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
28 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 32a 32b 32c 32d 32e"
29 # bug number for skipped test: LU-9897 LU-12032
30 ALWAYS_EXCEPT="$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()"
553 load_module ../libcfs/libcfs/libcfs ||
554 error "Failed to load modules libcfs rc = $?"
555 $LCTL set_param debug=all
558 run_test 5j "verify libcfs doesn't crash when setting debug flags"
563 mount_client $MOUNT || error "mount_client $MOUNT failed"
564 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
565 cleanup || error "cleanup failed with rc $?"
567 run_test 6 "manual umount, then mount again"
572 cleanup_nocli || error "cleanup_nocli failed with $?"
574 run_test 7 "manual umount, then cleanup"
578 mount_client $MOUNT2 || error "mount_client $MOUNT2 failed"
579 check_mount2 || error "check_mount2 failed"
580 umount_client $MOUNT2 || error "umount_client $MOUNT2 failed"
581 cleanup || error "cleanup failed with rc $?"
583 run_test 8 "double mount setup"
586 start_ost || error "OST start failed"
588 do_facet ost1 $LCTL set_param debug=\'inode trace\' ||
589 error "do_facet ost1 set_param inode trace failed."
590 do_facet ost1 $LCTL set_param subsystem_debug=\'mds ost\' ||
591 error "do_facet ost1 set_param debug mds ost failed."
593 CHECK_PTLDEBUG="`do_facet ost1 $LCTL get_param -n debug`"
594 if [ "$CHECK_PTLDEBUG" ] && { \
595 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
596 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
597 echo "lnet.debug success"
599 error "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
601 CHECK_SUBSYS="`do_facet ost1 $LCTL get_param -n subsystem_debug`"
602 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
603 echo "lnet.subsystem_debug success"
605 error "lnet.subsystem_debug: want 'mds ost' got '$CHECK_SUBSYS'"
607 stop_ost || error "Unable to stop OST1"
609 run_test 9 "test ptldebug and subsystem for mkfs"
614 if ! combined_mgs_mds; then
615 files=$(do_facet mgs "find /{proc/fs,sys/fs,sys/kernel/debug}/lustre -type l -exec test ! -e {} \; -print")
616 [ -z $files ] || echo "MGS $files is a broken symlink"
619 files=$(do_facet mds1 "find /{proc/fs,sys/fs,sys/kernel/debug}/lustre -type l -exec test ! -e {} \; -print")
620 [ -z $files ] || echo "MDS $files is a broken symlink"
622 files=$(do_facet ost1 "find /{proc/fs,sys/fs,sys/kernel/debug}/lustre -type l -exec test ! -e {} \; -print")
623 [ -z $files ] || echo "OSS $files is a broken symlink"
625 files=$(do_facet client "find /{proc/fs,sys/fs,sys/kernel/debug}/lustre -type l -exec test ! -e {} \; -print")
626 [ -z $files ] || echo "clients $files is a broken symlink"
628 cleanup || error "cleanup failed with rc $?"
630 run_test 10a "find lctl param broken symlinks"
633 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
634 # But with new MDS stack we don't care about the mode of local objects
635 # anymore, so this test is removed. See bug 22944 for more details.
639 if [ "$mds1_FSTYPE" != ldiskfs ]; then
640 skip "ldiskfs only test"
644 check_mount || error "check_mount failed"
645 cleanup || error "cleanup failed with rc $?"
647 echo "Remove mds config log"
648 if ! combined_mgs_mds ; then
652 do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' \
653 $(mgsdevname) || return \$?" ||
654 error "do_facet mgs failed with $?"
656 if ! combined_mgs_mds ; then
660 start_ost || error "OST start failed"
661 start_mds && error "MDS start succeeded, but should fail"
664 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
667 if [ "$mds1_FSTYPE" != ldiskfs ]; then
668 skip "ldiskfs only test"
671 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
676 # check if current MDSSIZE is large enough
677 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE &&
678 log "use MDSSIZE=$MDSSIZE"
680 # check if the global config has a large enough MDSSIZE
681 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] &&
682 [ $STORED_MDSSIZE -ge $MIN ] &&
683 OK=1 && myMDSSIZE=$STORED_MDSSIZE &&
684 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
686 # check if the block device is large enough
687 is_blkdev $SINGLEMDS $MDSDEV $MIN
688 local large_enough=$?
689 if [ -n "$OK" ]; then
690 [ $large_enough -ne 0 ] && OK=""
692 [ $large_enough -eq 0 ] && OK=1 && myMDSSIZE=$MIN &&
693 log "use device $MDSDEV with MIN=$MIN"
696 # check if a loopback device has enough space for fs metadata (5%)
698 if [ -z "$OK" ]; then
699 local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! \
700 -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
701 awk '($1 != "Filesystem") { print $4 }')
702 ! [ -z "$SPACE" ] && [ $SPACE -gt $((MIN / 20)) ] &&
703 OK=1 && myMDSSIZE=$MIN &&
704 log "use file $MDSDEV with MIN=$MIN"
707 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS"
709 echo "mount mds with large journal..."
711 local OLD_MDSSIZE=$MDSSIZE
715 echo "mount lustre system..."
717 check_mount || error "check_mount failed"
719 echo "check journal size..."
720 local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
721 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
722 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
724 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
727 cleanup || error "cleanup failed with rc $?"
732 run_test 18 "check mkfs creates large journals"
735 start_mds || error "MDS start failed"
736 stop_mds || error "Unable to stop MDS"
738 run_test 19a "start/stop MDS without OSTs"
741 start_ost || error "Unable to start OST1"
742 stop_ost -f || error "Unable to stop OST1"
744 run_test 19b "start/stop OSTs without MDS"
747 # first format the ost/mdt
748 start_mds || error "MDS start failed"
749 start_ost || error "Unable to start OST1"
750 mount_client $MOUNT || error "mount_client $MOUNT failed"
751 check_mount || error "check_mount failed"
752 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed."
753 mount_client $MOUNT remount,ro || error "remount client with ro failed"
754 touch $DIR/$tfile && error "$DIR/$tfile created incorrectly"
755 [ -e $DIR/$tfile ] && error "$DIR/$tfile exists incorrectly"
756 mount_client $MOUNT remount,rw || error "remount client with rw failed"
757 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
758 MCNT=$(grep -c $MOUNT' ' /etc/mtab)
759 [ "$MCNT" -ne 1 ] && error "$MOUNT in /etc/mtab $MCNT times"
761 stop_mds || error "Unable to stop MDS"
762 stop_ost || error "Unable to stop OST1"
764 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
767 start_mds || error "MDS start failed"
768 start_ost || error "unable to start OST1"
769 wait_osc_import_state mds ost FULL
770 stop_ost || error "unable to stop OST1"
771 stop_mds || error "unable to stop MDS"
773 run_test 21a "start mds before ost, stop ost first"
776 start_ost || error "unable to start OST1"
777 start_mds || error "MDS start failed"
778 wait_osc_import_state mds ost FULL
779 stop_mds || error "unable to stop MDS"
780 stop_ost || error "unable to stop OST1"
782 run_test 21b "start ost before mds, stop mds first"
785 start_ost || error "Unable to start OST1"
786 start_mds || error "MDS start failed"
787 start_ost2 || error "Unable to start OST2"
788 wait_osc_import_state mds ost2 FULL
789 stop_ost || error "Unable to stop OST1"
790 stop_ost2 || error "Unable to stop OST2"
791 stop_mds || error "Unable to stop MDS"
792 #writeconf to remove all ost2 traces for subsequent tests
793 writeconf_or_reformat
795 run_test 21c "start mds between two osts, stop mds last"
798 combined_mgs_mds && skip "need separate mgs device"
803 start_mgs || error "unable to start MGS"
804 start_ost || error "unable to start OST1"
805 start_ost2 || error "unable to start OST2"
806 start_mds || error "MDS start failed"
807 wait_osc_import_state mds ost2 FULL
809 local zkeeper=${KEEP_ZPOOL}
810 stack_trap "KEEP_ZPOOL=$zkeeper" EXIT
813 stop_ost || error "Unable to stop OST1"
814 stop_ost2 || error "Unable to stop OST2"
815 stop_mds || error "Unable to stop MDS"
817 #writeconf to remove all ost2 traces for subsequent tests
818 writeconf_or_reformat
819 KEEP_ZPOOL="${zkeeper}"
821 start_mgs || error "unable to start MGS"
823 run_test 21d "start mgs then ost and then mds"
826 MGSNID="$saved_mgsnid"
828 echo "stopping fs2mgs ..."
829 stop $fs2mgs -f || true
832 test_21e() { # LU-5863
833 if [[ -z "$fs3ost_DEV" || -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
834 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
835 skip_env "mixed loopback and real device not working"
838 local fs2mdsdev=$(mdsdevname 1_2)
839 local fs2ostdev=$(ostdevname 1_2)
840 local fs3ostdev=$(ostdevname 2_2)
842 local fs2mdsvdev=$(mdsvdevname 1_2)
843 local fs2ostvdev=$(ostvdevname 1_2)
844 local fs3ostvdev=$(ostvdevname 2_2)
846 # temporarily use fs3ost as fs2mgs
848 local fs2mgsdev=$fs3ostdev
849 local fs2mgsvdev=$fs3ostvdev
851 local fsname=test1234
853 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
854 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
855 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS && trap cleanup_21e EXIT INT ||
856 error "start fs2mgs failed"
858 local saved_mgsnid="$MGSNID"
859 MGSNID=$(do_facet $fs2mgs $LCTL list_nids | xargs | tr ' ' ,)
861 add fs2mds $(mkfs_opts mds1 $fs2mdsdev $fsname) \
862 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
863 add fs2ost $(mkfs_opts ost1 $fs2ostdev $fsname) \
864 --reformat $fs2ostdev $fs2ostvdev || error "add fs2ost failed"
866 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
867 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS || error "start fs2mds failed"
869 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
870 $MOUNT_CMD $MGSNID:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
871 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
875 run_test 21e "separate MGS and MDS"
878 start_mds || error "MDS start failed"
880 echo "Client mount with ost in logs, but none running"
881 start_ost || error "unable to start OST1"
882 # wait until mds connected to ost and open client connection
883 wait_osc_import_state mds ost FULL
884 stop_ost || error "unable to stop OST1"
885 mount_client $MOUNT || error "mount_client $MOUNT failed"
886 # check_mount will block trying to contact ost
887 mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed"
888 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed"
889 umount_client $MOUNT -f
892 echo "Client mount with a running ost"
893 start_ost || error "unable to start OST1"
895 # if gss enabled, wait full time to let connection from
896 # mds to ost be established, due to the mismatch between
897 # initial connect timeout and gss context negotiation timeout.
898 # This perhaps could be remove after AT landed.
899 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
900 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
902 mount_client $MOUNT || error "mount_client $MOUNT failed"
903 wait_osc_import_state mds ost FULL
904 wait_osc_import_ready client ost
905 check_mount || error "check_mount failed"
908 cleanup || error "cleanup failed with rc $?"
910 run_test 22 "start a client before osts (should return errs)"
912 test_23a() { # was test_23
915 stop $SINGLEMDS || error "failed to stop $SINGLEMDS"
916 # force down client so that recovering mds waits for reconnect
917 local running=$(grep -c $MOUNT /proc/mounts) || true
918 if [ $running -ne 0 ]; then
919 echo "Stopping client $MOUNT (opts: -f)"
923 # enter recovery on failed mds
924 local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/})
925 start $SINGLEMDS $MDT_DEV $MDS_MOUNT_OPTS || error "MDS start failed"
926 # try to start a new client
927 mount_client $MOUNT &
929 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
930 MOUNT_LUSTRE_PID=$(ps -ef | grep mount.lustre |
931 grep -v grep | awk '{print $2}')
932 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
934 ps --ppid $MOUNT_LUSTRE_PID
935 echo "waiting for mount to finish"
937 # "ctrl-c" sends SIGINT but it usually (in script) does not work on child process
938 # SIGTERM works but it does not spread to offspring processses
939 kill -s TERM $MOUNT_PID
940 kill -s TERM $MOUNT_LUSTRE_PID
941 # we can not wait $MOUNT_PID because it is not a child of this shell
947 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
949 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
950 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
953 [ -z "$PID1" -a -z "$PID2" ] && break
954 echo "waiting for mount to finish ... "
955 WAIT=$(( WAIT + sleep))
957 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
958 error "MOUNT_PID $MOUNT_PID and "\
959 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
962 cleanup || error "cleanup failed with rc $?"
964 run_test 23a "interrupt client during recovery mount delay"
966 test_23b() { # was test_23
967 start_mds || error "MDS start failed"
968 start_ost || error "Unable to start OST1"
969 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
970 $LCTL set_param fail_loc=0x80000313
972 cleanup || error "cleanup failed with rc $?"
974 run_test 23b "Simulate -EINTR during mount"
977 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
979 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
980 is_blkdev $SINGLEMDS $MDSDEV &&
981 skip_env "mixed loopback and real device not working"
984 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
986 local fs2mdsdev=$(mdsdevname 1_2)
987 local fs2ostdev=$(ostdevname 1_2)
988 local fs2mdsvdev=$(mdsvdevname 1_2)
989 local fs2ostvdev=$(ostvdevname 1_2)
992 # LU-9733 test fsname started with numbers as well
993 local FSNAME2=969362ae
995 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --nomgs --mgsnode=$MGSNID \
996 --fsname=${FSNAME2} --reformat $fs2mdsdev $fs2mdsvdev || exit 10
998 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --fsname=${FSNAME2} \
999 --reformat $fs2ostdev $fs2ostvdev || exit 10
1002 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
1003 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1004 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
1005 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
1007 # LU-9733 test fsname started with numbers
1008 cl_user=$(do_facet $SINGLEMDS lctl --device $FSNAME2-MDT0000 \
1009 changelog_register -n) ||
1010 error "register changelog failed"
1012 do_facet $SINGLEMDS lctl --device $FSNAME2-MDT0000 \
1013 changelog_deregister $cl_user ||
1014 error "deregister changelog failed"
1016 check_mount || error "check_mount failed"
1017 # files written on 1 should not show up on 2
1018 cp /etc/passwd $DIR/$tfile
1020 [ -e $MOUNT2/$tfile ] && error "File bleed"
1023 cp /etc/passwd $MOUNT2/$tfile ||
1024 error "cp /etc/passwd $MOUNT2/$tfile failed"
1025 rm $MOUNT2/$tfile || error "remove $MOUNT2/$tfile failed"
1026 # 2 is actually mounted
1027 grep $MOUNT2' ' /proc/mounts > /dev/null || error "$MOUNT2 not mounted"
1029 facet_failover fs2mds
1030 facet_failover fs2ost
1032 umount_client $MOUNT
1033 # the MDS must remain up until last MDT
1035 MDS=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
1036 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
1037 [ -z "$MDS" ] && error "No MDT"
1039 cleanup_nocli || error "cleanup_nocli failed with rc $?"
1041 run_test 24a "Multiple MDTs on a single node"
1044 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1046 skip "needs combined MGT and MDT device"
1048 if [ -z "$fs2mds_DEV" ]; then
1049 local dev=${SINGLEMDS}_dev
1050 local MDSDEV=${!dev}
1051 is_blkdev $SINGLEMDS $MDSDEV &&
1052 skip_env "mixed loopback and real device not working"
1055 local fs2mdsdev=$(mdsdevname 1_2)
1056 local fs2mdsvdev=$(mdsvdevname 1_2)
1058 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --mgs --fsname=${FSNAME}2 \
1059 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
1061 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && {
1063 error "start MDS should fail"
1066 cleanup || error "cleanup failed with rc $?"
1068 run_test 24b "Multiple MGSs on a single node (should return err)"
1072 check_mount || error "check_mount failed"
1073 local MODULES=$($LCTL modules | awk '{ print $2 }')
1074 rmmod $MODULES 2>/dev/null || true
1075 cleanup || error "cleanup failed with $?"
1077 run_test 25 "Verify modules are referenced"
1081 # we need modules before mount for sysctl, so make sure...
1082 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
1083 #define OBD_FAIL_MDS_FS_SETUP 0x135
1084 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000135"
1085 start_mds && error "MDS started but should not have started"
1086 $LCTL get_param -n devices
1087 DEVS=$($LCTL get_param -n devices | egrep -v MG | wc -l)
1088 [ $DEVS -gt 0 ] && error "number of devices is $DEVS, should be zero"
1089 # start mds to drop writeconf setting
1090 start_mds || error "Unable to start MDS"
1091 stop_mds || error "Unable to stop MDS"
1092 unload_modules_conf || error "unload_modules_conf failed with $?"
1094 run_test 26 "MDT startup failure cleans LOV (should return errs)"
1098 start_ost || error "Unable to start OST1"
1099 start_mds || error "Unable to start MDS"
1100 echo "Requeue thread should have started: "
1101 ps -e | grep ll_cfg_requeue
1102 set_persistent_param_and_check ost1 \
1103 "obdfilter.$FSNAME-OST0000.client_cache_seconds" \
1104 "$FSNAME-OST0000.ost.client_cache_seconds"
1105 cleanup_nocli || error "cleanup_nocli failed with rc $?"
1107 run_test 27a "Reacquire MGS lock if OST started first"
1112 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
1113 awk '($3 ~ "mdt" && $4 ~ "MDT0000") { print $4 }')
1115 facet_failover $SINGLEMDS
1116 set_persistent_param_and_check $SINGLEMDS \
1117 "mdt.$device.identity_acquire_expire" \
1118 "$device.mdt.identity_acquire_expire"
1119 set_persistent_param_and_check client \
1120 "mdc.$device-mdc-*.max_rpcs_in_flight" \
1121 "$device.mdc.max_rpcs_in_flight"
1123 cleanup || error "cleanup failed with $?"
1125 run_test 27b "Reacquire MGS lock after failover"
1127 test_28A() { # was test_28
1130 local TEST="llite.$FSNAME-*.max_read_ahead_whole_mb"
1131 local PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
1132 local orig=$($LCTL get_param -n $TEST)
1133 local max=$($LCTL get_param -n \
1134 llite.$FSNAME-*.max_read_ahead_per_file_mb)
1136 orig=${orig%%.[0-9]*}
1138 echo "ORIG:$orig MAX:$max"
1139 [[ $max -le $orig ]] && orig=$((max - 3))
1140 echo "ORIG:$orig MAX:$max"
1142 local final=$((orig + 1))
1144 set_persistent_param_and_check client "$TEST" "$PARAM" $final
1145 final=$((final + 1))
1146 set_persistent_param_and_check client "$TEST" "$PARAM" $final
1147 umount_client $MOUNT || error "umount_client $MOUNT failed"
1148 mount_client $MOUNT || error "mount_client $MOUNT failed"
1150 local result=$($LCTL get_param -n $TEST)
1152 if [ $result -ne $final ]; then
1153 error "New config not seen: wanted $final got $result"
1155 echo "New config success: got $result"
1157 set_persistent_param_and_check client "$TEST" "$PARAM" $orig
1158 cleanup || error "cleanup failed with rc $?"
1160 run_test 28A "permanent parameter setting"
1162 test_28a() { # LU-4221
1163 [[ "$OST1_VERSION" -ge $(version_code 2.5.52) ]] ||
1164 skip "Need OST version at least 2.5.52"
1165 [ "$ost1_FSTYPE" = zfs ] &&
1166 skip "LU-4221: no such proc params for ZFS OSTs"
1173 local device="$FSNAME-OST0000"
1177 # In this test we will set three kinds of proc parameters with
1178 # lctl set_param -P or lctl conf_param:
1179 # 1. non-symlink ones in the OFD
1180 # 2. non-symlink ones in the OSD
1183 # prepare a non-symlink parameter in the OFD
1184 name="client_cache_seconds"
1185 param="$device.ost.$name"
1186 cmd="obdfilter.$device.$name"
1188 # permanently setting the parameter in the OFD
1189 old=$(do_facet ost1 $LCTL get_param -n $cmd)
1191 set_persistent_param_and_check ost1 "$cmd" "$param" $new
1192 set_persistent_param_and_check ost1 "$cmd" "$param" $old
1195 # prepare a non-symlink parameter in the OSD
1197 param="$device.osd.$name"
1198 cmd="osd-*.$device.$name"
1200 # conf_param the parameter in the OSD
1201 old=$(do_facet ost1 $LCTL get_param -n $cmd)
1202 new=$(((old + 1) % 2))
1203 set_persistent_param_and_check ost1 "$cmd" "$param" $new
1204 set_persistent_param_and_check ost1 "$cmd" "$param" $old
1206 cleanup || error "cleanup failed with $?"
1208 run_test 28a "set symlink parameters permanently with lctl"
1211 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >= 2 OSTs"
1212 setup_noconfig > /dev/null 2>&1
1213 start_ost2 || error "Unable to start OST2"
1216 local PARAM="$FSNAME-OST0001.osc.active"
1217 # With lctl set_param -P the value $PROC_ACT will be sent to
1218 # all nodes. The [!M] filter out the ability to set active
1219 # on the MDS servers which is tested with wait_osp_* below.
1220 # For ost_server_uuid that only exist on client so filtering
1222 local PROC_ACT="os[cp].$FSNAME-OST0001-osc-[^M]*.active"
1223 local PROC_UUID="os[cp].$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
1225 ACTV=$($LCTL get_param -n $PROC_ACT)
1227 set_persistent_param_and_check client $PROC_ACT $PARAM $DEAC
1228 # also check ost_server_uuid status
1229 RESULT=$($LCTL get_param -n $PROC_UUID | grep DEACTIV)
1230 if [ -z "$RESULT" ]; then
1231 error "Client not deactivated: $($LCTL get_param \
1234 echo "Live client success: got $RESULT"
1238 wait_osp_active ost ${FSNAME}-OST0001 1 0
1240 # test new client starts deactivated
1241 umount_client $MOUNT || error "umount_client $MOUNT failed"
1242 mount_client $MOUNT || error "mount_client $MOUNT failed"
1244 # the 2nd and 3rd field of ost_server_uuid do not update at the same
1245 # time when using lctl set_param -P
1246 wait_update_facet client \
1247 "$LCTL get_param -n $PROC_UUID | awk '{print \\\$3 }'" \
1249 error "New client start active: $($LCTL get_param -n $PROC_UUID)"
1251 echo "New client success: got '$($LCTL get_param -n $PROC_UUID)'"
1253 # make sure it reactivates
1254 set_persistent_param_and_check client $PROC_ACT $PARAM $ACTV
1256 umount_client $MOUNT
1257 stop_ost2 || error "Unable to stop OST2"
1258 cleanup_nocli || error "cleanup_nocli failed with $?"
1259 #writeconf to remove all ost2 traces for subsequent tests
1260 writeconf_or_reformat
1262 run_test 29 "permanently remove an OST"
1267 echo Big config llog
1268 local path="llite.$FSNAME-*"
1269 local cpath="$FSNAME.llite"
1270 local param="max_read_ahead_whole_mb"
1272 local test="${path}.$param"
1273 local conf="${cpath}.$param"
1275 local orig=$($LCTL get_param -n $test)
1276 local list=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1277 for i in ${list[@]}; do
1278 set_persistent_param_and_check client $test $conf $i
1280 # make sure client restart still works
1281 umount_client $MOUNT
1282 mount_client $MOUNT || error "mount_client $MOUNT failed"
1283 [ "$($LCTL get_param -n $test)" -ne "$i" ] &&
1284 error "Param didn't stick across restart $($test) != $i"
1287 echo Erase parameter setting
1288 if [[ $PERM_CMD == *"set_param -P"* ]]; then
1289 do_facet mgs "$PERM_CMD -d $test" ||
1290 error "Erase param $test failed"
1292 do_facet mgs "$PERM_CMD -d $conf" ||
1293 error "Erase param $conf failed"
1295 umount_client $MOUNT
1296 mount_client $MOUNT || error "mount_client $MOUNT failed"
1297 local final=$($LCTL get_param -n $test)
1298 echo "deleted (default) value=$final, orig=$orig"
1299 orig=${orig%%.[0-9]*}
1300 final=${final%%.[0-9]*}
1301 # assumes this parameter started at the default value
1302 [ "$final" -eq "$orig" ] ||
1303 error "Deleted value=$final -ne orig=$orig"
1305 cleanup || error "cleanup failed with rc $?"
1307 run_test 30a "Big config llog and permanent parameter deletion"
1312 local orignids=$($LCTL get_param -n \
1313 osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1315 local orignidcount=$(echo "$orignids" | wc -w)
1317 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1318 # numerical part of it. Hopefully that's not already a failover address
1320 local OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | \
1322 local ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1323 local NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1324 local NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1325 echo "Using fake nid $NEW"
1327 local TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1328 grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1329 if [[ $PERM_CMD == *"set_param -P"* ]]; then
1330 PARAM="osc.$FSNAME-OST0000-osc-[^M]*.import"
1331 echo "Setting $PARAM from $TEST to $NEW"
1332 do_facet mgs "$PERM_CMD $PARAM='connection=$NEW'" ||
1333 error "$PERM_CMD $PARAM failed"
1335 PARAM="$FSNAME-OST0000.failover.node"
1336 echo "Setting $PARAM from $TEST to $NEW"
1337 do_facet mgs "$PERM_CMD $PARAM='$NEW'" ||
1338 error "$PARAM $PARAM failed"
1340 wait_update_facet client "$TEST" "$NEW" ||
1341 error "check $PARAM failed!"
