7 # bug number for skipped test:
8 ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT"
9 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
13 if [ -r /etc/SuSE-release ]
15 local vers=$(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 $SINGLEMDS) = "zfs" ] &&
81 # bug number for skipped test: LU-4444
82 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 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_mds || error "MDT start failed"
325 echo "start mds second time.."
326 start_mds && 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=$old failed"
1154 # prepare a non-symlink parameter in the OSD
1156 param="$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 $?"
1272 run_test 30a "Big config llog and conf_param deletion"
1277 local orignids=$($LCTL get_param -n \
1278 osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1280 local orignidcount=$(echo "$orignids" | wc -w)
1282 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1283 # numerical part of it. Hopefully that's not already a failover address
1285 local OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | \
1287 local ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1288 local NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1289 local NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1290 echo "Using fake nid $NEW"
1292 local TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1293 grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1294 set_conf_param_and_check client "$TEST" \
1295 "$FSNAME-OST0000.failover.node" $NEW ||
1296 error "didn't add failover nid $NEW"
1297 local NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1300 local NIDCOUNT=$(echo "$NIDS" | wc -w)
1301 echo "should have $((orignidcount + 1)) entries \
1302 in failover nids string, have $NIDCOUNT"
1303 [ $NIDCOUNT -eq $((orignidcount + 1)) ] ||
1304 error "Failover nid not added"
1306 do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" ||
1307 error "conf_param delete failed"
1308 umount_client $MOUNT
1309 mount_client $MOUNT || error "mount_client $MOUNT failed"
1311 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1314 NIDCOUNT=$(echo "$NIDS" | wc -w)
1315 echo "only $orignidcount final entries should remain \
1316 in failover nids string, have $NIDCOUNT"
1317 [ $NIDCOUNT -eq $orignidcount ] || error "Failover nids not removed"
1319 cleanup || error "cleanup failed with rc $?"
1321 run_test 30b "Remove failover nids"
1323 test_31() { # bug 10734
1324 # ipaddr must not exist
1325 $MOUNT_CMD 4.3.2.1@tcp:/lustre $MOUNT || true
1326 cleanup || error "cleanup failed with rc $?"
1328 run_test 31 "Connect to non-existent node (shouldn't crash)"
1332 T32_BLIMIT=20480 # Kbytes
1336 # This is not really a test but a tool to create new disk
1337 # image tarballs for the upgrade tests.
1339 # Disk image tarballs should be created on single-node
1340 # clusters by running this test with default configurations
1341 # plus a few mandatory environment settings that are verified
1342 # at the beginning of the test.
1344 test_32newtarball() {
1348 local tmp=$TMP/t32_image_create
1350 if [ $FSNAME != t32fs -o \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o \
1351 $OSTCOUNT -ne 1 -o -z "$OSTDEV1" ]; then
1352 error "Needs FSNAME=t32fs MDSCOUNT=2 " \
1353 "MDSDEV1=<nonexistent_file>" \
1354 "MDSDEV2=<nonexistent_file>" \
1355 "(or MDSDEV, in the case of b1_8)" \
1356 "OSTCOUNT=1 OSTDEV1=<nonexistent_file>"
1360 echo "Found stale $tmp"
1364 mkdir $tmp/src || return 1
1365 tar cf - -C $src . | tar xf - -C $tmp/src
1366 dd if=/dev/zero of=$tmp/src/t32_qf_old bs=1M \
1367 count=$(($T32_BLIMIT / 1024 / 2))
1368 chown $T32_QID.$T32_QID $tmp/src/t32_qf_old
1370 # format ost with comma-separated NIDs to verify LU-4460
1371 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
1372 MGSNID="$MGSNID,$MGSNID" OSTOPT="--failnode=$failnid" formatall
1376 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ] &&
1377 $LFS quotacheck -ug /mnt/$FSNAME
1378 $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
1381 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1384 mkdir $tmp/img || return 1
1388 ls -Rni --time-style=+%s >$tmp/img/list
1389 find ! -name .lustre -type f -exec sha1sum {} \; |
1390 sort -k 2 >$tmp/img/sha1sums
1392 $LCTL get_param -n version | head -n 1 |
1393 sed -e 's/^lustre: *//' >$tmp/img/commit
1395 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ] &&
1396 $LFS quotaon -ug /mnt/$FSNAME
1397 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1398 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1399 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1400 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1401 | tr -d "*" > $tmp/img/bspace
1402 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1403 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1404 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1405 | tr -d "*" > $tmp/img/ispace
1407 if [ $MDSCOUNT -ge 2 ]; then
1408 $LFS mkdir -i 1 /mnt/$FSNAME/remote_dir
1409 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME/remote_dir
1415 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1418 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1419 echo "Data verification failed"
1422 uname -r >$tmp/img/kernel
1423 uname -m >$tmp/img/arch
1425 mv ${MDSDEV1:-$MDSDEV} $tmp/img
1426 for num in $(seq 2 $MDSCOUNT); do
1427 local devname=$(mdsdevname $num)
1428 mv $devname $tmp/img
1430 mv $OSTDEV1 $tmp/img
1432 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1433 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1436 tar cjvf $dst/disk$version-$(facet_fstype $SINGLEMDS).tar.bz2 -S *
1441 #run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1444 # The list of applicable tarballs is returned via the caller's
1445 # variable "tarballs".
1448 local node=$(facet_active_host $SINGLEMDS)
1449 local r="do_node $node"
1451 if [ "$CLIENTONLY" ]; then
1452 skip "Client-only testing"
1456 if ! $r which $TUNEFS; then
1457 skip_env "tunefs.lustre required on $node"
1461 local IMGTYPE=$(facet_fstype $SINGLEMDS)
1463 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 -name \'disk*-$IMGTYPE.tar.bz2\')
1465 if [ -z "$tarballs" ]; then
1466 skip "No applicable tarballs found"
1471 t32_test_cleanup() {
1473 local fstype=$(facet_fstype $SINGLEMDS)
1476 if $shall_cleanup_lustre; then
1477 umount $tmp/mnt/lustre || rc=$?
1479 if $shall_cleanup_mdt; then
1480 $r umount -d $tmp/mnt/mdt || rc=$?
1482 if $shall_cleanup_mdt1; then
1483 $r umount -d $tmp/mnt/mdt1 || rc=$?
1485 if $shall_cleanup_ost; then
1486 $r umount -d $tmp/mnt/ost || rc=$?
1491 if [ $fstype == "zfs" ]; then
1492 $r $ZPOOL destroy t32fs-mdt1 || rc=$?
1493 $r $ZPOOL destroy t32fs-ost1 || rc=$?
1498 t32_bits_per_long() {
1500 # Yes, this is not meant to be perfect.
1510 t32_reload_modules() {
1512 local all_removed=false
1515 while ((i < 20)); do
1516 echo "Unloading modules on $node: Attempt $i"
1517 do_rpc_nodes $node $LUSTRE_RMMOD $(facet_fstype $SINGLEMDS) &&
1519 do_rpc_nodes $node check_mem_leak || return 1
1520 if $all_removed; then
1521 do_rpc_nodes $node load_modules
1527 echo "Unloading modules on $node: Given up"
1531 t32_wait_til_devices_gone() {
1537 echo wait for devices to go
1538 while ((i < 20)); do
1539 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1540 loops=$(do_rpc_nodes $node losetup -a | grep -c t32)
1541 ((devices == 0 && loops == 0)) && return 0
1545 echo "waiting for dev on $node: dev $devices loop $loops given up"
1546 do_rpc_nodes $node "losetup -a"
1547 do_rpc_nodes $node "$LCTL device_list"
1551 t32_verify_quota() {
1555 local fstype=$(facet_fstype $SINGLEMDS)
1559 $LFS quota -u $T32_QID -v $mnt
1561 qval=$($LFS quota -v -u $T32_QID $mnt |
1562 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1563 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1565 [ $qval -eq $img_bspace ] || {
1566 echo "bspace, act:$qval, exp:$img_bspace"
1570 qval=$($LFS quota -v -u $T32_QID $mnt |
1571 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1572 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1574 [ $qval -eq $img_ispace ] || {
1575 echo "ispace, act:$qval, exp:$img_ispace"
1579 qval=$($LFS quota -v -u $T32_QID $mnt |
1580 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1581 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1583 [ $qval -eq $T32_BLIMIT ] || {
1584 echo "blimit, act:$qval, exp:$T32_BLIMIT"
1588 qval=$($LFS quota -v -u $T32_QID $mnt |
1589 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1590 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1592 [ $qval -eq $T32_ILIMIT ] || {
1593 echo "ilimit, act:$qval, exp:$T32_ILIMIT"
1597 do_node $node $LCTL conf_param $fsname.quota.mdt=ug
1598 cmd="$LCTL get_param -n osd-$fstype.$fsname-MDT0000"
1599 cmd=$cmd.quota_slave.enabled
1600 wait_update $node "$cmd" "ug" || {
1601 echo "Enable mdt quota failed"
1605 do_node $node $LCTL conf_param $fsname.quota.ost=ug
1606 cmd="$LCTL get_param -n osd-$fstype.$fsname-OST0000"
1607 cmd=$cmd.quota_slave.enabled
1608 wait_update $node "$cmd" "ug" || {
1609 echo "Enable ost quota failed"
1614 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1615 bs=1M count=$(($T32_BLIMIT / 1024)) oflag=sync && {
1616 echo "Write succeed, but expect -EDQUOT"
1619 rm -f $mnt/t32_qf_new
1621 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1623 echo "Create succeed, but expect -EDQUOT"
1626 unlinkmany $mnt/t32_qf_ $T32_ILIMIT
1634 local dne_upgrade=${dne_upgrade:-"no"}
1635 local ff_convert=${ff_convert:-"no"}
1636 local shall_cleanup_mdt=false
1637 local shall_cleanup_mdt1=false
1638 local shall_cleanup_ost=false
1639 local shall_cleanup_lustre=false
1640 local node=$(facet_active_host $SINGLEMDS)
1641 local r="do_node $node"
1642 local node2=$(facet_active_host mds2)
1650 local nid=$($r $LCTL list_nids | head -1)
1656 local fstype=$(facet_fstype $SINGLEMDS)
1657 local mdt_dev=$tmp/mdt
1658 local mdt2_dev=$tmp/mdt2
1659 local ost_dev=$tmp/ost
1663 trap 'trap - RETURN; t32_test_cleanup' RETURN
1666 mkdir -p $tmp/mnt/lustre || error "mkdir $tmp/mnt/lustre failed"
1667 $r mkdir -p $tmp/mnt/{mdt,mdt1,ost}
1668 $r tar xjvf $tarball -S -C $tmp || {
1669 error_noexit "Unpacking the disk image tarball"
1672 img_commit=$($r cat $tmp/commit)
1673 img_kernel=$($r cat $tmp/kernel)
1674 img_arch=$($r cat $tmp/arch)
1675 img_bspace=$($r cat $tmp/bspace)
1676 img_ispace=$($r cat $tmp/ispace)
1677 echo "Upgrading from $(basename $tarball), created with:"
1678 echo " Commit: $img_commit"
1679 echo " Kernel: $img_kernel"
1680 echo " Arch: $img_arch"
1681 echo "OST version: $(lustre_build_version ost1)"
1683 # The conversion can be made only when both of the following
1684 # conditions are satisfied:
1685 # - ost device img version < 2.3.64
1686 # - ost server version >= 2.5
1687 [ $(version_code $img_commit) -ge $(version_code 2.3.64) -o \
1688 $(lustre_version_code ost1) -lt $(version_code 2.5.0) ] &&
1691 if [ $fstype == "zfs" ]; then
1693 $r $ZPOOL import -f -d $tmp t32fs-mdt1
1694 $r $ZPOOL import -f -d $tmp t32fs-ost1
1695 mdt_dev=t32fs-mdt1/mdt1
1696 ost_dev=t32fs-ost1/ost1
1697 wait_update_facet $SINGLEMDS "$ZPOOL list |
1698 awk '/^t32fs-mdt1/ { print \\\$1 }'" "t32fs-mdt1" || {
1699 error_noexit "import zfs pool failed"
1704 $r $LCTL set_param debug="$PTLDEBUG"
1706 $r $TUNEFS --dryrun $mdt_dev || {
1708 error_noexit "tunefs.lustre before mounting the MDT"
1712 if $r test -f $mdt2_dev; then
1713 $r $TUNEFS --dryrun $mdt2_dev || {
1715 error_noexit "tunefs.lustre before mounting the MDT"
1720 if [ "$writeconf" ]; then
1722 if [ $fstype == "ldiskfs" ]; then
1724 $r $TUNEFS --quota $mdt_dev || {
1726 error_noexit "Enable mdt quota feature"
1729 if $r test -f $mdt2_dev; then
1730 $r $TUNEFS --quota $mdt2_dev || {
1732 error_noexit "Enable mdt quota feature"
1738 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
1739 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
1740 { skip "LU-2200: Cannot run over Inifiniband w/o lctl replace_nids "
1741 "(Need MGS version at least 2.3.59)"; return 0; }
1743 local osthost=$(facet_active_host ost1)
1744 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
1747 if [ $fstype == "ldiskfs" ]; then
1750 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt
1751 $r $LCTL replace_nids $fsname-OST0000 $ostnid
1752 $r $LCTL replace_nids $fsname-MDT0000 $nid
1753 $r umount -d $tmp/mnt/mdt
1756 mopts=exclude=$fsname-OST0000
1757 if [ $fstype == "ldiskfs" ]; then
1762 t32_wait_til_devices_gone $node
1764 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
1766 error_noexit "Mounting the MDT"
1769 shall_cleanup_mdt=true
1771 if $r test -f $mdt2_dev; then
1772 mopts=mgsnode=$nid,$mopts
1773 $r $MOUNT_CMD -o $mopts $mdt2_dev $tmp/mnt/mdt1 || {
1775 error_noexit "Mounting the MDT"
1779 echo "mount new MDT....$mdt2_dev"
1780 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1781 error_noexit "enable remote dir create failed"
1783 shall_cleanup_mdt1=true
1784 elif [ "$dne_upgrade" != "no" ]; then
1785 local fs2mdsdev=$(mdsdevname 1_2)
1786 local fs2mdsvdev=$(mdsvdevname 1_2)
1788 echo "mkfs new MDT on ${fs2mdsdev}...."
1789 if [ $(facet_fstype mds1) == ldiskfs ]; then
1790 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
1793 add fs2mds $(mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
1794 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
1795 error_noexit "Mkfs new MDT failed"
1799 $r $TUNEFS --dryrun $fs2mdsdev || {
1800 error_noexit "tunefs.lustre before mounting the MDT"
1804 echo "mount new MDT....$fs2mdsdev"
1805 $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
1806 error_noexit "mount mdt1 failed"
1810 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1811 error_noexit "enable remote dir create failed"
1813 shall_cleanup_mdt1=true
1816 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
1817 error_noexit "Getting MDT UUID"
1820 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
1821 error_noexit "Unexpected MDT UUID: \"$uuid\""
1825 $r $TUNEFS --dryrun $ost_dev || {
1826 error_noexit "tunefs.lustre before mounting the OST"
1829 if [ "$writeconf" ]; then
1830 mopts=mgsnode=$nid,$writeconf
1831 if [ $fstype == "ldiskfs" ]; then
1833 $r $TUNEFS --quota $ost_dev || {
1835 error_noexit "Enable ost quota feature"
1841 if [ $fstype == "ldiskfs" ]; then
1845 $r $MOUNT_CMD -o $mopts $ost_dev $tmp/mnt/ost || {
1846 error_noexit "Mounting the OST"
1849 shall_cleanup_ost=true
1851 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
1852 error_noexit "Getting OST UUID"
1855 if [ "$uuid" != $fsname-OST0000_UUID ]; then
1856 error_noexit "Unexpected OST UUID: \"$uuid\""
1860 $r $LCTL conf_param $fsname-OST0000.osc.max_dirty_mb=15 || {
1861 error_noexit "Setting \"max_dirty_mb\""
1864 $r $LCTL conf_param $fsname-OST0000.failover.node=$nid || {
1865 error_noexit "Setting OST \"failover.node\""
1868 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
1869 error_noexit "Setting \"max_rpcs_in_flight\""
1872 $r $LCTL conf_param $fsname-MDT0000.failover.node=$nid || {
1873 error_noexit "Setting MDT \"failover.node\""
1876 $r $LCTL pool_new $fsname.interop || {
1877 error_noexit "Setting \"interop\""
1880 $r $LCTL conf_param $fsname-MDT0000.lov.stripesize=4M || {
1881 error_noexit "Setting \"lov.stripesize\""
1884 $r $LCTL conf_param $fsname-MDT0000.mdd.atime_diff=70 || {
1885 error_noexit "Setting \"mdd.atime_diff\""
1889 if [ "$ff_convert" != "no" -a $(facet_fstype ost1) == "ldiskfs" ]; then
1890 $r $LCTL lfsck_start -M $fsname-OST0000 || {
1891 error_noexit "Start OI scrub on OST0"
1895 # The oi_scrub should be on ost1, but for test_32(),
1896 # all on the SINGLEMDS.
