7 # bug number for skipped test: LU-2828
8 ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT 59 64"
9 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
13 if [ -r /etc/SuSE-release ]
15 local vers=$(grep VERSION /etc/SuSE-release | awk '{print $3}')
16 local patchlev=$(grep PATCHLEVEL /etc/SuSE-release |
18 if [ $vers -eq 11 ] && [ $patchlev -eq 2 ]
26 if is_sles11; then # LU-2181
27 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 23a 34b"
30 if [ "$FAILURE_MODE" = "HARD" ]; then
31 CONFIG_EXCEPTIONS="24a " &&
32 echo "Except the tests: $CONFIG_EXCEPTIONS for " \
33 "FAILURE_MODE=$FAILURE_MODE, b=23573" &&
34 ALWAYS_EXCEPT="$ALWAYS_EXCEPT $CONFIG_EXCEPTIONS"
37 # bug number for skipped test:
38 # a tool to create lustre filesystem images
39 ALWAYS_EXCEPT="32newtarball $ALWAYS_EXCEPT"
42 PATH=$PWD/$SRCDIR:$SRCDIR:$SRCDIR/../utils:$PATH
44 PTLDEBUG=${PTLDEBUG:--1}
46 LUSTRE=${LUSTRE:-$(dirname $0)/..}
47 RLUSTRE=${RLUSTRE:-$LUSTRE}
48 export MULTIOP=${MULTIOP:-multiop}
50 . $LUSTRE/tests/test-framework.sh
52 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
54 # use small MDS + OST size to speed formatting time
55 # do not use too small MDSSIZE/OSTSIZE, which affect the default jouranl size
56 # STORED_MDSSIZE is used in test_18
57 STORED_MDSSIZE=$MDSSIZE
58 STORED_OSTSIZE=$OSTSIZE
70 if ! combined_mgs_mds; then
71 # bug number for skipped test: 23954
72 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 24b"
75 # pass "-E lazy_itable_init" to mke2fs to speed up the formatting time
76 if [[ "$LDISKFS_MKFS_OPTS" != *lazy_itable_init* ]]; then
77 LDISKFS_MKFS_OPTS=$(csa_add "$LDISKFS_MKFS_OPTS" -E lazy_itable_init)
80 [ $(facet_fstype $SINGLEMDS) = "zfs" ] &&
81 # bug number for skipped test: LU-2778 LU-4444
82 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 57b 69"
87 require_dsh_mds || exit 0
88 require_dsh_ost || exit 0
90 [ "$SLOW" = "no" ] && EXCEPT_SLOW="30a 31 45 69"
95 # The MGS must be started before the OSTs for a new fs, so start
96 # and stop to generate the startup logs.
99 wait_osc_import_state mds ost FULL
104 reformat_and_config() {
106 if ! combined_mgs_mds ; then
112 writeconf_or_reformat() {
113 # There are at most 2 OSTs for write_conf test
114 # who knows if/where $TUNEFS is installed?
115 # Better reformat if it fails...
116 writeconf_all $MDSCOUNT 2 ||
117 { echo "tunefs failed, reformatting instead" &&
118 reformat_and_config && return 0; }
128 start mgs $(mgsdevname) $MGS_MOUNT_OPTS
134 local dev=$(mdsdevname $num)
137 echo "start mds service on `facet_active_host $facet`"
138 start $facet ${dev} $MDS_MOUNT_OPTS $@ || return 94
144 local dev=$(mdsdevname $num)
147 echo "stop mds service on `facet_active_host $facet`"
148 # These tests all use non-failover stop
149 stop $facet -f || return 97
155 for num in $(seq $MDSCOUNT); do
156 start_mdt $num $@ || return 94
161 if ! combined_mgs_mds ; then
169 for num in $(seq $MDSCOUNT); do
170 stop_mdt $num || return 97
175 echo "stop mgs service on `facet_active_host mgs`"
176 # These tests all use non-failover stop
177 stop mgs -f || return 97
181 echo "start ost1 service on `facet_active_host ost1`"
182 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS $@ || return 95
186 echo "stop ost1 service on `facet_active_host ost1`"
187 # These tests all use non-failover stop
188 stop ost1 -f || return 98
192 echo "start ost2 service on `facet_active_host ost2`"
193 start ost2 $(ostdevname 2) $OST_MOUNT_OPTS $@ || return 92
197 echo "stop ost2 service on `facet_active_host ost2`"
198 # These tests all use non-failover stop
199 stop ost2 -f || return 93
204 echo "mount $FSNAME on ${MOUNTPATH}....."
205 zconf_mount $(hostname) $MOUNTPATH || return 96
209 local mountopt="remount,$1"
211 echo "remount '$1' lustre on ${MOUNTPATH}....."
212 zconf_mount $(hostname) $MOUNTPATH "$mountopt" || return 96
217 echo "umount lustre on ${MOUNTPATH}....."
218 zconf_umount $(hostname) $MOUNTPATH || return 97
221 manual_umount_client(){
224 echo "manual umount lustre on ${MOUNT}...."
225 do_facet client "umount -d ${FORCE} $MOUNT"
231 start_mds || error "MDT start failed"
232 start_ost || error "Unable to start OST1"
233 mount_client $MOUNT || error "client start failed"
234 client_up || error "client_up failed"
243 unload_modules_conf () {
244 if combined_mgs_mds || ! local_mode; then
245 unload_modules || return 1
250 stop_ost || return 202
251 stop_mds || return 201
252 unload_modules_conf || return 203
256 umount_client $MOUNT || return 200
257 cleanup_nocli || return $?
262 echo "umount $MOUNT2 ..."
263 umount $MOUNT2 || true
264 echo "stopping fs2mds ..."
265 stop fs2mds -f || true
266 echo "stopping fs2ost ..."
267 stop fs2ost -f || true
271 do_facet client "cp /etc/passwd $DIR/a" || return 71
272 do_facet client "rm $DIR/a" || return 72
273 # make sure lustre is actually mounted (touch will block,
274 # but grep won't, so do it after)
275 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
276 echo "setup single mount lustre success"
280 do_facet client "touch $DIR/a" || return 71
281 do_facet client "rm $DIR/a" || return 72
282 do_facet client "touch $DIR2/a" || return 73
283 do_facet client "rm $DIR2/a" || return 74
284 echo "setup double mount lustre success"
289 if [ "$ONLY" == "setup" ]; then
294 if [ "$ONLY" == "cleanup" ]; then
301 #create single point mountpoint
307 check_mount || error "check_mount failed"
308 cleanup || error "cleanup failed with $?"
310 run_test 0 "single mount setup"
313 start_mds || error "MDS start failed"
314 start_ost || error "unable to start OST"
315 echo "start ost second time..."
316 start_ost && error "2nd OST start should fail"
317 mount_client $MOUNT || error "client start failed"
318 check_mount || error "check_mount failed"
319 cleanup || error "cleanup failed with $?"
321 run_test 1 "start up ost twice (should return errors)"
324 start_mdt 1 || error "MDT0 start fail"
325 echo "start mds second time.."
326 start_mdt 1 && error "2nd MDT start should fail"
327 start_ost || error "OST start failed"
328 mount_client $MOUNT || error "mount_client failed to start client"
329 check_mount || error "check_mount failed"
330 cleanup || error "cleanup failed with $?"
332 run_test 2 "start up mds twice (should return err)"
336 #mount.lustre returns an error if already in mtab
337 mount_client $MOUNT && error "2nd client mount should fail"
338 check_mount || error "check_mount failed"
339 cleanup || error "cleanup failed with $?"
341 run_test 3 "mount client twice (should return err)"
345 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
346 stop_ost || error "Unable to stop OST1"
349 # ok for ost to fail shutdown
350 if [ 202 -ne $eno ] && [ 0 -ne $eno ]; then
351 error "cleanup failed with $?"
354 run_test 4 "force cleanup ost, then cleanup"
356 test_5a() { # was test_5
358 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
359 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
361 stop_mds || error "Unable to stop MDS"
363 # cleanup may return an error from the failed
364 # disconnects; for now I'll consider this successful
365 # if all the modules have unloaded.
369 echo "killing umount"
370 kill -TERM $UMOUNT_PID
371 echo "waiting for umount to finish"
373 if grep " $MOUNT " /proc/mounts; then
374 echo "test 5: /proc/mounts after failed umount"
378 echo "killing umount"
379 kill -TERM $UMOUNT_PID
380 echo "waiting for umount to finish"
382 grep " $MOUNT " /proc/mounts &&
383 error "/proc/mounts after second umount"
386 # manual_mount_client may fail due to umount succeeding above
388 # stop_mds is a no-op here, and should not fail
389 cleanup_nocli || error "cleanup_nocli failed with $?"
390 # df may have lingering entry
392 # mtab may have lingering entry
396 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
398 grep -q $MOUNT" " /etc/mtab || break
399 echo "Waiting /etc/mtab updated ... "
400 WAIT=$(( WAIT + sleep))
402 [ "$WAIT" -eq "$MAX_WAIT" ] &&
403 error "/etc/mtab is not updated in $WAIT secs"
404 echo "/etc/mtab updated in $WAIT secs"
406 run_test 5a "force cleanup mds, then cleanup"
414 grep " $MOUNT " /etc/mtab &&
415 error false "unexpected entry in mtab before mount" && return 10
417 start_ost || error "OST start failed"
418 if ! combined_mgs_mds ; then
419 trap cleanup_5b EXIT ERR
420 start_mds || error "MDS start failed"
424 mount_client $MOUNT && error "mount_client $MOUNT should fail"
425 grep " $MOUNT " /etc/mtab &&
426 error "$MOUNT entry in mtab after failed mount"
428 # stop_mds is a no-op here, and should not fail
429 cleanup_nocli || error "cleanup_nocli failed with $?"
430 if ! combined_mgs_mds ; then
434 run_test 5b "Try to start a client with no MGS (should return errs)"
437 grep " $MOUNT " /etc/mtab &&
438 error false "unexpected entry in mtab before mount" && return 10
440 start_mds || error "MDS start failed"
441 start_ost || error "OST start failed"
442 local oldfs="${FSNAME}"
443 FSNAME="wrong.${FSNAME}"
444 mount_client $MOUNT || :
446 grep " $MOUNT " /etc/mtab &&
447 error "$MOUNT entry in mtab after failed mount"
449 cleanup_nocli || error "cleanup_nocli failed with $?"
451 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
454 grep " $MOUNT " /etc/mtab &&
455 error false "unexpected entry in mtab before mount" && return 10
457 start_ost || error "OST start failed"
458 start_mds || error "MDS start failed"
459 stop_ost || error "Unable to stop OST1"
460 mount_client $MOUNT || error "mount_client $MOUNT failed"
461 cleanup || error "cleanup_nocli failed with $?"
462 grep " $MOUNT " /etc/mtab &&
463 error "$MOUNT entry in mtab after unmount"
466 run_test 5d "mount with ost down"
469 grep " $MOUNT " /etc/mtab &&
470 error false "unexpected entry in mtab before mount" && return 10
472 start_mds || error "MDS start failed"
473 start_ost || error "OST start failed"
475 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
476 do_facet client "$LCTL set_param fail_loc=0x80000506"
477 mount_client $MOUNT || echo "mount failed (not fatal)"
478 cleanup || error "cleanup failed with $?"
479 grep " $MOUNT " /etc/mtab &&
480 error "$MOUNT entry in mtab after unmount"
483 run_test 5e "delayed connect, don't crash (bug 10268)"
486 if combined_mgs_mds ; then
487 skip "combined mgs and mds"
491 grep " $MOUNT " /etc/mtab &&
492 error false "unexpected entry in mtab before mount" && return 10
495 start_ost || error "OST start failed"
496 mount_client $MOUNT &
498 echo client_mount pid is $pid
502 if ! ps -f -p $pid >/dev/null; then
505 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
506 error "mount returns $rc, expected to hang"
508 cleanup || error "cleanup failed with $?"
513 start_mds || error "start MDS failed"
515 # mount should succeed after start mds
517 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
518 cleanup || error "final call to cleanup failed with rc $?"
520 run_test 5f "mds down, cleanup after failed mount (bug 2712)"
525 mount_client $MOUNT || error "mount_client $MOUNT failed"
526 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
527 cleanup || error "cleanup failed with rc $?"
529 run_test 6 "manual umount, then mount again"
534 cleanup_nocli || error "cleanup_nocli failed with $?"
536 run_test 7 "manual umount, then cleanup"
540 mount_client $MOUNT2 || error "mount_client $MOUNT2 failed"
541 check_mount2 || error "check_mount2 failed"
542 umount_client $MOUNT2 || error "umount_client $MOUNT2 failed"
543 cleanup || error "cleanup failed with rc $?"
545 run_test 8 "double mount setup"
548 start_ost || error "OST start failed"
550 do_facet ost1 $LCTL set_param debug=\'inode trace\' ||
551 error "do_facet ost1 set_param inode trace failed."
552 do_facet ost1 $LCTL set_param subsystem_debug=\'mds ost\' ||
553 error "do_facet ost1 set_param debug mds ost failed."
555 CHECK_PTLDEBUG="`do_facet ost1 $LCTL get_param -n debug`"
556 if [ "$CHECK_PTLDEBUG" ] && { \
557 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
558 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
559 echo "lnet.debug success"
561 error "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
563 CHECK_SUBSYS="`do_facet ost1 $LCTL get_param -n subsystem_debug`"
564 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
565 echo "lnet.subsystem_debug success"
567 error "lnet.subsystem_debug: want 'mds ost' got '$CHECK_SUBSYS'"
569 stop_ost || error "Unable to stop OST1"
571 run_test 9 "test ptldebug and subsystem for mkfs"
579 do_facet $facet "test -b $dev" || rc=1
580 if [[ "$size" ]]; then
581 local in=$(do_facet $facet "dd if=$dev of=/dev/null bs=1k \
582 count=1 skip=$size 2>&1" |
583 awk '($3 == "in") { print $1 }')
584 [[ $in = "1+0" ]] || rc=1
590 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
591 # But with new MDS stack we don't care about the mode of local objects
592 # anymore, so this test is removed. See bug 22944 for more details.
596 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
597 skip "Only applicable to ldiskfs-based MDTs"
602 check_mount || error "check_mount failed"
603 cleanup || error "cleanup failed with rc $?"
605 echo "Remove mds config log"
606 if ! combined_mgs_mds ; then
610 do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' \
611 $(mgsdevname) || return \$?" ||
612 error "do_facet mgs failed with $?"
614 if ! combined_mgs_mds ; then
618 start_ost || error "OST start failed"
619 start_mds && error "MDS start succeeded, but should fail"
622 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
625 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
626 skip "Only applicable to ldiskfs-based MDTs"
630 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
635 # check if current MDSSIZE is large enough
636 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE &&
637 log "use MDSSIZE=$MDSSIZE"
639 # check if the global config has a large enough MDSSIZE
640 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] &&
641 [ $STORED_MDSSIZE -ge $MIN ] &&
642 OK=1 && myMDSSIZE=$STORED_MDSSIZE &&
643 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
645 # check if the block device is large enough
646 is_blkdev $SINGLEMDS $MDSDEV $MIN
647 local large_enough=$?
648 if [ -n "$OK" ]; then
649 [ $large_enough -ne 0 ] && OK=""
651 [ $large_enough -eq 0 ] && OK=1 && myMDSSIZE=$MIN &&
652 log "use device $MDSDEV with MIN=$MIN"
655 # check if a loopback device has enough space for fs metadata (5%)
657 if [ -z "$OK" ]; then
658 local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! \
659 -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
660 awk '($1 != "Filesystem") { print $4 }')
661 ! [ -z "$SPACE" ] && [ $SPACE -gt $((MIN / 20)) ] &&
662 OK=1 && myMDSSIZE=$MIN &&
663 log "use file $MDSDEV with MIN=$MIN"
666 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS" && return
668 echo "mount mds with large journal..."
670 local OLD_MDSSIZE=$MDSSIZE
674 echo "mount lustre system..."
676 check_mount || error "check_mount failed"
678 echo "check journal size..."
679 local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
680 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
681 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
683 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
686 cleanup || error "cleanup failed with rc $?"
691 run_test 18 "check mkfs creates large journals"
694 start_mds || error "MDS start failed"
695 stop_mds || error "Unable to stop MDS"
697 run_test 19a "start/stop MDS without OSTs"
700 start_ost || error "Unable to start OST1"
701 stop_ost || error "Unable to stop OST1"
703 run_test 19b "start/stop OSTs without MDS"
706 # first format the ost/mdt
707 start_mds || error "MDS start failed"
708 start_ost || error "Unable to start OST1"
709 mount_client $MOUNT || error "mount_client $MOUNT failed"
710 check_mount || error "check_mount failed"
711 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed."
712 remount_client ro $MOUNT || error "remount_client with ro failed"
713 touch $DIR/$tfile && error "$DIR/$tfile created incorrectly"
714 [ -e $DIR/$tfile ] && error "$DIR/$tfile exists incorrectly"
715 remount_client rw $MOUNT || error "remount_client with rw failed"
716 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
717 MCNT=$(grep -c $MOUNT /etc/mtab)
718 [ "$MCNT" -ne 1 ] && error "$MOUNT in /etc/mtab $MCNT times"
720 stop_mds || error "Unable to stop MDS"
721 stop_ost || error "Unable to stop OST1"
723 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
726 start_mds || error "MDS start failed"
727 start_ost || error "unable to start OST1"
728 wait_osc_import_state mds ost FULL
729 stop_ost || error "unable to stop OST1"
730 stop_mds || error "unable to stop MDS"
732 run_test 21a "start mds before ost, stop ost first"
735 start_ost || error "unable to start OST1"
736 start_mds || error "MDS start failed"
737 wait_osc_import_state mds ost FULL
738 stop_mds || error "unable to stop MDS"
739 stop_ost || error "unable to stop OST1"
741 run_test 21b "start ost before mds, stop mds first"
744 start_ost || error "Unable to start OST1"
745 start_mds || error "MDS start failed"
746 start_ost2 || error "Unable to start OST2"
747 wait_osc_import_state mds ost2 FULL
748 stop_ost || error "Unable to stop OST1"
749 stop_ost2 || error "Unable to stop OST2"
750 stop_mds || error "Unable to stop MDS"
751 #writeconf to remove all ost2 traces for subsequent tests
752 writeconf_or_reformat
754 run_test 21c "start mds between two osts, stop mds last"
757 if combined_mgs_mds ; then
758 skip "need separate mgs device" && return 0
764 start_mgs || error "unable to start MGS"
765 start_ost || error "unable to start OST1"
766 start_ost2 || error "unable to start OST2"
767 start_mds || error "MDS start failed"
768 wait_osc_import_state mds ost2 FULL
770 stop_ost || error "Unable to stop OST1"
771 stop_ost2 || error "Unable to stop OST2"
772 stop_mds || error "Unable to stop MDS"
774 #writeconf to remove all ost2 traces for subsequent tests
775 writeconf_or_reformat
776 start_mgs || error "unable to start MGS"
778 run_test 21d "start mgs then ost and then mds"
781 MGSNID="$saved_mgsnid"
783 echo "stopping fs2mgs ..."