1343 local NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1345 local NIDCOUNT=$(echo "$NIDS" | wc -w)
1346 echo "should have $((orignidcount + 1)) entries \
1347 in failover nids string, have $NIDCOUNT"
1348 [ $NIDCOUNT -eq $((orignidcount + 1)) ] ||
1349 error "Failover nid not added"
1351 if [[ $PERM_CMD == *"set_param -P"* ]]; then
1352 do_facet mgs "$PERM_CMD -d osc.$FSNAME-OST0000-osc-*.import"
1354 do_facet mgs "$PERM_CMD -d $FSNAME-OST0000.failover.node" ||
1355 error "$PERM_CMD delete failed"
1357 umount_client $MOUNT
1358 mount_client $MOUNT || error "mount_client $MOUNT failed"
1360 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1362 NIDCOUNT=$(echo "$NIDS" | wc -w)
1363 echo "only $orignidcount final entries should remain \
1364 in failover nids string, have $NIDCOUNT"
1365 [ $NIDCOUNT -eq $orignidcount ] || error "Failover nids not removed"
1367 cleanup || error "cleanup failed with rc $?"
1369 run_test 30b "Remove failover nids"
1371 test_31() { # bug 10734
1372 # ipaddr must not exist
1373 $MOUNT_CMD 4.3.2.1@tcp:/lustre $MOUNT || true
1374 cleanup || error "cleanup failed with rc $?"
1376 run_test 31 "Connect to non-existent node (shouldn't crash)"
1380 T32_BLIMIT=40960 # Kbytes
1383 T32_PROLIMIT=$((T32_BLIMIT/10))
1385 # This is not really a test but a tool to create new disk
1386 # image tarballs for the upgrade tests.
1388 # Disk image tarballs should be created on single-node
1389 # clusters by running this test with default configurations
1390 # plus a few mandatory environment settings that are verified
1391 # at the beginning of the test.
1393 test_32newtarball() {
1397 local tmp=$TMP/t32_image_create
1408 local pj_quota_file_old
1411 if [ $FSNAME != t32fs -o \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o \
1412 $OSTCOUNT -ne 2 -o -z "$OSTDEV1" ]; then
1413 error "Needs FSNAME=t32fs MDSCOUNT=2 " \
1414 "MDSDEV1=<nonexistent_file> " \
1415 "MDSDEV2=<nonexistent_file> " \
1416 "(or MDSDEV, in the case of b1_8) " \
1417 "OSTCOUNT=2 OSTDEV1=<nonexistent_file> " \
1418 "OSTDEV2=<nonexistent_file>"
1422 echo "Found stale $tmp"
1426 mkdir $tmp/src || return 1
1427 tar cf - -C $src . | tar xf - -C $tmp/src
1429 # format ost with comma-separated NIDs to verify LU-4460
1430 local failnid="$(h2nettype 1.2.3.4),$(h2nettype 4.3.2.1)"
1431 MGSNID="$MGSNID,$MGSNID" OSTOPT="--failnode=$failnid" formatall
1435 [[ "$MDS1_VERSION" -ge $(version_code 2.3.50) ]] ||
1436 $LFS quotacheck -ug /mnt/$FSNAME
1437 $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
1440 if [[ $MDSCOUNT -ge 2 ]]; then
1441 remote_dir=/mnt/$FSNAME/remote_dir
1442 $LFS mkdir -i 1 $remote_dir
1443 tar cf - -C $tmp/src . | tar xf - -C $remote_dir
1445 target_dir=$remote_dir
1446 if [[ $MDS1_VERSION -ge $(version_code 2.7.0) ]]; then
1447 striped_dir=/mnt/$FSNAME/striped_dir_old
1448 $LFS mkdir -i 1 -c 2 $striped_dir
1449 tar cf - -C $tmp/src . | tar xf - -C $striped_dir
1452 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1453 target_dir=/mnt/$FSNAME
1456 # add project quota #
1457 [[ $mds1_FSTYPE == "ldiskfs" &&
1458 $MDS1_VERSION -gt $(version_code 2.9.55) ]] ||
1459 [[ $mds1_FSTYPE == "zfs" &&
1460 $MDS1_VERSION -gt $(version_code 2.10.53) ]] && {
1461 pj_quota_dir=$target_dir/project_quota_dir
1462 pj_quota_file_old=$pj_quota_dir/pj_quota_file_old
1464 enable_project_quota
1465 set_mdt_qtype ugp ||
1466 error "enable mdt quota failed"
1468 set_ost_qtype ugp ||
1469 error "enable ost quota failed"
1471 mkdir -p $pj_quota_dir
1472 $LFS setquota -p $T32_PRJID -b 0 -B $T32_PROLIMIT -i 0 \
1473 -I $T32_ILIMIT $pj_quota_dir ||
1474 error "setquota -p $T32_PRJID failed"
1476 $LFS setstripe $pj_quota_file_old -c 1 ||
1477 error "setstripe $pj_quota_file_old failed"
1479 chown $T32_QID:$T32_QID $pj_quota_file_old ||
1480 error "chown $pj_quota_file_old failed"
1482 change_project -p $T32_PRJID $pj_quota_file_old
1484 mkdir -p $tmp/src/project_quota_dir
1485 cp $pj_quota_file_old $tmp/src/project_quota_dir/
1488 #####################
1489 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1491 #if [[ $MDSCOUNT -ge 2 ]]; then
1492 # remote_dir=/mnt/$FSNAME/remote_dir
1493 # $LFS mkdir -i 1 $remote_dir
1494 # tar cf - -C $tmp/src . | tar xf - -C $remote_dir
1496 # if [[ "$MDS1_VERSION" -ge $(version_code 2.7.0) ]]; then
1497 # striped_dir=/mnt/$FSNAME/striped_dir_old
1498 # $LFS mkdir -i 1 -c 2 $striped_dir
1499 # tar cf - -C $tmp/src . | tar xf - -C $striped_dir
1504 if [[ $MDS1_VERSION -ge $(version_code 2.9.51) ]]; then
1505 pfl_dir=$target_dir/pfl_dir
1506 pfl_file=$pfl_dir/pfl_file
1508 $LFS setstripe -E 2M -c 1 -o 0 -E -1 -S 2M -c 1 -o 1 \
1510 error "Create PFL file failed"
1512 dd if=/dev/urandom of=$pfl_file bs=1k count=3k
1513 mkdir -p $tmp/src/pfl_dir
1514 cp $pfl_file $tmp/src/pfl_dir/
1519 if [[ $MDS1_VERSION -ge $(version_code 2.10.56) ]]; then
1520 dom_dir=$target_dir/dom_dir
1521 dom_file=$dom_dir/dom_file
1522 flr_dir=$target_dir/flr_dir
1523 flr_file=$flr_dir/flr_file
1526 $LFS setstripe -E 1M -L mdt -E -1 -S 4M $dom_file
1527 dd if=/dev/urandom of=$dom_file bs=1k count=2k
1528 mkdir -p $tmp/src/dom_dir
1529 cp $dom_file $tmp/src/dom_dir
1532 $LFS mirror create -N2 $flr_file
1533 dd if=/dev/urandom of=$flr_file bs=1k count=1
1534 mkdir -p $tmp/src/flr_dir
1535 cp $flr_file $tmp/src/flr_dir
1539 local large_xattr_dir=/mnt/$FSNAME/large_xattr_test_dir
1540 local xattr_file=$large_xattr_dir/large_xattr_file
1542 mkdir $large_xattr_dir
1544 setfattr -n user.fooattr -v $(printf "%c" {1..4096} ) $xattr_file ||
1549 mkdir $tmp/img || return 1
1553 pushd_dir=/mnt/$FSNAME
1554 if [[ $MDSCOUNT -ge 2 ]]; then
1555 pushd_dir=$remote_dir
1556 if [[ "$MDS1_VERSION" -ge $(version_code 2.7.0) ]]; then
1558 BLOCKSIZE=1024 ls -Rni --time-style=+%s >$tmp/img/list2
1564 BLOCKSIZE=1024 ls -Rni --time-style=+%s >$tmp/img/list
1565 find ! -name .lustre -type f -exec sha1sum {} \; |
1566 sort -k 2 >$tmp/img/sha1sums
1568 $LCTL get_param -n version | head -n 1 |
1569 sed -e 's/^lustre: *//' >$tmp/img/commit
1571 [[ "$MDS1_VERSION" -ge $(version_code 2.3.50) ]] ||
1572 $LFS quotaon -ug /mnt/$FSNAME
1573 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1574 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1575 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1576 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1577 | tr -d "*" > $tmp/img/bspace
1578 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1579 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1580 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1581 | tr -d "*" > $tmp/img/ispace
1582 echo $T32_BLIMIT > $tmp/img/blimit
1583 echo $T32_ILIMIT > $tmp/img/ilimit
1585 $MULTIOP /mnt/$FSNAME/orph_file Ouw_c&
1589 debugfs -R "ls /PENDING" ${MDSDEV1:-$MDSDEV}
1590 cp ${MDSDEV1:-$MDSDEV} $tmp/img
1599 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1602 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1603 echo "Data verification failed"
1606 uname -r >$tmp/img/kernel
1607 uname -m >$tmp/img/arch
1609 for ((num=1; num <= $MDSCOUNT; num++)); do
1610 local devname=$(mdsdevname $num)
1614 [[ num -eq 1 ]] && image_name=mdt || image_name=mdt$num
1615 [[ $(facet_fstype $facet) != zfs ]] ||
1616 devname=$(mdsvdevname $num)
1617 dd conv=sparse bs=4k if=$devname of=$tmp/img/$image_name
1620 for ((num=1; num <= $OSTCOUNT; num++)); do
1621 local devname=$(ostdevname $num)
1625 [[ num -eq 1 ]] && image_name=ost || image_name=ost$num
1626 [[ $(facet_fstype $facet) != zfs ]] ||
1627 devname=$(ostdevname $num)
1628 dd conv=sparse bs=4k if=$devname of=$tmp/img/$image_name
1631 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1632 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1635 tar cjvf $dst/disk$version-"$mds1_FSTYPE".tar.bz2 -S *
1640 # run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1643 # The list of applicable tarballs is returned via the caller's
1644 # variable "tarballs".
1647 [ "$CLIENTONLY" ] && skip "Client-only testing"
1649 check_versions || skip "do not upgrade with mismatched client/server"
1651 local node=$(facet_active_host $SINGLEMDS)
1652 local r="do_node $node"
1654 ! $r which "$TUNEFS" && skip_env "tunefs.lustre required on $node"
1656 local IMGTYPE="$mds1_FSTYPE"
1658 [[ -n "$T32IMAGE" ]] && tarballs=$T32IMAGE ||
1659 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 \
1660 -name \'disk*-$IMGTYPE.tar.bz2\')
1662 [ -z "$tarballs" ] && skip "No applicable tarballs found"
1665 t32_test_cleanup() {
1667 local facet=$SINGLEMDS
1670 echo "== cleanup with rc=$rc =="
1671 if $shall_cleanup_lustre; then
1672 umount $tmp/mnt/lustre || rc=$?
1674 if $shall_cleanup_mdt; then
1675 $r $UMOUNT $tmp/mnt/mdt || rc=$?
1677 if $shall_cleanup_mdt1; then
1678 $r $UMOUNT $tmp/mnt/mdt1 || rc=$?
1680 if $shall_cleanup_ost; then
1681 $r $UMOUNT $tmp/mnt/ost || rc=$?
1683 if $shall_cleanup_ost1; then
1684 $r $UMOUNT $tmp/mnt/ost1 || rc=$?
1689 if [[ "$mds1_FSTYPE" == zfs ]]; then
1691 local poolname_list="t32fs-mdt1 t32fs-ost1"
1693 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1695 for poolname in $poolname_list; do
1696 destroy_zpool $facet $poolname
1699 combined_mgs_mds || start_mgs || rc=$?
1703 t32_bits_per_long() {
1705 # Yes, this is not meant to be perfect.
1715 t32_reload_modules() {
1717 local all_removed=false
1720 [ "$mds1_FSTYPE" == zfs ] && do_rpc_nodes $node "service zed stop"
1722 while ((i < 20)); do
1723 echo "Unloading modules on $node: Attempt $i"
1724 do_rpc_nodes $node $LUSTRE_RMMOD "$mds1_FSTYPE" &&
1726 do_rpc_nodes $node check_mem_leak || return 1
1727 if $all_removed; then
1728 do_rpc_nodes $node load_modules
1731 if [ "$mds1_FSTYPE" == zfs ]; then
1732 do_rpc_nodes $node "$ZPOOL status -v"
1737 echo "Unloading modules on $node: Given up"
1741 t32_wait_til_devices_gone() {
1747 echo wait for devices to go
1748 while ((i < 20)); do
1749 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1750 loops=$(do_rpc_nodes $node losetup -a | grep -c t32)
1751 ((devices == 0 && loops == 0)) && return 0
1755 echo "waiting for dev on $node: dev $devices loop $loops given up"
1756 do_rpc_nodes $node "losetup -a"
1757 do_rpc_nodes $node "$LCTL device_list"
1761 t32_verify_quota() {
1768 # LU-2435: if the underlying zfs doesn't support userobj_accounting,
1769 # lustre will estimate the object count usage. This fails quota
1770 # verification in 32b. The object quota usage should be accurate after
1771 # zfs-0.7.0 is released.
1772 [ "$mds1_FSTYPE" == zfs ] && {
1773 local zfs_version=$(do_facet $facet cat /sys/module/zfs/version)
1775 [ $(version_code $zfs_version) -lt $(version_code 0.7.0) ] && {
1776 echo "Skip quota verify for zfs: $zfs_version"
1781 $LFS quota -u $T32_QID -v $mnt
1783 qval=$($LFS quota -v -u $T32_QID $mnt |
1784 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1785 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1787 [ $qval -eq $img_bspace ] || {
1788 echo "bspace, act:$qval, exp:$img_bspace"
1792 qval=$($LFS quota -v -u $T32_QID $mnt |
1793 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1794 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1796 [ $qval -eq $img_ispace ] || {
1797 echo "ispace, act:$qval, exp:$img_ispace"
1801 qval=$($LFS quota -v -u $T32_QID $mnt |
1802 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1803 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1805 [ $qval -eq $img_blimit ] || {
1806 echo "blimit, act:$qval, exp:$img_blimit"
1810 qval=$($LFS quota -v -u $T32_QID $mnt |
1811 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1812 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1814 [ $qval -eq $img_ilimit ] || {
1815 echo "ilimit, act:$qval, exp:$img_ilimit"
1819 set_persistent_param_and_check $facet \
1820 "osd-$mds1_FSTYPE.$fsname-MDT0000.quota_slave.enabled" \
1821 "$fsname.quota.mdt" ug
1823 set_persistent_param_and_check $facet \
1824 "osd-$mds1_FSTYPE.$fsname-OST0000.quota_slave.enabled" \
1825 "$fsname.quota.ost" ug
1828 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1829 bs=1M count=$((img_blimit / 1024)) oflag=sync && {
1830 echo "Write succeed, but expect -EDQUOT"
1833 rm -f $mnt/t32_qf_new
1835 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1837 echo "Create succeed, but expect -EDQUOT"
1840 unlinkmany $mnt/t32_qf_ $img_ilimit
1850 sync_all_data > /dev/null 2>&1 || true
1852 [ "$uuid" != "global" ] || uuid=$mnt
1854 $LFS quota -v "$1" "$2" $mnt |
1855 awk 'BEGIN { num='$spec' } { if ($1 == "'$uuid'") \
1856 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1863 local dne_upgrade=${dne_upgrade:-"no"}
1864 local dom_upgrade=${dom_upgrade:-"no"}
1865 local project_upgrade=${project_upgrade:-"no"}
1866 local ff_convert=${ff_convert:-"no"}
1867 local pfl_upgrade=${pfl_upgrade:-"no"}
1868 local project_quota_upgrade=${project_quota_upgrade:-"no"}
1869 local dom_new_upgrade=${dom_new_upgrade:-"no"}
1870 local flr_upgrade=${flr_upgrade:-"no"}
1871 local shall_cleanup_mdt=false
1872 local shall_cleanup_mdt1=false
1873 local shall_cleanup_ost=false
1874 local shall_cleanup_ost1=false
1875 local shall_cleanup_lustre=false
1876 local mdt2_is_available=false
1877 local ost2_is_available=false
1878 local node=$(facet_active_host $SINGLEMDS)
1879 local r="do_node $node"
1895 local mdt_dev=$tmp/mdt
1896 local mdt2_dev=$tmp/mdt2
1897 local ost_dev=$tmp/ost
1898 local ost2_dev=$tmp/ost2
1902 local pfl_file=$tmp/mnt/lustre/remote_dir/pfl_dir/pfl_file
1903 local flr_file=$tmp/mnt/lustre/remote_dir/flr_dir/flr_file
1904 local dom_file=$tmp/mnt/lustre/remote_dir/dom_dir/dom_file
1905 local quota_dir=$tmp/mnt/lustre/remote_dir/project_quota_dir
1907 combined_mgs_mds || stop_mgs || error "Unable to stop MGS"
1908 trap 'trap - RETURN; t32_test_cleanup' RETURN
1911 nid=$($r $LCTL list_nids | head -1)
1913 mkdir -p $tmp/mnt/lustre || error "mkdir $tmp/mnt/lustre failed"
1914 $r mkdir -p $tmp/mnt/{mdt,mdt1,ost,ost1}
1915 $r tar xjvf $tarball -S -C $tmp || {
1916 error_noexit "Unpacking the disk image tarball"
1919 img_commit=$($r cat $tmp/commit)
1920 img_kernel=$($r cat $tmp/kernel)
1921 img_arch=$($r cat $tmp/arch)
1922 img_bspace=$($r cat $tmp/bspace)
1923 img_ispace=$($r cat $tmp/ispace)
1925 # older images did not have "blimit" and "ilimit" files
1926 # use old values for T32_BLIMIT and T32_ILIMIT
1927 $r test -f $tmp/blimit && img_blimit=$($r cat $tmp/blimit) ||
1929 $r test -f $tmp/ilimit && img_ilimit=$($r cat $tmp/ilimit) ||
1932 echo "Upgrading from $(basename $tarball), created with:"
1933 echo " Commit: $img_commit"
1934 echo " Kernel: $img_kernel"
1935 echo " Arch: $img_arch"
1936 echo "OST version: $(lustre_build_version ost1)"
1938 # The conversion can be made only when both of the following
1939 # conditions are satisfied:
1940 # - ost device img version < 2.3.64
1941 # - ost server version >= 2.5
1942 [ $(version_code $img_commit) -ge $(version_code 2.3.64) -o \
1943 "$OST1_VERSION" -lt $(version_code 2.5.0) ] &&
1946 ! $r test -f $mdt2_dev || mdt2_is_available=true
1947 ! $r test -f $ost2_dev || ost2_is_available=true
1949 if [[ "$mds1_FSTYPE" == zfs ]]; then
1950 echo "== fstype is zfs =="
1953 local poolname_list="t32fs-mdt1 t32fs-ost1"
1955 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1957 for poolname in $poolname_list; do
1959 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
1960 $ZPOOL import -f -d $tmp $poolname"
1963 # upgrade zpool to latest supported features, including
1964 # dnode quota accounting in 0.7.0
1965 $r "$ZPOOL upgrade -a"
1967 mdt_dev=t32fs-mdt1/mdt1
1968 ost_dev=t32fs-ost1/ost1
1969 ! $mdt2_is_available || mdt2_dev=t32fs-mdt2/mdt2
1970 ! $ost2_is_available || ost2_dev=t32fs-ost2/ost2
1971 wait_update_facet $SINGLEMDS "$ZPOOL list |
1972 awk '/^t32fs-mdt1/ { print \\\$1 }'" "t32fs-mdt1" || {
1973 error_noexit "import zfs pool failed"
1976 elif [ "$project_upgrade" != "no" ]; then
1977 echo "== project upgrade =="
1978 ! $r tune2fs -O project $mdt_dev &&
1979 error_noexit "enable project on mdt0 failed" &&
1981 $mdt2_is_available && ! $r tune2fs "-O project" $mdt2_dev &&
1982 error_noexit "enable project on mdt failed" &&
1984 ! $r tune2fs -O project $ost_dev &&
1985 error_noexit "enable project on ost failed" &&
1987 $ost2_is_available && ! $r tune2fs -O project $ost2_dev &&
1988 error_noexit "enable project on ost2 failed" &&
1992 $r $LCTL set_param debug="$PTLDEBUG"
1994 $r $TUNEFS --dryrun $mdt_dev || {
1996 error_noexit "tunefs.lustre before mounting the MDT"
2000 if $mdt2_is_available; then
2001 $r $TUNEFS --dryrun $mdt2_dev || {
2003 error_noexit "tunefs.lustre before mounting the MDT"
2008 if [ "$writeconf" ]; then
2009 echo "== writeconf mdt tunefs and quota =="
2011 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2013 $r $TUNEFS --quota $mdt_dev || {
2015 error_noexit "Enable mdt quota feature"
2019 if [ "$project_quota_upgrade" != "no" ]; then
2020 echo "== enable mdt2 project quota =="
2021 $r $TUNE2FS -O project $mdt_dev || {
2023 error_noexit "tune2fs $mdt_dev failed"
2028 if $mdt2_is_available; then
2029 $r $TUNEFS --quota $mdt2_dev || {
2031 error_noexit "Enable mdt2 quota feature"
2034 if [ "$project_quota_upgrade" != "no" ]; then
2035 echo "== enable mdt2 project quota =="
2036 $r $TUNE2FS -O project $mdt2_dev || {
2039 "tune2fs $mdt2_dev failed"
2046 echo "== replace nids =="
2047 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
2048 [[ "$MGS_VERSION" -ge $(version_code 2.3.59) ]] ||
2049 skip "LU-2200: Cannot run over IB w/o lctl replace_nids "
2050 "(Need MGS version at least 2.3.59)"
2053 local osthost=$(facet_active_host ost1)
2054 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
2057 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2060 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt
2061 $r $LCTL replace_nids $fsname-OST0000 $ostnid || {
2063 error_noexit "replace_nids $fsname-OST0000 $ostnid failed"
2066 if $ost2_is_available; then
2067 $r $LCTL replace_nids $fsname-OST0001 $ostnid || {
2069 error_noexit "replace_nids $fsname-OST0001 $ostnid failed"
2073 $r $LCTL replace_nids $fsname-MDT0000 $nid || {
2075 error_noexit "replace_nids $fsname-MDT0000 $nid failed"
2078 if $mdt2_is_available; then
2079 $r $LCTL replace_nids $fsname-MDT0001 $nid || {
2081 error_noexit "replace_nids $fsname-MDT0001 $nid failed"
2085 $r $UMOUNT $tmp/mnt/mdt
2087 mopts=exclude=$fsname-OST0000
2088 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2093 t32_wait_til_devices_gone $node
2095 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2097 error_noexit "Mounting the MDT"
2100 shall_cleanup_mdt=true
2102 if $mdt2_is_available; then
2103 echo "== mdt2 available =="
2104 mopts=mgsnode=$nid,$mopts
2105 $r $MOUNT_CMD -o $mopts $mdt2_dev $tmp/mnt/mdt1 || {
2107 error_noexit "Mounting the MDT"
2110 shall_cleanup_mdt1=true
2111 echo "== mount new MDT $mdt2_dev =="
2113 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 || {
2114 error_noexit "enable remote dir create failed"
2118 elif [ "$dne_upgrade" != "no" ]; then
2119 local fs2mdsdev=$(mdsdevname 1_2)
2120 local fs2mdsvdev=$(mdsvdevname 1_2)
2122 echo "== mkfs new MDT on ${fs2mdsdev} =="
2123 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2124 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
2127 add $SINGLEMDS $(mds2failover_HOST="" \
2128 mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
2129 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
2130 error_noexit "Mkfs new MDT failed"
2134 [[ "$mds1_FSTYPE" != zfs ]] || import_zpool mds1
2136 $r $TUNEFS --dryrun $fs2mdsdev || {
2137 error_noexit "tunefs.lustre before mounting the MDT"
2141 echo "== mount new MDT....$fs2mdsdev =="
2142 $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
2143 error_noexit "mount mdt1 failed"
2147 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
2148 error_noexit "enable remote dir create failed"
2150 shall_cleanup_mdt1=true
2153 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
2154 error_noexit "Getting MDT UUID"
2157 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
2158 error_noexit "Unexpected MDT UUID: \"$uuid\""
2162 $r $TUNEFS --dryrun $ost_dev || {
2163 error_noexit "tunefs.lustre before mounting the OST"
2167 if $ost2_is_available; then
2168 $r $TUNEFS --dryrun $ost2_dev || {
2169 error_noexit "tunefs.lustre before mounting the OST"
2174 if [ "$writeconf" ]; then
2175 echo "== writeconf and ost quota tunefs =="
2176 mopts=mgsnode=$nid,$writeconf
2177 if [ "$ost1_FSTYPE" == ldiskfs ]; then
2179 $r $TUNEFS --quota $ost_dev || {
2181 error_noexit "Enable ost quota feature"
2185 if [ "$project_quota_upgrade" != "no" ]; then
2186 echo "== enable ost project quota =="
2187 $r $TUNE2FS -O project $ost_dev || {
2189 error_noexit "tune2fs $ost_dev failed"
2194 if $ost2_is_available; then
2195 $r $TUNEFS --quota $ost2_dev || {
2197 error_noexit "Enable ost2 quota feature"
2200 if [ "$project_quota_upgrade" != "no" ]; then
2201 echo "== enable ost2 project quota =="
2202 $r $TUNE2FS -O project $ost2_dev || {
2205 "tune2fs $ost2_dev failed"
2213 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2218 $r $MOUNT_CMD -onomgs -o$mopts $ost_dev $tmp/mnt/ost || {
2219 error_noexit "Mounting the OST"
2223 if $ost2_is_available; then
2224 $r $MOUNT_CMD -onomgs -o$mopts $ost2_dev $tmp/mnt/ost1 || {
2225 error_noexit "Mounting the OST2"
2228 shall_cleanup_ost1=true
2231 shall_cleanup_ost=true
2233 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
2234 error_noexit "Getting OST UUID"
2237 if [ "$uuid" != $fsname-OST0000_UUID ]; then
2238 error_noexit "Unexpected OST UUID: \"$uuid\""
2242 if $ost2_is_available; then
2243 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0001.uuid) ||{
2244 error_noexit "Getting OST1 UUID"
2247 if [ "$uuid" != $fsname-OST0001_UUID ]; then
2248 error_noexit "Unexpected OST1 UUID: \"$uuid\""
2253 if [[ $PERM_CMD =~ "set_param" ]]; then
2254 echo "== perm cmd = '$PERM_CMD' =="
2255 $r $PERM_CMD osc.$fsname-OST0000*.import=connection=$nid || {
2256 error_noexit "Setting OST \"failover.node\""
2259 $r $PERM_CMD mdc.$fsname-MDT0000*.import=connection=$nid || {
2260 error_noexit "Setting MDT \"failover.node\""
2263 $r $PERM_CMD osc.$fsname-OST0000-*.max_dirty_mb=15 || {
2264 error_noexit "Setting \"max_dirty_mb\""
2267 $r $PERM_CMD mdc.$fsname-MDT0000-*.max_rpcs_in_flight=9 || {
2268 error_noexit "Setting \"max_rpcs_in_flight\""
2271 $r $PERM_CMD lov.$fsname-MDT0000-*.stripesize=4M || {
2272 error_noexit "Setting \"lov.stripesize\""
2275 $r $PERM_CMD mdd.$fsname-MDT0000-*.atime_diff=70 || {
2276 error_noexit "Setting \"mdd.atime_diff\""
2280 echo "== perm cmd = '$PERM_CMD' =="
2281 $r $PERM_CMD $fsname-OST0000.failover.node=$nid || {
2282 error_noexit "Setting OST \"failover.node\""
2286 $r $PERM_CMD $fsname-MDT0000.failover.node=$nid || {
2287 error_noexit "Setting MDT \"failover.node\""
2291 $r $PERM_CMD $fsname-OST0000.osc.max_dirty_mb=15 || {
2292 error_noexit "Setting \"max_dirty_mb\""
2295 $r $PERM_CMD $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
2296 error_noexit "Setting \"max_rpcs_in_flight\""
2299 $r $PERM_CMD $fsname-MDT0000.lov.stripesize=4M || {
2300 error_noexit "Setting \"lov.stripesize\""
2303 $r $PERM_CMD $fsname-MDT0000.mdd.atime_diff=70 || {
2304 error_noexit "Setting \"mdd.atime_diff\""
2309 $r $LCTL pool_new $fsname.interop || {
2310 error_noexit "Setting \"interop\""
2314 if [ "$ff_convert" != "no" -a "$ost1_FSTYPE" == ldiskfs ]; then
2315 echo "== ff convert ($ff_convert) =="
2316 $r $LCTL lfsck_start -M $fsname-OST0000 || {
2317 error_noexit "Start OI scrub on OST0"
2321 # The oi_scrub should be on ost1, but for test_32(),
2322 # all on the SINGLEMDS.
