mdsvdev=$(mdsvdevname 1_2)
3315 # temporarily use fs2mds as fs2mgs
3317 local fs2mgsdev=$fs2mdsdev
3318 local fs2mgsvdev=$fs2mdsvdev
3320 local fsname=test1234
3322 load_module llite/lustre
3323 local client_ip=$(host_nids_address $HOSTNAME $NETTYPE)
3324 local host=${client_ip//*./}
3325 local net=${client_ip/%$host/}
3326 local nosquash_nids=$(h2nettype $net[$host,$host,$host])
3328 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
3329 --param mdt.root_squash=$RUNAS_ID:$RUNAS_ID \
3330 --param mdt.nosquash_nids=$nosquash_nids \
3331 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
3332 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS || error "start fs2mgs failed"
3333 stop $fs2mgs -f || error "stop fs2mgs failed"
3334 cleanup || error "cleanup failed with $?"
3336 run_test 43b "parse nosquash_nids with commas in expr_list"
3340 check_mount || error "check_mount"
3341 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
3343 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
3344 for VAL in $UUIDS; do
3345 NID=$(echo $VAL | cut -d= -f1)
3346 CLUUID=$(echo $VAL | cut -d= -f2)
3347 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
3349 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
3350 cleanup || error "cleanup failed with $?"
3352 run_test 44 "mounted client proc entry exists"
3356 check_mount || error "check_mount"
3357 stop_mds || error "Unable to stop MDS"
3361 #define OBD_FAIL_PTLRPC_LONG_REPL_UNLINK 0x50f
3362 do_facet client "$LCTL set_param fail_loc=0x8000050f"
3365 manual_umount_client --force || error "manual_umount_client failed"
3366 do_facet client "$LCTL set_param fail_loc=0x0"
3367 start_mds || error "unable to start MDS"
3368 mount_client $MOUNT || error "mount_client $MOUNT failed"
3369 cleanup || error "cleanup failed with $?"
3371 run_test 45 "long unlink handling in ptlrpcd"
3378 umount_client $MOUNT2 || rc=$?
3379 umount_client $MOUNT || rc=$?
3380 while [ $count -gt 0 ]; do
3381 stop ost${count} -f || rc=$?
3385 cleanup_nocli || rc=$?
3386 #writeconf to remove all ost2 traces for subsequent tests
3387 writeconf_or_reformat
3392 echo "Testing with $OSTCOUNT OSTs"
3394 start_mds || error "unable to start MDS"
3395 #first client should see only one ost
3396 start_ost || error "Unable to start OST1"
3397 wait_osc_import_state mds ost FULL
3399 mount_client $MOUNT || error "mount_client $MOUNT failed"
3400 trap "cleanup_46a $OSTCOUNT" EXIT ERR
3403 for (( i=2; i<=$OSTCOUNT; i++ )); do
3404 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
3405 error "start_ost$i $(ostdevname $i) failed"
3408 # wait until osts in sync
3409 for (( i=2; i<=$OSTCOUNT; i++ )); do
3410 wait_osc_import_state mds ost$i FULL
3411 wait_osc_import_ready client ost$i
3414 #second client see all ost's
3416 mount_client $MOUNT2 || error "mount_client failed"
3417 $SETSTRIPE -c -1 $MOUNT2 || error "$SETSTRIPE -c -1 $MOUNT2 failed"
3418 $GETSTRIPE $MOUNT2 || error "$GETSTRIPE $MOUNT2 failed"
3420 echo "ok" > $MOUNT2/widestripe
3421 $GETSTRIPE $MOUNT2/widestripe ||
3422 error "$GETSTRIPE $MOUNT2/widestripe failed"
3423 # fill acl buffer for avoid expand lsm to them
3424 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd |
3426 setfacl -m $acl $MOUNT2/widestripe
3430 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3432 cleanup_46a $OSTCOUNT || error "cleanup_46a failed"
3434 run_test 46a "handle ost additional - wide striped file"
3439 check_mount || error "check_mount failed"
3440 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
3444 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3445 if echo $ns | grep "MDT[[:digit:]]*"; then
3448 lrs=$(echo $ns | sed 's/.*lru_size=//')
3449 lru_size[count]=$lrs
3454 facet_failover $SINGLEMDS
3455 client_up || error "client_up failed"
3458 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3459 if echo $ns | grep "MDT[[:digit:]]*"; then
3462 lrs=$(echo $ns | sed 's/.*lru_size=//')
3463 if ! test "$lrs" -eq "${lru_size[count]}"; then
3464 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
3465 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
3470 cleanup || error "cleanup failed with $?"
3472 run_test 47 "server restart does not make client loss lru_resize settings"
3477 # reformat after this test is needed - if the test fails,
3478 # we will have unkillable file at FS
3482 test_48() { # bz-17636 LU-7473
3486 check_mount || error "check_mount failed"
3488 $SETSTRIPE -c -1 $MOUNT || error "$SETSTRIPE -c -1 $MOUNT failed"
3489 $GETSTRIPE $MOUNT || error "$GETSTRIPE $MOUNT failed"
3491 echo "ok" > $MOUNT/widestripe
3492 $GETSTRIPE $MOUNT/widestripe ||
3493 error "$GETSTRIPE $MOUNT/widestripe failed"
3495 # In the future, we may introduce more EAs, such as selinux, enlarged
3496 # LOV EA, and so on. These EA will use some EA space that is shared by
3497 # ACL entries. So here we only check some reasonable ACL entries count,
3498 # instead of the max number that is calculated from the max_ea_size.
3499 if [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.8.57) ];
3501 count=28 # hard coded of RPC protocol
3502 elif [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3503 count=4000 # max_num 4091 max_ea_size = 32768
3504 elif ! large_xattr_enabled; then
3505 count=450 # max_num 497 max_ea_size = 4012
3507 count=4500 # max_num 8187 max_ea_size = 1048492
3508 # not create too much (>5000) to save test time
3511 echo "It is expected to hold at least $count ACL entries"
3512 trap cleanup_48 EXIT ERR
3513 for ((i = 0; i < $count; i++)) do
3514 setfacl -m u:$((i + 100)):rw $MOUNT/widestripe ||
3515 error "Fail to setfacl for $MOUNT/widestripe at $i"
3518 cancel_lru_locks mdc
3519 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3520 local r_count=$(getfacl $MOUNT/widestripe | grep "user:" | wc -l)
3521 count=$((count + 1)) # for the entry "user::rw-"
3523 [ $count -eq $r_count ] ||
3524 error "Expected ACL entries $count, but got $r_count"
3528 run_test 48 "too many acls on file"
3530 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
3531 test_49a() { # bug 17710
3532 local timeout_orig=$TIMEOUT
3533 local ldlm_timeout_orig=$LDLM_TIMEOUT
3534 local LOCAL_TIMEOUT=20
3536 LDLM_TIMEOUT=$LOCAL_TIMEOUT
3537 TIMEOUT=$LOCAL_TIMEOUT
3541 check_mount || error "client mount failed"
3543 echo "check ldlm_timout..."
3544 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3545 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3546 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3548 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3549 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3552 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
3553 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
3556 umount_client $MOUNT || error "umount_client $MOUNT failed"
3557 stop_ost || error "problem stopping OSS"
3558 stop_mds || error "problem stopping MDS"
3560 LDLM_TIMEOUT=$ldlm_timeout_orig
3561 TIMEOUT=$timeout_orig
3563 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3565 test_49b() { # bug 17710
3566 local timeout_orig=$TIMEOUT
3567 local ldlm_timeout_orig=$LDLM_TIMEOUT
3568 local LOCAL_TIMEOUT=20
3570 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
3571 TIMEOUT=$LOCAL_TIMEOUT
3575 check_mount || error "client mount failed"
3577 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3578 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3579 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3581 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3582 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3585 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
3586 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
3589 cleanup || error "cleanup failed"
3591 LDLM_TIMEOUT=$ldlm_timeout_orig
3592 TIMEOUT=$timeout_orig
3594 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3597 # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1
3599 # Test both statfs and lfs df and fail if either one fails
3600 multiop_bg_pause $1 f_
3603 killall -USR1 multiop
3604 [ $RC -ne 0 ] && log "lazystatfs multiop failed"
3605 wait $PID || { RC=$?; log "multiop return error "; }
3607 # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1
3612 if kill -s 0 $PID; then
3615 log "lazystatfs lfs df failed to complete in 5s"
3623 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3624 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3626 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
3628 cleanup || error "cleanup failed with rc $?"
3630 run_test 50a "lazystatfs all servers available"
3634 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3635 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3637 # Wait for client to detect down OST
3638 stop_ost || error "Unable to stop OST1"
3639 wait_osc_import_state client ost DISCONN
3641 log "OSCs should all be DISCONN"
3643 lazystatfs $MOUNT || error "lazystatfs should not return EIO"
3645 umount_client $MOUNT || error "Unable to unmount client"
3646 stop_mds || error "Unable to stop MDS"
3648 run_test 50b "lazystatfs all servers down"
3651 start_mds || error "Unable to start MDS"
3652 start_ost || error "Unable to start OST1"
3653 start_ost2 || error "Unable to start OST2"
3654 mount_client $MOUNT || error "Unable to mount client"
3655 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3656 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3658 # Wait for client to detect down OST
3659 stop_ost || error "Unable to stop OST1"
3660 wait_osc_import_state mds ost DISCONN
3661 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3663 umount_client $MOUNT || error "Unable to unmount client"
3664 stop_ost2 || error "Unable to stop OST2"
3665 stop_mds || error "Unable to stop MDS"
3666 #writeconf to remove all ost2 traces for subsequent tests
3667 writeconf_or_reformat
3669 run_test 50c "lazystatfs one server down"
3672 start_mds || error "Unable to start MDS"
3673 start_ost || error "Unable to start OST1"
3674 start_ost2 || error "Unable to start OST2"
3675 mount_client $MOUNT || error "Unable to mount client"
3676 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3677 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3679 # Issue the statfs during the window where the client still
3680 # belives the OST to be available but it is in fact down.
3681 # No failure just a statfs which hangs for a timeout interval.
3682 stop_ost || error "Unable to stop OST1"
3683 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3685 umount_client $MOUNT || error "Unable to unmount client"
3686 stop_ost2 || error "Unable to stop OST2"
3687 stop_mds || error "Unable to stop MDS"
3688 #writeconf to remove all ost2 traces for subsequent tests
3689 writeconf_or_reformat
3691 run_test 50d "lazystatfs client/server conn race"
3698 start_mds || error "Unable to start MDS"
3699 #first client should see only one ost
3700 start_ost || error "Unable to start OST1"
3701 wait_osc_import_state mds ost FULL
3703 # Wait for client to detect down OST
3704 stop_ost || error "Unable to stop OST1"
3705 wait_osc_import_state mds ost DISCONN
3707 mount_client $MOUNT || error "Unable to mount client"
3708 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3710 multiop_bg_pause $MOUNT _f
3714 if [ $RC1 -ne 0 ]; then
3715 log "multiop failed $RC1"
3718 sleep $(( $TIMEOUT+1 ))
3720 [ $? -ne 0 ] && error "process isn't sleep"
3721 start_ost || error "Unable to start OST1"
3722 wait $pid || error "statfs failed"
3725 umount_client $MOUNT || error "Unable to unmount client"
3726 stop_ost || error "Unable to stop OST1"
3727 stop_mds || error "Unable to stop MDS"
3729 run_test 50e "normal statfs all servers down"
3734 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
3736 start_mds || error "Unable to start mds"
3737 #first client should see only one ost
3738 start_ost || error "Unable to start OST1"
3739 wait_osc_import_state mds ost FULL
3741 start_ost2 || error "Unable to start OST2"
3742 wait_osc_import_state mds ost2 FULL
3744 # Wait for client to detect down OST
3745 stop_ost2 || error "Unable to stop OST2"
3747 wait_osc_import_state mds ost2 DISCONN
3748 mount_client $MOUNT || error "Unable to mount client"
3749 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3751 multiop_bg_pause $MOUNT _f
3755 if [ $RC1 -ne 0 ]; then
3756 log "lazystatfs multiop failed $RC1"
3759 sleep $(( $TIMEOUT+1 ))
3761 [ $? -ne 0 ] && error "process isn't sleep"
3762 start_ost2 || error "Unable to start OST2"
3763 wait $pid || error "statfs failed"
3764 stop_ost2 || error "Unable to stop OST2"
3767 umount_client $MOUNT -f || error "Unable to unmount client"
3768 stop_ost || error "Unable to stop OST1"
3769 stop_mds || error "Unable to stop MDS"
3770 #writeconf to remove all ost2 traces for subsequent tests
3771 writeconf_or_reformat
3773 run_test 50f "normal statfs one server in down"
3776 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs"
3778 start_ost2 || error "Unable to start OST2"
3779 wait_osc_import_state mds ost2 FULL
3780 wait_osc_import_ready client ost2
3782 if [[ $PERM_CMD = *"set_param -P"* ]]; then
3783 local PARAM="osc.${FSNAME}-OST0001*.active"
3785 local PARAM="${FSNAME}-OST0001.osc.active"
3788 $SETSTRIPE -c -1 $DIR/$tfile || error "$SETSTRIPE failed"
3789 do_facet mgs $PERM_CMD $PARAM=0 || error "Unable to deactivate OST"
3791 umount_client $MOUNT || error "Unable to unmount client"
3792 mount_client $MOUNT || error "Unable to mount client"
3793 # This df should not cause a panic
3796 do_facet mgs $PERM_CMD $PARAM=1 || error "Unable to activate OST"
3797 rm -f $DIR/$tfile || error "unable to remove file $DIR/$tfile"
3798 umount_client $MOUNT || error "Unable to unmount client"
3799 stop_ost2 || error "Unable to stop OST2"
3800 stop_ost || error "Unable to stop OST1"
3801 stop_mds || error "Unable to stop MDS"
3802 #writeconf to remove all ost2 traces for subsequent tests
3803 writeconf_or_reformat
3805 run_test 50g "deactivated OST should not cause panic"
3809 # prepare MDT/OST, make OSC inactive for OST1
3810 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs"
3812 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3813 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
3814 error "tunefs OST1 failed"
3815 start_mds || error "Unable to start MDT"
3816 start_ost || error "Unable to start OST1"
3817 start_ost2 || error "Unable to start OST2"
3818 mount_client $MOUNT || error "client start failed"
3820 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3822 # activatate OSC for OST1
3823 set_persistent_param_and_check client \
3824 "osc.${FSNAME}-OST0000-osc-[!M]*.active" \
3825 "${FSNAME}-OST0000.osc.active" 1
3827 mkdir $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3828 $SETSTRIPE -c -1 -i 0 $DIR/$tdir/2 ||
3829 error "$SETSTRIPE $DIR/$tdir/2 failed"
3830 sleep 1 && echo "create a file after OST1 is activated"
3831 # doing some io, shouldn't crash
3832 dd if=/dev/zero of=$DIR/$tdir/2/$tfile-io bs=1M count=10
3834 # check OSC import is working
3835 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
3836 error "some OSC imports are still not connected"
3840 umount_client $MOUNT || error "Unable to umount client"
3841 stop_ost2 || error "Unable to stop OST2"
3842 cleanup_nocli || error "cleanup_nocli failed with $?"
3844 run_test 50h "LU-642: activate deactivated OST"
3847 # prepare MDT/OST, make OSC inactive for OST1
3848 [ "$MDSCOUNT" -lt "2" ] && skip "needs >= 2 MDTs"
3851 [ $(facet_fstype mds2) == zfs ] && import_zpool mds2
3852 do_facet mds2 "$TUNEFS --param mdc.active=0 $(mdsdevname 2)" ||
3853 error "tunefs MDT2 failed"
3854 start_mds || error "Unable to start MDT"
3855 start_ost || error "Unable to start OST1"
3856 start_ost2 || error "Unable to start OST2"
3857 mount_client $MOUNT || error "client start failed"
3859 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3861 if [[ $PERM_CMD = *"set_param -P"* ]]; then
3862 $PERM_CMD mdc.${FSNAME}-MDT0001-mdc-*.active=0 &&
3863 error "deactive MDC0 succeeds"
3865 $PERM_CMD ${FSNAME}-MDT0000.mdc.active=0 &&
3866 error "deactive MDC0 succeeds"
3869 # activate MDC for MDT2
3870 set_persistent_param_and_check client \
3871 "mdc.${FSNAME}-MDT0001-mdc-*.active" \
3872 "${FSNAME}-MDT0001.mdc.active" 1
3874 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
3875 if [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.60) ]
3877 wait_dne_interconnect
3879 $LFS mkdir -i1 $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3881 createmany -o $DIR/$tdir/2/$tfile-%d 1 || error "create files failed"
3883 rm -rf $DIR/$tdir/2 || error "unlink dir failed"
3885 # deactivate MDC for MDT2
3886 set_persistent_param_and_check client \
3887 "mdc.${FSNAME}-MDT0001-mdc-*.active" \
3888 "${FSNAME}-MDT0001.mdc.active" 0
3890 wait_osp_active mds ${FSNAME}-MDT0001 1 0
3892 $LFS mkdir -i1 $DIR/$tdir/2 &&
3893 error "mkdir $DIR/$tdir/2 succeeds after deactive MDT"
3895 $LFS mkdir -i0 -c$MDSCOUNT $DIR/$tdir/striped_dir ||
3896 error "mkdir $DIR/$tdir/striped_dir fails after deactive MDT2"
3898 local stripe_count=$($LFS getdirstripe -c $DIR/$tdir/striped_dir)
3899 [ $stripe_count -eq $((MDSCOUNT - 1)) ] ||
3900 error "wrong $stripe_count != $((MDSCOUNT -1)) for striped_dir"
3903 umount_client $MOUNT || error "Unable to umount client"
3908 run_test 50i "activate deactivated MDT"
3911 local LOCAL_TIMEOUT=20
3915 check_mount || error "check_mount failed"
3917 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
3918 $SETSTRIPE -c -1 $MOUNT/$tdir ||
3919 error "$SETSTRIPE -c -1 $MOUNT/$tdir failed"
3920 #define OBD_FAIL_MDS_REINT_DELAY 0x142
3921 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x142"
3922 touch $MOUNT/$tdir/$tfile &
3925 start_ost2 || error "Unable to start OST1"
3927 stop_ost2 || error "Unable to stop OST1"
3928 umount_client $MOUNT -f || error “unmount $MOUNT failed”
3929 cleanup_nocli || error “stop server failed”
3930 #writeconf to remove all ost2 traces for subsequent tests
3931 writeconf_or_reformat
3933 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
3942 do_node $node mkdir -p $dest
3943 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3945 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
3946 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3947 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
3949 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
3950 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
3960 local backup2=${TMP}/backup2
3962 do_node $node mkdir -p $backup2
3963 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3965 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
3966 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3967 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
3969 do_node $node "diff -rq $backup $backup2"
3970 [ $? -eq 0 ] || { error "contents differ"; return 3; }
3972 local xattrs2=${TMP}/xattrs2
3973 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
3974 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
3976 do_node $node "diff $xattrs $xattrs2"
3977 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
3979 do_node $node "rm -rf $backup2 $xattrs2"
3980 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
3984 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3985 skip "ldiskfs only test"
3988 start_mds || error "Unable to start MDS"
3989 start_ost || error "Unable to start OST1"
3990 mount_client $MOUNT || error "Unable to mount client"
3993 local ost1mnt=$(facet_mntpt ost1)
3994 local ost1node=$(facet_active_host ost1)
3995 local ost1tmp=$TMP/conf52
3998 mkdir $DIR/$tdir || error "Unable to create $DIR/$tdir"
3999 touch $TMP/modified_first || error "Unable to create temporary file"
4000 local mtime=$(stat -c %Y $TMP/modified_first)
4001 do_node $ost1node "mkdir -p $ost1tmp &&
4002 touch -m -d @$mtime $ost1tmp/modified_first" ||
4003 error "Unable to create temporary file"
4006 $SETSTRIPE -c -1 -S 1M $DIR/$tdir || error "$SETSTRIPE failed"
4008 for (( i=0; i < nrfiles; i++ )); do
4009 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c ||
4010 error "multiop failed"
4015 # sync all the data and make sure no pending data on the client,
4016 # thus the SOM xattr would not be changed any more.
