7 # bug number for skipped test: LU-7428
8 ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT 84"
9 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
13 if [ -r /etc/SuSE-release ]
15 local vers=$(grep VERSION /etc/SuSE-release | awk '{print $3}')
16 local patchlev=$(grep PATCHLEVEL /etc/SuSE-release |
18 if [ $vers -eq 11 ] && [ $patchlev -eq 2 ]
26 if is_sles11; then # LU-2181
27 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 23a 34b"
30 if [ "$FAILURE_MODE" = "HARD" ]; then
31 CONFIG_EXCEPTIONS="24a " &&
32 echo "Except the tests: $CONFIG_EXCEPTIONS for " \
33 "FAILURE_MODE=$FAILURE_MODE, b=23573" &&
34 ALWAYS_EXCEPT="$ALWAYS_EXCEPT $CONFIG_EXCEPTIONS"
37 # bug number for skipped test:
38 # a tool to create lustre filesystem images
39 ALWAYS_EXCEPT="32newtarball $ALWAYS_EXCEPT"
42 PATH=$PWD/$SRCDIR:$SRCDIR:$SRCDIR/../utils:$PATH
44 PTLDEBUG=${PTLDEBUG:--1}
46 LUSTRE=${LUSTRE:-$(dirname $0)/..}
47 RLUSTRE=${RLUSTRE:-$LUSTRE}
48 export MULTIOP=${MULTIOP:-multiop}
50 . $LUSTRE/tests/test-framework.sh
52 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
54 # use small MDS + OST size to speed formatting time
55 # do not use too small MDSSIZE/OSTSIZE, which affect the default journal size
56 # STORED_MDSSIZE is used in test_18
57 STORED_MDSSIZE=$MDSSIZE
58 STORED_OSTSIZE=$OSTSIZE
70 if ! combined_mgs_mds; then
71 # bug number for skipped test: 23954
72 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 24b"
75 # pass "-E lazy_itable_init" to mke2fs to speed up the formatting time
76 if [[ "$LDISKFS_MKFS_OPTS" != *lazy_itable_init* ]]; then
77 LDISKFS_MKFS_OPTS=$(csa_add "$LDISKFS_MKFS_OPTS" -E lazy_itable_init)
80 [ $(facet_fstype $SINGLEMDS) = "zfs" ] &&
81 # bug number for skipped test: LU-4444
82 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 69"
87 require_dsh_mds || exit 0
88 require_dsh_ost || exit 0
91 [ "$SLOW" = "no" ] && EXCEPT_SLOW="45 69"
96 # The MGS must be started before the OSTs for a new fs, so start
97 # and stop to generate the startup logs.
100 wait_osc_import_state mds ost FULL
105 reformat_and_config() {
107 if ! combined_mgs_mds ; then
113 writeconf_or_reformat() {
114 # There are at most 2 OSTs for write_conf test
115 # who knows if/where $TUNEFS is installed?
116 # Better reformat if it fails...
117 writeconf_all $MDSCOUNT 2 ||
118 { echo "tunefs failed, reformatting instead" &&
119 reformat_and_config && return 0; }
129 start mgs $(mgsdevname) $MGS_MOUNT_OPTS
135 local dev=$(mdsdevname $num)
138 echo "start mds service on `facet_active_host $facet`"
139 start $facet ${dev} $MDS_MOUNT_OPTS $@ || return 94
145 local dev=$(mdsdevname $num)
148 echo "stop mds service on `facet_active_host $facet`"
149 # These tests all use non-failover stop
150 stop $facet -f || return 97
156 for num in $(seq $MDSCOUNT); do
157 start_mdt $num $@ || return 94
162 if ! combined_mgs_mds ; then
170 for num in $(seq $MDSCOUNT); do
171 stop_mdt $num || return 97
176 echo "stop mgs service on `facet_active_host mgs`"
177 # These tests all use non-failover stop
178 stop mgs -f || return 97
182 echo "start ost1 service on `facet_active_host ost1`"
183 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS $@ || return 95
187 echo "stop ost1 service on `facet_active_host ost1`"
188 # These tests all use non-failover stop
189 stop ost1 -f || return 98
193 echo "start ost2 service on `facet_active_host ost2`"
194 start ost2 $(ostdevname 2) $OST_MOUNT_OPTS $@ || return 92
198 echo "stop ost2 service on `facet_active_host ost2`"
199 # These tests all use non-failover stop
200 stop ost2 -f || return 93
205 echo "mount $FSNAME on ${MOUNTPATH}....."
206 zconf_mount $(hostname) $MOUNTPATH || return 96
210 local mountopt="remount,$1"
212 echo "remount '$1' lustre on ${MOUNTPATH}....."
213 zconf_mount $(hostname) $MOUNTPATH "$mountopt" || return 96
219 echo "umount lustre on $mountpath....."
220 zconf_umount $HOSTNAME $mountpath $@ || return 97
223 manual_umount_client(){
226 echo "manual umount lustre on ${MOUNT}...."
227 do_facet client "umount ${FORCE} $MOUNT"
233 start_mds || error "MDT start failed"
234 start_ost || error "Unable to start OST1"
235 mount_client $MOUNT || error "client start failed"
236 client_up || error "client_up failed"
245 unload_modules_conf () {
246 if combined_mgs_mds || ! local_mode; then
247 unload_modules || return 1
252 stop_ost || return 202
253 stop_mds || return 201
254 unload_modules_conf || return 203
258 umount_client $MOUNT || return 200
259 cleanup_nocli || return $?
264 echo "umount $MOUNT2 ..."
265 umount $MOUNT2 || true
266 echo "stopping fs2mds ..."
267 stop fs2mds -f || true
268 echo "stopping fs2ost ..."
269 stop fs2ost -f || true
273 do_facet client "cp /etc/passwd $DIR/a" || return 71
274 do_facet client "rm $DIR/a" || return 72
275 # make sure lustre is actually mounted (touch will block,
276 # but grep won't, so do it after)
277 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
278 echo "setup single mount lustre success"
282 do_facet client "touch $DIR/a" || return 71
283 do_facet client "rm $DIR/a" || return 72
284 do_facet client "touch $DIR2/a" || return 73
285 do_facet client "rm $DIR2/a" || return 74
286 echo "setup double mount lustre success"
291 if [ "$ONLY" == "setup" ]; then
296 if [ "$ONLY" == "cleanup" ]; then
303 #create single point mountpoint
309 check_mount || error "check_mount failed"
310 cleanup || error "cleanup failed with $?"
312 run_test 0 "single mount setup"
315 start_mds || error "MDS start failed"
316 start_ost || error "unable to start OST"
317 echo "start ost second time..."
318 start_ost && error "2nd OST start should fail"
319 mount_client $MOUNT || error "client start failed"
320 check_mount || error "check_mount failed"
321 cleanup || error "cleanup failed with $?"
323 run_test 1 "start up ost twice (should return errors)"
326 start_mds || error "MDT start failed"
327 echo "start mds second time.."
328 start_mds && error "2nd MDT start should fail"
329 start_ost || error "OST start failed"
330 mount_client $MOUNT || error "mount_client failed to start client"
331 check_mount || error "check_mount failed"
332 cleanup || error "cleanup failed with $?"
334 run_test 2 "start up mds twice (should return err)"
338 #mount.lustre returns an error if already in mtab
339 mount_client $MOUNT && error "2nd client mount should fail"
340 check_mount || error "check_mount failed"
341 cleanup || error "cleanup failed with $?"
343 run_test 3 "mount client twice (should return err)"
347 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
348 stop_ost || error "Unable to stop OST1"
349 umount_client $MOUNT -f || error “unmount $MOUNT failed”
352 # ok for ost to fail shutdown
353 if [ 202 -ne $eno ] && [ 0 -ne $eno ]; then
354 error "cleanup failed with $?"
357 run_test 4 "force cleanup ost, then cleanup"
359 test_5a() { # was test_5
361 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
362 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
364 stop_mds || error "Unable to stop MDS"
366 # cleanup may return an error from the failed
367 # disconnects; for now I'll consider this successful
368 # if all the modules have unloaded.
372 echo "killing umount"
373 kill -TERM $UMOUNT_PID
374 echo "waiting for umount to finish"
376 if grep " $MOUNT " /proc/mounts; then
377 echo "test 5: /proc/mounts after failed umount"
381 echo "killing umount"
382 kill -TERM $UMOUNT_PID
383 echo "waiting for umount to finish"
385 grep " $MOUNT " /proc/mounts &&
386 error "/proc/mounts after second umount"
389 # manual_mount_client may fail due to umount succeeding above
391 # stop_mds is a no-op here, and should not fail
392 cleanup_nocli || error "cleanup_nocli failed with $?"
393 # df may have lingering entry
395 # mtab may have lingering entry
399 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
401 grep -q $MOUNT" " /etc/mtab || break
402 echo "Waiting /etc/mtab updated ... "
403 WAIT=$(( WAIT + sleep))
405 [ "$WAIT" -eq "$MAX_WAIT" ] &&
406 error "/etc/mtab is not updated in $WAIT secs"
407 echo "/etc/mtab updated in $WAIT secs"
409 run_test 5a "force cleanup mds, then cleanup"
417 grep " $MOUNT " /etc/mtab &&
418 error false "unexpected entry in mtab before mount" && return 10
420 start_ost || error "OST start failed"
421 if ! combined_mgs_mds ; then
422 trap cleanup_5b EXIT ERR
423 start_mds || error "MDS start failed"
427 mount_client $MOUNT && error "mount_client $MOUNT should fail"
428 grep " $MOUNT " /etc/mtab &&
429 error "$MOUNT entry in mtab after failed mount"
431 # stop_mds is a no-op here, and should not fail
432 cleanup_nocli || error "cleanup_nocli failed with $?"
433 if ! combined_mgs_mds ; then
437 run_test 5b "Try to start a client with no MGS (should return errs)"
440 grep " $MOUNT " /etc/mtab &&
441 error false "unexpected entry in mtab before mount" && return 10
443 start_mds || error "MDS start failed"
444 start_ost || error "OST start failed"
445 local oldfs="${FSNAME}"
446 FSNAME="wrong.${FSNAME}"
447 mount_client $MOUNT || :
449 grep " $MOUNT " /etc/mtab &&
450 error "$MOUNT entry in mtab after failed mount"
452 cleanup_nocli || error "cleanup_nocli failed with $?"
454 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
457 grep " $MOUNT " /etc/mtab &&
458 error false "unexpected entry in mtab before mount" && return 10
460 start_ost || error "OST start failed"
461 start_mds || error "MDS start failed"
462 stop_ost || error "Unable to stop OST1"
463 mount_client $MOUNT || error "mount_client $MOUNT failed"
464 umount_client $MOUNT -f || error "umount_client $MOUNT failed"
465 cleanup_nocli || error "cleanup_nocli failed with $?"
466 grep " $MOUNT " /etc/mtab &&
467 error "$MOUNT entry in mtab after unmount"
470 run_test 5d "mount with ost down"
473 grep " $MOUNT " /etc/mtab &&
474 error false "unexpected entry in mtab before mount" && return 10
476 start_mds || error "MDS start failed"
477 start_ost || error "OST start failed"
479 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
480 do_facet client "$LCTL set_param fail_loc=0x80000506"
481 mount_client $MOUNT || echo "mount failed (not fatal)"
482 cleanup || error "cleanup failed with $?"
483 grep " $MOUNT " /etc/mtab &&
484 error "$MOUNT entry in mtab after unmount"
487 run_test 5e "delayed connect, don't crash (bug 10268)"
490 if combined_mgs_mds ; then
491 skip "combined mgs and mds"
495 grep " $MOUNT " /etc/mtab &&
496 error false "unexpected entry in mtab before mount" && return 10
499 start_ost || error "OST start failed"
500 mount_client $MOUNT &
502 echo client_mount pid is $pid
506 if ! ps -f -p $pid >/dev/null; then
509 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
510 error "mount returns $rc, expected to hang"
512 cleanup || error "cleanup failed with $?"
517 start_mds || error "start MDS failed"
519 # mount should succeed after start mds
521 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
522 cleanup || error "final call to cleanup failed with rc $?"
524 run_test 5f "mds down, cleanup after failed mount (bug 2712)"
529 mount_client $MOUNT || error "mount_client $MOUNT failed"
530 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
531 cleanup || error "cleanup failed with rc $?"
533 run_test 6 "manual umount, then mount again"
538 cleanup_nocli || error "cleanup_nocli failed with $?"
540 run_test 7 "manual umount, then cleanup"
544 mount_client $MOUNT2 || error "mount_client $MOUNT2 failed"
545 check_mount2 || error "check_mount2 failed"
546 umount_client $MOUNT2 || error "umount_client $MOUNT2 failed"
547 cleanup || error "cleanup failed with rc $?"
549 run_test 8 "double mount setup"
552 start_ost || error "OST start failed"
554 do_facet ost1 $LCTL set_param debug=\'inode trace\' ||
555 error "do_facet ost1 set_param inode trace failed."
556 do_facet ost1 $LCTL set_param subsystem_debug=\'mds ost\' ||
557 error "do_facet ost1 set_param debug mds ost failed."
559 CHECK_PTLDEBUG="`do_facet ost1 $LCTL get_param -n debug`"
560 if [ "$CHECK_PTLDEBUG" ] && { \
561 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
562 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
563 echo "lnet.debug success"
565 error "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
567 CHECK_SUBSYS="`do_facet ost1 $LCTL get_param -n subsystem_debug`"
568 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
569 echo "lnet.subsystem_debug success"
571 error "lnet.subsystem_debug: want 'mds ost' got '$CHECK_SUBSYS'"
573 stop_ost || error "Unable to stop OST1"
575 run_test 9 "test ptldebug and subsystem for mkfs"
583 do_facet $facet "test -b $dev" || rc=1
584 if [[ "$size" ]]; then
585 local in=$(do_facet $facet "dd if=$dev of=/dev/null bs=1k \
586 count=1 skip=$size 2>&1" |
587 awk '($3 == "in") { print $1 }')
588 [[ $in = "1+0" ]] || rc=1
594 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
595 # But with new MDS stack we don't care about the mode of local objects
596 # anymore, so this test is removed. See bug 22944 for more details.
600 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
601 skip "Only applicable to ldiskfs-based MDTs"
606 check_mount || error "check_mount failed"
607 cleanup || error "cleanup failed with rc $?"
609 echo "Remove mds config log"
610 if ! combined_mgs_mds ; then
614 do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' \
615 $(mgsdevname) || return \$?" ||
616 error "do_facet mgs failed with $?"
618 if ! combined_mgs_mds ; then
622 start_ost || error "OST start failed"
623 start_mds && error "MDS start succeeded, but should fail"
626 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
629 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
630 skip "Only applicable to ldiskfs-based MDTs"
634 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
639 # check if current MDSSIZE is large enough
640 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE &&
641 log "use MDSSIZE=$MDSSIZE"
643 # check if the global config has a large enough MDSSIZE
644 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] &&
645 [ $STORED_MDSSIZE -ge $MIN ] &&
646 OK=1 && myMDSSIZE=$STORED_MDSSIZE &&
647 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
649 # check if the block device is large enough
650 is_blkdev $SINGLEMDS $MDSDEV $MIN
651 local large_enough=$?
652 if [ -n "$OK" ]; then
653 [ $large_enough -ne 0 ] && OK=""
655 [ $large_enough -eq 0 ] && OK=1 && myMDSSIZE=$MIN &&
656 log "use device $MDSDEV with MIN=$MIN"
659 # check if a loopback device has enough space for fs metadata (5%)
661 if [ -z "$OK" ]; then
662 local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! \
663 -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
664 awk '($1 != "Filesystem") { print $4 }')
665 ! [ -z "$SPACE" ] && [ $SPACE -gt $((MIN / 20)) ] &&
666 OK=1 && myMDSSIZE=$MIN &&
667 log "use file $MDSDEV with MIN=$MIN"
670 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS" && return
672 echo "mount mds with large journal..."
674 local OLD_MDSSIZE=$MDSSIZE
678 echo "mount lustre system..."
680 check_mount || error "check_mount failed"
682 echo "check journal size..."
683 local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
684 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
685 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
687 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
690 cleanup || error "cleanup failed with rc $?"
695 run_test 18 "check mkfs creates large journals"
698 start_mds || error "MDS start failed"
699 stop_mds || error "Unable to stop MDS"
701 run_test 19a "start/stop MDS without OSTs"
704 start_ost || error "Unable to start OST1"
705 stop_ost || error "Unable to stop OST1"
707 run_test 19b "start/stop OSTs without MDS"
710 # first format the ost/mdt
711 start_mds || error "MDS start failed"
712 start_ost || error "Unable to start OST1"
713 mount_client $MOUNT || error "mount_client $MOUNT failed"
714 check_mount || error "check_mount failed"
715 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed."
716 remount_client ro $MOUNT || error "remount_client with ro failed"
717 touch $DIR/$tfile && error "$DIR/$tfile created incorrectly"
718 [ -e $DIR/$tfile ] && error "$DIR/$tfile exists incorrectly"
719 remount_client rw $MOUNT || error "remount_client with rw failed"
720 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
721 MCNT=$(grep -c $MOUNT /etc/mtab)
722 [ "$MCNT" -ne 1 ] && error "$MOUNT in /etc/mtab $MCNT times"
724 stop_mds || error "Unable to stop MDS"
725 stop_ost || error "Unable to stop OST1"
727 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
730 start_mds || error "MDS start failed"
731 start_ost || error "unable to start OST1"
732 wait_osc_import_state mds ost FULL
733 stop_ost || error "unable to stop OST1"
734 stop_mds || error "unable to stop MDS"
736 run_test 21a "start mds before ost, stop ost first"
739 start_ost || error "unable to start OST1"
740 start_mds || error "MDS start failed"
741 wait_osc_import_state mds ost FULL
742 stop_mds || error "unable to stop MDS"
743 stop_ost || error "unable to stop OST1"
745 run_test 21b "start ost before mds, stop mds first"
748 start_ost || error "Unable to start OST1"
749 start_mds || error "MDS start failed"
750 start_ost2 || error "Unable to start OST2"
751 wait_osc_import_state mds ost2 FULL
752 stop_ost || error "Unable to stop OST1"
753 stop_ost2 || error "Unable to stop OST2"
754 stop_mds || error "Unable to stop MDS"
755 #writeconf to remove all ost2 traces for subsequent tests
756 writeconf_or_reformat
758 run_test 21c "start mds between two osts, stop mds last"
761 if combined_mgs_mds ; then
762 skip "need separate mgs device" && return 0
768 start_mgs || error "unable to start MGS"
769 start_ost || error "unable to start OST1"
770 start_ost2 || error "unable to start OST2"
771 start_mds || error "MDS start failed"
772 wait_osc_import_state mds ost2 FULL
774 stop_ost || error "Unable to stop OST1"
775 stop_ost2 || error "Unable to stop OST2"
776 stop_mds || error "Unable to stop MDS"
778 #writeconf to remove all ost2 traces for subsequent tests
779 writeconf_or_reformat
780 start_mgs || error "unable to start MGS"
782 run_test 21d "start mgs then ost and then mds"
785 MGSNID="$saved_mgsnid"
787 echo "stopping fs2mgs ..."
788 stop $fs2mgs -f || true
791 test_21e() { # LU-5863
792 if [[ -z "$fs3ost_DEV" || -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
793 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
794 skip_env "mixed loopback and real device not working" && return
797 local fs2mdsdev=$(mdsdevname 1_2)
798 local fs2ostdev=$(ostdevname 1_2)
799 local fs3ostdev=$(ostdevname 2_2)
801 local fs2mdsvdev=$(mdsvdevname 1_2)
802 local fs2ostvdev=$(ostvdevname 1_2)
803 local fs3ostvdev=$(ostvdevname 2_2)
805 # temporarily use fs3ost as fs2mgs
807 local fs2mgsdev=$fs3ostdev
808 local fs2mgsvdev=$fs3ostvdev
810 local fsname=test1234
812 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
813 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
814 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS && trap cleanup_21e EXIT INT ||
815 error "start fs2mgs failed"
817 local saved_mgsnid="$MGSNID"
818 MGSNID=$(do_facet $fs2mgs $LCTL list_nids | xargs | tr ' ' ,)
820 add fs2mds $(mkfs_opts mds1 $fs2mdsdev $fsname) \
821 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
822 add fs2ost $(mkfs_opts ost1 $fs2ostdev $fsname) \
823 --reformat $fs2ostdev $fs2ostvdev || error "add fs2ost failed"
825 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
826 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS || error "start fs2mds failed"
828 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
829 $MOUNT_CMD $MGSNID:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
830 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
834 run_test 21e "separate MGS and MDS"
837 start_mds || error "MDS start failed"
839 echo "Client mount with ost in logs, but none running"
840 start_ost || error "unable to start OST1"
841 # wait until mds connected to ost and open client connection
842 wait_osc_import_state mds ost FULL
843 stop_ost || error "unable to stop OST1"
844 mount_client $MOUNT || error "mount_client $MOUNT failed"
845 # check_mount will block trying to contact ost
846 mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed"
847 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed"
848 umount_client $MOUNT -f
851 echo "Client mount with a running ost"
852 start_ost || error "unable to start OST1"
854 # if gss enabled, wait full time to let connection from
855 # mds to ost be established, due to the mismatch between
856 # initial connect timeout and gss context negotiation timeout.
857 # This perhaps could be remove after AT landed.
858 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
859 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
861 mount_client $MOUNT || error "mount_client $MOUNT failed"
862 wait_osc_import_state mds ost FULL
863 wait_osc_import_state client ost FULL
864 check_mount || error "check_mount failed"
867 cleanup || error "cleanup failed with rc $?"
869 run_test 22 "start a client before osts (should return errs)"
871 test_23a() { # was test_23
874 stop $SINGLEMDS || error "failed to stop $SINGLEMDS"
875 # force down client so that recovering mds waits for reconnect
876 local running=$(grep -c $MOUNT /proc/mounts) || true
877 if [ $running -ne 0 ]; then
878 echo "Stopping client $MOUNT (opts: -f)"
882 # enter recovery on failed mds
883 local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/})
884 start $SINGLEMDS $MDT_DEV $MDS_MOUNT_OPTS || error "MDS start failed"
885 # try to start a new client
886 mount_client $MOUNT &
888 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
889 MOUNT_LUSTRE_PID=$(ps -ef | grep mount.lustre |
890 grep -v grep | awk '{print $2}')
891 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
893 ps --ppid $MOUNT_LUSTRE_PID
894 echo "waiting for mount to finish"
896 # "ctrl-c" sends SIGINT but it usually (in script) does not work on child process
897 # SIGTERM works but it does not spread to offspring processses
898 kill -s TERM $MOUNT_PID
899 kill -s TERM $MOUNT_LUSTRE_PID
900 # we can not wait $MOUNT_PID because it is not a child of this shell
906 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
908 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
909 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
912 [ -z "$PID1" -a -z "$PID2" ] && break
913 echo "waiting for mount to finish ... "
914 WAIT=$(( WAIT + sleep))
916 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
917 error "MOUNT_PID $MOUNT_PID and "\
918 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
921 stop_mds || error "stopping MDSes failed"
922 stop_ost || error "stopping OSSes failed"
924 run_test 23a "interrupt client during recovery mount delay"
929 test_23b() { # was test_23
930 start_mds || error "MDS start failed"
931 start_ost || error "Unable to start OST1"
932 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
933 $LCTL set_param fail_loc=0x80000313
935 cleanup || error "cleanup failed with rc $?"