784 stop $fs2mgs -f || true
787 test_21e() { # LU-5863
788 if [[ -z "$fs3ost_DEV" || -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
789 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
790 skip_env "mixed loopback and real device not working" && return
793 local fs2mdsdev=$(mdsdevname 1_2)
794 local fs2ostdev=$(ostdevname 1_2)
795 local fs3ostdev=$(ostdevname 2_2)
797 local fs2mdsvdev=$(mdsvdevname 1_2)
798 local fs2ostvdev=$(ostvdevname 1_2)
799 local fs3ostvdev=$(ostvdevname 2_2)
801 # temporarily use fs3ost as fs2mgs
803 local fs2mgsdev=$fs3ostdev
804 local fs2mgsvdev=$fs3ostvdev
806 local fsname=test1234
808 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
809 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
810 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS && trap cleanup_21e EXIT INT ||
811 error "start fs2mgs failed"
813 local saved_mgsnid="$MGSNID"
814 MGSNID=$(do_facet $fs2mgs $LCTL list_nids | xargs | tr ' ' ,)
816 add fs2mds $(mkfs_opts mds1 $fs2mdsdev $fsname) \
817 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
818 add fs2ost $(mkfs_opts ost1 $fs2ostdev $fsname) \
819 --reformat $fs2ostdev $fs2ostvdev || error "add fs2ost failed"
821 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
822 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS || error "start fs2mds failed"
824 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
825 $MOUNT_CMD $MGSNID:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
826 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
830 run_test 21e "separate MGS and MDS"
833 start_mds || error "MDS start failed"
835 echo "Client mount with ost in logs, but none running"
836 start_ost || error "unable to start OST1"
837 # wait until mds connected to ost and open client connection
838 wait_osc_import_state mds ost FULL
839 stop_ost || error "unable to stop OST1"
840 mount_client $MOUNT || error "mount_client $MOUNT failed"
841 # check_mount will block trying to contact ost
842 mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed"
843 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed"
847 echo "Client mount with a running ost"
848 start_ost || error "unable to start OST1"
850 # if gss enabled, wait full time to let connection from
851 # mds to ost be established, due to the mismatch between
852 # initial connect timeout and gss context negotiation timeout.
853 # This perhaps could be remove after AT landed.
854 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
855 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
857 mount_client $MOUNT || error "mount_client $MOUNT failed"
858 wait_osc_import_state mds ost FULL
859 wait_osc_import_state client ost FULL
860 check_mount || error "check_mount failed"
863 cleanup || error "cleanup failed with rc $?"
865 run_test 22 "start a client before osts (should return errs)"
867 test_23a() { # was test_23
870 stop $SINGLEMDS || error "failed to stop $SINGLEMDS"
871 # force down client so that recovering mds waits for reconnect
872 local running=$(grep -c $MOUNT /proc/mounts) || true
873 if [ $running -ne 0 ]; then
874 echo "Stopping client $MOUNT (opts: -f)"
878 # enter recovery on failed mds
879 local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/})
880 start $SINGLEMDS $MDT_DEV $MDS_MOUNT_OPTS || error "MDS start failed"
881 # try to start a new client
882 mount_client $MOUNT &
884 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
885 MOUNT_LUSTRE_PID=$(ps -ef | grep mount.lustre |
886 grep -v grep | awk '{print $2}')
887 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
889 ps --ppid $MOUNT_LUSTRE_PID
890 echo "waiting for mount to finish"
892 # "ctrl-c" sends SIGINT but it usually (in script) does not work on child process
893 # SIGTERM works but it does not spread to offspring processses
894 kill -s TERM $MOUNT_PID
895 kill -s TERM $MOUNT_LUSTRE_PID
896 # we can not wait $MOUNT_PID because it is not a child of this shell
902 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
904 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
905 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
908 [ -z "$PID1" -a -z "$PID2" ] && break
909 echo "waiting for mount to finish ... "
910 WAIT=$(( WAIT + sleep))
912 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
913 error "MOUNT_PID $MOUNT_PID and "\
914 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
917 stop_mds || error "stopping MDSes failed"
918 stop_ost || error "stopping OSSes failed"
920 run_test 23a "interrupt client during recovery mount delay"
925 test_23b() { # was test_23
926 start_mds || error "MDS start failed"
927 start_ost || error "Unable to start OST1"
928 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
929 $LCTL set_param fail_loc=0x80000313
931 cleanup || error "cleanup failed with rc $?"
933 run_test 23b "Simulate -EINTR during mount"
936 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
938 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
939 is_blkdev $SINGLEMDS $MDSDEV &&
940 skip_env "mixed loopback and real device not working" && return
943 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
945 local fs2mdsdev=$(mdsdevname 1_2)
946 local fs2ostdev=$(ostdevname 1_2)
947 local fs2mdsvdev=$(mdsvdevname 1_2)
948 local fs2ostvdev=$(ostvdevname 1_2)
950 # test 8-char fsname as well
951 local FSNAME2=test1234
953 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --nomgs --mgsnode=$MGSNID \
954 --fsname=${FSNAME2} --reformat $fs2mdsdev $fs2mdsvdev || exit 10
956 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --fsname=${FSNAME2} \
957 --reformat $fs2ostdev $fs2ostvdev || exit 10
960 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
961 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
962 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
963 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
965 check_mount || error "check_mount failed"
966 # files written on 1 should not show up on 2
967 cp /etc/passwd $DIR/$tfile
969 [ -e $MOUNT2/$tfile ] && error "File bleed"
972 cp /etc/passwd $MOUNT2/$tfile ||
973 error "cp /etc/passwd $MOUNT2/$tfile failed"
974 rm $MOUNT2/$tfile || error "remove $MOUNT2/$tfile failed"
975 # 2 is actually mounted
976 grep $MOUNT2' ' /proc/mounts > /dev/null || error "$MOUNT2 not mounted"
978 facet_failover fs2mds
979 facet_failover fs2ost
982 # the MDS must remain up until last MDT
984 MDS=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
985 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
986 [ -z "$MDS" ] && error "No MDT"
988 cleanup_nocli || error "cleanup_nocli failed with rc $?"
990 run_test 24a "Multiple MDTs on a single node"
993 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
995 if [ -z "$fs2mds_DEV" ]; then
996 local dev=${SINGLEMDS}_dev
998 is_blkdev $SINGLEMDS $MDSDEV &&
999 skip_env "mixed loopback and real device not working" && return
1002 local fs2mdsdev=$(mdsdevname 1_2)
1003 local fs2mdsvdev=$(mdsvdevname 1_2)
1005 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --mgs --fsname=${FSNAME}2 \
1006 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
1008 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS &&
1009 error "start MDS should fail"
1011 cleanup || error "cleanup failed with rc $?"
1013 run_test 24b "Multiple MGSs on a single node (should return err)"
1017 check_mount || error "check_mount failed"
1018 local MODULES=$($LCTL modules | awk '{ print $2 }')
1019 rmmod $MODULES 2>/dev/null || true
1020 cleanup || error "cleanup failed with $?"
1022 run_test 25 "Verify modules are referenced"
1026 # we need modules before mount for sysctl, so make sure...
1027 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
1028 #define OBD_FAIL_MDS_FS_SETUP 0x135
1029 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000135"
1030 start_mds && error "MDS started but should not have started"
1031 $LCTL get_param -n devices
1032 DEVS=$($LCTL get_param -n devices | egrep -v MG | wc -l)
1033 [ $DEVS -gt 0 ] && error "number of devices is $DEVS, should be zero"
1034 # start mds to drop writeconf setting
1035 start_mds || error "Unable to start MDS"
1036 stop_mds || error "Unable to stop MDS"
1037 unload_modules_conf || error "unload_modules_conf failed with $?"
1039 run_test 26 "MDT startup failure cleans LOV (should return errs)"
1042 start_ost || error "Unable to start OST1"
1043 start_mds || error "Unable to start MDS"
1044 echo "Requeue thread should have started: "
1045 ps -e | grep ll_cfg_requeue
1046 set_conf_param_and_check ost1 \
1047 "$LCTL get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" \
1048 "$FSNAME-OST0000.ost.client_cache_seconds" ||
1049 error "set_conf_param_and_check ost1 failed"
1050 cleanup_nocli || error "cleanup_nocli failed with rc $?"
1052 run_test 27a "Reacquire MGS lock if OST started first"
1057 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
1058 awk '($3 ~ "mdt" && $4 ~ "MDT0000") { print $4 }')
1060 facet_failover $SINGLEMDS
1061 set_conf_param_and_check $SINGLEMDS \
1062 "$LCTL get_param -n mdt.$device.identity_acquire_expire" \
1063 "$device.mdt.identity_acquire_expire" ||
1064 error "set_conf_param_and_check $SINGLEMDS failed"
1065 set_conf_param_and_check client \
1066 "$LCTL get_param -n mdc.$device-mdc-*.max_rpcs_in_flight"\
1067 "$device.mdc.max_rpcs_in_flight" ||
1068 error "set_conf_param_and_check client failed"
1070 cleanup || error "cleanup failed with $?"
1072 run_test 27b "Reacquire MGS lock after failover"
1076 TEST="$LCTL get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1077 PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
1079 FINAL=$(($ORIG + 1))
1080 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL ||
1081 error "first set_conf_param_and_check client failed"
1082 FINAL=$(($FINAL + 1))
1083 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL ||
1084 error "second set_conf_param_and_check client failed"
1085 umount_client $MOUNT || error "umount_client $MOUNT failed"
1086 mount_client $MOUNT || error "mount_client $MOUNT failed"
1088 if [ $RESULT -ne $FINAL ]; then
1089 error "New config not seen: wanted $FINAL got $RESULT"
1091 echo "New config success: got $RESULT"
1093 set_conf_param_and_check client "$TEST" "$PARAM" $ORIG ||
1094 error "third set_conf_param_and_check client failed"
1095 cleanup || error "cleanup failed with rc $?"
1097 run_test 28 "permanent parameter setting"
1099 test_28a() { # LU-4221
1100 [[ $(lustre_version_code ost1) -ge $(version_code 2.5.52) ]] ||
1101 { skip "Need OST version at least 2.5.52" && return 0; }
1102 [ "$(facet_fstype ost1)" = "zfs" ] &&
1103 skip "LU-4221: no such proc params for ZFS OSTs" && return
1110 local device="$FSNAME-OST0000"
1114 # In this test we will set three kinds of proc parameters with
1116 # 1. the ones moved from the OFD to the OSD, and only their
1117 # symlinks kept in obdfilter
1118 # 2. non-symlink ones in the OFD
1119 # 3. non-symlink ones in the OSD
1122 # prepare a symlink parameter in the OFD
1123 name="writethrough_cache_enable"
1124 param="$device.ost.$name"
1125 cmd="$LCTL get_param -n obdfilter.$device.$name"
1127 # conf_param the symlink parameter in the OFD
1128 old=$(do_facet ost1 $cmd)
1129 new=$(((old + 1) % 2))
1130 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1131 error "lctl conf_param $device.ost.$param=$new failed"
1133 # conf_param the target parameter in the OSD
1134 param="$device.osd.$name"
1135 cmd="$LCTL get_param -n osd-*.$device.$name"
1136 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1137 error "lctl conf_param $device.osd.$param=$old failed"
1140 # prepare a non-symlink parameter in the OFD
1141 name="client_cache_seconds"
1142 param="$device.ost.$name"
1143 cmd="$LCTL get_param -n obdfilter.$device.$name"
1145 # conf_param the parameter in the OFD
1146 old=$(do_facet ost1 $cmd)
1148 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1149 error "lctl conf_param $device.ost.$param=$new failed"
1150 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1151 error "lctl conf_param $device.ost.$param=$old failed"
1154 # prepare a non-symlink parameter in the OSD
1155 name="lma_self_repair"
1156 param="$device.osd.$name"
1157 cmd="$LCTL get_param -n osd-*.$device.$name"
1159 # conf_param the parameter in the OSD
1160 old=$(do_facet ost1 $cmd)
1161 new=$(((old + 1) % 2))
1162 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1163 error "lctl conf_param $device.osd.$param=$new failed"
1164 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1165 error "lctl conf_param $device.osd.$param=$old failed"
1167 cleanup || error "cleanup failed with $?"
1169 run_test 28a "set symlink parameters permanently with conf_param"
1172 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
1173 setup > /dev/null 2>&1
1174 start_ost2 || error "Unable to start OST2"
1177 local PARAM="$FSNAME-OST0001.osc.active"
1178 local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
1179 local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
1181 ACTV=$($LCTL get_param -n $PROC_ACT)
1183 set_conf_param_and_check client \
1184 "$LCTL get_param -n $PROC_ACT" "$PARAM" $DEAC ||
1185 error "set_conf_param_and_check client failed"
1186 # also check ost_server_uuid status
1187 RESULT=$($LCTL get_param -n $PROC_UUID | grep DEACTIV)
1188 if [ -z "$RESULT" ]; then
1189 error "Client not deactivated: $($LCTL get_param \
1192 echo "Live client success: got $RESULT"
1196 for num in $(seq $MDSCOUNT); do
1197 local mdtosc=$(get_mdtosc_proc_path mds${num} $FSNAME-OST0001)
1198 local MPROC="osc.$mdtosc.active"
1203 RESULT=$(do_facet mds${num} "$LCTL get_param -n $MPROC")
1204 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
1205 if [ $RESULT -eq $DEAC ]; then
1206 echo -n "MDT deactivated also after"
1207 echo "$WAIT sec (got $RESULT)"
1211 if [ $WAIT -eq $MAX ]; then
1212 error "MDT active: wanted $DEAC got $RESULT"
1214 echo "Waiting $(($MAX - $WAIT))secs for MDT deactivated"
1217 # test new client starts deactivated
1218 umount_client $MOUNT || error "umount_client $MOUNT failed"
1219 mount_client $MOUNT || error "mount_client $MOUNT failed"
1220 RESULT=$($LCTL get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
1221 if [ -z "$RESULT" ]; then
1222 error "New client start active: $(lctl get_param -n $PROC_UUID)"
1224 echo "New client success: got $RESULT"
1227 # make sure it reactivates
1228 set_conf_param_and_check client \
1229 "$LCTL get_param -n $PROC_ACT" "$PARAM" $ACTV ||
1230 error "lctl get_param $PROC_ACT $PARAM $ACTV failed"
1232 umount_client $MOUNT
1233 stop_ost2 || error "Unable to stop OST2"
1234 cleanup_nocli || error "cleanup_nocli failed with $?"
1235 #writeconf to remove all ost2 traces for subsequent tests
1236 writeconf_or_reformat
1238 run_test 29 "permanently remove an OST"
1243 echo Big config llog
1244 TEST="$LCTL get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1246 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1247 for i in ${LIST[@]}; do
1248 set_conf_param_and_check client "$TEST" \
1249 "$FSNAME.llite.max_read_ahead_whole_mb" $i ||
1250 error "Set $FSNAME.llite.max_read_ahead_whole_mb failed"
1252 # make sure client restart still works
1253 umount_client $MOUNT
1254 mount_client $MOUNT || error "mount_client $MOUNT failed"
1255 [ "$($TEST)" -ne "$i" ] &&
1256 error "Param didn't stick across restart $($TEST) != $i"
1259 echo Erase parameter setting
1260 do_facet mgs "$LCTL conf_param \
1261 -d $FSNAME.llite.max_read_ahead_whole_mb" ||
1262 error "Erase param $FSNAME.llite.max_read_ahead_whole_mb failed"
1263 umount_client $MOUNT
1264 mount_client $MOUNT || error "mount_client $MOUNT failed"
1266 echo "deleted (default) value=$FINAL, orig=$ORIG"
1267 # assumes this parameter started at the default value
1268 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
1270 cleanup || error "cleanup failed with rc $?"
1272 run_test 30a "Big config llog and conf_param deletion"
1277 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1278 # numerical part of it. Hopefully that's not already a failover address
1280 OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
1281 ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1282 NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1283 NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1284 echo "Using fake nid $NEW"
1286 TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1287 grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1288 set_conf_param_and_check client "$TEST" \
1289 "$FSNAME-OST0000.failover.node" $NEW ||
1290 error "didn't add failover nid $NEW"
1291 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1294 # The NIDS value is the failover nid strings and "[" and "]". So
1295 # we need to subtract the space taken by the delimiters. This has
1296 # changed from earlier version of Lustre but this test is run only
1297 # locally so this change will not break interop. See LU-3386
1298 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 3))
1299 echo "should have 2 failover nids: $NIDCOUNT"
1300 [ $NIDCOUNT -eq 2 ] || error "Failover nid not added"
1301 do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" ||
1302 error "conf_param delete failed"
1303 umount_client $MOUNT
1304 mount_client $MOUNT || error "mount_client $MOUNT failed"
1306 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1309 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 3))
1310 echo "only 1 final nid should remain: $NIDCOUNT"
1311 [ $NIDCOUNT -eq 1 ] || error "Failover nids not removed"
1313 cleanup || error "cleanup failed with rc $?"
1315 run_test 30b "Remove failover nids"
1317 test_31() { # bug 10734
1318 # ipaddr must not exist
1319 $MOUNT_CMD 4.3.2.1@tcp:/lustre $MOUNT || true
1320 cleanup || error "cleanup failed with rc $?"
1322 run_test 31 "Connect to non-existent node (shouldn't crash)"
1326 T32_BLIMIT=20480 # Kbytes
1330 # This is not really a test but a tool to create new disk
1331 # image tarballs for the upgrade tests.
1333 # Disk image tarballs should be created on single-node
1334 # clusters by running this test with default configurations
1335 # plus a few mandatory environment settings that are verified
1336 # at the beginning of the test.