4017 cancel_lru_locks osc
4020 echo backup files to $TMP/$tdir
4021 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
4022 copy_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
4023 error "Unable to copy files"
4025 umount_client $MOUNT || error "Unable to umount client"
4026 stop_ost || error "Unable to stop ost1"
4028 echo mount ost1 as ldiskfs
4029 do_node $ost1node mkdir -p $ost1mnt || error "Unable to create $ost1mnt"
4030 if ! do_node $ost1node test -b $ost1_dev; then
4033 do_node $ost1node mount -t $(facet_fstype ost1) $loop $ost1_dev \
4035 error "Unable to mount ost1 as ldiskfs"
4038 echo backup objects to $ost1tmp/objects
4039 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
4040 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
4041 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
4043 error "Unable to copy objects"
4045 # move objects to lost+found
4046 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
4047 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
4049 do_node $ost1node "umount $ost1mnt" ||
4050 error "Unable to umount ost1 as ldiskfs"
4052 start_ost || error "Unable to start OST1"
4053 mount_client $MOUNT || error "Unable to mount client"
4055 local REPAIRED=$(do_node $ost1node "$LCTL get_param \
4056 -n osd-ldiskfs.$FSNAME-OST0000.oi_scrub" |
4057 awk '/^lf_repa[ri]*ed/ { print $2 }')
4058 [ $REPAIRED -gt 0 ] ||
4059 error "Some entry under /lost+found should be repaired"
4062 diff_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
4063 error "Unable to diff files"
4065 rm -rf $TMP/$tdir $TMP/file_xattrs ||
4066 error "Unable to delete temporary files"
4067 do_node $ost1node "rm -rf $ost1tmp" ||
4068 error "Unable to delete temporary files"
4069 cleanup || error "cleanup failed with $?"
4071 run_test 52 "check recovering objects from lost+found"
4073 # Checks threads_min/max/started for some service
4075 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
4076 # parameter pattern prefix like 'ost.*.ost'.
4089 local msg="Insane $modname thread counts"
4090 local ncpts=$(check_cpt_number $facet)
4094 check_mount || return 41
4096 # We need to expand $parampat, but it may match multiple parameters, so
4097 # we'll pick the first one
4098 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
4099 error "Couldn't expand ${parampat}.threads_min parameter name"
4103 # Remove the .threads_min part
4104 paramp=${paramp%.threads_min}
4106 # Check for sanity in defaults
4107 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4109 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4111 tstarted=$(do_facet $facet "$LCTL get_param \
4112 -n ${paramp}.threads_started" || echo 0)
4113 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' ||
4115 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4117 nthrs=$(expr $tmax - $tmin)
4118 if [ $nthrs -lt $ncpts ]; then
4124 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
4125 skip_env "module parameter forced $facet thread count"
4127 # Check that we can change min/max
4128 do_facet $facet "$LCTL set_param \
4129 ${paramp}.threads_min=$((tmin + nthrs))"
4130 do_facet $facet "$LCTL set_param \
4131 ${paramp}.threads_max=$((tmax - nthrs))"
4132 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4134 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4136 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) &&
4137 $tmax2 == ($tmax - $nthrs)))' || return $?
4139 # Check that we can set min/max to the same value
4140 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4142 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$tmin"
4143 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4145 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4147 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
4149 # Check that we can't set max < min
4150 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$((tmin - 1))"
4151 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4153 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4155 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
4157 # We need to ensure that we get the module options desired; to do this
4158 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
4159 LOAD_MODULES_REMOTE=true
4162 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
4163 setmodopts -a $modname "$newvalue" oldvalue
4166 check_mount || return 41
4168 # Restore previous setting of MODOPTS_*
4169 setmodopts $modname "$oldvalue"
4171 # Check that $opts took
4172 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4174 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4176 tstarted=$(do_facet $facet \
4177 "$LCTL get_param -n ${paramp}.threads_started" || echo 0)
4178 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4187 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
4188 cleanup || error "cleanup failed with rc $?"
4190 run_test 53a "check OSS thread count params"
4194 local mds=$(do_facet $SINGLEMDS "$LCTL get_param \
4195 -N mds.*.*.threads_max 2>/dev/null")
4196 if [ -z "$mds" ]; then
4197 #running this on an old MDT
4198 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
4200 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
4202 cleanup || error "cleanup failed with $?"
4204 run_test 53b "check MDS thread count params"
4207 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4208 skip "ldiskfs only test"
4211 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p ||
4212 error "llverdev failed with rc=$?"
4215 run_test 54a "test llverdev and partial verify of device"
4218 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4219 skip "ldiskfs only test"
4223 run_llverfs $MOUNT -p || error "llverfs failed with rc=$?"
4224 cleanup || error "cleanup failed with rc=$?"
4226 run_test 54b "test llverfs and partial verify of filesystem"
4230 local max_ost_index=$1
4231 echo -n $(((max_ost_index + 1) * 8))
4235 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4236 skip "ldiskfs only test"
4239 local mdsdev=$(mdsdevname 1)
4240 local mdsvdev=$(mdsvdevname 1)
4244 if ! combined_mgs_mds; then
4245 stop_mgs || error "stopping MGS service failed"
4246 format_mgs || error "formatting MGT failed"
4248 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
4250 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
4251 --reformat $(ostdevname 1) $(ostvdevname 1)
4257 echo checking size of lov_objid for ost index $i
4258 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" |
4259 grep ^User | awk -F 'Size: ' '{print $2}')
4260 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
4261 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
4263 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
4270 run_test 55 "check lov_objid size"
4273 local server_version=$(lustre_version_code $SINGLEMDS)
4274 local mds_journal_size_orig=$MDSJOURNALSIZE
4280 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
4281 $(ostdevname 1) $(ostvdevname 1)
4282 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
4283 $(ostdevname 2) $(ostvdevname 2)
4286 start_ost || error "Unable to start first ost (idx 10000)"
4287 start_ost2 || error "Unable to start second ost (idx 1000)"
4288 mount_client $MOUNT || error "Unable to mount client"
4292 if [[ $server_version -ge $(version_code 2.6.54) ]] ||
4293 [[ $server_version -ge $(version_code 2.5.4) &&
4294 $server_version -lt $(version_code 2.5.11) ]]; then
4295 wait_osc_import_state mds ost1 FULL
4296 wait_osc_import_state mds ost2 FULL
4297 $SETSTRIPE --stripe-count=-1 $DIR/$tfile ||
4298 error "Unable to setstripe $DIR/$tfile"
4299 n=$($LFS getstripe --stripe-count $DIR/$tfile)
4300 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
4305 MDSJOURNALSIZE=$mds_journal_size_orig
4308 run_test 56a "check big OST indexes and out-of-index-order start"
4313 umount_client $MOUNT -f || error "unmount client failed"
4322 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs"
4324 trap cleanup_56b EXIT RETURN ERR
4327 if ! combined_mgs_mds ; then
4332 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) --index=0 --reformat \
4333 $(mdsdevname 1) $(mdsvdevname 1)
4334 add mds2 $(mkfs_opts mds2 $(mdsdevname 2)) --index=1 --reformat \
4335 $(mdsdevname 2) $(mdsvdevname 2)
4336 add mds3 $(mkfs_opts mds3 $(mdsdevname 3)) --index=1000 --reformat \
4337 $(mdsdevname 3) $(mdsvdevname 3)
4341 start_mdt 1 || error "MDT 1 (idx 0) start failed"
4342 start_mdt 2 || error "MDT 2 (idx 1) start failed"
4343 start_mdt 3 || error "MDT 3 (idx 1000) start failed"
4344 start_ost || error "Unable to start first ost"
4345 start_ost2 || error "Unable to start second ost"
4347 do_nodes $(comma_list $(mdts_nodes)) \
4348 "$LCTL set_param mdt.*.enable_remote_dir=1 \
4349 mdt.*.enable_remote_dir_gid=-1"
4351 mount_client $MOUNT || error "Unable to mount client"
4353 $LFS mkdir -c3 $MOUNT/$tdir || error "failed to make testdir"
4355 echo "This is test file 1!" > $MOUNT/$tdir/$tfile.1 ||
4356 error "failed to make test file 1"
4357 echo "This is test file 2!" > $MOUNT/$tdir/$tfile.2 ||
4358 error "failed to make test file 2"
4359 echo "This is test file 1000!" > $MOUNT/$tdir/$tfile.1000 ||
4360 error "failed to make test file 1000"
4362 rm -rf $MOUNT/$tdir || error "failed to remove testdir"
4364 $LFS mkdir -i1000 $MOUNT/$tdir.1000 ||
4365 error "create remote dir at idx 1000 failed"
4368 echo "=== START lfs df OUTPUT ==="
4370 echo "==== END lfs df OUTPUT ===="
4372 mdtcnt=$(echo -e "$output" | grep $FSNAME-MDT | wc -l)
4373 ostcnt=$(echo -e "$output" | grep $FSNAME-OST | wc -l)
4375 echo "lfs df returned mdt count $mdtcnt and ost count $ostcnt"
4376 [ $mdtcnt -eq 3 ] || error "lfs df returned wrong mdt count"
4377 [ $ostcnt -eq 2 ] || error "lfs df returned wrong ost count"
4379 echo "This is test file 1!" > $MOUNT/$tdir.1000/$tfile.1 ||
4380 error "failed to make test file 1"
4381 echo "This is test file 2!" > $MOUNT/$tdir.1000/$tfile.2 ||
4382 error "failed to make test file 2"
4383 echo "This is test file 1000!" > $MOUNT/$tdir.1000/$tfile.1000 ||
4384 error "failed to make test file 1000"
4385 rm -rf $MOUNT/$tdir.1000 || error "failed to remove remote_dir"
4388 echo "=== START lfs mdts OUTPUT ==="
4390 echo "==== END lfs mdts OUTPUT ===="
4392 echo -e "$output" | grep -v "MDTS:" | awk '{print $1}' |
4393 sed 's/://g' > $TMP/mdts-actual.txt
4394 sort $TMP/mdts-actual.txt -o $TMP/mdts-actual.txt
4396 echo -e "0\n1\n1000" > $TMP/mdts-expected.txt
4398 diff $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4401 rm $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4403 [ $result -eq 0 ] || error "target_obd proc file is incorrect!"
4405 run_test 56b "test target_obd correctness with nonconsecutive MDTs"
4407 test_57a() { # bug 22656
4408 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4409 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4410 tail -1 | awk '{print $1}')
4411 writeconf_or_reformat
4412 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4413 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" ||
4414 error "tunefs failed"
4416 start_ost && error "OST registration from failnode should fail"
4419 run_test 57a "initial registration from failnode should fail (should return errs)"
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 --servicenode=$NID `ostdevname 1`" ||
4428 error "tunefs failed"
4430 start_ost || error "OST registration from servicenode should not fail"
4433 run_test 57b "initial registration from servicenode should not fail"
4436 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
4439 test_58() { # bug 22658
4440 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
4442 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4443 createmany -o $DIR/$tdir/$tfile-%d 100
4444 unlinkmany $DIR/$tdir/$tfile-%d 100
4445 stop_mds || error "Unable to stop MDS"
4447 local MNTDIR=$(facet_mntpt $SINGLEMDS)
4448 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4450 # remove all files from the OBJECTS dir
4451 mount_fstype $SINGLEMDS
4453 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
4455 unmount_fstype $SINGLEMDS
4456 # restart MDS with missing llog files
4457 start_mds || error "unable to start MDS"
4458 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0"
4461 run_test 58 "missing llog files must not prevent MDT from mounting"
4464 start_mgsmds >> /dev/null
4465 local C1=$(count_osts)
4466 if [ $C1 -eq 0 ]; then
4467 start_ost >> /dev/null
4471 echo "original ost count: $C1 (expect > 0)"
4472 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
4473 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
4474 local C2=$(count_osts)
4475 echo "after mdt writeconf count: $C2 (expect 0)"
4476 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
4477 echo "OST start without writeconf should fail:"
4478 start_ost >> /dev/null &&
4479 error "OST start without writeconf didn't fail"
4480 echo "OST start with writeconf should succeed:"
4481 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
4482 local C3=$(count_osts)
4483 echo "after ost writeconf count: $C3 (expect 1)"
4484 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
4485 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
4486 local C4=$(count_osts)
4487 echo "after ost2 writeconf count: $C4 (expect 2)"
4488 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
4489 stop_ost2 >> /dev/null
4490 cleanup_nocli >> /dev/null
4491 #writeconf to remove all ost2 traces for subsequent tests
4492 writeconf_or_reformat
4494 run_test 59 "writeconf mount option"
4496 test_60() { # LU-471
4497 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4498 skip "ldiskfs only test"
4503 for num in $(seq $MDSCOUNT); do
4504 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
4505 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
4506 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4510 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
4511 [ ${PIPESTATUS[0]} -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
4513 # MDT default has dirdata feature
4514 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
4515 # we disable uninit_bg feature
4516 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
4517 # we set stride extended options
4518 echo $dump | grep stride > /dev/null || error "stride is not set"
4522 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
4527 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] ||
4528 skip "Need MDS version at least 2.1.53"
4530 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
4531 ! large_xattr_enabled; then
4534 for num in $(seq $MDSCOUNT); do
4535 do_facet mds${num} $TUNE2FS -O ea_inode \
4536 $(mdsdevname $num) ||
4537 error "tune2fs on mds $num failed"
4541 setup || error "setting up the filesystem failed"
4542 client_up || error "starting client failed"
4544 local file=$DIR/$tfile
4545 touch $file || error "touch $file failed"
4547 local large_value="$(generate_string $(max_xattr_size))"
4548 local small_value="bar"
4550 local name="trusted.big"
4551 log "save large xattr $name on $file"
4552 setfattr -n $name -v $large_value $file ||
4553 error "saving $name on $file failed"
4555 local new_value=$(get_xattr_value $name $file)
4556 [[ "$new_value" != "$large_value" ]] &&
4557 error "$name different after saving"
4559 log "shrink value of $name on $file"
4560 setfattr -n $name -v $small_value $file ||
4561 error "shrinking value of $name on $file failed"
4563 new_value=$(get_xattr_value $name $file)
4564 [[ "$new_value" != "$small_value" ]] &&
4565 error "$name different after shrinking"
4567 log "grow value of $name on $file"
4568 setfattr -n $name -v $large_value $file ||
4569 error "growing value of $name on $file failed"
4571 new_value=$(get_xattr_value $name $file)
4572 [[ "$new_value" != "$large_value" ]] &&
4573 error "$name different after growing"
4575 log "check value of $name on $file after remounting MDS"
4577 new_value=$(get_xattr_value $name $file)
4578 [[ "$new_value" != "$large_value" ]] &&
4579 error "$name different after remounting MDS"
4581 log "remove large xattr $name from $file"
4582 setfattr -x $name $file || error "removing $name from $file failed"
4585 stopall || error "stopping for e2fsck run"
4586 for num in $(seq $MDSCOUNT); do
4587 run_e2fsck $(facet_active_host mds$num) \
4588 $(mdsdevname $num) "-y" ||
4589 error "e2fsck MDT$num failed"
4591 setup_noconfig || error "remounting the filesystem failed"
4594 # need to delete this file to avoid problems in other tests
4596 cleanup || error "stopping systems failed"
4598 run_test 61 "large xattr"
4601 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4602 skip "ldiskfs only test"
4604 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] ||
4605 skip "Need MDS version at least 2.2.51"
4608 local mdsdev=$(mdsdevname 1)
4609 local ostdev=$(ostdevname 1)
4611 echo "disable journal for mds"
4612 do_facet mds1 $TUNE2FS -O ^has_journal $mdsdev || error "tune2fs failed"
4613 start_mds && error "MDT start should fail"
4614 echo "disable journal for ost"
4615 do_facet ost1 $TUNE2FS -O ^has_journal $ostdev || error "tune2fs failed"
4616 start_ost && error "OST start should fail"
4617 cleanup || error "cleanup failed with rc $?"
4620 run_test 62 "start with disabled journal"
4623 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4624 skip "ldiskfs only test"
4627 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ldiskfs
4628 local inode_slab=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
4629 awk '/ldiskfs_inode_cache/ { print $5 / $6 }')
4630 if [ -z "$inode_slab" ]; then
4631 skip "ldiskfs module has not been loaded"
4634 echo "$inode_slab ldiskfs inodes per page"
4635 [ "${inode_slab%.*}" -ge "3" ] && return 0
4637 # If kmalloc-128 is also 1 per page - this is a debug kernel
4638 # and so this is not an error.
4639 local kmalloc128=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
4640 awk '/^(kmalloc|size)-128 / { print $5 / $6 }')
4641 # 32 128-byte chunks in 4k
4642 [ "${kmalloc128%.*}" -lt "32" ] ||
4643 error "ldiskfs inode too big, only $inode_slab objs/page, " \
4644 "kmalloc128 = $kmalloc128 objs/page"
4646 run_test 63 "Verify each page can at least hold 3 ldiskfs inodes"
4649 start_mds || error "unable to start MDS"
4650 start_ost || error "Unable to start OST1"
4651 start_ost2 || error "Unable to start second ost"
4652 mount_client $MOUNT || error "Unable to mount client"
4653 stop_ost2 || error "Unable to stop second ost"
4656 umount_client $MOUNT -f || error “unmount $MOUNT failed”
4657 cleanup_nocli || error "cleanup_nocli failed with $?"
4658 #writeconf to remove all ost2 traces for subsequent tests
4659 writeconf_or_reformat
4661 run_test 64 "check lfs df --lazy "
4663 test_65() { # LU-2237
4664 # Currently, the test is only valid for ldiskfs backend
4665 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4666 skip "ldiskfs only test"
4668 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4669 local brpt=$(facet_mntpt brpt)
4672 if ! do_facet $SINGLEMDS "test -b $devname"; then
4676 stop_mds || error "Unable to stop MDS"
4677 local obj=$(do_facet $SINGLEMDS \
4678 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
4680 if [ -z "$obj" ]; then
4681 # The MDT may be just re-formatted, mount the MDT for the
4682 # first time to guarantee the "last_rcvd" file is there.
4683 start_mds || error "fail to mount the MDS for the first time"
4684 stop_mds || error "Unable to stop MDS"
4687 # remove the "last_rcvd" file
4688 do_facet $SINGLEMDS "mkdir -p $brpt"
4689 do_facet $SINGLEMDS \
4690 "mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
4691 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
4692 do_facet $SINGLEMDS "$UMOUNT $brpt"
4694 # restart MDS, the "last_rcvd" file should be recreated.
