8 PTLDEBUG=${PTLDEBUG:--1}
9 LUSTRE=${LUSTRE:-$(dirname $0)/..}
10 . $LUSTRE/tests/test-framework.sh
14 # tool to create lustre filesystem images
15 ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT 32newtarball"
17 # bug number for skipped test: LU-11915
18 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 110"
19 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
22 # bug number for skipped tests: LU-9795 (all below)
23 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 84 86 103"
26 if ! combined_mgs_mds; then
27 # bug number for skipped test: LU-11991 LU-11990
28 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 32a 32b 32c 32d 32e 66"
29 # bug number for skipped test: LU-9897 LU-12032
30 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 84 123F"
34 [ "$SLOW" = "no" ] && EXCEPT_SLOW="45 69 106 111 114"
38 # use small MDS + OST size to speed formatting time
39 # do not use too small MDSSIZE/OSTSIZE, which affect the default journal size
40 # STORED_MDSSIZE is used in test_18
41 STORED_MDSSIZE=$MDSSIZE
42 STORED_OSTSIZE=$OSTSIZE
44 [ "$mds1_FSTYPE" = zfs ] && MDSSIZE=400000
46 [ "$ost1_FSTYPE" = zfs ] && OSTSIZE=400000
56 # pass "-E lazy_itable_init" to mke2fs to speed up the formatting time
57 if [[ "$LDISKFS_MKFS_OPTS" != *lazy_itable_init* ]]; then
58 LDISKFS_MKFS_OPTS=$(csa_add "$LDISKFS_MKFS_OPTS" -E lazy_itable_init)
62 require_dsh_mds || exit 0
63 require_dsh_ost || exit 0
68 # The MGS must be started before the OSTs for a new fs, so start
69 # and stop to generate the startup logs.
72 wait_osc_import_state mds ost FULL
77 reformat_and_config() {
79 if ! combined_mgs_mds ; then
85 writeconf_or_reformat() {
86 # There are at most 2 OSTs for write_conf test
87 # who knows if/where $TUNEFS is installed?
88 # Better reformat if it fails...
89 writeconf_all $MDSCOUNT 2 ||
90 { echo "tunefs failed, reformatting instead" &&
91 reformat_and_config && return 0; }
100 echo "start mgs service on $(facet_active_host mgs)"
101 start mgs $(mgsdevname) $MGS_MOUNT_OPTS "$@"
107 local dev=$(mdsdevname $num)
110 echo "start mds service on `facet_active_host $facet`"
111 start $facet ${dev} $MDS_MOUNT_OPTS "$@" || return 94
114 stop_mdt_no_force() {
117 local dev=$(mdsdevname $num)
120 echo "stop mds service on `facet_active_host $facet`"
121 stop $facet || return 97
127 local dev=$(mdsdevname $num)
130 echo "stop mds service on `facet_active_host $facet`"
131 # These tests all use non-failover stop
132 stop $facet -f || return 97
136 local mdscount=$MDSCOUNT
139 [[ "$1" == "--mdscount" ]] && mdscount=$2 && shift 2
141 for ((num=1; num <= $mdscount; num++ )); do
142 start_mdt $num "$@" || return 94
144 for ((num=1; num <= $mdscount; num++ )); do
145 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds${num} FULL
150 if ! combined_mgs_mds ; then
158 for num in $(seq $MDSCOUNT); do
159 stop_mdt $num || return 97
164 echo "stop mgs service on `facet_active_host mgs`"
165 # These tests all use non-failover stop
166 stop mgs -f || return 97
170 echo "start ost1 service on `facet_active_host ost1`"
171 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS "$@" || return 95
172 wait_clients_import_ready ${CLIENTS:-$HOSTNAME} ost1
176 echo "stop ost1 service on `facet_active_host ost1`"
177 # These tests all use non-failover stop
178 stop ost1 -f || return 98
182 echo "start ost2 service on `facet_active_host ost2`"
183 start ost2 $(ostdevname 2) $OST_MOUNT_OPTS "$@" || return 92
184 wait_clients_import_ready ${CLIENTS:-$HOSTNAME} ost2
188 echo "stop ost2 service on `facet_active_host ost2`"
189 # These tests all use non-failover stop
190 stop ost2 -f || return 93
197 echo "mount $FSNAME ${mountopt:+with opts $mountopt} on $mountpath....."
198 zconf_mount $HOSTNAME $mountpath $mountopt || return 96
204 echo "umount lustre on $mountpath....."
205 zconf_umount $HOSTNAME $mountpath "$@" || return 97
208 manual_umount_client(){
211 echo "manual umount lustre on ${MOUNT}...."
212 do_facet client "umount ${FORCE} $MOUNT"
218 start_mds || error "MDT start failed"
219 start_ost || error "Unable to start OST1"
220 mount_client $MOUNT || error "client start failed"
221 client_up || error "client_up failed"
230 unload_modules_conf () {
231 if combined_mgs_mds || ! local_mode; then
232 unload_modules || return 1
237 stop_ost || return 202
238 stop_mds || return 201
239 unload_modules_conf || return 203
244 [ "x$1" != "x" ] && force='-f'
245 umount_client $MOUNT $force|| return 200
246 cleanup_nocli || return $?
251 echo "umount $MOUNT2 ..."
252 umount $MOUNT2 || true
253 echo "stopping fs2mds ..."
254 stop fs2mds -f || true
255 echo "stopping fs2ost ..."
256 stop fs2ost -f || true
260 do_facet client "cp /etc/passwd $DIR/a" || return 71
261 do_facet client "rm $DIR/a" || return 72
262 # make sure lustre is actually mounted (touch will block,
263 # but grep won't, so do it after)
264 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
265 echo "setup single mount lustre success"
269 do_facet client "touch $DIR/a" || return 71
270 do_facet client "rm $DIR/a" || return 72
271 do_facet client "touch $DIR2/a" || return 73
272 do_facet client "rm $DIR2/a" || return 74
273 echo "setup double mount lustre success"
277 cat /dev/urandom | tr -dc 'a-zA-Z0-9' | fold -w $1 | head -n 1
280 if [ "$ONLY" == "setup" ]; then
285 if [ "$ONLY" == "cleanup" ]; then
292 #create single point mountpoint
298 check_mount || error "check_mount failed"
299 cleanup || error "cleanup failed with $?"
301 run_test 0 "single mount setup"
304 start_mds || error "MDS start failed"
305 start_ost || error "unable to start OST"
306 echo "start ost second time..."
307 start_ost && error "2nd OST start should fail"
308 mount_client $MOUNT || error "client start failed"
309 check_mount || error "check_mount failed"
310 cleanup || error "cleanup failed with $?"
312 run_test 1 "start up ost twice (should return errors)"
315 start_mds || error "MDT start failed"
316 echo "start mds second time.."
317 start_mds && error "2nd MDT start should fail"
318 start_ost || error "OST start failed"
319 mount_client $MOUNT || error "mount_client failed to start client"
320 check_mount || error "check_mount failed"
321 cleanup || error "cleanup failed with $?"
323 run_test 2 "start up mds twice (should return err)"
327 #mount.lustre returns an error if already in mtab
328 mount_client $MOUNT && error "2nd client mount should fail"
329 check_mount || error "check_mount failed"
330 cleanup || error "cleanup failed with $?"
332 run_test 3 "mount client twice (should return err)"
336 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
337 stop_ost || error "Unable to stop OST1"
338 umount_client $MOUNT -f || error "unmount $MOUNT failed"
341 # ok for ost to fail shutdown
342 if [ 202 -ne $eno ] && [ 0 -ne $eno ]; then
343 error "cleanup failed with $?"
346 run_test 4 "force cleanup ost, then cleanup"
348 test_5a() { # was test_5
350 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
351 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
353 stop_mds || error "Unable to stop MDS"
355 # cleanup may return an error from the failed
356 # disconnects; for now I'll consider this successful
357 # if all the modules have unloaded.
361 echo "killing umount"
362 kill -TERM $UMOUNT_PID
363 echo "waiting for umount to finish"
365 if grep " $MOUNT " /proc/mounts; then
366 echo "test 5: /proc/mounts after failed umount"
370 echo "killing umount"
371 kill -TERM $UMOUNT_PID
372 echo "waiting for umount to finish"
374 grep " $MOUNT " /proc/mounts &&
375 error "/proc/mounts after second umount"
378 # manual_mount_client may fail due to umount succeeding above
380 # stop_mds is a no-op here, and should not fail
381 cleanup_nocli || error "cleanup_nocli failed with $?"
382 # df may have lingering entry
384 # mtab may have lingering entry
388 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
390 grep -q $MOUNT" " /etc/mtab || break
391 echo "Waiting /etc/mtab updated ... "
392 WAIT=$(( WAIT + sleep))
394 [ "$WAIT" -eq "$MAX_WAIT" ] &&
395 error "/etc/mtab is not updated in $WAIT secs"
396 echo "/etc/mtab updated in $WAIT secs"
398 run_test 5a "force cleanup mds, then cleanup"
406 grep " $MOUNT " /etc/mtab &&
407 error false "unexpected entry in mtab before mount" && return 10
409 start_ost || error "OST start failed"
410 if ! combined_mgs_mds ; then
411 trap cleanup_5b EXIT ERR
412 start_mds || error "MDS start failed"
416 mount_client $MOUNT && error "mount_client $MOUNT should fail"
417 grep " $MOUNT " /etc/mtab &&
418 error "$MOUNT entry in mtab after failed mount"
420 # stop_mds is a no-op here, and should not fail
421 cleanup_nocli || error "cleanup_nocli failed with $?"
422 if ! combined_mgs_mds ; then
426 run_test 5b "Try to start a client with no MGS (should return errs)"
429 grep " $MOUNT " /etc/mtab &&
430 error false "unexpected entry in mtab before mount" && return 10
432 start_mds || error "MDS start failed"
433 start_ost || error "OST start failed"
434 local oldfs="${FSNAME}"
435 FSNAME="wrong.${FSNAME}"
436 mount_client $MOUNT || :
438 grep " $MOUNT " /etc/mtab &&
439 error "$MOUNT entry in mtab after failed mount"
441 cleanup_nocli || error "cleanup_nocli failed with $?"
443 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
446 grep " $MOUNT " /etc/mtab &&
447 error "unexpected entry in mtab before mount"
449 start_ost || error "OST start failed"
450 start_mds || error "MDS start failed"
451 stop_ost -f || error "Unable to stop OST1"
452 mount_client $MOUNT || error "mount_client $MOUNT failed"
453 umount_client $MOUNT -f || error "umount_client $MOUNT failed"
454 cleanup_nocli || error "cleanup_nocli failed with $?"
455 ! grep " $MOUNT " /etc/mtab ||
456 error "$MOUNT entry in mtab after unmount"
458 run_test 5d "mount with ost down"
461 grep " $MOUNT " /etc/mtab &&
462 error false "unexpected entry in mtab before mount" && return 10
464 start_mds || error "MDS start failed"
465 start_ost || error "OST start failed"
467 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
468 do_facet client "$LCTL set_param fail_loc=0x80000506"
469 mount_client $MOUNT || echo "mount failed (not fatal)"
470 cleanup || error "cleanup failed with $?"
471 grep " $MOUNT " /etc/mtab &&
472 error "$MOUNT entry in mtab after unmount"
475 run_test 5e "delayed connect, don't crash (bug 10268)"
478 combined_mgs_mds && skip "needs separate mgs and mds"
480 grep " $MOUNT " /etc/mtab &&
481 error false "unexpected entry in mtab before mount" && return 10
484 start_ost || error "OST start failed"
485 mount_client $MOUNT &
487 echo client_mount pid is $pid
491 if ! ps -f -p $pid >/dev/null; then
494 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
495 error "mount returns $rc, expected to hang"
497 cleanup || error "cleanup failed with $?"
502 start_mds || error "start MDS failed"
504 # mount should succeed after start mds
506 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
507 cleanup || error "final call to cleanup failed with rc $?"
509 run_test 5f "mds down, cleanup after failed mount (bug 2712)"
513 [ "$CLIENT_VERSION" -lt $(version_code 2.9.53) ] &&
514 skip "automount of debugfs missing before 2.9.53"
515 umount /sys/kernel/debug
516 $LCTL get_param -n devices | egrep -v "error" && \
517 error "lctl can't access debugfs data"
518 grep " debugfs " /etc/mtab || error "debugfs failed to remount"
520 run_test 5g "handle missing debugfs"
526 #define OBD_FAIL_MDS_FS_SETUP 0x135
527 do_facet mds1 "$LCTL set_param fail_loc=0x80000135"
528 start_mdt 1 && error "start mdt should fail"
529 start_mdt 1 || error "start mdt failed"
530 client_up || error "client_up failed"
533 run_test 5h "start mdt failure at mdt_fs_setup()"
536 (( $MDS1_VERSION >= $(version_code 2.12.54) )) ||
537 skip "Need MDS version at least 2.12.54"
542 #define OBD_FAIL_QUOTA_INIT 0xA05
543 do_facet mds1 "$LCTL set_param fail_loc=0x80000A05"
544 start_mdt 1 && error "start mdt should fail"
545 start_mdt 1 || error "start mdt failed"
546 client_up || error "client_up failed"
549 run_test 5i "start mdt failure at mdt_quota_init()"
554 mount_client $MOUNT || error "mount_client $MOUNT failed"
555 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
556 cleanup || error "cleanup failed with rc $?"
558 run_test 6 "manual umount, then mount again"
563 cleanup_nocli || error "cleanup_nocli failed with $?"
565 run_test 7 "manual umount, then cleanup"
569 mount_client $MOUNT2 || error "mount_client $MOUNT2 failed"
570 check_mount2 || error "check_mount2 failed"
571 umount_client $MOUNT2 || error "umount_client $MOUNT2 failed"
572 cleanup || error "cleanup failed with rc $?"
574 run_test 8 "double mount setup"
577 start_ost || error "OST start failed"
579 do_facet ost1 $LCTL set_param debug=\'inode trace\' ||
580 error "do_facet ost1 set_param inode trace failed."
581 do_facet ost1 $LCTL set_param subsystem_debug=\'mds ost\' ||
582 error "do_facet ost1 set_param debug mds ost failed."
584 CHECK_PTLDEBUG="`do_facet ost1 $LCTL get_param -n debug`"
585 if [ "$CHECK_PTLDEBUG" ] && { \
586 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
587 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
588 echo "lnet.debug success"
590 error "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
592 CHECK_SUBSYS="`do_facet ost1 $LCTL get_param -n subsystem_debug`"
593 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
594 echo "lnet.subsystem_debug success"
596 error "lnet.subsystem_debug: want 'mds ost' got '$CHECK_SUBSYS'"
598 stop_ost || error "Unable to stop OST1"
600 run_test 9 "test ptldebug and subsystem for mkfs"
605 if ! combined_mgs_mds; then
606 files=$(do_facet mgs "find /{proc/fs,sys/fs,sys/kernel/debug}/lustre -type l -exec test ! -e {} \; -print")
607 [ -z $files ] || echo "MGS $files is a broken symlink"
610 files=$(do_facet mds1 "find /{proc/fs,sys/fs,sys/kernel/debug}/lustre -type l -exec test ! -e {} \; -print")
611 [ -z $files ] || echo "MDS $files is a broken symlink"
613 files=$(do_facet ost1 "find /{proc/fs,sys/fs,sys/kernel/debug}/lustre -type l -exec test ! -e {} \; -print")
614 [ -z $files ] || echo "OSS $files is a broken symlink"
616 files=$(do_facet client "find /{proc/fs,sys/fs,sys/kernel/debug}/lustre -type l -exec test ! -e {} \; -print")
617 [ -z $files ] || echo "clients $files is a broken symlink"
619 cleanup || error "cleanup failed with rc $?"
621 run_test 10a "find lctl param broken symlinks"
624 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
625 # But with new MDS stack we don't care about the mode of local objects
626 # anymore, so this test is removed. See bug 22944 for more details.
630 if [ "$mds1_FSTYPE" != ldiskfs ]; then
631 skip "ldiskfs only test"
635 check_mount || error "check_mount failed"
636 cleanup || error "cleanup failed with rc $?"
638 echo "Remove mds config log"
639 if ! combined_mgs_mds ; then
643 do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' \
644 $(mgsdevname) || return \$?" ||
645 error "do_facet mgs failed with $?"
647 if ! combined_mgs_mds ; then
651 start_ost || error "OST start failed"
652 start_mds && error "MDS start succeeded, but should fail"
655 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
658 if [ "$mds1_FSTYPE" != ldiskfs ]; then
659 skip "ldiskfs only test"
662 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
667 # check if current MDSSIZE is large enough
668 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE &&
669 log "use MDSSIZE=$MDSSIZE"
671 # check if the global config has a large enough MDSSIZE
672 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] &&
673 [ $STORED_MDSSIZE -ge $MIN ] &&
674 OK=1 && myMDSSIZE=$STORED_MDSSIZE &&
675 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
677 # check if the block device is large enough
678 is_blkdev $SINGLEMDS $MDSDEV $MIN
679 local large_enough=$?
680 if [ -n "$OK" ]; then
681 [ $large_enough -ne 0 ] && OK=""
683 [ $large_enough -eq 0 ] && OK=1 && myMDSSIZE=$MIN &&
684 log "use device $MDSDEV with MIN=$MIN"
687 # check if a loopback device has enough space for fs metadata (5%)
689 if [ -z "$OK" ]; then
690 local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! \
691 -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
692 awk '($1 != "Filesystem") { print $4 }')
693 ! [ -z "$SPACE" ] && [ $SPACE -gt $((MIN / 20)) ] &&
694 OK=1 && myMDSSIZE=$MIN &&
695 log "use file $MDSDEV with MIN=$MIN"
698 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS"
700 echo "mount mds with large journal..."
702 local OLD_MDSSIZE=$MDSSIZE
706 echo "mount lustre system..."
708 check_mount || error "check_mount failed"
710 echo "check journal size..."
711 local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
712 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
713 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
715 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
718 cleanup || error "cleanup failed with rc $?"
723 run_test 18 "check mkfs creates large journals"
726 start_mds || error "MDS start failed"
727 stop_mds || error "Unable to stop MDS"
729 run_test 19a "start/stop MDS without OSTs"
732 start_ost || error "Unable to start OST1"
733 stop_ost -f || error "Unable to stop OST1"
735 run_test 19b "start/stop OSTs without MDS"
738 # first format the ost/mdt
739 start_mds || error "MDS start failed"
740 start_ost || error "Unable to start OST1"
741 mount_client $MOUNT || error "mount_client $MOUNT failed"
742 check_mount || error "check_mount failed"
743 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed."
744 mount_client $MOUNT remount,ro || error "remount client with ro failed"
745 touch $DIR/$tfile && error "$DIR/$tfile created incorrectly"
746 [ -e $DIR/$tfile ] && error "$DIR/$tfile exists incorrectly"
747 mount_client $MOUNT remount,rw || error "remount client with rw failed"
748 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
749 MCNT=$(grep -c $MOUNT' ' /etc/mtab)
750 [ "$MCNT" -ne 1 ] && error "$MOUNT in /etc/mtab $MCNT times"
752 stop_mds || error "Unable to stop MDS"
753 stop_ost || error "Unable to stop OST1"
755 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
758 start_mds || error "MDS start failed"
759 start_ost || error "unable to start OST1"
760 wait_osc_import_state mds ost FULL
761 stop_ost || error "unable to stop OST1"
762 stop_mds || error "unable to stop MDS"
764 run_test 21a "start mds before ost, stop ost first"
767 start_ost || error "unable to start OST1"
768 start_mds || error "MDS start failed"
769 wait_osc_import_state mds ost FULL
770 stop_mds || error "unable to stop MDS"
771 stop_ost || error "unable to stop OST1"
773 run_test 21b "start ost before mds, stop mds first"
776 start_ost || error "Unable to start OST1"
777 start_mds || error "MDS start failed"
778 start_ost2 || error "Unable to start OST2"
779 wait_osc_import_state mds ost2 FULL
780 stop_ost || error "Unable to stop OST1"
781 stop_ost2 || error "Unable to stop OST2"
782 stop_mds || error "Unable to stop MDS"
783 #writeconf to remove all ost2 traces for subsequent tests
784 writeconf_or_reformat
786 run_test 21c "start mds between two osts, stop mds last"
789 combined_mgs_mds && skip "need separate mgs device"
794 start_mgs || error "unable to start MGS"
795 start_ost || error "unable to start OST1"
796 start_ost2 || error "unable to start OST2"
797 start_mds || error "MDS start failed"
798 wait_osc_import_state mds ost2 FULL
800 local zkeeper=${KEEP_ZPOOL}
801 stack_trap "KEEP_ZPOOL=$zkeeper" EXIT
804 stop_ost || error "Unable to stop OST1"
805 stop_ost2 || error "Unable to stop OST2"
806 stop_mds || error "Unable to stop MDS"
808 #writeconf to remove all ost2 traces for subsequent tests
809 writeconf_or_reformat
810 KEEP_ZPOOL="${zkeeper}"
812 start_mgs || error "unable to start MGS"
814 run_test 21d "start mgs then ost and then mds"
817 MGSNID="$saved_mgsnid"
819 echo "stopping fs2mgs ..."
820 stop $fs2mgs -f || true
823 test_21e() { # LU-5863
824 if [[ -z "$fs3ost_DEV" || -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
825 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
826 skip_env "mixed loopback and real device not working"
829 local fs2mdsdev=$(mdsdevname 1_2)
830 local fs2ostdev=$(ostdevname 1_2)
831 local fs3ostdev=$(ostdevname 2_2)
833 local fs2mdsvdev=$(mdsvdevname 1_2)
834 local fs2ostvdev=$(ostvdevname 1_2)
835 local fs3ostvdev=$(ostvdevname 2_2)
837 # temporarily use fs3ost as fs2mgs
839 local fs2mgsdev=$fs3ostdev
840 local fs2mgsvdev=$fs3ostvdev
842 local fsname=test1234
844 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
845 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
846 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS && trap cleanup_21e EXIT INT ||
847 error "start fs2mgs failed"
849 local saved_mgsnid="$MGSNID"
850 MGSNID=$(do_facet $fs2mgs $LCTL list_nids | xargs | tr ' ' ,)
852 add fs2mds $(mkfs_opts mds1 $fs2mdsdev $fsname) \
853 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
854 add fs2ost $(mkfs_opts ost1 $fs2ostdev $fsname) \
855 --reformat $fs2ostdev $fs2ostvdev || error "add fs2ost failed"
857 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
858 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS || error "start fs2mds failed"
860 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
861 $MOUNT_CMD $MGSNID:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
862 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
866 run_test 21e "separate MGS and MDS"
869 start_mds || error "MDS start failed"
871 echo "Client mount with ost in logs, but none running"
872 start_ost || error "unable to start OST1"
873 # wait until mds connected to ost and open client connection
874 wait_osc_import_state mds ost FULL
875 stop_ost || error "unable to stop OST1"
876 mount_client $MOUNT || error "mount_client $MOUNT failed"
877 # check_mount will block trying to contact ost
878 mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed"
879 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed"
880 umount_client $MOUNT -f
883 echo "Client mount with a running ost"
884 start_ost || error "unable to start OST1"
886 # if gss enabled, wait full time to let connection from
887 # mds to ost be established, due to the mismatch between
888 # initial connect timeout and gss context negotiation timeout.
889 # This perhaps could be remove after AT landed.
890 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
891 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
893 mount_client $MOUNT || error "mount_client $MOUNT failed"
894 wait_osc_import_state mds ost FULL
895 wait_osc_import_ready client ost
896 check_mount || error "check_mount failed"
899 cleanup || error "cleanup failed with rc $?"
901 run_test 22 "start a client before osts (should return errs)"
903 test_23a() { # was test_23
906 stop $SINGLEMDS || error "failed to stop $SINGLEMDS"
907 # force down client so that recovering mds waits for reconnect
908 local running=$(grep -c $MOUNT /proc/mounts) || true
909 if [ $running -ne 0 ]; then
910 echo "Stopping client $MOUNT (opts: -f)"
914 # enter recovery on failed mds
915 local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/})
916 start $SINGLEMDS $MDT_DEV $MDS_MOUNT_OPTS || error "MDS start failed"
917 # try to start a new client
918 mount_client $MOUNT &
920 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
921 MOUNT_LUSTRE_PID=$(ps -ef | grep mount.lustre |
922 grep -v grep | awk '{print $2}')
923 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
925 ps --ppid $MOUNT_LUSTRE_PID
926 echo "waiting for mount to finish"
928 # "ctrl-c" sends SIGINT but it usually (in script) does not work on child process
929 # SIGTERM works but it does not spread to offspring processses
930 kill -s TERM $MOUNT_PID
931 kill -s TERM $MOUNT_LUSTRE_PID
932 # we can not wait $MOUNT_PID because it is not a child of this shell
938 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
940 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
941 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
944 [ -z "$PID1" -a -z "$PID2" ] && break
945 echo "waiting for mount to finish ... "
946 WAIT=$(( WAIT + sleep))
948 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
949 error "MOUNT_PID $MOUNT_PID and "\
950 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
953 cleanup || error "cleanup failed with rc $?"
955 run_test 23a "interrupt client during recovery mount delay"
957 test_23b() { # was test_23
958 start_mds || error "MDS start failed"
959 start_ost || error "Unable to start OST1"
960 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
961 $LCTL set_param fail_loc=0x80000313
963 cleanup || error "cleanup failed with rc $?"
965 run_test 23b "Simulate -EINTR during mount"
968 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
970 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
971 is_blkdev $SINGLEMDS $MDSDEV &&
972 skip_env "mixed loopback and real device not working"
975 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
977 local fs2mdsdev=$(mdsdevname 1_2)
978 local fs2ostdev=$(ostdevname 1_2)
979 local fs2mdsvdev=$(mdsvdevname 1_2)
980 local fs2ostvdev=$(ostvdevname 1_2)
983 # LU-9733 test fsname started with numbers as well
984 local FSNAME2=969362ae
986 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --nomgs --mgsnode=$MGSNID \
987 --fsname=${FSNAME2} --reformat $fs2mdsdev $fs2mdsvdev || exit 10
989 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --fsname=${FSNAME2} \
990 --reformat $fs2ostdev $fs2ostvdev || exit 10
993 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
994 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
995 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
996 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
998 # LU-9733 test fsname started with numbers
999 cl_user=$(do_facet $SINGLEMDS lctl --device $FSNAME2-MDT0000 \
1000 changelog_register -n) ||
1001 error "register changelog failed"
1003 do_facet $SINGLEMDS lctl --device $FSNAME2-MDT0000 \
1004 changelog_deregister $cl_user ||
1005 error "deregister changelog failed"
1007 check_mount || error "check_mount failed"
1008 # files written on 1 should not show up on 2
1009 cp /etc/passwd $DIR/$tfile
1011 [ -e $MOUNT2/$tfile ] && error "File bleed"
1014 cp /etc/passwd $MOUNT2/$tfile ||
1015 error "cp /etc/passwd $MOUNT2/$tfile failed"
1016 rm $MOUNT2/$tfile || error "remove $MOUNT2/$tfile failed"
1017 # 2 is actually mounted
1018 grep $MOUNT2' ' /proc/mounts > /dev/null || error "$MOUNT2 not mounted"
1020 facet_failover fs2mds
1021 facet_failover fs2ost
1023 umount_client $MOUNT
1024 # the MDS must remain up until last MDT
1026 MDS=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
1027 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
1028 [ -z "$MDS" ] && error "No MDT"
1030 cleanup_nocli || error "cleanup_nocli failed with rc $?"
1032 run_test 24a "Multiple MDTs on a single node"
1035 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1037 skip "needs combined MGT and MDT device"
1039 if [ -z "$fs2mds_DEV" ]; then
1040 local dev=${SINGLEMDS}_dev
1041 local MDSDEV=${!dev}
1042 is_blkdev $SINGLEMDS $MDSDEV &&
1043 skip_env "mixed loopback and real device not working"
1046 local fs2mdsdev=$(mdsdevname 1_2)
1047 local fs2mdsvdev=$(mdsvdevname 1_2)
1049 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --mgs --fsname=${FSNAME}2 \
1050 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
1052 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && {
1054 error "start MDS should fail"
1057 cleanup || error "cleanup failed with rc $?"
1059 run_test 24b "Multiple MGSs on a single node (should return err)"
1063 check_mount || error "check_mount failed"
1064 local MODULES=$($LCTL modules | awk '{ print $2 }')
1065 rmmod $MODULES 2>/dev/null || true
1066 cleanup || error "cleanup failed with $?"
1068 run_test 25 "Verify modules are referenced"
1072 # we need modules before mount for sysctl, so make sure...
1073 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
1074 #define OBD_FAIL_MDS_FS_SETUP 0x135
1075 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000135"
1076 start_mds && error "MDS started but should not have started"
1077 $LCTL get_param -n devices
1078 DEVS=$($LCTL get_param -n devices | egrep -v MG | wc -l)
1079 [ $DEVS -gt 0 ] && error "number of devices is $DEVS, should be zero"
1080 # start mds to drop writeconf setting
1081 start_mds || error "Unable to start MDS"
1082 stop_mds || error "Unable to stop MDS"
1083 unload_modules_conf || error "unload_modules_conf failed with $?"
1085 run_test 26 "MDT startup failure cleans LOV (should return errs)"
1089 start_ost || error "Unable to start OST1"
1090 start_mds || error "Unable to start MDS"
1091 echo "Requeue thread should have started: "
1092 ps -e | grep ll_cfg_requeue
1093 set_persistent_param_and_check ost1 \
1094 "obdfilter.$FSNAME-OST0000.client_cache_seconds" \
1095 "$FSNAME-OST0000.ost.client_cache_seconds"
1096 cleanup_nocli || error "cleanup_nocli failed with rc $?"
1098 run_test 27a "Reacquire MGS lock if OST started first"
1103 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
1104 awk '($3 ~ "mdt" && $4 ~ "MDT0000") { print $4 }')
1106 facet_failover $SINGLEMDS
1107 set_persistent_param_and_check $SINGLEMDS \
1108 "mdt.$device.identity_acquire_expire" \
1109 "$device.mdt.identity_acquire_expire"
1110 set_persistent_param_and_check client \
1111 "mdc.$device-mdc-*.max_rpcs_in_flight" \
1112 "$device.mdc.max_rpcs_in_flight"
1114 cleanup || error "cleanup failed with $?"
1116 run_test 27b "Reacquire MGS lock after failover"
1118 test_28A() { # was test_28
1121 local TEST="llite.$FSNAME-*.max_read_ahead_whole_mb"
1122 local PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
1123 local orig=$($LCTL get_param -n $TEST)
1124 local max=$($LCTL get_param -n \
1125 llite.$FSNAME-*.max_read_ahead_per_file_mb)
1127 orig=${orig%%.[0-9]*}
1129 echo "ORIG:$orig MAX:$max"
1130 [[ $max -le $orig ]] && orig=$((max - 3))
1131 echo "ORIG:$orig MAX:$max"
1133 local final=$((orig + 1))
1135 set_persistent_param_and_check client "$TEST" "$PARAM" $final
1136 final=$((final + 1))
1137 set_persistent_param_and_check client "$TEST" "$PARAM" $final
1138 umount_client $MOUNT || error "umount_client $MOUNT failed"
1139 mount_client $MOUNT || error "mount_client $MOUNT failed"
1141 local result=$($LCTL get_param -n $TEST)
1143 if [ $result -ne $final ]; then
1144 error "New config not seen: wanted $final got $result"
1146 echo "New config success: got $result"
1148 set_persistent_param_and_check client "$TEST" "$PARAM" $orig
1149 cleanup || error "cleanup failed with rc $?"
1151 run_test 28A "permanent parameter setting"
1153 test_28a() { # LU-4221
1154 [[ "$OST1_VERSION" -ge $(version_code 2.5.52) ]] ||
1155 skip "Need OST version at least 2.5.52"
1156 [ "$ost1_FSTYPE" = zfs ] &&
1157 skip "LU-4221: no such proc params for ZFS OSTs"
1164 local device="$FSNAME-OST0000"
1168 # In this test we will set three kinds of proc parameters with
1169 # lctl set_param -P or lctl conf_param:
1170 # 1. non-symlink ones in the OFD
1171 # 2. non-symlink ones in the OSD
1174 # prepare a non-symlink parameter in the OFD
1175 name="client_cache_seconds"
1176 param="$device.ost.$name"
1177 cmd="obdfilter.$device.$name"
1179 # permanently setting the parameter in the OFD
1180 old=$(do_facet ost1 $LCTL get_param -n $cmd)
1182 set_persistent_param_and_check ost1 "$cmd" "$param" $new
1183 set_persistent_param_and_check ost1 "$cmd" "$param" $old
1186 # prepare a non-symlink parameter in the OSD
1188 param="$device.osd.$name"
1189 cmd="osd-*.$device.$name"
1191 # conf_param the parameter in the OSD
1192 old=$(do_facet ost1 $LCTL get_param -n $cmd)
1193 new=$(((old + 1) % 2))
1194 set_persistent_param_and_check ost1 "$cmd" "$param" $new
1195 set_persistent_param_and_check ost1 "$cmd" "$param" $old
1197 cleanup || error "cleanup failed with $?"
1199 run_test 28a "set symlink parameters permanently with lctl"
1202 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >= 2 OSTs"
1203 setup_noconfig > /dev/null 2>&1
1204 start_ost2 || error "Unable to start OST2"
1207 local PARAM="$FSNAME-OST0001.osc.active"
1208 # With lctl set_param -P the value $PROC_ACT will be sent to
1209 # all nodes. The [!M] filter out the ability to set active
1210 # on the MDS servers which is tested with wait_osp_* below.
1211 # For ost_server_uuid that only exist on client so filtering
1213 local PROC_ACT="os[cp].$FSNAME-OST0001-osc-[^M]*.active"
1214 local PROC_UUID="os[cp].$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
1216 ACTV=$($LCTL get_param -n $PROC_ACT)
1218 set_persistent_param_and_check client $PROC_ACT $PARAM $DEAC
1219 # also check ost_server_uuid status
1220 RESULT=$($LCTL get_param -n $PROC_UUID | grep DEACTIV)
1221 if [ -z "$RESULT" ]; then
1222 error "Client not deactivated: $($LCTL get_param \
1225 echo "Live client success: got $RESULT"
1229 wait_osp_active ost ${FSNAME}-OST0001 1 0
1231 # test new client starts deactivated
1232 umount_client $MOUNT || error "umount_client $MOUNT failed"
1233 mount_client $MOUNT || error "mount_client $MOUNT failed"
1235 # the 2nd and 3rd field of ost_server_uuid do not update at the same
1236 # time when using lctl set_param -P
1237 wait_update_facet client \
1238 "$LCTL get_param -n $PROC_UUID | awk '{print \\\$3 }'" \
1240 error "New client start active: $($LCTL get_param -n $PROC_UUID)"
1242 echo "New client success: got '$($LCTL get_param -n $PROC_UUID)'"
1244 # make sure it reactivates
1245 set_persistent_param_and_check client $PROC_ACT $PARAM $ACTV
1247 umount_client $MOUNT
1248 stop_ost2 || error "Unable to stop OST2"
1249 cleanup_nocli || error "cleanup_nocli failed with $?"
1250 #writeconf to remove all ost2 traces for subsequent tests
1251 writeconf_or_reformat
1253 run_test 29 "permanently remove an OST"
1258 echo Big config llog
1259 local path="llite.$FSNAME-*"
1260 local cpath="$FSNAME.llite"
1261 local param="max_read_ahead_whole_mb"
1263 local test="${path}.$param"
1264 local conf="${cpath}.$param"
1266 local orig=$($LCTL get_param -n $test)
1267 local list=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1268 for i in ${list[@]}; do
1269 set_persistent_param_and_check client $test $conf $i
1271 # make sure client restart still works
1272 umount_client $MOUNT
1273 mount_client $MOUNT || error "mount_client $MOUNT failed"
1274 [ "$($LCTL get_param -n $test)" -ne "$i" ] &&
1275 error "Param didn't stick across restart $($test) != $i"
1278 echo Erase parameter setting
1279 if [[ $PERM_CMD == *"set_param -P"* ]]; then
1280 do_facet mgs "$PERM_CMD -d $test" ||
1281 error "Erase param $test failed"
1283 do_facet mgs "$PERM_CMD -d $conf" ||
1284 error "Erase param $conf failed"
1286 umount_client $MOUNT
1287 mount_client $MOUNT || error "mount_client $MOUNT failed"
1288 local final=$($LCTL get_param -n $test)
1289 echo "deleted (default) value=$final, orig=$orig"
1290 orig=${orig%%.[0-9]*}
1291 final=${final%%.[0-9]*}
1292 # assumes this parameter started at the default value
1293 [ "$final" -eq "$orig" ] ||
1294 error "Deleted value=$final -ne orig=$orig"
1296 cleanup || error "cleanup failed with rc $?"
1298 run_test 30a "Big config llog and permanent parameter deletion"
1303 local orignids=$($LCTL get_param -n \
1304 osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1306 local orignidcount=$(echo "$orignids" | wc -w)
1308 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1309 # numerical part of it. Hopefully that's not already a failover address
1311 local OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | \
1313 local ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1314 local NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1315 local NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1316 echo "Using fake nid $NEW"
1318 local TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1319 grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1320 if [[ $PERM_CMD == *"set_param -P"* ]]; then
1321 PARAM="osc.$FSNAME-OST0000-osc-[^M]*.import"
1322 echo "Setting $PARAM from $TEST to $NEW"
1323 do_facet mgs "$PERM_CMD $PARAM='connection=$NEW'" ||
1324 error "$PERM_CMD $PARAM failed"
1326 PARAM="$FSNAME-OST0000.failover.node"
1327 echo "Setting $PARAM from $TEST to $NEW"
1328 do_facet mgs "$PERM_CMD $PARAM='$NEW'" ||
1329 error "$PARAM $PARAM failed"
1331 wait_update_facet client "$TEST" "$NEW" ||
1332 error "check $PARAM failed!"
1334 local NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1336 local NIDCOUNT=$(echo "$NIDS" | wc -w)
1337 echo "should have $((orignidcount + 1)) entries \
1338 in failover nids string, have $NIDCOUNT"
1339 [ $NIDCOUNT -eq $((orignidcount + 1)) ] ||
1340 error "Failover nid not added"
1342 if [[ $PERM_CMD == *"set_param -P"* ]]; then
1343 do_facet mgs "$PERM_CMD -d osc.$FSNAME-OST0000-osc-*.import"
1345 do_facet mgs "$PERM_CMD -d $FSNAME-OST0000.failover.node" ||
1346 error "$PERM_CMD delete failed"
1348 umount_client $MOUNT
1349 mount_client $MOUNT || error "mount_client $MOUNT failed"
1351 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1353 NIDCOUNT=$(echo "$NIDS" | wc -w)
1354 echo "only $orignidcount final entries should remain \
1355 in failover nids string, have $NIDCOUNT"
1356 [ $NIDCOUNT -eq $orignidcount ] || error "Failover nids not removed"
1358 cleanup || error "cleanup failed with rc $?"
1360 run_test 30b "Remove failover nids"
1362 test_31() { # bug 10734
1363 # ipaddr must not exist
1364 $MOUNT_CMD 4.3.2.1@tcp:/lustre $MOUNT || true
1365 cleanup || error "cleanup failed with rc $?"
1367 run_test 31 "Connect to non-existent node (shouldn't crash)"
1371 T32_BLIMIT=40960 # Kbytes
1374 T32_PROLIMIT=$((T32_BLIMIT/10))
1376 # This is not really a test but a tool to create new disk
1377 # image tarballs for the upgrade tests.
1379 # Disk image tarballs should be created on single-node
1380 # clusters by running this test with default configurations
1381 # plus a few mandatory environment settings that are verified
1382 # at the beginning of the test.
1384 test_32newtarball() {
1388 local tmp=$TMP/t32_image_create
1399 local pj_quota_file_old
1402 if [ $FSNAME != t32fs -o \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o \
1403 $OSTCOUNT -ne 2 -o -z "$OSTDEV1" ]; then
1404 error "Needs FSNAME=t32fs MDSCOUNT=2 " \
1405 "MDSDEV1=<nonexistent_file> " \
1406 "MDSDEV2=<nonexistent_file> " \
1407 "(or MDSDEV, in the case of b1_8) " \
1408 "OSTCOUNT=2 OSTDEV1=<nonexistent_file> " \
1409 "OSTDEV2=<nonexistent_file>"
1413 echo "Found stale $tmp"
1417 mkdir $tmp/src || return 1
1418 tar cf - -C $src . | tar xf - -C $tmp/src
1420 # format ost with comma-separated NIDs to verify LU-4460
1421 local failnid="$(h2nettype 1.2.3.4),$(h2nettype 4.3.2.1)"
1422 MGSNID="$MGSNID,$MGSNID" OSTOPT="--failnode=$failnid" formatall
1426 [[ "$MDS1_VERSION" -ge $(version_code 2.3.50) ]] ||
1427 $LFS quotacheck -ug /mnt/$FSNAME
1428 $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
1431 if [[ $MDSCOUNT -ge 2 ]]; then
1432 remote_dir=/mnt/$FSNAME/remote_dir
1433 $LFS mkdir -i 1 $remote_dir
1434 tar cf - -C $tmp/src . | tar xf - -C $remote_dir
1436 target_dir=$remote_dir
1437 if [[ $MDS1_VERSION -ge $(version_code 2.7.0) ]]; then
1438 striped_dir=/mnt/$FSNAME/striped_dir_old
1439 $LFS mkdir -i 1 -c 2 $striped_dir
1440 tar cf - -C $tmp/src . | tar xf - -C $striped_dir
1443 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1444 target_dir=/mnt/$FSNAME
1447 # add project quota #
1448 [[ $mds1_FSTYPE == "ldiskfs" &&
1449 $MDS1_VERSION -gt $(version_code 2.9.55) ]] ||
1450 [[ $mds1_FSTYPE == "zfs" &&
1451 $MDS1_VERSION -gt $(version_code 2.10.53) ]] && {
1452 pj_quota_dir=$target_dir/project_quota_dir
1453 pj_quota_file_old=$pj_quota_dir/pj_quota_file_old
1455 enable_project_quota
1456 set_mdt_qtype ugp ||
1457 error "enable mdt quota failed"
1459 set_ost_qtype ugp ||
1460 error "enable ost quota failed"
1462 mkdir -p $pj_quota_dir
1463 $LFS setquota -p $T32_PRJID -b 0 -B $T32_PROLIMIT -i 0 \
1464 -I $T32_ILIMIT $pj_quota_dir ||
1465 error "setquota -p $T32_PRJID failed"
1467 $LFS setstripe $pj_quota_file_old -c 1 ||
1468 error "setstripe $pj_quota_file_old failed"
1470 chown $T32_QID:$T32_QID $pj_quota_file_old ||
1471 error "chown $pj_quota_file_old failed"
1473 change_project -p $T32_PRJID $pj_quota_file_old
1475 mkdir -p $tmp/src/project_quota_dir
1476 cp $pj_quota_file_old $tmp/src/project_quota_dir/
1479 #####################
1480 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1482 #if [[ $MDSCOUNT -ge 2 ]]; then
1483 # remote_dir=/mnt/$FSNAME/remote_dir
1484 # $LFS mkdir -i 1 $remote_dir
1485 # tar cf - -C $tmp/src . | tar xf - -C $remote_dir
1487 # if [[ "$MDS1_VERSION" -ge $(version_code 2.7.0) ]]; then
1488 # striped_dir=/mnt/$FSNAME/striped_dir_old
1489 # $LFS mkdir -i 1 -c 2 $striped_dir
1490 # tar cf - -C $tmp/src . | tar xf - -C $striped_dir
1495 if [[ $MDS1_VERSION -ge $(version_code 2.9.51) ]]; then
1496 pfl_dir=$target_dir/pfl_dir
1497 pfl_file=$pfl_dir/pfl_file
1499 $LFS setstripe -E 2M -c 1 -o 0 -E -1 -S 2M -c 1 -o 1 \
1501 error "Create PFL file failed"
1503 dd if=/dev/urandom of=$pfl_file bs=1k count=3k
1504 mkdir -p $tmp/src/pfl_dir
1505 cp $pfl_file $tmp/src/pfl_dir/
1510 if [[ $MDS1_VERSION -ge $(version_code 2.10.56) ]]; then
1511 dom_dir=$target_dir/dom_dir
1512 dom_file=$dom_dir/dom_file
1513 flr_dir=$target_dir/flr_dir
1514 flr_file=$flr_dir/flr_file
1517 $LFS setstripe -E 1M -L mdt -E -1 -S 4M $dom_file
1518 dd if=/dev/urandom of=$dom_file bs=1k count=2k
1519 mkdir -p $tmp/src/dom_dir
1520 cp $dom_file $tmp/src/dom_dir
1523 $LFS mirror create -N2 $flr_file
1524 dd if=/dev/urandom of=$flr_file bs=1k count=1
1525 mkdir -p $tmp/src/flr_dir
1526 cp $flr_file $tmp/src/flr_dir
1530 local large_xattr_dir=/mnt/$FSNAME/large_xattr_test_dir
1531 local xattr_file=$large_xattr_dir/large_xattr_file
1533 mkdir $large_xattr_dir
1535 setfattr -n user.fooattr -v $(printf "%c" {1..4096} ) $xattr_file ||
1540 mkdir $tmp/img || return 1
1544 pushd_dir=/mnt/$FSNAME
1545 if [[ $MDSCOUNT -ge 2 ]]; then
1546 pushd_dir=$remote_dir
1547 if [[ "$MDS1_VERSION" -ge $(version_code 2.7.0) ]]; then
1549 BLOCKSIZE=1024 ls -Rni --time-style=+%s >$tmp/img/list2
1555 BLOCKSIZE=1024 ls -Rni --time-style=+%s >$tmp/img/list
1556 find ! -name .lustre -type f -exec sha1sum {} \; |
1557 sort -k 2 >$tmp/img/sha1sums
1559 $LCTL get_param -n version | head -n 1 |
1560 sed -e 's/^lustre: *//' >$tmp/img/commit
1562 [[ "$MDS1_VERSION" -ge $(version_code 2.3.50) ]] ||
1563 $LFS quotaon -ug /mnt/$FSNAME
1564 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1565 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1566 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1567 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1568 | tr -d "*" > $tmp/img/bspace
1569 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1570 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1571 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1572 | tr -d "*" > $tmp/img/ispace
1573 echo $T32_BLIMIT > $tmp/img/blimit
1574 echo $T32_ILIMIT > $tmp/img/ilimit
1576 $MULTIOP /mnt/$FSNAME/orph_file Ouw_c&
1580 debugfs -R "ls /PENDING" ${MDSDEV1:-$MDSDEV}
1581 cp ${MDSDEV1:-$MDSDEV} $tmp/img
1590 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1593 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1594 echo "Data verification failed"
1597 uname -r >$tmp/img/kernel
1598 uname -m >$tmp/img/arch
1600 for ((num=1; num <= $MDSCOUNT; num++)); do
1601 local devname=$(mdsdevname $num)
1605 [[ num -eq 1 ]] && image_name=mdt || image_name=mdt$num
1606 [[ $(facet_fstype $facet) != zfs ]] ||
1607 devname=$(mdsvdevname $num)
1608 dd conv=sparse bs=4k if=$devname of=$tmp/img/$image_name
1611 for ((num=1; num <= $OSTCOUNT; num++)); do
1612 local devname=$(ostdevname $num)
1616 [[ num -eq 1 ]] && image_name=ost || image_name=ost$num
1617 [[ $(facet_fstype $facet) != zfs ]] ||
1618 devname=$(ostdevname $num)
1619 dd conv=sparse bs=4k if=$devname of=$tmp/img/$image_name
1622 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1623 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1626 tar cjvf $dst/disk$version-"$mds1_FSTYPE".tar.bz2 -S *
1631 # run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1634 # The list of applicable tarballs is returned via the caller's
1635 # variable "tarballs".