937 run_test 23b "Simulate -EINTR during mount"
940 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
942 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
943 is_blkdev $SINGLEMDS $MDSDEV &&
944 skip_env "mixed loopback and real device not working" && return
947 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
949 local fs2mdsdev=$(mdsdevname 1_2)
950 local fs2ostdev=$(ostdevname 1_2)
951 local fs2mdsvdev=$(mdsvdevname 1_2)
952 local fs2ostvdev=$(ostvdevname 1_2)
954 # test 8-char fsname as well
955 local FSNAME2=test1234
957 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --nomgs --mgsnode=$MGSNID \
958 --fsname=${FSNAME2} --reformat $fs2mdsdev $fs2mdsvdev || exit 10
960 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --fsname=${FSNAME2} \
961 --reformat $fs2ostdev $fs2ostvdev || exit 10
964 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
965 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
966 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
967 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
969 check_mount || error "check_mount failed"
970 # files written on 1 should not show up on 2
971 cp /etc/passwd $DIR/$tfile
973 [ -e $MOUNT2/$tfile ] && error "File bleed"
976 cp /etc/passwd $MOUNT2/$tfile ||
977 error "cp /etc/passwd $MOUNT2/$tfile failed"
978 rm $MOUNT2/$tfile || error "remove $MOUNT2/$tfile failed"
979 # 2 is actually mounted
980 grep $MOUNT2' ' /proc/mounts > /dev/null || error "$MOUNT2 not mounted"
982 facet_failover fs2mds
983 facet_failover fs2ost
986 # the MDS must remain up until last MDT
988 MDS=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
989 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
990 [ -z "$MDS" ] && error "No MDT"
992 cleanup_nocli || error "cleanup_nocli failed with rc $?"
994 run_test 24a "Multiple MDTs on a single node"
997 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
999 if [ -z "$fs2mds_DEV" ]; then
1000 local dev=${SINGLEMDS}_dev
1001 local MDSDEV=${!dev}
1002 is_blkdev $SINGLEMDS $MDSDEV &&
1003 skip_env "mixed loopback and real device not working" && return
1006 local fs2mdsdev=$(mdsdevname 1_2)
1007 local fs2mdsvdev=$(mdsvdevname 1_2)
1009 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --mgs --fsname=${FSNAME}2 \
1010 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
1012 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS &&
1013 error "start MDS should fail"
1015 cleanup || error "cleanup failed with rc $?"
1017 run_test 24b "Multiple MGSs on a single node (should return err)"
1021 check_mount || error "check_mount failed"
1022 local MODULES=$($LCTL modules | awk '{ print $2 }')
1023 rmmod $MODULES 2>/dev/null || true
1024 cleanup || error "cleanup failed with $?"
1026 run_test 25 "Verify modules are referenced"
1030 # we need modules before mount for sysctl, so make sure...
1031 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
1032 #define OBD_FAIL_MDS_FS_SETUP 0x135
1033 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000135"
1034 start_mds && error "MDS started but should not have started"
1035 $LCTL get_param -n devices
1036 DEVS=$($LCTL get_param -n devices | egrep -v MG | wc -l)
1037 [ $DEVS -gt 0 ] && error "number of devices is $DEVS, should be zero"
1038 # start mds to drop writeconf setting
1039 start_mds || error "Unable to start MDS"
1040 stop_mds || error "Unable to stop MDS"
1041 unload_modules_conf || error "unload_modules_conf failed with $?"
1043 run_test 26 "MDT startup failure cleans LOV (should return errs)"
1046 start_ost || error "Unable to start OST1"
1047 start_mds || error "Unable to start MDS"
1048 echo "Requeue thread should have started: "
1049 ps -e | grep ll_cfg_requeue
1050 set_conf_param_and_check ost1 \
1051 "$LCTL get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" \
1052 "$FSNAME-OST0000.ost.client_cache_seconds" ||
1053 error "set_conf_param_and_check ost1 failed"
1054 cleanup_nocli || error "cleanup_nocli failed with rc $?"
1056 run_test 27a "Reacquire MGS lock if OST started first"
1061 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
1062 awk '($3 ~ "mdt" && $4 ~ "MDT0000") { print $4 }')
1064 facet_failover $SINGLEMDS
1065 set_conf_param_and_check $SINGLEMDS \
1066 "$LCTL get_param -n mdt.$device.identity_acquire_expire" \
1067 "$device.mdt.identity_acquire_expire" ||
1068 error "set_conf_param_and_check $SINGLEMDS failed"
1069 set_conf_param_and_check client \
1070 "$LCTL get_param -n mdc.$device-mdc-*.max_rpcs_in_flight"\
1071 "$device.mdc.max_rpcs_in_flight" ||
1072 error "set_conf_param_and_check client failed"
1074 cleanup || error "cleanup failed with $?"
1076 run_test 27b "Reacquire MGS lock after failover"
1080 TEST="$LCTL get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1081 PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
1083 FINAL=$(($ORIG + 1))
1084 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL ||
1085 error "first set_conf_param_and_check client failed"
1086 FINAL=$(($FINAL + 1))
1087 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL ||
1088 error "second set_conf_param_and_check client failed"
1089 umount_client $MOUNT || error "umount_client $MOUNT failed"
1090 mount_client $MOUNT || error "mount_client $MOUNT failed"
1092 if [ $RESULT -ne $FINAL ]; then
1093 error "New config not seen: wanted $FINAL got $RESULT"
1095 echo "New config success: got $RESULT"
1097 set_conf_param_and_check client "$TEST" "$PARAM" $ORIG ||
1098 error "third set_conf_param_and_check client failed"
1099 cleanup || error "cleanup failed with rc $?"
1101 run_test 28 "permanent parameter setting"
1103 test_28a() { # LU-4221
1104 [[ $(lustre_version_code ost1) -ge $(version_code 2.5.52) ]] ||
1105 { skip "Need OST version at least 2.5.52" && return 0; }
1106 [ "$(facet_fstype ost1)" = "zfs" ] &&
1107 skip "LU-4221: no such proc params for ZFS OSTs" && return
1114 local device="$FSNAME-OST0000"
1118 # In this test we will set three kinds of proc parameters with
1120 # 1. the ones moved from the OFD to the OSD, and only their
1121 # symlinks kept in obdfilter
1122 # 2. non-symlink ones in the OFD
1123 # 3. non-symlink ones in the OSD
1126 # prepare a symlink parameter in the OFD
1127 name="writethrough_cache_enable"
1128 param="$device.ost.$name"
1129 cmd="$LCTL get_param -n obdfilter.$device.$name"
1131 # conf_param the symlink parameter in the OFD
1132 old=$(do_facet ost1 $cmd)
1133 new=$(((old + 1) % 2))
1134 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1135 error "lctl conf_param $device.ost.$param=$new failed"
1137 # conf_param the target parameter in the OSD
1138 param="$device.osd.$name"
1139 cmd="$LCTL get_param -n osd-*.$device.$name"
1140 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1141 error "lctl conf_param $device.osd.$param=$old failed"
1144 # prepare a non-symlink parameter in the OFD
1145 name="client_cache_seconds"
1146 param="$device.ost.$name"
1147 cmd="$LCTL get_param -n obdfilter.$device.$name"
1149 # conf_param the parameter in the OFD
1150 old=$(do_facet ost1 $cmd)
1152 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1153 error "lctl conf_param $device.ost.$param=$new failed"
1154 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1155 error "lctl conf_param $device.ost.$param=$old failed"
1158 # prepare a non-symlink parameter in the OSD
1160 param="$device.osd.$name"
1161 cmd="$LCTL get_param -n osd-*.$device.$name"
1163 # conf_param the parameter in the OSD
1164 old=$(do_facet ost1 $cmd)
1165 new=$(((old + 1) % 2))
1166 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1167 error "lctl conf_param $device.osd.$param=$new failed"
1168 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1169 error "lctl conf_param $device.osd.$param=$old failed"
1171 cleanup || error "cleanup failed with $?"
1173 run_test 28a "set symlink parameters permanently with conf_param"
1176 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
1177 setup > /dev/null 2>&1
1178 start_ost2 || error "Unable to start OST2"
1181 local PARAM="$FSNAME-OST0001.osc.active"
1182 local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
1183 local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
1185 ACTV=$($LCTL get_param -n $PROC_ACT)
1187 set_conf_param_and_check client \
1188 "$LCTL get_param -n $PROC_ACT" "$PARAM" $DEAC ||
1189 error "set_conf_param_and_check client failed"
1190 # also check ost_server_uuid status
1191 RESULT=$($LCTL get_param -n $PROC_UUID | grep DEACTIV)
1192 if [ -z "$RESULT" ]; then
1193 error "Client not deactivated: $($LCTL get_param \
1196 echo "Live client success: got $RESULT"
1200 for num in $(seq $MDSCOUNT); do
1201 local mdtosc=$(get_mdtosc_proc_path mds${num} $FSNAME-OST0001)
1202 local MPROC="osc.$mdtosc.active"
1207 RESULT=$(do_facet mds${num} "$LCTL get_param -n $MPROC")
1208 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
1209 if [ $RESULT -eq $DEAC ]; then
1210 echo -n "MDT deactivated also after"
1211 echo "$WAIT sec (got $RESULT)"
1215 if [ $WAIT -eq $MAX ]; then
1216 error "MDT active: wanted $DEAC got $RESULT"
1218 echo "Waiting $(($MAX - $WAIT))secs for MDT deactivated"
1221 # test new client starts deactivated
1222 umount_client $MOUNT || error "umount_client $MOUNT failed"
1223 mount_client $MOUNT || error "mount_client $MOUNT failed"
1224 RESULT=$($LCTL get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
1225 if [ -z "$RESULT" ]; then
1226 error "New client start active: $(lctl get_param -n $PROC_UUID)"
1228 echo "New client success: got $RESULT"
1231 # make sure it reactivates
1232 set_conf_param_and_check client \
1233 "$LCTL get_param -n $PROC_ACT" "$PARAM" $ACTV ||
1234 error "lctl get_param $PROC_ACT $PARAM $ACTV failed"
1236 umount_client $MOUNT
1237 stop_ost2 || error "Unable to stop OST2"
1238 cleanup_nocli || error "cleanup_nocli failed with $?"
1239 #writeconf to remove all ost2 traces for subsequent tests
1240 writeconf_or_reformat
1242 run_test 29 "permanently remove an OST"
1247 echo Big config llog
1248 TEST="$LCTL get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1250 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1251 for i in ${LIST[@]}; do
1252 set_conf_param_and_check client "$TEST" \
1253 "$FSNAME.llite.max_read_ahead_whole_mb" $i ||
1254 error "Set $FSNAME.llite.max_read_ahead_whole_mb failed"
1256 # make sure client restart still works
1257 umount_client $MOUNT
1258 mount_client $MOUNT || error "mount_client $MOUNT failed"
1259 [ "$($TEST)" -ne "$i" ] &&
1260 error "Param didn't stick across restart $($TEST) != $i"
1263 echo Erase parameter setting
1264 do_facet mgs "$LCTL conf_param \
1265 -d $FSNAME.llite.max_read_ahead_whole_mb" ||
1266 error "Erase param $FSNAME.llite.max_read_ahead_whole_mb failed"
1267 umount_client $MOUNT
1268 mount_client $MOUNT || error "mount_client $MOUNT failed"
1270 echo "deleted (default) value=$FINAL, orig=$ORIG"
1271 # assumes this parameter started at the default value
1272 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
1274 cleanup || error "cleanup failed with rc $?"
1276 run_test 30a "Big config llog and conf_param deletion"
1281 local orignids=$($LCTL get_param -n \
1282 osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1284 local orignidcount=$(echo "$orignids" | wc -w)
1286 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1287 # numerical part of it. Hopefully that's not already a failover address
1289 local OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | \
1291 local ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1292 local NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1293 local NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1294 echo "Using fake nid $NEW"
1296 local TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1297 grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1298 set_conf_param_and_check client "$TEST" \
1299 "$FSNAME-OST0000.failover.node" $NEW ||
1300 error "didn't add failover nid $NEW"
1301 local NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1304 local NIDCOUNT=$(echo "$NIDS" | wc -w)
1305 echo "should have $((orignidcount + 1)) entries \
1306 in failover nids string, have $NIDCOUNT"
1307 [ $NIDCOUNT -eq $((orignidcount + 1)) ] ||
1308 error "Failover nid not added"
1310 do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" ||
1311 error "conf_param delete failed"
1312 umount_client $MOUNT
1313 mount_client $MOUNT || error "mount_client $MOUNT failed"
1315 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1318 NIDCOUNT=$(echo "$NIDS" | wc -w)
1319 echo "only $orignidcount final entries should remain \
1320 in failover nids string, have $NIDCOUNT"
1321 [ $NIDCOUNT -eq $orignidcount ] || error "Failover nids not removed"
1323 cleanup || error "cleanup failed with rc $?"
1325 run_test 30b "Remove failover nids"
1327 test_31() { # bug 10734
1328 # ipaddr must not exist
1329 $MOUNT_CMD 4.3.2.1@tcp:/lustre $MOUNT || true
1330 cleanup || error "cleanup failed with rc $?"
1332 run_test 31 "Connect to non-existent node (shouldn't crash)"
1336 T32_BLIMIT=40960 # Kbytes
1340 # This is not really a test but a tool to create new disk
1341 # image tarballs for the upgrade tests.
1343 # Disk image tarballs should be created on single-node
1344 # clusters by running this test with default configurations
1345 # plus a few mandatory environment settings that are verified
1346 # at the beginning of the test.
1348 test_32newtarball() {
1352 local tmp=$TMP/t32_image_create
1353 local server_version=$(lustre_version_code $SINGLEMDS)
1358 if [ $FSNAME != t32fs -o \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o \
1359 $OSTCOUNT -ne 1 -o -z "$OSTDEV1" ]; then
1360 error "Needs FSNAME=t32fs MDSCOUNT=2 " \
1361 "MDSDEV1=<nonexistent_file>" \
1362 "MDSDEV2=<nonexistent_file>" \
1363 "(or MDSDEV, in the case of b1_8)" \
1364 "OSTCOUNT=1 OSTDEV1=<nonexistent_file>"
1368 echo "Found stale $tmp"
1372 mkdir $tmp/src || return 1
1373 tar cf - -C $src . | tar xf - -C $tmp/src
1374 dd if=/dev/zero of=$tmp/src/t32_qf_old bs=1M \
1375 count=$(($T32_BLIMIT / 1024 / 4))
1376 chown $T32_QID.$T32_QID $tmp/src/t32_qf_old
1378 # format ost with comma-separated NIDs to verify LU-4460
1379 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
1380 MGSNID="$MGSNID,$MGSNID" OSTOPT="--failnode=$failnid" formatall
1384 [[ $server_version -ge $(version_code 2.3.50) ]] ||
1385 $LFS quotacheck -ug /mnt/$FSNAME
1386 $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
1389 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1391 if [[ $MDSCOUNT -ge 2 ]]; then
1392 remote_dir=/mnt/$FSNAME/remote_dir
1393 $LFS mkdir -i 1 $remote_dir
1394 tar cf - -C $tmp/src . | tar xf - -C $remote_dir
1396 if [[ $server_version -ge $(version_code 2.7.0) ]]; then
1397 striped_dir=/mnt/$FSNAME/striped_dir_old
1398 $LFS mkdir -i 1 -c 2 $striped_dir
1399 tar cf - -C $tmp/src . | tar xf - -C $striped_dir
1405 mkdir $tmp/img || return 1
1409 pushd_dir=/mnt/$FSNAME
1410 if [[ $MDSCOUNT -ge 2 ]]; then
1411 pushd_dir=$remote_dir
1412 if [[ $server_version -ge $(version_code 2.7.0) ]]; then
1414 ls -Rni --time-style=+%s >$tmp/img/list2
1420 ls -Rni --time-style=+%s >$tmp/img/list
1421 find ! -name .lustre -type f -exec sha1sum {} \; |
1422 sort -k 2 >$tmp/img/sha1sums
1424 $LCTL get_param -n version | head -n 1 |
1425 sed -e 's/^lustre: *//' >$tmp/img/commit
1427 [[ $server_version -ge $(version_code 2.3.50) ]] ||
1428 $LFS quotaon -ug /mnt/$FSNAME
1429 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1430 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1431 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1432 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1433 | tr -d "*" > $tmp/img/bspace
1434 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1435 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1436 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1437 | tr -d "*" > $tmp/img/ispace
1438 echo $T32_BLIMIT > $tmp/img/blimit
1439 echo $T32_ILIMIT > $tmp/img/ilimit
1444 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1447 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1448 echo "Data verification failed"
1451 uname -r >$tmp/img/kernel
1452 uname -m >$tmp/img/arch
1454 mv ${MDSDEV1:-$MDSDEV} $tmp/img
1455 for num in $(seq 2 $MDSCOUNT); do
1456 local devname=$(mdsdevname $num)
1458 [[ $(facet_fstype $facet) != zfs ]] ||
1459 devname=$(mdsvdevname $num)
1460 mv $devname $tmp/img
1462 mv $OSTDEV1 $tmp/img
1464 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1465 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1468 tar cjvf $dst/disk$version-$(facet_fstype $SINGLEMDS).tar.bz2 -S *
1473 #run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1476 # The list of applicable tarballs is returned via the caller's
1477 # variable "tarballs".
1480 local node=$(facet_active_host $SINGLEMDS)
1481 local r="do_node $node"
1483 if [ "$CLIENTONLY" ]; then
1484 skip "Client-only testing"
1488 if ! $r which $TUNEFS; then
1489 skip_env "tunefs.lustre required on $node"
1493 local IMGTYPE=$(facet_fstype $SINGLEMDS)
1495 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 -name \'disk*-$IMGTYPE.tar.bz2\')
1497 if [ -z "$tarballs" ]; then
1498 skip "No applicable tarballs found"
1503 t32_test_cleanup() {
1505 local facet=$SINGLEMDS
1506 local fstype=$(facet_fstype $facet)
1509 if $shall_cleanup_lustre; then
1510 umount $tmp/mnt/lustre || rc=$?
1512 if $shall_cleanup_mdt; then
1513 $r $UMOUNT $tmp/mnt/mdt || rc=$?
1515 if $shall_cleanup_mdt1; then
1516 $r $UMOUNT $tmp/mnt/mdt1 || rc=$?
1518 if $shall_cleanup_ost; then
1519 $r $UMOUNT $tmp/mnt/ost || rc=$?
1524 if [[ $fstype == zfs ]]; then
1526 local poolname_list="t32fs-mdt1 t32fs-ost1"
1528 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1530 for poolname in $poolname_list; do
1531 destroy_zpool $facet $poolname
1537 t32_bits_per_long() {
1539 # Yes, this is not meant to be perfect.
