7 # bug number for skipped test: LU-8972
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/$NAME.sh}
54 # use small MDS + OST size to speed formatting time
55 # do not use too small MDSSIZE/OSTSIZE, which affect the default journal size
56 # STORED_MDSSIZE is used in test_18
57 STORED_MDSSIZE=$MDSSIZE
58 STORED_OSTSIZE=$OSTSIZE
70 if ! combined_mgs_mds; then
71 # bug number for skipped test: 23954
72 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 24b"
75 # pass "-E lazy_itable_init" to mke2fs to speed up the formatting time
76 if [[ "$LDISKFS_MKFS_OPTS" != *lazy_itable_init* ]]; then
77 LDISKFS_MKFS_OPTS=$(csa_add "$LDISKFS_MKFS_OPTS" -E lazy_itable_init)
80 [ $(facet_fstype $SINGLEMDS) = "zfs" ] &&
81 # bug number for skipped test:
82 ALWAYS_EXCEPT="$ALWAYS_EXCEPT"
87 require_dsh_mds || exit 0
88 require_dsh_ost || exit 0
91 [ "$SLOW" = "no" ] && EXCEPT_SLOW="45 69"
96 # The MGS must be started before the OSTs for a new fs, so start
97 # and stop to generate the startup logs.
100 wait_osc_import_state mds ost FULL
105 reformat_and_config() {
107 if ! combined_mgs_mds ; then
113 writeconf_or_reformat() {
114 # There are at most 2 OSTs for write_conf test
115 # who knows if/where $TUNEFS is installed?
116 # Better reformat if it fails...
117 writeconf_all $MDSCOUNT 2 ||
118 { echo "tunefs failed, reformatting instead" &&
119 reformat_and_config && return 0; }
129 start mgs $(mgsdevname) $MGS_MOUNT_OPTS
135 local dev=$(mdsdevname $num)
138 echo "start mds service on `facet_active_host $facet`"
139 start $facet ${dev} $MDS_MOUNT_OPTS $@ || return 94
145 local dev=$(mdsdevname $num)
148 echo "stop mds service on `facet_active_host $facet`"
149 # These tests all use non-failover stop
150 stop $facet -f || return 97
156 for num in $(seq $MDSCOUNT); do
157 start_mdt $num $@ || return 94
162 if ! combined_mgs_mds ; then
170 for num in $(seq $MDSCOUNT); do
171 stop_mdt $num || return 97
176 echo "stop mgs service on `facet_active_host mgs`"
177 # These tests all use non-failover stop
178 stop mgs -f || return 97
182 echo "start ost1 service on `facet_active_host ost1`"
183 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS $@ || return 95
187 echo "stop ost1 service on `facet_active_host ost1`"
188 # These tests all use non-failover stop
189 stop ost1 -f || return 98
193 echo "start ost2 service on `facet_active_host ost2`"
194 start ost2 $(ostdevname 2) $OST_MOUNT_OPTS $@ || return 92
198 echo "stop ost2 service on `facet_active_host ost2`"
199 # These tests all use non-failover stop
200 stop ost2 -f || return 93
205 echo "mount $FSNAME on ${MOUNTPATH}....."
206 zconf_mount $(hostname) $MOUNTPATH || return 96
210 local mountopt="remount,$1"
212 echo "remount '$1' lustre on ${MOUNTPATH}....."
213 zconf_mount $(hostname) $MOUNTPATH "$mountopt" || return 96
219 echo "umount lustre on $mountpath....."
220 zconf_umount $HOSTNAME $mountpath $@ || return 97
223 manual_umount_client(){
226 echo "manual umount lustre on ${MOUNT}...."
227 do_facet client "umount ${FORCE} $MOUNT"
233 start_mds || error "MDT start failed"
234 start_ost || error "Unable to start OST1"
235 mount_client $MOUNT || error "client start failed"
236 client_up || error "client_up failed"
245 unload_modules_conf () {
246 if combined_mgs_mds || ! local_mode; then
247 unload_modules || return 1
252 stop_ost || return 202
253 stop_mds || return 201
254 unload_modules_conf || return 203
258 umount_client $MOUNT || return 200
259 cleanup_nocli || return $?
264 echo "umount $MOUNT2 ..."
265 umount $MOUNT2 || true
266 echo "stopping fs2mds ..."
267 stop fs2mds -f || true
268 echo "stopping fs2ost ..."
269 stop fs2ost -f || true
273 do_facet client "cp /etc/passwd $DIR/a" || return 71
274 do_facet client "rm $DIR/a" || return 72
275 # make sure lustre is actually mounted (touch will block,
276 # but grep won't, so do it after)
277 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
278 echo "setup single mount lustre success"
282 do_facet client "touch $DIR/a" || return 71
283 do_facet client "rm $DIR/a" || return 72
284 do_facet client "touch $DIR2/a" || return 73
285 do_facet client "rm $DIR2/a" || return 74
286 echo "setup double mount lustre success"
291 if [ "$ONLY" == "setup" ]; then
296 if [ "$ONLY" == "cleanup" ]; then
303 #create single point mountpoint
309 check_mount || error "check_mount failed"
310 cleanup || error "cleanup failed with $?"
312 run_test 0 "single mount setup"
315 start_mds || error "MDS start failed"
316 start_ost || error "unable to start OST"
317 echo "start ost second time..."
318 start_ost && error "2nd OST start should fail"
319 mount_client $MOUNT || error "client start failed"
320 check_mount || error "check_mount failed"
321 cleanup || error "cleanup failed with $?"
323 run_test 1 "start up ost twice (should return errors)"
326 start_mds || error "MDT start failed"
327 echo "start mds second time.."
328 start_mds && error "2nd MDT start should fail"
329 start_ost || error "OST start failed"
330 mount_client $MOUNT || error "mount_client failed to start client"
331 check_mount || error "check_mount failed"
332 cleanup || error "cleanup failed with $?"
334 run_test 2 "start up mds twice (should return err)"
338 #mount.lustre returns an error if already in mtab
339 mount_client $MOUNT && error "2nd client mount should fail"
340 check_mount || error "check_mount failed"
341 cleanup || error "cleanup failed with $?"
343 run_test 3 "mount client twice (should return err)"
347 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
348 stop_ost || error "Unable to stop OST1"
349 umount_client $MOUNT -f || error “unmount $MOUNT failed”
352 # ok for ost to fail shutdown
353 if [ 202 -ne $eno ] && [ 0 -ne $eno ]; then
354 error "cleanup failed with $?"
357 run_test 4 "force cleanup ost, then cleanup"
359 test_5a() { # was test_5
361 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
362 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
364 stop_mds || error "Unable to stop MDS"
366 # cleanup may return an error from the failed
367 # disconnects; for now I'll consider this successful
368 # if all the modules have unloaded.
372 echo "killing umount"
373 kill -TERM $UMOUNT_PID
374 echo "waiting for umount to finish"
376 if grep " $MOUNT " /proc/mounts; then
377 echo "test 5: /proc/mounts after failed umount"
381 echo "killing umount"
382 kill -TERM $UMOUNT_PID
383 echo "waiting for umount to finish"
385 grep " $MOUNT " /proc/mounts &&
386 error "/proc/mounts after second umount"
389 # manual_mount_client may fail due to umount succeeding above
391 # stop_mds is a no-op here, and should not fail
392 cleanup_nocli || error "cleanup_nocli failed with $?"
393 # df may have lingering entry
395 # mtab may have lingering entry
399 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
401 grep -q $MOUNT" " /etc/mtab || break
402 echo "Waiting /etc/mtab updated ... "
403 WAIT=$(( WAIT + sleep))
405 [ "$WAIT" -eq "$MAX_WAIT" ] &&
406 error "/etc/mtab is not updated in $WAIT secs"
407 echo "/etc/mtab updated in $WAIT secs"
409 run_test 5a "force cleanup mds, then cleanup"
417 grep " $MOUNT " /etc/mtab &&
418 error false "unexpected entry in mtab before mount" && return 10
420 start_ost || error "OST start failed"
421 if ! combined_mgs_mds ; then
422 trap cleanup_5b EXIT ERR
423 start_mds || error "MDS start failed"
427 mount_client $MOUNT && error "mount_client $MOUNT should fail"
428 grep " $MOUNT " /etc/mtab &&
429 error "$MOUNT entry in mtab after failed mount"
431 # stop_mds is a no-op here, and should not fail
432 cleanup_nocli || error "cleanup_nocli failed with $?"
433 if ! combined_mgs_mds ; then
437 run_test 5b "Try to start a client with no MGS (should return errs)"
440 grep " $MOUNT " /etc/mtab &&
441 error false "unexpected entry in mtab before mount" && return 10
443 start_mds || error "MDS start failed"
444 start_ost || error "OST start failed"
445 local oldfs="${FSNAME}"
446 FSNAME="wrong.${FSNAME}"
447 mount_client $MOUNT || :
449 grep " $MOUNT " /etc/mtab &&
450 error "$MOUNT entry in mtab after failed mount"
452 cleanup_nocli || error "cleanup_nocli failed with $?"
454 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
457 grep " $MOUNT " /etc/mtab &&
458 error "unexpected entry in mtab before mount"
460 start_ost || error "OST start failed"
461 start_mds || error "MDS start failed"
462 stop_ost -f || error "Unable to stop OST1"
463 mount_client $MOUNT || error "mount_client $MOUNT failed"
464 umount_client $MOUNT -f || error "umount_client $MOUNT failed"
465 cleanup_nocli || error "cleanup_nocli failed with $?"
466 ! grep " $MOUNT " /etc/mtab ||
467 error "$MOUNT entry in mtab after unmount"
469 run_test 5d "mount with ost down"
472 grep " $MOUNT " /etc/mtab &&
473 error false "unexpected entry in mtab before mount" && return 10
475 start_mds || error "MDS start failed"
476 start_ost || error "OST start failed"
478 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
479 do_facet client "$LCTL set_param fail_loc=0x80000506"
480 mount_client $MOUNT || echo "mount failed (not fatal)"
481 cleanup || error "cleanup failed with $?"
482 grep " $MOUNT " /etc/mtab &&
483 error "$MOUNT entry in mtab after unmount"
486 run_test 5e "delayed connect, don't crash (bug 10268)"
489 if combined_mgs_mds ; then
490 skip "combined mgs and mds"
494 grep " $MOUNT " /etc/mtab &&
495 error false "unexpected entry in mtab before mount" && return 10
498 start_ost || error "OST start failed"
499 mount_client $MOUNT &
501 echo client_mount pid is $pid
505 if ! ps -f -p $pid >/dev/null; then
508 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
509 error "mount returns $rc, expected to hang"
511 cleanup || error "cleanup failed with $?"
516 start_mds || error "start MDS failed"
518 # mount should succeed after start mds
520 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
521 cleanup || error "final call to cleanup failed with rc $?"
523 run_test 5f "mds down, cleanup after failed mount (bug 2712)"
528 mount_client $MOUNT || error "mount_client $MOUNT failed"
529 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
530 cleanup || error "cleanup failed with rc $?"
532 run_test 6 "manual umount, then mount again"
537 cleanup_nocli || error "cleanup_nocli failed with $?"
539 run_test 7 "manual umount, then cleanup"
543 mount_client $MOUNT2 || error "mount_client $MOUNT2 failed"
544 check_mount2 || error "check_mount2 failed"
545 umount_client $MOUNT2 || error "umount_client $MOUNT2 failed"
546 cleanup || error "cleanup failed with rc $?"
548 run_test 8 "double mount setup"
551 start_ost || error "OST start failed"
553 do_facet ost1 $LCTL set_param debug=\'inode trace\' ||
554 error "do_facet ost1 set_param inode trace failed."
555 do_facet ost1 $LCTL set_param subsystem_debug=\'mds ost\' ||
556 error "do_facet ost1 set_param debug mds ost failed."
558 CHECK_PTLDEBUG="`do_facet ost1 $LCTL get_param -n debug`"
559 if [ "$CHECK_PTLDEBUG" ] && { \
560 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
561 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
562 echo "lnet.debug success"
564 error "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
566 CHECK_SUBSYS="`do_facet ost1 $LCTL get_param -n subsystem_debug`"
567 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
568 echo "lnet.subsystem_debug success"
570 error "lnet.subsystem_debug: want 'mds ost' got '$CHECK_SUBSYS'"
572 stop_ost || error "Unable to stop OST1"
574 run_test 9 "test ptldebug and subsystem for mkfs"
582 do_facet $facet "test -b $dev" || rc=1
583 if [[ "$size" ]]; then
584 local in=$(do_facet $facet "dd if=$dev of=/dev/null bs=1k \
585 count=1 skip=$size 2>&1" |
586 awk '($3 == "in") { print $1 }')
587 [[ $in = "1+0" ]] || rc=1
593 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
594 # But with new MDS stack we don't care about the mode of local objects
595 # anymore, so this test is removed. See bug 22944 for more details.
599 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
600 skip "Only applicable to ldiskfs-based MDTs"
605 check_mount || error "check_mount failed"
606 cleanup || error "cleanup failed with rc $?"
608 echo "Remove mds config log"
609 if ! combined_mgs_mds ; then
613 do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' \
614 $(mgsdevname) || return \$?" ||
615 error "do_facet mgs failed with $?"
617 if ! combined_mgs_mds ; then
621 start_ost || error "OST start failed"
622 start_mds && error "MDS start succeeded, but should fail"
625 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
628 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
629 skip "Only applicable to ldiskfs-based MDTs"
633 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
638 # check if current MDSSIZE is large enough
639 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE &&
640 log "use MDSSIZE=$MDSSIZE"
642 # check if the global config has a large enough MDSSIZE
643 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] &&
644 [ $STORED_MDSSIZE -ge $MIN ] &&
645 OK=1 && myMDSSIZE=$STORED_MDSSIZE &&
646 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
648 # check if the block device is large enough
649 is_blkdev $SINGLEMDS $MDSDEV $MIN
650 local large_enough=$?
651 if [ -n "$OK" ]; then
652 [ $large_enough -ne 0 ] && OK=""
654 [ $large_enough -eq 0 ] && OK=1 && myMDSSIZE=$MIN &&
655 log "use device $MDSDEV with MIN=$MIN"
658 # check if a loopback device has enough space for fs metadata (5%)
660 if [ -z "$OK" ]; then
661 local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! \
662 -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
663 awk '($1 != "Filesystem") { print $4 }')
664 ! [ -z "$SPACE" ] && [ $SPACE -gt $((MIN / 20)) ] &&
665 OK=1 && myMDSSIZE=$MIN &&
666 log "use file $MDSDEV with MIN=$MIN"
669 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS" && return
671 echo "mount mds with large journal..."
673 local OLD_MDSSIZE=$MDSSIZE
677 echo "mount lustre system..."
679 check_mount || error "check_mount failed"
681 echo "check journal size..."
682 local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
683 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
684 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
686 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
689 cleanup || error "cleanup failed with rc $?"
694 run_test 18 "check mkfs creates large journals"
697 start_mds || error "MDS start failed"
698 stop_mds || error "Unable to stop MDS"
700 run_test 19a "start/stop MDS without OSTs"
703 start_ost || error "Unable to start OST1"
704 stop_ost -f || error "Unable to stop OST1"
706 run_test 19b "start/stop OSTs without MDS"
709 # first format the ost/mdt
710 start_mds || error "MDS start failed"
711 start_ost || error "Unable to start OST1"
712 mount_client $MOUNT || error "mount_client $MOUNT failed"
713 check_mount || error "check_mount failed"
714 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed."
715 remount_client ro $MOUNT || error "remount_client with ro failed"
716 touch $DIR/$tfile && error "$DIR/$tfile created incorrectly"
717 [ -e $DIR/$tfile ] && error "$DIR/$tfile exists incorrectly"
718 remount_client rw $MOUNT || error "remount_client with rw failed"
719 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
720 MCNT=$(grep -c $MOUNT /etc/mtab)
721 [ "$MCNT" -ne 1 ] && error "$MOUNT in /etc/mtab $MCNT times"
723 stop_mds || error "Unable to stop MDS"
724 stop_ost || error "Unable to stop OST1"
726 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
729 start_mds || error "MDS start failed"
730 start_ost || error "unable to start OST1"
731 wait_osc_import_state mds ost FULL
732 stop_ost || error "unable to stop OST1"
733 stop_mds || error "unable to stop MDS"
735 run_test 21a "start mds before ost, stop ost first"
738 start_ost || error "unable to start OST1"
739 start_mds || error "MDS start failed"
740 wait_osc_import_state mds ost FULL
741 stop_mds || error "unable to stop MDS"
742 stop_ost || error "unable to stop OST1"
744 run_test 21b "start ost before mds, stop mds first"
747 start_ost || error "Unable to start OST1"
748 start_mds || error "MDS start failed"
749 start_ost2 || error "Unable to start OST2"
750 wait_osc_import_state mds ost2 FULL
751 stop_ost || error "Unable to stop OST1"
752 stop_ost2 || error "Unable to stop OST2"
753 stop_mds || error "Unable to stop MDS"
754 #writeconf to remove all ost2 traces for subsequent tests
755 writeconf_or_reformat
757 run_test 21c "start mds between two osts, stop mds last"
760 if combined_mgs_mds ; then
761 skip "need separate mgs device" && return 0
767 start_mgs || error "unable to start MGS"
768 start_ost || error "unable to start OST1"
769 start_ost2 || error "unable to start OST2"
770 start_mds || error "MDS start failed"
771 wait_osc_import_state mds ost2 FULL
773 stop_ost || error "Unable to stop OST1"
774 stop_ost2 || error "Unable to stop OST2"
775 stop_mds || error "Unable to stop MDS"
777 #writeconf to remove all ost2 traces for subsequent tests
778 writeconf_or_reformat
779 start_mgs || error "unable to start MGS"
781 run_test 21d "start mgs then ost and then mds"
784 MGSNID="$saved_mgsnid"
786 echo "stopping fs2mgs ..."
787 stop $fs2mgs -f || true
790 test_21e() { # LU-5863
791 if [[ -z "$fs3ost_DEV" || -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
792 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
793 skip_env "mixed loopback and real device not working" && return
796 local fs2mdsdev=$(mdsdevname 1_2)
797 local fs2ostdev=$(ostdevname 1_2)
798 local fs3ostdev=$(ostdevname 2_2)
800 local fs2mdsvdev=$(mdsvdevname 1_2)
801 local fs2ostvdev=$(ostvdevname 1_2)
802 local fs3ostvdev=$(ostvdevname 2_2)
804 # temporarily use fs3ost as fs2mgs
806 local fs2mgsdev=$fs3ostdev
807 local fs2mgsvdev=$fs3ostvdev
809 local fsname=test1234
811 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
812 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
813 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS && trap cleanup_21e EXIT INT ||
814 error "start fs2mgs failed"
816 local saved_mgsnid="$MGSNID"
817 MGSNID=$(do_facet $fs2mgs $LCTL list_nids | xargs | tr ' ' ,)
819 add fs2mds $(mkfs_opts mds1 $fs2mdsdev $fsname) \
820 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
821 add fs2ost $(mkfs_opts ost1 $fs2ostdev $fsname) \
822 --reformat $fs2ostdev $fs2ostvdev || error "add fs2ost failed"
824 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
825 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS || error "start fs2mds failed"
827 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
828 $MOUNT_CMD $MGSNID:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
829 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
833 run_test 21e "separate MGS and MDS"
836 start_mds || error "MDS start failed"
838 echo "Client mount with ost in logs, but none running"
839 start_ost || error "unable to start OST1"
840 # wait until mds connected to ost and open client connection
841 wait_osc_import_state mds ost FULL
842 stop_ost || error "unable to stop OST1"
843 mount_client $MOUNT || error "mount_client $MOUNT failed"
844 # check_mount will block trying to contact ost
845 mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed"
846 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed"
847 umount_client $MOUNT -f
850 echo "Client mount with a running ost"
851 start_ost || error "unable to start OST1"
853 # if gss enabled, wait full time to let connection from
854 # mds to ost be established, due to the mismatch between
855 # initial connect timeout and gss context negotiation timeout.
856 # This perhaps could be remove after AT landed.
857 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
858 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
860 mount_client $MOUNT || error "mount_client $MOUNT failed"
861 wait_osc_import_state mds ost FULL
862 wait_osc_import_state client ost FULL
863 check_mount || error "check_mount failed"
866 cleanup || error "cleanup failed with rc $?"
868 run_test 22 "start a client before osts (should return errs)"
870 test_23a() { # was test_23
873 stop $SINGLEMDS || error "failed to stop $SINGLEMDS"
874 # force down client so that recovering mds waits for reconnect
875 local running=$(grep -c $MOUNT /proc/mounts) || true
876 if [ $running -ne 0 ]; then
877 echo "Stopping client $MOUNT (opts: -f)"
881 # enter recovery on failed mds
882 local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/})
883 start $SINGLEMDS $MDT_DEV $MDS_MOUNT_OPTS || error "MDS start failed"
884 # try to start a new client
885 mount_client $MOUNT &
887 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
888 MOUNT_LUSTRE_PID=$(ps -ef | grep mount.lustre |
889 grep -v grep | awk '{print $2}')
890 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
892 ps --ppid $MOUNT_LUSTRE_PID
893 echo "waiting for mount to finish"
895 # "ctrl-c" sends SIGINT but it usually (in script) does not work on child process
896 # SIGTERM works but it does not spread to offspring processses
897 kill -s TERM $MOUNT_PID
898 kill -s TERM $MOUNT_LUSTRE_PID
899 # we can not wait $MOUNT_PID because it is not a child of this shell
905 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
907 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
908 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
911 [ -z "$PID1" -a -z "$PID2" ] && break
912 echo "waiting for mount to finish ... "
913 WAIT=$(( WAIT + sleep))
915 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
916 error "MOUNT_PID $MOUNT_PID and "\
917 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
920 stop_mds || error "stopping MDSes failed"
921 stop_ost || error "stopping OSSes failed"
923 run_test 23a "interrupt client during recovery mount delay"
928 test_23b() { # was test_23
929 start_mds || error "MDS start failed"
930 start_ost || error "Unable to start OST1"
931 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
932 $LCTL set_param fail_loc=0x80000313
934 cleanup || error "cleanup failed with rc $?"
936 run_test 23b "Simulate -EINTR during mount"
939 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
941 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
942 is_blkdev $SINGLEMDS $MDSDEV &&
943 skip_env "mixed loopback and real device not working" && return
946 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
948 local fs2mdsdev=$(mdsdevname 1_2)
949 local fs2ostdev=$(ostdevname 1_2)
950 local fs2mdsvdev=$(mdsvdevname 1_2)
951 local fs2ostvdev=$(ostvdevname 1_2)
953 # test 8-char fsname as well
954 local FSNAME2=test1234
956 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --nomgs --mgsnode=$MGSNID \
957 --fsname=${FSNAME2} --reformat $fs2mdsdev $fs2mdsvdev || exit 10
959 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --fsname=${FSNAME2} \
960 --reformat $fs2ostdev $fs2ostvdev || exit 10
963 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
964 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
965 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
966 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
968 check_mount || error "check_mount failed"
969 # files written on 1 should not show up on 2
970 cp /etc/passwd $DIR/$tfile
972 [ -e $MOUNT2/$tfile ] && error "File bleed"
975 cp /etc/passwd $MOUNT2/$tfile ||
976 error "cp /etc/passwd $MOUNT2/$tfile failed"
977 rm $MOUNT2/$tfile || error "remove $MOUNT2/$tfile failed"
978 # 2 is actually mounted
979 grep $MOUNT2' ' /proc/mounts > /dev/null || error "$MOUNT2 not mounted"
981 facet_failover fs2mds
982 facet_failover fs2ost
985 # the MDS must remain up until last MDT
987 MDS=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
988 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
989 [ -z "$MDS" ] && error "No MDT"
991 cleanup_nocli || error "cleanup_nocli failed with rc $?"
993 run_test 24a "Multiple MDTs on a single node"
996 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
998 if [ -z "$fs2mds_DEV" ]; then
999 local dev=${SINGLEMDS}_dev
1000 local MDSDEV=${!dev}
1001 is_blkdev $SINGLEMDS $MDSDEV &&
1002 skip_env "mixed loopback and real device not working" && return
1005 local fs2mdsdev=$(mdsdevname 1_2)
1006 local fs2mdsvdev=$(mdsvdevname 1_2)
1008 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --mgs --fsname=${FSNAME}2 \
1009 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
1011 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS &&
1012 error "start MDS should fail"
1014 cleanup || error "cleanup failed with rc $?"