4695 start_mds || error "fail to restart the MDS"
4696 stop_mds || error "Unable to stop MDS"
4697 obj=$(do_facet $SINGLEMDS \
4698 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
4699 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
4701 run_test 65 "re-create the lost last_rcvd file when server mount"
4704 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
4705 skip "Need MGS version at least 2.3.59"
4708 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
4709 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
4711 # add EXCLUDE records to config log, they are not to be
4712 # removed by lctl replace_nids
4713 set_conf_param_and_check mds \
4714 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
4715 "$FSNAME-OST0000.osc.active" \
4718 echo "replace_nids should fail if MDS, OSTs and clients are UP"
4719 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4720 error "replace_nids fail"
4722 umount_client $MOUNT || error "unmounting client failed"
4723 echo "replace_nids should fail if MDS and OSTs are UP"
4724 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4725 error "replace_nids fail"
4727 stop_ost || error "Unable to stop OST1"
4728 echo "replace_nids should fail if MDS is UP"
4729 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4730 error "replace_nids fail"
4732 stop_mds || error "stopping mds failed"
4734 if combined_mgs_mds; then
4735 start_mdt 1 "-o nosvc" ||
4736 error "starting mds with nosvc option failed"
4739 echo "command should accept two parameters"
4740 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
4741 error "command should accept two params"
4743 echo "correct device name should be passed"
4744 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
4745 error "wrong devname"
4747 echo "wrong nids list should not destroy the system"
4748 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
4750 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "asdfasdf, asdfadf" &&
4753 echo "replace OST nid"
4754 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
4755 error "replace nids failed"
4757 echo "command should accept two parameters"
4758 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
4759 error "command should accept two params"
4761 echo "wrong nids list should not destroy the system"
4762 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
4765 local FAKE_NIDS="192.168.0.112@tcp1,192.168.0.112@tcp2"
4766 local FAKE_FAILOVER="192.168.0.113@tcp1,192.168.0.113@tcp2"
4767 local NIDS_AND_FAILOVER="$MDS_NID,$FAKE_NIDS:$FAKE_FAILOVER"
4768 echo "set NIDs with failover"
4769 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $NIDS_AND_FAILOVER ||
4770 error "replace nids failed"
4773 echo "replace MDS nid"
4774 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
4775 error "replace nids failed"
4777 if ! combined_mgs_mds ; then
4780 stop_mds || error "Unable to stop MDS"
4783 start_mgsmds || error "start mgsmds failed"
4784 set_conf_param_and_check mds \
4785 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
4786 "$FSNAME-OST0000.osc.active" \
4788 start_ost || error "unable to start OST"
4789 mount_client $MOUNT || error "mount client failed"
4791 check_mount || error "error after nid replace"
4792 cleanup || error "cleanup failed"
4795 run_test 66 "replace nids"
4797 test_67() { #LU-2950
4798 local legacy="$TMP/legacy_lnet_config"
4799 local new="$TMP/new_routes_test"
4800 local out="$TMP/config_out_file"
4801 local verify="$TMP/conv_verify"
4802 local verify_conf="$TMP/conf_verify"
4804 # Create the legacy file that will be run through the
4805 # lustre_routes_conversion script
4806 cat <<- LEGACY_LNET_CONFIG > $legacy
4807 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
4808 tcp2 54 10.1.3.2@tcp;
4809 tcp3 10.3.4.3@tcp:3;
4813 # Create the verification file to verify the output of
4814 # lustre_routes_conversion script against.
4815 cat <<- VERIFY_LNET_CONFIG > $verify
4816 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
4817 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
4818 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
4819 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
4820 tcp4: { gateway: 10.3.3.4@tcp }
4823 # Create the verification file to verify the output of
4824 # lustre_routes_config script against
4825 cat <<- VERIFY_LNET_CONFIG > $verify_conf
4826 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
4827 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
4828 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
4829 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
4830 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
4833 $LUSTRE_ROUTES_CONVERSION $legacy $new > /dev/null
4834 if [ -f $new ]; then
4835 # verify the conversion output
4836 cmp -s $new $verify > /dev/null
4837 if [ $? -eq 1 ]; then
4838 error "routes conversion failed"
4841 lustre_routes_config --dry-run --verbose $new > $out
4842 # check that the script succeeded
4843 cmp -s $out $verify_conf > /dev/null
4844 if [ $? -eq 1 ]; then
4845 error "routes config failed"
4848 error "routes conversion test failed"
4850 # remove generated files
4851 rm -f $new $legacy $verify $verify_conf $out
4853 run_test 67 "test routes conversion and configuration"
4861 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.4.53) ] ||
4862 skip "Need MDS version at least 2.4.53"
4864 umount_client $MOUNT || error "umount client failed"
4869 # START-END - the sequences we'll be reserving
4870 START=$(do_facet $SINGLEMDS \
4871 $LCTL get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
4872 END=$((START + (1 << 30)))
4873 do_facet $SINGLEMDS \
4874 $LCTL set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
4876 # reset the sequences MDT0000 has already assigned
4877 do_facet $SINGLEMDS \
4878 $LCTL set_param seq.srv*MDT0000.space=clear
4880 # remount to let the client allocate new sequence
4881 mount_client $MOUNT || error "mount client failed"
4883 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4884 do_facet $SINGLEMDS \
4885 $LCTL get_param seq.srv*MDT0000.space
4886 $LFS path2fid $DIR/$tfile
4888 local old_ifs="$IFS"
4890 fid=($($LFS path2fid $DIR/$tfile))
4894 if [[ $seq < $END ]]; then
4895 error "used reserved sequence $seq?"
4897 cleanup || error "cleanup failed with $?"
4899 run_test 68 "be able to reserve specific sequences in FLDB"
4901 # Test 69: is about the total number of objects ever created on an OST.
4902 # so that when it is reformatted the normal MDS->OST orphan recovery won't
4903 # just "precreate" the missing objects. In the past it might try to recreate
4904 # millions of objects after an OST was reformatted
4906 local server_version=$(lustre_version_code $SINGLEMDS)
4908 [[ $server_version -lt $(version_code 2.4.2) ]] &&
4909 skip "Need MDS version at least 2.4.2"
4911 [[ $server_version -ge $(version_code 2.4.50) ]] &&
4912 [[ $server_version -lt $(version_code 2.5.0) ]] &&
4913 skip "Need MDS version at least 2.5.0"
4916 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4918 # use OST0000 since it probably has the most creations
4919 local OSTNAME=$(ostname_from_index 0)
4920 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
4921 local last_id=$(do_facet mds1 $LCTL get_param -n \
4922 osp.$mdtosc_proc1.prealloc_last_id)
4924 # Want to have OST LAST_ID over 5 * OST_MAX_PRECREATE to
4925 # verify that the LAST_ID recovery is working properly. If
4926 # not, then the OST will refuse to allow the MDS connect
4927 # because the LAST_ID value is too different from the MDS
4928 #define OST_MAX_PRECREATE=20000
4929 local ost_max_pre=20000
4930 local num_create=$(( ost_max_pre * 5 + 1 - last_id))
4932 # If the LAST_ID is already over 5 * OST_MAX_PRECREATE, we don't
4933 # need to create any files. So, skip this section.
4934 if [ $num_create -gt 0 ]; then
4935 # Check the number of inodes available on OST0
4937 local ifree=$($LFS df -i $MOUNT | awk '/OST0000/ { print $4 }')
4938 log "On OST0, $ifree inodes available. Want $num_create."
4940 $SETSTRIPE -i 0 $DIR/$tdir ||
4941 error "$SETSTRIPE -i 0 $DIR/$tdir failed"
4942 if [ $ifree -lt 10000 ]; then
4943 files=$(( ifree - 50 ))
4948 local j=$((num_create / files + 1))
4949 for i in $(seq 1 $j); do
4950 createmany -o $DIR/$tdir/$tfile-$i- $files ||
4951 error "createmany fail create $files files: $?"
4952 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
4953 error "unlinkmany failed unlink $files files"
4957 # delete all of the files with objects on OST0 so the
4958 # filesystem is not inconsistent later on
4959 $LFS find $MOUNT --ost 0 -print0 | xargs -0 rm
4961 umount_client $MOUNT || error "umount client failed"
4962 stop_ost || error "OST0 stop failure"
4963 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
4964 $(ostdevname 1) $(ostvdevname 1) ||
4965 error "reformat and replace $ostdev failed"
4966 start_ost || error "OST0 restart failure"
4967 wait_osc_import_state mds ost FULL
4969 mount_client $MOUNT || error "mount client failed"
4970 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
4971 local idx=$($GETSTRIPE -i $DIR/$tdir/$tfile-last)
4972 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
4974 local iused=$($LFS df -i $MOUNT | awk '/OST0000/ { print $3 }')
4975 log "On OST0, $iused used inodes"
4976 [ $iused -ge $((ost_max_pre/2 + 1000)) ] &&
4977 error "OST replacement created too many inodes; $iused"
4978 cleanup || error "cleanup failed with $?"
4980 run_test 69 "replace an OST with the same index"
4983 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
4986 cleanup || error "cleanup failed with $?"
4988 start_mdt 1 || error "MDT0 start fail"
4990 start_ost || error "OST0 start fail"
4991 for num in $(seq 2 $MDSCOUNT); do
4992 start_mdt $num || return
4995 mount_client $MOUNT || error "mount client fails"
4997 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
4999 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5000 error "create remote dir fail"
5002 rm -rf $DIR/$tdir || error "delete dir fail"
5003 cleanup || error "cleanup failed with $?"
5005 run_test 70a "start MDT0, then OST, then MDT1"
5008 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5011 start_ost || error "OST0 start fail"
5013 start_mds || error "MDS start fail"
5015 mount_client $MOUNT || error "mount client fails"
5017 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
5019 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5020 error "create remote dir fail"
5022 rm -rf $DIR/$tdir || error "delete dir fail"
5024 cleanup || error "cleanup failed with $?"
5026 run_test 70b "start OST, MDT1, MDT0"
5029 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5032 start_mds || error "MDS start fail"
5033 start_ost || error "OST0 start fail"
5035 mount_client $MOUNT || error "mount client fails"
5036 stop_mdt 1 || error "MDT1 start fail"
5038 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
5039 echo "deactivate $mdc_for_mdt1"
5040 $LCTL --device $mdc_for_mdt1 deactivate ||
5041 error "set $mdc_for_mdt1 deactivate failed"
5043 mkdir $DIR/$tdir && error "mkdir succeed"
5045 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
5046 error "create remote dir succeed"
5048 cleanup || error "cleanup failed with $?"
5050 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
5053 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5056 start_mds || error "MDS start fail"
5057 start_ost || error "OST0 start fail"
5059 mount_client $MOUNT || error "mount client fails"
5061 stop_mdt 2 || error "MDT1 start fail"
5063 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
5065 echo "deactivate $mdc_for_mdt2"
5066 $LCTL --device $mdc_for_mdt2 deactivate ||
5067 error "set $mdc_for_mdt2 deactivate failed"
5069 mkdir $DIR/$tdir || error "mkdir fail"
5070 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
5071 error "create remote dir succeed"
5073 rm -rf $DIR/$tdir || error "delete dir fail"
5075 cleanup || error "cleanup failed with $?"
5077 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
5080 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5082 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.62) ] ||
5083 skip "Need MDS version at least 2.7.62"
5085 reformat || error "reformat failed with $?"
5089 local mdsdev=$(mdsdevname 1)
5090 local ostdev=$(ostdevname 1)
5091 local mdsvdev=$(mdsvdevname 1)
5092 local ostvdev=$(ostvdevname 1)
5093 local opts_mds="$(mkfs_opts mds1 $mdsdev) --reformat $mdsdev $mdsvdev"
5094 local opts_ost="$(mkfs_opts ost1 $ostdev) --reformat $ostdev $ostvdev"
5096 if ! combined_mgs_mds ; then
5100 add mds1 $opts_mds || error "add mds1 failed"
5101 start_mdt 1 || error "start mdt1 failed"
5102 add ost1 $opts_ost || error "add ost1 failed"
5103 start_ost || error "start ost failed"
5104 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
5106 local soc=$(do_facet mds1 "$LCTL get_param -n \
5107 mdt.*MDT0000.sync_lock_cancel")
5108 [ $soc == "never" ] || error "SoC enabled on single MDS"
5110 for i in $(seq 2 $MDSCOUNT); do
5111 mdsdev=$(mdsdevname $i)
5112 mdsvdev=$(mdsvdevname $i)
5113 opts_mds="$(mkfs_opts mds$i $mdsdev) --reformat $mdsdev \
5115 add mds$i $opts_mds || error "add mds$i failed"
5116 start_mdt $i || error "start mdt$i fail"
5119 wait_dne_interconnect
5121 for i in $(seq $MDSCOUNT); do
5122 soc=$(do_facet mds$i "$LCTL get_param -n \
5123 mdt.*MDT000$((i - 1)).sync_lock_cancel")
5124 [ $soc == "blocking" ] || error "SoC not enabled on DNE"
5127 for i in $(seq 2 $MDSCOUNT); do
5128 stop_mdt $i || error "stop mdt$i fail"
5130 soc=$(do_facet mds1 "$LCTL get_param -n \
5131 mdt.*MDT0000.sync_lock_cancel")
5132 [ $soc == "never" ] || error "SoC enabled on single MDS"
5133 umount_client $MOUNT -f > /dev/null
5135 cleanup || error "cleanup failed with $?"
5137 run_test 70e "Sync-on-Cancel will be enabled by default on DNE"
5140 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5141 if combined_mgs_mds; then
5142 skip "needs separate MGS/MDT"
5146 start_mdt 1 || error "MDT0 start fail"
5147 start_ost || error "OST0 start fail"
5148 for num in $(seq 2 $MDSCOUNT); do
5149 start_mdt $num || return
5152 start_ost2 || error "OST1 start fail"
5154 mount_client $MOUNT || error "mount client fails"
5156 mkdir $DIR/$tdir || error "mkdir fail"
5157 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5158 error "create remote dir succeed"
5160 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5161 rm -rf $DIR/$tdir || error "delete dir fail"
5163 umount_client $MOUNT || error "umount_client failed"
5164 stop_mds || error "MDS stop fail"
5165 stop_ost || error "OST0 stop fail"
5166 stop_ost2 || error "OST1 stop fail"
5168 run_test 71a "start MDT0 OST0, MDT1, OST1"
5171 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5172 if combined_mgs_mds; then
5173 skip "needs separate MGS/MDT"
5177 for num in $(seq 2 $MDSCOUNT); do
5178 start_mdt $num || return
5180 start_ost || error "OST0 start fail"
5181 start_mdt 1 || error "MDT0 start fail"
5182 start_ost2 || error "OST1 start fail"
5184 mount_client $MOUNT || error "mount client fails"
5186 mkdir $DIR/$tdir || error "mkdir fail"
5187 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5188 error "create remote dir succeed"
5190 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5191 rm -rf $DIR/$tdir || error "delete dir fail"
5193 umount_client $MOUNT || error "umount_client failed"
5194 stop_mds || error "MDT0 stop fail"
5195 stop_ost || error "OST0 stop fail"
5196 stop_ost2 || error "OST1 stop fail"
5198 run_test 71b "start MDT1, OST0, MDT0, OST1"
5201 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5202 combined_mgs_mds && skip "needs separate MGS/MDT"
5206 start_ost || error "OST0 start fail"
5207 start_ost2 || error "OST1 start fail"
5208 for num in $(seq 2 $MDSCOUNT); do
5209 start_mdt $num || return
5211 start_mdt 1 || error "MDT0 start fail"
5213 mount_client $MOUNT || error "mount client fails"
5215 mkdir $DIR/$tdir || error "mkdir fail"
5216 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5217 error "create remote dir succeed"
5219 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5220 rm -rf $DIR/$tdir || error "delete dir fail"
5222 umount_client $MOUNT || error "umount_client failed"
5223 stop_mds || error "MDS stop fail"
5224 stop_ost || error "OST0 stop fail"
5225 stop_ost2 || error "OST1 stop fail"
5228 run_test 71c "start OST0, OST1, MDT1, MDT0"
5231 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5232 combined_mgs_mds && skip "needs separate MGS/MDT"
5236 start_ost || error "OST0 start fail"
5237 for num in $(seq 2 $MDSCOUNT); do
5238 start_mdt $num || return
5240 start_mdt 1 || error "MDT0 start fail"
5241 start_ost2 || error "OST1 start fail"
5243 mount_client $MOUNT || error "mount client fails"
5245 mkdir $DIR/$tdir || error "mkdir fail"
5246 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5247 error "create remote dir succeed"
5249 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5250 rm -rf $DIR/$tdir || error "delete dir fail"
5252 umount_client $MOUNT || error "umount_client failed"
5253 stop_mds || error "MDS stop fail"
5254 stop_ost || error "OST0 stop fail"
5255 stop_ost2 || error "OST1 stop fail"
5258 run_test 71d "start OST0, MDT1, MDT0, OST1"
5261 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5262 combined_mgs_mds && skip "needs separate MGS/MDT"
5266 start_ost || error "OST0 start fail"
5267 for num in $(seq 2 $MDSCOUNT); do
5268 start_mdt $num || return
5270 start_ost2 || error "OST1 start fail"
5271 start_mdt 1 || error "MDT0 start fail"
5273 mount_client $MOUNT || error "mount client fails"
5275 mkdir $DIR/$tdir || error "mkdir fail"
5276 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5277 error "create remote dir succeed"
5279 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5280 rm -rf $DIR/$tdir || error "delete dir fail"
5282 umount_client $MOUNT || error "umount_client failed"
5283 stop_mds || error "MDS stop fail"
5284 stop_ost || error "OST0 stop fail"
5285 stop_ost2 || error "OST1 stop fail"
5288 run_test 71e "start OST0, MDT1, OST1, MDT0"
5290 test_72() { #LU-2634
5291 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
5292 skip "ldiskfs only test"
5294 local mdsdev=$(mdsdevname 1)
5295 local ostdev=$(ostdevname 1)
5296 local cmd="$E2FSCK -fnvd $mdsdev"
5303 if combined_mgs_mds; then
5304 add_options='--reformat'
5306 add_options='--reformat --replace'
5309 #tune MDT with "-O extents"
5311 for num in $(seq $MDSCOUNT); do
5312 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
5313 $add_options $(mdsdevname $num) $(mdsvdevname $num) ||
5314 error "add mds $num failed"
5315 do_facet mds${num} "$TUNE2FS -O extents $(mdsdevname $num)" ||
5316 error "$TUNE2FS failed on mds${num}"
5319 add ost1 $(mkfs_opts ost1 $ostdev) $add_options $ostdev ||
5320 error "add $ostdev failed"
5321 start_mds || error "start mds failed"
5322 start_ost || error "start ost failed"
5323 mount_client $MOUNT || error "mount client failed"
5325 #create some short symlinks
5326 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5327 createmany -o $DIR/$tdir/$tfile-%d $fn
5328 echo "create $fn short symlinks"
5329 for i in $(seq -w 1 $fn); do
5330 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
5335 umount_client $MOUNT || error "umount client failed"
5336 stop_mds || error "stop mds failed"
5337 stop_ost || error "stop ost failed"
5340 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
5342 run_test 72 "test fast symlink with extents flag enabled"
5344 test_73() { #LU-3006
5345 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
5346 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
5347 error "1st tunefs failed"
5348 start_mgsmds || error "start mds failed"
5349 start_ost || error "start ost failed"
5350 mount_client $MOUNT || error "mount client failed"
5351 $LCTL get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
5352 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
5353 umount_client $MOUNT || error "umount client failed"
5357 run_test 73 "failnode to update from mountdata properly"
5359 test_75() { # LU-2374
5360 [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] &&
5361 skip "Need MDS version at least 2.4.1"
5364 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
5365 --replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
5366 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
5367 --replace --reformat $(ostdevname 1) $(ostvdevname 1)"
5369 #check with default parameters
5370 add mds1 $opts_mds || error "add mds1 failed for default params"
5371 add ost1 $opts_ost || error "add ost1 failed for default params"
5373 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
5374 opts_mds=$(echo $opts_mds |
5375 sed -e "s/--index=$index/--index=$index --mdt/")
5376 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
5377 opts_ost=$(echo $opts_ost |
5378 sed -e "s/--index=$index/--index=$index --ost/")
5380 add mds1 $opts_mds || error "add mds1 failed for new params"
5381 add ost1 $opts_ost || error "add ost1 failed for new params"
5386 run_test 75 "The order of --index should be irrelevant"
5389 [[ $(lustre_version_code mgs) -ge $(version_code 2.4.52) ]] ||
5390 skip "Need MDS version at least 2.4.52"
5393 local MDMB_PARAM="osc.*.max_dirty_mb"
5394 echo "Change MGS params"