2323 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
2324 osd-ldiskfs.$fsname-OST0000.oi_scrub |
2325 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
2326 error_noexit "Failed to get the expected 'completed'"
2330 local UPDATED=$($r $LCTL get_param -n \
2331 osd-ldiskfs.$fsname-OST0000.oi_scrub |
2332 awk '/^updated/ { print $2 }')
2333 [ $UPDATED -ge 1 ] || {
2334 error_noexit "Only $UPDATED objects have been converted"
2339 if [[ "$dne_upgrade" != "no" ]] || $mdt2_is_available; then
2340 echo "== dne upgrate ($dne_upgrade) or mdt2 available =="
2341 if [[ $PERM_CMD =~ "set_param" ]]; then
2342 $r $PERM_CMD mdc.$fsname-MDT0001*.import=connection=$nid || {
2343 error_noexit "Setting MDT1 \"failover.node\""
2347 $r $PERM_CMD mdc.$fsname-MDT0001-*.max_rpcs_in_flight=9 || {
2348 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
2351 $r $PERM_CMD lov.$fsname-MDT0001-*.stripesize=4M || {
2352 error_noexit "Setting MDT1 \"lov.stripesize\""
2356 $r $PERM_CMD $fsname-MDT0001.failover.node=$nid || {
2357 error_noexit "Setting MDT1 \"failover.node\""
2360 $r $PERM_CMD $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
2361 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
2364 $r $PERM_CMD $fsname-MDT0001.lov.stripesize=4M || {
2365 error_noexit "Setting MDT1 \"lov.stripesize\""
2371 if [ "$writeconf" ]; then
2372 echo "== writeconf and client mount =="
2373 $MOUNT_CMD $nid:/$fsname -o user_xattr $tmp/mnt/lustre || {
2374 error_noexit "Mounting the client"
2378 shall_cleanup_lustre=true
2379 $r $LCTL set_param debug="$PTLDEBUG"
2381 t32_verify_quota $SINGLEMDS $fsname $tmp/mnt/lustre || {
2382 error_noexit "verify quota failed"
2386 if $r test -f $tmp/list; then
2387 echo "== list verification =="
2389 # There is not a Test Framework API to copy files to or
2390 # from a remote node.
2392 # LU-2393 - do both sorts on same node to ensure locale
2394 local list_file=$tmp/list
2396 if $mdt2_is_available; then
2397 if [[ -d $tmp/mnt/lustre/striped_dir_old ]] &&
2398 $r test -f $tmp/list2; then
2399 list_file=$tmp/list2
2400 pushd $tmp/mnt/lustre/striped_dir_old
2402 pushd $tmp/mnt/lustre/remote_dir
2405 pushd $tmp/mnt/lustre
2407 $r cat $list_file | sort -k 6 >$tmp/list.orig
2408 BLOCKSIZE=1024 ls -Rni --time-style=+%s | sort -k 6 |
2409 sed 's/\. / /' >$tmp/list || {
2415 # 32-bit and 64-bit clients use different algorithms to
2416 # convert FIDs into inode numbers. Hence, remove the
2417 # inode numbers from the lists, if the original list was
2418 # created on an architecture with different number of
2421 if [ $(t32_bits_per_long $(uname -m)) != \
2422 $(t32_bits_per_long $img_arch) ]; then
2423 echo "Different number of bits per \"long\"" \
2424 "from the disk image"
2425 for list in list.orig list; do
2426 sed -i -e 's/^[0-9]\+[ \t]\+//' \
2430 if ! diff -ub $tmp/list.orig $tmp/list; then
2431 error_noexit "list verification failed"
2435 echo "== list verification skipped =="
2438 #non-dom upgrade to dom
2439 if [ "$dom_upgrade" != "no" ]; then
2440 echo "== dom upgrade =="
2441 $LFS setstripe -E 1M -L mdt -E EOF $tmp/mnt/lustre/dom || {
2442 error_noexit "Verify DoM creation"
2445 [ $($LFS getstripe -L $tmp/mnt/lustre/dom) == "mdt" ] || {
2446 error_noexit "Verify a DoM file"
2449 dd if=/dev/urandom of=$tmp/mnt/lustre/dom bs=4096 \
2450 count=1 conv=fsync || {
2451 error_noexit "Cannot write to DoM file"
2454 [ $(stat -c%s $tmp/mnt/lustre/dom) == 4096 ] || {
2455 error_noexit "DoM: bad size after write"
2458 rm $tmp/mnt/lustre/dom
2460 set_persistent_param_and_check mds \
2461 "lod.*MDT0000*.dom_stripesize" \
2462 "$fsname-MDT0000.lod.dom_stripesize" 0 || {
2463 error_noexit "Changing \"dom_stripesize\""
2469 #$LFS setstripe -E 1M -L mdt -E -1 -S 4M $dom_file
2470 if [ "$dom_new_upgrade" != "no" ]; then
2471 if ! $mdt2_is_available; then
2472 dom_file=$tmp/mnt/lustre/dom_dir/dom_file
2474 echo "== check DoM file can be accessed =="
2475 [ $($LFS getstripe -I1 -L $dom_file) == "mdt" ] || {
2476 error_noexit "Verify a DoM file"
2479 [ $(stat -c%s $dom_file) == $((2 * 1024 * 1024)) ] || {
2480 error_noexit "DoM: bad size after write"
2485 if [ "$flr_upgrade" != "no" ]; then
2486 if ! $mdt2_is_available; then
2487 flr_file=$tmp/mnt/lustre/flr_dir/flr_file
2489 local mirror_count=$($LFS getstripe -N $flr_file)
2490 echo "== check FLR file =="
2491 [ $mirror_count == 2 ] || {
2492 error_noexit "FLR mirror count wrong"
2497 if [ "$dne_upgrade" != "no" ]; then
2498 echo "== dne upgrade striped_dir =="
2499 $LFS mkdir -i 1 -c2 $tmp/mnt/lustre/striped_dir || {
2500 error_noexit "set striped dir failed"
2503 $LFS setdirstripe -D -c2 $tmp/mnt/lustre/striped_dir
2505 pushd $tmp/mnt/lustre
2506 tar -c --exclude=./striped_dir \
2507 --exclude=./striped_dir_old \
2508 --exclude=./remote_dir -f - .|
2509 tar -xvf - -C striped_dir 1>/dev/null || {
2510 error_noexit "cp to striped dir failed"
2516 if [ "$pfl_upgrade" != "no" ]; then
2523 echo "== check PFL file =="
2524 if ! $mdt2_is_available; then
2525 pfl_file=$tmp/mnt/lustre/pfl_dir/pfl_file
2527 comp_cnt=$($LFS getstripe --component-count $pfl_file)
2528 [ $comp_cnt == 2 ] || {
2529 error_noexit "wrong comp_cnt $comp_cnt"
2533 comp_size=$($LFS getstripe -I1 -E $pfl_file)
2534 [ $comp_size == 2097152 ] || {
2535 error_noexit "wrong component size $comp_size"
2539 comp_id=$($LFS getstripe -I1 -i $pfl_file)
2540 [ $comp_id == 0 ] || {
2541 error_noexit "wrong comp id $comp_id"
2545 comp_id=$($LFS getstripe -I2 -i $pfl_file)
2546 [ $comp_id -eq 1 ] || {
2547 error_noexit "wrong comp id $comp_id"
2551 stripe_size=$($LFS getstripe -I1 -S $pfl_file)
2552 [ $stripe_size -eq 1048576 ] || {
2553 error_noexit "wrong stripe size $stripe_size"
2557 stripe_size=$($LFS getstripe -I2 -S $pfl_file)
2558 [ $comp_size -eq 2097152 ] || {
2559 error_noexit "wrong component size $comp_size"
2564 local large_xattr_dir=$tmp/mnt/lustre/large_xattr_test_dir
2566 if [[ -d $large_xattr_dir ]]; then
2567 echo "== check Large EA =="
2568 local xattr_file=$large_xattr_dir/large_xattr_file
2570 xattr_val=$(getfattr --only-values\
2571 -n user.fooattr $xattr_file) || {
2572 error_noexit "Large EA cannot be read"
2576 [[ $xattr_val == $(printf "%c" {1..4096}) ]] || {
2577 error_noexit "Wrong large EA value"
2582 if [ "$project_quota_upgrade" != "no" ]; then
2583 if ! $mdt2_is_available; then
2584 quota_dir=$tmp/mnt/lustre/project_quota_dir
2588 echo "== check Project Quota =="
2589 hardlimit=$(getquota -p $T32_PRJID global 3 \
2591 [ $hardlimit == $T32_PROLIMIT ] || {
2592 error_noexit "wrong hardlimit $hardlimit"
2597 # If it is upgrade from DNE (2.5), then rename the remote dir,
2598 # which is created in 2.5 to striped dir.
2599 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]]; then
2600 echo "== rename remote_dir =="
2601 stripe_index=$($LFS getdirstripe -i \
2602 $tmp/mnt/lustre/remote_dir)
2604 [[ $stripe_index -eq 1 ]] || {
2605 error_noexit "get index \"$stripe_index\"" \
2606 "from remote dir failed"
2609 mv $tmp/mnt/lustre/remote_dir \
2610 $tmp/mnt/lustre/striped_dir/ || {
2611 error_noexit "mv remote dir failed"
2616 # If it is upgraded from DNE (2.7), then move the striped dir
2617 # which was created in 2.7 to the new striped dir.
2618 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]] &&
2619 echo "== rename striped dir =="
2620 [[ -d $tmp/mnt/lustre/striped_dir_old ]]; then
2621 stripe_count=$($LFS getdirstripe -c \
2622 $tmp/mnt/lustre/striped_dir_old)
2623 [[ $stripe_count -eq 2 ]] || {
2624 error_noexit "get count $stripe_count" \
2625 "from striped dir failed"
2628 mv $tmp/mnt/lustre/striped_dir_old \
2629 $tmp/mnt/lustre/striped_dir/ || {
2630 error_noexit "mv striped dir failed"
2636 $r $LCTL set_param -n osd*.*.force_sync=1
2637 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
2638 error_noexit "dd failed"
2641 rm -rf $tmp/mnt/lustre/tmp_file || {
2642 error_noexit "rm failed"
2646 if $r test -f $tmp/sha1sums; then
2647 echo "== checking sha1sums =="
2648 # LU-2393 - do both sorts on same node to ensure locale
2650 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
2651 if [[ "$dne_upgrade" != "no" ]]; then
2652 if [[ -d $tmp/mnt/lustre/striped_dir/remote_dir ]]; then
2653 pushd $tmp/mnt/lustre/striped_dir/remote_dir
2655 pushd $tmp/mnt/lustre/striped_dir
2657 elif [[ "$pfl_upgrade" != "no" ]] ||
2658 [[ "$flr_upgrade" != "no" ]] ||
2659 [[ "$dom_new_upgrade" != "no" ]] ||
2660 [[ "$project_quota_upgrade" != "no" ]] ||
2661 [[ -d $tmp/mnt/lustre/remote_dir ]]; then
2662 pushd $tmp/mnt/lustre/remote_dir
2664 pushd $tmp/mnt/lustre
2667 find ! -path "*remote_dir*" ! -path "*striped_dir*" \
2668 ! -name .lustre -type f -exec sha1sum {} \; |
2669 sort -k 2 >$tmp/sha1sums || {
2671 error_noexit "sha1sum"
2677 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
2678 echo "==** find $save files start **=="
2679 find $save -type f -ls
2680 echo "==** find $save files finish **=="
2681 error_noexit "sha1sum verification failed"
2685 # if upgrade from DNE(2.5), then check remote directory
2686 # if upgrade from DNE(2.7), then check striped directory
2687 if $mdt2_is_available &&
2688 [[ "$dne_upgrade" != "no" ]]; then
2689 echo "== mdt2 upgrade sha1sum check =="
2690 local new_dir="$tmp/mnt/lustre/striped_dir"
2691 local striped_dir_old="$new_dir/striped_dir_old"
2693 local dir_list="$new_dir/remote_dir"
2694 [[ ! -d $triped_dir_old ]] ||
2695 dir_list+=" $striped_dir_old"
2697 for dir in $dir_list; do
2698 echo "== dir = $dir =="
2700 find ! -name .lustre -type f \
2701 -exec sha1sum {} \; |
2702 sort -k 2 >$tmp/sha1sums || {
2705 error_noexit "sha1sum"
2709 if ! diff -ub $tmp/sha1sums.orig \
2712 error_noexit "sha1sum $dir" \
2719 echo "== sha1sum verification skipped =="
2722 # PFL write test after sha1sum check
2723 if [ "$pfl_upgrade" != "no" ]; then
2724 local rw_len=$((3 * 1034 * 1024))
2726 if ! $mdt2_is_available; then
2727 pfl_file=$tmp/mnt/lustre/pfl_dir/pfl_file
2729 small_write $pfl_file $rw_len || {
2730 error_noexit "PFL RW Failed"
2735 if [ "$dne_upgrade" != "no" ]; then
2736 rm -rf $tmp/mnt/lustre/striped_dir || {
2737 error_noexit "remove remote dir failed"
2742 # migrate files/dirs to remote MDT, then move them back
2743 if [ "$MDS1_VERSION" -ge $(version_code 2.7.50) -a \
2744 $dne_upgrade != "no" ]; then
2745 $r $LCTL set_param -n \
2746 mdt.${fsname}*.enable_remote_dir=1 2>/dev/null
2748 echo "== test migration =="
2749 pushd $tmp/mnt/lustre
2750 for dir in $(find ! -name .lustre ! -name . -type d); do
2751 mdt_index=$($LFS getdirstripe -i $dir)
2752 stripe_cnt=$($LFS getdirstripe -c $dir)
2753 if [ $mdt_index = 0 -a $stripe_cnt -le 1 ]; then
2754 $LFS migrate -m 1 $dir || {
2756 error_noexit "migrate MDT1 failed"
2762 for dir in $(find ! -name . ! -name .lustre -type d); do
2763 mdt_index=$($LFS getdirstripe -i $dir)
2764 stripe_cnt=$($LFS getdirstripe -c $dir)
2765 if [ $mdt_index = 1 -a $stripe_cnt -le 1 ]; then
2766 $LFS migrate -m 0 $dir || {
2768 error_noexit "migrate MDT0 failed"
2777 # When adding new data verification tests, please check for
2778 # the presence of the required reference files first, like
2779 # the "sha1sums" and "list" tests above, to avoid the need to
2780 # regenerate every image for each test addition.
2783 nrpcs_orig=$($LCTL get_param \
2784 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
2785 error_noexit "Getting \"max_rpcs_in_flight\""
2788 nrpcs=$((nrpcs_orig + 5))
2790 set_persistent_param_and_check client \
2791 "mdc.$fsname-MDT0000*.max_rpcs_in_flight" \
2792 "$fsname-MDT0000.mdc.max_rpcs_in_flight" $nrpcs || {
2793 error_noexit "Changing \"max_rpcs_in_flight\""
2797 umount $tmp/mnt/lustre || {
2798 error_noexit "Unmounting the client"
2801 shall_cleanup_lustre=false
2803 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
2804 error_noexit "Mounting the client"
2808 if [ "$mds1_FSTYPE" == ldiskfs -a \
2809 "$project_upgrade" != "no" ]; then
2810 ! $LFS project -d -p 1 $tmp/mnt/lustre/* &&
2811 error_noexit "set project failed" &&
2815 [[ $(do_facet mds1 pgrep orph_.*-MDD | wc -l) == 0 ]] ||
2816 error "MDD orphan cleanup thread not quit"
2818 umount $tmp/mnt/lustre || {
2819 error_noexit "Unmounting the client"
2823 if [[ "$dne_upgrade" != "no" ]] || $mdt2_is_available; then
2824 $r $UMOUNT $tmp/mnt/mdt1 || {
2825 error_noexit "Unmounting the MDT2"
2828 if [[ "$mds1_FSTYPE" == zfs ]]; then
2829 $r "$ZPOOL export t32fs-mdt2"
2831 shall_cleanup_mdt1=false
2834 $r $UMOUNT $tmp/mnt/mdt || {
2835 error_noexit "Unmounting the MDT"
2838 if [[ "$mds1_FSTYPE" == zfs ]]; then
2839 $r "$ZPOOL export t32fs-mdt1"
2841 shall_cleanup_mdt=false
2843 $r $UMOUNT $tmp/mnt/ost || {
2844 error_noexit "Unmounting the OST"
2847 if $ost2_is_available; then
2848 $r $UMOUNT $tmp/mnt/ost1 || {
2849 error_noexit "Unmounting the OST1"
2852 shall_cleanup_ost1=false
2855 if [[ $ost1_FSTYPE == zfs ]]; then
2856 $r "$ZPOOL export t32fs-ost1"
2857 if $ost2_is_available; then
2858 $r "$ZPOOL export t32fs-ost2"
2861 shall_cleanup_ost=false
2863 t32_reload_modules $node || {
2864 error_noexit "Reloading modules"
2868 if [[ "$mds1_FSTYPE" == zfs ]]; then
2869 local poolname=t32fs-mdt1
2871 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
2872 $ZPOOL import -f -d $tmp $poolname"
2874 # upgrade zpool to latest supported features,
2875 # including dnode quota accounting in 0.7.0
2876 $r "$ZPOOL upgrade $poolname"
2879 # mount a second time to make sure we didnt leave upgrade flag on
2880 $r $TUNEFS --dryrun $mdt_dev || {
2882 error_noexit "tunefs.lustre before remounting the MDT"
2886 mopts=exclude=$fsname-OST0000
2887 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2890 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2891 error_noexit "Remounting the MDT"
2894 shall_cleanup_mdt=true
2904 for tarball in $tarballs; do
2905 banner "testing $tarball upgrade"
2906 t32_test $tarball || let "rc += $?"
2910 run_test 32a "Upgrade (not live)"
2918 for tarball in $tarballs; do
2919 banner "testing $tarball upgrade with writeconf"
2920 t32_test $tarball writeconf || let "rc += $?"
2924 run_test 32b "Upgrade with writeconf"
2932 for tarball in $tarballs; do
2933 banner "testing $tarball upgrade with DNE"
2935 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
2939 run_test 32c "dne upgrade test"
2947 for tarball in $tarballs; do
2948 banner "testing $tarball upgrade with ff and project upgrade"
2949 project_upgrade="no"
2950 [[ "$MDS1_VERSION" -ge $(version_code 2.13.54) ]] &&
2951 [[ "$tarball" =~ "disk2_4-ldiskfs" ]] &&
2952 project_upgrade="yes"
2953 project_upgrade=$project_upgrade ff_convert=yes t32_test \
2958 run_test 32d "convert ff and project quota upgrade test"
2961 [[ "$MDS1_VERSION" -ge $(version_code 2.10.56) ]] ||
2962 skip "Need MDS version at least 2.10.56"
2969 for tarball in $tarballs; do
2970 [[ "$tarball" =~ "2_9" ]] ||
2971 { echo "skip $(basename $tarball)"; continue; }
2973 banner "testing $tarball upgrade with DoM"
2974 dom_upgrade=yes t32_test $tarball writeconf || let "rc += $?"
2978 run_test 32e "dom upgrade test"
2981 [[ $MDS1_VERSION -ge $(version_code 2.10.56) ]] ||
2982 skip "Need MDS version at least 2.10.56"
2989 for tarball in $tarballs; do
2990 [[ "$tarball" =~ "2_10" ]] ||
2991 { echo "skip $(basename $tarball)"; continue; }
2992 pfl_upgrade=yes project_quota_upgrade=yes \
2993 t32_test $tarball writeconf || let "rc += $?"
2997 run_test 32f "pfl upgrade test"
3000 [[ $MDS1_VERSION -ge $(version_code 2.10.56) ]] ||
3001 skip "Need MDS version at least 2.10.56"
3008 for tarball in $tarballs; do
3009 [[ $tarball =~ "2_12" ]] ||
3010 { echo "skip $(basename $tarball)"; continue; }
3011 flr_upgrade=yes dom_new_upgrade=yes \
3012 t32_test $tarball writeconf || let "rc += $?"