1549 t32_reload_modules() {
1551 local all_removed=false
1554 while ((i < 20)); do
1555 echo "Unloading modules on $node: Attempt $i"
1556 do_rpc_nodes $node $LUSTRE_RMMOD $(facet_fstype $SINGLEMDS) &&
1558 do_rpc_nodes $node check_mem_leak || return 1
1559 if $all_removed; then
1560 do_rpc_nodes $node load_modules
1566 echo "Unloading modules on $node: Given up"
1570 t32_wait_til_devices_gone() {
1576 echo wait for devices to go
1577 while ((i < 20)); do
1578 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1579 loops=$(do_rpc_nodes $node losetup -a | grep -c t32)
1580 ((devices == 0 && loops == 0)) && return 0
1584 echo "waiting for dev on $node: dev $devices loop $loops given up"
1585 do_rpc_nodes $node "losetup -a"
1586 do_rpc_nodes $node "$LCTL device_list"
1590 t32_verify_quota() {
1594 local fstype=$(facet_fstype $SINGLEMDS)
1598 $LFS quota -u $T32_QID -v $mnt
1600 qval=$($LFS quota -v -u $T32_QID $mnt |
1601 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1602 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1604 [ $qval -eq $img_bspace ] || {
1605 echo "bspace, act:$qval, exp:$img_bspace"
1609 qval=$($LFS quota -v -u $T32_QID $mnt |
1610 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1611 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1613 [ $qval -eq $img_ispace ] || {
1614 echo "ispace, act:$qval, exp:$img_ispace"
1618 qval=$($LFS quota -v -u $T32_QID $mnt |
1619 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1620 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1622 [ $qval -eq $img_blimit ] || {
1623 echo "blimit, act:$qval, exp:$img_blimit"
1627 qval=$($LFS quota -v -u $T32_QID $mnt |
1628 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1629 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1631 [ $qval -eq $img_ilimit ] || {
1632 echo "ilimit, act:$qval, exp:$img_ilimit"
1636 do_node $node $LCTL conf_param $fsname.quota.mdt=ug
1637 cmd="$LCTL get_param -n osd-$fstype.$fsname-MDT0000"
1638 cmd=$cmd.quota_slave.enabled
1639 wait_update $node "$cmd" "ug" || {
1640 echo "Enable mdt quota failed"
1644 do_node $node $LCTL conf_param $fsname.quota.ost=ug
1645 cmd="$LCTL get_param -n osd-$fstype.$fsname-OST0000"
1646 cmd=$cmd.quota_slave.enabled
1647 wait_update $node "$cmd" "ug" || {
1648 echo "Enable ost quota failed"
1653 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1654 bs=1M count=$((img_blimit / 1024)) oflag=sync && {
1655 echo "Write succeed, but expect -EDQUOT"
1658 rm -f $mnt/t32_qf_new
1660 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1662 echo "Create succeed, but expect -EDQUOT"
1665 unlinkmany $mnt/t32_qf_ $img_ilimit
1673 local dne_upgrade=${dne_upgrade:-"no"}
1674 local ff_convert=${ff_convert:-"no"}
1675 local shall_cleanup_mdt=false
1676 local shall_cleanup_mdt1=false
1677 local shall_cleanup_ost=false
1678 local shall_cleanup_lustre=false
1679 local mdt2_is_available=false
1680 local node=$(facet_active_host $SINGLEMDS)
1681 local r="do_node $node"
1682 local node2=$(facet_active_host mds2)
1692 local nid=$($r $LCTL list_nids | head -1)
1698 local fstype=$(facet_fstype $SINGLEMDS)
1699 local mdt_dev=$tmp/mdt
1700 local mdt2_dev=$tmp/mdt2
1701 local ost_dev=$tmp/ost
1706 trap 'trap - RETURN; t32_test_cleanup' RETURN
1709 mkdir -p $tmp/mnt/lustre || error "mkdir $tmp/mnt/lustre failed"
1710 $r mkdir -p $tmp/mnt/{mdt,mdt1,ost}
1711 $r tar xjvf $tarball -S -C $tmp || {
1712 error_noexit "Unpacking the disk image tarball"
1715 img_commit=$($r cat $tmp/commit)
1716 img_kernel=$($r cat $tmp/kernel)
1717 img_arch=$($r cat $tmp/arch)
1718 img_bspace=$($r cat $tmp/bspace)
1719 img_ispace=$($r cat $tmp/ispace)
1721 # older images did not have "blimit" and "ilimit" files
1722 # use old values for T32_BLIMIT and T32_ILIMIT
1723 $r test -f $tmp/blimit && img_blimit=$($r cat $tmp/blimit) ||
1725 $r test -f $tmp/ilimit && img_ilimit=$($r cat $tmp/ilimit) ||
1728 echo "Upgrading from $(basename $tarball), created with:"
1729 echo " Commit: $img_commit"
1730 echo " Kernel: $img_kernel"
1731 echo " Arch: $img_arch"
1732 echo "OST version: $(lustre_build_version ost1)"
1734 # The conversion can be made only when both of the following
1735 # conditions are satisfied:
1736 # - ost device img version < 2.3.64
1737 # - ost server version >= 2.5
1738 [ $(version_code $img_commit) -ge $(version_code 2.3.64) -o \
1739 $(lustre_version_code ost1) -lt $(version_code 2.5.0) ] &&
1742 ! $r test -f $mdt2_dev || mdt2_is_available=true
1744 if [[ $fstype == zfs ]]; then
1747 local poolname_list="t32fs-mdt1 t32fs-ost1"
1749 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1751 for poolname in $poolname_list; do
1752 $r "$ZPOOL list -H $poolname >/dev/null 2>&1 ||
1753 $ZPOOL import -f -d $tmp $poolname"
1756 mdt_dev=t32fs-mdt1/mdt1
1757 ost_dev=t32fs-ost1/ost1
1758 ! $mdt2_is_available || mdt2_dev=t32fs-mdt2/mdt2
1759 wait_update_facet $SINGLEMDS "$ZPOOL list |
1760 awk '/^t32fs-mdt1/ { print \\\$1 }'" "t32fs-mdt1" || {
1761 error_noexit "import zfs pool failed"
1766 $r $LCTL set_param debug="$PTLDEBUG"
1768 $r $TUNEFS --dryrun $mdt_dev || {
1770 error_noexit "tunefs.lustre before mounting the MDT"
1774 if $mdt2_is_available; then
1775 $r $TUNEFS --dryrun $mdt2_dev || {
1777 error_noexit "tunefs.lustre before mounting the MDT"
1782 if [ "$writeconf" ]; then
1784 if [ $fstype == "ldiskfs" ]; then
1786 $r $TUNEFS --quota $mdt_dev || {
1788 error_noexit "Enable mdt quota feature"
1791 if $mdt2_is_available; then
1792 $r $TUNEFS --quota $mdt2_dev || {
1794 error_noexit "Enable mdt quota feature"
1800 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
1801 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
1802 { skip "LU-2200: Cannot run over Inifiniband w/o lctl replace_nids "
1803 "(Need MGS version at least 2.3.59)"; return 0; }
1805 local osthost=$(facet_active_host ost1)
1806 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
1809 if [ $fstype == "ldiskfs" ]; then
1812 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt
1813 $r $LCTL replace_nids $fsname-OST0000 $ostnid
1814 $r $LCTL replace_nids $fsname-MDT0000 $nid
1815 $r $UMOUNT $tmp/mnt/mdt
1818 mopts=exclude=$fsname-OST0000
1819 if [ $fstype == "ldiskfs" ]; then
1824 t32_wait_til_devices_gone $node
1826 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
1828 error_noexit "Mounting the MDT"
1831 shall_cleanup_mdt=true
1833 if $mdt2_is_available; then
1834 mopts=mgsnode=$nid,$mopts
1835 $r $MOUNT_CMD -o $mopts $mdt2_dev $tmp/mnt/mdt1 || {
1837 error_noexit "Mounting the MDT"
1841 echo "mount new MDT....$mdt2_dev"
1842 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1843 error_noexit "enable remote dir create failed"
1845 shall_cleanup_mdt1=true
1846 elif [ "$dne_upgrade" != "no" ]; then
1847 local fs2mdsdev=$(mdsdevname 1_2)
1848 local fs2mdsvdev=$(mdsvdevname 1_2)
1850 echo "mkfs new MDT on ${fs2mdsdev}...."
1851 if [ $(facet_fstype mds1) == ldiskfs ]; then
1852 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
1855 add fs2mds $(mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
1856 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
1857 error_noexit "Mkfs new MDT failed"
1861 [[ $(facet_fstype mds1) != zfs ]] || import_zpool fs2mds
1863 $r $TUNEFS --dryrun $fs2mdsdev || {
1864 error_noexit "tunefs.lustre before mounting the MDT"
1868 echo "mount new MDT....$fs2mdsdev"
1869 $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
1870 error_noexit "mount mdt1 failed"
1874 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1875 error_noexit "enable remote dir create failed"
1877 shall_cleanup_mdt1=true
1880 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
1881 error_noexit "Getting MDT UUID"
1884 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
1885 error_noexit "Unexpected MDT UUID: \"$uuid\""
1889 $r $TUNEFS --dryrun $ost_dev || {
1890 error_noexit "tunefs.lustre before mounting the OST"
1893 if [ "$writeconf" ]; then
1894 mopts=mgsnode=$nid,$writeconf
1895 if [ $fstype == "ldiskfs" ]; then
1897 $r $TUNEFS --quota $ost_dev || {
1899 error_noexit "Enable ost quota feature"
1905 if [ $fstype == "ldiskfs" ]; then
1909 $r $MOUNT_CMD -o $mopts $ost_dev $tmp/mnt/ost || {
1910 error_noexit "Mounting the OST"
1913 shall_cleanup_ost=true
1915 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
1916 error_noexit "Getting OST UUID"
1919 if [ "$uuid" != $fsname-OST0000_UUID ]; then
1920 error_noexit "Unexpected OST UUID: \"$uuid\""
1924 $r $LCTL conf_param $fsname-OST0000.osc.max_dirty_mb=15 || {
1925 error_noexit "Setting \"max_dirty_mb\""
1928 $r $LCTL conf_param $fsname-OST0000.failover.node=$nid || {
1929 error_noexit "Setting OST \"failover.node\""
1932 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
1933 error_noexit "Setting \"max_rpcs_in_flight\""
1936 $r $LCTL conf_param $fsname-MDT0000.failover.node=$nid || {
1937 error_noexit "Setting MDT \"failover.node\""
1940 $r $LCTL pool_new $fsname.interop || {
1941 error_noexit "Setting \"interop\""
1944 $r $LCTL conf_param $fsname-MDT0000.lov.stripesize=4M || {
1945 error_noexit "Setting \"lov.stripesize\""
1948 $r $LCTL conf_param $fsname-MDT0000.mdd.atime_diff=70 || {
1949 error_noexit "Setting \"mdd.atime_diff\""
1953 if [ "$ff_convert" != "no" -a $(facet_fstype ost1) == "ldiskfs" ]; then
1954 $r $LCTL lfsck_start -M $fsname-OST0000 || {
1955 error_noexit "Start OI scrub on OST0"
1959 # The oi_scrub should be on ost1, but for test_32(),
1960 # all on the SINGLEMDS.
1961 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
1962 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1963 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
1964 error_noexit "Failed to get the expected 'completed'"
1968 local UPDATED=$($r $LCTL get_param -n \
1969 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1970 awk '/^updated/ { print $2 }')
1971 [ $UPDATED -ge 1 ] || {
1972 error_noexit "Only $UPDATED objects have been converted"
1977 if [ "$dne_upgrade" != "no" ]; then
1978 $r $LCTL conf_param \
1979 $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
1980 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
1983 $r $LCTL conf_param $fsname-MDT0001.failover.node=$nid || {
1984 error_noexit "Setting MDT1 \"failover.node\""
1987 $r $LCTL conf_param $fsname-MDT0001.lov.stripesize=4M || {
1988 error_noexit "Setting MDT1 \"lov.stripesize\""
1993 if [ "$writeconf" ]; then
1994 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
1995 error_noexit "Mounting the client"
1999 shall_cleanup_lustre=true
2000 $r $LCTL set_param debug="$PTLDEBUG"
2002 t32_verify_quota $node $fsname $tmp/mnt/lustre || {
2003 error_noexit "verify quota failed"
2007 if $r test -f $tmp/list; then
2009 # There is not a Test Framework API to copy files to or
2010 # from a remote node.
2012 # LU-2393 - do both sorts on same node to ensure locale
2014 local list_file=$tmp/list
2016 if $mdt2_is_available; then
2017 if [[ -d $tmp/mnt/lustre/striped_dir_old ]] &&
2018 $r test -f $tmp/list2; then
2019 list_file=$tmp/list2
2020 pushd $tmp/mnt/lustre/striped_dir_old
2022 pushd $tmp/mnt/lustre/remote_dir
2025 pushd $tmp/mnt/lustre
2027 $r cat $list_file | sort -k 6 >$tmp/list.orig
2028 ls -Rni --time-style=+%s | sort -k 6 |
2029 sed 's/\. / /' >$tmp/list || {
2035 # 32-bit and 64-bit clients use different algorithms to
2036 # convert FIDs into inode numbers. Hence, remove the
2037 # inode numbers from the lists, if the original list was
2038 # created on an architecture with different number of
2041 if [ $(t32_bits_per_long $(uname -m)) != \
2042 $(t32_bits_per_long $img_arch) ]; then
2043 echo "Different number of bits per \"long\"" \
2044 "from the disk image"
2045 for list in list.orig list; do
2046 sed -i -e 's/^[0-9]\+[ \t]\+//' \
2050 if ! diff -ub $tmp/list.orig $tmp/list; then
2051 error_noexit "list verification failed"
2055 echo "list verification skipped"
2058 if [ "$dne_upgrade" != "no" ]; then
2059 $LFS mkdir -i 1 -c2 $tmp/mnt/lustre/striped_dir || {
2060 error_noexit "set striped dir failed"
2064 $LFS setdirstripe -D -c2 $tmp/mnt/lustre/striped_dir
2066 pushd $tmp/mnt/lustre
2067 tar -cf - . --exclude=./striped_dir \
2068 --exclude=./striped_dir_old \
2069 --exclude=./remote_dir |
2070 tar -xvf - -C striped_dir 1>/dev/null || {
2071 error_noexit "cp to striped dir failed"
2077 # If it is upgrade from DNE (2.5), then rename the remote dir,
2078 # which is created in 2.5 to striped dir.
2079 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]]; then
2080 stripe_index=$($LFS getdirstripe -i \
2081 $tmp/mnt/lustre/remote_dir)
2083 [[ $stripe_index -eq 1 ]] || {
2084 error_noexit "get index \"$stripe_index\"" \
2085 "from remote dir failed"
2088 mv $tmp/mnt/lustre/remote_dir \
2089 $tmp/mnt/lustre/striped_dir/ || {
2090 error_noexit "mv remote dir failed"
2095 # If it is upgraded from DNE (2.7), then move the striped dir
2096 # which was created in 2.7 to the new striped dir.
2097 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]] &&
2098 [[ -d $tmp/mnt/lustre/striped_dir_old ]]; then
2099 stripe_count=$($LFS getdirstripe -c \
2100 $tmp/mnt/lustre/striped_dir_old)
2101 [[ $stripe_count -eq 2 ]] || {
2102 error_noexit "get count $stripe_count" \
2103 "from striped dir failed"
2106 mv $tmp/mnt/lustre/striped_dir_old \
2107 $tmp/mnt/lustre/striped_dir/ || {
2108 error_noexit "mv striped dir failed"
2114 $r $LCTL set_param -n osd*.*.force_sync=1
2115 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
2116 error_noexit "dd failed"
2119 rm -rf $tmp/mnt/lustre/tmp_file || {
2120 error_noexit "rm failed"
2124 if $r test -f $tmp/sha1sums; then
2125 # LU-2393 - do both sorts on same node to ensure locale
2127 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
2128 if [ "$dne_upgrade" != "no" ]; then
2129 pushd $tmp/mnt/lustre/striped_dir
2131 pushd $tmp/mnt/lustre
2134 find ! -path "*remote_dir*" ! -path "*striped_dir*" \
2135 ! -name .lustre -type f -exec sha1sum {} \; |
2136 sort -k 2 >$tmp/sha1sums || {
2138 error_noexit "sha1sum"
2142 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
2143 error_noexit "sha1sum verification failed"
2147 # if upgrade from DNE(2.5), then check remote directory
2148 # if upgrade from DNE(2.7), then check striped directory
2149 if $mdt2_is_available &&
2150 [[ "$dne_upgrade" != "no" ]]; then
2151 local new_dir="$tmp/mnt/lustre/striped_dir"
2152 local striped_dir_old="$new_dir/striped_dir_old"
2154 local dir_list="$new_dir/remote_dir"
2155 [[ ! -d $triped_dir_old ]] ||
2156 dir_list+=" $striped_dir_old"
2158 for dir in $dir_list; do
2160 find ! -name .lustre -type f \
2161 -exec sha1sum {} \; |
2162 sort -k 2 >$tmp/sha1sums || {
2164 error_noexit "sha1sum"
2168 if ! diff -ub $tmp/sha1sums.orig \
2170 error_noexit "sha1sum $dir" \
2177 echo "sha1sum verification skipped"
2180 if [ "$dne_upgrade" != "no" ]; then
2181 rm -rf $tmp/mnt/lustre/striped_dir || {
2182 error_noexit "remove remote dir failed"
2187 # migrate files/dirs to remote MDT, then move them back
2188 if [ $(lustre_version_code mds1) -ge $(version_code 2.7.50) -a \
2189 $dne_upgrade != "no" ]; then
2190 $r $LCTL set_param -n \
2191 mdt.${fsname}*.enable_remote_dir=1 2>/dev/null
2193 echo "test migration"
2194 pushd $tmp/mnt/lustre
2195 for dir in $(find ! -name .lustre ! -name . -type d); do
2196 mdt_index=$($LFS getdirstripe -i $dir)
2197 stripe_cnt=$($LFS getdirstripe -c $dir)
2198 if [ $mdt_index = 0 -a $stripe_cnt -le 1 ]; then
2199 $LFS mv -M 1 $dir || {
2201 error_noexit "migrate MDT1 failed"
2207 for dir in $(find ! -name . ! -name .lustre -type d); do
2208 mdt_index=$($LFS getdirstripe -i $dir)
2209 stripe_cnt=$($LFS getdirstripe -c $dir)
2210 if [ $mdt_index = 1 -a $stripe_cnt -le 1 ]; then
2211 $LFS mv -M 0 $dir || {
2213 error_noexit "migrate MDT0 failed"
2222 # When adding new data verification tests, please check for
2223 # the presence of the required reference files first, like
2224 # the "sha1sums" and "list" tests above, to avoid the need to
2225 # regenerate every image for each test addition.
2228 nrpcs_orig=$($LCTL get_param \
2229 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
2230 error_noexit "Getting \"max_rpcs_in_flight\""
2233 nrpcs=$((nrpcs_orig + 5))
2234 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=$nrpcs || {
2235 error_noexit "Changing \"max_rpcs_in_flight\""
2238 wait_update $HOSTNAME "$LCTL get_param \
2239 -n mdc.*MDT0000*.max_rpcs_in_flight" $nrpcs || {
2240 error_noexit "Verifying \"max_rpcs_in_flight\""
2244 umount $tmp/mnt/lustre || {
2245 error_noexit "Unmounting the client"
2248 shall_cleanup_lustre=false
2250 if [[ "$dne_upgrade" != "no" ]] || $mdt2_is_available; then
2251 $r $UMOUNT $tmp/mnt/mdt1 || {
2252 error_noexit "Unmounting the MDT2"
2255 shall_cleanup_mdt1=false
2258 $r $UMOUNT $tmp/mnt/mdt || {
2259 error_noexit "Unmounting the MDT"
2262 shall_cleanup_mdt=false
2264 $r $UMOUNT $tmp/mnt/ost || {
2265 error_noexit "Unmounting the OST"
2268 shall_cleanup_ost=false
2270 t32_reload_modules $node || {
2271 error_noexit "Reloading modules"
2275 if [[ $fstype == zfs ]]; then
2276 local poolname=t32fs-mdt1
2277 $r "$ZPOOL list -H $poolname >/dev/null 2>&1 ||
2278 $ZPOOL import -f -d $tmp $poolname"
2281 # mount a second time to make sure we didnt leave upgrade flag on
2282 $r $TUNEFS --dryrun $mdt_dev || {
2284 error_noexit "tunefs.lustre before remounting the MDT"
2288 mopts=exclude=$fsname-OST0000
2289 if [ $fstype == "ldiskfs" ]; then
2292 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2293 error_noexit "Remounting the MDT"
2296 shall_cleanup_mdt=true
2306 for tarball in $tarballs; do
2307 t32_test $tarball || let "rc += $?"
2311 run_test 32a "Upgrade (not live)"
2319 for tarball in $tarballs; do
2320 t32_test $tarball writeconf || let "rc += $?"
2324 run_test 32b "Upgrade with writeconf"
2331 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
2333 for tarball in $tarballs; do
2334 # Do not support 1_8 and 2_1 direct upgrade to DNE2 anymore */
2335 echo $tarball | grep "1_8" && continue
2336 echo $tarball | grep "2_1" && continue
2338 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
2342 run_test 32c "dne upgrade test"
2350 for tarball in $tarballs; do
2351 ff_convert=yes t32_test $tarball || rc=$?
2355 run_test 32d "convert ff test"
2357 test_33a() { # bug 12333, was test_33
2358 local FSNAME2=test-123
2359 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2362 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
2364 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
2365 local dev=${SINGLEMDS}_dev
2366 local MDSDEV=${!dev}
2367 is_blkdev $SINGLEMDS $MDSDEV &&
2368 skip_env "mixed loopback and real device not working" &&
2372 local fs2mdsdev=$(mdsdevname 1_2)
2373 local fs2ostdev=$(ostdevname 1_2)
2374 local fs2mdsvdev=$(mdsvdevname 1_2)
2375 local fs2ostvdev=$(ostvdevname 1_2)
2377 if [ $(facet_fstype mds1) == ldiskfs ]; then
2378 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
2381 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2382 --reformat $mkfsoptions $fs2mdsdev $fs2mdsvdev || exit 10
2383 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2384 --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \
2385 $fs2ostvdev || exit 10
2387 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
2388 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2389 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" ||
2390 error "$LCTL conf_param $FSNAME2.sys.timeout=200 failed"
2391 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2392 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2395 cp /etc/hosts $MOUNT2/ || error "copy /etc/hosts $MOUNT2/ failed"
2396 $GETSTRIPE $MOUNT2/hosts || error "$GETSTRIPE $MOUNT2/hosts failed"
2401 cleanup_nocli || error "cleanup_nocli failed with $?"
2403 run_test 33a "Mount ost with a large index number"
2405 test_33b() { # was test_34
2408 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
2409 # Drop lock cancelation reply during umount
2410 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
2411 do_facet client $LCTL set_param fail_loc=0x80000304
2412 #lctl set_param debug=-1
2413 umount_client $MOUNT
2414 cleanup || error "cleanup failed with $?"
2416 run_test 33b "Drop cancel during umount"
2420 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
2421 manual_umount_client
2423 do_facet client killall -USR1 multiop
2424 if [ $rc -eq 0 ]; then
2425 error "umount not fail!"
2428 cleanup || error "cleanup failed with rc $?"
2430 run_test 34a "umount with opened file should be fail"
2434 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2435 stop_mds || error "Unable to stop MDS"
2437 manual_umount_client --force || error "mtab after failed umount with $?"
2439 cleanup || error "cleanup failed with $?"
2441 run_test 34b "force umount with failed mds should be normal"
2445 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2446 stop_ost || error "Unable to stop OST1"
2448 manual_umount_client --force || error "mtab after failed umount with $?"
2450 cleanup || error "cleanup failed with $?"
2452 run_test 34c "force umount with failed ost should be normal"
2454 test_35a() { # bug 12459
2457 DBG_SAVE="`$LCTL get_param -n debug`"
2458 $LCTL set_param debug="ha"
2460 log "Set up a fake failnode for the MDS"
2462 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2463 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2464 do_facet mgs "$LCTL conf_param \
2465 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" ||
2466 error "Setting ${device}.failover.node=\
2467 $(h2$NETTYPE $FAKENID) failed."
2469 log "Wait for RECONNECT_INTERVAL seconds (10s)"
2472 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
2475 log "Stopping the MDT: $device"
2476 stop_mdt 1 || error "MDT0 stop fail"
2478 df $MOUNT > /dev/null 2>&1 &
2480 log "Restarting the MDT: $device"
2481 start_mdt 1 || error "MDT0 start fail"
2482 log "Wait for df ($DFPID) ... "
2485 $LCTL set_param debug="$DBG_SAVE"
2487 # retrieve from the log the first server that the client tried to
2488 # contact after the connection loss
2489 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2490 NEXTCONN=`awk "/${MSG}/ {start = 1;}
2491 /import_select_connection.*$device-mdc.* using connection/ {
2493 if (\\\$NF ~ /$FAKENID/)
2499 }" $TMP/lustre-log-$TESTNAME.log`
2500 [ "$NEXTCONN" != "0" ] &&
2501 error "Tried to connect to ${NEXTCONN} not last active server"
2502 cleanup || error "cleanup failed with $?"
