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/$NAME.sh}
54 # use small MDS + OST size to speed formatting time
55 # do not use too small MDSSIZE/OSTSIZE, which affect the default jouranl size
56 # STORED_MDSSIZE is used in test_18
57 STORED_MDSSIZE=$MDSSIZE
58 STORED_OSTSIZE=$OSTSIZE
70 if ! combined_mgs_mds; then
71 # bug number for skipped test: 23954
72 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 24b"
75 # pass "-E lazy_itable_init" to mke2fs to speed up the formatting time
76 if [[ "$LDISKFS_MKFS_OPTS" != *lazy_itable_init* ]]; then
77 LDISKFS_MKFS_OPTS=$(csa_add "$LDISKFS_MKFS_OPTS" -E lazy_itable_init)
80 [ $(facet_fstype $SINGLEMDS) = "zfs" ] &&
81 # bug number for skipped test: LU-4444
82 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 69"
87 require_dsh_mds || exit 0
88 require_dsh_ost || exit 0
90 [ "$SLOW" = "no" ] && EXCEPT_SLOW="30a 31 45 69"
95 # The MGS must be started before the OSTs for a new fs, so start
96 # and stop to generate the startup logs.
99 wait_osc_import_state mds ost FULL
104 reformat_and_config() {
106 if ! combined_mgs_mds ; then
112 writeconf_or_reformat() {
113 # There are at most 2 OSTs for write_conf test
114 # who knows if/where $TUNEFS is installed?
115 # Better reformat if it fails...
116 writeconf_all $MDSCOUNT 2 ||
117 { echo "tunefs failed, reformatting instead" &&
118 reformat_and_config && return 0; }
128 start mgs $(mgsdevname) $MGS_MOUNT_OPTS
134 local dev=$(mdsdevname $num)
137 echo "start mds service on `facet_active_host $facet`"
138 start $facet ${dev} $MDS_MOUNT_OPTS $@ || return 94
144 local dev=$(mdsdevname $num)
147 echo "stop mds service on `facet_active_host $facet`"
148 # These tests all use non-failover stop
149 stop $facet -f || return 97
155 for num in $(seq $MDSCOUNT); do
156 start_mdt $num $@ || return 94
161 if ! combined_mgs_mds ; then
169 for num in $(seq $MDSCOUNT); do
170 stop_mdt $num || return 97
175 echo "stop mgs service on `facet_active_host mgs`"
176 # These tests all use non-failover stop
177 stop mgs -f || return 97
181 echo "start ost1 service on `facet_active_host ost1`"
182 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS $@ || return 95
186 echo "stop ost1 service on `facet_active_host ost1`"
187 # These tests all use non-failover stop
188 stop ost1 -f || return 98
192 echo "start ost2 service on `facet_active_host ost2`"
193 start ost2 $(ostdevname 2) $OST_MOUNT_OPTS $@ || return 92
197 echo "stop ost2 service on `facet_active_host ost2`"
198 # These tests all use non-failover stop
199 stop ost2 -f || return 93
204 echo "mount $FSNAME on ${MOUNTPATH}....."
205 zconf_mount $(hostname) $MOUNTPATH || return 96
209 local mountopt="remount,$1"
211 echo "remount '$1' lustre on ${MOUNTPATH}....."
212 zconf_mount $(hostname) $MOUNTPATH "$mountopt" || return 96
218 echo "umount lustre on $mountpath....."
219 zconf_umount $HOSTNAME $mountpath $@ || return 97
222 manual_umount_client(){
225 echo "manual umount lustre on ${MOUNT}...."
226 do_facet client "umount ${FORCE} $MOUNT"
232 start_mds || error "MDT start failed"
233 start_ost || error "Unable to start OST1"
234 mount_client $MOUNT || error "client start failed"
235 client_up || error "client_up failed"
244 unload_modules_conf () {
245 if combined_mgs_mds || ! local_mode; then
246 unload_modules || return 1
251 stop_ost || return 202
252 stop_mds || return 201
253 unload_modules_conf || return 203
257 umount_client $MOUNT || return 200
258 cleanup_nocli || return $?
263 echo "umount $MOUNT2 ..."
264 umount $MOUNT2 || true
265 echo "stopping fs2mds ..."
266 stop fs2mds -f || true
267 echo "stopping fs2ost ..."
268 stop fs2ost -f || true
272 do_facet client "cp /etc/passwd $DIR/a" || return 71
273 do_facet client "rm $DIR/a" || return 72
274 # make sure lustre is actually mounted (touch will block,
275 # but grep won't, so do it after)
276 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
277 echo "setup single mount lustre success"
281 do_facet client "touch $DIR/a" || return 71
282 do_facet client "rm $DIR/a" || return 72
283 do_facet client "touch $DIR2/a" || return 73
284 do_facet client "rm $DIR2/a" || return 74
285 echo "setup double mount lustre success"
290 if [ "$ONLY" == "setup" ]; then
295 if [ "$ONLY" == "cleanup" ]; then
302 #create single point mountpoint
308 check_mount || error "check_mount failed"
309 cleanup || error "cleanup failed with $?"
311 run_test 0 "single mount setup"
314 start_mds || error "MDS start failed"
315 start_ost || error "unable to start OST"
316 echo "start ost second time..."
317 start_ost && error "2nd OST start should fail"
318 mount_client $MOUNT || error "client start failed"
319 check_mount || error "check_mount failed"
320 cleanup || error "cleanup failed with $?"
322 run_test 1 "start up ost twice (should return errors)"
325 start_mds || error "MDT start failed"
326 echo "start mds second time.."
327 start_mds && error "2nd MDT start should fail"
328 start_ost || error "OST start failed"
329 mount_client $MOUNT || error "mount_client failed to start client"
330 check_mount || error "check_mount failed"
331 cleanup || error "cleanup failed with $?"
333 run_test 2 "start up mds twice (should return err)"
337 #mount.lustre returns an error if already in mtab
338 mount_client $MOUNT && error "2nd client mount should fail"
339 check_mount || error "check_mount failed"
340 cleanup || error "cleanup failed with $?"
342 run_test 3 "mount client twice (should return err)"
346 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
347 stop_ost || error "Unable to stop OST1"
348 umount_client $MOUNT -f || error “unmount $MOUNT failed”
351 # ok for ost to fail shutdown
352 if [ 202 -ne $eno ] && [ 0 -ne $eno ]; then
353 error "cleanup failed with $?"
356 run_test 4 "force cleanup ost, then cleanup"
358 test_5a() { # was test_5
360 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
361 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
363 stop_mds || error "Unable to stop MDS"
365 # cleanup may return an error from the failed
366 # disconnects; for now I'll consider this successful
367 # if all the modules have unloaded.
371 echo "killing umount"
372 kill -TERM $UMOUNT_PID
373 echo "waiting for umount to finish"
375 if grep " $MOUNT " /proc/mounts; then
376 echo "test 5: /proc/mounts after failed umount"
380 echo "killing umount"
381 kill -TERM $UMOUNT_PID
382 echo "waiting for umount to finish"
384 grep " $MOUNT " /proc/mounts &&
385 error "/proc/mounts after second umount"
388 # manual_mount_client may fail due to umount succeeding above
390 # stop_mds is a no-op here, and should not fail
391 cleanup_nocli || error "cleanup_nocli failed with $?"
392 # df may have lingering entry
394 # mtab may have lingering entry
398 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
400 grep -q $MOUNT" " /etc/mtab || break
401 echo "Waiting /etc/mtab updated ... "
402 WAIT=$(( WAIT + sleep))
404 [ "$WAIT" -eq "$MAX_WAIT" ] &&
405 error "/etc/mtab is not updated in $WAIT secs"
406 echo "/etc/mtab updated in $WAIT secs"
408 run_test 5a "force cleanup mds, then cleanup"
416 grep " $MOUNT " /etc/mtab &&
417 error false "unexpected entry in mtab before mount" && return 10
419 start_ost || error "OST start failed"
420 if ! combined_mgs_mds ; then
421 trap cleanup_5b EXIT ERR
422 start_mds || error "MDS start failed"
426 mount_client $MOUNT && error "mount_client $MOUNT should fail"
427 grep " $MOUNT " /etc/mtab &&
428 error "$MOUNT entry in mtab after failed mount"
430 # stop_mds is a no-op here, and should not fail
431 cleanup_nocli || error "cleanup_nocli failed with $?"
432 if ! combined_mgs_mds ; then
436 run_test 5b "Try to start a client with no MGS (should return errs)"
439 grep " $MOUNT " /etc/mtab &&
440 error false "unexpected entry in mtab before mount" && return 10
442 start_mds || error "MDS start failed"
443 start_ost || error "OST start failed"
444 local oldfs="${FSNAME}"
445 FSNAME="wrong.${FSNAME}"
446 mount_client $MOUNT || :
448 grep " $MOUNT " /etc/mtab &&
449 error "$MOUNT entry in mtab after failed mount"
451 cleanup_nocli || error "cleanup_nocli failed with $?"
453 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
456 grep " $MOUNT " /etc/mtab &&
457 error false "unexpected entry in mtab before mount" && return 10
459 start_ost || error "OST start failed"
460 start_mds || error "MDS start failed"
461 stop_ost || error "Unable to stop OST1"
462 mount_client $MOUNT || error "mount_client $MOUNT failed"
463 umount_client $MOUNT -f || error "umount_client $MOUNT failed"
464 cleanup_nocli || error "cleanup_nocli failed with $?"
465 grep " $MOUNT " /etc/mtab &&
466 error "$MOUNT entry in mtab after unmount"
469 run_test 5d "mount with ost down"
472 grep " $MOUNT " /etc/mtab &&
473 error false "unexpected entry in mtab before mount" && return 10
475 start_mds || error "MDS start failed"
476 start_ost || error "OST start failed"
478 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
479 do_facet client "$LCTL set_param fail_loc=0x80000506"
480 mount_client $MOUNT || echo "mount failed (not fatal)"
481 cleanup || error "cleanup failed with $?"
482 grep " $MOUNT " /etc/mtab &&
483 error "$MOUNT entry in mtab after unmount"
486 run_test 5e "delayed connect, don't crash (bug 10268)"
489 if combined_mgs_mds ; then
490 skip "combined mgs and mds"
494 grep " $MOUNT " /etc/mtab &&
495 error false "unexpected entry in mtab before mount" && return 10
498 start_ost || error "OST start failed"
499 mount_client $MOUNT &
501 echo client_mount pid is $pid
505 if ! ps -f -p $pid >/dev/null; then
508 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
509 error "mount returns $rc, expected to hang"
511 cleanup || error "cleanup failed with $?"
516 start_mds || error "start MDS failed"
518 # mount should succeed after start mds
520 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
521 cleanup || error "final call to cleanup failed with rc $?"
523 run_test 5f "mds down, cleanup after failed mount (bug 2712)"
528 mount_client $MOUNT || error "mount_client $MOUNT failed"
529 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
530 cleanup || error "cleanup failed with rc $?"
532 run_test 6 "manual umount, then mount again"
537 cleanup_nocli || error "cleanup_nocli failed with $?"
539 run_test 7 "manual umount, then cleanup"
543 mount_client $MOUNT2 || error "mount_client $MOUNT2 failed"
544 check_mount2 || error "check_mount2 failed"
545 umount_client $MOUNT2 || error "umount_client $MOUNT2 failed"
546 cleanup || error "cleanup failed with rc $?"
548 run_test 8 "double mount setup"
551 start_ost || error "OST start failed"
553 do_facet ost1 $LCTL set_param debug=\'inode trace\' ||
554 error "do_facet ost1 set_param inode trace failed."
555 do_facet ost1 $LCTL set_param subsystem_debug=\'mds ost\' ||
556 error "do_facet ost1 set_param debug mds ost failed."
558 CHECK_PTLDEBUG="`do_facet ost1 $LCTL get_param -n debug`"
559 if [ "$CHECK_PTLDEBUG" ] && { \
560 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
561 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
562 echo "lnet.debug success"
564 error "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
566 CHECK_SUBSYS="`do_facet ost1 $LCTL get_param -n subsystem_debug`"
567 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
568 echo "lnet.subsystem_debug success"
570 error "lnet.subsystem_debug: want 'mds ost' got '$CHECK_SUBSYS'"
572 stop_ost || error "Unable to stop OST1"
574 run_test 9 "test ptldebug and subsystem for mkfs"
582 do_facet $facet "test -b $dev" || rc=1
583 if [[ "$size" ]]; then
584 local in=$(do_facet $facet "dd if=$dev of=/dev/null bs=1k \
585 count=1 skip=$size 2>&1" |
586 awk '($3 == "in") { print $1 }')
587 [[ $in = "1+0" ]] || rc=1
593 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
594 # But with new MDS stack we don't care about the mode of local objects
595 # anymore, so this test is removed. See bug 22944 for more details.
599 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
600 skip "Only applicable to ldiskfs-based MDTs"
605 check_mount || error "check_mount failed"
606 cleanup || error "cleanup failed with rc $?"
608 echo "Remove mds config log"
609 if ! combined_mgs_mds ; then
613 do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' \
614 $(mgsdevname) || return \$?" ||
615 error "do_facet mgs failed with $?"
617 if ! combined_mgs_mds ; then
621 start_ost || error "OST start failed"
622 start_mds && error "MDS start succeeded, but should fail"
625 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
628 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
629 skip "Only applicable to ldiskfs-based MDTs"
633 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
638 # check if current MDSSIZE is large enough
639 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE &&
640 log "use MDSSIZE=$MDSSIZE"
642 # check if the global config has a large enough MDSSIZE
643 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] &&
644 [ $STORED_MDSSIZE -ge $MIN ] &&
645 OK=1 && myMDSSIZE=$STORED_MDSSIZE &&
646 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
648 # check if the block device is large enough
649 is_blkdev $SINGLEMDS $MDSDEV $MIN
650 local large_enough=$?
651 if [ -n "$OK" ]; then
652 [ $large_enough -ne 0 ] && OK=""
654 [ $large_enough -eq 0 ] && OK=1 && myMDSSIZE=$MIN &&
655 log "use device $MDSDEV with MIN=$MIN"
658 # check if a loopback device has enough space for fs metadata (5%)
660 if [ -z "$OK" ]; then
661 local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! \
662 -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
663 awk '($1 != "Filesystem") { print $4 }')
664 ! [ -z "$SPACE" ] && [ $SPACE -gt $((MIN / 20)) ] &&
665 OK=1 && myMDSSIZE=$MIN &&
666 log "use file $MDSDEV with MIN=$MIN"
669 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS" && return
671 echo "mount mds with large journal..."
673 local OLD_MDSSIZE=$MDSSIZE
677 echo "mount lustre system..."
679 check_mount || error "check_mount failed"
681 echo "check journal size..."
682 local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
683 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
684 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
686 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
689 cleanup || error "cleanup failed with rc $?"
694 run_test 18 "check mkfs creates large journals"
697 start_mds || error "MDS start failed"
698 stop_mds || error "Unable to stop MDS"
700 run_test 19a "start/stop MDS without OSTs"
703 start_ost || error "Unable to start OST1"
704 stop_ost || error "Unable to stop OST1"
706 run_test 19b "start/stop OSTs without MDS"
709 # first format the ost/mdt
710 start_mds || error "MDS start failed"
711 start_ost || error "Unable to start OST1"
712 mount_client $MOUNT || error "mount_client $MOUNT failed"
713 check_mount || error "check_mount failed"
714 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed."
715 remount_client ro $MOUNT || error "remount_client with ro failed"
716 touch $DIR/$tfile && error "$DIR/$tfile created incorrectly"
717 [ -e $DIR/$tfile ] && error "$DIR/$tfile exists incorrectly"
718 remount_client rw $MOUNT || error "remount_client with rw failed"
719 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
720 MCNT=$(grep -c $MOUNT /etc/mtab)
721 [ "$MCNT" -ne 1 ] && error "$MOUNT in /etc/mtab $MCNT times"
723 stop_mds || error "Unable to stop MDS"
724 stop_ost || error "Unable to stop OST1"
726 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
729 start_mds || error "MDS start failed"
730 start_ost || error "unable to start OST1"
731 wait_osc_import_state mds ost FULL
732 stop_ost || error "unable to stop OST1"
733 stop_mds || error "unable to stop MDS"
735 run_test 21a "start mds before ost, stop ost first"
738 start_ost || error "unable to start OST1"
739 start_mds || error "MDS start failed"
740 wait_osc_import_state mds ost FULL
741 stop_mds || error "unable to stop MDS"
742 stop_ost || error "unable to stop OST1"
744 run_test 21b "start ost before mds, stop mds first"
747 start_ost || error "Unable to start OST1"
748 start_mds || error "MDS start failed"
749 start_ost2 || error "Unable to start OST2"
750 wait_osc_import_state mds ost2 FULL
751 stop_ost || error "Unable to stop OST1"
752 stop_ost2 || error "Unable to stop OST2"
753 stop_mds || error "Unable to stop MDS"
754 #writeconf to remove all ost2 traces for subsequent tests
755 writeconf_or_reformat
757 run_test 21c "start mds between two osts, stop mds last"
760 if combined_mgs_mds ; then
761 skip "need separate mgs device" && return 0
767 start_mgs || error "unable to start MGS"
768 start_ost || error "unable to start OST1"
769 start_ost2 || error "unable to start OST2"
770 start_mds || error "MDS start failed"
771 wait_osc_import_state mds ost2 FULL
773 stop_ost || error "Unable to stop OST1"
774 stop_ost2 || error "Unable to stop OST2"
775 stop_mds || error "Unable to stop MDS"
777 #writeconf to remove all ost2 traces for subsequent tests
778 writeconf_or_reformat
779 start_mgs || error "unable to start MGS"
781 run_test 21d "start mgs then ost and then mds"
784 MGSNID="$saved_mgsnid"
786 echo "stopping fs2mgs ..."
787 stop $fs2mgs -f || true
790 test_21e() { # LU-5863
791 if [[ -z "$fs3ost_DEV" || -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
792 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
793 skip_env "mixed loopback and real device not working" && return
796 local fs2mdsdev=$(mdsdevname 1_2)
797 local fs2ostdev=$(ostdevname 1_2)
798 local fs3ostdev=$(ostdevname 2_2)
800 local fs2mdsvdev=$(mdsvdevname 1_2)
801 local fs2ostvdev=$(ostvdevname 1_2)
802 local fs3ostvdev=$(ostvdevname 2_2)
804 # temporarily use fs3ost as fs2mgs
806 local fs2mgsdev=$fs3ostdev
807 local fs2mgsvdev=$fs3ostvdev
809 local fsname=test1234
811 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
812 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
813 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS && trap cleanup_21e EXIT INT ||
814 error "start fs2mgs failed"
816 local saved_mgsnid="$MGSNID"
817 MGSNID=$(do_facet $fs2mgs $LCTL list_nids | xargs | tr ' ' ,)
819 add fs2mds $(mkfs_opts mds1 $fs2mdsdev $fsname) \
820 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
821 add fs2ost $(mkfs_opts ost1 $fs2ostdev $fsname) \
822 --reformat $fs2ostdev $fs2ostvdev || error "add fs2ost failed"
824 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
825 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS || error "start fs2mds failed"
827 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
828 $MOUNT_CMD $MGSNID:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
829 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
833 run_test 21e "separate MGS and MDS"
836 start_mds || error "MDS start failed"
838 echo "Client mount with ost in logs, but none running"
839 start_ost || error "unable to start OST1"
840 # wait until mds connected to ost and open client connection
841 wait_osc_import_state mds ost FULL
842 stop_ost || error "unable to stop OST1"
843 mount_client $MOUNT || error "mount_client $MOUNT failed"
844 # check_mount will block trying to contact ost
845 mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed"
846 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed"
847 umount_client $MOUNT -f
850 echo "Client mount with a running ost"
851 start_ost || error "unable to start OST1"
853 # if gss enabled, wait full time to let connection from
854 # mds to ost be established, due to the mismatch between
855 # initial connect timeout and gss context negotiation timeout.
856 # This perhaps could be remove after AT landed.
857 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
858 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
860 mount_client $MOUNT || error "mount_client $MOUNT failed"
861 wait_osc_import_state mds ost FULL
862 wait_osc_import_state client ost FULL
863 check_mount || error "check_mount failed"
866 cleanup || error "cleanup failed with rc $?"
868 run_test 22 "start a client before osts (should return errs)"
870 test_23a() { # was test_23
873 stop $SINGLEMDS || error "failed to stop $SINGLEMDS"
874 # force down client so that recovering mds waits for reconnect
875 local running=$(grep -c $MOUNT /proc/mounts) || true
876 if [ $running -ne 0 ]; then
877 echo "Stopping client $MOUNT (opts: -f)"
881 # enter recovery on failed mds
882 local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/})
883 start $SINGLEMDS $MDT_DEV $MDS_MOUNT_OPTS || error "MDS start failed"
884 # try to start a new client
885 mount_client $MOUNT &
887 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
888 MOUNT_LUSTRE_PID=$(ps -ef | grep mount.lustre |
889 grep -v grep | awk '{print $2}')
890 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
892 ps --ppid $MOUNT_LUSTRE_PID
893 echo "waiting for mount to finish"
895 # "ctrl-c" sends SIGINT but it usually (in script) does not work on child process
896 # SIGTERM works but it does not spread to offspring processses
897 kill -s TERM $MOUNT_PID
898 kill -s TERM $MOUNT_LUSTRE_PID
899 # we can not wait $MOUNT_PID because it is not a child of this shell
905 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
907 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
908 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
911 [ -z "$PID1" -a -z "$PID2" ] && break
912 echo "waiting for mount to finish ... "
913 WAIT=$(( WAIT + sleep))
915 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
916 error "MOUNT_PID $MOUNT_PID and "\
917 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
920 stop_mds || error "stopping MDSes failed"
921 stop_ost || error "stopping OSSes failed"
923 run_test 23a "interrupt client during recovery mount delay"
928 test_23b() { # was test_23
929 start_mds || error "MDS start failed"
930 start_ost || error "Unable to start OST1"
931 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
932 $LCTL set_param fail_loc=0x80000313
934 cleanup || error "cleanup failed with rc $?"