3016 run_test 32g "flr/dom upgrade test"
3018 test_33a() { # bug 12333, was test_33
3019 local FSNAME2=test-$testnum
3021 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3024 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
3026 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
3027 local dev=${SINGLEMDS}_dev
3028 local MDSDEV=${!dev}
3030 is_blkdev $SINGLEMDS $MDSDEV &&
3031 skip_env "mixed loopback and real device not working"
3034 local fs2mdsdev=$(mdsdevname 1_2)
3035 local fs2ostdev=$(ostdevname 1_2)
3036 local fs2mdsvdev=$(mdsvdevname 1_2)
3037 local fs2ostvdev=$(ostvdevname 1_2)
3039 if [ "$mds1_FSTYPE" == ldiskfs ]; then
3040 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
3043 if combined_mgs_mds; then
3044 local mgs_flag="--mgs"
3047 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --fsname=${FSNAME2} \
3048 --reformat $mgs_flag $mkfsoptions $fs2mdsdev $fs2mdsvdev ||
3050 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
3051 --fsname=${FSNAME2} --index=0x1fff --reformat $fs2ostdev \
3052 $fs2ostvdev || exit 10
3054 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
3055 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
3057 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3058 do_facet mgs "$PERM_CMD timeout=200" ||
3059 error "$PERM_CMD timeout=200 failed"
3061 do_facet mgs "$PERM_CMD $FSNAME2.sys.timeout=200" ||
3062 error "$PERM_CMD $FSNAME2.sys.timeout=200 failed"
3064 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
3065 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
3068 cp /etc/hosts $MOUNT2/ || error "copy /etc/hosts $MOUNT2/ failed"
3069 $LFS getstripe $MOUNT2/hosts ||
3070 error "$LFS getstripe $MOUNT2/hosts failed"
3074 if (( "$MGS_VERSION" >= $(version_code 2.15.51) &&
3075 "$MDS1_VERSION" >= $(version_code 2.15.51) )); then
3076 # test lctl del_ost on large index
3077 do_facet mgs "$LCTL del_ost -t ${FSNAME2}-OST1fff" ||
3078 error "del_ost failed with $?"
3079 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 ||
3080 error "$MOUNT_CMD failed"
3082 $LFS df | grep -q OST1fff &&
3083 error "del_ost did not remove OST1fff!"
3089 cleanup_nocli || error "cleanup_nocli failed with $?"
3091 run_test 33a "Mount ost with a large index number"
3093 test_33b() { # was test_34
3096 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
3097 # Drop lock cancelation reply during umount
3098 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
3099 do_facet client $LCTL set_param fail_loc=0x80000304
3100 #lctl set_param debug=-1
3101 umount_client $MOUNT
3102 cleanup || error "cleanup failed with $?"
3104 run_test 33b "Drop cancel during umount"
3107 (( MDS1_VERSION >= $(version_code 2.15.57) )) ||
3108 skip "Need MDS version at least 2.15.57"
3109 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3110 local tstid=${TSTID:-"$(id -u $TSTUSR)"}
3112 local qpool="qpool1"
3114 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
3116 local fs2mdsdev=$(mdsdevname 1_2)
3117 local fs2ostdev=$(ostdevname 1_2)
3118 local fs2mdsvdev=$(mdsvdevname 1_2)
3119 local fs2ostvdev=$(ostvdevname 1_2)
3121 if [ "$mds1_FSTYPE" == ldiskfs ]; then
3122 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
3125 if combined_mgs_mds; then
3126 local mgs_flag="--mgs"
3130 stack_trap unload_modules_conf
3132 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --fsname=${FSNAME} \
3133 --reformat $mgs_flag $mkfsoptions $fs2mdsdev $fs2mdsvdev ||
3135 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
3136 --fsname=${FSNAME} --index=0x7c6 --reformat $fs2ostdev \
3137 $fs2ostvdev || exit 10
3140 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
3141 stack_trap "stop fs2mds -f"
3142 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
3143 stack_trap "stop fs2ost -f"
3145 mount_client $MOUNT || error "client start failed"
3146 stack_trap "umount_client $MOUNT"
3147 mkdir_on_mdt0 $DIR/$tdir || error "cannot create $DIR/$tdir"
3148 chmod 0777 $DIR/$tdir || error "chown failed"
3149 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3150 do_facet mgs $PERM_CMD \
3151 osd-*.$FSNAME-OST*.quota_slave.enable=$QUOTA_TYPE
3153 do_facet mgs $PERM_CMD $FSNAME.quota.ost=$QUOTA_TYPE ||
3154 error "set ost quota type failed"
3157 local old_MDSCOUNT=$MDSCOUNT
3159 stack_trap "MDSCOUNT=$old_MDSCOUNT"
3161 pool_add $qpool || error "pool_add failed"
3162 pool_add_targets $qpool 0x7c6
3164 $LFS setquota -u $tstid -B20M -b 0 $MOUNT
3165 $LFS setquota -g $tstid -B20M -b 0 $MOUNT
3166 $LFS setquota -u $tstid -B20M -b 0 --pool $qpool $MOUNT
3167 $LFS setquota -g $tstid -B20M -b 0 --pool $qpool $MOUNT
3169 for i in {1..10}; do
3170 runas -u $tstid -g $tstid dd if=/dev/zero of=$DIR/$tdir/f1 \
3171 bs=1M count=30 oflag=direct
3177 #umount_client $MOUNT || error "client start failed"
3180 #cleanup_nocli || error "cleanup_nocli failed with $?"
3182 run_test 33c "Mount ost with a large index number"
3186 do_facet client "bash runmultiop_bg_pause $DIR/file O_c"
3187 manual_umount_client
3189 do_facet client killall -USR1 multiop
3190 if [ $rc -eq 0 ]; then
3191 error "umount not fail!"
3194 cleanup || error "cleanup failed with rc $?"
3196 run_test 34a "umount with opened file should be fail"
3200 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3201 stop_mds || error "Unable to stop MDS"
3203 manual_umount_client --force || error "mtab after failed umount with $?"
3205 cleanup || error "cleanup failed with $?"
3207 run_test 34b "force umount with failed mds should be normal"
3211 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3212 stop_ost || error "Unable to stop OST1"
3214 manual_umount_client --force || error "mtab after failed umount with $?"
3216 cleanup || error "cleanup failed with $?"
3218 run_test 34c "force umount with failed ost should be normal"
3220 test_35a() { # bug 12459
3223 DBG_SAVE="`$LCTL get_param -n debug`"
3224 $LCTL set_param debug="ha"
3226 log "Set up a fake failnode for the MDS"
3228 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
3229 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
3231 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3232 do_facet mgs "$PERM_CMD \
3233 mdc.*${device}*.import=connection=$(h2nettype $FAKENID)" ||
3234 error "Setting mdc.*${device}*.import=connection=\
3235 $(h2nettype $FAKENID) failed."
3237 do_facet mgs "$PERM_CMD \
3238 ${device}.failover.node=$(h2nettype $FAKENID)" ||
3239 error "Setting ${device}.failover.node=\
3240 $(h2nettype $FAKENID) failed."
3242 log "Wait for RECONNECT_INTERVAL seconds (10s)"
3245 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
3248 log "Stopping the MDT: $device"
3249 stop_mdt 1 || error "MDT0 stop fail"
3251 df $MOUNT > /dev/null 2>&1 &
3253 log "Restarting the MDT: $device"
3254 start_mdt 1 || error "MDT0 start fail"
3255 log "Wait for df ($DFPID) ... "
3258 $LCTL set_param debug="$DBG_SAVE"
3260 # retrieve from the log the first server that the client tried to
3261 # contact after the connection loss
3262 $LCTL dk $TMP/lustre-log-$TESTNAME.log
3263 NEXTCONN=`awk "/${MSG}/ {start = 1;}
3264 /import_select_connection.*$device-mdc.* using connection/ {
3266 if (\\\$NF ~ /$FAKENID/)
3272 }" $TMP/lustre-log-$TESTNAME.log`
3273 [ "$NEXTCONN" != "0" ] &&
3274 error "Tried to connect to ${NEXTCONN} not last active server"
3275 cleanup || error "cleanup failed with $?"
3276 # remove nid settings
3277 writeconf_or_reformat
3279 run_test 35a "Reconnect to the last active server first"
3281 test_35b() { # bug 18674
3282 remote_mds || skip "local MDS"
3286 $LCTL set_param debug="ha"
3288 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
3291 log "Set up a fake failnode for the MDS"
3293 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
3294 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
3296 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3297 do_facet mgs "$PERM_CMD \
3298 mdc.*${device}*.import=connection=$(h2nettype $FAKENID)" ||
3299 error "Set mdc.*${device}*.import=connection=\
3300 $(h2nettype $FAKENID) failed"
3302 do_facet mgs "$PERM_CMD \
3303 ${device}.failover.node=$(h2nettype $FAKENID)" ||
3304 error "Set ${device}.failover.node=\
3305 $(h2nettype $FAKENID) failed"
3308 local at_max_saved=0
3309 # adaptive timeouts may prevent seeing the issue
3310 if at_is_enabled; then
3311 at_max_saved=$(at_max_get mds)
3312 at_max_set 0 mds client
3315 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
3317 log "Injecting EBUSY on MDS"
3318 # Setting OBD_FAIL_MDS_RESEND=0x136
3319 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" ||
3320 error "unable to set param fail_loc=0x80000136"
3322 $LCTL set_param mdc.${FSNAME}*.stats=clear
3324 log "Creating a test file and stat it"
3325 touch $MOUNT/$tdir/$tfile || error "touch $MOUNT/$tdir/$tfile failed"
3326 stat $MOUNT/$tdir/$tfile
3328 log "Stop injecting EBUSY on MDS"
3329 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" ||
3330 error "unable to set param fail_loc=0"
3331 rm -f $MOUNT/$tdir/$tfile || error "remove $MOUNT/$tdir/$tfile failed"
3334 # restore adaptive timeout
3335 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
3337 $LCTL dk $TMP/lustre-log-$TESTNAME.log
3339 CONNCNT=$($LCTL get_param mdc.${FSNAME}*.stats |
3340 awk '/mds_connect/{print $2}')
3342 # retrieve from the log if the client has ever tried to
3343 # contact the fake server after the loss of connection
3344 FAILCONN=`awk "BEGIN {ret = 0;}
3345 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
3347 if (\\\$NF ~ /$FAKENID/) {
3352 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
3354 [ "$FAILCONN" == "0" ] &&
3355 error "The client reconnection has not been triggered"
3356 [ "$FAILCONN" == "2" ] &&
3357 error "Primary server busy, client reconnect to failover failed"
3360 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
3361 # Reconnects are supposed to be rate limited to one every 5s
3362 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] &&
3363 error "Too many reconnects $CONNCNT"
3365 cleanup || error "cleanup failed with $?"
3366 # remove nid settings
3367 writeconf_or_reformat
3369 run_test 35b "Continue reconnection retries, if the active server is busy"
3372 [ $OSTCOUNT -lt 2 ] && skip_env "needs >= 2 OSTs"
3374 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] ||
3378 local FSNAME2=test1234
3379 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3381 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
3383 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
3384 is_blkdev $SINGLEMDS $MDSDEV &&
3385 skip_env "mixed loopback and real device not working"
3388 local fs2mdsdev=$(mdsdevname 1_2)
3389 local fs2ostdev=$(ostdevname 1_2)
3390 local fs3ostdev=$(ostdevname 2_2)
3391 local fs2mdsvdev=$(mdsvdevname 1_2)
3392 local fs2ostvdev=$(ostvdevname 1_2)
3393 local fs3ostvdev=$(ostvdevname 2_2)
3396 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
3397 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
3398 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
3399 # different one than the default value here.
3400 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
3401 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
3402 add fs3ost $(mkfs_opts ost2 ${fs3ostdev}) --mgsnode=$MGSNID \
3403 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
3405 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
3406 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
3407 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
3408 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
3409 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
3411 sleep 5 # until 11778 fixed
3413 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || error "dd failed"
3415 BKTOTAL=$($LCTL get_param -n obdfilter.*.kbytestotal |
3416 awk 'BEGIN{total=0}; {total+=$1}; END{print total}')
3417 BKFREE=$($LCTL get_param -n obdfilter.*.kbytesfree |
3418 awk 'BEGIN{free=0}; {free+=$1}; END{print free}')
3419 BKAVAIL=$($LCTL get_param -n obdfilter.*.kbytesavail |
3420 awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}')
3421 STRING=$(df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}')
3422 DFTOTAL=$(echo $STRING | cut -d, -f1)
3423 DFUSED=$(echo $STRING | cut -d, -f2)
3424 DFAVAIL=$(echo $STRING | cut -d, -f3)
3425 DFFREE=$(($DFTOTAL - $DFUSED))
3427 ALLOWANCE=$((64 * $OSTCOUNT))
3429 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
3430 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
3431 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
3434 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
3435 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
3436 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
3439 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
3440 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
3441 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
3446 stop fs3ost -f || error "unable to stop OST3"
3447 stop fs2ost -f || error "unable to stop OST2"
3448 stop fs2mds -f || error "unable to stop second MDS"
3449 unload_modules_conf || error "unable unload modules"
3452 run_test 36 "df report consistency on OSTs with different block size"
3455 local mntpt=$(facet_mntpt $SINGLEMDS)
3456 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
3457 local mdsdev_sym="$TMP/sym_mdt.img"
3458 local opts=$MDS_MOUNT_OPTS
3461 if [ "$mds1_FSTYPE" != ldiskfs ]; then
3462 skip "ldiskfs only test"
3465 echo "MDS : $mdsdev"
3466 echo "SYMLINK : $mdsdev_sym"
3467 do_facet $SINGLEMDS rm -f $mdsdev_sym
3469 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
3471 echo "mount symlink device - $mdsdev_sym"
3473 if ! do_facet $SINGLEMDS test -b $mdsdev; then
3474 opts=$(csa_add "$opts" -o loop)
3478 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
3479 $mdsdev_sym $mntpt 2>&1)
3482 echo mount_op=$mount_op
3484 do_facet $SINGLEMDS "$UMOUNT $mntpt && rm -f $mdsdev_sym"
3486 if $(echo $mount_op | grep -q "unable to set tunable"); then
3487 error "set tunables failed for symlink device"
3490 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
3492 run_test 37 "verify set tunables works for symlink device"
3494 test_38() { # bug 14222
3495 local mntpt=$(facet_mntpt $SINGLEMDS)
3500 local SRC="/etc /bin"
3501 local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT)
3503 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
3504 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3505 tar cf - $FILES | tar xf - -C $DIR/$tdir ||
3506 error "copying $SRC to $DIR/$tdir"
3508 umount_client $MOUNT || error "umount_client $MOUNT failed"
3509 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3510 stop_mds || error "Unable to stop MDS"
3511 log "delete lov_objid file on MDS"
3513 mount_fstype $SINGLEMDS || error "mount MDS failed (1)"
3515 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid; rm $mntpt/lov_objid"
3517 unmount_fstype $SINGLEMDS || error "umount failed (1)"
3519 # check create in mds_lov_connect
3520 start_mds || error "unable to start MDS"
3521 mount_client $MOUNT || error "mount_client $MOUNT failed"
3523 [ $V ] && log "verifying $DIR/$tdir/$f"
3524 diff -q $f $DIR/$tdir/$f || ERROR=y
3526 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3527 if [ "$ERROR" = "y" ]; then
3528 # check it's updates in sync
3529 umount_client $MOUNT
3531 mount_fstype $SIGNLEMDS
3532 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3533 unmount_fstype $SINGLEMDS
3534 error "old and new files are different after connect" || true
3536 touch $DIR/$tdir/f2 || error "f2 file create failed"
3538 # check it's updates in sync
3539 umount_client $MOUNT || error "second umount_client $MOUNT failed"
3542 mount_fstype $SINGLEMDS || error "mount MDS failed (3)"
3544 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3545 do_facet $SINGLEMDS dd if=/dev/zero of=$mntpt/lov_objid.clear count=8
3547 unmount_fstype $SINGLEMDS || error "umount failed (3)"
3549 start_mds || error "unable to start MDS"
3550 mount_client $MOUNT || error "mount_client $MOUNT failed"
3552 [ $V ] && log "verifying $DIR/$tdir/$f"
3553 diff -q $f $DIR/$tdir/$f || ERROR=y
3555 touch $DIR/$tdir/f3 || error "f3 file create failed"
3556 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3557 umount_client $MOUNT || error "third umount_client $MOUNT failed"
3559 mount_fstype $SINGLEMDS || error "mount MDS failed (4)"
3560 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3561 unmount_fstype $SINGLEMDS || error "umount failed (4)"
3563 [[ "$ERROR" != "y" ]] ||
3564 error "old and new files are different after sync"
3566 log "files compared the same"
3567 cleanup || error "cleanup failed with $?"
3569 run_test 38 "MDS recreates missing lov_objid file from OST data"
3572 [[ -n "$(type -p perl)" ]] || skip_env "need perl for leak_finder.pl"
3576 cleanup || error "cleanup failed with $?"
3577 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
3578 error "memory leak detected" || true
3580 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
3582 test_40() { # bug 15759
3583 start_ost || error "Unable to start OST1"
3584 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
3585 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
3587 cleanup || error "cleanup failed with rc $?"
3589 run_test 40 "race during service thread startup"
3591 test_41a() { #bug 14134
3592 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3593 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
3594 skip "Loop devices does not work with nosvc option"
3597 combined_mgs_mds || skip "needs combined MGT and MDT device"
3599 start_mdt 1 -o nosvc -n
3600 if [ $MDSCOUNT -ge 2 ]; then
3601 for num in $(seq 2 $MDSCOUNT); do
3602 start_mdt $num || return
3605 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
3606 start_mdt 1 -o nomgs,force
3607 mount_client $MOUNT || error "mount_client $MOUNT failed"
3610 echo "blah blah" > $MOUNT/$tfile
3613 umount_client $MOUNT || error "umount_client $MOUNT failed"
3614 stop ost1 -f || error "unable to stop OST1"
3615 stop_mds || error "Unable to stop MDS"
3616 stop_mds || error "Unable to stop MDS on second try"
3618 run_test 41a "mount mds with --nosvc and --nomgs"
3621 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3622 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
3623 skip "Loop devices does not work with nosvc option"
3626 ! combined_mgs_mds && skip "needs combined mgs device"
3630 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3632 start_mdt 1 -o nosvc -n
3633 if [ $MDSCOUNT -ge 2 ]; then
3634 for num in $(seq 2 $MDSCOUNT); do
3635 start_mdt $num || return
3638 start_ost || error "Unable to start OST1"
3639 start_mdt 1 -o nomgs,force
3640 mount_client $MOUNT || error "mount_client $MOUNT failed"
3643 echo "blah blah" > $MOUNT/$tfile
3644 cat $MOUNT/$tfile || error "cat $MOUNT/$tfile failed"
3646 umount_client $MOUNT -f || error "umount_client $MOUNT failed"
3647 stop_ost || error "Unable to stop OST1"
3648 stop_mds || error "Unable to stop MDS"
3649 stop_mds || error "Unable to stop MDS on second try"
3651 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
3654 local oss_list=$(comma_list $(osts_nodes))
3656 [[ "$MDS1_VERSION" -ge $(version_code 2.6.52) ]] ||
3657 [[ "$MDS1_VERSION" -ge $(version_code 2.5.26) &&
3658 "$MDS1_VERSION" -lt $(version_code 2.5.50) ]] ||
3659 [[ "$MDS1_VERSION" -ge $(version_code 2.5.4) &&
3660 "$MDS1_VERSION" -lt $(version_code 2.5.11) ]] ||
3661 skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"
3663 # ensure mds1 ost1 have been created even if running sub-test standalone
3666 cleanup || error "cleanup failed"
3668 # using directly mount command instead of start() function to avoid
3669 # any side effect of // with others/externals tools/features
3670 # ("zpool import", ...)
3672 # MDT concurrent start
3674 LOAD_MODULES_REMOTE=true load_modules
3675 do_facet $SINGLEMDS "lsmod | grep -q libcfs" ||
3676 error "MDT concurrent start: libcfs module not loaded"
3678 local mds1dev=$(mdsdevname 1)
3679 local mds1mnt=$(facet_mntpt mds1)
3680 local mds1opts=$MDS_MOUNT_OPTS
3682 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3683 ! do_facet mds1 test -b $mds1dev; then
3684 mds1opts=$(csa_add "$mds1opts" -o loop)
3686 if [[ "$mds1_FSTYPE" == zfs ]]; then
3687 import_zpool mds1 || return ${PIPESTATUS[0]}
3690 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3691 do_facet mds1 "$LCTL set_param fail_loc=0x80000716"
3693 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts &
3696 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts
3701 do_facet mds1 "$LCTL set_param fail_loc=0x0"
3702 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3703 echo "1st MDT start succeed"
3704 echo "2nd MDT start failed with $rc2"
3705 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3706 echo "1st MDT start failed with $rc"
3707 echo "2nd MDT start succeed"
3710 error "unexpected concurrent MDT mounts result, rc=$rc rc2=$rc2"
3713 if [ $MDSCOUNT -ge 2 ]; then
3714 for num in $(seq 2 $MDSCOUNT); do
3715 start_mdt $num || return
3719 # OST concurrent start
3721 do_rpc_nodes $oss_list "lsmod | grep -q libcfs" ||
3722 error "OST concurrent start: libcfs module not loaded"
3724 local ost1dev=$(ostdevname 1)
3725 local ost1mnt=$(facet_mntpt ost1)
3726 local ost1opts=$OST_MOUNT_OPTS
3728 if [ "$ost1_FSTYPE" == ldiskfs ] &&
3729 ! do_facet ost1 test -b $ost1dev; then
3730 ost1opts=$(csa_add "$ost1opts" -o loop)
3732 if [[ "$ost1_FSTYPE" == zfs ]]; then
3733 import_zpool ost1 || return ${PIPESTATUS[0]}
3736 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3737 do_facet ost1 "$LCTL set_param fail_loc=0x80000716"
3739 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts &
3742 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts
3746 do_facet ost1 "$LCTL set_param fail_loc=0x0"
3747 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3748 echo "1st OST start succeed"
3749 echo "2nd OST start failed with $rc2"
3750 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3751 echo "1st OST start failed with $rc"
3752 echo "2nd OST start succeed"
3756 error "unexpected concurrent OST mounts result, rc=$rc rc2=$rc2"
3762 # verify everything ok
3767 error "MDT(s) start failed"
3775 error "OST(s) start failed"
3783 error "client start failed"
3790 error "client mount failed"
3794 run_test 41c "concurrent mounts of MDT/OST should all fail but one"
3796 test_42() { #bug 14693
3800 check_mount || error "client was not mounted"
3802 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3803 PARAM="llite.$FSNAME-*.some_wrong_param"
3805 PARAM="$FSNAME.llite.some_wrong_param"
3808 do_facet mgs $PERM_CMD $PARAM=10
3809 umount_client $MOUNT ||
3810 error "unmounting client failed with invalid llite param"
3811 mount_client $MOUNT ||
3812 error "mounting client failed with invalid llite param"
3814 do_facet mgs $PERM_CMD $PARAM=20
3815 cleanup || error "stopping $FSNAME failed with invalid sys param"
3817 check_mount || error "client was not mounted with invalid sys param"
3818 cleanup || error "stopping $FSNAME failed with invalid sys param"
3820 run_test 42 "allow client/server mount/unmount with invalid config param"
3823 [[ "$MGS_VERSION" -ge $(version_code 2.5.58) ]] ||
3824 skip "Need MDS version at least 2.5.58"
3825 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
3828 USER1=$(getent passwd | grep :$ID1:$ID1: | cut -d: -f1)
3829 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1"
3832 chmod ugo+x $DIR || error "chmod 0 failed"
3833 set_persistent_param_and_check mds1 \
3834 "mdt.$FSNAME-MDT0000.root_squash" \
3835 "$FSNAME.mdt.root_squash" \
3837 wait_update $HOSTNAME \
3838 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3840 error "check llite root_squash failed!"
3841 set_persistent_param_and_check mds1 \
3842 "mdt.$FSNAME-MDT0000.nosquash_nids" \
3843 "$FSNAME.mdt.nosquash_nids" \
3845 wait_update $HOSTNAME \
3846 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3848 error "check llite nosquash_nids failed!"
3851 # create set of test files
3853 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
3854 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
3855 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
3857 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
3858 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
3860 mkdir_on_mdt0 $DIR/$tdir-rootdir || error "mkdir failed"
3861 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
3862 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
3864 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
3865 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
3866 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
3869 # check root_squash:
3870 # set root squash UID:GID to RUNAS_ID
3871 # root should be able to access only files owned by RUNAS_ID
3873 set_persistent_param_and_check mds1 \
3874 "mdt.$FSNAME-MDT0000.root_squash" \
3875 "$FSNAME.mdt.root_squash" \
3876 "$RUNAS_ID:$RUNAS_ID"
3877 wait_update $HOSTNAME \
3878 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3879 "$RUNAS_ID:$RUNAS_ID" ||
3880 error "check llite root_squash failed!"