1638 [ "$CLIENTONLY" ] && skip "Client-only testing"
1640 check_versions || skip "do not upgrade with mismatched client/server"
1642 local node=$(facet_active_host $SINGLEMDS)
1643 local r="do_node $node"
1645 ! $r which "$TUNEFS" && skip_env "tunefs.lustre required on $node"
1647 local IMGTYPE="$mds1_FSTYPE"
1649 [[ -n "$T32IMAGE" ]] && tarballs=$T32IMAGE ||
1650 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 \
1651 -name \'disk*-$IMGTYPE.tar.bz2\')
1653 [ -z "$tarballs" ] && skip "No applicable tarballs found"
1656 t32_test_cleanup() {
1658 local facet=$SINGLEMDS
1661 echo "== cleanup with rc=$rc =="
1662 if $shall_cleanup_lustre; then
1663 umount $tmp/mnt/lustre || rc=$?
1665 if $shall_cleanup_mdt; then
1666 $r $UMOUNT $tmp/mnt/mdt || rc=$?
1668 if $shall_cleanup_mdt1; then
1669 $r $UMOUNT $tmp/mnt/mdt1 || rc=$?
1671 if $shall_cleanup_ost; then
1672 $r $UMOUNT $tmp/mnt/ost || rc=$?
1674 if $shall_cleanup_ost1; then
1675 $r $UMOUNT $tmp/mnt/ost1 || rc=$?
1680 if [[ "$mds1_FSTYPE" == zfs ]]; then
1682 local poolname_list="t32fs-mdt1 t32fs-ost1"
1684 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1686 for poolname in $poolname_list; do
1687 destroy_zpool $facet $poolname
1690 combined_mgs_mds || start_mgs || rc=$?
1694 t32_bits_per_long() {
1696 # Yes, this is not meant to be perfect.
1706 t32_reload_modules() {
1708 local all_removed=false
1711 [ "$mds1_FSTYPE" == zfs ] && do_rpc_nodes $node "service zed stop"
1713 while ((i < 20)); do
1714 echo "Unloading modules on $node: Attempt $i"
1715 do_rpc_nodes $node $LUSTRE_RMMOD "$mds1_FSTYPE" &&
1717 do_rpc_nodes $node check_mem_leak || return 1
1718 if $all_removed; then
1719 do_rpc_nodes $node load_modules
1722 if [ "$mds1_FSTYPE" == zfs ]; then
1723 do_rpc_nodes $node "$ZPOOL status -v"
1728 echo "Unloading modules on $node: Given up"
1732 t32_wait_til_devices_gone() {
1738 echo wait for devices to go
1739 while ((i < 20)); do
1740 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1741 loops=$(do_rpc_nodes $node losetup -a | grep -c t32)
1742 ((devices == 0 && loops == 0)) && return 0
1746 echo "waiting for dev on $node: dev $devices loop $loops given up"
1747 do_rpc_nodes $node "losetup -a"
1748 do_rpc_nodes $node "$LCTL device_list"
1752 t32_verify_quota() {
1759 # LU-2435: if the underlying zfs doesn't support userobj_accounting,
1760 # lustre will estimate the object count usage. This fails quota
1761 # verification in 32b. The object quota usage should be accurate after
1762 # zfs-0.7.0 is released.
1763 [ "$mds1_FSTYPE" == zfs ] && {
1764 local zfs_version=$(do_facet $facet cat /sys/module/zfs/version)
1766 [ $(version_code $zfs_version) -lt $(version_code 0.7.0) ] && {
1767 echo "Skip quota verify for zfs: $zfs_version"
1772 $LFS quota -u $T32_QID -v $mnt
1774 qval=$($LFS quota -v -u $T32_QID $mnt |
1775 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1776 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1778 [ $qval -eq $img_bspace ] || {
1779 echo "bspace, act:$qval, exp:$img_bspace"
1783 qval=$($LFS quota -v -u $T32_QID $mnt |
1784 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1785 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1787 [ $qval -eq $img_ispace ] || {
1788 echo "ispace, act:$qval, exp:$img_ispace"
1792 qval=$($LFS quota -v -u $T32_QID $mnt |
1793 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1794 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1796 [ $qval -eq $img_blimit ] || {
1797 echo "blimit, act:$qval, exp:$img_blimit"
1801 qval=$($LFS quota -v -u $T32_QID $mnt |
1802 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1803 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1805 [ $qval -eq $img_ilimit ] || {
1806 echo "ilimit, act:$qval, exp:$img_ilimit"
1810 set_persistent_param_and_check $facet \
1811 "osd-$mds1_FSTYPE.$fsname-MDT0000.quota_slave.enabled" \
1812 "$fsname.quota.mdt" ug
1814 set_persistent_param_and_check $facet \
1815 "osd-$mds1_FSTYPE.$fsname-OST0000.quota_slave.enabled" \
1816 "$fsname.quota.ost" ug
1819 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1820 bs=1M count=$((img_blimit / 1024)) oflag=sync && {
1821 echo "Write succeed, but expect -EDQUOT"
1824 rm -f $mnt/t32_qf_new
1826 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1828 echo "Create succeed, but expect -EDQUOT"
1831 unlinkmany $mnt/t32_qf_ $img_ilimit
1841 sync_all_data > /dev/null 2>&1 || true
1843 [ "$uuid" != "global" ] || uuid=$mnt
1845 $LFS quota -v "$1" "$2" $mnt |
1846 awk 'BEGIN { num='$spec' } { if ($1 == "'$uuid'") \
1847 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1854 local dne_upgrade=${dne_upgrade:-"no"}
1855 local dom_upgrade=${dom_upgrade:-"no"}
1856 local project_upgrade=${project_upgrade:-"no"}
1857 local ff_convert=${ff_convert:-"no"}
1858 local pfl_upgrade=${pfl_upgrade:-"no"}
1859 local project_quota_upgrade=${project_quota_upgrade:-"no"}
1860 local dom_new_upgrade=${dom_new_upgrade:-"no"}
1861 local flr_upgrade=${flr_upgrade:-"no"}
1862 local shall_cleanup_mdt=false
1863 local shall_cleanup_mdt1=false
1864 local shall_cleanup_ost=false
1865 local shall_cleanup_ost1=false
1866 local shall_cleanup_lustre=false
1867 local mdt2_is_available=false
1868 local ost2_is_available=false
1869 local node=$(facet_active_host $SINGLEMDS)
1870 local r="do_node $node"
1886 local mdt_dev=$tmp/mdt
1887 local mdt2_dev=$tmp/mdt2
1888 local ost_dev=$tmp/ost
1889 local ost2_dev=$tmp/ost2
1893 local pfl_file=$tmp/mnt/lustre/remote_dir/pfl_dir/pfl_file
1894 local flr_file=$tmp/mnt/lustre/remote_dir/flr_dir/flr_file
1895 local dom_file=$tmp/mnt/lustre/remote_dir/dom_dir/dom_file
1896 local quota_dir=$tmp/mnt/lustre/remote_dir/project_quota_dir
1898 combined_mgs_mds || stop_mgs || error "Unable to stop MGS"
1899 trap 'trap - RETURN; t32_test_cleanup' RETURN
1902 nid=$($r $LCTL list_nids | head -1)
1904 mkdir -p $tmp/mnt/lustre || error "mkdir $tmp/mnt/lustre failed"
1905 $r mkdir -p $tmp/mnt/{mdt,mdt1,ost,ost1}
1906 $r tar xjvf $tarball -S -C $tmp || {
1907 error_noexit "Unpacking the disk image tarball"
1910 img_commit=$($r cat $tmp/commit)
1911 img_kernel=$($r cat $tmp/kernel)
1912 img_arch=$($r cat $tmp/arch)
1913 img_bspace=$($r cat $tmp/bspace)
1914 img_ispace=$($r cat $tmp/ispace)
1916 # older images did not have "blimit" and "ilimit" files
1917 # use old values for T32_BLIMIT and T32_ILIMIT
1918 $r test -f $tmp/blimit && img_blimit=$($r cat $tmp/blimit) ||
1920 $r test -f $tmp/ilimit && img_ilimit=$($r cat $tmp/ilimit) ||
1923 echo "Upgrading from $(basename $tarball), created with:"
1924 echo " Commit: $img_commit"
1925 echo " Kernel: $img_kernel"
1926 echo " Arch: $img_arch"
1927 echo "OST version: $(lustre_build_version ost1)"
1929 # The conversion can be made only when both of the following
1930 # conditions are satisfied:
1931 # - ost device img version < 2.3.64
1932 # - ost server version >= 2.5
1933 [ $(version_code $img_commit) -ge $(version_code 2.3.64) -o \
1934 "$OST1_VERSION" -lt $(version_code 2.5.0) ] &&
1937 ! $r test -f $mdt2_dev || mdt2_is_available=true
1938 ! $r test -f $ost2_dev || ost2_is_available=true
1940 if [[ "$mds1_FSTYPE" == zfs ]]; then
1941 echo "== fstype is zfs =="
1944 local poolname_list="t32fs-mdt1 t32fs-ost1"
1946 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1948 for poolname in $poolname_list; do
1950 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
1951 $ZPOOL import -f -d $tmp $poolname"
1954 # upgrade zpool to latest supported features, including
1955 # dnode quota accounting in 0.7.0
1956 $r "$ZPOOL upgrade -a"
1958 mdt_dev=t32fs-mdt1/mdt1
1959 ost_dev=t32fs-ost1/ost1
1960 ! $mdt2_is_available || mdt2_dev=t32fs-mdt2/mdt2
1961 ! $ost2_is_available || ost2_dev=t32fs-ost2/ost2
1962 wait_update_facet $SINGLEMDS "$ZPOOL list |
1963 awk '/^t32fs-mdt1/ { print \\\$1 }'" "t32fs-mdt1" || {
1964 error_noexit "import zfs pool failed"
1967 elif [ "$project_upgrade" != "no" ]; then
1968 echo "== project upgrade =="
1969 ! $r tune2fs -O project $mdt_dev &&
1970 error_noexit "enable project on mdt0 failed" &&
1972 $mdt2_is_available && ! $r tune2fs "-O project" $mdt2_dev &&
1973 error_noexit "enable project on mdt failed" &&
1975 ! $r tune2fs -O project $ost_dev &&
1976 error_noexit "enable project on ost failed" &&
1978 $ost2_is_available && ! $r tune2fs -O project $ost2_dev &&
1979 error_noexit "enable project on ost2 failed" &&
1983 $r $LCTL set_param debug="$PTLDEBUG"
1985 $r $TUNEFS --dryrun $mdt_dev || {
1987 error_noexit "tunefs.lustre before mounting the MDT"
1991 if $mdt2_is_available; then
1992 $r $TUNEFS --dryrun $mdt2_dev || {
1994 error_noexit "tunefs.lustre before mounting the MDT"
1999 if [ "$writeconf" ]; then
2000 echo "== writeconf mdt tunefs and quota =="
2002 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2004 $r $TUNEFS --quota $mdt_dev || {
2006 error_noexit "Enable mdt quota feature"
2010 if [ "$project_quota_upgrade" != "no" ]; then
2011 echo "== enable mdt2 project quota =="
2012 $r $TUNE2FS -O project $mdt_dev || {
2014 error_noexit "tune2fs $mdt_dev failed"
2019 if $mdt2_is_available; then
2020 $r $TUNEFS --quota $mdt2_dev || {
2022 error_noexit "Enable mdt2 quota feature"
2025 if [ "$project_quota_upgrade" != "no" ]; then
2026 echo "== enable mdt2 project quota =="
2027 $r $TUNE2FS -O project $mdt2_dev || {
2030 "tune2fs $mdt2_dev failed"
2037 echo "== replace nids =="
2038 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
2039 [[ "$MGS_VERSION" -ge $(version_code 2.3.59) ]] ||
2040 skip "LU-2200: Cannot run over IB w/o lctl replace_nids "
2041 "(Need MGS version at least 2.3.59)"
2044 local osthost=$(facet_active_host ost1)
2045 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
2048 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2051 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt
2052 $r $LCTL replace_nids $fsname-OST0000 $ostnid || {
2054 error_noexit "replace_nids $fsname-OST0000 $ostnid failed"
2057 if $ost2_is_available; then
2058 $r $LCTL replace_nids $fsname-OST0001 $ostnid || {
2060 error_noexit "replace_nids $fsname-OST0001 $ostnid failed"
2064 $r $LCTL replace_nids $fsname-MDT0000 $nid || {
2066 error_noexit "replace_nids $fsname-MDT0000 $nid failed"
2069 if $mdt2_is_available; then
2070 $r $LCTL replace_nids $fsname-MDT0001 $nid || {
2072 error_noexit "replace_nids $fsname-MDT0001 $nid failed"
2076 $r $UMOUNT $tmp/mnt/mdt
2078 mopts=exclude=$fsname-OST0000
2079 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2084 t32_wait_til_devices_gone $node
2086 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2088 error_noexit "Mounting the MDT"
2091 shall_cleanup_mdt=true
2093 if $mdt2_is_available; then
2094 echo "== mdt2 available =="
2095 mopts=mgsnode=$nid,$mopts
2096 $r $MOUNT_CMD -o $mopts $mdt2_dev $tmp/mnt/mdt1 || {
2098 error_noexit "Mounting the MDT"
2101 shall_cleanup_mdt1=true
2102 echo "== mount new MDT $mdt2_dev =="
2104 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 || {
2105 error_noexit "enable remote dir create failed"
2109 elif [ "$dne_upgrade" != "no" ]; then
2110 local fs2mdsdev=$(mdsdevname 1_2)
2111 local fs2mdsvdev=$(mdsvdevname 1_2)
2113 echo "== mkfs new MDT on ${fs2mdsdev} =="
2114 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2115 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
2118 add $SINGLEMDS $(mds2failover_HOST="" \
2119 mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
2120 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
2121 error_noexit "Mkfs new MDT failed"
2125 [[ "$mds1_FSTYPE" != zfs ]] || import_zpool mds1
2127 $r $TUNEFS --dryrun $fs2mdsdev || {
2128 error_noexit "tunefs.lustre before mounting the MDT"
2132 echo "== mount new MDT....$fs2mdsdev =="
2133 $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
2134 error_noexit "mount mdt1 failed"
2138 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
2139 error_noexit "enable remote dir create failed"
2141 shall_cleanup_mdt1=true
2144 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
2145 error_noexit "Getting MDT UUID"
2148 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
2149 error_noexit "Unexpected MDT UUID: \"$uuid\""
2153 $r $TUNEFS --dryrun $ost_dev || {
2154 error_noexit "tunefs.lustre before mounting the OST"
2158 if $ost2_is_available; then
2159 $r $TUNEFS --dryrun $ost2_dev || {
2160 error_noexit "tunefs.lustre before mounting the OST"
2165 if [ "$writeconf" ]; then
2166 echo "== writeconf and ost quota tunefs =="
2167 mopts=mgsnode=$nid,$writeconf
2168 if [ "$ost1_FSTYPE" == ldiskfs ]; then
2170 $r $TUNEFS --quota $ost_dev || {
2172 error_noexit "Enable ost quota feature"
2176 if [ "$project_quota_upgrade" != "no" ]; then
2177 echo "== enable ost project quota =="
2178 $r $TUNE2FS -O project $ost_dev || {
2180 error_noexit "tune2fs $ost_dev failed"
2185 if $ost2_is_available; then
2186 $r $TUNEFS --quota $ost2_dev || {
2188 error_noexit "Enable ost2 quota feature"
2191 if [ "$project_quota_upgrade" != "no" ]; then
2192 echo "== enable ost2 project quota =="
2193 $r $TUNE2FS -O project $ost2_dev || {
2196 "tune2fs $ost2_dev failed"
2204 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2209 $r $MOUNT_CMD -onomgs -o$mopts $ost_dev $tmp/mnt/ost || {
2210 error_noexit "Mounting the OST"
2214 if $ost2_is_available; then
2215 $r $MOUNT_CMD -onomgs -o$mopts $ost2_dev $tmp/mnt/ost1 || {
2216 error_noexit "Mounting the OST2"
2219 shall_cleanup_ost1=true
2222 shall_cleanup_ost=true
2224 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
2225 error_noexit "Getting OST UUID"
2228 if [ "$uuid" != $fsname-OST0000_UUID ]; then
2229 error_noexit "Unexpected OST UUID: \"$uuid\""
2233 if $ost2_is_available; then
2234 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0001.uuid) ||{
2235 error_noexit "Getting OST1 UUID"
2238 if [ "$uuid" != $fsname-OST0001_UUID ]; then
2239 error_noexit "Unexpected OST1 UUID: \"$uuid\""
2244 if [[ $PERM_CMD =~ "set_param" ]]; then
2245 echo "== perm cmd = '$PERM_CMD' =="
2246 $r $PERM_CMD osc.$fsname-OST0000*.import=connection=$nid || {
2247 error_noexit "Setting OST \"failover.node\""
2250 $r $PERM_CMD mdc.$fsname-MDT0000*.import=connection=$nid || {
2251 error_noexit "Setting MDT \"failover.node\""
2254 $r $PERM_CMD osc.$fsname-OST0000-*.max_dirty_mb=15 || {
2255 error_noexit "Setting \"max_dirty_mb\""
2258 $r $PERM_CMD mdc.$fsname-MDT0000-*.max_rpcs_in_flight=9 || {
2259 error_noexit "Setting \"max_rpcs_in_flight\""
2262 $r $PERM_CMD lov.$fsname-MDT0000-*.stripesize=4M || {
2263 error_noexit "Setting \"lov.stripesize\""
2266 $r $PERM_CMD mdd.$fsname-MDT0000-*.atime_diff=70 || {
2267 error_noexit "Setting \"mdd.atime_diff\""
2271 echo "== perm cmd = '$PERM_CMD' =="
2272 $r $PERM_CMD $fsname-OST0000.failover.node=$nid || {
2273 error_noexit "Setting OST \"failover.node\""
2277 $r $PERM_CMD $fsname-MDT0000.failover.node=$nid || {
2278 error_noexit "Setting MDT \"failover.node\""
2282 $r $PERM_CMD $fsname-OST0000.osc.max_dirty_mb=15 || {
2283 error_noexit "Setting \"max_dirty_mb\""
2286 $r $PERM_CMD $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
2287 error_noexit "Setting \"max_rpcs_in_flight\""
2290 $r $PERM_CMD $fsname-MDT0000.lov.stripesize=4M || {
2291 error_noexit "Setting \"lov.stripesize\""
2294 $r $PERM_CMD $fsname-MDT0000.mdd.atime_diff=70 || {
2295 error_noexit "Setting \"mdd.atime_diff\""
2300 $r $LCTL pool_new $fsname.interop || {
2301 error_noexit "Setting \"interop\""
2305 if [ "$ff_convert" != "no" -a "$ost1_FSTYPE" == ldiskfs ]; then
2306 echo "== ff convert ($ff_convert) =="
2307 $r $LCTL lfsck_start -M $fsname-OST0000 || {
2308 error_noexit "Start OI scrub on OST0"
2312 # The oi_scrub should be on ost1, but for test_32(),
2313 # all on the SINGLEMDS.
2314 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
2315 osd-ldiskfs.$fsname-OST0000.oi_scrub |
2316 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
2317 error_noexit "Failed to get the expected 'completed'"
2321 local UPDATED=$($r $LCTL get_param -n \
2322 osd-ldiskfs.$fsname-OST0000.oi_scrub |
2323 awk '/^updated/ { print $2 }')
2324 [ $UPDATED -ge 1 ] || {
2325 error_noexit "Only $UPDATED objects have been converted"
2330 if [[ "$dne_upgrade" != "no" ]] || $mdt2_is_available; then
2331 echo "== dne upgrate ($dne_upgrade) or mdt2 available =="
2332 if [[ $PERM_CMD =~ "set_param" ]]; then
2333 $r $PERM_CMD mdc.$fsname-MDT0001*.import=connection=$nid || {
2334 error_noexit "Setting MDT1 \"failover.node\""
2338 $r $PERM_CMD mdc.$fsname-MDT0001-*.max_rpcs_in_flight=9 || {
2339 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
2342 $r $PERM_CMD lov.$fsname-MDT0001-*.stripesize=4M || {
2343 error_noexit "Setting MDT1 \"lov.stripesize\""
2347 $r $PERM_CMD $fsname-MDT0001.failover.node=$nid || {
2348 error_noexit "Setting MDT1 \"failover.node\""
2351 $r $PERM_CMD $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
2352 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
2355 $r $PERM_CMD $fsname-MDT0001.lov.stripesize=4M || {
2356 error_noexit "Setting MDT1 \"lov.stripesize\""
2362 if [ "$writeconf" ]; then
2363 echo "== writeconf and client mount =="
2364 $MOUNT_CMD $nid:/$fsname -o user_xattr $tmp/mnt/lustre || {
2365 error_noexit "Mounting the client"
2369 shall_cleanup_lustre=true
2370 $r $LCTL set_param debug="$PTLDEBUG"
2372 t32_verify_quota $SINGLEMDS $fsname $tmp/mnt/lustre || {
2373 error_noexit "verify quota failed"
2377 if $r test -f $tmp/list; then
2378 echo "== list verification =="
2380 # There is not a Test Framework API to copy files to or
2381 # from a remote node.
2383 # LU-2393 - do both sorts on same node to ensure locale
2385 local list_file=$tmp/list
2387 if $mdt2_is_available; then
2388 if [[ -d $tmp/mnt/lustre/striped_dir_old ]] &&
2389 $r test -f $tmp/list2; then
2390 list_file=$tmp/list2
2391 pushd $tmp/mnt/lustre/striped_dir_old
2393 pushd $tmp/mnt/lustre/remote_dir
2396 pushd $tmp/mnt/lustre
2398 $r cat $list_file | sort -k 6 >$tmp/list.orig
2399 BLOCKSIZE=1024 ls -Rni --time-style=+%s | sort -k 6 |
2400 sed 's/\. / /' >$tmp/list || {
2406 # 32-bit and 64-bit clients use different algorithms to
2407 # convert FIDs into inode numbers. Hence, remove the
2408 # inode numbers from the lists, if the original list was
2409 # created on an architecture with different number of
2412 if [ $(t32_bits_per_long $(uname -m)) != \
2413 $(t32_bits_per_long $img_arch) ]; then
2414 echo "Different number of bits per \"long\"" \
2415 "from the disk image"
2416 for list in list.orig list; do
2417 sed -i -e 's/^[0-9]\+[ \t]\+//' \
2421 if ! diff -ub $tmp/list.orig $tmp/list; then
2422 error_noexit "list verification failed"
2426 echo "== list verification skipped =="
2429 #non-dom upgrade to dom
2430 if [ "$dom_upgrade" != "no" ]; then
2431 echo "== dom upgrade =="
2432 $LFS setstripe -E 1M -L mdt -E EOF $tmp/mnt/lustre/dom || {
2433 error_noexit "Verify DoM creation"
2436 [ $($LFS getstripe -L $tmp/mnt/lustre/dom) == "mdt" ] || {
2437 error_noexit "Verify a DoM file"
2440 dd if=/dev/urandom of=$tmp/mnt/lustre/dom bs=4096 \
2441 count=1 conv=fsync || {
2442 error_noexit "Cannot write to DoM file"
2445 [ $(stat -c%s $tmp/mnt/lustre/dom) == 4096 ] || {
2446 error_noexit "DoM: bad size after write"
2449 rm $tmp/mnt/lustre/dom
2451 set_persistent_param_and_check mds \
2452 "lod.*MDT0000*.dom_stripesize" \
2453 "$fsname-MDT0000.lod.dom_stripesize" 0 || {
2454 error_noexit "Changing \"dom_stripesize\""
2460 #$LFS setstripe -E 1M -L mdt -E -1 -S 4M $dom_file
2461 if [ "$dom_new_upgrade" != "no" ]; then
2462 if ! $mdt2_is_available; then
2463 dom_file=$tmp/mnt/lustre/dom_dir/dom_file
2465 echo "== check DoM file can be accessed =="
2466 [ $($LFS getstripe -I1 -L $dom_file) == "mdt" ] || {
2467 error_noexit "Verify a DoM file"
2470 [ $(stat -c%s $dom_file) == $((2 * 1024 * 1024)) ] || {
2471 error_noexit "DoM: bad size after write"
2476 if [ "$flr_upgrade" != "no" ]; then
2477 if ! $mdt2_is_available; then
2478 flr_file=$tmp/mnt/lustre/flr_dir/flr_file
2480 local mirror_count=$($LFS getstripe -N $flr_file)
2481 echo "== check FLR file =="
2482 [ $mirror_count == 2 ] || {
2483 error_noexit "FLR mirror count wrong"
2488 if [ "$dne_upgrade" != "no" ]; then
2489 echo "== dne upgrade striped_dir =="
2490 $LFS mkdir -i 1 -c2 $tmp/mnt/lustre/striped_dir || {
2491 error_noexit "set striped dir failed"
2494 $LFS setdirstripe -D -c2 $tmp/mnt/lustre/striped_dir
2496 pushd $tmp/mnt/lustre
2497 tar -c --exclude=./striped_dir \
2498 --exclude=./striped_dir_old \
2499 --exclude=./remote_dir -f - .|
2500 tar -xvf - -C striped_dir 1>/dev/null || {
2501 error_noexit "cp to striped dir failed"
2507 if [ "$pfl_upgrade" != "no" ]; then
2514 echo "== check PFL file =="
2515 if ! $mdt2_is_available; then
2516 pfl_file=$tmp/mnt/lustre/pfl_dir/pfl_file
2518 comp_cnt=$($LFS getstripe --component-count $pfl_file)
2519 [ $comp_cnt == 2 ] || {
2520 error_noexit "wrong comp_cnt $comp_cnt"
2524 comp_size=$($LFS getstripe -I1 -E $pfl_file)
2525 [ $comp_size == 2097152 ] || {
2526 error_noexit "wrong component size $comp_size"
2530 comp_id=$($LFS getstripe -I1 -i $pfl_file)
2531 [ $comp_id == 0 ] || {
2532 error_noexit "wrong comp id $comp_id"
2536 comp_id=$($LFS getstripe -I2 -i $pfl_file)
2537 [ $comp_id -eq 1 ] || {
2538 error_noexit "wrong comp id $comp_id"
2542 stripe_size=$($LFS getstripe -I1 -S $pfl_file)
2543 [ $stripe_size -eq 1048576 ] || {
2544 error_noexit "wrong stripe size $stripe_size"
2548 stripe_size=$($LFS getstripe -I2 -S $pfl_file)
2549 [ $comp_size -eq 2097152 ] || {
2550 error_noexit "wrong component size $comp_size"
2555 local large_xattr_dir=$tmp/mnt/lustre/large_xattr_test_dir
2557 if [[ -d $large_xattr_dir ]]; then
2558 echo "== check Large EA =="
2559 local xattr_file=$large_xattr_dir/large_xattr_file
2561 xattr_val=$(getfattr --only-values\
2562 -n user.fooattr $xattr_file) || {
2563 error_noexit "Large EA cannot be read"
2567 [[ $xattr_val == $(printf "%c" {1..4096}) ]] || {
2568 error_noexit "Wrong large EA value"
2573 if [ "$project_quota_upgrade" != "no" ]; then
2574 if ! $mdt2_is_available; then
2575 quota_dir=$tmp/mnt/lustre/project_quota_dir
2579 echo "== check Project Quota =="
2580 hardlimit=$(getquota -p $T32_PRJID global 3 \
2582 [ $hardlimit == $T32_PROLIMIT ] || {
2583 error_noexit "wrong hardlimit $hardlimit"
2588 # If it is upgrade from DNE (2.5), then rename the remote dir,
2589 # which is created in 2.5 to striped dir.
2590 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]]; then
2591 echo "== rename remote_dir =="
2592 stripe_index=$($LFS getdirstripe -i \
2593 $tmp/mnt/lustre/remote_dir)
2595 [[ $stripe_index -eq 1 ]] || {
2596 error_noexit "get index \"$stripe_index\"" \
2597 "from remote dir failed"
2600 mv $tmp/mnt/lustre/remote_dir \
2601 $tmp/mnt/lustre/striped_dir/ || {
2602 error_noexit "mv remote dir failed"
2607 # If it is upgraded from DNE (2.7), then move the striped dir
2608 # which was created in 2.7 to the new striped dir.
2609 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]] &&
2610 echo "== rename striped dir =="
2611 [[ -d $tmp/mnt/lustre/striped_dir_old ]]; then
2612 stripe_count=$($LFS getdirstripe -c \
2613 $tmp/mnt/lustre/striped_dir_old)
2614 [[ $stripe_count -eq 2 ]] || {
2615 error_noexit "get count $stripe_count" \
2616 "from striped dir failed"
2619 mv $tmp/mnt/lustre/striped_dir_old \
2620 $tmp/mnt/lustre/striped_dir/ || {
2621 error_noexit "mv striped dir failed"
2627 $r $LCTL set_param -n osd*.*.force_sync=1
2628 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
2629 error_noexit "dd failed"
2632 rm -rf $tmp/mnt/lustre/tmp_file || {
2633 error_noexit "rm failed"
2637 if $r test -f $tmp/sha1sums; then
2638 echo "== checking sha1sums =="
2639 # LU-2393 - do both sorts on same node to ensure locale
2641 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
2642 if [[ "$dne_upgrade" != "no" ]]; then
2643 if [[ -d $tmp/mnt/lustre/striped_dir/remote_dir ]]; then
2644 pushd $tmp/mnt/lustre/striped_dir/remote_dir
2646 pushd $tmp/mnt/lustre/striped_dir
2648 elif [[ "$pfl_upgrade" != "no" ]] ||
2649 [[ "$flr_upgrade" != "no" ]] ||
2650 [[ "$dom_new_upgrade" != "no" ]] ||
2651 [[ "$project_quota_upgrade" != "no" ]] ||
2652 [[ -d $tmp/mnt/lustre/remote_dir ]]; then
2653 pushd $tmp/mnt/lustre/remote_dir
2655 pushd $tmp/mnt/lustre
2658 find ! -path "*remote_dir*" ! -path "*striped_dir*" \
2659 ! -name .lustre -type f -exec sha1sum {} \; |
2660 sort -k 2 >$tmp/sha1sums || {
2662 error_noexit "sha1sum"
2668 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
2669 echo "==** find $save files start **=="
2670 find $save -type f -ls
2671 echo "==** find $save files finish **=="
2672 error_noexit "sha1sum verification failed"
2676 # if upgrade from DNE(2.5), then check remote directory
2677 # if upgrade from DNE(2.7), then check striped directory
2678 if $mdt2_is_available &&
2679 [[ "$dne_upgrade" != "no" ]]; then
2680 echo "== mdt2 upgrade sha1sum check =="
2681 local new_dir="$tmp/mnt/lustre/striped_dir"
2682 local striped_dir_old="$new_dir/striped_dir_old"
2684 local dir_list="$new_dir/remote_dir"
2685 [[ ! -d $triped_dir_old ]] ||
2686 dir_list+=" $striped_dir_old"
2688 for dir in $dir_list; do
2689 echo "== dir = $dir =="
2691 find ! -name .lustre -type f \
2692 -exec sha1sum {} \; |
2693 sort -k 2 >$tmp/sha1sums || {
2696 error_noexit "sha1sum"
2700 if ! diff -ub $tmp/sha1sums.orig \
2703 error_noexit "sha1sum $dir" \
2710 echo "== sha1sum verification skipped =="
2713 # PFL write test after sha1sum check
2714 if [ "$pfl_upgrade" != "no" ]; then
2715 local rw_len=$((3 * 1034 * 1024))
2717 if ! $mdt2_is_available; then
2718 pfl_file=$tmp/mnt/lustre/pfl_dir/pfl_file
2720 small_write $pfl_file $rw_len || {
2721 error_noexit "PFL RW Failed"
2726 if [ "$dne_upgrade" != "no" ]; then
2727 rm -rf $tmp/mnt/lustre/striped_dir || {
2728 error_noexit "remove remote dir failed"
2733 # migrate files/dirs to remote MDT, then move them back
2734 if [ "$MDS1_VERSION" -ge $(version_code 2.7.50) -a \
2735 $dne_upgrade != "no" ]; then
2736 $r $LCTL set_param -n \
2737 mdt.${fsname}*.enable_remote_dir=1 2>/dev/null
2739 echo "== test migration =="
2740 pushd $tmp/mnt/lustre
2741 for dir in $(find ! -name .lustre ! -name . -type d); do
2742 mdt_index=$($LFS getdirstripe -i $dir)
2743 stripe_cnt=$($LFS getdirstripe -c $dir)
2744 if [ $mdt_index = 0 -a $stripe_cnt -le 1 ]; then
2745 $LFS migrate -m 1 $dir || {
2747 error_noexit "migrate MDT1 failed"
2753 for dir in $(find ! -name . ! -name .lustre -type d); do
2754 mdt_index=$($LFS getdirstripe -i $dir)
2755 stripe_cnt=$($LFS getdirstripe -c $dir)
2756 if [ $mdt_index = 1 -a $stripe_cnt -le 1 ]; then
2757 $LFS migrate -m 0 $dir || {
2759 error_noexit "migrate MDT0 failed"
2768 # When adding new data verification tests, please check for
2769 # the presence of the required reference files first, like
2770 # the "sha1sums" and "list" tests above, to avoid the need to
2771 # regenerate every image for each test addition.
2774 nrpcs_orig=$($LCTL get_param \
2775 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
2776 error_noexit "Getting \"max_rpcs_in_flight\""
2779 nrpcs=$((nrpcs_orig + 5))
2781 set_persistent_param_and_check client \
2782 "mdc.$fsname-MDT0000*.max_rpcs_in_flight" \
2783 "$fsname-MDT0000.mdc.max_rpcs_in_flight" $nrpcs || {
2784 error_noexit "Changing \"max_rpcs_in_flight\""
2788 umount $tmp/mnt/lustre || {
2789 error_noexit "Unmounting the client"
2792 shall_cleanup_lustre=false
2794 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
2795 error_noexit "Mounting the client"
2799 if [ "$mds1_FSTYPE" == ldiskfs -a \
2800 "$project_upgrade" != "no" ]; then
2801 ! $LFS project -d -p 1 $tmp/mnt/lustre/* &&
2802 error_noexit "set project failed" &&
2806 [[ $(do_facet mds1 pgrep orph_.*-MDD | wc -l) == 0 ]] ||
2807 error "MDD orphan cleanup thread not quit"
2809 umount $tmp/mnt/lustre || {
2810 error_noexit "Unmounting the client"
2814 if [[ "$dne_upgrade" != "no" ]] || $mdt2_is_available; then
2815 $r $UMOUNT $tmp/mnt/mdt1 || {
2816 error_noexit "Unmounting the MDT2"
2819 if [[ "$mds1_FSTYPE" == zfs ]]; then
2820 $r "$ZPOOL export t32fs-mdt2"
2822 shall_cleanup_mdt1=false
2825 $r $UMOUNT $tmp/mnt/mdt || {
2826 error_noexit "Unmounting the MDT"
2829 if [[ "$mds1_FSTYPE" == zfs ]]; then
2830 $r "$ZPOOL export t32fs-mdt1"
2832 shall_cleanup_mdt=false
2834 $r $UMOUNT $tmp/mnt/ost || {
2835 error_noexit "Unmounting the OST"
2838 if $ost2_is_available; then
2839 $r $UMOUNT $tmp/mnt/ost1 || {
2840 error_noexit "Unmounting the OST1"
2843 shall_cleanup_ost1=false
2846 if [[ $ost1_FSTYPE == zfs ]]; then
2847 $r "$ZPOOL export t32fs-ost1"
2848 if $ost2_is_available; then
2849 $r "$ZPOOL export t32fs-ost2"
2852 shall_cleanup_ost=false
2854 t32_reload_modules $node || {
2855 error_noexit "Reloading modules"
2859 if [[ "$mds1_FSTYPE" == zfs ]]; then
2860 local poolname=t32fs-mdt1
2862 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
2863 $ZPOOL import -f -d $tmp $poolname"
2865 # upgrade zpool to latest supported features,
2866 # including dnode quota accounting in 0.7.0
2867 $r "$ZPOOL upgrade $poolname"
2870 # mount a second time to make sure we didnt leave upgrade flag on
2871 $r $TUNEFS --dryrun $mdt_dev || {
2873 error_noexit "tunefs.lustre before remounting the MDT"
2877 mopts=exclude=$fsname-OST0000
2878 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2881 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2882 error_noexit "Remounting the MDT"
2885 shall_cleanup_mdt=true
2895 for tarball in $tarballs; do
2896 banner "testing $tarball upgrade"
2897 t32_test $tarball || let "rc += $?"
2901 run_test 32a "Upgrade (not live)"
2909 for tarball in $tarballs; do
2910 banner "testing $tarball upgrade with writeconf"
2911 t32_test $tarball writeconf || let "rc += $?"
2915 run_test 32b "Upgrade with writeconf"
2923 for tarball in $tarballs; do
2924 banner "testing $tarball upgrade with DNE"
2926 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
2930 run_test 32c "dne upgrade test"
2938 for tarball in $tarballs; do
2939 banner "testing $tarball upgrade with ff and project upgrade"
2940 project_upgrade="no"
2941 [[ "$MDS1_VERSION" -ge $(version_code 2.13.54) ]] &&
2942 [[ "$tarball" =~ "disk2_4-ldiskfs" ]] &&
2943 project_upgrade="yes"
2944 project_upgrade=$project_upgrade ff_convert=yes t32_test \
2949 run_test 32d "convert ff and project quota upgrade test"
2952 [[ "$MDS1_VERSION" -ge $(version_code 2.10.56) ]] ||
2953 skip "Need MDS version at least 2.10.56"
2960 for tarball in $tarballs; do
2961 [[ "$tarball" =~ "2_9" ]] ||
2962 { echo "skip $(basename $tarball)"; continue; }
2964 banner "testing $tarball upgrade with DoM"
2965 dom_upgrade=yes t32_test $tarball writeconf || let "rc += $?"
2969 run_test 32e "dom upgrade test"
2972 [[ $MDS1_VERSION -ge $(version_code 2.10.56) ]] ||
2973 skip "Need MDS version at least 2.10.56"
2980 for tarball in $tarballs; do
2981 [[ "$tarball" =~ "2_10" ]] ||
2982 { echo "skip $(basename $tarball)"; continue; }
2983 pfl_upgrade=yes project_quota_upgrade=yes \
2984 t32_test $tarball writeconf || let "rc += $?"
2988 run_test 32f "pfl upgrade test"
2991 [[ $MDS1_VERSION -ge $(version_code 2.10.56) ]] ||
2992 skip "Need MDS version at least 2.10.56"
2999 for tarball in $tarballs; do
3000 [[ $tarball =~ "2_12" ]] ||
3001 { echo "skip $(basename $tarball)"; continue; }
3002 flr_upgrade=yes dom_new_upgrade=yes \
3003 t32_test $tarball writeconf || let "rc += $?"
