7 # bug number for skipped test:
8 ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT"
9 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
11 # bug number for skipped test:
12 # a tool to create lustre filesystem images
13 ALWAYS_EXCEPT="32newtarball $ALWAYS_EXCEPT"
15 # bug number for skipped tests: LU-9795 (all below)
16 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 0 31 32a 32d 35a"
17 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 53a 53b 54b 76a 76b"
18 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 76c 76d 78 103"
22 PATH=$PWD/$SRCDIR:$SRCDIR:$SRCDIR/../utils:$PATH
24 PTLDEBUG=${PTLDEBUG:--1}
26 LUSTRE=${LUSTRE:-$(dirname $0)/..}
27 RLUSTRE=${RLUSTRE:-$LUSTRE}
28 export MULTIOP=${MULTIOP:-multiop}
30 . $LUSTRE/tests/test-framework.sh
32 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
34 # use small MDS + OST size to speed formatting time
35 # do not use too small MDSSIZE/OSTSIZE, which affect the default journal size
36 # STORED_MDSSIZE is used in test_18
37 STORED_MDSSIZE=$MDSSIZE
38 STORED_OSTSIZE=$OSTSIZE
40 [ $(facet_fstype $SINGLEMDS) = "zfs" ] && MDSSIZE=400000
42 [ $(facet_fstype ost1) = "zfs" ] && OSTSIZE=400000
52 if ! combined_mgs_mds; then
53 # bug number for skipped test: LU-9860 LU-9860 LU-9860
54 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 43b 53b 54b"
55 # bug number for skipped test: LU-9875 LU-9879 LU-9879 LU-9879 LU-9879
56 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 70e 80 84 87 100"
57 # bug number for skipped test: LU-8110 LU-9879 LU-9879 LU-9879
58 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 102 104 105 107"
61 # pass "-E lazy_itable_init" to mke2fs to speed up the formatting time
62 if [[ "$LDISKFS_MKFS_OPTS" != *lazy_itable_init* ]]; then
63 LDISKFS_MKFS_OPTS=$(csa_add "$LDISKFS_MKFS_OPTS" -E lazy_itable_init)
66 [ $(facet_fstype $SINGLEMDS) = "zfs" ] &&
67 # bug number for skipped test:
68 ALWAYS_EXCEPT="$ALWAYS_EXCEPT"
69 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
71 [ $MDSCOUNT -ge 2 ] &&
72 # bug number for skipped test: LU-11915
74 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
79 require_dsh_mds || exit 0
80 require_dsh_ost || exit 0
83 [ "$SLOW" = "no" ] && EXCEPT_SLOW="45 69 106 111"
88 # The MGS must be started before the OSTs for a new fs, so start
89 # and stop to generate the startup logs.
92 wait_osc_import_state mds ost FULL
97 reformat_and_config() {
99 if ! combined_mgs_mds ; then
105 writeconf_or_reformat() {
106 # There are at most 2 OSTs for write_conf test
107 # who knows if/where $TUNEFS is installed?
108 # Better reformat if it fails...
109 writeconf_all $MDSCOUNT 2 ||
110 { echo "tunefs failed, reformatting instead" &&
111 reformat_and_config && return 0; }
120 echo "start mgs service on $(facet_active_host mgs)"
121 start mgs $(mgsdevname) $MGS_MOUNT_OPTS $@
127 local dev=$(mdsdevname $num)
130 echo "start mds service on `facet_active_host $facet`"
131 start $facet ${dev} $MDS_MOUNT_OPTS $@ || return 94
137 local dev=$(mdsdevname $num)
140 echo "stop mds service on `facet_active_host $facet`"
141 # These tests all use non-failover stop
142 stop $facet -f || return 97
148 for num in $(seq $MDSCOUNT); do
149 start_mdt $num $@ || return 94
154 if ! combined_mgs_mds ; then
162 for num in $(seq $MDSCOUNT); do
163 stop_mdt $num || return 97
168 echo "stop mgs service on `facet_active_host mgs`"
169 # These tests all use non-failover stop
170 stop mgs -f || return 97
174 echo "start ost1 service on `facet_active_host ost1`"
175 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS $@ || return 95
179 echo "stop ost1 service on `facet_active_host ost1`"
180 # These tests all use non-failover stop
181 stop ost1 -f || return 98
185 echo "start ost2 service on `facet_active_host ost2`"
186 start ost2 $(ostdevname 2) $OST_MOUNT_OPTS $@ || return 92
190 echo "stop ost2 service on `facet_active_host ost2`"
191 # These tests all use non-failover stop
192 stop ost2 -f || return 93
197 echo "mount $FSNAME on ${MOUNTPATH}....."
198 zconf_mount $(hostname) $MOUNTPATH || return 96
202 local mountopt="remount,$1"
204 echo "remount '$1' lustre on ${MOUNTPATH}....."
205 zconf_mount $(hostname) $MOUNTPATH "$mountopt" || return 96
211 echo "umount lustre on $mountpath....."
212 zconf_umount $HOSTNAME $mountpath $@ || return 97
215 manual_umount_client(){
218 echo "manual umount lustre on ${MOUNT}...."
219 do_facet client "umount ${FORCE} $MOUNT"
225 start_mds || error "MDT start failed"
226 start_ost || error "Unable to start OST1"
227 mount_client $MOUNT || error "client start failed"
228 client_up || error "client_up failed"
237 unload_modules_conf () {
238 if combined_mgs_mds || ! local_mode; then
239 unload_modules || return 1
244 stop_ost || return 202
245 stop_mds || return 201
246 unload_modules_conf || return 203
251 [ "x$1" != "x" ] && force='-f'
252 umount_client $MOUNT $force|| return 200
253 cleanup_nocli || return $?
258 echo "umount $MOUNT2 ..."
259 umount $MOUNT2 || true
260 echo "stopping fs2mds ..."
261 stop fs2mds -f || true
262 echo "stopping fs2ost ..."
263 stop fs2ost -f || true
267 do_facet client "cp /etc/passwd $DIR/a" || return 71
268 do_facet client "rm $DIR/a" || return 72
269 # make sure lustre is actually mounted (touch will block,
270 # but grep won't, so do it after)
271 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
272 echo "setup single mount lustre success"
276 do_facet client "touch $DIR/a" || return 71
277 do_facet client "rm $DIR/a" || return 72
278 do_facet client "touch $DIR2/a" || return 73
279 do_facet client "rm $DIR2/a" || return 74
280 echo "setup double mount lustre success"
284 cat /dev/urandom | tr -dc 'a-zA-Z0-9' | fold -w $1 | head -n 1
289 if [ "$ONLY" == "setup" ]; then
294 if [ "$ONLY" == "cleanup" ]; then
301 #create single point mountpoint
307 check_mount || error "check_mount failed"
308 cleanup || error "cleanup failed with $?"
310 run_test 0 "single mount setup"
313 start_mds || error "MDS start failed"
314 start_ost || error "unable to start OST"
315 echo "start ost second time..."
316 start_ost && error "2nd OST start should fail"
317 mount_client $MOUNT || error "client start failed"
318 check_mount || error "check_mount failed"
319 cleanup || error "cleanup failed with $?"
321 run_test 1 "start up ost twice (should return errors)"
324 start_mds || error "MDT start failed"
325 echo "start mds second time.."
326 start_mds && error "2nd MDT start should fail"
327 start_ost || error "OST start failed"
328 mount_client $MOUNT || error "mount_client failed to start client"
329 check_mount || error "check_mount failed"
330 cleanup || error "cleanup failed with $?"
332 run_test 2 "start up mds twice (should return err)"
336 #mount.lustre returns an error if already in mtab
337 mount_client $MOUNT && error "2nd client mount should fail"
338 check_mount || error "check_mount failed"
339 cleanup || error "cleanup failed with $?"
341 run_test 3 "mount client twice (should return err)"
345 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
346 stop_ost || error "Unable to stop OST1"
347 umount_client $MOUNT -f || error “unmount $MOUNT failed”
350 # ok for ost to fail shutdown
351 if [ 202 -ne $eno ] && [ 0 -ne $eno ]; then
352 error "cleanup failed with $?"
355 run_test 4 "force cleanup ost, then cleanup"
357 test_5a() { # was test_5
359 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
360 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
362 stop_mds || error "Unable to stop MDS"
364 # cleanup may return an error from the failed
365 # disconnects; for now I'll consider this successful
366 # if all the modules have unloaded.
370 echo "killing umount"
371 kill -TERM $UMOUNT_PID
372 echo "waiting for umount to finish"
374 if grep " $MOUNT " /proc/mounts; then
375 echo "test 5: /proc/mounts after failed umount"
379 echo "killing umount"
380 kill -TERM $UMOUNT_PID
381 echo "waiting for umount to finish"
383 grep " $MOUNT " /proc/mounts &&
384 error "/proc/mounts after second umount"
387 # manual_mount_client may fail due to umount succeeding above
389 # stop_mds is a no-op here, and should not fail
390 cleanup_nocli || error "cleanup_nocli failed with $?"
391 # df may have lingering entry
393 # mtab may have lingering entry
397 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
399 grep -q $MOUNT" " /etc/mtab || break
400 echo "Waiting /etc/mtab updated ... "
401 WAIT=$(( WAIT + sleep))
403 [ "$WAIT" -eq "$MAX_WAIT" ] &&
404 error "/etc/mtab is not updated in $WAIT secs"
405 echo "/etc/mtab updated in $WAIT secs"
407 run_test 5a "force cleanup mds, then cleanup"
415 grep " $MOUNT " /etc/mtab &&
416 error false "unexpected entry in mtab before mount" && return 10
418 start_ost || error "OST start failed"
419 if ! combined_mgs_mds ; then
420 trap cleanup_5b EXIT ERR
421 start_mds || error "MDS start failed"
425 mount_client $MOUNT && error "mount_client $MOUNT should fail"
426 grep " $MOUNT " /etc/mtab &&
427 error "$MOUNT entry in mtab after failed mount"
429 # stop_mds is a no-op here, and should not fail
430 cleanup_nocli || error "cleanup_nocli failed with $?"
431 if ! combined_mgs_mds ; then
435 run_test 5b "Try to start a client with no MGS (should return errs)"
438 grep " $MOUNT " /etc/mtab &&
439 error false "unexpected entry in mtab before mount" && return 10
441 start_mds || error "MDS start failed"
442 start_ost || error "OST start failed"
443 local oldfs="${FSNAME}"
444 FSNAME="wrong.${FSNAME}"
445 mount_client $MOUNT || :
447 grep " $MOUNT " /etc/mtab &&
448 error "$MOUNT entry in mtab after failed mount"
450 cleanup_nocli || error "cleanup_nocli failed with $?"
452 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
455 grep " $MOUNT " /etc/mtab &&
456 error "unexpected entry in mtab before mount"
458 start_ost || error "OST start failed"
459 start_mds || error "MDS start failed"
460 stop_ost -f || error "Unable to stop OST1"
461 mount_client $MOUNT || error "mount_client $MOUNT failed"
462 umount_client $MOUNT -f || error "umount_client $MOUNT failed"
463 cleanup_nocli || error "cleanup_nocli failed with $?"
464 ! grep " $MOUNT " /etc/mtab ||
465 error "$MOUNT entry in mtab after unmount"
467 run_test 5d "mount with ost down"
470 grep " $MOUNT " /etc/mtab &&
471 error false "unexpected entry in mtab before mount" && return 10
473 start_mds || error "MDS start failed"
474 start_ost || error "OST start failed"
476 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
477 do_facet client "$LCTL set_param fail_loc=0x80000506"
478 mount_client $MOUNT || echo "mount failed (not fatal)"
479 cleanup || error "cleanup failed with $?"
480 grep " $MOUNT " /etc/mtab &&
481 error "$MOUNT entry in mtab after unmount"
484 run_test 5e "delayed connect, don't crash (bug 10268)"
487 combined_mgs_mds && skip "needs separate mgs and mds"
489 grep " $MOUNT " /etc/mtab &&
490 error false "unexpected entry in mtab before mount" && return 10
493 start_ost || error "OST start failed"
494 mount_client $MOUNT &
496 echo client_mount pid is $pid
500 if ! ps -f -p $pid >/dev/null; then
503 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
504 error "mount returns $rc, expected to hang"
506 cleanup || error "cleanup failed with $?"
511 start_mds || error "start MDS failed"
513 # mount should succeed after start mds
515 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
516 cleanup || error "final call to cleanup failed with rc $?"
518 run_test 5f "mds down, cleanup after failed mount (bug 2712)"
522 [ $(lustre_version_code client) -lt $(version_code 2.9.53) ] &&
523 skip "automount of debugfs missing before 2.9.53"
524 umount /sys/kernel/debug
525 $LCTL get_param -n devices | egrep -v "error" && \
526 error "lctl can't access debugfs data"
527 grep " debugfs " /etc/mtab || error "debugfs failed to remount"
529 run_test 5g "handle missing debugfs"
534 mount_client $MOUNT || error "mount_client $MOUNT failed"
535 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
536 cleanup || error "cleanup failed with rc $?"
538 run_test 6 "manual umount, then mount again"
543 cleanup_nocli || error "cleanup_nocli failed with $?"
545 run_test 7 "manual umount, then cleanup"
549 mount_client $MOUNT2 || error "mount_client $MOUNT2 failed"
550 check_mount2 || error "check_mount2 failed"
551 umount_client $MOUNT2 || error "umount_client $MOUNT2 failed"
552 cleanup || error "cleanup failed with rc $?"
554 run_test 8 "double mount setup"
557 start_ost || error "OST start failed"
559 do_facet ost1 $LCTL set_param debug=\'inode trace\' ||
560 error "do_facet ost1 set_param inode trace failed."
561 do_facet ost1 $LCTL set_param subsystem_debug=\'mds ost\' ||
562 error "do_facet ost1 set_param debug mds ost failed."
564 CHECK_PTLDEBUG="`do_facet ost1 $LCTL get_param -n debug`"
565 if [ "$CHECK_PTLDEBUG" ] && { \
566 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
567 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
568 echo "lnet.debug success"
570 error "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
572 CHECK_SUBSYS="`do_facet ost1 $LCTL get_param -n subsystem_debug`"
573 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
574 echo "lnet.subsystem_debug success"
576 error "lnet.subsystem_debug: want 'mds ost' got '$CHECK_SUBSYS'"
578 stop_ost || error "Unable to stop OST1"
580 run_test 9 "test ptldebug and subsystem for mkfs"
583 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
584 # But with new MDS stack we don't care about the mode of local objects
585 # anymore, so this test is removed. See bug 22944 for more details.
589 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
590 skip "ldiskfs only test"
594 check_mount || error "check_mount failed"
595 cleanup || error "cleanup failed with rc $?"
597 echo "Remove mds config log"
598 if ! combined_mgs_mds ; then
602 do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' \
603 $(mgsdevname) || return \$?" ||
604 error "do_facet mgs failed with $?"
606 if ! combined_mgs_mds ; then
610 start_ost || error "OST start failed"
611 start_mds && error "MDS start succeeded, but should fail"
614 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
617 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
618 skip "ldiskfs only test"
621 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
626 # check if current MDSSIZE is large enough
627 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE &&
628 log "use MDSSIZE=$MDSSIZE"
630 # check if the global config has a large enough MDSSIZE
631 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] &&
632 [ $STORED_MDSSIZE -ge $MIN ] &&
633 OK=1 && myMDSSIZE=$STORED_MDSSIZE &&
634 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
636 # check if the block device is large enough
637 is_blkdev $SINGLEMDS $MDSDEV $MIN
638 local large_enough=$?
639 if [ -n "$OK" ]; then
640 [ $large_enough -ne 0 ] && OK=""
642 [ $large_enough -eq 0 ] && OK=1 && myMDSSIZE=$MIN &&
643 log "use device $MDSDEV with MIN=$MIN"
646 # check if a loopback device has enough space for fs metadata (5%)
648 if [ -z "$OK" ]; then
649 local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! \
650 -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
651 awk '($1 != "Filesystem") { print $4 }')
652 ! [ -z "$SPACE" ] && [ $SPACE -gt $((MIN / 20)) ] &&
653 OK=1 && myMDSSIZE=$MIN &&
654 log "use file $MDSDEV with MIN=$MIN"
657 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS"
659 echo "mount mds with large journal..."
661 local OLD_MDSSIZE=$MDSSIZE
665 echo "mount lustre system..."
667 check_mount || error "check_mount failed"
669 echo "check journal size..."
670 local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
671 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
672 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
674 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
677 cleanup || error "cleanup failed with rc $?"
682 run_test 18 "check mkfs creates large journals"
685 start_mds || error "MDS start failed"
686 stop_mds || error "Unable to stop MDS"
688 run_test 19a "start/stop MDS without OSTs"
691 start_ost || error "Unable to start OST1"
692 stop_ost -f || error "Unable to stop OST1"
694 run_test 19b "start/stop OSTs without MDS"
697 # first format the ost/mdt
698 start_mds || error "MDS start failed"
699 start_ost || error "Unable to start OST1"
700 mount_client $MOUNT || error "mount_client $MOUNT failed"
701 check_mount || error "check_mount failed"
702 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed."
703 remount_client ro $MOUNT || error "remount_client with ro failed"
704 touch $DIR/$tfile && error "$DIR/$tfile created incorrectly"
705 [ -e $DIR/$tfile ] && error "$DIR/$tfile exists incorrectly"
706 remount_client rw $MOUNT || error "remount_client with rw failed"
707 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
708 MCNT=$(grep -c $MOUNT' ' /etc/mtab)
709 [ "$MCNT" -ne 1 ] && error "$MOUNT in /etc/mtab $MCNT times"
711 stop_mds || error "Unable to stop MDS"
712 stop_ost || error "Unable to stop OST1"
714 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
717 start_mds || error "MDS start failed"
718 start_ost || error "unable to start OST1"
719 wait_osc_import_state mds ost FULL
720 stop_ost || error "unable to stop OST1"
721 stop_mds || error "unable to stop MDS"
723 run_test 21a "start mds before ost, stop ost first"
726 start_ost || error "unable to start OST1"
727 start_mds || error "MDS start failed"
728 wait_osc_import_state mds ost FULL
729 stop_mds || error "unable to stop MDS"
730 stop_ost || error "unable to stop OST1"
732 run_test 21b "start ost before mds, stop mds first"
735 start_ost || error "Unable to start OST1"
736 start_mds || error "MDS start failed"
737 start_ost2 || error "Unable to start OST2"
738 wait_osc_import_state mds ost2 FULL
739 stop_ost || error "Unable to stop OST1"
740 stop_ost2 || error "Unable to stop OST2"
741 stop_mds || error "Unable to stop MDS"
742 #writeconf to remove all ost2 traces for subsequent tests
743 writeconf_or_reformat
745 run_test 21c "start mds between two osts, stop mds last"
748 combined_mgs_mds && skip "need separate mgs device"
753 start_mgs || error "unable to start MGS"
754 start_ost || error "unable to start OST1"
755 start_ost2 || error "unable to start OST2"
756 start_mds || error "MDS start failed"
757 wait_osc_import_state mds ost2 FULL
759 stop_ost || error "Unable to stop OST1"
760 stop_ost2 || error "Unable to stop OST2"
761 stop_mds || error "Unable to stop MDS"
763 #writeconf to remove all ost2 traces for subsequent tests
764 writeconf_or_reformat
765 start_mgs || error "unable to start MGS"
767 run_test 21d "start mgs then ost and then mds"
770 MGSNID="$saved_mgsnid"
772 echo "stopping fs2mgs ..."
773 stop $fs2mgs -f || true
776 test_21e() { # LU-5863
777 if [[ -z "$fs3ost_DEV" || -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
778 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
779 skip_env "mixed loopback and real device not working"
782 local fs2mdsdev=$(mdsdevname 1_2)
783 local fs2ostdev=$(ostdevname 1_2)
784 local fs3ostdev=$(ostdevname 2_2)
786 local fs2mdsvdev=$(mdsvdevname 1_2)
787 local fs2ostvdev=$(ostvdevname 1_2)
788 local fs3ostvdev=$(ostvdevname 2_2)
790 # temporarily use fs3ost as fs2mgs
792 local fs2mgsdev=$fs3ostdev
793 local fs2mgsvdev=$fs3ostvdev
795 local fsname=test1234
797 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
798 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
799 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS && trap cleanup_21e EXIT INT ||
800 error "start fs2mgs failed"
802 local saved_mgsnid="$MGSNID"
803 MGSNID=$(do_facet $fs2mgs $LCTL list_nids | xargs | tr ' ' ,)
805 add fs2mds $(mkfs_opts mds1 $fs2mdsdev $fsname) \
806 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
807 add fs2ost $(mkfs_opts ost1 $fs2ostdev $fsname) \
808 --reformat $fs2ostdev $fs2ostvdev || error "add fs2ost failed"
810 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
811 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS || error "start fs2mds failed"
813 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
814 $MOUNT_CMD $MGSNID:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
815 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
819 run_test 21e "separate MGS and MDS"
822 start_mds || error "MDS start failed"
824 echo "Client mount with ost in logs, but none running"
825 start_ost || error "unable to start OST1"
826 # wait until mds connected to ost and open client connection
827 wait_osc_import_state mds ost FULL
828 stop_ost || error "unable to stop OST1"
829 mount_client $MOUNT || error "mount_client $MOUNT failed"
830 # check_mount will block trying to contact ost
831 mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed"
832 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed"
833 umount_client $MOUNT -f
836 echo "Client mount with a running ost"
837 start_ost || error "unable to start OST1"
839 # if gss enabled, wait full time to let connection from
840 # mds to ost be established, due to the mismatch between
841 # initial connect timeout and gss context negotiation timeout.
842 # This perhaps could be remove after AT landed.
843 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
844 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
846 mount_client $MOUNT || error "mount_client $MOUNT failed"
847 wait_osc_import_state mds ost FULL
848 wait_osc_import_ready client ost
849 check_mount || error "check_mount failed"
852 cleanup || error "cleanup failed with rc $?"
854 run_test 22 "start a client before osts (should return errs)"
856 test_23a() { # was test_23
859 stop $SINGLEMDS || error "failed to stop $SINGLEMDS"
860 # force down client so that recovering mds waits for reconnect
861 local running=$(grep -c $MOUNT /proc/mounts) || true
862 if [ $running -ne 0 ]; then
863 echo "Stopping client $MOUNT (opts: -f)"
867 # enter recovery on failed mds
868 local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/})
869 start $SINGLEMDS $MDT_DEV $MDS_MOUNT_OPTS || error "MDS start failed"
870 # try to start a new client
871 mount_client $MOUNT &
873 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
874 MOUNT_LUSTRE_PID=$(ps -ef | grep mount.lustre |
875 grep -v grep | awk '{print $2}')
876 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
878 ps --ppid $MOUNT_LUSTRE_PID
879 echo "waiting for mount to finish"
881 # "ctrl-c" sends SIGINT but it usually (in script) does not work on child process
882 # SIGTERM works but it does not spread to offspring processses
883 kill -s TERM $MOUNT_PID
884 kill -s TERM $MOUNT_LUSTRE_PID
885 # we can not wait $MOUNT_PID because it is not a child of this shell
891 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
893 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
894 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
897 [ -z "$PID1" -a -z "$PID2" ] && break
898 echo "waiting for mount to finish ... "
899 WAIT=$(( WAIT + sleep))
901 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
902 error "MOUNT_PID $MOUNT_PID and "\
903 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
906 cleanup || error "cleanup failed with rc $?"
908 run_test 23a "interrupt client during recovery mount delay"
910 test_23b() { # was test_23
911 start_mds || error "MDS start failed"
912 start_ost || error "Unable to start OST1"
913 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
914 $LCTL set_param fail_loc=0x80000313
916 cleanup || error "cleanup failed with rc $?"
918 run_test 23b "Simulate -EINTR during mount"
921 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
923 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
924 is_blkdev $SINGLEMDS $MDSDEV &&
925 skip_env "mixed loopback and real device not working"
928 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
930 local fs2mdsdev=$(mdsdevname 1_2)
931 local fs2ostdev=$(ostdevname 1_2)
932 local fs2mdsvdev=$(mdsvdevname 1_2)
933 local fs2ostvdev=$(ostvdevname 1_2)
936 # LU-9733 test fsname started with numbers as well
937 local FSNAME2=969362ae
939 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --nomgs --mgsnode=$MGSNID \
940 --fsname=${FSNAME2} --reformat $fs2mdsdev $fs2mdsvdev || exit 10
942 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --fsname=${FSNAME2} \
943 --reformat $fs2ostdev $fs2ostvdev || exit 10
946 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
947 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
948 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
949 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
951 # LU-9733 test fsname started with numbers
952 cl_user=$(do_facet $SINGLEMDS lctl --device $FSNAME2-MDT0000 \
953 changelog_register -n) ||
954 error "register changelog failed"
956 do_facet $SINGLEMDS lctl --device $FSNAME2-MDT0000 \
957 changelog_deregister $cl_user ||
958 error "deregister changelog failed"
960 check_mount || error "check_mount failed"
961 # files written on 1 should not show up on 2
962 cp /etc/passwd $DIR/$tfile
964 [ -e $MOUNT2/$tfile ] && error "File bleed"
967 cp /etc/passwd $MOUNT2/$tfile ||
968 error "cp /etc/passwd $MOUNT2/$tfile failed"
969 rm $MOUNT2/$tfile || error "remove $MOUNT2/$tfile failed"
970 # 2 is actually mounted
971 grep $MOUNT2' ' /proc/mounts > /dev/null || error "$MOUNT2 not mounted"
973 facet_failover fs2mds
974 facet_failover fs2ost
977 # the MDS must remain up until last MDT
979 MDS=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
980 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
981 [ -z "$MDS" ] && error "No MDT"
983 cleanup_nocli || error "cleanup_nocli failed with rc $?"
985 run_test 24a "Multiple MDTs on a single node"
988 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
990 if [ -z "$fs2mds_DEV" ]; then
991 local dev=${SINGLEMDS}_dev
993 is_blkdev $SINGLEMDS $MDSDEV &&
994 skip_env "mixed loopback and real device not working"
997 local fs2mdsdev=$(mdsdevname 1_2)
998 local fs2mdsvdev=$(mdsvdevname 1_2)
1000 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --mgs --fsname=${FSNAME}2 \
1001 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
1003 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS &&
1004 error "start MDS should fail"
1006 cleanup || error "cleanup failed with rc $?"
1008 run_test 24b "Multiple MGSs on a single node (should return err)"
1012 check_mount || error "check_mount failed"
1013 local MODULES=$($LCTL modules | awk '{ print $2 }')
1014 rmmod $MODULES 2>/dev/null || true
1015 cleanup || error "cleanup failed with $?"
1017 run_test 25 "Verify modules are referenced"
1021 # we need modules before mount for sysctl, so make sure...
1022 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
1023 #define OBD_FAIL_MDS_FS_SETUP 0x135
1024 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000135"
1025 start_mds && error "MDS started but should not have started"
1026 $LCTL get_param -n devices
1027 DEVS=$($LCTL get_param -n devices | egrep -v MG | wc -l)
1028 [ $DEVS -gt 0 ] && error "number of devices is $DEVS, should be zero"
1029 # start mds to drop writeconf setting
1030 start_mds || error "Unable to start MDS"
1031 stop_mds || error "Unable to stop MDS"
1032 unload_modules_conf || error "unload_modules_conf failed with $?"
1034 run_test 26 "MDT startup failure cleans LOV (should return errs)"
1037 start_ost || error "Unable to start OST1"
1038 start_mds || error "Unable to start MDS"
1039 echo "Requeue thread should have started: "
1040 ps -e | grep ll_cfg_requeue
1041 set_persistent_param_and_check ost1 \
1042 "obdfilter.$FSNAME-OST0000.client_cache_seconds" \
1043 "$FSNAME-OST0000.ost.client_cache_seconds"
1044 cleanup_nocli || error "cleanup_nocli failed with rc $?"
1046 run_test 27a "Reacquire MGS lock if OST started first"
1051 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
1052 awk '($3 ~ "mdt" && $4 ~ "MDT0000") { print $4 }')
1054 facet_failover $SINGLEMDS
1055 set_persistent_param_and_check $SINGLEMDS \
1056 "mdt.$device.identity_acquire_expire" \
1057 "$device.mdt.identity_acquire_expire"
1058 set_persistent_param_and_check client \
1059 "mdc.$device-mdc-*.max_rpcs_in_flight" \
1060 "$device.mdc.max_rpcs_in_flight"
1062 cleanup || error "cleanup failed with $?"
1064 run_test 27b "Reacquire MGS lock after failover"
1066 test_28A() { # was test_28
1068 TEST="llite.$FSNAME-*.max_read_ahead_whole_mb"
1069 PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
1070 ORIG=$($LCTL get_param -n $TEST)
1071 FINAL=$(($ORIG + 1))
1072 set_persistent_param_and_check client "$TEST" "$PARAM" $FINAL
1073 FINAL=$(($FINAL + 1))
1074 set_persistent_param_and_check client "$TEST" "$PARAM" $FINAL
1075 umount_client $MOUNT || error "umount_client $MOUNT failed"
1076 mount_client $MOUNT || error "mount_client $MOUNT failed"
1077 RESULT=$($LCTL get_param -n $TEST)
1078 if [ $RESULT -ne $FINAL ]; then
1079 error "New config not seen: wanted $FINAL got $RESULT"
1081 echo "New config success: got $RESULT"
1083 set_persistent_param_and_check client "$TEST" "$PARAM" $ORIG
1084 cleanup || error "cleanup failed with rc $?"
1086 run_test 28A "permanent parameter setting"
1088 test_28a() { # LU-4221
1089 [[ $(lustre_version_code ost1) -ge $(version_code 2.5.52) ]] ||
1090 skip "Need OST version at least 2.5.52"
1091 [ "$(facet_fstype ost1)" = "zfs" ] &&
1092 skip "LU-4221: no such proc params for ZFS OSTs"
1099 local device="$FSNAME-OST0000"
1103 # In this test we will set three kinds of proc parameters with
1104 # lctl set_param -P or lctl conf_param:
1105 # 1. non-symlink ones in the OFD
1106 # 2. non-symlink ones in the OSD
1109 # prepare a non-symlink parameter in the OFD
1110 name="client_cache_seconds"
1111 param="$device.ost.$name"
1112 cmd="obdfilter.$device.$name"
1114 # permanently setting the parameter in the OFD
1115 old=$(do_facet ost1 $LCTL get_param -n $cmd)
1117 set_persistent_param_and_check ost1 "$cmd" "$param" $new
1118 set_persistent_param_and_check ost1 "$cmd" "$param" $old
1121 # prepare a non-symlink parameter in the OSD
1123 param="$device.osd.$name"
1124 cmd="osd-*.$device.$name"
1126 # conf_param the parameter in the OSD
1127 old=$(do_facet ost1 $LCTL get_param -n $cmd)
1128 new=$(((old + 1) % 2))
1129 set_persistent_param_and_check ost1 "$cmd" "$param" $new
1130 set_persistent_param_and_check ost1 "$cmd" "$param" $old
1132 cleanup || error "cleanup failed with $?"
1134 run_test 28a "set symlink parameters permanently with lctl"
1137 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >= 2 OSTs"
1138 setup > /dev/null 2>&1
1139 start_ost2 || error "Unable to start OST2"
1142 local PARAM="$FSNAME-OST0001.osc.active"
1143 # With lctl set_param -P the value $PROC_ACT will be sent to
1144 # all nodes. The [!M] filter out the ability to set active
1145 # on the MDS servers which is tested with wait_osp_* below.