3882 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
3883 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3884 error "$ST: root read permission is denied"
3885 echo "$ST: root read permission is granted - ok"
3888 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3889 error "$ST: root write permission is denied"
3890 echo "$ST: root write permission is granted - ok"
3892 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3893 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3894 error "$ST: root read permission is granted"
3895 echo "$ST: root read permission is denied - ok"
3898 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3899 error "$ST: root write permission is granted"
3900 echo "$ST: root write permission is denied - ok"
3902 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3903 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null &&
3904 error "$ST: root unlink permission is granted"
3905 echo "$ST: root unlink permission is denied - ok"
3907 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null &&
3908 error "$ST: root create permission is granted"
3909 echo "$ST: root create permission is denied - ok"
3912 # check root_squash is enforced independently
3913 # of client cache content
3915 # access file by USER1, keep access open
3916 # root should be denied access to user file
3918 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
3922 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
3923 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3924 { kill $pid; error "$ST: root read permission is granted"; }
3925 echo "$ST: root read permission is denied - ok"
3928 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3929 { kill $pid; error "$ST: root write permission is granted"; }
3930 echo "$ST: root write permission is denied - ok"
3936 # check nosquash_nids:
3937 # put client's NID into nosquash_nids list,
3938 # root should be able to access root file after that
3940 local NIDLIST=$($LCTL list_nids all | tr '\n' ' ')
3941 NIDLIST="2@gni $NIDLIST 192.168.0.[2,10]@tcp"
3942 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
3943 set_persistent_param_and_check mds1 \
3944 "mdt.$FSNAME-MDT0000.nosquash_nids" \
3945 "$FSNAME-MDTall.mdt.nosquash_nids" \
3947 wait_update $HOSTNAME \
3948 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3950 error "check llite nosquash_nids failed!"
3952 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3953 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3954 error "$ST: root read permission is denied"
3955 echo "$ST: root read permission is granted - ok"
3958 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3959 error "$ST: root write permission is denied"
3960 echo "$ST: root write permission is granted - ok"
3962 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3963 rm $DIR/$tdir-rootdir/tfile-1 ||
3964 error "$ST: root unlink permission is denied"
3965 echo "$ST: root unlink permission is granted - ok"
3966 touch $DIR/$tdir-rootdir/tfile-2 ||
3967 error "$ST: root create permission is denied"
3968 echo "$ST: root create permission is granted - ok"
3969 cleanup || error "cleanup failed with $?"
3971 run_test 43a "check root_squash and nosquash_nids"
3973 test_43b() { # LU-5690
3974 [[ "$MGS_VERSION" -ge $(version_code 2.7.62) ]] ||
3975 skip "Need MGS version 2.7.62+"
3977 if [[ -z "$fs2mds_DEV" ]]; then
3978 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
3979 skip_env "mixed loopback and real device not working"
3982 local fs2mdsdev=$(mdsdevname 1_2)
3983 local fs2mdsvdev=$(mdsvdevname 1_2)
3985 # temporarily use fs2mds as fs2mgs
3987 local fs2mgsdev=$fs2mdsdev
3988 local fs2mgsvdev=$fs2mdsvdev
3990 local fsname=test1234
3992 load_module llite/lustre
3993 local client_ip=$(host_nids_address $HOSTNAME $NETTYPE)
3994 local host=${client_ip//*./}
3995 local net=${client_ip/%$host/}
3996 local nosquash_nids=$(h2nettype $net[$host,$host,$host])
3998 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
3999 --param mdt.root_squash=$RUNAS_ID:$RUNAS_ID \
4000 --param mdt.nosquash_nids=$nosquash_nids \
4001 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
4002 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS || error "start fs2mgs failed"
4003 stop $fs2mgs -f || error "stop fs2mgs failed"
4004 cleanup || error "cleanup failed with $?"
4006 run_test 43b "parse nosquash_nids with commas in expr_list"
4010 check_mount || error "check_mount"
4011 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
4013 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
4014 for VAL in $UUIDS; do
4015 NID=$(echo $VAL | cut -d= -f1)
4016 CLUUID=$(echo $VAL | cut -d= -f2)
4017 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
4019 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
4020 cleanup || error "cleanup failed with $?"
4022 run_test 44 "mounted client proc entry exists"
4026 check_mount || error "check_mount"
4027 stop_mds || error "Unable to stop MDS"
4031 #define OBD_FAIL_PTLRPC_LONG_REPL_UNLINK 0x50f
4032 do_facet client "$LCTL set_param fail_loc=0x8000050f"
4035 manual_umount_client --force || error "manual_umount_client failed"
4036 do_facet client "$LCTL set_param fail_loc=0x0"
4037 start_mds || error "unable to start MDS"
4038 mount_client $MOUNT || error "mount_client $MOUNT failed"
4039 cleanup || error "cleanup failed with $?"
4041 run_test 45 "long unlink handling in ptlrpcd"
4048 umount_client $MOUNT2 || rc=$?
4049 umount_client $MOUNT || rc=$?
4050 while [ $count -gt 0 ]; do
4051 stop ost${count} -f || rc=$?
4055 cleanup_nocli || rc=$?
4056 #writeconf to remove all ost2 traces for subsequent tests
4057 writeconf_or_reformat
4062 echo "Testing with $OSTCOUNT OSTs"
4064 start_mds || error "unable to start MDS"
4065 #first client should see only one ost
4066 start_ost || error "Unable to start OST1"
4067 wait_osc_import_state mds ost FULL
4069 mount_client $MOUNT || error "mount_client $MOUNT failed"
4070 trap "cleanup_46a $OSTCOUNT" EXIT ERR
4073 for (( i=2; i<=$OSTCOUNT; i++ )); do
4074 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
4075 error "start_ost$i $(ostdevname $i) failed"
4078 # wait until osts in sync
4079 for (( i=2; i<=$OSTCOUNT; i++ )); do
4080 wait_osc_import_state mds ost$i FULL
4081 wait_osc_import_ready client ost$i
4084 #second client see all ost's
4086 mount_client $MOUNT2 || error "mount_client failed"
4087 $LFS setstripe -c -1 $MOUNT2 ||
4088 error "$LFS setstripe -c -1 $MOUNT2 failed"
4089 $LFS getstripe $MOUNT2 || error "$LFS getstripe $MOUNT2 failed"
4091 echo "ok" > $MOUNT2/widestripe
4092 $LFS getstripe $MOUNT2/widestripe ||
4093 error "$LFS getstripe $MOUNT2/widestripe failed"
4094 # fill acl buffer for avoid expand lsm to them
4095 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd |
4097 setfacl -m $acl $MOUNT2/widestripe
4101 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
4103 cleanup_46a $OSTCOUNT || error "cleanup_46a failed"
4105 run_test 46a "handle ost additional - wide striped file"
4110 check_mount || error "check_mount failed"
4111 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
4115 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
4116 if echo $ns | grep "MDT[[:digit:]]*"; then
4119 lrs=$(echo $ns | sed 's/.*lru_size=//')
4120 lru_size[count]=$lrs
4125 facet_failover $SINGLEMDS
4126 client_up || error "client_up failed"
4129 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
4130 if echo $ns | grep "MDT[[:digit:]]*"; then
4133 lrs=$(echo $ns | sed 's/.*lru_size=//')
4134 if ! test "$lrs" -eq "${lru_size[count]}"; then
4135 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
4136 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
4141 cleanup || error "cleanup failed with $?"
4143 run_test 47 "server restart does not make client loss lru_resize settings"
4148 # reformat after this test is needed - if the test fails,
4149 # we will have unkillable file at FS
4153 test_48() { # bz-17636 LU-7473
4157 check_mount || error "check_mount failed"
4159 $LFS setstripe -c -1 $MOUNT ||
4160 error "$LFS setstripe -c -1 $MOUNT failed"
4161 $LFS getstripe $MOUNT || error "$LFS getstripe $MOUNT failed"
4163 echo "ok" > $MOUNT/widestripe
4164 $LFS getstripe $MOUNT/widestripe ||
4165 error "$LFS getstripe $MOUNT/widestripe failed"
4167 # In the future, we may introduce more EAs, such as selinux, enlarged
4168 # LOV EA, and so on. These EA will use some EA space that is shared by
4169 # ACL entries. So here we only check some reasonable ACL entries count,
4170 # instead of the max number that is calculated from the max_ea_size.
4171 if [ "$MDS1_VERSION" -lt $(version_code 2.8.57) ]; then
4172 count=28 # hard coded of RPC protocol
4173 elif large_xattr_enabled; then
4174 count=4500 # max_num 8187 max_ea_size = 65452
4175 # not create too many (4500) to save test time
4177 count=450 # max_num 497 max_ea_size = 4012
4180 echo "It is expected to hold at least $count ACL entries"
4181 trap cleanup_48 EXIT ERR
4182 for ((i = 0; i < $count; i++)) do
4183 setfacl -m u:$((i + 100)):rw $MOUNT/widestripe ||
4184 error "Fail to setfacl for $MOUNT/widestripe at $i"
4187 cancel_lru_locks mdc
4188 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
4189 local r_count=$(getfacl $MOUNT/widestripe | grep "user:" | wc -l)
4191 count=$((count + 1)) # for the entry "user::rw-"
4193 [ $count -eq $r_count ] ||
4194 error "Expected ACL entries $count, but got $r_count"
4198 run_test 48 "too many acls on file"
4200 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
4201 test_49a() { # bug 17710
4202 local timeout_orig=$TIMEOUT
4203 local ldlm_timeout_orig=$LDLM_TIMEOUT
4204 local LOCAL_TIMEOUT=20
4206 LDLM_TIMEOUT=$LOCAL_TIMEOUT
4207 TIMEOUT=$LOCAL_TIMEOUT
4211 check_mount || error "client mount failed"
4213 echo "check ldlm_timout..."
4214 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
4215 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
4216 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
4218 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
4219 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
4222 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
4223 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
4226 umount_client $MOUNT || error "umount_client $MOUNT failed"
4227 stop_ost || error "problem stopping OSS"
4228 stop_mds || error "problem stopping MDS"
4230 LDLM_TIMEOUT=$ldlm_timeout_orig
4231 TIMEOUT=$timeout_orig
4233 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
4235 test_49b() { # bug 17710
4236 local timeout_orig=$TIMEOUT
4237 local ldlm_timeout_orig=$LDLM_TIMEOUT
4238 local LOCAL_TIMEOUT=20
4240 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
4241 TIMEOUT=$LOCAL_TIMEOUT
4245 check_mount || error "client mount failed"
4247 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
4248 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
4249 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
4251 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
4252 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
4255 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
4256 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
4259 cleanup || error "cleanup failed"
4261 LDLM_TIMEOUT=$ldlm_timeout_orig
4262 TIMEOUT=$timeout_orig
4264 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
4267 # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1
4269 # Test both statfs and lfs df and fail if either one fails
4270 multiop_bg_pause $1 f_
4273 killall -USR1 multiop
4274 [ $RC -ne 0 ] && log "lazystatfs multiop failed"
4275 wait $PID || { RC=$?; log "multiop return error "; }
4277 # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1
4282 if kill -s 0 $PID; then
4285 log "lazystatfs lfs df failed to complete in 5s"
4293 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4294 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4296 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
4298 cleanup || error "cleanup failed with rc $?"
4300 run_test 50a "lazystatfs all servers available"
4304 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4305 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4307 # Wait for client to detect down OST
4308 stop_ost || error "Unable to stop OST1"
4309 wait_osc_import_state client ost DISCONN
4311 log "OSCs should all be DISCONN"
4313 lazystatfs $MOUNT || error "lazystatfs should not return EIO"
4315 umount_client $MOUNT || error "Unable to unmount client"
4316 stop_mds || error "Unable to stop MDS"
4318 run_test 50b "lazystatfs all servers down"
4321 start_mds || error "Unable to start MDS"
4322 start_ost || error "Unable to start OST1"
4323 start_ost2 || error "Unable to start OST2"
4324 mount_client $MOUNT || error "Unable to mount client"
4325 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4326 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4328 # Wait for client to detect down OST
4329 stop_ost || error "Unable to stop OST1"
4330 wait_osc_import_state mds ost DISCONN
4331 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
4333 umount_client $MOUNT || error "Unable to unmount client"
4334 stop_ost2 || error "Unable to stop OST2"
4335 stop_mds || error "Unable to stop MDS"
4336 #writeconf to remove all ost2 traces for subsequent tests
4337 writeconf_or_reformat
4339 run_test 50c "lazystatfs one server down"
4342 start_mds || error "Unable to start MDS"
4343 start_ost || error "Unable to start OST1"
4344 start_ost2 || error "Unable to start OST2"
4345 mount_client $MOUNT || error "Unable to mount client"
4346 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4347 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4349 # Issue the statfs during the window where the client still
4350 # belives the OST to be available but it is in fact down.
4351 # No failure just a statfs which hangs for a timeout interval.
4352 stop_ost || error "Unable to stop OST1"
4353 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
4355 umount_client $MOUNT || error "Unable to unmount client"
4356 stop_ost2 || error "Unable to stop OST2"
4357 stop_mds || error "Unable to stop MDS"
4358 #writeconf to remove all ost2 traces for subsequent tests
4359 writeconf_or_reformat
4361 run_test 50d "lazystatfs client/server conn race"
4368 start_mds || error "Unable to start MDS"
4369 #first client should see only one ost
4370 start_ost || error "Unable to start OST1"
4371 wait_osc_import_state mds ost FULL
4373 # Wait for client to detect down OST
4374 stop_ost || error "Unable to stop OST1"
4375 wait_osc_import_state mds ost DISCONN
4377 mount_client $MOUNT || error "Unable to mount client"
4378 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
4380 multiop_bg_pause $MOUNT _f
4384 if [ $RC1 -ne 0 ]; then
4385 log "multiop failed $RC1"
4388 sleep $(( $TIMEOUT+1 ))
4390 [ $? -ne 0 ] && error "process isn't sleep"
4391 start_ost || error "Unable to start OST1"
4392 wait $pid || error "statfs failed"
4395 umount_client $MOUNT || error "Unable to unmount client"
4396 stop_ost || error "Unable to stop OST1"
4397 stop_mds || error "Unable to stop MDS"
4399 run_test 50e "normal statfs all servers down"
4404 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
4406 start_mds || error "Unable to start mds"
4407 #first client should see only one ost
4408 start_ost || error "Unable to start OST1"
4409 wait_osc_import_state mds ost FULL
4411 start_ost2 || error "Unable to start OST2"
4412 wait_osc_import_state mds ost2 FULL
4414 # Wait for client to detect down OST
4415 stop_ost2 || error "Unable to stop OST2"
4417 wait_osc_import_state mds ost2 DISCONN
4418 mount_client $MOUNT || error "Unable to mount client"
4419 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
4421 multiop_bg_pause $MOUNT _f
4425 if [ $RC1 -ne 0 ]; then
4426 log "lazystatfs multiop failed $RC1"
4429 sleep $(( $TIMEOUT+1 ))
4431 [ $? -ne 0 ] && error "process isn't sleep"
4432 start_ost2 || error "Unable to start OST2"
4433 wait $pid || error "statfs failed"
4434 stop_ost2 || error "Unable to stop OST2"
4437 umount_client $MOUNT -f || error "Unable to unmount client"
4438 stop_ost || error "Unable to stop OST1"
4439 stop_mds || error "Unable to stop MDS"
4440 #writeconf to remove all ost2 traces for subsequent tests
4441 writeconf_or_reformat
4443 run_test 50f "normal statfs one server in down"
4446 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs"
4448 start_ost2 || error "Unable to start OST2"
4449 wait_osc_import_state mds ost2 FULL
4450 wait_osc_import_ready client ost2
4452 if [[ $PERM_CMD == *"set_param -P"* ]]; then
4453 local PARAM="osc.${FSNAME}-OST0001*.active"
4455 local PARAM="${FSNAME}-OST0001.osc.active"
4458 $LFS setstripe -c -1 $DIR/$tfile || error "$LFS setstripe failed"
4459 do_facet mgs $PERM_CMD $PARAM=0 || error "Unable to deactivate OST"
4461 umount_client $MOUNT || error "Unable to unmount client"
4462 mount_client $MOUNT || error "Unable to mount client"
4463 # This df should not cause a panic
4466 do_facet mgs $PERM_CMD $PARAM=1 || error "Unable to activate OST"
4467 rm -f $DIR/$tfile || error "unable to remove file $DIR/$tfile"
4468 umount_client $MOUNT || error "Unable to unmount client"
4469 stop_ost2 || error "Unable to stop OST2"
4470 stop_ost || error "Unable to stop OST1"
4471 stop_mds || error "Unable to stop MDS"
4472 #writeconf to remove all ost2 traces for subsequent tests
4473 writeconf_or_reformat
4475 run_test 50g "deactivated OST should not cause panic"
4479 # prepare MDT/OST, make OSC inactive for OST1
4480 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs"
4482 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
4483 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
4484 error "tunefs OST1 failed"
4485 start_mds || error "Unable to start MDT"
4486 start_ost || error "Unable to start OST1"
4487 start_ost2 || error "Unable to start OST2"
4488 mount_client $MOUNT || error "client start failed"
4490 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4492 # activatate OSC for OST1
4493 set_persistent_param_and_check client \
4494 "osc.${FSNAME}-OST0000-osc-[!M]*.active" \
4495 "${FSNAME}-OST0000.osc.active" 1
4497 mkdir $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
4498 $LFS setstripe -c -1 -i 0 $DIR/$tdir/2 ||
4499 error "$LFS setstripe $DIR/$tdir/2 failed"
4500 sleep 1 && echo "create a file after OST1 is activated"
4501 # doing some io, shouldn't crash
4502 dd if=/dev/zero of=$DIR/$tdir/2/$tfile-io bs=1M count=10
4504 # check OSC import is working
4505 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
4506 error "some OSC imports are still not connected"
4510 umount_client $MOUNT || error "Unable to umount client"
4511 stop_ost2 || error "Unable to stop OST2"
4512 cleanup_nocli || error "cleanup_nocli failed with $?"
4514 run_test 50h "LU-642: activate deactivated OST"
4517 # prepare MDT/OST, make OSC inactive for OST1
4518 [ "$MDSCOUNT" -lt "2" ] && skip "needs >= 2 MDTs"
4521 [ $(facet_fstype mds2) == zfs ] && import_zpool mds2
4522 do_facet mds2 "$TUNEFS --param mdc.active=0 $(mdsdevname 2)" ||
4523 error "tunefs MDT2 failed"
4524 start_mds || error "Unable to start MDT"
4525 start_ost || error "Unable to start OST1"
4526 start_ost2 || error "Unable to start OST2"
4527 mount_client $MOUNT || error "client start failed"
4529 mkdir_on_mdt0 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4531 if [[ $PERM_CMD == *"set_param -P"* ]]; then
4532 $PERM_CMD mdc.${FSNAME}-MDT0001-mdc-*.active=0 &&
4533 error "deactive MDC0 succeeds"
4535 $PERM_CMD ${FSNAME}-MDT0000.mdc.active=0 &&
4536 error "deactive MDC0 succeeds"
4539 # activate MDC for MDT2
4540 set_persistent_param_and_check client \
4541 "mdc.${FSNAME}-MDT0001-mdc-*.active" \
4542 "${FSNAME}-MDT0001.mdc.active" 1
4544 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
4545 if [ "$MDS1_VERSION" -ge $(version_code 2.7.60) ]
4547 wait_dne_interconnect
4549 $LFS mkdir -i1 $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
4551 createmany -o $DIR/$tdir/2/$tfile-%d 1 || error "create files failed"
4553 rm -rf $DIR/$tdir/2 || error "unlink dir failed"
4555 # deactivate MDC for MDT2
4556 set_persistent_param_and_check client \
4557 "mdc.${FSNAME}-MDT0001-mdc-*.active" \
4558 "${FSNAME}-MDT0001.mdc.active" 0
4560 wait_osp_active mds ${FSNAME}-MDT0001 1 0
4562 $LFS mkdir -i1 $DIR/$tdir/2 &&
4563 error "mkdir $DIR/$tdir/2 succeeds after deactive MDT"
4565 $LFS mkdir -i0 -c$MDSCOUNT $DIR/$tdir/striped_dir ||
4566 error "mkdir $DIR/$tdir/striped_dir fails after deactive MDT2"
4568 local stripe_count=$($LFS getdirstripe -c $DIR/$tdir/striped_dir)
4569 [ $stripe_count -eq $((MDSCOUNT - 1)) ] ||
4570 error "wrong $stripe_count != $((MDSCOUNT -1)) for striped_dir"
4573 umount_client $MOUNT || error "Unable to umount client"
4578 run_test 50i "activate deactivated MDT"
4581 local LOCAL_TIMEOUT=20
4585 check_mount || error "check_mount failed"
4587 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
4588 $LFS setstripe -c -1 $MOUNT/$tdir ||
4589 error "$LFS setstripe -c -1 $MOUNT/$tdir failed"
4590 #define OBD_FAIL_MDS_REINT_DELAY 0x142
4591 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x142"
4592 touch $MOUNT/$tdir/$tfile &
4595 start_ost2 || error "Unable to start OST1"
4597 stop_ost2 || error "Unable to stop OST1"
4598 umount_client $MOUNT -f || error "unmount $MOUNT failed"
4599 cleanup_nocli || error "stop server failed"
4600 #writeconf to remove all ost2 traces for subsequent tests
4601 writeconf_or_reformat
4603 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
4612 do_node $node mkdir -p $dest
4613 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
4615 do_node $node 'tar cf - '$*' | tar xf - -C '$dest';
4616 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
4617 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
4619 do_node $node 'getfattr -d -m "[a-z]*\\." '$*' > '$xattrs
4620 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
4630 local backup2=${TMP}/backup2
4632 do_node $node mkdir -p $backup2
4633 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
4635 do_node $node 'tar cf - '$*' | tar xf - -C '$backup2';
4636 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
4637 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
4639 do_node $node "diff -rq $backup $backup2"
4640 [ $? -eq 0 ] || { error "contents differ"; return 3; }
4642 local xattrs2=${TMP}/xattrs2
4643 do_node $node 'getfattr -d -m "[a-z]*\\." '$*' > '$xattrs2
4644 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
4646 do_node $node "diff $xattrs $xattrs2"
4647 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
4649 do_node $node "rm -rf $backup2 $xattrs2"
4650 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
4654 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4655 skip "ldiskfs only test"
4658 start_mds || error "Unable to start MDS"
4659 start_ost || error "Unable to start OST1"
4660 mount_client $MOUNT || error "Unable to mount client"
4663 local ost1mnt=$(facet_mntpt ost1)
4664 local ost1node=$(facet_active_host ost1)
4665 local ost1tmp=$TMP/conf52
4668 mkdir $DIR/$tdir || error "Unable to create $DIR/$tdir"
4669 touch $TMP/modified_first || error "Unable to create temporary file"
4670 local mtime=$(stat -c %Y $TMP/modified_first)
4671 do_node $ost1node "mkdir -p $ost1tmp &&
4672 touch -m -d @$mtime $ost1tmp/modified_first" ||
4673 error "Unable to create temporary file"
4676 $LFS setstripe -c -1 -S 1M $DIR/$tdir || error "$LFS setstripe failed"
4678 for (( i=0; i < nrfiles; i++ )); do
4679 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c ||
4680 error "multiop failed"
4685 # sync all the data and make sure no pending data on the client,
4686 # thus the SOM xattr would not be changed any more.
4687 cancel_lru_locks osc
4690 echo backup files to $TMP/$tdir
4691 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
4692 copy_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
4693 error "Unable to copy files"
4695 umount_client $MOUNT || error "Unable to umount client"
4696 stop_ost || error "Unable to stop ost1"
4698 echo mount ost1 as ldiskfs
4699 do_node $ost1node mkdir -p $ost1mnt || error "Unable to create $ost1mnt"
4700 if ! do_node $ost1node test -b $ost1_dev; then
4703 do_node $ost1node mount -t "$ost1_FSTYPE" $loop $ost1_dev \
4705 error "Unable to mount ost1 as ldiskfs"
4708 echo backup objects to $ost1tmp/objects
4709 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
4710 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
4711 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
4713 error "Unable to copy objects"
4715 # move objects to lost+found
4716 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
4717 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
4719 do_node $ost1node "umount $ost1mnt" ||
4720 error "Unable to umount ost1 as ldiskfs"
4722 start_ost || error "Unable to start OST1"
4723 mount_client $MOUNT || error "Unable to mount client"
4725 local REPAIRED=$(do_node $ost1node "$LCTL get_param \
4726 -n osd-ldiskfs.$FSNAME-OST0000.oi_scrub" |
4727 awk '/^lf_repa[ri]*ed/ { print $2 }')
4728 [ $REPAIRED -gt 0 ] ||
4729 error "Some entry under /lost+found should be repaired"
4732 diff_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
4733 error "Unable to diff files"
4735 rm -rf $TMP/$tdir $TMP/file_xattrs ||
4736 error "Unable to delete temporary files"
4737 do_node $ost1node "rm -rf $ost1tmp" ||
4738 error "Unable to delete temporary files"
4739 cleanup || error "cleanup failed with $?"
4741 run_test 52 "check recovering objects from lost+found"
4743 # Checks threads_min/max/started for some service
4745 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
4746 # parameter pattern prefix like 'ost.*.ost'.