5395 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
5397 echo "max_dirty_mb: $MAX_DIRTY_MB"
5398 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB - 10))
5399 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
5400 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
5401 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5402 head -1" $NEW_MAX_DIRTY_MB
5403 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5404 echo "$MAX_DIRTY_MB"
5405 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5406 error "error while apply max_dirty_mb"
5408 echo "Check the value is stored after remount"
5411 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5412 head -1" $NEW_MAX_DIRTY_MB
5413 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5414 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5415 error "max_dirty_mb is not saved after remount"
5417 echo "Change OST params"
5418 CLIENT_PARAM="obdfilter.*.client_cache_count"
5419 local CLIENT_CACHE_COUNT
5420 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5422 echo "client_cache_count: $CLIENT_CACHE_COUNT"
5423 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
5424 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
5425 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
5426 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5427 head -1" $NEW_CLIENT_CACHE_COUNT
5428 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5430 echo "$CLIENT_CACHE_COUNT"
5431 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5432 error "error while apply client_cache_count"
5434 echo "Check the value is stored after remount"
5437 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5438 head -1" $NEW_CLIENT_CACHE_COUNT
5439 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5441 echo "$CLIENT_CACHE_COUNT"
5442 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5443 error "client_cache_count is not saved after remount"
5446 run_test 76a "set permanent params with lctl across mounts"
5448 test_76b() { # LU-4783
5449 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.57) ]] ||
5450 skip "Need MGS version at least 2.5.57"
5453 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
5454 error "start params log failed"
5457 run_test 76b "verify params log setup correctly"
5460 [[ $(lustre_version_code mgs) -ge $(version_code 2.8.54) ]] ||
5461 skip "Need MDS version at least 2.4.52"
5463 local MASK_PARAM="mdd.*.changelog_mask"
5464 echo "Change changelog_mask"
5465 do_facet mgs $LCTL set_param -P $MASK_PARAM=-CLOSE ||
5466 error "Can't change changlog_mask"
5467 wait_update $(facet_host mds) "$LCTL get_param -n $MASK_PARAM |
5470 echo "Check the value is stored after mds remount"
5471 stop_mds || error "Failed to stop MDS"
5472 start_mds || error "Failed to start MDS"
5473 local CHANGELOG_MASK=$(do_facet mgs $LCTL get_param -n $MASK_PARAM)
5474 echo $CHANGELOG_MASK | grep CLOSE > /dev/null &&
5475 error "changelog_mask is not changed"
5479 run_test 76c "verify changelog_mask is applied with lctl set_param -P"
5481 test_76d() { #LU-9399
5484 local xattr_cache="llite.*.xattr_cache"
5485 local cmd="$LCTL get_param -n $xattr_cache | head -1"
5486 local new=$((($(eval $cmd) + 1) % 2))
5488 echo "lctl set_param -P llite.*.xattr_cache=$new"
5489 do_facet mgs $LCTL set_param -P $xattr_cache=$new ||
5490 error "Can't change xattr_cache"
5491 wait_update $HOSTNAME "$cmd" "$new"
5493 echo "Check $xattr_cache on client $MOUNT"
5494 umount_client $MOUNT || error "umount $MOUNT failed"
5495 mount_client $MOUNT || error "mount $MOUNT failed"
5496 [ $(eval $cmd) -eq $new ] ||
5497 error "$xattr_cache != $new on client $MOUNT"
5499 echo "Check $xattr_cache on the new client $MOUNT2"
5500 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
5501 [ $(eval $cmd) -eq $new ] ||
5502 error "$xattr_cache != $new on client $MOUNT2"
5503 umount_client $MOUNT2 || error "umount $MOUNT2 failed"
5507 run_test 76d "verify llite.*.xattr_cache can be set by 'lctl set_param -P' correctly"
5509 test_77() { # LU-3445
5510 local server_version=$(lustre_version_code $SINGLEMDS)
5511 [[ $server_version -ge $(version_code 2.8.55) ]] ||
5512 skip "Need MDS version 2.8.55+ "
5514 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
5515 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
5516 skip_env "mixed loopback and real device not working"
5519 local fs2mdsdev=$(mdsdevname 1_2)
5520 local fs2ostdev=$(ostdevname 1_2)
5521 local fs2mdsvdev=$(mdsvdevname 1_2)
5522 local fs2ostvdev=$(ostvdevname 1_2)
5523 local fsname=test1234
5525 local failnid="$(h2nettype 1.2.3.4),$(h2nettype 4.3.2.1)"
5527 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
5529 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
5530 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
5531 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
5532 error "start fs2mds failed"
5534 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
5535 mgsnid="0.0.0.0@tcp,$mgsnid,$mgsnid:$mgsnid"
5537 add fs2ost --mgsnode=$mgsnid $(mkfs_opts ost1 $fs2ostdev) \
5538 --failnode=$failnid --fsname=$fsname \
5539 --reformat $fs2ostdev $fs2ostvdev ||
5540 error "add fs2ost failed"
5541 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
5543 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
5544 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
5545 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
5548 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
5551 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ||
5552 $(facet_fstype ost1) != ldiskfs ]] &&
5553 skip "ldiskfs only test"
5555 # reformat the Lustre filesystem with a smaller size
5556 local saved_MDSCOUNT=$MDSCOUNT
5557 local saved_MDSSIZE=$MDSSIZE
5558 local saved_OSTCOUNT=$OSTCOUNT
5559 local saved_OSTSIZE=$OSTSIZE
5562 MDSSIZE=$((MDSSIZE - 20000))
5563 OSTSIZE=$((OSTSIZE - 20000))
5564 reformat || error "(1) reformat Lustre filesystem failed"
5565 MDSSIZE=$saved_MDSSIZE
5566 OSTSIZE=$saved_OSTSIZE
5568 # mount the Lustre filesystem
5569 setup_noconfig || error "(2) setup Lustre filesystem failed"
5572 log "create test files"
5577 mkdir $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
5579 for i in $(seq $num_files); do
5580 file=$MOUNT/$tdir/$tfile-$i
5581 dd if=/dev/urandom of=$file count=1 bs=1M || {
5582 $LCTL get_param osc.*.cur*grant*
5583 $LFS df; $LFS df -i;
5584 # stop creating files if there is no more space
5585 if [ ! -e $file ]; then
5586 num_files=$((i - 1))
5590 $LFS getstripe -v $file
5591 local ost_idx=$(LFS getstripe -i $file)
5592 do_facet ost$((ost_idx + 1)) \
5593 $LCTL get_param obdfilter.*.*grant*
5594 error "(4) create $file failed"
5598 # unmount the Lustre filesystem
5599 cleanup || error "(5) cleanup Lustre filesystem failed"
5601 # run e2fsck on the MDT and OST devices
5602 local mds_host=$(facet_active_host $SINGLEMDS)
5603 local ost_host=$(facet_active_host ost1)
5604 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
5605 local ost_dev=$(ostdevname 1)
5607 run_e2fsck $mds_host $mds_dev "-y"
5608 run_e2fsck $ost_host $ost_dev "-y"
5610 # get the original block count of the MDT and OST filesystems
5611 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
5612 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
5614 # expand the MDT and OST filesystems to the device size
5615 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
5616 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
5618 # run e2fsck on the MDT and OST devices again
5619 run_e2fsck $mds_host $mds_dev "-y"
5620 run_e2fsck $ost_host $ost_dev "-y"
5622 # mount the Lustre filesystem
5626 log "check files after expanding the MDT and OST filesystems"
5627 for i in $(seq $num_files); do
5628 file=$MOUNT/$tdir/$tfile-$i
5629 $CHECKSTAT -t file -s 1048576 $file ||
5630 error "(6) checkstat $file failed"
5634 log "create more files after expanding the MDT and OST filesystems"
5635 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
5636 file=$MOUNT/$tdir/$tfile-$i
5637 dd if=/dev/urandom of=$file count=1 bs=1M ||
5638 error "(7) create $file failed"
5641 # unmount the Lustre filesystem
5642 cleanup || error "(8) cleanup Lustre filesystem failed"
5644 # run e2fsck on the MDT and OST devices
5645 run_e2fsck $mds_host $mds_dev "-y"
5646 run_e2fsck $ost_host $ost_dev "-y"
5648 # get the maximum block count of the MDT and OST filesystems
5649 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
5650 local ost_max_blks=$(get_block_count ost1 $ost_dev)
5652 # get the minimum block count of the MDT and OST filesystems
5653 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
5654 grep minimum | sed -e 's/^.*filesystem: //g')
5655 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
5656 grep minimum | sed -e 's/^.*filesystem: //g')
5658 # shrink the MDT and OST filesystems to a smaller size
5662 if [[ $mds_max_blks -gt $mds_min_blks &&
5663 $mds_max_blks -gt $mds_orig_blks ]]; then
5664 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
5665 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
5666 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
5667 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
5668 error "shrink $SINGLEMDS to $new_blks failed"
5672 if [[ $ost_max_blks -gt $ost_min_blks &&
5673 $ost_max_blks -gt $ost_orig_blks ]]; then
5674 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
5675 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
5676 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
5677 run_resize2fs ost1 $ost_dev $new_blks ||
5678 error "shrink ost1 to $new_blks failed"
5682 # check whether the MDT or OST filesystem was shrunk or not
5684 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
5685 reformat_and_config ||
5686 error "(10) reformat Lustre filesystem failed"
5690 # run e2fsck on the MDT and OST devices again
5691 run_e2fsck $mds_host $mds_dev "-y"
5692 run_e2fsck $ost_host $ost_dev "-y"
5694 # mount the Lustre filesystem again
5698 log "check files after shrinking the MDT and OST filesystems"
5699 for i in $(seq $((num_files + 10))); do
5700 file=$MOUNT/$tdir/$tfile-$i
5701 $CHECKSTAT -t file -s 1048576 $file ||
5702 error "(11) checkstat $file failed"
5705 # unmount and reformat the Lustre filesystem
5706 cleanup || error "(12) cleanup Lustre filesystem failed"
5707 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
5709 MDSCOUNT=$saved_MDSCOUNT
5710 OSTCOUNT=$saved_OSTCOUNT
5711 reformat_and_config || error "(14) reformat Lustre filesystem failed"
5713 run_test 78 "run resize2fs on MDT and OST filesystems"
5715 test_79() { # LU-4227
5716 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.5.59) ]] ||
5717 skip "Need MDS version at least 2.5.59"
5719 local mdsdev1=$(mdsdevname 1)
5720 local mdsvdev1=$(mdsvdevname 1)
5721 local mdsdev2=$(mdsdevname 2)
5722 local mdsvdev2=$(mdsvdevname 2)
5723 local ostdev1=$(ostdevname 1)
5724 local ostvdev1=$(ostvdevname 1)
5725 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
5726 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
5727 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
5730 # remove --mgs/--mgsnode from mkfs.lustre options
5731 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
5733 mgsnode_opt=$(echo $opts_mds2 |
5734 awk '{ for ( i = 1; i < NF; i++ )
5735 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5736 [ -n $mgsnode_opt ] &&
5737 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
5739 mgsnode_opt=$(echo $opts_ost1 |
5740 awk '{ for ( i = 1; i < NF; i++ )
5741 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5742 [ -n $mgsnode_opt ] &&
5743 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
5745 # -MGS, format a mdt without --mgs option
5746 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
5747 error "Must specify --mgs when formatting mdt combined with mgs"
5749 # +MGS, format a mdt/ost without --mgsnode option
5750 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
5751 > /dev/null || error "start mds1 failed"
5752 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
5753 error "Must specify --mgsnode when formatting a mdt"
5754 add ost1 $opts_ost1 $ostdev1 $ostvdev2 &&
5755 error "Must specify --mgsnode when formatting an ost"
5759 run_test 79 "format MDT/OST without mgs option (should return errors)"
5762 start_mds || error "Failed to start MDT"
5763 start_ost || error "Failed to start OST1"
5764 uuid=$(do_facet ost1 $LCTL get_param -n mgc.*.uuid)
5765 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
5766 do_facet ost1 "$LCTL set_param fail_val=10 fail_loc=0x906"
5767 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x906"
5768 do_facet mgs "$LCTL set_param -n mgs/MGS/evict_client $uuid"
5770 start_ost2 || error "Failed to start OST2"
5772 do_facet ost1 "$LCTL set_param fail_loc=0"
5777 run_test 80 "mgc import reconnect race"
5779 #Save the original values of $OSTCOUNT and $OSTINDEX$i.
5781 local new_ostcount=$1
5782 saved_ostcount=$OSTCOUNT
5783 OSTCOUNT=$new_ostcount
5787 for i in $(seq $OSTCOUNT); do
5789 eval saved_ostindex$i=${!index}
5794 # Restore the original values of $OSTCOUNT and $OSTINDEX$i.
5795 restore_ostindex() {
5800 for i in $(seq $OSTCOUNT); do
5801 index=saved_ostindex$i
5802 eval OSTINDEX$i=${!index}
5804 OSTCOUNT=$saved_ostcount
5807 if ! combined_mgs_mds ; then
5812 # The main purpose of this test is to ensure the OST_INDEX_LIST functions as
5813 # expected. This test uses OST_INDEX_LIST to format OSTs with a randomly
5814 # assigned index and ensures we can mount such a formatted file system
5815 test_81() { # LU-4665
5816 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5817 skip "Need MDS version at least 2.6.54"
5818 [[ $OSTCOUNT -ge 3 ]] || skip_env "needs >= 3 OSTs"
5822 # Each time RANDOM is referenced, a random integer between 0 and 32767
5825 local saved_ostindex1=$OSTINDEX1
5826 for i in 65535 $((RANDOM + 65536)); do
5827 echo -e "\nFormat ost1 with --index=$i, should fail"
5829 if add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat \
5830 $(ostdevname 1) $(ostvdevname 1); then
5831 OSTINDEX1=$saved_ostindex1
5832 error "format ost1 with --index=$i should fail"
5835 OSTINDEX1=$saved_ostindex1
5839 # Format OSTs with random sparse indices.
5840 trap "restore_ostindex" EXIT
5841 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices"
5842 OST_INDEX_LIST=[0,$((RANDOM * 2 % 65533 + 1)),65534] formatall
5844 # Setup and check Lustre filesystem.
5845 start_mgsmds || error "start_mgsmds failed"
5846 for i in $(seq $OSTCOUNT); do
5847 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5848 error "start ost$i failed"
5851 mount_client $MOUNT || error "mount client $MOUNT failed"
5852 check_mount || error "check client $MOUNT failed"
5855 local max_easize=$($LCTL get_param -n llite.*.max_easize)
5856 [[ $max_easize -eq 128 ]] ||
5857 error "max_easize is $max_easize, should be 128 bytes"
5861 run_test 81 "sparse OST indexing"
5863 # Here we exercise the stripe placement functionality on a file system that
5864 # has formatted the OST with a random index. With the file system the following
5865 # functionality is tested:
5867 # 1. Creating a new file with a specific stripe layout.
5869 # 2. Modifiy a existing empty file with a specific stripe layout.
5871 # 3. Ensure we fail to set the stripe layout of a file that already has one.
5873 # 4. If ost-index is defined we need to ensure it is the first entry in the
5874 # ost index list returned by lfs getstripe.
5876 # 5. Lastly ensure this functionality fails with directories.
5877 test_82a() { # LU-4665
5878 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5879 skip "Need MDS version at least 2.6.54"
5880 [[ $OSTCOUNT -ge 3 ]] || skip_env "needs >= 3 OSTs"
5886 # Format OSTs with random sparse indices.
5890 local LOV_V1_INSANE_STRIPE_COUNT=65532
5891 for i in $(seq $OSTCOUNT); do
5892 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
5893 ost_indices+=" $index"
5895 ost_indices=$(comma_list $ost_indices)
5897 stack_trap "restore_ostindex" EXIT
5898 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5899 OST_INDEX_LIST=[$ost_indices] formatall
5901 # Setup Lustre filesystem.
5902 start_mgsmds || error "start_mgsmds failed"
5903 for i in $(seq $OSTCOUNT); do
5904 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5905 error "start ost$i failed"
5908 # Collect debug information - start of test
5909 do_nodes $(comma_list $(mdts_nodes)) \
5910 $LCTL get_param osp.*.prealloc_*_id
5912 mount_client $MOUNT || error "mount client $MOUNT failed"
5915 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5916 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5918 stack_trap "do_nodes $(comma_list $(mdts_nodes)) \
5919 $LCTL get_param osp.*.prealloc_*_id" EXIT
5921 # 1. If the file does not exist, new file will be created
5922 # with specified OSTs.
5923 local file=$DIR/$tdir/$tfile-1
5924 local cmd="$SETSTRIPE -o $ost_indices $file"
5926 eval $cmd || error "$cmd failed"
5927 check_stripe_count $file $OSTCOUNT
5928 check_obdidx $file $ost_indices
5929 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5930 error "write $file failed"
5932 # 2. If the file already exists and is an empty file, the file
5933 # will be attached with specified layout.
5934 file=$DIR/$tdir/$tfile-2
5935 mcreate $file || error "mcreate $file failed"
5936 cmd="$SETSTRIPE -o $ost_indices $file"
5938 eval $cmd || error "$cmd failed"
5939 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5940 error "write $file failed"
5941 check_stripe_count $file $OSTCOUNT
5942 check_obdidx $file $ost_indices
5944 # 3. If the file already has a valid layout attached, the command
5945 # should fail with EBUSY.
5947 eval $cmd && error "stripe is already set on $file, $cmd should fail"
5949 # 4. If [--stripe-index|-i <start_ost_idx>] is used, the index must
5950 # be in the OST indices list.
5951 local start_ost_idx=${ost_indices##*,}
5952 file=$DIR/$tdir/$tfile-3
5953 cmd="$SETSTRIPE -o $ost_indices -i $start_ost_idx $file"
5955 eval $cmd || error "$cmd failed"
5956 check_stripe_count $file $OSTCOUNT
5957 check_obdidx $file $ost_indices
5958 check_start_ost_idx $file $start_ost_idx
5960 file=$DIR/$tdir/$tfile-4
5962 cmd+=" -o $(exclude_items_from_list $ost_indices $start_ost_idx)"
5963 cmd+=" -i $start_ost_idx $file"
5965 eval $cmd && error "index $start_ost_idx should be in $ost_indices"
5967 # 5. Specifying OST indices for directory should succeed.
5968 local dir=$DIR/$tdir/$tdir
5969 mkdir $dir || error "mkdir $dir failed"
5970 cmd="$SETSTRIPE -o $ost_indices $dir"
5971 if [[ $(lustre_version_code $SINGLEMDS) -gt $(version_code 2.11.53) &&
5972 $(lustre_version_code client -gt $(version_code 2.11.53)) ]]; then
5974 eval $cmd || error "unable to specify OST indices on directory"
5976 echo "need MDS+client version at least 2.11.53"
5981 run_test 82a "specify OSTs for file (succeed) or directory (succeed)"
5986 # Remove OSTs from a pool and destroy the pool.
5987 destroy_pool $ost_pool || true
5992 # Test 82b is run to ensure that if the user supplies a pool with a specific
5993 # stripe layout that it behaves proprerly. It should fail in the case that
5994 # the supplied OST index list points to OSTs not contained in the user
5996 test_82b() { # LU-4665
5997 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5998 skip "Need MDS version at least 2.6.54"
5999 [[ $OSTCOUNT -ge 4 ]] || skip_env "needs >= 4 OSTs"
6005 # Format OSTs with random sparse indices.
6009 local LOV_V1_INSANE_STRIPE_COUNT=65532
6010 for i in $(seq $OSTCOUNT); do
6011 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
6012 ost_indices+=" $index"
6014 ost_indices=$(comma_list $ost_indices)
6016 trap "restore_ostindex" EXIT
6017 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
6018 OST_INDEX_LIST=[$ost_indices] formatall
6020 # Setup Lustre filesystem.
6021 start_mgsmds || error "start_mgsmds failed"
6022 for i in $(seq $OSTCOUNT); do
6023 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6024 error "start ost$i failed"
6027 mount_client $MOUNT || error "mount client $MOUNT failed"
6030 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
6031 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6033 # Create a new pool and add OSTs into it.
6034 local ost_pool=$FSNAME.$TESTNAME
6035 create_pool $ost_pool || error "create OST pool $ost_pool failed"
6038 trap "cleanup_82b" EXIT
6040 local ost_idx_in_list=${ost_indices##*,}
6041 local ost_idx_in_pool=$(exclude_items_from_list $ost_indices \
6044 local ost_targets="$FSNAME-OST["
6045 for i in ${ost_idx_in_pool//,/ }; do
6046 ost_targets=$ost_targets$(printf "%04x," $i)
6048 ost_targets="${ost_targets%,}]"
6050 local ost_targets_uuid=$(for i in ${ost_idx_in_pool//,/ }; \
6051 do printf "$FSNAME-OST%04x_UUID\n" $i; done |
6052 sort -u | tr '\n' ' ')
6054 local cmd="$LCTL pool_add $ost_pool $ost_targets"
6055 do_facet mgs $cmd || error "$cmd failed"
6056 wait_update $HOSTNAME "$LCTL get_param -n lov.$FSNAME-*.pools.$TESTNAME|
6057 sort -u | tr '\n' ' ' " "$ost_targets_uuid" ||
6058 error "wait_update $ost_pool failed"
6059 wait_update_facet $SINGLEMDS "$LCTL pool_list $ost_pool | wc -l" 4 ||
6060 error "wait_update pool_list $ost_pool failed"
6062 # If [--pool|-p <pool_name>] is set with [--ost-list|-o <ost_indices>],
6063 # then the OSTs must be the members of the pool.