1146 # For ost_server_uuid that only exist on client so filtering
1148 local PROC_ACT="osc.$FSNAME-OST0001-osc-*.active"
1149 local PROC_UUID="osc.$FSNAME-OST0001-osc-[!M]*.ost_server_uuid"
1151 ACTV=$($LCTL get_param -n $PROC_ACT)
1153 set_persistent_param_and_check client $PROC_ACT $PARAM $DEAC
1154 # also check ost_server_uuid status
1155 RESULT=$($LCTL get_param -n $PROC_UUID | grep DEACTIV)
1156 if [ -z "$RESULT" ]; then
1157 error "Client not deactivated: $($LCTL get_param \
1160 echo "Live client success: got $RESULT"
1164 wait_osp_active ost ${FSNAME}-OST0001 1 0
1166 # test new client starts deactivated
1167 umount_client $MOUNT || error "umount_client $MOUNT failed"
1168 mount_client $MOUNT || error "mount_client $MOUNT failed"
1170 # the 2nd and 3rd field of ost_server_uuid do not update at the same
1171 # time when using lctl set_param -P
1172 wait_update_facet client \
1173 "$LCTL get_param -n $PROC_UUID | awk '{print \\\$3 }'" \
1175 error "New client start active: $($LCTL get_param -n $PROC_UUID)"
1177 echo "New client success: got '$($LCTL get_param -n $PROC_UUID)'"
1179 # make sure it reactivates
1180 set_persistent_param_and_check client $PROC_ACT $PARAM $ACTV
1182 umount_client $MOUNT
1183 stop_ost2 || error "Unable to stop OST2"
1184 cleanup_nocli || error "cleanup_nocli failed with $?"
1185 #writeconf to remove all ost2 traces for subsequent tests
1186 writeconf_or_reformat
1188 run_test 29 "permanently remove an OST"
1193 echo Big config llog
1194 TEST="llite.$FSNAME-*.max_read_ahead_whole_mb"
1195 ORIG=$($LCTL get_param -n $TEST)
1196 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1197 for i in ${LIST[@]}; do
1198 set_persistent_param_and_check client "$TEST" \
1199 "$FSNAME.llite.max_read_ahead_whole_mb" $i
1201 # make sure client restart still works
1202 umount_client $MOUNT
1203 mount_client $MOUNT || error "mount_client $MOUNT failed"
1204 [ "$($LCTL get_param -n $TEST)" -ne "$i" ] &&
1205 error "Param didn't stick across restart $($TEST) != $i"
1208 echo Erase parameter setting
1209 if [[ $PERM_CMD == *"set_param -P"* ]]; then
1210 do_facet mgs "$PERM_CMD -d $TEST" ||
1211 error "Erase param $TEST failed"
1213 do_facet mgs "$PERM_CMD \
1214 -d $FSNAME.llite.max_read_ahead_whole_mb" ||
1215 error "Erase param $FSNAME.llite.max_read_ahead_whole_mb failed"
1217 umount_client $MOUNT
1218 mount_client $MOUNT || error "mount_client $MOUNT failed"
1219 FINAL=$($LCTL get_param -n $TEST)
1220 echo "deleted (default) value=$FINAL, orig=$ORIG"
1221 # assumes this parameter started at the default value
1222 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
1224 cleanup || error "cleanup failed with rc $?"
1226 run_test 30a "Big config llog and permanent parameter deletion"
1231 local orignids=$($LCTL get_param -n \
1232 osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1234 local orignidcount=$(echo "$orignids" | wc -w)
1236 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1237 # numerical part of it. Hopefully that's not already a failover address
1239 local OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | \
1241 local ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1242 local NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1243 local NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1244 echo "Using fake nid $NEW"
1246 local TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1247 grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1248 if [[ $PERM_CMD == *"set_param -P"* ]]; then
1249 PARAM="osc.$FSNAME-OST0000-osc-[^M]*.import"
1250 echo "Setting $PARAM from $TEST to $NEW"
1251 do_facet mgs "$PERM_CMD $PARAM='connection=$NEW'" ||
1252 error "$PERM_CMD $PARAM failed"
1254 PARAM="$FSNAME-OST0000.failover.node"
1255 echo "Setting $PARAM from $TEST to $NEW"
1256 do_facet mgs "$PERM_CMD $PARAM='$NEW'" ||
1257 error "$PARAM $PARAM failed"
1259 wait_update_facet client "$TEST" "$NEW" ||
1260 error "check $PARAM failed!"
1262 local NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1264 local NIDCOUNT=$(echo "$NIDS" | wc -w)
1265 echo "should have $((orignidcount + 1)) entries \
1266 in failover nids string, have $NIDCOUNT"
1267 [ $NIDCOUNT -eq $((orignidcount + 1)) ] ||
1268 error "Failover nid not added"
1270 if [[ $PERM_CMD == *"set_param -P"* ]]; then
1271 do_facet mgs "$PERM_CMD -d osc.$FSNAME-OST0000-osc-*.import"
1273 do_facet mgs "$PERM_CMD -d $FSNAME-OST0000.failover.node" ||
1274 error "$PERM_CMD delete failed"
1276 umount_client $MOUNT
1277 mount_client $MOUNT || error "mount_client $MOUNT failed"
1279 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1281 NIDCOUNT=$(echo "$NIDS" | wc -w)
1282 echo "only $orignidcount final entries should remain \
1283 in failover nids string, have $NIDCOUNT"
1284 [ $NIDCOUNT -eq $orignidcount ] || error "Failover nids not removed"
1286 cleanup || error "cleanup failed with rc $?"
1288 run_test 30b "Remove failover nids"
1290 test_31() { # bug 10734
1291 # ipaddr must not exist
1292 $MOUNT_CMD 4.3.2.1@tcp:/lustre $MOUNT || true
1293 cleanup || error "cleanup failed with rc $?"
1295 run_test 31 "Connect to non-existent node (shouldn't crash)"
1299 T32_BLIMIT=40960 # Kbytes
1303 # This is not really a test but a tool to create new disk
1304 # image tarballs for the upgrade tests.
1306 # Disk image tarballs should be created on single-node
1307 # clusters by running this test with default configurations
1308 # plus a few mandatory environment settings that are verified
1309 # at the beginning of the test.
1311 test_32newtarball() {
1315 local tmp=$TMP/t32_image_create
1316 local server_version=$(lustre_version_code $SINGLEMDS)
1321 if [ $FSNAME != t32fs -o \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o \
1322 $OSTCOUNT -ne 1 -o -z "$OSTDEV1" ]; then
1323 error "Needs FSNAME=t32fs MDSCOUNT=2 " \
1324 "MDSDEV1=<nonexistent_file>" \
1325 "MDSDEV2=<nonexistent_file>" \
1326 "(or MDSDEV, in the case of b1_8)" \
1327 "OSTCOUNT=1 OSTDEV1=<nonexistent_file>"
1331 echo "Found stale $tmp"
1335 mkdir $tmp/src || return 1
1336 tar cf - -C $src . | tar xf - -C $tmp/src
1337 dd if=/dev/zero of=$tmp/src/t32_qf_old bs=1M \
1338 count=$(($T32_BLIMIT / 1024 / 4))
1339 chown $T32_QID.$T32_QID $tmp/src/t32_qf_old
1341 # format ost with comma-separated NIDs to verify LU-4460
1342 local failnid="$(h2nettype 1.2.3.4),$(h2nettype 4.3.2.1)"
1343 MGSNID="$MGSNID,$MGSNID" OSTOPT="--failnode=$failnid" formatall
1347 [[ $server_version -ge $(version_code 2.3.50) ]] ||
1348 $LFS quotacheck -ug /mnt/$FSNAME
1349 $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
1352 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1354 if [[ $MDSCOUNT -ge 2 ]]; then
1355 remote_dir=/mnt/$FSNAME/remote_dir
1356 $LFS mkdir -i 1 $remote_dir
1357 tar cf - -C $tmp/src . | tar xf - -C $remote_dir
1359 if [[ $server_version -ge $(version_code 2.7.0) ]]; then
1360 striped_dir=/mnt/$FSNAME/striped_dir_old
1361 $LFS mkdir -i 1 -c 2 $striped_dir
1362 tar cf - -C $tmp/src . | tar xf - -C $striped_dir
1368 mkdir $tmp/img || return 1
1372 pushd_dir=/mnt/$FSNAME
1373 if [[ $MDSCOUNT -ge 2 ]]; then
1374 pushd_dir=$remote_dir
1375 if [[ $server_version -ge $(version_code 2.7.0) ]]; then
1377 ls -Rni --time-style=+%s >$tmp/img/list2
1383 ls -Rni --time-style=+%s >$tmp/img/list
1384 find ! -name .lustre -type f -exec sha1sum {} \; |
1385 sort -k 2 >$tmp/img/sha1sums
1387 $LCTL get_param -n version | head -n 1 |
1388 sed -e 's/^lustre: *//' >$tmp/img/commit
1390 [[ $server_version -ge $(version_code 2.3.50) ]] ||
1391 $LFS quotaon -ug /mnt/$FSNAME
1392 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1393 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1394 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1395 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1396 | tr -d "*" > $tmp/img/bspace
1397 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1398 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1399 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1400 | tr -d "*" > $tmp/img/ispace
1401 echo $T32_BLIMIT > $tmp/img/blimit
1402 echo $T32_ILIMIT > $tmp/img/ilimit
1407 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1410 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1411 echo "Data verification failed"
1414 uname -r >$tmp/img/kernel
1415 uname -m >$tmp/img/arch
1417 mv ${MDSDEV1:-$MDSDEV} $tmp/img
1418 for num in $(seq 2 $MDSCOUNT); do
1419 local devname=$(mdsdevname $num)
1421 [[ $(facet_fstype $facet) != zfs ]] ||
1422 devname=$(mdsvdevname $num)
1423 mv $devname $tmp/img
1425 mv $OSTDEV1 $tmp/img
1427 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1428 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1431 tar cjvf $dst/disk$version-$(facet_fstype $SINGLEMDS).tar.bz2 -S *
1436 #run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1439 # The list of applicable tarballs is returned via the caller's
1440 # variable "tarballs".
1443 [ "$CLIENTONLY" ] && skip "Client-only testing"
1445 local node=$(facet_active_host $SINGLEMDS)
1446 local r="do_node $node"
1448 ! $r which "$TUNEFS" && skip_env "tunefs.lustre required on $node"
1450 local IMGTYPE=$(facet_fstype $SINGLEMDS)
1452 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 \
1453 -name \'disk*-$IMGTYPE.tar.bz2\')
1455 [ -z "$tarballs" ] && skip "No applicable tarballs found"
1458 t32_test_cleanup() {
1460 local facet=$SINGLEMDS
1461 local fstype=$(facet_fstype $facet)
1464 if $shall_cleanup_lustre; then
1465 umount $tmp/mnt/lustre || rc=$?
1467 if $shall_cleanup_mdt; then
1468 $r $UMOUNT $tmp/mnt/mdt || rc=$?
1470 if $shall_cleanup_mdt1; then
1471 $r $UMOUNT $tmp/mnt/mdt1 || rc=$?
1473 if $shall_cleanup_ost; then
1474 $r $UMOUNT $tmp/mnt/ost || rc=$?
1479 if [[ $fstype == zfs ]]; then
1481 local poolname_list="t32fs-mdt1 t32fs-ost1"
1483 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1485 for poolname in $poolname_list; do
1486 destroy_zpool $facet $poolname
1489 combined_mgs_mds || start_mgs || rc=$?
1493 t32_bits_per_long() {
1495 # Yes, this is not meant to be perfect.
1505 t32_reload_modules() {
1507 local all_removed=false
1509 local fstype=$(facet_fstype $SINGLEMDS)
1511 [ $fstype == "zfs" ] && do_rpc_nodes $node "service zed stop"
1513 while ((i < 20)); do
1514 echo "Unloading modules on $node: Attempt $i"
1515 do_rpc_nodes $node $LUSTRE_RMMOD $fstype &&
1517 do_rpc_nodes $node check_mem_leak || return 1
1518 if $all_removed; then
1519 do_rpc_nodes $node load_modules
1522 if [ $fstype == "zfs" ]; then
1523 do_rpc_nodes $node "$ZPOOL status -v"
1528 echo "Unloading modules on $node: Given up"
1532 t32_wait_til_devices_gone() {
1538 echo wait for devices to go
1539 while ((i < 20)); do
1540 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1541 loops=$(do_rpc_nodes $node losetup -a | grep -c t32)
1542 ((devices == 0 && loops == 0)) && return 0
1546 echo "waiting for dev on $node: dev $devices loop $loops given up"
1547 do_rpc_nodes $node "losetup -a"
1548 do_rpc_nodes $node "$LCTL device_list"
1552 t32_verify_quota() {
1556 local fstype=$(facet_fstype $SINGLEMDS)
1560 # LU-2435: if the underlying zfs doesn't support userobj_accounting,
1561 # lustre will estimate the object count usage. This fails quota
1562 # verification in 32b. The object quota usage should be accurate after
1563 # zfs-0.7.0 is released.
1564 [ $fstype == "zfs" ] && {
1565 local zfs_version=$(do_facet $facet cat /sys/module/zfs/version)
1567 [ $(version_code $zfs_version) -lt $(version_code 0.7.0) ] && {
1568 echo "Skip quota verify for zfs: $zfs_version"
1573 $LFS quota -u $T32_QID -v $mnt
1575 qval=$($LFS quota -v -u $T32_QID $mnt |
1576 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1577 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1579 [ $qval -eq $img_bspace ] || {
1580 echo "bspace, act:$qval, exp:$img_bspace"
1584 qval=$($LFS quota -v -u $T32_QID $mnt |
1585 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1586 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1588 [ $qval -eq $img_ispace ] || {
1589 echo "ispace, act:$qval, exp:$img_ispace"
1593 qval=$($LFS quota -v -u $T32_QID $mnt |
1594 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1595 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1597 [ $qval -eq $img_blimit ] || {
1598 echo "blimit, act:$qval, exp:$img_blimit"
1602 qval=$($LFS quota -v -u $T32_QID $mnt |
1603 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1604 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1606 [ $qval -eq $img_ilimit ] || {
1607 echo "ilimit, act:$qval, exp:$img_ilimit"
1611 set_persistent_param_and_check $facet \
1612 "osd-$fstype.$fsname-MDT0000.quota_slave.enabled" \
1613 "$fsname.quota.mdt" ug
1615 set_persistent_param_and_check $facet \
1616 "osd-$fstype.$fsname-OST0000.quota_slave.enabled" \
1617 "$fsname.quota.ost" ug
1620 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1621 bs=1M count=$((img_blimit / 1024)) oflag=sync && {
1622 echo "Write succeed, but expect -EDQUOT"
1625 rm -f $mnt/t32_qf_new
1627 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1629 echo "Create succeed, but expect -EDQUOT"
1632 unlinkmany $mnt/t32_qf_ $img_ilimit
1640 local dne_upgrade=${dne_upgrade:-"no"}
1641 local dom_upgrade=${dom_upgrade:-"no"}
1642 local ff_convert=${ff_convert:-"no"}
1643 local shall_cleanup_mdt=false
1644 local shall_cleanup_mdt1=false
1645 local shall_cleanup_ost=false
1646 local shall_cleanup_lustre=false
1647 local mdt2_is_available=false
1648 local node=$(facet_active_host $SINGLEMDS)
1649 local r="do_node $node"
1665 local fstype=$(facet_fstype $SINGLEMDS)
1666 local mdt_dev=$tmp/mdt
1667 local mdt2_dev=$tmp/mdt2
1668 local ost_dev=$tmp/ost
1673 combined_mgs_mds || stop_mgs || error "Unable to stop MGS"
1674 trap 'trap - RETURN; t32_test_cleanup' RETURN
1677 nid=$($r $LCTL list_nids | head -1)
1679 mkdir -p $tmp/mnt/lustre || error "mkdir $tmp/mnt/lustre failed"
1680 $r mkdir -p $tmp/mnt/{mdt,mdt1,ost}
1681 $r tar xjvf $tarball -S -C $tmp || {
1682 error_noexit "Unpacking the disk image tarball"
1685 img_commit=$($r cat $tmp/commit)
1686 img_kernel=$($r cat $tmp/kernel)
1687 img_arch=$($r cat $tmp/arch)
1688 img_bspace=$($r cat $tmp/bspace)
1689 img_ispace=$($r cat $tmp/ispace)
1691 # older images did not have "blimit" and "ilimit" files
1692 # use old values for T32_BLIMIT and T32_ILIMIT
1693 $r test -f $tmp/blimit && img_blimit=$($r cat $tmp/blimit) ||
1695 $r test -f $tmp/ilimit && img_ilimit=$($r cat $tmp/ilimit) ||
1698 echo "Upgrading from $(basename $tarball), created with:"
1699 echo " Commit: $img_commit"
1700 echo " Kernel: $img_kernel"
1701 echo " Arch: $img_arch"
1702 echo "OST version: $(lustre_build_version ost1)"
1704 # The conversion can be made only when both of the following
1705 # conditions are satisfied:
1706 # - ost device img version < 2.3.64
1707 # - ost server version >= 2.5
1708 [ $(version_code $img_commit) -ge $(version_code 2.3.64) -o \
1709 $(lustre_version_code ost1) -lt $(version_code 2.5.0) ] &&
1712 ! $r test -f $mdt2_dev || mdt2_is_available=true
1714 if [[ $fstype == zfs ]]; then
1717 local poolname_list="t32fs-mdt1 t32fs-ost1"
1719 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1721 for poolname in $poolname_list; do
1723 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
1724 $ZPOOL import -f -d $tmp $poolname"
1727 # upgrade zpool to latest supported features, including
1728 # dnode quota accounting in 0.7.0
1729 $r "$ZPOOL upgrade -a"
1731 mdt_dev=t32fs-mdt1/mdt1
1732 ost_dev=t32fs-ost1/ost1
1733 ! $mdt2_is_available || mdt2_dev=t32fs-mdt2/mdt2
1734 wait_update_facet $SINGLEMDS "$ZPOOL list |
1735 awk '/^t32fs-mdt1/ { print \\\$1 }'" "t32fs-mdt1" || {
1736 error_noexit "import zfs pool failed"
1741 $r $LCTL set_param debug="$PTLDEBUG"
1743 $r $TUNEFS --dryrun $mdt_dev || {
1745 error_noexit "tunefs.lustre before mounting the MDT"
1749 if $mdt2_is_available; then
1750 $r $TUNEFS --dryrun $mdt2_dev || {
1752 error_noexit "tunefs.lustre before mounting the MDT"
1757 if [ "$writeconf" ]; then
1759 if [ $fstype == "ldiskfs" ]; then
1761 $r $TUNEFS --quota $mdt_dev || {
1763 error_noexit "Enable mdt quota feature"
1766 if $mdt2_is_available; then
1767 $r $TUNEFS --quota $mdt2_dev || {
1769 error_noexit "Enable mdt quota feature"
1775 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
1776 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
1777 skip "LU-2200: Cannot run over IB w/o lctl replace_nids "
1778 "(Need MGS version at least 2.3.59)"
1780 local osthost=$(facet_active_host ost1)
1781 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
1784 if [ $fstype == "ldiskfs" ]; then
1787 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt
1788 $r $LCTL replace_nids $fsname-OST0000 $ostnid
1789 $r $LCTL replace_nids $fsname-MDT0000 $nid
1790 $r $UMOUNT $tmp/mnt/mdt
1793 mopts=exclude=$fsname-OST0000
1794 if [ $fstype == "ldiskfs" ]; then
1799 t32_wait_til_devices_gone $node
1801 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
1803 error_noexit "Mounting the MDT"
1806 shall_cleanup_mdt=true
1808 if $mdt2_is_available; then
1809 mopts=mgsnode=$nid,$mopts
1810 $r $MOUNT_CMD -o $mopts $mdt2_dev $tmp/mnt/mdt1 || {
1812 error_noexit "Mounting the MDT"
1816 echo "mount new MDT....$mdt2_dev"
1817 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1818 error_noexit "enable remote dir create failed"
1820 shall_cleanup_mdt1=true
1821 elif [ "$dne_upgrade" != "no" ]; then
1822 local fs2mdsdev=$(mdsdevname 1_2)
1823 local fs2mdsvdev=$(mdsvdevname 1_2)
1825 echo "mkfs new MDT on ${fs2mdsdev}...."
1826 if [ $(facet_fstype mds1) == ldiskfs ]; then
1827 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
1830 add $SINGLEMDS $(mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
1831 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
1832 error_noexit "Mkfs new MDT failed"
1836 [[ $(facet_fstype mds1) != zfs ]] || import_zpool mds1
1838 $r $TUNEFS --dryrun $fs2mdsdev || {
1839 error_noexit "tunefs.lustre before mounting the MDT"
1843 echo "mount new MDT....$fs2mdsdev"
1844 $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
1845 error_noexit "mount mdt1 failed"
1849 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1850 error_noexit "enable remote dir create failed"
1852 shall_cleanup_mdt1=true
1855 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
1856 error_noexit "Getting MDT UUID"
1859 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
1860 error_noexit "Unexpected MDT UUID: \"$uuid\""
1864 $r $TUNEFS --dryrun $ost_dev || {
1865 error_noexit "tunefs.lustre before mounting the OST"
1868 if [ "$writeconf" ]; then
1869 mopts=mgsnode=$nid,$writeconf
1870 if [ $fstype == "ldiskfs" ]; then
1872 $r $TUNEFS --quota $ost_dev || {
1874 error_noexit "Enable ost quota feature"
1880 if [ $fstype == "ldiskfs" ]; then
1885 $r $MOUNT_CMD -onomgs -o$mopts $ost_dev $tmp/mnt/ost || {
1886 error_noexit "Mounting the OST"
1889 shall_cleanup_ost=true
1891 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
1892 error_noexit "Getting OST UUID"
1895 if [ "$uuid" != $fsname-OST0000_UUID ]; then
1896 error_noexit "Unexpected OST UUID: \"$uuid\""
1900 if [[ $PERM_CMD == *"set_param -P"* ]]; then
1901 $r $PERM_CMD osc.$fsname-OST0000*.import=connection=$nid || {
1902 error_noexit "Setting OST \"failover.node\""
1905 $r $PERM_CMD mdc.$fsname-MDT0000*.import=connection=$nid || {
1906 error_noexit "Setting MDT \"failover.node\""
1909 $r $PERM_CMD osc.$fsname-OST0000-*.max_dirty_mb=15 || {
1910 error_noexit "Setting \"max_dirty_mb\""
1913 $r $PERM_CMD mdc.$fsname-MDT0000-*.max_rpcs_in_flight=9 || {
1914 error_noexit "Setting \"max_rpcs_in_flight\""
1917 $r $PERM_CMD lov.$fsname-MDT0000-*.stripesize=4M || {
1918 error_noexit "Setting \"lov.stripesize\""
1921 $r $PERM_CMD mdd.$fsname-MDT0000-*.atime_diff=70 || {
1922 error_noexit "Setting \"mdd.atime_diff\""
1926 $r $PERM_CMD $fsname-OST0000.failover.node=$nid || {
1927 error_noexit "Setting OST \"failover.node\""
1931 $r $PERM_CMD $fsname-MDT0000.failover.node=$nid || {
1932 error_noexit "Setting MDT \"failover.node\""
1936 $r $PERM_CMD $fsname-OST0000.osc.max_dirty_mb=15 || {
1937 error_noexit "Setting \"max_dirty_mb\""
1940 $r $PERM_CMD $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
1941 error_noexit "Setting \"max_rpcs_in_flight\""
1944 $r $PERM_CMD $fsname-MDT0000.lov.stripesize=4M || {
1945 error_noexit "Setting \"lov.stripesize\""
1948 $r $PERM_CMD $fsname-MDT0000.mdd.atime_diff=70 || {
1949 error_noexit "Setting \"mdd.atime_diff\""
1954 $r $LCTL pool_new $fsname.interop || {
1955 error_noexit "Setting \"interop\""
1959 if [ "$ff_convert" != "no" -a $(facet_fstype ost1) == "ldiskfs" ]; then
1960 $r $LCTL lfsck_start -M $fsname-OST0000 || {
1961 error_noexit "Start OI scrub on OST0"
1965 # The oi_scrub should be on ost1, but for test_32(),
1966 # all on the SINGLEMDS.
1967 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
1968 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1969 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
1970 error_noexit "Failed to get the expected 'completed'"
1974 local UPDATED=$($r $LCTL get_param -n \
1975 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1976 awk '/^updated/ { print $2 }')
1977 [ $UPDATED -ge 1 ] || {
1978 error_noexit "Only $UPDATED objects have been converted"
1983 if [ "$dne_upgrade" != "no" ]; then
1984 if [[ $PERM_CMD == *"set_param -P"* ]]; then
1985 $r $PERM_CMD mdc.$fsname-MDT0001*.import=connection=$nid || {
1986 error_noexit "Setting MDT1 \"failover.node\""
1990 $r $PERM_CMD mdc.$fsname-MDT0001-*.max_rpcs_in_flight=9 || {
1991 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
1994 $r $PERM_CMD lov.$fsname-MDT0001-*.stripesize=4M || {
1995 error_noexit "Setting MDT1 \"lov.stripesize\""
1999 $r $PERM_CMD $fsname-MDT0001.failover.node=$nid || {
2000 error_noexit "Setting MDT1 \"failover.node\""
2003 $r $PERM_CMD $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
2004 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
2007 $r $PERM_CMD $fsname-MDT0001.lov.stripesize=4M || {
2008 error_noexit "Setting MDT1 \"lov.stripesize\""
2014 if [ "$writeconf" ]; then
2015 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
2016 error_noexit "Mounting the client"
2020 shall_cleanup_lustre=true
2021 $r $LCTL set_param debug="$PTLDEBUG"
2023 # Leave re-enabling this to a separate patch for LU-11558
2024 # t32_verify_quota $SINGLEMDS $fsname $tmp/mnt/lustre || {
2025 # error_noexit "verify quota failed"
2029 if $r test -f $tmp/list; then
2031 # There is not a Test Framework API to copy files to or
2032 # from a remote node.
2034 # LU-2393 - do both sorts on same node to ensure locale
2036 local list_file=$tmp/list
2038 if $mdt2_is_available; then
2039 if [[ -d $tmp/mnt/lustre/striped_dir_old ]] &&
2040 $r test -f $tmp/list2; then
2041 list_file=$tmp/list2
2042 pushd $tmp/mnt/lustre/striped_dir_old
2044 pushd $tmp/mnt/lustre/remote_dir
2047 pushd $tmp/mnt/lustre
2049 $r cat $list_file | sort -k 6 >$tmp/list.orig
2050 ls -Rni --time-style=+%s | sort -k 6 |
2051 sed 's/\. / /' >$tmp/list || {
2057 # 32-bit and 64-bit clients use different algorithms to
2058 # convert FIDs into inode numbers. Hence, remove the
2059 # inode numbers from the lists, if the original list was
2060 # created on an architecture with different number of
2063 if [ $(t32_bits_per_long $(uname -m)) != \
2064 $(t32_bits_per_long $img_arch) ]; then
2065 echo "Different number of bits per \"long\"" \
2066 "from the disk image"
2067 for list in list.orig list; do
2068 sed -i -e 's/^[0-9]\+[ \t]\+//' \
2072 if ! diff -ub $tmp/list.orig $tmp/list; then
2073 error_noexit "list verification failed"
2077 echo "list verification skipped"
2080 if [ "$dom_upgrade" != "no" ]; then
2081 echo "Check DoM file can be created"
2082 $LFS setstripe -E 1M -L mdt -E EOF $tmp/mnt/lustre/dom || {
2083 error_noexit "Verify DoM creation"
2086 [ $($LFS getstripe -L $tmp/mnt/lustre/dom) == "mdt" ] || {
2087 error_noexit "Verify a DoM file"
2090 dd if=/dev/urandom of=$tmp/mnt/lustre/dom bs=4096 \
2091 count=1 conv=fsync || {
2092 error_noexit "Cannot write to DoM file"
2095 [ $(stat -c%s $tmp/mnt/lustre/dom) == 4096 ] || {
2096 error_noexit "DoM: bad size after write"
2099 rm $tmp/mnt/lustre/dom
2101 set_persistent_param_and_check mds \
2102 "lod.*MDT0000*.dom_stripesize" \
2103 "$fsname-MDT0000.lod.dom_stripesize" 0 || {
2104 error_noexit "Changing \"dom_stripesize\""
2109 if [ "$dne_upgrade" != "no" ]; then
2110 $LFS mkdir -i 1 -c2 $tmp/mnt/lustre/striped_dir || {
2111 error_noexit "set striped dir failed"
2115 $LFS setdirstripe -D -c2 $tmp/mnt/lustre/striped_dir
2117 pushd $tmp/mnt/lustre
2118 tar -cf - . --exclude=./striped_dir \
2119 --exclude=./striped_dir_old \
2120 --exclude=./remote_dir |
2121 tar -xvf - -C striped_dir 1>/dev/null || {
2122 error_noexit "cp to striped dir failed"
2128 # If it is upgrade from DNE (2.5), then rename the remote dir,
2129 # which is created in 2.5 to striped dir.
2130 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]]; then
2131 stripe_index=$($LFS getdirstripe -i \
2132 $tmp/mnt/lustre/remote_dir)
2134 [[ $stripe_index -eq 1 ]] || {
2135 error_noexit "get index \"$stripe_index\"" \
2136 "from remote dir failed"
2139 mv $tmp/mnt/lustre/remote_dir \
2140 $tmp/mnt/lustre/striped_dir/ || {
2141 error_noexit "mv remote dir failed"
2146 # If it is upgraded from DNE (2.7), then move the striped dir
2147 # which was created in 2.7 to the new striped dir.
2148 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]] &&
2149 [[ -d $tmp/mnt/lustre/striped_dir_old ]]; then
2150 stripe_count=$($LFS getdirstripe -c \
2151 $tmp/mnt/lustre/striped_dir_old)
2152 [[ $stripe_count -eq 2 ]] || {
2153 error_noexit "get count $stripe_count" \
2154 "from striped dir failed"
2157 mv $tmp/mnt/lustre/striped_dir_old \
2158 $tmp/mnt/lustre/striped_dir/ || {
2159 error_noexit "mv striped dir failed"
2165 $r $LCTL set_param -n osd*.*.force_sync=1
2166 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
2167 error_noexit "dd failed"
2170 rm -rf $tmp/mnt/lustre/tmp_file || {
2171 error_noexit "rm failed"
2175 if $r test -f $tmp/sha1sums; then
2176 # LU-2393 - do both sorts on same node to ensure locale
2178 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
2179 if [ "$dne_upgrade" != "no" ]; then
2180 pushd $tmp/mnt/lustre/striped_dir
2182 pushd $tmp/mnt/lustre
2185 find ! -path "*remote_dir*" ! -path "*striped_dir*" \
2186 ! -name .lustre -type f -exec sha1sum {} \; |
2187 sort -k 2 >$tmp/sha1sums || {
2189 error_noexit "sha1sum"
2193 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
2194 error_noexit "sha1sum verification failed"
2198 # if upgrade from DNE(2.5), then check remote directory
2199 # if upgrade from DNE(2.7), then check striped directory
2200 if $mdt2_is_available &&
2201 [[ "$dne_upgrade" != "no" ]]; then
2202 local new_dir="$tmp/mnt/lustre/striped_dir"
2203 local striped_dir_old="$new_dir/striped_dir_old"
2205 local dir_list="$new_dir/remote_dir"
2206 [[ ! -d $triped_dir_old ]] ||
2207 dir_list+=" $striped_dir_old"
2209 for dir in $dir_list; do
2211 find ! -name .lustre -type f \
2212 -exec sha1sum {} \; |
2213 sort -k 2 >$tmp/sha1sums || {
2215 error_noexit "sha1sum"
2219 if ! diff -ub $tmp/sha1sums.orig \
2221 error_noexit "sha1sum $dir" \
2228 echo "sha1sum verification skipped"
2231 if [ "$dne_upgrade" != "no" ]; then
2232 rm -rf $tmp/mnt/lustre/striped_dir || {
2233 error_noexit "remove remote dir failed"
2238 # migrate files/dirs to remote MDT, then move them back
2239 if [ $(lustre_version_code mds1) -ge $(version_code 2.7.50) -a \
2240 $dne_upgrade != "no" ]; then
2241 $r $LCTL set_param -n \
2242 mdt.${fsname}*.enable_remote_dir=1 2>/dev/null
2244 echo "test migration"
2245 pushd $tmp/mnt/lustre
2246 for dir in $(find ! -name .lustre ! -name . -type d); do
2247 mdt_index=$($LFS getdirstripe -i $dir)
2248 stripe_cnt=$($LFS getdirstripe -c $dir)
2249 if [ $mdt_index = 0 -a $stripe_cnt -le 1 ]; then
2250 $LFS migrate -m 1 $dir || {
2252 error_noexit "migrate MDT1 failed"
2258 for dir in $(find ! -name . ! -name .lustre -type d); do
2259 mdt_index=$($LFS getdirstripe -i $dir)
2260 stripe_cnt=$($LFS getdirstripe -c $dir)
2261 if [ $mdt_index = 1 -a $stripe_cnt -le 1 ]; then
2262 $LFS migrate -m 0 $dir || {
2264 error_noexit "migrate MDT0 failed"
2273 # When adding new data verification tests, please check for
2274 # the presence of the required reference files first, like
2275 # the "sha1sums" and "list" tests above, to avoid the need to
2276 # regenerate every image for each test addition.