1016 run_test 24b "Multiple MGSs on a single node (should return err)"
1020 check_mount || error "check_mount failed"
1021 local MODULES=$($LCTL modules | awk '{ print $2 }')
1022 rmmod $MODULES 2>/dev/null || true
1023 cleanup || error "cleanup failed with $?"
1025 run_test 25 "Verify modules are referenced"
1029 # we need modules before mount for sysctl, so make sure...
1030 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
1031 #define OBD_FAIL_MDS_FS_SETUP 0x135
1032 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000135"
1033 start_mds && error "MDS started but should not have started"
1034 $LCTL get_param -n devices
1035 DEVS=$($LCTL get_param -n devices | egrep -v MG | wc -l)
1036 [ $DEVS -gt 0 ] && error "number of devices is $DEVS, should be zero"
1037 # start mds to drop writeconf setting
1038 start_mds || error "Unable to start MDS"
1039 stop_mds || error "Unable to stop MDS"
1040 unload_modules_conf || error "unload_modules_conf failed with $?"
1042 run_test 26 "MDT startup failure cleans LOV (should return errs)"
1045 start_ost || error "Unable to start OST1"
1046 start_mds || error "Unable to start MDS"
1047 echo "Requeue thread should have started: "
1048 ps -e | grep ll_cfg_requeue
1049 set_conf_param_and_check ost1 \
1050 "$LCTL get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" \
1051 "$FSNAME-OST0000.ost.client_cache_seconds" ||
1052 error "set_conf_param_and_check ost1 failed"
1053 cleanup_nocli || error "cleanup_nocli failed with rc $?"
1055 run_test 27a "Reacquire MGS lock if OST started first"
1060 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
1061 awk '($3 ~ "mdt" && $4 ~ "MDT0000") { print $4 }')
1063 facet_failover $SINGLEMDS
1064 set_conf_param_and_check $SINGLEMDS \
1065 "$LCTL get_param -n mdt.$device.identity_acquire_expire" \
1066 "$device.mdt.identity_acquire_expire" ||
1067 error "set_conf_param_and_check $SINGLEMDS failed"
1068 set_conf_param_and_check client \
1069 "$LCTL get_param -n mdc.$device-mdc-*.max_rpcs_in_flight"\
1070 "$device.mdc.max_rpcs_in_flight" ||
1071 error "set_conf_param_and_check client failed"
1073 cleanup || error "cleanup failed with $?"
1075 run_test 27b "Reacquire MGS lock after failover"
1079 TEST="$LCTL get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1080 PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
1082 FINAL=$(($ORIG + 1))
1083 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL ||
1084 error "first set_conf_param_and_check client failed"
1085 FINAL=$(($FINAL + 1))
1086 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL ||
1087 error "second set_conf_param_and_check client failed"
1088 umount_client $MOUNT || error "umount_client $MOUNT failed"
1089 mount_client $MOUNT || error "mount_client $MOUNT failed"
1091 if [ $RESULT -ne $FINAL ]; then
1092 error "New config not seen: wanted $FINAL got $RESULT"
1094 echo "New config success: got $RESULT"
1096 set_conf_param_and_check client "$TEST" "$PARAM" $ORIG ||
1097 error "third set_conf_param_and_check client failed"
1098 cleanup || error "cleanup failed with rc $?"
1100 run_test 28 "permanent parameter setting"
1102 test_28a() { # LU-4221
1103 [[ $(lustre_version_code ost1) -ge $(version_code 2.5.52) ]] ||
1104 { skip "Need OST version at least 2.5.52" && return 0; }
1105 [ "$(facet_fstype ost1)" = "zfs" ] &&
1106 skip "LU-4221: no such proc params for ZFS OSTs" && return
1113 local device="$FSNAME-OST0000"
1117 # In this test we will set three kinds of proc parameters with
1119 # 1. the ones moved from the OFD to the OSD, and only their
1120 # symlinks kept in obdfilter
1121 # 2. non-symlink ones in the OFD
1122 # 3. non-symlink ones in the OSD
1125 # prepare a symlink parameter in the OFD
1126 name="writethrough_cache_enable"
1127 param="$device.ost.$name"
1128 cmd="$LCTL get_param -n obdfilter.$device.$name"
1130 # conf_param the symlink parameter in the OFD
1131 old=$(do_facet ost1 $cmd)
1132 new=$(((old + 1) % 2))
1133 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1134 error "lctl conf_param $device.ost.$param=$new failed"
1136 # conf_param the target parameter in the OSD
1137 param="$device.osd.$name"
1138 cmd="$LCTL get_param -n osd-*.$device.$name"
1139 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1140 error "lctl conf_param $device.osd.$param=$old failed"
1143 # prepare a non-symlink parameter in the OFD
1144 name="client_cache_seconds"
1145 param="$device.ost.$name"
1146 cmd="$LCTL get_param -n obdfilter.$device.$name"
1148 # conf_param the parameter in the OFD
1149 old=$(do_facet ost1 $cmd)
1151 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1152 error "lctl conf_param $device.ost.$param=$new failed"
1153 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1154 error "lctl conf_param $device.ost.$param=$old failed"
1157 # prepare a non-symlink parameter in the OSD
1159 param="$device.osd.$name"
1160 cmd="$LCTL get_param -n osd-*.$device.$name"
1162 # conf_param the parameter in the OSD
1163 old=$(do_facet ost1 $cmd)
1164 new=$(((old + 1) % 2))
1165 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1166 error "lctl conf_param $device.osd.$param=$new failed"
1167 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1168 error "lctl conf_param $device.osd.$param=$old failed"
1170 cleanup || error "cleanup failed with $?"
1172 run_test 28a "set symlink parameters permanently with conf_param"
1175 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
1176 setup > /dev/null 2>&1
1177 start_ost2 || error "Unable to start OST2"
1180 local PARAM="$FSNAME-OST0001.osc.active"
1181 local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
1182 local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
1184 ACTV=$($LCTL get_param -n $PROC_ACT)
1186 set_conf_param_and_check client \
1187 "$LCTL get_param -n $PROC_ACT" "$PARAM" $DEAC ||
1188 error "set_conf_param_and_check client failed"
1189 # also check ost_server_uuid status
1190 RESULT=$($LCTL get_param -n $PROC_UUID | grep DEACTIV)
1191 if [ -z "$RESULT" ]; then
1192 error "Client not deactivated: $($LCTL get_param \
1195 echo "Live client success: got $RESULT"
1199 for num in $(seq $MDSCOUNT); do
1200 local mdtosc=$(get_mdtosc_proc_path mds${num} $FSNAME-OST0001)
1201 local MPROC="osc.$mdtosc.active"
1206 RESULT=$(do_facet mds${num} "$LCTL get_param -n $MPROC")
1207 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
1208 if [ $RESULT -eq $DEAC ]; then
1209 echo -n "MDT deactivated also after"
1210 echo "$WAIT sec (got $RESULT)"
1214 if [ $WAIT -eq $MAX ]; then
1215 error "MDT active: wanted $DEAC got $RESULT"
1217 echo "Waiting $(($MAX - $WAIT))secs for MDT deactivated"
1220 # test new client starts deactivated
1221 umount_client $MOUNT || error "umount_client $MOUNT failed"
1222 mount_client $MOUNT || error "mount_client $MOUNT failed"
1223 RESULT=$($LCTL get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
1224 if [ -z "$RESULT" ]; then
1225 error "New client start active: $(lctl get_param -n $PROC_UUID)"
1227 echo "New client success: got $RESULT"
1230 # make sure it reactivates
1231 set_conf_param_and_check client \
1232 "$LCTL get_param -n $PROC_ACT" "$PARAM" $ACTV ||
1233 error "lctl get_param $PROC_ACT $PARAM $ACTV failed"
1235 umount_client $MOUNT
1236 stop_ost2 || error "Unable to stop OST2"
1237 cleanup_nocli || error "cleanup_nocli failed with $?"
1238 #writeconf to remove all ost2 traces for subsequent tests
1239 writeconf_or_reformat
1241 run_test 29 "permanently remove an OST"
1246 echo Big config llog
1247 TEST="$LCTL get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1249 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1250 for i in ${LIST[@]}; do
1251 set_conf_param_and_check client "$TEST" \
1252 "$FSNAME.llite.max_read_ahead_whole_mb" $i ||
1253 error "Set $FSNAME.llite.max_read_ahead_whole_mb failed"
1255 # make sure client restart still works
1256 umount_client $MOUNT
1257 mount_client $MOUNT || error "mount_client $MOUNT failed"
1258 [ "$($TEST)" -ne "$i" ] &&
1259 error "Param didn't stick across restart $($TEST) != $i"
1262 echo Erase parameter setting
1263 do_facet mgs "$LCTL conf_param \
1264 -d $FSNAME.llite.max_read_ahead_whole_mb" ||
1265 error "Erase param $FSNAME.llite.max_read_ahead_whole_mb failed"
1266 umount_client $MOUNT
1267 mount_client $MOUNT || error "mount_client $MOUNT failed"
1269 echo "deleted (default) value=$FINAL, orig=$ORIG"
1270 # assumes this parameter started at the default value
1271 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
1273 cleanup || error "cleanup failed with rc $?"
1275 run_test 30a "Big config llog and conf_param deletion"
1280 local orignids=$($LCTL get_param -n \
1281 osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1283 local orignidcount=$(echo "$orignids" | wc -w)
1285 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1286 # numerical part of it. Hopefully that's not already a failover address
1288 local OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | \
1290 local ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1291 local NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1292 local NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1293 echo "Using fake nid $NEW"
1295 local TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1296 grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1297 set_conf_param_and_check client "$TEST" \
1298 "$FSNAME-OST0000.failover.node" $NEW ||
1299 error "didn't add failover nid $NEW"
1300 local NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1303 local NIDCOUNT=$(echo "$NIDS" | wc -w)
1304 echo "should have $((orignidcount + 1)) entries \
1305 in failover nids string, have $NIDCOUNT"
1306 [ $NIDCOUNT -eq $((orignidcount + 1)) ] ||
1307 error "Failover nid not added"
1309 do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" ||
1310 error "conf_param delete failed"
1311 umount_client $MOUNT
1312 mount_client $MOUNT || error "mount_client $MOUNT failed"
1314 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1317 NIDCOUNT=$(echo "$NIDS" | wc -w)
1318 echo "only $orignidcount final entries should remain \
1319 in failover nids string, have $NIDCOUNT"
1320 [ $NIDCOUNT -eq $orignidcount ] || error "Failover nids not removed"
1322 cleanup || error "cleanup failed with rc $?"
1324 run_test 30b "Remove failover nids"
1326 test_31() { # bug 10734
1327 # ipaddr must not exist
1328 $MOUNT_CMD 4.3.2.1@tcp:/lustre $MOUNT || true
1329 cleanup || error "cleanup failed with rc $?"
1331 run_test 31 "Connect to non-existent node (shouldn't crash)"
1335 T32_BLIMIT=40960 # Kbytes
1339 # This is not really a test but a tool to create new disk
1340 # image tarballs for the upgrade tests.
1342 # Disk image tarballs should be created on single-node
1343 # clusters by running this test with default configurations
1344 # plus a few mandatory environment settings that are verified
1345 # at the beginning of the test.
1347 test_32newtarball() {
1351 local tmp=$TMP/t32_image_create
1352 local server_version=$(lustre_version_code $SINGLEMDS)
1357 if [ $FSNAME != t32fs -o \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o \
1358 $OSTCOUNT -ne 1 -o -z "$OSTDEV1" ]; then
1359 error "Needs FSNAME=t32fs MDSCOUNT=2 " \
1360 "MDSDEV1=<nonexistent_file>" \
1361 "MDSDEV2=<nonexistent_file>" \
1362 "(or MDSDEV, in the case of b1_8)" \
1363 "OSTCOUNT=1 OSTDEV1=<nonexistent_file>"
1367 echo "Found stale $tmp"
1371 mkdir $tmp/src || return 1
1372 tar cf - -C $src . | tar xf - -C $tmp/src
1373 dd if=/dev/zero of=$tmp/src/t32_qf_old bs=1M \
1374 count=$(($T32_BLIMIT / 1024 / 4))
1375 chown $T32_QID.$T32_QID $tmp/src/t32_qf_old
1377 # format ost with comma-separated NIDs to verify LU-4460
1378 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
1379 MGSNID="$MGSNID,$MGSNID" OSTOPT="--failnode=$failnid" formatall
1383 [[ $server_version -ge $(version_code 2.3.50) ]] ||
1384 $LFS quotacheck -ug /mnt/$FSNAME
1385 $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
1388 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1390 if [[ $MDSCOUNT -ge 2 ]]; then
1391 remote_dir=/mnt/$FSNAME/remote_dir
1392 $LFS mkdir -i 1 $remote_dir
1393 tar cf - -C $tmp/src . | tar xf - -C $remote_dir
1395 if [[ $server_version -ge $(version_code 2.7.0) ]]; then
1396 striped_dir=/mnt/$FSNAME/striped_dir_old
1397 $LFS mkdir -i 1 -c 2 $striped_dir
1398 tar cf - -C $tmp/src . | tar xf - -C $striped_dir
1404 mkdir $tmp/img || return 1
1408 pushd_dir=/mnt/$FSNAME
1409 if [[ $MDSCOUNT -ge 2 ]]; then
1410 pushd_dir=$remote_dir
1411 if [[ $server_version -ge $(version_code 2.7.0) ]]; then
1413 ls -Rni --time-style=+%s >$tmp/img/list2
1419 ls -Rni --time-style=+%s >$tmp/img/list
1420 find ! -name .lustre -type f -exec sha1sum {} \; |
1421 sort -k 2 >$tmp/img/sha1sums
1423 $LCTL get_param -n version | head -n 1 |
1424 sed -e 's/^lustre: *//' >$tmp/img/commit
1426 [[ $server_version -ge $(version_code 2.3.50) ]] ||
1427 $LFS quotaon -ug /mnt/$FSNAME
1428 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1429 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1430 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1431 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1432 | tr -d "*" > $tmp/img/bspace
1433 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1434 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1435 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1436 | tr -d "*" > $tmp/img/ispace
1437 echo $T32_BLIMIT > $tmp/img/blimit
1438 echo $T32_ILIMIT > $tmp/img/ilimit
1443 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1446 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1447 echo "Data verification failed"
1450 uname -r >$tmp/img/kernel
1451 uname -m >$tmp/img/arch
1453 mv ${MDSDEV1:-$MDSDEV} $tmp/img
1454 for num in $(seq 2 $MDSCOUNT); do
1455 local devname=$(mdsdevname $num)
1457 [[ $(facet_fstype $facet) != zfs ]] ||
1458 devname=$(mdsvdevname $num)
1459 mv $devname $tmp/img
1461 mv $OSTDEV1 $tmp/img
1463 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1464 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1467 tar cjvf $dst/disk$version-$(facet_fstype $SINGLEMDS).tar.bz2 -S *
1472 #run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1475 # The list of applicable tarballs is returned via the caller's
1476 # variable "tarballs".
1479 local node=$(facet_active_host $SINGLEMDS)
1480 local r="do_node $node"
1482 if [ "$CLIENTONLY" ]; then
1483 skip "Client-only testing"
1487 if ! $r which $TUNEFS; then
1488 skip_env "tunefs.lustre required on $node"
1492 local IMGTYPE=$(facet_fstype $SINGLEMDS)
1494 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 -name \'disk*-$IMGTYPE.tar.bz2\')
1496 if [ -z "$tarballs" ]; then
1497 skip "No applicable tarballs found"
1502 t32_test_cleanup() {
1504 local facet=$SINGLEMDS
1505 local fstype=$(facet_fstype $facet)
1508 if $shall_cleanup_lustre; then
1509 umount $tmp/mnt/lustre || rc=$?
1511 if $shall_cleanup_mdt; then
1512 $r $UMOUNT $tmp/mnt/mdt || rc=$?
1514 if $shall_cleanup_mdt1; then
1515 $r $UMOUNT $tmp/mnt/mdt1 || rc=$?
1517 if $shall_cleanup_ost; then
1518 $r $UMOUNT $tmp/mnt/ost || rc=$?
1523 if [[ $fstype == zfs ]]; then
1525 local poolname_list="t32fs-mdt1 t32fs-ost1"
1527 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1529 for poolname in $poolname_list; do
1530 destroy_zpool $facet $poolname
1536 t32_bits_per_long() {
1538 # Yes, this is not meant to be perfect.
1548 t32_reload_modules() {
1550 local all_removed=false
1553 while ((i < 20)); do
1554 echo "Unloading modules on $node: Attempt $i"
1555 do_rpc_nodes $node $LUSTRE_RMMOD $(facet_fstype $SINGLEMDS) &&
1557 do_rpc_nodes $node check_mem_leak || return 1
1558 if $all_removed; then
1559 do_rpc_nodes $node load_modules
1565 echo "Unloading modules on $node: Given up"
1569 t32_wait_til_devices_gone() {
1575 echo wait for devices to go
1576 while ((i < 20)); do
1577 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1578 loops=$(do_rpc_nodes $node losetup -a | grep -c t32)
1579 ((devices == 0 && loops == 0)) && return 0
1583 echo "waiting for dev on $node: dev $devices loop $loops given up"
1584 do_rpc_nodes $node "losetup -a"
1585 do_rpc_nodes $node "$LCTL device_list"
1589 t32_verify_quota() {
1593 local fstype=$(facet_fstype $SINGLEMDS)
1597 $LFS quota -u $T32_QID -v $mnt
1599 qval=$($LFS quota -v -u $T32_QID $mnt |
1600 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1601 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1603 [ $qval -eq $img_bspace ] || {
1604 echo "bspace, act:$qval, exp:$img_bspace"
1608 qval=$($LFS quota -v -u $T32_QID $mnt |
1609 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1610 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1612 [ $qval -eq $img_ispace ] || {
1613 echo "ispace, act:$qval, exp:$img_ispace"
1617 qval=$($LFS quota -v -u $T32_QID $mnt |
1618 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1619 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1621 [ $qval -eq $img_blimit ] || {
1622 echo "blimit, act:$qval, exp:$img_blimit"
1626 qval=$($LFS quota -v -u $T32_QID $mnt |
1627 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1628 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1630 [ $qval -eq $img_ilimit ] || {
1631 echo "ilimit, act:$qval, exp:$img_ilimit"
1635 do_node $node $LCTL conf_param $fsname.quota.mdt=ug
1636 cmd="$LCTL get_param -n osd-$fstype.$fsname-MDT0000"
1637 cmd=$cmd.quota_slave.enabled
1638 wait_update $node "$cmd" "ug" || {
1639 echo "Enable mdt quota failed"
1643 do_node $node $LCTL conf_param $fsname.quota.ost=ug
1644 cmd="$LCTL get_param -n osd-$fstype.$fsname-OST0000"
1645 cmd=$cmd.quota_slave.enabled
1646 wait_update $node "$cmd" "ug" || {
1647 echo "Enable ost quota failed"
1652 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1653 bs=1M count=$((img_blimit / 1024)) oflag=sync && {
1654 echo "Write succeed, but expect -EDQUOT"
1657 rm -f $mnt/t32_qf_new
1659 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1661 echo "Create succeed, but expect -EDQUOT"
1664 unlinkmany $mnt/t32_qf_ $img_ilimit
1672 local dne_upgrade=${dne_upgrade:-"no"}
1673 local ff_convert=${ff_convert:-"no"}
1674 local shall_cleanup_mdt=false
1675 local shall_cleanup_mdt1=false
1676 local shall_cleanup_ost=false
1677 local shall_cleanup_lustre=false
1678 local mdt2_is_available=false
1679 local node=$(facet_active_host $SINGLEMDS)
1680 local r="do_node $node"
1681 local node2=$(facet_active_host mds2)
1691 local nid=$($r $LCTL list_nids | head -1)
1697 local fstype=$(facet_fstype $SINGLEMDS)
1698 local mdt_dev=$tmp/mdt
1699 local mdt2_dev=$tmp/mdt2
1700 local ost_dev=$tmp/ost
1705 trap 'trap - RETURN; t32_test_cleanup' RETURN
1708 mkdir -p $tmp/mnt/lustre || error "mkdir $tmp/mnt/lustre failed"
1709 $r mkdir -p $tmp/mnt/{mdt,mdt1,ost}
1710 $r tar xjvf $tarball -S -C $tmp || {
1711 error_noexit "Unpacking the disk image tarball"
1714 img_commit=$($r cat $tmp/commit)
1715 img_kernel=$($r cat $tmp/kernel)
1716 img_arch=$($r cat $tmp/arch)
1717 img_bspace=$($r cat $tmp/bspace)
1718 img_ispace=$($r cat $tmp/ispace)
1720 # older images did not have "blimit" and "ilimit" files
1721 # use old values for T32_BLIMIT and T32_ILIMIT
1722 $r test -f $tmp/blimit && img_blimit=$($r cat $tmp/blimit) ||
1724 $r test -f $tmp/ilimit && img_ilimit=$($r cat $tmp/ilimit) ||
1727 echo "Upgrading from $(basename $tarball), created with:"
1728 echo " Commit: $img_commit"
1729 echo " Kernel: $img_kernel"
1730 echo " Arch: $img_arch"
1731 echo "OST version: $(lustre_build_version ost1)"
1733 # The conversion can be made only when both of the following
1734 # conditions are satisfied:
1735 # - ost device img version < 2.3.64
1736 # - ost server version >= 2.5
1737 [ $(version_code $img_commit) -ge $(version_code 2.3.64) -o \
1738 $(lustre_version_code ost1) -lt $(version_code 2.5.0) ] &&
1741 ! $r test -f $mdt2_dev || mdt2_is_available=true
1743 if [[ $fstype == zfs ]]; then
1746 local poolname_list="t32fs-mdt1 t32fs-ost1"
1748 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1750 for poolname in $poolname_list; do
1752 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
1753 $ZPOOL import -f -d $tmp $poolname"
1756 mdt_dev=t32fs-mdt1/mdt1
1757 ost_dev=t32fs-ost1/ost1
1758 ! $mdt2_is_available || mdt2_dev=t32fs-mdt2/mdt2
1759 wait_update_facet $SINGLEMDS "$ZPOOL list |
1760 awk '/^t32fs-mdt1/ { print \\\$1 }'" "t32fs-mdt1" || {
1761 error_noexit "import zfs pool failed"
1766 $r $LCTL set_param debug="$PTLDEBUG"
1768 $r $TUNEFS --dryrun $mdt_dev || {
1770 error_noexit "tunefs.lustre before mounting the MDT"
1774 if $mdt2_is_available; then
1775 $r $TUNEFS --dryrun $mdt2_dev || {
1777 error_noexit "tunefs.lustre before mounting the MDT"
1782 if [ "$writeconf" ]; then
1784 if [ $fstype == "ldiskfs" ]; then
1786 $r $TUNEFS --quota $mdt_dev || {
1788 error_noexit "Enable mdt quota feature"
1791 if $mdt2_is_available; then
1792 $r $TUNEFS --quota $mdt2_dev || {
1794 error_noexit "Enable mdt quota feature"
1800 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
1801 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
1802 { skip "LU-2200: Cannot run over Inifiniband w/o lctl replace_nids "
1803 "(Need MGS version at least 2.3.59)"; return 0; }
1805 local osthost=$(facet_active_host ost1)
1806 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
1809 if [ $fstype == "ldiskfs" ]; then
1812 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt
1813 $r $LCTL replace_nids $fsname-OST0000 $ostnid
1814 $r $LCTL replace_nids $fsname-MDT0000 $nid
1815 $r $UMOUNT $tmp/mnt/mdt
1818 mopts=exclude=$fsname-OST0000
1819 if [ $fstype == "ldiskfs" ]; then
1824 t32_wait_til_devices_gone $node
1826 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
1828 error_noexit "Mounting the MDT"
1831 shall_cleanup_mdt=true
1833 if $mdt2_is_available; then
1834 mopts=mgsnode=$nid,$mopts
1835 $r $MOUNT_CMD -o $mopts $mdt2_dev $tmp/mnt/mdt1 || {
1837 error_noexit "Mounting the MDT"
1841 echo "mount new MDT....$mdt2_dev"
1842 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1843 error_noexit "enable remote dir create failed"
1845 shall_cleanup_mdt1=true
1846 elif [ "$dne_upgrade" != "no" ]; then
1847 local fs2mdsdev=$(mdsdevname 1_2)
1848 local fs2mdsvdev=$(mdsvdevname 1_2)
1850 echo "mkfs new MDT on ${fs2mdsdev}...."