6064 local file=$DIR/$tdir/$tfile
6065 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
6067 eval $cmd && error "OST with index $ost_idx_in_list should be" \
6068 "in OST pool $ost_pool"
6070 # Only select OST $ost_idx_in_list from $ost_pool for file.
6071 ost_idx_in_list=${ost_idx_in_pool#*,}
6072 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
6074 eval $cmd || error "$cmd failed"
6075 cmd="$GETSTRIPE $file"
6077 eval $cmd || error "$cmd failed"
6078 check_stripe_count $file 2
6079 check_obdidx $file $ost_idx_in_list
6080 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6081 error "write $file failed"
6085 run_test 82b "specify OSTs for file with --pool and --ost-list options"
6088 [[ $(lustre_version_code ost1) -ge $(version_code 2.6.91) ]] ||
6089 skip "Need OST version at least 2.6.91"
6090 if [ $(facet_fstype ost1) != ldiskfs ]; then
6091 skip "ldiskfs only test"
6100 ostmnt=$(facet_mntpt ost1)
6101 fstype=$(facet_fstype ost1)
6103 # Mount the OST as an ldiskfs filesystem.
6104 log "mount the OST $dev as a $fstype filesystem"
6105 add ost1 $(mkfs_opts ost1 $dev) $FSTYPE_OPT \
6106 --reformat $dev > /dev/null ||
6107 error "format ost1 error"
6109 if ! test -b $dev; then
6110 mnt_opts=$(csa_add "$OST_MOUNT_OPTS" -o loop)
6112 echo "mnt_opts $mnt_opts"
6113 do_facet ost1 mount -t $fstype $dev \
6115 # Run llverfs on the mounted ldiskfs filesystem.
6116 # It is needed to get ENOSPACE.
6117 log "run llverfs in partial mode on the OST $fstype $ostmnt"
6118 do_rpc_nodes $(facet_host ost1) run_llverfs $ostmnt -vpl \
6119 "no" || error "run_llverfs error on $fstype"
6122 log "unmount the OST $dev"
6125 # Delete file IO_scrub. Later osd_scrub_setup will try to
6126 # create "IO_scrub" but will get ENOSPACE.
6128 echo "start ost1 service on `facet_active_host ost1`"
6129 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
6132 err=$(do_facet ost1 dmesg | grep "VFS: Busy inodes after unmount of")
6133 echo "string err $err"
6134 [ -z "$err" ] || error $err
6137 run_test 83 "ENOSPACE on OST doesn't cause message VFS: \
6138 Busy inodes after unmount ..."
6141 local facet=$SINGLEMDS
6142 local num=$(echo $facet | tr -d "mds")
6143 local dev=$(mdsdevname $num)
6144 local time_min=$(recovery_time_min)
6145 local recovery_duration
6146 local completed_clients
6147 local correct_clients
6150 echo "start mds service on $(facet_active_host $facet)"
6152 "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
6153 error "start MDS failed"
6155 start_ost || error "start OST0000 failed"
6156 start_ost2 || error "start OST0001 failed"
6158 echo "recovery_time=$time_min, timeout=$TIMEOUT, wrap_up=$wrap_up"
6160 mount_client $MOUNT1 || error "mount $MOUNT1 failed"
6161 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
6162 # make sure new superblock labels are sync'd before disabling writes
6166 replay_barrier $SINGLEMDS
6167 createmany -o $DIR1/$tfile-%d 1000
6169 # We need to catch the end of recovery window to extend it.
6170 # Skip 5 requests and add delay to request handling.
6171 #define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP
6172 do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5"
6174 facet_failover --fsck $SINGLEMDS || error "failover: $?"
6177 echo "recovery status"
6178 do_facet $SINGLEMDS \
6179 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status"
6181 recovery_duration=$(do_facet $SINGLEMDS \
6182 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
6183 awk '/recovery_duration/ { print $2 }')
6184 (( $recovery_duration > $time_min + $wrap_up )) &&
6185 error "recovery_duration > recovery_time_hard + wrap up"
6186 completed_clients=$(do_facet $SINGLEMDS \
6187 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
6188 awk '/completed_clients/ { print $2 }')
6190 correct_clients="$MDSCOUNT/$((MDSCOUNT+1))"
6191 [ "$completed_clients" = "${correct_clients}" ] ||
6192 error "$completed_clients != $correct_clients"
6194 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
6195 umount_client $MOUNT1
6196 umount_client $MOUNT2
6202 run_test 84 "check recovery_hard_time"
6205 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.55) ]] ||
6206 skip "Need OST version at least 2.7.55"
6207 ##define OBD_FAIL_OSD_OST_EA_FID_SET 0x197
6208 do_facet ost1 "lctl set_param fail_loc=0x197"
6212 run_test 85 "osd_ost init: fail ea_fid_set"
6217 # ost1 has already registered to the MGS before the reformat.
6218 # So after reformatting it with option "-G", it could not be
6219 # mounted to the MGS. Cleanup the system for subsequent tests.
6224 local server_version=$(lustre_version_code $SINGLEMDS)
6225 [ "$(facet_fstype ost1)" = "zfs" ] &&
6226 skip "LU-6442: no such mkfs params for ZFS OSTs"
6227 [[ $server_version -ge $(version_code 2.7.56) ]] ||
6228 skip "Need server version newer than 2.7.55"
6230 local OST_OPTS="$(mkfs_opts ost1 $(ostdevname 1)) \
6231 --reformat $(ostdevname 1) $(ostvdevname 1)"
6234 local OLDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6235 awk '/Flex block group size: / { print $NF; exit; }')
6238 if [[ ${!opts} != *mkfsoptions* ]]; then
6239 eval opts=\"${!opts} \
6240 --mkfsoptions='\\\"-O flex_bg -G $NEWSIZE\\\"'\"
6242 val=${!opts//--mkfsoptions=\\\"/ \
6243 --mkfsoptions=\\\"-O flex_bg -G $NEWSIZE }
6247 echo "params: $opts"
6249 trap cleanup_86 EXIT ERR
6252 add ost1 $opts || error "add ost1 failed with new params"
6254 local FOUNDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6255 awk '/Flex block group size: / { print $NF; exit; }')
6257 [[ $FOUNDSIZE == $NEWSIZE ]] ||
6258 error "Flex block group size: $FOUNDSIZE, expected: $NEWSIZE"
6262 run_test 86 "Replacing mkfs.lustre -G option"
6264 test_87() { #LU-6544
6265 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.9.51) ]] ||
6266 skip "Need MDS version at least 2.9.51"
6267 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ]] &&
6268 skip "ldiskfs only test"
6269 [[ $OSTCOUNT -gt 59 ]] &&
6270 skip "Ignore wide striping situation"
6272 local mdsdev=$(mdsdevname 1)
6273 local mdsvdev=$(mdsvdevname 1)
6274 local file=$DIR/$tfile
6275 local mntpt=$(facet_mntpt $SINGLEMDS)
6276 local used_xattr_blk=0
6277 local inode_size=${1:-1024}
6279 local xtest="trusted.test"
6283 local stripe_cnt=$(($OSTCOUNT + 2))
6285 #Please see ldiskfs_make_lustre() for MDT inode size calculation
6286 if [ $stripe_cnt -gt 16 ]; then
6289 left_size=$(expr $inode_size - \
6292 32 - 40 \* 3 - 32 \* 3 - $stripe_cnt \* 24 - 16 - 3 - \
6294 24 - 18 - $(expr length $tfile) - 16 - 4)
6295 if [ $left_size -le 0 ]; then
6296 echo "No space($left_size) is expected in inode."
6297 echo "Try 1-byte xattr instead to verify this."
6300 echo "Estimate: at most $left_size-byte space left in inode."
6306 if ! combined_mgs_mds ; then
6310 add mds1 $(mkfs_opts mds1 ${mdsdev}) --stripe-count-hint=$stripe_cnt \
6311 --reformat $mdsdev $mdsvdev || error "add mds1 failed"
6312 start_mdt 1 > /dev/null || error "start mdt1 failed"
6313 for i in $(seq $OSTCOUNT); do
6314 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS > /dev/null ||
6315 error "start ost$i failed"
6317 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
6318 check_mount || error "check client $MOUNT failed"
6321 $SETSTRIPE -E 1M -S 1M -c 1 -E 64M -c 1 -E -1 -c -1 $file ||
6322 error "Create file with 3 components failed"
6323 $TRUNCATE $file $((1024*1024*64+1)) || error "truncate file failed"
6324 i=$($GETSTRIPE -I3 -c $file) || error "get 3rd stripe count failed"
6325 if [ $i -ne $OSTCOUNT ]; then
6326 left_size=$(expr $left_size + $(expr $OSTCOUNT - $i) \* 24)
6327 echo -n "Since only $i out $OSTCOUNT OSTs are used, "
6328 echo -n "the expected left space is changed to "
6329 echo "$left_size bytes at most."
6331 value=$(generate_string $left_size)
6332 setfattr -n $xtest -v $value $file
6333 orig=$(get_xattr_value $xtest $file)
6334 [[ "$orig" != "$value" ]] && error "$xtest changed"
6336 #Verify if inode has some expected space left
6337 umount $MOUNT > /dev/null || error "umount $MOUNT failed"
6338 stop_mdt 1 > /dev/null || error "stop mdt1 failed"
6339 mount_ldiskfs $SINGLEMDS || error "mount -t ldiskfs $SINGLEMDS failed"
6341 do_facet $SINGLEMDS ls -sal $mntpt/ROOT/$tfile
6342 used_xattr_blk=$(do_facet $SINGLEMDS ls -s $mntpt/ROOT/$tfile |
6344 [[ $used_xattr_blk -eq 0 ]] &&
6345 error "Please check MDS inode size calculation: \
6346 more than $left_size-byte space left in inode."
6347 echo "Verified: at most $left_size-byte space left in inode."
6349 umount_ldiskfs $SINGLEMDS
6351 for i in $(seq $OSTCOUNT); do
6352 stop ost$i -f || error "stop ost$i failed"
6355 run_test 87 "check if MDT inode can hold EAs with N stripes properly"
6358 [ "$(facet_fstype mds1)" == "zfs" ] &&
6359 skip "LU-6662: no implementation for ZFS"
6363 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
6364 --reformat $(mdsdevname 1) || error "add mds1 failed"
6366 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
6367 grep -e \".*opts:.*errors=remount-ro.*\"" ||
6368 error "default mount options is missing"
6370 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
6371 --mountfsoptions="user_xattr,errors=panic" \
6372 --reformat $(mdsdevname 1) || error "add mds1 failed"
6374 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
6375 grep -e \".*opts:.*errors=panic.*\"" ||
6376 error "user can't override default mount options"
6378 run_test 88 "check the default mount options can be overridden"
6380 test_89() { # LU-7131
6381 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.9.54) ]] ||
6382 skip "Need MDT version at least 2.9.54"
6384 local key=failover.node
6385 local val1=192.0.2.254@tcp0 # Reserved IPs, see RFC 5735
6386 local val2=192.0.2.255@tcp0
6387 local mdsdev=$(mdsdevname 1)
6392 [ $(facet_fstype mds1) == zfs ] && import_zpool mds1
6393 # Check that parameters are added correctly
6394 echo "tunefs --param $key=$val1"
6395 do_facet mds "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
6396 error "tunefs --param $key=$val1 failed"
6397 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6398 error "tunefs --dryrun failed"
6399 params=${params##*Parameters:}
6400 params=${params%%exiting*}
6401 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val1) = "1" ] ||
6402 error "on-disk parameter not added correctly via tunefs"
6404 # Check that parameters replace existing instances when added
6405 echo "tunefs --param $key=$val2"
6406 do_facet mds "$TUNEFS --param $key=$val2 $mdsdev >/dev/null" ||
6407 error "tunefs --param $key=$val2 failed"
6408 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6409 error "tunefs --dryrun failed"
6410 params=${params##*Parameters:}
6411 params=${params%%exiting*}
6412 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "1" ] ||
6413 error "on-disk parameter not replaced via tunefs"
6414 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val2) = "1" ] ||
6415 error "on-disk parameter not replaced correctly via tunefs"
6417 # Check that a parameter is erased properly
6418 echo "tunefs --erase-param $key"
6419 do_facet mds "$TUNEFS --erase-param $key $mdsdev >/dev/null" ||
6420 error "tunefs --erase-param $key failed"
6421 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6422 error "tunefs --dryrun failed"
6423 params=${params##*Parameters:}
6424 params=${params%%exiting*}
6425 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "0" ] ||
6426 error "on-disk parameter not erased correctly via tunefs"
6428 # Check that all the parameters are erased
6429 echo "tunefs --erase-params"
6430 do_facet mds "$TUNEFS --erase-params $mdsdev >/dev/null" ||
6431 error "tunefs --erase-params failed"
6432 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6433 error "tunefs --dryrun failed"
6434 params=${params##*Parameters:}
6435 params=${params%%exiting*}
6437 error "all on-disk parameters not erased correctly via tunefs"
6439 # Check the order of options --erase-params and --param
6440 echo "tunefs --param $key=$val1 --erase-params"
6442 "$TUNEFS --param $key=$val1 --erase-params $mdsdev >/dev/null"||
6443 error "tunefs --param $key=$val1 --erase-params failed"
6444 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6445 error "tunefs --dryrun failed"
6446 params=${params##*Parameters:}
6447 params=${params%%exiting*}
6448 [ $(echo $params | tr ' ' '\n') == "$key=$val1" ] ||
6449 error "on-disk param not added correctly with --erase-params"
6453 run_test 89 "check tunefs --param and --erase-param{s} options"
6456 # $2 (optional) value of max_mod_rpcs_in_flight to set
6457 check_max_mod_rpcs_in_flight() {
6465 idx=$(printf "%04x" $($LFS getdirstripe -i $dir))
6466 facet="mds$((0x$idx + 1))"
6468 if [ -z "$mmr" ]; then
6469 # get value of max_mod_rcps_in_flight
6470 mmr=$($LCTL get_param -n \
6471 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) ||
6472 error "Unable to get max_mod_rpcs_in_flight"
6473 echo "max_mod_rcps_in_flight is $mmr"
6475 # set value of max_mod_rpcs_in_flight
6477 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight=$mmr ||
6478 error "Unable to set max_mod_rpcs_in_flight to $mmr"
6479 echo "max_mod_rpcs_in_flight set to $mmr"
6482 # create mmr+1 files
6483 echo "creating $((mmr + 1)) files ..."
6485 for i in $(seq $((mmr + 1))); do
6491 # consumes mmr-1 modify RPC slots
6492 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6493 # drop requests on MDT so that RPC slots are consumed
6494 # during all the request resend interval
6495 do_facet $facet "$LCTL set_param fail_loc=0x159"
6496 echo "launch $((mmr - 1)) chmod in parallel ..."
6497 for i in $(seq $((mmr - 1))); do
6498 chmod 0600 $dir/file-$i &
6502 # send one additional modify RPC
6503 do_facet $facet "$LCTL set_param fail_loc=0"
6504 echo "launch 1 additional chmod in parallel ..."
6505 chmod 0600 $dir/file-$mmr &
6508 # check this additional modify RPC get a modify RPC slot
6509 # and succeed its operation
6510 checkstat -vp 0600 $dir/file-$mmr ||
6511 error "Unable to send $mmr modify RPCs in parallel"
6516 # consumes mmr modify RPC slots
6517 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6518 # drop requests on MDT so that RPC slots are consumed
6519 # during all the request resend interval
6520 do_facet $facet "$LCTL set_param fail_loc=0x159"
6521 echo "launch $mmr chmod in parallel ..."
6522 for i in $(seq $mmr); do
6523 chmod 0666 $dir/file-$i &
6527 # send one additional modify RPC
6528 do_facet $facet "$LCTL set_param fail_loc=0"
6529 echo "launch 1 additional chmod in parallel ..."
6530 chmod 0666 $dir/file-$((mmr + 1)) &
6533 # check this additional modify RPC blocked getting a modify RPC slot
6534 checkstat -vp 0644 $dir/file-$((mmr + 1)) ||
6535 error "Unexpectedly send $(($mmr + 1)) modify RPCs in parallel"
6542 [[ $($LCTL get_param mdc.*.import |
6543 grep "connect_flags:.*multi_mod_rpc") ]] ||
6544 skip "Need MDC with 'multi_mod_rpcs' feature"
6546 # check default value
6547 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6548 check_max_mod_rpcs_in_flight $DIR/$tdir
6552 run_test 90a "check max_mod_rpcs_in_flight is enforced"
6562 [[ $($LCTL get_param mdc.*.import |
6563 grep "connect_flags:.*multi_mod_rpc") ]] ||
6564 skip "Need MDC with 'multi_mod_rpcs' feature"
6567 # update max_mod_rpcs_in_flight
6568 $LFS mkdir -c1 $DIR/${tdir}1 || error "mkdir $DIR/${tdir}1 failed"
6569 check_max_mod_rpcs_in_flight $DIR/${tdir}1 1
6572 # check client is able to send multiple modify RPCs in paralell
6573 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
6574 grep -c "multi_mod_rpcs")
6575 if [ "$tmp" -ne $MDSCOUNT ]; then
6577 skip "Client not able to send multiple modify RPCs in parallel"
6580 # update max_mod_rpcs_in_flight
6581 $LFS mkdir -c1 $DIR/${tdir}2 || error "mkdir $DIR/${tdir}2 failed"
6582 check_max_mod_rpcs_in_flight $DIR/${tdir}2 5
6585 $LFS mkdir -c1 $DIR/${tdir}3 || error "mkdir $DIR/${tdir}3 failed"
6586 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/${tdir}3))
6587 facet="mds$((0x$idx + 1))"
6589 # save MDT max_mod_rpcs_per_client
6590 mmrpc=$(do_facet $facet \
6591 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
6593 # update max_mod_rpcs_in_flight
6594 umount_client $MOUNT
6596 "echo 16 > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6598 $LCTL set_param mdc.$FSNAME-MDT$idx-mdc-*.max_rpcs_in_flight=17
6599 check_max_mod_rpcs_in_flight $DIR/${tdir}3 16
6601 # restore MDT max_mod_rpcs_per_client initial value
6603 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6605 rm -rf $DIR/${tdir}?
6608 run_test 90b "check max_mod_rpcs_in_flight is enforced after update"
6617 [[ $($LCTL get_param mdc.*.import |
6618 grep "connect_flags:.*multi_mod_rpc") ]] ||
6619 skip "Need MDC with 'multi_mod_rpcs' feature"
6621 # check client is able to send multiple modify RPCs in paralell
6622 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
6623 grep -c "multi_mod_rpcs")
6624 if [ "$tmp" -ne $MDSCOUNT ]; then
6626 skip "Client not able to send multiple modify RPCs in parallel"
6629 # get max_rpcs_in_flight value
6630 mrif=$($LCTL get_param -n mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
6631 echo "max_rpcs_in_flight is $mrif"
6633 # get MDT max_mod_rpcs_per_client
6634 mmrpc=$(do_facet mds1 \
6635 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
6636 echo "max_mod_rpcs_per_client is $mmrpc"
6639 # attempt to set max_mod_rpcs_in_flight to max_rpcs_in_flight value
6640 # prerequisite: set max_mod_rpcs_per_client to max_rpcs_in_flight value
6641 umount_client $MOUNT
6643 "echo $mrif > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6647 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif &&
6648 error "set max_mod_rpcs_in_flight to $mrif should fail"
6650 umount_client $MOUNT
6652 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6656 # attempt to set max_mod_rpcs_in_flight to max_mod_rpcs_per_client+1
6657 # prerequisite: set max_rpcs_in_flight to max_mod_rpcs_per_client+2
6659 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$((mmrpc + 2))
6662 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$((mmrpc + 1)) &&
6663 error "set max_mod_rpcs_in_flight to $((mmrpc + 1)) should fail"
6667 run_test 90c "check max_mod_rpcs_in_flight update limits"
6678 [[ $($LCTL get_param mdc.*.import |
6679 grep "connect_flags:.*multi_mod_rpc") ]] ||
6680 skip "Need MDC with 'multi_mod_rpcs' feature"
6682 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6683 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
6684 facet="mds$((0x$idx + 1))"
6686 # check client version supports multislots
6687 tmp=$($LCTL get_param -N \
6688 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
6689 if [ -z "$tmp" ]; then
6691 skip "Client does not support multiple modify RPCs in flight"
6694 # get current value of max_mod_rcps_in_flight
6695 mmr=$($LCTL get_param -n \
6696 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
6697 echo "max_mod_rcps_in_flight is $mmr"
6700 echo "creating $mmr files ..."