2279 nrpcs_orig=$($LCTL get_param \
2280 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
2281 error_noexit "Getting \"max_rpcs_in_flight\""
2284 nrpcs=$((nrpcs_orig + 5))
2286 set_persistent_param_and_check client \
2287 "mdc.$fsname-MDT0000*.max_rpcs_in_flight" \
2288 "$fsname-MDT0000.mdc.max_rpcs_in_flight" $nrpcs || {
2289 error_noexit "Changing \"max_rpcs_in_flight\""
2293 umount $tmp/mnt/lustre || {
2294 error_noexit "Unmounting the client"
2297 shall_cleanup_lustre=false
2299 if [[ "$dne_upgrade" != "no" ]] || $mdt2_is_available; then
2300 $r $UMOUNT $tmp/mnt/mdt1 || {
2301 error_noexit "Unmounting the MDT2"
2304 if [[ $fstype == zfs ]]; then
2305 $r "$ZPOOL export t32fs-mdt2"
2307 shall_cleanup_mdt1=false
2310 $r $UMOUNT $tmp/mnt/mdt || {
2311 error_noexit "Unmounting the MDT"
2314 if [[ $fstype == zfs ]]; then
2315 $r "$ZPOOL export t32fs-mdt1"
2317 shall_cleanup_mdt=false
2319 $r $UMOUNT $tmp/mnt/ost || {
2320 error_noexit "Unmounting the OST"
2323 if [[ $fstype == zfs ]]; then
2324 $r "$ZPOOL export t32fs-ost1"
2326 shall_cleanup_ost=false
2328 t32_reload_modules $node || {
2329 error_noexit "Reloading modules"
2333 if [[ $fstype == zfs ]]; then
2334 local poolname=t32fs-mdt1
2336 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
2337 $ZPOOL import -f -d $tmp $poolname"
2339 # upgrade zpool to latest supported features,
2340 # including dnode quota accounting in 0.7.0
2341 $r "$ZPOOL upgrade $poolname"
2344 # mount a second time to make sure we didnt leave upgrade flag on
2345 $r $TUNEFS --dryrun $mdt_dev || {
2347 error_noexit "tunefs.lustre before remounting the MDT"
2351 mopts=exclude=$fsname-OST0000
2352 if [ $fstype == "ldiskfs" ]; then
2355 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2356 error_noexit "Remounting the MDT"
2359 shall_cleanup_mdt=true
2369 for tarball in $tarballs; do
2370 t32_test $tarball || let "rc += $?"
2374 run_test 32a "Upgrade (not live)"
2382 for tarball in $tarballs; do
2383 t32_test $tarball writeconf || let "rc += $?"
2387 run_test 32b "Upgrade with writeconf"
2394 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
2396 for tarball in $tarballs; do
2397 # Do not support 1_8 and 2_1 direct upgrade to DNE2 anymore */
2398 echo $tarball | grep "1_8" && continue
2399 echo $tarball | grep "2_1" && continue
2401 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
2405 run_test 32c "dne upgrade test"
2413 for tarball in $tarballs; do
2414 ff_convert=yes t32_test $tarball || rc=$?
2418 run_test 32d "convert ff test"
2421 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.10.56) ]] ||
2422 skip "Need MDS version at least 2.10.56"
2429 for tarball in $tarballs; do
2430 echo $tarball | grep "2_9" || continue
2432 dom_upgrade=yes t32_test $tarball writeconf || let "rc += $?"
2436 run_test 32e "dom upgrade test"
2438 test_33a() { # bug 12333, was test_33
2439 local FSNAME2=test-123
2440 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2443 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
2445 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
2446 local dev=${SINGLEMDS}_dev
2447 local MDSDEV=${!dev}
2448 is_blkdev $SINGLEMDS $MDSDEV &&
2449 skip_env "mixed loopback and real device not working"
2452 local fs2mdsdev=$(mdsdevname 1_2)
2453 local fs2ostdev=$(ostdevname 1_2)
2454 local fs2mdsvdev=$(mdsvdevname 1_2)
2455 local fs2ostvdev=$(ostvdevname 1_2)
2457 if [ $(facet_fstype mds1) == ldiskfs ]; then
2458 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
2461 if combined_mgs_mds; then
2462 local mgs_flag="--mgs"
2465 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --fsname=${FSNAME2} \
2466 --reformat $mgs_flag $mkfsoptions $fs2mdsdev $fs2mdsvdev ||
2468 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2469 --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \
2470 $fs2ostvdev || exit 10
2472 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
2473 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2475 if [[ $PERM_CMD == *"set_param -P"* ]]; then
2476 do_facet mgs "$PERM_CMD timeout=200" ||
2477 error "$PERM_CMD timeout=200 failed"
2479 do_facet mgs "$PERM_CMD $FSNAME2.sys.timeout=200" ||
2480 error "$PERM_CMD $FSNAME2.sys.timeout=200 failed"
2482 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2483 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2486 cp /etc/hosts $MOUNT2/ || error "copy /etc/hosts $MOUNT2/ failed"
2487 $LFS getstripe $MOUNT2/hosts ||
2488 error "$LFS getstripe $MOUNT2/hosts failed"
2493 cleanup_nocli || error "cleanup_nocli failed with $?"
2495 run_test 33a "Mount ost with a large index number"
2497 test_33b() { # was test_34
2500 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
2501 # Drop lock cancelation reply during umount
2502 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
2503 do_facet client $LCTL set_param fail_loc=0x80000304
2504 #lctl set_param debug=-1
2505 umount_client $MOUNT
2506 cleanup || error "cleanup failed with $?"
2508 run_test 33b "Drop cancel during umount"
2512 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
2513 manual_umount_client
2515 do_facet client killall -USR1 multiop
2516 if [ $rc -eq 0 ]; then
2517 error "umount not fail!"
2520 cleanup || error "cleanup failed with rc $?"
2522 run_test 34a "umount with opened file should be fail"
2526 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2527 stop_mds || error "Unable to stop MDS"
2529 manual_umount_client --force || error "mtab after failed umount with $?"
2531 cleanup || error "cleanup failed with $?"
2533 run_test 34b "force umount with failed mds should be normal"
2537 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2538 stop_ost || error "Unable to stop OST1"
2540 manual_umount_client --force || error "mtab after failed umount with $?"
2542 cleanup || error "cleanup failed with $?"
2544 run_test 34c "force umount with failed ost should be normal"
2546 test_35a() { # bug 12459
2549 DBG_SAVE="`$LCTL get_param -n debug`"
2550 $LCTL set_param debug="ha"
2552 log "Set up a fake failnode for the MDS"
2554 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2555 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2557 if [[ $PERM_CMD == *"set_param -P"* ]]; then
2558 do_facet mgs "$PERM_CMD \
2559 mdc.*${device}*.import=connection=$(h2nettype $FAKENID)" ||
2560 error "Setting mdc.*${device}*.import=connection=\
2561 $(h2nettype $FAKENID) failed."
2563 do_facet mgs "$PERM_CMD \
2564 ${device}.failover.node=$(h2nettype $FAKENID)" ||
2565 error "Setting ${device}.failover.node=\
2566 $(h2nettype $FAKENID) failed."
2568 log "Wait for RECONNECT_INTERVAL seconds (10s)"
2571 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
2574 log "Stopping the MDT: $device"
2575 stop_mdt 1 || error "MDT0 stop fail"
2577 df $MOUNT > /dev/null 2>&1 &
2579 log "Restarting the MDT: $device"
2580 start_mdt 1 || error "MDT0 start fail"
2581 log "Wait for df ($DFPID) ... "
2584 $LCTL set_param debug="$DBG_SAVE"
2586 # retrieve from the log the first server that the client tried to
2587 # contact after the connection loss
2588 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2589 NEXTCONN=`awk "/${MSG}/ {start = 1;}
2590 /import_select_connection.*$device-mdc.* using connection/ {
2592 if (\\\$NF ~ /$FAKENID/)
2598 }" $TMP/lustre-log-$TESTNAME.log`
2599 [ "$NEXTCONN" != "0" ] &&
2600 error "Tried to connect to ${NEXTCONN} not last active server"
2601 cleanup || error "cleanup failed with $?"
2602 # remove nid settings
2603 writeconf_or_reformat
2605 run_test 35a "Reconnect to the last active server first"
2607 test_35b() { # bug 18674
2608 remote_mds || skip "local MDS"
2612 $LCTL set_param debug="ha"
2614 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
2617 log "Set up a fake failnode for the MDS"
2619 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2620 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2622 if [[ $PERM_CMD == *"set_param -P"* ]]; then
2623 do_facet mgs "$PERM_CMD \
2624 mdc.*${device}*.import=connection=$(h2nettype $FAKENID)" ||
2625 error "Set mdc.*${device}*.import=connection=\
2626 $(h2nettype $FAKENID) failed"
2628 do_facet mgs "$PERM_CMD \
2629 ${device}.failover.node=$(h2nettype $FAKENID)" ||
2630 error "Set ${device}.failover.node=\
2631 $(h2nettype $FAKENID) failed"
2634 local at_max_saved=0
2635 # adaptive timeouts may prevent seeing the issue
2636 if at_is_enabled; then
2637 at_max_saved=$(at_max_get mds)
2638 at_max_set 0 mds client
2641 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
2643 log "Injecting EBUSY on MDS"
2644 # Setting OBD_FAIL_MDS_RESEND=0x136
2645 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" ||
2646 error "unable to set param fail_loc=0x80000136"
2648 $LCTL set_param mdc.${FSNAME}*.stats=clear
2650 log "Creating a test file and stat it"
2651 touch $MOUNT/$tdir/$tfile || error "touch $MOUNT/$tdir/$tfile failed"
2652 stat $MOUNT/$tdir/$tfile
2654 log "Stop injecting EBUSY on MDS"
2655 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" ||
2656 error "unable to set param fail_loc=0"
2657 rm -f $MOUNT/$tdir/$tfile || error "remove $MOUNT/$tdir/$tfile failed"
2660 # restore adaptive timeout
2661 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
2663 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2665 CONNCNT=$($LCTL get_param mdc.${FSNAME}*.stats |
2666 awk '/mds_connect/{print $2}')
2668 # retrieve from the log if the client has ever tried to
2669 # contact the fake server after the loss of connection
2670 FAILCONN=`awk "BEGIN {ret = 0;}
2671 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
2673 if (\\\$NF ~ /$FAKENID/) {
2678 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
2680 [ "$FAILCONN" == "0" ] &&
2681 error "The client reconnection has not been triggered"
2682 [ "$FAILCONN" == "2" ] &&
2683 error "Primary server busy, client reconnect to failover failed"
2686 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
2687 # Reconnects are supposed to be rate limited to one every 5s
2688 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] &&
2689 error "Too many reconnects $CONNCNT"
2691 cleanup || error "cleanup failed with $?"
2692 # remove nid settings
2693 writeconf_or_reformat
2695 run_test 35b "Continue reconnection retries, if the active server is busy"
2698 [ $OSTCOUNT -lt 2 ] && skip_env "needs >= 2 OSTs"
2700 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] ||
2704 local FSNAME2=test1234
2705 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2707 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
2709 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
2710 is_blkdev $SINGLEMDS $MDSDEV &&
2711 skip_env "mixed loopback and real device not working"
2714 local fs2mdsdev=$(mdsdevname 1_2)
2715 local fs2ostdev=$(ostdevname 1_2)
2716 local fs3ostdev=$(ostdevname 2_2)
2717 local fs2mdsvdev=$(mdsvdevname 1_2)
2718 local fs2ostvdev=$(ostvdevname 1_2)
2719 local fs3ostvdev=$(ostvdevname 2_2)
2721 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2722 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
2723 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
2724 # different one than the default value here.
2725 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2726 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
2727 add fs3ost $(mkfs_opts ost2 ${fs3ostdev}) --mgsnode=$MGSNID \
2728 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
2730 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
2731 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2732 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
2733 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2734 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2736 sleep 5 # until 11778 fixed
2738 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || error "dd failed"
2740 BKTOTAL=$($LCTL get_param -n obdfilter.*.kbytestotal |
2741 awk 'BEGIN{total=0}; {total+=$1}; END{print total}')
2742 BKFREE=$($LCTL get_param -n obdfilter.*.kbytesfree |
2743 awk 'BEGIN{free=0}; {free+=$1}; END{print free}')
2744 BKAVAIL=$($LCTL get_param -n obdfilter.*.kbytesavail |
2745 awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}')
2746 STRING=$(df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}')
2747 DFTOTAL=$(echo $STRING | cut -d, -f1)
2748 DFUSED=$(echo $STRING | cut -d, -f2)
2749 DFAVAIL=$(echo $STRING | cut -d, -f3)
2750 DFFREE=$(($DFTOTAL - $DFUSED))
2752 ALLOWANCE=$((64 * $OSTCOUNT))
2754 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
2755 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
2756 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
2759 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
2760 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
2761 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
2764 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
2765 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
2766 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
2771 stop fs3ost -f || error "unable to stop OST3"
2772 stop fs2ost -f || error "unable to stop OST2"
2773 stop fs2mds -f || error "unable to stop second MDS"
2774 unload_modules_conf || error "unable unload modules"
2777 run_test 36 "df report consistency on OSTs with different block size"
2780 local mntpt=$(facet_mntpt $SINGLEMDS)
2781 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
2782 local mdsdev_sym="$TMP/sym_mdt.img"
2783 local opts=$MDS_MOUNT_OPTS
2786 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2787 skip "ldiskfs only test"
2790 echo "MDS : $mdsdev"
2791 echo "SYMLINK : $mdsdev_sym"
2792 do_facet $SINGLEMDS rm -f $mdsdev_sym
2794 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
2796 echo "mount symlink device - $mdsdev_sym"
2798 if ! do_facet $SINGLEMDS test -b $mdsdev; then
2799 opts=$(csa_add "$opts" -o loop)
2801 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
2802 $mdsdev_sym $mntpt 2>&1)
2805 echo mount_op=$mount_op
2807 do_facet $SINGLEMDS "$UMOUNT $mntpt && rm -f $mdsdev_sym"
2809 if $(echo $mount_op | grep -q "unable to set tunable"); then
2810 error "set tunables failed for symlink device"
2813 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
2815 run_test 37 "verify set tunables works for symlink device"
2817 test_38() { # bug 14222
2818 local fstype=$(facet_fstype $SINGLEMDS)
2819 local mntpt=$(facet_mntpt $SINGLEMDS)
2824 local SRC="/etc /bin"
2825 local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT)
2826 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
2827 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2828 tar cf - $FILES | tar xf - -C $DIR/$tdir ||
2829 error "copying $SRC to $DIR/$tdir"
2831 umount_client $MOUNT || error "umount_client $MOUNT failed"
2832 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2833 stop_mds || error "Unable to stop MDS"
2834 log "delete lov_objid file on MDS"
2836 mount_fstype $SINGLEMDS || error "mount MDS failed (1)"
2838 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid; rm $mntpt/lov_objid"
2840 unmount_fstype $SINGLEMDS || error "umount failed (1)"
2842 # check create in mds_lov_connect
2843 start_mds || error "unable to start MDS"
2844 mount_client $MOUNT || error "mount_client $MOUNT failed"
2846 [ $V ] && log "verifying $DIR/$tdir/$f"
2847 diff -q $f $DIR/$tdir/$f || ERROR=y
2849 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2850 if [ "$ERROR" = "y" ]; then
2851 # check it's updates in sync
2852 umount_client $MOUNT
2854 mount_fstype $SIGNLEMDS
2855 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2856 unmount_fstype $SINGLEMDS
2857 error "old and new files are different after connect" || true
2859 touch $DIR/$tdir/f2 || error "f2 file create failed"
2861 # check it's updates in sync
2862 umount_client $MOUNT || error "second umount_client $MOUNT failed"
2865 mount_fstype $SINGLEMDS || error "mount MDS failed (3)"
2867 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2868 do_facet $SINGLEMDS dd if=/dev/zero of=$mntpt/lov_objid.clear count=8
2870 unmount_fstype $SINGLEMDS || error "umount failed (3)"
2872 start_mds || error "unable to start MDS"
2873 mount_client $MOUNT || error "mount_client $MOUNT failed"
2875 [ $V ] && log "verifying $DIR/$tdir/$f"
2876 diff -q $f $DIR/$tdir/$f || ERROR=y
2878 touch $DIR/$tdir/f3 || error "f3 file create failed"
2879 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2880 umount_client $MOUNT || error "third umount_client $MOUNT failed"
2882 mount_fstype $SINGLEMDS || error "mount MDS failed (4)"
2883 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2884 unmount_fstype $SINGLEMDS || error "umount failed (4)"
2886 [ "$ERROR" = "y" ] &&
2887 error "old and new files are different after sync" || true
2889 log "files compared the same"
2890 cleanup || error "cleanup failed with $?"
2892 run_test 38 "MDS recreates missing lov_objid file from OST data"
2897 cleanup || error "cleanup failed with $?"
2898 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
2899 error "memory leak detected" || true
2901 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
2903 test_40() { # bug 15759
2904 start_ost || error "Unable to start OST1"
2905 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
2906 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
2908 cleanup || error "cleanup failed with rc $?"
2910 run_test 40 "race during service thread startup"
2912 test_41a() { #bug 14134
2913 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2914 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2915 skip "Loop devices does not work with nosvc option"
2918 combined_mgs_mds || skip "needs combined MGT and MDT device"
2920 start_mdt 1 -o nosvc -n
2921 if [ $MDSCOUNT -ge 2 ]; then
2922 for num in $(seq 2 $MDSCOUNT); do
2923 start_mdt $num || return
2926 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
2927 start_mdt 1 -o nomgs,force
2928 mount_client $MOUNT || error "mount_client $MOUNT failed"
2931 echo "blah blah" > $MOUNT/$tfile
2934 umount_client $MOUNT || error "umount_client $MOUNT failed"
2935 stop ost1 -f || error "unable to stop OST1"
2936 stop_mds || error "Unable to stop MDS"
2937 stop_mds || error "Unable to stop MDS on second try"
2939 run_test 41a "mount mds with --nosvc and --nomgs"
2942 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2943 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2944 skip "Loop devices does not work with nosvc option"
2947 ! combined_mgs_mds && skip "needs combined mgs device"
2951 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2953 start_mdt 1 -o nosvc -n
2954 if [ $MDSCOUNT -ge 2 ]; then
2955 for num in $(seq 2 $MDSCOUNT); do
2956 start_mdt $num || return
2959 start_ost || error "Unable to start OST1"
2960 start_mdt 1 -o nomgs,force
2961 mount_client $MOUNT || error "mount_client $MOUNT failed"
2964 echo "blah blah" > $MOUNT/$tfile
2965 cat $MOUNT/$tfile || error "cat $MOUNT/$tfile failed"
2967 umount_client $MOUNT -f || error "umount_client $MOUNT failed"
2968 stop_ost || error "Unable to stop OST1"
2969 stop_mds || error "Unable to stop MDS"
2970 stop_mds || error "Unable to stop MDS on second try"
2972 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
2975 local server_version=$(lustre_version_code $SINGLEMDS)
2976 local oss_list=$(comma_list $(osts_nodes))
2978 [[ $server_version -ge $(version_code 2.6.52) ]] ||
2979 [[ $server_version -ge $(version_code 2.5.26) &&
2980 $server_version -lt $(version_code 2.5.50) ]] ||
2981 [[ $server_version -ge $(version_code 2.5.4) &&
2982 $server_version -lt $(version_code 2.5.11) ]] ||
2983 skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"
2985 # ensure mds1 ost1 have been created even if running sub-test standalone
2988 cleanup || error "cleanup failed"
2990 # using directly mount command instead of start() function to avoid
2991 # any side effect of // with others/externals tools/features
2992 # ("zpool import", ...)
2994 # MDT concurrent start
2996 LOAD_MODULES_REMOTE=true load_modules
2997 do_facet $SINGLEMDS "lsmod | grep -q libcfs" ||
2998 error "MDT concurrent start: libcfs module not loaded"
3000 local mds1dev=$(mdsdevname 1)
3001 local mds1mnt=$(facet_mntpt mds1)
3002 local mds1fstype=$(facet_fstype mds1)
3003 local mds1opts=$MDS_MOUNT_OPTS
3005 if [ $mds1fstype == ldiskfs ] &&
3006 ! do_facet mds1 test -b $mds1dev; then
3007 mds1opts=$(csa_add "$mds1opts" -o loop)
3009 if [[ $mds1fstype == zfs ]]; then
3010 import_zpool mds1 || return ${PIPESTATUS[0]}
3013 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3014 do_facet mds1 "$LCTL set_param fail_loc=0x80000716"
3016 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts &
3019 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts
3023 do_facet mds1 "$LCTL set_param fail_loc=0x0"
3024 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3025 echo "1st MDT start succeed"
3026 echo "2nd MDT start failed with $rc2"
3027 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3028 echo "1st MDT start failed with $rc"
3029 echo "2nd MDT start succeed"
3032 error "unexpected concurrent MDT mounts result, rc=$rc rc2=$rc2"
3035 if [ $MDSCOUNT -ge 2 ]; then
3036 for num in $(seq 2 $MDSCOUNT); do
3037 start_mdt $num || return
3041 # OST concurrent start
3043 do_rpc_nodes $oss_list "lsmod | grep -q libcfs" ||
3044 error "OST concurrent start: libcfs module not loaded"
3046 local ost1dev=$(ostdevname 1)
3047 local ost1mnt=$(facet_mntpt ost1)
3048 local ost1fstype=$(facet_fstype ost1)
3049 local ost1opts=$OST_MOUNT_OPTS
3051 if [ $ost1fstype == ldiskfs ] &&
3052 ! do_facet ost1 test -b $ost1dev; then
3053 ost1opts=$(csa_add "$ost1opts" -o loop)
3055 if [[ $ost1fstype == zfs ]]; then
3056 import_zpool ost1 || return ${PIPESTATUS[0]}
3059 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3060 do_facet ost1 "$LCTL set_param fail_loc=0x80000716"
3062 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts &
3065 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts
3069 do_facet ost1 "$LCTL set_param fail_loc=0x0"
3070 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3071 echo "1st OST start succeed"
3072 echo "2nd OST start failed with $rc2"
3073 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3074 echo "1st OST start failed with $rc"
3075 echo "2nd OST start succeed"
3079 error "unexpected concurrent OST mounts result, rc=$rc rc2=$rc2"
3085 # verify everything ok
3090 error "MDT(s) start failed"
3098 error "OST(s) start failed"
3106 error "client start failed"
3113 error "client mount failed"
3117 run_test 41c "concurrent mounts of MDT/OST should all fail but one"
3119 test_42() { #bug 14693
3123 check_mount || error "client was not mounted"
3125 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3126 PARAM="llite.$FSNAME-*.some_wrong_param"
3128 PARAM="$FSNAME.llite.some_wrong_param"
3131 do_facet mgs $PERM_CMD $PARAM=10
3132 umount_client $MOUNT ||
3133 error "unmounting client failed with invalid llite param"
3134 mount_client $MOUNT ||
3135 error "mounting client failed with invalid llite param"
3137 do_facet mgs $PERM_CMD $PARAM=20
3138 cleanup || error "stopping $FSNAME failed with invalid sys param"
3140 check_mount || error "client was not mounted with invalid sys param"
3141 cleanup || error "stopping $FSNAME failed with invalid sys param"
3143 run_test 42 "allow client/server mount/unmount with invalid config param"
3146 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.58) ]] ||
3147 skip "Need MDS version at least 2.5.58"
3148 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
3151 USER1=$(getent passwd | grep :$ID1:$ID1: | cut -d: -f1)
3152 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1"
3155 chmod ugo+x $DIR || error "chmod 0 failed"
3156 set_persistent_param_and_check mds1 \
3157 "mdt.$FSNAME-MDT0000.root_squash" \
3158 "$FSNAME.mdt.root_squash" \
3160 wait_update $HOSTNAME \
3161 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3163 error "check llite root_squash failed!"
3164 set_persistent_param_and_check mds1 \
3165 "mdt.$FSNAME-MDT0000.nosquash_nids" \
3166 "$FSNAME.mdt.nosquash_nids" \
3168 wait_update $HOSTNAME \
3169 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3171 error "check llite nosquash_nids failed!"
3174 # create set of test files
3176 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
3177 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
3178 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
3180 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
3181 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
3183 mkdir $DIR/$tdir-rootdir || error "mkdir failed"
3184 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
3185 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
3187 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
3188 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
3189 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
3192 # check root_squash:
3193 # set root squash UID:GID to RUNAS_ID
3194 # root should be able to access only files owned by RUNAS_ID
3196 set_persistent_param_and_check mds1 \
3197 "mdt.$FSNAME-MDT0000.root_squash" \
3198 "$FSNAME.mdt.root_squash" \
3199 "$RUNAS_ID:$RUNAS_ID"
3200 wait_update $HOSTNAME \
3201 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3202 "$RUNAS_ID:$RUNAS_ID" ||
3203 error "check llite root_squash failed!"
3205 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
3206 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3207 error "$ST: root read permission is denied"
3208 echo "$ST: root read permission is granted - ok"
3211 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3212 error "$ST: root write permission is denied"
3213 echo "$ST: root write permission is granted - ok"
3215 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3216 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3217 error "$ST: root read permission is granted"
3218 echo "$ST: root read permission is denied - ok"
3221 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3222 error "$ST: root write permission is granted"
3223 echo "$ST: root write permission is denied - ok"
3225 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3226 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null &&
3227 error "$ST: root unlink permission is granted"
3228 echo "$ST: root unlink permission is denied - ok"
3230 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null &&
3231 error "$ST: root create permission is granted"
3232 echo "$ST: root create permission is denied - ok"
3235 # check root_squash is enforced independently
3236 # of client cache content
3238 # access file by USER1, keep access open
3239 # root should be denied access to user file
3241 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
3245 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
3246 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3247 { kill $pid; error "$ST: root read permission is granted"; }
3248 echo "$ST: root read permission is denied - ok"
3251 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3252 { kill $pid; error "$ST: root write permission is granted"; }
3253 echo "$ST: root write permission is denied - ok"
3259 # check nosquash_nids:
3260 # put client's NID into nosquash_nids list,
3261 # root should be able to access root file after that
3263 local NIDLIST=$($LCTL list_nids all | tr '\n' ' ')
3264 NIDLIST="2@gni $NIDLIST 192.168.0.[2,10]@tcp"
3265 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
3266 set_persistent_param_and_check mds1 \
3267 "mdt.$FSNAME-MDT0000.nosquash_nids" \
3268 "$FSNAME-MDTall.mdt.nosquash_nids" \
3270 wait_update $HOSTNAME \
3271 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3273 error "check llite nosquash_nids failed!"
3275 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3276 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3277 error "$ST: root read permission is denied"
3278 echo "$ST: root read permission is granted - ok"
3281 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3282 error "$ST: root write permission is denied"
3283 echo "$ST: root write permission is granted - ok"
3285 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3286 rm $DIR/$tdir-rootdir/tfile-1 ||
3287 error "$ST: root unlink permission is denied"
3288 echo "$ST: root unlink permission is granted - ok"
3289 touch $DIR/$tdir-rootdir/tfile-2 ||
3290 error "$ST: root create permission is denied"
3291 echo "$ST: root create permission is granted - ok"
3292 cleanup || error "cleanup failed with $?"
3294 run_test 43a "check root_squash and nosquash_nids"
3296 test_43b() { # LU-5690
3297 [[ $(lustre_version_code mgs) -ge $(version_code 2.7.62) ]] ||
3298 skip "Need MGS version 2.7.62+"
3300 if [[ -z "$fs2mds_DEV" ]]; then
3301 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
3302 skip_env "mixed loopback and real device not working"
3305 local fs2mdsdev=$(mdsdevname 1_2)
3306 local fs2mdsvdev=$(mdsvdevname 1_2)
3308 # temporarily use fs2mds as fs2mgs
3310 local fs2mgsdev=$fs2mdsdev
3311 local fs2mgsvdev=$fs2mdsvdev
3313 local fsname=test1234
3315 load_module llite/lustre
3316 local client_ip=$(host_nids_address $HOSTNAME $NETTYPE)
3317 local host=${client_ip//*./}
3318 local net=${client_ip/%$host/}
3319 local nosquash_nids=$(h2nettype $net[$host,$host,$host])
3321 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
3322 --param mdt.root_squash=$RUNAS_ID:$RUNAS_ID \
3323 --param mdt.nosquash_nids=$nosquash_nids \
3324 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
3325 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS || error "start fs2mgs failed"
3326 stop $fs2mgs -f || error "stop fs2mgs failed"
3327 cleanup || error "cleanup failed with $?"
3329 run_test 43b "parse nosquash_nids with commas in expr_list"
3333 check_mount || error "check_mount"
3334 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
3336 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
3337 for VAL in $UUIDS; do
3338 NID=$(echo $VAL | cut -d= -f1)
3339 CLUUID=$(echo $VAL | cut -d= -f2)
3340 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
3342 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
3343 cleanup || error "cleanup failed with $?"
3345 run_test 44 "mounted client proc entry exists"
3349 check_mount || error "check_mount"
3350 stop_mds || error "Unable to stop MDS"
3354 #define OBD_FAIL_PTLRPC_LONG_REPL_UNLINK 0x50f
3355 do_facet client "$LCTL set_param fail_loc=0x8000050f"
3358 manual_umount_client --force || error "manual_umount_client failed"
3359 do_facet client "$LCTL set_param fail_loc=0x0"
3360 start_mds || error "unable to start MDS"
3361 mount_client $MOUNT || error "mount_client $MOUNT failed"
3362 cleanup || error "cleanup failed with $?"
3364 run_test 45 "long unlink handling in ptlrpcd"
3371 umount_client $MOUNT2 || rc=$?
3372 umount_client $MOUNT || rc=$?
3373 while [ $count -gt 0 ]; do
3374 stop ost${count} -f || rc=$?
3378 cleanup_nocli || rc=$?