3007 run_test 32g "flr/dom upgrade test"
3009 test_33a() { # bug 12333, was test_33
3010 local FSNAME2=test-$testnum
3012 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3015 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
3017 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
3018 local dev=${SINGLEMDS}_dev
3019 local MDSDEV=${!dev}
3021 is_blkdev $SINGLEMDS $MDSDEV &&
3022 skip_env "mixed loopback and real device not working"
3025 local fs2mdsdev=$(mdsdevname 1_2)
3026 local fs2ostdev=$(ostdevname 1_2)
3027 local fs2mdsvdev=$(mdsvdevname 1_2)
3028 local fs2ostvdev=$(ostvdevname 1_2)
3030 if [ "$mds1_FSTYPE" == ldiskfs ]; then
3031 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
3034 if combined_mgs_mds; then
3035 local mgs_flag="--mgs"
3038 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --fsname=${FSNAME2} \
3039 --reformat $mgs_flag $mkfsoptions $fs2mdsdev $fs2mdsvdev ||
3041 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
3042 --fsname=${FSNAME2} --index=0x1fff --reformat $fs2ostdev \
3043 $fs2ostvdev || exit 10
3045 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
3046 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
3048 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3049 do_facet mgs "$PERM_CMD timeout=200" ||
3050 error "$PERM_CMD timeout=200 failed"
3052 do_facet mgs "$PERM_CMD $FSNAME2.sys.timeout=200" ||
3053 error "$PERM_CMD $FSNAME2.sys.timeout=200 failed"
3055 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
3056 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
3059 cp /etc/hosts $MOUNT2/ || error "copy /etc/hosts $MOUNT2/ failed"
3060 $LFS getstripe $MOUNT2/hosts ||
3061 error "$LFS getstripe $MOUNT2/hosts failed"
3065 if (( "$MGS_VERSION" >= $(version_code 2.15.51) &&
3066 "$MDS1_VERSION" >= $(version_code 2.15.51) )); then
3067 # test lctl del_ost on large index
3068 do_facet mgs "$LCTL del_ost -t ${FSNAME2}-OST1fff" ||
3069 error "del_ost failed with $?"
3070 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 ||
3071 error "$MOUNT_CMD failed"
3073 $LFS df | grep -q OST1fff &&
3074 error "del_ost did not remove OST1fff!"
3080 cleanup_nocli || error "cleanup_nocli failed with $?"
3082 run_test 33a "Mount ost with a large index number"
3084 test_33b() { # was test_34
3087 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
3088 # Drop lock cancelation reply during umount
3089 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
3090 do_facet client $LCTL set_param fail_loc=0x80000304
3091 #lctl set_param debug=-1
3092 umount_client $MOUNT
3093 cleanup || error "cleanup failed with $?"
3095 run_test 33b "Drop cancel during umount"
3098 (( MDS1_VERSION >= $(version_code 2.15.57) )) ||
3099 skip "Need MDS version at least 2.15.57"
3100 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3101 local tstid=${TSTID:-"$(id -u $TSTUSR)"}
3103 local qpool="qpool1"
3105 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
3107 local fs2mdsdev=$(mdsdevname 1_2)
3108 local fs2ostdev=$(ostdevname 1_2)
3109 local fs2mdsvdev=$(mdsvdevname 1_2)
3110 local fs2ostvdev=$(ostvdevname 1_2)
3112 if [ "$mds1_FSTYPE" == ldiskfs ]; then
3113 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
3116 if combined_mgs_mds; then
3117 local mgs_flag="--mgs"
3121 stack_trap unload_modules_conf
3123 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --fsname=${FSNAME} \
3124 --reformat $mgs_flag $mkfsoptions $fs2mdsdev $fs2mdsvdev ||
3126 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
3127 --fsname=${FSNAME} --index=0x7c6 --reformat $fs2ostdev \
3128 $fs2ostvdev || exit 10
3131 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
3132 stack_trap "stop fs2mds -f"
3133 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
3134 stack_trap "stop fs2ost -f"
3136 mount_client $MOUNT || error "client start failed"
3137 stack_trap "umount_client $MOUNT"
3138 mkdir_on_mdt0 $DIR/$tdir || error "cannot create $DIR/$tdir"
3139 chmod 0777 $DIR/$tdir || error "chown failed"
3140 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3141 do_facet mgs $PERM_CMD \
3142 osd-*.$FSNAME-OST*.quota_slave.enable=$QUOTA_TYPE
3144 do_facet mgs $PERM_CMD $FSNAME.quota.ost=$QUOTA_TYPE ||
3145 error "set ost quota type failed"
3148 local old_MDSCOUNT=$MDSCOUNT
3150 stack_trap "MDSCOUNT=$old_MDSCOUNT"
3152 pool_add $qpool || error "pool_add failed"
3153 pool_add_targets $qpool 0x7c6
3155 $LFS setquota -u $tstid -B20M -b 0 $MOUNT
3156 $LFS setquota -g $tstid -B20M -b 0 $MOUNT
3157 $LFS setquota -u $tstid -B20M -b 0 --pool $qpool $MOUNT
3158 $LFS setquota -g $tstid -B20M -b 0 --pool $qpool $MOUNT
3160 for i in {1..10}; do
3161 runas -u $tstid -g $tstid dd if=/dev/zero of=$DIR/$tdir/f1 \
3162 bs=1M count=30 oflag=direct
3168 #umount_client $MOUNT || error "client start failed"
3171 #cleanup_nocli || error "cleanup_nocli failed with $?"
3173 run_test 33c "Mount ost with a large index number"
3177 do_facet client "bash runmultiop_bg_pause $DIR/file O_c"
3178 manual_umount_client
3180 do_facet client killall -USR1 multiop
3181 if [ $rc -eq 0 ]; then
3182 error "umount not fail!"
3185 cleanup || error "cleanup failed with rc $?"
3187 run_test 34a "umount with opened file should be fail"
3191 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3192 stop_mds || error "Unable to stop MDS"
3194 manual_umount_client --force || error "mtab after failed umount with $?"
3196 cleanup || error "cleanup failed with $?"
3198 run_test 34b "force umount with failed mds should be normal"
3202 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3203 stop_ost || error "Unable to stop OST1"
3205 manual_umount_client --force || error "mtab after failed umount with $?"
3207 cleanup || error "cleanup failed with $?"
3209 run_test 34c "force umount with failed ost should be normal"
3211 test_35a() { # bug 12459
3214 DBG_SAVE="`$LCTL get_param -n debug`"
3215 $LCTL set_param debug="ha"
3217 log "Set up a fake failnode for the MDS"
3219 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
3220 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
3222 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3223 do_facet mgs "$PERM_CMD \
3224 mdc.*${device}*.import=connection=$(h2nettype $FAKENID)" ||
3225 error "Setting mdc.*${device}*.import=connection=\
3226 $(h2nettype $FAKENID) failed."
3228 do_facet mgs "$PERM_CMD \
3229 ${device}.failover.node=$(h2nettype $FAKENID)" ||
3230 error "Setting ${device}.failover.node=\
3231 $(h2nettype $FAKENID) failed."
3233 log "Wait for RECONNECT_INTERVAL seconds (10s)"
3236 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
3239 log "Stopping the MDT: $device"
3240 stop_mdt 1 || error "MDT0 stop fail"
3242 df $MOUNT > /dev/null 2>&1 &
3244 log "Restarting the MDT: $device"
3245 start_mdt 1 || error "MDT0 start fail"
3246 log "Wait for df ($DFPID) ... "
3249 $LCTL set_param debug="$DBG_SAVE"
3251 # retrieve from the log the first server that the client tried to
3252 # contact after the connection loss
3253 $LCTL dk $TMP/lustre-log-$TESTNAME.log
3254 NEXTCONN=`awk "/${MSG}/ {start = 1;}
3255 /import_select_connection.*$device-mdc.* using connection/ {
3257 if (\\\$NF ~ /$FAKENID/)
3263 }" $TMP/lustre-log-$TESTNAME.log`
3264 [ "$NEXTCONN" != "0" ] &&
3265 error "Tried to connect to ${NEXTCONN} not last active server"
3266 cleanup || error "cleanup failed with $?"
3267 # remove nid settings
3268 writeconf_or_reformat
3270 run_test 35a "Reconnect to the last active server first"
3272 test_35b() { # bug 18674
3273 remote_mds || skip "local MDS"
3277 $LCTL set_param debug="ha"
3279 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
3282 log "Set up a fake failnode for the MDS"
3284 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
3285 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
3287 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3288 do_facet mgs "$PERM_CMD \
3289 mdc.*${device}*.import=connection=$(h2nettype $FAKENID)" ||
3290 error "Set mdc.*${device}*.import=connection=\
3291 $(h2nettype $FAKENID) failed"
3293 do_facet mgs "$PERM_CMD \
3294 ${device}.failover.node=$(h2nettype $FAKENID)" ||
3295 error "Set ${device}.failover.node=\
3296 $(h2nettype $FAKENID) failed"
3299 local at_max_saved=0
3300 # adaptive timeouts may prevent seeing the issue
3301 if at_is_enabled; then
3302 at_max_saved=$(at_max_get mds)
3303 at_max_set 0 mds client
3306 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
3308 log "Injecting EBUSY on MDS"
3309 # Setting OBD_FAIL_MDS_RESEND=0x136
3310 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" ||
3311 error "unable to set param fail_loc=0x80000136"
3313 $LCTL set_param mdc.${FSNAME}*.stats=clear
3315 log "Creating a test file and stat it"
3316 touch $MOUNT/$tdir/$tfile || error "touch $MOUNT/$tdir/$tfile failed"
3317 stat $MOUNT/$tdir/$tfile
3319 log "Stop injecting EBUSY on MDS"
3320 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" ||
3321 error "unable to set param fail_loc=0"
3322 rm -f $MOUNT/$tdir/$tfile || error "remove $MOUNT/$tdir/$tfile failed"
3325 # restore adaptive timeout
3326 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
3328 $LCTL dk $TMP/lustre-log-$TESTNAME.log
3330 CONNCNT=$($LCTL get_param mdc.${FSNAME}*.stats |
3331 awk '/mds_connect/{print $2}')
3333 # retrieve from the log if the client has ever tried to
3334 # contact the fake server after the loss of connection
3335 FAILCONN=`awk "BEGIN {ret = 0;}
3336 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
3338 if (\\\$NF ~ /$FAKENID/) {
3343 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
3345 [ "$FAILCONN" == "0" ] &&
3346 error "The client reconnection has not been triggered"
3347 [ "$FAILCONN" == "2" ] &&
3348 error "Primary server busy, client reconnect to failover failed"
3351 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
3352 # Reconnects are supposed to be rate limited to one every 5s
3353 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] &&
3354 error "Too many reconnects $CONNCNT"
3356 cleanup || error "cleanup failed with $?"
3357 # remove nid settings
3358 writeconf_or_reformat
3360 run_test 35b "Continue reconnection retries, if the active server is busy"
3363 [ $OSTCOUNT -lt 2 ] && skip_env "needs >= 2 OSTs"
3365 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] ||
3369 local FSNAME2=test1234
3370 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3372 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
3374 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
3375 is_blkdev $SINGLEMDS $MDSDEV &&
3376 skip_env "mixed loopback and real device not working"
3379 local fs2mdsdev=$(mdsdevname 1_2)
3380 local fs2ostdev=$(ostdevname 1_2)
3381 local fs3ostdev=$(ostdevname 2_2)
3382 local fs2mdsvdev=$(mdsvdevname 1_2)
3383 local fs2ostvdev=$(ostvdevname 1_2)
3384 local fs3ostvdev=$(ostvdevname 2_2)
3387 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
3388 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
3389 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
3390 # different one than the default value here.
3391 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
3392 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
3393 add fs3ost $(mkfs_opts ost2 ${fs3ostdev}) --mgsnode=$MGSNID \
3394 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
3396 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
3397 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
3398 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
3399 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
3400 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
3402 sleep 5 # until 11778 fixed
3404 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || error "dd failed"
3406 BKTOTAL=$($LCTL get_param -n obdfilter.*.kbytestotal |
3407 awk 'BEGIN{total=0}; {total+=$1}; END{print total}')
3408 BKFREE=$($LCTL get_param -n obdfilter.*.kbytesfree |
3409 awk 'BEGIN{free=0}; {free+=$1}; END{print free}')
3410 BKAVAIL=$($LCTL get_param -n obdfilter.*.kbytesavail |
3411 awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}')
3412 STRING=$(df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}')
3413 DFTOTAL=$(echo $STRING | cut -d, -f1)
3414 DFUSED=$(echo $STRING | cut -d, -f2)
3415 DFAVAIL=$(echo $STRING | cut -d, -f3)
3416 DFFREE=$(($DFTOTAL - $DFUSED))
3418 ALLOWANCE=$((64 * $OSTCOUNT))
3420 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
3421 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
3422 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
3425 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
3426 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
3427 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
3430 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
3431 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
3432 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
3437 stop fs3ost -f || error "unable to stop OST3"
3438 stop fs2ost -f || error "unable to stop OST2"
3439 stop fs2mds -f || error "unable to stop second MDS"
3440 unload_modules_conf || error "unable unload modules"
3443 run_test 36 "df report consistency on OSTs with different block size"
3446 local mntpt=$(facet_mntpt $SINGLEMDS)
3447 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
3448 local mdsdev_sym="$TMP/sym_mdt.img"
3449 local opts=$MDS_MOUNT_OPTS
3452 if [ "$mds1_FSTYPE" != ldiskfs ]; then
3453 skip "ldiskfs only test"
3456 echo "MDS : $mdsdev"
3457 echo "SYMLINK : $mdsdev_sym"
3458 do_facet $SINGLEMDS rm -f $mdsdev_sym
3460 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
3462 echo "mount symlink device - $mdsdev_sym"
3464 if ! do_facet $SINGLEMDS test -b $mdsdev; then
3465 opts=$(csa_add "$opts" -o loop)
3469 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
3470 $mdsdev_sym $mntpt 2>&1)
3473 echo mount_op=$mount_op
3475 do_facet $SINGLEMDS "$UMOUNT $mntpt && rm -f $mdsdev_sym"
3477 if $(echo $mount_op | grep -q "unable to set tunable"); then
3478 error "set tunables failed for symlink device"
3481 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
3483 run_test 37 "verify set tunables works for symlink device"
3485 test_38() { # bug 14222
3486 local mntpt=$(facet_mntpt $SINGLEMDS)
3491 local SRC="/etc /bin"
3492 local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT)
3494 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
3495 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3496 tar cf - $FILES | tar xf - -C $DIR/$tdir ||
3497 error "copying $SRC to $DIR/$tdir"
3499 umount_client $MOUNT || error "umount_client $MOUNT failed"
3500 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3501 stop_mds || error "Unable to stop MDS"
3502 log "delete lov_objid file on MDS"
3504 mount_fstype $SINGLEMDS || error "mount MDS failed (1)"
3506 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid; rm $mntpt/lov_objid"
3508 unmount_fstype $SINGLEMDS || error "umount failed (1)"
3510 # check create in mds_lov_connect
3511 start_mds || error "unable to start MDS"
3512 mount_client $MOUNT || error "mount_client $MOUNT failed"
3514 [ $V ] && log "verifying $DIR/$tdir/$f"
3515 diff -q $f $DIR/$tdir/$f || ERROR=y
3517 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3518 if [ "$ERROR" = "y" ]; then
3519 # check it's updates in sync
3520 umount_client $MOUNT
3522 mount_fstype $SIGNLEMDS
3523 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3524 unmount_fstype $SINGLEMDS
3525 error "old and new files are different after connect" || true
3527 touch $DIR/$tdir/f2 || error "f2 file create failed"
3529 # check it's updates in sync
3530 umount_client $MOUNT || error "second umount_client $MOUNT failed"
3533 mount_fstype $SINGLEMDS || error "mount MDS failed (3)"
3535 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3536 do_facet $SINGLEMDS dd if=/dev/zero of=$mntpt/lov_objid.clear count=8
3538 unmount_fstype $SINGLEMDS || error "umount failed (3)"
3540 start_mds || error "unable to start MDS"
3541 mount_client $MOUNT || error "mount_client $MOUNT failed"
3543 [ $V ] && log "verifying $DIR/$tdir/$f"
3544 diff -q $f $DIR/$tdir/$f || ERROR=y
3546 touch $DIR/$tdir/f3 || error "f3 file create failed"
3547 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3548 umount_client $MOUNT || error "third umount_client $MOUNT failed"
3550 mount_fstype $SINGLEMDS || error "mount MDS failed (4)"
3551 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3552 unmount_fstype $SINGLEMDS || error "umount failed (4)"
3554 [[ "$ERROR" != "y" ]] ||
3555 error "old and new files are different after sync"
3557 log "files compared the same"
3558 cleanup || error "cleanup failed with $?"
3560 run_test 38 "MDS recreates missing lov_objid file from OST data"
3563 [[ -n "$(type -p perl)" ]] || skip_env "need perl for leak_finder.pl"
3567 cleanup || error "cleanup failed with $?"
3568 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
3569 error "memory leak detected" || true
3571 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
3573 test_40() { # bug 15759
3574 start_ost || error "Unable to start OST1"
3575 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
3576 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
3578 cleanup || error "cleanup failed with rc $?"
3580 run_test 40 "race during service thread startup"
3582 test_41a() { #bug 14134
3583 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3584 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
3585 skip "Loop devices does not work with nosvc option"
3588 combined_mgs_mds || skip "needs combined MGT and MDT device"
3590 start_mdt 1 -o nosvc -n
3591 if [ $MDSCOUNT -ge 2 ]; then
3592 for num in $(seq 2 $MDSCOUNT); do
3593 start_mdt $num || return
3596 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
3597 start_mdt 1 -o nomgs,force
3598 mount_client $MOUNT || error "mount_client $MOUNT failed"
3601 echo "blah blah" > $MOUNT/$tfile
3604 umount_client $MOUNT || error "umount_client $MOUNT failed"
3605 stop ost1 -f || error "unable to stop OST1"
3606 stop_mds || error "Unable to stop MDS"
3607 stop_mds || error "Unable to stop MDS on second try"
3609 run_test 41a "mount mds with --nosvc and --nomgs"
3612 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3613 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
3614 skip "Loop devices does not work with nosvc option"
3617 ! combined_mgs_mds && skip "needs combined mgs device"
3621 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3623 start_mdt 1 -o nosvc -n
3624 if [ $MDSCOUNT -ge 2 ]; then
3625 for num in $(seq 2 $MDSCOUNT); do
3626 start_mdt $num || return
3629 start_ost || error "Unable to start OST1"
3630 start_mdt 1 -o nomgs,force
3631 mount_client $MOUNT || error "mount_client $MOUNT failed"
3634 echo "blah blah" > $MOUNT/$tfile
3635 cat $MOUNT/$tfile || error "cat $MOUNT/$tfile failed"
3637 umount_client $MOUNT -f || error "umount_client $MOUNT failed"
3638 stop_ost || error "Unable to stop OST1"
3639 stop_mds || error "Unable to stop MDS"
3640 stop_mds || error "Unable to stop MDS on second try"
3642 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
3645 local oss_list=$(comma_list $(osts_nodes))
3647 [[ "$MDS1_VERSION" -ge $(version_code 2.6.52) ]] ||
3648 [[ "$MDS1_VERSION" -ge $(version_code 2.5.26) &&
3649 "$MDS1_VERSION" -lt $(version_code 2.5.50) ]] ||
3650 [[ "$MDS1_VERSION" -ge $(version_code 2.5.4) &&
3651 "$MDS1_VERSION" -lt $(version_code 2.5.11) ]] ||
3652 skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"
3654 # ensure mds1 ost1 have been created even if running sub-test standalone
3657 cleanup || error "cleanup failed"
3659 # using directly mount command instead of start() function to avoid
3660 # any side effect of // with others/externals tools/features
3661 # ("zpool import", ...)
3663 # MDT concurrent start
3665 LOAD_MODULES_REMOTE=true load_modules
3666 do_facet $SINGLEMDS "lsmod | grep -q libcfs" ||
3667 error "MDT concurrent start: libcfs module not loaded"
3669 local mds1dev=$(mdsdevname 1)
3670 local mds1mnt=$(facet_mntpt mds1)
3671 local mds1opts=$MDS_MOUNT_OPTS
3673 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3674 ! do_facet mds1 test -b $mds1dev; then
3675 mds1opts=$(csa_add "$mds1opts" -o loop)
3677 if [[ "$mds1_FSTYPE" == zfs ]]; then
3678 import_zpool mds1 || return ${PIPESTATUS[0]}
3681 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3682 do_facet mds1 "$LCTL set_param fail_loc=0x80000716"
3684 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts &
3687 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts
3692 do_facet mds1 "$LCTL set_param fail_loc=0x0"
3693 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3694 echo "1st MDT start succeed"
3695 echo "2nd MDT start failed with $rc2"
3696 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3697 echo "1st MDT start failed with $rc"
3698 echo "2nd MDT start succeed"
3701 error "unexpected concurrent MDT mounts result, rc=$rc rc2=$rc2"
3704 if [ $MDSCOUNT -ge 2 ]; then
3705 for num in $(seq 2 $MDSCOUNT); do
3706 start_mdt $num || return
3710 # OST concurrent start
3712 do_rpc_nodes $oss_list "lsmod | grep -q libcfs" ||
3713 error "OST concurrent start: libcfs module not loaded"
3715 local ost1dev=$(ostdevname 1)
3716 local ost1mnt=$(facet_mntpt ost1)
3717 local ost1opts=$OST_MOUNT_OPTS
3719 if [ "$ost1_FSTYPE" == ldiskfs ] &&
3720 ! do_facet ost1 test -b $ost1dev; then
3721 ost1opts=$(csa_add "$ost1opts" -o loop)
3723 if [[ "$ost1_FSTYPE" == zfs ]]; then
3724 import_zpool ost1 || return ${PIPESTATUS[0]}
3727 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3728 do_facet ost1 "$LCTL set_param fail_loc=0x80000716"
3730 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts &
3733 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts
3737 do_facet ost1 "$LCTL set_param fail_loc=0x0"
3738 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3739 echo "1st OST start succeed"
3740 echo "2nd OST start failed with $rc2"
3741 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3742 echo "1st OST start failed with $rc"
3743 echo "2nd OST start succeed"
3747 error "unexpected concurrent OST mounts result, rc=$rc rc2=$rc2"
3753 # verify everything ok
3758 error "MDT(s) start failed"
3766 error "OST(s) start failed"
3774 error "client start failed"
3781 error "client mount failed"
3785 run_test 41c "concurrent mounts of MDT/OST should all fail but one"
3787 test_42() { #bug 14693
3791 check_mount || error "client was not mounted"
3793 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3794 PARAM="llite.$FSNAME-*.some_wrong_param"
3796 PARAM="$FSNAME.llite.some_wrong_param"
3799 do_facet mgs $PERM_CMD $PARAM=10
3800 umount_client $MOUNT ||
3801 error "unmounting client failed with invalid llite param"
3802 mount_client $MOUNT ||
3803 error "mounting client failed with invalid llite param"
3805 do_facet mgs $PERM_CMD $PARAM=20
3806 cleanup || error "stopping $FSNAME failed with invalid sys param"
3808 check_mount || error "client was not mounted with invalid sys param"
3809 cleanup || error "stopping $FSNAME failed with invalid sys param"
3811 run_test 42 "allow client/server mount/unmount with invalid config param"
3814 [[ "$MGS_VERSION" -ge $(version_code 2.5.58) ]] ||
3815 skip "Need MDS version at least 2.5.58"
3816 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
3819 USER1=$(getent passwd | grep :$ID1:$ID1: | cut -d: -f1)
3820 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1"
3823 chmod ugo+x $DIR || error "chmod 0 failed"
3824 set_persistent_param_and_check mds1 \
3825 "mdt.$FSNAME-MDT0000.root_squash" \
3826 "$FSNAME.mdt.root_squash" \
3828 wait_update $HOSTNAME \
3829 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3831 error "check llite root_squash failed!"
3832 set_persistent_param_and_check mds1 \
3833 "mdt.$FSNAME-MDT0000.nosquash_nids" \
3834 "$FSNAME.mdt.nosquash_nids" \
3836 wait_update $HOSTNAME \
3837 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3839 error "check llite nosquash_nids failed!"
3842 # create set of test files
3844 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
3845 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
3846 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
3848 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
3849 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
3851 mkdir_on_mdt0 $DIR/$tdir-rootdir || error "mkdir failed"
3852 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
3853 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
3855 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
3856 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
3857 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
3860 # check root_squash:
3861 # set root squash UID:GID to RUNAS_ID
3862 # root should be able to access only files owned by RUNAS_ID
3864 set_persistent_param_and_check mds1 \
3865 "mdt.$FSNAME-MDT0000.root_squash" \
3866 "$FSNAME.mdt.root_squash" \
3867 "$RUNAS_ID:$RUNAS_ID"
3868 wait_update $HOSTNAME \
3869 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3870 "$RUNAS_ID:$RUNAS_ID" ||
3871 error "check llite root_squash failed!"
3873 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
3874 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3875 error "$ST: root read permission is denied"
3876 echo "$ST: root read permission is granted - ok"
3879 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3880 error "$ST: root write permission is denied"
3881 echo "$ST: root write permission is granted - ok"
3883 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3884 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3885 error "$ST: root read permission is granted"
3886 echo "$ST: root read permission is denied - ok"
3889 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3890 error "$ST: root write permission is granted"
3891 echo "$ST: root write permission is denied - ok"
3893 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3894 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null &&
3895 error "$ST: root unlink permission is granted"
3896 echo "$ST: root unlink permission is denied - ok"
3898 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null &&
3899 error "$ST: root create permission is granted"
3900 echo "$ST: root create permission is denied - ok"
3903 # check root_squash is enforced independently
3904 # of client cache content
3906 # access file by USER1, keep access open
3907 # root should be denied access to user file
3909 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
3913 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
3914 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3915 { kill $pid; error "$ST: root read permission is granted"; }
3916 echo "$ST: root read permission is denied - ok"
3919 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3920 { kill $pid; error "$ST: root write permission is granted"; }
3921 echo "$ST: root write permission is denied - ok"
3927 # check nosquash_nids:
3928 # put client's NID into nosquash_nids list,
3929 # root should be able to access root file after that
3931 local NIDLIST=$($LCTL list_nids all | tr '\n' ' ')
3932 NIDLIST="2@gni $NIDLIST 192.168.0.[2,10]@tcp"
3933 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
3934 set_persistent_param_and_check mds1 \
3935 "mdt.$FSNAME-MDT0000.nosquash_nids" \
3936 "$FSNAME-MDTall.mdt.nosquash_nids" \
3938 wait_update $HOSTNAME \
3939 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3941 error "check llite nosquash_nids failed!"
3943 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3944 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3945 error "$ST: root read permission is denied"
3946 echo "$ST: root read permission is granted - ok"
3949 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3950 error "$ST: root write permission is denied"
3951 echo "$ST: root write permission is granted - ok"
3953 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3954 rm $DIR/$tdir-rootdir/tfile-1 ||
3955 error "$ST: root unlink permission is denied"
3956 echo "$ST: root unlink permission is granted - ok"
3957 touch $DIR/$tdir-rootdir/tfile-2 ||
3958 error "$ST: root create permission is denied"
3959 echo "$ST: root create permission is granted - ok"
3960 cleanup || error "cleanup failed with $?"
3962 run_test 43a "check root_squash and nosquash_nids"
3964 test_43b() { # LU-5690
3965 [[ "$MGS_VERSION" -ge $(version_code 2.7.62) ]] ||
3966 skip "Need MGS version 2.7.62+"
3968 if [[ -z "$fs2mds_DEV" ]]; then
3969 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
3970 skip_env "mixed loopback and real device not working"
3973 local fs2mdsdev=$(mdsdevname 1_2)
3974 local fs2mdsvdev=$(mdsvdevname 1_2)
3976 # temporarily use fs2mds as fs2mgs
3978 local fs2mgsdev=$fs2mdsdev
3979 local fs2mgsvdev=$fs2mdsvdev
3981 local fsname=test1234
3983 load_module llite/lustre
3984 local client_ip=$(host_nids_address $HOSTNAME $NETTYPE)
3985 local host=${client_ip//*./}
3986 local net=${client_ip/%$host/}
3987 local nosquash_nids=$(h2nettype $net[$host,$host,$host])
3989 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
3990 --param mdt.root_squash=$RUNAS_ID:$RUNAS_ID \
3991 --param mdt.nosquash_nids=$nosquash_nids \
3992 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
3993 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS || error "start fs2mgs failed"
3994 stop $fs2mgs -f || error "stop fs2mgs failed"
3995 cleanup || error "cleanup failed with $?"
3997 run_test 43b "parse nosquash_nids with commas in expr_list"
4001 check_mount || error "check_mount"
4002 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
4004 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
4005 for VAL in $UUIDS; do
4006 NID=$(echo $VAL | cut -d= -f1)
4007 CLUUID=$(echo $VAL | cut -d= -f2)
4008 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
4010 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
4011 cleanup || error "cleanup failed with $?"
4013 run_test 44 "mounted client proc entry exists"
4017 check_mount || error "check_mount"
4018 stop_mds || error "Unable to stop MDS"
4022 #define OBD_FAIL_PTLRPC_LONG_REPL_UNLINK 0x50f
4023 do_facet client "$LCTL set_param fail_loc=0x8000050f"
4026 manual_umount_client --force || error "manual_umount_client failed"
4027 do_facet client "$LCTL set_param fail_loc=0x0"
4028 start_mds || error "unable to start MDS"
4029 mount_client $MOUNT || error "mount_client $MOUNT failed"
4030 cleanup || error "cleanup failed with $?"
4032 run_test 45 "long unlink handling in ptlrpcd"
4039 umount_client $MOUNT2 || rc=$?
4040 umount_client $MOUNT || rc=$?
4041 while [ $count -gt 0 ]; do
4042 stop ost${count} -f || rc=$?
4046 cleanup_nocli || rc=$?
4047 #writeconf to remove all ost2 traces for subsequent tests
4048 writeconf_or_reformat
4053 echo "Testing with $OSTCOUNT OSTs"
4055 start_mds || error "unable to start MDS"
4056 #first client should see only one ost
4057 start_ost || error "Unable to start OST1"
4058 wait_osc_import_state mds ost FULL
4060 mount_client $MOUNT || error "mount_client $MOUNT failed"
4061 trap "cleanup_46a $OSTCOUNT" EXIT ERR
4064 for (( i=2; i<=$OSTCOUNT; i++ )); do
4065 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
4066 error "start_ost$i $(ostdevname $i) failed"
4069 # wait until osts in sync
4070 for (( i=2; i<=$OSTCOUNT; i++ )); do
4071 wait_osc_import_state mds ost$i FULL
4072 wait_osc_import_ready client ost$i
4075 #second client see all ost's
4077 mount_client $MOUNT2 || error "mount_client failed"
4078 $LFS setstripe -c -1 $MOUNT2 ||
4079 error "$LFS setstripe -c -1 $MOUNT2 failed"
4080 $LFS getstripe $MOUNT2 || error "$LFS getstripe $MOUNT2 failed"
4082 echo "ok" > $MOUNT2/widestripe
4083 $LFS getstripe $MOUNT2/widestripe ||
4084 error "$LFS getstripe $MOUNT2/widestripe failed"
4085 # fill acl buffer for avoid expand lsm to them
4086 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd |
4088 setfacl -m $acl $MOUNT2/widestripe
4092 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
4094 cleanup_46a $OSTCOUNT || error "cleanup_46a failed"
4096 run_test 46a "handle ost additional - wide striped file"
4101 check_mount || error "check_mount failed"
4102 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
4106 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
4107 if echo $ns | grep "MDT[[:digit:]]*"; then
4110 lrs=$(echo $ns | sed 's/.*lru_size=//')
4111 lru_size[count]=$lrs
4116 facet_failover $SINGLEMDS
4117 client_up || error "client_up failed"
4120 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
4121 if echo $ns | grep "MDT[[:digit:]]*"; then
4124 lrs=$(echo $ns | sed 's/.*lru_size=//')
4125 if ! test "$lrs" -eq "${lru_size[count]}"; then
4126 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
4127 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
4132 cleanup || error "cleanup failed with $?"
4134 run_test 47 "server restart does not make client loss lru_resize settings"
4139 # reformat after this test is needed - if the test fails,
4140 # we will have unkillable file at FS
4144 test_48() { # bz-17636 LU-7473
4148 check_mount || error "check_mount failed"
4150 $LFS setstripe -c -1 $MOUNT ||
4151 error "$LFS setstripe -c -1 $MOUNT failed"
4152 $LFS getstripe $MOUNT || error "$LFS getstripe $MOUNT failed"
4154 echo "ok" > $MOUNT/widestripe
4155 $LFS getstripe $MOUNT/widestripe ||
4156 error "$LFS getstripe $MOUNT/widestripe failed"
4158 # In the future, we may introduce more EAs, such as selinux, enlarged
4159 # LOV EA, and so on. These EA will use some EA space that is shared by
4160 # ACL entries. So here we only check some reasonable ACL entries count,
4161 # instead of the max number that is calculated from the max_ea_size.
4162 if [ "$MDS1_VERSION" -lt $(version_code 2.8.57) ]; then
4163 count=28 # hard coded of RPC protocol
4164 elif large_xattr_enabled; then
4165 count=4500 # max_num 8187 max_ea_size = 65452
4166 # not create too many (4500) to save test time
4168 count=450 # max_num 497 max_ea_size = 4012
4171 echo "It is expected to hold at least $count ACL entries"
4172 trap cleanup_48 EXIT ERR
4173 for ((i = 0; i < $count; i++)) do
4174 setfacl -m u:$((i + 100)):rw $MOUNT/widestripe ||
4175 error "Fail to setfacl for $MOUNT/widestripe at $i"
4178 cancel_lru_locks mdc
4179 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
4180 local r_count=$(getfacl $MOUNT/widestripe | grep "user:" | wc -l)
4182 count=$((count + 1)) # for the entry "user::rw-"
4184 [ $count -eq $r_count ] ||
4185 error "Expected ACL entries $count, but got $r_count"
4189 run_test 48 "too many acls on file"
4191 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
4192 test_49a() { # bug 17710
4193 local timeout_orig=$TIMEOUT
4194 local ldlm_timeout_orig=$LDLM_TIMEOUT
4195 local LOCAL_TIMEOUT=20
4197 LDLM_TIMEOUT=$LOCAL_TIMEOUT
4198 TIMEOUT=$LOCAL_TIMEOUT
4202 check_mount || error "client mount failed"
4204 echo "check ldlm_timout..."
4205 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
4206 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
4207 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
4209 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
4210 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
4213 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
4214 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
4217 umount_client $MOUNT || error "umount_client $MOUNT failed"
4218 stop_ost || error "problem stopping OSS"
4219 stop_mds || error "problem stopping MDS"
4221 LDLM_TIMEOUT=$ldlm_timeout_orig
4222 TIMEOUT=$timeout_orig
4224 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
4226 test_49b() { # bug 17710
4227 local timeout_orig=$TIMEOUT
4228 local ldlm_timeout_orig=$LDLM_TIMEOUT
4229 local LOCAL_TIMEOUT=20
4231 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
4232 TIMEOUT=$LOCAL_TIMEOUT
4236 check_mount || error "client mount failed"
4238 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
4239 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
4240 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
4242 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
4243 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
4246 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
4247 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
4250 cleanup || error "cleanup failed"
4252 LDLM_TIMEOUT=$ldlm_timeout_orig
4253 TIMEOUT=$timeout_orig
4255 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
4258 # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1
4260 # Test both statfs and lfs df and fail if either one fails
4261 multiop_bg_pause $1 f_
4264 killall -USR1 multiop
4265 [ $RC -ne 0 ] && log "lazystatfs multiop failed"
4266 wait $PID || { RC=$?; log "multiop return error "; }
4268 # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1
4273 if kill -s 0 $PID; then
4276 log "lazystatfs lfs df failed to complete in 5s"
4284 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4285 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4287 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
4289 cleanup || error "cleanup failed with rc $?"
4291 run_test 50a "lazystatfs all servers available"
4295 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4296 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4298 # Wait for client to detect down OST
4299 stop_ost || error "Unable to stop OST1"
4300 wait_osc_import_state client ost DISCONN
4302 log "OSCs should all be DISCONN"
4304 lazystatfs $MOUNT || error "lazystatfs should not return EIO"
4306 umount_client $MOUNT || error "Unable to unmount client"
4307 stop_mds || error "Unable to stop MDS"
4309 run_test 50b "lazystatfs all servers down"
4312 start_mds || error "Unable to start MDS"
4313 start_ost || error "Unable to start OST1"
4314 start_ost2 || error "Unable to start OST2"
4315 mount_client $MOUNT || error "Unable to mount client"
4316 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4317 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4319 # Wait for client to detect down OST
4320 stop_ost || error "Unable to stop OST1"
4321 wait_osc_import_state mds ost DISCONN
4322 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
4324 umount_client $MOUNT || error "Unable to unmount client"
4325 stop_ost2 || error "Unable to stop OST2"
4326 stop_mds || error "Unable to stop MDS"
4327 #writeconf to remove all ost2 traces for subsequent tests
4328 writeconf_or_reformat
4330 run_test 50c "lazystatfs one server down"
4333 start_mds || error "Unable to start MDS"
4334 start_ost || error "Unable to start OST1"
4335 start_ost2 || error "Unable to start OST2"
4336 mount_client $MOUNT || error "Unable to mount client"
4337 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4338 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4340 # Issue the statfs during the window where the client still
4341 # belives the OST to be available but it is in fact down.
4342 # No failure just a statfs which hangs for a timeout interval.
4343 stop_ost || error "Unable to stop OST1"
4344 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
4346 umount_client $MOUNT || error "Unable to unmount client"
4347 stop_ost2 || error "Unable to stop OST2"
4348 stop_mds || error "Unable to stop MDS"
4349 #writeconf to remove all ost2 traces for subsequent tests
4350 writeconf_or_reformat
4352 run_test 50d "lazystatfs client/server conn race"
4359 start_mds || error "Unable to start MDS"
4360 #first client should see only one ost
4361 start_ost || error "Unable to start OST1"
4362 wait_osc_import_state mds ost FULL
4364 # Wait for client to detect down OST
4365 stop_ost || error "Unable to stop OST1"
4366 wait_osc_import_state mds ost DISCONN
4368 mount_client $MOUNT || error "Unable to mount client"
4369 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
4371 multiop_bg_pause $MOUNT _f
4375 if [ $RC1 -ne 0 ]; then
4376 log "multiop failed $RC1"
4379 sleep $(( $TIMEOUT+1 ))
4381 [ $? -ne 0 ] && error "process isn't sleep"
4382 start_ost || error "Unable to start OST1"
4383 wait $pid || error "statfs failed"
4386 umount_client $MOUNT || error "Unable to unmount client"
4387 stop_ost || error "Unable to stop OST1"
4388 stop_mds || error "Unable to stop MDS"
4390 run_test 50e "normal statfs all servers down"
4395 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
4397 start_mds || error "Unable to start mds"
4398 #first client should see only one ost
4399 start_ost || error "Unable to start OST1"
4400 wait_osc_import_state mds ost FULL
4402 start_ost2 || error "Unable to start OST2"
4403 wait_osc_import_state mds ost2 FULL
4405 # Wait for client to detect down OST
4406 stop_ost2 || error "Unable to stop OST2"
4408 wait_osc_import_state mds ost2 DISCONN
4409 mount_client $MOUNT || error "Unable to mount client"
4410 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
4412 multiop_bg_pause $MOUNT _f
4416 if [ $RC1 -ne 0 ]; then
4417 log "lazystatfs multiop failed $RC1"
4420 sleep $(( $TIMEOUT+1 ))
4422 [ $? -ne 0 ] && error "process isn't sleep"
4423 start_ost2 || error "Unable to start OST2"
4424 wait $pid || error "statfs failed"
4425 stop_ost2 || error "Unable to stop OST2"
4428 umount_client $MOUNT -f || error "Unable to unmount client"
4429 stop_ost || error "Unable to stop OST1"
4430 stop_mds || error "Unable to stop MDS"
4431 #writeconf to remove all ost2 traces for subsequent tests
4432 writeconf_or_reformat
4434 run_test 50f "normal statfs one server in down"
4437 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs"
4439 start_ost2 || error "Unable to start OST2"
4440 wait_osc_import_state mds ost2 FULL
4441 wait_osc_import_ready client ost2
4443 if [[ $PERM_CMD == *"set_param -P"* ]]; then
4444 local PARAM="osc.${FSNAME}-OST0001*.active"
4446 local PARAM="${FSNAME}-OST0001.osc.active"
4449 $LFS setstripe -c -1 $DIR/$tfile || error "$LFS setstripe failed"
4450 do_facet mgs $PERM_CMD $PARAM=0 || error "Unable to deactivate OST"
4452 umount_client $MOUNT || error "Unable to unmount client"
4453 mount_client $MOUNT || error "Unable to mount client"
4454 # This df should not cause a panic
4457 do_facet mgs $PERM_CMD $PARAM=1 || error "Unable to activate OST"
4458 rm -f $DIR/$tfile || error "unable to remove file $DIR/$tfile"
4459 umount_client $MOUNT || error "Unable to unmount client"
4460 stop_ost2 || error "Unable to stop OST2"
4461 stop_ost || error "Unable to stop OST1"
4462 stop_mds || error "Unable to stop MDS"
4463 #writeconf to remove all ost2 traces for subsequent tests
4464 writeconf_or_reformat
4466 run_test 50g "deactivated OST should not cause panic"
4470 # prepare MDT/OST, make OSC inactive for OST1
4471 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs"
4473 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
4474 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
4475 error "tunefs OST1 failed"
4476 start_mds || error "Unable to start MDT"
4477 start_ost || error "Unable to start OST1"
4478 start_ost2 || error "Unable to start OST2"
4479 mount_client $MOUNT || error "client start failed"
4481 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4483 # activatate OSC for OST1
4484 set_persistent_param_and_check client \
4485 "osc.${FSNAME}-OST0000-osc-[!M]*.active" \
4486 "${FSNAME}-OST0000.osc.active" 1
4488 mkdir $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
4489 $LFS setstripe -c -1 -i 0 $DIR/$tdir/2 ||
4490 error "$LFS setstripe $DIR/$tdir/2 failed"
4491 sleep 1 && echo "create a file after OST1 is activated"
4492 # doing some io, shouldn't crash
4493 dd if=/dev/zero of=$DIR/$tdir/2/$tfile-io bs=1M count=10
4495 # check OSC import is working
4496 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
4497 error "some OSC imports are still not connected"
4501 umount_client $MOUNT || error "Unable to umount client"
4502 stop_ost2 || error "Unable to stop OST2"
4503 cleanup_nocli || error "cleanup_nocli failed with $?"