2503 # remove nid settings
2504 writeconf_or_reformat
2506 run_test 35a "Reconnect to the last active server first"
2508 test_35b() { # bug 18674
2509 remote_mds || { skip "local MDS" && return 0; }
2513 $LCTL set_param debug="ha"
2515 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
2518 log "Set up a fake failnode for the MDS"
2520 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2521 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2522 do_facet mgs "$LCTL conf_param \
2523 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" ||
2524 error "Set ${device}.failover.node=\
2525 $(h2$NETTYPE $FAKENID) failed"
2527 local at_max_saved=0
2528 # adaptive timeouts may prevent seeing the issue
2529 if at_is_enabled; then
2530 at_max_saved=$(at_max_get mds)
2531 at_max_set 0 mds client
2534 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
2536 log "Injecting EBUSY on MDS"
2537 # Setting OBD_FAIL_MDS_RESEND=0x136
2538 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" ||
2539 error "unable to set param fail_loc=0x80000136"
2541 $LCTL set_param mdc.${FSNAME}*.stats=clear
2543 log "Creating a test file and stat it"
2544 touch $MOUNT/$tdir/$tfile || error "touch $MOUNT/$tdir/$tfile failed"
2545 stat $MOUNT/$tdir/$tfile
2547 log "Stop injecting EBUSY on MDS"
2548 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" ||
2549 error "unable to set param fail_loc=0"
2550 rm -f $MOUNT/$tdir/$tfile || error "remove $MOUNT/$tdir/$tfile failed"
2553 # restore adaptive timeout
2554 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
2556 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2558 CONNCNT=$($LCTL get_param mdc.${FSNAME}*.stats |
2559 awk '/mds_connect/{print $2}')
2561 # retrieve from the log if the client has ever tried to
2562 # contact the fake server after the loss of connection
2563 FAILCONN=`awk "BEGIN {ret = 0;}
2564 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
2566 if (\\\$NF ~ /$FAKENID/) {
2571 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
2573 [ "$FAILCONN" == "0" ] &&
2574 error "The client reconnection has not been triggered"
2575 [ "$FAILCONN" == "2" ] &&
2576 error "Primary server busy, client reconnect to failover failed"
2579 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
2580 # Reconnects are supposed to be rate limited to one every 5s
2581 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] &&
2582 error "Too many reconnects $CONNCNT"
2584 cleanup || error "cleanup failed with $?"
2585 # remove nid settings
2586 writeconf_or_reformat
2588 run_test 35b "Continue reconnection retries, if the active server is busy"
2591 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
2593 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] ||
2594 { skip "remote OST" && return 0; }
2597 local FSNAME2=test1234
2598 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2600 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
2602 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
2603 is_blkdev $SINGLEMDS $MDSDEV &&
2604 skip_env "mixed loopback and real device not working" && return
2607 local fs2mdsdev=$(mdsdevname 1_2)
2608 local fs2ostdev=$(ostdevname 1_2)
2609 local fs3ostdev=$(ostdevname 2_2)
2610 local fs2mdsvdev=$(mdsvdevname 1_2)
2611 local fs2ostvdev=$(ostvdevname 1_2)
2612 local fs3ostvdev=$(ostvdevname 2_2)
2614 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2615 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
2616 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
2617 # different one than the default value here.
2618 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2619 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
2620 add fs3ost $(mkfs_opts ost2 ${fs3ostdev}) --mgsnode=$MGSNID \
2621 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
2623 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
2624 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2625 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
2626 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2627 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2629 sleep 5 # until 11778 fixed
2631 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || error "dd failed"
2633 BKTOTAL=$($LCTL get_param -n obdfilter.*.kbytestotal |
2634 awk 'BEGIN{total=0}; {total+=$1}; END{print total}')
2635 BKFREE=$($LCTL get_param -n obdfilter.*.kbytesfree |
2636 awk 'BEGIN{free=0}; {free+=$1}; END{print free}')
2637 BKAVAIL=$($LCTL get_param -n obdfilter.*.kbytesavail |
2638 awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}')
2639 STRING=$(df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}')
2640 DFTOTAL=$(echo $STRING | cut -d, -f1)
2641 DFUSED=$(echo $STRING | cut -d, -f2)
2642 DFAVAIL=$(echo $STRING | cut -d, -f3)
2643 DFFREE=$(($DFTOTAL - $DFUSED))
2645 ALLOWANCE=$((64 * $OSTCOUNT))
2647 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
2648 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
2649 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
2652 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
2653 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
2654 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
2657 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
2658 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
2659 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
2664 stop fs3ost -f || error "unable to stop OST3"
2665 stop fs2ost -f || error "unable to stop OST2"
2666 stop fs2mds -f || error "unable to stop second MDS"
2667 unload_modules_conf || error "unable unload modules"
2670 run_test 36 "df report consistency on OSTs with different block size"
2673 local mntpt=$(facet_mntpt $SINGLEMDS)
2674 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
2675 local mdsdev_sym="$TMP/sym_mdt.img"
2676 local opts=$MDS_MOUNT_OPTS
2679 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2680 skip "Currently only applicable to ldiskfs-based MDTs"
2684 echo "MDS : $mdsdev"
2685 echo "SYMLINK : $mdsdev_sym"
2686 do_facet $SINGLEMDS rm -f $mdsdev_sym
2688 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
2690 echo "mount symlink device - $mdsdev_sym"
2692 if ! do_facet $SINGLEMDS test -b $mdsdev; then
2693 opts=$(csa_add "$opts" -o loop)
2695 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
2696 $mdsdev_sym $mntpt 2>&1)
2699 echo mount_op=$mount_op
2701 do_facet $SINGLEMDS "$UMOUNT $mntpt && rm -f $mdsdev_sym"
2703 if $(echo $mount_op | grep -q "unable to set tunable"); then
2704 error "set tunables failed for symlink device"
2707 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
2709 run_test 37 "verify set tunables works for symlink device"
2711 test_38() { # bug 14222
2712 local fstype=$(facet_fstype $SINGLEMDS)
2713 local mntpt=$(facet_mntpt $SINGLEMDS)
2718 local SRC="/etc /bin"
2719 local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT)
2720 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
2721 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2722 tar cf - $FILES | tar xf - -C $DIR/$tdir ||
2723 error "copying $SRC to $DIR/$tdir"
2725 umount_client $MOUNT || error "umount_client $MOUNT failed"
2726 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2727 stop_mds || error "Unable to stop MDS"
2728 log "delete lov_objid file on MDS"
2730 mount_fstype $SINGLEMDS || error "mount MDS failed (1)"
2732 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid; rm $mntpt/lov_objid"
2734 unmount_fstype $SINGLEMDS || error "umount failed (1)"
2736 # check create in mds_lov_connect
2737 start_mds || error "unable to start MDS"
2738 mount_client $MOUNT || error "mount_client $MOUNT failed"
2740 [ $V ] && log "verifying $DIR/$tdir/$f"
2741 diff -q $f $DIR/$tdir/$f || ERROR=y
2743 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2744 if [ "$ERROR" = "y" ]; then
2745 # check it's updates in sync
2746 umount_client $MOUNT
2748 mount_fstype $SIGNLEMDS
2749 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2750 unmount_fstype $SINGLEMDS
2751 error "old and new files are different after connect" || true
2753 touch $DIR/$tdir/f2 || error "f2 file create failed"
2755 # check it's updates in sync
2756 umount_client $MOUNT || error "second umount_client $MOUNT failed"
2759 mount_fstype $SINGLEMDS || error "mount MDS failed (3)"
2761 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2762 do_facet $SINGLEMDS dd if=/dev/zero of=$mntpt/lov_objid.clear count=8
2764 unmount_fstype $SINGLEMDS || error "umount failed (3)"
2766 start_mds || error "unable to start MDS"
2767 mount_client $MOUNT || error "mount_client $MOUNT failed"
2769 [ $V ] && log "verifying $DIR/$tdir/$f"
2770 diff -q $f $DIR/$tdir/$f || ERROR=y
2772 touch $DIR/$tdir/f3 || error "f3 file create failed"
2773 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2774 umount_client $MOUNT || error "third umount_client $MOUNT failed"
2776 mount_fstype $SINGLEMDS || error "mount MDS failed (4)"
2777 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2778 unmount_fstype $SINGLEMDS || error "umount failed (4)"
2780 [ "$ERROR" = "y" ] &&
2781 error "old and new files are different after sync" || true
2783 log "files compared the same"
2784 cleanup || error "cleanup failed with $?"
2786 run_test 38 "MDS recreates missing lov_objid file from OST data"
2791 cleanup || error "cleanup failed with $?"
2792 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
2793 error "memory leak detected" || true
2795 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
2797 test_40() { # bug 15759
2798 start_ost || error "Unable to start OST1"
2799 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
2800 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
2802 cleanup || error "cleanup failed with rc $?"
2804 run_test 40 "race during service thread startup"
2806 test_41a() { #bug 14134
2807 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2808 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2809 skip "Loop devices does not work with nosvc option"
2813 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2815 start_mdt 1 -o nosvc -n
2816 if [ $MDSCOUNT -ge 2 ]; then
2817 for num in $(seq 2 $MDSCOUNT); do
2818 start_mdt $num || return
2821 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
2822 start_mdt 1 -o nomgs,force
2823 mount_client $MOUNT || error "mount_client $MOUNT failed"
2826 echo "blah blah" > $MOUNT/$tfile
2829 umount_client $MOUNT || error "umount_client $MOUNT failed"
2830 stop ost1 -f || error "unable to stop OST1"
2831 stop_mds || error "Unable to stop MDS"
2832 stop_mds || error "Unable to stop MDS on second try"
2834 run_test 41a "mount mds with --nosvc and --nomgs"
2837 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2838 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2839 skip "Loop devices does not work with nosvc option"
2843 ! combined_mgs_mds && skip "needs combined mgs device" && return 0
2847 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2849 start_mdt 1 -o nosvc -n
2850 if [ $MDSCOUNT -ge 2 ]; then
2851 for num in $(seq 2 $MDSCOUNT); do
2852 start_mdt $num || return
2855 start_ost || error "Unable to start OST1"
2856 start_mdt 1 -o nomgs,force
2857 mount_client $MOUNT || error "mount_client $MOUNT failed"
2860 echo "blah blah" > $MOUNT/$tfile
2861 cat $MOUNT/$tfile || error "cat $MOUNT/$tfile failed"
2863 umount_client $MOUNT || error "umount_client $MOUNT failed"
2864 stop_ost || error "Unable to stop OST1"
2865 stop_mds || error "Unable to stop MDS"
2866 stop_mds || error "Unable to stop MDS on second try"
2868 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
2871 local server_version=$(lustre_version_code $SINGLEMDS)
2873 [[ $server_version -ge $(version_code 2.6.52) ]] ||
2874 [[ $server_version -ge $(version_code 2.5.26) &&
2875 $server_version -lt $(version_code 2.5.50) ]] ||
2876 [[ $server_version -ge $(version_code 2.5.4) &&
2877 $server_version -lt $(version_code 2.5.11) ]] ||
2878 { skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"; return; }
2881 # MDT concurrent start
2882 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
2883 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x716"
2884 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS &
2886 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS &
2887 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x0"
2893 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
2894 echo "1st MDT start succeed"
2895 echo "2nd MDT start failed with $rc2"
2896 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
2897 echo "1st MDT start failed with $rc"
2898 echo "2nd MDT start succeed"
2901 error "unexpected concurrent MDT mounts result, rc=$rc rc2=$rc2"
2904 if [ $MDSCOUNT -ge 2 ]; then
2905 for num in $(seq 2 $MDSCOUNT); do
2906 start_mdt $num || return
2910 # OST concurrent start
2912 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
2913 do_facet ost1 "$LCTL set_param fail_loc=0x716"
2914 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &
2916 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &
2917 do_facet ost1 "$LCTL set_param fail_loc=0x0"
2923 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
2924 echo "1st OST start succeed"
2925 echo "2nd OST start failed with $rc2"
2926 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
2927 echo "1st OST start failed with $rc"
2928 echo "2nd OST start succeed"
2932 error "unexpected concurrent OST mounts result, rc=$rc rc2=$rc2"
2938 # verify everything ok
2943 error "MDT(s) start failed"
2951 error "OST(s) start failed"
2959 error "client start failed"
2966 error "client mount failed"
2970 run_test 41c "concurrent mounts of MDT/OST should all fail but one"
2972 test_42() { #bug 14693
2974 check_mount || error "client was not mounted"
2976 do_facet mgs $LCTL conf_param $FSNAME.llite.some_wrong_param=10
2977 umount_client $MOUNT ||
2978 error "unmounting client failed with invalid llite param"
2979 mount_client $MOUNT ||
2980 error "mounting client failed with invalid llite param"
2982 do_facet mgs $LCTL conf_param $FSNAME.sys.some_wrong_param=20
2983 cleanup || error "stopping $FSNAME failed with invalid sys param"
2985 check_mount || error "client was not mounted with invalid sys param"
2986 cleanup || error "stopping $FSNAME failed with invalid sys param"
2988 run_test 42 "allow client/server mount/unmount with invalid config param"
2991 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.58) ]] ||
2992 { skip "Need MDS version at least 2.5.58" && return 0; }
2993 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
2996 USER1=$(getent passwd | grep :$ID1:$ID1: | cut -d: -f1)
2997 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1" &&
3001 chmod ugo+x $DIR || error "chmod 0 failed"
3002 set_conf_param_and_check mds \
3003 "$LCTL get_param -n mdt.$FSNAME-MDT0000.root_squash" \
3004 "$FSNAME.mdt.root_squash" \
3006 wait_update $HOSTNAME \
3007 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3009 error "check llite root_squash failed!"
3010 set_conf_param_and_check mds \
3011 "$LCTL get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
3012 "$FSNAME.mdt.nosquash_nids" \
3014 wait_update $HOSTNAME \
3015 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3017 error "check llite nosquash_nids failed!"
3020 # create set of test files
3022 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
3023 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
3024 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
3026 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
3027 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
3029 mkdir $DIR/$tdir-rootdir || error "mkdir failed"
3030 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
3031 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
3033 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
3034 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
3035 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
3038 # check root_squash:
3039 # set root squash UID:GID to RUNAS_ID
3040 # root should be able to access only files owned by RUNAS_ID
3042 set_conf_param_and_check mds \
3043 "$LCTL get_param -n mdt.$FSNAME-MDT0000.root_squash" \
3044 "$FSNAME.mdt.root_squash" \
3045 "$RUNAS_ID:$RUNAS_ID"
3046 wait_update $HOSTNAME \
3047 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3048 "$RUNAS_ID:$RUNAS_ID" ||
3049 error "check llite root_squash failed!"
3051 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
3052 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3053 error "$ST: root read permission is denied"
3054 echo "$ST: root read permission is granted - ok"
3057 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3058 error "$ST: root write permission is denied"
3059 echo "$ST: root write permission is granted - ok"
3061 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3062 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3063 error "$ST: root read permission is granted"
3064 echo "$ST: root read permission is denied - ok"
3067 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3068 error "$ST: root write permission is granted"
3069 echo "$ST: root write permission is denied - ok"
3071 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3072 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null &&
3073 error "$ST: root unlink permission is granted"
3074 echo "$ST: root unlink permission is denied - ok"
3076 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null &&
3077 error "$ST: root create permission is granted"
3078 echo "$ST: root create permission is denied - ok"
3081 # check root_squash is enforced independently
3082 # of client cache content
3084 # access file by USER1, keep access open
3085 # root should be denied access to user file
3087 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
3091 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
3092 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3093 { kill $pid; error "$ST: root read permission is granted"; }
3094 echo "$ST: root read permission is denied - ok"
3097 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3098 { kill $pid; error "$ST: root write permission is granted"; }
3099 echo "$ST: root write permission is denied - ok"
3105 # check nosquash_nids:
3106 # put client's NID into nosquash_nids list,
3107 # root should be able to access root file after that
3109 local NIDLIST=$($LCTL list_nids all | tr '\n' ' ')
3110 NIDLIST="2@gni $NIDLIST 192.168.0.[2,10]@tcp"
3111 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
3112 set_conf_param_and_check mds \
3113 "$LCTL get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
3114 "$FSNAME-MDTall.mdt.nosquash_nids" \
3116 wait_update $HOSTNAME \
3117 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3119 error "check llite nosquash_nids failed!"
3121 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3122 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3123 error "$ST: root read permission is denied"
3124 echo "$ST: root read permission is granted - ok"
3127 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3128 error "$ST: root write permission is denied"
3129 echo "$ST: root write permission is granted - ok"
3131 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3132 rm $DIR/$tdir-rootdir/tfile-1 ||
3133 error "$ST: root unlink permission is denied"
3134 echo "$ST: root unlink permission is granted - ok"
3135 touch $DIR/$tdir-rootdir/tfile-2 ||
3136 error "$ST: root create permission is denied"
3137 echo "$ST: root create permission is granted - ok"
3138 cleanup || error "cleanup failed with $?"
3140 run_test 43a "check root_squash and nosquash_nids"
3142 test_43b() { # LU-5690
3143 [[ $(lustre_version_code mgs) -ge $(version_code 2.7.62) ]] ||
3144 { skip "Need MGS version 2.7.62+"; return; }
3146 if [[ -z "$fs2mds_DEV" ]]; then
3147 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
3148 skip_env "mixed loopback and real device not working" && return
3151 local fs2mdsdev=$(mdsdevname 1_2)
3152 local fs2mdsvdev=$(mdsvdevname 1_2)
3154 # temporarily use fs2mds as fs2mgs
3156 local fs2mgsdev=$fs2mdsdev
3157 local fs2mgsvdev=$fs2mdsvdev
3159 local fsname=test1234
3161 load_module llite/lustre
3162 local client_ip=$(host_nids_address $HOSTNAME $NETTYPE)
3163 local host=${client_ip//*./}
3164 local net=${client_ip/%$host/}
3165 local nosquash_nids=$(h2$NETTYPE $net[$host,$host,$host])
3167 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
3168 --param mdt.root_squash=$RUNAS_ID:$RUNAS_ID \
3169 --param mdt.nosquash_nids=$nosquash_nids \
3170 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
3171 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS || error "start fs2mgs failed"
3172 stop $fs2mgs -f || error "stop fs2mgs failed"
3174 run_test 43b "parse nosquash_nids with commas in expr_list"
3176 umount_client $MOUNT
3181 check_mount || error "check_mount"
3182 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
3184 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
3185 for VAL in $UUIDS; do
3186 NID=$(echo $VAL | cut -d= -f1)
3187 CLUUID=$(echo $VAL | cut -d= -f2)
3188 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
3190 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
3191 cleanup || error "cleanup failed with $?"
3193 run_test 44 "mounted client proc entry exists"
3197 check_mount || error "check_mount"
3198 stop_mds || error "Unable to stop MDS"
3202 #define OBD_FAIL_PTLRPC_LONG_UNLINK 0x50f
3203 do_facet client "$LCTL set_param fail_loc=0x50f"
3206 manual_umount_client --force || error "manual_umount_client failed"
3207 do_facet client "$LCTL set_param fail_loc=0x0"
3208 start_mds || error "unable to start MDS"
3209 mount_client $MOUNT || error "mount_client $MOUNT failed"
3210 cleanup || error "cleanup failed with $?"
3212 run_test 45 "long unlink handling in ptlrpcd"
3219 umount_client $MOUNT2 || rc=$?
3220 umount_client $MOUNT || rc=$?
3221 while [ $count -gt 0 ]; do
3222 stop ost${count} -f || rc=$?
3226 cleanup_nocli || rc=$?
3227 #writeconf to remove all ost2 traces for subsequent tests
3228 writeconf_or_reformat
3233 echo "Testing with $OSTCOUNT OSTs"
3235 start_mds || error "unable to start MDS"
3236 #first client should see only one ost
3237 start_ost || error "Unable to start OST1"
3238 wait_osc_import_state mds ost FULL
3240 mount_client $MOUNT || error "mount_client $MOUNT failed"
3241 trap "cleanup_46a $OSTCOUNT" EXIT ERR
3244 for (( i=2; i<=$OSTCOUNT; i++ )); do
3245 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
3246 error "start_ost$i $(ostdevname $i) failed"
3249 # wait until osts in sync
3250 for (( i=2; i<=$OSTCOUNT; i++ )); do
3251 wait_osc_import_state mds ost$i FULL
3252 wait_osc_import_state client ost$i FULL
3255 #second client see all ost's
3257 mount_client $MOUNT2 || error "mount_client failed"
3258 $SETSTRIPE -c -1 $MOUNT2 || error "$SETSTRIPE -c -1 $MOUNT2 failed"
3259 $GETSTRIPE $MOUNT2 || error "$GETSTRIPE $MOUNT2 failed"
3261 echo "ok" > $MOUNT2/widestripe
3262 $GETSTRIPE $MOUNT2/widestripe ||
3263 error "$GETSTRIPE $MOUNT2/widestripe failed"
3264 # fill acl buffer for avoid expand lsm to them
3265 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd |
3267 setfacl -m $acl $MOUNT2/widestripe
3271 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3273 cleanup_46a $OSTCOUNT || error "cleanup_46a failed"
3275 run_test 46a "handle ost additional - wide striped file"
3280 check_mount || error "check_mount failed"
3281 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
3285 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3286 if echo $ns | grep "MDT[[:digit:]]*"; then
3289 lrs=$(echo $ns | sed 's/.*lru_size=//')
3290 lru_size[count]=$lrs
3295 facet_failover $SINGLEMDS
3296 client_up || error "client_up failed"
3299 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3300 if echo $ns | grep "MDT[[:digit:]]*"; then
3303 lrs=$(echo $ns | sed 's/.*lru_size=//')
3304 if ! test "$lrs" -eq "${lru_size[count]}"; then
3305 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
3306 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
3311 cleanup || error "cleanup failed with $?"
3313 run_test 47 "server restart does not make client loss lru_resize settings"
3318 # reformat after this test is needed - if the test fails,
3319 # we will have unkillable file at FS
3323 test_48() { # bug 17636
3326 check_mount || error "check_mount failed"
3328 $SETSTRIPE -c -1 $MOUNT || error "$SETSTRIPE -c -1 $MOUNT failed"
3329 $GETSTRIPE $MOUNT || error "$GETSTRIPE $MOUNT failed"
3331 echo "ok" > $MOUNT/widestripe
3332 $GETSTRIPE $MOUNT/widestripe ||
3333 error "$GETSTRIPE $MOUNT/widestripe failed"
3335 trap cleanup_48 EXIT ERR
3337 # fill acl buffer for avoid expand lsm to them
3338 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
3339 setfacl -m $acl $MOUNT/widestripe
3342 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3346 run_test 48 "too many acls on file"
3348 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
3349 test_49a() { # bug 17710
3350 local timeout_orig=$TIMEOUT
3351 local ldlm_timeout_orig=$LDLM_TIMEOUT
3352 local LOCAL_TIMEOUT=20
3354 LDLM_TIMEOUT=$LOCAL_TIMEOUT
3355 TIMEOUT=$LOCAL_TIMEOUT
3359 check_mount || error "client mount failed"
3361 echo "check ldlm_timout..."