3379 #writeconf to remove all ost2 traces for subsequent tests
3380 writeconf_or_reformat
3385 echo "Testing with $OSTCOUNT OSTs"
3387 start_mds || error "unable to start MDS"
3388 #first client should see only one ost
3389 start_ost || error "Unable to start OST1"
3390 wait_osc_import_state mds ost FULL
3392 mount_client $MOUNT || error "mount_client $MOUNT failed"
3393 trap "cleanup_46a $OSTCOUNT" EXIT ERR
3396 for (( i=2; i<=$OSTCOUNT; i++ )); do
3397 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
3398 error "start_ost$i $(ostdevname $i) failed"
3401 # wait until osts in sync
3402 for (( i=2; i<=$OSTCOUNT; i++ )); do
3403 wait_osc_import_state mds ost$i FULL
3404 wait_osc_import_ready client ost$i
3407 #second client see all ost's
3409 mount_client $MOUNT2 || error "mount_client failed"
3410 $LFS setstripe -c -1 $MOUNT2 ||
3411 error "$LFS setstripe -c -1 $MOUNT2 failed"
3412 $LFS getstripe $MOUNT2 || error "$LFS getstripe $MOUNT2 failed"
3414 echo "ok" > $MOUNT2/widestripe
3415 $LFS getstripe $MOUNT2/widestripe ||
3416 error "$LFS getstripe $MOUNT2/widestripe failed"
3417 # fill acl buffer for avoid expand lsm to them
3418 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd |
3420 setfacl -m $acl $MOUNT2/widestripe
3424 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3426 cleanup_46a $OSTCOUNT || error "cleanup_46a failed"
3428 run_test 46a "handle ost additional - wide striped file"
3433 check_mount || error "check_mount failed"
3434 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
3438 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3439 if echo $ns | grep "MDT[[:digit:]]*"; then
3442 lrs=$(echo $ns | sed 's/.*lru_size=//')
3443 lru_size[count]=$lrs
3448 facet_failover $SINGLEMDS
3449 client_up || error "client_up failed"
3452 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3453 if echo $ns | grep "MDT[[:digit:]]*"; then
3456 lrs=$(echo $ns | sed 's/.*lru_size=//')
3457 if ! test "$lrs" -eq "${lru_size[count]}"; then
3458 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
3459 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
3464 cleanup || error "cleanup failed with $?"
3466 run_test 47 "server restart does not make client loss lru_resize settings"
3471 # reformat after this test is needed - if the test fails,
3472 # we will have unkillable file at FS
3476 test_48() { # bz-17636 LU-7473
3480 check_mount || error "check_mount failed"
3482 $LFS setstripe -c -1 $MOUNT ||
3483 error "$LFS setstripe -c -1 $MOUNT failed"
3484 $LFS getstripe $MOUNT || error "$LFS getstripe $MOUNT failed"
3486 echo "ok" > $MOUNT/widestripe
3487 $LFS getstripe $MOUNT/widestripe ||
3488 error "$LFS getstripe $MOUNT/widestripe failed"
3490 # In the future, we may introduce more EAs, such as selinux, enlarged
3491 # LOV EA, and so on. These EA will use some EA space that is shared by
3492 # ACL entries. So here we only check some reasonable ACL entries count,
3493 # instead of the max number that is calculated from the max_ea_size.
3494 if [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.8.57) ];
3496 count=28 # hard coded of RPC protocol
3497 elif [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3498 count=4000 # max_num 4091 max_ea_size = 32768
3499 elif ! large_xattr_enabled; then
3500 count=450 # max_num 497 max_ea_size = 4012
3502 count=4500 # max_num 8187 max_ea_size = 1048492
3503 # not create too much (>5000) to save test time
3506 echo "It is expected to hold at least $count ACL entries"
3507 trap cleanup_48 EXIT ERR
3508 for ((i = 0; i < $count; i++)) do
3509 setfacl -m u:$((i + 100)):rw $MOUNT/widestripe ||
3510 error "Fail to setfacl for $MOUNT/widestripe at $i"
3513 cancel_lru_locks mdc
3514 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3515 local r_count=$(getfacl $MOUNT/widestripe | grep "user:" | wc -l)
3516 count=$((count + 1)) # for the entry "user::rw-"
3518 [ $count -eq $r_count ] ||
3519 error "Expected ACL entries $count, but got $r_count"
3523 run_test 48 "too many acls on file"
3525 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
3526 test_49a() { # bug 17710
3527 local timeout_orig=$TIMEOUT
3528 local ldlm_timeout_orig=$LDLM_TIMEOUT
3529 local LOCAL_TIMEOUT=20
3531 LDLM_TIMEOUT=$LOCAL_TIMEOUT
3532 TIMEOUT=$LOCAL_TIMEOUT
3536 check_mount || error "client mount failed"
3538 echo "check ldlm_timout..."
3539 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3540 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3541 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3543 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3544 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3547 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
3548 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
3551 umount_client $MOUNT || error "umount_client $MOUNT failed"
3552 stop_ost || error "problem stopping OSS"
3553 stop_mds || error "problem stopping MDS"
3555 LDLM_TIMEOUT=$ldlm_timeout_orig
3556 TIMEOUT=$timeout_orig
3558 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3560 test_49b() { # bug 17710
3561 local timeout_orig=$TIMEOUT
3562 local ldlm_timeout_orig=$LDLM_TIMEOUT
3563 local LOCAL_TIMEOUT=20
3565 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
3566 TIMEOUT=$LOCAL_TIMEOUT
3570 check_mount || error "client mount failed"
3572 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3573 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3574 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3576 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3577 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3580 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
3581 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
3584 cleanup || error "cleanup failed"
3586 LDLM_TIMEOUT=$ldlm_timeout_orig
3587 TIMEOUT=$timeout_orig
3589 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3592 # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1
3594 # Test both statfs and lfs df and fail if either one fails
3595 multiop_bg_pause $1 f_
3598 killall -USR1 multiop
3599 [ $RC -ne 0 ] && log "lazystatfs multiop failed"
3600 wait $PID || { RC=$?; log "multiop return error "; }
3602 # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1
3607 if kill -s 0 $PID; then
3610 log "lazystatfs lfs df failed to complete in 5s"
3618 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3619 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3621 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
3623 cleanup || error "cleanup failed with rc $?"
3625 run_test 50a "lazystatfs all servers available"
3629 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3630 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3632 # Wait for client to detect down OST
3633 stop_ost || error "Unable to stop OST1"
3634 wait_osc_import_state client ost DISCONN
3636 log "OSCs should all be DISCONN"
3638 lazystatfs $MOUNT || error "lazystatfs should not return EIO"
3640 umount_client $MOUNT || error "Unable to unmount client"
3641 stop_mds || error "Unable to stop MDS"
3643 run_test 50b "lazystatfs all servers down"
3646 start_mds || error "Unable to start MDS"
3647 start_ost || error "Unable to start OST1"
3648 start_ost2 || error "Unable to start OST2"
3649 mount_client $MOUNT || error "Unable to mount client"
3650 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3651 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3653 # Wait for client to detect down OST
3654 stop_ost || error "Unable to stop OST1"
3655 wait_osc_import_state mds ost DISCONN
3656 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3658 umount_client $MOUNT || error "Unable to unmount client"
3659 stop_ost2 || error "Unable to stop OST2"
3660 stop_mds || error "Unable to stop MDS"
3661 #writeconf to remove all ost2 traces for subsequent tests
3662 writeconf_or_reformat
3664 run_test 50c "lazystatfs one server down"
3667 start_mds || error "Unable to start MDS"
3668 start_ost || error "Unable to start OST1"
3669 start_ost2 || error "Unable to start OST2"
3670 mount_client $MOUNT || error "Unable to mount client"
3671 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3672 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3674 # Issue the statfs during the window where the client still
3675 # belives the OST to be available but it is in fact down.
3676 # No failure just a statfs which hangs for a timeout interval.
3677 stop_ost || error "Unable to stop OST1"
3678 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3680 umount_client $MOUNT || error "Unable to unmount client"
3681 stop_ost2 || error "Unable to stop OST2"
3682 stop_mds || error "Unable to stop MDS"
3683 #writeconf to remove all ost2 traces for subsequent tests
3684 writeconf_or_reformat
3686 run_test 50d "lazystatfs client/server conn race"
3693 start_mds || error "Unable to start MDS"
3694 #first client should see only one ost
3695 start_ost || error "Unable to start OST1"
3696 wait_osc_import_state mds ost FULL
3698 # Wait for client to detect down OST
3699 stop_ost || error "Unable to stop OST1"
3700 wait_osc_import_state mds ost DISCONN
3702 mount_client $MOUNT || error "Unable to mount client"
3703 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3705 multiop_bg_pause $MOUNT _f
3709 if [ $RC1 -ne 0 ]; then
3710 log "multiop failed $RC1"
3713 sleep $(( $TIMEOUT+1 ))
3715 [ $? -ne 0 ] && error "process isn't sleep"
3716 start_ost || error "Unable to start OST1"
3717 wait $pid || error "statfs failed"
3720 umount_client $MOUNT || error "Unable to unmount client"
3721 stop_ost || error "Unable to stop OST1"
3722 stop_mds || error "Unable to stop MDS"
3724 run_test 50e "normal statfs all servers down"
3729 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
3731 start_mds || error "Unable to start mds"
3732 #first client should see only one ost
3733 start_ost || error "Unable to start OST1"
3734 wait_osc_import_state mds ost FULL
3736 start_ost2 || error "Unable to start OST2"
3737 wait_osc_import_state mds ost2 FULL
3739 # Wait for client to detect down OST
3740 stop_ost2 || error "Unable to stop OST2"
3742 wait_osc_import_state mds ost2 DISCONN
3743 mount_client $MOUNT || error "Unable to mount client"
3744 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3746 multiop_bg_pause $MOUNT _f
3750 if [ $RC1 -ne 0 ]; then
3751 log "lazystatfs multiop failed $RC1"
3754 sleep $(( $TIMEOUT+1 ))
3756 [ $? -ne 0 ] && error "process isn't sleep"
3757 start_ost2 || error "Unable to start OST2"
3758 wait $pid || error "statfs failed"
3759 stop_ost2 || error "Unable to stop OST2"
3762 umount_client $MOUNT -f || error "Unable to unmount client"
3763 stop_ost || error "Unable to stop OST1"
3764 stop_mds || error "Unable to stop MDS"
3765 #writeconf to remove all ost2 traces for subsequent tests
3766 writeconf_or_reformat
3768 run_test 50f "normal statfs one server in down"
3771 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs"
3773 start_ost2 || error "Unable to start OST2"
3774 wait_osc_import_state mds ost2 FULL
3775 wait_osc_import_ready client ost2
3777 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3778 local PARAM="osc.${FSNAME}-OST0001*.active"
3780 local PARAM="${FSNAME}-OST0001.osc.active"
3783 $LFS setstripe -c -1 $DIR/$tfile || error "$LFS setstripe failed"
3784 do_facet mgs $PERM_CMD $PARAM=0 || error "Unable to deactivate OST"
3786 umount_client $MOUNT || error "Unable to unmount client"
3787 mount_client $MOUNT || error "Unable to mount client"
3788 # This df should not cause a panic
3791 do_facet mgs $PERM_CMD $PARAM=1 || error "Unable to activate OST"
3792 rm -f $DIR/$tfile || error "unable to remove file $DIR/$tfile"
3793 umount_client $MOUNT || error "Unable to unmount client"
3794 stop_ost2 || error "Unable to stop OST2"
3795 stop_ost || error "Unable to stop OST1"
3796 stop_mds || error "Unable to stop MDS"
3797 #writeconf to remove all ost2 traces for subsequent tests
3798 writeconf_or_reformat
3800 run_test 50g "deactivated OST should not cause panic"
3804 # prepare MDT/OST, make OSC inactive for OST1
3805 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs"
3807 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3808 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
3809 error "tunefs OST1 failed"
3810 start_mds || error "Unable to start MDT"
3811 start_ost || error "Unable to start OST1"
3812 start_ost2 || error "Unable to start OST2"
3813 mount_client $MOUNT || error "client start failed"
3815 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3817 # activatate OSC for OST1
3818 set_persistent_param_and_check client \
3819 "osc.${FSNAME}-OST0000-osc-[!M]*.active" \
3820 "${FSNAME}-OST0000.osc.active" 1
3822 mkdir $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3823 $LFS setstripe -c -1 -i 0 $DIR/$tdir/2 ||
3824 error "$LFS setstripe $DIR/$tdir/2 failed"
3825 sleep 1 && echo "create a file after OST1 is activated"
3826 # doing some io, shouldn't crash
3827 dd if=/dev/zero of=$DIR/$tdir/2/$tfile-io bs=1M count=10
3829 # check OSC import is working
3830 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
3831 error "some OSC imports are still not connected"
3835 umount_client $MOUNT || error "Unable to umount client"
3836 stop_ost2 || error "Unable to stop OST2"
3837 cleanup_nocli || error "cleanup_nocli failed with $?"
3839 run_test 50h "LU-642: activate deactivated OST"
3842 # prepare MDT/OST, make OSC inactive for OST1
3843 [ "$MDSCOUNT" -lt "2" ] && skip "needs >= 2 MDTs"
3846 [ $(facet_fstype mds2) == zfs ] && import_zpool mds2
3847 do_facet mds2 "$TUNEFS --param mdc.active=0 $(mdsdevname 2)" ||
3848 error "tunefs MDT2 failed"
3849 start_mds || error "Unable to start MDT"
3850 start_ost || error "Unable to start OST1"
3851 start_ost2 || error "Unable to start OST2"
3852 mount_client $MOUNT || error "client start failed"
3854 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3856 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3857 $PERM_CMD mdc.${FSNAME}-MDT0001-mdc-*.active=0 &&
3858 error "deactive MDC0 succeeds"
3860 $PERM_CMD ${FSNAME}-MDT0000.mdc.active=0 &&
3861 error "deactive MDC0 succeeds"
3864 # activate MDC for MDT2
3865 set_persistent_param_and_check client \
3866 "mdc.${FSNAME}-MDT0001-mdc-*.active" \
3867 "${FSNAME}-MDT0001.mdc.active" 1
3869 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
3870 if [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.60) ]
3872 wait_dne_interconnect
3874 $LFS mkdir -i1 $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3876 createmany -o $DIR/$tdir/2/$tfile-%d 1 || error "create files failed"
3878 rm -rf $DIR/$tdir/2 || error "unlink dir failed"
3880 # deactivate MDC for MDT2
3881 set_persistent_param_and_check client \
3882 "mdc.${FSNAME}-MDT0001-mdc-*.active" \
3883 "${FSNAME}-MDT0001.mdc.active" 0
3885 wait_osp_active mds ${FSNAME}-MDT0001 1 0
3887 $LFS mkdir -i1 $DIR/$tdir/2 &&
3888 error "mkdir $DIR/$tdir/2 succeeds after deactive MDT"
3890 $LFS mkdir -i0 -c$MDSCOUNT $DIR/$tdir/striped_dir ||
3891 error "mkdir $DIR/$tdir/striped_dir fails after deactive MDT2"
3893 local stripe_count=$($LFS getdirstripe -c $DIR/$tdir/striped_dir)
3894 [ $stripe_count -eq $((MDSCOUNT - 1)) ] ||
3895 error "wrong $stripe_count != $((MDSCOUNT -1)) for striped_dir"
3898 umount_client $MOUNT || error "Unable to umount client"
3903 run_test 50i "activate deactivated MDT"
3906 local LOCAL_TIMEOUT=20
3910 check_mount || error "check_mount failed"
3912 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
3913 $LFS setstripe -c -1 $MOUNT/$tdir ||
3914 error "$LFS setstripe -c -1 $MOUNT/$tdir failed"
3915 #define OBD_FAIL_MDS_REINT_DELAY 0x142
3916 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x142"
3917 touch $MOUNT/$tdir/$tfile &
3920 start_ost2 || error "Unable to start OST1"
3922 stop_ost2 || error "Unable to stop OST1"
3923 umount_client $MOUNT -f || error “unmount $MOUNT failed”
3924 cleanup_nocli || error “stop server failed”
3925 #writeconf to remove all ost2 traces for subsequent tests
3926 writeconf_or_reformat
3928 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
3937 do_node $node mkdir -p $dest
3938 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3940 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
3941 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3942 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
3944 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
3945 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
3955 local backup2=${TMP}/backup2
3957 do_node $node mkdir -p $backup2
3958 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3960 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
3961 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3962 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
3964 do_node $node "diff -rq $backup $backup2"
3965 [ $? -eq 0 ] || { error "contents differ"; return 3; }
3967 local xattrs2=${TMP}/xattrs2
3968 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
3969 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
3971 do_node $node "diff $xattrs $xattrs2"
3972 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
3974 do_node $node "rm -rf $backup2 $xattrs2"
3975 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
3979 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3980 skip "ldiskfs only test"
3983 start_mds || error "Unable to start MDS"
3984 start_ost || error "Unable to start OST1"
3985 mount_client $MOUNT || error "Unable to mount client"
3988 local ost1mnt=$(facet_mntpt ost1)
3989 local ost1node=$(facet_active_host ost1)
3990 local ost1tmp=$TMP/conf52
3993 mkdir $DIR/$tdir || error "Unable to create $DIR/$tdir"
3994 touch $TMP/modified_first || error "Unable to create temporary file"
3995 local mtime=$(stat -c %Y $TMP/modified_first)
3996 do_node $ost1node "mkdir -p $ost1tmp &&
3997 touch -m -d @$mtime $ost1tmp/modified_first" ||
3998 error "Unable to create temporary file"
4001 $LFS setstripe -c -1 -S 1M $DIR/$tdir || error "$LFS setstripe failed"
4003 for (( i=0; i < nrfiles; i++ )); do
4004 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c ||
4005 error "multiop failed"
4010 # sync all the data and make sure no pending data on the client,
4011 # thus the SOM xattr would not be changed any more.
4012 cancel_lru_locks osc
4015 echo backup files to $TMP/$tdir
4016 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
4017 copy_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
4018 error "Unable to copy files"
4020 umount_client $MOUNT || error "Unable to umount client"
4021 stop_ost || error "Unable to stop ost1"
4023 echo mount ost1 as ldiskfs
4024 do_node $ost1node mkdir -p $ost1mnt || error "Unable to create $ost1mnt"
4025 if ! do_node $ost1node test -b $ost1_dev; then
4028 do_node $ost1node mount -t $(facet_fstype ost1) $loop $ost1_dev \
4030 error "Unable to mount ost1 as ldiskfs"
4033 echo backup objects to $ost1tmp/objects
4034 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
4035 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
4036 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
4038 error "Unable to copy objects"
4040 # move objects to lost+found
4041 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
4042 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
4044 do_node $ost1node "umount $ost1mnt" ||
4045 error "Unable to umount ost1 as ldiskfs"
4047 start_ost || error "Unable to start OST1"
4048 mount_client $MOUNT || error "Unable to mount client"
4050 local REPAIRED=$(do_node $ost1node "$LCTL get_param \
4051 -n osd-ldiskfs.$FSNAME-OST0000.oi_scrub" |
4052 awk '/^lf_repa[ri]*ed/ { print $2 }')
4053 [ $REPAIRED -gt 0 ] ||
4054 error "Some entry under /lost+found should be repaired"
4057 diff_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
4058 error "Unable to diff files"
4060 rm -rf $TMP/$tdir $TMP/file_xattrs ||
4061 error "Unable to delete temporary files"
4062 do_node $ost1node "rm -rf $ost1tmp" ||
4063 error "Unable to delete temporary files"
4064 cleanup || error "cleanup failed with $?"
4066 run_test 52 "check recovering objects from lost+found"
4068 # Checks threads_min/max/started for some service
4070 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
4071 # parameter pattern prefix like 'ost.*.ost'.
4084 local msg="Insane $modname thread counts"
4085 local ncpts=$(check_cpt_number $facet)
4089 check_mount || return 41
4091 # We need to expand $parampat, but it may match multiple parameters, so
4092 # we'll pick the first one
4093 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
4094 error "Couldn't expand ${parampat}.threads_min parameter name"
4098 # Remove the .threads_min part
4099 paramp=${paramp%.threads_min}
4101 # Check for sanity in defaults
4102 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4104 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4106 tstarted=$(do_facet $facet "$LCTL get_param \
4107 -n ${paramp}.threads_started" || echo 0)
4108 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' ||
4110 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4112 nthrs=$(expr $tmax - $tmin)
4113 if [ $nthrs -lt $ncpts ]; then
4119 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
4120 skip_env "module parameter forced $facet thread count"
4122 # Check that we can change min/max
4123 do_facet $facet "$LCTL set_param \
4124 ${paramp}.threads_min=$((tmin + nthrs))"
4125 do_facet $facet "$LCTL set_param \
4126 ${paramp}.threads_max=$((tmax - nthrs))"
4127 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4129 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4131 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) &&
4132 $tmax2 == ($tmax - $nthrs)))' || return $?
4134 # Check that we can set min/max to the same value
4135 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4137 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$tmin"
4138 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4140 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4142 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
4144 # Check that we can't set max < min
4145 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$((tmin - 1))"
4146 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4148 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4150 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
4152 # We need to ensure that we get the module options desired; to do this
4153 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
4154 LOAD_MODULES_REMOTE=true
4157 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
4158 setmodopts -a $modname "$newvalue" oldvalue
4161 check_mount || return 41
4163 # Restore previous setting of MODOPTS_*
4164 setmodopts $modname "$oldvalue"
4166 # Check that $opts took
4167 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4169 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4171 tstarted=$(do_facet $facet \
4172 "$LCTL get_param -n ${paramp}.threads_started" || echo 0)
4173 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4182 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
4183 cleanup || error "cleanup failed with rc $?"
4185 run_test 53a "check OSS thread count params"
4189 local mds=$(do_facet $SINGLEMDS "$LCTL get_param \
4190 -N mds.*.*.threads_max 2>/dev/null")
4191 if [ -z "$mds" ]; then
4192 #running this on an old MDT
4193 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
4195 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
4197 cleanup || error "cleanup failed with $?"
4199 run_test 53b "check MDS thread count params"
4202 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4203 skip "ldiskfs only test"
4206 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p ||
4207 error "llverdev failed with rc=$?"
4210 run_test 54a "test llverdev and partial verify of device"
4213 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4214 skip "ldiskfs only test"
4218 run_llverfs $MOUNT -p || error "llverfs failed with rc=$?"
4219 cleanup || error "cleanup failed with rc=$?"
4221 run_test 54b "test llverfs and partial verify of filesystem"
4225 local max_ost_index=$1
4226 echo -n $(((max_ost_index + 1) * 8))
4230 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4231 skip "ldiskfs only test"
4234 local mdsdev=$(mdsdevname 1)
4235 local mdsvdev=$(mdsvdevname 1)
4239 if ! combined_mgs_mds; then
4240 stop_mgs || error "stopping MGS service failed"
4241 format_mgs || error "formatting MGT failed"
4243 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
4245 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
4246 --reformat $(ostdevname 1) $(ostvdevname 1)
4252 echo checking size of lov_objid for ost index $i
4253 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" |
4254 grep ^User | awk -F 'Size: ' '{print $2}')
4255 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
4256 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
4258 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
4265 run_test 55 "check lov_objid size"
4268 local server_version=$(lustre_version_code $SINGLEMDS)
4269 local mds_journal_size_orig=$MDSJOURNALSIZE
4275 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
4276 $(ostdevname 1) $(ostvdevname 1)
4277 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
4278 $(ostdevname 2) $(ostvdevname 2)
4281 start_ost || error "Unable to start first ost (idx 10000)"
4282 start_ost2 || error "Unable to start second ost (idx 1000)"
4283 mount_client $MOUNT || error "Unable to mount client"
4287 if [[ $server_version -ge $(version_code 2.6.54) ]] ||
4288 [[ $server_version -ge $(version_code 2.5.4) &&
4289 $server_version -lt $(version_code 2.5.11) ]]; then
4290 wait_osc_import_state mds ost1 FULL
4291 wait_osc_import_state mds ost2 FULL
4292 $LFS setstripe --stripe-count=-1 $DIR/$tfile ||
4293 error "Unable to setstripe $DIR/$tfile"
4294 n=$($LFS getstripe --stripe-count $DIR/$tfile)
4295 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
4300 MDSJOURNALSIZE=$mds_journal_size_orig
4303 run_test 56a "check big OST indexes and out-of-index-order start"
4308 umount_client $MOUNT -f || error "unmount client failed"
4317 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs"
4319 trap cleanup_56b EXIT RETURN ERR
4322 if ! combined_mgs_mds ; then
4327 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) --index=0 --reformat \
4328 $(mdsdevname 1) $(mdsvdevname 1)
4329 add mds2 $(mkfs_opts mds2 $(mdsdevname 2)) --index=1 --reformat \
4330 $(mdsdevname 2) $(mdsvdevname 2)
4331 add mds3 $(mkfs_opts mds3 $(mdsdevname 3)) --index=1000 --reformat \
4332 $(mdsdevname 3) $(mdsvdevname 3)
4336 start_mdt 1 || error "MDT 1 (idx 0) start failed"
4337 start_mdt 2 || error "MDT 2 (idx 1) start failed"
4338 start_mdt 3 || error "MDT 3 (idx 1000) start failed"
4339 start_ost || error "Unable to start first ost"
4340 start_ost2 || error "Unable to start second ost"
4342 do_nodes $(comma_list $(mdts_nodes)) \
4343 "$LCTL set_param mdt.*.enable_remote_dir=1 \
4344 mdt.*.enable_remote_dir_gid=-1"
4346 mount_client $MOUNT || error "Unable to mount client"
4348 $LFS mkdir -c3 $MOUNT/$tdir || error "failed to make testdir"
4350 echo "This is test file 1!" > $MOUNT/$tdir/$tfile.1 ||
4351 error "failed to make test file 1"
4352 echo "This is test file 2!" > $MOUNT/$tdir/$tfile.2 ||
4353 error "failed to make test file 2"
4354 echo "This is test file 1000!" > $MOUNT/$tdir/$tfile.1000 ||
4355 error "failed to make test file 1000"
4357 rm -rf $MOUNT/$tdir || error "failed to remove testdir"
4359 $LFS mkdir -i1000 $MOUNT/$tdir.1000 ||
4360 error "create remote dir at idx 1000 failed"
4363 echo "=== START lfs df OUTPUT ==="
4365 echo "==== END lfs df OUTPUT ===="
4367 mdtcnt=$(echo -e "$output" | grep $FSNAME-MDT | wc -l)
4368 ostcnt=$(echo -e "$output" | grep $FSNAME-OST | wc -l)
4370 echo "lfs df returned mdt count $mdtcnt and ost count $ostcnt"
4371 [ $mdtcnt -eq 3 ] || error "lfs df returned wrong mdt count"
4372 [ $ostcnt -eq 2 ] || error "lfs df returned wrong ost count"
4374 echo "This is test file 1!" > $MOUNT/$tdir.1000/$tfile.1 ||
4375 error "failed to make test file 1"
4376 echo "This is test file 2!" > $MOUNT/$tdir.1000/$tfile.2 ||
4377 error "failed to make test file 2"
4378 echo "This is test file 1000!" > $MOUNT/$tdir.1000/$tfile.1000 ||
4379 error "failed to make test file 1000"
4380 rm -rf $MOUNT/$tdir.1000 || error "failed to remove remote_dir"
4383 echo "=== START lfs mdts OUTPUT ==="
4385 echo "==== END lfs mdts OUTPUT ===="
4387 echo -e "$output" | grep -v "MDTS:" | awk '{print $1}' |
4388 sed 's/://g' > $TMP/mdts-actual.txt
4389 sort $TMP/mdts-actual.txt -o $TMP/mdts-actual.txt
4391 echo -e "0\n1\n1000" > $TMP/mdts-expected.txt
4393 diff $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4396 rm $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4398 [ $result -eq 0 ] || error "target_obd proc file is incorrect!"
4400 run_test 56b "test target_obd correctness with nonconsecutive MDTs"
4402 test_57a() { # bug 22656
4403 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4404 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4405 tail -1 | awk '{print $1}')
4406 writeconf_or_reformat
4407 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4408 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" ||
4409 error "tunefs failed"
4411 start_ost && error "OST registration from failnode should fail"
4414 run_test 57a "initial registration from failnode should fail (should return errs)"
4417 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4418 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4419 tail -1 | awk '{print $1}')
4420 writeconf_or_reformat
4421 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4422 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" ||
4423 error "tunefs failed"
4425 start_ost || error "OST registration from servicenode should not fail"
4428 run_test 57b "initial registration from servicenode should not fail"
4431 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
4434 test_58() { # bug 22658
4435 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
4437 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4438 createmany -o $DIR/$tdir/$tfile-%d 100
4439 unlinkmany $DIR/$tdir/$tfile-%d 100
4440 stop_mds || error "Unable to stop MDS"
4442 local MNTDIR=$(facet_mntpt $SINGLEMDS)
4443 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4445 # remove all files from the OBJECTS dir
4446 mount_fstype $SINGLEMDS
4448 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
4450 unmount_fstype $SINGLEMDS
4451 # restart MDS with missing llog files
4452 start_mds || error "unable to start MDS"
4453 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0"
4456 run_test 58 "missing llog files must not prevent MDT from mounting"
4459 start_mgsmds >> /dev/null
4460 local C1=$(count_osts)
4461 if [ $C1 -eq 0 ]; then
4462 start_ost >> /dev/null
4466 echo "original ost count: $C1 (expect > 0)"
4467 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
4468 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
4469 local C2=$(count_osts)
4470 echo "after mdt writeconf count: $C2 (expect 0)"
4471 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
4472 echo "OST start without writeconf should fail:"
4473 start_ost >> /dev/null &&
4474 error "OST start without writeconf didn't fail"
4475 echo "OST start with writeconf should succeed:"
4476 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
4477 local C3=$(count_osts)
4478 echo "after ost writeconf count: $C3 (expect 1)"
4479 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
4480 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
4481 local C4=$(count_osts)
4482 echo "after ost2 writeconf count: $C4 (expect 2)"
4483 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
4484 stop_ost2 >> /dev/null
4485 cleanup_nocli >> /dev/null
4486 #writeconf to remove all ost2 traces for subsequent tests
4487 writeconf_or_reformat
4489 run_test 59 "writeconf mount option"
4491 test_60() { # LU-471
4492 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4493 skip "ldiskfs only test"
4498 for num in $(seq $MDSCOUNT); do
4499 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
4500 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
4501 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4505 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
4506 [ ${PIPESTATUS[0]} -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
4508 # MDT default has dirdata feature
4509 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
4510 # we disable uninit_bg feature
4511 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
4512 # we set stride extended options
4513 echo $dump | grep stride > /dev/null || error "stride is not set"
4516 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
4521 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] ||
4522 skip "Need MDS version at least 2.1.53"
4524 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
4525 ! large_xattr_enabled; then
4528 for num in $(seq $MDSCOUNT); do
4529 do_facet mds${num} $TUNE2FS -O ea_inode \
4530 $(mdsdevname $num) ||
4531 error "tune2fs on mds $num failed"
4535 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
4536 setup_noconfig || error "setting up the filesystem failed"
4537 client_up || error "starting client failed"
4539 local file=$DIR/$tfile
4540 touch $file || error "touch $file failed"
4542 local large_value="$(generate_string $(max_xattr_size))"
4543 local small_value="bar"
4545 local name="trusted.big"
4546 log "save large xattr $name on $file"
4547 setfattr -n $name -v $large_value $file ||
4548 error "saving $name on $file failed"
4550 local new_value=$(get_xattr_value $name $file)
4551 [[ "$new_value" != "$large_value" ]] &&
4552 error "$name different after saving"
4554 log "shrink value of $name on $file"
4555 setfattr -n $name -v $small_value $file ||
4556 error "shrinking value of $name on $file failed"
4558 new_value=$(get_xattr_value $name $file)
4559 [[ "$new_value" != "$small_value" ]] &&
4560 error "$name different after shrinking"
4562 log "grow value of $name on $file"
4563 setfattr -n $name -v $large_value $file ||
4564 error "growing value of $name on $file failed"
4566 new_value=$(get_xattr_value $name $file)
4567 [[ "$new_value" != "$large_value" ]] &&
4568 error "$name different after growing"
4570 log "check value of $name on $file after remounting MDS"
4572 new_value=$(get_xattr_value $name $file)
4573 [[ "$new_value" != "$large_value" ]] &&
4574 error "$name different after remounting MDS"
4576 log "remove large xattr $name from $file"
4577 setfattr -x $name $file || error "removing $name from $file failed"
4580 stopall || error "stopping for e2fsck run"
4581 for num in $(seq $MDSCOUNT); do
4582 run_e2fsck $(facet_active_host mds$num) \
4583 $(mdsdevname $num) "-y" ||
4584 error "e2fsck MDT$num failed"
4586 setup_noconfig || error "remounting the filesystem failed"
4589 # need to delete this file to avoid problems in other tests
4591 stopall || error "stopping systems failed"
4593 run_test 61 "large xattr"
4596 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4597 skip "ldiskfs only test"
4599 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] ||
4600 skip "Need MDS version at least 2.2.51"
4603 local mdsdev=$(mdsdevname 1)
4604 local ostdev=$(ostdevname 1)
4606 echo "disable journal for mds"
4607 do_facet mds1 $TUNE2FS -O ^has_journal $mdsdev || error "tune2fs failed"
4608 start_mds && error "MDT start should fail"
4609 echo "disable journal for ost"
4610 do_facet ost1 $TUNE2FS -O ^has_journal $ostdev || error "tune2fs failed"
4611 start_ost && error "OST start should fail"
4612 cleanup || error "cleanup failed with rc $?"