4505 run_test 50h "LU-642: activate deactivated OST"
4508 # prepare MDT/OST, make OSC inactive for OST1
4509 [ "$MDSCOUNT" -lt "2" ] && skip "needs >= 2 MDTs"
4512 [ $(facet_fstype mds2) == zfs ] && import_zpool mds2
4513 do_facet mds2 "$TUNEFS --param mdc.active=0 $(mdsdevname 2)" ||
4514 error "tunefs MDT2 failed"
4515 start_mds || error "Unable to start MDT"
4516 start_ost || error "Unable to start OST1"
4517 start_ost2 || error "Unable to start OST2"
4518 mount_client $MOUNT || error "client start failed"
4520 mkdir_on_mdt0 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4522 if [[ $PERM_CMD == *"set_param -P"* ]]; then
4523 $PERM_CMD mdc.${FSNAME}-MDT0001-mdc-*.active=0 &&
4524 error "deactive MDC0 succeeds"
4526 $PERM_CMD ${FSNAME}-MDT0000.mdc.active=0 &&
4527 error "deactive MDC0 succeeds"
4530 # activate MDC for MDT2
4531 set_persistent_param_and_check client \
4532 "mdc.${FSNAME}-MDT0001-mdc-*.active" \
4533 "${FSNAME}-MDT0001.mdc.active" 1
4535 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
4536 if [ "$MDS1_VERSION" -ge $(version_code 2.7.60) ]
4538 wait_dne_interconnect
4540 $LFS mkdir -i1 $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
4542 createmany -o $DIR/$tdir/2/$tfile-%d 1 || error "create files failed"
4544 rm -rf $DIR/$tdir/2 || error "unlink dir failed"
4546 # deactivate MDC for MDT2
4547 set_persistent_param_and_check client \
4548 "mdc.${FSNAME}-MDT0001-mdc-*.active" \
4549 "${FSNAME}-MDT0001.mdc.active" 0
4551 wait_osp_active mds ${FSNAME}-MDT0001 1 0
4553 $LFS mkdir -i1 $DIR/$tdir/2 &&
4554 error "mkdir $DIR/$tdir/2 succeeds after deactive MDT"
4556 $LFS mkdir -i0 -c$MDSCOUNT $DIR/$tdir/striped_dir ||
4557 error "mkdir $DIR/$tdir/striped_dir fails after deactive MDT2"
4559 local stripe_count=$($LFS getdirstripe -c $DIR/$tdir/striped_dir)
4560 [ $stripe_count -eq $((MDSCOUNT - 1)) ] ||
4561 error "wrong $stripe_count != $((MDSCOUNT -1)) for striped_dir"
4564 umount_client $MOUNT || error "Unable to umount client"
4569 run_test 50i "activate deactivated MDT"
4572 local LOCAL_TIMEOUT=20
4576 check_mount || error "check_mount failed"
4578 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
4579 $LFS setstripe -c -1 $MOUNT/$tdir ||
4580 error "$LFS setstripe -c -1 $MOUNT/$tdir failed"
4581 #define OBD_FAIL_MDS_REINT_DELAY 0x142
4582 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x142"
4583 touch $MOUNT/$tdir/$tfile &
4586 start_ost2 || error "Unable to start OST1"
4588 stop_ost2 || error "Unable to stop OST1"
4589 umount_client $MOUNT -f || error "unmount $MOUNT failed"
4590 cleanup_nocli || error "stop server failed"
4591 #writeconf to remove all ost2 traces for subsequent tests
4592 writeconf_or_reformat
4594 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
4603 do_node $node mkdir -p $dest
4604 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
4606 do_node $node 'tar cf - '$*' | tar xf - -C '$dest';
4607 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
4608 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
4610 do_node $node 'getfattr -d -m "[a-z]*\\." '$*' > '$xattrs
4611 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
4621 local backup2=${TMP}/backup2
4623 do_node $node mkdir -p $backup2
4624 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
4626 do_node $node 'tar cf - '$*' | tar xf - -C '$backup2';
4627 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
4628 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
4630 do_node $node "diff -rq $backup $backup2"
4631 [ $? -eq 0 ] || { error "contents differ"; return 3; }
4633 local xattrs2=${TMP}/xattrs2
4634 do_node $node 'getfattr -d -m "[a-z]*\\." '$*' > '$xattrs2
4635 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
4637 do_node $node "diff $xattrs $xattrs2"
4638 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
4640 do_node $node "rm -rf $backup2 $xattrs2"
4641 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
4645 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4646 skip "ldiskfs only test"
4649 start_mds || error "Unable to start MDS"
4650 start_ost || error "Unable to start OST1"
4651 mount_client $MOUNT || error "Unable to mount client"
4654 local ost1mnt=$(facet_mntpt ost1)
4655 local ost1node=$(facet_active_host ost1)
4656 local ost1tmp=$TMP/conf52
4659 mkdir $DIR/$tdir || error "Unable to create $DIR/$tdir"
4660 touch $TMP/modified_first || error "Unable to create temporary file"
4661 local mtime=$(stat -c %Y $TMP/modified_first)
4662 do_node $ost1node "mkdir -p $ost1tmp &&
4663 touch -m -d @$mtime $ost1tmp/modified_first" ||
4664 error "Unable to create temporary file"
4667 $LFS setstripe -c -1 -S 1M $DIR/$tdir || error "$LFS setstripe failed"
4669 for (( i=0; i < nrfiles; i++ )); do
4670 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c ||
4671 error "multiop failed"
4676 # sync all the data and make sure no pending data on the client,
4677 # thus the SOM xattr would not be changed any more.
4678 cancel_lru_locks osc
4681 echo backup files to $TMP/$tdir
4682 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
4683 copy_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
4684 error "Unable to copy files"
4686 umount_client $MOUNT || error "Unable to umount client"
4687 stop_ost || error "Unable to stop ost1"
4689 echo mount ost1 as ldiskfs
4690 do_node $ost1node mkdir -p $ost1mnt || error "Unable to create $ost1mnt"
4691 if ! do_node $ost1node test -b $ost1_dev; then
4694 do_node $ost1node mount -t "$ost1_FSTYPE" $loop $ost1_dev \
4696 error "Unable to mount ost1 as ldiskfs"
4699 echo backup objects to $ost1tmp/objects
4700 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
4701 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
4702 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
4704 error "Unable to copy objects"
4706 # move objects to lost+found
4707 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
4708 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
4710 do_node $ost1node "umount $ost1mnt" ||
4711 error "Unable to umount ost1 as ldiskfs"
4713 start_ost || error "Unable to start OST1"
4714 mount_client $MOUNT || error "Unable to mount client"
4716 local REPAIRED=$(do_node $ost1node "$LCTL get_param \
4717 -n osd-ldiskfs.$FSNAME-OST0000.oi_scrub" |
4718 awk '/^lf_repa[ri]*ed/ { print $2 }')
4719 [ $REPAIRED -gt 0 ] ||
4720 error "Some entry under /lost+found should be repaired"
4723 diff_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
4724 error "Unable to diff files"
4726 rm -rf $TMP/$tdir $TMP/file_xattrs ||
4727 error "Unable to delete temporary files"
4728 do_node $ost1node "rm -rf $ost1tmp" ||
4729 error "Unable to delete temporary files"
4730 cleanup || error "cleanup failed with $?"
4732 run_test 52 "check recovering objects from lost+found"
4734 # Checks threads_min/max/started for some service
4736 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
4737 # parameter pattern prefix like 'ost.*.ost'.
4750 local msg="Insane $modname thread counts"
4751 local ncpts=$(check_cpt_number $facet)
4755 check_mount || return 41
4757 # We need to expand $parampat, but it may match multiple parameters, so
4758 # we'll pick the first one
4759 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
4760 error "Couldn't expand ${parampat}.threads_min parameter name"
4764 # Remove the .threads_min part
4765 paramp=${paramp%.threads_min}
4767 # Check for sanity in defaults
4768 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4770 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4772 tstarted=$(do_facet $facet "$LCTL get_param \
4773 -n ${paramp}.threads_started" || echo 0)
4774 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' ||
4776 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4778 nthrs=$(expr $tmax - $tmin)
4779 if [ $nthrs -lt $ncpts ]; then
4785 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
4786 skip_env "module parameter forced $facet thread count"
4788 # Check that we can change min/max
4789 do_facet $facet "$LCTL set_param \
4790 ${paramp}.threads_min=$((tmin + nthrs))"
4791 do_facet $facet "$LCTL set_param \
4792 ${paramp}.threads_max=$((tmax - nthrs))"
4793 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4795 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4797 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) &&
4798 $tmax2 == ($tmax - $nthrs)))' || return $?
4800 # Check that we can set min/max to the same value
4801 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4803 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$tmin"
4804 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4806 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4808 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
4810 # Check that we can't set max < min
4811 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$((tmin - 1))"
4812 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4814 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4816 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
4818 # We need to ensure that we get the module options desired; to do this
4819 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
4820 LOAD_MODULES_REMOTE=true
4823 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
4825 setmodopts -a $modname "$newvalue" oldvalue
4828 check_mount || return 41
4830 # Restore previous setting of MODOPTS_*
4831 setmodopts $modname "$oldvalue"
4833 (( $MDS1_VERSION > $(version_code 2.12.52.91) )) || {
4834 echo "skip interop for MDS < v2_12_52-91-g183cb1e3cdd2"
4838 # Check that $opts took
4839 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4841 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4843 tstarted=$(do_facet $facet \
4844 "$LCTL get_param -n ${paramp}.threads_started" || echo 0)
4845 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4851 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
4852 cleanup || error "cleanup failed with rc $?"
4854 run_test 53a "check OSS thread count params"
4858 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
4859 cleanup || error "cleanup failed with $?"
4861 run_test 53b "check MDS thread count params"
4864 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4865 skip "ldiskfs only test"
4868 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p ||
4869 error "llverdev failed with rc=$?"
4872 run_test 54a "test llverdev and partial verify of device"
4875 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4876 skip "ldiskfs only test"
4880 run_llverfs $MOUNT -p || error "llverfs failed with rc=$?"
4881 cleanup || error "cleanup failed with rc=$?"
4883 run_test 54b "test llverfs and partial verify of filesystem"
4887 local max_ost_index=$1
4889 echo -n $(((max_ost_index + 1) * 8))
4893 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4894 skip "ldiskfs only test"
4897 local mdsdev=$(mdsdevname 1)
4898 local mdsvdev=$(mdsvdevname 1)
4902 if ! combined_mgs_mds; then
4903 stop_mgs || error "stopping MGS service failed"
4904 format_mgs || error "formatting MGT failed"
4906 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
4908 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
4909 --reformat $(ostdevname 1) $(ostvdevname 1)
4915 echo checking size of lov_objid for ost index $i
4916 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" |
4917 grep ^User | awk -F 'Size: ' '{print $2}')
4918 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
4919 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
4921 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
4928 run_test 55 "check lov_objid size"
4931 local mds_journal_size_orig=$MDSJOURNALSIZE
4937 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
4938 $(ostdevname 1) $(ostvdevname 1)
4939 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
4940 $(ostdevname 2) $(ostvdevname 2)
4943 start_ost || error "Unable to start first ost (idx 10000)"
4944 start_ost2 || error "Unable to start second ost (idx 1000)"
4945 mount_client $MOUNT || error "Unable to mount client"
4949 # test instantiating PFL components with sparse index LU-15513
4950 mkdir -p $MOUNT/$tdir
4951 $LFS setstripe -E 4M -c 1 -E 1G -c 4 -S4M -E eof -c -1 $MOUNT/$tdir
4952 dd if=/dev/zero of=$MOUNT/$tdir/$tfile bs=4K count=1 seek=10k ||
4953 error "dd to second component failed"
4955 if [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
4956 [[ "$MDS1_VERSION" -ge $(version_code 2.5.4) &&
4957 "$MDS1_VERSION" -lt $(version_code 2.5.11) ]]; then
4958 wait_osc_import_state mds ost1 FULL
4959 wait_osc_import_state mds ost2 FULL
4960 $LFS setstripe --stripe-count=-1 $DIR/$tfile ||
4961 error "Unable to setstripe $DIR/$tfile"
4962 n=$($LFS getstripe --stripe-count $DIR/$tfile)
4963 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
4968 MDSJOURNALSIZE=$mds_journal_size_orig
4971 run_test 56a "check big OST indexes and out-of-index-order start"
4976 umount_client $MOUNT -f || error "unmount client failed"
4985 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs"
4987 trap cleanup_56b EXIT RETURN ERR
4990 if ! combined_mgs_mds ; then
4995 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) --index=0 --reformat \
4996 $(mdsdevname 1) $(mdsvdevname 1)
4997 add mds2 $(mkfs_opts mds2 $(mdsdevname 2)) --index=1 --reformat \
4998 $(mdsdevname 2) $(mdsvdevname 2)
4999 add mds3 $(mkfs_opts mds3 $(mdsdevname 3)) --index=1000 --reformat \
5000 $(mdsdevname 3) $(mdsvdevname 3)
5004 start_mdt 1 || error "MDT 1 (idx 0) start failed"
5005 start_mdt 2 || error "MDT 2 (idx 1) start failed"
5006 start_mdt 3 || error "MDT 3 (idx 1000) start failed"
5007 start_ost || error "Unable to start first ost"
5008 start_ost2 || error "Unable to start second ost"
5010 do_nodes $(comma_list $(mdts_nodes)) \
5011 "$LCTL set_param mdt.*.enable_remote_dir=1 \
5012 mdt.*.enable_remote_dir_gid=-1"
5014 mount_client $MOUNT || error "Unable to mount client"
5016 $LFS mkdir -c3 $MOUNT/$tdir || error "failed to make testdir"
5018 echo "This is test file 1!" > $MOUNT/$tdir/$tfile.1 ||
5019 error "failed to make test file 1"
5020 echo "This is test file 2!" > $MOUNT/$tdir/$tfile.2 ||
5021 error "failed to make test file 2"
5022 echo "This is test file 1000!" > $MOUNT/$tdir/$tfile.1000 ||
5023 error "failed to make test file 1000"
5025 rm -rf $MOUNT/$tdir || error "failed to remove testdir"
5027 $LFS mkdir -i1000 $MOUNT/$tdir.1000 ||
5028 error "create remote dir at idx 1000 failed"
5031 echo "=== START lfs df OUTPUT ==="
5033 echo "==== END lfs df OUTPUT ===="
5035 mdtcnt=$(echo -e "$output" | grep $FSNAME-MDT | wc -l)
5036 ostcnt=$(echo -e "$output" | grep $FSNAME-OST | wc -l)
5038 echo "lfs df returned mdt count $mdtcnt and ost count $ostcnt"
5039 [ $mdtcnt -eq 3 ] || error "lfs df returned wrong mdt count"
5040 [ $ostcnt -eq 2 ] || error "lfs df returned wrong ost count"
5042 echo "This is test file 1!" > $MOUNT/$tdir.1000/$tfile.1 ||
5043 error "failed to make test file 1"
5044 echo "This is test file 2!" > $MOUNT/$tdir.1000/$tfile.2 ||
5045 error "failed to make test file 2"
5046 echo "This is test file 1000!" > $MOUNT/$tdir.1000/$tfile.1000 ||
5047 error "failed to make test file 1000"
5048 rm -rf $MOUNT/$tdir.1000 || error "failed to remove remote_dir"
5051 echo "=== START lfs mdts OUTPUT ==="
5053 echo "==== END lfs mdts OUTPUT ===="
5055 echo -e "$output" | grep -v "MDTS:" | awk '{print $1}' |
5056 sed 's/://g' > $TMP/mdts-actual.txt
5057 sort $TMP/mdts-actual.txt -o $TMP/mdts-actual.txt
5059 echo -e "0\n1\n1000" > $TMP/mdts-expected.txt
5061 diff $TMP/mdts-expected.txt $TMP/mdts-actual.txt
5064 rm $TMP/mdts-expected.txt $TMP/mdts-actual.txt
5066 [ $result -eq 0 ] || error "target_obd proc file is incorrect!"
5068 run_test 56b "test target_obd correctness with nonconsecutive MDTs"
5070 test_57a() { # bug 22656
5071 do_rpc_nodes $(facet_active_host ost1) load_modules_local
5072 local NID=$(do_facet ost1 "$LCTL get_param nis" |
5073 tail -1 | awk '{print $1}')
5074 writeconf_or_reformat
5075 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
5076 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" ||
5077 error "tunefs failed"
5079 start_ost && error "OST registration from failnode should fail"
5082 run_test 57a "initial registration from failnode should fail (should return errs)"
5085 do_rpc_nodes $(facet_active_host ost1) load_modules_local
5086 local NID=$(do_facet ost1 "$LCTL get_param nis" |
5087 tail -1 | awk '{print $1}')
5088 writeconf_or_reformat
5089 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
5090 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" ||
5091 error "tunefs failed"
5093 start_ost || error "OST registration from servicenode should not fail"
5096 run_test 57b "initial registration from servicenode should not fail"
5099 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
5102 test_58() { # bug 22658
5103 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
5105 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5106 createmany -o $DIR/$tdir/$tfile-%d 100
5107 unlinkmany $DIR/$tdir/$tfile-%d 100
5108 stop_mds || error "Unable to stop MDS"
5110 local MNTDIR=$(facet_mntpt $SINGLEMDS)
5111 local devname=$(mdsdevname ${SINGLEMDS//mds/})
5113 # remove all files from the OBJECTS dir
5114 mount_fstype $SINGLEMDS
5116 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
5118 unmount_fstype $SINGLEMDS
5119 # restart MDS with missing llog files
5120 start_mds || error "unable to start MDS"
5121 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0"
5124 run_test 58 "missing llog files must not prevent MDT from mounting"
5127 start_mgsmds >> /dev/null
5128 local C1=$(count_osts)
5129 if [ $C1 -eq 0 ]; then
5130 start_ost >> /dev/null
5134 echo "original ost count: $C1 (expect > 0)"
5135 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
5136 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
5137 local C2=$(count_osts)
5138 echo "after mdt writeconf count: $C2 (expect 0)"
5139 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
5140 echo "OST start without writeconf should fail:"
5141 start_ost >> /dev/null &&
5142 error "OST start without writeconf didn't fail"
5143 echo "OST start with writeconf should succeed:"
5144 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
5145 local C3=$(count_osts)
5146 echo "after ost writeconf count: $C3 (expect 1)"
5147 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
5148 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
5149 local C4=$(count_osts)
5150 echo "after ost2 writeconf count: $C4 (expect 2)"
5151 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
5152 stop_ost2 >> /dev/null
5153 cleanup_nocli >> /dev/null
5154 #writeconf to remove all ost2 traces for subsequent tests
5155 writeconf_or_reformat
5157 run_test 59 "writeconf mount option"
5159 test_60a() { # LU-471
5160 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5161 skip "ldiskfs only test"
5166 for num in $(seq $MDSCOUNT); do
5167 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
5168 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
5169 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
5173 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
5174 [ ${PIPESTATUS[0]} -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
5176 # MDT default has dirdata feature
5177 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
5178 # we disable uninit_bg feature
5179 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
5180 # we set stride extended options
5181 echo $dump | grep stride > /dev/null || error "stride is not set"
5185 run_test 60a "check mkfs.lustre --mkfsoptions -E -O options setting"
5188 [[ "$mds1_FSTYPE" == ldiskfs ]] || skip "ldiskfs only test"
5190 local features=$(do_facet $SINGLEMDS $DUMPE2FS $(mdsdevname 1) |
5192 [ ${PIPESTATUS[0]} -eq 0 ] || error "$DUMPE2FS $(mdsdevname 1) failed"
5195 # ea_inode feature should be enabled by default for MDTs
5196 [[ "$features" =~ "ea_inode" ]] || error "ea_inode is not set"
5197 # large_dir feature should be enabled by default for MDTs
5198 [[ "$features" =~ "large_dir" ]] || error "large_dir is not set"
5200 run_test 60b "check mkfs.lustre MDT default features"
5202 test_61a() { # LU-80
5203 local lxattr=$(large_xattr_enabled)
5205 (( "$MDS1_VERSION" >= $(version_code 2.1.53) )) ||
5206 skip "Need MDS version at least 2.1.53 for large_xattr"
5208 if [[ "$mds1_FSTYPE" == ldiskfs ]] && ! large_xattr_enabled; then
5211 for ((num=1; num <= $MDSCOUNT; num++)); do
5212 do_facet mds${num} $TUNE2FS -O ea_inode \
5213 $(mdsdevname $num) ||
5214 error "tune2fs on mds $num failed"
5218 setup || error "setting up the filesystem failed"
5219 client_up || error "starting client failed"
5221 local file=$DIR/$tfile
5222 touch $file || error "touch $file failed"
5224 local large_value="$(generate_string $(max_xattr_size))"
5225 local small_value="bar"
5227 local name="trusted.big"
5228 log "save large xattr of $(max_xattr_size) bytes on $name on $file"
5229 setfattr -n $name -v $large_value $file ||
5230 error "saving $name on $file failed"
5232 local new_value=$(get_xattr_value $name $file)
5233 [[ "$new_value" != "$large_value" ]] &&
5234 error "$name different after saving"
5236 log "shrink value of $name on $file"
5237 setfattr -n $name -v $small_value $file ||
5238 error "shrinking value of $name on $file failed"
5240 new_value=$(get_xattr_value $name $file)
5241 [[ "$new_value" != "$small_value" ]] &&
5242 error "$name different after shrinking"
5244 log "grow value of $name on $file"
5245 setfattr -n $name -v $large_value $file ||
5246 error "growing value of $name on $file failed"
5248 new_value=$(get_xattr_value $name $file)
5249 [[ "$new_value" != "$large_value" ]] &&
5250 error "$name different after growing"
5252 log "check value of $name on $file after remounting MDS"
5254 new_value=$(get_xattr_value $name $file)
5255 [[ "$new_value" != "$large_value" ]] &&
5256 error "$name different after remounting MDS"
5258 log "remove large xattr $name from $file"
5259 setfattr -x $name $file || error "removing $name from $file failed"
5261 if $lxattr && [ "$mds1_FSTYPE" == ldiskfs ]; then
5262 stopall || error "stopping for e2fsck run"
5263 for num in $(seq $MDSCOUNT); do
5264 run_e2fsck $(facet_active_host mds$num) \
5265 $(mdsdevname $num) "-y" ||
5266 error "e2fsck MDT$num failed"
5268 setup_noconfig || error "remounting the filesystem failed"
5271 # need to delete this file to avoid problems in other tests
5273 cleanup || error "stopping systems failed"
5275 run_test 61a "large xattr"
5277 test_61b() { # LU-80
5278 local lxattr=$(large_xattr_enabled)
5280 (( $MDS1_VERSION >= $(version_code 2.15.51) )) ||
5281 skip "Need MDS version at least 2.15.51 for large_xattr fix"
5283 [[ "$mds1_FSTYPE" == "ldiskfs" ]] || skip "ldiskfs specific bug"
5285 if ! large_xattr_enabled; then
5288 for (( num=1; num <= $MDSCOUNT; num++ )); do
5289 do_facet mds${num} $TUNE2FS -O ea_inode \
5290 $(mdsdevname $num) ||
5291 error "tune2fs on mds $num failed"
5295 setup || error "setting up the filesystem failed"
5296 client_up || error "starting client failed"
5298 local _file=$MOUNT/panda
5299 local large_value="$(generate_string $(max_xattr_size))"
5300 local name="trusted.big"
5302 touch ${_file} || error "touch ${_file} failed"
5303 setfattr -n $name -v $large_value ${_file} ||
5304 error "saving $name on $file failed"
5306 MDT_DEV="${FSNAME}-MDT0000"
5307 MDT_DEVNAME=$(mdsdevname ${SINGLEMDS//mds/})
5309 stopall || error "stopping for e2fsck run"
5312 ino=$(do_facet $SINGLEMDS "$DEBUGFS -R 'stat /ROOT/panda' \
5313 ${MDT_DEVNAME} | grep trusted.big")
5314 ino=$(echo "${ino}" | awk '{print $2;}')
5315 echo "large ea "${ino}
5317 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"ln $ino /lost+found\\\" \
5320 setup_noconfig || error "remounting the filesystem failed"
5322 do_facet $SINGLEMDS $LCTL lfsck_start -M ${MDT_DEV} -t namespace || {
5323 error "can't start lfsck namespace"
5327 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
5328 mdd.${MDT_DEV}.lfsck_namespace |
5329 awk '/^status/ { print \\\$2 }'" "completed" 32 || {
5330 error "(2) unexpected status"
5333 stopall || error "stopping for e2fsck run"
5334 for num in $(seq $MDSCOUNT); do
5335 run_e2fsck $(facet_active_host mds$num) \
5336 $(mdsdevname $num) "-y" ||
5337 error "e2fsck MDT$num failed"
5339 setup_noconfig || error "remounting the filesystem failed"
5341 # need to delete this file to avoid problems in other tests
5343 cleanup || error "stopping systems failed"
5345 run_test 61b "large xattr"
5348 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5349 skip "ldiskfs only test"
5351 [[ "$MDS1_VERSION" -ge $(version_code 2.2.51) ]] ||
5352 skip "Need MDS version at least 2.2.51"
5355 local mdsdev=$(mdsdevname 1)
5356 local ostdev=$(ostdevname 1)
5358 echo "disable journal for mds"
5359 do_facet mds1 $TUNE2FS -O ^has_journal $mdsdev || error "tune2fs failed"
5360 start_mds && error "MDT start should fail"
5361 echo "disable journal for ost"
5362 do_facet ost1 $TUNE2FS -O ^has_journal $ostdev || error "tune2fs failed"
5363 start_ost && error "OST start should fail"
5364 cleanup || error "cleanup failed with rc $?"
5367 run_test 62 "start with disabled journal"
5370 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5371 skip "ldiskfs only test"
5374 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ldiskfs
5375 local inode_slab=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
5376 awk '/ldiskfs_inode_cache/ { print $5 / $6 }')
5377 if [ -z "$inode_slab" ]; then
5378 skip "ldiskfs module has not been loaded"
5381 if grep -q "CONFIG_DEBUG_LOCK_ALLOC=y" /boot/config-$(uname -r); then
5382 skip "test is not compatible with CONFIG_DEBUG_LOCK_ALLOC=y"
5385 echo "$inode_slab ldiskfs inodes per page"
5386 [ "${inode_slab%.*}" -ge "3" ] && return 0
5388 # If kmalloc-128 is also 1 per page - this is a debug kernel
5389 # and so this is not an error.
5390 local kmalloc128=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
5391 awk '/^(dma-kmalloc|size)-128 / { print $5 / $6 }')
5392 # 32 128-byte chunks in 4k
5393 [ "${kmalloc128%.*}" -lt "32" ] ||
5394 error "ldiskfs inode too big, only $inode_slab objs/page, " \
5395 "kmalloc128 = $kmalloc128 objs/page"
5397 run_test 63 "Verify each page can at least hold 3 ldiskfs inodes"
5400 start_mds || error "unable to start MDS"
5401 start_ost || error "Unable to start OST1"
5402 start_ost2 || error "Unable to start second ost"
5403 mount_client $MOUNT || error "Unable to mount client"
5404 stop_ost2 || error "Unable to stop second ost"
5407 umount_client $MOUNT -f || error "unmount $MOUNT failed"
5408 cleanup_nocli || error "cleanup_nocli failed with $?"
5409 #writeconf to remove all ost2 traces for subsequent tests
5410 writeconf_or_reformat
5412 run_test 64 "check lfs df --lazy "
5414 test_65() { # LU-2237
5415 # Currently, the test is only valid for ldiskfs backend
5416 [ "$mds1_FSTYPE" != ldiskfs ] &&
5417 skip "ldiskfs only test"
5419 local devname=$(mdsdevname ${SINGLEMDS//mds/})
5420 local brpt=$(facet_mntpt brpt)
5423 if ! do_facet $SINGLEMDS "test -b $devname"; then
5427 stop_mds || error "Unable to stop MDS"
5428 local obj=$(do_facet $SINGLEMDS \
5429 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
5431 if [ -z "$obj" ]; then
5432 # The MDT may be just re-formatted, mount the MDT for the
5433 # first time to guarantee the "last_rcvd" file is there.
5434 start_mds || error "fail to mount the MDS for the first time"
5435 stop_mds || error "Unable to stop MDS"
5438 # remove the "last_rcvd" file
5439 do_facet $SINGLEMDS "mkdir -p $brpt"
5440 do_facet $SINGLEMDS \
5441 "mount -t $mds1_FSTYPE $opts $devname $brpt"
5442 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
5443 do_facet $SINGLEMDS "$UMOUNT $brpt"
5445 # restart MDS, the "last_rcvd" file should be recreated.
5446 start_mds || error "fail to restart the MDS"
5447 stop_mds || error "Unable to stop MDS"
5448 obj=$(do_facet $SINGLEMDS \
5449 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
5450 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
5452 run_test 65 "re-create the lost last_rcvd file when server mount"
5455 [[ "$MGS_VERSION" -ge $(version_code 2.3.59) ]] ||
5456 skip "Need MGS version at least 2.3.59"
5458 check_versions || skip "do not replace_nids with mismatched versions"
5461 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
5462 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
5464 # add EXCLUDE records to config log, they are not to be
5465 # removed by lctl replace_nids
5466 set_conf_param_and_check mds \
5467 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
5468 "$FSNAME-OST0000.osc.active" \
5471 echo "replace_nids should fail if MDS, OSTs and clients are UP"
5472 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5473 error "replace_nids fail"
5475 umount_client $MOUNT || error "unmounting client failed"
5476 echo "replace_nids should fail if MDS and OSTs are UP"
5477 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5478 error "replace_nids fail"
5480 stop_ost || error "Unable to stop OST1"
5481 echo "replace_nids should fail if MDS is UP"
5482 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5483 error "replace_nids fail"
5485 stop_mds || error "stopping mds failed"
5487 if combined_mgs_mds; then
5488 start_mdt 1 "-o nosvc" ||
5489 error "starting mds with nosvc option failed"
5492 echo "command should accept two parameters"
5493 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
5494 error "command should accept two params"
5496 echo "correct device name should be passed"
5497 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
5498 error "wrong devname"
5500 echo "wrong nids list should not destroy the system"
5501 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
5503 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "asdfasdf, asdfadf" &&
5506 echo "replace OST nid"
5507 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
5508 error "replace nids failed"
5510 echo "command should accept two parameters"
5511 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
5512 error "command should accept two params"
5514 echo "wrong nids list should not destroy the system"
5515 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
5518 local FAKE_NIDS="192.168.0.112@tcp1,192.168.0.112@tcp2"
5519 local FAKE_FAILOVER="192.168.0.113@tcp1,192.168.0.113@tcp2"
5520 local NIDS_AND_FAILOVER="$MDS_NID,$FAKE_NIDS:$FAKE_FAILOVER"
5521 echo "set NIDs with failover"
5522 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $NIDS_AND_FAILOVER ||
5523 error "replace nids failed"
5526 echo "replace MDS nid"
5527 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
5528 error "replace nids failed"
5530 if ! combined_mgs_mds ; then
5533 stop_mds || error "Unable to stop MDS"
5536 start_mgsmds || error "start mgsmds failed"
5537 set_conf_param_and_check mds \
5538 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
5539 "$FSNAME-OST0000.osc.active" \
5541 start_ost || error "unable to start OST"
5542 mount_client $MOUNT || error "mount client failed"
5544 check_mount || error "error after nid replace"
5545 cleanup || error "cleanup failed"
5548 run_test 66 "replace nids"
5550 test_67() { #LU-2950
5551 local legacy="$TMP/legacy_lnet_config"
5552 local new="$TMP/new_routes_test"
5553 local out="$TMP/config_out_file"
5554 local verify="$TMP/conv_verify"
5555 local verify_conf="$TMP/conf_verify"
5557 # Create the legacy file that will be run through the
5558 # lustre_routes_conversion script
5559 cat <<- LEGACY_LNET_CONFIG > $legacy
5560 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
5561 tcp2 54 10.1.3.2@tcp;
5562 tcp3 10.3.4.3@tcp:3;
5566 # Create the verification file to verify the output of
5567 # lustre_routes_conversion script against.
5568 cat <<- VERIFY_LNET_CONFIG > $verify
5569 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
5570 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
5571 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
5572 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
5573 tcp4: { gateway: 10.3.3.4@tcp }
5576 # Create the verification file to verify the output of
5577 # lustre_routes_config script against
5578 cat <<- VERIFY_LNET_CONFIG > $verify_conf
5579 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
5580 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
5581 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
5582 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
5583 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
5586 $LUSTRE_ROUTES_CONVERSION $legacy $new > /dev/null
5587 if [ -f $new ]; then
5588 # verify the conversion output
5589 cmp -s $new $verify > /dev/null
5590 if [ $? -eq 1 ]; then
5591 error "routes conversion failed"
5594 lustre_routes_config --dry-run --verbose $new > $out
5595 # check that the script succeeded
5596 cmp -s $out $verify_conf > /dev/null
5597 if [ $? -eq 1 ]; then
5598 error "routes config failed"
5601 error "routes conversion test failed"
5603 # remove generated files
5604 rm -f $new $legacy $verify $verify_conf $out
5606 run_test 67 "test routes conversion and configuration"
5614 [ "$MDS1_VERSION" -ge $(version_code 2.4.53) ] ||
5615 skip "Need MDS version at least 2.4.53"
5617 umount_client $MOUNT || error "umount client failed"
5622 # START-END - the sequences we'll be reserving
5623 START=$(do_facet $SINGLEMDS \
5624 $LCTL get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
5625 END=$((START + (1 << 30)))
5626 do_facet $SINGLEMDS \
5627 $LCTL set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
5629 # reset the sequences MDT0000 has already assigned
5630 do_facet $SINGLEMDS \
5631 $LCTL set_param seq.srv*MDT0000.space=clear
5633 # remount to let the client allocate new sequence
5634 mount_client $MOUNT || error "mount client failed"
5636 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
5637 do_facet $SINGLEMDS \
5638 $LCTL get_param seq.srv*MDT0000.space
5639 $LFS path2fid $DIR/$tfile
5641 local old_ifs="$IFS"
5643 fid=($($LFS path2fid $DIR/$tfile))
5647 if [[ $seq < $END ]]; then
5648 error "used reserved sequence $seq?"
5650 cleanup || error "cleanup failed with $?"
5652 run_test 68 "be able to reserve specific sequences in FLDB"
5654 # Test 69: is about the total number of objects ever created on an OST.
5655 # so that when it is reformatted the normal MDS->OST orphan recovery won't
5656 # just "precreate" the missing objects. In the past it might try to recreate
5657 # millions of objects after an OST was reformatted
5659 [[ "$MDS1_VERSION" -lt $(version_code 2.4.2) ]] &&
5660 skip "Need MDS version at least 2.4.2"
5662 [[ "$MDS1_VERSION" -ge $(version_code 2.4.50) ]] &&
5663 [[ "$MDS1_VERSION" -lt $(version_code 2.5.0) ]] &&
5664 skip "Need MDS version at least 2.5.0"
5667 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5668 do_nodes $(comma_list $(osts_nodes)) $LCTL set_param \
5669 seq.*OST*-super.width=$DATA_SEQ_MAX_WIDTH
5671 # use OST0000 since it probably has the most creations
5672 local OSTNAME=$(ostname_from_index 0)
5673 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
5674 local last_id=$(do_facet mds1 $LCTL get_param -n \
5675 osp.$mdtosc_proc1.prealloc_last_id)
5677 # Want to have OST LAST_ID over 5 * OST_MAX_PRECREATE to
5678 # verify that the LAST_ID recovery is working properly. If
5679 # not, then the OST will refuse to allow the MDS connect
5680 # because the LAST_ID value is too different from the MDS
5681 #define OST_MAX_PRECREATE=20000
5682 local ost_max_pre=20000
5683 local num_create=$(( ost_max_pre * 5 + 1 - last_id))
5685 # If the LAST_ID is already over 5 * OST_MAX_PRECREATE, we don't
5686 # need to create any files. So, skip this section.
5687 if [ $num_create -gt 0 ]; then
5688 # Check the number of inodes available on OST0
5690 local ifree=$($LFS df -i $MOUNT |
5691 awk '/OST0000/ { print $4 }'; exit ${PIPESTATUS[0]})
5692 log "On OST0, $ifree inodes available. Want $num_create. rc=$?"
5694 $LFS setstripe -i 0 $DIR/$tdir ||
5695 error "$LFS setstripe -i 0 $DIR/$tdir failed"
5696 if [ $ifree -lt 10000 ]; then
5697 files=$(( ifree - 50 ))
5702 local j=$((num_create / files + 1))
5703 for i in $(seq 1 $j); do
5704 createmany -o $DIR/$tdir/$tfile-$i- $files ||
5705 error "createmany fail create $files files: $?"
5706 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
5707 error "unlinkmany failed unlink $files files"
5711 # delete all of the files with objects on OST0 so the
5712 # filesystem is not inconsistent later on
5713 $LFS find $MOUNT --ost 0 -print0 | xargs -0 rm
5715 umount_client $MOUNT || error "umount client failed"
5716 stop_ost || error "OST0 stop failure"
5717 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
5718 $(ostdevname 1) $(ostvdevname 1) ||
5719 error "reformat and replace $ostdev failed"
5720 start_ost || error "OST0 restart failure"
5721 wait_osc_import_state mds ost FULL
5723 mount_client $MOUNT || error "mount client failed"
5724 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
5725 local idx=$($LFS getstripe -i $DIR/$tdir/$tfile-last)
5726 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
5728 local iused=$($LFS df -i $MOUNT |
5729 awk '/OST0000/ { print $3 }'; exit ${PIPESTATUS[0]})
5730 log "On OST0, $iused used inodes rc=$?"
5731 [ $iused -ge $((ost_max_pre + 1000)) ] &&
5732 error "OST replacement created too many inodes; $iused"
5733 cleanup || error "cleanup failed with $?"
5735 run_test 69 "replace an OST with the same index"
5738 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5741 cleanup || error "cleanup failed with $?"
5743 start_mdt 1 || error "MDT0 start fail"
5745 start_ost || error "OST0 start fail"
5746 for num in $(seq 2 $MDSCOUNT); do
5747 start_mdt $num || return
5750 mount_client $MOUNT || error "mount client fails"
5752 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
5754 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5755 error "create remote dir fail"
5757 rm -rf $DIR/$tdir || error "delete dir fail"
5758 cleanup || error "cleanup failed with $?"
5760 run_test 70a "start MDT0, then OST, then MDT1"
5763 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5766 start_ost || error "OST0 start fail"
5768 start_mds || error "MDS start fail"
5770 mount_client $MOUNT || error "mount client fails"
5772 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
5774 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5775 error "create remote dir fail"
5777 rm -rf $DIR/$tdir || error "delete dir fail"
5779 cleanup || error "cleanup failed with $?"
5781 run_test 70b "start OST, MDT1, MDT0"
5784 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5787 start_mds || error "MDS start fail"
5788 start_ost || error "OST0 start fail"
5790 mount_client $MOUNT || error "mount client fails"
5791 stop_mdt 1 || error "MDT1 start fail"
5793 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
5794 echo "deactivate $mdc_for_mdt1"
5795 $LCTL --device $mdc_for_mdt1 deactivate ||
5796 error "set $mdc_for_mdt1 deactivate failed"
5798 mkdir $DIR/$tdir && error "mkdir succeed"
5800 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
5801 error "create remote dir succeed"
5803 cleanup || error "cleanup failed with $?"
5805 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
5808 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5811 start_mds || error "MDS start fail"
5812 start_ost || error "OST0 start fail"
5814 mount_client $MOUNT || error "mount client fails"
5816 stop_mdt 2 || error "MDT1 start fail"
5818 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
5820 echo "deactivate $mdc_for_mdt2"
5821 $LCTL --device $mdc_for_mdt2 deactivate ||
5822 error "set $mdc_for_mdt2 deactivate failed"
5824 mkdir $DIR/$tdir || error "mkdir fail"
5825 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
5826 error "create remote dir succeed"
5828 rm -rf $DIR/$tdir || error "delete dir fail"
5830 cleanup || error "cleanup failed with $?"
5832 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
5835 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5837 [ "$MDS1_VERSION" -ge $(version_code 2.7.62) ] ||
5838 skip "Need MDS version at least 2.7.62"
5840 reformat || error "reformat failed with $?"
5844 local mdsdev=$(mdsdevname 1)
5845 local ostdev=$(ostdevname 1)
5846 local mdsvdev=$(mdsvdevname 1)
5847 local ostvdev=$(ostvdevname 1)
5848 local opts_mds="$(mkfs_opts mds1 $mdsdev) --reformat $mdsdev $mdsvdev"
5849 local opts_ost="$(mkfs_opts ost1 $ostdev) --reformat $ostdev $ostvdev"
5851 if ! combined_mgs_mds ; then
5855 add mds1 $opts_mds || error "add mds1 failed"
5856 start_mdt 1 || error "start mdt1 failed"
5857 add ost1 $opts_ost || error "add ost1 failed"
5858 start_ost || error "start ost failed"
5859 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
5861 local soc=$(do_facet mds1 "$LCTL get_param -n \
5862 mdt.*MDT0000.sync_lock_cancel")
5863 [ $soc == "never" ] || error "SoC enabled on single MDS"
5865 for i in $(seq 2 $MDSCOUNT); do
5866 mdsdev=$(mdsdevname $i)
5867 mdsvdev=$(mdsvdevname $i)
5868 opts_mds="$(mkfs_opts mds$i $mdsdev) --reformat $mdsdev \
5870 add mds$i $opts_mds || error "add mds$i failed"
5871 start_mdt $i || error "start mdt$i fail"
5874 wait_dne_interconnect
5876 for i in $(seq $MDSCOUNT); do
5877 soc=$(do_facet mds$i "$LCTL get_param -n \
5878 mdt.*MDT000$((i - 1)).sync_lock_cancel")
5879 [ $soc == "blocking" ] || error "SoC not enabled on DNE"
5882 for i in $(seq 2 $MDSCOUNT); do
5883 stop_mdt $i || error "stop mdt$i fail"
5885 soc=$(do_facet mds1 "$LCTL get_param -n \
5886 mdt.*MDT0000.sync_lock_cancel")
5887 [ $soc == "never" ] || error "SoC enabled on single MDS"
5888 umount_client $MOUNT -f > /dev/null
5890 cleanup || error "cleanup failed with $?"