1897 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
1898 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1899 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
1900 error_noexit "Failed to get the expected 'completed'"
1904 local UPDATED=$($r $LCTL get_param -n \
1905 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1906 awk '/^updated/ { print $2 }')
1907 [ $UPDATED -ge 1 ] || {
1908 error_noexit "Only $UPDATED objects have been converted"
1913 if [ "$dne_upgrade" != "no" ]; then
1914 $r $LCTL conf_param \
1915 $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
1916 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
1919 $r $LCTL conf_param $fsname-MDT0001.failover.node=$nid || {
1920 error_noexit "Setting MDT1 \"failover.node\""
1923 $r $LCTL conf_param $fsname-MDT0001.lov.stripesize=4M || {
1924 error_noexit "Setting MDT1 \"lov.stripesize\""
1929 if [ "$writeconf" ]; then
1930 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
1931 error_noexit "Mounting the client"
1935 shall_cleanup_lustre=true
1936 $r $LCTL set_param debug="$PTLDEBUG"
1938 t32_verify_quota $node $fsname $tmp/mnt/lustre || {
1939 error_noexit "verify quota failed"
1943 if [ "$dne_upgrade" != "no" ]; then
1944 $LFS mkdir -i 1 -c2 $tmp/mnt/lustre/striped_dir || {
1945 error_noexit "set remote dir failed"
1949 $LFS setdirstripe -D -c2 $tmp/mnt/lustre/striped_dir
1951 pushd $tmp/mnt/lustre
1952 tar -cf - . --exclude=./striped_dir \
1953 --exclude=./remote_dir |
1954 tar -xvf - -C striped_dir 1>/dev/null || {
1955 error_noexit "cp to remote dir failed"
1961 # If it is upgrade from DNE (2.5), then rename the remote dir,
1962 # which is created in 2.5 to striped dir.
1963 if $r test -f $mdt2_dev; then
1964 stripe_index=$(LFS getdirstripe -i \
1965 $tmp/mnt/lustre/remote_dir)
1966 [ $stripe_index -eq 1 ] || {
1967 error_noexit "get index $striped_index failed"
1970 mv $tmp/mnt/lustre/remote_dir \
1971 $tmp/mnt/lustre/striped_dir/ || {
1972 error_noexit "mv failed"
1977 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
1978 error_noexit "dd failed"
1981 rm -rf $tmp/mnt/lustre/tmp_file || {
1982 error_noexit "rm failed"
1986 if $r test -f $tmp/sha1sums; then
1987 # LU-2393 - do both sorts on same node to ensure locale
1989 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
1990 if [ "$dne_upgrade" != "no" ]; then
1991 pushd $tmp/mnt/lustre/striped_dir
1993 pushd $tmp/mnt/lustre
1996 find ! -path "*remote_dir*" ! -name .lustre -type f \
1997 -exec sha1sum {} \; |
1998 sort -k 2 >$tmp/sha1sums || {
2000 error_noexit "sha1sum"
2004 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
2005 error_noexit "sha1sum verification failed"
2009 # if upgrade from DNE (2.5), then check remote directory
2010 if $r test -f $mdt2_dev; then
2011 pushd $tmp/mnt/lustre/striped_dir/remote_dir
2012 find ! -name .lustre -type f \
2013 -exec sha1sum {} \; |
2014 sort -k 2 >$tmp/sha1sums || {
2016 error_noexit "sha1sum"
2020 if ! diff -ub $tmp/sha1sums.orig \
2022 error_noexit "sha1sum dne failed"
2027 echo "sha1sum verification skipped"
2030 if [ "$dne_upgrade" != "no" ]; then
2031 rm -rf $tmp/mnt/lustre/striped_dir || {
2032 error_noexit "remove remote dir failed"
2037 if $r test -f $tmp/list; then
2039 # There is not a Test Framework API to copy files to or
2040 # from a remote node.
2042 # LU-2393 - do both sorts on same node to ensure locale
2044 $r cat $tmp/list | sort -k 6 >$tmp/list.orig
2045 pushd $tmp/mnt/lustre
2046 ls -Rni --time-style=+%s | sort -k 6 >$tmp/list || {
2052 # 32-bit and 64-bit clients use different algorithms to
2053 # convert FIDs into inode numbers. Hence, remove the inode
2054 # numbers from the lists, if the original list was created
2055 # on an architecture with different number of bits per
2058 if [ $(t32_bits_per_long $(uname -m)) != \
2059 $(t32_bits_per_long $img_arch) ]; then
2060 echo "Different number of bits per \"long\" from the disk image"
2061 for list in list.orig list; do
2062 sed -i -e 's/^[0-9]\+[ \t]\+//' $tmp/$list
2065 if ! diff -ub $tmp/list.orig $tmp/list; then
2066 error_noexit "list verification failed"
2070 echo "list verification skipped"
2073 # migrate files/dirs to remote MDT, then move them back
2074 if [ $(lustre_version_code mds1) -ge $(version_code 2.7.50) -a \
2075 $dne_upgrade != "no" ]; then
2076 $r $LCTL set_param -n \
2077 mdt.${fsname}*.enable_remote_dir=1 2>/dev/null
2079 echo "test migration"
2080 pushd $tmp/mnt/lustre
2081 for dir in $(find ! -name .lustre ! -name . -type d); do
2082 mdt_index=$($LFS getdirstripe -i $dir)
2083 stripe_cnt=$($LFS getdirstripe -c $dir)
2084 if [ $mdt_index = 0 -a $stripe_cnt -le 1 ]; then
2085 $LFS mv -M 1 $dir || {
2087 error_noexit "migrate MDT1 failed"
2093 for dir in $(find ! -name . ! -name .lustre -type d); do
2094 mdt_index=$($LFS getdirstripe -i $dir)
2095 stripe_cnt=$($LFS getdirstripe -c $dir)
2096 if [ $mdt_index = 1 -a $stripe_cnt -le 1 ]; then
2097 $LFS mv -M 0 $dir || {
2099 error_noexit "migrate MDT0 failed"
2108 # When adding new data verification tests, please check for
2109 # the presence of the required reference files first, like
2110 # the "sha1sums" and "list" tests above, to avoid the need to
2111 # regenerate every image for each test addition.
2114 nrpcs_orig=$($LCTL get_param \
2115 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
2116 error_noexit "Getting \"max_rpcs_in_flight\""
2119 nrpcs=$((nrpcs_orig + 5))
2120 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=$nrpcs || {
2121 error_noexit "Changing \"max_rpcs_in_flight\""
2124 wait_update $HOSTNAME "$LCTL get_param \
2125 -n mdc.*MDT0000*.max_rpcs_in_flight" $nrpcs || {
2126 error_noexit "Verifying \"max_rpcs_in_flight\""
2130 umount $tmp/mnt/lustre || {
2131 error_noexit "Unmounting the client"
2134 shall_cleanup_lustre=false
2136 if [ "$dne_upgrade" != "no" ]; then
2137 $r umount -d $tmp/mnt/mdt1 || {
2138 error_noexit "Unmounting the MDT2"
2141 shall_cleanup_mdt1=false
2144 $r umount -d $tmp/mnt/mdt || {
2145 error_noexit "Unmounting the MDT"
2148 shall_cleanup_mdt=false
2150 $r umount -d $tmp/mnt/ost || {
2151 error_noexit "Unmounting the OST"
2154 shall_cleanup_ost=false
2156 t32_reload_modules $node || {
2157 error_noexit "Reloading modules"
2161 # mount a second time to make sure we didnt leave upgrade flag on
2162 $r $TUNEFS --dryrun $mdt_dev || {
2164 error_noexit "tunefs.lustre before remounting the MDT"
2168 mopts=exclude=$fsname-OST0000
2169 if [ $fstype == "ldiskfs" ]; then
2172 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2173 error_noexit "Remounting the MDT"
2176 shall_cleanup_mdt=true
2186 for tarball in $tarballs; do
2187 t32_test $tarball || let "rc += $?"
2191 run_test 32a "Upgrade (not live)"
2199 for tarball in $tarballs; do
2200 t32_test $tarball writeconf || let "rc += $?"
2204 run_test 32b "Upgrade with writeconf"
2211 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
2213 for tarball in $tarballs; do
2214 # Do not support 1_8 and 2_1 direct upgrade to DNE2 anymore */
2215 echo $tarball | grep "1_8" && continue
2216 echo $tarball | grep "2_1" && continue
2218 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
2222 run_test 32c "dne upgrade test"
2230 for tarball in $tarballs; do
2231 ff_convert=yes t32_test $tarball || rc=$?
2235 run_test 32d "convert ff test"
2237 test_33a() { # bug 12333, was test_33
2238 local FSNAME2=test-123
2239 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2242 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
2244 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
2245 local dev=${SINGLEMDS}_dev
2246 local MDSDEV=${!dev}
2247 is_blkdev $SINGLEMDS $MDSDEV &&
2248 skip_env "mixed loopback and real device not working" &&
2252 local fs2mdsdev=$(mdsdevname 1_2)
2253 local fs2ostdev=$(ostdevname 1_2)
2254 local fs2mdsvdev=$(mdsvdevname 1_2)
2255 local fs2ostvdev=$(ostvdevname 1_2)
2257 if [ $(facet_fstype mds1) == ldiskfs ]; then
2258 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
2261 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2262 --reformat $mkfsoptions $fs2mdsdev $fs2mdsvdev || exit 10
2263 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2264 --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \
2265 $fs2ostvdev || exit 10
2267 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
2268 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2269 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" ||
2270 error "$LCTL conf_param $FSNAME2.sys.timeout=200 failed"
2271 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2272 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2275 cp /etc/hosts $MOUNT2/ || error "copy /etc/hosts $MOUNT2/ failed"
2276 $GETSTRIPE $MOUNT2/hosts || error "$GETSTRIPE $MOUNT2/hosts failed"
2281 cleanup_nocli || error "cleanup_nocli failed with $?"
2283 run_test 33a "Mount ost with a large index number"
2285 test_33b() { # was test_34
2288 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
2289 # Drop lock cancelation reply during umount
2290 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
2291 do_facet client $LCTL set_param fail_loc=0x80000304
2292 #lctl set_param debug=-1
2293 umount_client $MOUNT
2294 cleanup || error "cleanup failed with $?"
2296 run_test 33b "Drop cancel during umount"
2300 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
2301 manual_umount_client
2303 do_facet client killall -USR1 multiop
2304 if [ $rc -eq 0 ]; then
2305 error "umount not fail!"
2308 cleanup || error "cleanup failed with rc $?"
2310 run_test 34a "umount with opened file should be fail"
2314 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2315 stop_mds || error "Unable to stop MDS"
2317 manual_umount_client --force || error "mtab after failed umount with $?"
2319 cleanup || error "cleanup failed with $?"
2321 run_test 34b "force umount with failed mds should be normal"
2325 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2326 stop_ost || error "Unable to stop OST1"
2328 manual_umount_client --force || error "mtab after failed umount with $?"
2330 cleanup || error "cleanup failed with $?"
2332 run_test 34c "force umount with failed ost should be normal"
2334 test_35a() { # bug 12459
2337 DBG_SAVE="`$LCTL get_param -n debug`"
2338 $LCTL set_param debug="ha"
2340 log "Set up a fake failnode for the MDS"
2342 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2343 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2344 do_facet mgs "$LCTL conf_param \
2345 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" ||
2346 error "Setting ${device}.failover.node=\
2347 $(h2$NETTYPE $FAKENID) failed."
2349 log "Wait for RECONNECT_INTERVAL seconds (10s)"
2352 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
2355 log "Stopping the MDT: $device"
2356 stop_mdt 1 || error "MDT0 stop fail"
2358 df $MOUNT > /dev/null 2>&1 &
2360 log "Restarting the MDT: $device"
2361 start_mdt 1 || error "MDT0 start fail"
2362 log "Wait for df ($DFPID) ... "
2365 $LCTL set_param debug="$DBG_SAVE"
2367 # retrieve from the log the first server that the client tried to
2368 # contact after the connection loss
2369 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2370 NEXTCONN=`awk "/${MSG}/ {start = 1;}
2371 /import_select_connection.*$device-mdc.* using connection/ {
2373 if (\\\$NF ~ /$FAKENID/)
2379 }" $TMP/lustre-log-$TESTNAME.log`
2380 [ "$NEXTCONN" != "0" ] &&
2381 error "Tried to connect to ${NEXTCONN} not last active server"
2382 cleanup || error "cleanup failed with $?"
2383 # remove nid settings
2384 writeconf_or_reformat
2386 run_test 35a "Reconnect to the last active server first"
2388 test_35b() { # bug 18674
2389 remote_mds || { skip "local MDS" && return 0; }
2393 $LCTL set_param debug="ha"
2395 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
2398 log "Set up a fake failnode for the MDS"
2400 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2401 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2402 do_facet mgs "$LCTL conf_param \
2403 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" ||
2404 error "Set ${device}.failover.node=\
2405 $(h2$NETTYPE $FAKENID) failed"
2407 local at_max_saved=0
2408 # adaptive timeouts may prevent seeing the issue
2409 if at_is_enabled; then
2410 at_max_saved=$(at_max_get mds)
2411 at_max_set 0 mds client
2414 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
2416 log "Injecting EBUSY on MDS"
2417 # Setting OBD_FAIL_MDS_RESEND=0x136
2418 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" ||
2419 error "unable to set param fail_loc=0x80000136"
2421 $LCTL set_param mdc.${FSNAME}*.stats=clear
2423 log "Creating a test file and stat it"
2424 touch $MOUNT/$tdir/$tfile || error "touch $MOUNT/$tdir/$tfile failed"
2425 stat $MOUNT/$tdir/$tfile
2427 log "Stop injecting EBUSY on MDS"
2428 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" ||
2429 error "unable to set param fail_loc=0"
2430 rm -f $MOUNT/$tdir/$tfile || error "remove $MOUNT/$tdir/$tfile failed"
2433 # restore adaptive timeout
2434 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
2436 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2438 CONNCNT=$($LCTL get_param mdc.${FSNAME}*.stats |
2439 awk '/mds_connect/{print $2}')
2441 # retrieve from the log if the client has ever tried to
2442 # contact the fake server after the loss of connection
2443 FAILCONN=`awk "BEGIN {ret = 0;}
2444 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
2446 if (\\\$NF ~ /$FAKENID/) {
2451 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
2453 [ "$FAILCONN" == "0" ] &&
2454 error "The client reconnection has not been triggered"
2455 [ "$FAILCONN" == "2" ] &&
2456 error "Primary server busy, client reconnect to failover failed"
2459 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
2460 # Reconnects are supposed to be rate limited to one every 5s
2461 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] &&
2462 error "Too many reconnects $CONNCNT"
2464 cleanup || error "cleanup failed with $?"
2465 # remove nid settings
2466 writeconf_or_reformat
2468 run_test 35b "Continue reconnection retries, if the active server is busy"
2471 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
2473 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] ||
2474 { skip "remote OST" && return 0; }
2477 local FSNAME2=test1234
2478 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2480 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
2482 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
2483 is_blkdev $SINGLEMDS $MDSDEV &&
2484 skip_env "mixed loopback and real device not working" && return
2487 local fs2mdsdev=$(mdsdevname 1_2)
2488 local fs2ostdev=$(ostdevname 1_2)
2489 local fs3ostdev=$(ostdevname 2_2)
2490 local fs2mdsvdev=$(mdsvdevname 1_2)
2491 local fs2ostvdev=$(ostvdevname 1_2)
2492 local fs3ostvdev=$(ostvdevname 2_2)
2494 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2495 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
2496 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
2497 # different one than the default value here.