1338 test_32newtarball() {
1342 local tmp=$TMP/t32_image_create
1344 if [ $FSNAME != t32fs -o $MDSCOUNT -ne 1 -o \
1345 \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o $OSTCOUNT -ne 1 -o \
1346 -z "$OSTDEV1" ]; then
1347 error "Needs FSNAME=t32fs MDSCOUNT=1 MDSDEV1=<nonexistent_file>" \
1348 "(or MDSDEV, in the case of b1_8) OSTCOUNT=1" \
1349 "OSTDEV1=<nonexistent_file>"
1353 echo "Found stale $tmp"
1357 mkdir $tmp/src || return 1
1358 tar cf - -C $src . | tar xf - -C $tmp/src
1359 dd if=/dev/zero of=$tmp/src/t32_qf_old bs=1M \
1360 count=$(($T32_BLIMIT / 1024 / 2))
1361 chown $T32_QID.$T32_QID $tmp/src/t32_qf_old
1363 # format ost with comma-separated NIDs to verify LU-4460
1364 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
1365 MGSNID="$MGSNID,$MGSNID" OSTOPT="--failnode=$failnid" formatall
1369 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ] &&
1370 $LFS quotacheck -ug /mnt/$FSNAME
1371 $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
1374 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1377 mkdir $tmp/img || return 1
1381 ls -Rni --time-style=+%s >$tmp/img/list
1382 find . ! -name .lustre -type f -exec sha1sum {} \; |
1383 sort -k 2 >$tmp/img/sha1sums
1385 $LCTL get_param -n version | head -n 1 |
1386 sed -e 's/^lustre: *//' >$tmp/img/commit
1388 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ] &&
1389 $LFS quotaon -ug /mnt/$FSNAME
1390 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1391 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1392 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1393 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1394 | tr -d "*" > $tmp/img/bspace
1395 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1396 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1397 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1398 | tr -d "*" > $tmp/img/ispace
1403 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1406 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1407 echo "Data verification failed"
1410 uname -r >$tmp/img/kernel
1411 uname -m >$tmp/img/arch
1413 mv ${MDSDEV1:-$MDSDEV} $tmp/img
1414 mv $OSTDEV1 $tmp/img
1416 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1417 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1420 tar cjvf $dst/disk$version-$(facet_fstype $SINGLEMDS).tar.bz2 -S *
1425 #run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1428 # The list of applicable tarballs is returned via the caller's
1429 # variable "tarballs".
1432 local node=$(facet_active_host $SINGLEMDS)
1433 local r="do_node $node"
1435 if [ "$CLIENTONLY" ]; then
1436 skip "Client-only testing"
1440 if ! $r which $TUNEFS; then
1441 skip_env "tunefs.lustre required on $node"
1445 local IMGTYPE=$(facet_fstype $SINGLEMDS)
1447 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 -name \'disk*-$IMGTYPE.tar.bz2\')
1449 if [ -z "$tarballs" ]; then
1450 skip "No applicable tarballs found"
1455 t32_test_cleanup() {
1457 local fstype=$(facet_fstype $SINGLEMDS)
1460 if $shall_cleanup_lustre; then
1461 umount $tmp/mnt/lustre || rc=$?
1463 if $shall_cleanup_mdt; then
1464 $r umount -d $tmp/mnt/mdt || rc=$?
1466 if $shall_cleanup_mdt1; then
1467 $r umount -d $tmp/mnt/mdt1 || rc=$?
1469 if $shall_cleanup_ost; then
1470 $r umount -d $tmp/mnt/ost || rc=$?
1475 if [ $fstype == "zfs" ]; then
1476 $r $ZPOOL destroy t32fs-mdt1 || rc=$?
1477 $r $ZPOOL destroy t32fs-ost1 || rc=$?
1482 t32_bits_per_long() {
1484 # Yes, this is not meant to be perfect.
1494 t32_reload_modules() {
1496 local all_removed=false
1499 while ((i < 20)); do
1500 echo "Unloading modules on $node: Attempt $i"
1501 do_rpc_nodes $node $LUSTRE_RMMOD $(facet_fstype $SINGLEMDS) &&
1503 do_rpc_nodes $node check_mem_leak || return 1
1504 if $all_removed; then
1505 do_rpc_nodes $node load_modules
1511 echo "Unloading modules on $node: Given up"
1515 t32_wait_til_devices_gone() {
1521 echo wait for devices to go
1522 while ((i < 20)); do
1523 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1524 loops=$(do_rpc_nodes $node losetup -a | grep -c t32)
1525 ((devices == 0 && loops == 0)) && return 0
1529 echo "waiting for dev on $node: dev $devices loop $loops given up"
1530 do_rpc_nodes $node "losetup -a"
1531 do_rpc_nodes $node "$LCTL devices_list"
1535 t32_verify_quota() {
1539 local fstype=$(facet_fstype $SINGLEMDS)
1543 $LFS quota -u $T32_QID -v $mnt
1545 qval=$($LFS quota -v -u $T32_QID $mnt |
1546 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1547 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1549 [ $qval -eq $img_bspace ] || {
1550 echo "bspace, act:$qval, exp:$img_bspace"
1554 qval=$($LFS quota -v -u $T32_QID $mnt |
1555 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1556 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1558 [ $qval -eq $img_ispace ] || {
1559 echo "ispace, act:$qval, exp:$img_ispace"
1563 qval=$($LFS quota -v -u $T32_QID $mnt |
1564 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1565 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1567 [ $qval -eq $T32_BLIMIT ] || {
1568 echo "blimit, act:$qval, exp:$T32_BLIMIT"
1572 qval=$($LFS quota -v -u $T32_QID $mnt |
1573 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1574 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1576 [ $qval -eq $T32_ILIMIT ] || {
1577 echo "ilimit, act:$qval, exp:$T32_ILIMIT"
1581 do_node $node $LCTL conf_param $fsname.quota.mdt=ug
1582 cmd="$LCTL get_param -n osd-$fstype.$fsname-MDT0000"
1583 cmd=$cmd.quota_slave.enabled
1584 wait_update $node "$cmd" "ug" || {
1585 echo "Enable mdt quota failed"
1589 do_node $node $LCTL conf_param $fsname.quota.ost=ug
1590 cmd="$LCTL get_param -n osd-$fstype.$fsname-OST0000"
1591 cmd=$cmd.quota_slave.enabled
1592 wait_update $node "$cmd" "ug" || {
1593 echo "Enable ost quota failed"
1598 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1599 bs=1M count=$(($T32_BLIMIT / 1024)) oflag=sync && {
1600 echo "Write succeed, but expect -EDQUOT"
1603 rm -f $mnt/t32_qf_new
1605 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1607 echo "Create succeed, but expect -EDQUOT"
1610 unlinkmany $mnt/t32_qf_ $T32_ILIMIT
1618 local dne_upgrade=${dne_upgrade:-"no"}
1619 local ff_convert=${ff_convert:-"no"}
1620 local shall_cleanup_mdt=false
1621 local shall_cleanup_mdt1=false
1622 local shall_cleanup_ost=false
1623 local shall_cleanup_lustre=false
1624 local node=$(facet_active_host $SINGLEMDS)
1625 local r="do_node $node"
1626 local node2=$(facet_active_host mds2)
1634 local nid=$($r $LCTL list_nids | head -1)
1640 local fstype=$(facet_fstype $SINGLEMDS)
1641 local mdt_dev=$tmp/mdt
1642 local ost_dev=$tmp/ost
1644 trap 'trap - RETURN; t32_test_cleanup' RETURN
1646 mkdir -p $tmp/mnt/lustre || error "mkdir $tmp/mnt/lustre failed"
1647 $r mkdir -p $tmp/mnt/{mdt,ost}
1648 $r tar xjvf $tarball -S -C $tmp || {
1649 error_noexit "Unpacking the disk image tarball"
1652 img_commit=$($r cat $tmp/commit)
1653 img_kernel=$($r cat $tmp/kernel)
1654 img_arch=$($r cat $tmp/arch)
1655 img_bspace=$($r cat $tmp/bspace)
1656 img_ispace=$($r cat $tmp/ispace)
1657 echo "Upgrading from $(basename $tarball), created with:"
1658 echo " Commit: $img_commit"
1659 echo " Kernel: $img_kernel"
1660 echo " Arch: $img_arch"
1661 echo "OST version: $(get_lustre_version ost1)"
1663 # The conversion can be made only when both of the following
1664 # conditions are satisfied:
1665 # - ost device img version < 2.3.64
1666 # - ost server version >= 2.5
1667 [ $(version_code $img_commit) -ge $(version_code 2.3.64) -o \
1668 $(lustre_version_code ost1) -lt $(version_code 2.5.0) ] &&
1671 if [ $fstype == "zfs" ]; then
1673 $r $ZPOOL import -f -d $tmp t32fs-mdt1
1674 $r $ZPOOL import -f -d $tmp t32fs-ost1
1675 mdt_dev=t32fs-mdt1/mdt1
1676 ost_dev=t32fs-ost1/ost1
1677 wait_update_facet $SINGLEMDS "$ZPOOL list |
1678 awk '/^t32fs-mdt1/ { print \\\$1 }'" "t32fs-mdt1" || {
1679 error_noexit "import zfs pool failed"
1684 $r $LCTL set_param debug="$PTLDEBUG"
1686 $r $TUNEFS --dryrun $mdt_dev || {
1688 error_noexit "tunefs.lustre before mounting the MDT"
1691 if [ "$writeconf" ]; then
1693 if [ $fstype == "ldiskfs" ]; then
1695 $r $TUNEFS --quota $mdt_dev || {
1697 error_noexit "Enable mdt quota feature"
1702 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
1703 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
1704 { skip "LU-2200: Cannot run over Inifiniband w/o lctl replace_nids "
1705 "(Need MGS version at least 2.3.59)"; return 0; }
1707 local osthost=$(facet_active_host ost1)
1708 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
1711 if [ $fstype == "ldiskfs" ]; then
1714 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt
1715 $r $LCTL replace_nids $fsname-OST0000 $ostnid
1716 $r $LCTL replace_nids $fsname-MDT0000 $nid
1717 $r umount -d $tmp/mnt/mdt
1720 mopts=exclude=$fsname-OST0000
1721 if [ $fstype == "ldiskfs" ]; then
1726 t32_wait_til_devices_gone $node
1728 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
1730 error_noexit "Mounting the MDT"
1733 shall_cleanup_mdt=true
1735 if [ "$dne_upgrade" != "no" ]; then
1736 local fs2mdsdev=$(mdsdevname 1_2)
1737 local fs2mdsvdev=$(mdsvdevname 1_2)
1739 echo "mkfs new MDT on ${fs2mdsdev}...."
1740 if [ $(facet_fstype mds1) == ldiskfs ]; then
1741 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
1744 add fs2mds $(mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
1745 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
1746 error_noexit "Mkfs new MDT failed"
1750 $r $TUNEFS --dryrun $fs2mdsdev || {
1751 error_noexit "tunefs.lustre before mounting the MDT"
1755 echo "mount new MDT....$fs2mdsdev"
1756 $r mkdir -p $tmp/mnt/mdt1
1757 $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
1758 error_noexit "mount mdt1 failed"
1762 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1763 error_noexit "enable remote dir create failed"
1765 shall_cleanup_mdt1=true
1768 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
1769 error_noexit "Getting MDT UUID"
1772 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
1773 error_noexit "Unexpected MDT UUID: \"$uuid\""
1777 $r $TUNEFS --dryrun $ost_dev || {
1778 error_noexit "tunefs.lustre before mounting the OST"
1781 if [ "$writeconf" ]; then
1782 mopts=mgsnode=$nid,$writeconf
1783 if [ $fstype == "ldiskfs" ]; then
1785 $r $TUNEFS --quota $ost_dev || {
1787 error_noexit "Enable ost quota feature"
1793 if [ $fstype == "ldiskfs" ]; then
1797 $r $MOUNT_CMD -o $mopts $ost_dev $tmp/mnt/ost || {
1798 error_noexit "Mounting the OST"
1801 shall_cleanup_ost=true
1803 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
1804 error_noexit "Getting OST UUID"
1807 if [ "$uuid" != $fsname-OST0000_UUID ]; then
1808 error_noexit "Unexpected OST UUID: \"$uuid\""
1812 $r $LCTL conf_param $fsname-OST0000.osc.max_dirty_mb=15 || {
1813 error_noexit "Setting \"max_dirty_mb\""
1816 $r $LCTL conf_param $fsname-OST0000.failover.node=$nid || {
1817 error_noexit "Setting OST \"failover.node\""
1820 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
1821 error_noexit "Setting \"max_rpcs_in_flight\""
1824 $r $LCTL conf_param $fsname-MDT0000.failover.node=$nid || {
1825 error_noexit "Setting MDT \"failover.node\""
1828 $r $LCTL pool_new $fsname.interop || {
1829 error_noexit "Setting \"interop\""
1832 $r $LCTL conf_param $fsname-MDT0000.lov.stripesize=4M || {
1833 error_noexit "Setting \"lov.stripesize\""
1837 if [ "$ff_convert" != "no" -a $(facet_fstype ost1) == "ldiskfs" ]; then
1838 $r $LCTL lfsck_start -M $fsname-OST0000 || {
1839 error_noexit "Start OI scrub on OST0"
1843 # The oi_scrub should be on ost1, but for test_32(),
1844 # all on the SINGLEMDS.
1845 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
1846 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1847 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
1848 error_noexit "Failed to get the expected 'completed'"
1852 local UPDATED=$($r $LCTL get_param -n \
1853 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1854 awk '/^updated/ { print $2 }')
1855 [ $UPDATED -ge 1 ] || {
1856 error_noexit "Only $UPDATED objects have been converted"
1861 if [ "$dne_upgrade" != "no" ]; then
1862 $r $LCTL conf_param \
1863 $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
1864 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
1867 $r $LCTL conf_param $fsname-MDT0001.failover.node=$nid || {
1868 error_noexit "Setting MDT1 \"failover.node\""
1871 $r $LCTL conf_param $fsname-MDT0001.lov.stripesize=4M || {
1872 error_noexit "Setting MDT1 \"lov.stripesize\""
1878 if [ "$writeconf" ]; then
1879 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
1880 error_noexit "Mounting the client"
1883 shall_cleanup_lustre=true
1884 $LCTL set_param debug="$PTLDEBUG"
1886 t32_verify_quota $node $fsname $tmp/mnt/lustre || {
1887 error_noexit "verify quota failed"
1891 if [ "$dne_upgrade" != "no" ]; then
1892 $LFS mkdir -i 1 $tmp/mnt/lustre/remote_dir || {
1893 error_noexit "set remote dir failed"
1897 pushd $tmp/mnt/lustre
1898 tar -cf - . --exclude=./remote_dir |
1899 tar -xvf - -C remote_dir 1>/dev/null || {
1900 error_noexit "cp to remote dir failed"
1906 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
1907 error_noexit "dd failed"
1910 rm -rf $tmp/mnt/lustre/tmp_file || {
1911 error_noexit "rm failed"
1915 if $r test -f $tmp/sha1sums; then
1916 # LU-2393 - do both sorts on same node to ensure locale
1918 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
1919 if [ "$dne_upgrade" != "no" ]; then
1920 pushd $tmp/mnt/lustre/remote_dir
1922 pushd $tmp/mnt/lustre
1925 find ! -name .lustre -type f -exec sha1sum {} \; |
1926 sort -k 2 >$tmp/sha1sums || {
1927 error_noexit "sha1sum"
1931 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
1932 error_noexit "sha1sum verification failed"
1936 echo "sha1sum verification skipped"
1939 if [ "$dne_upgrade" != "no" ]; then
1940 rm -rf $tmp/mnt/lustre/remote_dir || {
1941 error_noexit "remove remote dir failed"
1946 if $r test -f $tmp/list; then
1948 # There is not a Test Framework API to copy files to or
1949 # from a remote node.
1951 # LU-2393 - do both sorts on same node to ensure locale
1953 $r cat $tmp/list | sort -k 6 >$tmp/list.orig
1954 pushd $tmp/mnt/lustre
1955 ls -Rni --time-style=+%s | sort -k 6 >$tmp/list || {
1961 # 32-bit and 64-bit clients use different algorithms to
1962 # convert FIDs into inode numbers. Hence, remove the inode
1963 # numbers from the lists, if the original list was created
1964 # on an architecture with different number of bits per
1967 if [ $(t32_bits_per_long $(uname -m)) != \
1968 $(t32_bits_per_long $img_arch) ]; then
1969 echo "Different number of bits per \"long\" from the disk image"
1970 for list in list.orig list; do
1971 sed -i -e 's/^[0-9]\+[ \t]\+//' $tmp/$list
1974 if ! diff -ub $tmp/list.orig $tmp/list; then
1975 error_noexit "list verification failed"
1979 echo "list verification skipped"
1983 # When adding new data verification tests, please check for
1984 # the presence of the required reference files first, like
1985 # the "sha1sums" and "list" tests above, to avoid the need to
1986 # regenerate every image for each test addition.
1989 nrpcs_orig=$($LCTL get_param \
1990 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
1991 error_noexit "Getting \"max_rpcs_in_flight\""
1994 nrpcs=$((nrpcs_orig + 5))
1995 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=$nrpcs || {
1996 error_noexit "Changing \"max_rpcs_in_flight\""
1999 wait_update $HOSTNAME "$LCTL get_param \
2000 -n mdc.*MDT0000*.max_rpcs_in_flight" $nrpcs || {
2001 error_noexit "Verifying \"max_rpcs_in_flight\""
2005 umount $tmp/mnt/lustre || {
2006 error_noexit "Unmounting the client"
2009 shall_cleanup_lustre=false
2011 if [ "$dne_upgrade" != "no" ]; then
2012 $r umount -d $tmp/mnt/mdt1 || {
2013 error_noexit "Unmounting the MDT2"
2016 shall_cleanup_mdt1=false
2019 $r umount -d $tmp/mnt/mdt || {
2020 error_noexit "Unmounting the MDT"
2023 shall_cleanup_mdt=false
2025 $r umount -d $tmp/mnt/ost || {
2026 error_noexit "Unmounting the OST"
2029 shall_cleanup_ost=false
2031 t32_reload_modules $node || {
2032 error_noexit "Reloading modules"
2036 # mount a second time to make sure we didnt leave upgrade flag on
2037 $r $TUNEFS --dryrun $mdt_dev || {
2039 error_noexit "tunefs.lustre before remounting the MDT"
2043 mopts=exclude=$fsname-OST0000
2044 if [ $fstype == "ldiskfs" ]; then
2047 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2048 error_noexit "Remounting the MDT"
2051 shall_cleanup_mdt=true
2061 for tarball in $tarballs; do
2062 t32_test $tarball || let "rc += $?"
2066 run_test 32a "Upgrade (not live)"
2074 for tarball in $tarballs; do
2075 t32_test $tarball writeconf || let "rc += $?"