3362 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3363 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3364 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3366 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3367 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3370 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
3371 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
3374 umount_client $MOUNT || error "umount_client $MOUNT failed"
3375 stop_ost || error "problem stopping OSS"
3376 stop_mds || error "problem stopping MDS"
3378 LDLM_TIMEOUT=$ldlm_timeout_orig
3379 TIMEOUT=$timeout_orig
3381 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3383 test_49b() { # bug 17710
3384 local timeout_orig=$TIMEOUT
3385 local ldlm_timeout_orig=$LDLM_TIMEOUT
3386 local LOCAL_TIMEOUT=20
3388 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
3389 TIMEOUT=$LOCAL_TIMEOUT
3393 check_mount || error "client mount failed"
3395 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3396 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3397 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3399 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3400 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3403 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
3404 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
3407 cleanup || error "cleanup failed"
3409 LDLM_TIMEOUT=$ldlm_timeout_orig
3410 TIMEOUT=$timeout_orig
3412 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3415 # Test both statfs and lfs df and fail if either one fails
3416 multiop_bg_pause $1 f_
3419 killall -USR1 multiop
3420 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
3421 wait $PID || { RC1=$?; log "multiop return error "; }
3428 if [ $RC2 -eq 0 ]; then
3430 log "lazystatfs df failed"
3434 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
3440 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3441 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3443 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
3445 cleanup || error "cleanup failed with rc $?"
3447 run_test 50a "lazystatfs all servers available"
3451 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3452 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3454 # Wait for client to detect down OST
3455 stop_ost || error "Unable to stop OST1"
3456 wait_osc_import_state mds ost DISCONN
3458 lazystatfs $MOUNT || error "lazystatfs should not return EIO"
3460 umount_client $MOUNT || error "Unable to unmount client"
3461 stop_mds || error "Unable to stop MDS"
3463 run_test 50b "lazystatfs all servers down"
3466 start_mds || error "Unable to start MDS"
3467 start_ost || error "Unable to start OST1"
3468 start_ost2 || error "Unable to start OST2"
3469 mount_client $MOUNT || error "Unable to mount client"
3470 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3471 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3473 # Wait for client to detect down OST
3474 stop_ost || error "Unable to stop OST1"
3475 wait_osc_import_state mds ost DISCONN
3476 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3478 umount_client $MOUNT || error "Unable to unmount client"
3479 stop_ost2 || error "Unable to stop OST2"
3480 stop_mds || error "Unable to stop MDS"
3481 #writeconf to remove all ost2 traces for subsequent tests
3482 writeconf_or_reformat
3484 run_test 50c "lazystatfs one server down"
3487 start_mds || error "Unable to start MDS"
3488 start_ost || error "Unable to start OST1"
3489 start_ost2 || error "Unable to start OST2"
3490 mount_client $MOUNT || error "Unable to mount client"
3491 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3492 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3494 # Issue the statfs during the window where the client still
3495 # belives the OST to be available but it is in fact down.
3496 # No failure just a statfs which hangs for a timeout interval.
3497 stop_ost || error "Unable to stop OST1"
3498 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3500 umount_client $MOUNT || error "Unable to unmount client"
3501 stop_ost2 || error "Unable to stop OST2"
3502 stop_mds || error "Unable to stop MDS"
3503 #writeconf to remove all ost2 traces for subsequent tests
3504 writeconf_or_reformat
3506 run_test 50d "lazystatfs client/server conn race"
3513 start_mds || error "Unable to start MDS"
3514 #first client should see only one ost
3515 start_ost || error "Unable to start OST1"
3516 wait_osc_import_state mds ost FULL
3518 # Wait for client to detect down OST
3519 stop_ost || error "Unable to stop OST1"
3520 wait_osc_import_state mds ost DISCONN
3522 mount_client $MOUNT || error "Unable to mount client"
3523 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3525 multiop_bg_pause $MOUNT _f
3529 if [ $RC1 -ne 0 ]; then
3530 log "multiop failed $RC1"
3533 sleep $(( $TIMEOUT+1 ))
3535 [ $? -ne 0 ] && error "process isn't sleep"
3536 start_ost || error "Unable to start OST1"
3537 wait $pid || error "statfs failed"
3540 umount_client $MOUNT || error "Unable to unmount client"
3541 stop_ost || error "Unable to stop OST1"
3542 stop_mds || error "Unable to stop MDS"
3544 run_test 50e "normal statfs all servers down"
3549 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
3551 start_mds || error "Unable to start mds"
3552 #first client should see only one ost
3553 start_ost || error "Unable to start OST1"
3554 wait_osc_import_state mds ost FULL
3556 start_ost2 || error "Unable to start OST2"
3557 wait_osc_import_state mds ost2 FULL
3559 # Wait for client to detect down OST
3560 stop_ost2 || error "Unable to stop OST2"
3562 wait_osc_import_state mds ost2 DISCONN
3563 mount_client $MOUNT || error "Unable to mount client"
3564 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3566 multiop_bg_pause $MOUNT _f
3570 if [ $RC1 -ne 0 ]; then
3571 log "lazystatfs multiop failed $RC1"
3574 sleep $(( $TIMEOUT+1 ))
3576 [ $? -ne 0 ] && error "process isn't sleep"
3577 start_ost2 || error "Unable to start OST2"
3578 wait $pid || error "statfs failed"
3579 stop_ost2 || error "Unable to stop OST2"
3582 umount_client $MOUNT -f || error "Unable to unmount client"
3583 stop_ost || error "Unable to stop OST1"
3584 stop_mds || error "Unable to stop MDS"
3585 #writeconf to remove all ost2 traces for subsequent tests
3586 writeconf_or_reformat
3588 run_test 50f "normal statfs one server in down"
3591 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3593 start_ost2 || error "Unable to start OST2"
3594 wait_osc_import_state mds ost2 FULL
3595 wait_osc_import_state client ost2 FULL
3597 local PARAM="${FSNAME}-OST0001.osc.active"
3599 $SETSTRIPE -c -1 $DIR/$tfile || error "$SETSTRIPE failed"
3600 do_facet mgs $LCTL conf_param $PARAM=0 ||
3601 error "Unable to deactivate OST"
3603 umount_client $MOUNT || error "Unable to unmount client"
3604 mount_client $MOUNT || error "Unable to mount client"
3605 # This df should not cause a panic
3608 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
3609 rm -f $DIR/$tfile || error "unable to remove file $DIR/$tfile"
3610 umount_client $MOUNT || error "Unable to unmount client"
3611 stop_ost2 || error "Unable to stop OST2"
3612 stop_ost || error "Unable to stop OST1"
3613 stop_mds || error "Unable to stop MDS"
3614 #writeconf to remove all ost2 traces for subsequent tests
3615 writeconf_or_reformat
3617 run_test 50g "deactivated OST should not cause panic"
3621 # prepare MDT/OST, make OSC inactive for OST1
3622 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3624 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3625 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
3626 error "tunefs OST1 failed"
3627 start_mds || error "Unable to start MDT"
3628 start_ost || error "Unable to start OST1"
3629 start_ost2 || error "Unable to start OST2"
3630 mount_client $MOUNT || error "client start failed"
3632 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3634 # activatate OSC for OST1
3635 local TEST="$LCTL get_param -n osc.${FSNAME}-OST0000-osc-[!M]*.active"
3636 set_conf_param_and_check client \
3637 "$TEST" "${FSNAME}-OST0000.osc.active" 1 ||
3638 error "Unable to activate OST1"
3640 mkdir $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3641 $SETSTRIPE -c -1 -i 0 $DIR/$tdir/2 ||
3642 error "$SETSTRIPE $DIR/$tdir/2 failed"
3643 sleep 1 && echo "create a file after OST1 is activated"
3645 createmany -o $DIR/$tdir/2/$tfile-%d 1
3647 # check OSC import is working
3648 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
3649 error "some OSC imports are still not connected"
3652 umount_client $MOUNT || error "Unable to umount client"
3653 stop_ost2 || error "Unable to stop OST2"
3654 cleanup_nocli || error "cleanup_nocli failed with $?"
3656 run_test 50h "LU-642: activate deactivated OST"
3659 # prepare MDT/OST, make OSC inactive for OST1
3660 [ "$MDSCOUNT" -lt "2" ] && skip_env "$MDSCOUNT < 2, skipping" && return
3662 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3663 do_facet mds2 "$TUNEFS --param mdc.active=0 $(mdsdevname 2)" ||
3664 error "tunefs MDT2 failed"
3665 start_mds || error "Unable to start MDT"
3666 start_ost || error "Unable to start OST1"
3667 start_ost2 || error "Unable to start OST2"
3668 mount_client $MOUNT || error "client start failed"
3670 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3672 $LCTL conf_param ${FSNAME}-MDT0000.mdc.active=0 &&
3673 error "deactive MDC0 succeeds"
3674 # activate MDC for MDT2
3675 local TEST="$LCTL get_param -n mdc.${FSNAME}-MDT0001-mdc-[!M]*.active"
3676 set_conf_param_and_check client \
3677 "$TEST" "${FSNAME}-MDT0001.mdc.active" 1 ||
3678 error "Unable to activate MDT2"
3680 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
3681 if [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.60) ]
3683 wait_dne_interconnect
3685 $LFS mkdir -i1 $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3687 createmany -o $DIR/$tdir/2/$tfile-%d 1 || error "create files failed"
3689 rm -rf $DIR/$tdir/2 || error "unlink dir failed"
3691 # deactivate MDC for MDT2
3692 local TEST="$LCTL get_param -n mdc.${FSNAME}-MDT0001-mdc-[!M]*.active"
3693 set_conf_param_and_check client \
3694 "$TEST" "${FSNAME}-MDT0001.mdc.active" 0 ||
3695 error "Unable to deactivate MDT2"
3697 $LFS mkdir -i1 $DIR/$tdir/2 &&
3698 error "mkdir $DIR/$tdir/2 succeeds after deactive MDT"
3701 umount_client $MOUNT || error "Unable to umount client"
3706 run_test 50i "activate deactivated MDT"
3709 local LOCAL_TIMEOUT=20
3713 check_mount || error "check_mount failed"
3715 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
3716 $SETSTRIPE -c -1 $MOUNT/$tdir ||
3717 error "$SETSTRIPE -c -1 $MOUNT/$tdir failed"
3718 #define OBD_FAIL_MDS_REINT_DELAY 0x142
3719 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x142"
3720 touch $MOUNT/$tdir/$tfile &
3723 start_ost2 || error "Unable to start OST1"
3725 stop_ost2 || error "Unable to stop OST1"
3726 umount_client $MOUNT -f || error “unmount $MOUNT failed”
3727 cleanup_nocli || error “stop server failed”
3728 #writeconf to remove all ost2 traces for subsequent tests
3729 writeconf_or_reformat
3731 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
3740 do_node $node mkdir -p $dest
3741 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3743 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
3744 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3745 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
3747 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
3748 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
3758 local backup2=${TMP}/backup2
3760 do_node $node mkdir -p $backup2
3761 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3763 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
3764 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3765 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
3767 do_node $node "diff -rq $backup $backup2"
3768 [ $? -eq 0 ] || { error "contents differ"; return 3; }
3770 local xattrs2=${TMP}/xattrs2
3771 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
3772 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
3774 do_node $node "diff $xattrs $xattrs2"
3775 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
3777 do_node $node "rm -rf $backup2 $xattrs2"
3778 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
3782 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3783 skip "Only applicable to ldiskfs-based MDTs"
3787 start_mds || error "Unable to start MDS"
3788 start_ost || error "Unable to start OST1"
3789 mount_client $MOUNT || error "Unable to mount client"
3792 local ost1mnt=$(facet_mntpt ost1)
3793 local ost1node=$(facet_active_host ost1)
3794 local ost1tmp=$TMP/conf52
3797 mkdir $DIR/$tdir || error "Unable to create $DIR/$tdir"
3798 touch $TMP/modified_first || error "Unable to create temporary file"
3799 local mtime=$(stat -c %Y $TMP/modified_first)
3800 do_node $ost1node "mkdir -p $ost1tmp &&
3801 touch -m -d @$mtime $ost1tmp/modified_first" ||
3802 error "Unable to create temporary file"
3805 $SETSTRIPE -c -1 -S 1M $DIR/$tdir || error "$SETSTRIPE failed"
3807 for (( i=0; i < nrfiles; i++ )); do
3808 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c ||
3809 error "multiop failed"
3815 echo backup files to $TMP/$tdir
3816 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
3817 copy_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
3818 error "Unable to copy files"
3820 umount_client $MOUNT || error "Unable to umount client"
3821 stop_ost || error "Unable to stop ost1"
3823 echo mount ost1 as ldiskfs
3824 do_node $ost1node mkdir -p $ost1mnt || error "Unable to create $ost1mnt"
3825 if ! do_node $ost1node test -b $ost1_dev; then
3828 do_node $ost1node mount -t $(facet_fstype ost1) $loop $ost1_dev \
3830 error "Unable to mount ost1 as ldiskfs"
3833 echo backup objects to $ost1tmp/objects
3834 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
3835 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
3836 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
3838 error "Unable to copy objects"
3840 # move objects to lost+found
3841 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
3842 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
3844 do_node $ost1node "umount $ost1mnt" ||
3845 error "Unable to umount ost1 as ldiskfs"
3847 start_ost || error "Unable to start OST1"
3848 mount_client $MOUNT || error "Unable to mount client"
3850 local REPAIRED=$(do_node $ost1node "$LCTL get_param \
3851 -n osd-ldiskfs.$FSNAME-OST0000.oi_scrub" |
3852 awk '/^lf_repa[ri]*ed/ { print $2 }')
3853 [ $REPAIRED -gt 0 ] ||
3854 error "Some entry under /lost+found should be repaired"
3857 diff_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
3858 error "Unable to diff files"
3860 rm -rf $TMP/$tdir $TMP/file_xattrs ||
3861 error "Unable to delete temporary files"
3862 do_node $ost1node "rm -rf $ost1tmp" ||
3863 error "Unable to delete temporary files"
3864 cleanup || error "cleanup failed with $?"
3866 run_test 52 "check recovering objects from lost+found"
3868 # Checks threads_min/max/started for some service
3870 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
3871 # parameter pattern prefix like 'ost.*.ost'.
3884 local msg="Insane $modname thread counts"
3885 local ncpts=$(check_cpt_number $facet)
3889 check_mount || return 41
3891 # We need to expand $parampat, but it may match multiple parameters, so
3892 # we'll pick the first one
3893 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
3894 error "Couldn't expand ${parampat}.threads_min parameter name"
3898 # Remove the .threads_min part
3899 paramp=${paramp%.threads_min}
3901 # Check for sanity in defaults
3902 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3904 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3906 tstarted=$(do_facet $facet "$LCTL get_param \
3907 -n ${paramp}.threads_started" || echo 0)
3908 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' ||
3910 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
3912 nthrs=$(expr $tmax - $tmin)
3913 if [ $nthrs -lt $ncpts ]; then
3919 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
3920 skip_env "module parameter forced $facet thread count" &&
3921 tmin=3 && tmax=$((3 * tmax))
3923 # Check that we can change min/max
3924 do_facet $facet "$LCTL set_param \
3925 ${paramp}.threads_min=$((tmin + nthrs))"
3926 do_facet $facet "$LCTL set_param \
3927 ${paramp}.threads_max=$((tmax - nthrs))"
3928 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3930 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3932 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) &&
3933 $tmax2 == ($tmax - $nthrs)))' || return $?
3935 # Check that we can set min/max to the same value
3936 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3938 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$tmin"
3939 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3941 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3943 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
3945 # Check that we can't set max < min
3946 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$((tmin - 1))"
3947 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3949 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3951 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
3953 # We need to ensure that we get the module options desired; to do this
3954 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
3955 LOAD_MODULES_REMOTE=true
3958 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
3959 setmodopts -a $modname "$newvalue" oldvalue
3962 check_mount || return 41
3964 # Restore previous setting of MODOPTS_*
3965 setmodopts $modname "$oldvalue"
3967 # Check that $opts took
3968 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min")
3969 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max")
3970 tstarted=$(do_facet $facet \
3971 "$LCTL get_param -n ${paramp}.threads_started")
3972 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
3981 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
3982 cleanup || error "cleanup failed with rc $?"
3984 run_test 53a "check OSS thread count params"
3988 local mds=$(do_facet $SINGLEMDS "$LCTL get_param \
3989 -N mds.*.*.threads_max 2>/dev/null")
3990 if [ -z "$mds" ]; then
3991 #running this on an old MDT
3992 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
3994 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
3996 cleanup || error "cleanup failed with $?"
3998 run_test 53b "check MDS thread count params"
4001 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4002 skip "Only applicable to ldiskfs-based MDTs"
4006 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p ||
4007 error "llverdev failed with rc=$?"
4010 run_test 54a "test llverdev and partial verify of device"
4013 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4014 skip "Only applicable to ldiskfs-based MDTs"
4019 run_llverfs $MOUNT -p || error "llverfs failed with rc=$?"
4020 cleanup || error "cleanup failed with rc=$?"
4022 run_test 54b "test llverfs and partial verify of filesystem"
4026 local max_ost_index=$1
4027 echo -n $(((max_ost_index + 1) * 8))
4031 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4032 skip "Only applicable to ldiskfs-based MDTs"
4036 local mdsdev=$(mdsdevname 1)
4037 local mdsvdev=$(mdsvdevname 1)
4041 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
4043 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
4044 --reformat $(ostdevname 1) $(ostvdevname 1)
4050 echo checking size of lov_objid for ost index $i
4051 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
4052 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
4053 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
4055 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
4062 run_test 55 "check lov_objid size"
4065 local server_version=$(lustre_version_code $SINGLEMDS)
4066 local mds_journal_size_orig=$MDSJOURNALSIZE
4071 for num in $(seq 1 $MDSCOUNT); do
4074 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
4075 $(ostdevname 1) $(ostvdevname 1)
4076 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
4077 $(ostdevname 2) $(ostvdevname 2)
4080 start_ost || error "Unable to start first ost (idx 10000)"
4081 start_ost2 || error "Unable to start second ost (idx 1000)"
4082 mount_client $MOUNT || error "Unable to mount client"
4086 if [[ $server_version -ge $(version_code 2.6.54) ]] ||
4087 [[ $server_version -ge $(version_code 2.5.4) &&
4088 $server_version -lt $(version_code 2.5.11) ]]; then
4089 wait_osc_import_state mds ost1 FULL
4090 wait_osc_import_state mds ost2 FULL
4091 $SETSTRIPE --stripe-count=-1 $DIR/$tfile ||
4092 error "Unable to setstripe $DIR/$tfile"
4093 n=$($LFS getstripe --stripe-count $DIR/$tfile)
4094 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
4099 MDSJOURNALSIZE=$mds_journal_size_orig
4102 run_test 56 "check big OST indexes and out-of-index-order start"
4104 test_57a() { # bug 22656
4105 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4106 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4107 tail -1 | awk '{print $1}')
4108 writeconf_or_reformat
4109 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4110 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" ||
4111 error "tunefs failed"
4113 start_ost && error "OST registration from failnode should fail"
4116 run_test 57a "initial registration from failnode should fail (should return errs)"
4119 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4120 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4121 tail -1 | awk '{print $1}')
4122 writeconf_or_reformat
4123 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4124 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" ||
4125 error "tunefs failed"
4127 start_ost || error "OST registration from servicenode should not fail"
4130 run_test 57b "initial registration from servicenode should not fail"
4133 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
4136 test_58() { # bug 22658
4138 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4139 createmany -o $DIR/$tdir/$tfile-%d 100
4140 # make sure that OSTs do not cancel llog cookies before we unmount the MDS
4141 #define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601
4142 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x601"
4143 unlinkmany $DIR/$tdir/$tfile-%d 100
4144 stop_mds || error "Unable to stop MDS"
4146 local MNTDIR=$(facet_mntpt $SINGLEMDS)
4147 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4149 # remove all files from the OBJECTS dir
4150 mount_fstype $SINGLEMDS
4152 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
4154 unmount_fstype $SINGLEMDS
4155 # restart MDS with missing llog files
4156 start_mds || error "unable to start MDS"
4157 do_facet mds "$LCTL set_param fail_loc=0"
4160 run_test 58 "missing llog files must not prevent MDT from mounting"
4163 start_mgsmds >> /dev/null
4164 local C1=$(count_osts)
4165 if [ $C1 -eq 0 ]; then
4166 start_ost >> /dev/null
4170 echo "original ost count: $C1 (expect > 0)"
4171 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
4172 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
4173 local C2=$(count_osts)
4174 echo "after mdt writeconf count: $C2 (expect 0)"
4175 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
4176 echo "OST start without writeconf should fail:"
4177 start_ost >> /dev/null &&
4178 error "OST start without writeconf didn't fail"
4179 echo "OST start with writeconf should succeed:"
4180 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
4181 local C3=$(count_osts)
4182 echo "after ost writeconf count: $C3 (expect 1)"
4183 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
4184 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
4185 local C4=$(count_osts)
4186 echo "after ost2 writeconf count: $C4 (expect 2)"
4187 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
4188 stop_ost2 >> /dev/null
4189 cleanup_nocli >> /dev/null
4190 #writeconf to remove all ost2 traces for subsequent tests
4191 writeconf_or_reformat
4193 run_test 59 "writeconf mount option"
4195 test_60() { # LU-471
4198 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4199 skip "Only applicable to ldiskfs-based MDTs"
4203 for num in $(seq $MDSCOUNT); do
4204 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
4205 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
4206 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4210 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
4211 [ ${PIPESTATUS[0]} -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
4213 # MDT default has dirdata feature
4214 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
4215 # we disable uninit_bg feature
4216 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
4217 # we set stride extended options
4218 echo $dump | grep stride > /dev/null || error "stride is not set"
4221 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
4226 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] ||
4227 { skip "Need MDS version at least 2.1.53"; return 0; }
4229 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
4230 ! large_xattr_enabled; then
4233 for num in $(seq $MDSCOUNT); do
4234 do_facet mds${num} $TUNE2FS -O large_xattr \
4235 $(mdsdevname $num) ||
4236 error "tune2fs on mds $num failed"
4240 setup_noconfig || error "setting up the filesystem failed"
4241 client_up || error "starting client failed"
4243 local file=$DIR/$tfile
4244 touch $file || error "touch $file failed"
4246 local large_value="$(generate_string $(max_xattr_size))"
4247 local small_value="bar"
4249 local name="trusted.big"
4250 log "save large xattr $name on $file"
4251 setfattr -n $name -v $large_value $file ||
4252 error "saving $name on $file failed"
4254 local new_value=$(get_xattr_value $name $file)
4255 [[ "$new_value" != "$large_value" ]] &&
4256 error "$name different after saving"
4258 log "shrink value of $name on $file"
4259 setfattr -n $name -v $small_value $file ||
4260 error "shrinking value of $name on $file failed"
4262 new_value=$(get_xattr_value $name $file)
4263 [[ "$new_value" != "$small_value" ]] &&
4264 error "$name different after shrinking"
4266 log "grow value of $name on $file"
4267 setfattr -n $name -v $large_value $file ||
4268 error "growing value of $name on $file failed"
4270 new_value=$(get_xattr_value $name $file)
4271 [[ "$new_value" != "$large_value" ]] &&
4272 error "$name different after growing"
4274 log "check value of $name on $file after remounting MDS"
4276 new_value=$(get_xattr_value $name $file)
4277 [[ "$new_value" != "$large_value" ]] &&
4278 error "$name different after remounting MDS"
4280 log "remove large xattr $name from $file"
4281 setfattr -x $name $file || error "removing $name from $file failed"
4284 stopall || error "stopping for e2fsck run"
4285 for num in $(seq $MDSCOUNT); do
4286 run_e2fsck $(facet_active_host mds$num) \
4287 $(mdsdevname $num) "-y" ||
4288 error "e2fsck MDT$num failed"
4290 setup_noconfig || error "remounting the filesystem failed"
4293 # need to delete this file to avoid problems in other tests
4295 stopall || error "stopping systems to turn off large_xattr"
4297 for num in $(seq $MDSCOUNT); do
4298 do_facet mds${num} $TUNE2FS -O ^large_xattr \
4299 $(mdsdevname $num) ||
4300 error "tune2fs on mds $num failed"
4304 run_test 61 "large xattr"
4307 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4308 skip "Only applicable to ldiskfs-based MDTs"
4313 local mdsdev=$(mdsdevname 1)
4314 local ostdev=$(ostdevname 1)
4316 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] ||
4317 { skip "Need MDS version at least 2.2.51"; return 0; }
4319 echo "disable journal for mds"
4320 do_facet mds $TUNE2FS -O ^has_journal $mdsdev || error "tune2fs failed"
4321 start_mds && error "MDT start should fail"
4322 echo "disable journal for ost"
4323 do_facet ost1 $TUNE2FS -O ^has_journal $ostdev || error "tune2fs failed"
4324 start_ost && error "OST start should fail"
4325 cleanup || error "cleanup failed with rc $?"