4615 run_test 62 "start with disabled journal"
4618 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4619 skip "ldiskfs only test"
4622 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ldiskfs
4623 local inode_slab=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
4624 awk '/ldiskfs_inode_cache/ { print $5 / $6 }')
4625 if [ -z "$inode_slab" ]; then
4626 skip "ldiskfs module has not been loaded"
4629 echo "$inode_slab ldiskfs inodes per page"
4630 [ "${inode_slab%.*}" -ge "3" ] && return 0
4632 # If kmalloc-128 is also 1 per page - this is a debug kernel
4633 # and so this is not an error.
4634 local kmalloc128=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
4635 awk '/^(dma-kmalloc|size)-128 / { print $5 / $6 }')
4636 # 32 128-byte chunks in 4k
4637 [ "${kmalloc128%.*}" -lt "32" ] ||
4638 error "ldiskfs inode too big, only $inode_slab objs/page, " \
4639 "kmalloc128 = $kmalloc128 objs/page"
4641 run_test 63 "Verify each page can at least hold 3 ldiskfs inodes"
4644 start_mds || error "unable to start MDS"
4645 start_ost || error "Unable to start OST1"
4646 start_ost2 || error "Unable to start second ost"
4647 mount_client $MOUNT || error "Unable to mount client"
4648 stop_ost2 || error "Unable to stop second ost"
4651 umount_client $MOUNT -f || error “unmount $MOUNT failed”
4652 cleanup_nocli || error "cleanup_nocli failed with $?"
4653 #writeconf to remove all ost2 traces for subsequent tests
4654 writeconf_or_reformat
4656 run_test 64 "check lfs df --lazy "
4658 test_65() { # LU-2237
4659 # Currently, the test is only valid for ldiskfs backend
4660 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4661 skip "ldiskfs only test"
4663 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4664 local brpt=$(facet_mntpt brpt)
4667 if ! do_facet $SINGLEMDS "test -b $devname"; then
4671 stop_mds || error "Unable to stop MDS"
4672 local obj=$(do_facet $SINGLEMDS \
4673 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
4675 if [ -z "$obj" ]; then
4676 # The MDT may be just re-formatted, mount the MDT for the
4677 # first time to guarantee the "last_rcvd" file is there.
4678 start_mds || error "fail to mount the MDS for the first time"
4679 stop_mds || error "Unable to stop MDS"
4682 # remove the "last_rcvd" file
4683 do_facet $SINGLEMDS "mkdir -p $brpt"
4684 do_facet $SINGLEMDS \
4685 "mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
4686 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
4687 do_facet $SINGLEMDS "$UMOUNT $brpt"
4689 # restart MDS, the "last_rcvd" file should be recreated.
4690 start_mds || error "fail to restart the MDS"
4691 stop_mds || error "Unable to stop MDS"
4692 obj=$(do_facet $SINGLEMDS \
4693 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
4694 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
4696 run_test 65 "re-create the lost last_rcvd file when server mount"
4699 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
4700 skip "Need MGS version at least 2.3.59"
4703 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
4704 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
4706 # add EXCLUDE records to config log, they are not to be
4707 # removed by lctl replace_nids
4708 set_conf_param_and_check mds \
4709 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
4710 "$FSNAME-OST0000.osc.active" \
4713 echo "replace_nids should fail if MDS, OSTs and clients are UP"
4714 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4715 error "replace_nids fail"
4717 umount_client $MOUNT || error "unmounting client failed"
4718 echo "replace_nids should fail if MDS and OSTs are UP"
4719 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4720 error "replace_nids fail"
4722 stop_ost || error "Unable to stop OST1"
4723 echo "replace_nids should fail if MDS is UP"
4724 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4725 error "replace_nids fail"
4727 stop_mds || error "stopping mds failed"
4729 if combined_mgs_mds; then
4730 start_mdt 1 "-o nosvc" ||
4731 error "starting mds with nosvc option failed"
4734 echo "command should accept two parameters"
4735 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
4736 error "command should accept two params"
4738 echo "correct device name should be passed"
4739 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
4740 error "wrong devname"
4742 echo "wrong nids list should not destroy the system"
4743 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
4745 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "asdfasdf, asdfadf" &&
4748 echo "replace OST nid"
4749 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
4750 error "replace nids failed"
4752 echo "command should accept two parameters"
4753 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
4754 error "command should accept two params"
4756 echo "wrong nids list should not destroy the system"
4757 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
4760 local FAKE_NIDS="192.168.0.112@tcp1,192.168.0.112@tcp2"
4761 local FAKE_FAILOVER="192.168.0.113@tcp1,192.168.0.113@tcp2"
4762 local NIDS_AND_FAILOVER="$MDS_NID,$FAKE_NIDS:$FAKE_FAILOVER"
4763 echo "set NIDs with failover"
4764 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $NIDS_AND_FAILOVER ||
4765 error "replace nids failed"
4768 echo "replace MDS nid"
4769 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
4770 error "replace nids failed"
4772 if ! combined_mgs_mds ; then
4775 stop_mds || error "Unable to stop MDS"
4778 start_mgsmds || error "start mgsmds failed"
4779 set_conf_param_and_check mds \
4780 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
4781 "$FSNAME-OST0000.osc.active" \
4783 start_ost || error "unable to start OST"
4784 mount_client $MOUNT || error "mount client failed"
4786 check_mount || error "error after nid replace"
4787 cleanup || error "cleanup failed"
4790 run_test 66 "replace nids"
4792 test_67() { #LU-2950
4793 local legacy="$TMP/legacy_lnet_config"
4794 local new="$TMP/new_routes_test"
4795 local out="$TMP/config_out_file"
4796 local verify="$TMP/conv_verify"
4797 local verify_conf="$TMP/conf_verify"
4799 # Create the legacy file that will be run through the
4800 # lustre_routes_conversion script
4801 cat <<- LEGACY_LNET_CONFIG > $legacy
4802 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
4803 tcp2 54 10.1.3.2@tcp;
4804 tcp3 10.3.4.3@tcp:3;
4808 # Create the verification file to verify the output of
4809 # lustre_routes_conversion script against.
4810 cat <<- VERIFY_LNET_CONFIG > $verify
4811 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
4812 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
4813 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
4814 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
4815 tcp4: { gateway: 10.3.3.4@tcp }
4818 # Create the verification file to verify the output of
4819 # lustre_routes_config script against
4820 cat <<- VERIFY_LNET_CONFIG > $verify_conf
4821 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
4822 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
4823 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
4824 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
4825 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
4828 $LUSTRE_ROUTES_CONVERSION $legacy $new > /dev/null
4829 if [ -f $new ]; then
4830 # verify the conversion output
4831 cmp -s $new $verify > /dev/null
4832 if [ $? -eq 1 ]; then
4833 error "routes conversion failed"
4836 lustre_routes_config --dry-run --verbose $new > $out
4837 # check that the script succeeded
4838 cmp -s $out $verify_conf > /dev/null
4839 if [ $? -eq 1 ]; then
4840 error "routes config failed"
4843 error "routes conversion test failed"
4845 # remove generated files
4846 rm -f $new $legacy $verify $verify_conf $out
4848 run_test 67 "test routes conversion and configuration"
4856 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.4.53) ] ||
4857 skip "Need MDS version at least 2.4.53"
4859 umount_client $MOUNT || error "umount client failed"
4861 if ! combined_mgs_mds; then
4862 start_mgs || error "start mgs failed"
4865 start_mdt 1 || error "MDT start failed"
4866 start_ost || error "Unable to start OST1"
4868 # START-END - the sequences we'll be reserving
4869 START=$(do_facet $SINGLEMDS \
4870 $LCTL get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
4871 END=$((START + (1 << 30)))
4872 do_facet $SINGLEMDS \
4873 $LCTL set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
4875 # reset the sequences MDT0000 has already assigned
4876 do_facet $SINGLEMDS \
4877 $LCTL set_param seq.srv*MDT0000.space=clear
4879 # remount to let the client allocate new sequence
4880 mount_client $MOUNT || error "mount client failed"
4882 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4883 do_facet $SINGLEMDS \
4884 $LCTL get_param seq.srv*MDT0000.space
4885 $LFS path2fid $DIR/$tfile
4887 local old_ifs="$IFS"
4889 fid=($($LFS path2fid $DIR/$tfile))
4893 if [[ $seq < $END ]]; then
4894 error "used reserved sequence $seq?"
4896 cleanup || error "cleanup failed with $?"
4898 run_test 68 "be able to reserve specific sequences in FLDB"
4900 # Test 69: is about the total number of objects ever created on an OST.
4901 # so that when it is reformatted the normal MDS->OST orphan recovery won't
4902 # just "precreate" the missing objects. In the past it might try to recreate
4903 # millions of objects after an OST was reformatted
4905 local server_version=$(lustre_version_code $SINGLEMDS)
4907 [[ $server_version -lt $(version_code 2.4.2) ]] &&
4908 skip "Need MDS version at least 2.4.2"
4910 [[ $server_version -ge $(version_code 2.4.50) ]] &&
4911 [[ $server_version -lt $(version_code 2.5.0) ]] &&
4912 skip "Need MDS version at least 2.5.0"
4915 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4917 # use OST0000 since it probably has the most creations
4918 local OSTNAME=$(ostname_from_index 0)
4919 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
4920 local last_id=$(do_facet mds1 $LCTL get_param -n \
4921 osc.$mdtosc_proc1.prealloc_last_id)
4923 # Want to have OST LAST_ID over 5 * OST_MAX_PRECREATE to
4924 # verify that the LAST_ID recovery is working properly. If
4925 # not, then the OST will refuse to allow the MDS connect
4926 # because the LAST_ID value is too different from the MDS
4927 #define OST_MAX_PRECREATE=20000
4928 local ost_max_pre=20000
4929 local num_create=$(( ost_max_pre * 5 + 1 - last_id))
4931 # If the LAST_ID is already over 5 * OST_MAX_PRECREATE, we don't
4932 # need to create any files. So, skip this section.
4933 if [ $num_create -gt 0 ]; then
4934 # Check the number of inodes available on OST0
4936 local ifree=$($LFS df -i $MOUNT | awk '/OST0000/ { print $4 }')
4937 log "On OST0, $ifree inodes available. Want $num_create."
4939 $LFS setstripe -i 0 $DIR/$tdir ||
4940 error "$LFS setstripe -i 0 $DIR/$tdir failed"
4941 if [ $ifree -lt 10000 ]; then
4942 files=$(( ifree - 50 ))
4947 local j=$((num_create / files + 1))
4948 for i in $(seq 1 $j); do
4949 createmany -o $DIR/$tdir/$tfile-$i- $files ||
4950 error "createmany fail create $files files: $?"
4951 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
4952 error "unlinkmany failed unlink $files files"
4956 # delete all of the files with objects on OST0 so the
4957 # filesystem is not inconsistent later on
4958 $LFS find $MOUNT --ost 0 -print0 | xargs -0 rm
4960 umount_client $MOUNT || error "umount client failed"
4961 stop_ost || error "OST0 stop failure"
4962 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
4963 $(ostdevname 1) $(ostvdevname 1) ||
4964 error "reformat and replace $ostdev failed"
4965 start_ost || error "OST0 restart failure"
4966 wait_osc_import_state mds ost FULL
4968 mount_client $MOUNT || error "mount client failed"
4969 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
4970 local idx=$($LFS getstripe -i $DIR/$tdir/$tfile-last)
4971 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
4973 local iused=$($LFS df -i $MOUNT | awk '/OST0000/ { print $3 }')
4974 log "On OST0, $iused used inodes"
4975 [ $iused -ge $((ost_max_pre/2 + 1000)) ] &&
4976 error "OST replacement created too many inodes; $iused"
4977 cleanup || error "cleanup failed with $?"
4979 run_test 69 "replace an OST with the same index"
4982 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
4985 cleanup || error "cleanup failed with $?"
4987 start_mdt 1 || error "MDT0 start fail"
4989 start_ost || error "OST0 start fail"
4990 for num in $(seq 2 $MDSCOUNT); do
4991 start_mdt $num || return
4994 mount_client $MOUNT || error "mount client fails"
4996 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
4998 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4999 error "create remote dir fail"
5001 rm -rf $DIR/$tdir || error "delete dir fail"
5002 cleanup || error "cleanup failed with $?"
5004 run_test 70a "start MDT0, then OST, then MDT1"
5007 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5010 start_ost || error "OST0 start fail"
5012 start_mds || error "MDS start fail"
5014 mount_client $MOUNT || error "mount client fails"
5016 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
5018 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5019 error "create remote dir fail"
5021 rm -rf $DIR/$tdir || error "delete dir fail"
5023 cleanup || error "cleanup failed with $?"
5025 run_test 70b "start OST, MDT1, MDT0"
5028 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5031 start_mds || error "MDS start fail"
5032 start_ost || error "OST0 start fail"
5034 mount_client $MOUNT || error "mount client fails"
5035 stop_mdt 1 || error "MDT1 start fail"
5037 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
5038 echo "deactivate $mdc_for_mdt1"
5039 $LCTL --device $mdc_for_mdt1 deactivate ||
5040 error "set $mdc_for_mdt1 deactivate failed"
5042 mkdir $DIR/$tdir && error "mkdir succeed"
5044 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
5045 error "create remote dir succeed"
5047 cleanup || error "cleanup failed with $?"
5049 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
5052 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5055 start_mds || error "MDS start fail"
5056 start_ost || error "OST0 start fail"
5058 mount_client $MOUNT || error "mount client fails"
5060 stop_mdt 2 || error "MDT1 start fail"
5062 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
5064 echo "deactivate $mdc_for_mdt2"
5065 $LCTL --device $mdc_for_mdt2 deactivate ||
5066 error "set $mdc_for_mdt2 deactivate failed"
5068 mkdir $DIR/$tdir || error "mkdir fail"
5069 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
5070 error "create remote dir succeed"
5072 rm -rf $DIR/$tdir || error "delete dir fail"
5074 cleanup || error "cleanup failed with $?"
5076 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
5079 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5081 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.62) ] ||
5082 skip "Need MDS version at least 2.7.62"
5084 cleanup || error "cleanup failed with $?"
5086 local mdsdev=$(mdsdevname 1)
5087 local ostdev=$(ostdevname 1)
5088 local mdsvdev=$(mdsvdevname 1)
5089 local ostvdev=$(ostvdevname 1)
5090 local opts_mds="$(mkfs_opts mds1 $mdsdev) --reformat $mdsdev $mdsvdev"
5091 local opts_ost="$(mkfs_opts ost1 $ostdev) --reformat $ostdev $ostvdev"
5093 add mds1 $opts_mds || error "add mds1 failed"
5094 start_mdt 1 || error "start mdt1 failed"
5095 add ost1 $opts_ost || error "add ost1 failed"
5096 start_ost || error "start ost failed"
5097 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
5099 local soc=$(do_facet mds1 "$LCTL get_param -n \
5100 mdt.*MDT0000.sync_lock_cancel")
5101 [ $soc == "never" ] || error "SoC enabled on single MDS"
5103 for i in $(seq 2 $MDSCOUNT); do
5104 mdsdev=$(mdsdevname $i)
5105 mdsvdev=$(mdsvdevname $i)
5106 opts_mds="$(mkfs_opts mds$i $mdsdev) --reformat $mdsdev \
5108 add mds$i $opts_mds || error "add mds$i failed"
5109 start_mdt $i || error "start mdt$i fail"
5112 wait_dne_interconnect
5114 for i in $(seq $MDSCOUNT); do
5115 soc=$(do_facet mds$i "$LCTL get_param -n \
5116 mdt.*MDT000$((i - 1)).sync_lock_cancel")
5117 [ $soc == "blocking" ] || error "SoC not enabled on DNE"
5120 for i in $(seq 2 $MDSCOUNT); do
5121 stop_mdt $i || error "stop mdt$i fail"
5123 soc=$(do_facet mds1 "$LCTL get_param -n \
5124 mdt.*MDT0000.sync_lock_cancel")
5125 [ $soc == "never" ] || error "SoC enabled on single MDS"
5126 umount_client $MOUNT -f > /dev/null
5128 cleanup || error "cleanup failed with $?"
5130 run_test 70e "Sync-on-Cancel will be enabled by default on DNE"
5133 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5134 if combined_mgs_mds; then
5135 skip "needs separate MGS/MDT"
5139 start_mdt 1 || error "MDT0 start fail"
5140 start_ost || error "OST0 start fail"
5141 for num in $(seq 2 $MDSCOUNT); do
5142 start_mdt $num || return
5145 start_ost2 || error "OST1 start fail"
5147 mount_client $MOUNT || error "mount client fails"
5149 mkdir $DIR/$tdir || error "mkdir fail"
5150 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5151 error "create remote dir succeed"
5153 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5154 rm -rf $DIR/$tdir || error "delete dir fail"
5156 umount_client $MOUNT || error "umount_client failed"
5157 stop_mds || error "MDS stop fail"
5158 stop_ost || error "OST0 stop fail"
5159 stop_ost2 || error "OST1 stop fail"
5161 run_test 71a "start MDT0 OST0, MDT1, OST1"
5164 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5165 if combined_mgs_mds; then
5166 skip "needs separate MGS/MDT"
5170 for num in $(seq 2 $MDSCOUNT); do
5171 start_mdt $num || return
5173 start_ost || error "OST0 start fail"
5174 start_mdt 1 || error "MDT0 start fail"
5175 start_ost2 || error "OST1 start fail"
5177 mount_client $MOUNT || error "mount client fails"
5179 mkdir $DIR/$tdir || error "mkdir fail"
5180 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5181 error "create remote dir succeed"
5183 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5184 rm -rf $DIR/$tdir || error "delete dir fail"
5186 umount_client $MOUNT || error "umount_client failed"
5187 stop_mds || error "MDT0 stop fail"
5188 stop_ost || error "OST0 stop fail"
5189 stop_ost2 || error "OST1 stop fail"
5191 run_test 71b "start MDT1, OST0, MDT0, OST1"
5194 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5195 combined_mgs_mds && skip "needs separate MGS/MDT"
5199 start_ost || error "OST0 start fail"
5200 start_ost2 || error "OST1 start fail"
5201 for num in $(seq 2 $MDSCOUNT); do
5202 start_mdt $num || return
5204 start_mdt 1 || error "MDT0 start fail"
5206 mount_client $MOUNT || error "mount client fails"
5208 mkdir $DIR/$tdir || error "mkdir fail"
5209 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5210 error "create remote dir succeed"
5212 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5213 rm -rf $DIR/$tdir || error "delete dir fail"
5215 umount_client $MOUNT || error "umount_client failed"
5216 stop_mds || error "MDS stop fail"
5217 stop_ost || error "OST0 stop fail"
5218 stop_ost2 || error "OST1 stop fail"
5221 run_test 71c "start OST0, OST1, MDT1, MDT0"
5224 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5225 combined_mgs_mds && skip "needs separate MGS/MDT"
5229 start_ost || error "OST0 start fail"
5230 for num in $(seq 2 $MDSCOUNT); do
5231 start_mdt $num || return
5233 start_mdt 1 || error "MDT0 start fail"
5234 start_ost2 || error "OST1 start fail"
5236 mount_client $MOUNT || error "mount client fails"
5238 mkdir $DIR/$tdir || error "mkdir fail"
5239 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5240 error "create remote dir succeed"
5242 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5243 rm -rf $DIR/$tdir || error "delete dir fail"
5245 umount_client $MOUNT || error "umount_client failed"
5246 stop_mds || error "MDS stop fail"
5247 stop_ost || error "OST0 stop fail"
5248 stop_ost2 || error "OST1 stop fail"
5251 run_test 71d "start OST0, MDT1, MDT0, OST1"
5254 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5255 combined_mgs_mds && skip "needs separate MGS/MDT"
5259 start_ost || error "OST0 start fail"
5260 for num in $(seq 2 $MDSCOUNT); do
5261 start_mdt $num || return
5263 start_ost2 || error "OST1 start fail"
5264 start_mdt 1 || error "MDT0 start fail"
5266 mount_client $MOUNT || error "mount client fails"
5268 mkdir $DIR/$tdir || error "mkdir fail"
5269 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5270 error "create remote dir succeed"
5272 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5273 rm -rf $DIR/$tdir || error "delete dir fail"
5275 umount_client $MOUNT || error "umount_client failed"
5276 stop_mds || error "MDS stop fail"
5277 stop_ost || error "OST0 stop fail"
5278 stop_ost2 || error "OST1 stop fail"
5281 run_test 71e "start OST0, MDT1, OST1, MDT0"
5283 test_72() { #LU-2634
5284 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
5285 skip "ldiskfs only test"
5287 local mdsdev=$(mdsdevname 1)
5288 local ostdev=$(ostdevname 1)
5289 local cmd="$E2FSCK -fnvd $mdsdev"
5293 if combined_mgs_mds; then
5294 add_options='--reformat'
5296 add_options='--reformat --replace'
5299 #tune MDT with "-O extents"
5301 for num in $(seq $MDSCOUNT); do
5302 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
5303 $add_options $(mdsdevname $num) $(mdsvdevname $num) ||
5304 error "add mds $num failed"
5305 do_facet mds${num} "$TUNE2FS -O extents $(mdsdevname $num)" ||
5306 error "$TUNE2FS failed on mds${num}"
5309 add ost1 $(mkfs_opts ost1 $ostdev) $add_options $ostdev ||
5310 error "add $ostdev failed"
5311 start_mds || error "start mds failed"
5312 start_ost || error "start ost failed"
5313 mount_client $MOUNT || error "mount client failed"
5315 #create some short symlinks
5316 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5317 createmany -o $DIR/$tdir/$tfile-%d $fn
5318 echo "create $fn short symlinks"
5319 for i in $(seq -w 1 $fn); do
5320 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
5325 umount_client $MOUNT || error "umount client failed"
5326 stop_mds || error "stop mds failed"
5327 stop_ost || error "stop ost failed"
5330 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
5332 run_test 72 "test fast symlink with extents flag enabled"
5334 test_73() { #LU-3006
5335 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
5336 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
5337 error "1st tunefs failed"
5338 start_mgsmds || error "start mds failed"
5339 start_ost || error "start ost failed"
5340 mount_client $MOUNT || error "mount client failed"
5341 $LCTL get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
5342 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
5343 umount_client $MOUNT || error "umount client failed"
5347 run_test 73 "failnode to update from mountdata properly"
5349 test_75() { # LU-2374
5350 [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] &&
5351 skip "Need MDS version at least 2.4.1"
5354 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
5355 --reformat $(mdsdevname 1) $(mdsvdevname 1)"
5356 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
5357 --reformat $(ostdevname 1) $(ostvdevname 1)"
5359 #check with default parameters
5360 add mds1 $opts_mds || error "add mds1 failed for default params"
5361 add ost1 $opts_ost || error "add ost1 failed for default params"
5363 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
5364 opts_mds=$(echo $opts_mds |
5365 sed -e "s/--index=$index/--index=$index --mdt/")
5366 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
5367 opts_ost=$(echo $opts_ost |
5368 sed -e "s/--index=$index/--index=$index --ost/")
5370 add mds1 $opts_mds || error "add mds1 failed for new params"
5371 add ost1 $opts_ost || error "add ost1 failed for new params"
5372 if ! combined_mgs_mds; then
5373 stop_mgs || error "stop mgs failed"
5378 run_test 75 "The order of --index should be irrelevant"
5381 [[ $(lustre_version_code mgs) -ge $(version_code 2.4.52) ]] ||
5382 skip "Need MDS version at least 2.4.52"
5384 if ! combined_mgs_mds; then
5385 start_mgs || error "start mgs failed"
5388 local MDMB_PARAM="osc.*.max_dirty_mb"
5389 echo "Change MGS params"
5390 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
5392 echo "max_dirty_mb: $MAX_DIRTY_MB"
5393 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB - 10))
5394 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
5395 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
5396 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5397 head -1" $NEW_MAX_DIRTY_MB
5398 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5399 echo "$MAX_DIRTY_MB"
5400 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5401 error "error while apply max_dirty_mb"
5403 echo "Check the value is stored after remount"
5406 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5407 head -1" $NEW_MAX_DIRTY_MB
5408 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5409 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5410 error "max_dirty_mb is not saved after remount"
5412 echo "Change OST params"
5413 CLIENT_PARAM="obdfilter.*.client_cache_count"
5414 local CLIENT_CACHE_COUNT
5415 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5417 echo "client_cache_count: $CLIENT_CACHE_COUNT"
5418 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
5419 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
5420 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
5421 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5422 head -1" $NEW_CLIENT_CACHE_COUNT
5423 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5425 echo "$CLIENT_CACHE_COUNT"
5426 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5427 error "error while apply client_cache_count"
5429 echo "Check the value is stored after remount"
5432 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5433 head -1" $NEW_CLIENT_CACHE_COUNT
5434 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5436 echo "$CLIENT_CACHE_COUNT"
5437 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5438 error "client_cache_count is not saved after remount"
5441 run_test 76a "set permanent params with lctl across mounts"
5443 test_76b() { # LU-4783
5444 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.57) ]] ||
5445 skip "Need MGS version at least 2.5.57"
5448 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
5449 error "start params log failed"
5452 run_test 76b "verify params log setup correctly"
5455 [[ $(lustre_version_code mgs) -ge $(version_code 2.8.54) ]] ||
5456 skip "Need MDS version at least 2.4.52"
5458 local MASK_PARAM="mdd.*.changelog_mask"
5459 echo "Change changelog_mask"
5460 do_facet mgs $LCTL set_param -P $MASK_PARAM=-CLOSE ||
5461 error "Can't change changlog_mask"
5462 wait_update $(facet_host mds) "$LCTL get_param -n $MASK_PARAM |
5465 echo "Check the value is stored after mds remount"
5466 stop_mds || error "Failed to stop MDS"
5467 start_mds || error "Failed to start MDS"
5468 local CHANGELOG_MASK=$(do_facet mgs $LCTL get_param -n $MASK_PARAM)
5469 echo $CHANGELOG_MASK | grep CLOSE > /dev/null &&
5470 error "changelog_mask is not changed"
5474 run_test 76c "verify changelog_mask is applied with lctl set_param -P"
5476 test_76d() { #LU-9399
5479 local xattr_cache="llite.*.xattr_cache"
5480 local cmd="$LCTL get_param -n $xattr_cache | head -1"
5481 local new=$((($(eval $cmd) + 1) % 2))
5483 echo "lctl set_param -P llite.*.xattr_cache=$new"
5484 do_facet mgs $LCTL set_param -P $xattr_cache=$new ||
5485 error "Can't change xattr_cache"
5486 wait_update $HOSTNAME "$cmd" "$new"
5488 echo "Check $xattr_cache on client $MOUNT"
5489 umount_client $MOUNT || error "umount $MOUNT failed"
5490 mount_client $MOUNT || error "mount $MOUNT failed"
5491 [ $(eval $cmd) -eq $new ] ||
5492 error "$xattr_cache != $new on client $MOUNT"
5494 echo "Check $xattr_cache on the new client $MOUNT2"
5495 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
5496 [ $(eval $cmd) -eq $new ] ||
5497 error "$xattr_cache != $new on client $MOUNT2"
5498 umount_client $MOUNT2 || error "umount $MOUNT2 failed"
5502 run_test 76d "verify llite.*.xattr_cache can be set by 'lctl set_param -P' correctly"
5504 test_77() { # LU-3445
5505 local server_version=$(lustre_version_code $SINGLEMDS)
5506 [[ $server_version -ge $(version_code 2.8.55) ]] ||
5507 skip "Need MDS version 2.8.55+ "
5509 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
5510 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
5511 skip_env "mixed loopback and real device not working"
5514 local fs2mdsdev=$(mdsdevname 1_2)
5515 local fs2ostdev=$(ostdevname 1_2)
5516 local fs2mdsvdev=$(mdsvdevname 1_2)
5517 local fs2ostvdev=$(ostvdevname 1_2)
5518 local fsname=test1234
5520 local failnid="$(h2nettype 1.2.3.4),$(h2nettype 4.3.2.1)"
5522 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
5524 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
5525 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
5526 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
5527 error "start fs2mds failed"
5529 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
5530 mgsnid="0.0.0.0@tcp,$mgsnid,$mgsnid:$mgsnid"
5532 add fs2ost --mgsnode=$mgsnid $(mkfs_opts ost1 $fs2ostdev) \
5533 --failnode=$failnid --fsname=$fsname \
5534 --reformat $fs2ostdev $fs2ostvdev ||
5535 error "add fs2ost failed"
5536 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
5538 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
5539 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
5540 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
5543 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
5546 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ||
5547 $(facet_fstype ost1) != ldiskfs ]] &&
5548 skip "ldiskfs only test"
5550 # reformat the Lustre filesystem with a smaller size
5551 local saved_MDSCOUNT=$MDSCOUNT
5552 local saved_MDSSIZE=$MDSSIZE
5553 local saved_OSTCOUNT=$OSTCOUNT
5554 local saved_OSTSIZE=$OSTSIZE
5557 MDSSIZE=$((MDSSIZE - 20000))
5558 OSTSIZE=$((OSTSIZE - 20000))
5559 reformat || error "(1) reformat Lustre filesystem failed"
5560 MDSSIZE=$saved_MDSSIZE
5561 OSTSIZE=$saved_OSTSIZE
5563 # mount the Lustre filesystem
5564 setup_noconfig || error "(2) setup Lustre filesystem failed"
5567 log "create test files"
5572 mkdir $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
5574 for i in $(seq $num_files); do
5575 file=$MOUNT/$tdir/$tfile-$i
5576 dd if=/dev/urandom of=$file count=1 bs=1M || {
5577 $LCTL get_param osc.*.cur*grant*
5578 $LFS df; $LFS df -i;
5579 # stop creating files if there is no more space
5580 if [ ! -e $file ]; then
5581 num_files=$((i - 1))
5585 $LFS getstripe -v $file
5586 local ost_idx=$(LFS getstripe -i $file)
5587 do_facet ost$((ost_idx + 1)) \
5588 $LCTL get_param obdfilter.*.*grant*
5589 error "(4) create $file failed"
5593 # unmount the Lustre filesystem
5594 cleanup || error "(5) cleanup Lustre filesystem failed"
5596 # run e2fsck on the MDT and OST devices
5597 local mds_host=$(facet_active_host $SINGLEMDS)
5598 local ost_host=$(facet_active_host ost1)
5599 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
5600 local ost_dev=$(ostdevname 1)
5602 run_e2fsck $mds_host $mds_dev "-y"
5603 run_e2fsck $ost_host $ost_dev "-y"
5605 # get the original block count of the MDT and OST filesystems
5606 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
5607 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
5609 # expand the MDT and OST filesystems to the device size
5610 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
5611 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
5613 # run e2fsck on the MDT and OST devices again
5614 run_e2fsck $mds_host $mds_dev "-y"
5615 run_e2fsck $ost_host $ost_dev "-y"
5617 # mount the Lustre filesystem
5621 log "check files after expanding the MDT and OST filesystems"
5622 for i in $(seq $num_files); do
5623 file=$MOUNT/$tdir/$tfile-$i
5624 $CHECKSTAT -t file -s 1048576 $file ||
5625 error "(6) checkstat $file failed"
5629 log "create more files after expanding the MDT and OST filesystems"
5630 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
5631 file=$MOUNT/$tdir/$tfile-$i
5632 dd if=/dev/urandom of=$file count=1 bs=1M ||
5633 error "(7) create $file failed"
5636 # unmount the Lustre filesystem
5637 cleanup || error "(8) cleanup Lustre filesystem failed"
5639 # run e2fsck on the MDT and OST devices
5640 run_e2fsck $mds_host $mds_dev "-y"
5641 run_e2fsck $ost_host $ost_dev "-y"
5643 # get the maximum block count of the MDT and OST filesystems
5644 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
5645 local ost_max_blks=$(get_block_count ost1 $ost_dev)
5647 # get the minimum block count of the MDT and OST filesystems
5648 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
5649 grep minimum | sed -e 's/^.*filesystem: //g')
5650 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
5651 grep minimum | sed -e 's/^.*filesystem: //g')
5653 # shrink the MDT and OST filesystems to a smaller size
5657 if [[ $mds_max_blks -gt $mds_min_blks &&
5658 $mds_max_blks -gt $mds_orig_blks ]]; then
5659 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
5660 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
5661 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
5662 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
5663 error "shrink $SINGLEMDS to $new_blks failed"
5667 if [[ $ost_max_blks -gt $ost_min_blks &&
5668 $ost_max_blks -gt $ost_orig_blks ]]; then
5669 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
5670 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
5671 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
5672 run_resize2fs ost1 $ost_dev $new_blks ||
5673 error "shrink ost1 to $new_blks failed"
5677 # check whether the MDT or OST filesystem was shrunk or not
5679 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
5680 reformat || error "(10) reformat Lustre filesystem failed"
5684 # run e2fsck on the MDT and OST devices again
5685 run_e2fsck $mds_host $mds_dev "-y"
5686 run_e2fsck $ost_host $ost_dev "-y"
5688 # mount the Lustre filesystem again
5692 log "check files after shrinking the MDT and OST filesystems"
5693 for i in $(seq $((num_files + 10))); do
5694 file=$MOUNT/$tdir/$tfile-$i
5695 $CHECKSTAT -t file -s 1048576 $file ||
5696 error "(11) checkstat $file failed"
5699 # unmount and reformat the Lustre filesystem
5700 cleanup || error "(12) cleanup Lustre filesystem failed"
5701 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
5703 MDSCOUNT=$saved_MDSCOUNT
5704 OSTCOUNT=$saved_OSTCOUNT
5705 reformat || error "(14) reformat Lustre filesystem failed"
5707 run_test 78 "run resize2fs on MDT and OST filesystems"
5709 test_79() { # LU-4227
5710 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.5.59) ]] ||
5711 skip "Need MDS version at least 2.5.59"
5713 local mdsdev1=$(mdsdevname 1)
5714 local mdsvdev1=$(mdsvdevname 1)
5715 local mdsdev2=$(mdsdevname 2)
5716 local mdsvdev2=$(mdsvdevname 2)
5717 local ostdev1=$(ostdevname 1)
5718 local ostvdev1=$(ostvdevname 1)
5719 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
5720 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
5721 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
5724 # remove --mgs/--mgsnode from mkfs.lustre options
5725 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
5727 mgsnode_opt=$(echo $opts_mds2 |
5728 awk '{ for ( i = 1; i < NF; i++ )
5729 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5730 [ -n $mgsnode_opt ] &&
5731 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
5733 mgsnode_opt=$(echo $opts_ost1 |
5734 awk '{ for ( i = 1; i < NF; i++ )
5735 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5736 [ -n $mgsnode_opt ] &&
5737 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
5739 # -MGS, format a mdt without --mgs option
5740 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
5741 error "Must specify --mgs when formatting mdt combined with mgs"
5743 # +MGS, format a mdt/ost without --mgsnode option
5744 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
5745 > /dev/null || error "start mds1 failed"
5746 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
5747 error "Must specify --mgsnode when formatting a mdt"
5748 add ost1 $opts_ost1 $ostdev1 $ostvdev2 &&
5749 error "Must specify --mgsnode when formatting an ost"
5753 run_test 79 "format MDT/OST without mgs option (should return errors)"
5756 start_mds || error "Failed to start MDT"
5757 start_ost || error "Failed to start OST1"
5758 uuid=$(do_facet ost1 $LCTL get_param -n mgc.*.uuid)
5759 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
5760 do_facet ost1 "$LCTL set_param fail_val=10 fail_loc=0x906"
5761 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x906"
5762 do_facet mgs "$LCTL set_param -n mgs/MGS/evict_client $uuid"
5764 start_ost2 || error "Failed to start OST2"
5766 do_facet ost1 "$LCTL set_param fail_loc=0"
5769 run_test 80 "mgc import reconnect race"
5771 #Save the original values of $OSTCOUNT and $OSTINDEX$i.