1851 if [ $(facet_fstype mds1) == ldiskfs ]; then
1852 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
1855 add fs2mds $(mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
1856 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
1857 error_noexit "Mkfs new MDT failed"
1861 [[ $(facet_fstype mds1) != zfs ]] || import_zpool fs2mds
1863 $r $TUNEFS --dryrun $fs2mdsdev || {
1864 error_noexit "tunefs.lustre before mounting the MDT"
1868 echo "mount new MDT....$fs2mdsdev"
1869 $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
1870 error_noexit "mount mdt1 failed"
1874 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1875 error_noexit "enable remote dir create failed"
1877 shall_cleanup_mdt1=true
1880 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
1881 error_noexit "Getting MDT UUID"
1884 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
1885 error_noexit "Unexpected MDT UUID: \"$uuid\""
1889 $r $TUNEFS --dryrun $ost_dev || {
1890 error_noexit "tunefs.lustre before mounting the OST"
1893 if [ "$writeconf" ]; then
1894 mopts=mgsnode=$nid,$writeconf
1895 if [ $fstype == "ldiskfs" ]; then
1897 $r $TUNEFS --quota $ost_dev || {
1899 error_noexit "Enable ost quota feature"
1905 if [ $fstype == "ldiskfs" ]; then
1909 $r $MOUNT_CMD -o $mopts $ost_dev $tmp/mnt/ost || {
1910 error_noexit "Mounting the OST"
1913 shall_cleanup_ost=true
1915 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
1916 error_noexit "Getting OST UUID"
1919 if [ "$uuid" != $fsname-OST0000_UUID ]; then
1920 error_noexit "Unexpected OST UUID: \"$uuid\""
1924 $r $LCTL conf_param $fsname-OST0000.osc.max_dirty_mb=15 || {
1925 error_noexit "Setting \"max_dirty_mb\""
1928 $r $LCTL conf_param $fsname-OST0000.failover.node=$nid || {
1929 error_noexit "Setting OST \"failover.node\""
1932 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
1933 error_noexit "Setting \"max_rpcs_in_flight\""
1936 $r $LCTL conf_param $fsname-MDT0000.failover.node=$nid || {
1937 error_noexit "Setting MDT \"failover.node\""
1940 $r $LCTL pool_new $fsname.interop || {
1941 error_noexit "Setting \"interop\""
1944 $r $LCTL conf_param $fsname-MDT0000.lov.stripesize=4M || {
1945 error_noexit "Setting \"lov.stripesize\""
1948 $r $LCTL conf_param $fsname-MDT0000.mdd.atime_diff=70 || {
1949 error_noexit "Setting \"mdd.atime_diff\""
1953 if [ "$ff_convert" != "no" -a $(facet_fstype ost1) == "ldiskfs" ]; then
1954 $r $LCTL lfsck_start -M $fsname-OST0000 || {
1955 error_noexit "Start OI scrub on OST0"
1959 # The oi_scrub should be on ost1, but for test_32(),
1960 # all on the SINGLEMDS.
1961 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
1962 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1963 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
1964 error_noexit "Failed to get the expected 'completed'"
1968 local UPDATED=$($r $LCTL get_param -n \
1969 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1970 awk '/^updated/ { print $2 }')
1971 [ $UPDATED -ge 1 ] || {
1972 error_noexit "Only $UPDATED objects have been converted"
1977 if [ "$dne_upgrade" != "no" ]; then
1978 $r $LCTL conf_param \
1979 $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
1980 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
1983 $r $LCTL conf_param $fsname-MDT0001.failover.node=$nid || {
1984 error_noexit "Setting MDT1 \"failover.node\""
1987 $r $LCTL conf_param $fsname-MDT0001.lov.stripesize=4M || {
1988 error_noexit "Setting MDT1 \"lov.stripesize\""
1993 if [ "$writeconf" ]; then
1994 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
1995 error_noexit "Mounting the client"
1999 shall_cleanup_lustre=true
2000 $r $LCTL set_param debug="$PTLDEBUG"
2002 t32_verify_quota $node $fsname $tmp/mnt/lustre || {
2003 error_noexit "verify quota failed"
2007 if $r test -f $tmp/list; then
2009 # There is not a Test Framework API to copy files to or
2010 # from a remote node.
2012 # LU-2393 - do both sorts on same node to ensure locale
2014 local list_file=$tmp/list
2016 if $mdt2_is_available; then
2017 if [[ -d $tmp/mnt/lustre/striped_dir_old ]] &&
2018 $r test -f $tmp/list2; then
2019 list_file=$tmp/list2
2020 pushd $tmp/mnt/lustre/striped_dir_old
2022 pushd $tmp/mnt/lustre/remote_dir
2025 pushd $tmp/mnt/lustre
2027 $r cat $list_file | sort -k 6 >$tmp/list.orig
2028 ls -Rni --time-style=+%s | sort -k 6 |
2029 sed 's/\. / /' >$tmp/list || {
2035 # 32-bit and 64-bit clients use different algorithms to
2036 # convert FIDs into inode numbers. Hence, remove the
2037 # inode numbers from the lists, if the original list was
2038 # created on an architecture with different number of
2041 if [ $(t32_bits_per_long $(uname -m)) != \
2042 $(t32_bits_per_long $img_arch) ]; then
2043 echo "Different number of bits per \"long\"" \
2044 "from the disk image"
2045 for list in list.orig list; do
2046 sed -i -e 's/^[0-9]\+[ \t]\+//' \
2050 if ! diff -ub $tmp/list.orig $tmp/list; then
2051 error_noexit "list verification failed"
2055 echo "list verification skipped"
2058 if [ "$dne_upgrade" != "no" ]; then
2059 $LFS mkdir -i 1 -c2 $tmp/mnt/lustre/striped_dir || {
2060 error_noexit "set striped dir failed"
2064 $LFS setdirstripe -D -c2 $tmp/mnt/lustre/striped_dir
2066 pushd $tmp/mnt/lustre
2067 tar -cf - . --exclude=./striped_dir \
2068 --exclude=./striped_dir_old \
2069 --exclude=./remote_dir |
2070 tar -xvf - -C striped_dir 1>/dev/null || {
2071 error_noexit "cp to striped dir failed"
2077 # If it is upgrade from DNE (2.5), then rename the remote dir,
2078 # which is created in 2.5 to striped dir.
2079 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]]; then
2080 stripe_index=$($LFS getdirstripe -i \
2081 $tmp/mnt/lustre/remote_dir)
2083 [[ $stripe_index -eq 1 ]] || {
2084 error_noexit "get index \"$stripe_index\"" \
2085 "from remote dir failed"
2088 mv $tmp/mnt/lustre/remote_dir \
2089 $tmp/mnt/lustre/striped_dir/ || {
2090 error_noexit "mv remote dir failed"
2095 # If it is upgraded from DNE (2.7), then move the striped dir
2096 # which was created in 2.7 to the new striped dir.
2097 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]] &&
2098 [[ -d $tmp/mnt/lustre/striped_dir_old ]]; then
2099 stripe_count=$($LFS getdirstripe -c \
2100 $tmp/mnt/lustre/striped_dir_old)
2101 [[ $stripe_count -eq 2 ]] || {
2102 error_noexit "get count $stripe_count" \
2103 "from striped dir failed"
2106 mv $tmp/mnt/lustre/striped_dir_old \
2107 $tmp/mnt/lustre/striped_dir/ || {
2108 error_noexit "mv striped dir failed"
2114 $r $LCTL set_param -n osd*.*.force_sync=1
2115 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
2116 error_noexit "dd failed"
2119 rm -rf $tmp/mnt/lustre/tmp_file || {
2120 error_noexit "rm failed"
2124 if $r test -f $tmp/sha1sums; then
2125 # LU-2393 - do both sorts on same node to ensure locale
2127 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
2128 if [ "$dne_upgrade" != "no" ]; then
2129 pushd $tmp/mnt/lustre/striped_dir
2131 pushd $tmp/mnt/lustre
2134 find ! -path "*remote_dir*" ! -path "*striped_dir*" \
2135 ! -name .lustre -type f -exec sha1sum {} \; |
2136 sort -k 2 >$tmp/sha1sums || {
2138 error_noexit "sha1sum"
2142 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
2143 error_noexit "sha1sum verification failed"
2147 # if upgrade from DNE(2.5), then check remote directory
2148 # if upgrade from DNE(2.7), then check striped directory
2149 if $mdt2_is_available &&
2150 [[ "$dne_upgrade" != "no" ]]; then
2151 local new_dir="$tmp/mnt/lustre/striped_dir"
2152 local striped_dir_old="$new_dir/striped_dir_old"
2154 local dir_list="$new_dir/remote_dir"
2155 [[ ! -d $triped_dir_old ]] ||
2156 dir_list+=" $striped_dir_old"
2158 for dir in $dir_list; do
2160 find ! -name .lustre -type f \
2161 -exec sha1sum {} \; |
2162 sort -k 2 >$tmp/sha1sums || {
2164 error_noexit "sha1sum"
2168 if ! diff -ub $tmp/sha1sums.orig \
2170 error_noexit "sha1sum $dir" \
2177 echo "sha1sum verification skipped"
2180 if [ "$dne_upgrade" != "no" ]; then
2181 rm -rf $tmp/mnt/lustre/striped_dir || {
2182 error_noexit "remove remote dir failed"
2187 # migrate files/dirs to remote MDT, then move them back
2188 if [ $(lustre_version_code mds1) -ge $(version_code 2.7.50) -a \
2189 $dne_upgrade != "no" ]; then
2190 $r $LCTL set_param -n \
2191 mdt.${fsname}*.enable_remote_dir=1 2>/dev/null
2193 echo "test migration"
2194 pushd $tmp/mnt/lustre
2195 for dir in $(find ! -name .lustre ! -name . -type d); do
2196 mdt_index=$($LFS getdirstripe -i $dir)
2197 stripe_cnt=$($LFS getdirstripe -c $dir)
2198 if [ $mdt_index = 0 -a $stripe_cnt -le 1 ]; then
2199 $LFS mv -M 1 $dir || {
2201 error_noexit "migrate MDT1 failed"
2207 for dir in $(find ! -name . ! -name .lustre -type d); do
2208 mdt_index=$($LFS getdirstripe -i $dir)
2209 stripe_cnt=$($LFS getdirstripe -c $dir)
2210 if [ $mdt_index = 1 -a $stripe_cnt -le 1 ]; then
2211 $LFS mv -M 0 $dir || {
2213 error_noexit "migrate MDT0 failed"
2222 # When adding new data verification tests, please check for
2223 # the presence of the required reference files first, like
2224 # the "sha1sums" and "list" tests above, to avoid the need to
2225 # regenerate every image for each test addition.
2228 nrpcs_orig=$($LCTL get_param \
2229 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
2230 error_noexit "Getting \"max_rpcs_in_flight\""
2233 nrpcs=$((nrpcs_orig + 5))
2234 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=$nrpcs || {
2235 error_noexit "Changing \"max_rpcs_in_flight\""
2238 wait_update $HOSTNAME "$LCTL get_param \
2239 -n mdc.*MDT0000*.max_rpcs_in_flight" $nrpcs || {
2240 error_noexit "Verifying \"max_rpcs_in_flight\""
2244 umount $tmp/mnt/lustre || {
2245 error_noexit "Unmounting the client"
2248 shall_cleanup_lustre=false
2250 if [[ "$dne_upgrade" != "no" ]] || $mdt2_is_available; then
2251 $r $UMOUNT $tmp/mnt/mdt1 || {
2252 error_noexit "Unmounting the MDT2"
2255 shall_cleanup_mdt1=false
2258 $r $UMOUNT $tmp/mnt/mdt || {
2259 error_noexit "Unmounting the MDT"
2262 shall_cleanup_mdt=false
2264 $r $UMOUNT $tmp/mnt/ost || {
2265 error_noexit "Unmounting the OST"
2268 shall_cleanup_ost=false
2270 t32_reload_modules $node || {
2271 error_noexit "Reloading modules"
2275 if [[ $fstype == zfs ]]; then
2276 local poolname=t32fs-mdt1
2278 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
2279 $ZPOOL import -f -d $tmp $poolname"
2282 # mount a second time to make sure we didnt leave upgrade flag on
2283 $r $TUNEFS --dryrun $mdt_dev || {
2285 error_noexit "tunefs.lustre before remounting the MDT"
2289 mopts=exclude=$fsname-OST0000
2290 if [ $fstype == "ldiskfs" ]; then
2293 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2294 error_noexit "Remounting the MDT"
2297 shall_cleanup_mdt=true
2307 for tarball in $tarballs; do
2308 t32_test $tarball || let "rc += $?"
2312 run_test 32a "Upgrade (not live)"
2320 for tarball in $tarballs; do
2321 t32_test $tarball writeconf || let "rc += $?"
2325 run_test 32b "Upgrade with writeconf"
2332 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
2334 for tarball in $tarballs; do
2335 # Do not support 1_8 and 2_1 direct upgrade to DNE2 anymore */
2336 echo $tarball | grep "1_8" && continue
2337 echo $tarball | grep "2_1" && continue
2339 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
2343 run_test 32c "dne upgrade test"
2351 for tarball in $tarballs; do
2352 ff_convert=yes t32_test $tarball || rc=$?
2356 run_test 32d "convert ff test"
2358 test_33a() { # bug 12333, was test_33
2359 local FSNAME2=test-123
2360 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2363 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
2365 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
2366 local dev=${SINGLEMDS}_dev
2367 local MDSDEV=${!dev}
2368 is_blkdev $SINGLEMDS $MDSDEV &&
2369 skip_env "mixed loopback and real device not working" &&
2373 local fs2mdsdev=$(mdsdevname 1_2)
2374 local fs2ostdev=$(ostdevname 1_2)
2375 local fs2mdsvdev=$(mdsvdevname 1_2)
2376 local fs2ostvdev=$(ostvdevname 1_2)
2378 if [ $(facet_fstype mds1) == ldiskfs ]; then
2379 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
2382 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2383 --reformat $mkfsoptions $fs2mdsdev $fs2mdsvdev || exit 10
2384 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2385 --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \
2386 $fs2ostvdev || exit 10
2388 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
2389 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2390 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" ||
2391 error "$LCTL conf_param $FSNAME2.sys.timeout=200 failed"
2392 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2393 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2396 cp /etc/hosts $MOUNT2/ || error "copy /etc/hosts $MOUNT2/ failed"
2397 $GETSTRIPE $MOUNT2/hosts || error "$GETSTRIPE $MOUNT2/hosts failed"
2402 cleanup_nocli || error "cleanup_nocli failed with $?"
2404 run_test 33a "Mount ost with a large index number"
2406 test_33b() { # was test_34
2409 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
2410 # Drop lock cancelation reply during umount
2411 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
2412 do_facet client $LCTL set_param fail_loc=0x80000304
2413 #lctl set_param debug=-1
2414 umount_client $MOUNT
2415 cleanup || error "cleanup failed with $?"
2417 run_test 33b "Drop cancel during umount"
2421 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
2422 manual_umount_client
2424 do_facet client killall -USR1 multiop
2425 if [ $rc -eq 0 ]; then
2426 error "umount not fail!"
2429 cleanup || error "cleanup failed with rc $?"
2431 run_test 34a "umount with opened file should be fail"
2435 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2436 stop_mds || error "Unable to stop MDS"
2438 manual_umount_client --force || error "mtab after failed umount with $?"
2440 cleanup || error "cleanup failed with $?"
2442 run_test 34b "force umount with failed mds should be normal"
2446 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2447 stop_ost || error "Unable to stop OST1"
2449 manual_umount_client --force || error "mtab after failed umount with $?"
2451 cleanup || error "cleanup failed with $?"
2453 run_test 34c "force umount with failed ost should be normal"
2455 test_35a() { # bug 12459
2458 DBG_SAVE="`$LCTL get_param -n debug`"
2459 $LCTL set_param debug="ha"
2461 log "Set up a fake failnode for the MDS"
2463 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2464 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2465 do_facet mgs "$LCTL conf_param \
2466 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" ||
2467 error "Setting ${device}.failover.node=\
2468 $(h2$NETTYPE $FAKENID) failed."
2470 log "Wait for RECONNECT_INTERVAL seconds (10s)"
2473 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
2476 log "Stopping the MDT: $device"
2477 stop_mdt 1 || error "MDT0 stop fail"
2479 df $MOUNT > /dev/null 2>&1 &
2481 log "Restarting the MDT: $device"
2482 start_mdt 1 || error "MDT0 start fail"
2483 log "Wait for df ($DFPID) ... "
2486 $LCTL set_param debug="$DBG_SAVE"
2488 # retrieve from the log the first server that the client tried to
2489 # contact after the connection loss
2490 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2491 NEXTCONN=`awk "/${MSG}/ {start = 1;}
2492 /import_select_connection.*$device-mdc.* using connection/ {
2494 if (\\\$NF ~ /$FAKENID/)
2500 }" $TMP/lustre-log-$TESTNAME.log`
2501 [ "$NEXTCONN" != "0" ] &&
2502 error "Tried to connect to ${NEXTCONN} not last active server"
2503 cleanup || error "cleanup failed with $?"
2504 # remove nid settings
2505 writeconf_or_reformat
2507 run_test 35a "Reconnect to the last active server first"
2509 test_35b() { # bug 18674
2510 remote_mds || { skip "local MDS" && return 0; }
2514 $LCTL set_param debug="ha"
2516 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
2519 log "Set up a fake failnode for the MDS"
2521 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2522 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2523 do_facet mgs "$LCTL conf_param \
2524 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" ||
2525 error "Set ${device}.failover.node=\
2526 $(h2$NETTYPE $FAKENID) failed"
2528 local at_max_saved=0
2529 # adaptive timeouts may prevent seeing the issue
2530 if at_is_enabled; then
2531 at_max_saved=$(at_max_get mds)
2532 at_max_set 0 mds client
2535 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
2537 log "Injecting EBUSY on MDS"
2538 # Setting OBD_FAIL_MDS_RESEND=0x136
2539 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" ||
2540 error "unable to set param fail_loc=0x80000136"
2542 $LCTL set_param mdc.${FSNAME}*.stats=clear
2544 log "Creating a test file and stat it"
2545 touch $MOUNT/$tdir/$tfile || error "touch $MOUNT/$tdir/$tfile failed"
2546 stat $MOUNT/$tdir/$tfile
2548 log "Stop injecting EBUSY on MDS"
2549 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" ||
2550 error "unable to set param fail_loc=0"
2551 rm -f $MOUNT/$tdir/$tfile || error "remove $MOUNT/$tdir/$tfile failed"
2554 # restore adaptive timeout
2555 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
2557 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2559 CONNCNT=$($LCTL get_param mdc.${FSNAME}*.stats |
2560 awk '/mds_connect/{print $2}')
2562 # retrieve from the log if the client has ever tried to
2563 # contact the fake server after the loss of connection
2564 FAILCONN=`awk "BEGIN {ret = 0;}
2565 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
2567 if (\\\$NF ~ /$FAKENID/) {
2572 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
2574 [ "$FAILCONN" == "0" ] &&
2575 error "The client reconnection has not been triggered"
2576 [ "$FAILCONN" == "2" ] &&
2577 error "Primary server busy, client reconnect to failover failed"
2580 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
2581 # Reconnects are supposed to be rate limited to one every 5s
2582 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] &&
2583 error "Too many reconnects $CONNCNT"
2585 cleanup || error "cleanup failed with $?"
2586 # remove nid settings
2587 writeconf_or_reformat
2589 run_test 35b "Continue reconnection retries, if the active server is busy"
2592 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
2594 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] ||
2595 { skip "remote OST" && return 0; }
2598 local FSNAME2=test1234
2599 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2601 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
2603 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
2604 is_blkdev $SINGLEMDS $MDSDEV &&
2605 skip_env "mixed loopback and real device not working" && return
2608 local fs2mdsdev=$(mdsdevname 1_2)
2609 local fs2ostdev=$(ostdevname 1_2)
2610 local fs3ostdev=$(ostdevname 2_2)
2611 local fs2mdsvdev=$(mdsvdevname 1_2)
2612 local fs2ostvdev=$(ostvdevname 1_2)
2613 local fs3ostvdev=$(ostvdevname 2_2)
2615 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2616 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
2617 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
2618 # different one than the default value here.
2619 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2620 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
2621 add fs3ost $(mkfs_opts ost2 ${fs3ostdev}) --mgsnode=$MGSNID \
2622 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
2624 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
2625 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2626 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
2627 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2628 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2630 sleep 5 # until 11778 fixed
2632 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || error "dd failed"
2634 BKTOTAL=$($LCTL get_param -n obdfilter.*.kbytestotal |
2635 awk 'BEGIN{total=0}; {total+=$1}; END{print total}')
2636 BKFREE=$($LCTL get_param -n obdfilter.*.kbytesfree |
2637 awk 'BEGIN{free=0}; {free+=$1}; END{print free}')
2638 BKAVAIL=$($LCTL get_param -n obdfilter.*.kbytesavail |
2639 awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}')
2640 STRING=$(df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}')
2641 DFTOTAL=$(echo $STRING | cut -d, -f1)
2642 DFUSED=$(echo $STRING | cut -d, -f2)
2643 DFAVAIL=$(echo $STRING | cut -d, -f3)
2644 DFFREE=$(($DFTOTAL - $DFUSED))
2646 ALLOWANCE=$((64 * $OSTCOUNT))
2648 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
2649 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
2650 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
2653 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
2654 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
2655 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
2658 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
2659 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
2660 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
2665 stop fs3ost -f || error "unable to stop OST3"
2666 stop fs2ost -f || error "unable to stop OST2"
2667 stop fs2mds -f || error "unable to stop second MDS"
2668 unload_modules_conf || error "unable unload modules"
2671 run_test 36 "df report consistency on OSTs with different block size"
2674 local mntpt=$(facet_mntpt $SINGLEMDS)
2675 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
2676 local mdsdev_sym="$TMP/sym_mdt.img"
2677 local opts=$MDS_MOUNT_OPTS
2680 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2681 skip "Currently only applicable to ldiskfs-based MDTs"
2685 echo "MDS : $mdsdev"
2686 echo "SYMLINK : $mdsdev_sym"
2687 do_facet $SINGLEMDS rm -f $mdsdev_sym
2689 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
2691 echo "mount symlink device - $mdsdev_sym"
2693 if ! do_facet $SINGLEMDS test -b $mdsdev; then
2694 opts=$(csa_add "$opts" -o loop)
2696 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
2697 $mdsdev_sym $mntpt 2>&1)
2700 echo mount_op=$mount_op
2702 do_facet $SINGLEMDS "$UMOUNT $mntpt && rm -f $mdsdev_sym"
2704 if $(echo $mount_op | grep -q "unable to set tunable"); then
2705 error "set tunables failed for symlink device"
2708 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
2710 run_test 37 "verify set tunables works for symlink device"
2712 test_38() { # bug 14222
2713 local fstype=$(facet_fstype $SINGLEMDS)
2714 local mntpt=$(facet_mntpt $SINGLEMDS)
2719 local SRC="/etc /bin"
2720 local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT)
2721 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
2722 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2723 tar cf - $FILES | tar xf - -C $DIR/$tdir ||
2724 error "copying $SRC to $DIR/$tdir"
2726 umount_client $MOUNT || error "umount_client $MOUNT failed"
2727 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2728 stop_mds || error "Unable to stop MDS"
2729 log "delete lov_objid file on MDS"
2731 mount_fstype $SINGLEMDS || error "mount MDS failed (1)"
2733 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid; rm $mntpt/lov_objid"
2735 unmount_fstype $SINGLEMDS || error "umount failed (1)"
2737 # check create in mds_lov_connect
2738 start_mds || error "unable to start MDS"
2739 mount_client $MOUNT || error "mount_client $MOUNT failed"
2741 [ $V ] && log "verifying $DIR/$tdir/$f"
2742 diff -q $f $DIR/$tdir/$f || ERROR=y
2744 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2745 if [ "$ERROR" = "y" ]; then
2746 # check it's updates in sync
2747 umount_client $MOUNT
2749 mount_fstype $SIGNLEMDS
2750 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2751 unmount_fstype $SINGLEMDS
2752 error "old and new files are different after connect" || true
2754 touch $DIR/$tdir/f2 || error "f2 file create failed"
2756 # check it's updates in sync
2757 umount_client $MOUNT || error "second umount_client $MOUNT failed"
2760 mount_fstype $SINGLEMDS || error "mount MDS failed (3)"
2762 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2763 do_facet $SINGLEMDS dd if=/dev/zero of=$mntpt/lov_objid.clear count=8
2765 unmount_fstype $SINGLEMDS || error "umount failed (3)"
2767 start_mds || error "unable to start MDS"
2768 mount_client $MOUNT || error "mount_client $MOUNT failed"
2770 [ $V ] && log "verifying $DIR/$tdir/$f"
2771 diff -q $f $DIR/$tdir/$f || ERROR=y
2773 touch $DIR/$tdir/f3 || error "f3 file create failed"
2774 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2775 umount_client $MOUNT || error "third umount_client $MOUNT failed"
2777 mount_fstype $SINGLEMDS || error "mount MDS failed (4)"
2778 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2779 unmount_fstype $SINGLEMDS || error "umount failed (4)"
2781 [ "$ERROR" = "y" ] &&
2782 error "old and new files are different after sync" || true
2784 log "files compared the same"
2785 cleanup || error "cleanup failed with $?"