4759 local msg="Insane $modname thread counts"
4760 local ncpts=$(check_cpt_number $facet)
4764 check_mount || return 41
4766 # We need to expand $parampat, but it may match multiple parameters, so
4767 # we'll pick the first one
4768 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
4769 error "Couldn't expand ${parampat}.threads_min parameter name"
4773 # Remove the .threads_min part
4774 paramp=${paramp%.threads_min}
4776 # Check for sanity in defaults
4777 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4779 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4781 tstarted=$(do_facet $facet "$LCTL get_param \
4782 -n ${paramp}.threads_started" || echo 0)
4783 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' ||
4785 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4787 nthrs=$(expr $tmax - $tmin)
4788 if [ $nthrs -lt $ncpts ]; then
4794 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
4795 skip_env "module parameter forced $facet thread count"
4797 # Check that we can change min/max
4798 do_facet $facet "$LCTL set_param \
4799 ${paramp}.threads_min=$((tmin + nthrs))"
4800 do_facet $facet "$LCTL set_param \
4801 ${paramp}.threads_max=$((tmax - nthrs))"
4802 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4804 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4806 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) &&
4807 $tmax2 == ($tmax - $nthrs)))' || return $?
4809 # Check that we can set min/max to the same value
4810 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4812 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$tmin"
4813 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4815 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4817 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
4819 # Check that we can't set max < min
4820 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$((tmin - 1))"
4821 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4823 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4825 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
4827 # We need to ensure that we get the module options desired; to do this
4828 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
4829 LOAD_MODULES_REMOTE=true
4832 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
4834 setmodopts -a $modname "$newvalue" oldvalue
4837 check_mount || return 41
4839 # Restore previous setting of MODOPTS_*
4840 setmodopts $modname "$oldvalue"
4842 (( $MDS1_VERSION > $(version_code 2.12.52.91) )) || {
4843 echo "skip interop for MDS < v2_12_52-91-g183cb1e3cdd2"
4847 # Check that $opts took
4848 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4850 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4852 tstarted=$(do_facet $facet \
4853 "$LCTL get_param -n ${paramp}.threads_started" || echo 0)
4854 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4860 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
4861 cleanup || error "cleanup failed with rc $?"
4863 run_test 53a "check OSS thread count params"
4867 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
4868 cleanup || error "cleanup failed with $?"
4870 run_test 53b "check MDS thread count params"
4873 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4874 skip "ldiskfs only test"
4877 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p ||
4878 error "llverdev failed with rc=$?"
4881 run_test 54a "test llverdev and partial verify of device"
4884 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4885 skip "ldiskfs only test"
4889 run_llverfs $MOUNT -p || error "llverfs failed with rc=$?"
4890 cleanup || error "cleanup failed with rc=$?"
4892 run_test 54b "test llverfs and partial verify of filesystem"
4896 local max_ost_index=$1
4898 echo -n $(((max_ost_index + 1) * 8))
4902 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4903 skip "ldiskfs only test"
4906 local mdsdev=$(mdsdevname 1)
4907 local mdsvdev=$(mdsvdevname 1)
4911 if ! combined_mgs_mds; then
4912 stop_mgs || error "stopping MGS service failed"
4913 format_mgs || error "formatting MGT failed"
4915 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
4917 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
4918 --reformat $(ostdevname 1) $(ostvdevname 1)
4924 echo checking size of lov_objid for ost index $i
4925 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" |
4926 grep ^User | awk -F 'Size: ' '{print $2}')
4927 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
4928 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
4930 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
4937 run_test 55 "check lov_objid size"
4940 local mds_journal_size_orig=$MDSJOURNALSIZE
4946 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
4947 $(ostdevname 1) $(ostvdevname 1)
4948 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
4949 $(ostdevname 2) $(ostvdevname 2)
4952 start_ost || error "Unable to start first ost (idx 10000)"
4953 start_ost2 || error "Unable to start second ost (idx 1000)"
4954 mount_client $MOUNT || error "Unable to mount client"
4958 # test instantiating PFL components with sparse index LU-15513
4959 mkdir -p $MOUNT/$tdir
4960 $LFS setstripe -E 4M -c 1 -E 1G -c 4 -S4M -E eof -c -1 $MOUNT/$tdir
4961 dd if=/dev/zero of=$MOUNT/$tdir/$tfile bs=4K count=1 seek=10k ||
4962 error "dd to second component failed"
4964 if [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
4965 [[ "$MDS1_VERSION" -ge $(version_code 2.5.4) &&
4966 "$MDS1_VERSION" -lt $(version_code 2.5.11) ]]; then
4967 wait_osc_import_state mds ost1 FULL
4968 wait_osc_import_state mds ost2 FULL
4969 $LFS setstripe --stripe-count=-1 $DIR/$tfile ||
4970 error "Unable to setstripe $DIR/$tfile"
4971 n=$($LFS getstripe --stripe-count $DIR/$tfile)
4972 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
4977 MDSJOURNALSIZE=$mds_journal_size_orig
4980 run_test 56a "check big OST indexes and out-of-index-order start"
4985 umount_client $MOUNT -f || error "unmount client failed"
4994 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs"
4996 trap cleanup_56b EXIT RETURN ERR
4999 if ! combined_mgs_mds ; then
5004 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) --index=0 --reformat \
5005 $(mdsdevname 1) $(mdsvdevname 1)
5006 add mds2 $(mkfs_opts mds2 $(mdsdevname 2)) --index=1 --reformat \
5007 $(mdsdevname 2) $(mdsvdevname 2)
5008 add mds3 $(mkfs_opts mds3 $(mdsdevname 3)) --index=1000 --reformat \
5009 $(mdsdevname 3) $(mdsvdevname 3)
5013 start_mdt 1 || error "MDT 1 (idx 0) start failed"
5014 start_mdt 2 || error "MDT 2 (idx 1) start failed"
5015 start_mdt 3 || error "MDT 3 (idx 1000) start failed"
5016 start_ost || error "Unable to start first ost"
5017 start_ost2 || error "Unable to start second ost"
5019 do_nodes $(comma_list $(mdts_nodes)) \
5020 "$LCTL set_param mdt.*.enable_remote_dir=1 \
5021 mdt.*.enable_remote_dir_gid=-1"
5023 mount_client $MOUNT || error "Unable to mount client"
5025 $LFS mkdir -c3 $MOUNT/$tdir || error "failed to make testdir"
5027 echo "This is test file 1!" > $MOUNT/$tdir/$tfile.1 ||
5028 error "failed to make test file 1"
5029 echo "This is test file 2!" > $MOUNT/$tdir/$tfile.2 ||
5030 error "failed to make test file 2"
5031 echo "This is test file 1000!" > $MOUNT/$tdir/$tfile.1000 ||
5032 error "failed to make test file 1000"
5034 rm -rf $MOUNT/$tdir || error "failed to remove testdir"
5036 $LFS mkdir -i1000 $MOUNT/$tdir.1000 ||
5037 error "create remote dir at idx 1000 failed"
5040 echo "=== START lfs df OUTPUT ==="
5042 echo "==== END lfs df OUTPUT ===="
5044 mdtcnt=$(echo -e "$output" | grep $FSNAME-MDT | wc -l)
5045 ostcnt=$(echo -e "$output" | grep $FSNAME-OST | wc -l)
5047 echo "lfs df returned mdt count $mdtcnt and ost count $ostcnt"
5048 [ $mdtcnt -eq 3 ] || error "lfs df returned wrong mdt count"
5049 [ $ostcnt -eq 2 ] || error "lfs df returned wrong ost count"
5051 echo "This is test file 1!" > $MOUNT/$tdir.1000/$tfile.1 ||
5052 error "failed to make test file 1"
5053 echo "This is test file 2!" > $MOUNT/$tdir.1000/$tfile.2 ||
5054 error "failed to make test file 2"
5055 echo "This is test file 1000!" > $MOUNT/$tdir.1000/$tfile.1000 ||
5056 error "failed to make test file 1000"
5057 rm -rf $MOUNT/$tdir.1000 || error "failed to remove remote_dir"
5060 echo "=== START lfs mdts OUTPUT ==="
5062 echo "==== END lfs mdts OUTPUT ===="
5064 echo -e "$output" | grep -v "MDTS:" | awk '{print $1}' |
5065 sed 's/://g' > $TMP/mdts-actual.txt
5066 sort $TMP/mdts-actual.txt -o $TMP/mdts-actual.txt
5068 echo -e "0\n1\n1000" > $TMP/mdts-expected.txt
5070 diff $TMP/mdts-expected.txt $TMP/mdts-actual.txt
5073 rm $TMP/mdts-expected.txt $TMP/mdts-actual.txt
5075 [ $result -eq 0 ] || error "target_obd proc file is incorrect!"
5077 run_test 56b "test target_obd correctness with nonconsecutive MDTs"
5079 test_57a() { # bug 22656
5080 do_rpc_nodes $(facet_active_host ost1) load_modules_local
5081 local NID=$(do_facet ost1 "$LCTL get_param nis" |
5082 tail -1 | awk '{print $1}')
5083 writeconf_or_reformat
5084 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
5085 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" ||
5086 error "tunefs failed"
5088 start_ost && error "OST registration from failnode should fail"
5091 run_test 57a "initial registration from failnode should fail (should return errs)"
5094 do_rpc_nodes $(facet_active_host ost1) load_modules_local
5095 local NID=$(do_facet ost1 "$LCTL get_param nis" |
5096 tail -1 | awk '{print $1}')
5097 writeconf_or_reformat
5098 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
5099 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" ||
5100 error "tunefs failed"
5102 start_ost || error "OST registration from servicenode should not fail"
5105 run_test 57b "initial registration from servicenode should not fail"
5108 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
5111 test_58() { # bug 22658
5112 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
5114 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5115 createmany -o $DIR/$tdir/$tfile-%d 100
5116 unlinkmany $DIR/$tdir/$tfile-%d 100
5117 stop_mds || error "Unable to stop MDS"
5119 local MNTDIR=$(facet_mntpt $SINGLEMDS)
5120 local devname=$(mdsdevname ${SINGLEMDS//mds/})
5122 # remove all files from the OBJECTS dir
5123 mount_fstype $SINGLEMDS
5125 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
5127 unmount_fstype $SINGLEMDS
5128 # restart MDS with missing llog files
5129 start_mds || error "unable to start MDS"
5130 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0"
5133 run_test 58 "missing llog files must not prevent MDT from mounting"
5136 start_mgsmds >> /dev/null
5137 local C1=$(count_osts)
5138 if [ $C1 -eq 0 ]; then
5139 start_ost >> /dev/null
5143 echo "original ost count: $C1 (expect > 0)"
5144 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
5145 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
5146 local C2=$(count_osts)
5147 echo "after mdt writeconf count: $C2 (expect 0)"
5148 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
5149 echo "OST start without writeconf should fail:"
5150 start_ost >> /dev/null &&
5151 error "OST start without writeconf didn't fail"
5152 echo "OST start with writeconf should succeed:"
5153 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
5154 local C3=$(count_osts)
5155 echo "after ost writeconf count: $C3 (expect 1)"
5156 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
5157 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
5158 local C4=$(count_osts)
5159 echo "after ost2 writeconf count: $C4 (expect 2)"
5160 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
5161 stop_ost2 >> /dev/null
5162 cleanup_nocli >> /dev/null
5163 #writeconf to remove all ost2 traces for subsequent tests
5164 writeconf_or_reformat
5166 run_test 59 "writeconf mount option"
5168 test_60a() { # LU-471
5169 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5170 skip "ldiskfs only test"
5175 for num in $(seq $MDSCOUNT); do
5176 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
5177 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
5178 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
5182 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
5183 [ ${PIPESTATUS[0]} -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
5185 # MDT default has dirdata feature
5186 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
5187 # we disable uninit_bg feature
5188 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
5189 # we set stride extended options
5190 echo $dump | grep stride > /dev/null || error "stride is not set"
5194 run_test 60a "check mkfs.lustre --mkfsoptions -E -O options setting"
5197 [[ "$mds1_FSTYPE" == ldiskfs ]] || skip "ldiskfs only test"
5199 local features=$(do_facet $SINGLEMDS $DUMPE2FS $(mdsdevname 1) |
5201 [ ${PIPESTATUS[0]} -eq 0 ] || error "$DUMPE2FS $(mdsdevname 1) failed"
5204 # ea_inode feature should be enabled by default for MDTs
5205 [[ "$features" =~ "ea_inode" ]] || error "ea_inode is not set"
5206 # large_dir feature should be enabled by default for MDTs
5207 [[ "$features" =~ "large_dir" ]] || error "large_dir is not set"
5209 run_test 60b "check mkfs.lustre MDT default features"
5211 test_61a() { # LU-80
5212 local lxattr=$(large_xattr_enabled)
5214 (( "$MDS1_VERSION" >= $(version_code 2.1.53) )) ||
5215 skip "Need MDS version at least 2.1.53 for large_xattr"
5217 if [[ "$mds1_FSTYPE" == ldiskfs ]] && ! large_xattr_enabled; then
5220 for ((num=1; num <= $MDSCOUNT; num++)); do
5221 do_facet mds${num} $TUNE2FS -O ea_inode \
5222 $(mdsdevname $num) ||
5223 error "tune2fs on mds $num failed"
5227 setup || error "setting up the filesystem failed"
5228 client_up || error "starting client failed"
5230 local file=$DIR/$tfile
5231 touch $file || error "touch $file failed"
5233 local large_value="$(generate_string $(max_xattr_size))"
5234 local small_value="bar"
5236 local name="trusted.big"
5237 log "save large xattr of $(max_xattr_size) bytes on $name on $file"
5238 setfattr -n $name -v $large_value $file ||
5239 error "saving $name on $file failed"
5241 local new_value=$(get_xattr_value $name $file)
5242 [[ "$new_value" != "$large_value" ]] &&
5243 error "$name different after saving"
5245 log "shrink value of $name on $file"
5246 setfattr -n $name -v $small_value $file ||
5247 error "shrinking value of $name on $file failed"
5249 new_value=$(get_xattr_value $name $file)
5250 [[ "$new_value" != "$small_value" ]] &&
5251 error "$name different after shrinking"
5253 log "grow value of $name on $file"
5254 setfattr -n $name -v $large_value $file ||
5255 error "growing value of $name on $file failed"
5257 new_value=$(get_xattr_value $name $file)
5258 [[ "$new_value" != "$large_value" ]] &&
5259 error "$name different after growing"
5261 log "check value of $name on $file after remounting MDS"
5263 new_value=$(get_xattr_value $name $file)
5264 [[ "$new_value" != "$large_value" ]] &&
5265 error "$name different after remounting MDS"
5267 log "remove large xattr $name from $file"
5268 setfattr -x $name $file || error "removing $name from $file failed"
5270 if $lxattr && [ "$mds1_FSTYPE" == ldiskfs ]; then
5271 stopall || error "stopping for e2fsck run"
5272 for num in $(seq $MDSCOUNT); do
5273 run_e2fsck $(facet_active_host mds$num) \
5274 $(mdsdevname $num) "-y" ||
5275 error "e2fsck MDT$num failed"
5277 setup_noconfig || error "remounting the filesystem failed"
5280 # need to delete this file to avoid problems in other tests
5282 cleanup || error "stopping systems failed"
5284 run_test 61a "large xattr"
5286 test_61b() { # LU-80
5287 local lxattr=$(large_xattr_enabled)
5289 (( $MDS1_VERSION >= $(version_code 2.15.51) )) ||
5290 skip "Need MDS version at least 2.15.51 for large_xattr fix"
5292 [[ "$mds1_FSTYPE" == "ldiskfs" ]] || skip "ldiskfs specific bug"
5294 if ! large_xattr_enabled; then
5297 for (( num=1; num <= $MDSCOUNT; num++ )); do
5298 do_facet mds${num} $TUNE2FS -O ea_inode \
5299 $(mdsdevname $num) ||
5300 error "tune2fs on mds $num failed"
5304 setup || error "setting up the filesystem failed"
5305 client_up || error "starting client failed"
5307 local _file=$MOUNT/panda
5308 local large_value="$(generate_string $(max_xattr_size))"
5309 local name="trusted.big"
5311 touch ${_file} || error "touch ${_file} failed"
5312 setfattr -n $name -v $large_value ${_file} ||
5313 error "saving $name on $file failed"
5315 MDT_DEV="${FSNAME}-MDT0000"
5316 MDT_DEVNAME=$(mdsdevname ${SINGLEMDS//mds/})
5318 stopall || error "stopping for e2fsck run"
5321 ino=$(do_facet $SINGLEMDS "$DEBUGFS -R 'stat /ROOT/panda' \
5322 ${MDT_DEVNAME} | grep trusted.big")
5323 ino=$(echo "${ino}" | awk '{print $2;}')
5324 echo "large ea "${ino}
5326 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"ln $ino /lost+found\\\" \
5329 setup_noconfig || error "remounting the filesystem failed"
5331 do_facet $SINGLEMDS $LCTL lfsck_start -M ${MDT_DEV} -t namespace || {
5332 error "can't start lfsck namespace"
5336 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
5337 mdd.${MDT_DEV}.lfsck_namespace |
5338 awk '/^status/ { print \\\$2 }'" "completed" 32 || {
5339 error "(2) unexpected status"
5342 stopall || error "stopping for e2fsck run"
5343 for num in $(seq $MDSCOUNT); do
5344 run_e2fsck $(facet_active_host mds$num) \
5345 $(mdsdevname $num) "-y" ||
5346 error "e2fsck MDT$num failed"
5348 setup_noconfig || error "remounting the filesystem failed"
5350 # need to delete this file to avoid problems in other tests
5352 cleanup || error "stopping systems failed"
5354 run_test 61b "large xattr"
5357 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5358 skip "ldiskfs only test"
5360 [[ "$MDS1_VERSION" -ge $(version_code 2.2.51) ]] ||
5361 skip "Need MDS version at least 2.2.51"
5364 local mdsdev=$(mdsdevname 1)
5365 local ostdev=$(ostdevname 1)
5367 echo "disable journal for mds"
5368 do_facet mds1 $TUNE2FS -O ^has_journal $mdsdev || error "tune2fs failed"
5369 start_mds && error "MDT start should fail"
5370 echo "disable journal for ost"
5371 do_facet ost1 $TUNE2FS -O ^has_journal $ostdev || error "tune2fs failed"
5372 start_ost && error "OST start should fail"
5373 cleanup || error "cleanup failed with rc $?"
5376 run_test 62 "start with disabled journal"
5379 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5380 skip "ldiskfs only test"
5383 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ldiskfs
5384 local inode_slab=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
5385 awk '/ldiskfs_inode_cache/ { print $5 / $6 }')
5386 if [ -z "$inode_slab" ]; then
5387 skip "ldiskfs module has not been loaded"
5390 if grep -q "CONFIG_DEBUG_LOCK_ALLOC=y" /boot/config-$(uname -r); then
5391 skip "test is not compatible with CONFIG_DEBUG_LOCK_ALLOC=y"
5394 echo "$inode_slab ldiskfs inodes per page"
5395 [ "${inode_slab%.*}" -ge "3" ] && return 0
5397 # If kmalloc-128 is also 1 per page - this is a debug kernel
5398 # and so this is not an error.
5399 local kmalloc128=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
5400 awk '/^(dma-kmalloc|size)-128 / { print $5 / $6 }')
5401 # 32 128-byte chunks in 4k
5402 [ "${kmalloc128%.*}" -lt "32" ] ||
5403 error "ldiskfs inode too big, only $inode_slab objs/page, " \
5404 "kmalloc128 = $kmalloc128 objs/page"
5406 run_test 63 "Verify each page can at least hold 3 ldiskfs inodes"
5409 start_mds || error "unable to start MDS"
5410 start_ost || error "Unable to start OST1"
5411 start_ost2 || error "Unable to start second ost"
5412 mount_client $MOUNT || error "Unable to mount client"
5413 stop_ost2 || error "Unable to stop second ost"
5416 umount_client $MOUNT -f || error "unmount $MOUNT failed"
5417 cleanup_nocli || error "cleanup_nocli failed with $?"
5418 #writeconf to remove all ost2 traces for subsequent tests
5419 writeconf_or_reformat
5421 run_test 64 "check lfs df --lazy "
5423 test_65() { # LU-2237
5424 # Currently, the test is only valid for ldiskfs backend
5425 [ "$mds1_FSTYPE" != ldiskfs ] &&
5426 skip "ldiskfs only test"
5428 local devname=$(mdsdevname ${SINGLEMDS//mds/})
5429 local brpt=$(facet_mntpt brpt)
5432 if ! do_facet $SINGLEMDS "test -b $devname"; then
5436 stop_mds || error "Unable to stop MDS"
5437 local obj=$(do_facet $SINGLEMDS \
5438 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
5440 if [ -z "$obj" ]; then
5441 # The MDT may be just re-formatted, mount the MDT for the
5442 # first time to guarantee the "last_rcvd" file is there.
5443 start_mds || error "fail to mount the MDS for the first time"
5444 stop_mds || error "Unable to stop MDS"
5447 # remove the "last_rcvd" file
5448 do_facet $SINGLEMDS "mkdir -p $brpt"
5449 do_facet $SINGLEMDS \
5450 "mount -t $mds1_FSTYPE $opts $devname $brpt"
5451 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
5452 do_facet $SINGLEMDS "$UMOUNT $brpt"
5454 # restart MDS, the "last_rcvd" file should be recreated.
5455 start_mds || error "fail to restart the MDS"
5456 stop_mds || error "Unable to stop MDS"
5457 obj=$(do_facet $SINGLEMDS \
5458 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
5459 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
5461 run_test 65 "re-create the lost last_rcvd file when server mount"
5464 [[ "$MGS_VERSION" -ge $(version_code 2.3.59) ]] ||
5465 skip "Need MGS version at least 2.3.59"
5467 check_versions || skip "do not replace_nids with mismatched versions"
5470 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
5471 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
5473 # add EXCLUDE records to config log, they are not to be
5474 # removed by lctl replace_nids
5475 set_conf_param_and_check mds \
5476 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
5477 "$FSNAME-OST0000.osc.active" \
5480 echo "replace_nids should fail if MDS, OSTs and clients are UP"
5481 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5482 error "replace_nids fail"
5484 umount_client $MOUNT || error "unmounting client failed"
5485 echo "replace_nids should fail if MDS and OSTs are UP"
5486 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5487 error "replace_nids fail"
5489 stop_ost || error "Unable to stop OST1"
5490 echo "replace_nids should fail if MDS is UP"
5491 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5492 error "replace_nids fail"
5494 stop_mds || error "stopping mds failed"
5496 if combined_mgs_mds; then
5497 start_mdt 1 "-o nosvc" ||
5498 error "starting mds with nosvc option failed"
5501 echo "command should accept two parameters"
5502 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
5503 error "command should accept two params"
5505 echo "correct device name should be passed"
5506 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
5507 error "wrong devname"
5509 echo "wrong nids list should not destroy the system"
5510 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
5512 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "asdfasdf, asdfadf" &&
5515 echo "replace OST nid"
5516 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
5517 error "replace nids failed"
5519 echo "command should accept two parameters"
5520 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
5521 error "command should accept two params"
5523 echo "wrong nids list should not destroy the system"
5524 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
5527 local FAKE_NIDS="192.168.0.112@tcp1,192.168.0.112@tcp2"
5528 local FAKE_FAILOVER="192.168.0.113@tcp1,192.168.0.113@tcp2"
5529 local NIDS_AND_FAILOVER="$MDS_NID,$FAKE_NIDS:$FAKE_FAILOVER"
5530 echo "set NIDs with failover"
5531 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $NIDS_AND_FAILOVER ||
5532 error "replace nids failed"
5535 echo "replace MDS nid"
5536 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
5537 error "replace nids failed"
5539 if ! combined_mgs_mds ; then
5542 stop_mds || error "Unable to stop MDS"
5545 start_mgsmds || error "start mgsmds failed"
5546 set_conf_param_and_check mds \
5547 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
5548 "$FSNAME-OST0000.osc.active" \
5550 start_ost || error "unable to start OST"
5551 mount_client $MOUNT || error "mount client failed"
5553 check_mount || error "error after nid replace"
5554 cleanup || error "cleanup failed"
5557 run_test 66 "replace nids"
5559 test_67() { #LU-2950
5560 local legacy="$TMP/legacy_lnet_config"
5561 local new="$TMP/new_routes_test"
5562 local out="$TMP/config_out_file"
5563 local verify="$TMP/conv_verify"
5564 local verify_conf="$TMP/conf_verify"
5566 # Create the legacy file that will be run through the
5567 # lustre_routes_conversion script
5568 cat <<- LEGACY_LNET_CONFIG > $legacy
5569 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
5570 tcp2 54 10.1.3.2@tcp;
5571 tcp3 10.3.4.3@tcp:3;
5575 # Create the verification file to verify the output of
5576 # lustre_routes_conversion script against.