2498 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2499 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
2500 add fs3ost $(mkfs_opts ost2 ${fs3ostdev}) --mgsnode=$MGSNID \
2501 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
2503 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
2504 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2505 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
2506 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2507 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2509 sleep 5 # until 11778 fixed
2511 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || error "dd failed"
2513 BKTOTAL=$($LCTL get_param -n obdfilter.*.kbytestotal |
2514 awk 'BEGIN{total=0}; {total+=$1}; END{print total}')
2515 BKFREE=$($LCTL get_param -n obdfilter.*.kbytesfree |
2516 awk 'BEGIN{free=0}; {free+=$1}; END{print free}')
2517 BKAVAIL=$($LCTL get_param -n obdfilter.*.kbytesavail |
2518 awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}')
2519 STRING=$(df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}')
2520 DFTOTAL=$(echo $STRING | cut -d, -f1)
2521 DFUSED=$(echo $STRING | cut -d, -f2)
2522 DFAVAIL=$(echo $STRING | cut -d, -f3)
2523 DFFREE=$(($DFTOTAL - $DFUSED))
2525 ALLOWANCE=$((64 * $OSTCOUNT))
2527 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
2528 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
2529 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
2532 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
2533 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
2534 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
2537 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
2538 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
2539 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
2544 stop fs3ost -f || error "unable to stop OST3"
2545 stop fs2ost -f || error "unable to stop OST2"
2546 stop fs2mds -f || error "unable to stop second MDS"
2547 unload_modules_conf || error "unable unload modules"
2550 run_test 36 "df report consistency on OSTs with different block size"
2553 local mntpt=$(facet_mntpt $SINGLEMDS)
2554 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
2555 local mdsdev_sym="$TMP/sym_mdt.img"
2556 local opts=$MDS_MOUNT_OPTS
2559 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2560 skip "Currently only applicable to ldiskfs-based MDTs"
2564 echo "MDS : $mdsdev"
2565 echo "SYMLINK : $mdsdev_sym"
2566 do_facet $SINGLEMDS rm -f $mdsdev_sym
2568 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
2570 echo "mount symlink device - $mdsdev_sym"
2572 if ! do_facet $SINGLEMDS test -b $mdsdev; then
2573 opts=$(csa_add "$opts" -o loop)
2575 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
2576 $mdsdev_sym $mntpt 2>&1)
2579 echo mount_op=$mount_op
2581 do_facet $SINGLEMDS "umount -d $mntpt && rm -f $mdsdev_sym"
2583 if $(echo $mount_op | grep -q "unable to set tunable"); then
2584 error "set tunables failed for symlink device"
2587 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
2589 run_test 37 "verify set tunables works for symlink device"
2591 test_38() { # bug 14222
2592 local fstype=$(facet_fstype $SINGLEMDS)
2593 local mntpt=$(facet_mntpt $SINGLEMDS)
2598 local SRC="/etc /bin"
2599 local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT)
2600 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
2601 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2602 tar cf - $FILES | tar xf - -C $DIR/$tdir ||
2603 error "copying $SRC to $DIR/$tdir"
2605 umount_client $MOUNT || error "umount_client $MOUNT failed"
2606 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2607 stop_mds || error "Unable to stop MDS"
2608 log "delete lov_objid file on MDS"
2610 mount_fstype $SINGLEMDS || error "mount MDS failed (1)"
2612 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid; rm $mntpt/lov_objid"
2614 unmount_fstype $SINGLEMDS || error "umount failed (1)"
2616 # check create in mds_lov_connect
2617 start_mds || error "unable to start MDS"
2618 mount_client $MOUNT || error "mount_client $MOUNT failed"
2620 [ $V ] && log "verifying $DIR/$tdir/$f"
2621 diff -q $f $DIR/$tdir/$f || ERROR=y
2623 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2624 if [ "$ERROR" = "y" ]; then
2625 # check it's updates in sync
2626 umount_client $MOUNT
2628 mount_fstype $SIGNLEMDS
2629 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2630 unmount_fstype $SINGLEMDS
2631 error "old and new files are different after connect" || true
2633 touch $DIR/$tdir/f2 || error "f2 file create failed"
2635 # check it's updates in sync
2636 umount_client $MOUNT || error "second umount_client $MOUNT failed"
2639 mount_fstype $SINGLEMDS || error "mount MDS failed (3)"
2641 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2642 do_facet $SINGLEMDS dd if=/dev/zero of=$mntpt/lov_objid.clear count=8
2644 unmount_fstype $SINGLEMDS || error "umount failed (3)"
2646 start_mds || error "unable to start MDS"
2647 mount_client $MOUNT || error "mount_client $MOUNT failed"
2649 [ $V ] && log "verifying $DIR/$tdir/$f"
2650 diff -q $f $DIR/$tdir/$f || ERROR=y
2652 touch $DIR/$tdir/f3 || error "f3 file create failed"
2653 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2654 umount_client $MOUNT || error "third umount_client $MOUNT failed"
2656 mount_fstype $SINGLEMDS || error "mount MDS failed (4)"
2657 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2658 unmount_fstype $SINGLEMDS || error "umount failed (4)"
2660 [ "$ERROR" = "y" ] &&
2661 error "old and new files are different after sync" || true
2663 log "files compared the same"
2664 cleanup || error "cleanup failed with $?"
2666 run_test 38 "MDS recreates missing lov_objid file from OST data"
2671 cleanup || error "cleanup failed with $?"
2672 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
2673 error "memory leak detected" || true
2675 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
2677 test_40() { # bug 15759
2678 start_ost || error "Unable to start OST1"
2679 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
2680 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
2682 cleanup || error "cleanup failed with rc $?"
2684 run_test 40 "race during service thread startup"
2686 test_41a() { #bug 14134
2687 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2688 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2689 skip "Loop devices does not work with nosvc option"
2693 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2695 start_mdt 1 -o nosvc -n
2696 if [ $MDSCOUNT -ge 2 ]; then
2697 for num in $(seq 2 $MDSCOUNT); do
2698 start_mdt $num || return
2701 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
2702 start_mdt 1 -o nomgs,force
2703 mount_client $MOUNT || error "mount_client $MOUNT failed"
2706 echo "blah blah" > $MOUNT/$tfile
2709 umount_client $MOUNT || error "umount_client $MOUNT failed"
2710 stop ost1 -f || error "unable to stop OST1"
2711 stop_mds || error "Unable to stop MDS"
2712 stop_mds || error "Unable to stop MDS on second try"
2713 unload_modules_conf || error "unload_modules_conf failed"
2715 run_test 41a "mount mds with --nosvc and --nomgs"
2718 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2719 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2720 skip "Loop devices does not work with nosvc option"
2724 ! combined_mgs_mds && skip "needs combined mgs device" && return 0
2728 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2730 start_mdt 1 -o nosvc -n
2731 if [ $MDSCOUNT -ge 2 ]; then
2732 for num in $(seq 2 $MDSCOUNT); do
2733 start_mdt $num || return
2736 start_ost || error "Unable to start OST1"
2737 start_mdt 1 -o nomgs,force
2738 mount_client $MOUNT || error "mount_client $MOUNT failed"
2741 echo "blah blah" > $MOUNT/$tfile
2742 cat $MOUNT/$tfile || error "cat $MOUNT/$tfile failed"
2744 umount_client $MOUNT || error "umount_client $MOUNT failed"
2745 stop_ost || error "Unable to stop OST1"
2746 stop_mds || error "Unable to stop MDS"
2747 stop_mds || error "Unable to stop MDS on second try"
2749 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
2752 local server_version=$(lustre_version_code $SINGLEMDS)
2754 [[ $server_version -ge $(version_code 2.6.52) ]] ||
2755 [[ $server_version -ge $(version_code 2.5.26) &&
2756 $server_version -lt $(version_code 2.5.50) ]] ||
2757 [[ $server_version -ge $(version_code 2.5.4) &&
2758 $server_version -lt $(version_code 2.5.11) ]] ||
2759 { skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"; return; }
2762 # MDT concurrent start
2763 #define OBD_FAIL_TGT_DELAY_CONNECT 0x703
2764 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x703"
2765 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS &
2768 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x0"
2769 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS &
2775 if [ $rc == 0 ] && [ $rc2 == 114 ]; then
2776 echo "1st MDT start succeed"
2777 echo "2nd MDT start failed with EALREADY"
2778 elif [ $rc2 == 0 ] && [ $rc == 114 ]; then
2779 echo "1st MDT start failed with EALREADY"
2780 echo "2nd MDT start succeed"
2783 error "unexpected concurrent MDT mounts result, rc=$rc rc2=$rc2"
2786 if [ $MDSCOUNT -ge 2 ]; then
2787 for num in $(seq 2 $MDSCOUNT); do
2788 start_mdt $num || return
2792 # OST concurrent start
2794 #define OBD_FAIL_TGT_DELAY_CONNECT 0x703
2795 do_facet ost1 "$LCTL set_param fail_loc=0x703"
2796 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &
2799 do_facet ost1 "$LCTL set_param fail_loc=0x0"
2800 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &
2806 if [ $rc == 0 ] && [ $rc2 == 114 ]; then
2807 echo "1st OST start succeed"
2808 echo "2nd OST start failed with EALREADY"
2809 elif [ $rc2 == 0 ] && [ $rc == 114 ]; then
2810 echo "1st OST start failed with EALREADY"
2811 echo "2nd OST start succeed"
2815 error "unexpected concurrent OST mounts result, rc=$rc rc2=$rc2"
2821 # verify everything ok
2826 error "MDT(s) start failed"
2834 error "OST(s) start failed"
2842 error "client start failed"
2849 error "client mount failed"
2853 run_test 41c "concurrent mounts of MDT/OST should all fail but one"
2855 test_42() { #bug 14693
2857 check_mount || error "client was not mounted"
2859 do_facet mgs $LCTL conf_param $FSNAME.llite.some_wrong_param=10
2860 umount_client $MOUNT ||
2861 error "unmounting client failed with invalid llite param"
2862 mount_client $MOUNT ||
2863 error "mounting client failed with invalid llite param"
2865 do_facet mgs $LCTL conf_param $FSNAME.sys.some_wrong_param=20
2866 cleanup || error "stopping $FSNAME failed with invalid sys param"
2869 check_mount || error "client was not mounted with invalid sys param"
2870 cleanup || error "stopping $FSNAME failed with invalid sys param"
2872 run_test 42 "allow client/server mount/unmount with invalid config param"
2875 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.58) ]] ||
2876 { skip "Need MDS version at least 2.5.58" && return 0; }
2877 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
2880 USER1=$(cat /etc/passwd | grep :$ID1:$ID1: | cut -d: -f1)
2881 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1" &&
2885 chmod ugo+x $DIR || error "chmod 0 failed"
2886 set_conf_param_and_check mds \
2887 "$LCTL get_param -n mdt.$FSNAME-MDT0000.root_squash" \
2888 "$FSNAME.mdt.root_squash" \
2890 wait_update $HOSTNAME \
2891 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
2893 error "check llite root_squash failed!"
2894 set_conf_param_and_check mds \
2895 "$LCTL get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
2896 "$FSNAME.mdt.nosquash_nids" \
2898 wait_update $HOSTNAME \
2899 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
2901 error "check llite nosquash_nids failed!"
2904 # create set of test files
2906 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
2907 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
2908 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
2910 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
2911 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
2913 mkdir $DIR/$tdir-rootdir || error "mkdir failed"
2914 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
2915 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
2917 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
2918 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
2919 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
2922 # check root_squash:
2923 # set root squash UID:GID to RUNAS_ID
2924 # root should be able to access only files owned by RUNAS_ID
2926 set_conf_param_and_check mds \
2927 "$LCTL get_param -n mdt.$FSNAME-MDT0000.root_squash" \
2928 "$FSNAME.mdt.root_squash" \
2929 "$RUNAS_ID:$RUNAS_ID"
2930 wait_update $HOSTNAME \
2931 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
2932 "$RUNAS_ID:$RUNAS_ID" ||
2933 error "check llite root_squash failed!"
2935 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
2936 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
2937 error "$ST: root read permission is denied"
2938 echo "$ST: root read permission is granted - ok"
2941 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
2942 error "$ST: root write permission is denied"
2943 echo "$ST: root write permission is granted - ok"
2945 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
2946 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
2947 error "$ST: root read permission is granted"
2948 echo "$ST: root read permission is denied - ok"
2951 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
2952 error "$ST: root write permission is granted"
2953 echo "$ST: root write permission is denied - ok"
2955 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
2956 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null &&
2957 error "$ST: root unlink permission is granted"
2958 echo "$ST: root unlink permission is denied - ok"
2960 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null &&
2961 error "$ST: root create permission is granted"
2962 echo "$ST: root create permission is denied - ok"
2965 # check root_squash is enforced independently
2966 # of client cache content
2968 # access file by USER1, keep access open
2969 # root should be denied access to user file
2971 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
2975 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
2976 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
2977 { kill $pid; error "$ST: root read permission is granted"; }
2978 echo "$ST: root read permission is denied - ok"
2981 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
2982 { kill $pid; error "$ST: root write permission is granted"; }
2983 echo "$ST: root write permission is denied - ok"
2989 # check nosquash_nids:
2990 # put client's NID into nosquash_nids list,
2991 # root should be able to access root file after that
2993 local NIDLIST=$($LCTL list_nids all | tr '\n' ' ')
2994 NIDLIST="2@gni $NIDLIST 192.168.0.[2,10]@tcp"
2995 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
2996 set_conf_param_and_check mds \
2997 "$LCTL get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
2998 "$FSNAME-MDTall.mdt.nosquash_nids" \
3000 wait_update $HOSTNAME \
3001 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3003 error "check llite nosquash_nids failed!"
3005 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3006 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3007 error "$ST: root read permission is denied"
3008 echo "$ST: root read permission is granted - ok"
3011 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3012 error "$ST: root write permission is denied"
3013 echo "$ST: root write permission is granted - ok"
3015 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3016 rm $DIR/$tdir-rootdir/tfile-1 ||
3017 error "$ST: root unlink permission is denied"
3018 echo "$ST: root unlink permission is granted - ok"
3019 touch $DIR/$tdir-rootdir/tfile-2 ||
3020 error "$ST: root create permission is denied"
3021 echo "$ST: root create permission is granted - ok"
3023 run_test 43 "check root_squash and nosquash_nids"
3025 umount_client $MOUNT
3030 check_mount || error "check_mount"
3031 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
3033 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
3034 for VAL in $UUIDS; do
3035 NID=$(echo $VAL | cut -d= -f1)
3036 CLUUID=$(echo $VAL | cut -d= -f2)
3037 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
3039 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
3040 cleanup || error "cleanup failed with $?"
3042 run_test 44 "mounted client proc entry exists"
3046 check_mount || error "check_mount"
3047 stop_mds || error "Unable to stop MDS"
3051 #define OBD_FAIL_PTLRPC_LONG_UNLINK 0x50f
3052 do_facet client "$LCTL set_param fail_loc=0x50f"
3055 manual_umount_client --force || error "manual_umount_client failed"
3056 do_facet client "$LCTL set_param fail_loc=0x0"
3057 start_mds || error "unable to start MDS"
3058 mount_client $MOUNT || error "mount_client $MOUNT failed"
3059 cleanup || error "cleanup failed with $?"
3061 run_test 45 "long unlink handling in ptlrpcd"
3068 umount_client $MOUNT2 || rc=$?
3069 umount_client $MOUNT || rc=$?
3070 while [ $count -gt 0 ]; do
3071 stop ost${count} -f || rc=$?
3075 cleanup_nocli || rc=$?
3076 #writeconf to remove all ost2 traces for subsequent tests
3077 writeconf_or_reformat
3082 echo "Testing with $OSTCOUNT OSTs"
3084 start_mds || error "unable to start MDS"
3085 #first client should see only one ost
3086 start_ost || error "Unable to start OST1"
3087 wait_osc_import_state mds ost FULL
3089 mount_client $MOUNT || error "mount_client $MOUNT failed"
3090 trap "cleanup_46a $OSTCOUNT" EXIT ERR
3093 for (( i=2; i<=$OSTCOUNT; i++ )); do
3094 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
3095 error "start_ost$i $(ostdevname $i) failed"
3098 # wait until osts in sync
3099 for (( i=2; i<=$OSTCOUNT; i++ )); do
3100 wait_osc_import_state mds ost$i FULL
3101 wait_osc_import_state client ost$i FULL
3104 #second client see all ost's
3106 mount_client $MOUNT2 || error "mount_client failed"
3107 $SETSTRIPE -c -1 $MOUNT2 || error "$SETSTRIPE -c -1 $MOUNT2 failed"
3108 $GETSTRIPE $MOUNT2 || error "$GETSTRIPE $MOUNT2 failed"
3110 echo "ok" > $MOUNT2/widestripe
3111 $GETSTRIPE $MOUNT2/widestripe ||
3112 error "$GETSTRIPE $MOUNT2/widestripe failed"
3113 # fill acl buffer for avoid expand lsm to them
3114 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd |
3116 setfacl -m $acl $MOUNT2/widestripe
3120 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3122 cleanup_46a $OSTCOUNT || error "cleanup_46a failed"
3124 run_test 46a "handle ost additional - wide striped file"
3129 check_mount || error "check_mount failed"
3130 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
3134 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3135 if echo $ns | grep "MDT[[:digit:]]*"; then
3138 lrs=$(echo $ns | sed 's/.*lru_size=//')
3139 lru_size[count]=$lrs
3144 facet_failover $SINGLEMDS
3145 client_up || error "client_up failed"
3148 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3149 if echo $ns | grep "MDT[[:digit:]]*"; then
3152 lrs=$(echo $ns | sed 's/.*lru_size=//')
3153 if ! test "$lrs" -eq "${lru_size[count]}"; then
3154 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
3155 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
3160 cleanup || error "cleanup failed with $?"