2079 run_test 32b "Upgrade with writeconf"
2086 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
2088 for tarball in $tarballs; do
2089 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
2093 run_test 32c "dne upgrade test"
2101 for tarball in $tarballs; do
2102 ff_convert=yes t32_test $tarball || rc=$?
2106 run_test 32d "convert ff test"
2108 test_33a() { # bug 12333, was test_33
2109 local FSNAME2=test-123
2110 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2113 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
2115 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
2116 local dev=${SINGLEMDS}_dev
2117 local MDSDEV=${!dev}
2118 is_blkdev $SINGLEMDS $MDSDEV &&
2119 skip_env "mixed loopback and real device not working" &&
2123 local fs2mdsdev=$(mdsdevname 1_2)
2124 local fs2ostdev=$(ostdevname 1_2)
2125 local fs2mdsvdev=$(mdsvdevname 1_2)
2126 local fs2ostvdev=$(ostvdevname 1_2)
2128 if [ $(facet_fstype mds1) == ldiskfs ]; then
2129 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
2132 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2133 --reformat $mkfsoptions $fs2mdsdev $fs2mdsvdev || exit 10
2134 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2135 --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \
2136 $fs2ostvdev || exit 10
2138 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
2139 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2140 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" ||
2141 error "$LCTL conf_param $FSNAME2.sys.timeout=200 failed"
2142 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2143 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2146 cp /etc/hosts $MOUNT2/ || error "copy /etc/hosts $MOUNT2/ failed"
2147 $GETSTRIPE $MOUNT2/hosts || error "$GETSTRIPE $MOUNT2/hosts failed"
2152 cleanup_nocli || error "cleanup_nocli failed with $?"
2154 run_test 33a "Mount ost with a large index number"
2156 test_33b() { # was test_34
2159 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
2160 # Drop lock cancelation reply during umount
2161 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
2162 do_facet client $LCTL set_param fail_loc=0x80000304
2163 #lctl set_param debug=-1
2164 umount_client $MOUNT
2165 cleanup || error "cleanup failed with $?"
2167 run_test 33b "Drop cancel during umount"
2171 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
2172 manual_umount_client
2174 do_facet client killall -USR1 multiop
2175 if [ $rc -eq 0 ]; then
2176 error "umount not fail!"
2179 cleanup || error "cleanup failed with rc $?"
2181 run_test 34a "umount with opened file should be fail"
2185 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2186 stop_mds || error "Unable to stop MDS"
2188 manual_umount_client --force || error "mtab after failed umount with $?"
2190 cleanup || error "cleanup failed with $?"
2192 run_test 34b "force umount with failed mds should be normal"
2196 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2197 stop_ost || error "Unable to stop OST1"
2199 manual_umount_client --force || error "mtab after failed umount with $?"
2201 cleanup || error "cleanup failed with $?"
2203 run_test 34c "force umount with failed ost should be normal"
2205 test_35a() { # bug 12459
2208 DBG_SAVE="`$LCTL get_param -n debug`"
2209 $LCTL set_param debug="ha"
2211 log "Set up a fake failnode for the MDS"
2213 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2214 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2215 do_facet mgs "$LCTL conf_param \
2216 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" ||
2217 error "Setting ${device}.failover.node=\
2218 $(h2$NETTYPE $FAKENID) failed."
2220 log "Wait for RECONNECT_INTERVAL seconds (10s)"
2223 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
2226 log "Stopping the MDT: $device"
2227 stop_mdt 1 || error "MDT0 stop fail"
2229 df $MOUNT > /dev/null 2>&1 &
2231 log "Restarting the MDT: $device"
2232 start_mdt 1 || error "MDT0 start fail"
2233 log "Wait for df ($DFPID) ... "
2236 $LCTL set_param debug="$DBG_SAVE"
2238 # retrieve from the log the first server that the client tried to
2239 # contact after the connection loss
2240 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2241 NEXTCONN=`awk "/${MSG}/ {start = 1;}
2242 /import_select_connection.*$device-mdc.* using connection/ {
2244 if (\\\$NF ~ /$FAKENID/)
2250 }" $TMP/lustre-log-$TESTNAME.log`
2251 [ "$NEXTCONN" != "0" ] &&
2252 error "Tried to connect to ${NEXTCONN} not last active server"
2253 cleanup || error "cleanup failed with $?"
2254 # remove nid settings
2255 writeconf_or_reformat
2257 run_test 35a "Reconnect to the last active server first"
2259 test_35b() { # bug 18674
2260 remote_mds || { skip "local MDS" && return 0; }
2264 $LCTL set_param debug="ha"
2266 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
2269 log "Set up a fake failnode for the MDS"
2271 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2272 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2273 do_facet mgs "$LCTL conf_param \
2274 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" ||
2275 error "Set ${device}.failover.node=\
2276 $(h2$NETTYPE $FAKENID) failed"
2278 local at_max_saved=0
2279 # adaptive timeouts may prevent seeing the issue
2280 if at_is_enabled; then
2281 at_max_saved=$(at_max_get mds)
2282 at_max_set 0 mds client
2285 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
2287 log "Injecting EBUSY on MDS"
2288 # Setting OBD_FAIL_MDS_RESEND=0x136
2289 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" ||
2290 error "unable to set param fail_loc=0x80000136"
2292 $LCTL set_param mdc.${FSNAME}*.stats=clear
2294 log "Creating a test file and stat it"
2295 touch $MOUNT/$tdir/$tfile || error "touch $MOUNT/$tdir/$tfile failed"
2296 stat $MOUNT/$tdir/$tfile
2298 log "Stop injecting EBUSY on MDS"
2299 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" ||
2300 error "unable to set param fail_loc=0"
2301 rm -f $MOUNT/$tdir/$tfile || error "remove $MOUNT/$tdir/$tfile failed"
2304 # restore adaptive timeout
2305 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
2307 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2309 CONNCNT=$($LCTL get_param mdc.${FSNAME}*.stats |
2310 awk '/mds_connect/{print $2}')
2312 # retrieve from the log if the client has ever tried to
2313 # contact the fake server after the loss of connection
2314 FAILCONN=`awk "BEGIN {ret = 0;}
2315 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
2317 if (\\\$NF ~ /$FAKENID/) {
2322 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
2324 [ "$FAILCONN" == "0" ] &&
2325 error "The client reconnection has not been triggered"
2326 [ "$FAILCONN" == "2" ] &&
2327 error "Primary server busy, client reconnect to failover failed"
2330 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
2331 # Reconnects are supposed to be rate limited to one every 5s
2332 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] &&
2333 error "Too many reconnects $CONNCNT"
2335 cleanup || error "cleanup failed with $?"
2336 # remove nid settings
2337 writeconf_or_reformat
2339 run_test 35b "Continue reconnection retries, if the active server is busy"
2342 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
2344 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] ||
2345 { skip "remote OST" && return 0; }
2348 local FSNAME2=test1234
2349 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2351 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
2353 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
2354 is_blkdev $SINGLEMDS $MDSDEV &&
2355 skip_env "mixed loopback and real device not working" && return
2358 local fs2mdsdev=$(mdsdevname 1_2)
2359 local fs2ostdev=$(ostdevname 1_2)
2360 local fs3ostdev=$(ostdevname 2_2)
2361 local fs2mdsvdev=$(mdsvdevname 1_2)
2362 local fs2ostvdev=$(ostvdevname 1_2)
2363 local fs3ostvdev=$(ostvdevname 2_2)
2365 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2366 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
2367 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
2368 # different one than the default value here.
2369 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2370 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
2371 add fs3ost $(mkfs_opts ost2 ${fs3ostdev}) --mgsnode=$MGSNID \
2372 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
2374 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
2375 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2376 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
2377 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2378 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2380 sleep 5 # until 11778 fixed
2382 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || error "dd failed"
2384 BKTOTAL=$($LCTL get_param -n obdfilter.*.kbytestotal |
2385 awk 'BEGIN{total=0}; {total+=$1}; END{print total}')
2386 BKFREE=$($LCTL get_param -n obdfilter.*.kbytesfree |
2387 awk 'BEGIN{free=0}; {free+=$1}; END{print free}')
2388 BKAVAIL=$($LCTL get_param -n obdfilter.*.kbytesavail |
2389 awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}')
2390 STRING=$(df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}')
2391 DFTOTAL=$(echo $STRING | cut -d, -f1)
2392 DFUSED=$(echo $STRING | cut -d, -f2)
2393 DFAVAIL=$(echo $STRING | cut -d, -f3)
2394 DFFREE=$(($DFTOTAL - $DFUSED))
2396 ALLOWANCE=$((64 * $OSTCOUNT))
2398 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
2399 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
2400 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
2403 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
2404 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
2405 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
2408 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
2409 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
2410 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
2415 stop fs3ost -f || error "unable to stop OST3"
2416 stop fs2ost -f || error "unable to stop OST2"
2417 stop fs2mds -f || error "unable to stop second MDS"
2418 unload_modules_conf || error "unable unload modules"
2421 run_test 36 "df report consistency on OSTs with different block size"
2424 local mntpt=$(facet_mntpt $SINGLEMDS)
2425 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
2426 local mdsdev_sym="$TMP/sym_mdt.img"
2427 local opts=$MDS_MOUNT_OPTS
2430 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2431 skip "Currently only applicable to ldiskfs-based MDTs"
2435 echo "MDS : $mdsdev"
2436 echo "SYMLINK : $mdsdev_sym"
2437 do_facet $SINGLEMDS rm -f $mdsdev_sym
2439 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
2441 echo "mount symlink device - $mdsdev_sym"
2443 if ! do_facet $SINGLEMDS test -b $mdsdev; then
2444 opts=$(csa_add "$opts" -o loop)
2446 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
2447 $mdsdev_sym $mntpt 2>&1)
2450 echo mount_op=$mount_op
2452 do_facet $SINGLEMDS "umount -d $mntpt && rm -f $mdsdev_sym"
2454 if $(echo $mount_op | grep -q "unable to set tunable"); then
2455 error "set tunables failed for symlink device"
2458 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
2460 run_test 37 "verify set tunables works for symlink device"
2462 test_38() { # bug 14222
2463 local fstype=$(facet_fstype $SINGLEMDS)
2464 local mntpt=$(facet_mntpt $SINGLEMDS)
2469 local SRC="/etc /bin"
2470 local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT)
2471 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
2472 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2473 tar cf - $FILES | tar xf - -C $DIR/$tdir ||
2474 error "copying $SRC to $DIR/$tdir"
2476 umount_client $MOUNT || error "umount_client $MOUNT failed"
2477 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2478 stop_mds || error "Unable to stop MDS"
2479 log "delete lov_objid file on MDS"
2481 mount_fstype $SINGLEMDS || error "mount MDS failed (1)"
2483 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid; rm $mntpt/lov_objid"
2485 unmount_fstype $SINGLEMDS || error "umount failed (1)"
2487 # check create in mds_lov_connect
2488 start_mds || error "unable to start MDS"
2489 mount_client $MOUNT || error "mount_client $MOUNT failed"
2491 [ $V ] && log "verifying $DIR/$tdir/$f"
2492 diff -q $f $DIR/$tdir/$f || ERROR=y
2494 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2495 if [ "$ERROR" = "y" ]; then
2496 # check it's updates in sync
2497 umount_client $MOUNT
2499 mount_fstype $SIGNLEMDS
2500 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2501 unmount_fstype $SINGLEMDS
2502 error "old and new files are different after connect" || true
2504 touch $DIR/$tdir/f2 || error "f2 file create failed"
2506 # check it's updates in sync
2507 umount_client $MOUNT || error "second umount_client $MOUNT failed"
2510 mount_fstype $SINGLEMDS || error "mount MDS failed (3)"
2512 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2513 do_facet $SINGLEMDS dd if=/dev/zero of=$mntpt/lov_objid.clear count=8
2515 unmount_fstype $SINGLEMDS || error "umount failed (3)"
2517 start_mds || error "unable to start MDS"
2518 mount_client $MOUNT || error "mount_client $MOUNT failed"
2520 [ $V ] && log "verifying $DIR/$tdir/$f"
2521 diff -q $f $DIR/$tdir/$f || ERROR=y
2523 touch $DIR/$tdir/f3 || error "f3 file create failed"
2524 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2525 umount_client $MOUNT || error "third umount_client $MOUNT failed"
2527 mount_fstype $SINGLEMDS || error "mount MDS failed (4)"
2528 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2529 unmount_fstype $SINGLEMDS || error "umount failed (4)"
2531 [ "$ERROR" = "y" ] &&
2532 error "old and new files are different after sync" || true
2534 log "files compared the same"
2535 cleanup || error "cleanup failed with $?"
2537 run_test 38 "MDS recreates missing lov_objid file from OST data"
2542 cleanup || error "cleanup failed with $?"
2543 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
2544 error "memory leak detected" || true
2546 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
2548 test_40() { # bug 15759
2549 start_ost || error "Unable to start OST1"
2550 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
2551 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
2553 cleanup || error "cleanup failed with rc $?"
2555 run_test 40 "race during service thread startup"
2557 test_41a() { #bug 14134
2558 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2559 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2560 skip "Loop devices does not work with nosvc option"
2564 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2566 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
2567 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
2568 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
2569 mount_client $MOUNT || error "mount_client $MOUNT failed"
2572 echo "blah blah" > $MOUNT/$tfile
2575 umount_client $MOUNT || error "umount_client $MOUNT failed"
2576 stop ost1 -f || error "unable to stop OST1"
2577 stop_mds || error "Unable to stop MDS"
2578 stop_mds || error "Unable to stop MDS on second try"
2579 unload_modules_conf || error "unload_modules_conf failed"
2581 run_test 41a "mount mds with --nosvc and --nomgs"
2584 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2585 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2586 skip "Loop devices does not work with nosvc option"
2590 ! combined_mgs_mds && skip "needs combined mgs device" && return 0
2594 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2596 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
2597 start_ost || error "Unable to start OST1"
2598 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
2599 mount_client $MOUNT || error "mount_client $MOUNT failed"
2602 echo "blah blah" > $MOUNT/$tfile
2603 cat $MOUNT/$tfile || error "cat $MOUNT/$tfile failed"
2605 umount_client $MOUNT || error "umount_client $MOUNT failed"
2606 stop_ost || error "Unable to stop OST1"
2607 stop_mds || error "Unable to stop MDS"
2608 stop_mds || error "Unable to stop MDS on second try"
2610 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
2613 local server_version=$(lustre_version_code $SINGLEMDS)
2615 [[ $server_version -ge $(version_code 2.6.52) ]] ||
2616 [[ $server_version -ge $(version_code 2.5.26) &&
2617 $server_version -lt $(version_code 2.5.50) ]] ||
2618 [[ $server_version -ge $(version_code 2.5.4) &&
2619 $server_version -lt $(version_code 2.5.11) ]] ||
2620 { skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"; return; }
2623 # MDT concurent start
2624 #define OBD_FAIL_TGT_DELAY_CONNECT 0x703
2625 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x703"
2626 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS &
2629 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x0"
2630 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS &
2636 if [ $rc == 0 ] && [ $rc2 == 114 ]; then
2637 echo "1st MDT start succeed"
2638 echo "2nd MDT start failed with EALREADY"
2639 elif [ $rc2 == 0 ] && [ $rc == 114 ]; then
2640 echo "1st MDT start failed with EALREADY"
2641 echo "2nd MDT start succeed"
2644 error "unexpected concurent MDT mounts result, rc=$rc rc2=$rc2"
2647 # OST concurent start
2648 #define OBD_FAIL_TGT_DELAY_CONNECT 0x703
2649 do_facet ost1 "$LCTL set_param fail_loc=0x703"
2650 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &
2653 do_facet ost1 "$LCTL set_param fail_loc=0x0"
2654 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &
2660 if [ $rc == 0 ] && [ $rc2 == 114 ]; then
2661 echo "1st OST start succeed"
2662 echo "2nd OST start failed with EALREADY"
2663 elif [ $rc2 == 0 ] && [ $rc == 114 ]; then
2664 echo "1st OST start failed with EALREADY"
2665 echo "2nd OST start succeed"
2669 error "unexpected concurent OST mounts result, rc=$rc rc2=$rc2"
2675 # verify everything ok
2680 error "MDT(s) start failed"
2688 error "OST(s) start failed"
2696 error "client start failed"
2703 error "client mount failed"
2707 run_test 41c "concurent mounts of MDT/OST should all fail but one"
2709 test_42() { #bug 14693
2711 check_mount || error "client was not mounted"
2713 do_facet mgs $LCTL conf_param $FSNAME.llite.some_wrong_param=10
2714 umount_client $MOUNT ||
2715 error "unmounting client failed with invalid llite param"
2716 mount_client $MOUNT ||
2717 error "mounting client failed with invalid llite param"
2719 do_facet mgs $LCTL conf_param $FSNAME.sys.some_wrong_param=20
2720 cleanup || error "stopping $FSNAME failed with invalid sys param"
2723 check_mount || error "client was not mounted with invalid sys param"
2724 cleanup || error "stopping $FSNAME failed with invalid sys param"
2726 run_test 42 "allow client/server mount/unmount with invalid config param"
2729 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.58) ]] ||
2730 { skip "Need MDS version at least 2.5.58" && return 0; }
2731 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
2734 USER1=$(cat /etc/passwd | grep :$ID1:$ID1: | cut -d: -f1)
2735 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1" &&
2739 chmod ugo+x $DIR || error "chmod 0 failed"
2740 set_conf_param_and_check mds \
2741 "$LCTL get_param -n mdt.$FSNAME-MDT0000.root_squash" \
2742 "$FSNAME.mdt.root_squash" \
2744 wait_update $HOSTNAME \
2745 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
2747 error "check llite root_squash failed!"
2748 set_conf_param_and_check mds \
2749 "$LCTL get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
2750 "$FSNAME.mdt.nosquash_nids" \
2752 wait_update $HOSTNAME \
2753 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
2755 error "check llite nosquash_nids failed!"
2758 # create set of test files
2760 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
2761 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
2762 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
2764 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
2765 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
2767 mkdir $DIR/$tdir-rootdir || error "mkdir failed"
2768 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
2769 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
2771 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
2772 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
2773 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
2776 # check root_squash:
2777 # set root squash UID:GID to RUNAS_ID
2778 # root should be able to access only files owned by RUNAS_ID
2780 set_conf_param_and_check mds \
2781 "$LCTL get_param -n mdt.$FSNAME-MDT0000.root_squash" \
2782 "$FSNAME.mdt.root_squash" \
2783 "$RUNAS_ID:$RUNAS_ID"
2784 wait_update $HOSTNAME \
2785 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
2786 "$RUNAS_ID:$RUNAS_ID" ||
2787 error "check llite root_squash failed!"