4328 run_test 62 "start with disabled journal"
4331 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4332 skip "Only applicable to ldiskfs-based MDTs"
4336 local inode_slab=$(do_facet $SINGLEMDS \
4337 "awk '/ldiskfs_inode_cache/ { print \\\$5 }' /proc/slabinfo")
4338 if [ -z "$inode_slab" ]; then
4339 skip "ldiskfs module has not been loaded"
4343 echo "$inode_slab ldisk inodes per page"
4344 [ "$inode_slab" -ge "3" ] ||
4345 error "ldisk inode size is too big, $inode_slab objs per page"
4348 run_test 63 "Verify each page can at least hold 3 ldisk inodes"
4351 start_mds || error "unable to start MDS"
4352 start_ost || error "Unable to start OST1"
4353 start_ost2 || error "Unable to start second ost"
4354 mount_client $MOUNT || error "Unable to mount client"
4355 stop_ost2 || error "Unable to stop second ost"
4357 $LFS df --lazy || error "lfs df failed"
4358 umount_client $MOUNT -f || error “unmount $MOUNT failed”
4359 cleanup_nocli || error "cleanup_nocli failed with $?"
4360 #writeconf to remove all ost2 traces for subsequent tests
4361 writeconf_or_reformat
4363 run_test 64 "check lfs df --lazy "
4365 test_65() { # LU-2237
4366 # Currently, the test is only valid for ldiskfs backend
4367 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4368 skip "non-ldiskfs backend" && return
4370 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4371 local brpt=$(facet_mntpt brpt)
4374 if ! do_facet $SINGLEMDS "test -b $devname"; then
4378 stop_mds || error "Unable to stop MDS"
4379 local obj=$(do_facet $SINGLEMDS \
4380 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
4382 if [ -z "$obj" ]; then
4383 # The MDT may be just re-formatted, mount the MDT for the
4384 # first time to guarantee the "last_rcvd" file is there.
4385 start_mds || error "fail to mount the MDS for the first time"
4386 stop_mds || error "Unable to stop MDS"
4389 # remove the "last_rcvd" file
4390 do_facet $SINGLEMDS "mkdir -p $brpt"
4391 do_facet $SINGLEMDS \
4392 "mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
4393 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
4394 do_facet $SINGLEMDS "$UMOUNT $brpt"
4396 # restart MDS, the "last_rcvd" file should be recreated.
4397 start_mds || error "fail to restart the MDS"
4398 stop_mds || error "Unable to stop MDS"
4399 obj=$(do_facet $SINGLEMDS \
4400 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
4401 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
4403 run_test 65 "re-create the lost last_rcvd file when server mount"
4406 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
4407 { skip "Need MGS version at least 2.3.59"; return 0; }
4410 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
4411 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
4413 echo "replace_nids should fail if MDS, OSTs and clients are UP"
4414 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4415 error "replace_nids fail"
4417 umount_client $MOUNT || error "unmounting client failed"
4418 echo "replace_nids should fail if MDS and OSTs are UP"
4419 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4420 error "replace_nids fail"
4422 stop_ost || error "Unable to stop OST1"
4423 echo "replace_nids should fail if MDS is UP"
4424 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4425 error "replace_nids fail"
4427 stop_mds || error "stopping mds failed"
4429 if combined_mgs_mds; then
4430 start_mdt 1 "-o nosvc" ||
4431 error "starting mds with nosvc option failed"
4434 echo "command should accept two parameters"
4435 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
4436 error "command should accept two params"
4438 echo "correct device name should be passed"
4439 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
4440 error "wrong devname"
4442 echo "wrong nids list should not destroy the system"
4443 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
4446 echo "replace OST nid"
4447 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
4448 error "replace nids failed"
4450 echo "command should accept two parameters"
4451 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
4452 error "command should accept two params"
4454 echo "wrong nids list should not destroy the system"
4455 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
4458 echo "replace MDS nid"
4459 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
4460 error "replace nids failed"
4462 if ! combined_mgs_mds ; then
4465 stop_mds || error "Unable to stop MDS"
4469 check_mount || error "error after nid replace"
4470 cleanup || error "cleanup failed"
4473 run_test 66 "replace nids"
4475 test_67() { #LU-2950
4476 local legacy="$TMP/legacy_lnet_config"
4477 local new="$TMP/new_routes_test"
4478 local out="$TMP/config_out_file"
4479 local verify="$TMP/conv_verify"
4480 local verify_conf="$TMP/conf_verify"
4482 # Create the legacy file that will be run through the
4483 # lustre_routes_conversion script
4484 cat <<- LEGACY_LNET_CONFIG > $legacy
4485 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
4486 tcp2 54 10.1.3.2@tcp;
4487 tcp3 10.3.4.3@tcp:3;
4491 # Create the verification file to verify the output of
4492 # lustre_routes_conversion script against.
4493 cat <<- VERIFY_LNET_CONFIG > $verify
4494 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
4495 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
4496 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
4497 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
4498 tcp4: { gateway: 10.3.3.4@tcp }
4501 # Create the verification file to verify the output of
4502 # lustre_routes_config script against
4503 cat <<- VERIFY_LNET_CONFIG > $verify_conf
4504 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
4505 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
4506 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
4507 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
4508 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
4511 lustre_routes_conversion $legacy $new > /dev/null
4512 if [ -f $new ]; then
4513 # verify the conversion output
4514 cmp -s $new $verify > /dev/null
4515 if [ $? -eq 1 ]; then
4516 error "routes conversion failed"
4519 lustre_routes_config --dry-run --verbose $new > $out
4520 # check that the script succeeded
4521 cmp -s $out $verify_conf > /dev/null
4522 if [ $? -eq 1 ]; then
4523 error "routes config failed"
4526 error "routes conversion test failed"
4528 # remove generated files
4529 rm -f $new $legacy $verify $verify_conf $out
4531 run_test 67 "test routes conversion and configuration"
4539 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.4.53) ] ||
4540 { skip "Need MDS version at least 2.4.53"; return 0; }
4542 umount_client $MOUNT || error "umount client failed"
4544 start_mdt 1 || error "MDT start failed"
4545 start_ost || error "Unable to start OST1"
4547 # START-END - the sequences we'll be reserving
4548 START=$(do_facet $SINGLEMDS \
4549 $LCTL get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
4550 END=$((START + (1 << 30)))
4551 do_facet $SINGLEMDS \
4552 $LCTL set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
4554 # reset the sequences MDT0000 has already assigned
4555 do_facet $SINGLEMDS \
4556 $LCTL set_param seq.srv*MDT0000.space=clear
4558 # remount to let the client allocate new sequence
4559 mount_client $MOUNT || error "mount client failed"
4561 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4562 do_facet $SINGLEMDS \
4563 $LCTL get_param seq.srv*MDT0000.space
4564 $LFS path2fid $DIR/$tfile
4566 local old_ifs="$IFS"
4568 fid=($($LFS path2fid $DIR/$tfile))
4572 if [[ $seq < $END ]]; then
4573 error "used reserved sequence $seq?"
4575 cleanup || error "cleanup failed with $?"
4577 run_test 68 "be able to reserve specific sequences in FLDB"
4579 # Test 69: is about the total number of objects ever created on an OST.
4580 # so that when it is reformatted the normal MDS->OST orphan recovery won't
4581 # just "precreate" the missing objects. In the past it might try to recreate
4582 # millions of objects after an OST was reformatted
4584 local server_version=$(lustre_version_code $SINGLEMDS)
4586 [[ $server_version -lt $(version_code 2.4.2) ]] &&
4587 skip "Need MDS version at least 2.4.2" && return
4589 [[ $server_version -ge $(version_code 2.4.50) ]] &&
4590 [[ $server_version -lt $(version_code 2.5.0) ]] &&
4591 skip "Need MDS version at least 2.5.0" && return
4594 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4596 # use OST0000 since it probably has the most creations
4597 local OSTNAME=$(ostname_from_index 0)
4598 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
4599 local last_id=$(do_facet mds1 $LCTL get_param -n \
4600 osc.$mdtosc_proc1.prealloc_last_id)
4602 # Want to have OST LAST_ID over 5 * OST_MAX_PRECREATE to
4603 # verify that the LAST_ID recovery is working properly. If
4604 # not, then the OST will refuse to allow the MDS connect
4605 # because the LAST_ID value is too different from the MDS
4606 #define OST_MAX_PRECREATE=20000
4607 local ost_max_pre=20000
4608 local num_create=$(( ost_max_pre * 5 + 1 - last_id))
4610 # If the LAST_ID is already over 5 * OST_MAX_PRECREATE, we don't
4611 # need to create any files. So, skip this section.
4612 if [ $num_create -gt 0 ]; then
4613 # Check the number of inodes available on OST0
4615 local ifree=$($LFS df -i $MOUNT | awk '/OST0000/ { print $4 }')
4616 log "On OST0, $ifree inodes available. Want $num_create."
4618 $SETSTRIPE -i 0 $DIR/$tdir ||
4619 error "$SETSTRIPE -i 0 $DIR/$tdir failed"
4620 if [ $ifree -lt 10000 ]; then
4621 files=$(( ifree - 50 ))
4626 local j=$((num_create / files + 1))
4627 for i in $(seq 1 $j); do
4628 createmany -o $DIR/$tdir/$tfile-$i- $files ||
4629 error "createmany fail create $files files: $?"
4630 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
4631 error "unlinkmany failed unlink $files files"
4635 # delete all of the files with objects on OST0 so the
4636 # filesystem is not inconsistent later on
4637 $LFS find $MOUNT --ost 0 -print0 | xargs -0 rm
4639 umount_client $MOUNT || error "umount client failed"
4640 stop_ost || error "OST0 stop failure"
4641 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
4642 $(ostdevname 1) $(ostvdevname 1) ||
4643 error "reformat and replace $ostdev failed"
4644 start_ost || error "OST0 restart failure"
4645 wait_osc_import_state mds ost FULL
4647 mount_client $MOUNT || error "mount client failed"
4648 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
4649 local idx=$($GETSTRIPE -i $DIR/$tdir/$tfile-last)
4650 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
4652 local iused=$($LFS df -i $MOUNT | awk '/OST0000/ { print $3 }')
4653 log "On OST0, $iused used inodes"
4654 [ $iused -ge $((ost_max_pre/2 + 1000)) ] &&
4655 error "OST replacement created too many inodes; $iused"
4656 cleanup || error "cleanup failed with $?"
4658 run_test 69 "replace an OST with the same index"
4661 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4664 cleanup || error "cleanup failed with $?"
4666 start_mdt 1 || error "MDT0 start fail"
4668 start_ost || error "OST0 start fail"
4669 for num in $(seq 2 $MDSCOUNT); do
4670 start_mdt $num || return
4673 mount_client $MOUNT || error "mount client fails"
4675 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
4677 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4678 error "create remote dir fail"
4680 rm -rf $DIR/$tdir || error "delete dir fail"
4681 cleanup || error "cleanup failed with $?"
4683 run_test 70a "start MDT0, then OST, then MDT1"
4686 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4689 start_ost || error "OST0 start fail"
4691 start_mds || error "MDS start fail"
4693 mount_client $MOUNT || error "mount client fails"
4695 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
4697 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4698 error "create remote dir fail"
4700 rm -rf $DIR/$tdir || error "delete dir fail"
4702 cleanup || error "cleanup failed with $?"
4704 run_test 70b "start OST, MDT1, MDT0"
4707 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4710 start_mds || error "MDS start fail"
4711 start_ost || error "OST0 start fail"
4713 mount_client $MOUNT || error "mount client fails"
4714 stop_mdt 1 || error "MDT1 start fail"
4716 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
4717 echo "deactivate $mdc_for_mdt1"
4718 $LCTL --device $mdc_for_mdt1 deactivate ||
4719 error "set $mdc_for_mdt1 deactivate failed"
4721 mkdir $DIR/$tdir && error "mkdir succeed"
4723 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4724 error "create remote dir succeed"
4726 cleanup || error "cleanup failed with $?"
4728 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
4731 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4734 start_mds || error "MDS start fail"
4735 start_ost || error "OST0 start fail"
4737 mount_client $MOUNT || error "mount client fails"
4739 stop_mdt 2 || error "MDT1 start fail"
4741 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
4743 echo "deactivate $mdc_for_mdt2"
4744 $LCTL --device $mdc_for_mdt2 deactivate ||
4745 error "set $mdc_for_mdt2 deactivate failed"
4747 mkdir $DIR/$tdir || error "mkdir fail"
4748 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4749 error "create remote dir succeed"
4751 rm -rf $DIR/$tdir || error "delete dir fail"
4753 cleanup || error "cleanup failed with $?"
4755 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
4758 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4760 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.62) ] ||
4761 { skip "Need MDS version at least 2.7.62"; return 0; }
4763 cleanup || error "cleanup failed with $?"
4765 local mdsdev=$(mdsdevname 1)
4766 local ostdev=$(ostdevname 1)
4767 local mdsvdev=$(mdsvdevname 1)
4768 local ostvdev=$(ostvdevname 1)
4769 local opts_mds="$(mkfs_opts mds1 $mdsdev) --reformat $mdsdev $mdsvdev"
4770 local opts_ost="$(mkfs_opts ost1 $ostdev) --reformat $ostdev $ostvdev"
4772 add mds1 $opts_mds || error "add mds1 failed"
4773 start_mdt 1 || error "start mdt1 failed"
4774 add ost1 $opts_ost || error "add ost1 failed"
4775 start_ost || error "start ost failed"
4776 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
4778 local soc=$(do_facet mds1 "$LCTL get_param -n \
4779 mdt.*MDT0000.sync_lock_cancel")
4780 [ $soc == "never" ] || error "SoC enabled on single MDS"
4782 for i in $(seq 2 $MDSCOUNT); do
4783 mdsdev=$(mdsdevname $i)
4784 mdsvdev=$(mdsvdevname $i)
4785 opts_mds="$(mkfs_opts mds$i $mdsdev) --reformat $mdsdev \
4787 add mds$i $opts_mds || error "add mds$i failed"
4788 start_mdt $i || error "start mdt$i fail"
4791 wait_dne_interconnect
4793 for i in $(seq $MDSCOUNT); do
4794 soc=$(do_facet mds$i "$LCTL get_param -n \
4795 mdt.*MDT000$((i - 1)).sync_lock_cancel")
4796 [ $soc == "blocking" ] || error "SoC not enabled on DNE"
4799 for i in $(seq 2 $MDSCOUNT); do
4800 stop_mdt $i || error "stop mdt$i fail"
4802 soc=$(do_facet mds1 "$LCTL get_param -n \
4803 mdt.*MDT0000.sync_lock_cancel")
4804 [ $soc == "never" ] || error "SoC enabled on single MDS"
4806 cleanup || error "cleanup failed with $?"