5577 cat <<- VERIFY_LNET_CONFIG > $verify
5578 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
5579 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
5580 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
5581 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
5582 tcp4: { gateway: 10.3.3.4@tcp }
5585 # Create the verification file to verify the output of
5586 # lustre_routes_config script against
5587 cat <<- VERIFY_LNET_CONFIG > $verify_conf
5588 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
5589 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
5590 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
5591 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
5592 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
5595 $LUSTRE_ROUTES_CONVERSION $legacy $new > /dev/null
5596 if [ -f $new ]; then
5597 # verify the conversion output
5598 cmp -s $new $verify > /dev/null
5599 if [ $? -eq 1 ]; then
5600 error "routes conversion failed"
5603 lustre_routes_config --dry-run --verbose $new > $out
5604 # check that the script succeeded
5605 cmp -s $out $verify_conf > /dev/null
5606 if [ $? -eq 1 ]; then
5607 error "routes config failed"
5610 error "routes conversion test failed"
5612 # remove generated files
5613 rm -f $new $legacy $verify $verify_conf $out
5615 run_test 67 "test routes conversion and configuration"
5623 [ "$MDS1_VERSION" -ge $(version_code 2.4.53) ] ||
5624 skip "Need MDS version at least 2.4.53"
5626 umount_client $MOUNT || error "umount client failed"
5631 # START-END - the sequences we'll be reserving
5632 START=$(do_facet $SINGLEMDS \
5633 $LCTL get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
5634 END=$((START + (1 << 30)))
5635 do_facet $SINGLEMDS \
5636 $LCTL set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
5638 # reset the sequences MDT0000 has already assigned
5639 do_facet $SINGLEMDS \
5640 $LCTL set_param seq.srv*MDT0000.space=clear
5642 # remount to let the client allocate new sequence
5643 mount_client $MOUNT || error "mount client failed"
5645 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
5646 do_facet $SINGLEMDS \
5647 $LCTL get_param seq.srv*MDT0000.space
5648 $LFS path2fid $DIR/$tfile
5650 local old_ifs="$IFS"
5652 fid=($($LFS path2fid $DIR/$tfile))
5656 if [[ $seq < $END ]]; then
5657 error "used reserved sequence $seq?"
5659 cleanup || error "cleanup failed with $?"
5661 run_test 68 "be able to reserve specific sequences in FLDB"
5663 # Test 69: is about the total number of objects ever created on an OST.
5664 # so that when it is reformatted the normal MDS->OST orphan recovery won't
5665 # just "precreate" the missing objects. In the past it might try to recreate
5666 # millions of objects after an OST was reformatted
5668 [[ "$MDS1_VERSION" -lt $(version_code 2.4.2) ]] &&
5669 skip "Need MDS version at least 2.4.2"
5671 [[ "$MDS1_VERSION" -ge $(version_code 2.4.50) ]] &&
5672 [[ "$MDS1_VERSION" -lt $(version_code 2.5.0) ]] &&
5673 skip "Need MDS version at least 2.5.0"
5676 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5677 do_nodes $(comma_list $(osts_nodes)) $LCTL set_param \
5678 seq.*OST*-super.width=$DATA_SEQ_MAX_WIDTH
5680 # use OST0000 since it probably has the most creations
5681 local OSTNAME=$(ostname_from_index 0)
5682 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
5683 local last_id=$(do_facet mds1 $LCTL get_param -n \
5684 osp.$mdtosc_proc1.prealloc_last_id)
5686 # Want to have OST LAST_ID over 5 * OST_MAX_PRECREATE to
5687 # verify that the LAST_ID recovery is working properly. If
5688 # not, then the OST will refuse to allow the MDS connect
5689 # because the LAST_ID value is too different from the MDS
5690 #define OST_MAX_PRECREATE=20000
5691 local ost_max_pre=20000
5692 local num_create=$(( ost_max_pre * 5 + 1 - last_id))
5694 # If the LAST_ID is already over 5 * OST_MAX_PRECREATE, we don't
5695 # need to create any files. So, skip this section.
5696 if [ $num_create -gt 0 ]; then
5697 # Check the number of inodes available on OST0
5699 local ifree=$($LFS df -i $MOUNT |
5700 awk '/OST0000/ { print $4 }'; exit ${PIPESTATUS[0]})
5701 log "On OST0, $ifree inodes available. Want $num_create. rc=$?"
5703 $LFS setstripe -i 0 $DIR/$tdir ||
5704 error "$LFS setstripe -i 0 $DIR/$tdir failed"
5705 if [ $ifree -lt 10000 ]; then
5706 files=$(( ifree - 50 ))
5711 local j=$((num_create / files + 1))
5712 for i in $(seq 1 $j); do
5713 createmany -o $DIR/$tdir/$tfile-$i- $files ||
5714 error "createmany fail create $files files: $?"
5715 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
5716 error "unlinkmany failed unlink $files files"
5720 # delete all of the files with objects on OST0 so the
5721 # filesystem is not inconsistent later on
5722 $LFS find $MOUNT --ost 0 -print0 | xargs -0 rm
5724 umount_client $MOUNT || error "umount client failed"
5725 stop_ost || error "OST0 stop failure"
5726 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
5727 $(ostdevname 1) $(ostvdevname 1) ||
5728 error "reformat and replace $ostdev failed"
5729 start_ost || error "OST0 restart failure"
5730 wait_osc_import_state mds ost FULL
5732 mount_client $MOUNT || error "mount client failed"
5733 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
5734 local idx=$($LFS getstripe -i $DIR/$tdir/$tfile-last)
5735 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
5737 local iused=$($LFS df -i $MOUNT |
5738 awk '/OST0000/ { print $3 }'; exit ${PIPESTATUS[0]})
5739 log "On OST0, $iused used inodes rc=$?"
5740 [ $iused -ge $((ost_max_pre + 1000)) ] &&
5741 error "OST replacement created too many inodes; $iused"
5742 cleanup || error "cleanup failed with $?"
5744 run_test 69 "replace an OST with the same index"
5747 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5750 cleanup || error "cleanup failed with $?"
5752 start_mdt 1 || error "MDT0 start fail"
5754 start_ost || error "OST0 start fail"
5755 for num in $(seq 2 $MDSCOUNT); do
5756 start_mdt $num || return
5759 mount_client $MOUNT || error "mount client fails"
5761 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
5763 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5764 error "create remote dir fail"
5766 rm -rf $DIR/$tdir || error "delete dir fail"
5767 cleanup || error "cleanup failed with $?"
5769 run_test 70a "start MDT0, then OST, then MDT1"
5772 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5775 start_ost || error "OST0 start fail"
5777 start_mds || error "MDS start fail"
5779 mount_client $MOUNT || error "mount client fails"
5781 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
5783 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5784 error "create remote dir fail"
5786 rm -rf $DIR/$tdir || error "delete dir fail"
5788 cleanup || error "cleanup failed with $?"
5790 run_test 70b "start OST, MDT1, MDT0"
5793 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5796 start_mds || error "MDS start fail"
5797 start_ost || error "OST0 start fail"
5799 mount_client $MOUNT || error "mount client fails"
5800 stop_mdt 1 || error "MDT1 start fail"
5802 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
5803 echo "deactivate $mdc_for_mdt1"
5804 $LCTL --device $mdc_for_mdt1 deactivate ||
5805 error "set $mdc_for_mdt1 deactivate failed"
5807 mkdir $DIR/$tdir && error "mkdir succeed"
5809 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
5810 error "create remote dir succeed"
5812 cleanup || error "cleanup failed with $?"
5814 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
5817 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5820 start_mds || error "MDS start fail"
5821 start_ost || error "OST0 start fail"
5823 mount_client $MOUNT || error "mount client fails"
5825 stop_mdt 2 || error "MDT1 start fail"
5827 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
5829 echo "deactivate $mdc_for_mdt2"
5830 $LCTL --device $mdc_for_mdt2 deactivate ||
5831 error "set $mdc_for_mdt2 deactivate failed"
5833 mkdir $DIR/$tdir || error "mkdir fail"
5834 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
5835 error "create remote dir succeed"
5837 rm -rf $DIR/$tdir || error "delete dir fail"
5839 cleanup || error "cleanup failed with $?"
5841 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
5844 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5846 [ "$MDS1_VERSION" -ge $(version_code 2.7.62) ] ||
5847 skip "Need MDS version at least 2.7.62"
5849 reformat || error "reformat failed with $?"
5853 local mdsdev=$(mdsdevname 1)
5854 local ostdev=$(ostdevname 1)
5855 local mdsvdev=$(mdsvdevname 1)
5856 local ostvdev=$(ostvdevname 1)
5857 local opts_mds="$(mkfs_opts mds1 $mdsdev) --reformat $mdsdev $mdsvdev"
5858 local opts_ost="$(mkfs_opts ost1 $ostdev) --reformat $ostdev $ostvdev"
5860 if ! combined_mgs_mds ; then
5864 add mds1 $opts_mds || error "add mds1 failed"
5865 start_mdt 1 || error "start mdt1 failed"
5866 add ost1 $opts_ost || error "add ost1 failed"
5867 start_ost || error "start ost failed"
5868 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
5870 local soc=$(do_facet mds1 "$LCTL get_param -n \
5871 mdt.*MDT0000.sync_lock_cancel")
5872 [ $soc == "never" ] || error "SoC enabled on single MDS"
5874 for i in $(seq 2 $MDSCOUNT); do
5875 mdsdev=$(mdsdevname $i)
5876 mdsvdev=$(mdsvdevname $i)
5877 opts_mds="$(mkfs_opts mds$i $mdsdev) --reformat $mdsdev \
5879 add mds$i $opts_mds || error "add mds$i failed"
5880 start_mdt $i || error "start mdt$i fail"
5883 wait_dne_interconnect
5885 for i in $(seq $MDSCOUNT); do
5886 soc=$(do_facet mds$i "$LCTL get_param -n \
5887 mdt.*MDT000$((i - 1)).sync_lock_cancel")
5888 [ $soc == "blocking" ] || error "SoC not enabled on DNE"
5891 for i in $(seq 2 $MDSCOUNT); do
5892 stop_mdt $i || error "stop mdt$i fail"
5894 soc=$(do_facet mds1 "$LCTL get_param -n \
5895 mdt.*MDT0000.sync_lock_cancel")
5896 [ $soc == "never" ] || error "SoC enabled on single MDS"
5897 umount_client $MOUNT -f > /dev/null
5899 cleanup || error "cleanup failed with $?"
5901 run_test 70e "Sync-on-Cancel will be enabled by default on DNE"
5904 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5905 if combined_mgs_mds; then
5906 skip "needs separate MGS/MDT"
5910 start_mdt 1 || error "MDT0 start fail"
5911 start_ost || error "OST0 start fail"
5912 for num in $(seq 2 $MDSCOUNT); do
5913 start_mdt $num || return
5916 start_ost2 || error "OST1 start fail"
5918 mount_client $MOUNT || error "mount client fails"
5920 mkdir $DIR/$tdir || error "mkdir fail"
5921 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5922 error "create remote dir succeed"
5924 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5925 rm -rf $DIR/$tdir || error "delete dir fail"
5927 umount_client $MOUNT || error "umount_client failed"
5928 stop_mds || error "MDS stop fail"
5929 stop_ost || error "OST0 stop fail"
5930 stop_ost2 || error "OST1 stop fail"
5932 run_test 71a "start MDT0 OST0, MDT1, OST1"
5935 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5936 if combined_mgs_mds; then
5937 skip "needs separate MGS/MDT"
5941 for num in $(seq 2 $MDSCOUNT); do
5942 start_mdt $num || return
5944 start_ost || error "OST0 start fail"
5945 start_mdt 1 || error "MDT0 start fail"
5946 start_ost2 || error "OST1 start fail"
5948 mount_client $MOUNT || error "mount client fails"
5950 mkdir $DIR/$tdir || error "mkdir fail"
5951 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5952 error "create remote dir succeed"
5954 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5955 rm -rf $DIR/$tdir || error "delete dir fail"
5957 umount_client $MOUNT || error "umount_client failed"
5958 stop_mds || error "MDT0 stop fail"
5959 stop_ost || error "OST0 stop fail"
5960 stop_ost2 || error "OST1 stop fail"
5962 run_test 71b "start MDT1, OST0, MDT0, OST1"
5965 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5966 combined_mgs_mds && skip "needs separate MGS/MDT"
5970 start_ost || error "OST0 start fail"
5971 start_ost2 || error "OST1 start fail"
5972 for num in $(seq 2 $MDSCOUNT); do
5973 start_mdt $num || return
5975 start_mdt 1 || error "MDT0 start fail"
5977 mount_client $MOUNT || error "mount client fails"
5979 mkdir $DIR/$tdir || error "mkdir fail"
5980 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5981 error "create remote dir succeed"
5983 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5984 rm -rf $DIR/$tdir || error "delete dir fail"
5986 umount_client $MOUNT || error "umount_client failed"
5987 stop_mds || error "MDS stop fail"
5988 stop_ost || error "OST0 stop fail"
5989 stop_ost2 || error "OST1 stop fail"
5992 run_test 71c "start OST0, OST1, MDT1, MDT0"
5995 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5996 combined_mgs_mds && skip "needs separate MGS/MDT"
6000 start_ost || error "OST0 start fail"
6001 for num in $(seq 2 $MDSCOUNT); do
6002 start_mdt $num || return
6004 start_mdt 1 || error "MDT0 start fail"
6005 start_ost2 || error "OST1 start fail"
6007 mount_client $MOUNT || error "mount client fails"
6009 mkdir $DIR/$tdir || error "mkdir fail"
6010 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
6011 error "create remote dir succeed"
6013 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
6014 rm -rf $DIR/$tdir || error "delete dir fail"
6016 umount_client $MOUNT || error "umount_client failed"
6017 stop_mds || error "MDS stop fail"
6018 stop_ost || error "OST0 stop fail"
6019 stop_ost2 || error "OST1 stop fail"
6022 run_test 71d "start OST0, MDT1, MDT0, OST1"
6025 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
6026 combined_mgs_mds && skip "needs separate MGS/MDT"
6030 start_ost || error "OST0 start fail"
6031 for num in $(seq 2 $MDSCOUNT); do
6032 start_mdt $num || return
6034 start_ost2 || error "OST1 start fail"
6035 start_mdt 1 || error "MDT0 start fail"
6037 mount_client $MOUNT || error "mount client fails"
6039 mkdir $DIR/$tdir || error "mkdir fail"
6040 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
6041 error "create remote dir succeed"
6043 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
6044 rm -rf $DIR/$tdir || error "delete dir fail"
6046 umount_client $MOUNT || error "umount_client failed"
6047 stop_mds || error "MDS stop fail"
6048 stop_ost || error "OST0 stop fail"
6049 stop_ost2 || error "OST1 stop fail"
6052 run_test 71e "start OST0, MDT1, OST1, MDT0"
6054 test_72() { #LU-2634
6055 [ "$mds1_FSTYPE" != ldiskfs ] &&
6056 skip "ldiskfs only test"
6058 local mdsdev=$(mdsdevname 1)
6059 local ostdev=$(ostdevname 1)
6060 local cmd="$E2FSCK -fnvd $mdsdev"
6067 if combined_mgs_mds; then
6068 add_options='--reformat'
6070 add_options='--reformat --replace'
6073 #tune MDT with "-O extents"
6075 for num in $(seq $MDSCOUNT); do
6076 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
6077 $add_options $(mdsdevname $num) $(mdsvdevname $num) ||
6078 error "add mds $num failed"
6079 do_facet mds${num} "$TUNE2FS -O extents $(mdsdevname $num)" ||
6080 error "$TUNE2FS failed on mds${num}"
6083 add ost1 $(mkfs_opts ost1 $ostdev) $add_options $ostdev ||
6084 error "add $ostdev failed"
6085 start_mds || error "start mds failed"
6086 start_ost || error "start ost failed"
6087 mount_client $MOUNT || error "mount client failed"
6089 #create some short symlinks
6090 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6091 createmany -o $DIR/$tdir/$tfile-%d $fn
6092 echo "create $fn short symlinks"
6093 for i in $(seq -w 1 $fn); do
6094 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
6099 umount_client $MOUNT || error "umount client failed"
6100 stop_mds || error "stop mds failed"
6101 stop_ost || error "stop ost failed"
6104 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
6106 run_test 72 "test fast symlink with extents flag enabled"
6108 test_73() { #LU-3006
6109 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
6110 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
6111 error "1st tunefs failed"
6112 start_mgsmds || error "start mds failed"
6113 start_ost || error "start ost failed"
6114 mount_client $MOUNT || error "mount client failed"
6115 $LCTL get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
6116 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
6117 umount_client $MOUNT || error "umount client failed"
6121 run_test 73 "failnode to update from mountdata properly"
6125 (( $MDS1_VERSION >= $(version_code 2.15.57.16) )) ||
6126 skip "need MDS version >= 2.15.57.16 for per-device timeouts"
6129 stack_trap "cleanup"
6131 # Prepare fs2, share the mgs of fs
6132 local FSNAME2=fs15246
6133 local fs2mdsdev=$(mdsdevname 1_2)
6134 local fs2ostdev=$(ostdevname 1_2)
6135 local fs2mdsvdev=$(mdsvdevname 1_2)
6136 local fs2ostvdev=$(ostvdevname 1_2)
6138 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --fsname=$FSNAME2 \
6139 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
6140 add fs2ost $(mkfs_opts ost1 $fs2ostdev) --fsname=$FSNAME2 \
6141 --reformat $fs2ostdev $fs2ostvdev || error "add fs2ost failed"
6143 stack_trap "cleanup_fs2"
6145 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
6146 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS || error "start fs2mds failed"
6148 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
6149 $MOUNT_CMD $MGSNID:/$FSNAME2 $MOUNT2 || error "mount $MOUNT2 failed"
6151 echo "========== All mounted lustre fs ===================="
6152 mount | grep 'type lustre'
6153 echo "====================================================="
6155 # Set and check osc/ldlm_enqueue_min
6156 $LCTL set_param osc.${FSNAME}-*.ldlm_enqueue_min=99 ||
6157 error "failed to set per-device adaptive parameters on client"
6158 stack_trap "$LCTL set_param osc.${FSNAME}-*.ldlm_enqueue_min=0"
6160 local ldlm_enqueue_min
6161 ldlm_enqueue_min=$($LCTL get_param -n osc.${FSNAME}-*.ldlm_enqueue_min \
6163 (( $ldlm_enqueue_min == 99 )) ||
6164 error "wrong ldlm_enqueue_min value for osc.${FSNAME}-*"
6166 # Check fs2 as reference
6167 ldlm_enqueue_min=$($LCTL get_param -n osc.${FSNAME2}-*.ldlm_enqueue_min\
6169 (( $ldlm_enqueue_min == 0 )) ||
6170 error "wrong ldlm_enqueue_min value for osc.${FSNAME2}-*"
6172 # Set and check obdfilter/at_min
6173 do_facet ost1 $LCTL set_param obdfilter.${FSNAME}-*.at_min=1 ||
6174 error "failed to set per-device adaptive parameters on ost"
6175 stack_trap "do_facet ost1 $LCTL set_param obdfilter.${FSNAME}-*.at_min=0"
6178 at_min=$(do_facet ost1 $LCTL get_param -n obdfilter.${FSNAME}-*.at_min \
6180 (( $at_min == 1 )) ||
6181 error "wrong at_min value for obdfilter.${FSNAME}-*"
6183 # set and check mdc/at_max
6184 $LCTL set_param mdc.${FSNAME}-*.at_max=599 ||
6185 error "failed to set per-device adaptive parameters on client"
6186 stack_trap "$LCTL set_param mdc.${FSNAME}-*.at_max=0"
6189 at_max=$($LCTL get_param -n mdc.${FSNAME}-*.at_max | uniq)
6190 (( $at_max == 599 )) ||
6191 error "wrong at_max value for osc.${FSNAME}-*"
6193 # Check fs2 as reference
6194 at_max=$($LCTL get_param -n mdc.${FSNAME2}-*.at_max | uniq)
6195 (( $at_max == 0 )) ||
6196 error "wrong at_max value for osc.${FSNAME2}-*"
6198 # Set and check mds/at_max
6199 do_facet mds1 $LCTL set_param *.${FSNAME}-*.at_max=599 ||
6200 error "failed to set per-device adaptive parameters on mds"
6201 stack_trap "do_facet mds1 $LCTL set_param *.${FSNAME}-*.at_max=0"
6204 at_max=$(do_facet mds1 $LCTL get_param -n mdt.${FSNAME}-*.at_max | uniq)
6205 (( at_max == 599 )) ||
6206 error "wrong at_max value for mdt.${FSNAME}-*"
6208 # Set and check mgs&mgc/at_history
6209 local mgs_nid=$(do_facet $SINGLEMDS $LCTL list_nids | tail -1)
6210 $LCTL set_param mgc.MGC$mgs_nid.at_history=588
6211 stack_trap "$LCTL set_param mgc.MGC$mgs_nid.at_history=0"
6214 at_history=$($LCTL get_param -n mgc.MGC$mgs_nid.at_history)
6215 (( $at_history == 588 )) ||
6216 error "wrong at_history value for mgc.MGC$mgs_nid"
6218 do_facet mgs $LCTL set_param mgs.MGS.at_history=588
6219 stack_trap "do_facet mgs $LCTL set_param mgs.MGS.at_history=0"
6220 at_history=$(do_facet mgs $LCTL get_param -n mgs.MGS.at_history)
6221 (( $at_history == 588 )) ||
6222 error "wrong at_history value for mgs.MGS"
6224 run_test 74 "Test per-device adaptive timeout parameters"
6226 test_75() { # LU-2374
6227 [[ "$MDS1_VERSION" -lt $(version_code 2.4.1) ]] &&
6228 skip "Need MDS version at least 2.4.1"
6231 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
6232 --replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
6233 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
6234 --replace --reformat $(ostdevname 1) $(ostvdevname 1)"
6237 #check with default parameters
6238 add mds1 $opts_mds || error "add mds1 failed for default params"
6239 add ost1 $opts_ost || error "add ost1 failed for default params"
6241 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
6242 opts_mds=$(echo $opts_mds |
6243 sed -e "s/--index=$index/--index=$index --mdt/")
6244 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
6245 opts_ost=$(echo $opts_ost |
6246 sed -e "s/--index=$index/--index=$index --ost/")
6248 add mds1 $opts_mds || error "add mds1 failed for new params"
6249 add ost1 $opts_ost || error "add ost1 failed for new params"
6254 run_test 75 "The order of --index should be irrelevant"
6257 [[ "$MGS_VERSION" -ge $(version_code 2.4.52) ]] ||
6258 skip "Need MDS version at least 2.4.52"
6261 local MDMB_PARAM="osc.*.max_dirty_mb"
6262 echo "Change MGS params"
6263 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
6265 echo "max_dirty_mb: $MAX_DIRTY_MB"
6266 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB - 10))
6267 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
6268 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
6269 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
6270 head -1" $NEW_MAX_DIRTY_MB
6271 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
6272 echo "$MAX_DIRTY_MB"
6273 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
6274 error "error while apply max_dirty_mb"
6276 echo "Check the value is stored after remount"
6279 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
6280 head -1" $NEW_MAX_DIRTY_MB
6281 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
6282 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
6283 error "max_dirty_mb is not saved after remount"
6285 echo "Change OST params"
6286 CLIENT_PARAM="obdfilter.*.client_cache_count"
6287 local CLIENT_CACHE_COUNT
6288 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
6290 echo "client_cache_count: $CLIENT_CACHE_COUNT"
6291 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
6292 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
6293 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
6294 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
6295 head -1" $NEW_CLIENT_CACHE_COUNT
6296 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
6298 echo "$CLIENT_CACHE_COUNT"
6299 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
6300 error "error while apply client_cache_count"
6302 echo "Check the value is stored after remount"
6305 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
6306 head -1" $NEW_CLIENT_CACHE_COUNT
6307 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
6309 echo "$CLIENT_CACHE_COUNT"
6310 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
6311 error "client_cache_count is not saved after remount"
6314 run_test 76a "set permanent params with lctl across mounts"
6316 test_76b() { # LU-4783
6317 [[ "$MGS_VERSION" -ge $(version_code 2.5.57) ]] ||
6318 skip "Need MGS version at least 2.5.57"
6321 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
6322 error "start params log failed"
6325 run_test 76b "verify params log setup correctly"
6328 [[ "$MGS_VERSION" -ge $(version_code 2.8.54) ]] ||
6329 skip "Need MDS version at least 2.4.52"
6331 local MASK_PARAM="mdd.*.changelog_mask"
6332 echo "Change changelog_mask"