936 run_test 23b "Simulate -EINTR during mount"
939 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
941 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
942 is_blkdev $SINGLEMDS $MDSDEV &&
943 skip_env "mixed loopback and real device not working" && return
946 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
948 local fs2mdsdev=$(mdsdevname 1_2)
949 local fs2ostdev=$(ostdevname 1_2)
950 local fs2mdsvdev=$(mdsvdevname 1_2)
951 local fs2ostvdev=$(ostvdevname 1_2)
953 # test 8-char fsname as well
954 local FSNAME2=test1234
956 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --nomgs --mgsnode=$MGSNID \
957 --fsname=${FSNAME2} --reformat $fs2mdsdev $fs2mdsvdev || exit 10
959 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --fsname=${FSNAME2} \
960 --reformat $fs2ostdev $fs2ostvdev || exit 10
963 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
964 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
965 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
966 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
968 check_mount || error "check_mount failed"
969 # files written on 1 should not show up on 2
970 cp /etc/passwd $DIR/$tfile
972 [ -e $MOUNT2/$tfile ] && error "File bleed"
975 cp /etc/passwd $MOUNT2/$tfile ||
976 error "cp /etc/passwd $MOUNT2/$tfile failed"
977 rm $MOUNT2/$tfile || error "remove $MOUNT2/$tfile failed"
978 # 2 is actually mounted
979 grep $MOUNT2' ' /proc/mounts > /dev/null || error "$MOUNT2 not mounted"
981 facet_failover fs2mds
982 facet_failover fs2ost
985 # the MDS must remain up until last MDT
987 MDS=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
988 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
989 [ -z "$MDS" ] && error "No MDT"
991 cleanup_nocli || error "cleanup_nocli failed with rc $?"
993 run_test 24a "Multiple MDTs on a single node"
996 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
998 if [ -z "$fs2mds_DEV" ]; then
999 local dev=${SINGLEMDS}_dev
1000 local MDSDEV=${!dev}
1001 is_blkdev $SINGLEMDS $MDSDEV &&
1002 skip_env "mixed loopback and real device not working" && return
1005 local fs2mdsdev=$(mdsdevname 1_2)
1006 local fs2mdsvdev=$(mdsvdevname 1_2)
1008 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --mgs --fsname=${FSNAME}2 \
1009 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
1011 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS &&
1012 error "start MDS should fail"
1014 cleanup || error "cleanup failed with rc $?"
1016 run_test 24b "Multiple MGSs on a single node (should return err)"
1020 check_mount || error "check_mount failed"
1021 local MODULES=$($LCTL modules | awk '{ print $2 }')
1022 rmmod $MODULES 2>/dev/null || true
1023 cleanup || error "cleanup failed with $?"
1025 run_test 25 "Verify modules are referenced"
1029 # we need modules before mount for sysctl, so make sure...
1030 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
1031 #define OBD_FAIL_MDS_FS_SETUP 0x135
1032 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000135"
1033 start_mds && error "MDS started but should not have started"
1034 $LCTL get_param -n devices
1035 DEVS=$($LCTL get_param -n devices | egrep -v MG | wc -l)
1036 [ $DEVS -gt 0 ] && error "number of devices is $DEVS, should be zero"
1037 # start mds to drop writeconf setting
1038 start_mds || error "Unable to start MDS"
1039 stop_mds || error "Unable to stop MDS"
1040 unload_modules_conf || error "unload_modules_conf failed with $?"
1042 run_test 26 "MDT startup failure cleans LOV (should return errs)"
1045 start_ost || error "Unable to start OST1"
1046 start_mds || error "Unable to start MDS"
1047 echo "Requeue thread should have started: "
1048 ps -e | grep ll_cfg_requeue
1049 set_conf_param_and_check ost1 \
1050 "$LCTL get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" \
1051 "$FSNAME-OST0000.ost.client_cache_seconds" ||
1052 error "set_conf_param_and_check ost1 failed"
1053 cleanup_nocli || error "cleanup_nocli failed with rc $?"
1055 run_test 27a "Reacquire MGS lock if OST started first"
1060 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
1061 awk '($3 ~ "mdt" && $4 ~ "MDT0000") { print $4 }')
1063 facet_failover $SINGLEMDS
1064 set_conf_param_and_check $SINGLEMDS \
1065 "$LCTL get_param -n mdt.$device.identity_acquire_expire" \
1066 "$device.mdt.identity_acquire_expire" ||
1067 error "set_conf_param_and_check $SINGLEMDS failed"
1068 set_conf_param_and_check client \
1069 "$LCTL get_param -n mdc.$device-mdc-*.max_rpcs_in_flight"\
1070 "$device.mdc.max_rpcs_in_flight" ||
1071 error "set_conf_param_and_check client failed"
1073 cleanup || error "cleanup failed with $?"
1075 run_test 27b "Reacquire MGS lock after failover"
1079 TEST="$LCTL get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1080 PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
1082 FINAL=$(($ORIG + 1))
1083 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL ||
1084 error "first set_conf_param_and_check client failed"
1085 FINAL=$(($FINAL + 1))
1086 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL ||
1087 error "second set_conf_param_and_check client failed"
1088 umount_client $MOUNT || error "umount_client $MOUNT failed"
1089 mount_client $MOUNT || error "mount_client $MOUNT failed"
1091 if [ $RESULT -ne $FINAL ]; then
1092 error "New config not seen: wanted $FINAL got $RESULT"
1094 echo "New config success: got $RESULT"
1096 set_conf_param_and_check client "$TEST" "$PARAM" $ORIG ||
1097 error "third set_conf_param_and_check client failed"
1098 cleanup || error "cleanup failed with rc $?"
1100 run_test 28 "permanent parameter setting"
1102 test_28a() { # LU-4221
1103 [[ $(lustre_version_code ost1) -ge $(version_code 2.5.52) ]] ||
1104 { skip "Need OST version at least 2.5.52" && return 0; }
1105 [ "$(facet_fstype ost1)" = "zfs" ] &&
1106 skip "LU-4221: no such proc params for ZFS OSTs" && return
1113 local device="$FSNAME-OST0000"
1117 # In this test we will set three kinds of proc parameters with
1119 # 1. the ones moved from the OFD to the OSD, and only their
1120 # symlinks kept in obdfilter
1121 # 2. non-symlink ones in the OFD
1122 # 3. non-symlink ones in the OSD
1125 # prepare a symlink parameter in the OFD
1126 name="writethrough_cache_enable"
1127 param="$device.ost.$name"
1128 cmd="$LCTL get_param -n obdfilter.$device.$name"
1130 # conf_param the symlink parameter in the OFD
1131 old=$(do_facet ost1 $cmd)
1132 new=$(((old + 1) % 2))
1133 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1134 error "lctl conf_param $device.ost.$param=$new failed"
1136 # conf_param the target parameter in the OSD
1137 param="$device.osd.$name"
1138 cmd="$LCTL get_param -n osd-*.$device.$name"
1139 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1140 error "lctl conf_param $device.osd.$param=$old failed"
1143 # prepare a non-symlink parameter in the OFD
1144 name="client_cache_seconds"
1145 param="$device.ost.$name"
1146 cmd="$LCTL get_param -n obdfilter.$device.$name"
1148 # conf_param the parameter in the OFD
1149 old=$(do_facet ost1 $cmd)
1151 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1152 error "lctl conf_param $device.ost.$param=$new failed"
1153 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1154 error "lctl conf_param $device.ost.$param=$old failed"
1157 # prepare a non-symlink parameter in the OSD
1159 param="$device.osd.$name"
1160 cmd="$LCTL get_param -n osd-*.$device.$name"
1162 # conf_param the parameter in the OSD
1163 old=$(do_facet ost1 $cmd)
1164 new=$(((old + 1) % 2))
1165 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1166 error "lctl conf_param $device.osd.$param=$new failed"
1167 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1168 error "lctl conf_param $device.osd.$param=$old failed"
1170 cleanup || error "cleanup failed with $?"
1172 run_test 28a "set symlink parameters permanently with conf_param"
1175 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
1176 setup > /dev/null 2>&1
1177 start_ost2 || error "Unable to start OST2"
1180 local PARAM="$FSNAME-OST0001.osc.active"
1181 local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
1182 local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
1184 ACTV=$($LCTL get_param -n $PROC_ACT)
1186 set_conf_param_and_check client \
1187 "$LCTL get_param -n $PROC_ACT" "$PARAM" $DEAC ||
1188 error "set_conf_param_and_check client failed"
1189 # also check ost_server_uuid status
1190 RESULT=$($LCTL get_param -n $PROC_UUID | grep DEACTIV)
1191 if [ -z "$RESULT" ]; then
1192 error "Client not deactivated: $($LCTL get_param \
1195 echo "Live client success: got $RESULT"
1199 for num in $(seq $MDSCOUNT); do
1200 local mdtosc=$(get_mdtosc_proc_path mds${num} $FSNAME-OST0001)
1201 local MPROC="osc.$mdtosc.active"
1206 RESULT=$(do_facet mds${num} "$LCTL get_param -n $MPROC")
1207 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
1208 if [ $RESULT -eq $DEAC ]; then
1209 echo -n "MDT deactivated also after"
1210 echo "$WAIT sec (got $RESULT)"
1214 if [ $WAIT -eq $MAX ]; then
1215 error "MDT active: wanted $DEAC got $RESULT"
1217 echo "Waiting $(($MAX - $WAIT))secs for MDT deactivated"
1220 # test new client starts deactivated
1221 umount_client $MOUNT || error "umount_client $MOUNT failed"
1222 mount_client $MOUNT || error "mount_client $MOUNT failed"
1223 RESULT=$($LCTL get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
1224 if [ -z "$RESULT" ]; then
1225 error "New client start active: $(lctl get_param -n $PROC_UUID)"
1227 echo "New client success: got $RESULT"
1230 # make sure it reactivates
1231 set_conf_param_and_check client \
1232 "$LCTL get_param -n $PROC_ACT" "$PARAM" $ACTV ||
1233 error "lctl get_param $PROC_ACT $PARAM $ACTV failed"
1235 umount_client $MOUNT
1236 stop_ost2 || error "Unable to stop OST2"
1237 cleanup_nocli || error "cleanup_nocli failed with $?"
1238 #writeconf to remove all ost2 traces for subsequent tests
1239 writeconf_or_reformat
1241 run_test 29 "permanently remove an OST"
1246 echo Big config llog
1247 TEST="$LCTL get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1249 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1250 for i in ${LIST[@]}; do
1251 set_conf_param_and_check client "$TEST" \
1252 "$FSNAME.llite.max_read_ahead_whole_mb" $i ||
1253 error "Set $FSNAME.llite.max_read_ahead_whole_mb failed"
1255 # make sure client restart still works
1256 umount_client $MOUNT
1257 mount_client $MOUNT || error "mount_client $MOUNT failed"
1258 [ "$($TEST)" -ne "$i" ] &&
1259 error "Param didn't stick across restart $($TEST) != $i"
1262 echo Erase parameter setting
1263 do_facet mgs "$LCTL conf_param \
1264 -d $FSNAME.llite.max_read_ahead_whole_mb" ||
1265 error "Erase param $FSNAME.llite.max_read_ahead_whole_mb failed"
1266 umount_client $MOUNT
1267 mount_client $MOUNT || error "mount_client $MOUNT failed"
1269 echo "deleted (default) value=$FINAL, orig=$ORIG"
1270 # assumes this parameter started at the default value
1271 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
1273 cleanup || error "cleanup failed with rc $?"
1275 run_test 30a "Big config llog and conf_param deletion"
1280 local orignids=$($LCTL get_param -n \
1281 osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1283 local orignidcount=$(echo "$orignids" | wc -w)
1285 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1286 # numerical part of it. Hopefully that's not already a failover address
1288 local OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | \
1290 local ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1291 local NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1292 local NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1293 echo "Using fake nid $NEW"
1295 local TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1296 grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1297 set_conf_param_and_check client "$TEST" \
1298 "$FSNAME-OST0000.failover.node" $NEW ||
1299 error "didn't add failover nid $NEW"
1300 local NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1303 local NIDCOUNT=$(echo "$NIDS" | wc -w)
1304 echo "should have $((orignidcount + 1)) entries \
1305 in failover nids string, have $NIDCOUNT"
1306 [ $NIDCOUNT -eq $((orignidcount + 1)) ] ||
1307 error "Failover nid not added"
1309 do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" ||
1310 error "conf_param delete failed"
1311 umount_client $MOUNT
1312 mount_client $MOUNT || error "mount_client $MOUNT failed"
1314 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1317 NIDCOUNT=$(echo "$NIDS" | wc -w)
1318 echo "only $orignidcount final entries should remain \
1319 in failover nids string, have $NIDCOUNT"
1320 [ $NIDCOUNT -eq $orignidcount ] || error "Failover nids not removed"
1322 cleanup || error "cleanup failed with rc $?"
1324 run_test 30b "Remove failover nids"
1326 test_31() { # bug 10734
1327 # ipaddr must not exist
1328 $MOUNT_CMD 4.3.2.1@tcp:/lustre $MOUNT || true
1329 cleanup || error "cleanup failed with rc $?"
1331 run_test 31 "Connect to non-existent node (shouldn't crash)"
1335 T32_BLIMIT=20480 # Kbytes
1339 # This is not really a test but a tool to create new disk
1340 # image tarballs for the upgrade tests.
1342 # Disk image tarballs should be created on single-node
1343 # clusters by running this test with default configurations
1344 # plus a few mandatory environment settings that are verified
1345 # at the beginning of the test.
1347 test_32newtarball() {
1351 local tmp=$TMP/t32_image_create
1353 if [ $FSNAME != t32fs -o \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o \
1354 $OSTCOUNT -ne 1 -o -z "$OSTDEV1" ]; then
1355 error "Needs FSNAME=t32fs MDSCOUNT=2 " \
1356 "MDSDEV1=<nonexistent_file>" \
1357 "MDSDEV2=<nonexistent_file>" \
1358 "(or MDSDEV, in the case of b1_8)" \
1359 "OSTCOUNT=1 OSTDEV1=<nonexistent_file>"
1363 echo "Found stale $tmp"
1367 mkdir $tmp/src || return 1
1368 tar cf - -C $src . | tar xf - -C $tmp/src
1369 dd if=/dev/zero of=$tmp/src/t32_qf_old bs=1M \
1370 count=$(($T32_BLIMIT / 1024 / 2))
1371 chown $T32_QID.$T32_QID $tmp/src/t32_qf_old
1373 # format ost with comma-separated NIDs to verify LU-4460
1374 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
1375 MGSNID="$MGSNID,$MGSNID" OSTOPT="--failnode=$failnid" formatall
1379 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ] &&
1380 $LFS quotacheck -ug /mnt/$FSNAME
1381 $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
1384 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1387 mkdir $tmp/img || return 1
1391 ls -Rni --time-style=+%s >$tmp/img/list
1392 find ! -name .lustre -type f -exec sha1sum {} \; |
1393 sort -k 2 >$tmp/img/sha1sums
1395 $LCTL get_param -n version | head -n 1 |
1396 sed -e 's/^lustre: *//' >$tmp/img/commit
1398 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ] &&
1399 $LFS quotaon -ug /mnt/$FSNAME
1400 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1401 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1402 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1403 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1404 | tr -d "*" > $tmp/img/bspace
1405 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1406 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1407 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1408 | tr -d "*" > $tmp/img/ispace
1410 if [ $MDSCOUNT -ge 2 ]; then
1411 $LFS mkdir -i 1 /mnt/$FSNAME/remote_dir
1412 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME/remote_dir
1418 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1421 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1422 echo "Data verification failed"
1425 uname -r >$tmp/img/kernel
1426 uname -m >$tmp/img/arch
1428 mv ${MDSDEV1:-$MDSDEV} $tmp/img
1429 for num in $(seq 2 $MDSCOUNT); do
1430 local devname=$(mdsdevname $num)
1431 mv $devname $tmp/img
1433 mv $OSTDEV1 $tmp/img
1435 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1436 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1439 tar cjvf $dst/disk$version-$(facet_fstype $SINGLEMDS).tar.bz2 -S *
1444 #run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1447 # The list of applicable tarballs is returned via the caller's
1448 # variable "tarballs".
1451 local node=$(facet_active_host $SINGLEMDS)
1452 local r="do_node $node"
1454 if [ "$CLIENTONLY" ]; then
1455 skip "Client-only testing"
1459 if ! $r which $TUNEFS; then
1460 skip_env "tunefs.lustre required on $node"
1464 local IMGTYPE=$(facet_fstype $SINGLEMDS)
1466 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 -name \'disk*-$IMGTYPE.tar.bz2\')
1468 if [ -z "$tarballs" ]; then
1469 skip "No applicable tarballs found"
1474 t32_test_cleanup() {
1476 local fstype=$(facet_fstype $SINGLEMDS)
1479 if $shall_cleanup_lustre; then
1480 umount $tmp/mnt/lustre || rc=$?
1482 if $shall_cleanup_mdt; then
1483 $r $UMOUNT $tmp/mnt/mdt || rc=$?
1485 if $shall_cleanup_mdt1; then
1486 $r $UMOUNT $tmp/mnt/mdt1 || rc=$?
1488 if $shall_cleanup_ost; then
1489 $r $UMOUNT $tmp/mnt/ost || rc=$?
1494 if [ $fstype == "zfs" ]; then
1495 $r $ZPOOL destroy t32fs-mdt1 || rc=$?
1496 $r $ZPOOL destroy t32fs-ost1 || rc=$?
1501 t32_bits_per_long() {
1503 # Yes, this is not meant to be perfect.