6702 for i in $(seq $mmr); do
6703 touch $DIR/$tdir/file-$i
6706 # prepare for close RPC
6707 multiop_bg_pause $DIR/$tdir/file-close O_c
6710 # consumes mmr modify RPC slots
6711 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6712 # drop requests on MDT so that RPC slots are consumed
6713 # during all the request resend interval
6714 do_facet $facet "$LCTL set_param fail_loc=0x159"
6715 echo "launch $mmr chmod in parallel ..."
6716 for i in $(seq $mmr); do
6717 chmod 0600 $DIR/$tdir/file-$i &
6720 # send one additional close RPC
6721 do_facet $facet "$LCTL set_param fail_loc=0"
6722 echo "launch 1 additional close in parallel ..."
6724 cancel_lru_locks mdc
6727 # check this additional close RPC get a modify RPC slot
6728 # and multiop process completed
6729 [ -d /proc/$pid ] &&
6730 error "Unable to send the additional close RPC in parallel"
6735 run_test 90d "check one close RPC is allowed above max_mod_rpcs_in_flight"
6737 check_uuid_on_ost() {
6739 do_facet ost1 "$LCTL get_param obdfilter.${FSNAME}*.exports.'$nid'.uuid"
6742 check_uuid_on_mdt() {
6744 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
6752 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.63) ]] ||
6753 skip "Need OST version at least 2.7.63"
6754 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.63) ]] ||
6755 skip "Need MDT version at least 2.7.63"
6757 start_mds || error "MDS start failed"
6758 start_ost || error "unable to start OST"
6759 mount_client $MOUNT || error "client start failed"
6760 check_mount || error "check_mount failed"
6763 nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g")
6767 uuid=$(get_client_uuid $MOUNT)
6769 echo "list nids on mdt:"
6770 do_facet $SINGLEMDS "$LCTL list_param mdt.${FSNAME}*.exports.*"
6771 echo "uuid from $nid:"
6772 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
6774 found=$(check_uuid_on_mdt $nid | grep $uuid)
6775 [ -z "$found" ] && error "can't find $uuid $nid on MDT"
6776 found=$(check_uuid_on_ost $nid | grep $uuid)
6777 [ -z "$found" ] && error "can't find $uuid $nid on OST"
6779 # umount the client so it won't reconnect
6780 manual_umount_client --force || error "failed to umount $?"
6781 # shouldn't disappear on MDS after forced umount
6782 found=$(check_uuid_on_mdt $nid | grep $uuid)
6783 [ -z "$found" ] && error "can't find $uuid $nid"
6786 do_facet $SINGLEMDS \
6787 "$LCTL set_param -n mdt.${mds1_svc}.evict_client nid:$nid"
6789 found=$(check_uuid_on_mdt $nid | grep $uuid)
6790 [ -n "$found" ] && error "found $uuid $nid on MDT"
6791 found=$(check_uuid_on_ost $nid | grep $uuid)
6792 [ -n "$found" ] && error "found $uuid $nid on OST"
6794 # check it didn't reconnect (being umounted)
6795 sleep $((TIMEOUT+1))
6796 found=$(check_uuid_on_mdt $nid | grep $uuid)
6797 [ -n "$found" ] && error "found $uuid $nid on MDT"
6798 found=$(check_uuid_on_ost $nid | grep $uuid)
6799 [ -n "$found" ] && error "found $uuid $nid on OST"
6803 run_test 91 "evict-by-nid support"
6805 generate_ldev_conf() {
6806 # generate an ldev.conf file
6807 local ldevconfpath=$1
6809 local fsldevformat=""
6812 fstype=$(facet_fstype mgs)
6813 if [ "$fstype" == "zfs" ]; then
6814 fsldevformat="$fstype:"
6819 printf "%s\t-\t%s-MGS0000\t%s%s\n" \
6823 $(mgsdevname) > $ldevconfpath
6825 local mdsfo_host=$mdsfailover_HOST;
6826 if [ -z "$mdsfo_host" ]; then
6830 for num in $(seq $MDSCOUNT); do
6831 fstype=$(facet_fstype mds$num)
6832 if [ "$fstype" == "zfs" ]; then
6833 fsldevformat="$fstype:"
6838 printf "%s\t%s\t%s-MDT%04d\t%s%s\n" \
6844 $(mdsdevname $num) >> $ldevconfpath
6847 local ostfo_host=$ostfailover_HOST;
6848 if [ -z "$ostfo_host" ]; then
6852 for num in $(seq $OSTCOUNT); do
6853 fstype=$(facet_fstype ost$num)
6854 if [ "$fstype" == "zfs" ]; then
6855 fsldevformat="$fstype:"
6860 printf "%s\t%s\t%s-OST%04d\t%s%s\n" \
6866 $(ostdevname $num) >> $ldevconfpath
6869 echo "----- $ldevconfpath -----"
6871 echo "--- END $ldevconfpath ---"
6876 # generate a nids file (mapping between hostname to nid)
6877 # looks like we only have the MGS nid available to us
6878 # so just echo that to a file
6880 echo -e "${mgs_HOST}\t${MGSNID}" > $nidspath
6882 echo "----- $nidspath -----"
6884 echo "--- END $nidspath ---"
6887 compare_ldev_output() {
6891 sort $expected_output -o $expected_output
6892 sort $ldev_output -o $ldev_output
6894 echo "-- START OF LDEV OUTPUT --"
6896 echo "--- END OF LDEV OUTPUT ---"
6898 echo "-- START OF EXPECTED OUTPUT --"
6899 cat $expected_output
6900 echo "--- END OF EXPECTED OUTPUT ---"
6902 diff $expected_output $ldev_output
6907 if [ -z "$LDEV" ]; then
6908 error "ldev is missing!"
6911 local LDEVCONFPATH=$TMP/ldev.conf
6912 local NIDSPATH=$TMP/nids
6914 echo "Host is $(hostname)"
6916 generate_ldev_conf $LDEVCONFPATH
6917 generate_nids $NIDSPATH
6919 # echo the mgs nid and compare it to environment variable MGSNID
6920 # also, ldev.conf and nids is a server side thing, use the OSS
6923 output=$($LDEV -c $LDEVCONFPATH -H $ost_HOST -n $NIDSPATH echo %m)
6925 echo "-- START OF LDEV OUTPUT --"
6927 echo "--- END OF LDEV OUTPUT ---"
6929 # ldev failed, error
6930 if [ $? -ne 0 ]; then
6931 rm $LDEVCONFPATH $NIDSPATH
6932 error "ldev failed to execute!"
6935 # need to process multiple lines because of combined MGS and MDS
6936 echo -e $output | awk '{ print $2 }' | while read -r line ; do
6937 if [ "$line" != "$MGSNID" ]; then
6938 rm $LDEVCONFPATH $NIDSPATH
6939 error "ldev failed mgs nid '$line', expected '$MGSNID'"
6943 rm $LDEVCONFPATH $NIDSPATH
6945 run_test 92 "ldev returns MGS NID correctly in command substitution"
6948 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs"
6951 #start mgs or mgs/mdt0
6952 if ! combined_mgs_mds ; then
6959 start_ost || error "OST0 start fail"
6961 #define OBD_FAIL_MGS_WRITE_TARGET_DELAY 0x90e
6962 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x8000090e"
6963 for num in $(seq 2 $MDSCOUNT); do
6967 mount_client $MOUNT || error "mount client fails"
6968 wait_osc_import_state mds ost FULL
6969 wait_osc_import_ready client ost
6970 check_mount || error "check_mount failed"
6972 cleanup || error "cleanup failed with $?"
6974 run_test 93 "register mulitple MDT at the same time"
6977 if [ -z "$LDEV" ]; then
6978 error "ldev is missing!"
6981 local LDEVCONFPATH=$TMP/ldev.conf
6982 local NIDSPATH=$TMP/nids
6984 generate_ldev_conf $LDEVCONFPATH
6985 generate_nids $NIDSPATH
6987 local LDEV_OUTPUT=$TMP/ldev-output.txt
6988 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME > $LDEV_OUTPUT
6990 # ldev failed, error
6991 if [ $? -ne 0 ]; then
6992 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
6993 error "ldev failed to execute!"
6997 local EXPECTED_OUTPUT=$TMP/ldev-expected.txt
6999 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
7001 for num in $(seq $MDSCOUNT); do
7002 printf "%s-MDT%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
7005 for num in $(seq $OSTCOUNT); do
7006 printf "%s-OST%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
7009 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7011 if [ $? -ne 0 ]; then
7012 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7013 error "ldev failed to produce the correct hostlist!"
7016 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7018 run_test 94 "ldev outputs correct labels for file system name query"
7021 if [ -z "$LDEV" ]; then
7022 error "ldev is missing!"
7025 local LDEVCONFPATH=$TMP/ldev.conf
7026 local NIDSPATH=$TMP/nids
7028 generate_ldev_conf $LDEVCONFPATH
7029 generate_nids $NIDSPATH
7033 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME &>/dev/null
7034 if [ $? -ne 0 ]; then
7035 rm $LDEVCONFPATH $NIDSPATH
7036 error "ldev label filtering w/ -F failed!"
7040 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -l &>/dev/null
7041 if [ $? -ne 0 ]; then
7042 rm $LDEVCONFPATH $NIDSPATH
7043 error "ldev label filtering w/ -l failed!"
7047 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f &>/dev/null
7048 if [ $? -ne 0 ]; then
7049 rm $LDEVCONFPATH $NIDSPATH
7050 error "ldev label filtering w/ -f failed!"
7054 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a &>/dev/null
7055 if [ $? -ne 0 ]; then
7056 rm $LDEVCONFPATH $NIDSPATH
7057 error "ldev label filtering w/ -a failed!"
7062 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -F $FSNAME &>/dev/null
7063 if [ $? -eq 0 ]; then
7064 rm $LDEVCONFPATH $NIDSPATH
7065 error "ldev label filtering w/ -a and -F incorrectly succeeded"
7069 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -f &>/dev/null
7070 if [ $? -eq 0 ]; then
7071 rm $LDEVCONFPATH $NIDSPATH
7072 error "ldev label filtering w/ -a and -f incorrectly succeeded"
7076 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -l &>/dev/null
7077 if [ $? -eq 0 ]; then
7078 rm $LDEVCONFPATH $NIDSPATH
7079 error "ldev label filtering w/ -a and -l incorrectly succeeded"
7083 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f -l &>/dev/null
7084 if [ $? -eq 0 ]; then
7085 rm $LDEVCONFPATH $NIDSPATH
7086 error "ldev label filtering w/ -f and -l incorrectly succeeded"
7090 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -l &>/dev/null
7091 if [ $? -eq 0 ]; then
7092 rm $LDEVCONFPATH $NIDSPATH
7093 error "ldev label filtering w/ -F and -l incorrectly succeeded"
7096 # file sys & foreign
7097 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -f &>/dev/null
7098 if [ $? -eq 0 ]; then
7099 rm $LDEVCONFPATH $NIDSPATH
7100 error "ldev label filtering w/ -F and -f incorrectly succeeded"
7103 rm $LDEVCONFPATH $NIDSPATH
7105 run_test 95 "ldev should only allow one label filter"
7108 if [ -z "$LDEV" ]; then
7109 error "ldev is missing!"
7112 local LDEVCONFPATH=$TMP/ldev.conf
7113 local NIDSPATH=$TMP/nids
7115 generate_ldev_conf $LDEVCONFPATH
7116 generate_nids $NIDSPATH
7118 local LDEV_OUTPUT=$TMP/ldev-output.txt
7119 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -H $mgs_HOST \
7121 awk '{print $2}' > $LDEV_OUTPUT
7123 # ldev failed, error
7124 if [ $? -ne 0 ]; then
7125 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7126 error "ldev failed to execute!"
7130 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
7132 echo "$mgs_HOST-$(facet_fstype mgs)" > $EXPECTED_OUTPUT
7134 if [ "$mgs_HOST" == "$mds_HOST" ]; then
7135 for num in $(seq $MDSCOUNT); do
7136 echo "$mds_HOST-$(facet_fstype mds$num)" \
7141 if [ "$mgs_HOST" == "$ost_HOST" ]; then
7142 for num in $(seq $OSTCOUNT); do
7143 echo "$ost_HOST-$(facet_fstype ost$num)" \
7148 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7150 if [ $? -ne 0 ]; then
7151 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7152 error "ldev failed to produce the correct output!"
7155 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7157 run_test 96 "ldev returns hostname and backend fs correctly in command sub"
7160 if [ -z "$LDEV" ]; then
7161 error "ldev is missing!"
7164 local LDEVCONFPATH=$TMP/ldev.conf
7165 local NIDSPATH=$TMP/nids
7167 generate_ldev_conf $LDEVCONFPATH
7168 generate_nids $NIDSPATH
7170 local LDEV_OUTPUT=$TMP/ldev-output.txt
7171 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
7173 echo -e "\nMDT role"
7174 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mdt > $LDEV_OUTPUT
7176 if [ $? -ne 0 ]; then
7177 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7178 error "ldev failed to execute for mdt role!"
7181 for num in $(seq $MDSCOUNT); do
7182 printf "%s-MDT%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
7185 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7187 if [ $? -ne 0 ]; then
7188 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7189 error "ldev failed to produce the correct output for mdt role!"
7192 echo -e "\nOST role"
7193 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R ost > $LDEV_OUTPUT
7195 if [ $? -ne 0 ]; then
7196 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
7197 error "ldev failed to execute for ost role!"
7201 for num in $(seq $OSTCOUNT); do
7202 printf "%s-OST%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
7205 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7207 if [ $? -ne 0 ]; then
7208 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7209 error "ldev failed to produce the correct output for ost role!"
7212 echo -e "\nMGS role"
7213 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mgs > $LDEV_OUTPUT
7215 if [ $? -ne 0 ]; then
7216 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
7217 error "ldev failed to execute for mgs role!"
7220 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
7222 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7224 if [ $? -ne 0 ]; then
7225 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7226 error "ldev failed to produce the correct output for mgs role!"
7229 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7231 run_test 97 "ldev returns correct ouput when querying based on role"
7236 local temp=$MDS_MOUNT_OPTS
7239 check_mount || error "mount failed"
7240 mountopt="user_xattr"
7241 for ((x = 1; x <= 400; x++)); do
7242 mountopt="$mountopt,user_xattr"
7244 remount_client $mountopt $MOUNT 2>&1 | grep "too long" ||
7245 error "Buffer overflow check failed"
7246 cleanup || error "cleanup failed"
7248 run_test 98 "Buffer-overflow check while parsing mount_opts"
7252 [[ $(facet_fstype ost1) != ldiskfs ]] &&
7253 skip "ldiskfs only test"
7254 [[ $(lustre_version_code ost1) -ge $(version_code 2.8.57) ]] ||
7255 skip "Need OST version at least 2.8.57"
7257 local ost_opts="$(mkfs_opts ost1 $(ostdevname 1)) \
7258 --reformat $(ostdevname 1) $(ostvdevname 1)"
7259 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" &&
7260 skip "meta_bg already set"
7263 if [[ ${!opts} != *mkfsoptions* ]]; then
7264 eval opts=\"${!opts} \
7265 --mkfsoptions='\\\"-O ^resize_inode,meta_bg\\\"'\"
7267 local val=${!opts//--mkfsoptions=\\\"/ \
7268 --mkfsoptions=\\\"-O ^resize_inode,meta_bg }
7272 echo "params: $opts"
7275 add ost1 $opts || error "add ost1 failed with new params"
7277 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" ||
7278 error "meta_bg is not set"
7280 run_test 99 "Adding meta_bg option"
7285 start_mgsmds || error "MDS start failed"
7286 start_ost || error "unable to start OST"
7287 mount_client $MOUNT || error "client start failed"
7288 check_mount || error "check_mount failed"
7297 do_facet mgs 'lshowmount -v' | awk 'BEGIN {NR == 0; rc=1} /MGS:/ {rc=0}
7298 END {exit rc}' || error "lshowmount have no output MGS"
7300 do_facet mds1 'lshowmount -v' | awk 'BEGIN {NR == 2; rc=1} /-MDT0000:/
7301 {rc=0} END {exit rc}' || error "lshowmount have no output MDT0"
7303 do_facet ost1 'lshowmount -v' | awk 'BEGIN {NR == 4; rc=1} /-OST0000:/
7304 {rc=0} END {exit rc}' || error "lshowmount have no output OST0"
7306 cleanup || error "cleanup failed with $?"
7308 run_test 100 "check lshowmount lists MGS, MDT, OST and 0@lo"
7311 local createmany_pid
7312 local dev=$FSNAME-OST0000-osc-MDT0000
7316 createmany -o $DIR1/$tdir/$tfile-%d 50000 &
7318 # MDT->OST reconnection causes MDT<->OST last_id synchornisation
7319 # via osp_precreate_cleanup_orphans.
7320 for ((i = 0; i < 100; i++)); do
7321 for ((k = 0; k < 10; k++)); do
7322 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;" \
7323 "$LCTL --device $dev activate"
7326 ls -asl $MOUNT | grep '???' &&
7327 { kill -9 $createmany_pid &>/dev/null;
7328 error "File has no object on OST"; }
7330 kill -s 0 $createmany_pid || break
7332 wait $createmany_pid
7334 unlinkmany $DIR1/$tdir/$tfile-%d 50000
7337 run_test 101 "Race MDT->OST reconnection with create"
7340 [[ $(lustre_version_code $SINGLEMDS) -gt $(version_code 2.9.53) ]] ||
7341 skip "Need server version greater than 2.9.53"
7342 [[ “$(mdsdevname 1)” != “$(mgsdevname)” ]] &&
7343 [[ “$(facet_host mds1)” = “$(facet_host mgs)” ]] &&
7344 skip "MGS must be on different node or combined"
7346 cleanup || error "cleanup failed with $?"
7348 local mds1dev=$(mdsdevname 1)
7349 local mds1mnt=$(facet_mntpt mds1)
7350 local mds1fstype=$(facet_fstype mds1)
7351 local mds1opts=$MDS_MOUNT_OPTS
7353 if [ $mds1fstype == ldiskfs ] &&
7354 ! do_facet mds1 test -b $mds1dev; then
7355 mds1opts=$(csa_add "$mds1opts" -o loop)
7357 if [[ $mds1fstype == zfs ]]; then
7358 import_zpool mds1 || return ${PIPESTATUS[0]}
7361 # unload all and only load libcfs to allow fail_loc setting
7362 do_facet mds1 $LUSTRE_RMMOD || error "unable to unload modules"
7363 do_facet mds1 modprobe libcfs || error "libcfs not loaded"
7364 do_facet mds1 lsmod \| grep libcfs || error "libcfs not loaded"
7366 #define OBD_FAIL_OBDCLASS_MODULE_LOAD 0x60a
7367 do_facet mds1 "$LCTL set_param fail_loc=0x8000060a"
7369 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts &&
7370 error "mdt start must fail"
7371 do_facet mds1 lsmod \| grep obdclass && error "obdclass must not load"
7373 do_facet mds1 "$LCTL set_param fail_loc=0x0"
7375 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts ||
7376 error "mdt start must not fail"
7378 cleanup || error "cleanup failed with $?"