2787 run_test 38 "MDS recreates missing lov_objid file from OST data"
2792 cleanup || error "cleanup failed with $?"
2793 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
2794 error "memory leak detected" || true
2796 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
2798 test_40() { # bug 15759
2799 start_ost || error "Unable to start OST1"
2800 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
2801 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
2803 cleanup || error "cleanup failed with rc $?"
2805 run_test 40 "race during service thread startup"
2807 test_41a() { #bug 14134
2808 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2809 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2810 skip "Loop devices does not work with nosvc option"
2815 { skip "needs combined MGT and MDT device" && return 0; }
2817 start_mdt 1 -o nosvc -n
2818 if [ $MDSCOUNT -ge 2 ]; then
2819 for num in $(seq 2 $MDSCOUNT); do
2820 start_mdt $num || return
2823 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
2824 start_mdt 1 -o nomgs,force
2825 mount_client $MOUNT || error "mount_client $MOUNT failed"
2828 echo "blah blah" > $MOUNT/$tfile
2831 umount_client $MOUNT || error "umount_client $MOUNT failed"
2832 stop ost1 -f || error "unable to stop OST1"
2833 stop_mds || error "Unable to stop MDS"
2834 stop_mds || error "Unable to stop MDS on second try"
2836 run_test 41a "mount mds with --nosvc and --nomgs"
2839 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2840 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2841 skip "Loop devices does not work with nosvc option"
2845 ! combined_mgs_mds && skip "needs combined mgs device" && return 0
2849 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2851 start_mdt 1 -o nosvc -n
2852 if [ $MDSCOUNT -ge 2 ]; then
2853 for num in $(seq 2 $MDSCOUNT); do
2854 start_mdt $num || return
2857 start_ost || error "Unable to start OST1"
2858 start_mdt 1 -o nomgs,force
2859 mount_client $MOUNT || error "mount_client $MOUNT failed"
2862 echo "blah blah" > $MOUNT/$tfile
2863 cat $MOUNT/$tfile || error "cat $MOUNT/$tfile failed"
2865 umount_client $MOUNT || error "umount_client $MOUNT failed"
2866 stop_ost || error "Unable to stop OST1"
2867 stop_mds || error "Unable to stop MDS"
2868 stop_mds || error "Unable to stop MDS on second try"
2870 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
2873 local server_version=$(lustre_version_code $SINGLEMDS)
2874 local oss_list=$(comma_list $(osts_nodes))
2876 [[ $server_version -ge $(version_code 2.6.52) ]] ||
2877 [[ $server_version -ge $(version_code 2.5.26) &&
2878 $server_version -lt $(version_code 2.5.50) ]] ||
2879 [[ $server_version -ge $(version_code 2.5.4) &&
2880 $server_version -lt $(version_code 2.5.11) ]] ||
2881 { skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"; return; }
2883 # ensure mds1 ost1 have been created even if running sub-test standalone
2886 cleanup || error "cleanup failed"
2888 # using directly mount command instead of start() function to avoid
2889 # any side effect of // with others/externals tools/features
2890 # ("zpool import", ...)
2892 # MDT concurrent start
2894 LOAD_MODULES_REMOTE=true load_modules
2895 do_facet $SINGLEMDS "lsmod | grep -q libcfs" ||
2896 error "MDT concurrent start: libcfs module not loaded"
2898 local mds1dev=$(mdsdevname 1)
2899 local mds1mnt=$(facet_mntpt mds1)
2900 local mds1fstype=$(facet_fstype mds1)
2901 local mds1opts=$MDS_MOUNT_OPTS
2903 if [ $mds1fstype == ldiskfs ] &&
2904 ! do_facet mds1 test -b $mds1dev; then
2905 mds1opts=$(csa_add "$mds1opts" -o loop)
2907 if [[ $mds1fstype == zfs ]]; then
2908 import_zpool mds1 || return ${PIPESTATUS[0]}
2911 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
2912 do_facet mds1 "$LCTL set_param fail_loc=0x80000716"
2914 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts &
2917 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts
2921 do_facet mds1 "$LCTL set_param fail_loc=0x0"
2922 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
2923 echo "1st MDT start succeed"
2924 echo "2nd MDT start failed with $rc2"
2925 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
2926 echo "1st MDT start failed with $rc"
2927 echo "2nd MDT start succeed"
2930 error "unexpected concurrent MDT mounts result, rc=$rc rc2=$rc2"
2933 if [ $MDSCOUNT -ge 2 ]; then
2934 for num in $(seq 2 $MDSCOUNT); do
2935 start_mdt $num || return
2939 # OST concurrent start
2941 do_rpc_nodes $oss_list "lsmod | grep -q libcfs" ||
2942 error "OST concurrent start: libcfs module not loaded"
2944 local ost1dev=$(ostdevname 1)
2945 local ost1mnt=$(facet_mntpt ost1)
2946 local ost1fstype=$(facet_fstype ost1)
2947 local ost1opts=$OST_MOUNT_OPTS
2949 if [ $ost1fstype == ldiskfs ] &&
2950 ! do_facet ost1 test -b $ost1dev; then
2951 ost1opts=$(csa_add "$ost1opts" -o loop)
2953 if [[ $ost1fstype == zfs ]]; then
2954 import_zpool ost1 || return ${PIPESTATUS[0]}
2957 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
2958 do_facet ost1 "$LCTL set_param fail_loc=0x80000716"
2960 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts &
2963 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts
2967 do_facet ost1 "$LCTL set_param fail_loc=0x0"
2968 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
2969 echo "1st OST start succeed"
2970 echo "2nd OST start failed with $rc2"
2971 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
2972 echo "1st OST start failed with $rc"
2973 echo "2nd OST start succeed"
2977 error "unexpected concurrent OST mounts result, rc=$rc rc2=$rc2"
2983 # verify everything ok
2988 error "MDT(s) start failed"
2996 error "OST(s) start failed"
3004 error "client start failed"
3011 error "client mount failed"
3015 run_test 41c "concurrent mounts of MDT/OST should all fail but one"
3017 test_42() { #bug 14693
3019 check_mount || error "client was not mounted"
3021 do_facet mgs $LCTL conf_param $FSNAME.llite.some_wrong_param=10
3022 umount_client $MOUNT ||
3023 error "unmounting client failed with invalid llite param"
3024 mount_client $MOUNT ||
3025 error "mounting client failed with invalid llite param"
3027 do_facet mgs $LCTL conf_param $FSNAME.sys.some_wrong_param=20
3028 cleanup || error "stopping $FSNAME failed with invalid sys param"
3030 check_mount || error "client was not mounted with invalid sys param"
3031 cleanup || error "stopping $FSNAME failed with invalid sys param"
3033 run_test 42 "allow client/server mount/unmount with invalid config param"
3036 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.58) ]] ||
3037 { skip "Need MDS version at least 2.5.58" && return 0; }
3038 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
3041 USER1=$(getent passwd | grep :$ID1:$ID1: | cut -d: -f1)
3042 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1" &&
3046 chmod ugo+x $DIR || error "chmod 0 failed"
3047 set_conf_param_and_check mds1 \
3048 "$LCTL get_param -n mdt.$FSNAME-MDT0000.root_squash" \
3049 "$FSNAME.mdt.root_squash" \
3051 wait_update $HOSTNAME \
3052 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3054 error "check llite root_squash failed!"
3055 set_conf_param_and_check mds1 \
3056 "$LCTL get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
3057 "$FSNAME.mdt.nosquash_nids" \
3059 wait_update $HOSTNAME \
3060 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3062 error "check llite nosquash_nids failed!"
3065 # create set of test files
3067 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
3068 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
3069 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
3071 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
3072 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
3074 mkdir $DIR/$tdir-rootdir || error "mkdir failed"
3075 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
3076 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
3078 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
3079 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
3080 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
3083 # check root_squash:
3084 # set root squash UID:GID to RUNAS_ID
3085 # root should be able to access only files owned by RUNAS_ID
3087 set_conf_param_and_check mds1 \
3088 "$LCTL get_param -n mdt.$FSNAME-MDT0000.root_squash" \
3089 "$FSNAME.mdt.root_squash" \
3090 "$RUNAS_ID:$RUNAS_ID"
3091 wait_update $HOSTNAME \
3092 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3093 "$RUNAS_ID:$RUNAS_ID" ||
3094 error "check llite root_squash failed!"
3096 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
3097 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3098 error "$ST: root read permission is denied"
3099 echo "$ST: root read permission is granted - ok"
3102 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3103 error "$ST: root write permission is denied"
3104 echo "$ST: root write permission is granted - ok"
3106 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3107 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3108 error "$ST: root read permission is granted"
3109 echo "$ST: root read permission is denied - ok"
3112 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3113 error "$ST: root write permission is granted"
3114 echo "$ST: root write permission is denied - ok"
3116 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3117 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null &&
3118 error "$ST: root unlink permission is granted"
3119 echo "$ST: root unlink permission is denied - ok"
3121 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null &&
3122 error "$ST: root create permission is granted"
3123 echo "$ST: root create permission is denied - ok"
3126 # check root_squash is enforced independently
3127 # of client cache content
3129 # access file by USER1, keep access open
3130 # root should be denied access to user file
3132 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
3136 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
3137 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3138 { kill $pid; error "$ST: root read permission is granted"; }
3139 echo "$ST: root read permission is denied - ok"
3142 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3143 { kill $pid; error "$ST: root write permission is granted"; }
3144 echo "$ST: root write permission is denied - ok"
3150 # check nosquash_nids:
3151 # put client's NID into nosquash_nids list,
3152 # root should be able to access root file after that
3154 local NIDLIST=$($LCTL list_nids all | tr '\n' ' ')
3155 NIDLIST="2@gni $NIDLIST 192.168.0.[2,10]@tcp"
3156 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
3157 set_conf_param_and_check mds1 \
3158 "$LCTL get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
3159 "$FSNAME-MDTall.mdt.nosquash_nids" \
3161 wait_update $HOSTNAME \
3162 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3164 error "check llite nosquash_nids failed!"
3166 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3167 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3168 error "$ST: root read permission is denied"
3169 echo "$ST: root read permission is granted - ok"
3172 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3173 error "$ST: root write permission is denied"
3174 echo "$ST: root write permission is granted - ok"
3176 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3177 rm $DIR/$tdir-rootdir/tfile-1 ||
3178 error "$ST: root unlink permission is denied"
3179 echo "$ST: root unlink permission is granted - ok"
3180 touch $DIR/$tdir-rootdir/tfile-2 ||
3181 error "$ST: root create permission is denied"
3182 echo "$ST: root create permission is granted - ok"
3183 cleanup || error "cleanup failed with $?"
3185 run_test 43a "check root_squash and nosquash_nids"
3187 test_43b() { # LU-5690
3188 [[ $(lustre_version_code mgs) -ge $(version_code 2.7.62) ]] ||
3189 { skip "Need MGS version 2.7.62+"; return; }
3191 if [[ -z "$fs2mds_DEV" ]]; then
3192 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
3193 skip_env "mixed loopback and real device not working" && return
3196 local fs2mdsdev=$(mdsdevname 1_2)
3197 local fs2mdsvdev=$(mdsvdevname 1_2)
3199 # temporarily use fs2mds as fs2mgs
3201 local fs2mgsdev=$fs2mdsdev
3202 local fs2mgsvdev=$fs2mdsvdev
3204 local fsname=test1234
3206 load_module llite/lustre
3207 local client_ip=$(host_nids_address $HOSTNAME $NETTYPE)
3208 local host=${client_ip//*./}
3209 local net=${client_ip/%$host/}
3210 local nosquash_nids=$(h2$NETTYPE $net[$host,$host,$host])
3212 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
3213 --param mdt.root_squash=$RUNAS_ID:$RUNAS_ID \
3214 --param mdt.nosquash_nids=$nosquash_nids \
3215 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
3216 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS || error "start fs2mgs failed"
3217 stop $fs2mgs -f || error "stop fs2mgs failed"
3219 run_test 43b "parse nosquash_nids with commas in expr_list"
3221 umount_client $MOUNT
3226 check_mount || error "check_mount"
3227 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
3229 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
3230 for VAL in $UUIDS; do
3231 NID=$(echo $VAL | cut -d= -f1)
3232 CLUUID=$(echo $VAL | cut -d= -f2)
3233 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
3235 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
3236 cleanup || error "cleanup failed with $?"
3238 run_test 44 "mounted client proc entry exists"
3242 check_mount || error "check_mount"
3243 stop_mds || error "Unable to stop MDS"
3247 #define OBD_FAIL_PTLRPC_LONG_REPL_UNLINK 0x50f
3248 do_facet client "$LCTL set_param fail_loc=0x8000050f"
3251 manual_umount_client --force || error "manual_umount_client failed"
3252 do_facet client "$LCTL set_param fail_loc=0x0"
3253 start_mds || error "unable to start MDS"
3254 mount_client $MOUNT || error "mount_client $MOUNT failed"
3255 cleanup || error "cleanup failed with $?"
3257 run_test 45 "long unlink handling in ptlrpcd"
3264 umount_client $MOUNT2 || rc=$?
3265 umount_client $MOUNT || rc=$?
3266 while [ $count -gt 0 ]; do
3267 stop ost${count} -f || rc=$?
3271 cleanup_nocli || rc=$?
3272 #writeconf to remove all ost2 traces for subsequent tests
3273 writeconf_or_reformat
3278 echo "Testing with $OSTCOUNT OSTs"
3280 start_mds || error "unable to start MDS"
3281 #first client should see only one ost
3282 start_ost || error "Unable to start OST1"
3283 wait_osc_import_state mds ost FULL
3285 mount_client $MOUNT || error "mount_client $MOUNT failed"
3286 trap "cleanup_46a $OSTCOUNT" EXIT ERR
3289 for (( i=2; i<=$OSTCOUNT; i++ )); do
3290 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
3291 error "start_ost$i $(ostdevname $i) failed"
3294 # wait until osts in sync
3295 for (( i=2; i<=$OSTCOUNT; i++ )); do
3296 wait_osc_import_state mds ost$i FULL
3297 wait_osc_import_state client ost$i FULL
3300 #second client see all ost's
3302 mount_client $MOUNT2 || error "mount_client failed"
3303 $SETSTRIPE -c -1 $MOUNT2 || error "$SETSTRIPE -c -1 $MOUNT2 failed"
3304 $GETSTRIPE $MOUNT2 || error "$GETSTRIPE $MOUNT2 failed"
3306 echo "ok" > $MOUNT2/widestripe
3307 $GETSTRIPE $MOUNT2/widestripe ||
3308 error "$GETSTRIPE $MOUNT2/widestripe failed"
3309 # fill acl buffer for avoid expand lsm to them
3310 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd |
3312 setfacl -m $acl $MOUNT2/widestripe
3316 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3318 cleanup_46a $OSTCOUNT || error "cleanup_46a failed"
3320 run_test 46a "handle ost additional - wide striped file"
3325 check_mount || error "check_mount failed"
3326 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
3330 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3331 if echo $ns | grep "MDT[[:digit:]]*"; then
3334 lrs=$(echo $ns | sed 's/.*lru_size=//')
3335 lru_size[count]=$lrs
3340 facet_failover $SINGLEMDS
3341 client_up || error "client_up failed"
3344 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3345 if echo $ns | grep "MDT[[:digit:]]*"; then
3348 lrs=$(echo $ns | sed 's/.*lru_size=//')
3349 if ! test "$lrs" -eq "${lru_size[count]}"; then
3350 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
3351 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
3356 cleanup || error "cleanup failed with $?"
3358 run_test 47 "server restart does not make client loss lru_resize settings"
3363 # reformat after this test is needed - if the test fails,
3364 # we will have unkillable file at FS
3368 test_48() { # bug 17636
3371 check_mount || error "check_mount failed"
3373 $SETSTRIPE -c -1 $MOUNT || error "$SETSTRIPE -c -1 $MOUNT failed"
3374 $GETSTRIPE $MOUNT || error "$GETSTRIPE $MOUNT failed"
3376 echo "ok" > $MOUNT/widestripe
3377 $GETSTRIPE $MOUNT/widestripe ||
3378 error "$GETSTRIPE $MOUNT/widestripe failed"
3380 trap cleanup_48 EXIT ERR
3382 # fill acl buffer for avoid expand lsm to them
3383 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
3384 setfacl -m $acl $MOUNT/widestripe
3387 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3391 run_test 48 "too many acls on file"
3393 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
3394 test_49a() { # bug 17710
3395 local timeout_orig=$TIMEOUT
3396 local ldlm_timeout_orig=$LDLM_TIMEOUT
3397 local LOCAL_TIMEOUT=20
3399 LDLM_TIMEOUT=$LOCAL_TIMEOUT
3400 TIMEOUT=$LOCAL_TIMEOUT
3404 check_mount || error "client mount failed"
3406 echo "check ldlm_timout..."
3407 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3408 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3409 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3411 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3412 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3415 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
3416 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
3419 umount_client $MOUNT || error "umount_client $MOUNT failed"
3420 stop_ost || error "problem stopping OSS"
3421 stop_mds || error "problem stopping MDS"
3423 LDLM_TIMEOUT=$ldlm_timeout_orig
3424 TIMEOUT=$timeout_orig
3426 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3428 test_49b() { # bug 17710
3429 local timeout_orig=$TIMEOUT
3430 local ldlm_timeout_orig=$LDLM_TIMEOUT
3431 local LOCAL_TIMEOUT=20
3433 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
3434 TIMEOUT=$LOCAL_TIMEOUT
3438 check_mount || error "client mount failed"
3440 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3441 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3442 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3444 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3445 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3448 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
3449 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
3452 cleanup || error "cleanup failed"
3454 LDLM_TIMEOUT=$ldlm_timeout_orig
3455 TIMEOUT=$timeout_orig
3457 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3460 # Test both statfs and lfs df and fail if either one fails
3461 multiop_bg_pause $1 f_
3464 killall -USR1 multiop
3465 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
3466 wait $PID || { RC1=$?; log "multiop return error "; }
3473 if [ $RC2 -eq 0 ]; then
3475 log "lazystatfs df failed"
3479 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
3485 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3486 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3488 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
3490 cleanup || error "cleanup failed with rc $?"
3492 run_test 50a "lazystatfs all servers available"
3496 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3497 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3499 # Wait for client to detect down OST
3500 stop_ost || error "Unable to stop OST1"
3501 wait_osc_import_state mds ost DISCONN
3503 lazystatfs $MOUNT || error "lazystatfs should not return EIO"
3505 umount_client $MOUNT || error "Unable to unmount client"
3506 stop_mds || error "Unable to stop MDS"
3508 run_test 50b "lazystatfs all servers down"
3511 start_mds || error "Unable to start MDS"
3512 start_ost || error "Unable to start OST1"
3513 start_ost2 || error "Unable to start OST2"
3514 mount_client $MOUNT || error "Unable to mount client"
3515 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3516 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3518 # Wait for client to detect down OST
3519 stop_ost || error "Unable to stop OST1"
3520 wait_osc_import_state mds ost DISCONN
3521 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3523 umount_client $MOUNT || error "Unable to unmount client"
3524 stop_ost2 || error "Unable to stop OST2"
3525 stop_mds || error "Unable to stop MDS"
3526 #writeconf to remove all ost2 traces for subsequent tests
3527 writeconf_or_reformat
3529 run_test 50c "lazystatfs one server down"
3532 start_mds || error "Unable to start MDS"
3533 start_ost || error "Unable to start OST1"
3534 start_ost2 || error "Unable to start OST2"
3535 mount_client $MOUNT || error "Unable to mount client"
3536 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3537 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3539 # Issue the statfs during the window where the client still
3540 # belives the OST to be available but it is in fact down.
3541 # No failure just a statfs which hangs for a timeout interval.
3542 stop_ost || error "Unable to stop OST1"
3543 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3545 umount_client $MOUNT || error "Unable to unmount client"
3546 stop_ost2 || error "Unable to stop OST2"
3547 stop_mds || error "Unable to stop MDS"
3548 #writeconf to remove all ost2 traces for subsequent tests
3549 writeconf_or_reformat
3551 run_test 50d "lazystatfs client/server conn race"
3558 start_mds || error "Unable to start MDS"
3559 #first client should see only one ost
3560 start_ost || error "Unable to start OST1"
3561 wait_osc_import_state mds ost FULL
3563 # Wait for client to detect down OST
3564 stop_ost || error "Unable to stop OST1"
3565 wait_osc_import_state mds ost DISCONN
3567 mount_client $MOUNT || error "Unable to mount client"
3568 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3570 multiop_bg_pause $MOUNT _f
3574 if [ $RC1 -ne 0 ]; then
3575 log "multiop failed $RC1"
3578 sleep $(( $TIMEOUT+1 ))
3580 [ $? -ne 0 ] && error "process isn't sleep"
3581 start_ost || error "Unable to start OST1"
3582 wait $pid || error "statfs failed"
3585 umount_client $MOUNT || error "Unable to unmount client"
3586 stop_ost || error "Unable to stop OST1"
3587 stop_mds || error "Unable to stop MDS"
3589 run_test 50e "normal statfs all servers down"
3594 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
3596 start_mds || error "Unable to start mds"
3597 #first client should see only one ost
3598 start_ost || error "Unable to start OST1"
3599 wait_osc_import_state mds ost FULL
3601 start_ost2 || error "Unable to start OST2"
3602 wait_osc_import_state mds ost2 FULL
3604 # Wait for client to detect down OST
3605 stop_ost2 || error "Unable to stop OST2"
3607 wait_osc_import_state mds ost2 DISCONN
3608 mount_client $MOUNT || error "Unable to mount client"
3609 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3611 multiop_bg_pause $MOUNT _f
3615 if [ $RC1 -ne 0 ]; then
3616 log "lazystatfs multiop failed $RC1"
3619 sleep $(( $TIMEOUT+1 ))
3621 [ $? -ne 0 ] && error "process isn't sleep"
3622 start_ost2 || error "Unable to start OST2"
3623 wait $pid || error "statfs failed"
3624 stop_ost2 || error "Unable to stop OST2"
3627 umount_client $MOUNT -f || error "Unable to unmount client"
3628 stop_ost || error "Unable to stop OST1"
3629 stop_mds || error "Unable to stop MDS"
3630 #writeconf to remove all ost2 traces for subsequent tests
3631 writeconf_or_reformat
3633 run_test 50f "normal statfs one server in down"
3636 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3638 start_ost2 || error "Unable to start OST2"
3639 wait_osc_import_state mds ost2 FULL
3640 wait_osc_import_state client ost2 FULL
3642 local PARAM="${FSNAME}-OST0001.osc.active"
3644 $SETSTRIPE -c -1 $DIR/$tfile || error "$SETSTRIPE failed"
3645 do_facet mgs $LCTL conf_param $PARAM=0 ||
3646 error "Unable to deactivate OST"
3648 umount_client $MOUNT || error "Unable to unmount client"
3649 mount_client $MOUNT || error "Unable to mount client"
3650 # This df should not cause a panic
3653 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
3654 rm -f $DIR/$tfile || error "unable to remove file $DIR/$tfile"
3655 umount_client $MOUNT || error "Unable to unmount client"
3656 stop_ost2 || error "Unable to stop OST2"
3657 stop_ost || error "Unable to stop OST1"
3658 stop_mds || error "Unable to stop MDS"
3659 #writeconf to remove all ost2 traces for subsequent tests
3660 writeconf_or_reformat
3662 run_test 50g "deactivated OST should not cause panic"
3666 # prepare MDT/OST, make OSC inactive for OST1
3667 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3669 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3670 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
3671 error "tunefs OST1 failed"
3672 start_mds || error "Unable to start MDT"
3673 start_ost || error "Unable to start OST1"
3674 start_ost2 || error "Unable to start OST2"
3675 mount_client $MOUNT || error "client start failed"
3677 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3679 # activatate OSC for OST1
3680 local TEST="$LCTL get_param -n osc.${FSNAME}-OST0000-osc-[!M]*.active"
3681 set_conf_param_and_check client \
3682 "$TEST" "${FSNAME}-OST0000.osc.active" 1 ||
3683 error "Unable to activate OST1"
3685 mkdir $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3686 $SETSTRIPE -c -1 -i 0 $DIR/$tdir/2 ||
3687 error "$SETSTRIPE $DIR/$tdir/2 failed"
3688 sleep 1 && echo "create a file after OST1 is activated"
3690 createmany -o $DIR/$tdir/2/$tfile-%d 1
3692 # check OSC import is working
3693 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
3694 error "some OSC imports are still not connected"
3697 umount_client $MOUNT || error "Unable to umount client"
3698 stop_ost2 || error "Unable to stop OST2"
3699 cleanup_nocli || error "cleanup_nocli failed with $?"