1513 t32_reload_modules() {
1515 local all_removed=false
1518 while ((i < 20)); do
1519 echo "Unloading modules on $node: Attempt $i"
1520 do_rpc_nodes $node $LUSTRE_RMMOD $(facet_fstype $SINGLEMDS) &&
1522 do_rpc_nodes $node check_mem_leak || return 1
1523 if $all_removed; then
1524 do_rpc_nodes $node load_modules
1530 echo "Unloading modules on $node: Given up"
1534 t32_wait_til_devices_gone() {
1540 echo wait for devices to go
1541 while ((i < 20)); do
1542 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1543 loops=$(do_rpc_nodes $node losetup -a | grep -c t32)
1544 ((devices == 0 && loops == 0)) && return 0
1548 echo "waiting for dev on $node: dev $devices loop $loops given up"
1549 do_rpc_nodes $node "losetup -a"
1550 do_rpc_nodes $node "$LCTL device_list"
1554 t32_verify_quota() {
1558 local fstype=$(facet_fstype $SINGLEMDS)
1562 $LFS quota -u $T32_QID -v $mnt
1564 qval=$($LFS quota -v -u $T32_QID $mnt |
1565 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1566 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1568 [ $qval -eq $img_bspace ] || {
1569 echo "bspace, act:$qval, exp:$img_bspace"
1573 qval=$($LFS quota -v -u $T32_QID $mnt |
1574 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1575 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1577 [ $qval -eq $img_ispace ] || {
1578 echo "ispace, act:$qval, exp:$img_ispace"
1582 qval=$($LFS quota -v -u $T32_QID $mnt |
1583 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1584 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1586 [ $qval -eq $T32_BLIMIT ] || {
1587 echo "blimit, act:$qval, exp:$T32_BLIMIT"
1591 qval=$($LFS quota -v -u $T32_QID $mnt |
1592 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1593 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1595 [ $qval -eq $T32_ILIMIT ] || {
1596 echo "ilimit, act:$qval, exp:$T32_ILIMIT"
1600 do_node $node $LCTL conf_param $fsname.quota.mdt=ug
1601 cmd="$LCTL get_param -n osd-$fstype.$fsname-MDT0000"
1602 cmd=$cmd.quota_slave.enabled
1603 wait_update $node "$cmd" "ug" || {
1604 echo "Enable mdt quota failed"
1608 do_node $node $LCTL conf_param $fsname.quota.ost=ug
1609 cmd="$LCTL get_param -n osd-$fstype.$fsname-OST0000"
1610 cmd=$cmd.quota_slave.enabled
1611 wait_update $node "$cmd" "ug" || {
1612 echo "Enable ost quota failed"
1617 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1618 bs=1M count=$(($T32_BLIMIT / 1024)) oflag=sync && {
1619 echo "Write succeed, but expect -EDQUOT"
1622 rm -f $mnt/t32_qf_new
1624 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1626 echo "Create succeed, but expect -EDQUOT"
1629 unlinkmany $mnt/t32_qf_ $T32_ILIMIT
1637 local dne_upgrade=${dne_upgrade:-"no"}
1638 local ff_convert=${ff_convert:-"no"}
1639 local shall_cleanup_mdt=false
1640 local shall_cleanup_mdt1=false
1641 local shall_cleanup_ost=false
1642 local shall_cleanup_lustre=false
1643 local node=$(facet_active_host $SINGLEMDS)
1644 local r="do_node $node"
1645 local node2=$(facet_active_host mds2)
1653 local nid=$($r $LCTL list_nids | head -1)
1659 local fstype=$(facet_fstype $SINGLEMDS)
1660 local mdt_dev=$tmp/mdt
1661 local mdt2_dev=$tmp/mdt2
1662 local ost_dev=$tmp/ost
1666 trap 'trap - RETURN; t32_test_cleanup' RETURN
1669 mkdir -p $tmp/mnt/lustre || error "mkdir $tmp/mnt/lustre failed"
1670 $r mkdir -p $tmp/mnt/{mdt,mdt1,ost}
1671 $r tar xjvf $tarball -S -C $tmp || {
1672 error_noexit "Unpacking the disk image tarball"
1675 img_commit=$($r cat $tmp/commit)
1676 img_kernel=$($r cat $tmp/kernel)
1677 img_arch=$($r cat $tmp/arch)
1678 img_bspace=$($r cat $tmp/bspace)
1679 img_ispace=$($r cat $tmp/ispace)
1680 echo "Upgrading from $(basename $tarball), created with:"
1681 echo " Commit: $img_commit"
1682 echo " Kernel: $img_kernel"
1683 echo " Arch: $img_arch"
1684 echo "OST version: $(lustre_build_version ost1)"
1686 # The conversion can be made only when both of the following
1687 # conditions are satisfied:
1688 # - ost device img version < 2.3.64
1689 # - ost server version >= 2.5
1690 [ $(version_code $img_commit) -ge $(version_code 2.3.64) -o \
1691 $(lustre_version_code ost1) -lt $(version_code 2.5.0) ] &&
1694 if [ $fstype == "zfs" ]; then
1696 $r $ZPOOL import -f -d $tmp t32fs-mdt1
1697 $r $ZPOOL import -f -d $tmp t32fs-ost1
1698 mdt_dev=t32fs-mdt1/mdt1
1699 ost_dev=t32fs-ost1/ost1
1700 wait_update_facet $SINGLEMDS "$ZPOOL list |
1701 awk '/^t32fs-mdt1/ { print \\\$1 }'" "t32fs-mdt1" || {
1702 error_noexit "import zfs pool failed"
1707 $r $LCTL set_param debug="$PTLDEBUG"
1709 $r $TUNEFS --dryrun $mdt_dev || {
1711 error_noexit "tunefs.lustre before mounting the MDT"
1715 if $r test -f $mdt2_dev; then
1716 $r $TUNEFS --dryrun $mdt2_dev || {
1718 error_noexit "tunefs.lustre before mounting the MDT"
1723 if [ "$writeconf" ]; then
1725 if [ $fstype == "ldiskfs" ]; then
1727 $r $TUNEFS --quota $mdt_dev || {
1729 error_noexit "Enable mdt quota feature"
1732 if $r test -f $mdt2_dev; then
1733 $r $TUNEFS --quota $mdt2_dev || {
1735 error_noexit "Enable mdt quota feature"
1741 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
1742 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
1743 { skip "LU-2200: Cannot run over Inifiniband w/o lctl replace_nids "
1744 "(Need MGS version at least 2.3.59)"; return 0; }
1746 local osthost=$(facet_active_host ost1)
1747 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
1750 if [ $fstype == "ldiskfs" ]; then
1753 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt
1754 $r $LCTL replace_nids $fsname-OST0000 $ostnid
1755 $r $LCTL replace_nids $fsname-MDT0000 $nid
1756 $r $UMOUNT $tmp/mnt/mdt
1759 mopts=exclude=$fsname-OST0000
1760 if [ $fstype == "ldiskfs" ]; then
1765 t32_wait_til_devices_gone $node
1767 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
1769 error_noexit "Mounting the MDT"
1772 shall_cleanup_mdt=true
1774 if $r test -f $mdt2_dev; then
1775 mopts=mgsnode=$nid,$mopts
1776 $r $MOUNT_CMD -o $mopts $mdt2_dev $tmp/mnt/mdt1 || {
1778 error_noexit "Mounting the MDT"
1782 echo "mount new MDT....$mdt2_dev"
1783 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1784 error_noexit "enable remote dir create failed"
1786 shall_cleanup_mdt1=true
1787 elif [ "$dne_upgrade" != "no" ]; then
1788 local fs2mdsdev=$(mdsdevname 1_2)
1789 local fs2mdsvdev=$(mdsvdevname 1_2)
1791 echo "mkfs new MDT on ${fs2mdsdev}...."
1792 if [ $(facet_fstype mds1) == ldiskfs ]; then
1793 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
1796 add fs2mds $(mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
1797 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
1798 error_noexit "Mkfs new MDT failed"
1802 $r $TUNEFS --dryrun $fs2mdsdev || {
1803 error_noexit "tunefs.lustre before mounting the MDT"
1807 echo "mount new MDT....$fs2mdsdev"
1808 $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
1809 error_noexit "mount mdt1 failed"
1813 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1814 error_noexit "enable remote dir create failed"
1816 shall_cleanup_mdt1=true
1819 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
1820 error_noexit "Getting MDT UUID"
1823 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
1824 error_noexit "Unexpected MDT UUID: \"$uuid\""
1828 $r $TUNEFS --dryrun $ost_dev || {
1829 error_noexit "tunefs.lustre before mounting the OST"
1832 if [ "$writeconf" ]; then
1833 mopts=mgsnode=$nid,$writeconf
1834 if [ $fstype == "ldiskfs" ]; then
1836 $r $TUNEFS --quota $ost_dev || {
1838 error_noexit "Enable ost quota feature"
1844 if [ $fstype == "ldiskfs" ]; then
1848 $r $MOUNT_CMD -o $mopts $ost_dev $tmp/mnt/ost || {
1849 error_noexit "Mounting the OST"
1852 shall_cleanup_ost=true
1854 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
1855 error_noexit "Getting OST UUID"
1858 if [ "$uuid" != $fsname-OST0000_UUID ]; then
1859 error_noexit "Unexpected OST UUID: \"$uuid\""
1863 $r $LCTL conf_param $fsname-OST0000.osc.max_dirty_mb=15 || {
1864 error_noexit "Setting \"max_dirty_mb\""
1867 $r $LCTL conf_param $fsname-OST0000.failover.node=$nid || {
1868 error_noexit "Setting OST \"failover.node\""
1871 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
1872 error_noexit "Setting \"max_rpcs_in_flight\""
1875 $r $LCTL conf_param $fsname-MDT0000.failover.node=$nid || {
1876 error_noexit "Setting MDT \"failover.node\""
1879 $r $LCTL pool_new $fsname.interop || {
1880 error_noexit "Setting \"interop\""
1883 $r $LCTL conf_param $fsname-MDT0000.lov.stripesize=4M || {
1884 error_noexit "Setting \"lov.stripesize\""
1887 $r $LCTL conf_param $fsname-MDT0000.mdd.atime_diff=70 || {
1888 error_noexit "Setting \"mdd.atime_diff\""
1892 if [ "$ff_convert" != "no" -a $(facet_fstype ost1) == "ldiskfs" ]; then
1893 $r $LCTL lfsck_start -M $fsname-OST0000 || {
1894 error_noexit "Start OI scrub on OST0"
1898 # The oi_scrub should be on ost1, but for test_32(),
1899 # all on the SINGLEMDS.
1900 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
1901 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1902 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
1903 error_noexit "Failed to get the expected 'completed'"
1907 local UPDATED=$($r $LCTL get_param -n \
1908 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1909 awk '/^updated/ { print $2 }')
1910 [ $UPDATED -ge 1 ] || {
1911 error_noexit "Only $UPDATED objects have been converted"
1916 if [ "$dne_upgrade" != "no" ]; then
1917 $r $LCTL conf_param \
1918 $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
1919 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
1922 $r $LCTL conf_param $fsname-MDT0001.failover.node=$nid || {
1923 error_noexit "Setting MDT1 \"failover.node\""
1926 $r $LCTL conf_param $fsname-MDT0001.lov.stripesize=4M || {
1927 error_noexit "Setting MDT1 \"lov.stripesize\""
1932 if [ "$writeconf" ]; then
1933 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
1934 error_noexit "Mounting the client"
1938 shall_cleanup_lustre=true
1939 $r $LCTL set_param debug="$PTLDEBUG"
1941 t32_verify_quota $node $fsname $tmp/mnt/lustre || {
1942 error_noexit "verify quota failed"
1946 if [ "$dne_upgrade" != "no" ]; then
1947 $LFS mkdir -i 1 -c2 $tmp/mnt/lustre/striped_dir || {
1948 error_noexit "set remote dir failed"
1952 $LFS setdirstripe -D -c2 $tmp/mnt/lustre/striped_dir
1954 pushd $tmp/mnt/lustre
1955 tar -cf - . --exclude=./striped_dir \
1956 --exclude=./remote_dir |
1957 tar -xvf - -C striped_dir 1>/dev/null || {
1958 error_noexit "cp to remote dir failed"
1964 # If it is upgrade from DNE (2.5), then rename the remote dir,
1965 # which is created in 2.5 to striped dir.
1966 if $r test -f $mdt2_dev; then
1967 stripe_index=$(LFS getdirstripe -i \
1968 $tmp/mnt/lustre/remote_dir)
1969 [ $stripe_index -eq 1 ] || {
1970 error_noexit "get index $striped_index failed"
1973 mv $tmp/mnt/lustre/remote_dir \
1974 $tmp/mnt/lustre/striped_dir/ || {
1975 error_noexit "mv failed"
1980 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
1981 error_noexit "dd failed"
1984 rm -rf $tmp/mnt/lustre/tmp_file || {
1985 error_noexit "rm failed"
1989 if $r test -f $tmp/sha1sums; then
1990 # LU-2393 - do both sorts on same node to ensure locale
1992 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
1993 if [ "$dne_upgrade" != "no" ]; then
1994 pushd $tmp/mnt/lustre/striped_dir
1996 pushd $tmp/mnt/lustre
1999 find ! -path "*remote_dir*" ! -name .lustre -type f \
2000 -exec sha1sum {} \; |
2001 sort -k 2 >$tmp/sha1sums || {
2003 error_noexit "sha1sum"
2007 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
2008 error_noexit "sha1sum verification failed"
2012 # if upgrade from DNE (2.5), then check remote directory
2013 if $r test -f $mdt2_dev; then
2014 pushd $tmp/mnt/lustre/striped_dir/remote_dir
2015 find ! -name .lustre -type f \
2016 -exec sha1sum {} \; |
2017 sort -k 2 >$tmp/sha1sums || {
2019 error_noexit "sha1sum"
2023 if ! diff -ub $tmp/sha1sums.orig \
2025 error_noexit "sha1sum dne failed"
2030 echo "sha1sum verification skipped"
2033 if [ "$dne_upgrade" != "no" ]; then
2034 rm -rf $tmp/mnt/lustre/striped_dir || {
2035 error_noexit "remove remote dir failed"
2040 if $r test -f $tmp/list; then
2042 # There is not a Test Framework API to copy files to or
2043 # from a remote node.
2045 # LU-2393 - do both sorts on same node to ensure locale
2047 $r cat $tmp/list | sort -k 6 >$tmp/list.orig
2048 pushd $tmp/mnt/lustre
2049 ls -Rni --time-style=+%s | sort -k 6 >$tmp/list || {
2055 # 32-bit and 64-bit clients use different algorithms to
2056 # convert FIDs into inode numbers. Hence, remove the inode
2057 # numbers from the lists, if the original list was created
2058 # on an architecture with different number of bits per
2061 if [ $(t32_bits_per_long $(uname -m)) != \
2062 $(t32_bits_per_long $img_arch) ]; then
2063 echo "Different number of bits per \"long\" from the disk image"
2064 for list in list.orig list; do
2065 sed -i -e 's/^[0-9]\+[ \t]\+//' $tmp/$list
2068 if ! diff -ub $tmp/list.orig $tmp/list; then
2069 error_noexit "list verification failed"
2073 echo "list verification skipped"
2076 # migrate files/dirs to remote MDT, then move them back
2077 if [ $(lustre_version_code mds1) -ge $(version_code 2.7.50) -a \
2078 $dne_upgrade != "no" ]; then
2079 $r $LCTL set_param -n \
2080 mdt.${fsname}*.enable_remote_dir=1 2>/dev/null
2082 echo "test migration"
2083 pushd $tmp/mnt/lustre
2084 for dir in $(find ! -name .lustre ! -name . -type d); do
2085 mdt_index=$($LFS getdirstripe -i $dir)
2086 stripe_cnt=$($LFS getdirstripe -c $dir)
2087 if [ $mdt_index = 0 -a $stripe_cnt -le 1 ]; then
2088 $LFS mv -M 1 $dir || {
2090 error_noexit "migrate MDT1 failed"
2096 for dir in $(find ! -name . ! -name .lustre -type d); do
2097 mdt_index=$($LFS getdirstripe -i $dir)
2098 stripe_cnt=$($LFS getdirstripe -c $dir)
2099 if [ $mdt_index = 1 -a $stripe_cnt -le 1 ]; then
2100 $LFS mv -M 0 $dir || {
2102 error_noexit "migrate MDT0 failed"
2111 # When adding new data verification tests, please check for
2112 # the presence of the required reference files first, like
2113 # the "sha1sums" and "list" tests above, to avoid the need to
2114 # regenerate every image for each test addition.
2117 nrpcs_orig=$($LCTL get_param \
2118 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
2119 error_noexit "Getting \"max_rpcs_in_flight\""
2122 nrpcs=$((nrpcs_orig + 5))
2123 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=$nrpcs || {
2124 error_noexit "Changing \"max_rpcs_in_flight\""
2127 wait_update $HOSTNAME "$LCTL get_param \
2128 -n mdc.*MDT0000*.max_rpcs_in_flight" $nrpcs || {
2129 error_noexit "Verifying \"max_rpcs_in_flight\""
2133 umount $tmp/mnt/lustre || {
2134 error_noexit "Unmounting the client"
2137 shall_cleanup_lustre=false
2139 if [ "$dne_upgrade" != "no" ]; then
2140 $r $UMOUNT $tmp/mnt/mdt1 || {
2141 error_noexit "Unmounting the MDT2"
2144 shall_cleanup_mdt1=false
2147 $r $UMOUNT $tmp/mnt/mdt || {
2148 error_noexit "Unmounting the MDT"
2151 shall_cleanup_mdt=false
2153 $r $UMOUNT $tmp/mnt/ost || {
2154 error_noexit "Unmounting the OST"
2157 shall_cleanup_ost=false
2159 t32_reload_modules $node || {
2160 error_noexit "Reloading modules"
2164 # mount a second time to make sure we didnt leave upgrade flag on
2165 $r $TUNEFS --dryrun $mdt_dev || {
2167 error_noexit "tunefs.lustre before remounting the MDT"
2171 mopts=exclude=$fsname-OST0000
2172 if [ $fstype == "ldiskfs" ]; then
2175 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2176 error_noexit "Remounting the MDT"
2179 shall_cleanup_mdt=true
2189 for tarball in $tarballs; do
2190 t32_test $tarball || let "rc += $?"
2194 run_test 32a "Upgrade (not live)"
2202 for tarball in $tarballs; do
2203 t32_test $tarball writeconf || let "rc += $?"
2207 run_test 32b "Upgrade with writeconf"
2214 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
2216 for tarball in $tarballs; do
2217 # Do not support 1_8 and 2_1 direct upgrade to DNE2 anymore */
2218 echo $tarball | grep "1_8" && continue
2219 echo $tarball | grep "2_1" && continue
2221 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
2225 run_test 32c "dne upgrade test"
2233 for tarball in $tarballs; do
2234 ff_convert=yes t32_test $tarball || rc=$?
2238 run_test 32d "convert ff test"
2240 test_33a() { # bug 12333, was test_33
2241 local FSNAME2=test-123
2242 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2245 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
2247 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
2248 local dev=${SINGLEMDS}_dev
2249 local MDSDEV=${!dev}
2250 is_blkdev $SINGLEMDS $MDSDEV &&
2251 skip_env "mixed loopback and real device not working" &&
2255 local fs2mdsdev=$(mdsdevname 1_2)
2256 local fs2ostdev=$(ostdevname 1_2)
2257 local fs2mdsvdev=$(mdsvdevname 1_2)
2258 local fs2ostvdev=$(ostvdevname 1_2)
2260 if [ $(facet_fstype mds1) == ldiskfs ]; then
2261 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
2264 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2265 --reformat $mkfsoptions $fs2mdsdev $fs2mdsvdev || exit 10
2266 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2267 --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \
2268 $fs2ostvdev || exit 10
2270 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
2271 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2272 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" ||
2273 error "$LCTL conf_param $FSNAME2.sys.timeout=200 failed"
2274 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2275 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2278 cp /etc/hosts $MOUNT2/ || error "copy /etc/hosts $MOUNT2/ failed"
2279 $GETSTRIPE $MOUNT2/hosts || error "$GETSTRIPE $MOUNT2/hosts failed"
2284 cleanup_nocli || error "cleanup_nocli failed with $?"
2286 run_test 33a "Mount ost with a large index number"
2288 test_33b() { # was test_34
2291 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
2292 # Drop lock cancelation reply during umount
2293 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
2294 do_facet client $LCTL set_param fail_loc=0x80000304
2295 #lctl set_param debug=-1
2296 umount_client $MOUNT
2297 cleanup || error "cleanup failed with $?"
2299 run_test 33b "Drop cancel during umount"
2303 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
2304 manual_umount_client
2306 do_facet client killall -USR1 multiop
2307 if [ $rc -eq 0 ]; then
2308 error "umount not fail!"
2311 cleanup || error "cleanup failed with rc $?"
2313 run_test 34a "umount with opened file should be fail"
2317 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2318 stop_mds || error "Unable to stop MDS"
2320 manual_umount_client --force || error "mtab after failed umount with $?"
2322 cleanup || error "cleanup failed with $?"
2324 run_test 34b "force umount with failed mds should be normal"
2328 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2329 stop_ost || error "Unable to stop OST1"
2331 manual_umount_client --force || error "mtab after failed umount with $?"
2333 cleanup || error "cleanup failed with $?"
2335 run_test 34c "force umount with failed ost should be normal"
2337 test_35a() { # bug 12459
2340 DBG_SAVE="`$LCTL get_param -n debug`"
2341 $LCTL set_param debug="ha"
2343 log "Set up a fake failnode for the MDS"
2345 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2346 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2347 do_facet mgs "$LCTL conf_param \
2348 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" ||
2349 error "Setting ${device}.failover.node=\
2350 $(h2$NETTYPE $FAKENID) failed."
2352 log "Wait for RECONNECT_INTERVAL seconds (10s)"
2355 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
2358 log "Stopping the MDT: $device"
2359 stop_mdt 1 || error "MDT0 stop fail"
2361 df $MOUNT > /dev/null 2>&1 &
2363 log "Restarting the MDT: $device"
2364 start_mdt 1 || error "MDT0 start fail"
2365 log "Wait for df ($DFPID) ... "
2368 $LCTL set_param debug="$DBG_SAVE"
2370 # retrieve from the log the first server that the client tried to
2371 # contact after the connection loss
2372 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2373 NEXTCONN=`awk "/${MSG}/ {start = 1;}
2374 /import_select_connection.*$device-mdc.* using connection/ {
2376 if (\\\$NF ~ /$FAKENID/)
2382 }" $TMP/lustre-log-$TESTNAME.log`
2383 [ "$NEXTCONN" != "0" ] &&
2384 error "Tried to connect to ${NEXTCONN} not last active server"
2385 cleanup || error "cleanup failed with $?"