5892 run_test 70e "Sync-on-Cancel will be enabled by default on DNE"
5895 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5896 if combined_mgs_mds; then
5897 skip "needs separate MGS/MDT"
5901 start_mdt 1 || error "MDT0 start fail"
5902 start_ost || error "OST0 start fail"
5903 for num in $(seq 2 $MDSCOUNT); do
5904 start_mdt $num || return
5907 start_ost2 || error "OST1 start fail"
5909 mount_client $MOUNT || error "mount client fails"
5911 mkdir $DIR/$tdir || error "mkdir fail"
5912 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5913 error "create remote dir succeed"
5915 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5916 rm -rf $DIR/$tdir || error "delete dir fail"
5918 umount_client $MOUNT || error "umount_client failed"
5919 stop_mds || error "MDS stop fail"
5920 stop_ost || error "OST0 stop fail"
5921 stop_ost2 || error "OST1 stop fail"
5923 run_test 71a "start MDT0 OST0, MDT1, OST1"
5926 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5927 if combined_mgs_mds; then
5928 skip "needs separate MGS/MDT"
5932 for num in $(seq 2 $MDSCOUNT); do
5933 start_mdt $num || return
5935 start_ost || error "OST0 start fail"
5936 start_mdt 1 || error "MDT0 start fail"
5937 start_ost2 || error "OST1 start fail"
5939 mount_client $MOUNT || error "mount client fails"
5941 mkdir $DIR/$tdir || error "mkdir fail"
5942 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5943 error "create remote dir succeed"
5945 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5946 rm -rf $DIR/$tdir || error "delete dir fail"
5948 umount_client $MOUNT || error "umount_client failed"
5949 stop_mds || error "MDT0 stop fail"
5950 stop_ost || error "OST0 stop fail"
5951 stop_ost2 || error "OST1 stop fail"
5953 run_test 71b "start MDT1, OST0, MDT0, OST1"
5956 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5957 combined_mgs_mds && skip "needs separate MGS/MDT"
5961 start_ost || error "OST0 start fail"
5962 start_ost2 || error "OST1 start fail"
5963 for num in $(seq 2 $MDSCOUNT); do
5964 start_mdt $num || return
5966 start_mdt 1 || error "MDT0 start fail"
5968 mount_client $MOUNT || error "mount client fails"
5970 mkdir $DIR/$tdir || error "mkdir fail"
5971 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5972 error "create remote dir succeed"
5974 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5975 rm -rf $DIR/$tdir || error "delete dir fail"
5977 umount_client $MOUNT || error "umount_client failed"
5978 stop_mds || error "MDS stop fail"
5979 stop_ost || error "OST0 stop fail"
5980 stop_ost2 || error "OST1 stop fail"
5983 run_test 71c "start OST0, OST1, MDT1, MDT0"
5986 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5987 combined_mgs_mds && skip "needs separate MGS/MDT"
5991 start_ost || error "OST0 start fail"
5992 for num in $(seq 2 $MDSCOUNT); do
5993 start_mdt $num || return
5995 start_mdt 1 || error "MDT0 start fail"
5996 start_ost2 || error "OST1 start fail"
5998 mount_client $MOUNT || error "mount client fails"
6000 mkdir $DIR/$tdir || error "mkdir fail"
6001 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
6002 error "create remote dir succeed"
6004 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
6005 rm -rf $DIR/$tdir || error "delete dir fail"
6007 umount_client $MOUNT || error "umount_client failed"
6008 stop_mds || error "MDS stop fail"
6009 stop_ost || error "OST0 stop fail"
6010 stop_ost2 || error "OST1 stop fail"
6013 run_test 71d "start OST0, MDT1, MDT0, OST1"
6016 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
6017 combined_mgs_mds && skip "needs separate MGS/MDT"
6021 start_ost || error "OST0 start fail"
6022 for num in $(seq 2 $MDSCOUNT); do
6023 start_mdt $num || return
6025 start_ost2 || error "OST1 start fail"
6026 start_mdt 1 || error "MDT0 start fail"
6028 mount_client $MOUNT || error "mount client fails"
6030 mkdir $DIR/$tdir || error "mkdir fail"
6031 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
6032 error "create remote dir succeed"
6034 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
6035 rm -rf $DIR/$tdir || error "delete dir fail"
6037 umount_client $MOUNT || error "umount_client failed"
6038 stop_mds || error "MDS stop fail"
6039 stop_ost || error "OST0 stop fail"
6040 stop_ost2 || error "OST1 stop fail"
6043 run_test 71e "start OST0, MDT1, OST1, MDT0"
6045 test_72() { #LU-2634
6046 [ "$mds1_FSTYPE" != ldiskfs ] &&
6047 skip "ldiskfs only test"
6049 local mdsdev=$(mdsdevname 1)
6050 local ostdev=$(ostdevname 1)
6051 local cmd="$E2FSCK -fnvd $mdsdev"
6058 if combined_mgs_mds; then
6059 add_options='--reformat'
6061 add_options='--reformat --replace'
6064 #tune MDT with "-O extents"
6066 for num in $(seq $MDSCOUNT); do
6067 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
6068 $add_options $(mdsdevname $num) $(mdsvdevname $num) ||
6069 error "add mds $num failed"
6070 do_facet mds${num} "$TUNE2FS -O extents $(mdsdevname $num)" ||
6071 error "$TUNE2FS failed on mds${num}"
6074 add ost1 $(mkfs_opts ost1 $ostdev) $add_options $ostdev ||
6075 error "add $ostdev failed"
6076 start_mds || error "start mds failed"
6077 start_ost || error "start ost failed"
6078 mount_client $MOUNT || error "mount client failed"
6080 #create some short symlinks
6081 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6082 createmany -o $DIR/$tdir/$tfile-%d $fn
6083 echo "create $fn short symlinks"
6084 for i in $(seq -w 1 $fn); do
6085 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
6090 umount_client $MOUNT || error "umount client failed"
6091 stop_mds || error "stop mds failed"
6092 stop_ost || error "stop ost failed"
6095 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
6097 run_test 72 "test fast symlink with extents flag enabled"
6099 test_73() { #LU-3006
6100 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
6101 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
6102 error "1st tunefs failed"
6103 start_mgsmds || error "start mds failed"
6104 start_ost || error "start ost failed"
6105 mount_client $MOUNT || error "mount client failed"
6106 $LCTL get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
6107 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
6108 umount_client $MOUNT || error "umount client failed"
6112 run_test 73 "failnode to update from mountdata properly"
6116 (( $MDS1_VERSION >= $(version_code 2.15.57.16) )) ||
6117 skip "need MDS version >= 2.15.57.16 for per-device timeouts"
6120 stack_trap "cleanup"
6122 # Prepare fs2, share the mgs of fs
6123 local FSNAME2=fs15246
6124 local fs2mdsdev=$(mdsdevname 1_2)
6125 local fs2ostdev=$(ostdevname 1_2)
6126 local fs2mdsvdev=$(mdsvdevname 1_2)
6127 local fs2ostvdev=$(ostvdevname 1_2)
6129 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --fsname=$FSNAME2 \
6130 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
6131 add fs2ost $(mkfs_opts ost1 $fs2ostdev) --fsname=$FSNAME2 \
6132 --reformat $fs2ostdev $fs2ostvdev || error "add fs2ost failed"
6134 stack_trap "cleanup_fs2"
6136 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
6137 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS || error "start fs2mds failed"
6139 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
6140 $MOUNT_CMD $MGSNID:/$FSNAME2 $MOUNT2 || error "mount $MOUNT2 failed"
6142 echo "========== All mounted lustre fs ===================="
6143 mount | grep 'type lustre'
6144 echo "====================================================="
6146 # Set and check osc/ldlm_enqueue_min
6147 $LCTL set_param osc.${FSNAME}-*.ldlm_enqueue_min=99 ||
6148 error "failed to set per-device adaptive parameters on client"
6149 stack_trap "$LCTL set_param osc.${FSNAME}-*.ldlm_enqueue_min=0"
6151 local ldlm_enqueue_min
6152 ldlm_enqueue_min=$($LCTL get_param -n osc.${FSNAME}-*.ldlm_enqueue_min \
6154 (( $ldlm_enqueue_min == 99 )) ||
6155 error "wrong ldlm_enqueue_min value for osc.${FSNAME}-*"
6157 # Check fs2 as reference
6158 ldlm_enqueue_min=$($LCTL get_param -n osc.${FSNAME2}-*.ldlm_enqueue_min\
6160 (( $ldlm_enqueue_min == 0 )) ||
6161 error "wrong ldlm_enqueue_min value for osc.${FSNAME2}-*"
6163 # Set and check obdfilter/at_min
6164 do_facet ost1 $LCTL set_param obdfilter.${FSNAME}-*.at_min=1 ||
6165 error "failed to set per-device adaptive parameters on ost"
6166 stack_trap "do_facet ost1 $LCTL set_param obdfilter.${FSNAME}-*.at_min=0"
6169 at_min=$(do_facet ost1 $LCTL get_param -n obdfilter.${FSNAME}-*.at_min \
6171 (( $at_min == 1 )) ||
6172 error "wrong at_min value for obdfilter.${FSNAME}-*"
6174 # set and check mdc/at_max
6175 $LCTL set_param mdc.${FSNAME}-*.at_max=599 ||
6176 error "failed to set per-device adaptive parameters on client"
6177 stack_trap "$LCTL set_param mdc.${FSNAME}-*.at_max=0"
6180 at_max=$($LCTL get_param -n mdc.${FSNAME}-*.at_max | uniq)
6181 (( $at_max == 599 )) ||
6182 error "wrong at_max value for osc.${FSNAME}-*"
6184 # Check fs2 as reference
6185 at_max=$($LCTL get_param -n mdc.${FSNAME2}-*.at_max | uniq)
6186 (( $at_max == 0 )) ||
6187 error "wrong at_max value for osc.${FSNAME2}-*"
6189 # Set and check mds/at_max
6190 do_facet mds1 $LCTL set_param *.${FSNAME}-*.at_max=599 ||
6191 error "failed to set per-device adaptive parameters on mds"
6192 stack_trap "do_facet mds1 $LCTL set_param *.${FSNAME}-*.at_max=0"
6195 at_max=$(do_facet mds1 $LCTL get_param -n mdt.${FSNAME}-*.at_max | uniq)
6196 (( at_max == 599 )) ||
6197 error "wrong at_max value for mdt.${FSNAME}-*"
6199 # Set and check mgs&mgc/at_history
6200 local mgs_nid=$(do_facet $SINGLEMDS $LCTL list_nids | tail -1)
6201 $LCTL set_param mgc.MGC$mgs_nid.at_history=588
6202 stack_trap "$LCTL set_param mgc.MGC$mgs_nid.at_history=0"
6205 at_history=$($LCTL get_param -n mgc.MGC$mgs_nid.at_history)
6206 (( $at_history == 588 )) ||
6207 error "wrong at_history value for mgc.MGC$mgs_nid"
6209 do_facet mgs $LCTL set_param mgs.MGS.at_history=588
6210 stack_trap "do_facet mgs $LCTL set_param mgs.MGS.at_history=0"
6211 at_history=$(do_facet mgs $LCTL get_param -n mgs.MGS.at_history)
6212 (( $at_history == 588 )) ||
6213 error "wrong at_history value for mgs.MGS"
6215 run_test 74 "Test per-device adaptive timeout parameters"
6217 test_75() { # LU-2374
6218 [[ "$MDS1_VERSION" -lt $(version_code 2.4.1) ]] &&
6219 skip "Need MDS version at least 2.4.1"
6222 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
6223 --replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
6224 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
6225 --replace --reformat $(ostdevname 1) $(ostvdevname 1)"
6228 #check with default parameters
6229 add mds1 $opts_mds || error "add mds1 failed for default params"
6230 add ost1 $opts_ost || error "add ost1 failed for default params"
6232 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
6233 opts_mds=$(echo $opts_mds |
6234 sed -e "s/--index=$index/--index=$index --mdt/")
6235 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
6236 opts_ost=$(echo $opts_ost |
6237 sed -e "s/--index=$index/--index=$index --ost/")
6239 add mds1 $opts_mds || error "add mds1 failed for new params"
6240 add ost1 $opts_ost || error "add ost1 failed for new params"
6245 run_test 75 "The order of --index should be irrelevant"
6248 [[ "$MGS_VERSION" -ge $(version_code 2.4.52) ]] ||
6249 skip "Need MDS version at least 2.4.52"
6252 local MDMB_PARAM="osc.*.max_dirty_mb"
6253 echo "Change MGS params"
6254 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
6256 echo "max_dirty_mb: $MAX_DIRTY_MB"
6257 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB - 10))
6258 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
6259 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
6260 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
6261 head -1" $NEW_MAX_DIRTY_MB
6262 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
6263 echo "$MAX_DIRTY_MB"
6264 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
6265 error "error while apply max_dirty_mb"
6267 echo "Check the value is stored after remount"
6270 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
6271 head -1" $NEW_MAX_DIRTY_MB
6272 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
6273 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
6274 error "max_dirty_mb is not saved after remount"
6276 echo "Change OST params"
6277 CLIENT_PARAM="obdfilter.*.client_cache_count"
6278 local CLIENT_CACHE_COUNT
6279 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
6281 echo "client_cache_count: $CLIENT_CACHE_COUNT"
6282 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
6283 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
6284 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
6285 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
6286 head -1" $NEW_CLIENT_CACHE_COUNT
6287 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
6289 echo "$CLIENT_CACHE_COUNT"
6290 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
6291 error "error while apply client_cache_count"
6293 echo "Check the value is stored after remount"
6296 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
6297 head -1" $NEW_CLIENT_CACHE_COUNT
6298 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
6300 echo "$CLIENT_CACHE_COUNT"
6301 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
6302 error "client_cache_count is not saved after remount"
6305 run_test 76a "set permanent params with lctl across mounts"
6307 test_76b() { # LU-4783
6308 [[ "$MGS_VERSION" -ge $(version_code 2.5.57) ]] ||
6309 skip "Need MGS version at least 2.5.57"
6312 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
6313 error "start params log failed"
6316 run_test 76b "verify params log setup correctly"
6319 [[ "$MGS_VERSION" -ge $(version_code 2.8.54) ]] ||
6320 skip "Need MDS version at least 2.4.52"
6322 local MASK_PARAM="mdd.*.changelog_mask"
6323 echo "Change changelog_mask"
6324 do_facet mgs $LCTL set_param -P $MASK_PARAM=-CLOSE ||
6325 error "Can't change changlog_mask"
6326 wait_update $(facet_host mds) "$LCTL get_param -n $MASK_PARAM |
6329 echo "Check the value is stored after mds remount"
6330 stop_mds || error "Failed to stop MDS"
6331 start_mds || error "Failed to start MDS"
6332 local CHANGELOG_MASK=$(do_facet mgs $LCTL get_param -n $MASK_PARAM)
6333 echo $CHANGELOG_MASK | grep CLOSE > /dev/null &&
6334 error "changelog_mask is not changed"
6338 run_test 76c "verify changelog_mask is applied with lctl set_param -P"
6340 test_76d() { #LU-9399
6343 local xattr_cache="llite.*.xattr_cache"
6344 local cmd="$LCTL get_param -n $xattr_cache | head -1"
6345 local new=$((($(eval $cmd) + 1) % 2))
6347 echo "lctl set_param -P llite.*.xattr_cache=$new"
6348 do_facet mgs $LCTL set_param -P $xattr_cache=$new ||
6349 error "Can't change xattr_cache"
6350 wait_update $HOSTNAME "$cmd" "$new"
6352 echo "Check $xattr_cache on client $MOUNT"
6353 umount_client $MOUNT || error "umount $MOUNT failed"
6354 mount_client $MOUNT || error "mount $MOUNT failed"
6355 [ $(eval $cmd) -eq $new ] ||
6356 error "$xattr_cache != $new on client $MOUNT"
6358 echo "Check $xattr_cache on the new client $MOUNT2"
6359 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
6360 [ $(eval $cmd) -eq $new ] ||
6361 error "$xattr_cache != $new on client $MOUNT2"
6362 umount_client $MOUNT2 || error "umount $MOUNT2 failed"
6366 run_test 76d "verify llite.*.xattr_cache can be set by 'lctl set_param -P' correctly"
6368 test_77() { # LU-3445
6369 [[ "$MDS1_VERSION" -ge $(version_code 2.8.55) ]] ||
6370 skip "Need MDS version 2.8.55+ "
6372 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
6373 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
6374 skip_env "mixed loopback and real device not working"
6377 local fs2mdsdev=$(mdsdevname 1_2)
6378 local fs2ostdev=$(ostdevname 1_2)
6379 local fs2mdsvdev=$(mdsvdevname 1_2)
6380 local fs2ostvdev=$(ostvdevname 1_2)
6381 local fsname=test1234
6383 local failnid="$(h2nettype 1.2.3.4),$(h2nettype 4.3.2.1)"
6385 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
6387 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
6388 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
6389 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
6390 error "start fs2mds failed"
6392 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
6393 mgsnid="0.0.0.0@tcp,$mgsnid,$mgsnid:$mgsnid"
6395 add fs2ost --mgsnode=$mgsnid $(mkfs_opts ost1 $fs2ostdev) \
6396 --failnode=$failnid --fsname=$fsname \
6397 --reformat $fs2ostdev $fs2ostvdev ||
6398 error "add fs2ost failed"
6399 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
6401 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
6402 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
6403 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
6406 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
6409 [[ "$mds1_FSTYPE" != ldiskfs ||
6410 "$ost1_FSTYPE" != ldiskfs ]] &&
6411 skip "ldiskfs only test"
6413 # reformat the Lustre filesystem with a smaller size
6414 local saved_MDSCOUNT=$MDSCOUNT
6415 local saved_MDSSIZE=$MDSSIZE
6416 local saved_OSTCOUNT=$OSTCOUNT
6417 local saved_OSTSIZE=$OSTSIZE
6420 MDSSIZE=$((MDSSIZE - 20000))
6421 OSTSIZE=$((OSTSIZE - 20000))
6422 reformat || error "(1) reformat Lustre filesystem failed"
6423 MDSSIZE=$saved_MDSSIZE
6424 OSTSIZE=$saved_OSTSIZE
6426 # mount the Lustre filesystem
6427 setup_noconfig || error "(2) setup Lustre filesystem failed"
6430 log "create test files"
6435 mkdir $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
6437 for i in $(seq $num_files); do
6438 file=$MOUNT/$tdir/$tfile-$i
6439 dd if=/dev/urandom of=$file count=1 bs=1M || {
6440 $LCTL get_param osc.*.cur*grant*
6441 $LFS df; $LFS df -i;
6442 # stop creating files if there is no more space
6443 if [ ! -e $file ]; then
6444 num_files=$((i - 1))
6448 $LFS getstripe -v $file
6449 local ost_idx=$(LFS getstripe -i $file)
6450 do_facet ost$((ost_idx + 1)) \
6451 $LCTL get_param obdfilter.*.*grant*
6452 error "(4) create $file failed"
6456 # unmount the Lustre filesystem
6457 cleanup || error "(5) cleanup Lustre filesystem failed"
6459 # run e2fsck on the MDT and OST devices
6460 local mds_host=$(facet_active_host $SINGLEMDS)
6461 local ost_host=$(facet_active_host ost1)
6462 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
6463 local ost_dev=$(ostdevname 1)
6465 run_e2fsck $mds_host $mds_dev "-y"
6466 run_e2fsck $ost_host $ost_dev "-y"
6468 # get the original block count of the MDT and OST filesystems
6469 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
6470 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
6472 # expand the MDT and OST filesystems to the device size
6473 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
6474 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
6476 # run e2fsck on the MDT and OST devices again
6477 run_e2fsck $mds_host $mds_dev "-y"
6478 run_e2fsck $ost_host $ost_dev "-y"
6480 # mount the Lustre filesystem
6484 log "check files after expanding the MDT and OST filesystems"
6485 for i in $(seq $num_files); do
6486 file=$MOUNT/$tdir/$tfile-$i
6487 $CHECKSTAT -t file -s 1048576 $file ||
6488 error "(6) checkstat $file failed"
6492 log "create more files after expanding the MDT and OST filesystems"
6493 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
6494 file=$MOUNT/$tdir/$tfile-$i
6495 dd if=/dev/urandom of=$file count=1 bs=1M ||
6496 error "(7) create $file failed"
6499 # unmount the Lustre filesystem
6500 cleanup || error "(8) cleanup Lustre filesystem failed"
6502 # run e2fsck on the MDT and OST devices
6503 run_e2fsck $mds_host $mds_dev "-y"
6504 run_e2fsck $ost_host $ost_dev "-y"
6506 # get the maximum block count of the MDT and OST filesystems
6507 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
6508 local ost_max_blks=$(get_block_count ost1 $ost_dev)
6510 # get the minimum block count of the MDT and OST filesystems
6511 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
6512 grep minimum | sed -e 's/^.*filesystem: //g')
6513 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
6514 grep minimum | sed -e 's/^.*filesystem: //g')
6516 # shrink the MDT and OST filesystems to a smaller size
6520 if [[ $mds_max_blks -gt $mds_min_blks &&
6521 $mds_max_blks -gt $mds_orig_blks ]]; then
6522 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
6523 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
6524 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
6525 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
6526 error "shrink $SINGLEMDS to $new_blks failed"
6530 if [[ $ost_max_blks -gt $ost_min_blks &&
6531 $ost_max_blks -gt $ost_orig_blks ]]; then
6532 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
6533 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
6534 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
6535 run_resize2fs ost1 $ost_dev $new_blks ||
6536 error "shrink ost1 to $new_blks failed"
6540 # check whether the MDT or OST filesystem was shrunk or not
6542 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
6543 reformat_and_config ||
6544 error "(10) reformat Lustre filesystem failed"
6548 # run e2fsck on the MDT and OST devices again
6549 run_e2fsck $mds_host $mds_dev "-y"
6550 run_e2fsck $ost_host $ost_dev "-y"
6552 # mount the Lustre filesystem again
6556 log "check files after shrinking the MDT and OST filesystems"
6557 for i in $(seq $((num_files + 10))); do
6558 file=$MOUNT/$tdir/$tfile-$i
6559 $CHECKSTAT -t file -s 1048576 $file ||
6560 error "(11) checkstat $file failed"
6563 # unmount and reformat the Lustre filesystem
6564 cleanup || error "(12) cleanup Lustre filesystem failed"
6565 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
6567 MDSCOUNT=$saved_MDSCOUNT
6568 OSTCOUNT=$saved_OSTCOUNT
6569 reformat_and_config || error "(14) reformat Lustre filesystem failed"
6571 run_test 78 "run resize2fs on MDT and OST filesystems"
6573 test_79() { # LU-4227
6574 [[ "$MDS1_VERSION" -ge $(version_code 2.5.59) ]] ||
6575 skip "Need MDS version at least 2.5.59"
6577 local mdsdev1=$(mdsdevname 1)
6578 local mdsvdev1=$(mdsvdevname 1)
6579 local mdsdev2=$(mdsdevname 2)
6580 local mdsvdev2=$(mdsvdevname 2)
6581 local ostdev1=$(ostdevname 1)
6582 local ostvdev1=$(ostvdevname 1)
6583 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
6584 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
6585 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
6588 # remove --mgs/--mgsnode from mkfs.lustre options
6589 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
6591 mgsnode_opt=$(echo $opts_mds2 |
6592 awk '{ for ( i = 1; i < NF; i++ )
6593 if ( $i ~ "--mgsnode" ) { print $i; break } }')
6594 [ -n "$mgsnode_opt" ] &&
6595 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
6597 mgsnode_opt=$(echo $opts_ost1 |
6598 awk '{ for ( i = 1; i < NF; i++ )
6599 if ( $i ~ "--mgsnode" ) { print $i; break } }')
6600 [ -n "$mgsnode_opt" ] &&
6601 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
6603 # -MGS, format a mdt without --mgs option
6604 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
6605 error "Must specify --mgs when formatting mdt combined with mgs"
6607 # +MGS, format a mdt/ost without --mgsnode option
6608 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
6609 > /dev/null || error "start mds1 failed"
6610 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
6611 error "Must specify --mgsnode when formatting a mdt"
6612 add ost1 $opts_ost1 $ostdev1 $ostvdev1 &&
6613 error "Must specify --mgsnode when formatting an ost"
6617 run_test 79 "format MDT/OST without mgs option (should return errors)"
6620 start_mds || error "Failed to start MDT"
6621 start_ost || error "Failed to start OST1"
6622 uuid=$(do_facet ost1 $LCTL get_param -n mgc.*.uuid)
6623 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
6624 do_facet ost1 "$LCTL set_param fail_val=10 fail_loc=0x906"
6625 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x906"
6626 do_facet mgs "$LCTL set_param -n mgs/MGS/evict_client $uuid"
6628 start_ost2 || error "Failed to start OST2"
6630 do_facet ost1 "$LCTL set_param fail_loc=0"
6635 run_test 80 "mgc import reconnect race"
6637 #Save the original values of $OSTCOUNT and $OSTINDEX$i.
6639 local new_ostcount=$1
6640 saved_ostcount=$OSTCOUNT
6641 OSTCOUNT=$new_ostcount
6645 for ((i = 1; i <= $OSTCOUNT; i++ )); do
6647 eval saved_ostindex$i=${!index}
6652 # Restore the original values of $OSTCOUNT and $OSTINDEX$i.
6653 restore_ostindex() {
6657 echo "restoring OSTCOUNT=$saved_ostcount and OSTINDEXn"
6659 for ((i = 1; i <= $OSTCOUNT; i++ )); do
6660 index=saved_ostindex$i
6661 eval OSTINDEX$i=${!index}
6663 OSTCOUNT=$saved_ostcount
6666 if ! combined_mgs_mds ; then
6671 # The main purpose of this test is to ensure the OST_INDEX_LIST functions as
6672 # expected. This test uses OST_INDEX_LIST to format OSTs with a randomly
6673 # assigned index and ensures we can mount such a formatted file system
6674 test_81() { # LU-4665
6675 (( MDS1_VERSION >= $(version_code 2.6.54) )) ||
6676 skip "Need MDS version at least 2.6.54"
6677 (( OSTCOUNT >= 3 )) || skip_env "needs >= 3 OSTs"
6681 # Each time RANDOM is referenced, a random integer between 0 and 32767
6684 local saved_ostindex1=$OSTINDEX1
6685 for i in 65535 $((RANDOM + 65536)); do
6686 echo -e "\nFormat ost1 with --index=$i, should fail"
6688 if add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat \
6689 $(ostdevname 1) $(ostvdevname 1); then
6690 OSTINDEX1=$saved_ostindex1
6691 error "format ost1 with --index=$i should fail"
6694 OSTINDEX1=$saved_ostindex1
6697 stack_trap restore_ostindex
6699 # Format OSTs with random sparse indices.
6700 local rand_ost=$(((RANDOM * 2 % 65533) + 1))
6701 echo "Format $OSTCOUNT OSTs with OST_INDEX_LIST=[0,$rand_ost,65534]"
6702 OST_INDEX_LIST=[0,$rand_ost,65534] formatall ||
6703 error "formatall failed with $?"
6705 # Setup and check Lustre filesystem.
6706 start_mgsmds || error "start_mgsmds failed"
6707 for ((i = 1; i <= $OSTCOUNT; i++ )); do
6708 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6709 error "start ost$i failed"
6712 mount_client $MOUNT || error "mount client $MOUNT failed"
6713 check_mount || error "check client $MOUNT failed"
6717 local max_easize=$($LCTL get_param -n llite.*.max_easize)
6718 local xattr_size_max=$((4096 - 84)) # 4096 less ldiskfs ea overhead
6720 # XATTR_SIZE_MAX less ldiskfs ea overhead
6721 large_xattr_enabled && xattr_size_max=$((65536 - 84))
6722 (( max_easize >= xattr_size_max )) ||
6723 error "max_easize $max_easize < $xattr_size_max bytes"
6725 test_mkdir $DIR/$tdir
6726 $LFS setstripe -i $rand_ost $DIR/$tdir/$tfile ||
6727 error "error creating $tfile on ost$rand_ost"
6728 $LFS getstripe $DIR/$tdir/$tfile
6735 cmd="getstripe -i $DIR/$tdir/$tfile"
6738 (( $found == $rand_ost )) || error "index $found is not $rand_ost"
6740 cmd="find $DIR/$tdir -i $rand_ost"
6744 [[ "$found" == "$DIR/$tdir/$tfile" ]] ||
6745 error "'lfs find' returned '$found', not '$tfile' by index"
6748 uuid=$(ostuuid_from_index $rand_ost)
6750 cmd="find $DIR/$tdir -O $uuid"
6755 [[ "$found" == "$DIR/$tdir/$tfile" ]] ||
6756 error "'lfs find' returned '$found', not '$tfile' by UUID"
6758 run_test 81 "sparse OST indexing"
6760 # Here we exercise the stripe placement functionality on a file system that
6761 # has formatted the OST with a random index. With the file system the following
6762 # functionality is tested:
6764 # 1. Creating a new file with a specific stripe layout.
6766 # 2. Modifiy a existing empty file with a specific stripe layout.
6768 # 3. Ensure we fail to set the stripe layout of a file that already has one.
6770 # 4. If ost-index is defined we need to ensure it is the first entry in the
6771 # ost index list returned by lfs getstripe.
6773 # 5. Lastly ensure this functionality fails with directories.
6774 test_82a() { # LU-4665
6775 [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
6776 skip "Need MDS version at least 2.6.54"
6777 [[ $OSTCOUNT -ge 3 ]] || skip_env "needs >= 3 OSTs"
6783 # Format OSTs with random sparse indices.
6787 local LOV_V1_INSANE_STRIPE_COUNT=65532
6788 for i in $(seq $OSTCOUNT); do
6789 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
6790 ost_indices+=" $index"
6792 ost_indices=$(comma_list $ost_indices)
6794 stack_trap "restore_ostindex" EXIT
6795 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
6796 OST_INDEX_LIST=[$ost_indices] formatall
6798 # Setup Lustre filesystem.
6799 start_mgsmds || error "start_mgsmds failed"
6800 for i in $(seq $OSTCOUNT); do
6801 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6802 error "start ost$i failed"
6805 # Collect debug information - start of test
6806 do_nodes $(comma_list $(mdts_nodes)) \
6807 $LCTL get_param osp.*.prealloc_*_id
6809 mount_client $MOUNT || error "mount client $MOUNT failed"
6813 check_lfs_df_ret_val $? || error "$LFS df $MOUNT failed"
6814 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6816 stack_trap "do_nodes $(comma_list $(mdts_nodes)) \
6817 $LCTL get_param osp.*.prealloc_*_id || true" EXIT
6819 # 1. If the file does not exist, new file will be created
6820 # with specified OSTs.
6821 local file=$DIR/$tdir/$tfile-1
6822 local cmd="$LFS setstripe -o $ost_indices $file"
6824 eval $cmd || error "$cmd failed"
6825 check_stripe_count $file $OSTCOUNT
6826 check_obdidx $file $ost_indices
6827 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6828 error "write $file failed"
6830 # 2. If the file already exists and is an empty file, the file
6831 # will be attached with specified layout.
6832 file=$DIR/$tdir/$tfile-2
6833 mcreate $file || error "mcreate $file failed"
6834 cmd="$LFS setstripe -o $ost_indices $file"
6836 eval $cmd || error "$cmd failed"
6837 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6838 error "write $file failed"
6839 check_stripe_count $file $OSTCOUNT
6840 check_obdidx $file $ost_indices
6842 # 3. If the file already has a valid layout attached, the command
6843 # should fail with EBUSY.
6845 eval $cmd && error "stripe is already set on $file, $cmd should fail"
6847 # 4. If [--stripe-index|-i <start_ost_idx>] is used, the index must
6848 # be in the OST indices list.
6849 local start_ost_idx=${ost_indices##*,}
6850 file=$DIR/$tdir/$tfile-3
6851 cmd="$LFS setstripe -o $ost_indices -i $start_ost_idx $file"
6853 eval $cmd || error "$cmd failed"
6854 check_stripe_count $file $OSTCOUNT
6855 check_obdidx $file $ost_indices
6856 check_start_ost_idx $file $start_ost_idx
6858 file=$DIR/$tdir/$tfile-4
6859 cmd="$LFS setstripe"
6860 cmd+=" -o $(exclude_items_from_list $ost_indices $start_ost_idx)"
6861 cmd+=" -i $start_ost_idx $file"
6863 eval $cmd && error "index $start_ost_idx should be in $ost_indices"
6865 # 5. Specifying OST indices for directory should succeed.
6866 local dir=$DIR/$tdir/$tdir
6867 mkdir $dir || error "mkdir $dir failed"
6868 cmd="$LFS setstripe -o $ost_indices $dir"
6869 if [ "$MDS1_VERSION" -gt $(version_code 2.11.53) ] &&
6870 [ "$CLIENT_VERSION" -gt $(version_code 2.11.53) ]; then
6872 eval $cmd || error "unable to specify OST indices on directory"
6874 echo "need MDS+client version at least 2.11.53"
6877 run_test 82a "specify OSTs for file (succeed) or directory (succeed)"
6879 # Test 82b is run to ensure that if the user supplies a pool with a specific
6880 # stripe layout that it behaves proprerly. It should fail in the case that
6881 # the supplied OST index list points to OSTs not contained in the user
6883 test_82b() { # LU-4665
6884 [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
6885 skip "Need MDS version at least 2.6.54"
6886 [[ $OSTCOUNT -ge 4 ]] || skip_env "needs >= 4 OSTs"
6892 # Format OSTs with random sparse indices.
6896 local LOV_V1_INSANE_STRIPE_COUNT=65532
6897 for i in $(seq $OSTCOUNT); do
6898 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
6899 ost_indices+=" $index"
6901 ost_indices=$(comma_list $ost_indices)
6903 stack_trap "restore_ostindex" EXIT
6904 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
6905 OST_INDEX_LIST=[$ost_indices] formatall
6907 # Setup Lustre filesystem.
6908 start_mgsmds || error "start_mgsmds failed"
6909 for i in $(seq $OSTCOUNT); do
6910 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6911 error "start ost$i failed"
6914 mount_client $MOUNT || error "mount client $MOUNT failed"
6918 check_lfs_df_ret_val $? || error "$LFS df $MOUNT failed"
6919 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6921 # Create a new pool and add OSTs into it.
6922 local ost_pool=$FSNAME.$TESTNAME
6923 create_pool $ost_pool || error "create OST pool $ost_pool failed"
6925 local ost_idx_in_list=${ost_indices##*,}
6926 local ost_idx_in_pool=$(exclude_items_from_list $ost_indices \
6929 local ost_targets="$FSNAME-OST["
6930 for i in ${ost_idx_in_pool//,/ }; do
6931 ost_targets=$ost_targets$(printf "%04x," $i)
6933 ost_targets="${ost_targets%,}]"
6935 local ost_targets_uuid=$(for i in ${ost_idx_in_pool//,/ }; \
6936 do printf "$FSNAME-OST%04x_UUID\n" $i; done |
6937 sort -u | tr '\n' ' ')
6939 local cmd="$LCTL pool_add $ost_pool $ost_targets"
6940 do_facet mgs $cmd || error "$cmd failed"
6941 wait_update $HOSTNAME "$LCTL get_param -n lov.$FSNAME-*.pools.$TESTNAME|
6942 sort -u | tr '\n' ' ' " "$ost_targets_uuid" ||
6943 error "wait_update $ost_pool failed"
6944 wait_update_facet $SINGLEMDS "$LCTL pool_list $ost_pool | wc -l" 4 ||
6945 error "wait_update pool_list $ost_pool failed"
6947 # If [--pool|-p <pool_name>] is set with [--ost-list|-o <ost_indices>],
6948 # then the OSTs must be the members of the pool.
6949 local file=$DIR/$tdir/$tfile
6950 cmd="$LFS setstripe -p $ost_pool -o $ost_idx_in_list $file"
6952 eval $cmd && error "OST with index $ost_idx_in_list should be" \
6953 "in OST pool $ost_pool"
6955 # Only select OST $ost_idx_in_list from $ost_pool for file.
6956 ost_idx_in_list=${ost_idx_in_pool#*,}
6957 cmd="$LFS setstripe -p $ost_pool -o $ost_idx_in_list $file"
6959 eval $cmd || error "$cmd failed"
6960 cmd="$LFS getstripe $file"
6962 eval $cmd || error "$cmd failed"
6963 check_stripe_count $file 2
6964 check_obdidx $file $ost_idx_in_list
6965 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6966 error "write $file failed"
6968 run_test 82b "specify OSTs for file with --pool and --ost-list options"
6971 [[ "$OST1_VERSION" -ge $(version_code 2.6.91) ]] ||
6972 skip "Need OST version at least 2.6.91"
6973 if [ "$ost1_FSTYPE" != ldiskfs ]; then
6974 skip "ldiskfs only test"
6983 ostmnt=$(facet_mntpt ost1)
6985 # Mount the OST as an ldiskfs filesystem.
6986 log "mount the OST $dev as a $ost1_FSTYPE filesystem"
6987 add ost1 $(mkfs_opts ost1 $dev) $FSTYPE_OPT \
6988 --reformat $dev > /dev/null ||
6989 error "format ost1 error"
6991 if ! test -b $dev; then
6992 mnt_opts=$(csa_add "$OST_MOUNT_FS_OPTS" -o loop)
6994 echo "mnt_opts $mnt_opts"
6995 do_facet ost1 mount -t "$ost1_FSTYPE" $dev \
6997 # Run llverfs on the mounted ldiskfs filesystem.
6998 # It is needed to get ENOSPACE.
6999 log "run llverfs in partial mode on the OST $ost1_FSTYPE $ostmnt"
7000 do_rpc_nodes $(facet_host ost1) run_llverfs $ostmnt -vpl \
7001 "no" || error "run_llverfs error on $ost1_FSTYPE"
7004 log "unmount the OST $dev"
7007 # Delete file IO_scrub. Later osd_scrub_setup will try to
7008 # create "IO_scrub" but will get ENOSPACE.
7010 echo "start ost1 service on `facet_active_host ost1`"
7011 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
7014 err=$(do_facet ost1 dmesg | grep "VFS: Busy inodes after unmount of")
7015 echo "string err $err"
7016 [ -z "$err" ] || error $err
7019 run_test 83 "ENOSPACE on OST doesn't cause message VFS: \
7020 Busy inodes after unmount ..."
7023 local facet=$SINGLEMDS
7024 local num=$(echo $facet | tr -d "mds")
7025 local dev=$(mdsdevname $num)
7026 local time_min=$(recovery_time_min)
7027 local recovery_duration
7028 local completed_clients
7029 local correct_clients
7032 echo "start mds service on $(facet_active_host $facet)"
7034 "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
7035 error "start MDS failed"
7037 start_ost || error "start OST0000 failed"
7038 wait_osc_import_state mds ost1 FULL
7039 start_ost2 || error "start OST0001 failed"
7040 wait_osc_import_state mds ost2 FULL
7042 echo "recovery_time=$time_min, timeout=$TIMEOUT, wrap_up=$wrap_up"
7044 mount_client $MOUNT1 || error "mount $MOUNT1 failed"
7045 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
7046 # make sure new superblock labels are sync'd before disabling writes
7050 replay_barrier $SINGLEMDS
7051 createmany -o $DIR1/$tfile-%d 1000
7053 # We need to catch the end of recovery window to extend it.
7054 # Skip 5 requests and add delay to request handling.
7055 #define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP
7056 do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5"
7058 facet_failover --fsck $SINGLEMDS || error "failover: $?"
7061 echo "recovery status"
7062 do_facet $SINGLEMDS \
7063 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status"
7065 recovery_duration=$(do_facet $SINGLEMDS \
7066 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
7067 awk '/recovery_duration/ { print $2 }')
7068 (( $recovery_duration > $time_min + $wrap_up )) &&
7069 error "recovery_duration > recovery_time_hard + wrap up"
7070 completed_clients=$(do_facet $SINGLEMDS \
7071 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
7072 awk '/completed_clients/ { print $2 }')
7074 correct_clients="$MDSCOUNT/$((MDSCOUNT+1))"
7075 [ "$completed_clients" = "${correct_clients}" ] ||
7076 error "$completed_clients != $correct_clients"
7078 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
7079 umount_client $MOUNT1
7080 umount_client $MOUNT2
7086 run_test 84 "check recovery_hard_time"
7089 [[ "$OST1_VERSION" -ge $(version_code 2.7.55) ]] ||
7090 skip "Need OST version at least 2.7.55"
7091 ##define OBD_FAIL_OSD_OST_EA_FID_SET 0x197
7092 do_facet ost1 "lctl set_param fail_loc=0x197"
7096 run_test 85 "osd_ost init: fail ea_fid_set"
7101 # ost1 has already registered to the MGS before the reformat.
7102 # So after reformatting it with option "-G", it could not be
7103 # mounted to the MGS. Cleanup the system for subsequent tests.
7108 [ "$ost1_FSTYPE" = zfs ] &&
7109 skip "LU-6442: no such mkfs params for ZFS OSTs"
7110 [[ "$MDS1_VERSION" -ge $(version_code 2.7.56) ]] ||
7111 skip "Need server version newer than 2.7.55"
7114 local OLDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
7115 awk '/Flex block group size: / { print $NF; exit; }')
7117 [ "$OLDSIZE" == "$NEWSIZE" ] && skip "$NEWSIZE groups already"
7119 local opts=" -O flex_bg -G $NEWSIZE"
7120 opts=$(OST_FS_MKFS_OPTS+="$opts" mkfs_opts ost1 $(ostdevname 1))
7121 opts+=" --reformat $(ostdevname 1) $(ostvdevname 1)"
7122 echo "params: $opts"
7124 trap cleanup_86 EXIT ERR
7127 add ost1 $opts || error "add ost1 failed with new params"
7129 local FOUNDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
7130 awk '/Flex block group size: / { print $NF; exit; }')
7132 [[ $FOUNDSIZE == $NEWSIZE ]] ||
7133 error "Flex block group size: $FOUNDSIZE, expected: $NEWSIZE"
7137 run_test 86 "Replacing mkfs.lustre -G option"
7139 test_87() { #LU-6544
7140 [[ "$MDS1_VERSION" -ge $(version_code 2.9.51) ]] ||
7141 skip "Need MDS version at least 2.9.51"
7142 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
7143 skip "ldiskfs only test"
7144 [[ $OSTCOUNT -gt 59 ]] &&
7145 skip "Ignore wide striping situation"
7146 [ -n "$FILESET" ] && skip "Not functional for FILESET set"
7148 local mdsdev=$(mdsdevname 1)
7149 local mdsvdev=$(mdsvdevname 1)
7150 local file=$DIR/$tfile
7151 local mntpt=$(facet_mntpt $SINGLEMDS)
7152 local used_xattr_blk=0
7153 local inode_size=${1:-1024}
7155 local xtest="trusted.test"
7159 local stripe_cnt=$(($OSTCOUNT + 2))
7161 #Please see ldiskfs_make_lustre() for MDT inode size calculation
7162 if [ $stripe_cnt -gt 16 ]; then
7165 left_size=$(expr $inode_size - \
7168 32 - 40 \* 3 - 32 \* 3 - $stripe_cnt \* 24 - 16 - 3 - \
7170 24 - 18 - $(expr length $tfile) - 16 - 4)
7171 if [ $left_size -le 0 ]; then
7172 echo "No space($left_size) is expected in inode."
7173 echo "Try 1-byte xattr instead to verify this."
7176 echo "Estimate: at most $left_size-byte space left in inode."
7182 if ! combined_mgs_mds ; then
7186 add mds1 $(mkfs_opts mds1 ${mdsdev}) --stripe-count-hint=$stripe_cnt \
7187 --reformat $mdsdev $mdsvdev || error "add mds1 failed"
7188 start_mdt 1 > /dev/null || error "start mdt1 failed"
7189 for i in $(seq $OSTCOUNT); do
7190 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS > /dev/null ||
7191 error "start ost$i failed"
7193 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
7194 check_mount || error "check client $MOUNT failed"
7197 $LFS setstripe -E 1M -S 1M -c 1 -E 64M -c 1 -E -1 -c -1 $file ||
7198 error "Create file with 3 components failed"
7199 $TRUNCATE $file $((1024*1024*64+1)) || error "truncate file failed"
7200 i=$($LFS getstripe -I3 -c $file) || error "get 3rd stripe count failed"
7201 if [ $i -ne $OSTCOUNT ]; then
7202 left_size=$(expr $left_size + $(expr $OSTCOUNT - $i) \* 24)
7203 echo -n "Since only $i out $OSTCOUNT OSTs are used, "
7204 echo -n "the expected left space is changed to "
7205 echo "$left_size bytes at most."
7207 value=$(generate_string $left_size)
7208 setfattr -n $xtest -v $value $file
7209 orig=$(get_xattr_value $xtest $file)
7210 [[ "$orig" != "$value" ]] && error "$xtest changed"
7212 #Verify if inode has some expected space left
7213 umount $MOUNT > /dev/null || error "umount $MOUNT failed"
7214 stop_mdt 1 > /dev/null || error "stop mdt1 failed"
7215 mount_ldiskfs $SINGLEMDS || error "mount -t ldiskfs $SINGLEMDS failed"
7217 do_facet $SINGLEMDS ls -sal $mntpt/ROOT/$tfile
7218 used_xattr_blk=$(do_facet $SINGLEMDS ls -s $mntpt/ROOT/$tfile |
7220 [[ $used_xattr_blk -eq 0 ]] &&
7221 error "Please check MDS inode size calculation: \
7222 more than $left_size-byte space left in inode."
7223 echo "Verified: at most $left_size-byte space left in inode."