2789 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
2790 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
2791 error "$ST: root read permission is denied"
2792 echo "$ST: root read permission is granted - ok"
2795 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
2796 error "$ST: root write permission is denied"
2797 echo "$ST: root write permission is granted - ok"
2799 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
2800 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
2801 error "$ST: root read permission is granted"
2802 echo "$ST: root read permission is denied - ok"
2805 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
2806 error "$ST: root write permission is granted"
2807 echo "$ST: root write permission is denied - ok"
2809 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
2810 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null &&
2811 error "$ST: root unlink permission is granted"
2812 echo "$ST: root unlink permission is denied - ok"
2814 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null &&
2815 error "$ST: root create permission is granted"
2816 echo "$ST: root create permission is denied - ok"
2819 # check root_squash is enforced independently
2820 # of client cache content
2822 # access file by USER1, keep access open
2823 # root should be denied access to user file
2825 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
2829 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
2830 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
2831 { kill $pid; error "$ST: root read permission is granted"; }
2832 echo "$ST: root read permission is denied - ok"
2835 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
2836 { kill $pid; error "$ST: root write permission is granted"; }
2837 echo "$ST: root write permission is denied - ok"
2843 # check nosquash_nids:
2844 # put client's NID into nosquash_nids list,
2845 # root should be able to access root file after that
2847 local NIDLIST=$($LCTL list_nids all | tr '\n' ' ')
2848 NIDLIST="2@elan $NIDLIST 192.168.0.[2,10]@tcp"
2849 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
2850 set_conf_param_and_check mds \
2851 "$LCTL get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
2852 "$FSNAME-MDTall.mdt.nosquash_nids" \
2854 wait_update $HOSTNAME \
2855 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
2857 error "check llite nosquash_nids failed!"
2859 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
2860 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
2861 error "$ST: root read permission is denied"
2862 echo "$ST: root read permission is granted - ok"
2865 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
2866 error "$ST: root write permission is denied"
2867 echo "$ST: root write permission is granted - ok"
2869 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
2870 rm $DIR/$tdir-rootdir/tfile-1 ||
2871 error "$ST: root unlink permission is denied"
2872 echo "$ST: root unlink permission is granted - ok"
2873 touch $DIR/$tdir-rootdir/tfile-2 ||
2874 error "$ST: root create permission is denied"
2875 echo "$ST: root create permission is granted - ok"
2877 run_test 43 "check root_squash and nosquash_nids"
2879 umount_client $MOUNT
2884 check_mount || error "check_mount"
2885 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
2887 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
2888 for VAL in $UUIDS; do
2889 NID=$(echo $VAL | cut -d= -f1)
2890 CLUUID=$(echo $VAL | cut -d= -f2)
2891 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
2893 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
2894 cleanup || error "cleanup failed with $?"
2896 run_test 44 "mounted client proc entry exists"
2900 check_mount || error "check_mount"
2901 stop_mds || error "Unable to stop MDS"
2905 #define OBD_FAIL_PTLRPC_LONG_UNLINK 0x50f
2906 do_facet client "$LCTL set_param fail_loc=0x50f"
2909 manual_umount_client --force || error "manual_umount_client failed"
2910 do_facet client "$LCTL set_param fail_loc=0x0"
2911 start_mds || error "unable to start MDS"
2912 mount_client $MOUNT || error "mount_client $MOUNT failed"
2913 cleanup || error "cleanup failed with $?"
2915 run_test 45 "long unlink handling in ptlrpcd"
2922 umount_client $MOUNT2 || rc=$?
2923 umount_client $MOUNT || rc=$?
2924 while [ $count -gt 0 ]; do
2925 stop ost${count} -f || rc=$?
2929 cleanup_nocli || rc=$?
2930 #writeconf to remove all ost2 traces for subsequent tests
2931 writeconf_or_reformat
2936 echo "Testing with $OSTCOUNT OSTs"
2938 start_mds || error "unable to start MDS"
2939 #first client should see only one ost
2940 start_ost || error "Unable to start OST1"
2941 wait_osc_import_state mds ost FULL
2943 mount_client $MOUNT || error "mount_client $MOUNT failed"
2944 trap "cleanup_46a $OSTCOUNT" EXIT ERR
2947 for (( i=2; i<=$OSTCOUNT; i++ )); do
2948 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
2949 error "start_ost$i $(ostdevname $i) failed"
2952 # wait until osts in sync
2953 for (( i=2; i<=$OSTCOUNT; i++ )); do
2954 wait_osc_import_state mds ost$i FULL
2955 wait_osc_import_state client ost$i FULL
2958 #second client see all ost's
2960 mount_client $MOUNT2 || error "mount_client failed"
2961 $SETSTRIPE -c -1 $MOUNT2 || error "$SETSTRIPE -c -1 $MOUNT2 failed"
2962 $GETSTRIPE $MOUNT2 || error "$GETSTRIPE $MOUNT2 failed"
2964 echo "ok" > $MOUNT2/widestripe
2965 $GETSTRIPE $MOUNT2/widestripe ||
2966 error "$GETSTRIPE $MOUNT2/widestripe failed"
2967 # fill acl buffer for avoid expand lsm to them
2968 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd |
2970 setfacl -m $acl $MOUNT2/widestripe
2974 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
2976 cleanup_46a $OSTCOUNT || error "cleanup_46a failed"
2978 run_test 46a "handle ost additional - wide striped file"
2983 check_mount || error "check_mount failed"
2984 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
2988 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
2989 if echo $ns | grep "MDT[[:digit:]]*"; then
2992 lrs=$(echo $ns | sed 's/.*lru_size=//')
2993 lru_size[count]=$lrs
2998 facet_failover $SINGLEMDS
2999 client_up || error "client_up failed"
3002 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3003 if echo $ns | grep "MDT[[:digit:]]*"; then
3006 lrs=$(echo $ns | sed 's/.*lru_size=//')
3007 if ! test "$lrs" -eq "${lru_size[count]}"; then
3008 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
3009 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
3014 cleanup || error "cleanup failed with $?"
3016 run_test 47 "server restart does not make client loss lru_resize settings"
3021 # reformat after this test is needed - if test will failed
3022 # we will have unkillable file at FS
3026 test_48() { # bug 17636
3029 check_mount || error "check_mount failed"
3031 $SETSTRIPE -c -1 $MOUNT || error "$SETSTRIPE -c -1 $MOUNT failed"
3032 $GETSTRIPE $MOUNT || error "$GETSTRIPE $MOUNT failed"
3034 echo "ok" > $MOUNT/widestripe
3035 $GETSTRIPE $MOUNT/widestripe ||
3036 error "$GETSTRIPE $MOUNT/widestripe failed"
3038 trap cleanup_48 EXIT ERR
3040 # fill acl buffer for avoid expand lsm to them
3041 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
3042 setfacl -m $acl $MOUNT/widestripe
3045 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3049 run_test 48 "too many acls on file"
3051 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
3052 test_49a() { # bug 17710
3053 local timeout_orig=$TIMEOUT
3054 local ldlm_timeout_orig=$LDLM_TIMEOUT
3055 local LOCAL_TIMEOUT=20
3057 LDLM_TIMEOUT=$LOCAL_TIMEOUT
3058 TIMEOUT=$LOCAL_TIMEOUT
3062 check_mount || error "client mount failed"
3064 echo "check ldlm_timout..."
3065 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3066 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3067 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3069 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3070 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3073 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
3074 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
3077 umount_client $MOUNT || error "umount_client $MOUNT failed"
3078 stop_ost || error "problem stopping OSS"
3079 stop_mds || error "problem stopping MDS"
3081 LDLM_TIMEOUT=$ldlm_timeout_orig
3082 TIMEOUT=$timeout_orig
3084 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3086 test_49b() { # bug 17710
3087 local timeout_orig=$TIMEOUT
3088 local ldlm_timeout_orig=$LDLM_TIMEOUT
3089 local LOCAL_TIMEOUT=20
3091 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
3092 TIMEOUT=$LOCAL_TIMEOUT
3096 check_mount || error "client mount failed"
3098 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3099 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3100 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3102 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3103 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3106 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
3107 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
3110 cleanup || error "cleanup failed"
3112 LDLM_TIMEOUT=$ldlm_timeout_orig
3113 TIMEOUT=$timeout_orig
3115 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3118 # Test both statfs and lfs df and fail if either one fails
3119 multiop_bg_pause $1 f_
3122 killall -USR1 multiop
3123 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
3124 wait $PID || { RC1=$?; log "multiop return error "; }
3131 if [ $RC2 -eq 0 ]; then
3133 log "lazystatfs df failed"
3137 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
3143 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3144 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3146 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
3148 cleanup || error "cleanup failed with rc $?"
3150 run_test 50a "lazystatfs all servers available"
3154 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3155 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3157 # Wait for client to detect down OST
3158 stop_ost || error "Unable to stop OST1"
3159 wait_osc_import_state mds ost DISCONN
3161 lazystatfs $MOUNT || error "lazystatfs should not return EIO"
3163 umount_client $MOUNT || error "Unable to unmount client"
3164 stop_mds || error "Unable to stop MDS"
3166 run_test 50b "lazystatfs all servers down"
3169 start_mds || error "Unable to start MDS"
3170 start_ost || error "Unable to start OST1"
3171 start_ost2 || error "Unable to start OST2"
3172 mount_client $MOUNT || error "Unable to mount client"
3173 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3174 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3176 # Wait for client to detect down OST
3177 stop_ost || error "Unable to stop OST1"
3178 wait_osc_import_state mds ost DISCONN
3179 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3181 umount_client $MOUNT || error "Unable to unmount client"
3182 stop_ost2 || error "Unable to stop OST2"
3183 stop_mds || error "Unable to stop MDS"
3184 #writeconf to remove all ost2 traces for subsequent tests
3185 writeconf_or_reformat
3187 run_test 50c "lazystatfs one server down"
3190 start_mds || error "Unable to start MDS"
3191 start_ost || error "Unable to start OST1"
3192 start_ost2 || error "Unable to start OST2"
3193 mount_client $MOUNT || error "Unable to mount client"
3194 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3195 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3197 # Issue the statfs during the window where the client still
3198 # belives the OST to be available but it is in fact down.
3199 # No failure just a statfs which hangs for a timeout interval.
3200 stop_ost || error "Unable to stop OST1"
3201 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3203 umount_client $MOUNT || error "Unable to unmount client"
3204 stop_ost2 || error "Unable to stop OST2"
3205 stop_mds || error "Unable to stop MDS"
3206 #writeconf to remove all ost2 traces for subsequent tests
3207 writeconf_or_reformat
3209 run_test 50d "lazystatfs client/server conn race"
3216 start_mds || error "Unable to start MDS"
3217 #first client should see only one ost
3218 start_ost || error "Unable to start OST1"
3219 wait_osc_import_state mds ost FULL
3221 # Wait for client to detect down OST
3222 stop_ost || error "Unable to stop OST1"
3223 wait_osc_import_state mds ost DISCONN
3225 mount_client $MOUNT || error "Unable to mount client"
3226 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3228 multiop_bg_pause $MOUNT _f
3232 if [ $RC1 -ne 0 ]; then
3233 log "multiop failed $RC1"
3236 sleep $(( $TIMEOUT+1 ))
3238 [ $? -ne 0 ] && error "process isn't sleep"
3239 start_ost || error "Unable to start OST1"
3240 wait $pid || error "statfs failed"
3243 umount_client $MOUNT || error "Unable to unmount client"
3244 stop_ost || error "Unable to stop OST1"
3245 stop_mds || error "Unable to stop MDS"
3247 run_test 50e "normal statfs all servers down"
3252 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
3254 start_mds || error "Unable to start mds"
3255 #first client should see only one ost
3256 start_ost || error "Unable to start OST1"
3257 wait_osc_import_state mds ost FULL
3259 start_ost2 || error "Unable to start OST2"
3260 wait_osc_import_state mds ost2 FULL
3262 # Wait for client to detect down OST
3263 stop_ost2 || error "Unable to stop OST2"
3265 wait_osc_import_state mds ost2 DISCONN
3266 mount_client $MOUNT || error "Unable to mount client"
3267 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3269 multiop_bg_pause $MOUNT _f
3273 if [ $RC1 -ne 0 ]; then
3274 log "lazystatfs multiop failed $RC1"
3277 sleep $(( $TIMEOUT+1 ))
3279 [ $? -ne 0 ] && error "process isn't sleep"
3280 start_ost2 || error "Unable to start OST2"
3281 wait $pid || error "statfs failed"
3282 stop_ost2 || error "Unable to stop OST2"
3285 umount_client $MOUNT || error "Unable to unmount client"
3286 stop_ost || error "Unable to stop OST1"
3287 stop_mds || error "Unable to stop MDS"
3288 #writeconf to remove all ost2 traces for subsequent tests
3289 writeconf_or_reformat
3291 run_test 50f "normal statfs one server in down"
3294 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3296 start_ost2 || error "Unable to start OST2"
3297 wait_osc_import_state mds ost2 FULL
3298 wait_osc_import_state client ost2 FULL
3300 local PARAM="${FSNAME}-OST0001.osc.active"
3302 $SETSTRIPE -c -1 $DIR/$tfile || error "$SETSTRIPE failed"
3303 do_facet mgs $LCTL conf_param $PARAM=0 ||
3304 error "Unable to deactivate OST"
3306 umount_client $MOUNT || error "Unable to unmount client"
3307 mount_client $MOUNT || error "Unable to mount client"
3308 # This df should not cause a panic
3311 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
3312 rm -f $DIR/$tfile || error "unable to remove file $DIR/$tfile"
3313 umount_client $MOUNT || error "Unable to unmount client"
3314 stop_ost2 || error "Unable to stop OST2"
3315 stop_ost || error "Unable to stop OST1"
3316 stop_mds || error "Unable to stop MDS"
3317 #writeconf to remove all ost2 traces for subsequent tests
3318 writeconf_or_reformat
3320 run_test 50g "deactivated OST should not cause panic"
3324 # prepare MDT/OST, make OSC inactive for OST1
3325 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3327 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3328 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
3329 error "tunefs OST1 failed"
3330 start_mds || error "Unable to start MDT"
3331 start_ost || error "Unable to start OST1"
3332 start_ost2 || error "Unable to start OST2"
3333 mount_client $MOUNT || error "client start failed"
3335 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3337 # activatate OSC for OST1
3338 local TEST="$LCTL get_param -n osc.${FSNAME}-OST0000-osc-[!M]*.active"
3339 set_conf_param_and_check client \
3340 "$TEST" "${FSNAME}-OST0000.osc.active" 1 ||
3341 error "Unable to activate OST1"
3343 mkdir $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3344 $SETSTRIPE -c -1 -i 0 $DIR/$tdir/2 ||
3345 error "$SETSTRIPE $DIR/$tdir/2 failed"
3346 sleep 1 && echo "create a file after OST1 is activated"
3348 createmany -o $DIR/$tdir/2/$tfile-%d 1
3350 # check OSC import is working
3351 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
3352 error "some OSC imports are still not connected"
3355 umount_client $MOUNT || error "Unable to umount client"
3356 stop_ost2 || error "Unable to stop OST2"
3357 cleanup_nocli || error "cleanup_nocli failed with $?"
3359 run_test 50h "LU-642: activate deactivated OST"
3362 local LOCAL_TIMEOUT=20
3366 check_mount || error "check_mount failed"
3368 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
3369 $SETSTRIPE -c -1 $MOUNT/$tdir ||
3370 error "$SETSTRIPE -c -1 $MOUNT/$tdir failed"
3371 #define OBD_FAIL_MDS_REINT_DELAY 0x142
3372 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x142"
3373 touch $MOUNT/$tdir/$tfile &
3376 start_ost2 || error "Unable to start OST1"
3378 stop_ost2 || error "Unable to stop OST1"
3379 cleanup || error "cleanup failed with $?"
3380 #writeconf to remove all ost2 traces for subsequent tests
3381 writeconf_or_reformat
3383 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
3392 do_node $node mkdir -p $dest
3393 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3395 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
3396 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3397 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
3399 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
3400 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
3410 local backup2=${TMP}/backup2
3412 do_node $node mkdir -p $backup2
3413 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3415 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
3416 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3417 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
3419 do_node $node "diff -rq $backup $backup2"
3420 [ $? -eq 0 ] || { error "contents differ"; return 3; }
3422 local xattrs2=${TMP}/xattrs2
3423 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
3424 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
3426 do_node $node "diff $xattrs $xattrs2"
3427 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
3429 do_node $node "rm -rf $backup2 $xattrs2"
3430 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
3434 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3435 skip "Only applicable to ldiskfs-based MDTs"
3439 start_mds || error "Unable to start MDS"
3440 start_ost || error "Unable to start OST1"
3441 mount_client $MOUNT || error "Unable to mount client"
3444 local ost1mnt=$(facet_mntpt ost1)
3445 local ost1node=$(facet_active_host ost1)
3446 local ost1tmp=$TMP/conf52
3449 mkdir $DIR/$tdir || error "Unable to create $DIR/$tdir"
3450 touch $TMP/modified_first || error "Unable to create temporary file"
3451 local mtime=$(stat -c %Y $TMP/modified_first)
3452 do_node $ost1node "mkdir -p $ost1tmp &&
3453 touch -m -d @$mtime $ost1tmp/modified_first" ||
3454 error "Unable to create temporary file"
3457 $SETSTRIPE -c -1 -S 1M $DIR/$tdir || error "$SETSTRIPE failed"
3459 for (( i=0; i < nrfiles; i++ )); do
3460 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c ||
3461 error "multiop failed"
3467 echo backup files to $TMP/files
3468 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
3469 copy_files_xattrs $(hostname) $TMP/files $TMP/file_xattrs $files ||
3470 error "Unable to copy files"
3472 umount_client $MOUNT || error "Unable to umount client"
3473 stop_ost || error "Unable to stop ost1"
3475 echo mount ost1 as ldiskfs
3476 do_node $ost1node mkdir -p $ost1mnt || error "Unable to create $ost1mnt"
3477 if ! do_node $ost1node test -b $ost1_dev; then
3480 do_node $ost1node mount -t $(facet_fstype ost1) $loop $ost1_dev \
3482 error "Unable to mount ost1 as ldiskfs"
3485 echo backup objects to $ost1tmp/objects
3486 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
3487 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
3488 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
3490 error "Unable to copy objects"
3492 # move objects to lost+found
3493 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
3494 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
3496 # recover objects dry-run
3497 echo "ll_recover_lost_found_objs dry_run"
3498 do_node $ost1node "ll_recover_lost_found_objs -n -d $ost1mnt/O" ||
3499 error "ll_recover_lost_found_objs failed"
3502 echo "ll_recover_lost_found_objs fix run"
3503 do_node $ost1node "ll_recover_lost_found_objs -d $ost1mnt/lost+found" ||
3504 error "ll_recover_lost_found_objs failed"
3506 # compare restored objects against saved ones
3507 diff_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
3508 [ $? -eq 0 ] || error "Unable to diff objects"
3510 do_node $ost1node "umount $ost1mnt" ||
3511 error "Unable to umount ost1 as ldiskfs"
3513 start_ost || error "Unable to start OST1"
3514 mount_client $MOUNT || error "Unable to mount client"
3517 diff_files_xattrs $(hostname) $TMP/files $TMP/file_xattrs $files ||
3518 error "Unable to diff files"
3520 rm -rf $TMP/files $TMP/file_xattrs ||
3521 error "Unable to delete temporary files"
3522 do_node $ost1node "rm -rf $ost1tmp" ||
3523 error "Unable to delete temporary files"
3524 cleanup || error "cleanup failed with $?"