5773 local new_ostcount=$1
5774 saved_ostcount=$OSTCOUNT
5775 OSTCOUNT=$new_ostcount
5779 for i in $(seq $OSTCOUNT); do
5781 eval saved_ostindex$i=${!index}
5786 # Restore the original values of $OSTCOUNT and $OSTINDEX$i.
5787 restore_ostindex() {
5792 for i in $(seq $OSTCOUNT); do
5793 index=saved_ostindex$i
5794 eval OSTINDEX$i=${!index}
5796 OSTCOUNT=$saved_ostcount
5801 # The main purpose of this test is to ensure the OST_INDEX_LIST functions as
5802 # expected. This test uses OST_INDEX_LIST to format OSTs with a randomly
5803 # assigned index and ensures we can mount such a formatted file system
5804 test_81() { # LU-4665
5805 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5806 skip "Need MDS version at least 2.6.54"
5807 [[ $OSTCOUNT -ge 3 ]] || skip_env "needs >= 3 OSTs"
5811 # Each time RANDOM is referenced, a random integer between 0 and 32767
5814 local saved_ostindex1=$OSTINDEX1
5815 for i in 65535 $((RANDOM + 65536)); do
5816 echo -e "\nFormat ost1 with --index=$i, should fail"
5818 if add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat \
5819 $(ostdevname 1) $(ostvdevname 1); then
5820 OSTINDEX1=$saved_ostindex1
5821 error "format ost1 with --index=$i should fail"
5824 OSTINDEX1=$saved_ostindex1
5828 # Format OSTs with random sparse indices.
5829 trap "restore_ostindex" EXIT
5830 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices"
5831 OST_INDEX_LIST=[0,$((RANDOM * 2 % 65533 + 1)),65534] formatall
5833 # Setup and check Lustre filesystem.
5834 start_mgsmds || error "start_mgsmds failed"
5835 for i in $(seq $OSTCOUNT); do
5836 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5837 error "start ost$i failed"
5840 mount_client $MOUNT || error "mount client $MOUNT failed"
5841 check_mount || error "check client $MOUNT failed"
5844 local max_easize=$($LCTL get_param -n llite.*.max_easize)
5845 [[ $max_easize -eq 128 ]] ||
5846 error "max_easize is $max_easize, should be 128 bytes"
5850 run_test 81 "sparse OST indexing"
5852 # Here we exercise the stripe placement functionality on a file system that
5853 # has formatted the OST with a random index. With the file system the following
5854 # functionality is tested:
5856 # 1. Creating a new file with a specific stripe layout.
5858 # 2. Modifiy a existing empty file with a specific stripe layout.
5860 # 3. Ensure we fail to set the stripe layout of a file that already has one.
5862 # 4. If ost-index is defined we need to ensure it is the first entry in the
5863 # ost index list returned by lfs getstripe.
5865 # 5. Lastly ensure this functionality fails with directories.
5866 test_82a() { # LU-4665
5867 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5868 skip "Need MDS version at least 2.6.54"
5869 [[ $OSTCOUNT -ge 3 ]] || skip_env "needs >= 3 OSTs"
5875 # Format OSTs with random sparse indices.
5879 local LOV_V1_INSANE_STRIPE_COUNT=65532
5880 for i in $(seq $OSTCOUNT); do
5881 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
5882 ost_indices+=" $index"
5884 ost_indices=$(comma_list $ost_indices)
5886 stack_trap "restore_ostindex" EXIT
5887 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5888 OST_INDEX_LIST=[$ost_indices] formatall
5890 # Setup Lustre filesystem.
5891 start_mgsmds || error "start_mgsmds failed"
5892 for i in $(seq $OSTCOUNT); do
5893 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5894 error "start ost$i failed"
5897 # Collect debug information - start of test
5898 do_nodes $(comma_list $(mdts_nodes)) \
5899 $LCTL get_param osc.*.prealloc_*_id
5901 mount_client $MOUNT || error "mount client $MOUNT failed"
5904 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5905 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5907 stack_trap "do_nodes $(comma_list $(mdts_nodes)) \
5908 $LCTL get_param osc.*.prealloc_*_id" EXIT
5910 # 1. If the file does not exist, new file will be created
5911 # with specified OSTs.
5912 local file=$DIR/$tdir/$tfile-1
5913 local cmd="$LFS setstripe -o $ost_indices $file"
5915 eval $cmd || error "$cmd failed"
5916 check_stripe_count $file $OSTCOUNT
5917 check_obdidx $file $ost_indices
5918 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5919 error "write $file failed"
5921 # 2. If the file already exists and is an empty file, the file
5922 # will be attached with specified layout.
5923 file=$DIR/$tdir/$tfile-2
5924 mcreate $file || error "mcreate $file failed"
5925 cmd="$LFS setstripe -o $ost_indices $file"
5927 eval $cmd || error "$cmd failed"
5928 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5929 error "write $file failed"
5930 check_stripe_count $file $OSTCOUNT
5931 check_obdidx $file $ost_indices
5933 # 3. If the file already has a valid layout attached, the command
5934 # should fail with EBUSY.
5936 eval $cmd && error "stripe is already set on $file, $cmd should fail"
5938 # 4. If [--stripe-index|-i <start_ost_idx>] is used, the index must
5939 # be in the OST indices list.
5940 local start_ost_idx=${ost_indices##*,}
5941 file=$DIR/$tdir/$tfile-3
5942 cmd="$LFS setstripe -o $ost_indices -i $start_ost_idx $file"
5944 eval $cmd || error "$cmd failed"
5945 check_stripe_count $file $OSTCOUNT
5946 check_obdidx $file $ost_indices
5947 check_start_ost_idx $file $start_ost_idx
5949 file=$DIR/$tdir/$tfile-4
5950 cmd="$LFS setstripe"
5951 cmd+=" -o $(exclude_items_from_list $ost_indices $start_ost_idx)"
5952 cmd+=" -i $start_ost_idx $file"
5954 eval $cmd && error "index $start_ost_idx should be in $ost_indices"
5956 # 5. Specifying OST indices for directory should succeed.
5957 local dir=$DIR/$tdir/$tdir
5958 mkdir $dir || error "mkdir $dir failed"
5959 cmd="$LFS setstripe -o $ost_indices $dir"
5960 if [[ $(lustre_version_code $SINGLEMDS) -gt $(version_code 2.11.53) &&
5961 $(lustre_version_code client -gt $(version_code 2.11.53)) ]]; then
5963 eval $cmd || error "unable to specify OST indices on directory"
5965 echo "need MDS+client version at least 2.11.53"
5970 run_test 82a "specify OSTs for file (succeed) or directory (succeed)"
5975 # Remove OSTs from a pool and destroy the pool.
5976 destroy_pool $ost_pool || true
5978 if ! combined_mgs_mds ; then
5984 # Test 82b is run to ensure that if the user supplies a pool with a specific
5985 # stripe layout that it behaves proprerly. It should fail in the case that
5986 # the supplied OST index list points to OSTs not contained in the user
5988 test_82b() { # LU-4665
5989 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5990 skip "Need MDS version at least 2.6.54"
5991 [[ $OSTCOUNT -ge 4 ]] || skip_env "needs >= 4 OSTs"
5997 # Format OSTs with random sparse indices.
6001 local LOV_V1_INSANE_STRIPE_COUNT=65532
6002 for i in $(seq $OSTCOUNT); do
6003 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
6004 ost_indices+=" $index"
6006 ost_indices=$(comma_list $ost_indices)
6008 trap "restore_ostindex" EXIT
6009 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
6010 OST_INDEX_LIST=[$ost_indices] formatall
6012 # Setup Lustre filesystem.
6013 start_mgsmds || error "start_mgsmds failed"
6014 for i in $(seq $OSTCOUNT); do
6015 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6016 error "start ost$i failed"
6019 mount_client $MOUNT || error "mount client $MOUNT failed"
6020 if ! combined_mgs_mds ; then
6025 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
6026 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6028 # Create a new pool and add OSTs into it.
6029 local ost_pool=$FSNAME.$TESTNAME
6030 create_pool $ost_pool || error "create OST pool $ost_pool failed"
6033 trap "cleanup_82b" EXIT
6035 local ost_idx_in_list=${ost_indices##*,}
6036 local ost_idx_in_pool=$(exclude_items_from_list $ost_indices \
6039 local ost_targets="$FSNAME-OST["
6040 for i in ${ost_idx_in_pool//,/ }; do
6041 ost_targets=$ost_targets$(printf "%04x," $i)
6043 ost_targets="${ost_targets%,}]"
6045 local ost_targets_uuid=$(for i in ${ost_idx_in_pool//,/ }; \
6046 do printf "$FSNAME-OST%04x_UUID\n" $i; done |
6047 sort -u | tr '\n' ' ')
6049 local cmd="$LCTL pool_add $ost_pool $ost_targets"
6050 do_facet mgs $cmd || error "$cmd failed"
6051 wait_update $HOSTNAME "$LCTL get_param -n lov.$FSNAME-*.pools.$TESTNAME|
6052 sort -u | tr '\n' ' ' " "$ost_targets_uuid" ||
6053 error "wait_update $ost_pool failed"
6054 wait_update_facet $SINGLEMDS "$LCTL pool_list $ost_pool | wc -l" 4 ||
6055 error "wait_update pool_list $ost_pool failed"
6057 # If [--pool|-p <pool_name>] is set with [--ost-list|-o <ost_indices>],
6058 # then the OSTs must be the members of the pool.
6059 local file=$DIR/$tdir/$tfile
6060 cmd="$LFS setstripe -p $ost_pool -o $ost_idx_in_list $file"
6062 eval $cmd && error "OST with index $ost_idx_in_list should be" \
6063 "in OST pool $ost_pool"
6065 # Only select OST $ost_idx_in_list from $ost_pool for file.
6066 ost_idx_in_list=${ost_idx_in_pool#*,}
6067 cmd="$LFS setstripe -p $ost_pool -o $ost_idx_in_list $file"
6069 eval $cmd || error "$cmd failed"
6070 cmd="$LFS getstripe $file"
6072 eval $cmd || error "$cmd failed"
6073 check_stripe_count $file 2
6074 check_obdidx $file $ost_idx_in_list
6075 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6076 error "write $file failed"
6080 run_test 82b "specify OSTs for file with --pool and --ost-list options"
6083 [[ $(lustre_version_code ost1) -ge $(version_code 2.6.91) ]] ||
6084 skip "Need OST version at least 2.6.91"
6085 if [ $(facet_fstype ost1) != ldiskfs ]; then
6086 skip "ldiskfs only test"
6095 ostmnt=$(facet_mntpt ost1)
6096 fstype=$(facet_fstype ost1)
6098 # Mount the OST as an ldiskfs filesystem.
6099 log "mount the OST $dev as a $fstype filesystem"
6100 add ost1 $(mkfs_opts ost1 $dev) $FSTYPE_OPT \
6101 --reformat $dev > /dev/null ||
6102 error "format ost1 error"
6104 if ! test -b $dev; then
6105 mnt_opts=$(csa_add "$OST_MOUNT_OPTS" -o loop)
6107 echo "mnt_opts $mnt_opts"
6108 do_facet ost1 mount -t $fstype $dev \
6110 # Run llverfs on the mounted ldiskfs filesystem.
6111 # It is needed to get ENOSPACE.
6112 log "run llverfs in partial mode on the OST $fstype $ostmnt"
6113 do_rpc_nodes $(facet_host ost1) run_llverfs $ostmnt -vpl \
6114 "no" || error "run_llverfs error on $fstype"
6117 log "unmount the OST $dev"
6120 # Delete file IO_scrub. Later osd_scrub_setup will try to
6121 # create "IO_scrub" but will get ENOSPACE.
6123 echo "start ost1 service on `facet_active_host ost1`"
6124 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
6127 err=$(do_facet ost1 dmesg | grep "VFS: Busy inodes after unmount of")
6128 echo "string err $err"
6129 [ -z "$err" ] || error $err
6132 run_test 83 "ENOSPACE on OST doesn't cause message VFS: \
6133 Busy inodes after unmount ..."
6136 local facet=$SINGLEMDS
6137 local num=$(echo $facet | tr -d "mds")
6138 local dev=$(mdsdevname $num)
6139 local time_min=$(recovery_time_min)
6140 local recovery_duration
6141 local completed_clients
6142 local correct_clients
6145 echo "start mds service on $(facet_active_host $facet)"
6147 "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
6148 error "start MDS failed"
6150 start_ost || error "start OST0000 failed"
6151 start_ost2 || error "start OST0001 failed"
6153 echo "recovery_time=$time_min, timeout=$TIMEOUT, wrap_up=$wrap_up"
6155 mount_client $MOUNT1 || error "mount $MOUNT1 failed"
6156 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
6157 # make sure new superblock labels are sync'd before disabling writes
6161 replay_barrier $SINGLEMDS
6162 createmany -o $DIR1/$tfile-%d 1000
6164 # We need to catch the end of recovery window to extend it.
6165 # Skip 5 requests and add delay to request handling.
6166 #define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP
6167 do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5"
6169 facet_failover --fsck $SINGLEMDS || error "failover: $?"
6172 echo "recovery status"
6173 do_facet $SINGLEMDS \
6174 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status"
6176 recovery_duration=$(do_facet $SINGLEMDS \
6177 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
6178 awk '/recovery_duration/ { print $2 }')
6179 (( $recovery_duration > $time_min + $wrap_up )) &&
6180 error "recovery_duration > recovery_time_hard + wrap up"
6181 completed_clients=$(do_facet $SINGLEMDS \
6182 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
6183 awk '/completed_clients/ { print $2 }')
6185 correct_clients="$MDSCOUNT/$((MDSCOUNT+1))"
6186 [ "$completed_clients" = "${correct_clients}" ] ||
6187 error "$completed_clients != $correct_clients"
6189 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
6190 umount_client $MOUNT1
6191 umount_client $MOUNT2
6197 run_test 84 "check recovery_hard_time"
6200 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.55) ]] ||
6201 skip "Need OST version at least 2.7.55"
6202 ##define OBD_FAIL_OSD_OST_EA_FID_SET 0x197
6203 do_facet ost1 "lctl set_param fail_loc=0x197"
6207 run_test 85 "osd_ost init: fail ea_fid_set"
6212 # ost1 has already registered to the MGS before the reformat.
6213 # So after reformatting it with option "-G", it could not be
6214 # mounted to the MGS. Cleanup the system for subsequent tests.
6219 local server_version=$(lustre_version_code $SINGLEMDS)
6220 [ "$(facet_fstype ost1)" = "zfs" ] &&
6221 skip "LU-6442: no such mkfs params for ZFS OSTs"
6222 [[ $server_version -ge $(version_code 2.7.56) ]] ||
6223 skip "Need server version newer than 2.7.55"
6225 local OST_OPTS="$(mkfs_opts ost1 $(ostdevname 1)) \
6226 --reformat $(ostdevname 1) $(ostvdevname 1)"
6229 local OLDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6230 awk '/Flex block group size: / { print $NF; exit; }')
6233 if [[ ${!opts} != *mkfsoptions* ]]; then
6234 eval opts=\"${!opts} \
6235 --mkfsoptions='\\\"-O flex_bg -G $NEWSIZE\\\"'\"
6237 val=${!opts//--mkfsoptions=\\\"/ \
6238 --mkfsoptions=\\\"-O flex_bg -G $NEWSIZE }
6242 echo "params: $opts"
6244 trap cleanup_86 EXIT ERR
6247 add ost1 $opts || error "add ost1 failed with new params"
6249 local FOUNDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6250 awk '/Flex block group size: / { print $NF; exit; }')
6252 [[ $FOUNDSIZE == $NEWSIZE ]] ||
6253 error "Flex block group size: $FOUNDSIZE, expected: $NEWSIZE"
6257 run_test 86 "Replacing mkfs.lustre -G option"
6259 test_87() { #LU-6544
6260 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.9.51) ]] ||
6261 skip "Need MDS version at least 2.9.51"
6262 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ]] &&
6263 skip "ldiskfs only test"
6264 [[ $OSTCOUNT -gt 59 ]] &&
6265 skip "Ignore wide striping situation"
6267 local mdsdev=$(mdsdevname 1)
6268 local mdsvdev=$(mdsvdevname 1)
6269 local file=$DIR/$tfile
6270 local mntpt=$(facet_mntpt $SINGLEMDS)
6271 local used_xattr_blk=0
6272 local inode_size=${1:-1024}
6274 local xtest="trusted.test"
6278 local stripe_cnt=$(($OSTCOUNT + 2))
6280 #Please see ldiskfs_make_lustre() for MDT inode size calculation
6281 if [ $stripe_cnt -gt 16 ]; then
6284 left_size=$(expr $inode_size - \
6287 32 - 40 \* 3 - 32 \* 3 - $stripe_cnt \* 24 - 16 - 3 - \
6289 24 - 18 - $(expr length $tfile) - 16 - 4)
6290 if [ $left_size -le 0 ]; then
6291 echo "No space($left_size) is expected in inode."
6292 echo "Try 1-byte xattr instead to verify this."
6295 echo "Estimate: at most $left_size-byte space left in inode."
6301 add mds1 $(mkfs_opts mds1 ${mdsdev}) --stripe-count-hint=$stripe_cnt \
6302 --reformat $mdsdev $mdsvdev || error "add mds1 failed"
6303 start_mdt 1 > /dev/null || error "start mdt1 failed"
6304 for i in $(seq $OSTCOUNT); do
6305 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS > /dev/null ||
6306 error "start ost$i failed"
6308 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
6309 check_mount || error "check client $MOUNT failed"
6312 $LFS setstripe -E 1M -S 1M -c 1 -E 64M -c 1 -E -1 -c -1 $file ||
6313 error "Create file with 3 components failed"
6314 $TRUNCATE $file $((1024*1024*64+1)) || error "truncate file failed"
6315 i=$($LFS getstripe -I3 -c $file) || error "get 3rd stripe count failed"
6316 if [ $i -ne $OSTCOUNT ]; then
6317 left_size=$(expr $left_size + $(expr $OSTCOUNT - $i) \* 24)
6318 echo -n "Since only $i out $OSTCOUNT OSTs are used, "
6319 echo -n "the expected left space is changed to "
6320 echo "$left_size bytes at most."
6322 value=$(generate_string $left_size)
6323 setfattr -n $xtest -v $value $file
6324 orig=$(get_xattr_value $xtest $file)
6325 [[ "$orig" != "$value" ]] && error "$xtest changed"
6327 #Verify if inode has some expected space left
6328 umount $MOUNT > /dev/null || error "umount $MOUNT failed"
6329 stop_mdt 1 > /dev/null || error "stop mdt1 failed"
6330 mount_ldiskfs $SINGLEMDS || error "mount -t ldiskfs $SINGLEMDS failed"
6332 do_facet $SINGLEMDS ls -sal $mntpt/ROOT/$tfile
6333 used_xattr_blk=$(do_facet $SINGLEMDS ls -s $mntpt/ROOT/$tfile |
6335 [[ $used_xattr_blk -eq 0 ]] &&
6336 error "Please check MDS inode size calculation: \
6337 more than $left_size-byte space left in inode."
6338 echo "Verified: at most $left_size-byte space left in inode."
6342 run_test 87 "check if MDT inode can hold EAs with N stripes properly"
6345 [ "$(facet_fstype mds1)" == "zfs" ] &&
6346 skip "LU-6662: no implementation for ZFS"
6350 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
6351 --reformat $(mdsdevname 1) || error "add mds1 failed"
6353 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
6354 grep -e \".*opts:.*errors=remount-ro.*\"" ||
6355 error "default mount options is missing"
6357 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
6358 --mountfsoptions="user_xattr,errors=panic" \
6359 --reformat $(mdsdevname 1) || error "add mds1 failed"
6361 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
6362 grep -e \".*opts:.*errors=panic.*\"" ||
6363 error "user can't override default mount options"
6365 run_test 88 "check the default mount options can be overridden"
6367 test_89() { # LU-7131
6368 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.9.54) ]] ||
6369 skip "Need MDT version at least 2.9.54"
6371 local key=failover.node
6372 local val1=192.0.2.254@tcp0 # Reserved IPs, see RFC 5735
6373 local val2=192.0.2.255@tcp0
6374 local mdsdev=$(mdsdevname 1)
6379 [ $(facet_fstype mds1) == zfs ] && import_zpool mds1
6380 # Check that parameters are added correctly
6381 echo "tunefs --param $key=$val1"
6382 do_facet mds "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
6383 error "tunefs --param $key=$val1 failed"
6384 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6385 error "tunefs --dryrun failed"
6386 params=${params##*Parameters:}
6387 params=${params%%exiting*}
6388 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val1) = "1" ] ||
6389 error "on-disk parameter not added correctly via tunefs"
6391 # Check that parameters replace existing instances when added
6392 echo "tunefs --param $key=$val2"
6393 do_facet mds "$TUNEFS --param $key=$val2 $mdsdev >/dev/null" ||
6394 error "tunefs --param $key=$val2 failed"
6395 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6396 error "tunefs --dryrun failed"
6397 params=${params##*Parameters:}
6398 params=${params%%exiting*}
6399 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "1" ] ||
6400 error "on-disk parameter not replaced via tunefs"
6401 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val2) = "1" ] ||
6402 error "on-disk parameter not replaced correctly via tunefs"
6404 # Check that a parameter is erased properly
6405 echo "tunefs --erase-param $key"
6406 do_facet mds "$TUNEFS --erase-param $key $mdsdev >/dev/null" ||
6407 error "tunefs --erase-param $key failed"
6408 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6409 error "tunefs --dryrun failed"
6410 params=${params##*Parameters:}
6411 params=${params%%exiting*}
6412 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "0" ] ||
6413 error "on-disk parameter not erased correctly via tunefs"
6415 # Check that all the parameters are erased
6416 echo "tunefs --erase-params"
6417 do_facet mds "$TUNEFS --erase-params $mdsdev >/dev/null" ||
6418 error "tunefs --erase-params failed"
6419 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6420 error "tunefs --dryrun failed"
6421 params=${params##*Parameters:}
6422 params=${params%%exiting*}
6424 error "all on-disk parameters not erased correctly via tunefs"
6426 # Check the order of options --erase-params and --param
6427 echo "tunefs --param $key=$val1 --erase-params"
6429 "$TUNEFS --param $key=$val1 --erase-params $mdsdev >/dev/null"||
6430 error "tunefs --param $key=$val1 --erase-params failed"
6431 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6432 error "tunefs --dryrun failed"
6433 params=${params##*Parameters:}
6434 params=${params%%exiting*}
6435 [ $(echo $params | tr ' ' '\n') == "$key=$val1" ] ||
6436 error "on-disk param not added correctly with --erase-params"
6440 run_test 89 "check tunefs --param and --erase-param{s} options"
6443 # $2 (optional) value of max_mod_rpcs_in_flight to set
6444 check_max_mod_rpcs_in_flight() {
6452 idx=$(printf "%04x" $($LFS getdirstripe -i $dir))
6453 facet="mds$((0x$idx + 1))"
6455 if [ -z "$mmr" ]; then
6456 # get value of max_mod_rcps_in_flight
6457 mmr=$($LCTL get_param -n \
6458 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) ||
6459 error "Unable to get max_mod_rpcs_in_flight"
6460 echo "max_mod_rcps_in_flight is $mmr"
6462 # set value of max_mod_rpcs_in_flight
6464 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight=$mmr ||
6465 error "Unable to set max_mod_rpcs_in_flight to $mmr"
6466 echo "max_mod_rpcs_in_flight set to $mmr"
6469 # create mmr+1 files
6470 echo "creating $((mmr + 1)) files ..."
6472 for i in $(seq $((mmr + 1))); do
6478 # consumes mmr-1 modify RPC slots
6479 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6480 # drop requests on MDT so that RPC slots are consumed
6481 # during all the request resend interval
6482 do_facet $facet "$LCTL set_param fail_loc=0x159"
6483 echo "launch $((mmr - 1)) chmod in parallel ..."
6484 for i in $(seq $((mmr - 1))); do
6485 chmod 0600 $dir/file-$i &
6489 # send one additional modify RPC
6490 do_facet $facet "$LCTL set_param fail_loc=0"
6491 echo "launch 1 additional chmod in parallel ..."
6492 chmod 0600 $dir/file-$mmr &
6495 # check this additional modify RPC get a modify RPC slot
6496 # and succeed its operation
6497 checkstat -vp 0600 $dir/file-$mmr ||
6498 error "Unable to send $mmr modify RPCs in parallel"
6503 # consumes mmr modify RPC slots
6504 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6505 # drop requests on MDT so that RPC slots are consumed
6506 # during all the request resend interval
6507 do_facet $facet "$LCTL set_param fail_loc=0x159"
6508 echo "launch $mmr chmod in parallel ..."
6509 for i in $(seq $mmr); do
6510 chmod 0666 $dir/file-$i &
6514 # send one additional modify RPC
6515 do_facet $facet "$LCTL set_param fail_loc=0"
6516 echo "launch 1 additional chmod in parallel ..."
6517 chmod 0666 $dir/file-$((mmr + 1)) &
6520 # check this additional modify RPC blocked getting a modify RPC slot
6521 checkstat -vp 0644 $dir/file-$((mmr + 1)) ||
6522 error "Unexpectedly send $(($mmr + 1)) modify RPCs in parallel"
6528 if ! combined_mgs_mds ; then
6533 [[ $($LCTL get_param mdc.*.import |
6534 grep "connect_flags:.*multi_mod_rpc") ]] ||
6535 skip "Need MDC with 'multi_mod_rpcs' feature"
6537 # check default value
6538 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6539 check_max_mod_rpcs_in_flight $DIR/$tdir
6543 run_test 90a "check max_mod_rpcs_in_flight is enforced"
6553 [[ $($LCTL get_param mdc.*.import |
6554 grep "connect_flags:.*multi_mod_rpc") ]] ||
6555 skip "Need MDC with 'multi_mod_rpcs' feature"
6558 # update max_mod_rpcs_in_flight
6559 $LFS mkdir -c1 $DIR/${tdir}1 || error "mkdir $DIR/${tdir}1 failed"
6560 check_max_mod_rpcs_in_flight $DIR/${tdir}1 1
6563 # check client is able to send multiple modify RPCs in paralell
6564 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
6565 grep -c "multi_mod_rpcs")
6566 if [ "$tmp" -ne $MDSCOUNT ]; then
6568 skip "Client not able to send multiple modify RPCs in parallel"
6571 # update max_mod_rpcs_in_flight
6572 $LFS mkdir -c1 $DIR/${tdir}2 || error "mkdir $DIR/${tdir}2 failed"
6573 check_max_mod_rpcs_in_flight $DIR/${tdir}2 5
6576 $LFS mkdir -c1 $DIR/${tdir}3 || error "mkdir $DIR/${tdir}3 failed"
6577 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/${tdir}3))
6578 facet="mds$((0x$idx + 1))"
6580 # save MDT max_mod_rpcs_per_client
6581 mmrpc=$(do_facet $facet \
6582 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
6584 # update max_mod_rpcs_in_flight
6585 umount_client $MOUNT
6587 "echo 16 > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6589 $LCTL set_param mdc.$FSNAME-MDT$idx-mdc-*.max_rpcs_in_flight=17
6590 check_max_mod_rpcs_in_flight $DIR/${tdir}3 16
6592 # restore MDT max_mod_rpcs_per_client initial value
6594 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6596 rm -rf $DIR/${tdir}?