3162 run_test 47 "server restart does not make client loss lru_resize settings"
3167 # reformat after this test is needed - if the test fails,
3168 # we will have unkillable file at FS
3172 test_48() { # bug 17636
3175 check_mount || error "check_mount failed"
3177 $SETSTRIPE -c -1 $MOUNT || error "$SETSTRIPE -c -1 $MOUNT failed"
3178 $GETSTRIPE $MOUNT || error "$GETSTRIPE $MOUNT failed"
3180 echo "ok" > $MOUNT/widestripe
3181 $GETSTRIPE $MOUNT/widestripe ||
3182 error "$GETSTRIPE $MOUNT/widestripe failed"
3184 trap cleanup_48 EXIT ERR
3186 # fill acl buffer for avoid expand lsm to them
3187 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
3188 setfacl -m $acl $MOUNT/widestripe
3191 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3195 run_test 48 "too many acls on file"
3197 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
3198 test_49a() { # bug 17710
3199 local timeout_orig=$TIMEOUT
3200 local ldlm_timeout_orig=$LDLM_TIMEOUT
3201 local LOCAL_TIMEOUT=20
3203 LDLM_TIMEOUT=$LOCAL_TIMEOUT
3204 TIMEOUT=$LOCAL_TIMEOUT
3208 check_mount || error "client mount failed"
3210 echo "check ldlm_timout..."
3211 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3212 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3213 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3215 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3216 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3219 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
3220 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
3223 umount_client $MOUNT || error "umount_client $MOUNT failed"
3224 stop_ost || error "problem stopping OSS"
3225 stop_mds || error "problem stopping MDS"
3227 LDLM_TIMEOUT=$ldlm_timeout_orig
3228 TIMEOUT=$timeout_orig
3230 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3232 test_49b() { # bug 17710
3233 local timeout_orig=$TIMEOUT
3234 local ldlm_timeout_orig=$LDLM_TIMEOUT
3235 local LOCAL_TIMEOUT=20
3237 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
3238 TIMEOUT=$LOCAL_TIMEOUT
3242 check_mount || error "client mount failed"
3244 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3245 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3246 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3248 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3249 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3252 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
3253 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
3256 cleanup || error "cleanup failed"
3258 LDLM_TIMEOUT=$ldlm_timeout_orig
3259 TIMEOUT=$timeout_orig
3261 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3264 # Test both statfs and lfs df and fail if either one fails
3265 multiop_bg_pause $1 f_
3268 killall -USR1 multiop
3269 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
3270 wait $PID || { RC1=$?; log "multiop return error "; }
3277 if [ $RC2 -eq 0 ]; then
3279 log "lazystatfs df failed"
3283 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
3289 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3290 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3292 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
3294 cleanup || error "cleanup failed with rc $?"
3296 run_test 50a "lazystatfs all servers available"
3300 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3301 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3303 # Wait for client to detect down OST
3304 stop_ost || error "Unable to stop OST1"
3305 wait_osc_import_state mds ost DISCONN
3307 lazystatfs $MOUNT || error "lazystatfs should not return EIO"
3309 umount_client $MOUNT || error "Unable to unmount client"
3310 stop_mds || error "Unable to stop MDS"
3312 run_test 50b "lazystatfs all servers down"
3315 start_mds || error "Unable to start MDS"
3316 start_ost || error "Unable to start OST1"
3317 start_ost2 || error "Unable to start OST2"
3318 mount_client $MOUNT || error "Unable to mount client"
3319 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3320 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3322 # Wait for client to detect down OST
3323 stop_ost || error "Unable to stop OST1"
3324 wait_osc_import_state mds ost DISCONN
3325 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3327 umount_client $MOUNT || error "Unable to unmount client"
3328 stop_ost2 || error "Unable to stop OST2"
3329 stop_mds || error "Unable to stop MDS"
3330 #writeconf to remove all ost2 traces for subsequent tests
3331 writeconf_or_reformat
3333 run_test 50c "lazystatfs one server down"
3336 start_mds || error "Unable to start MDS"
3337 start_ost || error "Unable to start OST1"
3338 start_ost2 || error "Unable to start OST2"
3339 mount_client $MOUNT || error "Unable to mount client"
3340 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3341 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3343 # Issue the statfs during the window where the client still
3344 # belives the OST to be available but it is in fact down.
3345 # No failure just a statfs which hangs for a timeout interval.
3346 stop_ost || error "Unable to stop OST1"
3347 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3349 umount_client $MOUNT || error "Unable to unmount client"
3350 stop_ost2 || error "Unable to stop OST2"
3351 stop_mds || error "Unable to stop MDS"
3352 #writeconf to remove all ost2 traces for subsequent tests
3353 writeconf_or_reformat
3355 run_test 50d "lazystatfs client/server conn race"
3362 start_mds || error "Unable to start MDS"
3363 #first client should see only one ost
3364 start_ost || error "Unable to start OST1"
3365 wait_osc_import_state mds ost FULL
3367 # Wait for client to detect down OST
3368 stop_ost || error "Unable to stop OST1"
3369 wait_osc_import_state mds ost DISCONN
3371 mount_client $MOUNT || error "Unable to mount client"
3372 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3374 multiop_bg_pause $MOUNT _f
3378 if [ $RC1 -ne 0 ]; then
3379 log "multiop failed $RC1"
3382 sleep $(( $TIMEOUT+1 ))
3384 [ $? -ne 0 ] && error "process isn't sleep"
3385 start_ost || error "Unable to start OST1"
3386 wait $pid || error "statfs failed"
3389 umount_client $MOUNT || error "Unable to unmount client"
3390 stop_ost || error "Unable to stop OST1"
3391 stop_mds || error "Unable to stop MDS"
3393 run_test 50e "normal statfs all servers down"
3398 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
3400 start_mds || error "Unable to start mds"
3401 #first client should see only one ost
3402 start_ost || error "Unable to start OST1"
3403 wait_osc_import_state mds ost FULL
3405 start_ost2 || error "Unable to start OST2"
3406 wait_osc_import_state mds ost2 FULL
3408 # Wait for client to detect down OST
3409 stop_ost2 || error "Unable to stop OST2"
3411 wait_osc_import_state mds ost2 DISCONN
3412 mount_client $MOUNT || error "Unable to mount client"
3413 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3415 multiop_bg_pause $MOUNT _f
3419 if [ $RC1 -ne 0 ]; then
3420 log "lazystatfs multiop failed $RC1"
3423 sleep $(( $TIMEOUT+1 ))
3425 [ $? -ne 0 ] && error "process isn't sleep"
3426 start_ost2 || error "Unable to start OST2"
3427 wait $pid || error "statfs failed"
3428 stop_ost2 || error "Unable to stop OST2"
3431 umount_client $MOUNT || error "Unable to unmount client"
3432 stop_ost || error "Unable to stop OST1"
3433 stop_mds || error "Unable to stop MDS"
3434 #writeconf to remove all ost2 traces for subsequent tests
3435 writeconf_or_reformat
3437 run_test 50f "normal statfs one server in down"
3440 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3442 start_ost2 || error "Unable to start OST2"
3443 wait_osc_import_state mds ost2 FULL
3444 wait_osc_import_state client ost2 FULL
3446 local PARAM="${FSNAME}-OST0001.osc.active"
3448 $SETSTRIPE -c -1 $DIR/$tfile || error "$SETSTRIPE failed"
3449 do_facet mgs $LCTL conf_param $PARAM=0 ||
3450 error "Unable to deactivate OST"
3452 umount_client $MOUNT || error "Unable to unmount client"
3453 mount_client $MOUNT || error "Unable to mount client"
3454 # This df should not cause a panic
3457 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
3458 rm -f $DIR/$tfile || error "unable to remove file $DIR/$tfile"
3459 umount_client $MOUNT || error "Unable to unmount client"
3460 stop_ost2 || error "Unable to stop OST2"
3461 stop_ost || error "Unable to stop OST1"
3462 stop_mds || error "Unable to stop MDS"
3463 #writeconf to remove all ost2 traces for subsequent tests
3464 writeconf_or_reformat
3466 run_test 50g "deactivated OST should not cause panic"
3470 # prepare MDT/OST, make OSC inactive for OST1
3471 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3473 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3474 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
3475 error "tunefs OST1 failed"
3476 start_mds || error "Unable to start MDT"
3477 start_ost || error "Unable to start OST1"
3478 start_ost2 || error "Unable to start OST2"
3479 mount_client $MOUNT || error "client start failed"
3481 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3483 # activatate OSC for OST1
3484 local TEST="$LCTL get_param -n osc.${FSNAME}-OST0000-osc-[!M]*.active"
3485 set_conf_param_and_check client \
3486 "$TEST" "${FSNAME}-OST0000.osc.active" 1 ||
3487 error "Unable to activate OST1"
3489 mkdir $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3490 $SETSTRIPE -c -1 -i 0 $DIR/$tdir/2 ||
3491 error "$SETSTRIPE $DIR/$tdir/2 failed"
3492 sleep 1 && echo "create a file after OST1 is activated"
3494 createmany -o $DIR/$tdir/2/$tfile-%d 1
3496 # check OSC import is working
3497 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
3498 error "some OSC imports are still not connected"
3501 umount_client $MOUNT || error "Unable to umount client"
3502 stop_ost2 || error "Unable to stop OST2"
3503 cleanup_nocli || error "cleanup_nocli failed with $?"
3505 run_test 50h "LU-642: activate deactivated OST"
3508 # prepare MDT/OST, make OSC inactive for OST1
3509 [ "$MDSCOUNT" -lt "2" ] && skip_env "$MDSCOUNT < 2, skipping" && return
3511 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3513 do_facet mds2 "$TUNEFS --param mdc.active=0 $(mdsdevname 2)" ||
3514 error "tunefs MDT2 failed"
3515 start_mds || error "Unable to start MDT"
3516 start_ost || error "Unable to start OST1"
3517 start_ost2 || error "Unable to start OST2"
3518 mount_client $MOUNT || error "client start failed"
3520 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3522 $LCTL conf_param ${FSNAME}-MDT0000.mdc.active=0 &&
3523 error "deactive MDC0 succeeds"
3524 # activate MDC for MDT2
3525 local TEST="$LCTL get_param -n mdc.${FSNAME}-MDT0001-mdc-[!M]*.active"
3526 set_conf_param_and_check client \
3527 "$TEST" "${FSNAME}-MDT0001.mdc.active" 1 ||
3528 error "Unable to activate MDT2"
3530 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
3531 if [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.60) ]
3533 wait_dne_interconnect
3535 $LFS mkdir -i1 $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3537 createmany -o $DIR/$tdir/2/$tfile-%d 1 || error "create files failed"
3539 rm -rf $DIR/$tdir/2 || error "unlink dir failed"
3541 # deactivate MDC for MDT2
3542 local TEST="$LCTL get_param -n mdc.${FSNAME}-MDT0001-mdc-[!M]*.active"
3543 set_conf_param_and_check client \
3544 "$TEST" "${FSNAME}-MDT0001.mdc.active" 0 ||
3545 error "Unable to deactivate MDT2"
3547 $LFS mkdir -i1 $DIR/$tdir/2 &&
3548 error "mkdir $DIR/$tdir/2 succeeds after deactive MDT"
3551 umount_client $MOUNT || error "Unable to umount client"
3556 run_test 50i "activate deactivated MDT"
3559 local LOCAL_TIMEOUT=20
3563 check_mount || error "check_mount failed"
3565 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
3566 $SETSTRIPE -c -1 $MOUNT/$tdir ||
3567 error "$SETSTRIPE -c -1 $MOUNT/$tdir failed"
3568 #define OBD_FAIL_MDS_REINT_DELAY 0x142
3569 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x142"
3570 touch $MOUNT/$tdir/$tfile &
3573 start_ost2 || error "Unable to start OST1"
3575 stop_ost2 || error "Unable to stop OST1"
3576 cleanup || error "cleanup failed with $?"
3577 #writeconf to remove all ost2 traces for subsequent tests
3578 writeconf_or_reformat
3580 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
3589 do_node $node mkdir -p $dest
3590 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3592 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
3593 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3594 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
3596 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
3597 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
3607 local backup2=${TMP}/backup2
3609 do_node $node mkdir -p $backup2
3610 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3612 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
3613 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3614 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
3616 do_node $node "diff -rq $backup $backup2"
3617 [ $? -eq 0 ] || { error "contents differ"; return 3; }
3619 local xattrs2=${TMP}/xattrs2
3620 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
3621 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
3623 do_node $node "diff $xattrs $xattrs2"
3624 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
3626 do_node $node "rm -rf $backup2 $xattrs2"
3627 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
3631 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3632 skip "Only applicable to ldiskfs-based MDTs"
3636 start_mds || error "Unable to start MDS"
3637 start_ost || error "Unable to start OST1"
3638 mount_client $MOUNT || error "Unable to mount client"
3641 local ost1mnt=$(facet_mntpt ost1)
3642 local ost1node=$(facet_active_host ost1)
3643 local ost1tmp=$TMP/conf52
3646 mkdir $DIR/$tdir || error "Unable to create $DIR/$tdir"
3647 touch $TMP/modified_first || error "Unable to create temporary file"
3648 local mtime=$(stat -c %Y $TMP/modified_first)
3649 do_node $ost1node "mkdir -p $ost1tmp &&
3650 touch -m -d @$mtime $ost1tmp/modified_first" ||
3651 error "Unable to create temporary file"
3654 $SETSTRIPE -c -1 -S 1M $DIR/$tdir || error "$SETSTRIPE failed"
3656 for (( i=0; i < nrfiles; i++ )); do
3657 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c ||
3658 error "multiop failed"
3664 echo backup files to $TMP/files
3665 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
3666 copy_files_xattrs $(hostname) $TMP/files $TMP/file_xattrs $files ||
3667 error "Unable to copy files"
3669 umount_client $MOUNT || error "Unable to umount client"
3670 stop_ost || error "Unable to stop ost1"
3672 echo mount ost1 as ldiskfs
3673 do_node $ost1node mkdir -p $ost1mnt || error "Unable to create $ost1mnt"
3674 if ! do_node $ost1node test -b $ost1_dev; then
3677 do_node $ost1node mount -t $(facet_fstype ost1) $loop $ost1_dev \
3679 error "Unable to mount ost1 as ldiskfs"
3682 echo backup objects to $ost1tmp/objects
3683 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
3684 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
3685 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
3687 error "Unable to copy objects"
3689 # move objects to lost+found
3690 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
3691 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
3693 # recover objects dry-run
3694 if [ $(lustre_version_code ost1) -ge $(version_code 2.5.56) ]; then
3695 echo "ll_recover_lost_found_objs dry_run"
3697 "ll_recover_lost_found_objs -n -d $ost1mnt/O" ||
3698 error "ll_recover_lost_found_objs failed"
3702 echo "ll_recover_lost_found_objs fix run"
3703 do_node $ost1node "ll_recover_lost_found_objs -d $ost1mnt/lost+found" ||
3704 error "ll_recover_lost_found_objs failed"
3706 # compare restored objects against saved ones
3707 diff_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
3708 [ $? -eq 0 ] || error "Unable to diff objects"
3710 do_node $ost1node "umount $ost1mnt" ||
3711 error "Unable to umount ost1 as ldiskfs"
3713 start_ost || error "Unable to start OST1"
3714 mount_client $MOUNT || error "Unable to mount client"
3717 diff_files_xattrs $(hostname) $TMP/files $TMP/file_xattrs $files ||
3718 error "Unable to diff files"
3720 rm -rf $TMP/files $TMP/file_xattrs ||
3721 error "Unable to delete temporary files"
3722 do_node $ost1node "rm -rf $ost1tmp" ||
3723 error "Unable to delete temporary files"
3724 cleanup || error "cleanup failed with $?"
3726 run_test 52 "check recovering objects from lost+found"
3728 # Checks threads_min/max/started for some service
3730 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
3731 # parameter pattern prefix like 'ost.*.ost'.
3744 local msg="Insane $modname thread counts"
3745 local ncpts=$(check_cpt_number $facet)
3749 check_mount || return 41
3751 # We need to expand $parampat, but it may match multiple parameters, so
3752 # we'll pick the first one
3753 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
3754 error "Couldn't expand ${parampat}.threads_min parameter name"
3758 # Remove the .threads_min part
3759 paramp=${paramp%.threads_min}
3761 # Check for sanity in defaults
3762 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3764 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3766 tstarted=$(do_facet $facet "$LCTL get_param \
3767 -n ${paramp}.threads_started" || echo 0)
3768 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' ||
3770 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
3772 nthrs=$(expr $tmax - $tmin)
3773 if [ $nthrs -lt $ncpts ]; then
3779 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
3780 skip_env "module parameter forced $facet thread count" &&
3781 tmin=3 && tmax=$((3 * tmax))
3783 # Check that we can change min/max
3784 do_facet $facet "$LCTL set_param \
3785 ${paramp}.threads_min=$((tmin + nthrs))"
3786 do_facet $facet "$LCTL set_param \
3787 ${paramp}.threads_max=$((tmax - nthrs))"
3788 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3790 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3792 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) &&
3793 $tmax2 == ($tmax - $nthrs)))' || return $?