3526 run_test 52 "check recovering objects from lost+found"
3528 # Checks threads_min/max/started for some service
3530 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
3531 # parameter pattern prefix like 'ost.*.ost'.
3544 local msg="Insane $modname thread counts"
3545 local ncpts=$(check_cpt_number $facet)
3549 check_mount || return 41
3551 # We need to expand $parampat, but it may match multiple parameters, so
3552 # we'll pick the first one
3553 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
3554 error "Couldn't expand ${parampat}.threads_min parameter name"
3558 # Remove the .threads_min part
3559 paramp=${paramp%.threads_min}
3561 # Check for sanity in defaults
3562 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3564 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3566 tstarted=$(do_facet $facet "$LCTL get_param \
3567 -n ${paramp}.threads_started" || echo 0)
3568 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' ||
3570 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
3572 nthrs=$(expr $tmax - $tmin)
3573 if [ $nthrs -lt $ncpts ]; then
3579 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
3580 skip_env "module parameter forced $facet thread count" &&
3581 tmin=3 && tmax=$((3 * tmax))
3583 # Check that we can change min/max
3584 do_facet $facet "$LCTL set_param \
3585 ${paramp}.threads_min=$((tmin + nthrs))"
3586 do_facet $facet "$LCTL set_param \
3587 ${paramp}.threads_max=$((tmax - nthrs))"
3588 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3590 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3592 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) &&
3593 $tmax2 == ($tmax - $nthrs)))' || return $?
3595 # Check that we can set min/max to the same value
3596 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3598 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$tmin"
3599 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3601 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3603 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
3605 # Check that we can't set max < min
3606 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$((tmin - 1))"
3607 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3609 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3611 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
3613 # We need to ensure that we get the module options desired; to do this
3614 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
3615 LOAD_MODULES_REMOTE=true
3618 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
3619 setmodopts -a $modname "$newvalue" oldvalue
3623 check_mount || return 41
3625 # Restore previous setting of MODOPTS_*
3626 setmodopts $modname "$oldvalue"
3628 # Check that $opts took
3629 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min")
3630 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max")
3631 tstarted=$(do_facet $facet \
3632 "$LCTL get_param -n ${paramp}.threads_started")
3633 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
3643 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
3644 cleanup || error "cleanup failed with rc $?"
3646 run_test 53a "check OSS thread count params"
3650 local mds=$(do_facet $SINGLEMDS "$LCTL get_param \
3651 -N mds.*.*.threads_max 2>/dev/null")
3652 if [ -z "$mds" ]; then
3653 #running this on an old MDT
3654 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
3656 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
3658 cleanup || error "cleanup failed with $?"
3660 run_test 53b "check MDS thread count params"
3663 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3664 skip "Only applicable to ldiskfs-based MDTs"
3668 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p ||
3669 error "llverdev failed with rc=$?"
3672 run_test 54a "test llverdev and partial verify of device"
3675 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3676 skip "Only applicable to ldiskfs-based MDTs"
3681 run_llverfs $MOUNT -p || error "llverfs failed with rc=$?"
3682 cleanup || error "cleanup failed with rc=$?"
3684 run_test 54b "test llverfs and partial verify of filesystem"
3688 local max_ost_index=$1
3689 echo -n $(((max_ost_index + 1) * 8))
3693 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3694 skip "Only applicable to ldiskfs-based MDTs"
3698 local mdsdev=$(mdsdevname 1)
3699 local mdsvdev=$(mdsvdevname 1)
3703 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
3705 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
3706 --reformat $(ostdevname 1) $(ostvdevname 1)
3712 echo checking size of lov_objid for ost index $i
3713 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
3714 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
3715 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
3717 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
3724 run_test 55 "check lov_objid size"
3727 local server_version=$(lustre_version_code $SINGLEMDS)
3728 local mds_journal_size_orig=$MDSJOURNALSIZE
3733 for num in $(seq 1 $MDSCOUNT); do
3734 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
3735 --reformat $(mdsdevname $num) $(mdsvdevname $num)
3737 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
3738 $(ostdevname 1) $(ostvdevname 1)
3739 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
3740 $(ostdevname 2) $(ostvdevname 2)
3743 start_ost || error "Unable to start first ost (idx 10000)"
3744 start_ost2 || error "Unable to start second ost (idx 1000)"
3745 mount_client $MOUNT || error "Unable to mount client"
3749 if [[ $server_version -ge $(version_code 2.6.54) ]] ||
3750 [[ $server_version -ge $(version_code 2.5.4) &&
3751 $server_version -lt $(version_code 2.5.11) ]]; then
3752 wait_osc_import_state mds ost1 FULL
3753 wait_osc_import_state mds ost2 FULL
3754 $SETSTRIPE --stripe-count=-1 $DIR/$tfile ||
3755 error "Unable to setstripe $DIR/$tfile"
3756 n=$($LFS getstripe --stripe-count $DIR/$tfile)
3757 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
3762 MDSJOURNALSIZE=$mds_journal_size_orig
3765 run_test 56 "check big OST indexes and out-of-index-order start"
3767 test_57a() { # bug 22656
3768 do_rpc_nodes $(facet_active_host ost1) load_modules_local
3769 local NID=$(do_facet ost1 "$LCTL get_param nis" |
3770 tail -1 | awk '{print $1}')
3771 writeconf_or_reformat
3772 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3773 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" ||
3774 error "tunefs failed"
3776 start_ost && error "OST registration from failnode should fail"
3779 run_test 57a "initial registration from failnode should fail (should return errs)"
3782 do_rpc_nodes $(facet_active_host ost1) load_modules_local
3783 local NID=$(do_facet ost1 "$LCTL get_param nis" |
3784 tail -1 | awk '{print $1}')
3785 writeconf_or_reformat
3786 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3787 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" ||
3788 error "tunefs failed"
3790 start_ost || error "OST registration from servicenode should not fail"
3793 run_test 57b "initial registration from servicenode should not fail"
3796 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
3799 test_58() { # bug 22658
3801 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3802 createmany -o $DIR/$tdir/$tfile-%d 100
3803 # make sure that OSTs do not cancel llog cookies before we unmount the MDS
3804 #define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601
3805 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x601"
3806 unlinkmany $DIR/$tdir/$tfile-%d 100
3807 stop_mds || error "Unable to stop MDS"
3809 local MNTDIR=$(facet_mntpt $SINGLEMDS)
3810 local devname=$(mdsdevname ${SINGLEMDS//mds/})
3812 # remove all files from the OBJECTS dir
3813 mount_fstype $SINGLEMDS
3815 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
3817 unmount_fstype $SINGLEMDS
3818 # restart MDS with missing llog files
3819 start_mds || error "unable to start MDS"
3820 do_facet mds "$LCTL set_param fail_loc=0"
3823 run_test 58 "missing llog files must not prevent MDT from mounting"
3826 start_mgsmds >> /dev/null
3827 local C1=$(count_osts)
3828 if [ $C1 -eq 0 ]; then
3829 start_ost >> /dev/null
3833 echo "original ost count: $C1 (expect > 0)"
3834 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
3835 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
3836 local C2=$(count_osts)
3837 echo "after mdt writeconf count: $C2 (expect 0)"
3838 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
3839 echo "OST start without writeconf should fail:"
3840 start_ost >> /dev/null &&
3841 error "OST start without writeconf didn't fail"
3842 echo "OST start with writeconf should succeed:"
3843 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
3844 local C3=$(count_osts)
3845 echo "after ost writeconf count: $C3 (expect 1)"
3846 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
3847 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
3848 local C4=$(count_osts)
3849 echo "after ost2 writeconf count: $C4 (expect 2)"
3850 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
3851 stop_ost2 >> /dev/null
3852 cleanup_nocli >> /dev/null
3853 #writeconf to remove all ost2 traces for subsequent tests
3854 writeconf_or_reformat
3856 run_test 59 "writeconf mount option"
3858 test_60() { # LU-471
3861 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3862 skip "Only applicable to ldiskfs-based MDTs"
3866 for num in $(seq $MDSCOUNT); do
3867 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
3868 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
3869 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
3873 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
3874 [ ${PIPESTATUS[0]} -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
3876 # MDT default has dirdata feature
3877 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
3878 # we disable uninit_bg feature
3879 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
3880 # we set stride extended options
3881 echo $dump | grep stride > /dev/null || error "stride is not set"
3884 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
3887 local reformat=false
3889 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] ||
3890 { skip "Need MDS version at least 2.1.53"; return 0; }
3892 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
3893 ! large_xattr_enabled; then
3895 LDISKFS_MKFS_OPTS+=" -O large_xattr"
3897 for num in $(seq $MDSCOUNT); do
3898 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
3899 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
3900 error "add mds $num failed"
3904 setup_noconfig || error "setting up the filesystem failed"
3905 client_up || error "starting client failed"
3907 local file=$DIR/$tfile
3908 touch $file || error "touch $file failed"
3910 local large_value="$(generate_string $(max_xattr_size))"
3911 local small_value="bar"
3913 local name="trusted.big"
3914 log "save large xattr $name on $file"
3915 setfattr -n $name -v $large_value $file ||
3916 error "saving $name on $file failed"
3918 local new_value=$(get_xattr_value $name $file)
3919 [[ "$new_value" != "$large_value" ]] &&
3920 error "$name different after saving"
3922 log "shrink value of $name on $file"
3923 setfattr -n $name -v $small_value $file ||
3924 error "shrinking value of $name on $file failed"
3926 new_value=$(get_xattr_value $name $file)
3927 [[ "$new_value" != "$small_value" ]] &&
3928 error "$name different after shrinking"
3930 log "grow value of $name on $file"
3931 setfattr -n $name -v $large_value $file ||
3932 error "growing value of $name on $file failed"
3934 new_value=$(get_xattr_value $name $file)
3935 [[ "$new_value" != "$large_value" ]] &&
3936 error "$name different after growing"
3938 log "check value of $name on $file after remounting MDS"
3940 new_value=$(get_xattr_value $name $file)
3941 [[ "$new_value" != "$large_value" ]] &&
3942 error "$name different after remounting MDS"
3944 log "remove large xattr $name from $file"
3945 setfattr -x $name $file || error "removing $name from $file failed"
3950 LDISKFS_MKFS_OPTS=${LDISKFS_MKFS_OPTS% -O large_xattr}
3954 run_test 61 "large xattr"
3957 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3958 skip "Only applicable to ldiskfs-based MDTs"
3963 local mdsdev=$(mdsdevname 1)
3964 local ostdev=$(ostdevname 1)
3966 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] ||
3967 { skip "Need MDS version at least 2.2.51"; return 0; }
3969 echo "disable journal for mds"
3970 do_facet mds tune2fs -O ^has_journal $mdsdev || error "tune2fs failed"
3971 start_mds && error "MDT start should fail"
3972 echo "disable journal for ost"
3973 do_facet ost1 tune2fs -O ^has_journal $ostdev || error "tune2fs failed"
3974 start_ost && error "OST start should fail"
3975 cleanup || error "cleanup failed with rc $?"
3978 run_test 62 "start with disabled journal"
3981 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3982 skip "Only applicable to ldiskfs-based MDTs"
3986 local inode_slab=$(do_facet $SINGLEMDS \
3987 "awk '/ldiskfs_inode_cache/ { print \\\$5 }' /proc/slabinfo")
3988 if [ -z "$inode_slab" ]; then
3989 skip "ldiskfs module has not been loaded"
3993 echo "$inode_slab ldisk inodes per page"
3994 [ "$inode_slab" -ge "3" ] ||
3995 error "ldisk inode size is too big, $inode_slab objs per page"
3998 run_test 63 "Verify each page can at least hold 3 ldisk inodes"
4001 start_mds || error "unable to start MDS"
4002 start_ost || error "Unable to start OST1"
4003 start_ost2 || error "Unable to start second ost"
4004 mount_client $MOUNT || error "Unable to mount client"
4005 stop_ost2 || error "Unable to stop second ost"
4007 $LFS df --lazy || error "lfs df failed"
4008 cleanup || error "cleanup failed with $?"
4009 #writeconf to remove all ost2 traces for subsequent tests
4010 writeconf_or_reformat
4012 run_test 64 "check lfs df --lazy "
4014 test_65() { # LU-2237
4015 # Currently, the test is only valid for ldiskfs backend
4016 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4017 skip "non-ldiskfs backend" && return
4019 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4020 local brpt=$(facet_mntpt brpt)
4023 if ! do_facet $SINGLEMDS "test -b $devname"; then
4027 stop_mds || error "Unable to stop MDS"
4028 local obj=$(do_facet $SINGLEMDS \
4029 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
4031 if [ -z "$obj" ]; then
4032 # The MDT may be just re-formatted, mount the MDT for the
4033 # first time to guarantee the "last_rcvd" file is there.
4034 start_mds || error "fail to mount the MDS for the first time"
4035 stop_mds || error "Unable to stop MDS"
4038 # remove the "last_rcvd" file
4039 do_facet $SINGLEMDS "mkdir -p $brpt"
4040 do_facet $SINGLEMDS \
4041 "mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
4042 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
4043 do_facet $SINGLEMDS "umount -d $brpt"
4045 # restart MDS, the "last_rcvd" file should be recreated.
4046 start_mds || error "fail to restart the MDS"
4047 stop_mds || error "Unable to stop MDS"
4048 obj=$(do_facet $SINGLEMDS \
4049 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
4050 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
4052 run_test 65 "re-create the lost last_rcvd file when server mount"
4055 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
4056 { skip "Need MGS version at least 2.3.59"; return 0; }
4059 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
4060 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
4062 echo "replace_nids should fail if MDS, OSTs and clients are UP"
4063 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4064 error "replace_nids fail"
4066 umount_client $MOUNT || error "unmounting client failed"
4067 echo "replace_nids should fail if MDS and OSTs are UP"
4068 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4069 error "replace_nids fail"
4071 stop_ost || error "Unable to stop OST1"
4072 echo "replace_nids should fail if MDS is UP"
4073 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4074 error "replace_nids fail"
4076 stop_mds || error "stopping mds failed"
4078 if combined_mgs_mds; then
4079 start_mdt 1 "-o nosvc" ||
4080 error "starting mds with nosvc option failed"
4083 echo "command should accept two parameters"
4084 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
4085 error "command should accept two params"
4087 echo "correct device name should be passed"
4088 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
4089 error "wrong devname"
4091 echo "wrong nids list should not destroy the system"
4092 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
4095 echo "replace OST nid"
4096 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
4097 error "replace nids failed"
4099 echo "command should accept two parameters"
4100 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
4101 error "command should accept two params"
4103 echo "wrong nids list should not destroy the system"
4104 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
4107 echo "replace MDS nid"
4108 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
4109 error "replace nids failed"
4111 if ! combined_mgs_mds ; then
4114 stop_mds || error "Unable to stop MDS"
4118 check_mount || error "error after nid replace"
4119 cleanup || error "cleanup failed"
4122 run_test 66 "replace nids"
4124 test_67() { #LU-2950
4125 local legacy="$TMP/legacy_lnet_config"
4126 local new="$TMP/new_routes_test"
4127 local out="$TMP/config_out_file"
4128 local verify="$TMP/conv_verify"
4129 local verify_conf="$TMP/conf_verify"
4131 # Create the legacy file that will be run through the
4132 # lustre_routes_conversion script
4133 cat <<- LEGACY_LNET_CONFIG > $legacy
4134 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
4135 tcp2 54 10.1.3.2@tcp;
4136 tcp3 10.3.4.3@tcp:3;
4140 # Create the verification file to verify the output of
4141 # lustre_routes_conversion script against.
4142 cat <<- VERIFY_LNET_CONFIG > $verify
4143 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
4144 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
4145 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
4146 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
4147 tcp4: { gateway: 10.3.3.4@tcp }
4150 # Create the verification file to verify the output of
4151 # lustre_routes_config script against
4152 cat <<- VERIFY_LNET_CONFIG > $verify_conf
4153 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
4154 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
4155 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
4156 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
4157 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
4160 lustre_routes_conversion $legacy $new > /dev/null
4161 if [ -f $new ]; then
4162 # verify the conversion output
4163 cmp -s $new $verify > /dev/null
4164 if [ $? -eq 1 ]; then
4165 error "routes conversion failed"
4168 lustre_routes_config --dry-run --verbose $new > $out
4169 # check that the script succeeded
4170 cmp -s $out $verify_conf > /dev/null
4171 if [ $? -eq 1 ]; then
4172 error "routes config failed"
4175 error "routes conversion test failed"
4177 # remove generated files
4178 rm -f $new $legacy $verify $verify_conf $out
4180 run_test 67 "test routes conversion and configuration"
4188 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.4.53) ] ||
4189 { skip "Need MDS version at least 2.4.53"; return 0; }
4191 umount_client $MOUNT || error "umount client failed"
4193 start_mdt 1 || error "MDT start failed"
4194 start_ost || error "Unable to start OST1"
4196 # START-END - the sequences we'll be reserving
4197 START=$(do_facet $SINGLEMDS \
4198 $LCTL get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
4199 END=$((START + (1 << 30)))
4200 do_facet $SINGLEMDS \
4201 $LCTL set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
4203 # reset the sequences MDT0000 has already assigned
4204 do_facet $SINGLEMDS \
4205 $LCTL set_param seq.srv*MDT0000.space=clear
4207 # remount to let the client allocate new sequence
4208 mount_client $MOUNT || error "mount client failed"
4210 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4211 do_facet $SINGLEMDS \
4212 $LCTL get_param seq.srv*MDT0000.space
4213 $LFS path2fid $DIR/$tfile
4215 local old_ifs="$IFS"
4217 fid=($($LFS path2fid $DIR/$tfile))
4221 if [[ $seq < $END ]]; then
4222 error "used reserved sequence $seq?"