3701 run_test 50h "LU-642: activate deactivated OST"
3704 # prepare MDT/OST, make OSC inactive for OST1
3705 [ "$MDSCOUNT" -lt "2" ] && skip_env "$MDSCOUNT < 2, skipping" && return
3707 [ $(facet_fstype mds2) == zfs ] && import_zpool mds2
3708 do_facet mds2 "$TUNEFS --param mdc.active=0 $(mdsdevname 2)" ||
3709 error "tunefs MDT2 failed"
3710 start_mds || error "Unable to start MDT"
3711 start_ost || error "Unable to start OST1"
3712 start_ost2 || error "Unable to start OST2"
3713 mount_client $MOUNT || error "client start failed"
3715 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3717 $LCTL conf_param ${FSNAME}-MDT0000.mdc.active=0 &&
3718 error "deactive MDC0 succeeds"
3719 # activate MDC for MDT2
3720 local TEST="$LCTL get_param -n mdc.${FSNAME}-MDT0001-mdc-[!M]*.active"
3721 set_conf_param_and_check client \
3722 "$TEST" "${FSNAME}-MDT0001.mdc.active" 1 ||
3723 error "Unable to activate MDT2"
3725 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
3726 if [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.60) ]
3728 wait_dne_interconnect
3730 $LFS mkdir -i1 $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3732 createmany -o $DIR/$tdir/2/$tfile-%d 1 || error "create files failed"
3734 rm -rf $DIR/$tdir/2 || error "unlink dir failed"
3736 # deactivate MDC for MDT2
3737 local TEST="$LCTL get_param -n mdc.${FSNAME}-MDT0001-mdc-[!M]*.active"
3738 set_conf_param_and_check client \
3739 "$TEST" "${FSNAME}-MDT0001.mdc.active" 0 ||
3740 error "Unable to deactivate MDT2"
3742 $LFS mkdir -i1 $DIR/$tdir/2 &&
3743 error "mkdir $DIR/$tdir/2 succeeds after deactive MDT"
3746 umount_client $MOUNT || error "Unable to umount client"
3751 run_test 50i "activate deactivated MDT"
3754 local LOCAL_TIMEOUT=20
3758 check_mount || error "check_mount failed"
3760 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
3761 $SETSTRIPE -c -1 $MOUNT/$tdir ||
3762 error "$SETSTRIPE -c -1 $MOUNT/$tdir failed"
3763 #define OBD_FAIL_MDS_REINT_DELAY 0x142
3764 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x142"
3765 touch $MOUNT/$tdir/$tfile &
3768 start_ost2 || error "Unable to start OST1"
3770 stop_ost2 || error "Unable to stop OST1"
3771 umount_client $MOUNT -f || error “unmount $MOUNT failed”
3772 cleanup_nocli || error “stop server failed”
3773 #writeconf to remove all ost2 traces for subsequent tests
3774 writeconf_or_reformat
3776 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
3785 do_node $node mkdir -p $dest
3786 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3788 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
3789 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3790 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
3792 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
3793 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
3803 local backup2=${TMP}/backup2
3805 do_node $node mkdir -p $backup2
3806 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3808 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
3809 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3810 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
3812 do_node $node "diff -rq $backup $backup2"
3813 [ $? -eq 0 ] || { error "contents differ"; return 3; }
3815 local xattrs2=${TMP}/xattrs2
3816 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
3817 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
3819 do_node $node "diff $xattrs $xattrs2"
3820 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
3822 do_node $node "rm -rf $backup2 $xattrs2"
3823 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
3827 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3828 skip "Only applicable to ldiskfs-based MDTs"
3832 start_mds || error "Unable to start MDS"
3833 start_ost || error "Unable to start OST1"
3834 mount_client $MOUNT || error "Unable to mount client"
3837 local ost1mnt=$(facet_mntpt ost1)
3838 local ost1node=$(facet_active_host ost1)
3839 local ost1tmp=$TMP/conf52
3842 mkdir $DIR/$tdir || error "Unable to create $DIR/$tdir"
3843 touch $TMP/modified_first || error "Unable to create temporary file"
3844 local mtime=$(stat -c %Y $TMP/modified_first)
3845 do_node $ost1node "mkdir -p $ost1tmp &&
3846 touch -m -d @$mtime $ost1tmp/modified_first" ||
3847 error "Unable to create temporary file"
3850 $SETSTRIPE -c -1 -S 1M $DIR/$tdir || error "$SETSTRIPE failed"
3852 for (( i=0; i < nrfiles; i++ )); do
3853 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c ||
3854 error "multiop failed"
3860 echo backup files to $TMP/$tdir
3861 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
3862 copy_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
3863 error "Unable to copy files"
3865 umount_client $MOUNT || error "Unable to umount client"
3866 stop_ost || error "Unable to stop ost1"
3868 echo mount ost1 as ldiskfs
3869 do_node $ost1node mkdir -p $ost1mnt || error "Unable to create $ost1mnt"
3870 if ! do_node $ost1node test -b $ost1_dev; then
3873 do_node $ost1node mount -t $(facet_fstype ost1) $loop $ost1_dev \
3875 error "Unable to mount ost1 as ldiskfs"
3878 echo backup objects to $ost1tmp/objects
3879 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
3880 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
3881 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
3883 error "Unable to copy objects"
3885 # move objects to lost+found
3886 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
3887 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
3889 do_node $ost1node "umount $ost1mnt" ||
3890 error "Unable to umount ost1 as ldiskfs"
3892 start_ost || error "Unable to start OST1"
3893 mount_client $MOUNT || error "Unable to mount client"
3895 local REPAIRED=$(do_node $ost1node "$LCTL get_param \
3896 -n osd-ldiskfs.$FSNAME-OST0000.oi_scrub" |
3897 awk '/^lf_repa[ri]*ed/ { print $2 }')
3898 [ $REPAIRED -gt 0 ] ||
3899 error "Some entry under /lost+found should be repaired"
3902 diff_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
3903 error "Unable to diff files"
3905 rm -rf $TMP/$tdir $TMP/file_xattrs ||
3906 error "Unable to delete temporary files"
3907 do_node $ost1node "rm -rf $ost1tmp" ||
3908 error "Unable to delete temporary files"
3909 cleanup || error "cleanup failed with $?"
3911 run_test 52 "check recovering objects from lost+found"
3913 # Checks threads_min/max/started for some service
3915 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
3916 # parameter pattern prefix like 'ost.*.ost'.
3929 local msg="Insane $modname thread counts"
3930 local ncpts=$(check_cpt_number $facet)
3934 check_mount || return 41
3936 # We need to expand $parampat, but it may match multiple parameters, so
3937 # we'll pick the first one
3938 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
3939 error "Couldn't expand ${parampat}.threads_min parameter name"
3943 # Remove the .threads_min part
3944 paramp=${paramp%.threads_min}
3946 # Check for sanity in defaults
3947 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3949 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3951 tstarted=$(do_facet $facet "$LCTL get_param \
3952 -n ${paramp}.threads_started" || echo 0)
3953 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' ||
3955 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
3957 nthrs=$(expr $tmax - $tmin)
3958 if [ $nthrs -lt $ncpts ]; then
3964 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
3965 skip_env "module parameter forced $facet thread count" &&
3966 tmin=3 && tmax=$((3 * tmax))
3968 # Check that we can change min/max
3969 do_facet $facet "$LCTL set_param \
3970 ${paramp}.threads_min=$((tmin + nthrs))"
3971 do_facet $facet "$LCTL set_param \
3972 ${paramp}.threads_max=$((tmax - nthrs))"
3973 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3975 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3977 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) &&
3978 $tmax2 == ($tmax - $nthrs)))' || return $?
3980 # Check that we can set min/max to the same value
3981 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3983 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$tmin"
3984 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3986 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3988 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
3990 # Check that we can't set max < min
3991 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$((tmin - 1))"
3992 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3994 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3996 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
3998 # We need to ensure that we get the module options desired; to do this
3999 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
4000 LOAD_MODULES_REMOTE=true
4003 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
4004 setmodopts -a $modname "$newvalue" oldvalue
4007 check_mount || return 41
4009 # Restore previous setting of MODOPTS_*
4010 setmodopts $modname "$oldvalue"
4012 # Check that $opts took
4013 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min")
4014 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max")
4015 tstarted=$(do_facet $facet \
4016 "$LCTL get_param -n ${paramp}.threads_started")
4017 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4026 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
4027 cleanup || error "cleanup failed with rc $?"
4029 run_test 53a "check OSS thread count params"
4033 local mds=$(do_facet $SINGLEMDS "$LCTL get_param \
4034 -N mds.*.*.threads_max 2>/dev/null")
4035 if [ -z "$mds" ]; then
4036 #running this on an old MDT
4037 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
4039 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
4041 cleanup || error "cleanup failed with $?"
4043 run_test 53b "check MDS thread count params"
4046 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4047 skip "Only applicable to ldiskfs-based MDTs"
4051 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p ||
4052 error "llverdev failed with rc=$?"
4055 run_test 54a "test llverdev and partial verify of device"
4058 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4059 skip "Only applicable to ldiskfs-based MDTs"
4064 run_llverfs $MOUNT -p || error "llverfs failed with rc=$?"
4065 cleanup || error "cleanup failed with rc=$?"
4067 run_test 54b "test llverfs and partial verify of filesystem"
4071 local max_ost_index=$1
4072 echo -n $(((max_ost_index + 1) * 8))
4076 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4077 skip "Only applicable to ldiskfs-based MDTs"
4081 local mdsdev=$(mdsdevname 1)
4082 local mdsvdev=$(mdsvdevname 1)
4086 if ! combined_mgs_mds; then
4087 stop_mgs || error "stopping MGS service failed"
4088 format_mgs || error "formatting MGT failed"
4090 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
4092 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
4093 --reformat $(ostdevname 1) $(ostvdevname 1)
4099 echo checking size of lov_objid for ost index $i
4100 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
4101 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
4102 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
4104 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
4111 run_test 55 "check lov_objid size"
4114 local server_version=$(lustre_version_code $SINGLEMDS)
4115 local mds_journal_size_orig=$MDSJOURNALSIZE
4120 for num in $(seq 1 $MDSCOUNT); do
4123 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
4124 $(ostdevname 1) $(ostvdevname 1)
4125 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
4126 $(ostdevname 2) $(ostvdevname 2)
4129 start_ost || error "Unable to start first ost (idx 10000)"
4130 start_ost2 || error "Unable to start second ost (idx 1000)"
4131 mount_client $MOUNT || error "Unable to mount client"
4135 if [[ $server_version -ge $(version_code 2.6.54) ]] ||
4136 [[ $server_version -ge $(version_code 2.5.4) &&
4137 $server_version -lt $(version_code 2.5.11) ]]; then
4138 wait_osc_import_state mds ost1 FULL
4139 wait_osc_import_state mds ost2 FULL
4140 $SETSTRIPE --stripe-count=-1 $DIR/$tfile ||
4141 error "Unable to setstripe $DIR/$tfile"
4142 n=$($LFS getstripe --stripe-count $DIR/$tfile)
4143 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
4148 MDSJOURNALSIZE=$mds_journal_size_orig
4151 run_test 56a "check big OST indexes and out-of-index-order start"
4156 umount_client $MOUNT -f || error "unmount client failed"
4165 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs" && return
4167 trap cleanup_56b EXIT RETURN ERR
4170 if ! combined_mgs_mds ; then
4175 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) --index=0 --reformat \
4176 $(mdsdevname 1) $(mdsvdevname 1)
4177 add mds2 $(mkfs_opts mds2 $(mdsdevname 2)) --index=1 --reformat \
4178 $(mdsdevname 2) $(mdsvdevname 2)
4179 add mds3 $(mkfs_opts mds3 $(mdsdevname 3)) --index=1000 --reformat \
4180 $(mdsdevname 3) $(mdsvdevname 3)
4184 start_mdt 1 || error "MDT 1 (idx 0) start failed"
4185 start_mdt 2 || error "MDT 2 (idx 1) start failed"
4186 start_mdt 3 || error "MDT 3 (idx 1000) start failed"
4187 start_ost || error "Unable to start first ost"
4188 start_ost2 || error "Unable to start second ost"
4190 do_nodes $(comma_list $(mdts_nodes)) \
4191 "$LCTL set_param mdt.*.enable_remote_dir=1 \
4192 mdt.*.enable_remote_dir_gid=-1"
4194 mount_client $MOUNT || error "Unable to mount client"
4196 $LFS mkdir -c3 $MOUNT/$tdir || error "failed to make testdir"
4198 echo "This is test file 1!" > $MOUNT/$tdir/$tfile.1 ||
4199 error "failed to make test file 1"
4200 echo "This is test file 2!" > $MOUNT/$tdir/$tfile.2 ||
4201 error "failed to make test file 2"
4202 echo "This is test file 1000!" > $MOUNT/$tdir/$tfile.1000 ||
4203 error "failed to make test file 1000"
4205 rm -rf $MOUNT/$tdir || error "failed to remove testdir"
4207 $LFS mkdir -i1000 $MOUNT/$tdir.1000 ||
4208 error "create remote dir at idx 1000 failed"
4211 echo "=== START lfs df OUTPUT ==="
4213 echo "==== END lfs df OUTPUT ===="
4215 mdtcnt=$(echo -e "$output" | grep $FSNAME-MDT | wc -l)
4216 ostcnt=$(echo -e "$output" | grep $FSNAME-OST | wc -l)
4218 echo "lfs df returned mdt count $mdtcnt and ost count $ostcnt"
4219 [ $mdtcnt -eq 3 ] || error "lfs df returned wrong mdt count"
4220 [ $ostcnt -eq 2 ] || error "lfs df returned wrong ost count"
4222 echo "This is test file 1!" > $MOUNT/$tdir.1000/$tfile.1 ||
4223 error "failed to make test file 1"
4224 echo "This is test file 2!" > $MOUNT/$tdir.1000/$tfile.2 ||
4225 error "failed to make test file 2"
4226 echo "This is test file 1000!" > $MOUNT/$tdir.1000/$tfile.1000 ||
4227 error "failed to make test file 1000"
4228 rm -rf $MOUNT/$tdir.1000 || error "failed to remove remote_dir"
4231 echo "=== START lfs mdts OUTPUT ==="
4233 echo "==== END lfs mdts OUTPUT ===="
4235 echo -e "$output" | grep -v "MDTS:" | awk '{print $1}' |
4236 sed 's/://g' > $TMP/mdts-actual.txt
4237 sort $TMP/mdts-actual.txt -o $TMP/mdts-actual.txt
4239 echo -e "0\n1\n1000" > $TMP/mdts-expected.txt
4241 diff $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4244 rm $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4246 [ $result -eq 0 ] || error "target_obd proc file is incorrect!"
4248 run_test 56b "test target_obd correctness with nonconsecutive MDTs"
4250 test_57a() { # bug 22656
4251 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4252 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4253 tail -1 | awk '{print $1}')
4254 writeconf_or_reformat
4255 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4256 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" ||
4257 error "tunefs failed"
4259 start_ost && error "OST registration from failnode should fail"
4262 run_test 57a "initial registration from failnode should fail (should return errs)"
4265 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4266 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4267 tail -1 | awk '{print $1}')
4268 writeconf_or_reformat
4269 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4270 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" ||
4271 error "tunefs failed"
4273 start_ost || error "OST registration from servicenode should not fail"
4276 run_test 57b "initial registration from servicenode should not fail"
4279 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
4282 test_58() { # bug 22658
4283 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
4285 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4286 createmany -o $DIR/$tdir/$tfile-%d 100
4287 # make sure that OSTs do not cancel llog cookies before we unmount the MDS
4288 #define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601
4289 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x601"
4290 unlinkmany $DIR/$tdir/$tfile-%d 100
4291 stop_mds || error "Unable to stop MDS"
4293 local MNTDIR=$(facet_mntpt $SINGLEMDS)
4294 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4296 # remove all files from the OBJECTS dir
4297 mount_fstype $SINGLEMDS
4299 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
4301 unmount_fstype $SINGLEMDS
4302 # restart MDS with missing llog files
4303 start_mds || error "unable to start MDS"
4304 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0"
4307 run_test 58 "missing llog files must not prevent MDT from mounting"
4310 start_mgsmds >> /dev/null
4311 local C1=$(count_osts)
4312 if [ $C1 -eq 0 ]; then
4313 start_ost >> /dev/null
4317 echo "original ost count: $C1 (expect > 0)"
4318 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
4319 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
4320 local C2=$(count_osts)
4321 echo "after mdt writeconf count: $C2 (expect 0)"
4322 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
4323 echo "OST start without writeconf should fail:"
4324 start_ost >> /dev/null &&
4325 error "OST start without writeconf didn't fail"
4326 echo "OST start with writeconf should succeed:"
4327 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
4328 local C3=$(count_osts)
4329 echo "after ost writeconf count: $C3 (expect 1)"
4330 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
4331 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
4332 local C4=$(count_osts)
4333 echo "after ost2 writeconf count: $C4 (expect 2)"
4334 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
4335 stop_ost2 >> /dev/null
4336 cleanup_nocli >> /dev/null
4337 #writeconf to remove all ost2 traces for subsequent tests
4338 writeconf_or_reformat
4340 run_test 59 "writeconf mount option"
4342 test_60() { # LU-471
4345 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4346 skip "Only applicable to ldiskfs-based MDTs"
4350 for num in $(seq $MDSCOUNT); do
4351 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
4352 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
4353 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4357 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
4358 [ ${PIPESTATUS[0]} -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
4360 # MDT default has dirdata feature
4361 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
4362 # we disable uninit_bg feature
4363 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
4364 # we set stride extended options
4365 echo $dump | grep stride > /dev/null || error "stride is not set"
4368 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
4373 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] ||
4374 { skip "Need MDS version at least 2.1.53"; return 0; }
4376 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
4377 ! large_xattr_enabled; then
4380 for num in $(seq $MDSCOUNT); do
4381 do_facet mds${num} $TUNE2FS -O large_xattr \
4382 $(mdsdevname $num) ||
4383 error "tune2fs on mds $num failed"
4387 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
4388 setup_noconfig || error "setting up the filesystem failed"
4389 client_up || error "starting client failed"
4391 local file=$DIR/$tfile
4392 touch $file || error "touch $file failed"
4394 local large_value="$(generate_string $(max_xattr_size))"
4395 local small_value="bar"
4397 local name="trusted.big"
4398 log "save large xattr $name on $file"
4399 setfattr -n $name -v $large_value $file ||
4400 error "saving $name on $file failed"
4402 local new_value=$(get_xattr_value $name $file)
4403 [[ "$new_value" != "$large_value" ]] &&
4404 error "$name different after saving"
4406 log "shrink value of $name on $file"
4407 setfattr -n $name -v $small_value $file ||
4408 error "shrinking value of $name on $file failed"
4410 new_value=$(get_xattr_value $name $file)
4411 [[ "$new_value" != "$small_value" ]] &&
4412 error "$name different after shrinking"
4414 log "grow value of $name on $file"
4415 setfattr -n $name -v $large_value $file ||
4416 error "growing value of $name on $file failed"
4418 new_value=$(get_xattr_value $name $file)
4419 [[ "$new_value" != "$large_value" ]] &&
4420 error "$name different after growing"
4422 log "check value of $name on $file after remounting MDS"
4424 new_value=$(get_xattr_value $name $file)
4425 [[ "$new_value" != "$large_value" ]] &&
4426 error "$name different after remounting MDS"
4428 log "remove large xattr $name from $file"
4429 setfattr -x $name $file || error "removing $name from $file failed"
4432 stopall || error "stopping for e2fsck run"
4433 for num in $(seq $MDSCOUNT); do
4434 run_e2fsck $(facet_active_host mds$num) \
4435 $(mdsdevname $num) "-y" ||
4436 error "e2fsck MDT$num failed"
4438 setup_noconfig || error "remounting the filesystem failed"
4441 # need to delete this file to avoid problems in other tests
4443 stopall || error "stopping systems to turn off large_xattr"
4445 for num in $(seq $MDSCOUNT); do
4446 do_facet mds${num} $TUNE2FS -O ^large_xattr \
4447 $(mdsdevname $num) ||
4448 error "tune2fs on mds $num failed"
4452 run_test 61 "large xattr"
4455 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4456 skip "Only applicable to ldiskfs-based MDTs"
4461 local mdsdev=$(mdsdevname 1)
4462 local ostdev=$(ostdevname 1)
4464 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] ||
4465 { skip "Need MDS version at least 2.2.51"; return 0; }
4467 echo "disable journal for mds"
4468 do_facet mds1 $TUNE2FS -O ^has_journal $mdsdev || error "tune2fs failed"
4469 start_mds && error "MDT start should fail"
4470 echo "disable journal for ost"
4471 do_facet ost1 $TUNE2FS -O ^has_journal $ostdev || error "tune2fs failed"
4472 start_ost && error "OST start should fail"
4473 cleanup || error "cleanup failed with rc $?"
4476 run_test 62 "start with disabled journal"
4479 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4480 skip "Only applicable to ldiskfs-based MDTs"
4484 local inode_slab=$(do_facet $SINGLEMDS \
4485 "awk '/ldiskfs_inode_cache/ { print \\\$5 }' /proc/slabinfo")
4486 if [ -z "$inode_slab" ]; then
4487 skip "ldiskfs module has not been loaded"
4491 echo "$inode_slab ldisk inodes per page"
4492 if [ "$inode_slab" -ge "3" ] ; then
4493 # If kmalloc-128 is also 1 per page - this is a debug kernel
4494 # and so this is not an error.
4495 local kmalloc128=$(do_facet $SINGLEMDS \
4496 "awk '/^(kmalloc|size)-128 / { print \\\$5 }' /proc/slabinfo")
4497 # 32 128-byte chunks in 4k
4498 [ "$kmalloc128" -eq "32" ] ||
4499 error "ldisk inode size is too big, $inode_slab objs per page"
4504 run_test 63 "Verify each page can at least hold 3 ldisk inodes"
4507 start_mds || error "unable to start MDS"
4508 start_ost || error "Unable to start OST1"
4509 start_ost2 || error "Unable to start second ost"
4510 mount_client $MOUNT || error "Unable to mount client"
4511 stop_ost2 || error "Unable to stop second ost"
4514 umount_client $MOUNT -f || error “unmount $MOUNT failed”
4515 cleanup_nocli || error "cleanup_nocli failed with $?"
4516 #writeconf to remove all ost2 traces for subsequent tests
4517 writeconf_or_reformat
4519 run_test 64 "check lfs df --lazy "
4521 test_65() { # LU-2237
4522 # Currently, the test is only valid for ldiskfs backend
4523 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4524 skip "non-ldiskfs backend" && return
4526 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4527 local brpt=$(facet_mntpt brpt)
4530 if ! do_facet $SINGLEMDS "test -b $devname"; then
4534 stop_mds || error "Unable to stop MDS"
4535 local obj=$(do_facet $SINGLEMDS \
4536 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
4538 if [ -z "$obj" ]; then
4539 # The MDT may be just re-formatted, mount the MDT for the
4540 # first time to guarantee the "last_rcvd" file is there.
4541 start_mds || error "fail to mount the MDS for the first time"
4542 stop_mds || error "Unable to stop MDS"
4545 # remove the "last_rcvd" file
4546 do_facet $SINGLEMDS "mkdir -p $brpt"
4547 do_facet $SINGLEMDS \
4548 "mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
4549 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
4550 do_facet $SINGLEMDS "$UMOUNT $brpt"
4552 # restart MDS, the "last_rcvd" file should be recreated.