4808 run_test 70e "Sync-on-Cancel will be enabled by default on DNE"
4811 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4812 if combined_mgs_mds; then
4813 skip "needs separate MGS/MDT" && return
4817 start_mdt 1 || error "MDT0 start fail"
4818 start_ost || error "OST0 start fail"
4819 for num in $(seq 2 $MDSCOUNT); do
4820 start_mdt $num || return
4823 start_ost2 || error "OST1 start fail"
4825 mount_client $MOUNT || error "mount client fails"
4827 mkdir $DIR/$tdir || error "mkdir fail"
4828 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4829 error "create remote dir succeed"
4831 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4832 rm -rf $DIR/$tdir || error "delete dir fail"
4834 umount_client $MOUNT || error "umount_client failed"
4835 stop_mds || error "MDS stop fail"
4836 stop_ost || error "OST0 stop fail"
4837 stop_ost2 || error "OST1 stop fail"
4839 run_test 71a "start MDT0 OST0, MDT1, OST1"
4842 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4843 if combined_mgs_mds; then
4844 skip "needs separate MGS/MDT" && return
4848 for num in $(seq 2 $MDSCOUNT); do
4849 start_mdt $num || return
4851 start_ost || error "OST0 start fail"
4852 start_mdt 1 || error "MDT0 start fail"
4853 start_ost2 || error "OST1 start fail"
4855 mount_client $MOUNT || error "mount client fails"
4857 mkdir $DIR/$tdir || error "mkdir fail"
4858 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4859 error "create remote dir succeed"
4861 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4862 rm -rf $DIR/$tdir || error "delete dir fail"
4864 umount_client $MOUNT || error "umount_client failed"
4865 stop_mds || error "MDT0 stop fail"
4866 stop_ost || error "OST0 stop fail"
4867 stop_ost2 || error "OST1 stop fail"
4869 run_test 71b "start MDT1, OST0, MDT0, OST1"
4872 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4873 if combined_mgs_mds; then
4874 skip "needs separate MGS/MDT" && return
4878 start_ost || error "OST0 start fail"
4879 start_ost2 || error "OST1 start fail"
4880 for num in $(seq 2 $MDSCOUNT); do
4881 start_mdt $num || return
4883 start_mdt 1 || error "MDT0 start fail"
4885 mount_client $MOUNT || error "mount client fails"
4887 mkdir $DIR/$tdir || error "mkdir fail"
4888 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4889 error "create remote dir succeed"
4891 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4892 rm -rf $DIR/$tdir || error "delete dir fail"
4894 umount_client $MOUNT || error "umount_client failed"
4895 stop_mds || error "MDS stop fail"
4896 stop_ost || error "OST0 stop fail"
4897 stop_ost2 || error "OST1 stop fail"
4900 run_test 71c "start OST0, OST1, MDT1, MDT0"
4903 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4904 if combined_mgs_mds; then
4905 skip "needs separate MGS/MDT" && return
4909 start_ost || error "OST0 start fail"
4910 for num in $(seq 2 $MDSCOUNT); do
4911 start_mdt $num || return
4913 start_mdt 1 || error "MDT0 start fail"
4914 start_ost2 || error "OST1 start fail"
4916 mount_client $MOUNT || error "mount client fails"
4918 mkdir $DIR/$tdir || error "mkdir fail"
4919 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4920 error "create remote dir succeed"
4922 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4923 rm -rf $DIR/$tdir || error "delete dir fail"
4925 umount_client $MOUNT || error "umount_client failed"
4926 stop_mds || error "MDS stop fail"
4927 stop_ost || error "OST0 stop fail"
4928 stop_ost2 || error "OST1 stop fail"
4931 run_test 71d "start OST0, MDT1, MDT0, OST1"
4934 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4935 if combined_mgs_mds; then
4936 skip "needs separate MGS/MDT" && return
4940 start_ost || error "OST0 start fail"
4941 for num in $(seq 2 $MDSCOUNT); do
4942 start_mdt $num || return
4944 start_ost2 || error "OST1 start fail"
4945 start_mdt 1 || error "MDT0 start fail"
4947 mount_client $MOUNT || error "mount client fails"
4949 mkdir $DIR/$tdir || error "mkdir fail"
4950 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4951 error "create remote dir succeed"
4953 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4954 rm -rf $DIR/$tdir || error "delete dir fail"
4956 umount_client $MOUNT || error "umount_client failed"
4957 stop_mds || error "MDS stop fail"
4958 stop_ost || error "OST0 stop fail"
4959 stop_ost2 || error "OST1 stop fail"
4962 run_test 71e "start OST0, MDT1, OST1, MDT0"
4964 test_72() { #LU-2634
4965 local mdsdev=$(mdsdevname 1)
4966 local ostdev=$(ostdevname 1)
4967 local cmd="$E2FSCK -fnvd $mdsdev"
4970 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4971 skip "ldiskfs only test" && return
4973 #tune MDT with "-O extents"
4975 for num in $(seq $MDSCOUNT); do
4976 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
4977 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4978 error "add mds $num failed"
4979 do_facet mds${num} "$TUNE2FS -O extents $(mdsdevname $num)" ||
4980 error "$TUNE2FS failed on mds${num}"
4983 add ost1 $(mkfs_opts ost1 $ostdev) --reformat $ostdev ||
4984 error "add $ostdev failed"
4985 start_mgsmds || error "start mds failed"
4986 start_ost || error "start ost failed"
4987 mount_client $MOUNT || error "mount client failed"
4989 #create some short symlinks
4990 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4991 createmany -o $DIR/$tdir/$tfile-%d $fn
4992 echo "create $fn short symlinks"
4993 for i in $(seq -w 1 $fn); do
4994 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
4999 umount_client $MOUNT || error "umount client failed"
5000 stop_mds || error "stop mds failed"
5001 stop_ost || error "stop ost failed"
5004 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
5006 run_test 72 "test fast symlink with extents flag enabled"
5008 test_73() { #LU-3006
5009 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
5010 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
5011 error "1st tunefs failed"
5012 start_mgsmds || error "start mds failed"
5013 start_ost || error "start ost failed"
5014 mount_client $MOUNT || error "mount client failed"
5015 $LCTL get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
5016 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
5017 umount_client $MOUNT || error "umount client failed"
5021 run_test 73 "failnode to update from mountdata properly"
5023 test_75() { # LU-2374
5024 [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] &&
5025 skip "Need MDS version at least 2.4.1" && return
5028 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
5029 --reformat $(mdsdevname 1) $(mdsvdevname 1)"
5030 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
5031 --reformat $(ostdevname 1) $(ostvdevname 1)"
5033 #check with default parameters
5034 add mds1 $opts_mds || error "add mds1 failed for default params"
5035 add ost1 $opts_ost || error "add ost1 failed for default params"
5037 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
5038 opts_mds=$(echo $opts_mds |
5039 sed -e "s/--index=$index/--index=$index --mdt/")
5040 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
5041 opts_ost=$(echo $opts_ost |
5042 sed -e "s/--index=$index/--index=$index --ost/")
5044 add mds1 $opts_mds || error "add mds1 failed for new params"
5045 add ost1 $opts_ost || error "add ost1 failed for new params"
5048 run_test 75 "The order of --index should be irrelevant"
5051 [[ $(lustre_version_code mgs) -ge $(version_code 2.4.52) ]] ||
5052 { skip "Need MDS version at least 2.4.52" && return 0; }
5054 local MDMB_PARAM="osc.*.max_dirty_mb"
5055 echo "Change MGS params"
5056 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
5058 echo "max_dirty_mb: $MAX_DIRTY_MB"
5059 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB + MAX_DIRTY_MB))
5060 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
5061 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
5062 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5063 head -1" $NEW_MAX_DIRTY_MB
5064 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5065 echo "$MAX_DIRTY_MB"
5066 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5067 error "error while apply max_dirty_mb"
5069 echo "Check the value is stored after remount"
5072 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5073 head -1" $NEW_MAX_DIRTY_MB
5074 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5075 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5076 error "max_dirty_mb is not saved after remount"
5078 echo "Change OST params"
5079 CLIENT_PARAM="obdfilter.*.client_cache_count"
5080 local CLIENT_CACHE_COUNT
5081 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5083 echo "client_cache_count: $CLIENT_CACHE_COUNT"
5084 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
5085 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
5086 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
5087 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5088 head -1" $NEW_CLIENT_CACHE_COUNT
5089 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5091 echo "$CLIENT_CACHE_COUNT"
5092 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5093 error "error while apply client_cache_count"
5095 echo "Check the value is stored after remount"
5098 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5099 head -1" $NEW_CLIENT_CACHE_COUNT
5100 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5102 echo "$CLIENT_CACHE_COUNT"
5103 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5104 error "client_cache_count is not saved after remount"
5107 run_test 76a "set permanent params set_param -P"
5109 test_76b() { # LU-4783
5110 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.57) ]] ||
5111 { skip "Need MGS version at least 2.5.57" && return 0; }
5114 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
5115 error "start params log failed"
5118 run_test 76b "verify params log setup correctly"
5120 test_77() { # LU-3445
5121 local server_version=$(lustre_version_code $SINGLEMDS)
5123 [[ $server_version -ge $(version_code 2.2.60) ]] &&
5124 [[ $server_version -le $(version_code 2.4.0) ]] &&
5125 skip "Need MDS version < 2.2.60 or > 2.4.0" && return
5127 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
5128 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
5129 skip_env "mixed loopback and real device not working" && return
5132 local fs2mdsdev=$(mdsdevname 1_2)
5133 local fs2ostdev=$(ostdevname 1_2)
5134 local fs2mdsvdev=$(mdsvdevname 1_2)
5135 local fs2ostvdev=$(ostvdevname 1_2)
5136 local fsname=test1234
5138 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
5140 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
5141 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
5142 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
5143 error "start fs2mds failed"
5145 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
5146 [[ $mgsnid = *,* ]] || mgsnid+=",$mgsnid"
5148 add fs2ost $(mkfs_opts ost1 $fs2ostdev) --mgsnode=$mgsnid \
5149 --failnode=$failnid --fsname=$fsname \
5150 --reformat $fs2ostdev $fs2ostvdev ||
5151 error "add fs2ost failed"
5152 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
5154 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
5155 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
5156 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
5159 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
5162 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ||
5163 $(facet_fstype ost1) != ldiskfs ]] &&
5164 skip "only applicable to ldiskfs-based MDTs and OSTs" && return
5166 # reformat the Lustre filesystem with a smaller size
5167 local saved_MDSCOUNT=$MDSCOUNT
5168 local saved_MDSSIZE=$MDSSIZE
5169 local saved_OSTCOUNT=$OSTCOUNT
5170 local saved_OSTSIZE=$OSTSIZE
5173 MDSSIZE=$((MDSSIZE - 20000))
5174 OSTSIZE=$((OSTSIZE - 20000))
5175 reformat || error "(1) reformat Lustre filesystem failed"
5176 MDSSIZE=$saved_MDSSIZE
5177 OSTSIZE=$saved_OSTSIZE
5179 # mount the Lustre filesystem
5180 setup_noconfig || error "(2) setup Lustre filesystem failed"
5183 log "create test files"
5188 mkdir $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
5190 for i in $(seq $num_files); do
5191 file=$MOUNT/$tdir/$tfile-$i
5192 dd if=/dev/urandom of=$file count=1 bs=1M || {
5193 $LCTL get_param osc.*.cur*grant*
5194 $LFS df; $LFS df -i;
5195 # stop creating files if there is no more space
5196 if [ ! -e $file ]; then
5197 num_files=$((i - 1))
5201 $LFS getstripe -v $file
5202 local ost_idx=$(LFS getstripe -i $file)
5203 do_facet ost$((ost_idx + 1)) \
5204 $LCTL get_param obdfilter.*.*grant*
5205 error "(4) create $file failed"
5209 # unmount the Lustre filesystem
5210 cleanup || error "(5) cleanup Lustre filesystem failed"
5212 # run e2fsck on the MDT and OST devices
5213 local mds_host=$(facet_active_host $SINGLEMDS)
5214 local ost_host=$(facet_active_host ost1)
5215 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
5216 local ost_dev=$(ostdevname 1)
5218 run_e2fsck $mds_host $mds_dev "-y"
5219 run_e2fsck $ost_host $ost_dev "-y"
5221 # get the original block count of the MDT and OST filesystems
5222 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
5223 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
5225 # expand the MDT and OST filesystems to the device size
5226 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
5227 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
5229 # run e2fsck on the MDT and OST devices again
5230 run_e2fsck $mds_host $mds_dev "-y"
5231 run_e2fsck $ost_host $ost_dev "-y"
5233 # mount the Lustre filesystem
5237 log "check files after expanding the MDT and OST filesystems"
5238 for i in $(seq $num_files); do
5239 file=$MOUNT/$tdir/$tfile-$i
5240 $CHECKSTAT -t file -s 1048576 $file ||
5241 error "(6) checkstat $file failed"
5245 log "create more files after expanding the MDT and OST filesystems"
5246 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
5247 file=$MOUNT/$tdir/$tfile-$i
5248 dd if=/dev/urandom of=$file count=1 bs=1M ||
5249 error "(7) create $file failed"
5252 # unmount the Lustre filesystem
5253 cleanup || error "(8) cleanup Lustre filesystem failed"
5255 # run e2fsck on the MDT and OST devices
5256 run_e2fsck $mds_host $mds_dev "-y"
5257 run_e2fsck $ost_host $ost_dev "-y"
5259 # get the maximum block count of the MDT and OST filesystems
5260 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
5261 local ost_max_blks=$(get_block_count ost1 $ost_dev)
5263 # get the minimum block count of the MDT and OST filesystems
5264 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
5265 grep minimum | sed -e 's/^.*filesystem: //g')
5266 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
5267 grep minimum | sed -e 's/^.*filesystem: //g')
5269 # shrink the MDT and OST filesystems to a smaller size
5273 if [[ $mds_max_blks -gt $mds_min_blks &&
5274 $mds_max_blks -gt $mds_orig_blks ]]; then
5275 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
5276 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
5277 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
5278 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
5279 error "shrink $SINGLEMDS to $new_blks failed"
5283 if [[ $ost_max_blks -gt $ost_min_blks &&
5284 $ost_max_blks -gt $ost_orig_blks ]]; then
5285 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
5286 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
5287 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
5288 run_resize2fs ost1 $ost_dev $new_blks ||
5289 error "shrink ost1 to $new_blks failed"
5293 # check whether the MDT or OST filesystem was shrunk or not
5295 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
5296 reformat || error "(10) reformat Lustre filesystem failed"
5300 # run e2fsck on the MDT and OST devices again
5301 run_e2fsck $mds_host $mds_dev "-y"
5302 run_e2fsck $ost_host $ost_dev "-y"
5304 # mount the Lustre filesystem again
5308 log "check files after shrinking the MDT and OST filesystems"
5309 for i in $(seq $((num_files + 10))); do
5310 file=$MOUNT/$tdir/$tfile-$i
5311 $CHECKSTAT -t file -s 1048576 $file ||
5312 error "(11) checkstat $file failed"
5315 # unmount and reformat the Lustre filesystem
5316 cleanup || error "(12) cleanup Lustre filesystem failed"
5317 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
5319 MDSCOUNT=$saved_MDSCOUNT
5320 OSTCOUNT=$saved_OSTCOUNT
5321 reformat || error "(14) reformat Lustre filesystem failed"
5323 run_test 78 "run resize2fs on MDT and OST filesystems"
5325 test_79() { # LU-4227
5326 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.5.59) ]] ||
5327 { skip "Need MDS version at least 2.5.59"; return 0; }
5329 local mdsdev1=$(mdsdevname 1)
5330 local mdsvdev1=$(mdsvdevname 1)
5331 local mdsdev2=$(mdsdevname 2)
5332 local mdsvdev2=$(mdsvdevname 2)
5333 local ostdev1=$(ostdevname 1)
5334 local ostvdev1=$(ostvdevname 1)
5335 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
5336 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
5337 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
5340 # remove --mgs/--mgsnode from mkfs.lustre options
5341 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
5343 mgsnode_opt=$(echo $opts_mds2 |
5344 awk '{ for ( i = 1; i < NF; i++ )
5345 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5346 [ -n $mgsnode_opt ] &&
5347 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
5349 mgsnode_opt=$(echo $opts_ost1 |
5350 awk '{ for ( i = 1; i < NF; i++ )
5351 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5352 [ -n $mgsnode_opt ] &&
5353 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
5355 # -MGS, format a mdt without --mgs option
5356 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
5357 error "Must specify --mgs when formatting mdt combined with mgs"
5359 # +MGS, format a mdt/ost without --mgsnode option
5360 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
5361 > /dev/null || error "start mds1 failed"
5362 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
5363 error "Must specify --mgsnode when formatting a mdt"
5364 add ost1 $opts_ost1 $ostdev1 $ostvdev2 &&
5365 error "Must specify --mgsnode when formatting an ost"
5369 run_test 79 "format MDT/OST without mgs option (should return errors)"
5372 start_mds || error "Failed to start MDT"
5373 start_ost || error "Failed to start OST1"
5374 uuid=$(do_facet ost1 $LCTL get_param -n mgc.*.uuid)
5375 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
5376 do_facet ost1 "$LCTL set_param fail_val=10 fail_loc=0x906"
5377 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x906"
5378 do_facet mgs "$LCTL set_param -n mgs/MGS/evict_client $uuid"
5380 start_ost2 || error "Failed to start OST2"
5382 do_facet ost1 "$LCTL set_param fail_loc=0"
5385 run_test 80 "mgc import reconnect race"
5387 #Save the original values of $OSTCOUNT and $OSTINDEX$i.
5389 local new_ostcount=$1
5390 saved_ostcount=$OSTCOUNT
5391 OSTCOUNT=$new_ostcount
5395 for i in $(seq $OSTCOUNT); do
5397 eval saved_ostindex$i=${!index}
5402 # Restore the original values of $OSTCOUNT and $OSTINDEX$i.
5403 restore_ostindex() {
5408 for i in $(seq $OSTCOUNT); do
5409 index=saved_ostindex$i
5410 eval OSTINDEX$i=${!index}
5412 OSTCOUNT=$saved_ostcount
5417 # The main purpose of this test is to ensure the OST_INDEX_LIST functions as
5418 # expected. This test uses OST_INDEX_LIST to format OSTs with a randomly
5419 # assigned index and ensures we can mount such a formatted file system
5420 test_81() { # LU-4665
5421 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5422 { skip "Need MDS version at least 2.6.54" && return; }
5423 [[ $OSTCOUNT -ge 3 ]] || { skip_env "Need at least 3 OSTs" && return; }
5427 # Each time RANDOM is referenced, a random integer between 0 and 32767
5430 local saved_ostindex1=$OSTINDEX1
5431 for i in 65535 $((RANDOM + 65536)); do
5432 echo -e "\nFormat ost1 with --index=$i, should fail"
5434 if add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat \
5435 $(ostdevname 1) $(ostvdevname 1); then
5436 OSTINDEX1=$saved_ostindex1
5437 error "format ost1 with --index=$i should fail"
5440 OSTINDEX1=$saved_ostindex1
5444 # Format OSTs with random sparse indices.
5445 trap "restore_ostindex" EXIT
5446 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices"
5447 OST_INDEX_LIST=[0,$((RANDOM * 2 % 65533 + 1)),65534] formatall
5449 # Setup and check Lustre filesystem.
5450 start_mgsmds || error "start_mgsmds failed"
5451 for i in $(seq $OSTCOUNT); do
5452 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5453 error "start ost$i failed"
5456 mount_client $MOUNT || error "mount client $MOUNT failed"
5457 check_mount || error "check client $MOUNT failed"
5460 local max_easize=$($LCTL get_param -n llite.*.max_easize)
5461 [[ $max_easize -eq 128 ]] ||
5462 error "max_easize is $max_easize, should be 128 bytes"
5466 run_test 81 "sparse OST indexing"
5468 # Wait OSTs to be active on both client and MDT side.
5470 local cmd="$LCTL get_param -n lov.$FSNAME-clilov-*.target_obd |
5471 awk 'BEGIN {c = 0} /ACTIVE/{c += 1} END {printf \\\"%d\\\", c}'"
5472 wait_update $HOSTNAME "eval $cmd" $OSTCOUNT ||
5473 error "wait_update OSTs up on client failed"
5475 cmd="$LCTL get_param -n lod.$FSNAME-MDT*-*.target_obd | sort -u |
5476 awk 'BEGIN {c = 0} /ACTIVE/{c += 1} END {printf \\\"%d\\\", c}'"
5477 wait_update_facet $SINGLEMDS "eval $cmd" $OSTCOUNT ||
5478 error "wait_update OSTs up on MDT failed"
5481 # Here we exercise the stripe placement functionality on a file system that
5482 # has formatted the OST with a random index. With the file system the following
5483 # functionality is tested:
5485 # 1. Creating a new file with a specific stripe layout.
5487 # 2. Modifiy a existing empty file with a specific stripe layout.
5489 # 3. Ensure we fail to set the stripe layout of a file that already has one.
5491 # 4. If ost-index is defined we need to ensure it is the first entry in the
5492 # ost index list returned by lfs getstripe.
5494 # 5. Lastly ensure this functionality fails with directories.
5495 test_82a() { # LU-4665
5496 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5497 { skip "Need MDS version at least 2.6.54" && return; }
5498 [[ $OSTCOUNT -ge 3 ]] || { skip_env "Need at least 3 OSTs" && return; }
5504 # Format OSTs with random sparse indices.
5508 local LOV_V1_INSANE_STRIPE_COUNT=65532
5509 for i in $(seq $OSTCOUNT); do
5510 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
5511 ost_indices+=" $index"
5513 ost_indices=$(comma_list $ost_indices)
5515 trap "restore_ostindex" EXIT
5516 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5517 OST_INDEX_LIST=[$ost_indices] formatall
5519 # Setup Lustre filesystem.
5520 start_mgsmds || error "start_mgsmds failed"
5521 for i in $(seq $OSTCOUNT); do
5522 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5523 error "start ost$i failed"
5526 mount_client $MOUNT || error "mount client $MOUNT failed"
5529 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5530 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5532 # 1. If the file does not exist, new file will be created
5533 # with specified OSTs.
5534 local file=$DIR/$tdir/$tfile-1
5535 local cmd="$SETSTRIPE -o $ost_indices $file"
5537 eval $cmd || error "$cmd failed"
5538 check_stripe_count $file $OSTCOUNT
5539 check_obdidx $file $ost_indices
5540 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5541 error "write $file failed"
5543 # 2. If the file already exists and is an empty file, the file
5544 # will be attached with specified layout.
5545 file=$DIR/$tdir/$tfile-2
5546 mcreate $file || error "mcreate $file failed"
5547 cmd="$SETSTRIPE -o $ost_indices $file"
5549 eval $cmd || error "$cmd failed"
5550 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5551 error "write $file failed"
5552 check_stripe_count $file $OSTCOUNT
5553 check_obdidx $file $ost_indices
5555 # 3. If the file already has a valid layout attached, the command
5556 # should fail with EBUSY.
5558 eval $cmd && error "stripe is already set on $file, $cmd should fail"
5560 # 4. If [--stripe-index|-i <start_ost_idx>] is used, the index must
5561 # be in the OST indices list.
5562 local start_ost_idx=${ost_indices##*,}
5563 file=$DIR/$tdir/$tfile-3
5564 cmd="$SETSTRIPE -o $ost_indices -i $start_ost_idx $file"
5566 eval $cmd || error "$cmd failed"
5567 check_stripe_count $file $OSTCOUNT
5568 check_obdidx $file $ost_indices
5569 check_start_ost_idx $file $start_ost_idx
5571 file=$DIR/$tdir/$tfile-4
5573 cmd+=" -o $(exclude_items_from_list $ost_indices $start_ost_idx)"
5574 cmd+=" -i $start_ost_idx $file"
5576 eval $cmd && error "index $start_ost_idx should be in $ost_indices"
5578 # 5. Specifying OST indices for directory should fail with ENOSUPP.
5579 local dir=$DIR/$tdir/$tdir
5580 mkdir $dir || error "mkdir $dir failed"
5581 cmd="$SETSTRIPE -o $ost_indices $dir"
5583 eval $cmd && error "$cmd should fail, specifying OST indices" \
5584 "for directory is not supported"
5588 run_test 82a "specify OSTs for file (succeed) or directory (fail)"
5593 # Remove OSTs from a pool and destroy the pool.
5594 destroy_pool $ost_pool || true
5599 # Test 82b is run to ensure that if the user supplies a pool with a specific
5600 # stripe layout that it behaves proprerly. It should fail in the case that
5601 # the supplied OST index list points to OSTs not contained in the user
5603 test_82b() { # LU-4665
5604 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5605 { skip "Need MDS version at least 2.6.54" && return; }
5606 [[ $OSTCOUNT -ge 4 ]] || { skip_env "Need at least 4 OSTs" && return; }
5612 # Format OSTs with random sparse indices.
5616 local LOV_V1_INSANE_STRIPE_COUNT=65532
5617 for i in $(seq $OSTCOUNT); do
5618 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
5619 ost_indices+=" $index"
5621 ost_indices=$(comma_list $ost_indices)
5623 trap "restore_ostindex" EXIT
5624 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5625 OST_INDEX_LIST=[$ost_indices] formatall
5627 # Setup Lustre filesystem.
5628 start_mgsmds || error "start_mgsmds failed"
5629 for i in $(seq $OSTCOUNT); do
5630 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5631 error "start ost$i failed"
5634 mount_client $MOUNT || error "mount client $MOUNT failed"
5636 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5637 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5639 # Create a new pool and add OSTs into it.
5640 local ost_pool=$FSNAME.$TESTNAME
5641 create_pool $ost_pool || error "create OST pool $ost_pool failed"
5644 trap "cleanup_82b" EXIT
5646 local ost_idx_in_list=${ost_indices##*,}
5647 local ost_idx_in_pool=$(exclude_items_from_list $ost_indices \
5650 local ost_targets="$FSNAME-OST["
5651 for i in ${ost_idx_in_pool//,/ }; do
5652 ost_targets=$ost_targets$(printf "%04x," $i)
5654 ost_targets="${ost_targets%,}]"
5656 local ost_targets_uuid=$(for i in ${ost_idx_in_pool//,/ }; \
5657 do printf "$FSNAME-OST%04x_UUID\n" $i; done |
5658 sort -u | tr '\n' ' ')
5660 local cmd="$LCTL pool_add $ost_pool $ost_targets"
5661 do_facet mgs $cmd || error "$cmd failed"
5662 wait_update $HOSTNAME "$LCTL get_param -n lov.$FSNAME-*.pools.$TESTNAME|
5663 sort -u | tr '\n' ' ' " "$ost_targets_uuid" ||
5664 error "wait_update $ost_pool failed"
5665 pool_list $ost_pool || error "list OST pool $ost_pool failed"
5667 # If [--pool|-p <pool_name>] is set with [--ost-list|-o <ost_indices>],
5668 # then the OSTs must be the members of the pool.
5669 local file=$DIR/$tdir/$tfile
5670 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
5672 eval $cmd && error "OST with index $ost_idx_in_list should be" \
5673 "in OST pool $ost_pool"
5675 # Only select OST $ost_idx_in_list from $ost_pool for file.
5676 ost_idx_in_list=${ost_idx_in_pool#*,}
5677 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
5679 eval $cmd || error "$cmd failed"
5680 cmd="$GETSTRIPE $file"
5682 eval $cmd || error "$cmd failed"
5683 check_stripe_count $file 2
5684 check_obdidx $file $ost_idx_in_list
5685 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5686 error "write $file failed"
5690 run_test 82b "specify OSTs for file with --pool and --ost-list options"
5693 [[ $(lustre_version_code ost1) -ge $(version_code 2.6.91) ]] ||
5694 { skip "Need OST version at least 2.6.91" && return 0; }
5695 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
5696 skip "Only applicable to ldiskfs-based MDTs"
5706 ostmnt=$(facet_mntpt ost1)
5707 fstype=$(facet_fstype ost1)
5709 # Mount the OST as an ldiskfs filesystem.