3795 # Check that we can set min/max to the same value
3796 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3798 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$tmin"
3799 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3801 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3803 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
3805 # Check that we can't set max < min
3806 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$((tmin - 1))"
3807 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3809 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3811 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
3813 # We need to ensure that we get the module options desired; to do this
3814 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
3815 LOAD_MODULES_REMOTE=true
3818 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
3819 setmodopts -a $modname "$newvalue" oldvalue
3823 check_mount || return 41
3825 # Restore previous setting of MODOPTS_*
3826 setmodopts $modname "$oldvalue"
3828 # Check that $opts took
3829 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min")
3830 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max")
3831 tstarted=$(do_facet $facet \
3832 "$LCTL get_param -n ${paramp}.threads_started")
3833 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
3843 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
3844 cleanup || error "cleanup failed with rc $?"
3846 run_test 53a "check OSS thread count params"
3850 local mds=$(do_facet $SINGLEMDS "$LCTL get_param \
3851 -N mds.*.*.threads_max 2>/dev/null")
3852 if [ -z "$mds" ]; then
3853 #running this on an old MDT
3854 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
3856 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
3858 cleanup || error "cleanup failed with $?"
3860 run_test 53b "check MDS thread count params"
3863 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3864 skip "Only applicable to ldiskfs-based MDTs"
3868 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p ||
3869 error "llverdev failed with rc=$?"
3872 run_test 54a "test llverdev and partial verify of device"
3875 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3876 skip "Only applicable to ldiskfs-based MDTs"
3881 run_llverfs $MOUNT -p || error "llverfs failed with rc=$?"
3882 cleanup || error "cleanup failed with rc=$?"
3884 run_test 54b "test llverfs and partial verify of filesystem"
3888 local max_ost_index=$1
3889 echo -n $(((max_ost_index + 1) * 8))
3893 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3894 skip "Only applicable to ldiskfs-based MDTs"
3898 local mdsdev=$(mdsdevname 1)
3899 local mdsvdev=$(mdsvdevname 1)
3903 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
3905 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
3906 --reformat $(ostdevname 1) $(ostvdevname 1)
3912 echo checking size of lov_objid for ost index $i
3913 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
3914 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
3915 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
3917 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
3924 run_test 55 "check lov_objid size"
3927 local server_version=$(lustre_version_code $SINGLEMDS)
3928 local mds_journal_size_orig=$MDSJOURNALSIZE
3933 for num in $(seq 1 $MDSCOUNT); do
3936 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
3937 $(ostdevname 1) $(ostvdevname 1)
3938 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
3939 $(ostdevname 2) $(ostvdevname 2)
3942 start_ost || error "Unable to start first ost (idx 10000)"
3943 start_ost2 || error "Unable to start second ost (idx 1000)"
3944 mount_client $MOUNT || error "Unable to mount client"
3948 if [[ $server_version -ge $(version_code 2.6.54) ]] ||
3949 [[ $server_version -ge $(version_code 2.5.4) &&
3950 $server_version -lt $(version_code 2.5.11) ]]; then
3951 wait_osc_import_state mds ost1 FULL
3952 wait_osc_import_state mds ost2 FULL
3953 $SETSTRIPE --stripe-count=-1 $DIR/$tfile ||
3954 error "Unable to setstripe $DIR/$tfile"
3955 n=$($LFS getstripe --stripe-count $DIR/$tfile)
3956 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
3961 MDSJOURNALSIZE=$mds_journal_size_orig
3964 run_test 56 "check big OST indexes and out-of-index-order start"
3966 test_57a() { # bug 22656
3967 do_rpc_nodes $(facet_active_host ost1) load_modules_local
3968 local NID=$(do_facet ost1 "$LCTL get_param nis" |
3969 tail -1 | awk '{print $1}')
3970 writeconf_or_reformat
3971 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3972 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" ||
3973 error "tunefs failed"
3975 start_ost && error "OST registration from failnode should fail"
3978 run_test 57a "initial registration from failnode should fail (should return errs)"
3981 do_rpc_nodes $(facet_active_host ost1) load_modules_local
3982 local NID=$(do_facet ost1 "$LCTL get_param nis" |
3983 tail -1 | awk '{print $1}')
3984 writeconf_or_reformat
3985 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3986 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" ||
3987 error "tunefs failed"
3989 start_ost || error "OST registration from servicenode should not fail"
3992 run_test 57b "initial registration from servicenode should not fail"
3995 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
3998 test_58() { # bug 22658
4000 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4001 createmany -o $DIR/$tdir/$tfile-%d 100
4002 # make sure that OSTs do not cancel llog cookies before we unmount the MDS
4003 #define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601
4004 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x601"
4005 unlinkmany $DIR/$tdir/$tfile-%d 100
4006 stop_mds || error "Unable to stop MDS"
4008 local MNTDIR=$(facet_mntpt $SINGLEMDS)
4009 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4011 # remove all files from the OBJECTS dir
4012 mount_fstype $SINGLEMDS
4014 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
4016 unmount_fstype $SINGLEMDS
4017 # restart MDS with missing llog files
4018 start_mds || error "unable to start MDS"
4019 do_facet mds "$LCTL set_param fail_loc=0"
4022 run_test 58 "missing llog files must not prevent MDT from mounting"
4025 start_mgsmds >> /dev/null
4026 local C1=$(count_osts)
4027 if [ $C1 -eq 0 ]; then
4028 start_ost >> /dev/null
4032 echo "original ost count: $C1 (expect > 0)"
4033 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
4034 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
4035 local C2=$(count_osts)
4036 echo "after mdt writeconf count: $C2 (expect 0)"
4037 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
4038 echo "OST start without writeconf should fail:"
4039 start_ost >> /dev/null &&
4040 error "OST start without writeconf didn't fail"
4041 echo "OST start with writeconf should succeed:"
4042 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
4043 local C3=$(count_osts)
4044 echo "after ost writeconf count: $C3 (expect 1)"
4045 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
4046 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
4047 local C4=$(count_osts)
4048 echo "after ost2 writeconf count: $C4 (expect 2)"
4049 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
4050 stop_ost2 >> /dev/null
4051 cleanup_nocli >> /dev/null
4052 #writeconf to remove all ost2 traces for subsequent tests
4053 writeconf_or_reformat
4055 run_test 59 "writeconf mount option"
4057 test_60() { # LU-471
4060 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4061 skip "Only applicable to ldiskfs-based MDTs"
4065 for num in $(seq $MDSCOUNT); do
4066 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
4067 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
4068 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4072 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
4073 [ ${PIPESTATUS[0]} -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
4075 # MDT default has dirdata feature
4076 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
4077 # we disable uninit_bg feature
4078 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
4079 # we set stride extended options
4080 echo $dump | grep stride > /dev/null || error "stride is not set"
4083 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
4088 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] ||
4089 { skip "Need MDS version at least 2.1.53"; return 0; }
4091 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
4092 ! large_xattr_enabled; then
4095 for num in $(seq $MDSCOUNT); do
4096 do_facet mds${num} $TUNE2FS -O large_xattr \
4097 $(mdsdevname $num) ||
4098 error "tune2fs on mds $num failed"
4102 setup_noconfig || error "setting up the filesystem failed"
4103 client_up || error "starting client failed"
4105 local file=$DIR/$tfile
4106 touch $file || error "touch $file failed"
4108 local large_value="$(generate_string $(max_xattr_size))"
4109 local small_value="bar"
4111 local name="trusted.big"
4112 log "save large xattr $name on $file"
4113 setfattr -n $name -v $large_value $file ||
4114 error "saving $name on $file failed"
4116 local new_value=$(get_xattr_value $name $file)
4117 [[ "$new_value" != "$large_value" ]] &&
4118 error "$name different after saving"
4120 log "shrink value of $name on $file"
4121 setfattr -n $name -v $small_value $file ||
4122 error "shrinking value of $name on $file failed"
4124 new_value=$(get_xattr_value $name $file)
4125 [[ "$new_value" != "$small_value" ]] &&
4126 error "$name different after shrinking"
4128 log "grow value of $name on $file"
4129 setfattr -n $name -v $large_value $file ||
4130 error "growing value of $name on $file failed"
4132 new_value=$(get_xattr_value $name $file)
4133 [[ "$new_value" != "$large_value" ]] &&
4134 error "$name different after growing"
4136 log "check value of $name on $file after remounting MDS"
4138 new_value=$(get_xattr_value $name $file)
4139 [[ "$new_value" != "$large_value" ]] &&
4140 error "$name different after remounting MDS"
4142 log "remove large xattr $name from $file"
4143 setfattr -x $name $file || error "removing $name from $file failed"
4148 for num in $(seq $MDSCOUNT); do
4149 do_facet mds${num} $TUNE2FS -O ^large_xattr \
4150 $(mdsdevname $num) ||
4151 error "tune2fs on mds $num failed"
4155 run_test 61 "large xattr"
4158 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4159 skip "Only applicable to ldiskfs-based MDTs"
4164 local mdsdev=$(mdsdevname 1)
4165 local ostdev=$(ostdevname 1)
4167 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] ||
4168 { skip "Need MDS version at least 2.2.51"; return 0; }
4170 echo "disable journal for mds"
4171 do_facet mds $TUNE2FS -O ^has_journal $mdsdev || error "tune2fs failed"
4172 start_mds && error "MDT start should fail"
4173 echo "disable journal for ost"
4174 do_facet ost1 $TUNE2FS -O ^has_journal $ostdev || error "tune2fs failed"
4175 start_ost && error "OST start should fail"
4176 cleanup || error "cleanup failed with rc $?"
4179 run_test 62 "start with disabled journal"
4182 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4183 skip "Only applicable to ldiskfs-based MDTs"
4187 local inode_slab=$(do_facet $SINGLEMDS \
4188 "awk '/ldiskfs_inode_cache/ { print \\\$5 }' /proc/slabinfo")
4189 if [ -z "$inode_slab" ]; then
4190 skip "ldiskfs module has not been loaded"
4194 echo "$inode_slab ldisk inodes per page"
4195 [ "$inode_slab" -ge "3" ] ||
4196 error "ldisk inode size is too big, $inode_slab objs per page"
4199 run_test 63 "Verify each page can at least hold 3 ldisk inodes"
4202 start_mds || error "unable to start MDS"
4203 start_ost || error "Unable to start OST1"
4204 start_ost2 || error "Unable to start second ost"
4205 mount_client $MOUNT || error "Unable to mount client"
4206 stop_ost2 || error "Unable to stop second ost"
4208 $LFS df --lazy || error "lfs df failed"
4209 cleanup || error "cleanup failed with $?"
4210 #writeconf to remove all ost2 traces for subsequent tests
4211 writeconf_or_reformat
4213 run_test 64 "check lfs df --lazy "
4215 test_65() { # LU-2237
4216 # Currently, the test is only valid for ldiskfs backend
4217 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4218 skip "non-ldiskfs backend" && return
4220 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4221 local brpt=$(facet_mntpt brpt)
4224 if ! do_facet $SINGLEMDS "test -b $devname"; then
4228 stop_mds || error "Unable to stop MDS"
4229 local obj=$(do_facet $SINGLEMDS \
4230 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
4232 if [ -z "$obj" ]; then
4233 # The MDT may be just re-formatted, mount the MDT for the
4234 # first time to guarantee the "last_rcvd" file is there.
4235 start_mds || error "fail to mount the MDS for the first time"
4236 stop_mds || error "Unable to stop MDS"
4239 # remove the "last_rcvd" file
4240 do_facet $SINGLEMDS "mkdir -p $brpt"
4241 do_facet $SINGLEMDS \
4242 "mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
4243 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
4244 do_facet $SINGLEMDS "umount -d $brpt"
4246 # restart MDS, the "last_rcvd" file should be recreated.
4247 start_mds || error "fail to restart the MDS"
4248 stop_mds || error "Unable to stop MDS"
4249 obj=$(do_facet $SINGLEMDS \
4250 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
4251 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
4253 run_test 65 "re-create the lost last_rcvd file when server mount"
4256 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
4257 { skip "Need MGS version at least 2.3.59"; return 0; }
4260 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
4261 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
4263 echo "replace_nids should fail if MDS, OSTs and clients are UP"
4264 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4265 error "replace_nids fail"
4267 umount_client $MOUNT || error "unmounting client failed"
4268 echo "replace_nids should fail if MDS and OSTs are UP"
4269 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4270 error "replace_nids fail"
4272 stop_ost || error "Unable to stop OST1"
4273 echo "replace_nids should fail if MDS is UP"
4274 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4275 error "replace_nids fail"
4277 stop_mds || error "stopping mds failed"
4279 if combined_mgs_mds; then
4280 start_mdt 1 "-o nosvc" ||
4281 error "starting mds with nosvc option failed"
4284 echo "command should accept two parameters"
4285 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
4286 error "command should accept two params"
4288 echo "correct device name should be passed"
4289 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
4290 error "wrong devname"
4292 echo "wrong nids list should not destroy the system"
4293 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
4296 echo "replace OST nid"
4297 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
4298 error "replace nids failed"
4300 echo "command should accept two parameters"
4301 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
4302 error "command should accept two params"
4304 echo "wrong nids list should not destroy the system"
4305 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
4308 echo "replace MDS nid"
4309 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
4310 error "replace nids failed"
4312 if ! combined_mgs_mds ; then
4315 stop_mds || error "Unable to stop MDS"
4319 check_mount || error "error after nid replace"
4320 cleanup || error "cleanup failed"
4323 run_test 66 "replace nids"
4325 test_67() { #LU-2950
4326 local legacy="$TMP/legacy_lnet_config"
4327 local new="$TMP/new_routes_test"
4328 local out="$TMP/config_out_file"
4329 local verify="$TMP/conv_verify"
4330 local verify_conf="$TMP/conf_verify"
4332 # Create the legacy file that will be run through the
4333 # lustre_routes_conversion script
4334 cat <<- LEGACY_LNET_CONFIG > $legacy
4335 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
4336 tcp2 54 10.1.3.2@tcp;
4337 tcp3 10.3.4.3@tcp:3;
4341 # Create the verification file to verify the output of
4342 # lustre_routes_conversion script against.
4343 cat <<- VERIFY_LNET_CONFIG > $verify
4344 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
4345 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
4346 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
4347 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
4348 tcp4: { gateway: 10.3.3.4@tcp }
4351 # Create the verification file to verify the output of
4352 # lustre_routes_config script against
4353 cat <<- VERIFY_LNET_CONFIG > $verify_conf
4354 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
4355 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
4356 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
4357 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
4358 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
4361 lustre_routes_conversion $legacy $new > /dev/null
4362 if [ -f $new ]; then
4363 # verify the conversion output
4364 cmp -s $new $verify > /dev/null
4365 if [ $? -eq 1 ]; then
4366 error "routes conversion failed"
4369 lustre_routes_config --dry-run --verbose $new > $out
4370 # check that the script succeeded
4371 cmp -s $out $verify_conf > /dev/null
4372 if [ $? -eq 1 ]; then
4373 error "routes config failed"
4376 error "routes conversion test failed"
4378 # remove generated files
4379 rm -f $new $legacy $verify $verify_conf $out
4381 run_test 67 "test routes conversion and configuration"
4389 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.4.53) ] ||
4390 { skip "Need MDS version at least 2.4.53"; return 0; }
4392 umount_client $MOUNT || error "umount client failed"
4394 start_mdt 1 || error "MDT start failed"
4395 start_ost || error "Unable to start OST1"
4397 # START-END - the sequences we'll be reserving
4398 START=$(do_facet $SINGLEMDS \
4399 $LCTL get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
4400 END=$((START + (1 << 30)))
4401 do_facet $SINGLEMDS \
4402 $LCTL set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
4404 # reset the sequences MDT0000 has already assigned
4405 do_facet $SINGLEMDS \
4406 $LCTL set_param seq.srv*MDT0000.space=clear
4408 # remount to let the client allocate new sequence
4409 mount_client $MOUNT || error "mount client failed"
4411 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4412 do_facet $SINGLEMDS \
4413 $LCTL get_param seq.srv*MDT0000.space
4414 $LFS path2fid $DIR/$tfile
4416 local old_ifs="$IFS"
4418 fid=($($LFS path2fid $DIR/$tfile))
4422 if [[ $seq < $END ]]; then
4423 error "used reserved sequence $seq?"
4425 cleanup || error "cleanup failed with $?"
4427 run_test 68 "be able to reserve specific sequences in FLDB"
4429 # Test 69: is about the total number of objects ever created on an OST.