2386 # remove nid settings
2387 writeconf_or_reformat
2389 run_test 35a "Reconnect to the last active server first"
2391 test_35b() { # bug 18674
2392 remote_mds || { skip "local MDS" && return 0; }
2396 $LCTL set_param debug="ha"
2398 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
2401 log "Set up a fake failnode for the MDS"
2403 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2404 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2405 do_facet mgs "$LCTL conf_param \
2406 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" ||
2407 error "Set ${device}.failover.node=\
2408 $(h2$NETTYPE $FAKENID) failed"
2410 local at_max_saved=0
2411 # adaptive timeouts may prevent seeing the issue
2412 if at_is_enabled; then
2413 at_max_saved=$(at_max_get mds)
2414 at_max_set 0 mds client
2417 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
2419 log "Injecting EBUSY on MDS"
2420 # Setting OBD_FAIL_MDS_RESEND=0x136
2421 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" ||
2422 error "unable to set param fail_loc=0x80000136"
2424 $LCTL set_param mdc.${FSNAME}*.stats=clear
2426 log "Creating a test file and stat it"
2427 touch $MOUNT/$tdir/$tfile || error "touch $MOUNT/$tdir/$tfile failed"
2428 stat $MOUNT/$tdir/$tfile
2430 log "Stop injecting EBUSY on MDS"
2431 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" ||
2432 error "unable to set param fail_loc=0"
2433 rm -f $MOUNT/$tdir/$tfile || error "remove $MOUNT/$tdir/$tfile failed"
2436 # restore adaptive timeout
2437 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
2439 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2441 CONNCNT=$($LCTL get_param mdc.${FSNAME}*.stats |
2442 awk '/mds_connect/{print $2}')
2444 # retrieve from the log if the client has ever tried to
2445 # contact the fake server after the loss of connection
2446 FAILCONN=`awk "BEGIN {ret = 0;}
2447 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
2449 if (\\\$NF ~ /$FAKENID/) {
2454 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
2456 [ "$FAILCONN" == "0" ] &&
2457 error "The client reconnection has not been triggered"
2458 [ "$FAILCONN" == "2" ] &&
2459 error "Primary server busy, client reconnect to failover failed"
2462 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
2463 # Reconnects are supposed to be rate limited to one every 5s
2464 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] &&
2465 error "Too many reconnects $CONNCNT"
2467 cleanup || error "cleanup failed with $?"
2468 # remove nid settings
2469 writeconf_or_reformat
2471 run_test 35b "Continue reconnection retries, if the active server is busy"
2474 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
2476 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] ||
2477 { skip "remote OST" && return 0; }
2480 local FSNAME2=test1234
2481 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2483 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
2485 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
2486 is_blkdev $SINGLEMDS $MDSDEV &&
2487 skip_env "mixed loopback and real device not working" && return
2490 local fs2mdsdev=$(mdsdevname 1_2)
2491 local fs2ostdev=$(ostdevname 1_2)
2492 local fs3ostdev=$(ostdevname 2_2)
2493 local fs2mdsvdev=$(mdsvdevname 1_2)
2494 local fs2ostvdev=$(ostvdevname 1_2)
2495 local fs3ostvdev=$(ostvdevname 2_2)
2497 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2498 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
2499 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
2500 # different one than the default value here.
2501 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2502 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
2503 add fs3ost $(mkfs_opts ost2 ${fs3ostdev}) --mgsnode=$MGSNID \
2504 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
2506 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
2507 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2508 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
2509 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2510 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2512 sleep 5 # until 11778 fixed
2514 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || error "dd failed"
2516 BKTOTAL=$($LCTL get_param -n obdfilter.*.kbytestotal |
2517 awk 'BEGIN{total=0}; {total+=$1}; END{print total}')
2518 BKFREE=$($LCTL get_param -n obdfilter.*.kbytesfree |
2519 awk 'BEGIN{free=0}; {free+=$1}; END{print free}')
2520 BKAVAIL=$($LCTL get_param -n obdfilter.*.kbytesavail |
2521 awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}')
2522 STRING=$(df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}')
2523 DFTOTAL=$(echo $STRING | cut -d, -f1)
2524 DFUSED=$(echo $STRING | cut -d, -f2)
2525 DFAVAIL=$(echo $STRING | cut -d, -f3)
2526 DFFREE=$(($DFTOTAL - $DFUSED))
2528 ALLOWANCE=$((64 * $OSTCOUNT))
2530 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
2531 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
2532 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
2535 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
2536 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
2537 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
2540 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
2541 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
2542 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
2547 stop fs3ost -f || error "unable to stop OST3"
2548 stop fs2ost -f || error "unable to stop OST2"
2549 stop fs2mds -f || error "unable to stop second MDS"
2550 unload_modules_conf || error "unable unload modules"
2553 run_test 36 "df report consistency on OSTs with different block size"
2556 local mntpt=$(facet_mntpt $SINGLEMDS)
2557 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
2558 local mdsdev_sym="$TMP/sym_mdt.img"
2559 local opts=$MDS_MOUNT_OPTS
2562 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2563 skip "Currently only applicable to ldiskfs-based MDTs"
2567 echo "MDS : $mdsdev"
2568 echo "SYMLINK : $mdsdev_sym"
2569 do_facet $SINGLEMDS rm -f $mdsdev_sym
2571 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
2573 echo "mount symlink device - $mdsdev_sym"
2575 if ! do_facet $SINGLEMDS test -b $mdsdev; then
2576 opts=$(csa_add "$opts" -o loop)
2578 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
2579 $mdsdev_sym $mntpt 2>&1)
2582 echo mount_op=$mount_op
2584 do_facet $SINGLEMDS "$UMOUNT $mntpt && rm -f $mdsdev_sym"
2586 if $(echo $mount_op | grep -q "unable to set tunable"); then
2587 error "set tunables failed for symlink device"
2590 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
2592 run_test 37 "verify set tunables works for symlink device"
2594 test_38() { # bug 14222
2595 local fstype=$(facet_fstype $SINGLEMDS)
2596 local mntpt=$(facet_mntpt $SINGLEMDS)
2601 local SRC="/etc /bin"
2602 local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT)
2603 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
2604 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2605 tar cf - $FILES | tar xf - -C $DIR/$tdir ||
2606 error "copying $SRC to $DIR/$tdir"
2608 umount_client $MOUNT || error "umount_client $MOUNT failed"
2609 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2610 stop_mds || error "Unable to stop MDS"
2611 log "delete lov_objid file on MDS"
2613 mount_fstype $SINGLEMDS || error "mount MDS failed (1)"
2615 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid; rm $mntpt/lov_objid"
2617 unmount_fstype $SINGLEMDS || error "umount failed (1)"
2619 # check create in mds_lov_connect
2620 start_mds || error "unable to start MDS"
2621 mount_client $MOUNT || error "mount_client $MOUNT failed"
2623 [ $V ] && log "verifying $DIR/$tdir/$f"
2624 diff -q $f $DIR/$tdir/$f || ERROR=y
2626 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2627 if [ "$ERROR" = "y" ]; then
2628 # check it's updates in sync
2629 umount_client $MOUNT
2631 mount_fstype $SIGNLEMDS
2632 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2633 unmount_fstype $SINGLEMDS
2634 error "old and new files are different after connect" || true
2636 touch $DIR/$tdir/f2 || error "f2 file create failed"
2638 # check it's updates in sync
2639 umount_client $MOUNT || error "second umount_client $MOUNT failed"
2642 mount_fstype $SINGLEMDS || error "mount MDS failed (3)"
2644 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2645 do_facet $SINGLEMDS dd if=/dev/zero of=$mntpt/lov_objid.clear count=8
2647 unmount_fstype $SINGLEMDS || error "umount failed (3)"
2649 start_mds || error "unable to start MDS"
2650 mount_client $MOUNT || error "mount_client $MOUNT failed"
2652 [ $V ] && log "verifying $DIR/$tdir/$f"
2653 diff -q $f $DIR/$tdir/$f || ERROR=y
2655 touch $DIR/$tdir/f3 || error "f3 file create failed"
2656 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2657 umount_client $MOUNT || error "third umount_client $MOUNT failed"
2659 mount_fstype $SINGLEMDS || error "mount MDS failed (4)"
2660 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2661 unmount_fstype $SINGLEMDS || error "umount failed (4)"
2663 [ "$ERROR" = "y" ] &&
2664 error "old and new files are different after sync" || true
2666 log "files compared the same"
2667 cleanup || error "cleanup failed with $?"
2669 run_test 38 "MDS recreates missing lov_objid file from OST data"
2674 cleanup || error "cleanup failed with $?"
2675 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
2676 error "memory leak detected" || true
2678 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
2680 test_40() { # bug 15759
2681 start_ost || error "Unable to start OST1"
2682 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
2683 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
2685 cleanup || error "cleanup failed with rc $?"
2687 run_test 40 "race during service thread startup"
2689 test_41a() { #bug 14134
2690 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2691 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2692 skip "Loop devices does not work with nosvc option"
2696 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2698 start_mdt 1 -o nosvc -n
2699 if [ $MDSCOUNT -ge 2 ]; then
2700 for num in $(seq 2 $MDSCOUNT); do
2701 start_mdt $num || return
2704 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
2705 start_mdt 1 -o nomgs,force
2706 mount_client $MOUNT || error "mount_client $MOUNT failed"
2709 echo "blah blah" > $MOUNT/$tfile
2712 umount_client $MOUNT || error "umount_client $MOUNT failed"
2713 stop ost1 -f || error "unable to stop OST1"
2714 stop_mds || error "Unable to stop MDS"
2715 stop_mds || error "Unable to stop MDS on second try"
2716 unload_modules_conf || error "unload_modules_conf failed"
2718 run_test 41a "mount mds with --nosvc and --nomgs"
2721 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2722 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2723 skip "Loop devices does not work with nosvc option"
2727 ! combined_mgs_mds && skip "needs combined mgs device" && return 0
2731 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2733 start_mdt 1 -o nosvc -n
2734 if [ $MDSCOUNT -ge 2 ]; then
2735 for num in $(seq 2 $MDSCOUNT); do
2736 start_mdt $num || return
2739 start_ost || error "Unable to start OST1"
2740 start_mdt 1 -o nomgs,force
2741 mount_client $MOUNT || error "mount_client $MOUNT failed"
2744 echo "blah blah" > $MOUNT/$tfile
2745 cat $MOUNT/$tfile || error "cat $MOUNT/$tfile failed"
2747 umount_client $MOUNT || error "umount_client $MOUNT failed"
2748 stop_ost || error "Unable to stop OST1"
2749 stop_mds || error "Unable to stop MDS"
2750 stop_mds || error "Unable to stop MDS on second try"
2752 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
2755 local server_version=$(lustre_version_code $SINGLEMDS)
2757 [[ $server_version -ge $(version_code 2.6.52) ]] ||
2758 [[ $server_version -ge $(version_code 2.5.26) &&
2759 $server_version -lt $(version_code 2.5.50) ]] ||
2760 [[ $server_version -ge $(version_code 2.5.4) &&
2761 $server_version -lt $(version_code 2.5.11) ]] ||
2762 { skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"; return; }
2765 # MDT concurrent start
2766 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
2767 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x716"
2768 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS &
2770 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS &
2771 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x0"
2777 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
2778 echo "1st MDT start succeed"
2779 echo "2nd MDT start failed with $rc2"
2780 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
2781 echo "1st MDT start failed with $rc"
2782 echo "2nd MDT start succeed"
2785 error "unexpected concurrent MDT mounts result, rc=$rc rc2=$rc2"
2788 if [ $MDSCOUNT -ge 2 ]; then
2789 for num in $(seq 2 $MDSCOUNT); do
2790 start_mdt $num || return
2794 # OST concurrent start
2796 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
2797 do_facet ost1 "$LCTL set_param fail_loc=0x716"
2798 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &
2800 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &
2801 do_facet ost1 "$LCTL set_param fail_loc=0x0"
2807 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
2808 echo "1st OST start succeed"
2809 echo "2nd OST start failed with $rc2"
2810 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
2811 echo "1st OST start failed with $rc"
2812 echo "2nd OST start succeed"
2816 error "unexpected concurrent OST mounts result, rc=$rc rc2=$rc2"
2822 # verify everything ok
2827 error "MDT(s) start failed"
2835 error "OST(s) start failed"
2843 error "client start failed"
2850 error "client mount failed"
2854 run_test 41c "concurrent mounts of MDT/OST should all fail but one"
2856 test_42() { #bug 14693
2858 check_mount || error "client was not mounted"
2860 do_facet mgs $LCTL conf_param $FSNAME.llite.some_wrong_param=10
2861 umount_client $MOUNT ||
2862 error "unmounting client failed with invalid llite param"
2863 mount_client $MOUNT ||
2864 error "mounting client failed with invalid llite param"
2866 do_facet mgs $LCTL conf_param $FSNAME.sys.some_wrong_param=20
2867 cleanup || error "stopping $FSNAME failed with invalid sys param"
2870 check_mount || error "client was not mounted with invalid sys param"
2871 cleanup || error "stopping $FSNAME failed with invalid sys param"
2873 run_test 42 "allow client/server mount/unmount with invalid config param"
2876 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.58) ]] ||
2877 { skip "Need MDS version at least 2.5.58" && return 0; }
2878 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
2881 USER1=$(cat /etc/passwd | grep :$ID1:$ID1: | cut -d: -f1)
2882 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1" &&
2886 chmod ugo+x $DIR || error "chmod 0 failed"
2887 set_conf_param_and_check mds \
2888 "$LCTL get_param -n mdt.$FSNAME-MDT0000.root_squash" \
2889 "$FSNAME.mdt.root_squash" \
2891 wait_update $HOSTNAME \
2892 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
2894 error "check llite root_squash failed!"
2895 set_conf_param_and_check mds \
2896 "$LCTL get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
2897 "$FSNAME.mdt.nosquash_nids" \
2899 wait_update $HOSTNAME \
2900 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
2902 error "check llite nosquash_nids failed!"
2905 # create set of test files
2907 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
2908 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
2909 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
2911 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
2912 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
2914 mkdir $DIR/$tdir-rootdir || error "mkdir failed"
2915 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
2916 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
2918 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
2919 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
2920 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
2923 # check root_squash:
2924 # set root squash UID:GID to RUNAS_ID
2925 # root should be able to access only files owned by RUNAS_ID
2927 set_conf_param_and_check mds \
2928 "$LCTL get_param -n mdt.$FSNAME-MDT0000.root_squash" \
2929 "$FSNAME.mdt.root_squash" \
2930 "$RUNAS_ID:$RUNAS_ID"
2931 wait_update $HOSTNAME \
2932 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
2933 "$RUNAS_ID:$RUNAS_ID" ||
2934 error "check llite root_squash failed!"
2936 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
2937 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
2938 error "$ST: root read permission is denied"
2939 echo "$ST: root read permission is granted - ok"
2942 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
2943 error "$ST: root write permission is denied"
2944 echo "$ST: root write permission is granted - ok"
2946 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
2947 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
2948 error "$ST: root read permission is granted"
2949 echo "$ST: root read permission is denied - ok"
2952 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
2953 error "$ST: root write permission is granted"
2954 echo "$ST: root write permission is denied - ok"
2956 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
2957 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null &&
2958 error "$ST: root unlink permission is granted"
2959 echo "$ST: root unlink permission is denied - ok"
2961 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null &&
2962 error "$ST: root create permission is granted"
2963 echo "$ST: root create permission is denied - ok"
2966 # check root_squash is enforced independently
2967 # of client cache content
2969 # access file by USER1, keep access open
2970 # root should be denied access to user file
2972 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
2976 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
2977 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
2978 { kill $pid; error "$ST: root read permission is granted"; }
2979 echo "$ST: root read permission is denied - ok"
2982 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
2983 { kill $pid; error "$ST: root write permission is granted"; }
2984 echo "$ST: root write permission is denied - ok"
2990 # check nosquash_nids:
2991 # put client's NID into nosquash_nids list,
2992 # root should be able to access root file after that
2994 local NIDLIST=$($LCTL list_nids all | tr '\n' ' ')
2995 NIDLIST="2@gni $NIDLIST 192.168.0.[2,10]@tcp"
2996 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
2997 set_conf_param_and_check mds \
2998 "$LCTL get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
2999 "$FSNAME-MDTall.mdt.nosquash_nids" \
3001 wait_update $HOSTNAME \
3002 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3004 error "check llite nosquash_nids failed!"
3006 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3007 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3008 error "$ST: root read permission is denied"
3009 echo "$ST: root read permission is granted - ok"
3012 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3013 error "$ST: root write permission is denied"
3014 echo "$ST: root write permission is granted - ok"
3016 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3017 rm $DIR/$tdir-rootdir/tfile-1 ||
3018 error "$ST: root unlink permission is denied"
3019 echo "$ST: root unlink permission is granted - ok"
3020 touch $DIR/$tdir-rootdir/tfile-2 ||
3021 error "$ST: root create permission is denied"
3022 echo "$ST: root create permission is granted - ok"
3024 run_test 43 "check root_squash and nosquash_nids"
3026 umount_client $MOUNT
3031 check_mount || error "check_mount"
3032 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
3034 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
3035 for VAL in $UUIDS; do
3036 NID=$(echo $VAL | cut -d= -f1)
3037 CLUUID=$(echo $VAL | cut -d= -f2)
3038 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
3040 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
3041 cleanup || error "cleanup failed with $?"
3043 run_test 44 "mounted client proc entry exists"
3047 check_mount || error "check_mount"
3048 stop_mds || error "Unable to stop MDS"
3052 #define OBD_FAIL_PTLRPC_LONG_UNLINK 0x50f
3053 do_facet client "$LCTL set_param fail_loc=0x50f"
3056 manual_umount_client --force || error "manual_umount_client failed"
3057 do_facet client "$LCTL set_param fail_loc=0x0"
3058 start_mds || error "unable to start MDS"
3059 mount_client $MOUNT || error "mount_client $MOUNT failed"
3060 cleanup || error "cleanup failed with $?"
3062 run_test 45 "long unlink handling in ptlrpcd"
3069 umount_client $MOUNT2 || rc=$?
3070 umount_client $MOUNT || rc=$?
3071 while [ $count -gt 0 ]; do
3072 stop ost${count} -f || rc=$?
3076 cleanup_nocli || rc=$?
3077 #writeconf to remove all ost2 traces for subsequent tests
3078 writeconf_or_reformat
3083 echo "Testing with $OSTCOUNT OSTs"
3085 start_mds || error "unable to start MDS"
3086 #first client should see only one ost
3087 start_ost || error "Unable to start OST1"
3088 wait_osc_import_state mds ost FULL
3090 mount_client $MOUNT || error "mount_client $MOUNT failed"
3091 trap "cleanup_46a $OSTCOUNT" EXIT ERR
3094 for (( i=2; i<=$OSTCOUNT; i++ )); do
3095 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
3096 error "start_ost$i $(ostdevname $i) failed"
3099 # wait until osts in sync
3100 for (( i=2; i<=$OSTCOUNT; i++ )); do
3101 wait_osc_import_state mds ost$i FULL
3102 wait_osc_import_state client ost$i FULL
3105 #second client see all ost's
3107 mount_client $MOUNT2 || error "mount_client failed"
3108 $SETSTRIPE -c -1 $MOUNT2 || error "$SETSTRIPE -c -1 $MOUNT2 failed"
3109 $GETSTRIPE $MOUNT2 || error "$GETSTRIPE $MOUNT2 failed"
3111 echo "ok" > $MOUNT2/widestripe
3112 $GETSTRIPE $MOUNT2/widestripe ||
3113 error "$GETSTRIPE $MOUNT2/widestripe failed"
3114 # fill acl buffer for avoid expand lsm to them
3115 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd |
3117 setfacl -m $acl $MOUNT2/widestripe
3121 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3123 cleanup_46a $OSTCOUNT || error "cleanup_46a failed"
3125 run_test 46a "handle ost additional - wide striped file"
3130 check_mount || error "check_mount failed"
3131 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
3135 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3136 if echo $ns | grep "MDT[[:digit:]]*"; then
3139 lrs=$(echo $ns | sed 's/.*lru_size=//')
3140 lru_size[count]=$lrs
3145 facet_failover $SINGLEMDS
3146 client_up || error "client_up failed"
3149 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3150 if echo $ns | grep "MDT[[:digit:]]*"; then
3153 lrs=$(echo $ns | sed 's/.*lru_size=//')
3154 if ! test "$lrs" -eq "${lru_size[count]}"; then
3155 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
3156 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
3161 cleanup || error "cleanup failed with $?"
3163 run_test 47 "server restart does not make client loss lru_resize settings"
3168 # reformat after this test is needed - if the test fails,
3169 # we will have unkillable file at FS
3173 test_48() { # bug 17636
3176 check_mount || error "check_mount failed"
3178 $SETSTRIPE -c -1 $MOUNT || error "$SETSTRIPE -c -1 $MOUNT failed"
3179 $GETSTRIPE $MOUNT || error "$GETSTRIPE $MOUNT failed"
3181 echo "ok" > $MOUNT/widestripe
3182 $GETSTRIPE $MOUNT/widestripe ||
3183 error "$GETSTRIPE $MOUNT/widestripe failed"
3185 trap cleanup_48 EXIT ERR
3187 # fill acl buffer for avoid expand lsm to them
3188 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
3189 setfacl -m $acl $MOUNT/widestripe
3192 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3196 run_test 48 "too many acls on file"
3198 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
3199 test_49a() { # bug 17710
3200 local timeout_orig=$TIMEOUT
3201 local ldlm_timeout_orig=$LDLM_TIMEOUT
3202 local LOCAL_TIMEOUT=20
3204 LDLM_TIMEOUT=$LOCAL_TIMEOUT
3205 TIMEOUT=$LOCAL_TIMEOUT
3209 check_mount || error "client mount failed"
3211 echo "check ldlm_timout..."