7225 unmount_ldiskfs $SINGLEMDS
7227 for i in $(seq $OSTCOUNT); do
7228 stop ost$i -f || error "stop ost$i failed"
7231 run_test 87 "check if MDT inode can hold EAs with N stripes properly"
7234 [ "$mds1_FSTYPE" == zfs ] &&
7235 skip "LU-6662: no implementation for ZFS"
7239 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
7240 --reformat $(mdsdevname 1) || error "add mds1 failed"
7242 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
7243 grep -e \".*opts:.*errors=remount-ro.*\"" ||
7244 error "default mount options is missing"
7246 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
7247 --mountfsoptions="user_xattr,errors=panic" \
7248 --reformat $(mdsdevname 1) || error "add mds1 failed"
7250 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
7251 grep -e \".*opts:.*errors=panic.*\"" ||
7252 error "user can't override default mount options"
7254 run_test 88 "check the default mount options can be overridden"
7256 test_89() { # LU-7131
7257 [[ "$MDS1_VERSION" -ge $(version_code 2.9.54) ]] ||
7258 skip "Need MDT version at least 2.9.54"
7260 local key=failover.node
7261 local val1=192.0.2.254@tcp0 # Reserved IPs, see RFC 5735
7262 local val2=192.0.2.255@tcp0
7263 local mdsdev=$(mdsdevname 1)
7268 if [[ "$mds1_FSTYPE" == zfs ]]; then
7269 import_zpool mds1 || return ${PIPESTATUS[0]}
7272 # Check that parameters are added correctly
7273 echo "tunefs --param $key=$val1"
7274 do_facet mds1 "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
7275 error "tunefs --param $key=$val1 failed"
7276 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7277 error "tunefs --dryrun failed"
7278 params=${params##*Parameters:}
7279 params=${params%%exiting*}
7280 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val1) = "1" ] ||
7281 error "on-disk parameter not added correctly via tunefs"
7283 # Check that parameters replace existing instances when added
7284 echo "tunefs --param $key=$val2"
7285 do_facet mds1 "$TUNEFS --param $key=$val2 $mdsdev >/dev/null" ||
7286 error "tunefs --param $key=$val2 failed"
7287 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7288 error "tunefs --dryrun failed"
7289 params=${params##*Parameters:}
7290 params=${params%%exiting*}
7291 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "1" ] ||
7292 error "on-disk parameter not replaced via tunefs"
7293 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val2) = "1" ] ||
7294 error "on-disk parameter not replaced correctly via tunefs"
7296 # Check that a parameter is erased properly
7297 echo "tunefs --erase-param $key"
7298 do_facet mds1 "$TUNEFS --erase-param $key $mdsdev >/dev/null" ||
7299 error "tunefs --erase-param $key failed"
7300 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7301 error "tunefs --dryrun failed"
7302 params=${params##*Parameters:}
7303 params=${params%%exiting*}
7304 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "0" ] ||
7305 error "on-disk parameter not erased correctly via tunefs"
7307 # Check that all the parameters are erased
7308 do_facet mds1 "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
7309 error "tunefs --param $key=$val1 failed"
7310 echo "tunefs --erase-params"
7311 do_facet mds1 "$TUNEFS --erase-params $mdsdev >/dev/null" ||
7312 error "tunefs --erase-params failed"
7313 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7314 error "tunefs --dryrun failed"
7315 params=${params##*Parameters:}
7316 params=${params%%exiting*}
7317 params=$(echo $params | tr ' ' '\n')
7319 error "all on-disk parameters not erased correctly via tunefs $params"
7321 # Check the order of options --erase-params and --param
7322 echo "tunefs --param $key=$val1 --erase-params"
7324 "$TUNEFS --param $key=$val1 --erase-params $mdsdev >/dev/null"||
7325 error "tunefs --param $key=$val1 --erase-params failed"
7326 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7327 error "tunefs --dryrun failed"
7328 params=${params##*Parameters:}
7329 params=${params%%exiting*}
7330 [ $(echo $params | tr ' ' '\n') == "$key=$val1" ] ||
7331 error "on-disk param not added correctly with --erase-params"
7335 run_test 89 "check tunefs --param and --erase-param{s} options"
7338 # $2 (optional) value of max_mod_rpcs_in_flight to set
7339 check_max_mod_rpcs_in_flight() {
7347 idx=$(printf "%04x" $($LFS getdirstripe -i $dir))
7348 facet="mds$((0x$idx + 1))"
7350 if [ -z "$mmr" ]; then
7351 # get value of max_mod_rcps_in_flight
7352 mmr=$($LCTL get_param -n \
7353 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) ||
7354 error "Unable to get max_mod_rpcs_in_flight"
7355 echo "max_mod_rcps_in_flight is $mmr"
7357 # set value of max_mod_rpcs_in_flight
7359 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight=$mmr ||
7360 error "Unable to set max_mod_rpcs_in_flight to $mmr"
7361 echo "max_mod_rpcs_in_flight set to $mmr"
7364 # create mmr+1 files
7365 echo "creating $((mmr + 1)) files ..."
7367 for i in $(seq $((mmr + 1))); do
7373 # consumes mmr-1 modify RPC slots
7374 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7375 # drop requests on MDT so that RPC slots are consumed
7376 # during all the request resend interval
7377 do_facet $facet "$LCTL set_param fail_loc=0x159"
7378 echo "launch $((mmr - 1)) chmod in parallel ..."
7379 for i in $(seq $((mmr - 1))); do
7380 chmod 0600 $dir/file-$i &
7384 # send one additional modify RPC
7385 do_facet $facet "$LCTL set_param fail_loc=0"
7386 echo "launch 1 additional chmod in parallel ..."
7387 chmod 0600 $dir/file-$mmr &
7390 # check this additional modify RPC get a modify RPC slot
7391 # and succeed its operation
7392 checkstat -vp 0600 $dir/file-$mmr ||
7393 error "Unable to send $mmr modify RPCs in parallel"
7398 # consumes mmr modify RPC slots
7399 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7400 # drop requests on MDT so that RPC slots are consumed
7401 # during all the request resend interval
7402 do_facet $facet "$LCTL set_param fail_loc=0x159"
7403 echo "launch $mmr chmod in parallel ..."
7404 for i in $(seq $mmr); do
7405 chmod 0666 $dir/file-$i &
7409 # send one additional modify RPC
7410 do_facet $facet "$LCTL set_param fail_loc=0"
7411 echo "launch 1 additional chmod in parallel ..."
7412 chmod 0666 $dir/file-$((mmr + 1)) &
7415 # check this additional modify RPC blocked getting a modify RPC slot
7416 checkstat -vp 0644 $dir/file-$((mmr + 1)) ||
7417 error "Unexpectedly send $(($mmr + 1)) modify RPCs in parallel"
7421 get_mdt_max_mod_rpcs_in_flight_val() {
7425 # It will be enough to get one value from one MDT
7426 max_mrif=$(do_facet $facet \
7427 "$LCTL get_param -n \
7428 mdt.$FSNAME-MDT*.max_mod_rpcs_in_flight | head -n 1")
7429 if [[ $max_mrif ]]; then
7432 max_mrif=$(do_facet $facet \
7433 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
7438 set_mdt_max_mod_rpcs_in_flight() {
7443 lctl_op=$($LCTL get_param \
7444 mdt.*.max_mod_rpcs_in_flight)
7445 if [[ $lctl_op ]]; then
7448 mdt.$FSNAME-MDT*.max_mod_rpcs_in_flight=$max_mrif"
7452 /sys/module/mdt/parameters/max_mod_rpcs_per_client"
7453 echo "the deprecated max_mod_rpcs_per_client \
7454 parameter was involved"
7461 [[ $($LCTL get_param mdc.*.import |
7462 grep "connect_flags:.*multi_mod_rpc") ]] ||
7463 skip "Need MDC with 'multi_mod_rpcs' feature"
7465 # check default value
7466 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
7467 check_max_mod_rpcs_in_flight $DIR/$tdir
7471 run_test 90a "check max_mod_rpcs_in_flight is enforced"
7481 [[ $($LCTL get_param mdc.*.import |
7482 grep "connect_flags:.*multi_mod_rpc") ]] ||
7483 skip "Need MDC with 'multi_mod_rpcs' feature"
7486 # update max_mod_rpcs_in_flight
7487 $LFS mkdir -c1 $DIR/${tdir}1 || error "mkdir $DIR/${tdir}1 failed"
7488 check_max_mod_rpcs_in_flight $DIR/${tdir}1 1
7491 # check client is able to send multiple modify RPCs in paralell
7492 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
7493 grep -c "multi_mod_rpcs")
7494 if [ "$tmp" -ne $MDSCOUNT ]; then
7496 skip "Client not able to send multiple modify RPCs in parallel"
7499 # update max_mod_rpcs_in_flight
7500 $LFS mkdir -c1 $DIR/${tdir}2 || error "mkdir $DIR/${tdir}2 failed"
7501 check_max_mod_rpcs_in_flight $DIR/${tdir}2 5
7504 $LFS mkdir -c1 $DIR/${tdir}3 || error "mkdir $DIR/${tdir}3 failed"
7505 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/${tdir}3))
7506 facet="mds$((0x$idx + 1))"
7508 mmrpc=$(get_mdt_max_mod_rpcs_in_flight_val $facet)
7509 echo "mdt_max_mod_rpcs_in_flight is $mmrpc"
7510 # update max_mod_rpcs_in_flight
7511 umount_client $MOUNT
7512 set_mdt_max_mod_rpcs_in_flight 16 $facet
7514 $LCTL set_param mdc.$FSNAME-MDT$idx-mdc-*.max_rpcs_in_flight=17
7515 check_max_mod_rpcs_in_flight $DIR/${tdir}3 16
7517 # restore MDT max_mod_rpcs_in_flight initial value
7518 set_mdt_max_mod_rpcs_in_flight $mmrpc $facet
7520 rm -rf $DIR/${tdir}?
7523 run_test 90b "check max_mod_rpcs_in_flight is enforced after update"
7526 # get max_rpcs_in_flight value
7527 mrif_90c=$($LCTL get_param -n \
7528 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
7529 echo "max_rpcs_in_flight is $mrif_90c"
7531 # get MDC max_mod_rpcs_in_flight value
7532 mmrif_90c=$($LCTL get_param -n \
7533 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight)
7534 echo "MDC max_mod_rpcs_in_flight is $mmrif_90c"
7536 # get MDT max_mod_rpcs_in_flight value
7537 mmrpc_90c=$(get_mdt_max_mod_rpcs_in_flight_val "mds1")
7538 echo "mdt_max_mod_rpcs_in_flight is $mmrpc_90c"
7541 restore_params_90c() {
7544 # restore max_rpcs_in_flight value
7545 do_facet mgs $LCTL set_param -P \
7546 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$mrif_90c
7548 # restore max_mod_rpcs_in_flight value
7549 do_facet mgs $LCTL set_param -P \
7550 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mmrif_90c
7552 # restore MDT max_mod_rpcs_in_flight value
7553 set_mdt_max_mod_rpcs_in_flight $mmrpc_90c "mds1"
7561 [[ $($LCTL get_param mdc.*.import |
7562 grep "connect_flags:.*multi_mod_rpc") ]] ||
7563 skip "Need MDC with 'multi_mod_rpcs' feature"
7565 # check client is able to send multiple modify RPCs in paralell
7566 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
7567 grep -c "multi_mod_rpcs")
7568 if [ "$tmp" -ne $MDSCOUNT ]; then
7570 skip "Client not able to send multiple modify RPCs in parallel"
7574 stack_trap restore_params_90c
7577 # attempt to set max_mod_rpcs_in_flight to max_rpcs_in_flight value
7578 # prerequisite: set MDT max_mod_rpcs_in_flight to
7579 # max_rpcs_in_flight value
7580 set_mdt_max_mod_rpcs_in_flight $mrif_90c "mds1"
7582 # if max_mod_rpcs_in_flight is set to be equal to or larger than
7583 # max_rpcs_in_flight, then max_rpcs_in_flight will be increased
7584 if [[ "$CLIENT_VERSION" -ge $(version_code 2.13.53) ]]; then
7586 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif_90c ||
7587 error "set max_mod_rpcs_in_flight to $mrif_90c failed"
7589 local new_mrif=$($LCTL get_param -n \
7590 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
7591 ((new_mrif == mrif_90c + 1)) ||
7592 error "max_rpcs_in_flight was not increased"
7595 umount_client $MOUNT
7596 set_mdt_max_mod_rpcs_in_flight $mmrpc_90c "mds1"
7600 # attempt to set max_mod_rpcs_in_flight to MDT max_mod_rpcs_in_flight+1
7601 # prerequisite: set max_rpcs_in_flight to MDT max_mod_rpcs_in_flight+2
7603 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$((mmrpc_90c + 2))
7606 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$((mmrpc_90c + 1)) &&
7607 error "set max_mod_rpcs_in_flight to $((mmrpc_90c + 1)) should fail"
7610 # attempt to set max_mod_rpcs_in_flight permanently
7611 do_facet mgs $LCTL set_param -P \
7612 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$mrif_90c
7614 do_facet mgs $LCTL set_param -P \
7615 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif_90c
7617 remount_client $MOUNT
7619 wait_update_facet --verbose client "$LCTL get_param -n \
7620 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight" \
7621 "$((mrif_90c + 1))" ||
7622 error "expected '$((mrif_90c + 1))' for max_rpcs_in_flight"
7624 wait_update_facet --verbose client "$LCTL get_param -n \
7625 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight" \
7627 error "expected '$mrif_90c' for max_mod_rpcs_in_flight"
7632 run_test 90c "check max_mod_rpcs_in_flight update limits"
7643 [[ $($LCTL get_param mdc.*.import |
7644 grep "connect_flags:.*multi_mod_rpc") ]] ||
7645 skip "Need MDC with 'multi_mod_rpcs' feature"
7647 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
7648 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
7649 facet="mds$((0x$idx + 1))"
7651 # check client version supports multislots
7652 tmp=$($LCTL get_param -N \
7653 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
7654 if [ -z "$tmp" ]; then
7656 skip "Client does not support multiple modify RPCs in flight"
7659 # get current value of max_mod_rcps_in_flight
7660 mmr=$($LCTL get_param -n \
7661 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
7662 echo "max_mod_rcps_in_flight is $mmr"
7665 echo "creating $mmr files ..."
7667 for i in $(seq $mmr); do
7668 touch $DIR/$tdir/file-$i
7671 # prepare for close RPC
7672 multiop_bg_pause $DIR/$tdir/file-close O_c
7675 # consumes mmr modify RPC slots
7676 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7677 # drop requests on MDT so that RPC slots are consumed
7678 # during all the request resend interval
7679 do_facet $facet "$LCTL set_param fail_loc=0x159"
7680 echo "launch $mmr chmod in parallel ..."
7681 for i in $(seq $mmr); do
7682 chmod 0600 $DIR/$tdir/file-$i &
7685 # send one additional close RPC
7686 do_facet $facet "$LCTL set_param fail_loc=0"
7687 echo "launch 1 additional close in parallel ..."
7689 cancel_lru_locks mdc
7692 # check this additional close RPC get a modify RPC slot
7693 # and multiop process completed
7694 [ -d /proc/$pid ] &&
7695 error "Unable to send the additional close RPC in parallel"
7700 run_test 90d "check one close RPC is allowed above max_mod_rpcs_in_flight"
7702 check_uuid_on_ost() {
7704 do_facet ost1 "$LCTL get_param obdfilter.${FSNAME}*.exports.'$nid'.uuid"
7707 check_uuid_on_mdt() {
7709 do_facet $SINGLEMDS "$LCTL get_param mdt.${mds1_svc}*.exports.'$nid'.uuid"
7717 [[ "$OST1_VERSION" -ge $(version_code 2.7.63) ]] ||
7718 skip "Need OST version at least 2.7.63"
7719 [[ "$MDS1_VERSION" -ge $(version_code 2.7.63) ]] ||
7720 skip "Need MDT version at least 2.7.63"
7722 start_mds || error "MDS start failed"
7723 start_ost || error "unable to start OST"
7724 mount_client $MOUNT || error "client start failed"
7725 check_mount || error "check_mount failed"
7728 nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g")
7732 uuid=$(get_client_uuid $MOUNT)
7734 echo "list nids on mdt:"
7735 do_facet $SINGLEMDS "$LCTL list_param mdt.${FSNAME}*.exports.*"
7736 echo "uuid from $nid:"
7737 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
7739 found=$(check_uuid_on_mdt $nid | grep $uuid)
7740 [ -z "$found" ] && error "can't find $uuid $nid on MDT"
7741 found=$(check_uuid_on_ost $nid | grep $uuid)
7742 [ -z "$found" ] && error "can't find $uuid $nid on OST"
7744 # umount the client so it won't reconnect
7745 manual_umount_client --force || error "failed to umount $?"
7746 # shouldn't disappear on MDS after forced umount
7747 found=$(check_uuid_on_mdt $nid | grep $uuid)
7748 [ -z "$found" ] && error "can't find $uuid $nid"
7751 do_facet $SINGLEMDS \
7752 "$LCTL set_param -n mdt.${mds1_svc}.evict_client nid:$nid"
7753 sleep 1 # eviction above is async, give it some time to proceed
7755 found=$(check_uuid_on_mdt $nid | grep $uuid)
7756 [ -n "$found" ] && error "found $uuid $nid on MDT"
7757 found=$(check_uuid_on_ost $nid | grep $uuid)
7758 [ -n "$found" ] && error "found $uuid $nid on OST"
7760 # check it didn't reconnect (being umounted)
7761 sleep $((TIMEOUT+1))
7762 found=$(check_uuid_on_mdt $nid | grep $uuid)
7763 [ -n "$found" ] && error "found $uuid $nid on MDT"
7764 found=$(check_uuid_on_ost $nid | grep $uuid)
7765 [ -n "$found" ] && error "found $uuid $nid on OST"
7769 run_test 91 "evict-by-nid support"
7771 generate_ldev_conf() {
7772 # generate an ldev.conf file
7773 local ldevconfpath=$1
7775 local fsldevformat=""
7778 local facets="mgs,$(get_facets OST),$(get_facets MDS)"
7779 for facet in ${facets//,/ }; do
7781 fstype=$(facet_fstype $facet)
7782 [ "$fstype" = zfs ] && fsldevformat="$fstype:"
7783 local host=$(facet_host $facet)
7785 local varfo=${facet}failover_HOST
7786 if [ $facet == mgs ] && combined_mgs_mds; then
7787 varfo=mds1failover_HOST
7789 [ -n "${!varfo}" ] && fo=${!varfo}
7790 local type=$(echo $facet | tr -d "[:digit:]" | \
7791 tr "[:lower:]" "[:upper:]" | sed s/MDS/MDT/ )
7793 [ ${facet} == mgs ] ||
7794 num=$(facet_number $facet)
7795 printf "%s\t%s\t%s-%s%04x\t%s%s\n" \
7802 $(facet_device $facet) >> $ldevconfpath
7805 echo "----- $ldevconfpath -----"
7807 echo "--- END $ldevconfpath ---"
7812 # generate a nids file (mapping between hostname to nid)
7813 # looks like we only have the MGS nid available to us
7814 # so just echo that to a file
7816 echo -e "${mgs_HOST}\t${MGSNID}" > $nidspath
7818 echo "----- $nidspath -----"
7820 echo "--- END $nidspath ---"
7823 compare_ldev_output() {
7824 local ldev_output=$1
7825 local expected_output=$2
7827 sort $expected_output -o $expected_output
7828 sort $ldev_output -o $ldev_output
7830 echo "-- START OF LDEV OUTPUT --"
7832 echo "--- END OF LDEV OUTPUT ---"
7834 echo "-- START OF EXPECTED OUTPUT --"
7835 cat $expected_output
7836 echo "--- END OF EXPECTED OUTPUT ---"
7838 diff $expected_output $ldev_output
7843 if [ -z "$LDEV" ]; then
7844 error "ldev is missing!"
7847 local LDEVCONFPATH=$TMP/ldev.conf
7848 local NIDSPATH=$TMP/nids
7850 echo "Host is $(hostname)"
7852 generate_ldev_conf $LDEVCONFPATH
7853 generate_nids $NIDSPATH
7855 # echo the mgs nid and compare it to environment variable MGSNID
7856 local facets="$(get_facets OST),$(get_facets MDS),mgs"
7857 for facet in ${facets//,/ }; do
7858 local host=$(facet_host $facet)
7859 local output=$($LDEV -c $LDEVCONFPATH -H $host -n $NIDSPATH echo %m)
7861 echo "-- START OF LDEV OUTPUT --"
7863 echo "--- END OF LDEV OUTPUT ---"
7866 error "ldev failed to execute!"
7868 # need to process multiple lines because of
7869 # several targets on host
7870 echo -e $output | awk '{ print $2 }' | while read -r line ; do
7871 [ "$line" = "$MGSNID" ] ||
7872 error "ldev failed mgs nid '$line', \
7876 rm -f $LDEVCONFPATH $NIDSPATH
7878 run_test 92 "ldev returns MGS NID correctly in command substitution"
7881 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs"
7884 #start mgs or mgs/mdt0
7885 if ! combined_mgs_mds ; then
7892 start_ost || error "OST0 start fail"
7894 #define OBD_FAIL_MGS_WRITE_TARGET_DELAY 0x90e
7895 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x8000090e"
7896 for num in $(seq 2 $MDSCOUNT); do
7900 mount_client $MOUNT || error "mount client fails"
7901 wait_osc_import_state mds ost FULL
7902 wait_osc_import_ready client ost
7903 check_mount || error "check_mount failed"
7905 cleanup || error "cleanup failed with $?"
7907 run_test 93 "register mulitple MDT at the same time"
7910 if [ -z "$LDEV" ]; then
7911 error "ldev is missing!"
7914 local LDEVCONFPATH=$TMP/ldev.conf
7915 local NIDSPATH=$TMP/nids
7917 generate_ldev_conf $LDEVCONFPATH
7918 generate_nids $NIDSPATH
7920 local LDEV_OUTPUT=$TMP/ldev-output.txt
7921 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME > $LDEV_OUTPUT
7923 # ldev failed, error
7924 if [ $? -ne 0 ]; then
7925 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7926 error "ldev failed to execute!"
7930 local EXPECTED_OUTPUT=$TMP/ldev-expected.txt
7932 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
7934 for num in $(seq $MDSCOUNT); do
7935 printf "%s-MDT%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7938 for num in $(seq $OSTCOUNT); do
7939 printf "%s-OST%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7942 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7944 if [ $? -ne 0 ]; then
7945 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7946 error "ldev failed to produce the correct hostlist!"
7949 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7951 run_test 94 "ldev outputs correct labels for file system name query"
7954 if [ -z "$LDEV" ]; then
7955 error "ldev is missing!"
7958 local LDEVCONFPATH=$TMP/ldev.conf
7959 local NIDSPATH=$TMP/nids
7961 generate_ldev_conf $LDEVCONFPATH
7962 generate_nids $NIDSPATH
7966 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME &>/dev/null
7967 if [ $? -ne 0 ]; then
7968 rm $LDEVCONFPATH $NIDSPATH
7969 error "ldev label filtering w/ -F failed!"
7973 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -l &>/dev/null
7974 if [ $? -ne 0 ]; then
7975 rm $LDEVCONFPATH $NIDSPATH
7976 error "ldev label filtering w/ -l failed!"
7980 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f &>/dev/null
7981 if [ $? -ne 0 ]; then
7982 rm $LDEVCONFPATH $NIDSPATH
7983 error "ldev label filtering w/ -f failed!"
7987 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a &>/dev/null
7988 if [ $? -ne 0 ]; then
7989 rm $LDEVCONFPATH $NIDSPATH
7990 error "ldev label filtering w/ -a failed!"
7995 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -F $FSNAME &>/dev/null
7996 if [ $? -eq 0 ]; then
7997 rm $LDEVCONFPATH $NIDSPATH
7998 error "ldev label filtering w/ -a and -F incorrectly succeeded"
8002 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -f &>/dev/null
8003 if [ $? -eq 0 ]; then
8004 rm $LDEVCONFPATH $NIDSPATH
8005 error "ldev label filtering w/ -a and -f incorrectly succeeded"
8009 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -l &>/dev/null
8010 if [ $? -eq 0 ]; then
8011 rm $LDEVCONFPATH $NIDSPATH
8012 error "ldev label filtering w/ -a and -l incorrectly succeeded"
8016 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f -l &>/dev/null
8017 if [ $? -eq 0 ]; then
8018 rm $LDEVCONFPATH $NIDSPATH
8019 error "ldev label filtering w/ -f and -l incorrectly succeeded"
8023 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -l &>/dev/null
8024 if [ $? -eq 0 ]; then
8025 rm $LDEVCONFPATH $NIDSPATH
8026 error "ldev label filtering w/ -F and -l incorrectly succeeded"
8029 # file sys & foreign
8030 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -f &>/dev/null
8031 if [ $? -eq 0 ]; then
8032 rm $LDEVCONFPATH $NIDSPATH
8033 error "ldev label filtering w/ -F and -f incorrectly succeeded"
8036 rm $LDEVCONFPATH $NIDSPATH
8038 run_test 95 "ldev should only allow one label filter"
8041 if [ -z "$LDEV" ]; then
8042 error "ldev is missing!"
8045 local LDEVCONFPATH=$TMP/ldev.conf
8046 local NIDSPATH=$TMP/nids
8048 generate_ldev_conf $LDEVCONFPATH
8049 generate_nids $NIDSPATH
8051 local LDEV_OUTPUT=$TMP/ldev-output.txt
8052 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -H $mgs_HOST \
8054 awk '{print $2}' > $LDEV_OUTPUT
8056 # ldev failed, error
8057 if [ $? -ne 0 ]; then
8058 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
8059 error "ldev failed to execute!"
8063 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
8065 echo "$mgs_HOST-$(facet_fstype mgs)" > $EXPECTED_OUTPUT
8067 local facets="$(get_facets OST),$(get_facets MDS)"
8068 for facet in ${facets//,/ }; do
8069 local host=$(facet_host $facet)
8070 [ "$mgs_HOST" == "$host" ] &&
8071 echo "$host-$(facet_fstype $facet)" \
8075 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
8077 if [ $? -ne 0 ]; then
8078 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8079 error "ldev failed to produce the correct output!"
8082 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8084 run_test 96 "ldev returns hostname and backend fs correctly in command sub"
8087 if [ -z "$LDEV" ]; then
8088 error "ldev is missing!"
8091 local LDEVCONFPATH=$TMP/ldev.conf
8092 local NIDSPATH=$TMP/nids
8094 generate_ldev_conf $LDEVCONFPATH
8095 generate_nids $NIDSPATH
8097 local LDEV_OUTPUT=$TMP/ldev-output.txt
8098 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
8100 echo -e "\nMDT role"
8101 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mdt > $LDEV_OUTPUT
8103 if [ $? -ne 0 ]; then
8104 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
8105 error "ldev failed to execute for mdt role!"
8108 for num in $(seq $MDSCOUNT); do
8109 printf "%s-MDT%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
8112 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
8114 if [ $? -ne 0 ]; then
8115 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8116 error "ldev failed to produce the correct output for mdt role!"
8119 echo -e "\nOST role"
8120 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R ost > $LDEV_OUTPUT
8122 if [ $? -ne 0 ]; then
8123 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
8124 error "ldev failed to execute for ost role!"
8128 for num in $(seq $OSTCOUNT); do
8129 printf "%s-OST%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
8132 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
8134 if [ $? -ne 0 ]; then
8135 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8136 error "ldev failed to produce the correct output for ost role!"
8139 echo -e "\nMGS role"
8140 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mgs > $LDEV_OUTPUT
8142 if [ $? -ne 0 ]; then
8143 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
8144 error "ldev failed to execute for mgs role!"
8147 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
8149 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
8151 if [ $? -ne 0 ]; then
8152 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8153 error "ldev failed to produce the correct output for mgs role!"
8156 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8158 run_test 97 "ldev returns correct ouput when querying based on role"
8165 check_mount || error "mount failed"
8166 mountopt="user_xattr"
8167 for ((x = 1; x <= PAGE_SIZE/11; x++)); do
8168 mountopt="$mountopt,user_xattr"
8170 mount_client $MOUNT remount,$mountopt 2>&1 | grep "too long" ||
8171 error "Buffer overflow check failed"
8172 cleanup || error "cleanup failed"
8174 run_test 98 "Buffer-overflow check while parsing mount_opts"
8178 [[ "$ost1_FSTYPE" != ldiskfs ]] &&
8179 skip "ldiskfs only test"
8180 [[ "$OST1_VERSION" -ge $(version_code 2.8.57) ]] ||
8181 skip "Need OST version at least 2.8.57"
8183 local ost_opts="$(mkfs_opts ost1 $(ostdevname 1)) \
8184 --reformat $(ostdevname 1) $(ostvdevname 1)"
8185 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" &&
8186 skip "meta_bg already set"
8189 if [[ ${!opts} != *mkfsoptions* ]]; then
8190 eval opts=\"${!opts} \
8191 --mkfsoptions='\\\"-O ^resize_inode,meta_bg\\\"'\"
8193 local val=${!opts//--mkfsoptions=\\\"/ \
8194 --mkfsoptions=\\\"-O ^resize_inode,meta_bg }
8198 echo "params: $opts"
8201 add ost1 $opts || error "add ost1 failed with new params"
8203 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" ||
8204 error "meta_bg is not set"
8206 run_test 99 "Adding meta_bg option"
8211 start_mgsmds || error "MDS start failed"
8212 start_ost || error "unable to start OST"
8213 mount_client $MOUNT || error "client start failed"
8214 check_mount || error "check_mount failed"
8223 do_facet mgs 'lshowmount -v' | awk 'BEGIN {NR == 0; rc=1} /MGS:/ {rc=0}
8224 END {exit rc}' || error "lshowmount have no output MGS"
8226 do_facet mds1 'lshowmount -v' | awk 'BEGIN {NR == 2; rc=1} /-MDT0000:/
8227 {rc=0} END {exit rc}' || error "lshowmount have no output MDT0"
8229 do_facet ost1 'lshowmount -v' | awk 'BEGIN {NR == 4; rc=1} /-OST0000:/
8230 {rc=0} END {exit rc}' || error "lshowmount have no output OST0"
8232 cleanup || error "cleanup failed with $?"
8234 run_test 100 "check lshowmount lists MGS, MDT, OST and 0@lo"
8237 local createmany_pid
8238 local dev=$FSNAME-OST0000-osc-MDT0000
8242 do_nodes $(comma_list $(osts_nodes)) $LCTL set_param \
8243 seq.*OST*-super.width=$DATA_SEQ_MAX_WIDTH
8244 createmany -o $DIR1/$tdir/$tfile-%d 50000 &
8246 # MDT->OST reconnection causes MDT<->OST last_id synchornisation
8247 # via osp_precreate_cleanup_orphans.
8248 for ((i = 0; i < 100; i++)); do
8249 for ((k = 0; k < 10; k++)); do
8250 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;" \
8251 "$LCTL --device $dev activate"
8254 ls -asl $MOUNT | grep '???' &&
8255 { kill -9 $createmany_pid &>/dev/null;
8256 error "File has no object on OST"; }
8258 kill -s 0 $createmany_pid || break
8260 wait $createmany_pid
8262 unlinkmany $DIR1/$tdir/$tfile-%d 50000
8265 run_test 101a "Race MDT->OST reconnection with create"
8268 local dev=$FSNAME-OST0000-osc-MDT0000
8269 local dir=$DIR1/$tdir
8273 $LFS setstripe -c 1 -i 0 $dir
8274 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;"
8275 #define OBD_FAIL_OSP_CON_EVENT_DELAY 0x2107
8276 do_facet mds1 "$LCTL set_param fail_loc=0x80002107 fail_val=20"
8277 do_facet $SINGLEMDS "$LCTL --device $dev activate;"
8282 wait_osc_import_ready client ost1
8283 touch $dir/$tfile || error "Can't create file"
8287 run_test 101b "Race events DISCONNECT and ACTIVE in osp"
8290 [[ "$MDS1_VERSION" -gt $(version_code 2.9.53) ]] ||
8291 skip "Need server version greater than 2.9.53"
8292 [[ “$(mdsdevname 1)” != “$(mgsdevname)” ]] &&
8293 [[ “$(facet_host mds1)” = “$(facet_host mgs)” ]] &&
8294 skip "MGS must be on different node or combined"
8296 cleanup || error "cleanup failed with $?"
8298 local mds1dev=$(mdsdevname 1)
8299 local mds1mnt=$(facet_mntpt mds1)
8300 local mds1opts=$MDS_MOUNT_OPTS
8302 if [ "$mds1_FSTYPE" == ldiskfs ] &&
8303 ! do_facet mds1 test -b $mds1dev; then
8304 mds1opts=$(csa_add "$mds1opts" -o loop)
8306 if [[ "$mds1_FSTYPE" == zfs ]]; then
8307 import_zpool mds1 || return ${PIPESTATUS[0]}
8310 # unload all and only load libcfs to allow fail_loc setting
8311 do_facet mds1 $LUSTRE_RMMOD || error "unable to unload modules"
8312 do_facet mds1 modprobe libcfs || error "libcfs not loaded"
8313 do_facet mds1 lsmod \| grep libcfs || error "libcfs not loaded"
8315 #define OBD_FAIL_OBDCLASS_MODULE_LOAD 0x60a
8316 do_facet mds1 "$LCTL set_param fail_loc=0x8000060a"
8318 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts &&
8319 error "mdt start must fail"
8320 do_facet mds1 lsmod \| grep obdclass && error "obdclass must not load"
8322 do_facet mds1 "$LCTL set_param fail_loc=0x0"
8324 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts ||
8325 error "mdt start must not fail"
8327 cleanup || error "cleanup failed with $?"
8329 run_test 102 "obdclass module cleanup upon error"
8334 echo "rename $FSNAME to $newname"
8336 if ! combined_mgs_mds ; then
8337 local dev=$(mgsdevname)
8340 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
8341 error "(7) Fail to rename MGS"
8342 if [ "$(facet_fstype mgs)" = "zfs" ]; then
8343 reimport_zpool mgs $newname-mgs
8347 for num in $(seq $MDSCOUNT); do
8348 local dev=$(mdsdevname $num)
8350 do_facet mds${num} \
8351 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
8352 error "(8) Fail to rename MDT $num"
8353 if [ "$(facet_fstype mds${num})" = "zfs" ]; then
8354 reimport_zpool mds${num} $newname-mdt${num}
8358 for num in $(seq $OSTCOUNT); do
8359 local dev=$(ostdevname $num)
8361 do_facet ost${num} \
8362 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
8363 error "(9) Fail to rename OST $num"
8364 if [ "$(facet_fstype ost${num})" = "zfs" ]; then
8365 reimport_zpool ost${num} $newname-ost${num}
8370 test_103_set_pool() {
8374 do_facet mgs $LCTL pool_add $FSNAME.$pname ${FSNAME}-$ost_x ||
8375 error "Fail to add $ost_x to $FSNAME.$pname"
8376 wait_update $HOSTNAME \
8377 "lctl get_param -n lov.$FSNAME-clilov-*.pools.$pname |
8378 grep $ost_x" "$FSNAME-${ost_x}_UUID" ||
8379 error "$ost_x is NOT in pool $FSNAME.$pname"
8382 test_103_check_pool() {
8383 local save_fsname=$1
8386 stat $DIR/$tdir/test-framework.sh ||
8387 error "($errno) Fail to stat"
8388 do_facet mgs $LCTL pool_list $FSNAME.pool1 ||
8389 error "($errno) Fail to list $FSNAME.pool1"
8390 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname ||
8391 error "($errno) Fail to list $FSNAME.$save_fsname"
8392 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname |
8393 grep ${FSNAME}-OST0000 ||
8394 error "($errno) List $FSNAME.$save_fsname is invalid"
8396 local pname=$($LFS getstripe --pool $DIR/$tdir/d0)
8397 [ "$pname" = "$save_fsname" ] ||
8398 error "($errno) Unexpected pool name $pname"
8402 check_mount_and_prep
8404 mkdir $DIR/$tdir || error "(1) Fail to mkdir $DIR/$tdir"
8405 cp $LUSTRE/tests/test-framework.sh $DIR/$tdir ||
8406 error "(2) Fail to copy test-framework.sh"
8408 do_facet mgs $LCTL pool_new $FSNAME.pool1 ||
8409 error "(3) Fail to create $FSNAME.pool1"
8410 # name the pool name as the fsname
8411 do_facet mgs $LCTL pool_new $FSNAME.$FSNAME ||
8412 error "(4) Fail to create $FSNAME.$FSNAME"
8414 test_103_set_pool $FSNAME OST0000
8416 $LFS setstripe -p $FSNAME $DIR/$tdir/d0 ||
8417 error "(6) Fail to setstripe on $DIR/$tdir/d0"
8422 test_renamefs mylustre
8424 local save_fsname=$FSNAME
8428 test_103_check_pool $save_fsname 7
8430 if [ $OSTCOUNT -ge 2 ]; then
8431 test_103_set_pool $save_fsname OST0001
8434 $LFS setstripe -p $save_fsname $DIR/$tdir/f0 ||
8435 error "(16) Fail to setstripe on $DIR/$tdir/f0"
8444 test_103_check_pool $save_fsname 17
8448 test_renamefs $save_fsname
8454 run_test 103 "rename filesystem name"
8456 test_104a() { # LU-6952
8457 local mds_mountopts=$MDS_MOUNT_OPTS
8458 local ost_mountopts=$OST_MOUNT_OPTS
8459 local mds_mountfsopts=$MDS_MOUNT_FS_OPTS
8460 local lctl_ver=$(do_facet $SINGLEMDS $LCTL --version |
8463 [[ $(version_code $lctl_ver) -lt $(version_code 2.9.55) ]] &&
8464 skip "this test needs utils above 2.9.55"
8466 # specify "acl" in mount options used by mkfs.lustre
8467 if [ -z "$MDS_MOUNT_FS_OPTS" ]; then
8468 MDS_MOUNT_FS_OPTS="acl,user_xattr"
8471 MDS_MOUNT_FS_OPTS="${MDS_MOUNT_FS_OPTS},acl,user_xattr"
8474 echo "mountfsopt: $MDS_MOUNT_FS_OPTS"
8476 #reformat/remount the MDT to apply the MDT_MOUNT_FS_OPT options
8478 if ! combined_mgs_mds ; then
8482 if [ -z "$MDS_MOUNT_OPTS" ]; then
8483 MDS_MOUNT_OPTS="-o noacl"
8485 MDS_MOUNT_OPTS="${MDS_MOUNT_OPTS},noacl"
8488 for num in $(seq $MDSCOUNT); do
8489 start mds$num $(mdsdevname $num) $MDS_MOUNT_OPTS ||
8490 error "Failed to start MDS"
8493 for num in $(seq $OSTCOUNT); do
8494 start ost$num $(ostdevname $num) $OST_MOUNT_OPTS ||
8495 error "Failed to start OST"
8499 setfacl -m "d:$RUNAS_ID:rwx" $MOUNT &&
8500 error "ACL is applied when FS is mounted with noacl."
8502 MDS_MOUNT_OPTS=$mds_mountopts
8503 OST_MOUNT_OPTS=$ost_mountopts
8504 MDS_MOUNT_FS_OPTS=$mds_mountfsopts
8506 run_test 104a "Make sure user defined options are reflected in mount"
8508 test_104b() { # LU-12859
8509 mount_client $MOUNT3 flock,localflock
8510 stack_trap "umount_client $MOUNT3" EXIT
8511 mount | grep "$MOUNT3 .*,flock" && error "flock is still set"
8512 mount | grep "$MOUNT3 .*,localflock" || error "localflock is not set"
8513 umount_client $MOUNT3
8514 mount_client $MOUNT3 localflock,flock
8515 mount | grep "$MOUNT3 .*,localflock" && error "localflock is still set"
8516 mount | grep "$MOUNT3 .*,flock" || error "flock is not set"
8517 umount_client $MOUNT3
8518 mount_client $MOUNT3 localflock,flock,noflock
8519 flock_is_enabled $MOUNT3 && error "some flock is still enabled" || true
8521 run_test 104b "Mount uses last flock argument"
8523 error_and_umount() {
8534 mount --bind $DIR $TMP/$tdir || error "mount bind mnt pt failed"
8535 rm -f $TMP/$tdir/$tfile
8536 rm -f $TMP/$tdir/${tfile}1
8538 # Files should not be created in ro bind mount point
8539 # remounting from rw to ro
8540 mount -o remount,ro $TMP/$tdir ||
8541 error_and_umount "readonly remount of bind mnt pt failed"
8542 touch $TMP/$tdir/$tfile &&
8543 error_and_umount "touch succeeds on ro bind mnt pt"
8544 [ -e $TMP/$tdir/$tfile ] &&
8545 error_and_umount "file created on ro bind mnt pt"
8547 # Files should be created in rw bind mount point
8548 # remounting from ro to rw
8549 mount -o remount,rw $TMP/$tdir ||
8550 error_and_umount "read-write remount of bind mnt pt failed"
8551 touch $TMP/$tdir/${tfile}1 ||
8552 error_and_umount "touch fails on rw bind mnt pt"
8553 [ -e $TMP/$tdir/${tfile}1 ] ||
8554 error_and_umount "file not created on rw bind mnt pt"
8555 umount $TMP/$tdir || error "umount of bind mnt pt failed"
8557 cleanup || error "cleanup failed with $?"
8559 run_test 105 "check file creation for ro and rw bind mnt pt"
8563 local creates=64768 # one full plain llog
8565 # ensure there are enough inodes in the filesystem
8566 (( OSTSIZE < (creates + 1024) * 8)) && OSTSIZE=$(((creates + 1024) * 8))
8570 mkdir -p $DIR/$tdir || error "create $tdir failed"
8571 do_nodes $(comma_list $(osts_nodes)) $LCTL set_param \
8572 seq.*OST*-super.width=$DATA_SEQ_MAX_WIDTH
8573 lfs setstripe -c 1 -i 0 $DIR/$tdir
8574 #define OBD_FAIL_CAT_RECORDS 0x1312
8575 do_facet mds1 $LCTL set_param fail_loc=0x1312 fail_val=$repeat
8577 for ((i = 1; i <= $repeat; i++)); do
8578 createmany -o $DIR/$tdir/f- $creates || lfs df -i
8579 createmany -u $DIR/$tdir/f- $creates
8580 wait_delete_completed $((TIMEOUT * 7))
8582 #ASSERTION osp_sync_thread() ( thread->t_flags != SVC_RUNNING ) failed
8583 #shows that osp code is buggy
8584 do_facet mds1 $LCTL set_param fail_loc=0 fail_val=0
8588 run_test 106 "check osp llog processing when catalog is wrapped"
8591 [[ "$MDS1_VERSION" -ge $(version_code 2.10.50) ]] ||
8592 skip "Need MDS version > 2.10.50"
8595 start_mgsmds || error "start_mgsmds failed"
8596 start_ost || error "unable to start OST"
8598 # add unknown configuration parameter.
8599 if [[ $PERM_CMD == *"set_param -P"* ]]; then
8600 cmd="$PERM_CMD ost.$FSNAME-OST0000*.unknown_param"
8602 cmd="$PERM_CMD $FSNAME-OST0000*.ost.unknown_param"
8604 do_facet mgs "$cmd=50"
8605 cleanup_nocli || error "cleanup_nocli failed with $?"
8608 # unknown param should be ignored while mounting.
8609 start_ost || error "unable to start OST after unknown param set"
8611 cleanup || error "cleanup failed with $?"
8613 run_test 107 "Unknown config param should not fail target mounting"
8618 $rcmd rm -rf $tmp > /dev/null 2>&1
8619 $rcmd mkdir -p $tmp/{mnt,images} || error "failed to mkdir remotely"
8621 for facet in $facets; do
8622 [ "$mds1_FSTYPE" = zfs ] &&
8623 $rcmd $ZPOOL -f export lustre-$facet > /dev/null 2>&1
8624 $rcmd mkdir $tmp/mnt/$facet ||
8625 error "failed to mkdir $tmp/mnt/$facet"
8626 $rcmd dd if=/dev/zero of=$tmp/images/$facet \
8627 seek=199 bs=1M count=1 ||
8628 error "failed to create $tmp/images/$facet"
8637 local facet=${role}$((idx + 1))
8639 [ $# -eq 5 ] && pool=$5
8641 do_facet $SINGLEMDS $MKFS --fsname=lustre --$mgs \
8642 --$role --index=$idx --replace --backfstype=$bkfs \
8643 --device-size=200000 --reformat $pool $tmp/images/$facet ||
8644 error "failed to mkfs for $facet"
8648 echo "mounting client..."