4430 # so that when it is reformatted the normal MDS->OST orphan recovery won't
4431 # just "precreate" the missing objects. In the past it might try to recreate
4432 # millions of objects after an OST was reformatted
4434 local server_version=$(lustre_version_code $SINGLEMDS)
4436 [[ $server_version -lt $(version_code 2.4.2) ]] &&
4437 skip "Need MDS version at least 2.4.2" && return
4439 [[ $server_version -ge $(version_code 2.4.50) ]] &&
4440 [[ $server_version -lt $(version_code 2.5.0) ]] &&
4441 skip "Need MDS version at least 2.5.0" && return
4444 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4446 # use OST0000 since it probably has the most creations
4447 local OSTNAME=$(ostname_from_index 0)
4448 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
4449 local last_id=$(do_facet mds1 $LCTL get_param -n \
4450 osc.$mdtosc_proc1.prealloc_last_id)
4452 # Want to have OST LAST_ID over 5 * OST_MAX_PRECREATE to
4453 # verify that the LAST_ID recovery is working properly. If
4454 # not, then the OST will refuse to allow the MDS connect
4455 # because the LAST_ID value is too different from the MDS
4456 #define OST_MAX_PRECREATE=20000
4457 local ost_max_pre=20000
4458 local num_create=$(( ost_max_pre * 5 + 1 - last_id))
4460 # If the LAST_ID is already over 5 * OST_MAX_PRECREATE, we don't
4461 # need to create any files. So, skip this section.
4462 if [ $num_create -gt 0 ]; then
4463 # Check the number of inodes available on OST0
4465 local ifree=$($LFS df -i $MOUNT | awk '/OST0000/ { print $4 }')
4466 log "On OST0, $ifree inodes available. Want $num_create."
4468 $SETSTRIPE -i 0 $DIR/$tdir ||
4469 error "$SETSTRIPE -i 0 $DIR/$tdir failed"
4470 if [ $ifree -lt 10000 ]; then
4471 files=$(( ifree - 50 ))
4476 local j=$((num_create / files + 1))
4477 for i in $(seq 1 $j); do
4478 createmany -o $DIR/$tdir/$tfile-$i- $files ||
4479 error "createmany fail create $files files: $?"
4480 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
4481 error "unlinkmany failed unlink $files files"
4485 # delete all of the files with objects on OST0 so the
4486 # filesystem is not inconsistent later on
4487 $LFS find $MOUNT --ost 0 -print0 | xargs -0 rm
4489 umount_client $MOUNT || error "umount client failed"
4490 stop_ost || error "OST0 stop failure"
4491 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
4492 $(ostdevname 1) $(ostvdevname 1) ||
4493 error "reformat and replace $ostdev failed"
4494 start_ost || error "OST0 restart failure"
4495 wait_osc_import_state mds ost FULL
4497 mount_client $MOUNT || error "mount client failed"
4498 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
4499 local idx=$($GETSTRIPE -i $DIR/$tdir/$tfile-last)
4500 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
4502 local iused=$($LFS df -i $MOUNT | awk '/OST0000/ { print $3 }')
4503 log "On OST0, $iused used inodes"
4504 [ $iused -ge $((ost_max_pre/2 + 1000)) ] &&
4505 error "OST replacement created too many inodes; $iused"
4506 cleanup || error "cleanup failed with $?"
4508 run_test 69 "replace an OST with the same index"
4511 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4514 cleanup || error "cleanup failed with $?"
4516 start_mdt 1 || error "MDT0 start fail"
4518 start_ost || error "OST0 start fail"
4519 for num in $(seq 2 $MDSCOUNT); do
4520 start_mdt $num || return
4523 mount_client $MOUNT || error "mount client fails"
4525 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
4527 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4528 error "create remote dir fail"
4530 rm -rf $DIR/$tdir || error "delete dir fail"
4531 cleanup || error "cleanup failed with $?"
4533 run_test 70a "start MDT0, then OST, then MDT1"
4536 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4539 start_ost || error "OST0 start fail"
4541 start_mds || error "MDS start fail"
4543 mount_client $MOUNT || error "mount client fails"
4545 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
4547 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4548 error "create remote dir fail"
4550 rm -rf $DIR/$tdir || error "delete dir fail"
4552 cleanup || error "cleanup failed with $?"
4554 run_test 70b "start OST, MDT1, MDT0"
4557 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4560 start_mds || error "MDS start fail"
4561 start_ost || error "OST0 start fail"
4563 mount_client $MOUNT || error "mount client fails"
4564 stop_mdt 1 || error "MDT1 start fail"
4566 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
4567 echo "deactivate $mdc_for_mdt1"
4568 $LCTL --device $mdc_for_mdt1 deactivate ||
4569 error "set $mdc_for_mdt1 deactivate failed"
4571 mkdir $DIR/$tdir && error "mkdir succeed"
4573 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4574 error "create remote dir succeed"
4576 cleanup || error "cleanup failed with $?"
4578 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
4581 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4584 start_mds || error "MDS start fail"
4585 start_ost || error "OST0 start fail"
4587 mount_client $MOUNT || error "mount client fails"
4589 stop_mdt 2 || error "MDT1 start fail"
4591 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
4593 echo "deactivate $mdc_for_mdt2"
4594 $LCTL --device $mdc_for_mdt2 deactivate ||
4595 error "set $mdc_for_mdt2 deactivate failed"
4597 mkdir $DIR/$tdir || error "mkdir fail"
4598 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4599 error "create remote dir succeed"
4601 rm -rf $DIR/$tdir || error "delete dir fail"
4603 cleanup || error "cleanup failed with $?"
4605 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
4608 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4609 if combined_mgs_mds; then
4610 skip "needs separate MGS/MDT" && return
4614 start_mdt 1 || error "MDT0 start fail"
4615 start_ost || error "OST0 start fail"
4616 for num in $(seq 2 $MDSCOUNT); do
4617 start_mdt $num || return
4620 start_ost2 || error "OST1 start fail"
4622 mount_client $MOUNT || error "mount client fails"
4624 mkdir $DIR/$tdir || error "mkdir fail"
4625 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4626 error "create remote dir succeed"
4628 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4629 rm -rf $DIR/$tdir || error "delete dir fail"
4631 umount_client $MOUNT || error "umount_client failed"
4632 stop_mds || error "MDS stop fail"
4633 stop_ost || error "OST0 stop fail"
4634 stop_ost2 || error "OST1 stop fail"
4636 run_test 71a "start MDT0 OST0, MDT1, OST1"
4639 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4640 if combined_mgs_mds; then
4641 skip "needs separate MGS/MDT" && return
4645 for num in $(seq 2 $MDSCOUNT); do
4646 start_mdt $num || return
4648 start_ost || error "OST0 start fail"
4649 start_mdt 1 || error "MDT0 start fail"
4650 start_ost2 || error "OST1 start fail"
4652 mount_client $MOUNT || error "mount client fails"
4654 mkdir $DIR/$tdir || error "mkdir fail"
4655 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4656 error "create remote dir succeed"
4658 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4659 rm -rf $DIR/$tdir || error "delete dir fail"
4661 umount_client $MOUNT || error "umount_client failed"
4662 stop_mds || error "MDT0 stop fail"
4663 stop_ost || error "OST0 stop fail"
4664 stop_ost2 || error "OST1 stop fail"
4666 run_test 71b "start MDT1, OST0, MDT0, OST1"
4669 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4670 if combined_mgs_mds; then
4671 skip "needs separate MGS/MDT" && return
4675 start_ost || error "OST0 start fail"
4676 start_ost2 || error "OST1 start fail"
4677 for num in $(seq 2 $MDSCOUNT); do
4678 start_mdt $num || return
4680 start_mdt 1 || error "MDT0 start fail"
4682 mount_client $MOUNT || error "mount client fails"
4684 mkdir $DIR/$tdir || error "mkdir fail"
4685 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4686 error "create remote dir succeed"
4688 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4689 rm -rf $DIR/$tdir || error "delete dir fail"
4691 umount_client $MOUNT || error "umount_client failed"
4692 stop_mds || error "MDS stop fail"
4693 stop_ost || error "OST0 stop fail"
4694 stop_ost2 || error "OST1 stop fail"
4697 run_test 71c "start OST0, OST1, MDT1, MDT0"
4700 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4701 if combined_mgs_mds; then
4702 skip "needs separate MGS/MDT" && return
4706 start_ost || error "OST0 start fail"
4707 for num in $(seq 2 $MDSCOUNT); do
4708 start_mdt $num || return
4710 start_mdt 1 || error "MDT0 start fail"
4711 start_ost2 || error "OST1 start fail"
4713 mount_client $MOUNT || error "mount client fails"
4715 mkdir $DIR/$tdir || error "mkdir fail"
4716 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4717 error "create remote dir succeed"
4719 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4720 rm -rf $DIR/$tdir || error "delete dir fail"
4722 umount_client $MOUNT || error "umount_client failed"
4723 stop_mds || error "MDS stop fail"
4724 stop_ost || error "OST0 stop fail"
4725 stop_ost2 || error "OST1 stop fail"
4728 run_test 71d "start OST0, MDT1, MDT0, OST1"
4731 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4732 if combined_mgs_mds; then
4733 skip "needs separate MGS/MDT" && return
4737 start_ost || error "OST0 start fail"
4738 for num in $(seq 2 $MDSCOUNT); do
4739 start_mdt $num || return
4741 start_ost2 || error "OST1 start fail"
4742 start_mdt 1 || error "MDT0 start fail"
4744 mount_client $MOUNT || error "mount client fails"
4746 mkdir $DIR/$tdir || error "mkdir fail"
4747 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4748 error "create remote dir succeed"
4750 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4751 rm -rf $DIR/$tdir || error "delete dir fail"
4753 umount_client $MOUNT || error "umount_client failed"
4754 stop_mds || error "MDS stop fail"
4755 stop_ost || error "OST0 stop fail"
4756 stop_ost2 || error "OST1 stop fail"
4759 run_test 71e "start OST0, MDT1, OST1, MDT0"
4761 test_72() { #LU-2634
4762 local mdsdev=$(mdsdevname 1)
4763 local ostdev=$(ostdevname 1)
4764 local cmd="$E2FSCK -fnvd $mdsdev"
4767 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4768 skip "ldiskfs only test" && return
4770 #tune MDT with "-O extents"
4772 for num in $(seq $MDSCOUNT); do
4773 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
4774 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4775 error "add mds $num failed"
4776 do_facet mds${num} "$TUNE2FS -O extents $(mdsdevname $num)" ||
4777 error "$TUNE2FS failed on mds${num}"
4780 add ost1 $(mkfs_opts ost1 $ostdev) --reformat $ostdev ||
4781 error "add $ostdev failed"
4782 start_mgsmds || error "start mds failed"
4783 start_ost || error "start ost failed"
4784 mount_client $MOUNT || error "mount client failed"
4786 #create some short symlinks
4787 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4788 createmany -o $DIR/$tdir/$tfile-%d $fn
4789 echo "create $fn short symlinks"
4790 for i in $(seq -w 1 $fn); do
4791 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
4796 umount_client $MOUNT || error "umount client failed"
4797 stop_mds || error "stop mds failed"
4798 stop_ost || error "stop ost failed"
4801 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
4803 run_test 72 "test fast symlink with extents flag enabled"
4805 test_73() { #LU-3006
4807 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4808 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
4809 error "1st tunefs failed"
4810 start_mgsmds || error "start mds failed"
4811 start_ost || error "start ost failed"
4812 mount_client $MOUNT || error "mount client failed"
4813 $LCTL get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
4814 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
4815 umount_client $MOUNT || error "umount client failed"
4819 run_test 73 "failnode to update from mountdata properly"
4821 test_75() { # LU-2374
4822 [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] &&
4823 skip "Need MDS version at least 2.4.1" && return
4826 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
4827 --reformat $(mdsdevname 1) $(mdsvdevname 1)"
4828 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
4829 --reformat $(ostdevname 1) $(ostvdevname 1)"
4831 #check with default parameters
4832 add mds1 $opts_mds || error "add mds1 failed for default params"
4833 add ost1 $opts_ost || error "add ost1 failed for default params"
4835 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
4836 opts_mds=$(echo $opts_mds |
4837 sed -e "s/--index=$index/--index=$index --mdt/")
4838 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
4839 opts_ost=$(echo $opts_ost |
4840 sed -e "s/--index=$index/--index=$index --ost/")
4842 add mds1 $opts_mds || error "add mds1 failed for new params"
4843 add ost1 $opts_ost || error "add ost1 failed for new params"
4846 run_test 75 "The order of --index should be irrelevant"
4849 [[ $(lustre_version_code mgs) -ge $(version_code 2.4.52) ]] ||
4850 { skip "Need MDS version at least 2.4.52" && return 0; }
4852 local MDMB_PARAM="osc.*.max_dirty_mb"
4853 echo "Change MGS params"
4854 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
4856 echo "max_dirty_mb: $MAX_DIRTY_MB"
4857 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB + MAX_DIRTY_MB))
4858 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
4859 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
4860 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
4861 head -1" $NEW_MAX_DIRTY_MB
4862 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
4863 echo "$MAX_DIRTY_MB"
4864 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
4865 error "error while apply max_dirty_mb"
4867 echo "Check the value is stored after remount"
4870 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
4871 head -1" $NEW_MAX_DIRTY_MB
4872 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
4873 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
4874 error "max_dirty_mb is not saved after remount"
4876 echo "Change OST params"
4877 CLIENT_PARAM="obdfilter.*.client_cache_count"
4878 local CLIENT_CACHE_COUNT
4879 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4881 echo "client_cache_count: $CLIENT_CACHE_COUNT"
4882 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
4883 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
4884 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
4885 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
4886 head -1" $NEW_CLIENT_CACHE_COUNT
4887 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4889 echo "$CLIENT_CACHE_COUNT"
4890 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
4891 error "error while apply client_cache_count"
4893 echo "Check the value is stored after remount"
4896 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
4897 head -1" $NEW_CLIENT_CACHE_COUNT
4898 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4900 echo "$CLIENT_CACHE_COUNT"
4901 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
4902 error "client_cache_count is not saved after remount"
4905 run_test 76a "set permanent params set_param -P"
4907 test_76b() { # LU-4783
4908 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.57) ]] ||
4909 { skip "Need MGS version at least 2.5.57" && return 0; }
4912 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
4913 error "start params log failed"
4916 run_test 76b "verify params log setup correctly"
4918 test_77() { # LU-3445
4919 local server_version=$(lustre_version_code $SINGLEMDS)
4921 [[ $server_version -ge $(version_code 2.2.60) ]] &&
4922 [[ $server_version -le $(version_code 2.4.0) ]] &&
4923 skip "Need MDS version < 2.2.60 or > 2.4.0" && return
4925 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
4926 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
4927 skip_env "mixed loopback and real device not working" && return
4930 local fs2mdsdev=$(mdsdevname 1_2)
4931 local fs2ostdev=$(ostdevname 1_2)
4932 local fs2mdsvdev=$(mdsvdevname 1_2)
4933 local fs2ostvdev=$(ostvdevname 1_2)
4934 local fsname=test1234
4936 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
4938 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
4939 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
4940 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
4941 error "start fs2mds failed"
4943 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
4944 [[ $mgsnid = *,* ]] || mgsnid+=",$mgsnid"
4946 add fs2ost $(mkfs_opts ost1 $fs2ostdev) --mgsnode=$mgsnid \
4947 --failnode=$failnid --fsname=$fsname \
4948 --reformat $fs2ostdev $fs2ostvdev ||
4949 error "add fs2ost failed"
4950 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
4952 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
4953 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
4954 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
4957 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
4960 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ||
4961 $(facet_fstype ost1) != ldiskfs ]] &&
4962 skip "only applicable to ldiskfs-based MDTs and OSTs" && return