3212 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3213 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3214 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3216 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3217 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3220 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
3221 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
3224 umount_client $MOUNT || error "umount_client $MOUNT failed"
3225 stop_ost || error "problem stopping OSS"
3226 stop_mds || error "problem stopping MDS"
3228 LDLM_TIMEOUT=$ldlm_timeout_orig
3229 TIMEOUT=$timeout_orig
3231 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3233 test_49b() { # bug 17710
3234 local timeout_orig=$TIMEOUT
3235 local ldlm_timeout_orig=$LDLM_TIMEOUT
3236 local LOCAL_TIMEOUT=20
3238 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
3239 TIMEOUT=$LOCAL_TIMEOUT
3243 check_mount || error "client mount failed"
3245 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3246 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3247 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3249 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3250 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3253 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
3254 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
3257 cleanup || error "cleanup failed"
3259 LDLM_TIMEOUT=$ldlm_timeout_orig
3260 TIMEOUT=$timeout_orig
3262 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3265 # Test both statfs and lfs df and fail if either one fails
3266 multiop_bg_pause $1 f_
3269 killall -USR1 multiop
3270 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
3271 wait $PID || { RC1=$?; log "multiop return error "; }
3278 if [ $RC2 -eq 0 ]; then
3280 log "lazystatfs df failed"
3284 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
3290 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3291 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3293 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
3295 cleanup || error "cleanup failed with rc $?"
3297 run_test 50a "lazystatfs all servers available"
3301 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3302 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3304 # Wait for client to detect down OST
3305 stop_ost || error "Unable to stop OST1"
3306 wait_osc_import_state mds ost DISCONN
3308 lazystatfs $MOUNT || error "lazystatfs should not return EIO"
3310 umount_client $MOUNT || error "Unable to unmount client"
3311 stop_mds || error "Unable to stop MDS"
3313 run_test 50b "lazystatfs all servers down"
3316 start_mds || error "Unable to start MDS"
3317 start_ost || error "Unable to start OST1"
3318 start_ost2 || error "Unable to start OST2"
3319 mount_client $MOUNT || error "Unable to mount client"
3320 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3321 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3323 # Wait for client to detect down OST
3324 stop_ost || error "Unable to stop OST1"
3325 wait_osc_import_state mds ost DISCONN
3326 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3328 umount_client $MOUNT || error "Unable to unmount client"
3329 stop_ost2 || error "Unable to stop OST2"
3330 stop_mds || error "Unable to stop MDS"
3331 #writeconf to remove all ost2 traces for subsequent tests
3332 writeconf_or_reformat
3334 run_test 50c "lazystatfs one server down"
3337 start_mds || error "Unable to start MDS"
3338 start_ost || error "Unable to start OST1"
3339 start_ost2 || error "Unable to start OST2"
3340 mount_client $MOUNT || error "Unable to mount client"
3341 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3342 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3344 # Issue the statfs during the window where the client still
3345 # belives the OST to be available but it is in fact down.
3346 # No failure just a statfs which hangs for a timeout interval.
3347 stop_ost || error "Unable to stop OST1"
3348 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3350 umount_client $MOUNT || error "Unable to unmount client"
3351 stop_ost2 || error "Unable to stop OST2"
3352 stop_mds || error "Unable to stop MDS"
3353 #writeconf to remove all ost2 traces for subsequent tests
3354 writeconf_or_reformat
3356 run_test 50d "lazystatfs client/server conn race"
3363 start_mds || error "Unable to start MDS"
3364 #first client should see only one ost
3365 start_ost || error "Unable to start OST1"
3366 wait_osc_import_state mds ost FULL
3368 # Wait for client to detect down OST
3369 stop_ost || error "Unable to stop OST1"
3370 wait_osc_import_state mds ost DISCONN
3372 mount_client $MOUNT || error "Unable to mount client"
3373 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3375 multiop_bg_pause $MOUNT _f
3379 if [ $RC1 -ne 0 ]; then
3380 log "multiop failed $RC1"
3383 sleep $(( $TIMEOUT+1 ))
3385 [ $? -ne 0 ] && error "process isn't sleep"
3386 start_ost || error "Unable to start OST1"
3387 wait $pid || error "statfs failed"
3390 umount_client $MOUNT || error "Unable to unmount client"
3391 stop_ost || error "Unable to stop OST1"
3392 stop_mds || error "Unable to stop MDS"
3394 run_test 50e "normal statfs all servers down"
3399 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
3401 start_mds || error "Unable to start mds"
3402 #first client should see only one ost
3403 start_ost || error "Unable to start OST1"
3404 wait_osc_import_state mds ost FULL
3406 start_ost2 || error "Unable to start OST2"
3407 wait_osc_import_state mds ost2 FULL
3409 # Wait for client to detect down OST
3410 stop_ost2 || error "Unable to stop OST2"
3412 wait_osc_import_state mds ost2 DISCONN
3413 mount_client $MOUNT || error "Unable to mount client"
3414 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3416 multiop_bg_pause $MOUNT _f
3420 if [ $RC1 -ne 0 ]; then
3421 log "lazystatfs multiop failed $RC1"
3424 sleep $(( $TIMEOUT+1 ))
3426 [ $? -ne 0 ] && error "process isn't sleep"
3427 start_ost2 || error "Unable to start OST2"
3428 wait $pid || error "statfs failed"
3429 stop_ost2 || error "Unable to stop OST2"
3432 umount_client $MOUNT -f || error "Unable to unmount client"
3433 stop_ost || error "Unable to stop OST1"
3434 stop_mds || error "Unable to stop MDS"
3435 #writeconf to remove all ost2 traces for subsequent tests
3436 writeconf_or_reformat
3438 run_test 50f "normal statfs one server in down"
3441 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3443 start_ost2 || error "Unable to start OST2"
3444 wait_osc_import_state mds ost2 FULL
3445 wait_osc_import_state client ost2 FULL
3447 local PARAM="${FSNAME}-OST0001.osc.active"
3449 $SETSTRIPE -c -1 $DIR/$tfile || error "$SETSTRIPE failed"
3450 do_facet mgs $LCTL conf_param $PARAM=0 ||
3451 error "Unable to deactivate OST"
3453 umount_client $MOUNT || error "Unable to unmount client"
3454 mount_client $MOUNT || error "Unable to mount client"
3455 # This df should not cause a panic
3458 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
3459 rm -f $DIR/$tfile || error "unable to remove file $DIR/$tfile"
3460 umount_client $MOUNT || error "Unable to unmount client"
3461 stop_ost2 || error "Unable to stop OST2"
3462 stop_ost || error "Unable to stop OST1"
3463 stop_mds || error "Unable to stop MDS"
3464 #writeconf to remove all ost2 traces for subsequent tests
3465 writeconf_or_reformat
3467 run_test 50g "deactivated OST should not cause panic"
3471 # prepare MDT/OST, make OSC inactive for OST1
3472 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3474 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3475 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
3476 error "tunefs OST1 failed"
3477 start_mds || error "Unable to start MDT"
3478 start_ost || error "Unable to start OST1"
3479 start_ost2 || error "Unable to start OST2"
3480 mount_client $MOUNT || error "client start failed"
3482 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3484 # activatate OSC for OST1
3485 local TEST="$LCTL get_param -n osc.${FSNAME}-OST0000-osc-[!M]*.active"
3486 set_conf_param_and_check client \
3487 "$TEST" "${FSNAME}-OST0000.osc.active" 1 ||
3488 error "Unable to activate OST1"
3490 mkdir $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3491 $SETSTRIPE -c -1 -i 0 $DIR/$tdir/2 ||
3492 error "$SETSTRIPE $DIR/$tdir/2 failed"
3493 sleep 1 && echo "create a file after OST1 is activated"
3495 createmany -o $DIR/$tdir/2/$tfile-%d 1
3497 # check OSC import is working
3498 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
3499 error "some OSC imports are still not connected"
3502 umount_client $MOUNT || error "Unable to umount client"
3503 stop_ost2 || error "Unable to stop OST2"
3504 cleanup_nocli || error "cleanup_nocli failed with $?"
3506 run_test 50h "LU-642: activate deactivated OST"
3509 # prepare MDT/OST, make OSC inactive for OST1
3510 [ "$MDSCOUNT" -lt "2" ] && skip_env "$MDSCOUNT < 2, skipping" && return
3512 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3514 do_facet mds2 "$TUNEFS --param mdc.active=0 $(mdsdevname 2)" ||
3515 error "tunefs MDT2 failed"
3516 start_mds || error "Unable to start MDT"
3517 start_ost || error "Unable to start OST1"
3518 start_ost2 || error "Unable to start OST2"
3519 mount_client $MOUNT || error "client start failed"
3521 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3523 $LCTL conf_param ${FSNAME}-MDT0000.mdc.active=0 &&
3524 error "deactive MDC0 succeeds"
3525 # activate MDC for MDT2
3526 local TEST="$LCTL get_param -n mdc.${FSNAME}-MDT0001-mdc-[!M]*.active"
3527 set_conf_param_and_check client \
3528 "$TEST" "${FSNAME}-MDT0001.mdc.active" 1 ||
3529 error "Unable to activate MDT2"
3531 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
3532 if [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.60) ]
3534 wait_dne_interconnect
3536 $LFS mkdir -i1 $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3538 createmany -o $DIR/$tdir/2/$tfile-%d 1 || error "create files failed"
3540 rm -rf $DIR/$tdir/2 || error "unlink dir failed"
3542 # deactivate MDC for MDT2
3543 local TEST="$LCTL get_param -n mdc.${FSNAME}-MDT0001-mdc-[!M]*.active"
3544 set_conf_param_and_check client \
3545 "$TEST" "${FSNAME}-MDT0001.mdc.active" 0 ||
3546 error "Unable to deactivate MDT2"
3548 $LFS mkdir -i1 $DIR/$tdir/2 &&
3549 error "mkdir $DIR/$tdir/2 succeeds after deactive MDT"
3552 umount_client $MOUNT || error "Unable to umount client"
3557 run_test 50i "activate deactivated MDT"
3560 local LOCAL_TIMEOUT=20
3564 check_mount || error "check_mount failed"
3566 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
3567 $SETSTRIPE -c -1 $MOUNT/$tdir ||
3568 error "$SETSTRIPE -c -1 $MOUNT/$tdir failed"
3569 #define OBD_FAIL_MDS_REINT_DELAY 0x142
3570 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x142"
3571 touch $MOUNT/$tdir/$tfile &
3574 start_ost2 || error "Unable to start OST1"
3576 stop_ost2 || error "Unable to stop OST1"
3577 umount_client $MOUNT -f || error “unmount $MOUNT failed”
3578 cleanup_nocli || error “stop server failed”
3579 #writeconf to remove all ost2 traces for subsequent tests
3580 writeconf_or_reformat
3582 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
3591 do_node $node mkdir -p $dest
3592 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3594 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
3595 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3596 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
3598 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
3599 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
3609 local backup2=${TMP}/backup2
3611 do_node $node mkdir -p $backup2
3612 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3614 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
3615 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3616 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
3618 do_node $node "diff -rq $backup $backup2"
3619 [ $? -eq 0 ] || { error "contents differ"; return 3; }
3621 local xattrs2=${TMP}/xattrs2
3622 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
3623 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
3625 do_node $node "diff $xattrs $xattrs2"
3626 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
3628 do_node $node "rm -rf $backup2 $xattrs2"
3629 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
3633 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3634 skip "Only applicable to ldiskfs-based MDTs"
3638 start_mds || error "Unable to start MDS"
3639 start_ost || error "Unable to start OST1"
3640 mount_client $MOUNT || error "Unable to mount client"
3643 local ost1mnt=$(facet_mntpt ost1)
3644 local ost1node=$(facet_active_host ost1)
3645 local ost1tmp=$TMP/conf52
3648 mkdir $DIR/$tdir || error "Unable to create $DIR/$tdir"
3649 touch $TMP/modified_first || error "Unable to create temporary file"
3650 local mtime=$(stat -c %Y $TMP/modified_first)
3651 do_node $ost1node "mkdir -p $ost1tmp &&
3652 touch -m -d @$mtime $ost1tmp/modified_first" ||
3653 error "Unable to create temporary file"
3656 $SETSTRIPE -c -1 -S 1M $DIR/$tdir || error "$SETSTRIPE failed"
3658 for (( i=0; i < nrfiles; i++ )); do
3659 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c ||
3660 error "multiop failed"
3666 echo backup files to $TMP/files
3667 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
3668 copy_files_xattrs $(hostname) $TMP/files $TMP/file_xattrs $files ||
3669 error "Unable to copy files"
3671 umount_client $MOUNT || error "Unable to umount client"
3672 stop_ost || error "Unable to stop ost1"
3674 echo mount ost1 as ldiskfs
3675 do_node $ost1node mkdir -p $ost1mnt || error "Unable to create $ost1mnt"
3676 if ! do_node $ost1node test -b $ost1_dev; then
3679 do_node $ost1node mount -t $(facet_fstype ost1) $loop $ost1_dev \
3681 error "Unable to mount ost1 as ldiskfs"
3684 echo backup objects to $ost1tmp/objects
3685 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
3686 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
3687 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
3689 error "Unable to copy objects"
3691 # move objects to lost+found
3692 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
3693 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
3695 do_node $ost1node "umount $ost1mnt" ||
3696 error "Unable to umount ost1 as ldiskfs"
3698 start_ost || error "Unable to start OST1"
3699 mount_client $MOUNT || error "Unable to mount client"
3701 local REPAIRED=$(do_node $ost1node "$LCTL get_param \
3702 -n osd-ldiskfs.$FSNAME-OST0000.oi_scrub" |
3703 awk '/^lf_repa[ri]*ed/ { print $2 }')
3704 [ $REPAIRED -gt 0 ] ||
3705 error "Some entry under /lost+found should be repaired"
3708 diff_files_xattrs $(hostname) $TMP/files $TMP/file_xattrs $files ||
3709 error "Unable to diff files"
3711 rm -rf $TMP/files $TMP/file_xattrs ||
3712 error "Unable to delete temporary files"
3713 do_node $ost1node "rm -rf $ost1tmp" ||
3714 error "Unable to delete temporary files"
3715 cleanup || error "cleanup failed with $?"
3717 run_test 52 "check recovering objects from lost+found"
3719 # Checks threads_min/max/started for some service
3721 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
3722 # parameter pattern prefix like 'ost.*.ost'.
3735 local msg="Insane $modname thread counts"
3736 local ncpts=$(check_cpt_number $facet)
3740 check_mount || return 41
3742 # We need to expand $parampat, but it may match multiple parameters, so
3743 # we'll pick the first one
3744 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
3745 error "Couldn't expand ${parampat}.threads_min parameter name"
3749 # Remove the .threads_min part
3750 paramp=${paramp%.threads_min}
3752 # Check for sanity in defaults
3753 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3755 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3757 tstarted=$(do_facet $facet "$LCTL get_param \
3758 -n ${paramp}.threads_started" || echo 0)
3759 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' ||
3761 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
3763 nthrs=$(expr $tmax - $tmin)
3764 if [ $nthrs -lt $ncpts ]; then
3770 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
3771 skip_env "module parameter forced $facet thread count" &&
3772 tmin=3 && tmax=$((3 * tmax))
3774 # Check that we can change min/max
3775 do_facet $facet "$LCTL set_param \
3776 ${paramp}.threads_min=$((tmin + nthrs))"
3777 do_facet $facet "$LCTL set_param \
3778 ${paramp}.threads_max=$((tmax - nthrs))"
3779 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3781 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3783 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) &&
3784 $tmax2 == ($tmax - $nthrs)))' || return $?
3786 # Check that we can set min/max to the same value
3787 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3789 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$tmin"
3790 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3792 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3794 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
3796 # Check that we can't set max < min
3797 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$((tmin - 1))"
3798 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3800 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3802 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
3804 # We need to ensure that we get the module options desired; to do this
3805 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
3806 LOAD_MODULES_REMOTE=true
3809 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
3810 setmodopts -a $modname "$newvalue" oldvalue
3814 check_mount || return 41
3816 # Restore previous setting of MODOPTS_*
3817 setmodopts $modname "$oldvalue"
3819 # Check that $opts took
3820 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min")
3821 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max")
3822 tstarted=$(do_facet $facet \
3823 "$LCTL get_param -n ${paramp}.threads_started")
3824 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
3834 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
3835 cleanup || error "cleanup failed with rc $?"
3837 run_test 53a "check OSS thread count params"
3841 local mds=$(do_facet $SINGLEMDS "$LCTL get_param \
3842 -N mds.*.*.threads_max 2>/dev/null")
3843 if [ -z "$mds" ]; then
3844 #running this on an old MDT
3845 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
3847 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
3849 cleanup || error "cleanup failed with $?"
3851 run_test 53b "check MDS thread count params"
3854 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3855 skip "Only applicable to ldiskfs-based MDTs"
3859 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p ||
3860 error "llverdev failed with rc=$?"
3863 run_test 54a "test llverdev and partial verify of device"
3866 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3867 skip "Only applicable to ldiskfs-based MDTs"
3872 run_llverfs $MOUNT -p || error "llverfs failed with rc=$?"
3873 cleanup || error "cleanup failed with rc=$?"
3875 run_test 54b "test llverfs and partial verify of filesystem"
3879 local max_ost_index=$1
3880 echo -n $(((max_ost_index + 1) * 8))
3884 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3885 skip "Only applicable to ldiskfs-based MDTs"
3889 local mdsdev=$(mdsdevname 1)
3890 local mdsvdev=$(mdsvdevname 1)
3894 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
3896 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
3897 --reformat $(ostdevname 1) $(ostvdevname 1)
3903 echo checking size of lov_objid for ost index $i
3904 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
3905 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
3906 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
3908 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
3915 run_test 55 "check lov_objid size"
3918 local server_version=$(lustre_version_code $SINGLEMDS)
3919 local mds_journal_size_orig=$MDSJOURNALSIZE
3924 for num in $(seq 1 $MDSCOUNT); do
3927 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
3928 $(ostdevname 1) $(ostvdevname 1)
3929 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
3930 $(ostdevname 2) $(ostvdevname 2)
3933 start_ost || error "Unable to start first ost (idx 10000)"
3934 start_ost2 || error "Unable to start second ost (idx 1000)"
3935 mount_client $MOUNT || error "Unable to mount client"
3939 if [[ $server_version -ge $(version_code 2.6.54) ]] ||
3940 [[ $server_version -ge $(version_code 2.5.4) &&
3941 $server_version -lt $(version_code 2.5.11) ]]; then
3942 wait_osc_import_state mds ost1 FULL
3943 wait_osc_import_state mds ost2 FULL
3944 $SETSTRIPE --stripe-count=-1 $DIR/$tfile ||
3945 error "Unable to setstripe $DIR/$tfile"
3946 n=$($LFS getstripe --stripe-count $DIR/$tfile)
3947 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
3952 MDSJOURNALSIZE=$mds_journal_size_orig
3955 run_test 56 "check big OST indexes and out-of-index-order start"
3957 test_57a() { # bug 22656
3958 do_rpc_nodes $(facet_active_host ost1) load_modules_local
3959 local NID=$(do_facet ost1 "$LCTL get_param nis" |
3960 tail -1 | awk '{print $1}')
3961 writeconf_or_reformat
3962 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3963 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" ||
3964 error "tunefs failed"
3966 start_ost && error "OST registration from failnode should fail"
3969 run_test 57a "initial registration from failnode should fail (should return errs)"
3972 do_rpc_nodes $(facet_active_host ost1) load_modules_local
3973 local NID=$(do_facet ost1 "$LCTL get_param nis" |
3974 tail -1 | awk '{print $1}')
3975 writeconf_or_reformat
3976 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3977 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" ||
3978 error "tunefs failed"
3980 start_ost || error "OST registration from servicenode should not fail"
3983 run_test 57b "initial registration from servicenode should not fail"
3986 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
3989 test_58() { # bug 22658
3991 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3992 createmany -o $DIR/$tdir/$tfile-%d 100
3993 # make sure that OSTs do not cancel llog cookies before we unmount the MDS
3994 #define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601
3995 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x601"
3996 unlinkmany $DIR/$tdir/$tfile-%d 100
3997 stop_mds || error "Unable to stop MDS"
3999 local MNTDIR=$(facet_mntpt $SINGLEMDS)
4000 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4002 # remove all files from the OBJECTS dir
4003 mount_fstype $SINGLEMDS
4005 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
4007 unmount_fstype $SINGLEMDS
4008 # restart MDS with missing llog files
4009 start_mds || error "unable to start MDS"
4010 do_facet mds "$LCTL set_param fail_loc=0"
4013 run_test 58 "missing llog files must not prevent MDT from mounting"
4016 start_mgsmds >> /dev/null
4017 local C1=$(count_osts)
4018 if [ $C1 -eq 0 ]; then
4019 start_ost >> /dev/null
4023 echo "original ost count: $C1 (expect > 0)"
4024 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
4025 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
4026 local C2=$(count_osts)
4027 echo "after mdt writeconf count: $C2 (expect 0)"
4028 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
4029 echo "OST start without writeconf should fail:"
4030 start_ost >> /dev/null &&
4031 error "OST start without writeconf didn't fail"
4032 echo "OST start with writeconf should succeed:"
4033 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
4034 local C3=$(count_osts)
4035 echo "after ost writeconf count: $C3 (expect 1)"
4036 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
4037 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
4038 local C4=$(count_osts)
4039 echo "after ost2 writeconf count: $C4 (expect 2)"
4040 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
4041 stop_ost2 >> /dev/null
4042 cleanup_nocli >> /dev/null
4043 #writeconf to remove all ost2 traces for subsequent tests
4044 writeconf_or_reformat
4046 run_test 59 "writeconf mount option"
4048 test_60() { # LU-471
4051 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4052 skip "Only applicable to ldiskfs-based MDTs"
4056 for num in $(seq $MDSCOUNT); do
4057 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
4058 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
4059 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4063 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
4064 [ ${PIPESTATUS[0]} -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
4066 # MDT default has dirdata feature
4067 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
4068 # we disable uninit_bg feature
4069 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
4070 # we set stride extended options
4071 echo $dump | grep stride > /dev/null || error "stride is not set"
4074 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
4079 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] ||
4080 { skip "Need MDS version at least 2.1.53"; return 0; }
4082 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
4083 ! large_xattr_enabled; then
4086 for num in $(seq $MDSCOUNT); do
4087 do_facet mds${num} $TUNE2FS -O large_xattr \
4088 $(mdsdevname $num) ||
4089 error "tune2fs on mds $num failed"
4093 setup_noconfig || error "setting up the filesystem failed"
4094 client_up || error "starting client failed"
4096 local file=$DIR/$tfile
4097 touch $file || error "touch $file failed"
4099 local large_value="$(generate_string $(max_xattr_size))"
4100 local small_value="bar"
4102 local name="trusted.big"
4103 log "save large xattr $name on $file"
4104 setfattr -n $name -v $large_value $file ||
4105 error "saving $name on $file failed"
4107 local new_value=$(get_xattr_value $name $file)
4108 [[ "$new_value" != "$large_value" ]] &&
4109 error "$name different after saving"
4111 log "shrink value of $name on $file"
4112 setfattr -n $name -v $small_value $file ||
4113 error "shrinking value of $name on $file failed"
4115 new_value=$(get_xattr_value $name $file)
4116 [[ "$new_value" != "$small_value" ]] &&
4117 error "$name different after shrinking"
4119 log "grow value of $name on $file"
4120 setfattr -n $name -v $large_value $file ||
4121 error "growing value of $name on $file failed"
4123 new_value=$(get_xattr_value $name $file)
4124 [[ "$new_value" != "$large_value" ]] &&
4125 error "$name different after growing"
4127 log "check value of $name on $file after remounting MDS"
4129 new_value=$(get_xattr_value $name $file)
4130 [[ "$new_value" != "$large_value" ]] &&
4131 error "$name different after remounting MDS"
4133 log "remove large xattr $name from $file"
4134 setfattr -x $name $file || error "removing $name from $file failed"
4137 stopall || error "stopping for e2fsck run"
4138 for num in $(seq $MDSCOUNT); do
4139 run_e2fsck $(facet_active_host mds$num) \
4140 $(mdsdevname $num) "-y" ||
4141 error "e2fsck MDT$num failed"
4143 setup_noconfig || error "remounting the filesystem failed"
4146 # need to delete this file to avoid problems in other tests
4148 stopall || error "stopping systems to turn off large_xattr"
4150 for num in $(seq $MDSCOUNT); do
4151 do_facet mds${num} $TUNE2FS -O ^large_xattr \
4152 $(mdsdevname $num) ||
4153 error "tune2fs on mds $num failed"
4157 run_test 61 "large xattr"
4160 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4161 skip "Only applicable to ldiskfs-based MDTs"
4166 local mdsdev=$(mdsdevname 1)
4167 local ostdev=$(ostdevname 1)
4169 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] ||
4170 { skip "Need MDS version at least 2.2.51"; return 0; }
4172 echo "disable journal for mds"
4173 do_facet mds $TUNE2FS -O ^has_journal $mdsdev || error "tune2fs failed"
4174 start_mds && error "MDT start should fail"
4175 echo "disable journal for ost"
4176 do_facet ost1 $TUNE2FS -O ^has_journal $ostdev || error "tune2fs failed"
4177 start_ost && error "OST start should fail"
4178 cleanup || error "cleanup failed with rc $?"