4553 start_mds || error "fail to restart the MDS"
4554 stop_mds || error "Unable to stop MDS"
4555 obj=$(do_facet $SINGLEMDS \
4556 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
4557 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
4559 run_test 65 "re-create the lost last_rcvd file when server mount"
4562 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
4563 { skip "Need MGS version at least 2.3.59"; return 0; }
4566 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
4567 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
4569 echo "replace_nids should fail if MDS, OSTs and clients are UP"
4570 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4571 error "replace_nids fail"
4573 umount_client $MOUNT || error "unmounting client failed"
4574 echo "replace_nids should fail if MDS and OSTs are UP"
4575 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4576 error "replace_nids fail"
4578 stop_ost || error "Unable to stop OST1"
4579 echo "replace_nids should fail if MDS is UP"
4580 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4581 error "replace_nids fail"
4583 stop_mds || error "stopping mds failed"
4585 if combined_mgs_mds; then
4586 start_mdt 1 "-o nosvc" ||
4587 error "starting mds with nosvc option failed"
4590 echo "command should accept two parameters"
4591 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
4592 error "command should accept two params"
4594 echo "correct device name should be passed"
4595 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
4596 error "wrong devname"
4598 echo "wrong nids list should not destroy the system"
4599 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
4602 echo "replace OST nid"
4603 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
4604 error "replace nids failed"
4606 echo "command should accept two parameters"
4607 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
4608 error "command should accept two params"
4610 echo "wrong nids list should not destroy the system"
4611 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
4614 echo "replace MDS nid"
4615 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
4616 error "replace nids failed"
4618 if ! combined_mgs_mds ; then
4621 stop_mds || error "Unable to stop MDS"
4625 check_mount || error "error after nid replace"
4626 cleanup || error "cleanup failed"
4629 run_test 66 "replace nids"
4631 test_67() { #LU-2950
4632 local legacy="$TMP/legacy_lnet_config"
4633 local new="$TMP/new_routes_test"
4634 local out="$TMP/config_out_file"
4635 local verify="$TMP/conv_verify"
4636 local verify_conf="$TMP/conf_verify"
4638 # Create the legacy file that will be run through the
4639 # lustre_routes_conversion script
4640 cat <<- LEGACY_LNET_CONFIG > $legacy
4641 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
4642 tcp2 54 10.1.3.2@tcp;
4643 tcp3 10.3.4.3@tcp:3;
4647 # Create the verification file to verify the output of
4648 # lustre_routes_conversion script against.
4649 cat <<- VERIFY_LNET_CONFIG > $verify
4650 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
4651 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
4652 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
4653 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
4654 tcp4: { gateway: 10.3.3.4@tcp }
4657 # Create the verification file to verify the output of
4658 # lustre_routes_config script against
4659 cat <<- VERIFY_LNET_CONFIG > $verify_conf
4660 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
4661 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
4662 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
4663 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
4664 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
4667 lustre_routes_conversion $legacy $new > /dev/null
4668 if [ -f $new ]; then
4669 # verify the conversion output
4670 cmp -s $new $verify > /dev/null
4671 if [ $? -eq 1 ]; then
4672 error "routes conversion failed"
4675 lustre_routes_config --dry-run --verbose $new > $out
4676 # check that the script succeeded
4677 cmp -s $out $verify_conf > /dev/null
4678 if [ $? -eq 1 ]; then
4679 error "routes config failed"
4682 error "routes conversion test failed"
4684 # remove generated files
4685 rm -f $new $legacy $verify $verify_conf $out
4687 run_test 67 "test routes conversion and configuration"
4695 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.4.53) ] ||
4696 { skip "Need MDS version at least 2.4.53"; return 0; }
4698 umount_client $MOUNT || error "umount client failed"
4700 start_mdt 1 || error "MDT start failed"
4701 start_ost || error "Unable to start OST1"
4703 # START-END - the sequences we'll be reserving
4704 START=$(do_facet $SINGLEMDS \
4705 $LCTL get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
4706 END=$((START + (1 << 30)))
4707 do_facet $SINGLEMDS \
4708 $LCTL set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
4710 # reset the sequences MDT0000 has already assigned
4711 do_facet $SINGLEMDS \
4712 $LCTL set_param seq.srv*MDT0000.space=clear
4714 # remount to let the client allocate new sequence
4715 mount_client $MOUNT || error "mount client failed"
4717 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4718 do_facet $SINGLEMDS \
4719 $LCTL get_param seq.srv*MDT0000.space
4720 $LFS path2fid $DIR/$tfile
4722 local old_ifs="$IFS"
4724 fid=($($LFS path2fid $DIR/$tfile))
4728 if [[ $seq < $END ]]; then
4729 error "used reserved sequence $seq?"
4731 cleanup || error "cleanup failed with $?"
4733 run_test 68 "be able to reserve specific sequences in FLDB"
4735 # Test 69: is about the total number of objects ever created on an OST.
4736 # so that when it is reformatted the normal MDS->OST orphan recovery won't
4737 # just "precreate" the missing objects. In the past it might try to recreate
4738 # millions of objects after an OST was reformatted
4740 local server_version=$(lustre_version_code $SINGLEMDS)
4742 [[ $server_version -lt $(version_code 2.4.2) ]] &&
4743 skip "Need MDS version at least 2.4.2" && return
4745 [[ $server_version -ge $(version_code 2.4.50) ]] &&
4746 [[ $server_version -lt $(version_code 2.5.0) ]] &&
4747 skip "Need MDS version at least 2.5.0" && return
4750 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4752 # use OST0000 since it probably has the most creations
4753 local OSTNAME=$(ostname_from_index 0)
4754 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
4755 local last_id=$(do_facet mds1 $LCTL get_param -n \
4756 osc.$mdtosc_proc1.prealloc_last_id)
4758 # Want to have OST LAST_ID over 5 * OST_MAX_PRECREATE to
4759 # verify that the LAST_ID recovery is working properly. If
4760 # not, then the OST will refuse to allow the MDS connect
4761 # because the LAST_ID value is too different from the MDS
4762 #define OST_MAX_PRECREATE=20000
4763 local ost_max_pre=20000
4764 local num_create=$(( ost_max_pre * 5 + 1 - last_id))
4766 # If the LAST_ID is already over 5 * OST_MAX_PRECREATE, we don't
4767 # need to create any files. So, skip this section.
4768 if [ $num_create -gt 0 ]; then
4769 # Check the number of inodes available on OST0
4771 local ifree=$($LFS df -i $MOUNT | awk '/OST0000/ { print $4 }')
4772 log "On OST0, $ifree inodes available. Want $num_create."
4774 $SETSTRIPE -i 0 $DIR/$tdir ||
4775 error "$SETSTRIPE -i 0 $DIR/$tdir failed"
4776 if [ $ifree -lt 10000 ]; then
4777 files=$(( ifree - 50 ))
4782 local j=$((num_create / files + 1))
4783 for i in $(seq 1 $j); do
4784 createmany -o $DIR/$tdir/$tfile-$i- $files ||
4785 error "createmany fail create $files files: $?"
4786 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
4787 error "unlinkmany failed unlink $files files"
4791 # delete all of the files with objects on OST0 so the
4792 # filesystem is not inconsistent later on
4793 $LFS find $MOUNT --ost 0 -print0 | xargs -0 rm
4795 umount_client $MOUNT || error "umount client failed"
4796 stop_ost || error "OST0 stop failure"
4797 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
4798 $(ostdevname 1) $(ostvdevname 1) ||
4799 error "reformat and replace $ostdev failed"
4800 start_ost || error "OST0 restart failure"
4801 wait_osc_import_state mds ost FULL
4803 mount_client $MOUNT || error "mount client failed"
4804 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
4805 local idx=$($GETSTRIPE -i $DIR/$tdir/$tfile-last)
4806 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
4808 local iused=$($LFS df -i $MOUNT | awk '/OST0000/ { print $3 }')
4809 log "On OST0, $iused used inodes"
4810 [ $iused -ge $((ost_max_pre/2 + 1000)) ] &&
4811 error "OST replacement created too many inodes; $iused"
4812 cleanup || error "cleanup failed with $?"
4814 run_test 69 "replace an OST with the same index"
4817 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4820 cleanup || error "cleanup failed with $?"
4822 start_mdt 1 || error "MDT0 start fail"
4824 start_ost || error "OST0 start fail"
4825 for num in $(seq 2 $MDSCOUNT); do
4826 start_mdt $num || return
4829 mount_client $MOUNT || error "mount client fails"
4831 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
4833 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4834 error "create remote dir fail"
4836 rm -rf $DIR/$tdir || error "delete dir fail"
4837 cleanup || error "cleanup failed with $?"
4839 run_test 70a "start MDT0, then OST, then MDT1"
4842 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4845 start_ost || error "OST0 start fail"
4847 start_mds || error "MDS start fail"
4849 mount_client $MOUNT || error "mount client fails"
4851 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
4853 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4854 error "create remote dir fail"
4856 rm -rf $DIR/$tdir || error "delete dir fail"
4858 cleanup || error "cleanup failed with $?"
4860 run_test 70b "start OST, MDT1, MDT0"
4863 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4866 start_mds || error "MDS start fail"
4867 start_ost || error "OST0 start fail"
4869 mount_client $MOUNT || error "mount client fails"
4870 stop_mdt 1 || error "MDT1 start fail"
4872 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
4873 echo "deactivate $mdc_for_mdt1"
4874 $LCTL --device $mdc_for_mdt1 deactivate ||
4875 error "set $mdc_for_mdt1 deactivate failed"
4877 mkdir $DIR/$tdir && error "mkdir succeed"
4879 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4880 error "create remote dir succeed"
4882 cleanup || error "cleanup failed with $?"
4884 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
4887 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4890 start_mds || error "MDS start fail"
4891 start_ost || error "OST0 start fail"
4893 mount_client $MOUNT || error "mount client fails"
4895 stop_mdt 2 || error "MDT1 start fail"
4897 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
4899 echo "deactivate $mdc_for_mdt2"
4900 $LCTL --device $mdc_for_mdt2 deactivate ||
4901 error "set $mdc_for_mdt2 deactivate failed"
4903 mkdir $DIR/$tdir || error "mkdir fail"
4904 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4905 error "create remote dir succeed"
4907 rm -rf $DIR/$tdir || error "delete dir fail"
4909 cleanup || error "cleanup failed with $?"
4911 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
4914 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4916 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.62) ] ||
4917 { skip "Need MDS version at least 2.7.62"; return 0; }
4919 cleanup || error "cleanup failed with $?"
4921 local mdsdev=$(mdsdevname 1)
4922 local ostdev=$(ostdevname 1)
4923 local mdsvdev=$(mdsvdevname 1)
4924 local ostvdev=$(ostvdevname 1)
4925 local opts_mds="$(mkfs_opts mds1 $mdsdev) --reformat $mdsdev $mdsvdev"
4926 local opts_ost="$(mkfs_opts ost1 $ostdev) --reformat $ostdev $ostvdev"
4928 add mds1 $opts_mds || error "add mds1 failed"
4929 start_mdt 1 || error "start mdt1 failed"
4930 add ost1 $opts_ost || error "add ost1 failed"
4931 start_ost || error "start ost failed"
4932 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
4934 local soc=$(do_facet mds1 "$LCTL get_param -n \
4935 mdt.*MDT0000.sync_lock_cancel")
4936 [ $soc == "never" ] || error "SoC enabled on single MDS"
4938 for i in $(seq 2 $MDSCOUNT); do
4939 mdsdev=$(mdsdevname $i)
4940 mdsvdev=$(mdsvdevname $i)
4941 opts_mds="$(mkfs_opts mds$i $mdsdev) --reformat $mdsdev \
4943 add mds$i $opts_mds || error "add mds$i failed"
4944 start_mdt $i || error "start mdt$i fail"
4947 wait_dne_interconnect
4949 for i in $(seq $MDSCOUNT); do
4950 soc=$(do_facet mds$i "$LCTL get_param -n \
4951 mdt.*MDT000$((i - 1)).sync_lock_cancel")
4952 [ $soc == "blocking" ] || error "SoC not enabled on DNE"
4955 for i in $(seq 2 $MDSCOUNT); do
4956 stop_mdt $i || error "stop mdt$i fail"
4958 soc=$(do_facet mds1 "$LCTL get_param -n \
4959 mdt.*MDT0000.sync_lock_cancel")
4960 [ $soc == "never" ] || error "SoC enabled on single MDS"
4962 cleanup || error "cleanup failed with $?"
4964 run_test 70e "Sync-on-Cancel will be enabled by default on DNE"
4967 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4968 if combined_mgs_mds; then
4969 skip "needs separate MGS/MDT" && return
4973 start_mdt 1 || error "MDT0 start fail"
4974 start_ost || error "OST0 start fail"
4975 for num in $(seq 2 $MDSCOUNT); do
4976 start_mdt $num || return
4979 start_ost2 || error "OST1 start fail"
4981 mount_client $MOUNT || error "mount client fails"
4983 mkdir $DIR/$tdir || error "mkdir fail"
4984 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4985 error "create remote dir succeed"
4987 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4988 rm -rf $DIR/$tdir || error "delete dir fail"
4990 umount_client $MOUNT || error "umount_client failed"
4991 stop_mds || error "MDS stop fail"
4992 stop_ost || error "OST0 stop fail"
4993 stop_ost2 || error "OST1 stop fail"
4995 run_test 71a "start MDT0 OST0, MDT1, OST1"
4998 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4999 if combined_mgs_mds; then
5000 skip "needs separate MGS/MDT" && return
5004 for num in $(seq 2 $MDSCOUNT); do
5005 start_mdt $num || return
5007 start_ost || error "OST0 start fail"
5008 start_mdt 1 || error "MDT0 start fail"
5009 start_ost2 || error "OST1 start fail"
5011 mount_client $MOUNT || error "mount client fails"
5013 mkdir $DIR/$tdir || error "mkdir fail"
5014 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5015 error "create remote dir succeed"
5017 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5018 rm -rf $DIR/$tdir || error "delete dir fail"
5020 umount_client $MOUNT || error "umount_client failed"
5021 stop_mds || error "MDT0 stop fail"
5022 stop_ost || error "OST0 stop fail"
5023 stop_ost2 || error "OST1 stop fail"
5025 run_test 71b "start MDT1, OST0, MDT0, OST1"
5028 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5029 if combined_mgs_mds; then
5030 skip "needs separate MGS/MDT" && return
5034 start_ost || error "OST0 start fail"
5035 start_ost2 || error "OST1 start fail"
5036 for num in $(seq 2 $MDSCOUNT); do
5037 start_mdt $num || return
5039 start_mdt 1 || error "MDT0 start fail"
5041 mount_client $MOUNT || error "mount client fails"
5043 mkdir $DIR/$tdir || error "mkdir fail"
5044 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5045 error "create remote dir succeed"
5047 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5048 rm -rf $DIR/$tdir || error "delete dir fail"
5050 umount_client $MOUNT || error "umount_client failed"
5051 stop_mds || error "MDS stop fail"
5052 stop_ost || error "OST0 stop fail"
5053 stop_ost2 || error "OST1 stop fail"
5056 run_test 71c "start OST0, OST1, MDT1, MDT0"
5059 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5060 if combined_mgs_mds; then
5061 skip "needs separate MGS/MDT" && return
5065 start_ost || error "OST0 start fail"
5066 for num in $(seq 2 $MDSCOUNT); do
5067 start_mdt $num || return
5069 start_mdt 1 || error "MDT0 start fail"
5070 start_ost2 || error "OST1 start fail"
5072 mount_client $MOUNT || error "mount client fails"
5074 mkdir $DIR/$tdir || error "mkdir fail"
5075 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5076 error "create remote dir succeed"
5078 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5079 rm -rf $DIR/$tdir || error "delete dir fail"
5081 umount_client $MOUNT || error "umount_client failed"
5082 stop_mds || error "MDS stop fail"
5083 stop_ost || error "OST0 stop fail"
5084 stop_ost2 || error "OST1 stop fail"
5087 run_test 71d "start OST0, MDT1, MDT0, OST1"
5090 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5091 if combined_mgs_mds; then
5092 skip "needs separate MGS/MDT" && return
5096 start_ost || error "OST0 start fail"
5097 for num in $(seq 2 $MDSCOUNT); do
5098 start_mdt $num || return
5100 start_ost2 || error "OST1 start fail"
5101 start_mdt 1 || error "MDT0 start fail"
5103 mount_client $MOUNT || error "mount client fails"
5105 mkdir $DIR/$tdir || error "mkdir fail"
5106 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5107 error "create remote dir succeed"
5109 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5110 rm -rf $DIR/$tdir || error "delete dir fail"
5112 umount_client $MOUNT || error "umount_client failed"
5113 stop_mds || error "MDS stop fail"
5114 stop_ost || error "OST0 stop fail"
5115 stop_ost2 || error "OST1 stop fail"
5118 run_test 71e "start OST0, MDT1, OST1, MDT0"
5120 test_72() { #LU-2634
5121 local mdsdev=$(mdsdevname 1)
5122 local ostdev=$(ostdevname 1)
5123 local cmd="$E2FSCK -fnvd $mdsdev"
5126 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
5127 skip "ldiskfs only test" && return
5129 #tune MDT with "-O extents"
5131 for num in $(seq $MDSCOUNT); do
5132 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
5133 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
5134 error "add mds $num failed"
5135 do_facet mds${num} "$TUNE2FS -O extents $(mdsdevname $num)" ||
5136 error "$TUNE2FS failed on mds${num}"
5139 add ost1 $(mkfs_opts ost1 $ostdev) --reformat $ostdev ||
5140 error "add $ostdev failed"
5141 start_mgsmds || error "start mds failed"
5142 start_ost || error "start ost failed"
5143 mount_client $MOUNT || error "mount client failed"
5145 #create some short symlinks
5146 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5147 createmany -o $DIR/$tdir/$tfile-%d $fn
5148 echo "create $fn short symlinks"
5149 for i in $(seq -w 1 $fn); do
5150 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
5155 umount_client $MOUNT || error "umount client failed"
5156 stop_mds || error "stop mds failed"
5157 stop_ost || error "stop ost failed"
5160 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
5162 run_test 72 "test fast symlink with extents flag enabled"
5164 test_73() { #LU-3006
5165 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
5166 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
5167 error "1st tunefs failed"
5168 start_mgsmds || error "start mds failed"
5169 start_ost || error "start ost failed"
5170 mount_client $MOUNT || error "mount client failed"
5171 $LCTL get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
5172 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
5173 umount_client $MOUNT || error "umount client failed"
5177 run_test 73 "failnode to update from mountdata properly"
5179 test_75() { # LU-2374
5180 [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] &&
5181 skip "Need MDS version at least 2.4.1" && return
5184 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
5185 --reformat $(mdsdevname 1) $(mdsvdevname 1)"
5186 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
5187 --reformat $(ostdevname 1) $(ostvdevname 1)"
5189 #check with default parameters
5190 add mds1 $opts_mds || error "add mds1 failed for default params"
5191 add ost1 $opts_ost || error "add ost1 failed for default params"
5193 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
5194 opts_mds=$(echo $opts_mds |
5195 sed -e "s/--index=$index/--index=$index --mdt/")
5196 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
5197 opts_ost=$(echo $opts_ost |
5198 sed -e "s/--index=$index/--index=$index --ost/")
5200 add mds1 $opts_mds || error "add mds1 failed for new params"
5201 add ost1 $opts_ost || error "add ost1 failed for new params"
5204 run_test 75 "The order of --index should be irrelevant"
5207 [[ $(lustre_version_code mgs) -ge $(version_code 2.4.52) ]] ||
5208 { skip "Need MDS version at least 2.4.52" && return 0; }
5210 local MDMB_PARAM="osc.*.max_dirty_mb"
5211 echo "Change MGS params"
5212 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
5214 echo "max_dirty_mb: $MAX_DIRTY_MB"
5215 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB + MAX_DIRTY_MB))
5216 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
5217 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
5218 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5219 head -1" $NEW_MAX_DIRTY_MB
5220 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5221 echo "$MAX_DIRTY_MB"
5222 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5223 error "error while apply max_dirty_mb"
5225 echo "Check the value is stored after remount"
5228 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5229 head -1" $NEW_MAX_DIRTY_MB
5230 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5231 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5232 error "max_dirty_mb is not saved after remount"
5234 echo "Change OST params"
5235 CLIENT_PARAM="obdfilter.*.client_cache_count"
5236 local CLIENT_CACHE_COUNT
5237 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5239 echo "client_cache_count: $CLIENT_CACHE_COUNT"
5240 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
5241 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
5242 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
5243 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5244 head -1" $NEW_CLIENT_CACHE_COUNT
5245 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5247 echo "$CLIENT_CACHE_COUNT"
5248 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5249 error "error while apply client_cache_count"
5251 echo "Check the value is stored after remount"
5254 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5255 head -1" $NEW_CLIENT_CACHE_COUNT
5256 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5258 echo "$CLIENT_CACHE_COUNT"
5259 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5260 error "client_cache_count is not saved after remount"
5263 run_test 76a "set permanent params set_param -P"
5265 test_76b() { # LU-4783
5266 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.57) ]] ||
5267 { skip "Need MGS version at least 2.5.57" && return 0; }
5270 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
5271 error "start params log failed"
5274 run_test 76b "verify params log setup correctly"
5277 [[ $(lustre_version_code mgs) -ge $(version_code 2.8.54) ]] ||
5278 { skip "Need MDS version at least 2.4.52" && return 0; }
5280 local MASK_PARAM="mdd.*.changelog_mask"
5281 echo "Change changelog_mask"
5282 do_facet mgs $LCTL set_param -P $MASK_PARAM=-CLOSE ||
5283 error "Can't change changlog_mask"
5284 wait_update $(facet_host mds) "$LCTL get_param -n $MASK_PARAM |
5287 echo "Check the value is stored after mds remount"
5288 stop_mds || error "Failed to stop MDS"
5289 start_mds || error "Failed to start MDS"
5290 local CHANGELOG_MASK=$(do_facet mgs $LCTL get_param -n $MASK_PARAM)
5291 echo $CHANGELOG_MASK | grep CLOSE > /dev/null &&
5292 error "changelog_mask is not changed"
5296 run_test 76c "verify changelog_mask is applied with set_param -P"
5298 test_77() { # LU-3445
5299 local server_version=$(lustre_version_code $SINGLEMDS)
5300 [[ $server_version -ge $(version_code 2.8.55) ]] ||
5301 { skip "Need MDS version 2.8.55+ "; return; }
5303 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
5304 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
5305 skip_env "mixed loopback and real device not working" && return
5308 local fs2mdsdev=$(mdsdevname 1_2)
5309 local fs2ostdev=$(ostdevname 1_2)
5310 local fs2mdsvdev=$(mdsvdevname 1_2)
5311 local fs2ostvdev=$(ostvdevname 1_2)
5312 local fsname=test1234
5314 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
5316 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
5317 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
5318 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
5319 error "start fs2mds failed"
5321 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
5322 mgsnid="$mgsnid,$mgsnid:$mgsnid"
5324 add fs2ost $(mkfs_opts ost1 $fs2ostdev) --mgsnode=$mgsnid \
5325 --failnode=$failnid --fsname=$fsname \
5326 --reformat $fs2ostdev $fs2ostvdev ||
5327 error "add fs2ost failed"
5328 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
5330 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
5331 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
5332 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
5335 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
5338 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ||
5339 $(facet_fstype ost1) != ldiskfs ]] &&
5340 skip "only applicable to ldiskfs-based MDTs and OSTs" && return
5342 # reformat the Lustre filesystem with a smaller size
5343 local saved_MDSCOUNT=$MDSCOUNT
5344 local saved_MDSSIZE=$MDSSIZE
5345 local saved_OSTCOUNT=$OSTCOUNT
5346 local saved_OSTSIZE=$OSTSIZE
5349 MDSSIZE=$((MDSSIZE - 20000))
5350 OSTSIZE=$((OSTSIZE - 20000))
5351 reformat || error "(1) reformat Lustre filesystem failed"
5352 MDSSIZE=$saved_MDSSIZE
5353 OSTSIZE=$saved_OSTSIZE
5355 # mount the Lustre filesystem
5356 setup_noconfig || error "(2) setup Lustre filesystem failed"
5359 log "create test files"
5364 mkdir $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
5366 for i in $(seq $num_files); do
5367 file=$MOUNT/$tdir/$tfile-$i
5368 dd if=/dev/urandom of=$file count=1 bs=1M || {
5369 $LCTL get_param osc.*.cur*grant*
5370 $LFS df; $LFS df -i;
5371 # stop creating files if there is no more space
5372 if [ ! -e $file ]; then
5373 num_files=$((i - 1))
5377 $LFS getstripe -v $file
5378 local ost_idx=$(LFS getstripe -i $file)
5379 do_facet ost$((ost_idx + 1)) \
5380 $LCTL get_param obdfilter.*.*grant*
5381 error "(4) create $file failed"
5385 # unmount the Lustre filesystem
5386 cleanup || error "(5) cleanup Lustre filesystem failed"
5388 # run e2fsck on the MDT and OST devices
5389 local mds_host=$(facet_active_host $SINGLEMDS)
5390 local ost_host=$(facet_active_host ost1)
5391 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
5392 local ost_dev=$(ostdevname 1)
5394 run_e2fsck $mds_host $mds_dev "-y"
5395 run_e2fsck $ost_host $ost_dev "-y"
5397 # get the original block count of the MDT and OST filesystems
5398 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
5399 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
5401 # expand the MDT and OST filesystems to the device size
5402 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
5403 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
5405 # run e2fsck on the MDT and OST devices again
5406 run_e2fsck $mds_host $mds_dev "-y"
5407 run_e2fsck $ost_host $ost_dev "-y"
5409 # mount the Lustre filesystem
5413 log "check files after expanding the MDT and OST filesystems"
5414 for i in $(seq $num_files); do
5415 file=$MOUNT/$tdir/$tfile-$i
5416 $CHECKSTAT -t file -s 1048576 $file ||
5417 error "(6) checkstat $file failed"
5421 log "create more files after expanding the MDT and OST filesystems"
5422 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
5423 file=$MOUNT/$tdir/$tfile-$i
5424 dd if=/dev/urandom of=$file count=1 bs=1M ||
5425 error "(7) create $file failed"
5428 # unmount the Lustre filesystem
5429 cleanup || error "(8) cleanup Lustre filesystem failed"
5431 # run e2fsck on the MDT and OST devices
5432 run_e2fsck $mds_host $mds_dev "-y"
5433 run_e2fsck $ost_host $ost_dev "-y"
5435 # get the maximum block count of the MDT and OST filesystems
5436 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
5437 local ost_max_blks=$(get_block_count ost1 $ost_dev)
5439 # get the minimum block count of the MDT and OST filesystems
5440 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
5441 grep minimum | sed -e 's/^.*filesystem: //g')
5442 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
5443 grep minimum | sed -e 's/^.*filesystem: //g')
5445 # shrink the MDT and OST filesystems to a smaller size
5449 if [[ $mds_max_blks -gt $mds_min_blks &&
5450 $mds_max_blks -gt $mds_orig_blks ]]; then
5451 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
5452 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
5453 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
5454 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
5455 error "shrink $SINGLEMDS to $new_blks failed"
5459 if [[ $ost_max_blks -gt $ost_min_blks &&
5460 $ost_max_blks -gt $ost_orig_blks ]]; then
5461 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
5462 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
5463 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
5464 run_resize2fs ost1 $ost_dev $new_blks ||
5465 error "shrink ost1 to $new_blks failed"
5469 # check whether the MDT or OST filesystem was shrunk or not
5471 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
5472 reformat || error "(10) reformat Lustre filesystem failed"
5476 # run e2fsck on the MDT and OST devices again
5477 run_e2fsck $mds_host $mds_dev "-y"
5478 run_e2fsck $ost_host $ost_dev "-y"
5480 # mount the Lustre filesystem again
5484 log "check files after shrinking the MDT and OST filesystems"
5485 for i in $(seq $((num_files + 10))); do
5486 file=$MOUNT/$tdir/$tfile-$i
5487 $CHECKSTAT -t file -s 1048576 $file ||
5488 error "(11) checkstat $file failed"
5491 # unmount and reformat the Lustre filesystem
5492 cleanup || error "(12) cleanup Lustre filesystem failed"
5493 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
5495 MDSCOUNT=$saved_MDSCOUNT
5496 OSTCOUNT=$saved_OSTCOUNT
5497 reformat || error "(14) reformat Lustre filesystem failed"
5499 run_test 78 "run resize2fs on MDT and OST filesystems"
5501 test_79() { # LU-4227
5502 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.5.59) ]] ||
5503 { skip "Need MDS version at least 2.5.59"; return 0; }
5505 local mdsdev1=$(mdsdevname 1)
5506 local mdsvdev1=$(mdsvdevname 1)
5507 local mdsdev2=$(mdsdevname 2)
5508 local mdsvdev2=$(mdsvdevname 2)
5509 local ostdev1=$(ostdevname 1)
5510 local ostvdev1=$(ostvdevname 1)
5511 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
5512 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
5513 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
5516 # remove --mgs/--mgsnode from mkfs.lustre options
5517 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
5519 mgsnode_opt=$(echo $opts_mds2 |
5520 awk '{ for ( i = 1; i < NF; i++ )
5521 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5522 [ -n $mgsnode_opt ] &&
5523 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
5525 mgsnode_opt=$(echo $opts_ost1 |
5526 awk '{ for ( i = 1; i < NF; i++ )
5527 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5528 [ -n $mgsnode_opt ] &&
5529 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
5531 # -MGS, format a mdt without --mgs option
5532 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
5533 error "Must specify --mgs when formatting mdt combined with mgs"
5535 # +MGS, format a mdt/ost without --mgsnode option
5536 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
5537 > /dev/null || error "start mds1 failed"
5538 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
5539 error "Must specify --mgsnode when formatting a mdt"
5540 add ost1 $opts_ost1 $ostdev1 $ostvdev2 &&
5541 error "Must specify --mgsnode when formatting an ost"
5545 run_test 79 "format MDT/OST without mgs option (should return errors)"
5548 start_mds || error "Failed to start MDT"
5549 start_ost || error "Failed to start OST1"
5550 uuid=$(do_facet ost1 $LCTL get_param -n mgc.*.uuid)
5551 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
5552 do_facet ost1 "$LCTL set_param fail_val=10 fail_loc=0x906"
5553 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x906"
5554 do_facet mgs "$LCTL set_param -n mgs/MGS/evict_client $uuid"
5556 start_ost2 || error "Failed to start OST2"
5558 do_facet ost1 "$LCTL set_param fail_loc=0"
5561 run_test 80 "mgc import reconnect race"
5563 #Save the original values of $OSTCOUNT and $OSTINDEX$i.