8649 mount -t lustre ${nid}:/lustre $MOUNT ||
8650 error "failed to mount lustre"
8653 ls -l $MOUNT/local_dir || error "failed to list"
8655 echo "check truncate && write"
8656 echo "dummmmmmmmmmmmm" > $MOUNT/remote_dir/fsx.c ||
8657 error "failed to tuncate & write"
8660 touch $MOUNT/foooo ||
8661 error "failed to create"
8663 echo "check read && write && append"
8664 sha1sum $MOUNT/conf-sanity.sh |
8665 awk '{ print $1 }' > $MOUNT/checksum.new ||
8666 error "failed to read(1)"
8667 sha1sum $MOUNT/remote_dir/unlinkmany.c |
8668 awk '{ print $1 }' >> $MOUNT/checksum.new ||
8669 error "failed to read(2)"
8670 sha1sum $MOUNT/striped_dir/lockahead_test.o |
8671 awk '{ print $1 }' >> $MOUNT/checksum.new ||
8672 error "failed to read(3)"
8675 diff $MOUNT/checksum.new $MOUNT/checksum.src ||
8676 error "failed to verify data"
8686 umount -f $MOUNT || error "failed to umount client"
8687 for facet in $facets; do
8688 $rcmd umount -f $tmp/mnt/$facet ||
8689 error "failed to umount $facet"
8690 if [ "$mds1_FSTYPE" = zfs ]; then
8691 $rcmd $ZPOOL export -f lustre-$facet ||
8692 error "failed to export lustre-$facet"
8696 $rcmd rm -rf $tmp || error "failed to rm the dir $tmp"
8700 [ "$CLIENTONLY" ] && skip "Client-only testing"
8701 [ "$mds1_FSTYPE" != zfs ] && skip "zfs only test"
8702 [ "$MDS1_VERSION" -lt $(version_code 2.10.58) ] &&
8703 skip "Need server version at least 2.10.58"
8708 local tmp=$TMP/$tdir
8709 local rcmd="do_facet $SINGLEMDS"
8710 local facets="mdt1 mdt2 ost1 ost2"
8711 local nid=$($rcmd $LCTL list_nids | head -1)
8714 trap t_108_cleanup EXIT ERR
8717 t_108_mkfs mdt 0 zfs mgs lustre-mdt1/mdt1
8718 t_108_mkfs mdt 1 zfs mgsnode=$nid lustre-mdt2/mdt2
8719 t_108_mkfs ost 0 zfs mgsnode=$nid lustre-ost1/ost1
8720 t_108_mkfs ost 1 zfs mgsnode=$nid lustre-ost2/ost2
8722 for facet in $facets; do
8723 $rcmd zfs set mountpoint=$tmp/mnt/$facet canmount=on \
8724 lustre-$facet/$facet ||
8725 error "failed to zfs set for $facet (1)"
8726 $rcmd zfs mount lustre-$facet/$facet ||
8727 error "failed to local mount $facet"
8728 $rcmd tar jxf $LUSTRE/tests/ldiskfs_${facet}_2_11.tar.bz2 \
8729 --xattrs --xattrs-include="trusted.*" \
8730 -C $tmp/mnt/$facet/ > /dev/null 2>&1 ||
8731 error "failed to untar image for $facet"
8732 $rcmd "cd $tmp/mnt/$facet && rm -rf oi.* OI_* lfsck_* LFSCK" ||
8733 error "failed to cleanup for $facet"
8734 $rcmd zfs umount lustre-$facet/$facet ||
8735 error "failed to local umount $facet"
8736 $rcmd zfs set canmount=off lustre-$facet/$facet ||
8737 error "failed to zfs set $facet (2)"
8740 echo "changing server nid..."
8741 $rcmd mount -t lustre -o nosvc lustre-mdt1/mdt1 $tmp/mnt/mdt1
8742 $rcmd lctl replace_nids $FSNAME-MDT0000 $nid
8743 $rcmd lctl replace_nids $FSNAME-MDT0001 $nid
8744 $rcmd lctl replace_nids $FSNAME-OST0000 $nid
8745 $rcmd lctl replace_nids $FSNAME-OST0001 $nid
8746 $rcmd umount $tmp/mnt/mdt1
8748 for facet in $facets; do
8749 echo "mounting $facet from backup..."
8750 $rcmd mount -t lustre -o abort_recov lustre-$facet/$facet \
8751 $tmp/mnt/$facet || error "failed to mount $facet"
8754 # ZFS backend can detect migration and trigger OI scrub automatically
8755 # sleep 3 seconds for scrub done
8761 run_test 108a "migrate from ldiskfs to ZFS"
8764 [ "$CLIENTONLY" ] && skip "Client-only testing"
8765 [ "$mds1_FSTYPE" != ldiskfs ] && skip "ldiskfs only test"
8766 [ "$MDS1_VERSION" -lt $(version_code 2.10.58) ] &&
8767 skip "Need server version at least 2.10.58"
8772 local tmp=$TMP/$tdir
8773 local rcmd="do_facet $SINGLEMDS"
8774 local facets="mdt1 mdt2 ost1 ost2"
8775 local scrub_list="MDT0000 MDT0001 OST0000 OST0001"
8776 local nid=$($rcmd $LCTL list_nids | head -1)
8779 trap t_108_cleanup EXIT ERR
8782 t_108_mkfs mdt 0 ldiskfs mgs
8783 t_108_mkfs mdt 1 ldiskfs mgsnode=$nid
8784 t_108_mkfs ost 0 ldiskfs mgsnode=$nid
8785 t_108_mkfs ost 1 ldiskfs mgsnode=$nid
8787 for facet in $facets; do
8788 $rcmd mount -t ldiskfs -o loop $tmp/images/$facet \
8790 error "failed to local mount $facet"
8792 $rcmd tar jxf $LUSTRE/tests/zfs_${facet}_2_11.tar.bz2 \
8793 --xattrs --xattrs-include="*.*" \
8794 -C $tmp/mnt/$facet/ > /dev/null 2>&1 ||
8795 error "failed to untar image for $facet"
8796 $rcmd "cd $tmp/mnt/$facet && rm -rf oi.* OI_* lfsck_* LFSCK" ||
8797 error "failed to cleanup for $facet"
8798 $rcmd umount $tmp/mnt/$facet ||
8799 error "failed to local umount $facet"
8802 echo "changing server nid..."
8803 $rcmd mount -t lustre -o nosvc,loop $tmp/images/mdt1 $tmp/mnt/mdt1
8804 $rcmd lctl replace_nids $FSNAME-MDT0000 $nid
8805 $rcmd lctl replace_nids $FSNAME-MDT0001 $nid
8806 $rcmd lctl replace_nids $FSNAME-OST0000 $nid
8807 $rcmd lctl replace_nids $FSNAME-OST0001 $nid
8808 $rcmd umount $tmp/mnt/mdt1
8810 for facet in $facets; do
8811 echo "mounting $facet from backup..."
8812 $rcmd mount -t lustre -o loop,abort_recov $tmp/images/$facet \
8813 $tmp/mnt/$facet || error "failed to mount $facet"
8816 for facet in $scrub_list; do
8817 $rcmd $LCTL lfsck_start -M $FSNAME-$facet -t scrub ||
8818 error "failed to start OI scrub on $facet"
8821 # sleep 3 seconds for scrub done
8827 run_test 108b "migrate from ZFS to ldiskfs"
8830 # set number of permanent parameters
8832 test_109_set_params() {
8835 set_persistent_param_and_check mds \
8836 "mdd.$fsname-MDT0000.atime_diff" \
8837 "$fsname-MDT0000.mdd.atime_diff" \
8839 set_persistent_param_and_check mds \
8840 "mdd.$fsname-MDT0000.atime_diff" \
8841 "$fsname-MDT0000.mdd.atime_diff" \
8843 set_persistent_param_and_check client \
8844 "llite.$fsname*.max_read_ahead_mb" \
8845 "$fsname.llite.max_read_ahead_mb" \
8847 set_persistent_param_and_check client \
8848 "llite.$fsname*.max_read_ahead_mb" \
8849 "$fsname.llite.max_read_ahead_mb" \
8851 create_pool $fsname.pool1 || error "create pool failed"
8852 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
8853 error "pool_add failed"
8854 do_facet mgs $LCTL pool_remove $fsname.pool1 OST0000 ||
8855 error "pool_remove failed"
8856 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
8857 error "pool_add failed"
8861 # check permanent parameters
8863 test_109_test_params() {
8866 local atime_diff=$(do_facet mds $LCTL \
8867 get_param -n mdd.$fsname-MDT0000.atime_diff)
8868 [ $atime_diff == 63 ] || error "wrong mdd parameter after clear_conf"
8869 local max_read_ahead_mb=$(do_facet client $LCTL \
8870 get_param -n llite.$fsname*.max_read_ahead_mb)
8871 [ $max_read_ahead_mb == 64 ] ||
8872 error "wrong llite parameter after clear_conf"
8873 local ost_in_pool=$(do_facet mds $LCTL pool_list $fsname.pool1 |
8874 grep -v "^Pool:" | sed 's/_UUID//')
8875 [ $ost_in_pool = "$fsname-OST0000" ] ||
8876 error "wrong pool after clear_conf"
8880 # run lctl clear_conf, store CONFIGS before and after that
8882 test_109_clear_conf()
8884 local clear_conf_arg=$1
8887 if ! combined_mgs_mds ; then
8888 mgsdev=$(mgsdevname)
8889 stop_mgs || error "stop_mgs failed"
8890 start_mgs "-o nosvc" || error "start_mgs nosvc failed"
8892 mgsdev=$(mdsdevname 1)
8893 start_mdt 1 "-o nosvc" || error "start_mdt 1 nosvc failed"
8896 do_facet mgs "rm -rf $TMP/${tdir}/conf1; mkdir -p $TMP/${tdir}/conf1;" \
8897 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf1\\\" \
8901 # the command being tested
8903 do_facet mgs $LCTL clear_conf $clear_conf_arg ||
8904 error "clear_conf failed"
8905 if ! combined_mgs_mds ; then
8906 stop_mgs || error "stop_mgs failed"
8908 stop_mdt 1 || error "stop_mdt 1 failed"
8911 do_facet mgs "rm -rf $TMP/${tdir}/conf2; mkdir -p $TMP/${tdir}/conf2;" \
8912 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf2\\\" \
8916 test_109_file_shortened() {
8918 local sizes=($(do_facet mgs "stat -c %s " \
8919 "$TMP/${tdir}/conf1/CONFIGS/$file" \
8920 "$TMP/${tdir}/conf2/CONFIGS/$file"))
8921 [ ${sizes[1]} -lt ${sizes[0]} ] && return 0
8927 [ $MDS1_VERSION -lt $(version_code 2.10.59) ] &&
8928 skip "Needs MDS version 2.10.59 or later."
8929 [ "$(facet_fstype mgs)" == zfs ] &&
8930 skip "LU-8727: no implementation for ZFS"
8935 client_up || error "client_up failed"
8938 # set number of permanent parameters
8940 test_109_set_params $FSNAME
8942 umount_client $MOUNT || error "umount_client failed"
8943 stop_ost || error "stop_ost failed"
8944 stop_mds || error "stop_mds failed"
8946 test_109_clear_conf $FSNAME
8948 # make sure that all configs are cleared
8950 test_109_file_shortened $FSNAME-MDT0000 ||
8951 error "failed to clear MDT0000 config"
8952 test_109_file_shortened $FSNAME-client ||
8953 error "failed to clear client config"
8958 # check that configurations are intact
8960 test_109_test_params $FSNAME
8965 destroy_test_pools || error "destroy test pools failed"
8969 run_test 109a "test lctl clear_conf fsname"
8973 [ $MDS1_VERSION -lt $(version_code 2.10.59) ] &&
8974 skip "Needs MDS version 2.10.59 or later."
8975 [ "$(facet_fstype mgs)" == zfs ] &&
8976 skip "LU-8727: no implementation for ZFS"
8981 client_up || error "client_up failed"
8984 # set number of permanent parameters
8986 test_109_set_params $FSNAME
8988 umount_client $MOUNT || error "umount_client failed"
8989 stop_ost || error "stop_ost failed"
8990 stop_mds || error "stop_mds failed"
8992 test_109_clear_conf $FSNAME-MDT0000
8994 # make sure that only one config is cleared
8996 test_109_file_shortened $FSNAME-MDT0000 ||
8997 error "failed to clear MDT0000 config"
8998 test_109_file_shortened $FSNAME-client &&
8999 error "failed to clear client config"
9003 # check that configurations are intact
9005 test_109_test_params $FSNAME
9010 destroy_test_pools || error "destroy test pools failed"
9014 run_test 109b "test lctl clear_conf one config"
9018 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
9019 skip "Only applicable to ldiskfs-based MDTs"
9021 do_facet $SINGLEMDS $DEBUGFS -w -R supported_features |grep large_dir ||
9022 skip "large_dir option is not supported on MDS"
9023 do_facet ost1 $DEBUGFS -w -R supported_features | grep large_dir ||
9024 skip "large_dir option is not supported on OSS"
9026 stopall # stop all targets before modifying the target counts
9027 local old_mdscount=$MDSCOUNT
9028 local old_ostcount=$OSTCOUNT
9030 stack_trap "MDSCOUNT=$old_mdscount OSTCOUNT=$old_ostcount; reformat"
9034 # ext4_dir_entry_2 struct size:264
9035 # dx_root struct size:8
9036 # dx_node struct size:8
9037 # dx_entry struct size:8
9038 # For 1024 bytes block size.
9039 # First level directory entries: 126
9040 # Second level directory entries: 127
9041 # Entries in leaf: 3
9042 # For 2 levels limit: 48006
9043 # For 3 levels limit : 6096762
9044 # Create 80000 files to safely exceed 2-level htree limit.
9045 CONF_SANITY_110_LINKS=${CONF_SANITY_110_LINKS:-80000}
9047 # can fit at most 3 filenames per 1KB leaf block, but each
9048 # leaf/index block will only be 3/4 full before split at each level
9049 (( MDSSIZE < CONF_SANITY_110_LINKS / 3 * 4/3 * 4/3 )) &&
9050 CONF_SANITY_110_LINKS=$((MDSSIZE * 3 * 3/4 * 3/4))
9052 combined_mgs_mds || replace=" --replace "
9053 local opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
9054 $replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
9055 if [[ $opts != *large_dir* ]]; then
9056 if [[ $opts != *mkfsoptions* ]]; then
9057 opts+=" --mkfsoptions=\\\"-O large_dir -b 1024 -i 65536\\\""
9059 opts="${opts//--mkfsoptions=\\\"/ \
9060 --mkfsoptions=\\\"-O large_dir -b 1024 -i 65536 }"
9063 echo "MDT params: $opts"
9065 combined_mgs_mds || start_mgs
9066 add mds1 $opts || error "add mds1 failed with new params"
9067 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
9069 opts="$(mkfs_opts ost1 $(ostdevname 1)) \
9070 $replace --reformat $(ostdevname 1) $(ostvdevname 1)"
9072 if [[ $opts != *large_dir* ]]; then
9073 if [[ $opts != *mkfsoptions* ]]; then
9074 opts+=" --mkfsoptions=\\\"-O large_dir\\\" "
9076 opts="${opts//--mkfsoptions=\\\"/ \
9077 --mkfsoptions=\\\"-O large_dir }"
9080 echo "OST params: $opts"
9081 add ost1 $opts || error "add ost1 failed with new params"
9082 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
9084 MOUNT_2=yes mountcli || error "mount clients failed"
9086 mkdir -v $DIR/$tdir || error "cannot create $DIR/$tdir"
9087 local pids count=0 group=0
9089 echo "creating $CONF_SANITY_110_LINKS in total"
9090 while (( count < CONF_SANITY_110_LINKS )); do
9091 local len=$((253 - $(wc -c <<<"$tfile-$group-40000-")))
9092 local dir=DIR$((group % 2 + 1))
9093 local target=${!dir}/$tdir/$tfile-$group
9094 local long=$target-$(generate_name $len)-
9095 local create=$((CONF_SANITY_110_LINKS - count))
9097 (( create > 40000 )) && create=40000
9098 touch $target || error "creating $target failed"
9099 echo "creating $create hard links to $target"
9100 createmany -l $target $long $create &
9103 count=$((count + create))
9104 group=$((group + 1))
9106 echo "waiting for PIDs$pids to complete"
9107 wait $pids || error "createmany failed after $group groups"
9109 umount_client $MOUNT2 -f
9112 run_e2fsck $(facet_active_host mds1) $(mdsdevname 1) -n
9113 MDSCOUNT=$old_mdscount
9114 OSTCOUNT=$old_ostcount
9116 run_test 110 "Adding large_dir with 3-level htree"
9119 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
9120 skip "Only applicable to ldiskfs-based MDTs"
9122 is_dm_flakey_dev $SINGLEMDS $(mdsdevname 1) &&
9123 skip "This test can not be executed on flakey dev"
9125 do_facet $SINGLEMDS $DEBUGFS -w -R supported_features |grep large_dir ||
9126 skip "large_dir option is not supported on MDS"
9128 do_facet ost1 $DEBUGFS -w -R supported_features | grep large_dir ||
9129 skip "large_dir option is not supported on OSS"
9131 # cleanup before changing target counts
9133 local old_mdscount=$MDSCOUNT
9134 local old_ostcount=$OSTCOUNT
9135 local old_mdssize=$MDSSIZE
9137 stack_trap "MDSSIZE=$old_mdssize MDSCOUNT=$old_mdscount \
9138 OSTCOUNT=$old_ostcount; reformat"
9141 (( MDSSIZE < 2400000 )) && MDSSIZE=2400000 # need at least 2.4GB
9143 local mdsdev=$(mdsdevname 1)
9144 combined_mgs_mds || replace=" --replace "
9145 local opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
9146 $replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
9147 if [[ $opts != *large_dir* ]]; then
9148 if [[ $opts != *mkfsoptions* ]]; then
9149 opts+=" --mkfsoptions=\\\"-O large_dir -i 1048576 \\\" "
9151 opts="${opts//--mkfsoptions=\\\"/ \
9152 --mkfsoptions=\\\"-O large_dir -i 1048576 }"
9155 echo "MDT params: $opts"
9157 combined_mgs_mds || start_mgs
9158 __touch_device mds 1
9159 add mds1 $opts || error "add mds1 failed with new params"
9160 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
9162 opts="$(mkfs_opts ost1 $(ostdevname 1)) \
9163 $replace --reformat $(ostdevname 1) $(ostvdevname 1)"
9164 if [[ $opts != *large_dir* ]]; then
9165 if [[ $opts != *mkfsoptions* ]]; then
9166 opts+=" --mkfsoptions=\\\"-O large_dir \\\""
9168 opts="${opts//--mkfsoptions=\\\"/ --mkfsoptions=\\\"-O large_dir }"
9171 echo "OST params: $opts"
9172 __touch_device ost 1
9173 add ost1 $opts || error "add ost1 failed with new params"
9174 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
9176 MOUNT_2=yes mountcli
9177 mkdir $DIR/$tdir || error "cannot create $DIR/$tdir"
9179 lfs df -i $DIR/$tdir
9183 local start=$SECONDS
9185 local dirmax=$((2 << 30))
9191 while (( !needskip & dirsize < dirmax )); do
9194 for cli in ${CLIENTS//,/ }; do
9195 local len=$((253 - $(wc -c <<<"$cli-$group-60000-")))
9196 local target=$cli-$group
9197 local long=$DIR/$tdir/$target-$(generate_name $len)-
9199 RPWD=$DIR/$tdir do_node $cli touch $target ||
9200 error "creating $target failed"
9201 echo "creating 60000 hardlinks to $target"
9202 RPWD=$DIR/$tdir do_node $cli createmany -l $target $long 60000 &
9205 group=$((group + 1))
9207 long=$DIR2/$tdir/$target-$(generate_name $len)-
9209 RPWD=$DIR2/$tdir do_node $cli touch $target ||
9210 error "creating $target failed"
9211 echo "creating 60000 hardlinks to $target"
9212 RPWD=$DIR2/$tdir do_node $cli createmany -l $target $long 60000 &
9215 group=$((group + 1))
9217 echo "waiting for PIDs$pids to complete"
9218 wait $pids || error "createmany failed after $group groups"
9219 dirsize=$(stat -c %s $DIR/$tdir)
9220 taken=$((SECONDS - start))
9221 rate=$((dirsize / taken))
9222 left=$(((dirmax - dirsize) / rate))
9223 num=$((group * 60000))
9224 echo "estimate ${left}s left after $num files / ${taken}s"
9225 # if the estimated time remaining is too large (it may change
9226 # over time as the create rate is not constant) then exit
9227 # without declaring a failure.
9228 (( left > 1200 )) && needskip=1
9231 umount_client $MOUNT2 -f
9234 ! (( $needskip )) ||
9235 echo "ETA ${left}s after $num files / ${taken}s is too long"
9237 run_e2fsck $(facet_active_host mds1) $(mdsdevname 1) -n
9238 MDSCOUNT=$old_mdscount
9239 OSTCOUNT=$old_ostcount
9240 MDSSIZE=$old_mdssize
9242 run_test 111 "Adding large_dir with over 2GB directory"
9245 local param="no_create"
9247 (( $OST1_VERSION > $(version_code 2.14.0) )) ||
9248 skip "need OSS at least 2.14.0"
9249 (( $OST1_VERSION >= $(version_code 2.15.56.125) )) ||
9250 param="no_precreate"
9252 start_mds || error "MDS start failed"
9253 start_ost || error "OSS start failed"
9254 echo "start ost2 service on $(facet_active_host ost2)"
9255 start ost2 $(ostdevname 2) $(csa_add "$OST_MOUNT_OPTS" -o $param) ||
9256 error "start ost2 facet failed"
9257 local val=$(do_facet ost2 \
9258 "$LCTL get_param -n obdfilter.$FSNAME-OST0001*.$param")
9259 (( $val == 1 )) || error "obdfilter.$FSNAME-OST0001*.$param=$val"
9261 mount_client $MOUNT || error "mount client failed"
9262 wait_osc_import_state mds1 ost1 FULL
9263 wait_osc_import_ready client ost1
9264 wait_osc_import_ready client ost2
9266 $LFS setstripe -i 0 $DIR/$tfile.0 ||
9267 error "problem creating $tfile.0 on OST0000"
9268 $LFS setstripe -i 1 $DIR/$tfile.1 && $LFS getstripe $DIR/$tfile.1 &&
9269 (( $($LFS getstripe -i $DIR/$tfile.1) == 1 )) &&
9270 error "allowed to create $tfile.1 on OST0001"
9272 $LFS df -v $MOUNT | grep -q "OST:1.*N" ||
9273 error "NOCREATE not in 'lfs df'"
9275 do_facet ost2 $LCTL set_param obdfilter.$FSNAME-OST0001*.$param=0
9277 $LFS setstripe -i 1 $DIR/$tfile.2 ||
9278 error "failed to create $tfile.2 on ost1 facet"
9279 # files not cleaned with ONLY_REPEAT because of client unmount below
9280 rm -f $DIR/$tfile.[012]
9281 stop_ost2 || error "stop ost2 facet failed"
9284 run_test 112a "mount OST with no_create option"
9287 (( MDSCOUNT >= 2 )) || skip "need at least 2 MDTs"
9288 (( $MDS1_VERSION >= $(version_code 2.15.56.125) )) ||
9289 skip "need MDS >= 2.15.56.125"
9290 local mdsnum=$MDSCOUNT
9291 local facet=mds$mdsnum
9292 local mdtidx=$((mdsnum - 1))
9293 local mdtname=$FSNAME-MDT$(printf %04x $mdtidx)
9295 start_mds --mdscount $((mdsnum - 1)) || error "MDS start failed"
9296 start_mdt $mdsnum -o no_create || error "start $facet failed"
9297 local val=$(do_facet $facet \
9298 "$LCTL get_param -n mdt.$mdtname*.no_create")
9299 (( $val == 1 )) || error "mdt.$mdtname*.no_create=$val"
9300 start_ost || error "ost1 start failed"
9301 start_ost2 || error "ost1 start failed"
9303 mount_client $MOUNT || error "mount client failed"
9304 wait_osc_import_ready $facet ost2
9307 $LFS df -v $MOUNT | grep -q "MDT:$mdtidx.*N" ||
9308 error "NOCREATE not in 'lfs df'"
9310 $LFS mkdir -i $mdtidx $DIR/$tdir ||
9311 $LFS setdirstripe -D -c 1 -i -1 --max-inherit-rr 2 $DIR/$tdir ||
9312 error "error creating $tdir on $mdtname"
9313 stack_trap "rm -rf $DIR/$tdir"
9315 mkdir $DIR/$tdir/d1.{1..100} || error "mkdir $tdir/d1.{1..100} failed"
9316 $LFS getdirstripe -i $DIR/$tdir/d1.* | sort | uniq -c
9317 do_facet $facet $LCTL set_param mdt.$mdtname*.no_create=0
9318 # allow one initial create for delayed statfs on client
9319 (( $($LFS getdirstripe -i $DIR/$tdir/d1.* | grep -c $mdtidx) < 2 )) ||
9320 error "allowed create on $mdtname"
9323 mkdir $DIR/$tdir/d2.{1..100} || error "mkdir $tdir/d2.{1..100} failed"
9324 $LFS getdirstripe -i $DIR/$tdir/d2.{1..100} | sort | uniq -c
9325 (( $($LFS getdirstripe -i $DIR/$tdir/d2.* | grep -c $mdtidx) > 10 )) ||
9326 error "no create on $mdtname"
9327 # files not cleaned with ONLY_REPEAT because of client unmount below
9329 stop_ost2 || error "ost1 start failed"
9332 run_test 112b "mount MDT with no_create option"
9335 SAVE_MGS_MOUNT_OPTS=$MGS_MOUNT_OPTS
9336 SAVE_MDS_MOUNT_OPTS=$MDS_MOUNT_OPTS
9337 SAVE_OST_MOUNT_OPTS=$OST_MOUNT_OPTS
9343 MGS_MOUNT_OPTS=$SAVE_MGS_MOUNT_OPTS
9344 MDS_MOUNT_OPTS=$SAVE_MDS_MOUNT_OPTS
9345 OST_MOUNT_OPTS=$SAVE_OST_MOUNT_OPTS
9346 # Revert old mount options back
9348 # Subsequent following test requires
9349 # conf-sanity to be in stopall state.
9350 # Force 'stopall' so others following
9355 # Error out with mount info
9357 local server_nodes=$(comma_list $(mdts_nodes) $(osts_nodes))
9360 echo "--Client Mount Info--"
9361 mount | grep -i lustre
9362 echo "--Server Mount Info--"
9363 do_nodes $server_nodes mount | grep -i lustre
9369 local ost_version="2.15.51" # Minimum version required
9371 (( OST1_VERSION >= $(version_code $ost_version) )) ||
9372 skip "Need server version at least $ost_version"
9374 stack_trap cleanup_113 EXIT
9376 # Reset before starting
9380 # Verify MDS's should start with "rw"
9381 do_facet $SINGLEMDS mount | grep "lustre.*rw,.*MDT" ||
9382 error_113 "$SINGLEMDS should be read-write"
9384 # Verify OST's should start with "rw"
9385 for (( i=1; i <= OSTCOUNT; i++ )); do
9386 do_facet ost$i mount | grep "lustre.*rw,.*OST" ||
9387 error_113 "ost$i should be read-write"
9390 # rdonly_dev does not currently work for ldiskfs
9391 # We skip the rdonly_dev check until then.
9392 if [[ $ost1_FSTYPE == ldiskfs ]]; then
9393 echo "Shadow Mountpoint correctly reports rw for ldiskfs"
9398 # Only ZFS specific tests below.
9401 # Must stop all (server+client) and restart to verify new
9405 # add rdonly_dev to mount option
9406 MGS_MOUNT_OPTS=$(csa_add "$MGS_MOUNT_OPTS" -o rdonly_dev)
9407 MDS_MOUNT_OPTS=$(csa_add "$MDS_MOUNT_OPTS" -o rdonly_dev)
9408 OST_MOUNT_OPTS=$(csa_add "$OST_MOUNT_OPTS" -o rdonly_dev)
9410 # Only restart server(mds/ost). Sufficient for test
9411 setupall server_only || error "Fail to start servers"
9413 # Verify MDS's should be "ro"
9414 do_facet $SINGLEMDS mount | grep "lustre.*ro,.*MDT.*rdonly_dev" ||
9415 error_113 "$SINGLEMDS should be read-only"
9417 # Verify OST's should be "ro"
9418 for (( i=1; i <= OSTCOUNT; i++ )); do
9419 do_facet ost$i mount | grep "lustre.*ro,.*OST.*rdonly_dev" ||
9420 error_113 "ost$i should be read-only"
9423 run_test 113 "Shadow mountpoint correctly report ro/rw for mounts"
9426 # Purpose: To verify dynamic thread (OSS) creation.
9427 # (This was sanity/115)
9430 [ $PARALLEL == "yes" ] && skip "skip parallel run"
9431 remote_ost_nodsh && skip "remote OST with nodsh"
9433 # Lustre does not stop service threads once they are started.
9434 # Reset number of running threads to default.
9439 local save_params="$TMP/sanity-$TESTNAME.parameters"
9441 # Get ll_ost_io count before I/O
9442 OSTIO_pre=$(do_facet ost1 \
9443 "$LCTL get_param ost.OSS.ost_io.threads_started | cut -d= -f2")
9444 # Exit if lustre is not running (ll_ost_io not running).
9445 [ -z "$OSTIO_pre" ] && error "no OSS threads"
9447 echo "Starting with $OSTIO_pre threads"
9448 local thread_max=$((OSTIO_pre * 2))
9449 local rpc_in_flight=$((thread_max * 2))
9451 # this is limited to OSC_MAX_RIF_MAX (256)
9452 [ $rpc_in_flight -gt 256 ] && rpc_in_flight=256
9453 thread_max=$((rpc_in_flight / 2))
9454 [ $thread_max -le $OSTIO_pre ] && skip "Too many ost_io threads" &&
9457 # Number of I/O Process proposed to be started.
9459 local facets=$(get_facets OST)
9461 save_lustre_params client "osc.*OST*.max_rpcs_in_flight" > $save_params
9462 save_lustre_params $facets "ost.OSS.ost_io.threads_max" >> $save_params
9464 # Set in_flight to $rpc_in_flight
9465 $LCTL set_param osc.*OST*.max_rpcs_in_flight=$rpc_in_flight ||
9466 error "Failed to set max_rpcs_in_flight to $rpc_in_flight"
9467 nfiles=${rpc_in_flight}
9468 # Set ost thread_max to $thread_max
9469 do_facet ost1 "$LCTL set_param ost.OSS.ost_io.threads_max=$thread_max"
9471 # 5 Minutes should be sufficient for max number of OSS
9472 # threads(thread_max) to be created.
9476 local wtl=${WTL:-"$LUSTRE/tests/write_time_limit"}
9478 test_mkdir $DIR/$tdir
9479 for ((i = 1; i <= nfiles; i++)); do
9480 local file=$DIR/$tdir/${tfile}-$i
9482 $LFS setstripe -c -1 -i 0 $file
9483 ($wtl $file $timeout)&
9486 # I/O Started - Wait for thread_started to reach thread_max or report
9487 # error if thread_started is more than thread_max.
9488 echo "Waiting for thread_started to reach thread_max"
9489 local thread_started=0
9490 local end_time=$((SECONDS + timeout))
9492 while [ $SECONDS -le $end_time ] ; do
9494 # Get ost i/o thread_started count.
9495 thread_started=$(do_facet ost1 \
9497 ost.OSS.ost_io.threads_started | cut -d= -f2")
9498 # Break out if thread_started is equal/greater than thread_max
9499 if (( $thread_started >= $thread_max )); then
9500 echo ll_ost_io thread_started $thread_started, \
9501 equal/greater than thread_max $thread_max
9507 # Cleanup - We have the numbers, Kill i/o jobs if running.
9508 jobcount=($(jobs -p))
9510 for ((i=0; i < ${#jobcount[*]}; i++)); do
9511 kill -9 ${jobcount[$i]}
9512 if [ $? -ne 0 ] ; then
9513 echo "warning: cannot kill WTL pid ${jobcount[$i]}"
9517 # Cleanup files left by WTL binary.
9518 for ((i = 1; i <= nfiles; i++)); do
9519 local file=$DIR/$tdir/${tfile}-$i
9522 if [ $? -ne 0 ] ; then
9523 echo "Warning: Failed to delete file $file"
9527 restore_lustre_params <$save_params
9528 rm -f $save_params || echo "Warning: delete file '$save_params' failed"
9530 # Error out if no new thread has started or Thread started is greater
9532 if (( $thread_started <= $OSTIO_pre ||
9533 $thread_started > $thread_max )); then
9534 error "ll_ost_io: thread_started $thread_started" \
9535 "OSTIO_pre $OSTIO_pre, thread_max $thread_max." \
9536 "No new thread started or thread started greater " \
9540 run_test 114 "verify dynamic thread creation===================="
9546 do_facet mds1 rm -f $1
9550 if [ "$mds1_FSTYPE" != ldiskfs ]; then
9551 skip "Only applicable to ldiskfs-based MDTs"
9553 [ -n "$FILESET" ] && skip "Not functional for FILESET set"
9555 local dbfs_ver=$(do_facet mds1 $DEBUGFS -V 2>&1)
9557 echo "debugfs version: $dbfs_ver"
9558 echo "$dbfs_ver" | egrep -w "1.44.3.wc1|1.44.5.wc1|1.45.2.wc1" &&
9559 skip_env "This version of debugfs doesn't show inode number"
9561 local IMAGESIZE=$((3072 << 30)) # 3072 GiB
9571 do_facet mds1 "hostname; $LCTL dl; mount"
9573 do_facet ost1 "hostname; $LCTL dl; mount"
9574 # We need MDT size 3072GB, because it is smallest
9575 # partition that can store 2B inodes
9576 do_facet mds1 "mkdir -p $TMP/$tdir"
9577 local mdsimgname=$TMP/$tdir/lustre-mdt
9579 do_facet mds1 "rm -f $mdsimgname"
9580 do_facet mds1 "touch $mdsimgname"
9581 stack_trap "cleanup_115 $mdsimgname" EXIT
9582 do_facet mds1 "$TRUNCATE $mdsimgname $IMAGESIZE" ||
9583 skip "Backend FS doesn't support sparse files"
9584 local mdsdev=$(do_facet mds1 "losetup -f")
9586 do_facet mds1 "losetup $mdsdev $mdsimgname"
9588 local mds_opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
9589 --mkfsoptions='-O ea_inode,^resize_inode,meta_bg \
9590 -N 2247484000 -E lazy_itable_init' --device-size=$IMAGESIZE"
9591 add mds1 $mds_opts --mgs --reformat $mdsdev || {
9592 do_facet $SINGLEMDS \
9593 "losetup -d $mdsdev && rm -f $mdsimgname"
9594 skip_env "format large MDT failed"
9597 local ostdev=$(ostdevname 1)
9599 local opts="$(mkfs_opts ost1 $ostdev) \
9600 --reformat $ostdev $ostdev"
9601 add ost1 $opts || error "add ost1 failed with new params"
9602 start mds1 $mdsdev $MDS_MOUNT_OPTS || error "start MDS failed"
9603 start_ost || error "start OSS failed"
9604 mount_client $MOUNT || error "mount client failed"
9606 mkdir -p $DIR/$tdir || error "mkdir $DIR/$tdir fail"
9607 local goal="/sys/fs/ldiskfs/$(basename $mdsdev)/inode_goal"
9609 # 2147483648 is 0x80000000
9610 do_facet mds1 "echo 2147483648 >> $goal; grep . $goal"
9611 touch $DIR/$tdir/$tfile
9613 # attrs from 1 to 16 go to block, 17th - to inode
9616 for i in {1..17}; do
9617 local nm="trusted.ea$i"
9618 setfattr -n $nm -v $(printf "xattr%0250d" $i) $DIR/$tdir/$tfile
9621 do_facet mds1 "$DEBUGFS -c -R 'stat ROOT/$tdir/$tfile' $mdsdev"
9623 # inode <2147483649> trusted.ea16 (255)
9624 local inode_num=$(do_facet mds1 \
9625 "$DEBUGFS -c -R 'stat ROOT/$tdir/$tfile' $mdsdev" |
9626 awk '/ea17/ { print $2 }' |
9627 sed -e 's/>//' -e 's/<//' -e 's/\"//')
9628 echo "inode num: $inode_num"
9629 [ $inode_num -ge 2147483648 ] || error "inode $inode_num too small"
9630 do_facet mds1 "losetup -d $mdsdev"
9632 run_test 115 "Access large xattr with inodes number over 2TB"
9635 [ "$mds1_FSTYPE" != ldiskfs ] && skip "ldiskfs only test"
9636 [ "$MDS1_VERSION" -lt $(version_code 2.10.59) ] &&
9637 skip "Need server version at least 2.10.59"
9638 do_facet $SINGLEMDS which mkfs.xfs ||
9639 skip_env "No mkfs.xfs installed"
9644 local tmpmnt=$TMP/$tdir
9645 local mdtimg=$tfile-mdt0
9647 do_facet $SINGLEMDS mkdir -p $tmpmnt
9648 stack_trap "do_facet $SINGLEMDS rmdir $tmpmnt" EXIT
9650 do_facet $SINGLEMDS touch $TMP/$mdtimg
9651 stack_trap "do_facet $SINGLEMDS rm -f $TMP/$mdtimg" EXIT
9652 do_facet $SINGLEMDS mkfs -t xfs -d file,size=1t,name=$TMP/$mdtimg ||
9653 error "mkfs temporary xfs image"
9655 do_facet $SINGLEMDS mount $TMP/$mdtimg $tmpmnt ||
9656 error "mount temporary xfs image"
9657 stack_trap "do_facet $SINGLEMDS umount $tmpmnt" EXIT
9658 local old_mdssize=$MDSSIZE
9659 local old_mdsisize=$MDSISIZE
9661 MDSSIZE=$((17 * 1024 * 1024 * 1024)) # 17T MDT
9662 MDSISIZE=$((16 << 20))
9663 local opts17t="$(mkfs_opts $SINGLEMDS)"
9665 MDSSIZE=$old_mdssize
9666 MDSISIZE=$old_mdsisize
9668 do_facet $SINGLEMDS $MKFS \
9669 ${opts17t/-E /-E lazy_itable_init,lazy_journal_init,} \
9671 error "failed to mkfs for $tmpmnt/$mdtimg"
9673 do_facet $SINGLEMDS $TUNE2FS -l $tmpmnt/$mdtimg |
9674 grep -qw 'features.*extent' || error "extent should be enabled"
9677 run_test 116 "big size MDT support"
9680 # Call setup only if LustreFS is not mounted
9681 check_mount || setup
9683 do_facet ost1 "$LCTL set_param ost.OSS.ost_io.nrs_policies=fifo"
9684 do_facet ost1 "$LCTL get_param -n ost.OSS.ost_io.nrs_tbf_rule" &&
9685 error "get_param should fail"
9686 cleanup || error "cleanup failed with rc $?"
9688 run_test 117 "lctl get_param return errors properly"
9694 [[ "$MDSCOUNT" -ge 2 ]] || skip "Need more at least 2 MDTs"
9696 had_config=$(do_facet mds1 "$LCTL get_param debug | grep config")
9697 do_facet mds1 "$LCTL set_param debug=+config"
9698 do_facet mds1 "$LCTL clear"
9701 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)" &>/dev/null
9702 # mount after writeconf will make "add osp" added to mdt0 config:
9703 # 53 (224)marker 60 (flags=0x01, v2.5.1.0) lustre-MDT0001 'add osp'
9704 # 54 (080)add_uuid nid=... 0: 1:...
9705 # 55 (144)attach 0:lustre-MDT0001-osp-MDT0000 1:osp 2:...
9706 # 56 (144)setup 0:lustre-MDT0001-osp-MDT0000 1:... 2:...