4181 run_test 62 "start with disabled journal"
4184 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4185 skip "Only applicable to ldiskfs-based MDTs"
4189 local inode_slab=$(do_facet $SINGLEMDS \
4190 "awk '/ldiskfs_inode_cache/ { print \\\$5 }' /proc/slabinfo")
4191 if [ -z "$inode_slab" ]; then
4192 skip "ldiskfs module has not been loaded"
4196 echo "$inode_slab ldisk inodes per page"
4197 [ "$inode_slab" -ge "3" ] ||
4198 error "ldisk inode size is too big, $inode_slab objs per page"
4201 run_test 63 "Verify each page can at least hold 3 ldisk inodes"
4204 start_mds || error "unable to start MDS"
4205 start_ost || error "Unable to start OST1"
4206 start_ost2 || error "Unable to start second ost"
4207 mount_client $MOUNT || error "Unable to mount client"
4208 stop_ost2 || error "Unable to stop second ost"
4210 $LFS df --lazy || error "lfs df failed"
4211 umount_client $MOUNT -f || error “unmount $MOUNT failed”
4212 cleanup_nocli || error "cleanup_nocli failed with $?"
4213 #writeconf to remove all ost2 traces for subsequent tests
4214 writeconf_or_reformat
4216 run_test 64 "check lfs df --lazy "
4218 test_65() { # LU-2237
4219 # Currently, the test is only valid for ldiskfs backend
4220 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4221 skip "non-ldiskfs backend" && return
4223 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4224 local brpt=$(facet_mntpt brpt)
4227 if ! do_facet $SINGLEMDS "test -b $devname"; then
4231 stop_mds || error "Unable to stop MDS"
4232 local obj=$(do_facet $SINGLEMDS \
4233 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
4235 if [ -z "$obj" ]; then
4236 # The MDT may be just re-formatted, mount the MDT for the
4237 # first time to guarantee the "last_rcvd" file is there.
4238 start_mds || error "fail to mount the MDS for the first time"
4239 stop_mds || error "Unable to stop MDS"
4242 # remove the "last_rcvd" file
4243 do_facet $SINGLEMDS "mkdir -p $brpt"
4244 do_facet $SINGLEMDS \
4245 "mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
4246 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
4247 do_facet $SINGLEMDS "$UMOUNT $brpt"
4249 # restart MDS, the "last_rcvd" file should be recreated.
4250 start_mds || error "fail to restart the MDS"
4251 stop_mds || error "Unable to stop MDS"
4252 obj=$(do_facet $SINGLEMDS \
4253 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
4254 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
4256 run_test 65 "re-create the lost last_rcvd file when server mount"
4259 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
4260 { skip "Need MGS version at least 2.3.59"; return 0; }
4263 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
4264 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
4266 echo "replace_nids should fail if MDS, OSTs and clients are UP"
4267 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4268 error "replace_nids fail"
4270 umount_client $MOUNT || error "unmounting client failed"
4271 echo "replace_nids should fail if MDS and OSTs are UP"
4272 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4273 error "replace_nids fail"
4275 stop_ost || error "Unable to stop OST1"
4276 echo "replace_nids should fail if MDS is UP"
4277 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4278 error "replace_nids fail"
4280 stop_mds || error "stopping mds failed"
4282 if combined_mgs_mds; then
4283 start_mdt 1 "-o nosvc" ||
4284 error "starting mds with nosvc option failed"
4287 echo "command should accept two parameters"
4288 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
4289 error "command should accept two params"
4291 echo "correct device name should be passed"
4292 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
4293 error "wrong devname"
4295 echo "wrong nids list should not destroy the system"
4296 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
4299 echo "replace OST nid"
4300 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
4301 error "replace nids failed"
4303 echo "command should accept two parameters"
4304 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
4305 error "command should accept two params"
4307 echo "wrong nids list should not destroy the system"
4308 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
4311 echo "replace MDS nid"
4312 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
4313 error "replace nids failed"
4315 if ! combined_mgs_mds ; then
4318 stop_mds || error "Unable to stop MDS"
4322 check_mount || error "error after nid replace"
4323 cleanup || error "cleanup failed"
4326 run_test 66 "replace nids"
4328 test_67() { #LU-2950
4329 local legacy="$TMP/legacy_lnet_config"
4330 local new="$TMP/new_routes_test"
4331 local out="$TMP/config_out_file"
4332 local verify="$TMP/conv_verify"
4333 local verify_conf="$TMP/conf_verify"
4335 # Create the legacy file that will be run through the
4336 # lustre_routes_conversion script
4337 cat <<- LEGACY_LNET_CONFIG > $legacy
4338 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
4339 tcp2 54 10.1.3.2@tcp;
4340 tcp3 10.3.4.3@tcp:3;
4344 # Create the verification file to verify the output of
4345 # lustre_routes_conversion script against.
4346 cat <<- VERIFY_LNET_CONFIG > $verify
4347 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
4348 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
4349 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
4350 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
4351 tcp4: { gateway: 10.3.3.4@tcp }
4354 # Create the verification file to verify the output of
4355 # lustre_routes_config script against
4356 cat <<- VERIFY_LNET_CONFIG > $verify_conf
4357 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
4358 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
4359 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
4360 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
4361 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
4364 lustre_routes_conversion $legacy $new > /dev/null
4365 if [ -f $new ]; then
4366 # verify the conversion output
4367 cmp -s $new $verify > /dev/null
4368 if [ $? -eq 1 ]; then
4369 error "routes conversion failed"
4372 lustre_routes_config --dry-run --verbose $new > $out
4373 # check that the script succeeded
4374 cmp -s $out $verify_conf > /dev/null
4375 if [ $? -eq 1 ]; then
4376 error "routes config failed"
4379 error "routes conversion test failed"
4381 # remove generated files
4382 rm -f $new $legacy $verify $verify_conf $out
4384 run_test 67 "test routes conversion and configuration"
4392 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.4.53) ] ||
4393 { skip "Need MDS version at least 2.4.53"; return 0; }
4395 umount_client $MOUNT || error "umount client failed"
4397 start_mdt 1 || error "MDT start failed"
4398 start_ost || error "Unable to start OST1"
4400 # START-END - the sequences we'll be reserving
4401 START=$(do_facet $SINGLEMDS \
4402 $LCTL get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
4403 END=$((START + (1 << 30)))
4404 do_facet $SINGLEMDS \
4405 $LCTL set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
4407 # reset the sequences MDT0000 has already assigned
4408 do_facet $SINGLEMDS \
4409 $LCTL set_param seq.srv*MDT0000.space=clear
4411 # remount to let the client allocate new sequence
4412 mount_client $MOUNT || error "mount client failed"
4414 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4415 do_facet $SINGLEMDS \
4416 $LCTL get_param seq.srv*MDT0000.space
4417 $LFS path2fid $DIR/$tfile
4419 local old_ifs="$IFS"
4421 fid=($($LFS path2fid $DIR/$tfile))
4425 if [[ $seq < $END ]]; then
4426 error "used reserved sequence $seq?"
4428 cleanup || error "cleanup failed with $?"
4430 run_test 68 "be able to reserve specific sequences in FLDB"
4432 # Test 69: is about the total number of objects ever created on an OST.
4433 # so that when it is reformatted the normal MDS->OST orphan recovery won't
4434 # just "precreate" the missing objects. In the past it might try to recreate
4435 # millions of objects after an OST was reformatted
4437 local server_version=$(lustre_version_code $SINGLEMDS)
4439 [[ $server_version -lt $(version_code 2.4.2) ]] &&
4440 skip "Need MDS version at least 2.4.2" && return
4442 [[ $server_version -ge $(version_code 2.4.50) ]] &&
4443 [[ $server_version -lt $(version_code 2.5.0) ]] &&
4444 skip "Need MDS version at least 2.5.0" && return
4447 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4449 # use OST0000 since it probably has the most creations
4450 local OSTNAME=$(ostname_from_index 0)
4451 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
4452 local last_id=$(do_facet mds1 $LCTL get_param -n \
4453 osc.$mdtosc_proc1.prealloc_last_id)
4455 # Want to have OST LAST_ID over 5 * OST_MAX_PRECREATE to
4456 # verify that the LAST_ID recovery is working properly. If
4457 # not, then the OST will refuse to allow the MDS connect
4458 # because the LAST_ID value is too different from the MDS
4459 #define OST_MAX_PRECREATE=20000
4460 local ost_max_pre=20000
4461 local num_create=$(( ost_max_pre * 5 + 1 - last_id))
4463 # If the LAST_ID is already over 5 * OST_MAX_PRECREATE, we don't
4464 # need to create any files. So, skip this section.
4465 if [ $num_create -gt 0 ]; then
4466 # Check the number of inodes available on OST0
4468 local ifree=$($LFS df -i $MOUNT | awk '/OST0000/ { print $4 }')
4469 log "On OST0, $ifree inodes available. Want $num_create."
4471 $SETSTRIPE -i 0 $DIR/$tdir ||
4472 error "$SETSTRIPE -i 0 $DIR/$tdir failed"
4473 if [ $ifree -lt 10000 ]; then
4474 files=$(( ifree - 50 ))
4479 local j=$((num_create / files + 1))
4480 for i in $(seq 1 $j); do
4481 createmany -o $DIR/$tdir/$tfile-$i- $files ||
4482 error "createmany fail create $files files: $?"
4483 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
4484 error "unlinkmany failed unlink $files files"
4488 # delete all of the files with objects on OST0 so the
4489 # filesystem is not inconsistent later on
4490 $LFS find $MOUNT --ost 0 -print0 | xargs -0 rm
4492 umount_client $MOUNT || error "umount client failed"
4493 stop_ost || error "OST0 stop failure"
4494 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
4495 $(ostdevname 1) $(ostvdevname 1) ||
4496 error "reformat and replace $ostdev failed"
4497 start_ost || error "OST0 restart failure"
4498 wait_osc_import_state mds ost FULL
4500 mount_client $MOUNT || error "mount client failed"
4501 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
4502 local idx=$($GETSTRIPE -i $DIR/$tdir/$tfile-last)
4503 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
4505 local iused=$($LFS df -i $MOUNT | awk '/OST0000/ { print $3 }')
4506 log "On OST0, $iused used inodes"
4507 [ $iused -ge $((ost_max_pre/2 + 1000)) ] &&
4508 error "OST replacement created too many inodes; $iused"
4509 cleanup || error "cleanup failed with $?"
4511 run_test 69 "replace an OST with the same index"
4514 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4517 cleanup || error "cleanup failed with $?"
4519 start_mdt 1 || error "MDT0 start fail"
4521 start_ost || error "OST0 start fail"
4522 for num in $(seq 2 $MDSCOUNT); do
4523 start_mdt $num || return
4526 mount_client $MOUNT || error "mount client fails"
4528 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
4530 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4531 error "create remote dir fail"
4533 rm -rf $DIR/$tdir || error "delete dir fail"
4534 cleanup || error "cleanup failed with $?"
4536 run_test 70a "start MDT0, then OST, then MDT1"
4539 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4542 start_ost || error "OST0 start fail"
4544 start_mds || error "MDS start fail"
4546 mount_client $MOUNT || error "mount client fails"
4548 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
4550 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4551 error "create remote dir fail"
4553 rm -rf $DIR/$tdir || error "delete dir fail"
4555 cleanup || error "cleanup failed with $?"
4557 run_test 70b "start OST, MDT1, MDT0"
4560 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4563 start_mds || error "MDS start fail"
4564 start_ost || error "OST0 start fail"
4566 mount_client $MOUNT || error "mount client fails"
4567 stop_mdt 1 || error "MDT1 start fail"
4569 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
4570 echo "deactivate $mdc_for_mdt1"
4571 $LCTL --device $mdc_for_mdt1 deactivate ||
4572 error "set $mdc_for_mdt1 deactivate failed"
4574 mkdir $DIR/$tdir && error "mkdir succeed"
4576 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4577 error "create remote dir succeed"
4579 cleanup || error "cleanup failed with $?"
4581 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
4584 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4587 start_mds || error "MDS start fail"
4588 start_ost || error "OST0 start fail"
4590 mount_client $MOUNT || error "mount client fails"
4592 stop_mdt 2 || error "MDT1 start fail"
4594 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
4596 echo "deactivate $mdc_for_mdt2"
4597 $LCTL --device $mdc_for_mdt2 deactivate ||
4598 error "set $mdc_for_mdt2 deactivate failed"
4600 mkdir $DIR/$tdir || error "mkdir fail"
4601 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4602 error "create remote dir succeed"
4604 rm -rf $DIR/$tdir || error "delete dir fail"
4606 cleanup || error "cleanup failed with $?"