5565 local new_ostcount=$1
5566 saved_ostcount=$OSTCOUNT
5567 OSTCOUNT=$new_ostcount
5571 for i in $(seq $OSTCOUNT); do
5573 eval saved_ostindex$i=${!index}
5578 # Restore the original values of $OSTCOUNT and $OSTINDEX$i.
5579 restore_ostindex() {
5584 for i in $(seq $OSTCOUNT); do
5585 index=saved_ostindex$i
5586 eval OSTINDEX$i=${!index}
5588 OSTCOUNT=$saved_ostcount
5593 # The main purpose of this test is to ensure the OST_INDEX_LIST functions as
5594 # expected. This test uses OST_INDEX_LIST to format OSTs with a randomly
5595 # assigned index and ensures we can mount such a formatted file system
5596 test_81() { # LU-4665
5597 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5598 { skip "Need MDS version at least 2.6.54" && return; }
5599 [[ $OSTCOUNT -ge 3 ]] || { skip_env "Need at least 3 OSTs" && return; }
5603 # Each time RANDOM is referenced, a random integer between 0 and 32767
5606 local saved_ostindex1=$OSTINDEX1
5607 for i in 65535 $((RANDOM + 65536)); do
5608 echo -e "\nFormat ost1 with --index=$i, should fail"
5610 if add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat \
5611 $(ostdevname 1) $(ostvdevname 1); then
5612 OSTINDEX1=$saved_ostindex1
5613 error "format ost1 with --index=$i should fail"
5616 OSTINDEX1=$saved_ostindex1
5620 # Format OSTs with random sparse indices.
5621 trap "restore_ostindex" EXIT
5622 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices"
5623 OST_INDEX_LIST=[0,$((RANDOM * 2 % 65533 + 1)),65534] formatall
5625 # Setup and check Lustre filesystem.
5626 start_mgsmds || error "start_mgsmds failed"
5627 for i in $(seq $OSTCOUNT); do
5628 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5629 error "start ost$i failed"
5632 mount_client $MOUNT || error "mount client $MOUNT failed"
5633 check_mount || error "check client $MOUNT failed"
5636 local max_easize=$($LCTL get_param -n llite.*.max_easize)
5637 [[ $max_easize -eq 128 ]] ||
5638 error "max_easize is $max_easize, should be 128 bytes"
5642 run_test 81 "sparse OST indexing"
5644 # Here we exercise the stripe placement functionality on a file system that
5645 # has formatted the OST with a random index. With the file system the following
5646 # functionality is tested:
5648 # 1. Creating a new file with a specific stripe layout.
5650 # 2. Modifiy a existing empty file with a specific stripe layout.
5652 # 3. Ensure we fail to set the stripe layout of a file that already has one.
5654 # 4. If ost-index is defined we need to ensure it is the first entry in the
5655 # ost index list returned by lfs getstripe.
5657 # 5. Lastly ensure this functionality fails with directories.
5658 test_82a() { # LU-4665
5659 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5660 { skip "Need MDS version at least 2.6.54" && return; }
5661 [[ $OSTCOUNT -ge 3 ]] || { skip_env "Need at least 3 OSTs" && return; }
5667 # Format OSTs with random sparse indices.
5671 local LOV_V1_INSANE_STRIPE_COUNT=65532
5672 for i in $(seq $OSTCOUNT); do
5673 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
5674 ost_indices+=" $index"
5676 ost_indices=$(comma_list $ost_indices)
5678 trap "restore_ostindex" EXIT
5679 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5680 OST_INDEX_LIST=[$ost_indices] formatall
5682 # Setup Lustre filesystem.
5683 start_mgsmds || error "start_mgsmds failed"
5684 for i in $(seq $OSTCOUNT); do
5685 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5686 error "start ost$i failed"
5689 mount_client $MOUNT || error "mount client $MOUNT failed"
5692 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5693 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5695 # 1. If the file does not exist, new file will be created
5696 # with specified OSTs.
5697 local file=$DIR/$tdir/$tfile-1
5698 local cmd="$SETSTRIPE -o $ost_indices $file"
5700 eval $cmd || error "$cmd failed"
5701 check_stripe_count $file $OSTCOUNT
5702 check_obdidx $file $ost_indices
5703 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5704 error "write $file failed"
5706 # 2. If the file already exists and is an empty file, the file
5707 # will be attached with specified layout.
5708 file=$DIR/$tdir/$tfile-2
5709 mcreate $file || error "mcreate $file failed"
5710 cmd="$SETSTRIPE -o $ost_indices $file"
5712 eval $cmd || error "$cmd failed"
5713 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5714 error "write $file failed"
5715 check_stripe_count $file $OSTCOUNT
5716 check_obdidx $file $ost_indices
5718 # 3. If the file already has a valid layout attached, the command
5719 # should fail with EBUSY.
5721 eval $cmd && error "stripe is already set on $file, $cmd should fail"
5723 # 4. If [--stripe-index|-i <start_ost_idx>] is used, the index must
5724 # be in the OST indices list.
5725 local start_ost_idx=${ost_indices##*,}
5726 file=$DIR/$tdir/$tfile-3
5727 cmd="$SETSTRIPE -o $ost_indices -i $start_ost_idx $file"
5729 eval $cmd || error "$cmd failed"
5730 check_stripe_count $file $OSTCOUNT
5731 check_obdidx $file $ost_indices
5732 check_start_ost_idx $file $start_ost_idx
5734 file=$DIR/$tdir/$tfile-4
5736 cmd+=" -o $(exclude_items_from_list $ost_indices $start_ost_idx)"
5737 cmd+=" -i $start_ost_idx $file"
5739 eval $cmd && error "index $start_ost_idx should be in $ost_indices"
5741 # 5. Specifying OST indices for directory should fail with ENOSUPP.
5742 local dir=$DIR/$tdir/$tdir
5743 mkdir $dir || error "mkdir $dir failed"
5744 cmd="$SETSTRIPE -o $ost_indices $dir"
5746 eval $cmd && error "$cmd should fail, specifying OST indices" \
5747 "for directory is not supported"
5751 run_test 82a "specify OSTs for file (succeed) or directory (fail)"
5756 # Remove OSTs from a pool and destroy the pool.
5757 destroy_pool $ost_pool || true
5762 # Test 82b is run to ensure that if the user supplies a pool with a specific
5763 # stripe layout that it behaves proprerly. It should fail in the case that
5764 # the supplied OST index list points to OSTs not contained in the user
5766 test_82b() { # LU-4665
5767 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5768 { skip "Need MDS version at least 2.6.54" && return; }
5769 [[ $OSTCOUNT -ge 4 ]] || { skip_env "Need at least 4 OSTs" && return; }
5775 # Format OSTs with random sparse indices.
5779 local LOV_V1_INSANE_STRIPE_COUNT=65532
5780 for i in $(seq $OSTCOUNT); do
5781 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
5782 ost_indices+=" $index"
5784 ost_indices=$(comma_list $ost_indices)
5786 trap "restore_ostindex" EXIT
5787 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5788 OST_INDEX_LIST=[$ost_indices] formatall
5790 # Setup Lustre filesystem.
5791 start_mgsmds || error "start_mgsmds failed"
5792 for i in $(seq $OSTCOUNT); do
5793 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5794 error "start ost$i failed"
5797 mount_client $MOUNT || error "mount client $MOUNT failed"
5799 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5800 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5802 # Create a new pool and add OSTs into it.
5803 local ost_pool=$FSNAME.$TESTNAME
5804 create_pool $ost_pool || error "create OST pool $ost_pool failed"
5807 trap "cleanup_82b" EXIT
5809 local ost_idx_in_list=${ost_indices##*,}
5810 local ost_idx_in_pool=$(exclude_items_from_list $ost_indices \
5813 local ost_targets="$FSNAME-OST["
5814 for i in ${ost_idx_in_pool//,/ }; do
5815 ost_targets=$ost_targets$(printf "%04x," $i)
5817 ost_targets="${ost_targets%,}]"
5819 local ost_targets_uuid=$(for i in ${ost_idx_in_pool//,/ }; \
5820 do printf "$FSNAME-OST%04x_UUID\n" $i; done |
5821 sort -u | tr '\n' ' ')
5823 local cmd="$LCTL pool_add $ost_pool $ost_targets"
5824 do_facet mgs $cmd || error "$cmd failed"
5825 wait_update $HOSTNAME "$LCTL get_param -n lov.$FSNAME-*.pools.$TESTNAME|
5826 sort -u | tr '\n' ' ' " "$ost_targets_uuid" ||
5827 error "wait_update $ost_pool failed"
5828 [[ -z $(list_pool $ost_pool) ]] &&
5829 error "list OST pool $ost_pool failed"
5831 # If [--pool|-p <pool_name>] is set with [--ost-list|-o <ost_indices>],
5832 # then the OSTs must be the members of the pool.
5833 local file=$DIR/$tdir/$tfile
5834 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
5836 eval $cmd && error "OST with index $ost_idx_in_list should be" \
5837 "in OST pool $ost_pool"
5839 # Only select OST $ost_idx_in_list from $ost_pool for file.
5840 ost_idx_in_list=${ost_idx_in_pool#*,}
5841 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
5843 eval $cmd || error "$cmd failed"
5844 cmd="$GETSTRIPE $file"
5846 eval $cmd || error "$cmd failed"
5847 check_stripe_count $file 2
5848 check_obdidx $file $ost_idx_in_list
5849 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5850 error "write $file failed"
5854 run_test 82b "specify OSTs for file with --pool and --ost-list options"
5857 [[ $(lustre_version_code ost1) -ge $(version_code 2.6.91) ]] ||
5858 { skip "Need OST version at least 2.6.91" && return 0; }
5859 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
5860 skip "Only applicable to ldiskfs-based MDTs"
5870 ostmnt=$(facet_mntpt ost1)
5871 fstype=$(facet_fstype ost1)
5873 # Mount the OST as an ldiskfs filesystem.
5874 log "mount the OST $dev as a $fstype filesystem"
5875 add ost1 $(mkfs_opts ost1 $dev) $FSTYPE_OPT \
5876 --reformat $dev $dev > /dev/null ||
5877 error "format ost1 error"
5879 if ! test -b $dev; then
5880 mnt_opts=$(csa_add "$OST_MOUNT_OPTS" -o loop)
5882 echo "mnt_opts $mnt_opts"
5883 do_facet ost1 mount -t $fstype $dev \
5885 # Run llverfs on the mounted ldiskfs filesystem.
5886 # It is needed to get ENOSPACE.
5887 log "run llverfs in partial mode on the OST $fstype $ostmnt"
5888 do_rpc_nodes $(facet_host ost1) run_llverfs $ostmnt -vpl \
5889 "no" || error "run_llverfs error on $fstype"
5892 log "unmount the OST $dev"
5895 # Delete file IO_scrub. Later osd_scrub_setup will try to
5896 # create "IO_scrub" but will get ENOSPACE.
5898 echo "start ost1 service on `facet_active_host ost1`"
5899 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
5902 err=$(do_facet ost1 dmesg | grep "VFS: Busy inodes after unmount of")
5903 echo "string err $err"
5904 [ -z "$err" ] || error $err
5907 run_test 83 "ENOSPACE on OST doesn't cause message VFS: \
5908 Busy inodes after unmount ..."
5911 local facet=$SINGLEMDS
5912 local num=$(echo $facet | tr -d "mds")
5913 local dev=$(mdsdevname $num)
5914 local time_min=$(recovery_time_min)
5915 local recovery_duration
5916 local completed_clients
5917 local correct_clients
5920 echo "start mds service on $(facet_active_host $facet)"
5922 "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
5923 error "start MDS failed"
5925 start_ost || error "start OST0000 failed"
5926 start_ost2 || error "start OST0001 failed"
5928 echo "recovery_time=$time_min, timeout=$TIMEOUT, wrap_up=$wrap_up"
5930 mount_client $MOUNT1 || error "mount $MOUNT1 failed"
5931 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
5932 # make sure new superblock labels are sync'd before disabling writes
5936 replay_barrier $SINGLEMDS
5937 createmany -o $DIR1/$tfile-%d 1000
5939 # We need to catch the end of recovery window to extend it.
5940 # Skip 5 requests and add delay to request handling.
5941 #define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP
5942 do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5"
5944 facet_failover --fsck $SINGLEMDS || error "failover: $?"
5947 echo "recovery status"
5948 do_facet $SINGLEMDS \
5949 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status"
5951 recovery_duration=$(do_facet $SINGLEMDS \
5952 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
5953 awk '/recovery_duration/ { print $2 }')
5954 (( $recovery_duration > $time_min + $wrap_up )) &&
5955 error "recovery_duration > recovery_time_hard + wrap up"
5956 completed_clients=$(do_facet $SINGLEMDS \
5957 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
5958 awk '/completed_clients/ { print $2 }')
5960 correct_clients="$MDSCOUNT/$((MDSCOUNT+1))"
5961 [ "$completed_clients" = "${correct_clients}" ] ||
5962 error "$completed_clients != $correct_clients"
5964 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
5965 umount_client $MOUNT1
5966 umount_client $MOUNT2
5972 run_test 84 "check recovery_hard_time"
5975 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.55) ]] ||
5976 { skip "Need OST version at least 2.7.55" && return 0; }
5977 ##define OBD_FAIL_OSD_OST_EA_FID_SET 0x197
5978 do_facet ost1 "lctl set_param fail_loc=0x197"
5982 run_test 85 "osd_ost init: fail ea_fid_set"
5987 # ost1 has already registered to the MGS before the reformat.
5988 # So after reformatting it with option "-G", it could not be
5989 # mounted to the MGS. Cleanup the system for subsequent tests.
5994 local server_version=$(lustre_version_code $SINGLEMDS)
5995 [ "$(facet_fstype ost1)" = "zfs" ] &&
5996 skip "LU-6442: no such mkfs params for ZFS OSTs" && return
5997 [[ $server_version -ge $(version_code 2.7.56) ]] ||
5998 { skip "Need server version newer than 2.7.55"; return 0; }
6000 local OST_OPTS="$(mkfs_opts ost1 $(ostdevname 1)) \
6001 --reformat $(ostdevname 1) $(ostvdevname 1)"
6004 local OLDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6005 awk '/Flex block group size: / { print $NF; exit; }')
6008 if [[ ${!opts} != *mkfsoptions* ]]; then
6009 eval opts=\"${!opts} \
6010 --mkfsoptions='\\\"-O flex_bg -G $NEWSIZE\\\"'\"
6012 val=${!opts//--mkfsoptions=\\\"/ \
6013 --mkfsoptions=\\\"-O flex_bg -G $NEWSIZE }
6017 echo "params: $opts"
6019 trap cleanup_86 EXIT ERR
6022 add ost1 $opts || error "add ost1 failed with new params"
6024 local FOUNDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6025 awk '/Flex block group size: / { print $NF; exit; }')
6027 [[ $FOUNDSIZE == $NEWSIZE ]] ||
6028 error "Flex block group size: $FOUNDSIZE, expected: $NEWSIZE"
6032 run_test 86 "Replacing mkfs.lustre -G option"
6034 test_87() { #LU-6544
6035 [[ $(lustre_version_code $SINGLEMDS1) -ge $(version_code 2.7.56) ]] ||
6036 { skip "Need MDS version at least 2.7.56" && return; }
6037 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ]] &&
6038 { skip "Only applicable to ldiskfs-based MDTs" && return; }
6039 [[ $OSTCOUNT -gt 69 ]] &&
6040 { skip "Ignore wide striping situation" && return; }
6042 local mdsdev=$(mdsdevname 1)
6043 local mdsvdev=$(mdsvdevname 1)
6044 local file=$DIR/$tfile
6045 local mntpt=$(facet_mntpt $SINGLEMDS)
6046 local used_xattr_blk=0
6047 local inode_size=${1:-512}
6049 local xtest="trusted.test"
6054 #Please see LU-6544 for MDT inode size calculation
6055 if [ $OSTCOUNT -gt 26 ]; then
6057 elif [ $OSTCOUNT -gt 5 ]; then
6060 left_size=$(expr $inode_size - \
6063 32 - $OSTCOUNT \* 24 - 16 - 3 - \
6065 24 - 18 - $(expr length $tfile) - 16 - 4)
6066 if [ $left_size -le 0 ]; then
6067 echo "No space($left_size) is expected in inode."
6068 echo "Try 1-byte xattr instead to verify this."
6071 echo "Estimate: at most $left_size-byte space left in inode."
6077 add mds1 $(mkfs_opts mds1 ${mdsdev}) --stripe-count-hint=$OSTCOUNT \
6078 --reformat $mdsdev $mdsvdev || error "add mds1 failed"
6079 start_mdt 1 > /dev/null || error "start mdt1 failed"
6080 for i in $(seq $OSTCOUNT); do
6081 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS > /dev/null ||
6082 error "start ost$i failed"
6084 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
6085 check_mount || error "check client $MOUNT failed"
6088 $SETSTRIPE -c -1 $file || error "$SETSTRIPE -c -1 $file failed"
6089 $GETSTRIPE $file || error "$GETSTRIPE $file failed"
6090 i=$($GETSTRIPE -c $file)
6091 if [ $i -ne $OSTCOUNT ]; then
6092 left_size=$(expr $left_size + $(expr $OSTCOUNT - $i) \* 24)
6093 echo -n "Since only $i out $OSTCOUNT OSTs are used, "
6094 echo -n "the expected left space is changed to "
6095 echo "$left_size bytes at most."
6097 value=$(generate_string $left_size)
6098 setfattr -n $xtest -v $value $file
6099 orig=$(get_xattr_value $xtest $file)
6100 [[ "$orig" != "$value" ]] && error "$xtest changed"
6102 #Verify if inode has some expected space left
6103 umount $MOUNT > /dev/null || error "umount $MOUNT failed"
6104 stop_mdt 1 > /dev/null || error "stop mdt1 failed"
6105 mount_ldiskfs $SINGLEMDS || error "mount -t ldiskfs $SINGLEMDS failed"
6107 do_facet $SINGLEMDS ls -sal $mntpt/ROOT/$tfile
6108 used_xattr_blk=$(do_facet $SINGLEMDS ls -s $mntpt/ROOT/$tfile |
6110 [[ $used_xattr_blk -eq 0 ]] &&
6111 error "Please check MDS inode size calculation: \
6112 more than $left_size-byte space left in inode."
6113 echo "Verified: at most $left_size-byte space left in inode."