4224 cleanup || error "cleanup failed with $?"
4226 run_test 68 "be able to reserve specific sequences in FLDB"
4229 local server_version=$(lustre_version_code $SINGLEMDS)
4231 [[ $server_version -lt $(version_code 2.4.2) ]] &&
4232 skip "Need MDS version at least 2.4.2" && return
4234 [[ $server_version -ge $(version_code 2.4.50) ]] &&
4235 [[ $server_version -lt $(version_code 2.5.0) ]] &&
4236 skip "Need MDS version at least 2.5.0" && return
4240 # use OST0000 since it probably has the most creations
4241 local OSTNAME=$(ostname_from_index 0)
4242 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
4243 local last_id=$(do_facet mds1 $LCTL get_param -n \
4244 osc.$mdtosc_proc1.prealloc_last_id)
4246 # Want to have OST LAST_ID over 1.5 * OST_MAX_PRECREATE to
4247 # verify that the LAST_ID recovery is working properly. If
4248 # not, then the OST will refuse to allow the MDS connect
4249 # because the LAST_ID value is too different from the MDS
4250 #define OST_MAX_PRECREATE=20000
4251 local num_create=$((20000 * 5))
4253 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4254 $SETSTRIPE -i 0 $DIR/$tdir || error "$SETSTRIPE -i 0 $DIR/$tdir failed"
4255 createmany -o $DIR/$tdir/$tfile- $num_create ||
4256 error "createmany: failed to create $num_create files: $?"
4257 # delete all of the files with objects on OST0 so the
4258 # filesystem is not inconsistent later on
4259 $LFS find $MOUNT --ost 0 | xargs rm
4261 stop_ost || error "OST0 stop failure"
4262 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
4263 $(ostdevname 1) $(ostvdevname 1) ||
4264 error "reformat and replace $ostdev failed"
4265 start_ost || error "OST0 restart failure"
4266 wait_osc_import_state mds ost FULL
4268 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
4269 local idx=$($GETSTRIPE -i $DIR/$tdir/$tfile-last)
4270 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
4272 cleanup || error "cleanup failed with $?"
4274 run_test 69 "replace an OST with the same index"
4277 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4280 cleanup || error "cleanup failed with $?"
4282 start_mdt 1 || error "MDT0 start fail"
4284 start_ost || error "OST0 start fail"
4286 start_mdt 2 || error "MDT1 start fail"
4288 mount_client $MOUNT || error "mount client fails"
4290 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
4292 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4293 error "create remote dir fail"
4295 rm -rf $DIR/$tdir || error "delete dir fail"
4296 cleanup || error "cleanup failed with $?"
4298 run_test 70a "start MDT0, then OST, then MDT1"
4301 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4304 start_ost || error "OST0 start fail"
4306 start_mdt 1 || error "MDT0 start fail"
4307 start_mdt 2 || error "MDT1 start fail"
4309 mount_client $MOUNT || error "mount client fails"
4311 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
4313 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4314 error "create remote dir fail"
4316 rm -rf $DIR/$tdir || error "delete dir fail"
4318 cleanup || error "cleanup failed with $?"
4320 run_test 70b "start OST, MDT1, MDT0"
4323 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4326 start_mdt 1 || error "MDT0 start fail"
4327 start_mdt 2 || error "MDT1 start fail"
4328 start_ost || error "OST0 start fail"
4330 mount_client $MOUNT || error "mount client fails"
4331 stop_mdt 1 || error "MDT1 start fail"
4333 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
4334 echo "deactivate $mdc_for_mdt1"
4335 $LCTL --device $mdc_for_mdt1 deactivate ||
4336 error "set $mdc_for_mdt1 deactivate failed"
4338 mkdir $DIR/$tdir && error "mkdir succeed"
4340 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4341 error "create remote dir succeed"
4343 cleanup || error "cleanup failed with $?"
4345 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
4348 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4351 start_mdt 1 || error "MDT0 start fail"
4352 start_mdt 2 || error "MDT1 start fail"
4353 start_ost || error "OST0 start fail"
4355 mount_client $MOUNT || error "mount client fails"
4357 stop_mdt 2 || error "MDT1 start fail"
4359 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
4361 echo "deactivate $mdc_for_mdt2"
4362 $LCTL --device $mdc_for_mdt2 deactivate ||
4363 error "set $mdc_for_mdt2 deactivate failed"
4365 mkdir $DIR/$tdir || error "mkdir fail"
4366 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4367 error "create remote dir succeed"
4369 rm -rf $DIR/$tdir || error "delete dir fail"
4371 cleanup || error "cleanup failed with $?"
4373 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
4376 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4377 if combined_mgs_mds; then
4378 skip "needs separate MGS/MDT" && return
4382 start_mdt 1 || error "MDT0 start fail"
4383 start_ost || error "OST0 start fail"
4384 start_mdt 2 || error "MDT1 start fail"
4385 start_ost2 || error "OST1 start fail"
4387 mount_client $MOUNT || error "mount client fails"
4389 mkdir $DIR/$tdir || error "mkdir fail"
4390 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4391 error "create remote dir succeed"
4393 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4394 rm -rf $DIR/$tdir || error "delete dir fail"
4396 umount_client $MOUNT || error "umount_client failed"
4397 stop_mdt 1 || error "MDT0 stop fail"
4398 stop_mdt 2 || error "MDT1 stop fail"
4399 stop_ost || error "OST0 stop fail"
4400 stop_ost2 || error "OST1 stop fail"
4402 run_test 71a "start MDT0 OST0, MDT1, OST1"
4405 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4406 if combined_mgs_mds; then
4407 skip "needs separate MGS/MDT" && return
4411 start_mdt 2 || error "MDT1 start fail"
4412 start_ost || error "OST0 start fail"
4413 start_mdt 1 || error "MDT0 start fail"
4414 start_ost2 || error "OST1 start fail"
4416 mount_client $MOUNT || error "mount client fails"
4418 mkdir $DIR/$tdir || error "mkdir fail"
4419 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4420 error "create remote dir succeed"
4422 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4423 rm -rf $DIR/$tdir || error "delete dir fail"
4425 umount_client $MOUNT || error "umount_client failed"
4426 stop_mdt 1 || error "MDT0 stop fail"
4427 stop_mdt 2 || error "MDT1 stop fail"
4428 stop_ost || error "OST0 stop fail"
4429 stop_ost2 || error "OST1 stop fail"
4431 run_test 71b "start MDT1, OST0, MDT0, OST1"
4434 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4435 if combined_mgs_mds; then
4436 skip "needs separate MGS/MDT" && return
4440 start_ost || error "OST0 start fail"
4441 start_ost2 || error "OST1 start fail"
4442 start_mdt 2 || error "MDT1 start fail"
4443 start_mdt 1 || error "MDT0 start fail"
4445 mount_client $MOUNT || error "mount client fails"
4447 mkdir $DIR/$tdir || error "mkdir fail"
4448 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4449 error "create remote dir succeed"
4451 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4452 rm -rf $DIR/$tdir || error "delete dir fail"
4454 umount_client $MOUNT || error "umount_client failed"
4455 stop_mdt 1 || error "MDT0 stop fail"
4456 stop_mdt 2 || error "MDT1 stop fail"
4457 stop_ost || error "OST0 stop fail"
4458 stop_ost2 || error "OST1 stop fail"
4461 run_test 71c "start OST0, OST1, MDT1, MDT0"
4464 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4465 if combined_mgs_mds; then
4466 skip "needs separate MGS/MDT" && return
4470 start_ost || error "OST0 start fail"
4471 start_mdt 2 || error "MDT0 start fail"
4472 start_mdt 1 || error "MDT0 start fail"
4473 start_ost2 || error "OST1 start fail"
4475 mount_client $MOUNT || error "mount client fails"
4477 mkdir $DIR/$tdir || error "mkdir fail"
4478 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4479 error "create remote dir succeed"
4481 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4482 rm -rf $DIR/$tdir || error "delete dir fail"
4484 umount_client $MOUNT || error "umount_client failed"
4485 stop_mdt 1 || error "MDT0 stop fail"
4486 stop_mdt 2 || error "MDT1 stop fail"
4487 stop_ost || error "OST0 stop fail"
4488 stop_ost2 || error "OST1 stop fail"
4491 run_test 71d "start OST0, MDT1, MDT0, OST1"
4494 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4495 if combined_mgs_mds; then
4496 skip "needs separate MGS/MDT" && return
4500 start_ost || error "OST0 start fail"
4501 start_mdt 2 || error "MDT1 start fail"
4502 start_ost2 || error "OST1 start fail"
4503 start_mdt 1 || error "MDT0 start fail"
4505 mount_client $MOUNT || error "mount client fails"
4507 mkdir $DIR/$tdir || error "mkdir fail"
4508 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4509 error "create remote dir succeed"
4511 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4512 rm -rf $DIR/$tdir || error "delete dir fail"
4514 umount_client $MOUNT || error "umount_client failed"
4515 stop_mdt 1 || error "MDT0 stop fail"
4516 stop_mdt 2 || error "MDT1 stop fail"
4517 stop_ost || error "OST0 stop fail"
4518 stop_ost2 || error "OST1 stop fail"
4521 run_test 71e "start OST0, MDT1, OST1, MDT0"
4523 test_72() { #LU-2634
4524 local mdsdev=$(mdsdevname 1)
4525 local ostdev=$(ostdevname 1)
4526 local cmd="$E2FSCK -fnvd $mdsdev"
4529 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4530 skip "ldiskfs only test" && return
4532 #tune MDT with "-O extents"
4534 for num in $(seq $MDSCOUNT); do
4535 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
4536 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4537 error "add mds $num failed"
4538 $TUNE2FS -O extents $(mdsdevname $num)
4541 add ost1 $(mkfs_opts ost1 $ostdev) --reformat $ostdev ||
4542 error "add $ostdev failed"
4543 start_mgsmds || error "start mds failed"
4544 start_ost || error "start ost failed"
4545 mount_client $MOUNT || error "mount client failed"
4547 #create some short symlinks
4548 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4549 createmany -o $DIR/$tdir/$tfile-%d $fn
4550 echo "create $fn short symlinks"
4551 for i in $(seq -w 1 $fn); do
4552 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
4557 umount_client $MOUNT || error "umount client failed"
4558 stop_mds || error "stop mds failed"
4559 stop_ost || error "stop ost failed"
4562 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
4564 run_test 72 "test fast symlink with extents flag enabled"
4566 test_73() { #LU-3006
4568 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4569 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
4570 error "1st tunefs failed"
4571 start_mgsmds || error "start mds failed"
4572 start_ost || error "start ost failed"
4573 mount_client $MOUNT || error "mount client failed"
4574 $LCTL get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
4575 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
4576 umount_client $MOUNT || error "umount client failed"
4580 run_test 73 "failnode to update from mountdata properly"
4582 test_75() { # LU-2374
4583 [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] &&
4584 skip "Need MDS version at least 2.4.1" && return
4587 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
4588 --reformat $(mdsdevname 1) $(mdsvdevname 1)"
4589 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
4590 --reformat $(ostdevname 1) $(ostvdevname 1)"
4592 #check with default parameters
4593 add mds1 $opts_mds || error "add mds1 failed for default params"
4594 add ost1 $opts_ost || error "add ost1 failed for default params"
4596 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
4597 opts_mds=$(echo $opts_mds |
4598 sed -e "s/--index=$index/--index=$index --mdt/")
4599 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
4600 opts_ost=$(echo $opts_ost |
4601 sed -e "s/--index=$index/--index=$index --ost/")
4603 add mds1 $opts_mds || error "add mds1 failed for new params"
4604 add ost1 $opts_ost || error "add ost1 failed for new params"
4607 run_test 75 "The order of --index should be irrelevant"
4610 [[ $(lustre_version_code mgs) -ge $(version_code 2.4.52) ]] ||
4611 { skip "Need MDS version at least 2.4.52" && return 0; }
4613 local MDMB_PARAM="osc.*.max_dirty_mb"
4614 echo "Change MGS params"
4615 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
4617 echo "max_dirty_mb: $MAX_DIRTY_MB"
4618 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB + MAX_DIRTY_MB))
4619 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
4620 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
4621 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
4622 head -1" $NEW_MAX_DIRTY_MB
4623 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
4624 echo "$MAX_DIRTY_MB"
4625 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
4626 error "error while apply max_dirty_mb"
4628 echo "Check the value is stored after remount"
4631 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
4632 head -1" $NEW_MAX_DIRTY_MB
4633 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
4634 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
4635 error "max_dirty_mb is not saved after remount"
4637 echo "Change OST params"
4638 CLIENT_PARAM="obdfilter.*.client_cache_count"
4639 local CLIENT_CACHE_COUNT
4640 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4642 echo "client_cache_count: $CLIENT_CACHE_COUNT"
4643 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
4644 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
4645 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
4646 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
4647 head -1" $NEW_CLIENT_CACHE_COUNT
4648 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4650 echo "$CLIENT_CACHE_COUNT"
4651 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
4652 error "error while apply client_cache_count"
4654 echo "Check the value is stored after remount"
4657 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
4658 head -1" $NEW_CLIENT_CACHE_COUNT
4659 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4661 echo "$CLIENT_CACHE_COUNT"
4662 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
4663 error "client_cache_count is not saved after remount"
4666 run_test 76a "set permanent params set_param -P"
4668 test_76b() { # LU-4783
4669 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.57) ]] ||
4670 { skip "Need MGS version at least 2.5.57" && return 0; }
4673 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
4674 error "start params log failed"
4677 run_test 76b "verify params log setup correctly"
4679 test_77() { # LU-3445
4680 local server_version=$(lustre_version_code $SINGLEMDS)
4682 [[ $server_version -ge $(version_code 2.2.60) ]] &&
4683 [[ $server_version -le $(version_code 2.4.0) ]] &&
4684 skip "Need MDS version < 2.2.60 or > 2.4.0" && return
4686 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
4687 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
4688 skip_env "mixed loopback and real device not working" && return
4691 local fs2mdsdev=$(mdsdevname 1_2)
4692 local fs2ostdev=$(ostdevname 1_2)
4693 local fs2mdsvdev=$(mdsvdevname 1_2)
4694 local fs2ostvdev=$(ostvdevname 1_2)
4695 local fsname=test1234
4697 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
4699 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
4700 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
4701 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
4702 error "start fs2mds failed"
4704 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
4705 [[ $mgsnid = *,* ]] || mgsnid+=",$mgsnid"
4707 add fs2ost $(mkfs_opts ost1 $fs2ostdev) --mgsnode=$mgsnid \
4708 --failnode=$failnid --fsname=$fsname \
4709 --reformat $fs2ostdev $fs2ostvdev ||
4710 error "add fs2ost failed"
4711 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
4713 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
4714 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
4715 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
4718 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
4721 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ||
4722 $(facet_fstype ost1) != ldiskfs ]] &&
4723 skip "only applicable to ldiskfs-based MDTs and OSTs" && return
4725 # reformat the Lustre filesystem with a smaller size
4726 local saved_MDSSIZE=$MDSSIZE
4727 local saved_OSTSIZE=$OSTSIZE
4728 MDSSIZE=$((MDSSIZE - 20000))
4729 OSTSIZE=$((OSTSIZE - 20000))
4730 reformat || error "(1) reformat Lustre filesystem failed"
4731 MDSSIZE=$saved_MDSSIZE
4732 OSTSIZE=$saved_OSTSIZE
4734 # mount the Lustre filesystem
4735 setup_noconfig || error "(2) setup Lustre filesystem failed"
4738 log "create test files"
4742 mkdir $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
4743 for i in $(seq $num_files); do
4744 file=$MOUNT/$tdir/$tfile-$i
4745 dd if=/dev/urandom of=$file count=1 bs=1M ||
4746 error "(4) create $file failed"
4749 # unmount the Lustre filesystem
4750 cleanup || error "(5) cleanup Lustre filesystem failed"
4752 # run e2fsck on the MDT and OST devices
4753 local mds_host=$(facet_active_host $SINGLEMDS)
4754 local ost_host=$(facet_active_host ost1)
4755 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
4756 local ost_dev=$(ostdevname 1)
4758 run_e2fsck $mds_host $mds_dev "-y"
4759 run_e2fsck $ost_host $ost_dev "-y"
4761 # get the original block count of the MDT and OST filesystems
4762 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
4763 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
4765 # expand the MDT and OST filesystems to the device size
4766 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
4767 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
4769 # run e2fsck on the MDT and OST devices again
4770 run_e2fsck $mds_host $mds_dev "-y"
4771 run_e2fsck $ost_host $ost_dev "-y"
4773 # mount the Lustre filesystem
4777 log "check files after expanding the MDT and OST filesystems"
4778 for i in $(seq $num_files); do
4779 file=$MOUNT/$tdir/$tfile-$i
4780 $CHECKSTAT -t file -s 1048576 $file ||
4781 error "(6) checkstat $file failed"
4785 log "create more files after expanding the MDT and OST filesystems"
4786 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
4787 file=$MOUNT/$tdir/$tfile-$i
4788 dd if=/dev/urandom of=$file count=1 bs=1M ||
4789 error "(7) create $file failed"
4792 # unmount the Lustre filesystem
4793 cleanup || error "(8) cleanup Lustre filesystem failed"
4795 # run e2fsck on the MDT and OST devices
4796 run_e2fsck $mds_host $mds_dev "-y"
4797 run_e2fsck $ost_host $ost_dev "-y"
4799 # get the maximum block count of the MDT and OST filesystems
4800 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
4801 local ost_max_blks=$(get_block_count ost1 $ost_dev)
4803 # get the minimum block count of the MDT and OST filesystems
4804 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
4805 grep minimum | sed -e 's/^.*filesystem: //g')
4806 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
4807 grep minimum | sed -e 's/^.*filesystem: //g')
4809 # shrink the MDT and OST filesystems to a smaller size
4813 if [[ $mds_max_blks -gt $mds_min_blks &&
4814 $mds_max_blks -gt $mds_orig_blks ]]; then
4815 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
4816 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
4817 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
4818 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
4819 error "shrink $SINGLEMDS to $new_blks failed"
4823 if [[ $ost_max_blks -gt $ost_min_blks &&
4824 $ost_max_blks -gt $ost_orig_blks ]]; then
4825 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
4826 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
4827 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
4828 run_resize2fs ost1 $ost_dev $new_blks ||
4829 error "shrink ost1 to $new_blks failed"
4833 # check whether the MDT or OST filesystem was shrunk or not
4835 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
4836 reformat || error "(10) reformat Lustre filesystem failed"
4840 # run e2fsck on the MDT and OST devices again
4841 run_e2fsck $mds_host $mds_dev "-y"
4842 run_e2fsck $ost_host $ost_dev "-y"
4844 # mount the Lustre filesystem again
4848 log "check files after shrinking the MDT and OST filesystems"
4849 for i in $(seq $((num_files + 10))); do
4850 file=$MOUNT/$tdir/$tfile-$i
4851 $CHECKSTAT -t file -s 1048576 $file ||
4852 error "(11) checkstat $file failed"
4855 # unmount and reformat the Lustre filesystem
4856 cleanup || error "(12) cleanup Lustre filesystem failed"
4857 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
4858 reformat || error "(14) reformat Lustre filesystem failed"
4860 run_test 78 "run resize2fs on MDT and OST filesystems"
4862 test_79() { # LU-4227
4863 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.5.59) ]] ||
4864 { skip "Need MDS version at least 2.5.59"; return 0; }
4866 local mdsdev1=$(mdsdevname 1)
4867 local mdsvdev1=$(mdsvdevname 1)
4868 local mdsdev2=$(mdsdevname 2)
4869 local mdsvdev2=$(mdsvdevname 2)
4870 local ostdev1=$(ostdevname 1)
4871 local ostvdev1=$(ostvdevname 1)
4872 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
4873 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
4874 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
4877 # remove --mgs/--mgsnode from mkfs.lustre options
4878 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
4880 mgsnode_opt=$(echo $opts_mds2 |
4881 awk '{ for ( i = 1; i < NF; i++ )
4882 if ( $i ~ "--mgsnode" ) { print $i; break } }')
4883 [ -n $mgsnode_opt ] &&
4884 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
4886 mgsnode_opt=$(echo $opts_ost1 |
4887 awk '{ for ( i = 1; i < NF; i++ )
4888 if ( $i ~ "--mgsnode" ) { print $i; break } }')
4889 [ -n $mgsnode_opt ] &&
4890 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
4892 # -MGS, format a mdt without --mgs option
4893 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
4894 error "Must specify --mgs when formatting mdt combined with mgs"
4896 # +MGS, format a mdt/ost without --mgsnode option
4897 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
4898 > /dev/null || error "start mds1 failed"
4899 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
4900 error "Must specify --mgsnode when formatting a mdt"
4901 add ost1 $opts_ost1 $ostdev1 $ostvdev2 &&
4902 error "Must specify --mgsnode when formatting an ost"
4906 run_test 79 "format MDT/OST without mgs option (should return errors)"
4909 start_mds || error "Failed to start MDT"
4910 start_ost || error "Failed to start OST1"
4911 uuid=$(do_facet ost1 $LCTL get_param -n mgc.*.uuid)
4912 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
4913 do_facet ost1 "$LCTL set_param fail_val=10 fail_loc=0x906"
4914 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x906"
4915 do_facet mgs "$LCTL set_param -n mgs/MGS/evict_client $uuid"
4917 start_ost2 || error "Failed to start OST2"
4919 do_facet ost1 "$LCTL set_param fail_loc=0"
4922 run_test 80 "mgc import reconnect race"
4924 # Save the original values of $OSTCOUNT and $OSTINDEX$i.