4608 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
4611 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4612 if combined_mgs_mds; then
4613 skip "needs separate MGS/MDT" && return
4617 start_mdt 1 || error "MDT0 start fail"
4618 start_ost || error "OST0 start fail"
4619 for num in $(seq 2 $MDSCOUNT); do
4620 start_mdt $num || return
4623 start_ost2 || error "OST1 start fail"
4625 mount_client $MOUNT || error "mount client fails"
4627 mkdir $DIR/$tdir || error "mkdir fail"
4628 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4629 error "create remote dir succeed"
4631 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4632 rm -rf $DIR/$tdir || error "delete dir fail"
4634 umount_client $MOUNT || error "umount_client failed"
4635 stop_mds || error "MDS stop fail"
4636 stop_ost || error "OST0 stop fail"
4637 stop_ost2 || error "OST1 stop fail"
4639 run_test 71a "start MDT0 OST0, MDT1, OST1"
4642 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4643 if combined_mgs_mds; then
4644 skip "needs separate MGS/MDT" && return
4648 for num in $(seq 2 $MDSCOUNT); do
4649 start_mdt $num || return
4651 start_ost || error "OST0 start fail"
4652 start_mdt 1 || error "MDT0 start fail"
4653 start_ost2 || error "OST1 start fail"
4655 mount_client $MOUNT || error "mount client fails"
4657 mkdir $DIR/$tdir || error "mkdir fail"
4658 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4659 error "create remote dir succeed"
4661 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4662 rm -rf $DIR/$tdir || error "delete dir fail"
4664 umount_client $MOUNT || error "umount_client failed"
4665 stop_mds || error "MDT0 stop fail"
4666 stop_ost || error "OST0 stop fail"
4667 stop_ost2 || error "OST1 stop fail"
4669 run_test 71b "start MDT1, OST0, MDT0, OST1"
4672 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4673 if combined_mgs_mds; then
4674 skip "needs separate MGS/MDT" && return
4678 start_ost || error "OST0 start fail"
4679 start_ost2 || error "OST1 start fail"
4680 for num in $(seq 2 $MDSCOUNT); do
4681 start_mdt $num || return
4683 start_mdt 1 || error "MDT0 start fail"
4685 mount_client $MOUNT || error "mount client fails"
4687 mkdir $DIR/$tdir || error "mkdir fail"
4688 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4689 error "create remote dir succeed"
4691 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4692 rm -rf $DIR/$tdir || error "delete dir fail"
4694 umount_client $MOUNT || error "umount_client failed"
4695 stop_mds || error "MDS stop fail"
4696 stop_ost || error "OST0 stop fail"
4697 stop_ost2 || error "OST1 stop fail"
4700 run_test 71c "start OST0, OST1, MDT1, MDT0"
4703 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4704 if combined_mgs_mds; then
4705 skip "needs separate MGS/MDT" && return
4709 start_ost || error "OST0 start fail"
4710 for num in $(seq 2 $MDSCOUNT); do
4711 start_mdt $num || return
4713 start_mdt 1 || error "MDT0 start fail"
4714 start_ost2 || error "OST1 start fail"
4716 mount_client $MOUNT || error "mount client fails"
4718 mkdir $DIR/$tdir || error "mkdir fail"
4719 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4720 error "create remote dir succeed"
4722 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4723 rm -rf $DIR/$tdir || error "delete dir fail"
4725 umount_client $MOUNT || error "umount_client failed"
4726 stop_mds || error "MDS stop fail"
4727 stop_ost || error "OST0 stop fail"
4728 stop_ost2 || error "OST1 stop fail"
4731 run_test 71d "start OST0, MDT1, MDT0, OST1"
4734 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4735 if combined_mgs_mds; then
4736 skip "needs separate MGS/MDT" && return
4740 start_ost || error "OST0 start fail"
4741 for num in $(seq 2 $MDSCOUNT); do
4742 start_mdt $num || return
4744 start_ost2 || error "OST1 start fail"
4745 start_mdt 1 || error "MDT0 start fail"
4747 mount_client $MOUNT || error "mount client fails"
4749 mkdir $DIR/$tdir || error "mkdir fail"
4750 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4751 error "create remote dir succeed"
4753 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4754 rm -rf $DIR/$tdir || error "delete dir fail"
4756 umount_client $MOUNT || error "umount_client failed"
4757 stop_mds || error "MDS stop fail"
4758 stop_ost || error "OST0 stop fail"
4759 stop_ost2 || error "OST1 stop fail"
4762 run_test 71e "start OST0, MDT1, OST1, MDT0"
4764 test_72() { #LU-2634
4765 local mdsdev=$(mdsdevname 1)
4766 local ostdev=$(ostdevname 1)
4767 local cmd="$E2FSCK -fnvd $mdsdev"
4770 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4771 skip "ldiskfs only test" && return
4773 #tune MDT with "-O extents"
4775 for num in $(seq $MDSCOUNT); do
4776 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
4777 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4778 error "add mds $num failed"
4779 do_facet mds${num} "$TUNE2FS -O extents $(mdsdevname $num)" ||
4780 error "$TUNE2FS failed on mds${num}"
4783 add ost1 $(mkfs_opts ost1 $ostdev) --reformat $ostdev ||
4784 error "add $ostdev failed"
4785 start_mgsmds || error "start mds failed"
4786 start_ost || error "start ost failed"
4787 mount_client $MOUNT || error "mount client failed"
4789 #create some short symlinks
4790 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4791 createmany -o $DIR/$tdir/$tfile-%d $fn
4792 echo "create $fn short symlinks"
4793 for i in $(seq -w 1 $fn); do
4794 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
4799 umount_client $MOUNT || error "umount client failed"
4800 stop_mds || error "stop mds failed"
4801 stop_ost || error "stop ost failed"
4804 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
4806 run_test 72 "test fast symlink with extents flag enabled"
4808 test_73() { #LU-3006
4810 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4811 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
4812 error "1st tunefs failed"
4813 start_mgsmds || error "start mds failed"
4814 start_ost || error "start ost failed"
4815 mount_client $MOUNT || error "mount client failed"
4816 $LCTL get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
4817 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
4818 umount_client $MOUNT || error "umount client failed"
4822 run_test 73 "failnode to update from mountdata properly"
4824 test_75() { # LU-2374
4825 [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] &&
4826 skip "Need MDS version at least 2.4.1" && return
4829 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
4830 --reformat $(mdsdevname 1) $(mdsvdevname 1)"
4831 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
4832 --reformat $(ostdevname 1) $(ostvdevname 1)"
4834 #check with default parameters
4835 add mds1 $opts_mds || error "add mds1 failed for default params"
4836 add ost1 $opts_ost || error "add ost1 failed for default params"
4838 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
4839 opts_mds=$(echo $opts_mds |
4840 sed -e "s/--index=$index/--index=$index --mdt/")
4841 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
4842 opts_ost=$(echo $opts_ost |
4843 sed -e "s/--index=$index/--index=$index --ost/")
4845 add mds1 $opts_mds || error "add mds1 failed for new params"
4846 add ost1 $opts_ost || error "add ost1 failed for new params"
4849 run_test 75 "The order of --index should be irrelevant"
4852 [[ $(lustre_version_code mgs) -ge $(version_code 2.4.52) ]] ||
4853 { skip "Need MDS version at least 2.4.52" && return 0; }
4855 local MDMB_PARAM="osc.*.max_dirty_mb"
4856 echo "Change MGS params"
4857 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
4859 echo "max_dirty_mb: $MAX_DIRTY_MB"
4860 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB + MAX_DIRTY_MB))
4861 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
4862 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
4863 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
4864 head -1" $NEW_MAX_DIRTY_MB
4865 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
4866 echo "$MAX_DIRTY_MB"
4867 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
4868 error "error while apply max_dirty_mb"
4870 echo "Check the value is stored after remount"
4873 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
4874 head -1" $NEW_MAX_DIRTY_MB
4875 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
4876 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
4877 error "max_dirty_mb is not saved after remount"
4879 echo "Change OST params"
4880 CLIENT_PARAM="obdfilter.*.client_cache_count"
4881 local CLIENT_CACHE_COUNT
4882 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4884 echo "client_cache_count: $CLIENT_CACHE_COUNT"
4885 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
4886 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
4887 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
4888 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
4889 head -1" $NEW_CLIENT_CACHE_COUNT
4890 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4892 echo "$CLIENT_CACHE_COUNT"
4893 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
4894 error "error while apply client_cache_count"
4896 echo "Check the value is stored after remount"
4899 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
4900 head -1" $NEW_CLIENT_CACHE_COUNT
4901 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4903 echo "$CLIENT_CACHE_COUNT"
4904 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
4905 error "client_cache_count is not saved after remount"
4908 run_test 76a "set permanent params set_param -P"
4910 test_76b() { # LU-4783
4911 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.57) ]] ||
4912 { skip "Need MGS version at least 2.5.57" && return 0; }
4915 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
4916 error "start params log failed"
4919 run_test 76b "verify params log setup correctly"
4921 test_77() { # LU-3445
4922 local server_version=$(lustre_version_code $SINGLEMDS)
4924 [[ $server_version -ge $(version_code 2.2.60) ]] &&
4925 [[ $server_version -le $(version_code 2.4.0) ]] &&
4926 skip "Need MDS version < 2.2.60 or > 2.4.0" && return
4928 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
4929 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
4930 skip_env "mixed loopback and real device not working" && return
4933 local fs2mdsdev=$(mdsdevname 1_2)
4934 local fs2ostdev=$(ostdevname 1_2)
4935 local fs2mdsvdev=$(mdsvdevname 1_2)
4936 local fs2ostvdev=$(ostvdevname 1_2)
4937 local fsname=test1234
4939 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
4941 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
4942 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
4943 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
4944 error "start fs2mds failed"
4946 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
4947 [[ $mgsnid = *,* ]] || mgsnid+=",$mgsnid"
4949 add fs2ost $(mkfs_opts ost1 $fs2ostdev) --mgsnode=$mgsnid \
4950 --failnode=$failnid --fsname=$fsname \
4951 --reformat $fs2ostdev $fs2ostvdev ||
4952 error "add fs2ost failed"
4953 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
4955 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
4956 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
4957 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
4960 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
4963 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ||
4964 $(facet_fstype ost1) != ldiskfs ]] &&
4965 skip "only applicable to ldiskfs-based MDTs and OSTs" && return
4967 # reformat the Lustre filesystem with a smaller size
4968 local saved_MDSSIZE=$MDSSIZE
4969 local saved_OSTSIZE=$OSTSIZE
4970 MDSSIZE=$((MDSSIZE - 20000))
4971 OSTSIZE=$((OSTSIZE - 20000))
4972 reformat || error "(1) reformat Lustre filesystem failed"
4973 MDSSIZE=$saved_MDSSIZE
4974 OSTSIZE=$saved_OSTSIZE
4976 # mount the Lustre filesystem
4977 setup_noconfig || error "(2) setup Lustre filesystem failed"
4980 log "create test files"
4984 mkdir $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
4985 for i in $(seq $num_files); do
4986 file=$MOUNT/$tdir/$tfile-$i
4987 dd if=/dev/urandom of=$file count=1 bs=1M ||
4988 error "(4) create $file failed"
4991 # unmount the Lustre filesystem
4992 cleanup || error "(5) cleanup Lustre filesystem failed"
4994 # run e2fsck on the MDT and OST devices
4995 local mds_host=$(facet_active_host $SINGLEMDS)
4996 local ost_host=$(facet_active_host ost1)
4997 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
4998 local ost_dev=$(ostdevname 1)
5000 run_e2fsck $mds_host $mds_dev "-y"
5001 run_e2fsck $ost_host $ost_dev "-y"
5003 # get the original block count of the MDT and OST filesystems
5004 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
5005 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
5007 # expand the MDT and OST filesystems to the device size
5008 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
5009 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
5011 # run e2fsck on the MDT and OST devices again
5012 run_e2fsck $mds_host $mds_dev "-y"
5013 run_e2fsck $ost_host $ost_dev "-y"
5015 # mount the Lustre filesystem
5019 log "check files after expanding the MDT and OST filesystems"
5020 for i in $(seq $num_files); do
5021 file=$MOUNT/$tdir/$tfile-$i
5022 $CHECKSTAT -t file -s 1048576 $file ||
5023 error "(6) checkstat $file failed"
5027 log "create more files after expanding the MDT and OST filesystems"
5028 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
5029 file=$MOUNT/$tdir/$tfile-$i
5030 dd if=/dev/urandom of=$file count=1 bs=1M ||
5031 error "(7) create $file failed"
5034 # unmount the Lustre filesystem
5035 cleanup || error "(8) cleanup Lustre filesystem failed"
5037 # run e2fsck on the MDT and OST devices
5038 run_e2fsck $mds_host $mds_dev "-y"
5039 run_e2fsck $ost_host $ost_dev "-y"
5041 # get the maximum block count of the MDT and OST filesystems
5042 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
5043 local ost_max_blks=$(get_block_count ost1 $ost_dev)
5045 # get the minimum block count of the MDT and OST filesystems
5046 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
5047 grep minimum | sed -e 's/^.*filesystem: //g')
5048 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
5049 grep minimum | sed -e 's/^.*filesystem: //g')
5051 # shrink the MDT and OST filesystems to a smaller size
5055 if [[ $mds_max_blks -gt $mds_min_blks &&
5056 $mds_max_blks -gt $mds_orig_blks ]]; then
5057 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
5058 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
5059 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
5060 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
5061 error "shrink $SINGLEMDS to $new_blks failed"
5065 if [[ $ost_max_blks -gt $ost_min_blks &&
5066 $ost_max_blks -gt $ost_orig_blks ]]; then
5067 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
5068 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
5069 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
5070 run_resize2fs ost1 $ost_dev $new_blks ||
5071 error "shrink ost1 to $new_blks failed"
5075 # check whether the MDT or OST filesystem was shrunk or not
5077 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
5078 reformat || error "(10) reformat Lustre filesystem failed"
5082 # run e2fsck on the MDT and OST devices again
5083 run_e2fsck $mds_host $mds_dev "-y"
5084 run_e2fsck $ost_host $ost_dev "-y"
5086 # mount the Lustre filesystem again
5090 log "check files after shrinking the MDT and OST filesystems"
5091 for i in $(seq $((num_files + 10))); do
5092 file=$MOUNT/$tdir/$tfile-$i
5093 $CHECKSTAT -t file -s 1048576 $file ||
5094 error "(11) checkstat $file failed"
5097 # unmount and reformat the Lustre filesystem
5098 cleanup || error "(12) cleanup Lustre filesystem failed"
5099 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
5100 reformat || error "(14) reformat Lustre filesystem failed"
5102 run_test 78 "run resize2fs on MDT and OST filesystems"
5104 test_79() { # LU-4227
5105 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.5.59) ]] ||
5106 { skip "Need MDS version at least 2.5.59"; return 0; }
5108 local mdsdev1=$(mdsdevname 1)
5109 local mdsvdev1=$(mdsvdevname 1)
5110 local mdsdev2=$(mdsdevname 2)
5111 local mdsvdev2=$(mdsvdevname 2)
5112 local ostdev1=$(ostdevname 1)
5113 local ostvdev1=$(ostvdevname 1)
5114 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
5115 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
5116 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
5119 # remove --mgs/--mgsnode from mkfs.lustre options
5120 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
5122 mgsnode_opt=$(echo $opts_mds2 |
5123 awk '{ for ( i = 1; i < NF; i++ )
5124 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5125 [ -n $mgsnode_opt ] &&
5126 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
5128 mgsnode_opt=$(echo $opts_ost1 |
5129 awk '{ for ( i = 1; i < NF; i++ )
5130 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5131 [ -n $mgsnode_opt ] &&
5132 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
5134 # -MGS, format a mdt without --mgs option
5135 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
5136 error "Must specify --mgs when formatting mdt combined with mgs"
5138 # +MGS, format a mdt/ost without --mgsnode option
5139 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
5140 > /dev/null || error "start mds1 failed"
5141 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
5142 error "Must specify --mgsnode when formatting a mdt"
5143 add ost1 $opts_ost1 $ostdev1 $ostvdev2 &&
5144 error "Must specify --mgsnode when formatting an ost"
5148 run_test 79 "format MDT/OST without mgs option (should return errors)"
5151 start_mds || error "Failed to start MDT"
5152 start_ost || error "Failed to start OST1"
5153 uuid=$(do_facet ost1 $LCTL get_param -n mgc.*.uuid)
5154 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
5155 do_facet ost1 "$LCTL set_param fail_val=10 fail_loc=0x906"
5156 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x906"
5157 do_facet mgs "$LCTL set_param -n mgs/MGS/evict_client $uuid"
5159 start_ost2 || error "Failed to start OST2"
5161 do_facet ost1 "$LCTL set_param fail_loc=0"
5164 run_test 80 "mgc import reconnect race"
5166 #Save the original values of $OSTCOUNT and $OSTINDEX$i.
5168 local new_ostcount=$1
5169 saved_ostcount=$OSTCOUNT
5170 OSTCOUNT=$new_ostcount
5174 for i in $(seq $OSTCOUNT); do
5176 eval saved_ostindex$i=${!index}
5181 # Restore the original values of $OSTCOUNT and $OSTINDEX$i.
5182 restore_ostindex() {
5187 for i in $(seq $OSTCOUNT); do
5188 index=saved_ostindex$i
5189 eval OSTINDEX$i=${!index}
5191 OSTCOUNT=$saved_ostcount
5196 # The main purpose of this test is to ensure the OST_INDEX_LIST functions as
5197 # expected. This test uses OST_INDEX_LIST to format OSTs with a randomly
5198 # assigned index and ensures we can mount such a formatted file system
5199 test_81() { # LU-4665
5200 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5201 { skip "Need MDS version at least 2.6.54" && return; }
5202 [[ $OSTCOUNT -ge 3 ]] || { skip_env "Need at least 3 OSTs" && return; }
5206 # Each time RANDOM is referenced, a random integer between 0 and 32767
5209 local saved_ostindex1=$OSTINDEX1
5210 for i in 65535 $((RANDOM + 65536)); do
5211 echo -e "\nFormat ost1 with --index=$i, should fail"
5213 if add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat \
5214 $(ostdevname 1) $(ostvdevname 1); then
5215 OSTINDEX1=$saved_ostindex1
5216 error "format ost1 with --index=$i should fail"
5219 OSTINDEX1=$saved_ostindex1
5223 # Format OSTs with random sparse indices.
5224 trap "restore_ostindex" EXIT
5225 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices"
5226 OST_INDEX_LIST=[0,$((RANDOM * 2 % 65533 + 1)),65534] formatall
5228 # Setup and check Lustre filesystem.
5229 start_mgsmds || error "start_mgsmds failed"
5230 for i in $(seq $OSTCOUNT); do
5231 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5232 error "start ost$i failed"
5235 mount_client $MOUNT || error "mount client $MOUNT failed"
5236 check_mount || error "check client $MOUNT failed"
5239 local max_easize=$($LCTL get_param -n llite.*.max_easize)
5240 [[ $max_easize -eq 128 ]] ||
5241 error "max_easize is $max_easize, should be 128 bytes"
5245 run_test 81 "sparse OST indexing"
5247 # Wait OSTs to be active on both client and MDT side.
5249 local cmd="$LCTL get_param -n lov.$FSNAME-clilov-*.target_obd |
5250 awk 'BEGIN {c = 0} /ACTIVE/{c += 1} END {printf \\\"%d\\\", c}'"
5251 wait_update $HOSTNAME "eval $cmd" $OSTCOUNT ||
5252 error "wait_update OSTs up on client failed"
5254 cmd="$LCTL get_param -n lod.$FSNAME-MDT*-*.target_obd | sort -u |
5255 awk 'BEGIN {c = 0} /ACTIVE/{c += 1} END {printf \\\"%d\\\", c}'"
5256 wait_update_facet $SINGLEMDS "eval $cmd" $OSTCOUNT ||
5257 error "wait_update OSTs up on MDT failed"
5260 # Here we exercise the stripe placement functionality on a file system that
5261 # has formatted the OST with a random index. With the file system the following
5262 # functionality is tested:
5264 # 1. Creating a new file with a specific stripe layout.
5266 # 2. Modifiy a existing empty file with a specific stripe layout.
5268 # 3. Ensure we fail to set the stripe layout of a file that already has one.
5270 # 4. If ost-index is defined we need to ensure it is the first entry in the
5271 # ost index list returned by lfs getstripe.
5273 # 5. Lastly ensure this functionality fails with directories.
5274 test_82a() { # LU-4665
5275 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5276 { skip "Need MDS version at least 2.6.54" && return; }
5277 [[ $OSTCOUNT -ge 3 ]] || { skip_env "Need at least 3 OSTs" && return; }
5283 # Format OSTs with random sparse indices.
5287 for i in $(seq $OSTCOUNT); do
5288 index=$((RANDOM * 2))
5289 ost_indices+=" $index"
5291 ost_indices=$(comma_list $ost_indices)
5293 trap "restore_ostindex" EXIT
5294 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5295 OST_INDEX_LIST=[$ost_indices] formatall
5297 # Setup Lustre filesystem.
5298 start_mgsmds || error "start_mgsmds failed"
5299 for i in $(seq $OSTCOUNT); do
5300 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5301 error "start ost$i failed"
5304 mount_client $MOUNT || error "mount client $MOUNT failed"
5307 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5308 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5310 # 1. If the file does not exist, new file will be created
5311 # with specified OSTs.
5312 local file=$DIR/$tdir/$tfile-1
5313 local cmd="$SETSTRIPE -o $ost_indices $file"
5315 eval $cmd || error "$cmd failed"
5316 check_stripe_count $file $OSTCOUNT
5317 check_obdidx $file $ost_indices
5318 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5319 error "write $file failed"
5321 # 2. If the file already exists and is an empty file, the file
5322 # will be attached with specified layout.
5323 file=$DIR/$tdir/$tfile-2
5324 mcreate $file || error "mcreate $file failed"
5325 cmd="$SETSTRIPE -o $ost_indices $file"
5327 eval $cmd || error "$cmd failed"
5328 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5329 error "write $file failed"
5330 check_stripe_count $file $OSTCOUNT
5331 check_obdidx $file $ost_indices
5333 # 3. If the file already has a valid layout attached, the command
5334 # should fail with EBUSY.
5336 eval $cmd && error "stripe is already set on $file, $cmd should fail"
5338 # 4. If [--stripe-index|-i <start_ost_idx>] is used, the index must
5339 # be in the OST indices list.
5340 local start_ost_idx=${ost_indices##*,}
5341 file=$DIR/$tdir/$tfile-3
5342 cmd="$SETSTRIPE -o $ost_indices -i $start_ost_idx $file"
5344 eval $cmd || error "$cmd failed"
5345 check_stripe_count $file $OSTCOUNT
5346 check_obdidx $file $ost_indices
5347 check_start_ost_idx $file $start_ost_idx
5349 file=$DIR/$tdir/$tfile-4
5351 cmd+=" -o $(exclude_items_from_list $ost_indices $start_ost_idx)"
5352 cmd+=" -i $start_ost_idx $file"
5354 eval $cmd && error "index $start_ost_idx should be in $ost_indices"
5356 # 5. Specifying OST indices for directory should fail with ENOSUPP.
5357 local dir=$DIR/$tdir/$tdir
5358 mkdir $dir || error "mkdir $dir failed"
5359 cmd="$SETSTRIPE -o $ost_indices $dir"
5361 eval $cmd && error "$cmd should fail, specifying OST indices" \
5362 "for directory is not supported"
5366 run_test 82a "specify OSTs for file (succeed) or directory (fail)"
5371 # Remove OSTs from a pool and destroy the pool.
5372 destroy_pool $ost_pool || true
5377 # Test 82b is run to ensure that if the user supplies a pool with a specific
5378 # stripe layout that it behaves proprerly. It should fail in the case that
5379 # the supplied OST index list points to OSTs not contained in the user
5381 test_82b() { # LU-4665
5382 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5383 { skip "Need MDS version at least 2.6.54" && return; }
5384 [[ $OSTCOUNT -ge 4 ]] || { skip_env "Need at least 4 OSTs" && return; }
5390 # Format OSTs with random sparse indices.
5394 for i in $(seq $OSTCOUNT); do
5395 index=$((RANDOM * 2))
5396 ost_indices+=" $index"
5398 ost_indices=$(comma_list $ost_indices)
5400 trap "restore_ostindex" EXIT
5401 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5402 OST_INDEX_LIST=[$ost_indices] formatall
5404 # Setup Lustre filesystem.
5405 start_mgsmds || error "start_mgsmds failed"
5406 for i in $(seq $OSTCOUNT); do
5407 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5408 error "start ost$i failed"
5411 mount_client $MOUNT || error "mount client $MOUNT failed"
5413 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5414 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5416 # Create a new pool and add OSTs into it.
5417 local ost_pool=$FSNAME.$TESTNAME
5418 create_pool $ost_pool || error "create OST pool $ost_pool failed"
5421 trap "cleanup_82b" EXIT
5423 local ost_idx_in_list=${ost_indices##*,}
5424 local ost_idx_in_pool=$(exclude_items_from_list $ost_indices \
5427 local ost_targets="$FSNAME-OST["
5428 for i in ${ost_idx_in_pool//,/ }; do
5429 ost_targets=$ost_targets$(printf "%04x," $i)
5431 ost_targets="${ost_targets%,}]"
5433 local ost_targets_uuid=$(for i in ${ost_idx_in_pool//,/ }; \
5434 do printf "$FSNAME-OST%04x_UUID\n" $i; done |
5435 sort -u | tr '\n' ' ')
5437 local cmd="$LCTL pool_add $ost_pool $ost_targets"
5438 do_facet mgs $cmd || error "$cmd failed"
5439 wait_update $HOSTNAME "$LCTL get_param -n lov.$FSNAME-*.pools.$TESTNAME|
5440 sort -u | tr '\n' ' ' " "$ost_targets_uuid" ||
5441 error "wait_update $ost_pool failed"
5442 pool_list $ost_pool || error "list OST pool $ost_pool failed"
5444 # If [--pool|-p <pool_name>] is set with [--ost-list|-o <ost_indices>],
5445 # then the OSTs must be the members of the pool.