4964 # reformat the Lustre filesystem with a smaller size
4965 local saved_MDSSIZE=$MDSSIZE
4966 local saved_OSTSIZE=$OSTSIZE
4967 MDSSIZE=$((MDSSIZE - 20000))
4968 OSTSIZE=$((OSTSIZE - 20000))
4969 reformat || error "(1) reformat Lustre filesystem failed"
4970 MDSSIZE=$saved_MDSSIZE
4971 OSTSIZE=$saved_OSTSIZE
4973 # mount the Lustre filesystem
4974 setup_noconfig || error "(2) setup Lustre filesystem failed"
4977 log "create test files"
4981 mkdir $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
4982 for i in $(seq $num_files); do
4983 file=$MOUNT/$tdir/$tfile-$i
4984 dd if=/dev/urandom of=$file count=1 bs=1M ||
4985 error "(4) create $file failed"
4988 # unmount the Lustre filesystem
4989 cleanup || error "(5) cleanup Lustre filesystem failed"
4991 # run e2fsck on the MDT and OST devices
4992 local mds_host=$(facet_active_host $SINGLEMDS)
4993 local ost_host=$(facet_active_host ost1)
4994 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
4995 local ost_dev=$(ostdevname 1)
4997 run_e2fsck $mds_host $mds_dev "-y"
4998 run_e2fsck $ost_host $ost_dev "-y"
5000 # get the original block count of the MDT and OST filesystems
5001 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
5002 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
5004 # expand the MDT and OST filesystems to the device size
5005 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
5006 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
5008 # run e2fsck on the MDT and OST devices again
5009 run_e2fsck $mds_host $mds_dev "-y"
5010 run_e2fsck $ost_host $ost_dev "-y"
5012 # mount the Lustre filesystem
5016 log "check files after expanding the MDT and OST filesystems"
5017 for i in $(seq $num_files); do
5018 file=$MOUNT/$tdir/$tfile-$i
5019 $CHECKSTAT -t file -s 1048576 $file ||
5020 error "(6) checkstat $file failed"
5024 log "create more files after expanding the MDT and OST filesystems"
5025 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
5026 file=$MOUNT/$tdir/$tfile-$i
5027 dd if=/dev/urandom of=$file count=1 bs=1M ||
5028 error "(7) create $file failed"
5031 # unmount the Lustre filesystem
5032 cleanup || error "(8) cleanup Lustre filesystem failed"
5034 # run e2fsck on the MDT and OST devices
5035 run_e2fsck $mds_host $mds_dev "-y"
5036 run_e2fsck $ost_host $ost_dev "-y"
5038 # get the maximum block count of the MDT and OST filesystems
5039 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
5040 local ost_max_blks=$(get_block_count ost1 $ost_dev)
5042 # get the minimum block count of the MDT and OST filesystems
5043 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
5044 grep minimum | sed -e 's/^.*filesystem: //g')
5045 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
5046 grep minimum | sed -e 's/^.*filesystem: //g')
5048 # shrink the MDT and OST filesystems to a smaller size
5052 if [[ $mds_max_blks -gt $mds_min_blks &&
5053 $mds_max_blks -gt $mds_orig_blks ]]; then
5054 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
5055 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
5056 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
5057 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
5058 error "shrink $SINGLEMDS to $new_blks failed"
5062 if [[ $ost_max_blks -gt $ost_min_blks &&
5063 $ost_max_blks -gt $ost_orig_blks ]]; then
5064 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
5065 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
5066 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
5067 run_resize2fs ost1 $ost_dev $new_blks ||
5068 error "shrink ost1 to $new_blks failed"
5072 # check whether the MDT or OST filesystem was shrunk or not
5074 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
5075 reformat || error "(10) reformat Lustre filesystem failed"
5079 # run e2fsck on the MDT and OST devices again
5080 run_e2fsck $mds_host $mds_dev "-y"
5081 run_e2fsck $ost_host $ost_dev "-y"
5083 # mount the Lustre filesystem again
5087 log "check files after shrinking the MDT and OST filesystems"
5088 for i in $(seq $((num_files + 10))); do
5089 file=$MOUNT/$tdir/$tfile-$i
5090 $CHECKSTAT -t file -s 1048576 $file ||
5091 error "(11) checkstat $file failed"
5094 # unmount and reformat the Lustre filesystem
5095 cleanup || error "(12) cleanup Lustre filesystem failed"
5096 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
5097 reformat || error "(14) reformat Lustre filesystem failed"
5099 run_test 78 "run resize2fs on MDT and OST filesystems"
5101 test_79() { # LU-4227
5102 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.5.59) ]] ||
5103 { skip "Need MDS version at least 2.5.59"; return 0; }
5105 local mdsdev1=$(mdsdevname 1)
5106 local mdsvdev1=$(mdsvdevname 1)
5107 local mdsdev2=$(mdsdevname 2)
5108 local mdsvdev2=$(mdsvdevname 2)
5109 local ostdev1=$(ostdevname 1)
5110 local ostvdev1=$(ostvdevname 1)
5111 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
5112 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
5113 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
5116 # remove --mgs/--mgsnode from mkfs.lustre options
5117 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
5119 mgsnode_opt=$(echo $opts_mds2 |
5120 awk '{ for ( i = 1; i < NF; i++ )
5121 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5122 [ -n $mgsnode_opt ] &&
5123 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
5125 mgsnode_opt=$(echo $opts_ost1 |
5126 awk '{ for ( i = 1; i < NF; i++ )
5127 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5128 [ -n $mgsnode_opt ] &&
5129 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
5131 # -MGS, format a mdt without --mgs option
5132 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
5133 error "Must specify --mgs when formatting mdt combined with mgs"
5135 # +MGS, format a mdt/ost without --mgsnode option
5136 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
5137 > /dev/null || error "start mds1 failed"
5138 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
5139 error "Must specify --mgsnode when formatting a mdt"
5140 add ost1 $opts_ost1 $ostdev1 $ostvdev2 &&
5141 error "Must specify --mgsnode when formatting an ost"
5145 run_test 79 "format MDT/OST without mgs option (should return errors)"
5148 start_mds || error "Failed to start MDT"
5149 start_ost || error "Failed to start OST1"
5150 uuid=$(do_facet ost1 $LCTL get_param -n mgc.*.uuid)
5151 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
5152 do_facet ost1 "$LCTL set_param fail_val=10 fail_loc=0x906"
5153 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x906"
5154 do_facet mgs "$LCTL set_param -n mgs/MGS/evict_client $uuid"
5156 start_ost2 || error "Failed to start OST2"
5158 do_facet ost1 "$LCTL set_param fail_loc=0"
5161 run_test 80 "mgc import reconnect race"
5163 #Save the original values of $OSTCOUNT and $OSTINDEX$i.
5165 local new_ostcount=$1
5166 saved_ostcount=$OSTCOUNT
5167 OSTCOUNT=$new_ostcount
5171 for i in $(seq $OSTCOUNT); do
5173 eval saved_ostindex$i=${!index}
5178 # Restore the original values of $OSTCOUNT and $OSTINDEX$i.
5179 restore_ostindex() {
5184 for i in $(seq $OSTCOUNT); do
5185 index=saved_ostindex$i
5186 eval OSTINDEX$i=${!index}
5188 OSTCOUNT=$saved_ostcount
5193 # The main purpose of this test is to ensure the OST_INDEX_LIST functions as
5194 # expected. This test uses OST_INDEX_LIST to format OSTs with a randomly
5195 # assigned index and ensures we can mount such a formatted file system
5196 test_81() { # LU-4665
5197 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5198 { skip "Need MDS version at least 2.6.54" && return; }
5199 [[ $OSTCOUNT -ge 3 ]] || { skip_env "Need at least 3 OSTs" && return; }
5203 # Each time RANDOM is referenced, a random integer between 0 and 32767
5206 local saved_ostindex1=$OSTINDEX1
5207 for i in 65535 $((RANDOM + 65536)); do
5208 echo -e "\nFormat ost1 with --index=$i, should fail"
5210 if add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat \
5211 $(ostdevname 1) $(ostvdevname 1); then
5212 OSTINDEX1=$saved_ostindex1
5213 error "format ost1 with --index=$i should fail"
5216 OSTINDEX1=$saved_ostindex1
5220 # Format OSTs with random sparse indices.
5221 trap "restore_ostindex" EXIT
5222 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices"
5223 OST_INDEX_LIST=[0,$((RANDOM * 2 % 65533 + 1)),65534] formatall
5225 # Setup and check Lustre filesystem.
5226 start_mgsmds || error "start_mgsmds failed"
5227 for i in $(seq $OSTCOUNT); do
5228 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5229 error "start ost$i failed"
5232 mount_client $MOUNT || error "mount client $MOUNT failed"
5233 check_mount || error "check client $MOUNT failed"
5236 local max_easize=$($LCTL get_param -n llite.*.max_easize)
5237 [[ $max_easize -eq 128 ]] ||
5238 error "max_easize is $max_easize, should be 128 bytes"
5242 run_test 81 "sparse OST indexing"
5244 # Wait OSTs to be active on both client and MDT side.
5246 local cmd="$LCTL get_param -n lov.$FSNAME-clilov-*.target_obd |
5247 awk 'BEGIN {c = 0} /ACTIVE/{c += 1} END {printf \\\"%d\\\", c}'"
5248 wait_update $HOSTNAME "eval $cmd" $OSTCOUNT ||
5249 error "wait_update OSTs up on client failed"
5251 cmd="$LCTL get_param -n lod.$FSNAME-MDT*-*.target_obd | sort -u |
5252 awk 'BEGIN {c = 0} /ACTIVE/{c += 1} END {printf \\\"%d\\\", c}'"
5253 wait_update_facet $SINGLEMDS "eval $cmd" $OSTCOUNT ||
5254 error "wait_update OSTs up on MDT failed"
5257 # Here we exercise the stripe placement functionality on a file system that
5258 # has formatted the OST with a random index. With the file system the following
5259 # functionality is tested:
5261 # 1. Creating a new file with a specific stripe layout.
5263 # 2. Modifiy a existing empty file with a specific stripe layout.
5265 # 3. Ensure we fail to set the stripe layout of a file that already has one.
5267 # 4. If ost-index is defined we need to ensure it is the first entry in the
5268 # ost index list returned by lfs getstripe.
5270 # 5. Lastly ensure this functionality fails with directories.
5271 test_82a() { # LU-4665
5272 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5273 { skip "Need MDS version at least 2.6.54" && return; }
5274 [[ $OSTCOUNT -ge 3 ]] || { skip_env "Need at least 3 OSTs" && return; }
5280 # Format OSTs with random sparse indices.
5284 for i in $(seq $OSTCOUNT); do
5285 index=$((RANDOM * 2))
5286 ost_indices+=" $index"
5288 ost_indices=$(comma_list $ost_indices)
5290 trap "restore_ostindex" EXIT
5291 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5292 OST_INDEX_LIST=[$ost_indices] formatall
5294 # Setup Lustre filesystem.
5295 start_mgsmds || error "start_mgsmds failed"
5296 for i in $(seq $OSTCOUNT); do
5297 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5298 error "start ost$i failed"
5301 mount_client $MOUNT || error "mount client $MOUNT failed"
5304 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5305 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5307 # 1. If the file does not exist, new file will be created
5308 # with specified OSTs.
5309 local file=$DIR/$tdir/$tfile-1
5310 local cmd="$SETSTRIPE -o $ost_indices $file"
5312 eval $cmd || error "$cmd failed"
5313 check_stripe_count $file $OSTCOUNT
5314 check_obdidx $file $ost_indices
5315 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5316 error "write $file failed"
5318 # 2. If the file already exists and is an empty file, the file
5319 # will be attached with specified layout.
5320 file=$DIR/$tdir/$tfile-2
5321 mcreate $file || error "mcreate $file failed"
5322 cmd="$SETSTRIPE -o $ost_indices $file"
5324 eval $cmd || error "$cmd failed"
5325 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5326 error "write $file failed"
5327 check_stripe_count $file $OSTCOUNT
5328 check_obdidx $file $ost_indices
5330 # 3. If the file already has a valid layout attached, the command
5331 # should fail with EBUSY.
5333 eval $cmd && error "stripe is already set on $file, $cmd should fail"
5335 # 4. If [--stripe-index|-i <start_ost_idx>] is used, the index must
5336 # be in the OST indices list.
5337 local start_ost_idx=${ost_indices##*,}
5338 file=$DIR/$tdir/$tfile-3
5339 cmd="$SETSTRIPE -o $ost_indices -i $start_ost_idx $file"
5341 eval $cmd || error "$cmd failed"
5342 check_stripe_count $file $OSTCOUNT
5343 check_obdidx $file $ost_indices
5344 check_start_ost_idx $file $start_ost_idx
5346 file=$DIR/$tdir/$tfile-4
5348 cmd+=" -o $(exclude_items_from_list $ost_indices $start_ost_idx)"
5349 cmd+=" -i $start_ost_idx $file"
5351 eval $cmd && error "index $start_ost_idx should be in $ost_indices"
5353 # 5. Specifying OST indices for directory should fail with ENOSUPP.
5354 local dir=$DIR/$tdir/$tdir
5355 mkdir $dir || error "mkdir $dir failed"
5356 cmd="$SETSTRIPE -o $ost_indices $dir"
5358 eval $cmd && error "$cmd should fail, specifying OST indices" \
5359 "for directory is not supported"
5363 run_test 82a "specify OSTs for file (succeed) or directory (fail)"
5368 # Remove OSTs from a pool and destroy the pool.
5369 destroy_pool $ost_pool || true
5374 # Test 82b is run to ensure that if the user supplies a pool with a specific
5375 # stripe layout that it behaves proprerly. It should fail in the case that
5376 # the supplied OST index list points to OSTs not contained in the user
5378 test_82b() { # LU-4665
5379 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5380 { skip "Need MDS version at least 2.6.54" && return; }
5381 [[ $OSTCOUNT -ge 4 ]] || { skip_env "Need at least 4 OSTs" && return; }
5387 # Format OSTs with random sparse indices.
5391 for i in $(seq $OSTCOUNT); do
5392 index=$((RANDOM * 2))
5393 ost_indices+=" $index"
5395 ost_indices=$(comma_list $ost_indices)
5397 trap "restore_ostindex" EXIT
5398 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5399 OST_INDEX_LIST=[$ost_indices] formatall
5401 # Setup Lustre filesystem.
5402 start_mgsmds || error "start_mgsmds failed"
5403 for i in $(seq $OSTCOUNT); do
5404 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5405 error "start ost$i failed"
5408 mount_client $MOUNT || error "mount client $MOUNT failed"
5410 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5411 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5413 # Create a new pool and add OSTs into it.
5414 local ost_pool=$FSNAME.$TESTNAME
5415 create_pool $ost_pool || error "create OST pool $ost_pool failed"
5418 trap "cleanup_82b" EXIT
5420 local ost_idx_in_list=${ost_indices##*,}
5421 local ost_idx_in_pool=$(exclude_items_from_list $ost_indices \
5424 local ost_targets="$FSNAME-OST["
5425 for i in ${ost_idx_in_pool//,/ }; do
5426 ost_targets=$ost_targets$(printf "%04x," $i)
5428 ost_targets="${ost_targets%,}]"
5430 local ost_targets_uuid=$(for i in ${ost_idx_in_pool//,/ }; \
5431 do printf "$FSNAME-OST%04x_UUID\n" $i; done |
5432 sort -u | tr '\n' ' ')
5434 local cmd="$LCTL pool_add $ost_pool $ost_targets"
5435 do_facet mgs $cmd || error "$cmd failed"
5436 wait_update $HOSTNAME "$LCTL get_param -n lov.$FSNAME-*.pools.$TESTNAME|
5437 sort -u | tr '\n' ' ' " "$ost_targets_uuid" ||
5438 error "wait_update $ost_pool failed"
5439 pool_list $ost_pool || error "list OST pool $ost_pool failed"
5441 # If [--pool|-p <pool_name>] is set with [--ost-list|-o <ost_indices>],
5442 # then the OSTs must be the members of the pool.
5443 local file=$DIR/$tdir/$tfile
5444 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
5446 eval $cmd && error "OST with index $ost_idx_in_list should be" \
5447 "in OST pool $ost_pool"
5449 # Only select OST $ost_idx_in_list from $ost_pool for file.
5450 ost_idx_in_list=${ost_idx_in_pool#*,}
5451 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
5453 eval $cmd || error "$cmd failed"
5454 cmd="$GETSTRIPE $file"
5456 eval $cmd || error "$cmd failed"
5457 check_stripe_count $file 2
5458 check_obdidx $file $ost_idx_in_list
5459 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5460 error "write $file failed"
5464 run_test 82b "specify OSTs for file with --pool and --ost-list options"
5467 [[ $(lustre_version_code ost1) -ge $(version_code 2.6.91) ]] ||
5468 { skip "Need OST version at least 2.6.91" && return 0; }
5469 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
5470 skip "Only applicable to ldiskfs-based MDTs"
5480 ostmnt=$(facet_mntpt ost1)
5481 fstype=$(facet_fstype ost1)
5483 # Mount the OST as an ldiskfs filesystem.
5484 log "mount the OST $dev as a $fstype filesystem"
5485 add ost1 $(mkfs_opts ost1 $dev) $FSTYPE_OPT \
5486 --reformat $dev $dev > /dev/null ||
5487 error "format ost1 error"
5489 if ! test -b $dev; then
5490 mnt_opts=$(csa_add "$OST_MOUNT_OPTS" -o loop)
5492 echo "mnt_opts $mnt_opts"
5493 do_facet ost1 mount -t $fstype $dev \
5495 # Run llverfs on the mounted ldiskfs filesystem.
5496 # It is needed to get ENOSPACE.
5497 log "run llverfs in partial mode on the OST $fstype $ostmnt"
5498 do_rpc_nodes $(facet_host ost1) run_llverfs $ostmnt -vpl \
5499 "no" || error "run_llverfs error on $fstype"
5502 log "unmount the OST $dev"
5505 # Delete file IO_scrub. Later osd_scrub_setup will try to
5506 # create "IO_scrub" but will get ENOSPACE.