6599 run_test 90b "check max_mod_rpcs_in_flight is enforced after update"
6608 [[ $($LCTL get_param mdc.*.import |
6609 grep "connect_flags:.*multi_mod_rpc") ]] ||
6610 skip "Need MDC with 'multi_mod_rpcs' feature"
6612 # check client is able to send multiple modify RPCs in paralell
6613 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
6614 grep -c "multi_mod_rpcs")
6615 if [ "$tmp" -ne $MDSCOUNT ]; then
6617 skip "Client not able to send multiple modify RPCs in parallel"
6620 # get max_rpcs_in_flight value
6621 mrif=$($LCTL get_param -n mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
6622 echo "max_rpcs_in_flight is $mrif"
6624 # get MDT max_mod_rpcs_per_client
6625 mmrpc=$(do_facet mds1 \
6626 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
6627 echo "max_mod_rpcs_per_client is $mmrpc"
6630 # attempt to set max_mod_rpcs_in_flight to max_rpcs_in_flight value
6631 # prerequisite: set max_mod_rpcs_per_client to max_rpcs_in_flight value
6632 umount_client $MOUNT
6634 "echo $mrif > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6638 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif &&
6639 error "set max_mod_rpcs_in_flight to $mrif should fail"
6641 umount_client $MOUNT
6643 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6647 # attempt to set max_mod_rpcs_in_flight to max_mod_rpcs_per_client+1
6648 # prerequisite: set max_rpcs_in_flight to max_mod_rpcs_per_client+2
6650 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$((mmrpc + 2))
6653 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$((mmrpc + 1)) &&
6654 error "set max_mod_rpcs_in_flight to $((mmrpc + 1)) should fail"
6658 run_test 90c "check max_mod_rpcs_in_flight update limits"
6669 [[ $($LCTL get_param mdc.*.import |
6670 grep "connect_flags:.*multi_mod_rpc") ]] ||
6671 skip "Need MDC with 'multi_mod_rpcs' feature"
6673 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6674 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
6675 facet="mds$((0x$idx + 1))"
6677 # check client version supports multislots
6678 tmp=$($LCTL get_param -N \
6679 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
6680 if [ -z "$tmp" ]; then
6682 skip "Client does not support multiple modify RPCs in flight"
6685 # get current value of max_mod_rcps_in_flight
6686 mmr=$($LCTL get_param -n \
6687 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
6688 echo "max_mod_rcps_in_flight is $mmr"
6691 echo "creating $mmr files ..."
6693 for i in $(seq $mmr); do
6694 touch $DIR/$tdir/file-$i
6697 # prepare for close RPC
6698 multiop_bg_pause $DIR/$tdir/file-close O_c
6701 # consumes mmr modify RPC slots
6702 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6703 # drop requests on MDT so that RPC slots are consumed
6704 # during all the request resend interval
6705 do_facet $facet "$LCTL set_param fail_loc=0x159"
6706 echo "launch $mmr chmod in parallel ..."
6707 for i in $(seq $mmr); do
6708 chmod 0600 $DIR/$tdir/file-$i &
6711 # send one additional close RPC
6712 do_facet $facet "$LCTL set_param fail_loc=0"
6713 echo "launch 1 additional close in parallel ..."
6715 cancel_lru_locks mdc
6718 # check this additional close RPC get a modify RPC slot
6719 # and multiop process completed
6720 [ -d /proc/$pid ] &&
6721 error "Unable to send the additional close RPC in parallel"
6726 run_test 90d "check one close RPC is allowed above max_mod_rpcs_in_flight"
6728 check_uuid_on_ost() {
6730 do_facet ost1 "$LCTL get_param obdfilter.${FSNAME}*.exports.'$nid'.uuid"
6733 check_uuid_on_mdt() {
6735 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
6743 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.63) ]] ||
6744 skip "Need OST version at least 2.7.63"
6745 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.63) ]] ||
6746 skip "Need MDT version at least 2.7.63"
6748 start_mds || error "MDS start failed"
6749 start_ost || error "unable to start OST"
6750 mount_client $MOUNT || error "client start failed"
6751 check_mount || error "check_mount failed"
6754 nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g")
6758 uuid=$(get_client_uuid $MOUNT)
6760 echo "list nids on mdt:"
6761 do_facet $SINGLEMDS "$LCTL list_param mdt.${FSNAME}*.exports.*"
6762 echo "uuid from $nid:"
6763 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
6765 found=$(check_uuid_on_mdt $nid | grep $uuid)
6766 [ -z "$found" ] && error "can't find $uuid $nid on MDT"
6767 found=$(check_uuid_on_ost $nid | grep $uuid)
6768 [ -z "$found" ] && error "can't find $uuid $nid on OST"
6770 # umount the client so it won't reconnect
6771 manual_umount_client --force || error "failed to umount $?"
6772 # shouldn't disappear on MDS after forced umount
6773 found=$(check_uuid_on_mdt $nid | grep $uuid)
6774 [ -z "$found" ] && error "can't find $uuid $nid"
6777 do_facet $SINGLEMDS \
6778 "$LCTL set_param -n mdt.${mds1_svc}.evict_client nid:$nid"
6780 found=$(check_uuid_on_mdt $nid | grep $uuid)
6781 [ -n "$found" ] && error "found $uuid $nid on MDT"
6782 found=$(check_uuid_on_ost $nid | grep $uuid)
6783 [ -n "$found" ] && error "found $uuid $nid on OST"
6785 # check it didn't reconnect (being umounted)
6786 sleep $((TIMEOUT+1))
6787 found=$(check_uuid_on_mdt $nid | grep $uuid)
6788 [ -n "$found" ] && error "found $uuid $nid on MDT"
6789 found=$(check_uuid_on_ost $nid | grep $uuid)
6790 [ -n "$found" ] && error "found $uuid $nid on OST"
6794 run_test 91 "evict-by-nid support"
6796 generate_ldev_conf() {
6797 # generate an ldev.conf file
6798 local ldevconfpath=$1
6800 local fsldevformat=""
6803 fstype=$(facet_fstype mgs)
6804 if [ "$fstype" == "zfs" ]; then
6805 fsldevformat="$fstype:"
6810 printf "%s\t-\t%s-MGS0000\t%s%s\n" \
6814 $(mgsdevname) > $ldevconfpath
6816 local mdsfo_host=$mdsfailover_HOST;
6817 if [ -z "$mdsfo_host" ]; then
6821 for num in $(seq $MDSCOUNT); do
6822 fstype=$(facet_fstype mds$num)
6823 if [ "$fstype" == "zfs" ]; then
6824 fsldevformat="$fstype:"
6829 printf "%s\t%s\t%s-MDT%04d\t%s%s\n" \
6835 $(mdsdevname $num) >> $ldevconfpath
6838 local ostfo_host=$ostfailover_HOST;
6839 if [ -z "$ostfo_host" ]; then
6843 for num in $(seq $OSTCOUNT); do
6844 fstype=$(facet_fstype ost$num)
6845 if [ "$fstype" == "zfs" ]; then
6846 fsldevformat="$fstype:"
6851 printf "%s\t%s\t%s-OST%04d\t%s%s\n" \
6857 $(ostdevname $num) >> $ldevconfpath
6860 echo "----- $ldevconfpath -----"
6862 echo "--- END $ldevconfpath ---"
6867 # generate a nids file (mapping between hostname to nid)
6868 # looks like we only have the MGS nid available to us
6869 # so just echo that to a file
6871 echo -e "${mgs_HOST}\t${MGSNID}" > $nidspath
6873 echo "----- $nidspath -----"
6875 echo "--- END $nidspath ---"
6878 compare_ldev_output() {
6882 sort $expected_output -o $expected_output
6883 sort $ldev_output -o $ldev_output
6885 echo "-- START OF LDEV OUTPUT --"
6887 echo "--- END OF LDEV OUTPUT ---"
6889 echo "-- START OF EXPECTED OUTPUT --"
6890 cat $expected_output
6891 echo "--- END OF EXPECTED OUTPUT ---"
6893 diff $expected_output $ldev_output
6898 if [ -z "$LDEV" ]; then
6899 error "ldev is missing!"
6902 local LDEVCONFPATH=$TMP/ldev.conf
6903 local NIDSPATH=$TMP/nids
6905 echo "Host is $(hostname)"
6907 generate_ldev_conf $LDEVCONFPATH
6908 generate_nids $NIDSPATH
6910 # echo the mgs nid and compare it to environment variable MGSNID
6911 # also, ldev.conf and nids is a server side thing, use the OSS
6914 output=$($LDEV -c $LDEVCONFPATH -H $ost_HOST -n $NIDSPATH echo %m)
6916 echo "-- START OF LDEV OUTPUT --"
6918 echo "--- END OF LDEV OUTPUT ---"
6920 # ldev failed, error
6921 if [ $? -ne 0 ]; then
6922 rm $LDEVCONFPATH $NIDSPATH
6923 error "ldev failed to execute!"
6926 # need to process multiple lines because of combined MGS and MDS
6927 echo -e $output | awk '{ print $2 }' | while read -r line ; do
6928 if [ "$line" != "$MGSNID" ]; then
6929 rm $LDEVCONFPATH $NIDSPATH
6930 error "ldev failed mgs nid '$line', expected '$MGSNID'"
6934 rm $LDEVCONFPATH $NIDSPATH
6936 run_test 92 "ldev returns MGS NID correctly in command substitution"
6939 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs"
6942 #start mgs or mgs/mdt0
6943 if ! combined_mgs_mds ; then
6950 start_ost || error "OST0 start fail"
6952 #define OBD_FAIL_MGS_WRITE_TARGET_DELAY 0x90e
6953 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x8000090e"
6954 for num in $(seq 2 $MDSCOUNT); do
6958 mount_client $MOUNT || error "mount client fails"
6959 wait_osc_import_state mds ost FULL
6960 wait_osc_import_ready client ost
6961 check_mount || error "check_mount failed"
6963 cleanup || error "cleanup failed with $?"
6965 run_test 93 "register mulitple MDT at the same time"
6968 if [ -z "$LDEV" ]; then
6969 error "ldev is missing!"
6972 local LDEVCONFPATH=$TMP/ldev.conf
6973 local NIDSPATH=$TMP/nids
6975 generate_ldev_conf $LDEVCONFPATH
6976 generate_nids $NIDSPATH
6978 local LDEV_OUTPUT=$TMP/ldev-output.txt
6979 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME > $LDEV_OUTPUT
6981 # ldev failed, error
6982 if [ $? -ne 0 ]; then
6983 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
6984 error "ldev failed to execute!"
6988 local EXPECTED_OUTPUT=$TMP/ldev-expected.txt
6990 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
6992 for num in $(seq $MDSCOUNT); do
6993 printf "%s-MDT%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
6996 for num in $(seq $OSTCOUNT); do
6997 printf "%s-OST%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
7000 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7002 if [ $? -ne 0 ]; then
7003 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7004 error "ldev failed to produce the correct hostlist!"
7007 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7009 run_test 94 "ldev outputs correct labels for file system name query"
7012 if [ -z "$LDEV" ]; then
7013 error "ldev is missing!"
7016 local LDEVCONFPATH=$TMP/ldev.conf
7017 local NIDSPATH=$TMP/nids
7019 generate_ldev_conf $LDEVCONFPATH
7020 generate_nids $NIDSPATH
7024 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME &>/dev/null
7025 if [ $? -ne 0 ]; then
7026 rm $LDEVCONFPATH $NIDSPATH
7027 error "ldev label filtering w/ -F failed!"
7031 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -l &>/dev/null
7032 if [ $? -ne 0 ]; then
7033 rm $LDEVCONFPATH $NIDSPATH
7034 error "ldev label filtering w/ -l failed!"
7038 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f &>/dev/null
7039 if [ $? -ne 0 ]; then
7040 rm $LDEVCONFPATH $NIDSPATH
7041 error "ldev label filtering w/ -f failed!"
7045 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a &>/dev/null
7046 if [ $? -ne 0 ]; then
7047 rm $LDEVCONFPATH $NIDSPATH
7048 error "ldev label filtering w/ -a failed!"
7053 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -F $FSNAME &>/dev/null
7054 if [ $? -eq 0 ]; then
7055 rm $LDEVCONFPATH $NIDSPATH
7056 error "ldev label filtering w/ -a and -F incorrectly succeeded"
7060 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -f &>/dev/null
7061 if [ $? -eq 0 ]; then
7062 rm $LDEVCONFPATH $NIDSPATH
7063 error "ldev label filtering w/ -a and -f incorrectly succeeded"
7067 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -l &>/dev/null
7068 if [ $? -eq 0 ]; then
7069 rm $LDEVCONFPATH $NIDSPATH
7070 error "ldev label filtering w/ -a and -l incorrectly succeeded"
7074 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f -l &>/dev/null
7075 if [ $? -eq 0 ]; then
7076 rm $LDEVCONFPATH $NIDSPATH
7077 error "ldev label filtering w/ -f and -l incorrectly succeeded"
7081 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -l &>/dev/null
7082 if [ $? -eq 0 ]; then
7083 rm $LDEVCONFPATH $NIDSPATH
7084 error "ldev label filtering w/ -F and -l incorrectly succeeded"
7087 # file sys & foreign
7088 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -f &>/dev/null
7089 if [ $? -eq 0 ]; then
7090 rm $LDEVCONFPATH $NIDSPATH
7091 error "ldev label filtering w/ -F and -f incorrectly succeeded"
7094 rm $LDEVCONFPATH $NIDSPATH
7096 run_test 95 "ldev should only allow one label filter"
7099 if [ -z "$LDEV" ]; then
7100 error "ldev is missing!"
7103 local LDEVCONFPATH=$TMP/ldev.conf
7104 local NIDSPATH=$TMP/nids
7106 generate_ldev_conf $LDEVCONFPATH
7107 generate_nids $NIDSPATH
7109 local LDEV_OUTPUT=$TMP/ldev-output.txt
7110 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -H $mgs_HOST \
7112 awk '{print $2}' > $LDEV_OUTPUT
7114 # ldev failed, error
7115 if [ $? -ne 0 ]; then
7116 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7117 error "ldev failed to execute!"
7121 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
7123 echo "$mgs_HOST-$(facet_fstype mgs)" > $EXPECTED_OUTPUT
7125 if [ "$mgs_HOST" == "$mds_HOST" ]; then
7126 for num in $(seq $MDSCOUNT); do
7127 echo "$mds_HOST-$(facet_fstype mds$num)" \
7132 if [ "$mgs_HOST" == "$ost_HOST" ]; then
7133 for num in $(seq $OSTCOUNT); do
7134 echo "$ost_HOST-$(facet_fstype ost$num)" \
7139 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7141 if [ $? -ne 0 ]; then
7142 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7143 error "ldev failed to produce the correct output!"
7146 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7148 run_test 96 "ldev returns hostname and backend fs correctly in command sub"
7151 if [ -z "$LDEV" ]; then
7152 error "ldev is missing!"
7155 local LDEVCONFPATH=$TMP/ldev.conf
7156 local NIDSPATH=$TMP/nids
7158 generate_ldev_conf $LDEVCONFPATH
7159 generate_nids $NIDSPATH
7161 local LDEV_OUTPUT=$TMP/ldev-output.txt
7162 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
7164 echo -e "\nMDT role"
7165 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mdt > $LDEV_OUTPUT
7167 if [ $? -ne 0 ]; then
7168 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7169 error "ldev failed to execute for mdt role!"
7172 for num in $(seq $MDSCOUNT); do
7173 printf "%s-MDT%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
7176 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7178 if [ $? -ne 0 ]; then
7179 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7180 error "ldev failed to produce the correct output for mdt role!"
7183 echo -e "\nOST role"
7184 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R ost > $LDEV_OUTPUT
7186 if [ $? -ne 0 ]; then
7187 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
7188 error "ldev failed to execute for ost role!"
7192 for num in $(seq $OSTCOUNT); do
7193 printf "%s-OST%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
7196 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7198 if [ $? -ne 0 ]; then
7199 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7200 error "ldev failed to produce the correct output for ost role!"
7203 echo -e "\nMGS role"
7204 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mgs > $LDEV_OUTPUT
7206 if [ $? -ne 0 ]; then
7207 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
7208 error "ldev failed to execute for mgs role!"
7211 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
7213 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7215 if [ $? -ne 0 ]; then
7216 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7217 error "ldev failed to produce the correct output for mgs role!"
7220 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7222 run_test 97 "ldev returns correct ouput when querying based on role"
7227 local temp=$MDS_MOUNT_OPTS
7230 check_mount || error "mount failed"
7231 mountopt="user_xattr"
7232 for ((x = 1; x <= 400; x++)); do
7233 mountopt="$mountopt,user_xattr"
7235 remount_client $mountopt $MOUNT 2>&1 | grep "too long" ||
7236 error "Buffer overflow check failed"
7237 cleanup || error "cleanup failed"
7239 run_test 98 "Buffer-overflow check while parsing mount_opts"
7243 [[ $(facet_fstype ost1) != ldiskfs ]] &&
7244 skip "ldiskfs only test"
7245 [[ $(lustre_version_code ost1) -ge $(version_code 2.8.57) ]] ||
7246 skip "Need OST version at least 2.8.57"
7248 local ost_opts="$(mkfs_opts ost1 $(ostdevname 1)) \
7249 --reformat $(ostdevname 1) $(ostvdevname 1)"
7250 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" &&
7251 skip "meta_bg already set"
7254 if [[ ${!opts} != *mkfsoptions* ]]; then
7255 eval opts=\"${!opts} \
7256 --mkfsoptions='\\\"-O ^resize_inode,meta_bg\\\"'\"
7258 local val=${!opts//--mkfsoptions=\\\"/ \
7259 --mkfsoptions=\\\"-O ^resize_inode,meta_bg }
7263 echo "params: $opts"
7265 add ost1 $opts || error "add ost1 failed with new params"
7267 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" ||
7268 error "meta_bg is not set"
7272 run_test 99 "Adding meta_bg option"
7276 start_mds || error "MDS start failed"
7277 start_ost || error "unable to start OST"
7278 mount_client $MOUNT || error "client start failed"
7279 check_mount || error "check_mount failed"
7288 do_facet mgs 'lshowmount -v' | awk 'BEGIN {NR == 0; rc=1} /MGS:/ {rc=0}
7289 END {exit rc}' || error "lshowmount have no output MGS"
7291 do_facet mds1 'lshowmount -v' | awk 'BEGIN {NR == 2; rc=1} /-MDT0000:/
7292 {rc=0} END {exit rc}' || error "lshowmount have no output MDT0"
7294 do_facet ost1 'lshowmount -v' | awk 'BEGIN {NR == 4; rc=1} /-OST0000:/
7295 {rc=0} END {exit rc}' || error "lshowmount have no output OST0"
7297 cleanup || error "cleanup failed with $?"
7299 run_test 100 "check lshowmount lists MGS, MDT, OST and 0@lo"
7302 local createmany_pid
7303 local dev=$FSNAME-OST0000-osc-MDT0000
7307 createmany -o $DIR1/$tdir/$tfile-%d 50000 &
7309 # MDT->OST reconnection causes MDT<->OST last_id synchornisation
7310 # via osp_precreate_cleanup_orphans.
7311 for ((i = 0; i < 100; i++)); do
7312 for ((k = 0; k < 10; k++)); do
7313 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;" \
7314 "$LCTL --device $dev activate"
7317 ls -asl $MOUNT | grep '???' &&
7318 { kill -9 $createmany_pid &>/dev/null;
7319 error "File has no object on OST"; }
7321 kill -s 0 $createmany_pid || break
7323 wait $createmany_pid
7325 unlinkmany $DIR1/$tdir/$tfile-%d 50000
7328 run_test 101 "Race MDT->OST reconnection with create"
7331 [[ $(lustre_version_code $SINGLEMDS) -gt $(version_code 2.9.53) ]] ||
7332 skip "Need server version greater than 2.9.53"
7333 cleanup || error "cleanup failed with $?"
7335 local mds1dev=$(mdsdevname 1)
7336 local mds1mnt=$(facet_mntpt mds1)
7337 local mds1fstype=$(facet_fstype mds1)
7338 local mds1opts=$MDS_MOUNT_OPTS
7340 if [ $mds1fstype == ldiskfs ] &&
7341 ! do_facet mds1 test -b $mds1dev; then
7342 mds1opts=$(csa_add "$mds1opts" -o loop)
7344 if [[ $mds1fstype == zfs ]]; then
7345 import_zpool mds1 || return ${PIPESTATUS[0]}
7348 # unload all and only load libcfs to allow fail_loc setting
7349 do_facet mds1 $LUSTRE_RMMOD || error "unable to unload modules"
7350 do_facet mds1 modprobe libcfs || error "libcfs not loaded"
7351 do_facet mds1 lsmod \| grep libcfs || error "libcfs not loaded"
7353 #define OBD_FAIL_OBDCLASS_MODULE_LOAD 0x60a
7354 do_facet mds1 "$LCTL set_param fail_loc=0x8000060a"
7356 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts &&
7357 error "mdt start must fail"
7358 do_facet mds1 lsmod \| grep obdclass && error "obdclass must not load"
7360 do_facet mds1 "$LCTL set_param fail_loc=0x0"
7362 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts ||
7363 error "mdt start must not fail"
7365 cleanup || error "cleanup failed with $?"
7367 run_test 102 "obdclass module cleanup upon error"
7372 echo "rename $FSNAME to $newname"
7374 if ! combined_mgs_mds ; then
7375 local facet=$(mgsdevname)
7378 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $facet"||
7379 error "(7) Fail to rename MGS"
7380 if [ "$(facet_fstype $facet)" = "zfs" ]; then
7381 reimport_zpool mgs $newname-mgs
7385 for num in $(seq $MDSCOUNT); do
7386 local facet=$(mdsdevname $num)
7388 do_facet mds${num} \
7389 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $facet"||
7390 error "(8) Fail to rename MDT $num"
7391 if [ "$(facet_fstype $facet)" = "zfs" ]; then
7392 reimport_zpool mds${num} $newname-mdt${num}
7396 for num in $(seq $OSTCOUNT); do
7397 local facet=$(ostdevname $num)
7399 do_facet ost${num} \
7400 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $facet"||
7401 error "(9) Fail to rename OST $num"
7402 if [ "$(facet_fstype $facet)" = "zfs" ]; then
7403 reimport_zpool ost${num} $newname-ost${num}
7408 test_103_set_pool() {
7412 do_facet mgs $LCTL pool_add $FSNAME.$pname ${FSNAME}-$ost_x ||
7413 error "Fail to add $ost_x to $FSNAME.$pname"
7414 wait_update $HOSTNAME \
7415 "lctl get_param -n lov.$FSNAME-clilov-*.pools.$pname |
7416 grep $ost_x" "$FSNAME-${ost_x}_UUID" ||
7417 error "$ost_x is NOT in pool $FSNAME.$pname"
7420 test_103_check_pool() {
7421 local save_fsname=$1
7424 stat $DIR/$tdir/test-framework.sh ||
7425 error "($errno) Fail to stat"
7426 do_facet mgs $LCTL pool_list $FSNAME.pool1 ||
7427 error "($errno) Fail to list $FSNAME.pool1"
7428 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname ||
7429 error "($errno) Fail to list $FSNAME.$save_fsname"
7430 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname |
7431 grep ${FSNAME}-OST0000 ||
7432 error "($errno) List $FSNAME.$save_fsname is invalid"
7434 local pname=$($LFS getstripe --pool $DIR/$tdir/d0)
7435 [ "$pname" = "$save_fsname" ] ||
7436 error "($errno) Unexpected pool name $pname"
7440 check_mount_and_prep
7442 mkdir $DIR/$tdir || error "(1) Fail to mkdir $DIR/$tdir"
7443 cp $LUSTRE/tests/test-framework.sh $DIR/$tdir ||
7444 error "(2) Fail to copy test-framework.sh"
7446 if ! combined_mgs_mds ; then
7449 do_facet mgs $LCTL pool_new $FSNAME.pool1 ||
7450 error "(3) Fail to create $FSNAME.pool1"
7451 # name the pool name as the fsname
7452 do_facet mgs $LCTL pool_new $FSNAME.$FSNAME ||
7453 error "(4) Fail to create $FSNAME.$FSNAME"
7455 test_103_set_pool $FSNAME OST0000
7457 $LFS setstripe -p $FSNAME $DIR/$tdir/d0 ||
7458 error "(6) Fail to setstripe on $DIR/$tdir/d0"
7460 if ! combined_mgs_mds ; then
7466 test_renamefs mylustre
7468 local save_fsname=$FSNAME
7472 if ! combined_mgs_mds ; then
7475 test_103_check_pool $save_fsname 7
7477 if [ $OSTCOUNT -ge 2 ]; then
7478 test_103_set_pool $save_fsname OST0001
7481 $LFS setstripe -p $save_fsname $DIR/$tdir/f0 ||
7482 error "(16) Fail to setstripe on $DIR/$tdir/f0"
7483 if ! combined_mgs_mds ; then
7494 if ! combined_mgs_mds ; then
7497 test_103_check_pool $save_fsname 17
7499 if ! combined_mgs_mds ; then
7504 test_renamefs $save_fsname
7510 run_test 103 "rename filesystem name"
7512 test_104() { # LU-6952
7513 local mds_mountopts=$MDS_MOUNT_OPTS
7514 local ost_mountopts=$OST_MOUNT_OPTS
7515 local mds_mountfsopts=$MDS_MOUNT_FS_OPTS
7516 local lctl_ver=$(do_facet $SINGLEMDS $LCTL --version |
7519 [[ $(version_code $lctl_ver) -lt $(version_code 2.9.55) ]] &&
7520 skip "this test needs utils above 2.9.55"
7522 # specify "acl" in mount options used by mkfs.lustre
7523 if [ -z "$MDS_MOUNT_FS_OPTS" ]; then
7524 MDS_MOUNT_FS_OPTS="acl,user_xattr"
7527 MDS_MOUNT_FS_OPTS="${MDS_MOUNT_FS_OPTS},acl,user_xattr"
7530 echo "mountfsopt: $MDS_MOUNT_FS_OPTS"
7532 #reformat/remount the MDT to apply the MDT_MOUNT_FS_OPT options
7534 if [ -z "$MDS_MOUNT_OPTS" ]; then
7535 MDS_MOUNT_OPTS="-o noacl"
7537 MDS_MOUNT_OPTS="${MDS_MOUNT_OPTS},noacl"
7540 for num in $(seq $MDSCOUNT); do
7541 start mds$num $(mdsdevname $num) $MDS_MOUNT_OPTS ||
7542 error "Failed to start MDS"
7545 for num in $(seq $OSTCOUNT); do
7546 start ost$num $(ostdevname $num) $OST_MOUNT_OPTS ||
7547 error "Failed to start OST"
7551 setfacl -m "d:$RUNAS_ID:rwx" $MOUNT &&
7552 error "ACL is applied when FS is mounted with noacl."
7554 MDS_MOUNT_OPTS=$mds_mountopts
7555 OST_MOUNT_OPTS=$ost_mountopts
7556 MDS_MOUNT_FS_OPTS=$mds_mountfsopts
7561 run_test 104 "Make sure user defined options are reflected in mount"
7563 error_and_umount() {
7574 mount --bind $DIR $TMP/$tdir || error "mount bind mnt pt failed"
7575 rm -f $TMP/$tdir/$tfile
7576 rm -f $TMP/$tdir/${tfile}1
7578 # Files should not be created in ro bind mount point
7579 # remounting from rw to ro
7580 mount -o remount,ro $TMP/$tdir ||
7581 error_and_umount "readonly remount of bind mnt pt failed"
7582 touch $TMP/$tdir/$tfile &&
7583 error_and_umount "touch succeeds on ro bind mnt pt"
7584 [ -e $TMP/$tdir/$tfile ] &&
7585 error_and_umount "file created on ro bind mnt pt"
7587 # Files should be created in rw bind mount point
7588 # remounting from ro to rw
7589 mount -o remount,rw $TMP/$tdir ||
7590 error_and_umount "read-write remount of bind mnt pt failed"
7591 touch $TMP/$tdir/${tfile}1 ||
7592 error_and_umount "touch fails on rw bind mnt pt"
7593 [ -e $TMP/$tdir/${tfile}1 ] ||
7594 error_and_umount "file not created on rw bind mnt pt"
7595 umount $TMP/$tdir || error "umount of bind mnt pt failed"
7597 cleanup || error "cleanup failed with $?"
7599 run_test 105 "check file creation for ro and rw bind mnt pt"
7606 mkdir -p $DIR/$tdir || error "create $tdir failed"
7607 lfs setstripe -c 1 -i 0 $DIR/$tdir
7608 #define OBD_FAIL_CAT_RECORDS 0x1312
7609 do_facet mds1 $LCTL set_param fail_loc=0x1312 fail_val=$repeat
7611 for ((i = 1; i <= $repeat; i++)); do
7613 #one full plain llog
7614 createmany -o $DIR/$tdir/f- 64768
7616 createmany -u $DIR/$tdir/f- 64768
7618 wait_delete_completed $((TIMEOUT * 7))
7619 #ASSERTION osp_sync_thread() ( thread->t_flags != SVC_RUNNING ) failed
7620 #shows that osp code is buggy
7621 do_facet mds1 $LCTL set_param fail_loc=0 fail_val=0
7625 run_test 106 "check osp llog processing when catalog is wrapped"
7628 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.10.50) ]] ||
7629 skip "Need MDS version > 2.10.50"
7632 start_mgsmds || error "start_mgsmds failed"
7633 start_ost || error "unable to start OST"
7635 # add unknown configuration parameter.
7636 if [[ $PERM_CMD == *"set_param -P"* ]]; then
7637 cmd="$PERM_CMD ost.$FSNAME-OST0000*.unknown_param"
7639 cmd="$PERM_CMD $FSNAME-OST0000*.ost.unknown_param"
7641 do_facet mgs "$cmd=50"
7642 cleanup_nocli || error "cleanup_nocli failed with $?"
7645 # unknown param should be ignored while mounting.
7646 start_ost || error "unable to start OST after unknown param set"
7648 cleanup || error "cleanup failed with $?"
7650 run_test 107 "Unknown config param should not fail target mounting"
7655 $rcmd rm -rf $tmp > /dev/null 2>&1
7656 $rcmd mkdir -p $tmp/{mnt,images} || error "failed to mkdir remotely"
7658 for facet in $facets; do
7659 [ $(facet_fstype $SINGLEMDS) = "zfs" ] &&
7660 $rcmd $ZPOOL -f export lustre-$facet > /dev/null 2>&1
7661 $rcmd mkdir $tmp/mnt/$facet ||
7662 error "failed to mkdir $tmp/mnt/$facet"
7663 $rcmd dd if=/dev/zero of=$tmp/images/$facet \
7664 seek=199 bs=1M count=1 ||
7665 error "failed to create $tmp/images/$facet"
7674 local facet=${role}$((idx + 1))
7676 [ $# -eq 5 ] && pool=$5
7678 do_facet $SINGLEMDS $MKFS --fsname=lustre --$mgs \
7679 --$role --index=$idx --replace --backfstype=$bkfs \
7680 --device-size=200000 --reformat $pool $tmp/images/$facet ||
7681 error "failed to mkfs for $facet"
7685 echo "mounting client..."