5446 local file=$DIR/$tdir/$tfile
5447 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
5449 eval $cmd && error "OST with index $ost_idx_in_list should be" \
5450 "in OST pool $ost_pool"
5452 # Only select OST $ost_idx_in_list from $ost_pool for file.
5453 ost_idx_in_list=${ost_idx_in_pool#*,}
5454 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
5456 eval $cmd || error "$cmd failed"
5457 cmd="$GETSTRIPE $file"
5459 eval $cmd || error "$cmd failed"
5460 check_stripe_count $file 2
5461 check_obdidx $file $ost_idx_in_list
5462 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5463 error "write $file failed"
5467 run_test 82b "specify OSTs for file with --pool and --ost-list options"
5470 [[ $(lustre_version_code ost1) -ge $(version_code 2.6.91) ]] ||
5471 { skip "Need OST version at least 2.6.91" && return 0; }
5472 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
5473 skip "Only applicable to ldiskfs-based MDTs"
5483 ostmnt=$(facet_mntpt ost1)
5484 fstype=$(facet_fstype ost1)
5486 # Mount the OST as an ldiskfs filesystem.
5487 log "mount the OST $dev as a $fstype filesystem"
5488 add ost1 $(mkfs_opts ost1 $dev) $FSTYPE_OPT \
5489 --reformat $dev $dev > /dev/null ||
5490 error "format ost1 error"
5492 if ! test -b $dev; then
5493 mnt_opts=$(csa_add "$OST_MOUNT_OPTS" -o loop)
5495 echo "mnt_opts $mnt_opts"
5496 do_facet ost1 mount -t $fstype $dev \
5498 # Run llverfs on the mounted ldiskfs filesystem.
5499 # It is needed to get ENOSPACE.
5500 log "run llverfs in partial mode on the OST $fstype $ostmnt"
5501 do_rpc_nodes $(facet_host ost1) run_llverfs $ostmnt -vpl \
5502 "no" || error "run_llverfs error on $fstype"
5505 log "unmount the OST $dev"
5508 # Delete file IO_scrub. Later osd_scrub_setup will try to
5509 # create "IO_scrub" but will get ENOSPACE.
5511 echo "start ost1 service on `facet_active_host ost1`"
5512 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
5515 err=$(do_facet ost1 dmesg | grep "VFS: Busy inodes after unmount of")
5516 echo "string err $err"
5517 [ -z "$err" ] || error $err
5520 run_test 83 "ENOSPACE on OST doesn't cause message VFS: \
5521 Busy inodes after unmount ..."
5523 recovery_time_min() {
5524 local CONNECTION_SWITCH_MIN=5
5525 local CONNECTION_SWITCH_INC=5
5526 local CONNECTION_SWITCH_MAX
5527 local RECONNECT_DELAY_MAX
5528 local INITIAL_CONNECT_TIMEOUT
5532 #CONNECTION_SWITCH_MAX=min(50, max($CONNECTION_SWITCH_MIN,$TIMEOUT)
5533 (($CONNECTION_SWITCH_MIN>$TIMEOUT)) && \
5534 max=$CONNECTION_SWITCH_MIN || max=$TIMEOUT
5535 (($max<50)) && CONNECTION_SWITCH_MAX=$max || CONNECTION_SWITCH_MAX=50
5537 #INITIAL_CONNECT_TIMEOUT = max(CONNECTION_SWITCH_MIN, \
5539 TO_20=$(($TIMEOUT/20))
5540 (($CONNECTION_SWITCH_MIN>$TO_20)) && \
5541 INITIAL_CONNECT_TIMEOUT=$CONNECTION_SWITCH_MIN || \
5542 INITIAL_CONNECT_TIMEOUT=$TO_20
5544 RECONNECT_DELAY_MAX=$(($CONNECTION_SWITCH_MAX+$CONNECTION_SWITCH_INC+ \
5545 $INITIAL_CONNECT_TIMEOUT))
5546 echo $((2*$RECONNECT_DELAY_MAX))
5550 local facet=$SINGLEMDS
5551 local num=$(echo $facet | tr -d "mds")
5552 local dev=$(mdsdevname $num)
5553 local time_min=$(recovery_time_min)
5554 local recovery_duration
5555 local completed_clients
5556 local correct_clients
5560 echo "start mds service on $(facet_active_host $facet)"
5562 "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
5563 error "start MDS failed"
5565 start_ost || error "start OST0000 failed"
5566 start_ost2 || error "start OST0001 failed"
5568 echo "recovery_time=$time_min, timeout=$TIMEOUT, wrap_up=$wrap_up"
5570 mount_client $MOUNT1 || error "mount $MOUNT1 failed"
5571 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
5572 # make sure new superblock labels are sync'd before disabling writes
5576 replay_barrier $SINGLEMDS
5577 createmany -o $DIR1/$tfile-%d 1000
5579 # We need to catch the end of recovery window to extend it.
5580 # Skip 5 requests and add delay to request handling.
5581 #define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP
5582 do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5"
5584 facet_failover --fsck $SINGLEMDS || error "failover: $?"
5587 echo "recovery status"
5588 do_facet $SINGLEMDS \
5589 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status"
5591 recovery_duration=$(do_facet $SINGLEMDS \
5592 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
5593 awk '/recovery_duration/ { print $2 }')
5594 (( $recovery_duration > $time_min + $wrap_up )) &&
5595 error "recovery_duration > recovery_time_hard + wrap up"
5596 completed_clients=$(do_facet $SINGLEMDS \
5597 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
5598 awk '/completed_clients/ { print $2 }')
5600 correct_clients="$MDSCOUNT/$((MDSCOUNT+1))"
5601 [ "$completed_clients" = "${correct_clients}" ] ||
5602 error "$completed_clients != $correct_clients"
5604 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
5605 umount_client $MOUNT1
5606 umount_client $MOUNT2
5612 run_test 84 "check recovery_hard_time"
5615 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.55) ]] ||
5616 { skip "Need OST version at least 2.7.55" && return 0; }
5617 ##define OBD_FAIL_OSD_OST_EA_FID_SET 0x197
5618 do_facet ost1 "lctl set_param fail_loc=0x197"
5622 run_test 85 "osd_ost init: fail ea_fid_set"
5625 [ "$(facet_fstype ost1)" = "zfs" ] &&
5626 skip "LU-6442: no such mkfs params for ZFS OSTs" && return
5628 local OST_OPTS="$(mkfs_opts ost1 $(ostdevname 1)) \
5629 --reformat $(ostdevname 1) $(ostvdevname 1)"
5632 local OLDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
5633 awk '/Flex block group size: / { print $NF; exit; }')
5636 if [[ ${!opts} != *mkfsoptions* ]]; then
5637 eval opts=\"${!opts} \
5638 --mkfsoptions='\\\"-O flex_bg -G $NEWSIZE\\\"'\"
5640 val=${!opts//--mkfsoptions=\\\"/ \
5641 --mkfsoptions=\\\"-O flex_bg -G $NEWSIZE }
5645 echo "params: $opts"
5647 add ost1 $opts || error "add ost1 failed with new params"
5649 local FOUNDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
5650 awk '/Flex block group size: / { print $NF; exit; }')
5652 [[ $FOUNDSIZE == $NEWSIZE ]] ||
5653 error "Flex block group size: $FOUNDSIZE, expected: $NEWSIZE"
5656 run_test 86 "Replacing mkfs.lustre -G option"
5658 test_87() { #LU-6544
5659 [[ $(lustre_version_code $SINGLEMDS1) -ge $(version_code 2.7.56) ]] ||
5660 { skip "Need MDS version at least 2.7.56" && return; }
5661 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ]] &&
5662 { skip "Only applicable to ldiskfs-based MDTs" && return; }
5663 [[ $OSTCOUNT -gt 69 ]] &&
5664 { skip "Ignore wide striping situation" && return; }
5666 local mdsdev=$(mdsdevname 1)
5667 local mdsvdev=$(mdsvdevname 1)
5668 local file=$DIR/$tfile
5669 local mntpt=$(facet_mntpt $SINGLEMDS)
5670 local used_xattr_blk=0
5671 local inode_size=${1:-512}
5673 local xtest="trusted.test"
5678 #Please see LU-6544 for MDT inode size calculation
5679 if [ $OSTCOUNT -gt 26 ]; then
5681 elif [ $OSTCOUNT -gt 5 ]; then
5684 left_size=$(expr $inode_size - \
5687 32 - $OSTCOUNT \* 24 - 16 - 3 - \
5689 24 - 18 - $(expr length $tfile) - 16 - 4)
5690 if [ $left_size -le 0 ]; then
5691 echo "No space($left_size) is expected in inode."
5692 echo "Try 1-byte xattr instead to verify this."
5695 echo "Estimate: at most $left_size-byte space left in inode."
5701 add mds1 $(mkfs_opts mds1 ${mdsdev}) --stripe-count-hint=$OSTCOUNT \
5702 --reformat $mdsdev $mdsvdev || error "add mds1 failed"
5703 start_mdt 1 > /dev/null || error "start mdt1 failed"
5704 for i in $(seq $OSTCOUNT); do
5705 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS > /dev/null ||
5706 error "start ost$i failed"
5708 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
5709 check_mount || error "check client $MOUNT failed"
5712 $SETSTRIPE -c -1 $file || error "$SETSTRIPE -c -1 $file failed"
5713 $GETSTRIPE $file || error "$GETSTRIPE $file failed"
5714 i=$($GETSTRIPE -c $file)
5715 if [ $i -ne $OSTCOUNT ]; then
5716 left_size=$(expr $left_size + $(expr $OSTCOUNT - $i) \* 24)
5717 echo -n "Since only $i out $OSTCOUNT OSTs are used, "
5718 echo -n "the expected left space is changed to "
5719 echo "$left_size bytes at most."
5721 value=$(generate_string $left_size)
5722 setfattr -n $xtest -v $value $file
5723 orig=$(get_xattr_value $xtest $file)
5724 [[ "$orig" != "$value" ]] && error "$xtest changed"
5726 #Verify if inode has some expected space left
5727 umount $MOUNT > /dev/null || error "umount $MOUNT failed"
5728 stop_mdt 1 > /dev/null || error "stop mdt1 failed"
5729 mount_ldiskfs $SINGLEMDS || error "mount -t ldiskfs $SINGLEMDS failed"
5731 do_facet $SINGLEMDS ls -sal $mntpt/ROOT/$tfile
5732 used_xattr_blk=$(do_facet $SINGLEMDS ls -s $mntpt/ROOT/$tfile |
5734 [[ $used_xattr_blk -eq 0 ]] &&
5735 error "Please check MDS inode size calculation: \
5736 more than $left_size-byte space left in inode."
5737 echo "Verified: at most $left_size-byte space left in inode."
5741 run_test 87 "check if MDT inode can hold EAs with N stripes properly"
5744 # $2 (optional) value of max_mod_rpcs_in_flight to set
5745 check_max_mod_rpcs_in_flight() {
5753 idx=$(printf "%04x" $($LFS getdirstripe -i $dir))
5754 facet="mds$((0x$idx + 1))"
5756 if [ -z "$mmr" ]; then
5757 # get value of max_mod_rcps_in_flight
5758 mmr=$($LCTL get_param -n \
5759 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) ||
5760 error "Unable to get max_mod_rpcs_in_flight"
5761 echo "max_mod_rcps_in_flight is $mmr"
5763 # set value of max_mod_rpcs_in_flight
5765 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight=$mmr ||
5766 error "Unable to set max_mod_rpcs_in_flight to $mmr"
5767 echo "max_mod_rpcs_in_flight set to $mmr"
5770 # create mmr+1 files
5771 echo "creating $((mmr + 1)) files ..."
5773 for i in $(seq $((mmr + 1))); do
5779 # consumes mmr-1 modify RPC slots
5780 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
5781 # drop requests on MDT so that RPC slots are consumed
5782 # during all the request resend interval
5783 do_facet $facet "$LCTL set_param fail_loc=0x159"
5784 echo "launch $((mmr - 1)) chmod in parallel ..."
5785 for i in $(seq $((mmr - 1))); do
5786 chmod 0600 $dir/file-$i &
5790 # send one additional modify RPC
5791 do_facet $facet "$LCTL set_param fail_loc=0"
5792 echo "launch 1 additional chmod in parallel ..."
5793 chmod 0600 $dir/file-$mmr &
5796 # check this additional modify RPC get a modify RPC slot
5797 # and succeed its operation
5798 checkstat -vp 0600 $dir/file-$mmr ||
5799 error "Unable to send $mmr modify RPCs in parallel"
5804 # consumes mmr modify RPC slots
5805 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
5806 # drop requests on MDT so that RPC slots are consumed
5807 # during all the request resend interval
5808 do_facet $facet "$LCTL set_param fail_loc=0x159"
5809 echo "launch $mmr chmod in parallel ..."
5810 for i in $(seq $mmr); do
5811 chmod 0666 $dir/file-$i &
5815 # send one additional modify RPC
5816 do_facet $facet "$LCTL set_param fail_loc=0"
5817 echo "launch 1 additional chmod in parallel ..."
5818 chmod 0666 $dir/file-$((mmr + 1)) &
5821 # check this additional modify RPC blocked getting a modify RPC slot
5822 checkstat -vp 0644 $dir/file-$((mmr + 1)) ||
5823 error "Unexpectedly send $(($mmr + 1)) modify RPCs in parallel"
5829 if ! combined_mgs_mds ; then
5834 [[ $($LCTL get_param mdc.*.import |
5835 grep "connect_flags:.*multi_mod_rpc") ]] ||
5836 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
5838 # check default value
5839 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5840 check_max_mod_rpcs_in_flight $DIR/$tdir
5844 run_test 90a "check max_mod_rpcs_in_flight is enforced"
5854 [[ $($LCTL get_param mdc.*.import |
5855 grep "connect_flags:.*multi_mod_rpc") ]] ||
5856 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
5859 # update max_mod_rpcs_in_flight
5860 $LFS mkdir -c1 $DIR/${tdir}1 || error "mkdir $DIR/${tdir}1 failed"
5861 check_max_mod_rpcs_in_flight $DIR/${tdir}1 1
5864 # check client is able to send multiple modify RPCs in paralell
5865 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
5866 grep -c "multi_mod_rpcs")
5867 if [ "$tmp" -ne $MDSCOUNT ]; then
5868 echo "Client not able to send multiple modify RPCs in parallel"
5873 # update max_mod_rpcs_in_flight
5874 $LFS mkdir -c1 $DIR/${tdir}2 || error "mkdir $DIR/${tdir}2 failed"
5875 check_max_mod_rpcs_in_flight $DIR/${tdir}2 5
5878 $LFS mkdir -c1 $DIR/${tdir}3 || error "mkdir $DIR/${tdir}3 failed"
5879 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/${tdir}3))
5880 facet="mds$((0x$idx + 1))"
5882 # save MDT max_mod_rpcs_per_client
5883 mmrpc=$(do_facet $facet \
5884 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
5886 # update max_mod_rpcs_in_flight
5887 umount_client $MOUNT
5889 "echo 16 > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
5891 $LCTL set_param mdc.$FSNAME-MDT$idx-mdc-*.max_rpcs_in_flight=17
5892 check_max_mod_rpcs_in_flight $DIR/${tdir}3 16
5894 # restore MDT max_mod_rpcs_per_client initial value
5896 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
5898 rm -rf $DIR/${tdir}?
5901 run_test 90b "check max_mod_rpcs_in_flight is enforced after update"
5910 [[ $($LCTL get_param mdc.*.import |
5911 grep "connect_flags:.*multi_mod_rpc") ]] ||
5912 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
5914 # check client is able to send multiple modify RPCs in paralell
5915 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
5916 grep -c "multi_mod_rpcs")
5917 if [ "$tmp" -ne $MDSCOUNT ]; then
5918 skip "Client not able to send multiple modify RPCs in parallel"
5923 # get max_rpcs_in_flight value
5924 mrif=$($LCTL get_param -n mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
5925 echo "max_rpcs_in_flight is $mrif"
5927 # get MDT max_mod_rpcs_per_client
5928 mmrpc=$(do_facet mds1 \
5929 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
5930 echo "max_mod_rpcs_per_client is $mmrpc"
5933 # attempt to set max_mod_rpcs_in_flight to max_rpcs_in_flight value
5934 # prerequisite: set max_mod_rpcs_per_client to max_rpcs_in_flight value
5935 umount_client $MOUNT
5937 "echo $mrif > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
5941 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif &&
5942 error "set max_mod_rpcs_in_flight to $mrif should fail"
5944 umount_client $MOUNT
5946 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
5950 # attempt to set max_mod_rpcs_in_flight to max_mod_rpcs_per_client+1
5951 # prerequisite: set max_rpcs_in_flight to max_mod_rpcs_per_client+2
5953 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$((mmrpc + 2))
5956 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$((mmrpc + 1)) &&
5957 error "set max_mod_rpcs_in_flight to $((mmrpc + 1)) should fail"
5961 run_test 90c "check max_mod_rpcs_in_flight update limits"
5972 [[ $($LCTL get_param mdc.*.import |
5973 grep "connect_flags:.*multi_mod_rpc") ]] ||
5974 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
5976 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5977 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
5978 facet="mds$((0x$idx + 1))"
5980 # check client version supports multislots
5981 tmp=$($LCTL get_param -N \
5982 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
5983 if [ -z "$tmp" ]; then
5984 skip "Client does not support multiple modify RPCs in flight"
5989 # get current value of max_mod_rcps_in_flight
5990 mmr=$($LCTL get_param -n \
5991 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
5992 echo "max_mod_rcps_in_flight is $mmr"
5995 echo "creating $mmr files ..."
5997 for i in $(seq $mmr); do
5998 touch $DIR/$tdir/file-$i
6001 # prepare for close RPC
6002 multiop_bg_pause $DIR/$tdir/file-close O_c
6005 # consumes mmr modify RPC slots
6006 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6007 # drop requests on MDT so that RPC slots are consumed
6008 # during all the request resend interval
6009 do_facet $facet "$LCTL set_param fail_loc=0x159"
6010 echo "launch $mmr chmod in parallel ..."
6011 for i in $(seq $mmr); do
6012 chmod 0600 $DIR/$tdir/file-$i &
6015 # send one additional close RPC
6016 do_facet $facet "$LCTL set_param fail_loc=0"
6017 echo "launch 1 additional close in parallel ..."
6019 cancel_lru_locks mdc
6022 # check this additional close RPC get a modify RPC slot
6023 # and multiop process completed
6024 [ -d /proc/$pid ] &&
6025 error "Unable to send the additional close RPC in parallel"
6030 run_test 90d "check one close RPC is allowed above max_mod_rpcs_in_flight"
6032 check_uuid_on_ost() {
6034 do_facet ost1 "$LCTL get_param obdfilter.${FSNAME}*.exports.'$nid'.uuid"
6037 check_uuid_on_mdt() {
6039 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
6049 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.63) ]] ||
6050 { skip "Need OST version at least 2.7.63" && return 0; }
6051 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.63) ]] ||
6052 { skip "Need MDT version at least 2.7.63" && return 0; }
6054 start_mds || error "MDS start failed"
6055 start_ost || error "unable to start OST"
6056 mount_client $MOUNT || error "client start failed"
6057 check_mount || error "check_mount failed"
6060 nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g")
6064 uuid=$(get_client_uuid $MOUNT)
6066 echo "list nids on mdt:"
6067 do_facet $SINGLEMDS "$LCTL list_param mdt.${FSNAME}*.exports.*"
6068 echo "uuid from $nid:"
6069 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
6071 found=$(check_uuid_on_mdt $nid | grep $uuid)
6072 [ -z "$found" ] && error "can't find $uuid $nid on MDT"
6073 found=$(check_uuid_on_ost $nid | grep $uuid)
6074 [ -z "$found" ] && error "can't find $uuid $nid on OST"
6076 # umount the client so it won't reconnect
6077 manual_umount_client --force || error "failed to umount $?"
6078 # shouldn't disappear on MDS after forced umount
6079 found=$(check_uuid_on_mdt $nid | grep $uuid)
6080 [ -z "$found" ] && error "can't find $uuid $nid"
6083 do_facet $SINGLEMDS \
6084 "$LCTL set_param -n mdt.${mds1_svc}.evict_client nid:$nid"
6086 found=$(check_uuid_on_mdt $nid | grep $uuid)
6087 [ -n "$found" ] && error "found $uuid $nid on MDT"
6088 found=$(check_uuid_on_ost $nid | grep $uuid)
6089 [ -n "$found" ] && error "found $uuid $nid on OST"
6091 # check it didn't reconnect (being umounted)
6092 sleep $((TIMEOUT+1))
6093 found=$(check_uuid_on_mdt $nid | grep $uuid)
6094 [ -n "$found" ] && error "found $uuid $nid on MDT"
6095 found=$(check_uuid_on_ost $nid | grep $uuid)
6096 [ -n "$found" ] && error "found $uuid $nid on OST"
6100 run_test 91 "evict-by-nid support"
6102 if ! combined_mgs_mds ; then
6108 # restore the values of MDSSIZE and OSTSIZE
6109 MDSSIZE=$STORED_MDSSIZE
6110 OSTSIZE=$STORED_OSTSIZE