7380 run_test 102 "obdclass module cleanup upon error"
7385 echo "rename $FSNAME to $newname"
7387 if ! combined_mgs_mds ; then
7388 local dev=$(mgsdevname)
7391 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
7392 error "(7) Fail to rename MGS"
7393 if [ "$(facet_fstype mgs)" = "zfs" ]; then
7394 reimport_zpool mgs $newname-mgs
7398 for num in $(seq $MDSCOUNT); do
7399 local dev=$(mdsdevname $num)
7401 do_facet mds${num} \
7402 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
7403 error "(8) Fail to rename MDT $num"
7404 if [ "$(facet_fstype mds${num})" = "zfs" ]; then
7405 reimport_zpool mds${num} $newname-mdt${num}
7409 for num in $(seq $OSTCOUNT); do
7410 local dev=$(ostdevname $num)
7412 do_facet ost${num} \
7413 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
7414 error "(9) Fail to rename OST $num"
7415 if [ "$(facet_fstype ost${num})" = "zfs" ]; then
7416 reimport_zpool ost${num} $newname-ost${num}
7421 test_103_set_pool() {
7425 do_facet mgs $LCTL pool_add $FSNAME.$pname ${FSNAME}-$ost_x ||
7426 error "Fail to add $ost_x to $FSNAME.$pname"
7427 wait_update $HOSTNAME \
7428 "lctl get_param -n lov.$FSNAME-clilov-*.pools.$pname |
7429 grep $ost_x" "$FSNAME-${ost_x}_UUID" ||
7430 error "$ost_x is NOT in pool $FSNAME.$pname"
7433 test_103_check_pool() {
7434 local save_fsname=$1
7437 stat $DIR/$tdir/test-framework.sh ||
7438 error "($errno) Fail to stat"
7439 do_facet mgs $LCTL pool_list $FSNAME.pool1 ||
7440 error "($errno) Fail to list $FSNAME.pool1"
7441 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname ||
7442 error "($errno) Fail to list $FSNAME.$save_fsname"
7443 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname |
7444 grep ${FSNAME}-OST0000 ||
7445 error "($errno) List $FSNAME.$save_fsname is invalid"
7447 local pname=$($LFS getstripe --pool $DIR/$tdir/d0)
7448 [ "$pname" = "$save_fsname" ] ||
7449 error "($errno) Unexpected pool name $pname"
7453 check_mount_and_prep
7455 mkdir $DIR/$tdir || error "(1) Fail to mkdir $DIR/$tdir"
7456 cp $LUSTRE/tests/test-framework.sh $DIR/$tdir ||
7457 error "(2) Fail to copy test-framework.sh"
7459 do_facet mgs $LCTL pool_new $FSNAME.pool1 ||
7460 error "(3) Fail to create $FSNAME.pool1"
7461 # name the pool name as the fsname
7462 do_facet mgs $LCTL pool_new $FSNAME.$FSNAME ||
7463 error "(4) Fail to create $FSNAME.$FSNAME"
7465 test_103_set_pool $FSNAME OST0000
7467 $SETSTRIPE -p $FSNAME $DIR/$tdir/d0 ||
7468 error "(6) Fail to setstripe on $DIR/$tdir/d0"
7473 test_renamefs mylustre
7475 local save_fsname=$FSNAME
7479 test_103_check_pool $save_fsname 7
7481 if [ $OSTCOUNT -ge 2 ]; then
7482 test_103_set_pool $save_fsname OST0001
7485 $SETSTRIPE -p $save_fsname $DIR/$tdir/f0 ||
7486 error "(16) Fail to setstripe on $DIR/$tdir/f0"
7495 test_103_check_pool $save_fsname 17
7499 test_renamefs $save_fsname
7505 run_test 103 "rename filesystem name"
7507 test_104() { # LU-6952
7508 local mds_mountopts=$MDS_MOUNT_OPTS
7509 local ost_mountopts=$OST_MOUNT_OPTS
7510 local mds_mountfsopts=$MDS_MOUNT_FS_OPTS
7511 local lctl_ver=$(do_facet $SINGLEMDS $LCTL --version |
7514 [[ $(version_code $lctl_ver) -lt $(version_code 2.9.55) ]] &&
7515 skip "this test needs utils above 2.9.55"
7517 # specify "acl" in mount options used by mkfs.lustre
7518 if [ -z "$MDS_MOUNT_FS_OPTS" ]; then
7519 MDS_MOUNT_FS_OPTS="acl,user_xattr"
7522 MDS_MOUNT_FS_OPTS="${MDS_MOUNT_FS_OPTS},acl,user_xattr"
7525 echo "mountfsopt: $MDS_MOUNT_FS_OPTS"
7527 #reformat/remount the MDT to apply the MDT_MOUNT_FS_OPT options
7529 if ! combined_mgs_mds ; then
7533 if [ -z "$MDS_MOUNT_OPTS" ]; then
7534 MDS_MOUNT_OPTS="-o noacl"
7536 MDS_MOUNT_OPTS="${MDS_MOUNT_OPTS},noacl"
7539 for num in $(seq $MDSCOUNT); do
7540 start mds$num $(mdsdevname $num) $MDS_MOUNT_OPTS ||
7541 error "Failed to start MDS"
7544 for num in $(seq $OSTCOUNT); do
7545 start ost$num $(ostdevname $num) $OST_MOUNT_OPTS ||
7546 error "Failed to start OST"
7550 setfacl -m "d:$RUNAS_ID:rwx" $MOUNT &&
7551 error "ACL is applied when FS is mounted with noacl."
7553 MDS_MOUNT_OPTS=$mds_mountopts
7554 OST_MOUNT_OPTS=$ost_mountopts
7555 MDS_MOUNT_FS_OPTS=$mds_mountfsopts
7557 run_test 104 "Make sure user defined options are reflected in mount"
7559 error_and_umount() {
7570 mount --bind $DIR $TMP/$tdir || error "mount bind mnt pt failed"
7571 rm -f $TMP/$tdir/$tfile
7572 rm -f $TMP/$tdir/${tfile}1
7574 # Files should not be created in ro bind mount point
7575 # remounting from rw to ro
7576 mount -o remount,ro $TMP/$tdir ||
7577 error_and_umount "readonly remount of bind mnt pt failed"
7578 touch $TMP/$tdir/$tfile &&
7579 error_and_umount "touch succeeds on ro bind mnt pt"
7580 [ -e $TMP/$tdir/$tfile ] &&
7581 error_and_umount "file created on ro bind mnt pt"
7583 # Files should be created in rw bind mount point
7584 # remounting from ro to rw
7585 mount -o remount,rw $TMP/$tdir ||
7586 error_and_umount "read-write remount of bind mnt pt failed"
7587 touch $TMP/$tdir/${tfile}1 ||
7588 error_and_umount "touch fails on rw bind mnt pt"
7589 [ -e $TMP/$tdir/${tfile}1 ] ||
7590 error_and_umount "file not created on rw bind mnt pt"
7591 umount $TMP/$tdir || error "umount of bind mnt pt failed"
7593 cleanup || error "cleanup failed with $?"
7595 run_test 105 "check file creation for ro and rw bind mnt pt"
7602 mkdir -p $DIR/$tdir || error "create $tdir failed"
7603 lfs setstripe -c 1 -i 0 $DIR/$tdir
7604 #define OBD_FAIL_CAT_RECORDS 0x1312
7605 do_facet mds1 $LCTL set_param fail_loc=0x1312 fail_val=$repeat
7607 for ((i = 1; i <= $repeat; i++)); do
7609 #one full plain llog
7610 createmany -o $DIR/$tdir/f- 64768
7612 createmany -u $DIR/$tdir/f- 64768
7614 wait_delete_completed $((TIMEOUT * 7))
7615 #ASSERTION osp_sync_thread() ( thread->t_flags != SVC_RUNNING ) failed
7616 #shows that osp code is buggy
7617 do_facet mds1 $LCTL set_param fail_loc=0 fail_val=0
7621 run_test 106 "check osp llog processing when catalog is wrapped"
7624 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.10.50) ]] ||
7625 skip "Need MDS version > 2.10.50"
7628 start_mgsmds || error "start_mgsmds failed"
7629 start_ost || error "unable to start OST"
7631 # add unknown configuration parameter.
7632 if [[ $PERM_CMD = *"set_param -P"* ]]; then
7633 cmd="$PERM_CMD ost.$FSNAME-OST0000*.unknown_param"
7635 cmd="$PERM_CMD $FSNAME-OST0000*.ost.unknown_param"
7637 do_facet mgs "$cmd=50"
7638 cleanup_nocli || error "cleanup_nocli failed with $?"
7641 # unknown param should be ignored while mounting.
7642 start_ost || error "unable to start OST after unknown param set"
7644 cleanup || error "cleanup failed with $?"
7646 run_test 107 "Unknown config param should not fail target mounting"
7651 $rcmd rm -rf $tmp > /dev/null 2>&1
7652 $rcmd mkdir -p $tmp/{mnt,images} || error "failed to mkdir remotely"
7654 for facet in $facets; do
7655 [ $(facet_fstype $SINGLEMDS) = "zfs" ] &&
7656 $rcmd $ZPOOL -f export lustre-$facet > /dev/null 2>&1
7657 $rcmd mkdir $tmp/mnt/$facet ||
7658 error "failed to mkdir $tmp/mnt/$facet"
7659 $rcmd dd if=/dev/zero of=$tmp/images/$facet \
7660 seek=199 bs=1M count=1 ||
7661 error "failed to create $tmp/images/$facet"
7670 local facet=${role}$((idx + 1))
7672 [ $# -eq 5 ] && pool=$5
7674 do_facet $SINGLEMDS $MKFS --fsname=lustre --$mgs \
7675 --$role --index=$idx --replace --backfstype=$bkfs \
7676 --device-size=200000 --reformat $pool $tmp/images/$facet ||
7677 error "failed to mkfs for $facet"
7681 echo "mounting client..."
7682 mount -t lustre ${nid}:/lustre $MOUNT ||
7683 error "failed to mount lustre"
7686 ls -l $MOUNT/local_dir || error "failed to list"
7688 echo "check truncate && write"
7689 echo "dummmmmmmmmmmmm" > $MOUNT/remote_dir/fsx.c ||
7690 error "failed to tuncate & write"
7693 touch $MOUNT/foooo ||
7694 error "failed to create"
7696 echo "check read && write && append"
7697 sha1sum $MOUNT/conf-sanity.sh |
7698 awk '{ print $1 }' > $MOUNT/checksum.new ||
7699 error "failed to read(1)"
7700 sha1sum $MOUNT/remote_dir/unlinkmany.c |
7701 awk '{ print $1 }' >> $MOUNT/checksum.new ||
7702 error "failed to read(2)"
7703 sha1sum $MOUNT/striped_dir/lockahead_test.o |
7704 awk '{ print $1 }' >> $MOUNT/checksum.new ||
7705 error "failed to read(3)"
7708 diff $MOUNT/checksum.new $MOUNT/checksum.src ||
7709 error "failed to verify data"
7719 umount -f $MOUNT || error "failed to umount client"
7720 for facet in $facets; do
7721 $rcmd umount -f $tmp/mnt/$facet ||
7722 error "failed to umount $facet"
7723 if [ $(facet_fstype $SINGLEMDS) = "zfs" ]; then
7724 $rcmd $ZPOOL export -f lustre-$facet ||
7725 error "failed to export lustre-$facet"
7729 $rcmd rm -rf $tmp || error "failed to rm the dir $tmp"
7733 [ "$CLIENTONLY" ] && skip "Client-only testing"
7734 [ $(facet_fstype $SINGLEMDS) != "zfs" ] && skip "zfs only test"
7735 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.58) ] &&
7736 skip "Need server version at least 2.10.58"
7741 local tmp=$TMP/$tdir
7742 local rcmd="do_facet $SINGLEMDS"
7743 local facets="mdt1 mdt2 ost1 ost2"
7744 local nid=$($rcmd $LCTL list_nids | head -1)
7747 trap t_108_cleanup EXIT ERR
7750 t_108_mkfs mdt 0 zfs mgs lustre-mdt1/mdt1
7751 t_108_mkfs mdt 1 zfs mgsnode=$nid lustre-mdt2/mdt2
7752 t_108_mkfs ost 0 zfs mgsnode=$nid lustre-ost1/ost1
7753 t_108_mkfs ost 1 zfs mgsnode=$nid lustre-ost2/ost2
7755 for facet in $facets; do
7756 $rcmd zfs set mountpoint=$tmp/mnt/$facet canmount=on \
7757 lustre-$facet/$facet ||
7758 error "failed to zfs set for $facet (1)"
7759 $rcmd zfs mount lustre-$facet/$facet ||
7760 error "failed to local mount $facet"
7761 $rcmd tar jxf $LUSTRE/tests/ldiskfs_${facet}_2_11.tar.bz2 \
7762 --xattrs --xattrs-include="trusted.*" \
7763 -C $tmp/mnt/$facet/ > /dev/null 2>&1 ||
7764 error "failed to untar image for $facet"
7765 $rcmd "cd $tmp/mnt/$facet && rm -rf oi.* OI_* lfsck_* LFSCK" ||
7766 error "failed to cleanup for $facet"
7767 $rcmd zfs umount lustre-$facet/$facet ||
7768 error "failed to local umount $facet"
7769 $rcmd zfs set canmount=off lustre-$facet/$facet ||
7770 error "failed to zfs set $facet (2)"
7773 echo "changing server nid..."
7774 $rcmd mount -t lustre -o nosvc lustre-mdt1/mdt1 $tmp/mnt/mdt1
7775 $rcmd lctl replace_nids lustre-MDT0000 $nid
7776 $rcmd lctl replace_nids lustre-MDT0001 $nid
7777 $rcmd lctl replace_nids lustre-OST0000 $nid
7778 $rcmd lctl replace_nids lustre-OST0001 $nid
7779 $rcmd umount $tmp/mnt/mdt1
7781 for facet in $facets; do
7782 echo "mounting $facet from backup..."
7783 $rcmd mount -t lustre -o abort_recov lustre-$facet/$facet \
7784 $tmp/mnt/$facet || error "failed to mount $facet"
7787 # ZFS backend can detect migration and trigger OI scrub automatically
7788 # sleep 3 seconds for scrub done
7794 run_test 108a "migrate from ldiskfs to ZFS"
7797 [ "$CLIENTONLY" ] && skip "Client-only testing"
7798 [ $(facet_fstype $SINGLEMDS) != "ldiskfs" ] && skip "ldiskfs only test"
7799 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.58) ] &&
7800 skip "Need server version at least 2.10.58"
7805 local tmp=$TMP/$tdir
7806 local rcmd="do_facet $SINGLEMDS"
7807 local facets="mdt1 mdt2 ost1 ost2"
7808 local scrub_list="MDT0000 MDT0001 OST0000 OST0001"
7809 local nid=$($rcmd $LCTL list_nids | head -1)
7812 trap t_108_cleanup EXIT ERR
7815 t_108_mkfs mdt 0 ldiskfs mgs
7816 t_108_mkfs mdt 1 ldiskfs mgsnode=$nid
7817 t_108_mkfs ost 0 ldiskfs mgsnode=$nid
7818 t_108_mkfs ost 1 ldiskfs mgsnode=$nid
7820 for facet in $facets; do
7821 $rcmd mount -t ldiskfs -o loop $tmp/images/$facet \
7823 error "failed to local mount $facet"
7825 $rcmd tar jxf $LUSTRE/tests/zfs_${facet}_2_11.tar.bz2 \
7826 --xattrs --xattrs-include="*.*" \
7827 -C $tmp/mnt/$facet/ > /dev/null 2>&1 ||
7828 error "failed to untar image for $facet"
7829 $rcmd "cd $tmp/mnt/$facet && rm -rf oi.* OI_* lfsck_* LFSCK" ||
7830 error "failed to cleanup for $facet"
7831 $rcmd umount $tmp/mnt/$facet ||
7832 error "failed to local umount $facet"
7835 echo "changing server nid..."
7836 $rcmd mount -t lustre -o nosvc,loop $tmp/images/mdt1 $tmp/mnt/mdt1
7837 $rcmd lctl replace_nids lustre-MDT0000 $nid
7838 $rcmd lctl replace_nids lustre-MDT0001 $nid
7839 $rcmd lctl replace_nids lustre-OST0000 $nid
7840 $rcmd lctl replace_nids lustre-OST0001 $nid
7841 $rcmd umount $tmp/mnt/mdt1
7843 for facet in $facets; do
7844 echo "mounting $facet from backup..."