5710 log "mount the OST $dev as a $fstype filesystem"
5711 add ost1 $(mkfs_opts ost1 $dev) $FSTYPE_OPT \
5712 --reformat $dev $dev > /dev/null ||
5713 error "format ost1 error"
5715 if ! test -b $dev; then
5716 mnt_opts=$(csa_add "$OST_MOUNT_OPTS" -o loop)
5718 echo "mnt_opts $mnt_opts"
5719 do_facet ost1 mount -t $fstype $dev \
5721 # Run llverfs on the mounted ldiskfs filesystem.
5722 # It is needed to get ENOSPACE.
5723 log "run llverfs in partial mode on the OST $fstype $ostmnt"
5724 do_rpc_nodes $(facet_host ost1) run_llverfs $ostmnt -vpl \
5725 "no" || error "run_llverfs error on $fstype"
5728 log "unmount the OST $dev"
5731 # Delete file IO_scrub. Later osd_scrub_setup will try to
5732 # create "IO_scrub" but will get ENOSPACE.
5734 echo "start ost1 service on `facet_active_host ost1`"
5735 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
5738 err=$(do_facet ost1 dmesg | grep "VFS: Busy inodes after unmount of")
5739 echo "string err $err"
5740 [ -z "$err" ] || error $err
5743 run_test 83 "ENOSPACE on OST doesn't cause message VFS: \
5744 Busy inodes after unmount ..."
5747 local facet=$SINGLEMDS
5748 local num=$(echo $facet | tr -d "mds")
5749 local dev=$(mdsdevname $num)
5750 local time_min=$(recovery_time_min)
5751 local recovery_duration
5752 local completed_clients
5753 local correct_clients
5756 echo "start mds service on $(facet_active_host $facet)"
5758 "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
5759 error "start MDS failed"
5761 start_ost || error "start OST0000 failed"
5762 start_ost2 || error "start OST0001 failed"
5764 echo "recovery_time=$time_min, timeout=$TIMEOUT, wrap_up=$wrap_up"
5766 mount_client $MOUNT1 || error "mount $MOUNT1 failed"
5767 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
5768 # make sure new superblock labels are sync'd before disabling writes
5772 replay_barrier $SINGLEMDS
5773 createmany -o $DIR1/$tfile-%d 1000
5775 # We need to catch the end of recovery window to extend it.
5776 # Skip 5 requests and add delay to request handling.
5777 #define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP
5778 do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5"
5780 facet_failover --fsck $SINGLEMDS || error "failover: $?"
5783 echo "recovery status"
5784 do_facet $SINGLEMDS \
5785 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status"
5787 recovery_duration=$(do_facet $SINGLEMDS \
5788 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
5789 awk '/recovery_duration/ { print $2 }')
5790 (( $recovery_duration > $time_min + $wrap_up )) &&
5791 error "recovery_duration > recovery_time_hard + wrap up"
5792 completed_clients=$(do_facet $SINGLEMDS \
5793 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
5794 awk '/completed_clients/ { print $2 }')
5796 correct_clients="$MDSCOUNT/$((MDSCOUNT+1))"
5797 [ "$completed_clients" = "${correct_clients}" ] ||
5798 error "$completed_clients != $correct_clients"
5800 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
5801 umount_client $MOUNT1
5802 umount_client $MOUNT2
5808 run_test 84 "check recovery_hard_time"
5811 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.55) ]] ||
5812 { skip "Need OST version at least 2.7.55" && return 0; }
5813 ##define OBD_FAIL_OSD_OST_EA_FID_SET 0x197
5814 do_facet ost1 "lctl set_param fail_loc=0x197"
5818 run_test 85 "osd_ost init: fail ea_fid_set"
5821 [ "$(facet_fstype ost1)" = "zfs" ] &&
5822 skip "LU-6442: no such mkfs params for ZFS OSTs" && return
5824 local OST_OPTS="$(mkfs_opts ost1 $(ostdevname 1)) \
5825 --reformat $(ostdevname 1) $(ostvdevname 1)"
5828 local OLDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
5829 awk '/Flex block group size: / { print $NF; exit; }')
5832 if [[ ${!opts} != *mkfsoptions* ]]; then
5833 eval opts=\"${!opts} \
5834 --mkfsoptions='\\\"-O flex_bg -G $NEWSIZE\\\"'\"
5836 val=${!opts//--mkfsoptions=\\\"/ \
5837 --mkfsoptions=\\\"-O flex_bg -G $NEWSIZE }
5841 echo "params: $opts"
5843 add ost1 $opts || error "add ost1 failed with new params"
5845 local FOUNDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
5846 awk '/Flex block group size: / { print $NF; exit; }')
5848 [[ $FOUNDSIZE == $NEWSIZE ]] ||
5849 error "Flex block group size: $FOUNDSIZE, expected: $NEWSIZE"
5852 run_test 86 "Replacing mkfs.lustre -G option"
5854 test_87() { #LU-6544
5855 [[ $(lustre_version_code $SINGLEMDS1) -ge $(version_code 2.7.56) ]] ||
5856 { skip "Need MDS version at least 2.7.56" && return; }
5857 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ]] &&
5858 { skip "Only applicable to ldiskfs-based MDTs" && return; }
5859 [[ $OSTCOUNT -gt 69 ]] &&
5860 { skip "Ignore wide striping situation" && return; }
5862 local mdsdev=$(mdsdevname 1)
5863 local mdsvdev=$(mdsvdevname 1)
5864 local file=$DIR/$tfile
5865 local mntpt=$(facet_mntpt $SINGLEMDS)
5866 local used_xattr_blk=0
5867 local inode_size=${1:-512}
5869 local xtest="trusted.test"
5874 #Please see LU-6544 for MDT inode size calculation
5875 if [ $OSTCOUNT -gt 26 ]; then
5877 elif [ $OSTCOUNT -gt 5 ]; then
5880 left_size=$(expr $inode_size - \
5883 32 - $OSTCOUNT \* 24 - 16 - 3 - \
5885 24 - 18 - $(expr length $tfile) - 16 - 4)
5886 if [ $left_size -le 0 ]; then
5887 echo "No space($left_size) is expected in inode."
5888 echo "Try 1-byte xattr instead to verify this."
5891 echo "Estimate: at most $left_size-byte space left in inode."
5897 add mds1 $(mkfs_opts mds1 ${mdsdev}) --stripe-count-hint=$OSTCOUNT \
5898 --reformat $mdsdev $mdsvdev || error "add mds1 failed"
5899 start_mdt 1 > /dev/null || error "start mdt1 failed"
5900 for i in $(seq $OSTCOUNT); do
5901 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS > /dev/null ||
5902 error "start ost$i failed"
5904 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
5905 check_mount || error "check client $MOUNT failed"
5908 $SETSTRIPE -c -1 $file || error "$SETSTRIPE -c -1 $file failed"
5909 $GETSTRIPE $file || error "$GETSTRIPE $file failed"
5910 i=$($GETSTRIPE -c $file)
5911 if [ $i -ne $OSTCOUNT ]; then
5912 left_size=$(expr $left_size + $(expr $OSTCOUNT - $i) \* 24)
5913 echo -n "Since only $i out $OSTCOUNT OSTs are used, "
5914 echo -n "the expected left space is changed to "
5915 echo "$left_size bytes at most."
5917 value=$(generate_string $left_size)
5918 setfattr -n $xtest -v $value $file
5919 orig=$(get_xattr_value $xtest $file)
5920 [[ "$orig" != "$value" ]] && error "$xtest changed"
5922 #Verify if inode has some expected space left
5923 umount $MOUNT > /dev/null || error "umount $MOUNT failed"
5924 stop_mdt 1 > /dev/null || error "stop mdt1 failed"
5925 mount_ldiskfs $SINGLEMDS || error "mount -t ldiskfs $SINGLEMDS failed"
5927 do_facet $SINGLEMDS ls -sal $mntpt/ROOT/$tfile
5928 used_xattr_blk=$(do_facet $SINGLEMDS ls -s $mntpt/ROOT/$tfile |
5930 [[ $used_xattr_blk -eq 0 ]] &&
5931 error "Please check MDS inode size calculation: \
5932 more than $left_size-byte space left in inode."
5933 echo "Verified: at most $left_size-byte space left in inode."
5937 run_test 87 "check if MDT inode can hold EAs with N stripes properly"
5940 [ "$(facet_fstype mds1)" == "zfs" ] &&
5941 skip "LU-6662: no implementation for ZFS" && return
5945 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
5946 --reformat $(mdsdevname 1) || error "add mds1 failed"
5948 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
5949 grep -e \".*opts:.*errors=remount-ro.*\"" ||
5950 error "default mount options is missing"
5952 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
5953 --mountfsoptions="user_xattr,errors=panic" \
5954 --reformat $(mdsdevname 1) || error "add mds1 failed"
5956 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
5957 grep -e \".*opts:.*errors=panic.*\"" ||
5958 error "user can't override default mount options"
5960 run_test 88 "check the default mount options can be overridden"
5963 # $2 (optional) value of max_mod_rpcs_in_flight to set
5964 check_max_mod_rpcs_in_flight() {
5972 idx=$(printf "%04x" $($LFS getdirstripe -i $dir))
5973 facet="mds$((0x$idx + 1))"
5975 if [ -z "$mmr" ]; then
5976 # get value of max_mod_rcps_in_flight
5977 mmr=$($LCTL get_param -n \
5978 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) ||
5979 error "Unable to get max_mod_rpcs_in_flight"
5980 echo "max_mod_rcps_in_flight is $mmr"
5982 # set value of max_mod_rpcs_in_flight
5984 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight=$mmr ||
5985 error "Unable to set max_mod_rpcs_in_flight to $mmr"
5986 echo "max_mod_rpcs_in_flight set to $mmr"
5989 # create mmr+1 files
5990 echo "creating $((mmr + 1)) files ..."
5992 for i in $(seq $((mmr + 1))); do
5998 # consumes mmr-1 modify RPC slots
5999 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6000 # drop requests on MDT so that RPC slots are consumed
6001 # during all the request resend interval
6002 do_facet $facet "$LCTL set_param fail_loc=0x159"
6003 echo "launch $((mmr - 1)) chmod in parallel ..."
6004 for i in $(seq $((mmr - 1))); do
6005 chmod 0600 $dir/file-$i &
6009 # send one additional modify RPC
6010 do_facet $facet "$LCTL set_param fail_loc=0"
6011 echo "launch 1 additional chmod in parallel ..."
6012 chmod 0600 $dir/file-$mmr &
6015 # check this additional modify RPC get a modify RPC slot
6016 # and succeed its operation
6017 checkstat -vp 0600 $dir/file-$mmr ||
6018 error "Unable to send $mmr modify RPCs in parallel"
6023 # consumes mmr modify RPC slots
6024 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6025 # drop requests on MDT so that RPC slots are consumed
6026 # during all the request resend interval
6027 do_facet $facet "$LCTL set_param fail_loc=0x159"
6028 echo "launch $mmr chmod in parallel ..."
6029 for i in $(seq $mmr); do
6030 chmod 0666 $dir/file-$i &
6034 # send one additional modify RPC
6035 do_facet $facet "$LCTL set_param fail_loc=0"
6036 echo "launch 1 additional chmod in parallel ..."
6037 chmod 0666 $dir/file-$((mmr + 1)) &
6040 # check this additional modify RPC blocked getting a modify RPC slot
6041 checkstat -vp 0644 $dir/file-$((mmr + 1)) ||
6042 error "Unexpectedly send $(($mmr + 1)) modify RPCs in parallel"
6048 if ! combined_mgs_mds ; then
6053 [[ $($LCTL get_param mdc.*.import |
6054 grep "connect_flags:.*multi_mod_rpc") ]] ||
6055 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6057 # check default value
6058 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6059 check_max_mod_rpcs_in_flight $DIR/$tdir
6063 run_test 90a "check max_mod_rpcs_in_flight is enforced"
6073 [[ $($LCTL get_param mdc.*.import |
6074 grep "connect_flags:.*multi_mod_rpc") ]] ||
6075 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6078 # update max_mod_rpcs_in_flight
6079 $LFS mkdir -c1 $DIR/${tdir}1 || error "mkdir $DIR/${tdir}1 failed"
6080 check_max_mod_rpcs_in_flight $DIR/${tdir}1 1
6083 # check client is able to send multiple modify RPCs in paralell
6084 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
6085 grep -c "multi_mod_rpcs")
6086 if [ "$tmp" -ne $MDSCOUNT ]; then
6087 echo "Client not able to send multiple modify RPCs in parallel"
6092 # update max_mod_rpcs_in_flight
6093 $LFS mkdir -c1 $DIR/${tdir}2 || error "mkdir $DIR/${tdir}2 failed"
6094 check_max_mod_rpcs_in_flight $DIR/${tdir}2 5
6097 $LFS mkdir -c1 $DIR/${tdir}3 || error "mkdir $DIR/${tdir}3 failed"
6098 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/${tdir}3))
6099 facet="mds$((0x$idx + 1))"
6101 # save MDT max_mod_rpcs_per_client
6102 mmrpc=$(do_facet $facet \
6103 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
6105 # update max_mod_rpcs_in_flight
6106 umount_client $MOUNT
6108 "echo 16 > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6110 $LCTL set_param mdc.$FSNAME-MDT$idx-mdc-*.max_rpcs_in_flight=17
6111 check_max_mod_rpcs_in_flight $DIR/${tdir}3 16
6113 # restore MDT max_mod_rpcs_per_client initial value
6115 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6117 rm -rf $DIR/${tdir}?
6120 run_test 90b "check max_mod_rpcs_in_flight is enforced after update"
6129 [[ $($LCTL get_param mdc.*.import |
6130 grep "connect_flags:.*multi_mod_rpc") ]] ||
6131 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6133 # check client is able to send multiple modify RPCs in paralell
6134 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
6135 grep -c "multi_mod_rpcs")
6136 if [ "$tmp" -ne $MDSCOUNT ]; then
6137 skip "Client not able to send multiple modify RPCs in parallel"
6142 # get max_rpcs_in_flight value
6143 mrif=$($LCTL get_param -n mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
6144 echo "max_rpcs_in_flight is $mrif"
6146 # get MDT max_mod_rpcs_per_client
6147 mmrpc=$(do_facet mds1 \
6148 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
6149 echo "max_mod_rpcs_per_client is $mmrpc"
6152 # attempt to set max_mod_rpcs_in_flight to max_rpcs_in_flight value
6153 # prerequisite: set max_mod_rpcs_per_client to max_rpcs_in_flight value
6154 umount_client $MOUNT
6156 "echo $mrif > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6160 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif &&
6161 error "set max_mod_rpcs_in_flight to $mrif should fail"
6163 umount_client $MOUNT
6165 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6169 # attempt to set max_mod_rpcs_in_flight to max_mod_rpcs_per_client+1
6170 # prerequisite: set max_rpcs_in_flight to max_mod_rpcs_per_client+2
6172 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$((mmrpc + 2))
6175 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$((mmrpc + 1)) &&
6176 error "set max_mod_rpcs_in_flight to $((mmrpc + 1)) should fail"
6180 run_test 90c "check max_mod_rpcs_in_flight update limits"
6191 [[ $($LCTL get_param mdc.*.import |
6192 grep "connect_flags:.*multi_mod_rpc") ]] ||
6193 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6195 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6196 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
6197 facet="mds$((0x$idx + 1))"
6199 # check client version supports multislots
6200 tmp=$($LCTL get_param -N \
6201 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
6202 if [ -z "$tmp" ]; then
6203 skip "Client does not support multiple modify RPCs in flight"
6208 # get current value of max_mod_rcps_in_flight
6209 mmr=$($LCTL get_param -n \
6210 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
6211 echo "max_mod_rcps_in_flight is $mmr"
6214 echo "creating $mmr files ..."
6216 for i in $(seq $mmr); do
6217 touch $DIR/$tdir/file-$i
6220 # prepare for close RPC
6221 multiop_bg_pause $DIR/$tdir/file-close O_c
6224 # consumes mmr modify RPC slots
6225 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6226 # drop requests on MDT so that RPC slots are consumed
6227 # during all the request resend interval
6228 do_facet $facet "$LCTL set_param fail_loc=0x159"
6229 echo "launch $mmr chmod in parallel ..."
6230 for i in $(seq $mmr); do
6231 chmod 0600 $DIR/$tdir/file-$i &
6234 # send one additional close RPC
6235 do_facet $facet "$LCTL set_param fail_loc=0"
6236 echo "launch 1 additional close in parallel ..."
6238 cancel_lru_locks mdc
6241 # check this additional close RPC get a modify RPC slot
6242 # and multiop process completed
6243 [ -d /proc/$pid ] &&
6244 error "Unable to send the additional close RPC in parallel"
6249 run_test 90d "check one close RPC is allowed above max_mod_rpcs_in_flight"
6251 check_uuid_on_ost() {
6253 do_facet ost1 "$LCTL get_param obdfilter.${FSNAME}*.exports.'$nid'.uuid"
6256 check_uuid_on_mdt() {
6258 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
6266 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.63) ]] ||
6267 { skip "Need OST version at least 2.7.63" && return 0; }
6268 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.63) ]] ||
6269 { skip "Need MDT version at least 2.7.63" && return 0; }
6271 start_mds || error "MDS start failed"
6272 start_ost || error "unable to start OST"
6273 mount_client $MOUNT || error "client start failed"
6274 check_mount || error "check_mount failed"
6277 nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g")
6281 uuid=$(get_client_uuid $MOUNT)
6283 echo "list nids on mdt:"
6284 do_facet $SINGLEMDS "$LCTL list_param mdt.${FSNAME}*.exports.*"
6285 echo "uuid from $nid:"
6286 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
6288 found=$(check_uuid_on_mdt $nid | grep $uuid)
6289 [ -z "$found" ] && error "can't find $uuid $nid on MDT"
6290 found=$(check_uuid_on_ost $nid | grep $uuid)
6291 [ -z "$found" ] && error "can't find $uuid $nid on OST"
6293 # umount the client so it won't reconnect
6294 manual_umount_client --force || error "failed to umount $?"
6295 # shouldn't disappear on MDS after forced umount
6296 found=$(check_uuid_on_mdt $nid | grep $uuid)
6297 [ -z "$found" ] && error "can't find $uuid $nid"
6300 do_facet $SINGLEMDS \
6301 "$LCTL set_param -n mdt.${mds1_svc}.evict_client nid:$nid"
6303 found=$(check_uuid_on_mdt $nid | grep $uuid)
6304 [ -n "$found" ] && error "found $uuid $nid on MDT"
6305 found=$(check_uuid_on_ost $nid | grep $uuid)
6306 [ -n "$found" ] && error "found $uuid $nid on OST"
6308 # check it didn't reconnect (being umounted)
6309 sleep $((TIMEOUT+1))
6310 found=$(check_uuid_on_mdt $nid | grep $uuid)
6311 [ -n "$found" ] && error "found $uuid $nid on MDT"
6312 found=$(check_uuid_on_ost $nid | grep $uuid)
6313 [ -n "$found" ] && error "found $uuid $nid on OST"
6317 run_test 91 "evict-by-nid support"
6319 generate_ldev_conf() {
6320 # generate an ldev.conf file
6321 local ldevconfpath=$1
6323 printf "%s\t-\t%s-MGS0000\t%s\n" \
6326 $(mgsdevname) >> $ldevconfpath
6328 local mdsfo_host=$mdsfailover_HOST;
6329 if [ -z "$mdsfo_host" ]; then
6333 for num in $(seq $MDSCOUNT); do
6334 printf "%s\t%s\t%s-MDT%04d\t%s\n" \
6339 $(mdsdevname $num) >> $ldevconfpath
6342 local ostfo_host=$ostfailover_HOST;
6343 if [ -z "$ostfo_host" ]; then
6347 for num in $(seq $OSTCOUNT); do
6348 printf "%s\t%s\t%s-OST%04d\t%s\n" \
6353 $(ostdevname $num) >> $ldevconfpath
6358 # generate a nids file (mapping between hostname to nid)
6359 # looks like we only have the MGS nid available to us
6360 # so just echo that to a file
6363 echo -e "${mgs_HOST}\t${MGSNID}" >> $nidspath
6367 local LDEVCONFPATH=$TMP/ldev.conf
6368 local NIDSPATH=$TMP/nids
6370 echo "Host is $(hostname)"
6372 generate_ldev_conf $LDEVCONFPATH
6373 generate_nids $NIDSPATH
6375 echo "----- ldev.conf -----"
6377 echo "--- END ldev.conf ---"
6379 echo "----- /etc/nids -----"
6381 echo "--- END /etc/nids ---"
6383 # ldev can be in our build tree and if we aren't in a
6384 # build tree, use 'which' to try and find it
6385 local LDEV=$LUSTRE/scripts/ldev
6386 [ ! -f "$LDEV" ] && local LDEV=$(which ldev 2> /dev/null)
6388 echo "ldev path is $LDEV"
6390 if [ ! -f "$LDEV" ]; then
6391 rm $LDEVCONFPATH $NIDSPATH
6392 error "failed to find ldev!"
6395 # echo the mgs nid and compare it to environment variable MGSNID
6396 # also, ldev.conf and nids is a server side thing, use the OSS
6399 output=$(perl $LDEV -c $LDEVCONFPATH -H \
6400 $ost_HOST -n $NIDSPATH echo %m)
6402 echo "-- START OF LDEV OUTPUT --"
6404 echo "--- END OF LDEV OUTPUT ---"
6406 # ldev failed, error
6407 if [ $? -ne 0 ]; then
6408 rm $LDEVCONFPATH $NIDSPATH
6409 error "ldev failed to execute!"
6412 # need to process multiple lines because of combined MGS and MDS
6413 echo -e $output | awk '{ print $2 }' | while read -r line ; do
6414 if [ "$line" != "$MGSNID" ]; then
6415 rm $LDEVCONFPATH $NIDSPATH
6416 error "ldev failed mgs nid '$line', expected '$MGSNID'"
6420 rm $LDEVCONFPATH $NIDSPATH
6422 run_test 92 "ldev returns MGS NID correctly in command substitution"
6424 if ! combined_mgs_mds ; then
6430 # restore the values of MDSSIZE and OSTSIZE
6431 MDSSIZE=$STORED_MDSSIZE
6432 OSTSIZE=$STORED_OSTSIZE