7686 mount -t lustre ${nid}:/lustre $MOUNT ||
7687 error "failed to mount lustre"
7690 ls -l $MOUNT/local_dir || error "failed to list"
7692 echo "check truncate && write"
7693 echo "dummmmmmmmmmmmm" > $MOUNT/remote_dir/fsx.c ||
7694 error "failed to tuncate & write"
7697 touch $MOUNT/foooo ||
7698 error "failed to create"
7700 echo "check read && write && append"
7701 sha1sum $MOUNT/conf-sanity.sh |
7702 awk '{ print $1 }' > $MOUNT/checksum.new ||
7703 error "failed to read(1)"
7704 sha1sum $MOUNT/remote_dir/unlinkmany.c |
7705 awk '{ print $1 }' >> $MOUNT/checksum.new ||
7706 error "failed to read(2)"
7707 sha1sum $MOUNT/striped_dir/lockahead_test.o |
7708 awk '{ print $1 }' >> $MOUNT/checksum.new ||
7709 error "failed to read(3)"
7712 diff $MOUNT/checksum.new $MOUNT/checksum.src ||
7713 error "failed to verify data"
7723 umount -f $MOUNT || error "failed to umount client"
7724 for facet in $facets; do
7725 $rcmd umount -f $tmp/mnt/$facet ||
7726 error "failed to umount $facet"
7727 if [ $(facet_fstype $SINGLEMDS) = "zfs" ]; then
7728 $rcmd $ZPOOL export -f lustre-$facet ||
7729 error "failed to export lustre-$facet"
7733 $rcmd rm -rf $tmp || error "failed to rm the dir $tmp"
7737 [ "$CLIENTONLY" ] && skip "Client-only testing"
7738 [ $(facet_fstype $SINGLEMDS) != "zfs" ] && skip "zfs only test"
7739 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.58) ] &&
7740 skip "Need server version at least 2.10.58"
7745 local tmp=$TMP/$tdir
7746 local rcmd="do_facet $SINGLEMDS"
7747 local facets="mdt1 mdt2 ost1 ost2"
7748 local nid=$($rcmd $LCTL list_nids | head -1)
7751 trap t_108_cleanup EXIT ERR
7754 t_108_mkfs mdt 0 zfs mgs lustre-mdt1/mdt1
7755 t_108_mkfs mdt 1 zfs mgsnode=$nid lustre-mdt2/mdt2
7756 t_108_mkfs ost 0 zfs mgsnode=$nid lustre-ost1/ost1
7757 t_108_mkfs ost 1 zfs mgsnode=$nid lustre-ost2/ost2
7759 for facet in $facets; do
7760 $rcmd zfs set mountpoint=$tmp/mnt/$facet canmount=on \
7761 lustre-$facet/$facet ||
7762 error "failed to zfs set for $facet (1)"
7763 $rcmd zfs mount lustre-$facet/$facet ||
7764 error "failed to local mount $facet"
7765 $rcmd tar jxf $LUSTRE/tests/ldiskfs_${facet}_2_11.tar.bz2 \
7766 --xattrs --xattrs-include="trusted.*" \
7767 -C $tmp/mnt/$facet/ > /dev/null 2>&1 ||
7768 error "failed to untar image for $facet"
7769 $rcmd "cd $tmp/mnt/$facet && rm -rf oi.* OI_* lfsck_* LFSCK" ||
7770 error "failed to cleanup for $facet"
7771 $rcmd zfs umount lustre-$facet/$facet ||
7772 error "failed to local umount $facet"
7773 $rcmd zfs set canmount=off lustre-$facet/$facet ||
7774 error "failed to zfs set $facet (2)"
7777 echo "changing server nid..."
7778 $rcmd mount -t lustre -o nosvc lustre-mdt1/mdt1 $tmp/mnt/mdt1
7779 $rcmd lctl replace_nids lustre-MDT0000 $nid
7780 $rcmd lctl replace_nids lustre-MDT0001 $nid
7781 $rcmd lctl replace_nids lustre-OST0000 $nid
7782 $rcmd lctl replace_nids lustre-OST0001 $nid
7783 $rcmd umount $tmp/mnt/mdt1
7785 for facet in $facets; do
7786 echo "mounting $facet from backup..."
7787 $rcmd mount -t lustre -o abort_recov lustre-$facet/$facet \
7788 $tmp/mnt/$facet || error "failed to mount $facet"
7791 # ZFS backend can detect migration and trigger OI scrub automatically
7792 # sleep 3 seconds for scrub done
7798 run_test 108a "migrate from ldiskfs to ZFS"
7801 [ "$CLIENTONLY" ] && skip "Client-only testing"
7802 [ $(facet_fstype $SINGLEMDS) != "ldiskfs" ] && skip "ldiskfs only test"
7803 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.58) ] &&
7804 skip "Need server version at least 2.10.58"
7809 local tmp=$TMP/$tdir
7810 local rcmd="do_facet $SINGLEMDS"
7811 local facets="mdt1 mdt2 ost1 ost2"
7812 local scrub_list="MDT0000 MDT0001 OST0000 OST0001"
7813 local nid=$($rcmd $LCTL list_nids | head -1)
7816 trap t_108_cleanup EXIT ERR
7819 t_108_mkfs mdt 0 ldiskfs mgs
7820 t_108_mkfs mdt 1 ldiskfs mgsnode=$nid
7821 t_108_mkfs ost 0 ldiskfs mgsnode=$nid
7822 t_108_mkfs ost 1 ldiskfs mgsnode=$nid
7824 for facet in $facets; do
7825 $rcmd mount -t ldiskfs -o loop $tmp/images/$facet \
7827 error "failed to local mount $facet"
7829 $rcmd tar jxf $LUSTRE/tests/zfs_${facet}_2_11.tar.bz2 \
7830 --xattrs --xattrs-include="*.*" \
7831 -C $tmp/mnt/$facet/ > /dev/null 2>&1 ||
7832 error "failed to untar image for $facet"
7833 $rcmd "cd $tmp/mnt/$facet && rm -rf oi.* OI_* lfsck_* LFSCK" ||
7834 error "failed to cleanup for $facet"
7835 $rcmd umount $tmp/mnt/$facet ||
7836 error "failed to local umount $facet"
7839 echo "changing server nid..."
7840 $rcmd mount -t lustre -o nosvc,loop $tmp/images/mdt1 $tmp/mnt/mdt1
7841 $rcmd lctl replace_nids lustre-MDT0000 $nid
7842 $rcmd lctl replace_nids lustre-MDT0001 $nid
7843 $rcmd lctl replace_nids lustre-OST0000 $nid
7844 $rcmd lctl replace_nids lustre-OST0001 $nid
7845 $rcmd umount $tmp/mnt/mdt1
7847 for facet in $facets; do
7848 echo "mounting $facet from backup..."
7849 $rcmd mount -t lustre -o loop,abort_recov $tmp/images/$facet \
7850 $tmp/mnt/$facet || error "failed to mount $facet"
7853 for facet in $scrub_list; do
7854 $rcmd $LCTL lfsck_start -M lustre-$facet -t scrub ||
7855 error "failed to start OI scrub on $facet"
7858 # sleep 3 seconds for scrub done
7864 run_test 108b "migrate from ZFS to ldiskfs"
7868 # set number of permanent parameters
7870 test_109_set_params() {
7873 set_persistent_param_and_check mds \
7874 "mdd.$fsname-MDT0000.atime_diff" \
7875 "$fsname-MDT0000.mdd.atime_diff" \
7877 set_persistent_param_and_check mds \
7878 "mdd.$fsname-MDT0000.atime_diff" \
7879 "$fsname-MDT0000.mdd.atime_diff" \
7881 set_persistent_param_and_check client \
7882 "llite.$fsname*.max_read_ahead_mb" \
7883 "$fsname.llite.max_read_ahead_mb" \
7885 set_persistent_param_and_check client \
7886 "llite.$fsname*.max_read_ahead_mb" \
7887 "$fsname.llite.max_read_ahead_mb" \
7889 create_pool $fsname.pool1 || error "create pool failed"
7890 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
7891 error "pool_add failed"
7892 do_facet mgs $LCTL pool_remove $fsname.pool1 OST0000 ||
7893 error "pool_remove failed"
7894 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
7895 error "pool_add failed"
7899 # check permanent parameters
7901 test_109_test_params() {
7904 local atime_diff=$(do_facet mds $LCTL \
7905 get_param -n mdd.$fsname-MDT0000.atime_diff)
7906 [ $atime_diff == 63 ] || error "wrong mdd parameter after clear_conf"
7907 local max_read_ahead_mb=$(do_facet client $LCTL \
7908 get_param -n llite.$fsname*.max_read_ahead_mb)
7909 [ $max_read_ahead_mb == 64 ] ||
7910 error "wrong llite parameter after clear_conf"
7911 local ost_in_pool=$(do_facet mds $LCTL pool_list $fsname.pool1 |
7912 grep -v "^Pool:" | sed 's/_UUID//')
7913 [ $ost_in_pool = "$fsname-OST0000" ] ||
7914 error "wrong pool after clear_conf"
7918 # run lctl clear_conf, store CONFIGS before and after that
7920 test_109_clear_conf()
7922 local clear_conf_arg=$1
7925 if ! combined_mgs_mds ; then
7927 stop_mgs || error "stop_mgs failed"
7928 start_mgs "-o nosvc" || error "start_mgs nosvc failed"
7930 mgsdev=$(mdsdevname 1)
7931 start_mdt 1 "-o nosvc" || error "start_mdt 1 nosvc failed"
7934 do_facet mgs "rm -rf $TMP/${tdir}/conf1; mkdir -p $TMP/${tdir}/conf1;" \
7935 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf1\\\" \
7939 # the command being tested
7941 do_facet mgs $LCTL clear_conf $clear_conf_arg ||
7942 error "clear_conf failed"
7943 if ! combined_mgs_mds ; then
7944 stop_mgs || error "stop_mgs failed"
7946 stop_mdt 1 || error "stop_mdt 1 failed"
7949 do_facet mgs "rm -rf $TMP/${tdir}/conf2; mkdir -p $TMP/${tdir}/conf2;" \
7950 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf2\\\" \
7954 test_109_file_shortened() {
7956 local sizes=($(do_facet mgs "stat -c %s " \
7957 "$TMP/${tdir}/conf1/CONFIGS/$file" \
7958 "$TMP/${tdir}/conf2/CONFIGS/$file"))
7959 [ ${sizes[1]} -lt ${sizes[0]} ] && return 0
7965 [ "$(facet_fstype mgs)" == "zfs" ] &&
7966 skip "LU-8727: no implementation for ZFS"
7970 client_up || error "client_up failed"
7973 # set number of permanent parameters
7975 test_109_set_params $FSNAME
7977 umount_client $MOUNT || error "umount_client failed"
7978 stop_ost || error "stop_ost failed"
7979 stop_mds || error "stop_mds failed"
7981 test_109_clear_conf $FSNAME
7983 # make sure that all configs are cleared
7985 test_109_file_shortened $FSNAME-MDT0000 ||
7986 error "failed to clear MDT0000 config"
7987 test_109_file_shortened $FSNAME-client ||
7988 error "failed to clear client config"
7993 # check that configurations are intact
7995 test_109_test_params $FSNAME
8000 destroy_test_pools || error "destroy test pools failed"
8004 run_test 109a "test lctl clear_conf fsname"
8008 [ "$(facet_fstype mgs)" == "zfs" ] &&
8009 skip "LU-8727: no implementation for ZFS"
8013 client_up || error "client_up failed"
8016 # set number of permanent parameters
8018 test_109_set_params $FSNAME
8020 umount_client $MOUNT || error "umount_client failed"
8021 stop_ost || error "stop_ost failed"
8022 stop_mds || error "stop_mds failed"
8024 test_109_clear_conf $FSNAME-MDT0000
8026 # make sure that only one config is cleared
8028 test_109_file_shortened $FSNAME-MDT0000 ||
8029 error "failed to clear MDT0000 config"
8030 test_109_file_shortened $FSNAME-client &&
8031 error "failed to clear client config"
8036 # check that configurations are intact
8038 test_109_test_params $FSNAME
8043 destroy_test_pools || error "destroy test pools failed"
8047 run_test 109b "test lctl clear_conf one config"
8051 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ]] &&
8052 skip "Only applicable to ldiskfs-based MDTs"
8054 do_facet $SINGLEMDS $DEBUGFS -w -R supported_features |grep large_dir ||
8055 skip "large_dir option is not supported on MDS"
8056 do_facet ost1 $DEBUGFS -w -R supported_features | grep large_dir ||
8057 skip "large_dir option is not supported on OSS"
8059 stopall # stop all targets before modifying the target counts
8060 stack_trap "MDSCOUNT=$MDSCOUNT OSTCOUNT=$OSTCOUNT" EXIT
8064 # ext4_dir_entry_2 struct size:264
8065 # dx_root struct size:8
8066 # dx_node struct size:8
8067 # dx_entry struct size:8
8068 # For 1024 bytes block size.
8069 # First level directory entries: 126
8070 # Second level directory entries: 127
8071 # Entries in leaf: 3
8072 # For 2 levels limit: 48006
8073 # For 3 levels limit : 6096762
8074 # Create 80000 files to safely exceed 2-level htree limit.
8075 CONF_SANITY_110_LINKS=${CONF_SANITY_110_LINKS:-80000}
8077 # can fit at most 3 filenames per 1KB leaf block, but each
8078 # leaf/index block will only be 3/4 full before split at each level
8079 (( MDSSIZE < CONF_SANITY_110_LINKS / 3 * 4/3 * 4/3 )) &&
8080 CONF_SANITY_110_LINKS=$((MDSSIZE * 3 * 3/4 * 3/4))
8082 local opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
8083 --reformat $(mdsdevname 1) $(mdsvdevname 1)"
8084 if [[ $opts != *mkfsoptions* ]]; then
8085 opts+=" --mkfsoptions=\\\"-O large_dir -b 1024 -i 65536\\\""
8087 opts="${opts//--mkfsoptions=\\\"/ \
8088 --mkfsoptions=\\\"-O large_dir -b 1024 -i 65536 }"
8090 echo "MDT params: $opts"
8091 add mds1 $opts || error "add mds1 failed with new params"
8092 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
8094 opts="$(mkfs_opts ost1 $(ostdevname 1)) \
8095 --reformat $(ostdevname 1) $(ostvdevname 1)"
8097 if [[ $opts != *mkfsoptions* ]]; then
8098 opts+=" --mkfsoptions=\\\"-O large_dir\\\" "
8100 opts="${opts//--mkfsoptions=\\\"/ \
8101 --mkfsoptions=\\\"-O large_dir }"
8103 echo "OST params: $opts"
8104 add ost1 $opts || error "add ost1 failed with new params"
8105 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
8107 MOUNT_2=yes mountcli || error "mount clients failed"
8109 mkdir -v $DIR/$tdir || error "cannot create $DIR/$tdir"
8110 local pids count=0 group=0
8112 echo "creating $CONF_SANITY_110_LINKS in total"
8113 while (( count < CONF_SANITY_110_LINKS )); do
8114 local len=$((253 - $(wc -c <<<"$tfile-$group-40000-")))
8115 local dir=DIR$((group % 2 + 1))
8116 local target=${!dir}/$tdir/$tfile-$group
8117 local long=$target-$(generate_name $len)-
8118 local create=$((CONF_SANITY_110_LINKS - count))
8120 (( create > 40000 )) && create=40000
8121 touch $target || error "creating $target failed"
8122 echo "creating $create hard links to $target"
8123 createmany -l $target $long $create &
8126 count=$((count + create))
8127 group=$((group + 1))
8129 echo "waiting for PIDs$pids to complete"
8130 wait $pids || error "createmany failed after $group groups"
8134 run_e2fsck $(facet_active_host mds1) $(mdsdevname 1) -n
8136 run_test 110 "Adding large_dir with 3-level htree"
8139 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ]] &&
8140 skip "Only applicable to ldiskfs-based MDTs"
8142 is_dm_flakey_dev $SINGLEMDS $(mdsdevname 1) &&
8143 skip "This test can not be executed on flakey dev"
8145 do_facet $SINGLEMDS $DEBUGFS -w -R supported_features |grep large_dir ||
8146 skip "large_dir option is not supported on MDS"
8148 do_facet ost1 $DEBUGFS -w -R supported_features | grep large_dir ||
8149 skip "large_dir option is not supported on OSS"
8151 # cleanup before changing target counts
8153 stack_trap "MDSSIZE=$MDSSIZE MDSCOUNT=$MDSCOUNT OSTCOUNT=$OSTCOUNT" EXIT
8156 (( MDSSIZE < 2400000 )) && MDSSIZE=2400000 # need at least 2.4GB
8158 local mdsdev=$(mdsdevname 1)
8160 local opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
8161 --reformat $(mdsdevname 1) $(mdsvdevname 1)"
8162 if [[ $opts != *mkfsoptions* ]]; then
8163 opts+=" --mkfsoptions=\\\"-O large_dir -i 1048576 \\\" "
8165 opts="${opts//--mkfsoptions=\\\"/ \
8166 --mkfsoptions=\\\"-O large_dir -i 1048576 }"
8168 echo "MDT params: $opts"
8169 __touch_device mds 1
8170 add mds1 $opts || error "add mds1 failed with new params"
8171 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
8173 opts="$(mkfs_opts ost1 $(ostdevname 1)) \
8174 --reformat $(ostdevname 1) $(ostvdevname 1)"
8175 if [[ $opts != *mkfsoptions* ]]; then
8176 opts+=" --mkfsoptions=\\\"-O large_dir \\\""
8178 opts="${opts//--mkfsoptions=\\\"/ --mkfsoptions=\\\"-O large_dir }"
8180 echo "OST params: $opts"
8181 __touch_device ost 1
8182 add ost1 $opts || error "add ost1 failed with new params"
8183 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
8185 MOUNT_2=yes mountcli
8186 mkdir $DIR/$tdir || error "cannot create $DIR/$tdir"
8188 lfs df -i $DIR/$tdir
8192 local start=$SECONDS
8194 local dirmax=$((2 << 30))
8200 while (( !needskip & dirsize < dirmax )); do
8203 for cli in ${CLIENTS//,/ }; do
8204 local len=$((253 - $(wc -c <<<"$cli-$group-60000-")))
8205 local target=$cli-$group
8206 local long=$DIR/$tdir/$target-$(generate_name $len)-
8208 RPWD=$DIR/$tdir do_node $cli touch $target ||
8209 error "creating $target failed"
8210 echo "creating 60000 hardlinks to $target"
8211 RPWD=$DIR/$tdir do_node $cli createmany -l $target $long 60000 &
8214 group=$((group + 1))
8216 long=$DIR2/$tdir/$target-$(generate_name $len)-
8218 RPWD=$DIR2/$tdir do_node $cli touch $target ||
8219 error "creating $target failed"
8220 echo "creating 60000 hardlinks to $target"
8221 RPWD=$DIR2/$tdir do_node $cli createmany -l $target $long 60000 &
8224 group=$((group + 1))
8226 echo "waiting for PIDs$pids to complete"
8227 wait $pids || error "createmany failed after $group groups"
8228 dirsize=$(stat -c %s $DIR/$tdir)
8229 taken=$((SECONDS - start))
8230 rate=$((dirsize / taken))
8231 left=$(((dirmax - dirsize) / rate))
8232 num=$((group * 60000))
8233 echo "estimate ${left}s left after $num files / ${taken}s"
8234 # if the estimated time remaining is too large (it may change
8235 # over time as the create rate is not constant) then exit
8236 # without declaring a failure.
8237 (( left > 1200 )) && needskip=1
8242 (( $needskip )) && skip "ETA ${left}s after $num files / ${taken}s is too long"
8244 run_e2fsck $(facet_active_host mds1) $(mdsdevname 1) -n
8246 run_test 111 "Adding large_dir with over 2GB directory"
8253 rm -f $TMP/$tdir/lustre-mdt
8258 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
8259 skip "Only applicable to ldiskfs-based MDTs"
8262 IMAGESIZE=$((3072 << 30)) # 3072 GiB
8265 # We need MDT size 3072GB, because it is smallest
8266 # partition that can store 2B inodes
8267 do_facet $SINGLEMDS "mkdir -p $TMP/$tdir"
8268 local mdsimgname=$TMP/$tdir/lustre-mdt
8269 do_facet $SINGLEMDS "rm -f $mdsimgname"
8270 do_facet $SINGLEMDS "touch $mdsimgname"
8271 trap cleanup_115 RETURN EXIT
8272 do_facet $SINGLEMDS "$TRUNCATE $mdsimgname $IMAGESIZE" ||
8273 skip "Backend FS doesn't support sparse files"
8274 local mdsdev=$(do_facet $SINGLEMDS "losetup -f")
8275 do_facet $SINGLEMDS "losetup $mdsdev $mdsimgname"
8277 local mds_opts="$(mkfs_opts mds1 ${mdsdev}) --device-size=$IMAGESIZE \
8278 --mkfsoptions='-O lazy_itable_init,ea_inode,^resize_inode,meta_bg \
8280 add mds1 $mds_opts --mgs --reformat $mdsdev ||
8281 skip_env "format large MDT failed"
8282 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
8283 --reformat $(ostdevname 1) $(ostvdevname 1)
8285 start $SINGLEMDS ${mdsdev} $MDS_MOUNT_OPTS || error "start MDS failed"
8286 start_ost || error "start OSS failed"
8287 mount_client $MOUNT || error "mount client failed"
8289 mkdir -p $DIR/$tdir || error "mkdir $DIR/$tdir fail"
8290 for goal in $(do_facet $SINGLEMDS "ls /sys/fs/ldiskfs/*/inode_goal"); do
8291 do_facet $SINGLEMDS "echo 2147483947 >> $goal; grep . $goal"
8294 touch $DIR/$tdir/$tfile
8296 # Add > 5k bytes to xattr
8297 for i in {1..30}; do
8298 ln $DIR/$tdir/$tfile $DIR/$tdir/$(printf "link%0250d" $i) ||
8299 error "Can't make link"
8304 local inode_num=$(do_facet $SINGLEMDS \
8305 "$DEBUGFS -c -R 'stat ROOT/$tdir/$tfile' $mdsimgname" |
8306 awk '/link =/ { print $4 }' |
8307 sed -e 's/>//' -e 's/<//' -e 's/\"//')
8308 echo "inode num: $inode_num"
8309 [ $inode_num -ge 2147483947 ] || error "inode $inode_num too small"
8310 do_facet $SINGLEMDS "losetup -d $mdsdev"
8313 run_test 115 "Access large xattr with inodes number over 2TB"
8316 [ $(facet_fstype $SINGLEMDS) != "ldiskfs" ] && skip "ldiskfs only test"
8317 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.59) ] &&
8318 skip "Need server version at least 2.10.59"
8319 do_facet $SINGLEMDS which mkfs.xfs ||
8320 skip_env "No mkfs.xfs installed"
8325 local tmpmnt=$TMP/$tdir
8326 local mdtimg=$tfile-mdt0
8328 do_facet $SINGLEMDS mkdir -p $tmpmnt
8329 stack_trap "do_facet $SINGLEMDS rmdir $tmpmnt" EXIT
8331 do_facet $SINGLEMDS touch $TMP/$mdtimg
8332 stack_trap "do_facet $SINGLEMDS rm -f $TMP/$mdtimg" EXIT
8333 do_facet $SINGLEMDS mkfs -t xfs -d file,size=1t,name=$TMP/$mdtimg ||
8334 error "mkfs temporary xfs image"
8336 do_facet $SINGLEMDS mount $TMP/$mdtimg $tmpmnt ||
8337 error "mount temporary xfs image"
8338 stack_trap "do_facet $SINGLEMDS umount $tmpmnt" EXIT
8339 local old_mdssize=$MDSSIZE
8340 local old_mdsisize=$MDSISIZE
8342 MDSSIZE=$((17 * 1024 * 1024 * 1024)) # 17T MDT
8343 MDSISIZE=$((16 << 20))
8344 local opts17t="$(mkfs_opts $SINGLEMDS)"
8346 MDSSIZE=$old_mdssize
8347 MDSISIZE=$old_mdsisize
8348 do_facet $SINGLEMDS $MKFS $opts17t $tmpmnt/$mdtimg ||
8349 error "failed to mkfs for $tmpmnt/$mdtimg"
8351 do_facet $SINGLEMDS $TUNE2FS -l $tmpmnt/$mdtimg |
8352 grep -qw 'features.*extent' || error "extent should be enabled"
8354 run_test 116 "big size MDT support"
8358 do_facet ost1 "$LCTL set_param ost.OSS.ost_io.nrs_policies=fifo"
8359 do_facet ost1 "$LCTL get_param -n ost.OSS.ost_io.nrs_tbf_rule" &&
8360 error "get_param should fail"
8361 cleanup || error "cleanup failed with rc $?"
8363 run_test 117 "lctl get_param return errors properly"
8365 test_120() { # LU-11130
8366 [ "$MDSCOUNT" -lt 2 ] && skip "mdt count < 2"
8367 [ $(facet_fstype $SINGLEMDS) != "ldiskfs" ] &&
8368 skip "ldiskfs only test"
8369 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.56) ] &&
8370 skip "Need DNE2 capable MD target with LU-11130 fix"
8374 local mds1host=$(facet_active_host mds1)
8375 local mds1dev=$(mdsdevname 1)
8377 $LFS mkdir -i 1 $DIR/$tdir
8378 $LFS mkdir -i 0 $DIR/$tdir/mds1dir
8380 ln -s foo $DIR/$tdir/bar
8381 mv $DIR/$tdir/bar $DIR/$tdir/mds1dir/bar2 ||
8382 error "cross-target rename failed"
8386 run_e2fsck $mds1host $mds1dev "-n"
8388 run_test 120 "cross-target rename should not create bad symlinks"
8391 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
8392 [[ $(lustre_version_code ost1) -ge $(version_code 2.11.53) ]] ||
8393 skip "Need OST version at least 2.11.53"
8396 LOAD_MODULES_REMOTE=true load_modules
8397 #define OBD_FAIL_OFD_SET_OID 0x1e0
8398 do_facet ost1 $LCTL set_param fail_loc=0x00001e0
8401 $LFS mkdir -i1 -c1 $DIR/$tdir
8402 $LFS setstripe -i0 -c1 $DIR/$tdir
8403 do_facet ost1 $LCTL set_param fail_loc=0
8404 createmany -o $DIR/$tdir/file_ 1000 ||
8405 error "Fail to create a new sequence"
8409 run_test 122 "Check OST sequence update"
8412 remote_mgs_nodsh && skip "remote MGS with nodsh"
8413 [ -d $MOUNT/.lustre ] || setupall
8415 # test old logid format until removal from llog_ioctl.c::str2logid()
8416 if [ $MGS_VERSION -lt $(version_code 3.1.53) ]; then
8417 do_facet mgs $LCTL dl | grep MGS
8418 do_facet mgs "$LCTL --device %MGS llog_print \
8419 \\\\\\\$$FSNAME-client 1 10" ||
8420 error "old llog_print failed"
8423 # test new logid format
8424 if [ $MGS_VERSION -ge $(version_code 2.9.53) ]; then
8425 do_facet mgs "$LCTL --device MGS llog_print $FSNAME-client" ||
8426 error "new llog_print failed"
8429 run_test 123aa "llog_print works with FIDs and simple names"
8432 remote_mgs_nodsh && skip "remote MGS with nodsh"
8433 [[ $MGS_VERSION -gt $(version_code 2.11.51) ]] ||
8434 skip "Need server with working llog_print support"
8436 [ -d $MOUNT/.lustre ] || setupall
8441 orig_val=$(do_facet mgs $LCTL get_param jobid_name)
8442 do_facet mgs $LCTL set_param -P jobid_name="testname"
8444 yaml=$(do_facet mgs $LCTL --device MGS llog_print params |
8445 grep jobid_name | tail -n 1)
8447 local param=$(awk '{ print $10 }' <<< "$yaml")
8448 local val=$(awk '{ print $12 }' <<< "$yaml")
8449 #return to the default
8450 do_facet mgs $LCTL set_param -P jobid_name=$orig_val
8451 [ $val = "testname" ] || error "bad value: $val"
8452 [ $param = "jobid_name," ] || error "Bad param: $param"
8454 run_test 123ab "llog_print params output values from set_param -P"
8456 test_123ac() { # LU-11566
8457 remote_mgs_nodsh && skip "remote MGS with nodsh"
8458 do_facet mgs "$LCTL help llog_print" 2>&1 | grep -q -- --start ||
8459 skip "Need 'lctl llog_print --start' on MGS"
8464 [ -d $MOUNT/.lustre ] || setupall
8466 # - { index: 10, event: add_uuid, nid: 192.168.20.1@tcp(0x20000c0a81401,
8467 # node: 192.168.20.1@tcp }
8468 do_facet mgs $LCTL --device MGS \
8469 llog_print --start $start --end $end $FSNAME-client | tr -d , |
8470 while read DASH BRACE INDEX idx EVENT BLAH BLAH BLAH; do
8471 (( idx >= start )) || error "llog_print index $idx < $start"
8472 (( idx <= end )) || error "llog_print index $idx > $end"
8475 run_test 123ac "llog_print with --start and --end"
8477 test_123ad() { # LU-11566
8478 remote_mgs_nodsh && skip "remote MGS with nodsh"
8479 # older versions of lctl may not print all records properly
8480 do_facet mgs "$LCTL help llog_print" 2>&1 | grep -q -- --start ||
8481 skip "Need 'lctl llog_print --start' on MGS"
8483 [ -d $MOUNT/.lustre ] || setupall
8485 # append a new record, to avoid issues if last record was cancelled
8486 local old=$($LCTL get_param -n osc.*-OST0000-*.max_dirty_mb | head -1)
8487 do_facet mgs $LCTL conf_param $FSNAME-OST0000.osc.max_dirty_mb=$old
8489 # logid: [0x3:0xa:0x0]:0
8493 local num=$(do_facet mgs $LCTL --device MGS llog_info $FSNAME-client |
8494 awk '/last_index:/ { print $2 - 1 }')
8496 # - { index: 71, event: set_timeout, num: 0x14, param: sys.timeout=20 }
8497 local last=$(do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
8498 tail -1 | awk '{ print $4 }' | tr -d , )
8499 (( last == num )) || error "llog_print only showed $last/$num records"
8501 run_test 123ad "llog_print shows all records"
8505 local yaml_file="$TMP/$tfile.yaml"
8506 do_facet mgs rm "$yaml_file"
8507 local cfgfiles=$(do_facet mgs "lctl --device MGS llog_catlist" |
8508 sed 's/config_log://')
8510 # set jobid_var to a different value for test
8511 local orig_val=$(do_facet mgs $LCTL get_param jobid_var)
8512 do_facet mgs $LCTL set_param -P jobid_var="testname"
8514 for i in $cfgfiles params; do
8515 do_facet mgs "lctl --device MGS llog_print ${i} >> $yaml_file"
8518 echo "Unmounting FS"
8528 # Reapply the config from before
8529 echo "Setting configuration parameters"
8530 do_facet mgs "lctl set_param -F $yaml_file"
8532 local set_val=$(do_facet mgs $LCTL get_param jobid_var)
8533 do_facet mgs $LCTL set_param -P $orig_val
8535 [ $set_val == "jobid_var=testname" ] ||
8536 error "$set_val is not testname"
8538 do_facet mgs rm "$yaml_file"
8540 run_test 123F "clear and reset all parameters using set_param -F"
8544 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
8545 [ -z $mds2failover_HOST ] && skip "needs MDT failover setup"
8551 if combined_mgs_mds; then
8552 start_mdt 1 "-o nosvc" ||
8553 error "starting mds with nosvc option failed"
8555 local nid=$(do_facet mds2 $LCTL list_nids | head -1)
8556 local failover_nid=$(do_node $mds2failover_HOST $LCTL list_nids | head -1)
8557 do_facet mgs $LCTL replace_nids $FSNAME-MDT0001 $nid:$failover_nid ||
8558 error "replace_nids execution error"
8560 if combined_mgs_mds; then
8566 echo "lfs setdirstripe"
8567 $LFS setdirstripe -i 1 $MOUNT/$tdir || error "setdirstirpe error"
8570 run_test 124 "check failover after replace_nids"
8572 if ! combined_mgs_mds ; then
8578 # restore the values of MDSSIZE and OSTSIZE
8579 MDSSIZE=$STORED_MDSSIZE
8580 OSTSIZE=$STORED_OSTSIZE
8584 check_and_cleanup_lustre