4926 local new_ostcount=$1
4927 saved_ostcount=$OSTCOUNT
4928 OSTCOUNT=$new_ostcount
4932 for i in $(seq $OSTCOUNT); do
4934 eval saved_ostindex$i=${!index}
4939 # Restore the original values of $OSTCOUNT and $OSTINDEX$i.
4940 restore_ostindex() {
4945 for i in $(seq $OSTCOUNT); do
4946 index=saved_ostindex$i
4947 eval OSTINDEX$i=${!index}
4949 OSTCOUNT=$saved_ostcount
4954 # The main purpose of this test is to ensure the OST_INDEX_LIST functions as
4955 # expected. This test uses OST_INDEX_LIST to format OSTs with a randomly
4956 # assigned index and ensures we can mount such a formatted file system
4957 test_81() { # LU-4665
4958 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
4959 { skip "Need MDS version at least 2.6.54" && return; }
4960 [[ $OSTCOUNT -ge 3 ]] || { skip_env "Need at least 3 OSTs" && return; }
4964 # Each time RANDOM is referenced, a random integer between 0 and 32767
4967 local saved_ostindex1=$OSTINDEX1
4968 for i in 65535 $((RANDOM + 65536)); do
4969 echo -e "\nFormat ost1 with --index=$i, should fail"
4971 if add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat \
4972 $(ostdevname 1) $(ostvdevname 1); then
4973 OSTINDEX1=$saved_ostindex1
4974 error "format ost1 with --index=$i should fail"
4977 OSTINDEX1=$saved_ostindex1
4981 # Format OSTs with random sparse indices.
4982 trap "restore_ostindex" EXIT
4983 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices"
4984 OST_INDEX_LIST=[0,$((RANDOM * 2 % 65533 + 1)),65534] formatall
4986 # Setup and check Lustre filesystem.
4987 start_mgsmds || error "start_mgsmds failed"
4988 for i in $(seq $OSTCOUNT); do
4989 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
4990 error "start ost$i failed"
4993 mount_client $MOUNT || error "mount client $MOUNT failed"
4994 check_mount || error "check client $MOUNT failed"
4997 local max_easize=$($LCTL get_param -n llite.*.max_easize)
4998 [[ $max_easize -eq 128 ]] ||
4999 error "max_easize is $max_easize, should be 128 bytes"
5003 run_test 81 "sparse OST indexing"
5005 # Wait OSTs to be active on both client and MDT side.
5007 local cmd="$LCTL get_param -n lov.$FSNAME-clilov-*.target_obd |
5008 awk 'BEGIN {c = 0} /ACTIVE/{c += 1} END {printf \\\"%d\\\", c}'"
5009 wait_update $HOSTNAME "eval $cmd" $OSTCOUNT ||
5010 error "wait_update OSTs up on client failed"
5012 cmd="$LCTL get_param -n lod.$FSNAME-MDT*-*.target_obd | sort -u |
5013 awk 'BEGIN {c = 0} /ACTIVE/{c += 1} END {printf \\\"%d\\\", c}'"
5014 wait_update_facet $SINGLEMDS "eval $cmd" $OSTCOUNT ||
5015 error "wait_update OSTs up on MDT failed"
5018 # Here we exercise the stripe placement functionality on a file system that
5019 # has formatted the OST with a random index. With the file system the following
5020 # functionality is tested:
5022 # 1. Creating a new file with a specific stripe layout.
5024 # 2. Modifiy a existing empty file with a specific stripe layout.
5026 # 3. Ensure we fail to set the stripe layout of a file that already has one.
5028 # 4. If ost-index is defined we need to ensure it is the first entry in the
5029 # ost index list returned by lfs getstripe.
5031 # 5. Lastly ensure this functionality fails with directories.
5032 test_82a() { # LU-4665
5033 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5034 { skip "Need MDS version at least 2.6.54" && return; }
5035 [[ $OSTCOUNT -ge 3 ]] || { skip_env "Need at least 3 OSTs" && return; }
5041 # Format OSTs with random sparse indices.
5045 for i in $(seq $OSTCOUNT); do
5046 index=$((RANDOM * 2))
5047 ost_indices+=" $index"
5049 ost_indices=$(comma_list $ost_indices)
5051 trap "restore_ostindex" EXIT
5052 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5053 OST_INDEX_LIST=[$ost_indices] formatall
5055 # Setup Lustre filesystem.
5056 start_mgsmds || error "start_mgsmds failed"
5057 for i in $(seq $OSTCOUNT); do
5058 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5059 error "start ost$i failed"
5062 mount_client $MOUNT || error "mount client $MOUNT failed"
5065 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5066 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5068 # 1. If the file does not exist, new file will be created
5069 # with specified OSTs.
5070 local file=$DIR/$tdir/$tfile-1
5071 local cmd="$SETSTRIPE -o $ost_indices $file"
5073 eval $cmd || error "$cmd failed"
5074 check_stripe_count $file $OSTCOUNT
5075 check_obdidx $file $ost_indices
5076 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5077 error "write $file failed"
5079 # 2. If the file already exists and is an empty file, the file
5080 # will be attached with specified layout.
5081 file=$DIR/$tdir/$tfile-2
5082 mcreate $file || error "mcreate $file failed"
5083 cmd="$SETSTRIPE -o $ost_indices $file"
5085 eval $cmd || error "$cmd failed"
5086 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5087 error "write $file failed"
5088 check_stripe_count $file $OSTCOUNT
5089 check_obdidx $file $ost_indices
5091 # 3. If the file already has a valid layout attached, the command
5092 # should fail with EBUSY.
5094 eval $cmd && error "stripe is already set on $file, $cmd should fail"
5096 # 4. If [--stripe-index|-i <start_ost_idx>] is used, the index must
5097 # be in the OST indices list.
5098 local start_ost_idx=${ost_indices##*,}
5099 file=$DIR/$tdir/$tfile-3
5100 cmd="$SETSTRIPE -o $ost_indices -i $start_ost_idx $file"
5102 eval $cmd || error "$cmd failed"
5103 check_stripe_count $file $OSTCOUNT
5104 check_obdidx $file $ost_indices
5105 check_start_ost_idx $file $start_ost_idx
5107 file=$DIR/$tdir/$tfile-4
5109 cmd+=" -o $(exclude_items_from_list $ost_indices $start_ost_idx)"
5110 cmd+=" -i $start_ost_idx $file"
5112 eval $cmd && error "index $start_ost_idx should be in $ost_indices"
5114 # 5. Specifying OST indices for directory should fail with ENOSUPP.
5115 local dir=$DIR/$tdir/$tdir
5116 mkdir $dir || error "mkdir $dir failed"
5117 cmd="$SETSTRIPE -o $ost_indices $dir"
5119 eval $cmd && error "$cmd should fail, specifying OST indices" \
5120 "for directory is not supported"
5124 run_test 82a "specify OSTs for file (succeed) or directory (fail)"
5129 # Remove OSTs from a pool and destroy the pool.
5130 destroy_pool $ost_pool || true
5135 # Test 82b is run to ensure that if the user supplies a pool with a specific
5136 # stripe layout that it behaves proprerly. It should fail in the case that
5137 # the supplied OST index list points to OSTs not contained in the user
5139 test_82b() { # LU-4665
5140 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5141 { skip "Need MDS version at least 2.6.54" && return; }
5142 [[ $OSTCOUNT -ge 4 ]] || { skip_env "Need at least 4 OSTs" && return; }
5148 # Format OSTs with random sparse indices.
5152 for i in $(seq $OSTCOUNT); do
5153 index=$((RANDOM * 2))
5154 ost_indices+=" $index"
5156 ost_indices=$(comma_list $ost_indices)
5158 trap "restore_ostindex" EXIT
5159 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5160 OST_INDEX_LIST=[$ost_indices] formatall
5162 # Setup Lustre filesystem.
5163 start_mgsmds || error "start_mgsmds failed"
5164 for i in $(seq $OSTCOUNT); do
5165 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5166 error "start ost$i failed"
5169 mount_client $MOUNT || error "mount client $MOUNT failed"
5171 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5172 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5174 # Create a new pool and add OSTs into it.
5175 local ost_pool=$FSNAME.$TESTNAME
5176 create_pool $ost_pool || error "create OST pool $ost_pool failed"
5179 trap "cleanup_82b" EXIT
5181 local ost_idx_in_list=${ost_indices##*,}
5182 local ost_idx_in_pool=$(exclude_items_from_list $ost_indices \
5185 local ost_targets="$FSNAME-OST["
5186 for i in ${ost_idx_in_pool//,/ }; do
5187 ost_targets=$ost_targets$(printf "%04x," $i)
5189 ost_targets="${ost_targets%,}]"
5191 local ost_targets_uuid=$(for i in ${ost_idx_in_pool//,/ }; \
5192 do printf "$FSNAME-OST%04x_UUID\n" $i; done |
5193 sort -u | tr '\n' ' ')
5195 local cmd="$LCTL pool_add $ost_pool $ost_targets"
5196 do_facet mgs $cmd || error "$cmd failed"
5197 wait_update $HOSTNAME "$LCTL get_param -n lov.$FSNAME-*.pools.$TESTNAME|
5198 sort -u | tr '\n' ' ' " "$ost_targets_uuid" ||
5199 error "wait_update $ost_pool failed"
5200 pool_list $ost_pool || error "list OST pool $ost_pool failed"
5202 # If [--pool|-p <pool_name>] is set with [--ost-list|-o <ost_indices>],
5203 # then the OSTs must be the members of the pool.
5204 local file=$DIR/$tdir/$tfile
5205 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
5207 eval $cmd && error "OST with index $ost_idx_in_list should be" \
5208 "in OST pool $ost_pool"
5210 # Only select OST $ost_idx_in_list from $ost_pool for file.
5211 ost_idx_in_list=${ost_idx_in_pool#*,}
5212 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
5214 eval $cmd || error "$cmd failed"
5215 cmd="$GETSTRIPE $file"
5217 eval $cmd || error "$cmd failed"
5218 check_stripe_count $file 2
5219 check_obdidx $file $ost_idx_in_list
5220 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5221 error "write $file failed"
5225 run_test 82b "specify OSTs for file with --pool and --ost-list options"
5233 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
5234 skip "Only applicable to ldiskfs-based MDTs"
5239 ostmnt=$(facet_mntpt ost1)
5240 fstype=$(facet_fstype ost1)
5242 # Mount the OST as an ldiskfs filesystem.
5243 log "mount the OST $dev as a $fstype filesystem"
5244 add ost1 $(mkfs_opts ost1 $dev) $FSTYPE_OPT \
5245 --reformat $dev $dev > /dev/null ||
5246 error "format ost1 error"
5248 if ! test -b $dev; then
5249 mnt_opts=$(csa_add "$OST_MOUNT_OPTS" -o loop)
5251 echo "mnt_opts $mnt_opts"
5252 do_facet ost1 mount -t $fstype $dev \
5254 # Run llverfs on the mounted ldiskfs filesystem.
5255 # It is needed to get ENOSPACE.
5256 log "run llverfs in partial mode on the OST $fstype $ostmnt"
5257 do_rpc_nodes $(facet_host ost1) run_llverfs $ostmnt -vpl \
5258 "no" || error "run_llverfs error on $fstype"
5261 log "unmount the OST $dev"
5264 # Delete file IO_scrub. Later osd_scrub_setup will try to
5265 # create "IO_scrub" but will get ENOSPACE.
5267 echo "start ost1 service on `facet_active_host ost1`"
5268 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
5271 err=$(do_facet ost1 dmesg | grep "VFS: Busy inodes after unmount of")
5272 echo "string err $err"
5273 [ -z "$err" ] || error $err
5276 run_test 83 "ENOSPACE on OST doesn't cause message VFS: \
5277 Busy inodes after unmount ..."
5279 recovery_time_min() {
5280 local CONNECTION_SWITCH_MIN=5
5281 local CONNECTION_SWITCH_INC=5
5282 local CONNECTION_SWITCH_MAX
5283 local RECONNECT_DELAY_MAX
5284 local INITIAL_CONNECT_TIMEOUT
5288 #CONNECTION_SWITCH_MAX=min(50, max($CONNECTION_SWITCH_MIN,$TIMEOUT)
5289 (($CONNECTION_SWITCH_MIN>$TIMEOUT)) && \
5290 max=$CONNECTION_SWITCH_MIN || max=$TIMEOUT
5291 (($max<50)) && CONNECTION_SWITCH_MAX=$max || CONNECTION_SWITCH_MAX=50
5293 #INITIAL_CONNECT_TIMEOUT = max(CONNECTION_SWITCH_MIN, \
5295 TO_20=$(($TIMEOUT/20))
5296 (($CONNECTION_SWITCH_MIN>$TO_20)) && \
5297 INITIAL_CONNECT_TIMEOUT=$CONNECTION_SWITCH_MIN || \
5298 INITIAL_CONNECT_TIMEOUT=$TO_20
5300 RECONNECT_DELAY_MAX=$(($CONNECTION_SWITCH_MAX+$CONNECTION_SWITCH_INC+ \
5301 $INITIAL_CONNECT_TIMEOUT))
5302 echo $((2*$RECONNECT_DELAY_MAX))
5306 local facet=$SINGLEMDS
5307 local num=$(echo $facet | tr -d "mds")
5308 local dev=$(mdsdevname $num)
5309 local time_min=$(recovery_time_min)
5310 local recovery_duration
5311 local completed_clients
5314 echo "start mds service on $(facet_active_host $facet)"
5315 start $facet ${dev} $MDS_MOUNT_OPTS \
5316 "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
5317 error "start MDS failed"
5322 echo "recovery_time=$time_min, timeout=$TIMEOUT, wrap_up=$wrap_up"
5324 mount_client $MOUNT1 || error "mount failed"
5325 mount_client $MOUNT2 || error "mount failed"
5327 replay_barrier $SINGLEMDS
5328 createmany -o $DIR1/$tfile-%d 1000
5330 # We need to catch the end of recovery window to extend it.
5331 # Skip 5 requests and add delay to request handling.
5332 #define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP
5333 do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5"
5335 facet_failover $SINGLEMDS || error "failover: $?"
5338 echo "recovery status"
5339 do_facet $SINGLEMDS \
5340 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status"
5342 recovery_duration=$(do_facet $SINGLEMDS \
5343 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
5344 awk '/recovery_duration/ { print $2 }')
5345 (( $recovery_duration > $time_min + $wrap_up )) &&
5346 error "recovery_duration > recovery_time_hard + wrap up"
5347 completed_clients=$(do_facet $SINGLEMDS \
5348 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
5349 awk '/completed_clients/ { print $2 }')
5350 [ "$completed_clients" = "1/2" ] ||
5351 error "completed_clients != 1/2: $completed_clients"
5353 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
5354 umount_client $MOUNT1
5355 umount_client $MOUNT2
5361 run_test 84 "check recovery_hard_time"
5363 if ! combined_mgs_mds ; then
5369 # restore the values of MDSSIZE and OSTSIZE
5370 MDSSIZE=$STORED_MDSSIZE
5371 OSTSIZE=$STORED_OSTSIZE