5508 echo "start ost1 service on `facet_active_host ost1`"
5509 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
5512 err=$(do_facet ost1 dmesg | grep "VFS: Busy inodes after unmount of")
5513 echo "string err $err"
5514 [ -z "$err" ] || error $err
5517 run_test 83 "ENOSPACE on OST doesn't cause message VFS: \
5518 Busy inodes after unmount ..."
5520 recovery_time_min() {
5521 local CONNECTION_SWITCH_MIN=5
5522 local CONNECTION_SWITCH_INC=5
5523 local CONNECTION_SWITCH_MAX
5524 local RECONNECT_DELAY_MAX
5525 local INITIAL_CONNECT_TIMEOUT
5529 #CONNECTION_SWITCH_MAX=min(50, max($CONNECTION_SWITCH_MIN,$TIMEOUT)
5530 (($CONNECTION_SWITCH_MIN>$TIMEOUT)) && \
5531 max=$CONNECTION_SWITCH_MIN || max=$TIMEOUT
5532 (($max<50)) && CONNECTION_SWITCH_MAX=$max || CONNECTION_SWITCH_MAX=50
5534 #INITIAL_CONNECT_TIMEOUT = max(CONNECTION_SWITCH_MIN, \
5536 TO_20=$(($TIMEOUT/20))
5537 (($CONNECTION_SWITCH_MIN>$TO_20)) && \
5538 INITIAL_CONNECT_TIMEOUT=$CONNECTION_SWITCH_MIN || \
5539 INITIAL_CONNECT_TIMEOUT=$TO_20
5541 RECONNECT_DELAY_MAX=$(($CONNECTION_SWITCH_MAX+$CONNECTION_SWITCH_INC+ \
5542 $INITIAL_CONNECT_TIMEOUT))
5543 echo $((2*$RECONNECT_DELAY_MAX))
5547 local facet=$SINGLEMDS
5548 local num=$(echo $facet | tr -d "mds")
5549 local dev=$(mdsdevname $num)
5550 local time_min=$(recovery_time_min)
5551 local recovery_duration
5552 local completed_clients
5553 local correct_clients
5557 echo "start mds service on $(facet_active_host $facet)"
5559 "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
5560 error "start MDS failed"
5565 echo "recovery_time=$time_min, timeout=$TIMEOUT, wrap_up=$wrap_up"
5567 mount_client $MOUNT1 || error "mount failed"
5568 mount_client $MOUNT2 || error "mount failed"
5570 replay_barrier $SINGLEMDS
5571 createmany -o $DIR1/$tfile-%d 1000
5573 # We need to catch the end of recovery window to extend it.
5574 # Skip 5 requests and add delay to request handling.
5575 #define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP
5576 do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5"
5578 facet_failover $SINGLEMDS || error "failover: $?"
5581 echo "recovery status"
5582 do_facet $SINGLEMDS \
5583 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status"
5585 recovery_duration=$(do_facet $SINGLEMDS \
5586 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
5587 awk '/recovery_duration/ { print $2 }')
5588 (( $recovery_duration > $time_min + $wrap_up )) &&
5589 error "recovery_duration > recovery_time_hard + wrap up"
5590 completed_clients=$(do_facet $SINGLEMDS \
5591 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
5592 awk '/completed_clients/ { print $2 }')
5594 correct_clients="$MDSCOUNT/$((MDSCOUNT+1))"
5595 [ "$completed_clients" = "${correct_clients}" ] ||
5596 error "$completed_clients != $correct_clients"
5598 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
5599 umount_client $MOUNT1
5600 umount_client $MOUNT2
5606 run_test 84 "check recovery_hard_time"
5609 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.55) ]] ||
5610 { skip "Need OST version at least 2.7.55" && return 0; }
5611 ##define OBD_FAIL_OSD_OST_EA_FID_SET 0x197
5612 do_facet ost1 "lctl set_param fail_loc=0x197"
5616 run_test 85 "osd_ost init: fail ea_fid_set"
5619 [ "$(facet_fstype ost1)" = "zfs" ] &&
5620 skip "LU-6442: no such mkfs params for ZFS OSTs" && return
5622 local OST_OPTS="$(mkfs_opts ost1 $(ostdevname 1)) \
5623 --reformat $(ostdevname 1) $(ostvdevname 1)"
5626 local OLDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
5627 awk '/Flex block group size: / { print $NF; exit; }')
5630 if [[ ${!opts} != *mkfsoptions* ]]; then
5631 eval opts=\"${!opts} \
5632 --mkfsoptions='\\\"-O flex_bg -G $NEWSIZE\\\"'\"
5634 val=${!opts//--mkfsoptions=\\\"/ \
5635 --mkfsoptions=\\\"-O flex_bg -G $NEWSIZE }
5639 echo "params: $opts"
5641 add ost1 $opts || error "add ost1 failed with new params"
5643 local FOUNDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
5644 awk '/Flex block group size: / { print $NF; exit; }')
5646 [[ $FOUNDSIZE == $NEWSIZE ]] ||
5647 error "Flex block group size: $FOUNDSIZE, expected: $NEWSIZE"
5650 run_test 86 "Replacing mkfs.lustre -G option"
5652 test_87() { #LU-6544
5653 [[ $(lustre_version_code $SINGLEMDS1) -ge $(version_code 2.7.56) ]] ||
5654 { skip "Need MDS version at least 2.7.56" && return; }
5655 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ]] &&
5656 { skip "Only applicable to ldiskfs-based MDTs" && return; }
5657 [[ $OSTCOUNT -gt 69 ]] &&
5658 { skip "Ignore wide striping situation" && return; }
5660 local mdsdev=$(mdsdevname 1)
5661 local mdsvdev=$(mdsvdevname 1)
5662 local file=$DIR/$tfile
5663 local mntpt=$(facet_mntpt $SINGLEMDS)
5664 local used_xattr_blk=0
5665 local inode_size=${1:-512}
5667 local xtest="trusted.test"
5672 #Please see LU-6544 for MDT inode size calculation
5673 if [ $OSTCOUNT -gt 26 ]; then
5675 elif [ $OSTCOUNT -gt 5 ]; then
5678 left_size=$(expr $inode_size - \
5681 32 - $OSTCOUNT \* 24 - 16 - 3 - \
5683 24 - 18 - $(expr length $tfile) - 16 - 4)
5684 if [ $left_size -le 0 ]; then
5685 echo "No space($left_size) is expected in inode."
5686 echo "Try 1-byte xattr instead to verify this."
5689 echo "Estimate: at most $left_size-byte space left in inode."
5695 add mds1 $(mkfs_opts mds1 ${mdsdev}) --stripe-count-hint=$OSTCOUNT \
5696 --reformat $mdsdev $mdsvdev || error "add mds1 failed"
5697 start_mdt 1 > /dev/null || error "start mdt1 failed"
5698 for i in $(seq $OSTCOUNT); do
5699 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS > /dev/null ||
5700 error "start ost$i failed"
5702 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
5703 check_mount || error "check client $MOUNT failed"
5706 $SETSTRIPE -c -1 $file || error "$SETSTRIPE -c -1 $file failed"
5707 $GETSTRIPE $file || error "$GETSTRIPE $file failed"
5708 i=$($GETSTRIPE -c $file)
5709 if [ $i -ne $OSTCOUNT ]; then
5710 left_size=$(expr $left_size + $(expr $OSTCOUNT - $i) \* 24)
5711 echo -n "Since only $i out $OSTCOUNT OSTs are used, "
5712 echo -n "the expected left space is changed to "
5713 echo "$left_size bytes at most."
5715 value=$(generate_string $left_size)
5716 setfattr -n $xtest -v $value $file
5717 orig=$(get_xattr_value $xtest $file)
5718 [[ "$orig" != "$value" ]] && error "$xtest changed"
5720 #Verify if inode has some expected space left
5721 umount $MOUNT > /dev/null || error "umount $MOUNT failed"
5722 stop_mdt 1 > /dev/null || error "stop mdt1 failed"
5723 mount_ldiskfs $SINGLEMDS || error "mount -t ldiskfs $SINGLEMDS failed"
5725 do_facet $SINGLEMDS ls -sal $mntpt/ROOT/$tfile
5726 used_xattr_blk=$(do_facet $SINGLEMDS ls -s $mntpt/ROOT/$tfile |
5728 [[ $used_xattr_blk -eq 0 ]] &&
5729 error "Please check MDS inode size calculation: \
5730 more than $left_size-byte space left in inode."
5731 echo "Verified: at most $left_size-byte space left in inode."
5735 run_test 87 "check if MDT inode can hold EAs with N stripes properly"
5738 # $2 (optional) value of max_mod_rpcs_in_flight to set
5739 check_max_mod_rpcs_in_flight() {
5747 idx=$(printf "%04x" $($LFS getdirstripe -i $dir))
5748 facet="mds$((0x$idx + 1))"
5750 if [ -z "$mmr" ]; then
5751 # get value of max_mod_rcps_in_flight
5752 mmr=$($LCTL get_param -n \
5753 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) ||
5754 error "Unable to get max_mod_rpcs_in_flight"
5755 echo "max_mod_rcps_in_flight is $mmr"
5757 # set value of max_mod_rpcs_in_flight
5759 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight=$mmr ||
5760 error "Unable to set max_mod_rpcs_in_flight to $mmr"
5761 echo "max_mod_rpcs_in_flight set to $mmr"
5764 # create mmr+1 files
5765 echo "creating $((mmr + 1)) files ..."
5767 for i in $(seq $((mmr + 1))); do
5773 # consumes mmr-1 modify RPC slots
5774 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
5775 # drop requests on MDT so that RPC slots are consumed
5776 # during all the request resend interval
5777 do_facet $facet "$LCTL set_param fail_loc=0x159"
5778 echo "launch $((mmr - 1)) chmod in parallel ..."
5779 for i in $(seq $((mmr - 1))); do
5780 chmod 0600 $dir/file-$i &
5784 # send one additional modify RPC
5785 do_facet $facet "$LCTL set_param fail_loc=0"
5786 echo "launch 1 additional chmod in parallel ..."
5787 chmod 0600 $dir/file-$mmr &
5790 # check this additional modify RPC get a modify RPC slot
5791 # and succeed its operation
5792 checkstat -vp 0600 $dir/file-$mmr ||
5793 error "Unable to send $mmr modify RPCs in parallel"
5798 # consumes mmr modify RPC slots
5799 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
5800 # drop requests on MDT so that RPC slots are consumed
5801 # during all the request resend interval
5802 do_facet $facet "$LCTL set_param fail_loc=0x159"
5803 echo "launch $mmr chmod in parallel ..."
5804 for i in $(seq $mmr); do
5805 chmod 0666 $dir/file-$i &
5809 # send one additional modify RPC
5810 do_facet $facet "$LCTL set_param fail_loc=0"
5811 echo "launch 1 additional chmod in parallel ..."
5812 chmod 0666 $dir/file-$((mmr + 1)) &
5815 # check this additional modify RPC blocked getting a modify RPC slot
5816 checkstat -vp 0644 $dir/file-$((mmr + 1)) ||
5817 error "Unexpectedly send $(($mmr + 1)) modify RPCs in parallel"
5823 if ! combined_mgs_mds ; then
5828 [[ $($LCTL get_param mdc.*.import |
5829 grep "connect_flags:.*multi_mod_rpc") ]] ||
5830 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
5832 # check default value
5833 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5834 check_max_mod_rpcs_in_flight $DIR/$tdir
5838 run_test 90a "check max_mod_rpcs_in_flight is enforced"
5848 [[ $($LCTL get_param mdc.*.import |
5849 grep "connect_flags:.*multi_mod_rpc") ]] ||
5850 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
5853 # update max_mod_rpcs_in_flight
5854 $LFS mkdir -c1 $DIR/${tdir}1 || error "mkdir $DIR/${tdir}1 failed"
5855 check_max_mod_rpcs_in_flight $DIR/${tdir}1 1
5858 # check client is able to send multiple modify RPCs in paralell
5859 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
5860 grep -c "multi_mod_rpcs")
5861 if [ "$tmp" -ne $MDSCOUNT ]; then
5862 echo "Client not able to send multiple modify RPCs in parallel"
5867 # update max_mod_rpcs_in_flight
5868 $LFS mkdir -c1 $DIR/${tdir}2 || error "mkdir $DIR/${tdir}2 failed"
5869 check_max_mod_rpcs_in_flight $DIR/${tdir}2 5
5872 $LFS mkdir -c1 $DIR/${tdir}3 || error "mkdir $DIR/${tdir}3 failed"
5873 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/${tdir}3))
5874 facet="mds$((0x$idx + 1))"
5876 # save MDT max_mod_rpcs_per_client
5877 mmrpc=$(do_facet $facet \
5878 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
5880 # update max_mod_rpcs_in_flight
5881 umount_client $MOUNT
5883 "echo 16 > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
5885 $LCTL set_param mdc.$FSNAME-MDT$idx-mdc-*.max_rpcs_in_flight=17
5886 check_max_mod_rpcs_in_flight $DIR/${tdir}3 16
5888 # restore MDT max_mod_rpcs_per_client initial value
5890 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
5892 rm -rf $DIR/${tdir}?
5895 run_test 90b "check max_mod_rpcs_in_flight is enforced after update"
5904 [[ $($LCTL get_param mdc.*.import |
5905 grep "connect_flags:.*multi_mod_rpc") ]] ||
5906 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
5908 # check client is able to send multiple modify RPCs in paralell
5909 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
5910 grep -c "multi_mod_rpcs")
5911 if [ "$tmp" -ne $MDSCOUNT ]; then
5912 skip "Client not able to send multiple modify RPCs in parallel"
5917 # get max_rpcs_in_flight value
5918 mrif=$($LCTL get_param -n mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
5919 echo "max_rpcs_in_flight is $mrif"
5921 # get MDT max_mod_rpcs_per_client
5922 mmrpc=$(do_facet mds1 \
5923 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
5924 echo "max_mod_rpcs_per_client is $mmrpc"
5927 # attempt to set max_mod_rpcs_in_flight to max_rpcs_in_flight value
5928 # prerequisite: set max_mod_rpcs_per_client to max_rpcs_in_flight value
5929 umount_client $MOUNT
5931 "echo $mrif > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
5935 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif &&
5936 error "set max_mod_rpcs_in_flight to $mrif should fail"
5938 umount_client $MOUNT
5940 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
5944 # attempt to set max_mod_rpcs_in_flight to max_mod_rpcs_per_client+1
5945 # prerequisite: set max_rpcs_in_flight to max_mod_rpcs_per_client+2
5947 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$((mmrpc + 2))
5950 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$((mmrpc + 1)) &&
5951 error "set max_mod_rpcs_in_flight to $((mmrpc + 1)) should fail"
5955 run_test 90c "check max_mod_rpcs_in_flight update limits"
5966 [[ $($LCTL get_param mdc.*.import |
5967 grep "connect_flags:.*multi_mod_rpc") ]] ||
5968 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
5970 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5971 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
5972 facet="mds$((0x$idx + 1))"
5974 # check client version supports multislots
5975 tmp=$($LCTL get_param -N \
5976 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
5977 if [ -z "$tmp" ]; then
5978 skip "Client does not support multiple modify RPCs in flight"
5983 # get current value of max_mod_rcps_in_flight
5984 mmr=$($LCTL get_param -n \
5985 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
5986 echo "max_mod_rcps_in_flight is $mmr"
5989 echo "creating $mmr files ..."
5991 for i in $(seq $mmr); do
5992 touch $DIR/$tdir/file-$i
5995 # prepare for close RPC
5996 multiop_bg_pause $DIR/$tdir/file-close O_c
5999 # consumes mmr modify RPC slots
6000 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6001 # drop requests on MDT so that RPC slots are consumed
6002 # during all the request resend interval
6003 do_facet $facet "$LCTL set_param fail_loc=0x159"
6004 echo "launch $mmr chmod in parallel ..."
6005 for i in $(seq $mmr); do
6006 chmod 0600 $DIR/$tdir/file-$i &
6009 # send one additional close RPC
6010 do_facet $facet "$LCTL set_param fail_loc=0"
6011 echo "launch 1 additional close in parallel ..."
6013 cancel_lru_locks mdc
6016 # check this additional close RPC get a modify RPC slot
6017 # and multiop process completed
6018 [ -d /proc/$pid ] &&
6019 error "Unable to send the additional close RPC in parallel"
6024 run_test 90d "check one close RPC is allowed above max_mod_rpcs_in_flight"
6026 if ! combined_mgs_mds ; then
6032 # restore the values of MDSSIZE and OSTSIZE
6033 MDSSIZE=$STORED_MDSSIZE
6034 OSTSIZE=$STORED_OSTSIZE