9707 # 57 (136)modify_mdc_tgts add 0:lustre-MDT0000-mdtlov 1:... 2:1 3:1
9708 # duplicate modify_mdc_tgts caused crashes
9711 # using larger debug_mb size to avoid lctl dk log truncation
9712 size_mb=$((DEBUG_SIZE_SAVED * 4))
9715 # though config processing stops after failed attach and setup
9716 # it will proceed after the failed command after each writeconf
9717 # this is the original scenario of the issue
9718 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)" &>/dev/null
9719 do_facet mds1 "$LCTL set_param debug_mb=$size_mb"
9722 wait_update_facet_cond mds1 \
9723 "$LCTL dk | grep -c Processed.log.$FSNAME-MDT0000" \
9728 [[ -z "$had_config" ]] && do_facet mds1 lctl set_param debug=-config
9732 run_test 119 "writeconf on slave mdt shouldn't duplicate mdc/osp and crash"
9734 test_120() { # LU-11130
9735 [ "$MDSCOUNT" -lt 2 ] && skip "mdt count < 2"
9736 [ "$mds1_FSTYPE" != ldiskfs ] &&
9737 skip "ldiskfs only test"
9738 [ "$MDS1_VERSION" -lt $(version_code 2.11.56) ] &&
9739 skip "Need DNE2 capable MD target with LU-11130 fix"
9743 local mds1host=$(facet_active_host mds1)
9744 local mds1dev=$(mdsdevname 1)
9746 $LFS mkdir -i 1 $DIR/$tdir
9747 $LFS mkdir -i 0 $DIR/$tdir/mds1dir
9749 ln -s foo $DIR/$tdir/bar
9750 mv $DIR/$tdir/bar $DIR/$tdir/mds1dir/bar2 ||
9751 error "cross-target rename failed"
9755 run_e2fsck $mds1host $mds1dev "-n"
9757 run_test 120 "cross-target rename should not create bad symlinks"
9761 start_mgsmds || error "MGS MDS Start failed"
9763 stop_mds || error "Stopping MDSes failed"
9767 stop_mds || error "Stopping MDSes failed"
9769 run_test 121 "failover MGS"
9772 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
9773 [[ "$OST1_VERSION" -ge $(version_code 2.11.53) ]] ||
9774 skip "Need OST version at least 2.11.53"
9777 LOAD_MODULES_REMOTE=true load_modules
9778 #define OBD_FAIL_OFD_SET_OID 0x1e0
9779 do_facet ost1 $LCTL set_param fail_loc=0x00001e0
9782 $LFS mkdir -i1 -c1 $DIR/$tdir
9783 $LFS setstripe -i0 -c1 $DIR/$tdir
9784 do_facet ost1 $LCTL set_param fail_loc=0
9785 createmany -o $DIR/$tdir/file_ 1000 ||
9786 error "Fail to create a new sequence"
9790 run_test 122a "Check OST sequence update"
9793 remote_mgs_nodsh && skip "remote MGS with nodsh"
9794 [ -d $MOUNT/.lustre ] || setup
9796 # test old logid format until removal from llog_ioctl.c::str2logid()
9797 if (( $MGS_VERSION < $(version_code 2.99.53) )); then
9798 do_facet mgs $LCTL dl | grep MGS
9799 do_facet mgs "$LCTL --device %MGS llog_print \
9800 \\\\\\\$$FSNAME-client 1 10" ||
9801 error "old llog_print failed"
9804 # test new logid format
9805 if [ $MGS_VERSION -ge $(version_code 2.9.53) ]; then
9806 do_facet mgs "$LCTL --device MGS llog_print $FSNAME-client" ||
9807 error "new llog_print failed"
9810 run_test 123aa "llog_print works with FIDs and simple names"
9813 remote_mgs_nodsh && skip "remote MGS with nodsh"
9814 [[ $MGS_VERSION -gt $(version_code 2.11.51) ]] ||
9815 skip "Need server with working llog_print support"
9817 [ -d $MOUNT/.lustre ] || setup
9823 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9824 mgs_arg="--device MGS"
9826 orig_val=$(do_facet mgs $LCTL get_param jobid_name)
9827 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME"
9829 yaml=$(do_facet mgs $LCTL $mgs_arg llog_print params |
9830 grep jobid_name | tail -n 1)
9832 local param=$(awk '{ print $10 }' <<< "$yaml")
9833 local val=$(awk '{ print $12 }' <<< "$yaml")
9834 #return to the default
9835 do_facet mgs $LCTL set_param -P jobid_name=$orig_val
9836 [ $val = "TESTNAME" ] || error "bad value: $val"
9837 [ $param = "jobid_name," ] || error "Bad param: $param"
9839 run_test 123ab "llog_print params output values from set_param -P"
9841 test_123ac() { # LU-11566
9842 remote_mgs_nodsh && skip "remote MGS with nodsh"
9843 do_facet mgs "$LCTL help llog_print" 2>&1 | grep -q -- --start ||
9844 skip "Need 'lctl llog_print --start' on MGS"
9850 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9851 mgs_arg="--device MGS"
9853 [ -d $MOUNT/.lustre ] || setup
9855 # - { index: 10, event: add_uuid, nid: 192.168.20.1@tcp(0x20000c0a81401,
9856 # node: 192.168.20.1@tcp }
9857 do_facet mgs $LCTL $mgs_arg \
9858 llog_print --start $start --end $end $FSNAME-client | tr -d , |
9859 while read DASH BRACE INDEX idx EVENT BLAH BLAH BLAH; do
9860 (( idx >= start )) || error "llog_print index $idx < $start"
9861 (( idx <= end )) || error "llog_print index $idx > $end"
9864 run_test 123ac "llog_print with --start and --end"
9866 test_123ad() { # LU-11566
9867 remote_mgs_nodsh && skip "remote MGS with nodsh"
9868 # older versions of lctl may not print all records properly
9869 do_facet mgs "$LCTL help llog_print" 2>&1 | grep -q -- --start ||
9870 skip "Need 'lctl llog_print --start' on MGS"
9872 [ -d $MOUNT/.lustre ] || setup
9874 # append a new record, to avoid issues if last record was cancelled
9875 local old=$($LCTL get_param -n osc.*-OST0000-*.max_dirty_mb | head -1)
9876 do_facet mgs $LCTL conf_param $FSNAME-OST0000.osc.max_dirty_mb=$old
9878 # logid: [0x3:0xa:0x0]:0
9882 local num=$(do_facet mgs $LCTL --device MGS llog_info $FSNAME-client |
9883 awk '/last_index:/ { print $2 - 1 }')
9885 # - { index: 71, event: set_timeout, num: 0x14, param: sys.timeout=20 }
9886 local last=$(do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
9887 tail -1 | awk '{ print $4 }' | tr -d , )
9888 (( last == num )) || error "llog_print only showed $last/$num records"
9890 run_test 123ad "llog_print shows all records"
9892 test_123ae() { # LU-11566
9900 remote_mgs_nodsh && skip "remote MGS with nodsh"
9901 [ -d $MOUNT/.lustre ] || setupall
9903 max=$($LCTL get_param -n osc.*-OST0000-*.max_dirty_mb | head -1)
9904 pgs=$($LCTL get_param -n osc.*-OST0000-*.max_pages_per_rpc | head -1)
9905 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9906 mgs_arg="--device MGS"
9908 if do_facet mgs "$LCTL help llog_cancel" 2>&1| grep -q -- --log_id; then
9909 # save one set_param -P record in case none exist
9911 do_facet mgs $LCTL set_param -P osc.*.max_pages_per_rpc=$pgs
9912 stack_trap "do_facet mgs $LCTL set_param -P -d \
9913 osc.*.max_pages_per_rpc"
9916 orig=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9917 tail -1 | awk '{ print $4 }' | tr -d , )
9918 do_facet mgs $LCTL set_param -P osc.*.max_dirty_mb=$max
9919 do_facet mgs $LCTL $mgs_arg llog_print $log | tail -1 |
9920 grep "parameter: osc.*.max_dirty_mb" ||
9921 error "new set_param -P wasn't stored in params log"
9923 # - { index: 71, event: set_param, device: general,
9924 # param: osc.*.max_dirty_mb, value: 256 }
9925 id=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9926 tail -1 | awk '{ print $4 }' | tr -d , )
9927 do_facet mgs $LCTL $mgs_arg llog_cancel $log --log_idx=$id
9928 local new=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9929 tail -1 | awk '{ print $4 }' | tr -d , )
9930 (( new == orig )) ||
9931 error "new llog_cancel now $new, not at $orig records"
9934 # test old positional parameters for a while still
9935 if [ "$MGS_VERSION" -le $(version_code 3.1.53) ]; then
9938 do_facet mgs $LCTL conf_param \
9939 $FSNAME-OST0000.osc.max_pages_per_rpc=$pgs
9940 stack_trap "do_facet mgs $LCTL conf_param -d \
9941 $FSNAME-OST0000.osc.max_pages_per_rpc"
9943 orig=$(do_facet mgs $LCTL --device MGS llog_print $log |
9944 tail -1 | awk '{ print $4 }' | tr -d , )
9945 do_facet mgs $LCTL conf_param $FSNAME-OST0000.osc.max_dirty_mb=$max
9946 do_facet mgs $LCTL --device MGS llog_print $log |
9947 tail -1 | grep "parameter: osc.max_dirty_mb" ||
9948 error "old conf_param wasn't stored in params log"
9949 do_facet mgs $LCTL --device MGS llog_print $log
9950 # - { index: 71, event: conf_param, device: testfs-OST0000-osc,
9951 # param: osc.max_dirty_mb=256 }
9952 id=$(do_facet mgs $LCTL --device MGS llog_print $log |
9953 tail -1 | awk '{ print $4 }' | tr -d , )
9954 do_facet mgs $LCTL --device MGS llog_cancel $log $id
9955 do_facet mgs $LCTL --device MGS llog_print $log
9956 new=$(do_facet mgs $LCTL --device MGS llog_print $log |
9957 tail -1 | awk '{ print $4 }' | tr -d , )
9958 (( new == orig )) ||
9959 error "old llog_cancel now $new, not at $orig records"
9962 run_test 123ae "llog_cancel can cancel requested record"
9964 test_123af() { #LU-13609
9965 [ "$MGS_VERSION" -ge $(version_code 2.13.54) -a \
9966 "$MDS1_VERSION" -ge $(version_code 2.13.54) ] ||
9967 skip "Need both MGS and MDS version at least 2.13.54"
9969 [ -d $MOUNT/.lustre ] || setupall
9970 stack_trap "do_facet mds1 $LCTL set_param fail_loc=0" EXIT
9980 for device in "MGS" "$FSNAME-MDT0000"; do
9981 cmd="--device $device llog_catlist"
9982 echo "lctl $cmd ..."
9983 if [ "$device" = "MGS" ]; then
9988 orig_clist=($(do_facet $facet $LCTL $cmd | awk '{ print $2 }'))
9989 orig_count=${#orig_clist[@]}
9990 echo "orig_clist: ${orig_clist[*]}"
9992 #define OBD_FAIL_CATLIST 0x131b
9993 #fetch to llog records from the second one
9994 do_facet $facet $LCTL set_param fail_loc=0x131b fail_val=2
9996 new_clist=($(do_facet $facet $LCTL $cmd | awk '{ print $2 }'))
9997 new_count=${#new_clist[@]}
9998 echo "new_clist: ${new_clist[*]}"
10000 [ $new_count -eq $((orig_count - 1)) ] ||
10001 error "$new_count != $orig_count - 1"
10002 for i in $(seq 0 $new_count); do
10004 [ "${orig_clist[$j]}" = "${new_clist[$i]}" ] ||
10005 error "${orig_clist[$j]} != ${new_clist[$i]}"
10007 do_facet mds1 $LCTL set_param fail_loc=0
10011 run_test 123af "llog_catlist can show all config files correctly"
10013 test_123ag() { # LU-15142
10017 remote_mgs_nodsh && skip "remote MGS with nodsh"
10018 (( $MGS_VERSION >= $(version_code 2.14.55) )) ||
10019 skip "Need server version least 2.14.55"
10021 [ -d $MOUNT/.lustre ] || setup
10023 orig_val=$(do_facet mgs $LCTL get_param jobid_name)
10024 stack_trap "do_facet mgs $LCTL set_param -P jobid_name=$orig_val"
10026 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME1"
10027 do_facet mgs $LCTL set_param -P -d jobid_name
10028 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
10029 grep -c jobid_name)
10030 (( rec == 0 )) || error "parameter was not deleted, check #1"
10031 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME1"
10032 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
10033 grep -c jobid_name)
10034 (( rec == 1)) || error "parameter is not set"
10035 # usage with ordinary set_param format works too
10036 do_facet mgs $LCTL set_param -P -d jobid_name="ANY"
10037 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
10038 grep -c jobid_name)
10039 (( rec == 0 )) || error "parameter was not deleted, check #2"
10041 run_test 123ag "llog_print skips values deleted by set_param -P -d"
10043 test_123ah() { #LU-7668 del_ost
10044 (( "$MGS_VERSION" >= $(version_code 2.15.51) &&
10045 "$MDS1_VERSION" >= $(version_code 2.15.51) )) ||
10046 skip "Need MGS/MDS version >= 2.15.51 for del_ost"
10048 [ -d $MOUNT/.lustre ] || setupall
10049 stack_trap "do_facet mds1 $LCTL set_param fail_loc=0" EXIT
10051 local cmd="--device MGS llog_print"
10053 cli_llogcnt_orig=$(do_facet mgs $LCTL $cmd $FSNAME-client |
10054 grep -c $FSNAME-OST0000)
10055 mds1_llogcnt_orig=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 |
10056 grep -c $FSNAME-OST0000)
10058 [ $cli_llogcnt_orig -gt 0 ] ||
10059 error "$FSNAME-OST0000 not found (client)"
10060 [ $mds1_llogcnt_orig -gt 0 ] || error "$FSNAME-OST0000 not found (MDT)"
10062 # -n/--dryrun should NOT modify catalog
10063 do_facet mgs "$LCTL del_ost -n -t $FSNAME-OST0000" ||
10064 error "del_ost --dryrun failed with $?"
10066 local cli_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-client |
10067 grep -c $FSNAME-OST0000)
10068 local mds1_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 |
10069 grep -c $FSNAME-OST0000)
10071 [ $cli_llogcnt -eq $cli_llogcnt_orig ] ||
10072 error "--dryrun error: $cli_llogcnt != $cli_llogcnt_orig"
10073 [ $mds1_llogcnt -eq $mds1_llogcnt_orig ] ||
10074 error "--dryrun error: $mds1_llogcnt != $mds1_llogcnt_orig"
10077 do_facet mgs "$LCTL del_ost --target $FSNAME-OST0000" ||
10078 error "del_ost failed with $?"
10080 local cli_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-client |
10081 grep -c $FSNAME-OST0000)
10082 local mds1_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 |
10083 grep -c $FSNAME-OST0000)
10085 # every catalog entry for OST0000 should have been cancelled
10086 [ $cli_llogcnt -eq 0 ] || error "$FSNAME-OST0000 not cancelled (cli)"
10087 [ $mds1_llogcnt -eq 0 ] || error "$FSNAME-OST0000 not cancelled (MDT)"
10089 umount_client $MOUNT
10090 mount_client $MOUNT
10092 $LFS df | grep -q OST0000 && error "del_ost did not remove OST0000!"
10094 reformat_and_config
10096 run_test 123ah "del_ost cancels config log entries correctly"
10101 echo "cleanup test 123ai"
10103 # cancel last timeout record
10104 do_facet mgs "$LCTL set_param -P -d timeout"
10106 # restore timeout value
10107 do_nodes $(comma_list $(all_nodes)) "$LCTL set_param timeout=$timeout"
10110 test_123ai() { #LU-16167
10116 remote_mgs_nodsh && skip "remote MGS with nodsh"
10117 (( MDS1_VERSION >= $(version_code 2.15.51) )) ||
10118 skip "Need MDS version at least 2.15.51"
10120 [ -d $MOUNT/.lustre ] || setup
10122 old_count=$(do_facet mgs "$LCTL --device MGS llog_print -r params" |
10123 grep -c "parameter: timeout")
10125 old_timeout=$($LCTL get_param -n timeout)
10126 stack_trap "cleanup_123ai $old_timeout" EXIT
10128 # add and cancel params (2 * MAX_IOC_BUFLEN / 128 + 1 = 129)
10129 for i in {1..129}; do
10130 do_facet mgs "$LCTL set_param -P timeout=$i" ||
10131 error "fail to set timeout=$i on MGS"
10134 count=$(do_facet mgs "$LCTL --device MGS llog_print -r params" |
10135 grep -c "parameter: timeout")
10137 ((count - old_count == 129)) ||
10138 error "number timeout record missmatch ($count - $old_count != 129)"
10140 do_facet mgs "$LCTL --device MGS llog_print params" |
10141 tail -1 | grep "timeout, value: 129" ||
10142 error "llog_print could not display the last record (timeout=129)"
10145 run_test 123ai "llog_print display all non skipped records"
10148 remote_mgs_nodsh && skip "remote MGS with nodsh"
10150 [ -d $MOUNT/.lustre ] || setup
10151 yaml_file="$TMP/$tfile.yaml"
10153 do_facet mgs rm "$yaml_file"
10154 cfgfiles=$(do_facet mgs "lctl --device MGS llog_catlist" |
10155 sed 's/config_log://')
10157 # set jobid_var to a different value for test
10158 orig_val=$(do_facet mgs $LCTL get_param jobid_var)
10160 do_facet mgs $LCTL set_param -P jobid_var="TESTNAME"
10162 for i in $cfgfiles params; do
10163 do_facet mgs "lctl --device MGS llog_print ${i} >> $yaml_file"
10166 echo "Unmounting FS"
10174 test_123_restore() {
10175 set_val=$(do_facet mgs $LCTL get_param jobid_var)
10177 do_facet mgs $LCTL set_param -P $orig_val
10179 [ $set_val == "jobid_var=TESTNAME" ] ||
10180 error "$set_val is not TESTNAME"
10182 do_facet mgs rm "$yaml_file"
10194 # Reapply the config from before
10195 echo "Setting configuration parameters"
10196 do_facet mgs "lctl set_param -F $yaml_file"
10200 run_test 123F "clear and reset all parameters using set_param -F"
10210 # Reapply the config from before
10211 echo "Setting configuration parameters"
10212 do_facet mgs "lctl apply_yaml $yaml_file"
10216 run_test 123G "clear and reset all parameters using apply_yaml"
10221 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
10222 [ -z $mds2failover_HOST ] && skip "needs MDT failover setup"
10225 do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
10226 grep 1.2.3.4@tcp && error "Should not be fake nid"
10227 do_facet mgs $LCTL conf_param $FSNAME-MDT0001.failover.node=1.2.3.4@tcp\
10228 || error "Set params error"
10229 do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
10230 grep 1.2.3.4@tcp || error "Fake nid should be added"
10234 if combined_mgs_mds; then
10235 start_mdt 1 "-o nosvc" ||
10236 error "starting mds with nosvc option failed"
10238 local nid=$(do_facet mds2 $LCTL list_nids | head -1)
10239 local failover_nid=$(do_node $mds2failover_HOST $LCTL list_nids | head -1)
10241 do_facet mgs $LCTL replace_nids $FSNAME-MDT0001 $nid:$failover_nid ||
10242 error "replace_nids execution error"
10244 if combined_mgs_mds; then
10250 echo "lfs setdirstripe"
10251 $LFS setdirstripe -i 1 $MOUNT/$tdir || error "setdirstirpe error"
10254 run_test 124 "check failover after replace_nids"
10256 get_max_sectors_kb() {
10259 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
10260 local max_sectors_path="/sys/block/${dev_base}/queue/max_sectors_kb"
10262 do_facet ${facet} "[[ -e ${max_sectors_path} ]] &&
10263 cat ${max_sectors_path}"
10266 get_max_hw_sectors_kb() {
10269 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
10270 local max_hw_path="/sys/block/${dev_base}/queue/max_hw_sectors_kb"
10272 do_facet ${facet} "[[ -e ${max_hw_path} ]] && cat ${max_hw_path}"
10275 set_max_sectors_kb() {
10279 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
10280 local max_sectors_path="/sys/block/${dev_base}/queue/max_sectors_kb"
10282 do_facet ${facet} "[[ -e ${max_sectors_path} ]] &&
10283 echo ${value} > ${max_sectors_path}"
10286 [[ $rc -ne 0 ]] && echo "Failed to set ${max_sectors_path} to ${value}"
10291 # Return 0 if all slave devices have max_sectors_kb == max_hw_sectors_kb
10292 # Otherwise return > 0
10293 check_slaves_max_sectors_kb()
10297 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
10298 local slaves_dir=/sys/block/${dev_base}/slaves
10299 local slave_devices=$(do_facet ${facet} "ls ${slaves_dir} 2>/dev/null")
10300 [[ -z ${slave_devices} ]] && return 0
10302 local slave max_sectors new_max_sectors max_hw_sectors path
10305 for slave in ${slave_devices}; do
10306 path="/dev/${slave}"
10307 ! is_blkdev ${facet} ${path} && continue
10308 max_sectors=$(get_max_sectors_kb ${facet} ${path})
10309 max_hw_sectors=$(get_max_hw_sectors_kb ${facet} ${path})
10310 new_max_sectors=${max_hw_sectors}
10311 [[ ${new_max_sectors} -gt ${RQ_SIZE_LIMIT} ]] &&
10312 new_max_sectors=${RQ_SIZE_LIMIT}
10314 if [[ ${max_sectors} -ne ${new_max_sectors} ]]; then
10315 echo "${path} ${max_sectors} ${new_max_sectors}"
10318 check_slaves_max_sectors_kb ${facet} ${path}
10327 local facet_list="mgs mds1 ost1"
10328 combined_mgs_mds && facet_list="mgs ost1"
10331 for facet in ${facet_list}; do
10332 [[ $(facet_fstype ${facet}) != ldiskfs ]] &&
10333 skip "ldiskfs only test" &&
10335 ! is_blkdev ${facet} $(facet_device ${facet}) &&
10336 skip "requires all real devices" &&
10341 # We don't increase IO request size limit past 16MB. See comments in
10342 # lustre/utils/libmount_utils_ldiskfs.c:tune_max_sectors_kb()
10343 RQ_SIZE_LIMIT=$((16 * 1024))
10344 local device old_max_sectors new_max_sectors max_hw_sectors
10345 for facet in ${facet_list}; do
10346 device=$(facet_device ${facet})
10347 old_max_sectors=$(get_max_sectors_kb ${facet} ${device})
10348 max_hw_sectors=$(get_max_hw_sectors_kb ${facet} ${device})
10350 # The expected value after l_tunedisk is executed
10351 new_max_sectors=$old_max_sectors
10352 [[ ${new_max_sectors_kb} -gt ${RQ_SIZE_LIMIT} ]] &&
10353 new_max_sectors_kb=${RQ_SIZE_LIMIT}
10355 # Ensure the current value of max_sectors_kb does not equal
10356 # max_hw_sectors_kb, so we can tell whether l_tunedisk did
10358 set_max_sectors_kb ${facet} ${device} $((new_max_sectors - 1))
10360 # Value before l_tunedisk
10361 local pre_max_sectors=$(get_max_sectors_kb ${facet} ${device})
10362 if [[ ${pre_max_sectors} -ne $((new_max_sectors - 1)) ]]; then
10363 echo "unable to satsify test pre-condition:"
10364 echo "${pre_max_sectors} != $((new_max_sectors - 1))"
10369 echo "Before: ${facet} ${device} ${pre_max_sectors} ${max_hw_sectors}"
10371 do_facet ${facet} "l_tunedisk ${device}"
10373 # Value after l_tunedisk
10374 local post_max_sectors=$(get_max_sectors_kb ${facet} ${device})
10376 echo "After: ${facet} ${device} ${post_max_sectors} ${max_hw_sectors}"
10378 if [[ ${facet} != ost1 ]]; then
10379 if [[ ${post_max_sectors} -ne ${pre_max_sectors} ]]; then
10380 echo "l_tunedisk modified max_sectors_kb of ${facet}"
10384 set_max_sectors_kb ${facet} ${device} ${old_max_sectors}
10386 if [[ ${post_max_sectors} -eq ${pre_max_sectors} ]]; then
10387 echo "l_tunedisk failed to modify max_sectors_kb of ${facet}"
10391 check_slaves_max_sectors_kb ${facet} ${device} ||
10398 run_test 125 "check l_tunedisk only tunes OSTs and their slave devices"
10401 [[ "$MDS1_VERSION" -ge $(version_code 2.13.52) ]] ||
10402 skip "Need MDS version at least 2.13.52"
10405 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ../libcfs/libcfs/libcfs
10406 #define OBD_FAIL_OBD_SETUP 0x60d
10407 do_facet mds1 $LCTL set_param fail_loc=0x60d
10408 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_modules &
10409 for i in {1..40}; do
10410 do_facet mds1 lsmod | grep -q osd_$mds1_FSTYPE && break
10413 clear_failloc $SINGLEMDS 20 &
10414 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
10416 run_test 126 "mount in parallel shouldn't cause a crash"
10419 [[ "$ost1_FSTYPE" == ldiskfs ]] || skip "ldiskfs only test"
10423 zconf_umount_clients $RCLIENTS $MOUNT
10425 wait_osp_active ost ${FSNAME}-OST0000 0 1
10426 local osc_tgt="$FSNAME-OST0000-osc-$($LFS getname -i $DIR)"
10427 local avail1=($($LCTL get_param -n osc.${osc_tgt}.kbytesavail))
10429 $LFS setstripe -i 0 $DIR/$tfile || error "failed creating $DIR/$tfile"
10430 dd if=/dev/zero of=$DIR/$tfile bs=1M oflag=direct || true
10432 local avail2=($($LCTL get_param -n osc.${osc_tgt}.kbytesavail))
10434 if ((avail2 * 100 / avail1 > 1)); then
10437 error "more than 1% space left: before=$avail1 after=$avail2"
10440 local mbs=$(($(stat -c %s $DIR/$tfile) / (1024 * 1024)))
10442 dd if=/dev/zero of=$DIR/$tfile bs=1M count=$mbs conv=notrunc \
10443 oflag=direct || error "overwrite failed"
10445 run_test 127 "direct io overwrite on full ost"
10449 combined_mgs_mds && skip "need separate mgs device"
10450 [ "$ost2_FSTYPE" == zfs ] && import_zpool ost2
10453 # Try to apply nolocallogs to the virgin OST. Should fail.
10454 do_facet ost2 "$TUNEFS --nolocallogs $(ostdevname 2)" &&
10455 error "nolocallogs should not be allowed on the virgin target"
10460 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
10461 # Start OST without MGS (local configs)
10462 do_facet ost1 "$TUNEFS --dryrun $(ostdevname 1)"
10463 start_ost || error "unable to start OST1"
10464 stop_ost || error "Unable to stop OST1"
10466 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
10467 # Do not allow reading local configs, should fail
10468 do_facet ost1 "$TUNEFS --nolocallogs $(ostdevname 1)" ||
10469 error "Can not set nolocallogs"
10470 start_ost && error "OST1 started, but should fail"
10472 # Connect to MGS successfully, reset nolocallogs flag
10473 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
10474 start_mgs || error "unable to start MGS"
10475 start_ost || error "unable to start OST1"
10477 do_facet ost1 "$TUNEFS --dryrun $(ostdevname 1)" | grep "nolocallogs" &&
10478 error "nolocallogs expected to be reset"
10480 stop_ost || error "Unable to stop OST1"
10482 run_test 128 "Force using remote logs with --nolocallogs"
10486 (( MDS1_VERSION >= $(version_code 2.14.57) )) ||
10487 skip "Need MDS version at least 2.14.57"
10489 start_mgsmds || error "MGS/MDS start failed"
10491 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &&
10492 error "start ost1 should fail" || true
10493 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &&
10494 error "second start ost1 should fail" || true
10495 do_facet ost1 "$TUNEFS --writeconf $(ostdevname 1)"
10496 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS ||
10497 error "start ost1 failed"
10501 run_test 129 "attempt to connect an OST with the same index should fail"
10505 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
10507 stop_mdt 2 || error "mdt2 stop failed"
10508 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)"
10509 start_mdt 2 || error "mdt2 start failed"
10510 do_facet mds2 "$LCTL dl" | grep MDT0001-osp-MDT0001 &&
10511 error "Illegal OSP device created" || true
10513 run_test 130 "re-register an MDT after writeconf"
10516 [ "$mds1_FSTYPE" == "ldiskfs" ] || skip "ldiskfs only test"
10517 (( $MDS1_VERSION >= $(version_code 2.14.56.35) )) ||
10518 skip "Need MDS version at least 2.14.56.35"
10519 do_facet mds1 $DEBUGFS -R features $(mdsdevname 1) |
10520 grep -q project || skip "skip project quota not supported"
10525 test_mkdir -c $MDSCOUNT -p $DIR/$tdir
10526 $LFS project -p 1000 $DIR/$tdir || error "set dir project id failed"
10527 createmany -o $DIR/$tdir/f 512
10528 for ((i = 0; i < 512; ++i)); do
10529 $LFS project -p $i $DIR/$tdir/f${i} ||
10530 error "set f${i} project id failed"
10533 test_mkdir -c $MDSCOUNT -p $DIR/$tdir.inherit
10534 $LFS project -p 1001 -s $DIR/$tdir.inherit
10535 createmany -o $DIR/$tdir.inherit/f 128
10536 (( $($LFS project $DIR/$tdir.inherit/f* |
10537 awk '$1 == 1001 { print }' | wc -l) == 128 )) ||
10538 error "files did not inherit projid 1001"
10542 for i in $(seq $MDSCOUNT); do
10543 mds_backup_restore mds$i ||
10544 error "Backup/restore on mds$i failed"
10549 projid=($($LFS project -d $DIR/$tdir))
10550 [ ${projid[0]} == "1000" ] ||
10551 error "projid expected 1000 not ${projid[0]}"
10552 for ((i = 0; i < 512; ++i)); do
10553 projid=($($LFS project $DIR/$tdir/f${i}))
10554 [ ${projid[0]} == "$i" ] ||
10555 error "projid expected $i not ${projid[0]}"
10558 (( $($LFS project $DIR/$tdir.inherit/f* |
10559 awk '$1 == 1001 { print }' | wc -l) == 128 )) ||
10560 error "restore did not copy projid 1001"
10562 run_test 131 "MDT backup restore with project ID"
10568 (( MDS1_VERSION >= $(version_code 2.14.57) )) ||
10569 skip "Need MDS version at least 2.14.57"
10571 combined_mgs_mds || start_mgs || error "unable to start MGS"
10572 start_mdt 1 || error "unable to start mdt1"
10574 err_cnt=$(do_facet mds1 dmesg | grep -c "cannot take the layout locks")
10575 stop_mdt 1 || error "stop mdt1 failed"
10577 [ "$mds1_FSTYPE" == zfs ] && import_zpool mds1
10578 do_facet mds1 $TUNEFS --param mdt.hsm_control=enabled $(mdsdevname 1) ||
10579 error "tunefs failed"
10580 start_mdt 1 || error "cannot start mdt1"
10582 err_cnt2=$(do_facet mds1 dmesg | grep -c "cannot take the layout locks")
10583 [ $err_cnt -eq $err_cnt2 ] || error "Can not take the layout lock"
10584 stop_mdt 1 || error "stop mdt1 failed"
10586 run_test 132 "hsm_actions processed after failover"
10588 # This test verifies we do RR allocation within a pool even if there is a
10589 # significant imbalance vs an OST outside the pool
10591 [[ $OSTCOUNT -lt 4 ]] && skip_env "needs >= 4 OSTs"
10592 [[ "$OST1_VERSION" -ge $(version_code 2.15.51) ]] ||
10593 skip "Need OST version at least 2.15.51"
10594 # This is the easiest way to ensure OSTs start out balanced
10595 reformat_and_config
10598 check_set_fallocate_or_skip
10600 local testfile=$DIR/$tdir/$tfile
10601 local pool="testpool"
10602 local ostrange=$((OSTCOUNT - 1))
10603 # Select all but the last OST to add to the pool
10604 local poolostrange=$((OSTCOUNT - 2))
10612 [ $MINV -eq 0 ] && error "no free space in OST$MINI"
10613 [ $MAXV -gt $((2 * $MINV)) ] &&
10614 error "OSTs badly unbalanced after reformat"
10616 create_pool $FSNAME.$pool || error "failed to create a pool"
10617 do_facet mgs $LCTL pool_add $FSNAME.$pool OST[0-$poolostrange] ||
10618 error "failed to add OST[0-$poolostrange] to the pool"
10620 test_mkdir -p $DIR/$tdir || error "failed to mkdir $DIR/$tdir"
10621 # Consume space on the OSTs in the pool so they are unbalanced with the
10622 # OST outside of the pool
10623 # fill each OST 90% with fallocate so they are widely
10625 local size=$(((MINV * 9 / 10) * 1024))
10626 for ((i = 0; i <= poolostrange; i++)); do
10627 $LFS setstripe -c 1 -i $i $testfile$i ||
10628 error "failed to setstripe $testfile$i"
10629 fallocate -l $size $testfile$i || error "fallocate failed"
10635 # Create files in the pool now that there is an imbalance
10636 filecount=$(((OSTCOUNT - 1) * filenum))
10637 for ((i = 0; i < filecount; i++)); do
10638 $LFS setstripe -p $pool $testfile-$i ||
10639 error "failed to setstripe -p $pool $testfile-$i"
10641 $LFS getstripe -i $testfile-* > /tmp/$tfile.log
10642 # Count the number of files with a stripe on each OST to verify the
10643 # pool allocated with round-robin
10644 for ((i = 0; i <= poolostrange; i++)); do
10645 stripecount=$(grep -c $i /tmp/$tfile.log)
10646 # Allow a little leeway
10647 if (( stripecount < filenum - 1 ||
10648 stripecount > filenum + 1 )); then
10649 cat /tmp/$tfile.log
10650 error "$stripecount != $filenum files on OST$i"
10654 # Create files across the system now that there is an imbalance
10655 filecount=$((OSTCOUNT * filenum))
10656 for ((i = 1; i < filecount; i++)); do
10657 $LFS setstripe $testfile-$i.2 ||
10658 error "failed to setstripe $testilfe-$i.2"
10660 $LFS getstripe -i $testfile-*.2 > /tmp/$tfile.log
10662 # Count the number of files with a stripe on each OST to verify the
10663 # files are *NOT* allocated with round-robin
10664 for ((i = 0; i <= ostrange; i++)); do
10665 stripecount=$(grep -c $i /tmp/$tfile.log)
10666 if [[ $stripecount -ne $filenum ]]; then
10668 echo "QOS: $stripecount != $filenum files on OST$i"
10671 if [ -z "$qos_used" ]; then
10672 error "QOS not used on imbalanced OSTs!"
10675 rm -rf /tmp/$tfile.log $DIR/$tdir
10676 do_facet mgs $LCTL pool_remove $FSNAME.$pool OST[0-$poolostrange] ||
10677 "failed to remove OST[0-$poolostrange] from the pool"
10678 do_facet mgs $LCTL pool_destroy $FSNAME.$pool ||
10679 error "failed to destroy pool"
10681 run_test 133 "stripe QOS: free space balance in a pool"
10684 [ "$mds1_FSTYPE" == "ldiskfs" ] || skip "ldiskfs only test"
10687 local mdt_dev=$(facet_device mds1)
10688 local tmp_dir=$TMP/$tdir
10689 local out=$tmp_dir/check_iam.txt
10690 local CHECK_IAM=${CHECK_IAM:-$(do_facet mds1 "which check_iam 2> /dev/null || true")}
10692 [[ -n "$CHECK_IAM" ]] || skip "check_iam not found"
10695 for ((i=0; i<64; i++)); do
10697 #cmd introduce a random corruption to IAM file
10698 local cmd="dd if=/dev/urandom of=$tmp_dir/$f bs=2 conv=notrunc count=1 seek=$((RANDOM % 36))"
10699 do_facet mds1 "mkdir -p $tmp_dir; \
10700 $DEBUGFS -c -R 'dump $f $tmp_dir/$f' $mdt_dev 2>&1; \
10701 $CHECK_IAM -v $tmp_dir/$f 2>&1; \
10702 echo $cmd; eval $cmd 2>/dev/null;
10703 $CHECK_IAM -v $tmp_dir/$f 2>&1; echo \\\$?" >> $out 2>&1
10708 stack_trap "rm -rf $tmp_dir && do_facet mds1 rm -rf $tmp_dir" EXIT
10710 rc=$(grep -c "fault\|except" $out)
10711 (( rc == 0 )) || { cat $out &&
10712 error "check_iam failed with fault or exception $rc"; }
10714 rc=$(grep -c "^255" $out)
10715 errors=$(grep -c "FINISHED WITH ERRORS" $out)
10717 (( rc == errors )) || { cat $out &&
10718 error "check_iam errcode does not fit with errors $rc $errors"; }
10720 run_test 134 "check_iam works without faults"
10722 cleanup_test_135(){
10725 printf "\nCleanup test_135\n" >&2
10726 do_facet mds1 "$LCTL set_param -n $oldgc"
10727 rm -rf $DIR/$tdir &> /dev/null
10731 __test_135_file_thread() {
10733 local init_time=$(awk '{print $1}' /proc/uptime)
10734 local awkcmd="/crosses index zero/ {if (\$1 > $init_time) exit(1);}"
10736 #Generate a full plain llogs
10737 while dmesg | sed -r 's/(\[|\])//g' | awk "$awkcmd" ; do
10738 createmany -o $DIR/$tdir/f 4500 >&2
10739 createmany -u $DIR/$tdir/f 4500 >&2
10743 __test_135_reader() {
10746 local firstidx=$(changelog_user_rec mds1 $cl_user)
10747 local oldidx=$firstidx
10752 while read -t10 -u$fd newidx other; do
10753 (( (newidx - oldidx) == 1 )) ||
10754 error "changelog jump detected (last: $oldidx, current: $newidx)"
10756 if (( (newidx - firstidx + 1) % 13000 == 0 )); then
10757 [[ $pid -eq 0 ]] ||
10758 wait $pid || error "changelog_clear failed"
10759 changelog_clear $((newidx - 1)) mds1 >&2 & pid=$!
10764 [[ $pid -eq 0 ]] ||
10765 wait $pid || error "changelog_clear failed"
10771 (( MDS1_VERSION >= $(version_code 2.15.52) )) ||
10772 skip "need MDS version at least 2.15.52"
10774 local service=$(facet_svc mds1)
10776 local lastread lastidx
10777 local files_pid reader_pid
10782 # Need to reformat because we are changing llog catalog sizes to 5.
10783 # Otherwise, processing could fail with existing catalogs (last_idx>5).
10787 # Disable changelog garbage colector
10788 local oldgc=$(do_facet mds1 "$LCTL get_param mdd.${service}.changelog_gc")
10789 do_facet mds1 "$LCTL set_param -n mdd.${service}.changelog_gc=0"
10790 stack_trap "cleanup_test_135 $oldgc" EXIT INT
10792 # change the changelog_catalog size to 5 entries for everybody
10793 #define OBD_FAIL_CAT_RECORDS 0x1312
10794 do_node $(comma_list $(all_nodes)) $LCTL set_param fail_loc=0x1312 fail_val=5
10796 # disable console ratelimit
10797 local rl=$(cat /sys/module/libcfs/parameters/libcfs_console_ratelimit)
10798 echo 0 > /sys/module/libcfs/parameters/libcfs_console_ratelimit
10799 stack_trap "echo $rl > /sys/module/libcfs/parameters/libcfs_console_ratelimit" EXIT
10801 test_mkdir -c 1 -i 0 $DIR/$tdir || error "Failed to create directory"
10802 changelog_chmask "ALL" || error "changelog_chmask failed"
10803 changelog_register || error "changelog_register failed"
10805 cl_user="${CL_USERS[mds1]%% *}"
10806 changelog_users mds1 | grep -q $cl_user ||
10807 error "User $cl_user not found in changelog_users"
10809 # Start reader thread
10810 coproc $LFS changelog --follow $service
10813 stack_trap "echo kill changelog reader; kill $reader_pid" EXIT
10815 echo -e "\nWrap arround changelog catalog"
10817 # Start file writer thread
10818 __test_135_file_thread "$service" & files_pid=$!
10819 stack_trap "(pkill -P$files_pid; kill $files_pid) &> /dev/null || true" EXIT
10821 # Check changelog entries
10822 lastread=$(__test_135_reader $fd $cl_user) || exit $?
10823 ! kill -0 $files_pid 2>/dev/null ||
10824 error "creation thread is running. Is changelog reader stuck?"
10826 lastidx=$(changelog_users mds1 | awk '/current_index/ {print $NF}' )
10827 [[ "$lastread" -eq "$lastidx" ]] ||
10828 error "invalid changelog lastidx (read: $lastread, mds: $lastidx)"
10830 run_test 135 "check the behavior when changelog is wrapped around"
10833 (( MDS1_VERSION >= $(version_code 2.15.55) )) ||
10834 skip "need MDS version at least 2.15.55"
10835 (( MDSCOUNT >= 2 )) || skip "needs >= 2 MDTs"
10837 RM_UPDATELOG=$(do_facet mds2 "which remove_updatelog 2> /dev/null")
10838 RM_UPDATELOG=${RM_UPDATELOG:-"$LUSTRE/scripts/remove_updatelog"}
10840 [ -f "$RM_UPDATELOG" ] ||
10841 skip_env "remove_updatelog is not found on mds2"
10843 local mntpt=$(facet_mntpt mds2)
10846 test_mkdir -c $MDSCOUNT -p $DIR/$tdir ||
10847 error "mkdir $DIR/$tdir failed"
10848 mkdir $DIR/$tdir/d{1..256}
10853 mount_fstype mds2 || error "mount as fstype mds2 failed"
10854 do_facet mds2 $RM_UPDATELOG -n $mntpt
10855 MDTS=($(do_facet mds2 $RM_UPDATELOG -n $mntpt |
10856 grep -o "Processing MDT[0-9]*" | awk -F'MDT' '{print $2}'))
10857 (( ${#MDTS[@]} == MDSCOUNT )) ||
10858 error "Processed ${#MDTS[@]} from $MDSCOUNT"
10860 do_facet mds2 $RM_UPDATELOG -n -m 1,0 $mntpt
10861 MDTS=($(do_facet mds2 $RM_UPDATELOG -n -m 1,0 $mntpt |
10862 grep -o "Processing MDT[0-9]*" | awk -F'MDT' '{print $2}'))
10863 (( ${#MDTS[@]} == 2 )) ||
10864 error "Processed ${#MDTS[@]} instead of 2"
10865 (( ${MDTS[0]} == 1 && ${MDTS[1]} == 0 )) ||
10866 error "Processed: ${MDTS[*]}, expected: 1 0"
10868 do_facet mds2 $RM_UPDATELOG -m 0 $mntpt
10869 unmount_fstype mds2
10870 start_mdt 2 || error "mds2 start fail"
10871 start_mdt 1 || error "mds1 start fail"
10872 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds1 FULL
10873 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
10875 rm -rf $DIR/$tdir || error "Can't remove $tdir"
10877 reformat_and_config
10879 run_test 140 "remove_updatelog script actions"
10884 local max_cached_mb=$($LCTL get_param llite.*.max_cached_mb |
10885 awk '/^max_cached_mb/ { print $2 }')
10886 stack_trap "$LCTL set_param -n llite.*.max_cached_mb=$max_cached_mb"
10888 $LCTL set_param llite.*.max_cached_mb='100%'
10890 local new_max_cached_mb=$($LCTL get_param llite.*.max_cached_mb |
10891 awk '/^max_cached_mb/ { print $2 }')
10892 local total_ram_mb=$(free -m | grep 'Mem:' | awk '{print $2}')
10894 $LCTL get_param llite.*.max_cached_mb
10895 echo "total ram mb: $total_ram_mb"
10896 (( new_max_cached_mb == total_ram_mb )) ||
10897 error "setting cache to 100% not equal to total RAM"
10899 $LCTL set_param llite.*.max_cached_mb='50%'
10900 new_max_cached_mb=$($LCTL get_param llite.*.max_cached_mb |
10901 awk '/^max_cached_mb/ { print $2 }')
10903 $LCTL get_param llite.*.max_cached_mb
10904 (( new_max_cached_mb == $((total_ram_mb / 2)) )) ||
10905 error "setting cache to 50% not equal to 50% of RAM"
10907 $LCTL set_param llite.*.max_cached_mb='105%' &&
10908 error "should not be able to set insane value"
10910 $LCTL set_param llite.*.max_cached_mb='0%'
10911 new_max_cached_mb=$($LCTL get_param llite.*.max_cached_mb |
10912 awk '/^max_cached_mb/ { print $2 }')
10913 # Minimum cache size is 64 MiB
10914 $LCTL get_param llite.*.max_cached_mb
10915 (( new_max_cached_mb == 64 )) ||
10916 error "setting cache to 0% != minimum cache size"
10918 run_test 150 "test setting max_cached_mb to a %"
10921 (( MDS1_VERSION >= $(version_code 2.15.58) )) ||
10922 skip "need MDS version at least 2.15.58"
10923 [[ "$ost1_FSTYPE" == ldiskfs ]] || skip "ldiskfs only test"
10926 if ! combined_mgs_mds ; then
10930 echo "Damage ost1 local config log"
10931 do_facet ost1 "$DEBUGFS -w -R 'punch CONFIGS/$FSNAME-OST0000 0 1' \
10932 $(ostdevname 1) || return \$?" ||
10933 error "do_facet ost1 failed with $?"
10935 # expect OST to fail mount with no MGS and bad local config
10936 start_ost && error "OST start should fail"
10938 if ! combined_mgs_mds ; then
10941 start_mds || error "MDS start failed"
10942 # now it should start with MGS config
10943 start_ost || error "OST start failed"
10944 reformat_and_config
10946 run_test 151 "damaged local config doesn't prevent mounting"
10949 # (This was sanity/802a)
10951 saved_MGS_MOUNT_OPTS=$MGS_MOUNT_OPTS
10952 saved_MDS_MOUNT_OPTS=$MDS_MOUNT_OPTS
10953 saved_OST_MOUNT_OPTS=$OST_MOUNT_OPTS
10954 saved_MOUNT_OPTS=$MOUNT_OPTS
10958 MGS_MOUNT_OPTS=$saved_MGS_MOUNT_OPTS
10959 MDS_MOUNT_OPTS=$saved_MDS_MOUNT_OPTS
10960 OST_MOUNT_OPTS=$saved_OST_MOUNT_OPTS
10961 MOUNT_OPTS=$saved_MOUNT_OPTS
10966 [[ $mds1_FSTYPE = zfs ]] || skip "ZFS specific test"
10967 [[ $MDS1_VERSION -lt $(version_code 2.9.55) ]] ||
10968 [[ $OST1_VERSION -lt $(version_code 2.9.55) ]] &&
10969 skip "Need server version at least 2.9.55"
10971 [[ $ENABLE_QUOTA ]] && skip "Quota enabled for read-only test"
10973 # Reset before starting
10977 mkdir $DIR/$tdir || error "(1) fail to mkdir"
10979 cp $LUSTRE/tests/test-framework.sh $DIR/$tdir/ ||
10980 error "(2) Fail to copy"
10982 stack_trap cleanup_802a EXIT
10984 # sync by force before remount as readonly
10985 sync; sync_all_data; sleep 3; sync_all_data
10989 MGS_MOUNT_OPTS=$(csa_add "$MGS_MOUNT_OPTS" -o rdonly_dev)
10990 MDS_MOUNT_OPTS=$(csa_add "$MDS_MOUNT_OPTS" -o rdonly_dev)
10991 OST_MOUNT_OPTS=$(csa_add "$OST_MOUNT_OPTS" -o rdonly_dev)
10993 echo "Mount the server as read only"
10994 setupall server_only || error "(3) Fail to start servers"
10996 echo "Mount client without ro should fail"
10997 mount_client $MOUNT &&
10998 error "(4) Mount client without 'ro' should fail"
11000 echo "Mount client with ro should succeed"
11001 MOUNT_OPTS=$(csa_add "$MOUNT_OPTS" -o ro)
11002 mount_client $MOUNT ||
11003 error "(5) Mount client with 'ro' should succeed"
11005 echo "Modify should be refused"
11006 touch $DIR/$tdir/guard && error "(6) Touch should fail under ro mode"
11008 echo "Read should be allowed"
11009 diff $LUSTRE/tests/test-framework.sh $DIR/$tdir/test-framework.sh ||
11010 error "(7) Read should succeed under ro mode"
11012 run_test 802a "simulate readonly device"
11014 if ! combined_mgs_mds ; then
11020 # restore the values of MDSSIZE and OSTSIZE
11021 MDSSIZE=$STORED_MDSSIZE
11022 OSTSIZE=$STORED_OSTSIZE
11025 complete_test $SECONDS
11026 check_and_cleanup_lustre