6117 run_test 87 "check if MDT inode can hold EAs with N stripes properly"
6120 [ "$(facet_fstype mds1)" == "zfs" ] &&
6121 skip "LU-6662: no implementation for ZFS" && return
6125 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
6126 --reformat $(mdsdevname 1) || error "add mds1 failed"
6128 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
6129 grep -e \".*opts:.*errors=remount-ro.*\"" ||
6130 error "default mount options is missing"
6132 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
6133 --mountfsoptions="user_xattr,errors=panic" \
6134 --reformat $(mdsdevname 1) || error "add mds1 failed"
6136 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
6137 grep -e \".*opts:.*errors=panic.*\"" ||
6138 error "user can't override default mount options"
6140 run_test 88 "check the default mount options can be overridden"
6143 # $2 (optional) value of max_mod_rpcs_in_flight to set
6144 check_max_mod_rpcs_in_flight() {
6152 idx=$(printf "%04x" $($LFS getdirstripe -i $dir))
6153 facet="mds$((0x$idx + 1))"
6155 if [ -z "$mmr" ]; then
6156 # get value of max_mod_rcps_in_flight
6157 mmr=$($LCTL get_param -n \
6158 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) ||
6159 error "Unable to get max_mod_rpcs_in_flight"
6160 echo "max_mod_rcps_in_flight is $mmr"
6162 # set value of max_mod_rpcs_in_flight
6164 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight=$mmr ||
6165 error "Unable to set max_mod_rpcs_in_flight to $mmr"
6166 echo "max_mod_rpcs_in_flight set to $mmr"
6169 # create mmr+1 files
6170 echo "creating $((mmr + 1)) files ..."
6172 for i in $(seq $((mmr + 1))); do
6178 # consumes mmr-1 modify RPC slots
6179 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6180 # drop requests on MDT so that RPC slots are consumed
6181 # during all the request resend interval
6182 do_facet $facet "$LCTL set_param fail_loc=0x159"
6183 echo "launch $((mmr - 1)) chmod in parallel ..."
6184 for i in $(seq $((mmr - 1))); do
6185 chmod 0600 $dir/file-$i &
6189 # send one additional modify RPC
6190 do_facet $facet "$LCTL set_param fail_loc=0"
6191 echo "launch 1 additional chmod in parallel ..."
6192 chmod 0600 $dir/file-$mmr &
6195 # check this additional modify RPC get a modify RPC slot
6196 # and succeed its operation
6197 checkstat -vp 0600 $dir/file-$mmr ||
6198 error "Unable to send $mmr modify RPCs in parallel"
6203 # consumes mmr modify RPC slots
6204 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6205 # drop requests on MDT so that RPC slots are consumed
6206 # during all the request resend interval
6207 do_facet $facet "$LCTL set_param fail_loc=0x159"
6208 echo "launch $mmr chmod in parallel ..."
6209 for i in $(seq $mmr); do
6210 chmod 0666 $dir/file-$i &
6214 # send one additional modify RPC
6215 do_facet $facet "$LCTL set_param fail_loc=0"
6216 echo "launch 1 additional chmod in parallel ..."
6217 chmod 0666 $dir/file-$((mmr + 1)) &
6220 # check this additional modify RPC blocked getting a modify RPC slot
6221 checkstat -vp 0644 $dir/file-$((mmr + 1)) ||
6222 error "Unexpectedly send $(($mmr + 1)) modify RPCs in parallel"
6228 if ! combined_mgs_mds ; then
6233 [[ $($LCTL get_param mdc.*.import |
6234 grep "connect_flags:.*multi_mod_rpc") ]] ||
6235 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6237 # check default value
6238 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6239 check_max_mod_rpcs_in_flight $DIR/$tdir
6243 run_test 90a "check max_mod_rpcs_in_flight is enforced"
6253 [[ $($LCTL get_param mdc.*.import |
6254 grep "connect_flags:.*multi_mod_rpc") ]] ||
6255 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6258 # update max_mod_rpcs_in_flight
6259 $LFS mkdir -c1 $DIR/${tdir}1 || error "mkdir $DIR/${tdir}1 failed"
6260 check_max_mod_rpcs_in_flight $DIR/${tdir}1 1
6263 # check client is able to send multiple modify RPCs in paralell
6264 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
6265 grep -c "multi_mod_rpcs")
6266 if [ "$tmp" -ne $MDSCOUNT ]; then
6267 echo "Client not able to send multiple modify RPCs in parallel"
6272 # update max_mod_rpcs_in_flight
6273 $LFS mkdir -c1 $DIR/${tdir}2 || error "mkdir $DIR/${tdir}2 failed"
6274 check_max_mod_rpcs_in_flight $DIR/${tdir}2 5
6277 $LFS mkdir -c1 $DIR/${tdir}3 || error "mkdir $DIR/${tdir}3 failed"
6278 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/${tdir}3))
6279 facet="mds$((0x$idx + 1))"
6281 # save MDT max_mod_rpcs_per_client
6282 mmrpc=$(do_facet $facet \
6283 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
6285 # update max_mod_rpcs_in_flight
6286 umount_client $MOUNT
6288 "echo 16 > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6290 $LCTL set_param mdc.$FSNAME-MDT$idx-mdc-*.max_rpcs_in_flight=17
6291 check_max_mod_rpcs_in_flight $DIR/${tdir}3 16
6293 # restore MDT max_mod_rpcs_per_client initial value
6295 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6297 rm -rf $DIR/${tdir}?
6300 run_test 90b "check max_mod_rpcs_in_flight is enforced after update"
6309 [[ $($LCTL get_param mdc.*.import |
6310 grep "connect_flags:.*multi_mod_rpc") ]] ||
6311 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6313 # check client is able to send multiple modify RPCs in paralell
6314 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
6315 grep -c "multi_mod_rpcs")
6316 if [ "$tmp" -ne $MDSCOUNT ]; then
6317 skip "Client not able to send multiple modify RPCs in parallel"
6322 # get max_rpcs_in_flight value
6323 mrif=$($LCTL get_param -n mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
6324 echo "max_rpcs_in_flight is $mrif"
6326 # get MDT max_mod_rpcs_per_client
6327 mmrpc=$(do_facet mds1 \
6328 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
6329 echo "max_mod_rpcs_per_client is $mmrpc"
6332 # attempt to set max_mod_rpcs_in_flight to max_rpcs_in_flight value
6333 # prerequisite: set max_mod_rpcs_per_client to max_rpcs_in_flight value
6334 umount_client $MOUNT
6336 "echo $mrif > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6340 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif &&
6341 error "set max_mod_rpcs_in_flight to $mrif should fail"
6343 umount_client $MOUNT
6345 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6349 # attempt to set max_mod_rpcs_in_flight to max_mod_rpcs_per_client+1
6350 # prerequisite: set max_rpcs_in_flight to max_mod_rpcs_per_client+2
6352 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$((mmrpc + 2))
6355 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$((mmrpc + 1)) &&
6356 error "set max_mod_rpcs_in_flight to $((mmrpc + 1)) should fail"
6360 run_test 90c "check max_mod_rpcs_in_flight update limits"
6371 [[ $($LCTL get_param mdc.*.import |
6372 grep "connect_flags:.*multi_mod_rpc") ]] ||
6373 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6375 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6376 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
6377 facet="mds$((0x$idx + 1))"
6379 # check client version supports multislots
6380 tmp=$($LCTL get_param -N \
6381 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
6382 if [ -z "$tmp" ]; then
6383 skip "Client does not support multiple modify RPCs in flight"
6388 # get current value of max_mod_rcps_in_flight
6389 mmr=$($LCTL get_param -n \
6390 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
6391 echo "max_mod_rcps_in_flight is $mmr"
6394 echo "creating $mmr files ..."
6396 for i in $(seq $mmr); do
6397 touch $DIR/$tdir/file-$i
6400 # prepare for close RPC
6401 multiop_bg_pause $DIR/$tdir/file-close O_c
6404 # consumes mmr modify RPC slots
6405 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6406 # drop requests on MDT so that RPC slots are consumed
6407 # during all the request resend interval
6408 do_facet $facet "$LCTL set_param fail_loc=0x159"
6409 echo "launch $mmr chmod in parallel ..."
6410 for i in $(seq $mmr); do
6411 chmod 0600 $DIR/$tdir/file-$i &
6414 # send one additional close RPC
6415 do_facet $facet "$LCTL set_param fail_loc=0"
6416 echo "launch 1 additional close in parallel ..."
6418 cancel_lru_locks mdc
6421 # check this additional close RPC get a modify RPC slot
6422 # and multiop process completed
6423 [ -d /proc/$pid ] &&
6424 error "Unable to send the additional close RPC in parallel"
6429 run_test 90d "check one close RPC is allowed above max_mod_rpcs_in_flight"
6431 check_uuid_on_ost() {
6433 do_facet ost1 "$LCTL get_param obdfilter.${FSNAME}*.exports.'$nid'.uuid"
6436 check_uuid_on_mdt() {
6438 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
6446 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.63) ]] ||
6447 { skip "Need OST version at least 2.7.63" && return 0; }
6448 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.63) ]] ||
6449 { skip "Need MDT version at least 2.7.63" && return 0; }
6451 start_mds || error "MDS start failed"
6452 start_ost || error "unable to start OST"
6453 mount_client $MOUNT || error "client start failed"
6454 check_mount || error "check_mount failed"
6457 nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g")
6461 uuid=$(get_client_uuid $MOUNT)
6463 echo "list nids on mdt:"
6464 do_facet $SINGLEMDS "$LCTL list_param mdt.${FSNAME}*.exports.*"
6465 echo "uuid from $nid:"
6466 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
6468 found=$(check_uuid_on_mdt $nid | grep $uuid)
6469 [ -z "$found" ] && error "can't find $uuid $nid on MDT"
6470 found=$(check_uuid_on_ost $nid | grep $uuid)
6471 [ -z "$found" ] && error "can't find $uuid $nid on OST"
6473 # umount the client so it won't reconnect
6474 manual_umount_client --force || error "failed to umount $?"
6475 # shouldn't disappear on MDS after forced umount
6476 found=$(check_uuid_on_mdt $nid | grep $uuid)
6477 [ -z "$found" ] && error "can't find $uuid $nid"
6480 do_facet $SINGLEMDS \
6481 "$LCTL set_param -n mdt.${mds1_svc}.evict_client nid:$nid"
6483 found=$(check_uuid_on_mdt $nid | grep $uuid)
6484 [ -n "$found" ] && error "found $uuid $nid on MDT"
6485 found=$(check_uuid_on_ost $nid | grep $uuid)
6486 [ -n "$found" ] && error "found $uuid $nid on OST"
6488 # check it didn't reconnect (being umounted)
6489 sleep $((TIMEOUT+1))
6490 found=$(check_uuid_on_mdt $nid | grep $uuid)
6491 [ -n "$found" ] && error "found $uuid $nid on MDT"
6492 found=$(check_uuid_on_ost $nid | grep $uuid)
6493 [ -n "$found" ] && error "found $uuid $nid on OST"
6497 run_test 91 "evict-by-nid support"
6499 generate_ldev_conf() {
6500 # generate an ldev.conf file
6501 local ldevconfpath=$1
6503 local fsldevformat=""
6506 fstype=$(facet_fstype mgs)
6507 if [ "$fstype" == "zfs" ]; then
6508 fsldevformat="$fstype:"
6513 printf "%s\t-\t%s-MGS0000\t%s%s\n" \
6517 $(mgsdevname) > $ldevconfpath
6519 local mdsfo_host=$mdsfailover_HOST;
6520 if [ -z "$mdsfo_host" ]; then
6524 for num in $(seq $MDSCOUNT); do
6525 fstype=$(facet_fstype mds$num)
6526 if [ "$fstype" == "zfs" ]; then
6527 fsldevformat="$fstype:"
6532 printf "%s\t%s\t%s-MDT%04d\t%s%s\n" \
6538 $(mdsdevname $num) >> $ldevconfpath
6541 local ostfo_host=$ostfailover_HOST;
6542 if [ -z "$ostfo_host" ]; then
6546 for num in $(seq $OSTCOUNT); do
6547 fstype=$(facet_fstype ost$num)
6548 if [ "$fstype" == "zfs" ]; then
6549 fsldevformat="$fstype:"
6554 printf "%s\t%s\t%s-OST%04d\t%s%s\n" \
6560 $(ostdevname $num) >> $ldevconfpath
6563 echo "----- $ldevconfpath -----"
6565 echo "--- END $ldevconfpath ---"
6570 # generate a nids file (mapping between hostname to nid)
6571 # looks like we only have the MGS nid available to us
6572 # so just echo that to a file
6574 echo -e "${mgs_HOST}\t${MGSNID}" > $nidspath
6576 echo "----- $nidspath -----"
6578 echo "--- END $nidspath ---"
6581 compare_ldev_output() {
6585 sort $expected_output -o $expected_output
6586 sort $ldev_output -o $ldev_output
6588 echo "-- START OF LDEV OUTPUT --"
6590 echo "--- END OF LDEV OUTPUT ---"
6592 echo "-- START OF EXPECTED OUTPUT --"
6593 cat $expected_output
6594 echo "--- END OF EXPECTED OUTPUT ---"
6596 diff $expected_output $ldev_output
6601 if [ -z "$LDEV" ]; then
6602 error "ldev is missing!"
6605 local LDEVCONFPATH=$TMP/ldev.conf
6606 local NIDSPATH=$TMP/nids
6608 echo "Host is $(hostname)"
6610 generate_ldev_conf $LDEVCONFPATH
6611 generate_nids $NIDSPATH
6613 # echo the mgs nid and compare it to environment variable MGSNID
6614 # also, ldev.conf and nids is a server side thing, use the OSS
6617 output=$($LDEV -c $LDEVCONFPATH -H $ost_HOST -n $NIDSPATH echo %m)
6619 echo "-- START OF LDEV OUTPUT --"
6621 echo "--- END OF LDEV OUTPUT ---"
6623 # ldev failed, error
6624 if [ $? -ne 0 ]; then
6625 rm $LDEVCONFPATH $NIDSPATH
6626 error "ldev failed to execute!"
6629 # need to process multiple lines because of combined MGS and MDS
6630 echo -e $output | awk '{ print $2 }' | while read -r line ; do
6631 if [ "$line" != "$MGSNID" ]; then
6632 rm $LDEVCONFPATH $NIDSPATH
6633 error "ldev failed mgs nid '$line', expected '$MGSNID'"
6637 rm $LDEVCONFPATH $NIDSPATH
6639 run_test 92 "ldev returns MGS NID correctly in command substitution"
6642 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs" && return
6645 #start mgs or mgs/mdt0
6646 if ! combined_mgs_mds ; then
6653 start_ost || error "OST0 start fail"
6655 #define OBD_FAIL_MGS_WRITE_TARGET_DELAY 0x90e
6656 do_facet mgs "$LCTL set_param fail_val = 10 fail_loc=0x8000090e"
6657 for num in $(seq 2 $MDSCOUNT); do
6661 mount_client $MOUNT || error "mount client fails"
6662 wait_osc_import_state mds ost FULL
6663 wait_osc_import_state client ost FULL
6664 check_mount || error "check_mount failed"
6666 cleanup || error "cleanup failed with $?"
6668 run_test 93 "register mulitple MDT at the same time"
6671 if [ -z "$LDEV" ]; then
6672 error "ldev is missing!"
6675 local LDEVCONFPATH=$TMP/ldev.conf
6676 local NIDSPATH=$TMP/nids
6678 generate_ldev_conf $LDEVCONFPATH
6679 generate_nids $NIDSPATH
6681 local LDEV_OUTPUT=$TMP/ldev-output.txt
6682 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME > $LDEV_OUTPUT
6684 # ldev failed, error
6685 if [ $? -ne 0 ]; then
6686 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
6687 error "ldev failed to execute!"
6691 local EXPECTED_OUTPUT=$TMP/ldev-expected.txt
6693 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
6695 for num in $(seq $MDSCOUNT); do
6696 printf "%s-MDT%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
6699 for num in $(seq $OSTCOUNT); do
6700 printf "%s-OST%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
6703 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6705 if [ $? -ne 0 ]; then
6706 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6707 error "ldev failed to produce the correct hostlist!"
6710 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6712 run_test 94 "ldev outputs correct labels for file system name query"
6715 if [ -z "$LDEV" ]; then
6716 error "ldev is missing!"
6719 local LDEVCONFPATH=$TMP/ldev.conf
6720 local NIDSPATH=$TMP/nids
6722 generate_ldev_conf $LDEVCONFPATH
6723 generate_nids $NIDSPATH
6727 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME &>/dev/null
6728 if [ $? -ne 0 ]; then
6729 rm $LDEVCONFPATH $NIDSPATH
6730 error "ldev label filtering w/ -F failed!"
6734 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -l &>/dev/null
6735 if [ $? -ne 0 ]; then
6736 rm $LDEVCONFPATH $NIDSPATH
6737 error "ldev label filtering w/ -l failed!"
6741 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f &>/dev/null
6742 if [ $? -ne 0 ]; then
6743 rm $LDEVCONFPATH $NIDSPATH
6744 error "ldev label filtering w/ -f failed!"
6748 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a &>/dev/null
6749 if [ $? -ne 0 ]; then
6750 rm $LDEVCONFPATH $NIDSPATH
6751 error "ldev label filtering w/ -a failed!"
6756 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -F $FSNAME &>/dev/null
6757 if [ $? -eq 0 ]; then
6758 rm $LDEVCONFPATH $NIDSPATH
6759 error "ldev label filtering w/ -a and -F incorrectly succeeded"
6763 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -f &>/dev/null
6764 if [ $? -eq 0 ]; then
6765 rm $LDEVCONFPATH $NIDSPATH
6766 error "ldev label filtering w/ -a and -f incorrectly succeeded"
6770 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -l &>/dev/null
6771 if [ $? -eq 0 ]; then
6772 rm $LDEVCONFPATH $NIDSPATH
6773 error "ldev label filtering w/ -a and -l incorrectly succeeded"
6777 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f -l &>/dev/null
6778 if [ $? -eq 0 ]; then
6779 rm $LDEVCONFPATH $NIDSPATH
6780 error "ldev label filtering w/ -f and -l incorrectly succeeded"
6784 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -l &>/dev/null
6785 if [ $? -eq 0 ]; then
6786 rm $LDEVCONFPATH $NIDSPATH
6787 error "ldev label filtering w/ -F and -l incorrectly succeeded"
6790 # file sys & foreign
6791 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -f &>/dev/null
6792 if [ $? -eq 0 ]; then
6793 rm $LDEVCONFPATH $NIDSPATH
6794 error "ldev label filtering w/ -F and -f incorrectly succeeded"
6797 rm $LDEVCONFPATH $NIDSPATH
6799 run_test 95 "ldev should only allow one label filter"
6802 if [ -z "$LDEV" ]; then
6803 error "ldev is missing!"
6806 local LDEVCONFPATH=$TMP/ldev.conf
6807 local NIDSPATH=$TMP/nids
6809 generate_ldev_conf $LDEVCONFPATH
6810 generate_nids $NIDSPATH
6812 local LDEV_OUTPUT=$TMP/ldev-output.txt
6813 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -H $mgs_HOST \
6815 awk '{print $2}' > $LDEV_OUTPUT
6817 # ldev failed, error
6818 if [ $? -ne 0 ]; then
6819 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
6820 error "ldev failed to execute!"
6824 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
6826 echo "$mgs_HOST-$(facet_fstype mgs)" > $EXPECTED_OUTPUT
6828 if [ "$mgs_HOST" == "$mds_HOST" ]; then
6829 for num in $(seq $MDSCOUNT); do
6830 echo "$mds_HOST-$(facet_fstype mds$num)" \
6835 if [ "$mgs_HOST" == "$ost_HOST" ]; then
6836 for num in $(seq $OSTCOUNT); do
6837 echo "$ost_HOST-$(facet_fstype ost$num)" \
6842 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6844 if [ $? -ne 0 ]; then
6845 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6846 error "ldev failed to produce the correct output!"
6849 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6851 run_test 96 "ldev returns hostname and backend fs correctly in command sub"
6854 if [ -z "$LDEV" ]; then
6855 error "ldev is missing!"
6858 local LDEVCONFPATH=$TMP/ldev.conf
6859 local NIDSPATH=$TMP/nids
6861 generate_ldev_conf $LDEVCONFPATH
6862 generate_nids $NIDSPATH
6864 local LDEV_OUTPUT=$TMP/ldev-output.txt
6865 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
6867 echo -e "\nMDT role"
6868 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mdt > $LDEV_OUTPUT
6870 if [ $? -ne 0 ]; then
6871 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
6872 error "ldev failed to execute for mdt role!"
6875 for num in $(seq $MDSCOUNT); do
6876 printf "%s-MDT%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
6879 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6881 if [ $? -ne 0 ]; then
6882 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6883 error "ldev failed to produce the correct output for mdt role!"
6886 echo -e "\nOST role"
6887 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R ost > $LDEV_OUTPUT
6889 if [ $? -ne 0 ]; then
6890 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
6891 error "ldev failed to execute for ost role!"
6895 for num in $(seq $OSTCOUNT); do
6896 printf "%s-OST%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
6899 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6901 if [ $? -ne 0 ]; then
6902 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6903 error "ldev failed to produce the correct output for ost role!"
6906 echo -e "\nMGS role"
6907 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mgs > $LDEV_OUTPUT
6909 if [ $? -ne 0 ]; then
6910 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
6911 error "ldev failed to execute for mgs role!"
6914 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
6916 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6918 if [ $? -ne 0 ]; then
6919 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6920 error "ldev failed to produce the correct output for mgs role!"
6923 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6925 run_test 97 "ldev returns correct ouput when querying based on role"
6930 local temp=$MDS_MOUNT_OPTS
6933 check_mount || error "mount failed"
6934 mountopt="user_xattr"
6935 for ((x = 1; x <= 400; x++)); do
6936 mountopt="$mountopt,user_xattr"
6938 remount_client $mountopt $MOUNT 2>&1 | grep "too long" ||
6939 error "Buffer overflow check failed"
6940 cleanup || error "cleanup failed"
6942 run_test 98 "Buffer-overflow check while parsing mount_opts"
6946 [[ $(facet_fstype ost1) != ldiskfs ]] &&
6947 { skip "Only applicable to ldiskfs-based OSTs" && return; }
6948 [[ $(lustre_version_code ost1) -ge $(version_code 2.8.57) ]] ||
6949 { skip "Need OST version at least 2.8.57" && return 0; }
6951 local ost_opts="$(mkfs_opts ost1 $(ostdevname 1)) \
6952 --reformat $(ostdevname 1) $(ostvdevname 1)"
6953 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" &&
6954 skip "meta_bg already set" && return
6957 if [[ ${!opts} != *mkfsoptions* ]]; then
6958 eval opts=\"${!opts} \
6959 --mkfsoptions='\\\"-O ^resize_inode,meta_bg\\\"'\"
6961 local val=${!opts//--mkfsoptions=\\\"/ \
6962 --mkfsoptions=\\\"-O ^resize_inode,meta_bg }
6966 echo "params: $opts"
6968 add ost1 $opts || error "add ost1 failed with new params"
6970 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" ||
6971 error "meta_bg is not set"
6975 run_test 99 "Adding meta_bg option"
6979 start_mds || error "MDS start failed"
6980 start_ost || error "unable to start OST"
6981 mount_client $MOUNT || error "client start failed"
6982 check_mount || error "check_mount failed"
6991 do_facet mgs 'lshowmount -v' | awk 'BEGIN {NR == 0; rc=1} /MGS:/ {rc=0}
6992 END {exit rc}' || error "lshowmount have no output MGS"
6994 do_facet mds1 'lshowmount -v' | awk 'BEGIN {NR == 2; rc=1} /-MDT0000:/
6995 {rc=0} END {exit rc}' || error "lshowmount have no output MDT0"
6997 do_facet ost1 'lshowmount -v' | awk 'BEGIN {NR == 4; rc=1} /-OST0000:/
6998 {rc=0} END {exit rc}' || error "lshowmount have no output OST0"
7000 cleanup || error "cleanup failed with $?"
7002 run_test 100 "check lshowmount lists MGS, MDT, OST and 0@lo"
7005 local createmany_oid
7006 local dev=$FSNAME-OST0000-osc-MDT0000
7009 createmany -o $DIR1/$tfile-%d 50000 &
7011 # MDT->OST reconnection causes MDT<->OST last_id synchornisation
7012 # via osp_precreate_cleanup_orphans.
7013 for ((i = 0; i < 100; i++)); do
7014 for ((k = 0; k < 10; k++)); do
7015 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;" \
7016 "$LCTL --device $dev activate"
7019 ls -asl $MOUNT | grep '???' &&
7020 (kill -9 $createmany_oid &>/dev/null; \
7021 error "File hasn't object on OST")
7023 kill -s 0 $createmany_oid || break
7025 wait $createmany_oid
7028 run_test 101 "Race MDT->OST reconnection with create"
7030 if ! combined_mgs_mds ; then
7036 # restore the values of MDSSIZE and OSTSIZE
7037 MDSSIZE=$STORED_MDSSIZE
7038 OSTSIZE=$STORED_OSTSIZE