7845 $rcmd mount -t lustre -o loop,abort_recov $tmp/images/$facet \
7846 $tmp/mnt/$facet || error "failed to mount $facet"
7849 for facet in $scrub_list; do
7850 $rcmd $LCTL lfsck_start -M lustre-$facet -t scrub ||
7851 error "failed to start OI scrub on $facet"
7854 # sleep 3 seconds for scrub done
7860 run_test 108b "migrate from ZFS to ldiskfs"
7864 # set number of permanent parameters
7866 test_109_set_params() {
7869 set_persistent_param_and_check mds \
7870 "mdd.$fsname-MDT0000.atime_diff" \
7871 "$fsname-MDT0000.mdd.atime_diff" \
7873 set_persistent_param_and_check mds \
7874 "mdd.$fsname-MDT0000.atime_diff" \
7875 "$fsname-MDT0000.mdd.atime_diff" \
7877 set_persistent_param_and_check client \
7878 "llite.$fsname*.max_read_ahead_mb" \
7879 "$fsname.llite.max_read_ahead_mb" \
7881 set_persistent_param_and_check client \
7882 "llite.$fsname*.max_read_ahead_mb" \
7883 "$fsname.llite.max_read_ahead_mb" \
7885 create_pool $fsname.pool1 || error "create pool failed"
7886 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
7887 error "pool_add failed"
7888 do_facet mgs $LCTL pool_remove $fsname.pool1 OST0000 ||
7889 error "pool_remove failed"
7890 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
7891 error "pool_add failed"
7895 # check permanent parameters
7897 test_109_test_params() {
7900 local atime_diff=$(do_facet mds $LCTL \
7901 get_param -n mdd.$fsname-MDT0000.atime_diff)
7902 [ $atime_diff == 63 ] || error "wrong mdd parameter after clear_conf"
7903 local max_read_ahead_mb=$(do_facet client $LCTL \
7904 get_param -n llite.$fsname*.max_read_ahead_mb)
7905 [ $max_read_ahead_mb == 64 ] ||
7906 error "wrong llite parameter after clear_conf"
7907 local ost_in_pool=$(do_facet mds $LCTL pool_list $fsname.pool1 |
7908 grep -v "^Pool:" | sed 's/_UUID//')
7909 [ $ost_in_pool = "$fsname-OST0000" ] ||
7910 error "wrong pool after clear_conf"
7914 # run lctl clear_conf, store CONFIGS before and after that
7916 test_109_clear_conf()
7918 local clear_conf_arg=$1
7921 if ! combined_mgs_mds ; then
7922 mgsdev=$(mgsdevname)
7923 stop_mgs || error "stop_mgs failed"
7924 start_mgs "-o nosvc" || error "start_mgs nosvc failed"
7926 mgsdev=$(mdsdevname 1)
7927 start_mdt 1 "-o nosvc" || error "start_mdt 1 nosvc failed"
7930 do_facet mgs "rm -rf $TMP/${tdir}/conf1; mkdir -p $TMP/${tdir}/conf1;" \
7931 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf1\\\" \
7935 # the command being tested
7937 do_facet mgs $LCTL clear_conf $clear_conf_arg ||
7938 error "clear_conf failed"
7939 if ! combined_mgs_mds ; then
7940 stop_mgs || error "stop_mgs failed"
7942 stop_mdt 1 || error "stop_mdt 1 failed"
7945 do_facet mgs "rm -rf $TMP/${tdir}/conf2; mkdir -p $TMP/${tdir}/conf2;" \
7946 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf2\\\" \
7950 test_109_file_shortened() {
7952 local sizes=($(do_facet mgs "stat -c %s " \
7953 "$TMP/${tdir}/conf1/CONFIGS/$file" \
7954 "$TMP/${tdir}/conf2/CONFIGS/$file"))
7955 [ ${sizes[1]} -lt ${sizes[0]} ] && return 0
7961 [ "$(facet_fstype mgs)" == "zfs" ] &&
7962 skip "LU-8727: no implementation for ZFS"
7967 client_up || error "client_up failed"
7970 # set number of permanent parameters
7972 test_109_set_params $FSNAME
7974 umount_client $MOUNT || error "umount_client failed"
7975 stop_ost || error "stop_ost failed"
7976 stop_mds || error "stop_mds failed"
7978 test_109_clear_conf $FSNAME
7980 # make sure that all configs are cleared
7982 test_109_file_shortened $FSNAME-MDT0000 ||
7983 error "failed to clear MDT0000 config"
7984 test_109_file_shortened $FSNAME-client ||
7985 error "failed to clear client config"
7990 # check that configurations are intact
7992 test_109_test_params $FSNAME
7997 destroy_test_pools || error "destroy test pools failed"
8001 run_test 109a "test lctl clear_conf fsname"
8005 [ "$(facet_fstype mgs)" == "zfs" ] &&
8006 skip "LU-8727: no implementation for ZFS"
8011 client_up || error "client_up failed"
8014 # set number of permanent parameters
8016 test_109_set_params $FSNAME
8018 umount_client $MOUNT || error "umount_client failed"
8019 stop_ost || error "stop_ost failed"
8020 stop_mds || error "stop_mds failed"
8022 test_109_clear_conf $FSNAME-MDT0000
8024 # make sure that only one config is cleared
8026 test_109_file_shortened $FSNAME-MDT0000 ||
8027 error "failed to clear MDT0000 config"
8028 test_109_file_shortened $FSNAME-client &&
8029 error "failed to clear client config"
8033 # check that configurations are intact
8035 test_109_test_params $FSNAME
8040 destroy_test_pools || error "destroy test pools failed"
8044 run_test 109b "test lctl clear_conf one config"
8050 rm -f $TMP/$tdir/lustre-mdt
8055 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
8056 skip "Only applicable to ldiskfs-based MDTs"
8059 IMAGESIZE=$((3072 << 30)) # 3072 GiB
8062 # We need MDT size 3072GB, because it is smallest
8063 # partition that can store 2B inodes
8064 do_facet $SINGLEMDS "mkdir -p $TMP/$tdir"
8065 local mdsimgname=$TMP/$tdir/lustre-mdt
8066 do_facet $SINGLEMDS "rm -f $mdsimgname"
8067 do_facet $SINGLEMDS "touch $mdsimgname"
8068 trap cleanup_115 RETURN EXIT
8069 do_facet $SINGLEMDS "$TRUNCATE $mdsimgname $IMAGESIZE" ||
8070 skip "Backend FS doesn't support sparse files"
8071 local mdsdev=$(do_facet $SINGLEMDS "losetup -f")
8072 do_facet $SINGLEMDS "losetup $mdsdev $mdsimgname"
8074 local mds_opts="$(mkfs_opts mds1 ${mdsdev}) --device-size=$IMAGESIZE \
8075 --mkfsoptions='-O lazy_itable_init,ea_inode,^resize_inode,meta_bg \
8077 add mds1 $mds_opts --mgs --reformat $mdsdev ||
8078 skip_env "format large MDT failed"
8079 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
8080 --reformat $(ostdevname 1) $(ostvdevname 1)
8082 start $SINGLEMDS ${mdsdev} $MDS_MOUNT_OPTS || error "start MDS failed"
8083 start_ost || error "start OSS failed"
8084 mount_client $MOUNT || error "mount client failed"
8086 mkdir -p $DIR/$tdir || error "mkdir $DIR/$tdir fail"
8087 for goal in $(do_facet $SINGLEMDS "ls /sys/fs/ldiskfs/*/inode_goal"); do
8088 do_facet $SINGLEMDS "echo 2147483947 >> $goal; grep . $goal"
8091 touch $DIR/$tdir/$tfile
8093 # Add > 5k bytes to xattr
8094 for i in {1..30}; do
8095 ln $DIR/$tdir/$tfile $DIR/$tdir/$(printf "link%0250d" $i) ||
8096 error "Can't make link"
8101 local inode_num=$(do_facet $SINGLEMDS \
8102 "$DEBUGFS -c -R 'stat ROOT/$tdir/$tfile' $mdsimgname" |
8103 awk '/link =/ { print $4 }' |
8104 sed -e 's/>//' -e 's/<//' -e 's/\"//')
8105 echo "inode num: $inode_num"
8106 [ $inode_num -ge 2147483947 ] || error "inode $inode_num too small"
8107 do_facet $SINGLEMDS "losetup -d $mdsdev"
8110 run_test 115 "Access large xattr with inodes number over 2TB"
8113 [ $(facet_fstype $SINGLEMDS) != "ldiskfs" ] && skip "ldiskfs only test"
8114 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.59) ] &&
8115 skip "Need server version at least 2.10.59"
8116 do_facet $SINGLEMDS which mkfs.xfs ||
8117 skip_env "No mkfs.xfs installed"
8122 local tmpmnt=$TMP/$tdir
8123 local mdtimg=$tfile-mdt0
8125 do_facet $SINGLEMDS mkdir -p $tmpmnt
8126 stack_trap "do_facet $SINGLEMDS rmdir $tmpmnt" EXIT
8128 do_facet $SINGLEMDS touch $TMP/$mdtimg
8129 stack_trap "do_facet $SINGLEMDS rm -f $TMP/$mdtimg" EXIT
8130 do_facet $SINGLEMDS mkfs -t xfs -d file,size=1t,name=$TMP/$mdtimg ||
8131 error "mkfs temporary xfs image"
8133 do_facet $SINGLEMDS mount $TMP/$mdtimg $tmpmnt ||
8134 error "mount temporary xfs image"
8135 stack_trap "do_facet $SINGLEMDS umount $tmpmnt" EXIT
8136 local old_mdssize=$MDSSIZE
8137 local old_mdsisize=$MDSISIZE
8139 MDSSIZE=$((17 * 1024 * 1024 * 1024)) # 17T MDT
8140 MDSISIZE=$((16 << 20))
8141 local opts17t="$(mkfs_opts $SINGLEMDS)"
8143 MDSSIZE=$old_mdssize
8144 MDSISIZE=$old_mdsisize
8145 do_facet $SINGLEMDS $MKFS $opts17t $tmpmnt/$mdtimg ||
8146 error "failed to mkfs for $tmpmnt/$mdtimg"
8148 do_facet $SINGLEMDS $TUNE2FS -l $tmpmnt/$mdtimg |
8149 grep -qw 'features.*extent' || error "extent should be enabled"
8152 run_test 116 "big size MDT support"
8156 do_facet ost1 "$LCTL set_param ost.OSS.ost_io.nrs_policies=fifo"
8157 do_facet ost1 "$LCTL get_param -n ost.OSS.ost_io.nrs_tbf_rule" &&
8158 error "get_param should fail"
8159 cleanup || error "cleanup failed with rc $?"
8161 run_test 117 "lctl get_param return errors properly"
8163 test_120() { # LU-11130
8164 [ "$MDSCOUNT" -lt 2 ] && skip "mdt count < 2"
8165 [ $(facet_fstype $SINGLEMDS) != "ldiskfs" ] &&
8166 skip "ldiskfs only test"
8167 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.56) ] &&
8168 skip "Need DNE2 capable MD target with LU-11130 fix"
8172 local mds1host=$(facet_active_host mds1)
8173 local mds1dev=$(mdsdevname 1)
8175 $LFS mkdir -i 1 $DIR/$tdir
8176 $LFS mkdir -i 0 $DIR/$tdir/mds1dir
8178 ln -s foo $DIR/$tdir/bar
8179 mv $DIR/$tdir/bar $DIR/$tdir/mds1dir/bar2 ||
8180 error "cross-target rename failed"
8184 run_e2fsck $mds1host $mds1dev "-n"
8186 run_test 120 "cross-target rename should not create bad symlinks"
8189 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
8190 [[ $(lustre_version_code ost1) -ge $(version_code 2.11.53) ]] ||
8191 skip "Need OST version at least 2.11.53"
8194 LOAD_MODULES_REMOTE=true load_modules
8195 #define OBD_FAIL_OFD_SET_OID 0x1e0
8196 do_facet ost1 $LCTL set_param fail_loc=0x00001e0
8199 $LFS mkdir -i1 -c1 $DIR/$tdir
8200 $LFS setstripe -i0 -c1 $DIR/$tdir
8201 do_facet ost1 $LCTL set_param fail_loc=0
8202 createmany -o $DIR/$tdir/file_ 1000 ||
8203 error "Fail to create a new sequence"
8207 run_test 122 "Check OST sequence update"
8210 remote_mgs_nodsh && skip "remote MGS with nodsh"
8211 [ -d $MOUNT/.lustre ] || setup
8213 # test old logid format until removal from llog_ioctl.c::str2logid()
8214 if [ $MGS_VERSION -lt $(version_code 3.1.53) ]; then
8215 do_facet mgs $LCTL dl | grep MGS
8216 do_facet mgs "$LCTL --device %MGS llog_print \
8217 \\\\\\\$$FSNAME-client 1 10" ||
8218 error "old llog_print failed"
8221 # test new logid format
8222 if [ $MGS_VERSION -ge $(version_code 2.9.53) ]; then
8223 do_facet mgs "$LCTL --device MGS llog_print $FSNAME-client" ||
8224 error "new llog_print failed"
8227 run_test 123aa "llog_print works with FIDs and simple names"
8230 remote_mgs_nodsh && skip "remote MGS with nodsh"
8231 [[ $MGS_VERSION -gt $(version_code 2.11.51) ]] ||
8232 skip "Need server with working llog_print support"
8234 [ -d $MOUNT/.lustre ] || setup
8239 orig_val=$(do_facet mgs $LCTL get_param jobid_name)
8240 do_facet mgs $LCTL set_param -P jobid_name="testname"
8242 yaml=$(do_facet mgs $LCTL --device MGS llog_print params |
8243 grep jobid_name | tail -n 1)
8245 local param=$(awk '{ print $10 }' <<< "$yaml")
8246 local val=$(awk '{ print $12 }' <<< "$yaml")
8247 #return to the default
8248 do_facet mgs $LCTL set_param -P jobid_name=$orig_val
8249 [ $val = "testname" ] || error "bad value: $val"
8250 [ $param = "jobid_name," ] || error "Bad param: $param"
8252 run_test 123ab "llog_print params output values from set_param -P"
8254 test_123ac() { # LU-11566
8255 remote_mgs_nodsh && skip "remote MGS with nodsh"
8256 do_facet mgs "$LCTL help llog_print" 2>&1 | grep -q -- --start ||
8257 skip "Need 'lctl llog_print --start' on MGS"
8262 [ -d $MOUNT/.lustre ] || setup
8264 # - { index: 10, event: add_uuid, nid: 192.168.20.1@tcp(0x20000c0a81401,
8265 # node: 192.168.20.1@tcp }
8266 do_facet mgs $LCTL --device MGS \
8267 llog_print --start $start --end $end $FSNAME-client | tr -d , |
8268 while read DASH BRACE INDEX idx EVENT BLAH BLAH BLAH; do
8269 (( idx >= start )) || error "llog_print index $idx < $start"
8270 (( idx <= end )) || error "llog_print index $idx > $end"
8273 run_test 123ac "llog_print with --start and --end"
8275 test_123ad() { # LU-11566
8276 remote_mgs_nodsh && skip "remote MGS with nodsh"
8277 # older versions of lctl may not print all records properly
8278 do_facet mgs "$LCTL help llog_print" 2>&1 | grep -q -- --start ||
8279 skip "Need 'lctl llog_print --start' on MGS"
8281 [ -d $MOUNT/.lustre ] || setup
8283 # append a new record, to avoid issues if last record was cancelled
8284 local old=$($LCTL get_param -n osc.*-OST0000-*.max_dirty_mb | head -1)
8285 do_facet mgs $LCTL conf_param $FSNAME-OST0000.osc.max_dirty_mb=$old
8287 # logid: [0x3:0xa:0x0]:0
8291 local num=$(do_facet mgs $LCTL --device MGS llog_info $FSNAME-client |
8292 awk '/last_index:/ { print $2 - 1 }')
8294 # - { index: 71, event: set_timeout, num: 0x14, param: sys.timeout=20 }
8295 local last=$(do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
8296 tail -1 | awk '{ print $4 }' | tr -d , )
8297 (( last == num )) || error "llog_print only showed $last/$num records"
8299 run_test 123ad "llog_print shows all records"
8302 remote_mgs_nodsh && skip "remote MGS with nodsh"
8304 [ -d $MOUNT/.lustre ] || setup
8305 local yaml_file="$TMP/$tfile.yaml"
8306 do_facet mgs rm "$yaml_file"
8307 local cfgfiles=$(do_facet mgs "lctl --device MGS llog_catlist" |
8308 sed 's/config_log://')
8310 # set jobid_var to a different value for test
8311 local orig_val=$(do_facet mgs $LCTL get_param jobid_var)
8312 do_facet mgs $LCTL set_param -P jobid_var="testname"
8314 for i in $cfgfiles params; do
8315 do_facet mgs "lctl --device MGS llog_print ${i} >> $yaml_file"
8318 echo "Unmounting FS"
8328 # Reapply the config from before
8329 echo "Setting configuration parameters"
8330 do_facet mgs "lctl set_param -F $yaml_file"
8332 local set_val=$(do_facet mgs $LCTL get_param jobid_var)
8333 do_facet mgs $LCTL set_param -P $orig_val
8335 [ $set_val == "jobid_var=testname" ] ||
8336 error "$set_val is not testname"
8338 do_facet mgs rm "$yaml_file"
8341 run_test 123F "clear and reset all parameters using set_param -F"
8345 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
8346 [ -z $mds2failover_HOST ] && skip "needs MDT failover setup"
8352 if combined_mgs_mds; then
8353 start_mdt 1 "-o nosvc" ||
8354 error "starting mds with nosvc option failed"
8356 local nid=$(do_facet mds2 $LCTL list_nids | head -1)
8357 local failover_nid=$(do_node $mds2failover_HOST $LCTL list_nids | head -1)
8358 do_facet mgs $LCTL replace_nids $FSNAME-MDT0001 $nid:$failover_nid ||
8359 error "replace_nids execution error"
8361 if combined_mgs_mds; then
8367 echo "lfs setdirstripe"
8368 $LFS setdirstripe -i 1 $MOUNT/$tdir || error "setdirstirpe error"
8371 run_test 124 "check failover after replace_nids"
8373 get_max_sectors_kb() {
8376 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
8377 local max_sectors_path="/sys/block/${dev_base}/queue/max_sectors_kb"
8379 do_facet ${facet} "[[ -e ${max_sectors_path} ]] &&
8380 cat ${max_sectors_path}"
8383 get_max_hw_sectors_kb() {
8386 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
8387 local max_hw_path="/sys/block/${dev_base}/queue/max_hw_sectors_kb"
8389 do_facet ${facet} "[[ -e ${max_hw_path} ]] && cat ${max_hw_path}"
8392 set_max_sectors_kb() {
8396 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
8397 local max_sectors_path="/sys/block/${dev_base}/queue/max_sectors_kb"
8399 do_facet ${facet} "[[ -e ${max_sectors_path} ]] &&
8400 echo ${value} > ${max_sectors_path}"
8403 [[ $rc -ne 0 ]] && echo "Failed to set ${max_sectors_path} to ${value}"
8408 # Return 0 if all slave devices have max_sectors_kb == max_hw_sectors_kb
8409 # Otherwise return > 0
8410 check_slaves_max_sectors_kb()
8414 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
8415 local slaves_dir=/sys/block/${dev_base}/slaves
8416 local slave_devices=$(do_facet ${facet} "ls ${slaves_dir} 2>/dev/null")
8417 [[ -z ${slave_devices} ]] && return 0
8419 local slave max_sectors new_max_sectors max_hw_sectors path
8421 for slave in ${slave_devices}; do
8422 path="/dev/${slave}"
8423 ! is_blkdev ${facet} ${path} && continue
8424 max_sectors=$(get_max_sectors_kb ${facet} ${path})
8425 max_hw_sectors=$(get_max_hw_sectors_kb ${facet} ${path})
8426 new_max_sectors=${max_hw_sectors}
8427 [[ ${new_max_sectors} -gt ${RQ_SIZE_LIMIT} ]] &&
8428 new_max_sectors=${RQ_SIZE_LIMIT}
8430 if [[ ${max_sectors} -ne ${new_max_sectors} ]]; then
8431 echo "${path} ${max_sectors} ${new_max_sectors}"
8434 check_slaves_max_sectors_kb ${facet} ${path}
8443 local facet_list="mgs mds1 ost1"
8444 combined_mgs_mds && facet_list="mgs ost1"
8447 for facet in ${facet_list}; do
8448 [[ $(facet_fstype ${facet}) != ldiskfs ]] &&
8449 skip "ldiskfs only test" &&
8451 ! is_blkdev ${facet} $(facet_device ${facet}) &&
8452 skip "requires all real devices" &&
8457 # We don't increase IO request size limit past 16MB. See comments in
8458 # lustre/utils/libmount_utils_ldiskfs.c:tune_max_sectors_kb()
8459 RQ_SIZE_LIMIT=$((16 * 1024))
8460 local device old_max_sectors new_max_sectors max_hw_sectors
8461 for facet in ${facet_list}; do
8462 device=$(facet_device ${facet})
8463 old_max_sectors=$(get_max_sectors_kb ${facet} ${device})
8464 max_hw_sectors=$(get_max_hw_sectors_kb ${facet} ${device})
8466 # The expected value after l_tunedisk is executed
8467 new_max_sectors=$old_max_sectors
8468 [[ ${new_max_sectors_kb} -gt ${RQ_SIZE_LIMIT} ]] &&
8469 new_max_sectors_kb=${RQ_SIZE_LIMIT}
8471 # Ensure the current value of max_sectors_kb does not equal
8472 # max_hw_sectors_kb, so we can tell whether l_tunedisk did
8474 set_max_sectors_kb ${facet} ${device} $((new_max_sectors - 1))
8476 # Value before l_tunedisk
8477 local pre_max_sectors=$(get_max_sectors_kb ${facet} ${device})
8478 if [[ ${pre_max_sectors} -ne $((new_max_sectors - 1)) ]]; then
8479 echo "unable to satsify test pre-condition:"
8480 echo "${pre_max_sectors} != $((new_max_sectors - 1))"
8485 echo "Before: ${facet} ${device} ${pre_max_sectors} ${max_hw_sectors}"
8487 do_facet ${facet} "libtool execute l_tunedisk ${device}"
8489 # Value after l_tunedisk
8490 local post_max_sectors=$(get_max_sectors_kb ${facet} ${device})
8492 echo "After: ${facet} ${device} ${post_max_sectors} ${max_hw_sectors}"
8494 if [[ ${facet} != ost1 ]]; then
8495 if [[ ${post_max_sectors} -ne ${pre_max_sectors} ]]; then
8496 echo "l_tunedisk modified max_sectors_kb of ${facet}"
8500 set_max_sectors_kb ${facet} ${device} ${old_max_sectors}
8502 if [[ ${post_max_sectors} -eq ${pre_max_sectors} ]]; then
8503 echo "l_tunedisk failed to modify max_sectors_kb of ${facet}"
8507 check_slaves_max_sectors_kb ${facet} ${device} ||
8514 run_test 125 "check l_tunedisk only tunes OSTs and their slave devices"
8516 if ! combined_mgs_mds ; then
8522 # restore the values of MDSSIZE and OSTSIZE
8523 MDSSIZE=$STORED_MDSSIZE
8524 OSTSIZE=$STORED_OSTSIZE
8528 check_and_cleanup_lustre