6333 do_facet mgs $LCTL set_param -P $MASK_PARAM=-CLOSE ||
6334 error "Can't change changlog_mask"
6335 wait_update $(facet_host mds) "$LCTL get_param -n $MASK_PARAM |
6338 echo "Check the value is stored after mds remount"
6339 stop_mds || error "Failed to stop MDS"
6340 start_mds || error "Failed to start MDS"
6341 local CHANGELOG_MASK=$(do_facet mgs $LCTL get_param -n $MASK_PARAM)
6342 echo $CHANGELOG_MASK | grep CLOSE > /dev/null &&
6343 error "changelog_mask is not changed"
6347 run_test 76c "verify changelog_mask is applied with lctl set_param -P"
6349 test_76d() { #LU-9399
6352 local xattr_cache="llite.*.xattr_cache"
6353 local cmd="$LCTL get_param -n $xattr_cache | head -1"
6354 local new=$((($(eval $cmd) + 1) % 2))
6356 echo "lctl set_param -P llite.*.xattr_cache=$new"
6357 do_facet mgs $LCTL set_param -P $xattr_cache=$new ||
6358 error "Can't change xattr_cache"
6359 wait_update $HOSTNAME "$cmd" "$new"
6361 echo "Check $xattr_cache on client $MOUNT"
6362 umount_client $MOUNT || error "umount $MOUNT failed"
6363 mount_client $MOUNT || error "mount $MOUNT failed"
6364 [ $(eval $cmd) -eq $new ] ||
6365 error "$xattr_cache != $new on client $MOUNT"
6367 echo "Check $xattr_cache on the new client $MOUNT2"
6368 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
6369 [ $(eval $cmd) -eq $new ] ||
6370 error "$xattr_cache != $new on client $MOUNT2"
6371 umount_client $MOUNT2 || error "umount $MOUNT2 failed"
6375 run_test 76d "verify llite.*.xattr_cache can be set by 'lctl set_param -P' correctly"
6377 test_77() { # LU-3445
6378 [[ "$MDS1_VERSION" -ge $(version_code 2.8.55) ]] ||
6379 skip "Need MDS version 2.8.55+ "
6381 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
6382 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
6383 skip_env "mixed loopback and real device not working"
6386 local fs2mdsdev=$(mdsdevname 1_2)
6387 local fs2ostdev=$(ostdevname 1_2)
6388 local fs2mdsvdev=$(mdsvdevname 1_2)
6389 local fs2ostvdev=$(ostvdevname 1_2)
6390 local fsname=test1234
6392 local failnid="$(h2nettype 1.2.3.4),$(h2nettype 4.3.2.1)"
6394 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
6396 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
6397 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
6398 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
6399 error "start fs2mds failed"
6401 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
6402 mgsnid="0.0.0.0@tcp,$mgsnid,$mgsnid:$mgsnid"
6404 add fs2ost --mgsnode=$mgsnid $(mkfs_opts ost1 $fs2ostdev) \
6405 --failnode=$failnid --fsname=$fsname \
6406 --reformat $fs2ostdev $fs2ostvdev ||
6407 error "add fs2ost failed"
6408 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
6410 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
6411 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
6412 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
6415 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
6418 [[ "$mds1_FSTYPE" != ldiskfs ||
6419 "$ost1_FSTYPE" != ldiskfs ]] &&
6420 skip "ldiskfs only test"
6422 # reformat the Lustre filesystem with a smaller size
6423 local saved_MDSCOUNT=$MDSCOUNT
6424 local saved_MDSSIZE=$MDSSIZE
6425 local saved_OSTCOUNT=$OSTCOUNT
6426 local saved_OSTSIZE=$OSTSIZE
6429 MDSSIZE=$((MDSSIZE - 20000))
6430 OSTSIZE=$((OSTSIZE - 20000))
6431 reformat || error "(1) reformat Lustre filesystem failed"
6432 MDSSIZE=$saved_MDSSIZE
6433 OSTSIZE=$saved_OSTSIZE
6435 # mount the Lustre filesystem
6436 setup_noconfig || error "(2) setup Lustre filesystem failed"
6439 log "create test files"
6444 mkdir $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
6446 for i in $(seq $num_files); do
6447 file=$MOUNT/$tdir/$tfile-$i
6448 dd if=/dev/urandom of=$file count=1 bs=1M || {
6449 $LCTL get_param osc.*.cur*grant*
6450 $LFS df; $LFS df -i;
6451 # stop creating files if there is no more space
6452 if [ ! -e $file ]; then
6453 num_files=$((i - 1))
6457 $LFS getstripe -v $file
6458 local ost_idx=$(LFS getstripe -i $file)
6459 do_facet ost$((ost_idx + 1)) \
6460 $LCTL get_param obdfilter.*.*grant*
6461 error "(4) create $file failed"
6465 # unmount the Lustre filesystem
6466 cleanup || error "(5) cleanup Lustre filesystem failed"
6468 # run e2fsck on the MDT and OST devices
6469 local mds_host=$(facet_active_host $SINGLEMDS)
6470 local ost_host=$(facet_active_host ost1)
6471 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
6472 local ost_dev=$(ostdevname 1)
6474 run_e2fsck $mds_host $mds_dev "-y"
6475 run_e2fsck $ost_host $ost_dev "-y"
6477 # get the original block count of the MDT and OST filesystems
6478 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
6479 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
6481 # expand the MDT and OST filesystems to the device size
6482 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
6483 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
6485 # run e2fsck on the MDT and OST devices again
6486 run_e2fsck $mds_host $mds_dev "-y"
6487 run_e2fsck $ost_host $ost_dev "-y"
6489 # mount the Lustre filesystem
6493 log "check files after expanding the MDT and OST filesystems"
6494 for i in $(seq $num_files); do
6495 file=$MOUNT/$tdir/$tfile-$i
6496 $CHECKSTAT -t file -s 1048576 $file ||
6497 error "(6) checkstat $file failed"
6501 log "create more files after expanding the MDT and OST filesystems"
6502 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
6503 file=$MOUNT/$tdir/$tfile-$i
6504 dd if=/dev/urandom of=$file count=1 bs=1M ||
6505 error "(7) create $file failed"
6508 # unmount the Lustre filesystem
6509 cleanup || error "(8) cleanup Lustre filesystem failed"
6511 # run e2fsck on the MDT and OST devices
6512 run_e2fsck $mds_host $mds_dev "-y"
6513 run_e2fsck $ost_host $ost_dev "-y"
6515 # get the maximum block count of the MDT and OST filesystems
6516 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
6517 local ost_max_blks=$(get_block_count ost1 $ost_dev)
6519 # get the minimum block count of the MDT and OST filesystems
6520 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
6521 grep minimum | sed -e 's/^.*filesystem: //g')
6522 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
6523 grep minimum | sed -e 's/^.*filesystem: //g')
6525 # shrink the MDT and OST filesystems to a smaller size
6529 if [[ $mds_max_blks -gt $mds_min_blks &&
6530 $mds_max_blks -gt $mds_orig_blks ]]; then
6531 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
6532 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
6533 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
6534 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
6535 error "shrink $SINGLEMDS to $new_blks failed"
6539 if [[ $ost_max_blks -gt $ost_min_blks &&
6540 $ost_max_blks -gt $ost_orig_blks ]]; then
6541 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
6542 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
6543 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
6544 run_resize2fs ost1 $ost_dev $new_blks ||
6545 error "shrink ost1 to $new_blks failed"
6549 # check whether the MDT or OST filesystem was shrunk or not
6551 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
6552 reformat_and_config ||
6553 error "(10) reformat Lustre filesystem failed"
6557 # run e2fsck on the MDT and OST devices again
6558 run_e2fsck $mds_host $mds_dev "-y"
6559 run_e2fsck $ost_host $ost_dev "-y"
6561 # mount the Lustre filesystem again
6565 log "check files after shrinking the MDT and OST filesystems"
6566 for i in $(seq $((num_files + 10))); do
6567 file=$MOUNT/$tdir/$tfile-$i
6568 $CHECKSTAT -t file -s 1048576 $file ||
6569 error "(11) checkstat $file failed"
6572 # unmount and reformat the Lustre filesystem
6573 cleanup || error "(12) cleanup Lustre filesystem failed"
6574 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
6576 MDSCOUNT=$saved_MDSCOUNT
6577 OSTCOUNT=$saved_OSTCOUNT
6578 reformat_and_config || error "(14) reformat Lustre filesystem failed"
6580 run_test 78 "run resize2fs on MDT and OST filesystems"
6582 test_79() { # LU-4227
6583 [[ "$MDS1_VERSION" -ge $(version_code 2.5.59) ]] ||
6584 skip "Need MDS version at least 2.5.59"
6586 local mdsdev1=$(mdsdevname 1)
6587 local mdsvdev1=$(mdsvdevname 1)
6588 local mdsdev2=$(mdsdevname 2)
6589 local mdsvdev2=$(mdsvdevname 2)
6590 local ostdev1=$(ostdevname 1)
6591 local ostvdev1=$(ostvdevname 1)
6592 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
6593 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
6594 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
6597 # remove --mgs/--mgsnode from mkfs.lustre options
6598 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
6600 mgsnode_opt=$(echo $opts_mds2 |
6601 awk '{ for ( i = 1; i < NF; i++ )
6602 if ( $i ~ "--mgsnode" ) { print $i; break } }')
6603 [ -n "$mgsnode_opt" ] &&
6604 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
6606 mgsnode_opt=$(echo $opts_ost1 |
6607 awk '{ for ( i = 1; i < NF; i++ )
6608 if ( $i ~ "--mgsnode" ) { print $i; break } }')
6609 [ -n "$mgsnode_opt" ] &&
6610 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
6612 # -MGS, format a mdt without --mgs option
6613 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
6614 error "Must specify --mgs when formatting mdt combined with mgs"
6616 # +MGS, format a mdt/ost without --mgsnode option
6617 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
6618 > /dev/null || error "start mds1 failed"
6619 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
6620 error "Must specify --mgsnode when formatting a mdt"
6621 add ost1 $opts_ost1 $ostdev1 $ostvdev1 &&
6622 error "Must specify --mgsnode when formatting an ost"
6626 run_test 79 "format MDT/OST without mgs option (should return errors)"
6629 start_mds || error "Failed to start MDT"
6630 start_ost || error "Failed to start OST1"
6631 uuid=$(do_facet ost1 $LCTL get_param -n mgc.*.uuid)
6632 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
6633 do_facet ost1 "$LCTL set_param fail_val=10 fail_loc=0x906"
6634 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x906"
6635 do_facet mgs "$LCTL set_param -n mgs/MGS/evict_client $uuid"
6637 start_ost2 || error "Failed to start OST2"
6639 do_facet ost1 "$LCTL set_param fail_loc=0"
6644 run_test 80 "mgc import reconnect race"
6646 #Save the original values of $OSTCOUNT and $OSTINDEX$i.
6648 local new_ostcount=$1
6649 saved_ostcount=$OSTCOUNT
6650 OSTCOUNT=$new_ostcount
6654 for ((i = 1; i <= $OSTCOUNT; i++ )); do
6656 eval saved_ostindex$i=${!index}
6661 # Restore the original values of $OSTCOUNT and $OSTINDEX$i.
6662 restore_ostindex() {
6666 echo "restoring OSTCOUNT=$saved_ostcount and OSTINDEXn"
6668 for ((i = 1; i <= $OSTCOUNT; i++ )); do
6669 index=saved_ostindex$i
6670 eval OSTINDEX$i=${!index}
6672 OSTCOUNT=$saved_ostcount
6675 if ! combined_mgs_mds ; then
6680 # The main purpose of this test is to ensure the OST_INDEX_LIST functions as
6681 # expected. This test uses OST_INDEX_LIST to format OSTs with a randomly
6682 # assigned index and ensures we can mount such a formatted file system
6683 test_81() { # LU-4665
6684 (( MDS1_VERSION >= $(version_code 2.6.54) )) ||
6685 skip "Need MDS version at least 2.6.54"
6686 (( OSTCOUNT >= 3 )) || skip_env "needs >= 3 OSTs"
6690 # Each time RANDOM is referenced, a random integer between 0 and 32767
6693 local saved_ostindex1=$OSTINDEX1
6694 for i in 65535 $((RANDOM + 65536)); do
6695 echo -e "\nFormat ost1 with --index=$i, should fail"
6697 if add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat \
6698 $(ostdevname 1) $(ostvdevname 1); then
6699 OSTINDEX1=$saved_ostindex1
6700 error "format ost1 with --index=$i should fail"
6703 OSTINDEX1=$saved_ostindex1
6706 stack_trap restore_ostindex
6708 # Format OSTs with random sparse indices.
6709 local rand_ost=$(((RANDOM * 2 % 65533) + 1))
6710 echo "Format $OSTCOUNT OSTs with OST_INDEX_LIST=[0,$rand_ost,65534]"
6711 OST_INDEX_LIST=[0,$rand_ost,65534] formatall ||
6712 error "formatall failed with $?"
6714 # Setup and check Lustre filesystem.
6715 start_mgsmds || error "start_mgsmds failed"
6716 for ((i = 1; i <= $OSTCOUNT; i++ )); do
6717 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6718 error "start ost$i failed"
6721 mount_client $MOUNT || error "mount client $MOUNT failed"
6722 check_mount || error "check client $MOUNT failed"
6726 local max_easize=$($LCTL get_param -n llite.*.max_easize)
6727 local xattr_size_max=$((4096 - 84)) # 4096 less ldiskfs ea overhead
6729 # XATTR_SIZE_MAX less ldiskfs ea overhead
6730 large_xattr_enabled && xattr_size_max=$((65536 - 84))
6731 (( max_easize >= xattr_size_max )) ||
6732 error "max_easize $max_easize < $xattr_size_max bytes"
6734 test_mkdir $DIR/$tdir
6735 $LFS setstripe -i $rand_ost $DIR/$tdir/$tfile ||
6736 error "error creating $tfile on ost$rand_ost"
6737 $LFS getstripe $DIR/$tdir/$tfile
6744 cmd="getstripe -i $DIR/$tdir/$tfile"
6747 (( $found == $rand_ost )) || error "index $found is not $rand_ost"
6749 cmd="find $DIR/$tdir -i $rand_ost"
6753 [[ "$found" == "$DIR/$tdir/$tfile" ]] ||
6754 error "'lfs find' returned '$found', not '$tfile' by index"
6757 uuid=$(ostuuid_from_index $rand_ost)
6759 cmd="find $DIR/$tdir -O $uuid"
6764 [[ "$found" == "$DIR/$tdir/$tfile" ]] ||
6765 error "'lfs find' returned '$found', not '$tfile' by UUID"
6767 run_test 81 "sparse OST indexing"
6769 # Here we exercise the stripe placement functionality on a file system that
6770 # has formatted the OST with a random index. With the file system the following
6771 # functionality is tested:
6773 # 1. Creating a new file with a specific stripe layout.
6775 # 2. Modifiy a existing empty file with a specific stripe layout.
6777 # 3. Ensure we fail to set the stripe layout of a file that already has one.
6779 # 4. If ost-index is defined we need to ensure it is the first entry in the
6780 # ost index list returned by lfs getstripe.
6782 # 5. Lastly ensure this functionality fails with directories.
6783 test_82a() { # LU-4665
6784 [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
6785 skip "Need MDS version at least 2.6.54"
6786 [[ $OSTCOUNT -ge 3 ]] || skip_env "needs >= 3 OSTs"
6792 # Format OSTs with random sparse indices.
6796 local LOV_V1_INSANE_STRIPE_COUNT=65532
6797 for i in $(seq $OSTCOUNT); do
6798 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
6799 ost_indices+=" $index"
6801 ost_indices=$(comma_list $ost_indices)
6803 stack_trap "restore_ostindex" EXIT
6804 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
6805 OST_INDEX_LIST=[$ost_indices] formatall
6807 # Setup Lustre filesystem.
6808 start_mgsmds || error "start_mgsmds failed"
6809 for i in $(seq $OSTCOUNT); do
6810 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6811 error "start ost$i failed"
6814 # Collect debug information - start of test
6815 do_nodes $(comma_list $(mdts_nodes)) \
6816 $LCTL get_param osp.*.prealloc_*_id
6818 mount_client $MOUNT || error "mount client $MOUNT failed"
6822 check_lfs_df_ret_val $? || error "$LFS df $MOUNT failed"
6823 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6825 stack_trap "do_nodes $(comma_list $(mdts_nodes)) \
6826 $LCTL get_param osp.*.prealloc_*_id || true" EXIT
6828 # 1. If the file does not exist, new file will be created
6829 # with specified OSTs.
6830 local file=$DIR/$tdir/$tfile-1
6831 local cmd="$LFS setstripe -o $ost_indices $file"
6833 eval $cmd || error "$cmd failed"
6834 check_stripe_count $file $OSTCOUNT
6835 check_obdidx $file $ost_indices
6836 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6837 error "write $file failed"
6839 # 2. If the file already exists and is an empty file, the file
6840 # will be attached with specified layout.
6841 file=$DIR/$tdir/$tfile-2
6842 mcreate $file || error "mcreate $file failed"
6843 cmd="$LFS setstripe -o $ost_indices $file"
6845 eval $cmd || error "$cmd failed"
6846 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6847 error "write $file failed"
6848 check_stripe_count $file $OSTCOUNT
6849 check_obdidx $file $ost_indices
6851 # 3. If the file already has a valid layout attached, the command
6852 # should fail with EBUSY.
6854 eval $cmd && error "stripe is already set on $file, $cmd should fail"
6856 # 4. If [--stripe-index|-i <start_ost_idx>] is used, the index must
6857 # be in the OST indices list.
6858 local start_ost_idx=${ost_indices##*,}
6859 file=$DIR/$tdir/$tfile-3
6860 cmd="$LFS setstripe -o $ost_indices -i $start_ost_idx $file"
6862 eval $cmd || error "$cmd failed"
6863 check_stripe_count $file $OSTCOUNT
6864 check_obdidx $file $ost_indices
6865 check_start_ost_idx $file $start_ost_idx
6867 file=$DIR/$tdir/$tfile-4
6868 cmd="$LFS setstripe"
6869 cmd+=" -o $(exclude_items_from_list $ost_indices $start_ost_idx)"
6870 cmd+=" -i $start_ost_idx $file"
6872 eval $cmd && error "index $start_ost_idx should be in $ost_indices"
6874 # 5. Specifying OST indices for directory should succeed.
6875 local dir=$DIR/$tdir/$tdir
6876 mkdir $dir || error "mkdir $dir failed"
6877 cmd="$LFS setstripe -o $ost_indices $dir"
6878 if [ "$MDS1_VERSION" -gt $(version_code 2.11.53) ] &&
6879 [ "$CLIENT_VERSION" -gt $(version_code 2.11.53) ]; then
6881 eval $cmd || error "unable to specify OST indices on directory"
6883 echo "need MDS+client version at least 2.11.53"
6886 run_test 82a "specify OSTs for file (succeed) or directory (succeed)"
6888 # Test 82b is run to ensure that if the user supplies a pool with a specific
6889 # stripe layout that it behaves proprerly. It should fail in the case that
6890 # the supplied OST index list points to OSTs not contained in the user
6892 test_82b() { # LU-4665
6893 [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
6894 skip "Need MDS version at least 2.6.54"
6895 [[ $OSTCOUNT -ge 4 ]] || skip_env "needs >= 4 OSTs"
6901 # Format OSTs with random sparse indices.
6905 local LOV_V1_INSANE_STRIPE_COUNT=65532
6906 for i in $(seq $OSTCOUNT); do
6907 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
6908 ost_indices+=" $index"
6910 ost_indices=$(comma_list $ost_indices)
6912 stack_trap "restore_ostindex" EXIT
6913 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
6914 OST_INDEX_LIST=[$ost_indices] formatall
6916 # Setup Lustre filesystem.
6917 start_mgsmds || error "start_mgsmds failed"
6918 for i in $(seq $OSTCOUNT); do
6919 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6920 error "start ost$i failed"
6923 mount_client $MOUNT || error "mount client $MOUNT failed"
6927 check_lfs_df_ret_val $? || error "$LFS df $MOUNT failed"
6928 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6930 # Create a new pool and add OSTs into it.
6931 local ost_pool=$FSNAME.$TESTNAME
6932 create_pool $ost_pool || error "create OST pool $ost_pool failed"
6934 local ost_idx_in_list=${ost_indices##*,}
6935 local ost_idx_in_pool=$(exclude_items_from_list $ost_indices \
6938 local ost_targets="$FSNAME-OST["
6939 for i in ${ost_idx_in_pool//,/ }; do
6940 ost_targets=$ost_targets$(printf "%04x," $i)
6942 ost_targets="${ost_targets%,}]"
6944 local ost_targets_uuid=$(for i in ${ost_idx_in_pool//,/ }; \
6945 do printf "$FSNAME-OST%04x_UUID\n" $i; done |
6946 sort -u | tr '\n' ' ')
6948 local cmd="$LCTL pool_add $ost_pool $ost_targets"
6949 do_facet mgs $cmd || error "$cmd failed"
6950 wait_update $HOSTNAME "$LCTL get_param -n lov.$FSNAME-*.pools.$TESTNAME|
6951 sort -u | tr '\n' ' ' " "$ost_targets_uuid" ||
6952 error "wait_update $ost_pool failed"
6953 wait_update_facet $SINGLEMDS "$LCTL pool_list $ost_pool | wc -l" 4 ||
6954 error "wait_update pool_list $ost_pool failed"
6956 # If [--pool|-p <pool_name>] is set with [--ost-list|-o <ost_indices>],
6957 # then the OSTs must be the members of the pool.
6958 local file=$DIR/$tdir/$tfile
6959 cmd="$LFS setstripe -p $ost_pool -o $ost_idx_in_list $file"
6961 eval $cmd && error "OST with index $ost_idx_in_list should be" \
6962 "in OST pool $ost_pool"
6964 # Only select OST $ost_idx_in_list from $ost_pool for file.
6965 ost_idx_in_list=${ost_idx_in_pool#*,}
6966 cmd="$LFS setstripe -p $ost_pool -o $ost_idx_in_list $file"
6968 eval $cmd || error "$cmd failed"
6969 cmd="$LFS getstripe $file"
6971 eval $cmd || error "$cmd failed"
6972 check_stripe_count $file 2
6973 check_obdidx $file $ost_idx_in_list
6974 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6975 error "write $file failed"
6977 run_test 82b "specify OSTs for file with --pool and --ost-list options"
6980 [[ "$OST1_VERSION" -ge $(version_code 2.6.91) ]] ||
6981 skip "Need OST version at least 2.6.91"
6982 if [ "$ost1_FSTYPE" != ldiskfs ]; then
6983 skip "ldiskfs only test"
6992 ostmnt=$(facet_mntpt ost1)
6994 # Mount the OST as an ldiskfs filesystem.
6995 log "mount the OST $dev as a $ost1_FSTYPE filesystem"
6996 add ost1 $(mkfs_opts ost1 $dev) $FSTYPE_OPT \
6997 --reformat $dev > /dev/null ||
6998 error "format ost1 error"
7000 if ! test -b $dev; then
7001 mnt_opts=$(csa_add "$OST_MOUNT_FS_OPTS" -o loop)
7003 echo "mnt_opts $mnt_opts"
7004 do_facet ost1 mount -t "$ost1_FSTYPE" $dev \
7006 # Run llverfs on the mounted ldiskfs filesystem.
7007 # It is needed to get ENOSPACE.
7008 log "run llverfs in partial mode on the OST $ost1_FSTYPE $ostmnt"
7009 do_rpc_nodes $(facet_host ost1) run_llverfs $ostmnt -vpl \
7010 "no" || error "run_llverfs error on $ost1_FSTYPE"
7013 log "unmount the OST $dev"
7016 # Delete file IO_scrub. Later osd_scrub_setup will try to
7017 # create "IO_scrub" but will get ENOSPACE.
7019 echo "start ost1 service on `facet_active_host ost1`"
7020 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
7023 err=$(do_facet ost1 dmesg | grep "VFS: Busy inodes after unmount of")
7024 echo "string err $err"
7025 [ -z "$err" ] || error $err
7028 run_test 83 "ENOSPACE on OST doesn't cause message VFS: \
7029 Busy inodes after unmount ..."
7032 local facet=$SINGLEMDS
7033 local num=$(echo $facet | tr -d "mds")
7034 local dev=$(mdsdevname $num)
7035 local time_min=$(recovery_time_min)
7036 local recovery_duration
7037 local completed_clients
7038 local correct_clients
7041 echo "start mds service on $(facet_active_host $facet)"
7043 "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
7044 error "start MDS failed"
7046 start_ost || error "start OST0000 failed"
7047 wait_osc_import_state mds ost1 FULL
7048 start_ost2 || error "start OST0001 failed"
7049 wait_osc_import_state mds ost2 FULL
7051 echo "recovery_time=$time_min, timeout=$TIMEOUT, wrap_up=$wrap_up"
7053 mount_client $MOUNT1 || error "mount $MOUNT1 failed"
7054 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
7055 # make sure new superblock labels are sync'd before disabling writes
7059 replay_barrier $SINGLEMDS
7060 createmany -o $DIR1/$tfile-%d 1000
7062 # We need to catch the end of recovery window to extend it.
7063 # Skip 5 requests and add delay to request handling.
7064 #define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP
7065 do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5"
7067 facet_failover --fsck $SINGLEMDS || error "failover: $?"
7070 echo "recovery status"
7071 do_facet $SINGLEMDS \
7072 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status"
7074 recovery_duration=$(do_facet $SINGLEMDS \
7075 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
7076 awk '/recovery_duration/ { print $2 }')
7077 (( $recovery_duration > $time_min + $wrap_up )) &&
7078 error "recovery_duration > recovery_time_hard + wrap up"
7079 completed_clients=$(do_facet $SINGLEMDS \
7080 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
7081 awk '/completed_clients/ { print $2 }')
7083 correct_clients="$MDSCOUNT/$((MDSCOUNT+1))"
7084 [ "$completed_clients" = "${correct_clients}" ] ||
7085 error "$completed_clients != $correct_clients"
7087 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
7088 umount_client $MOUNT1
7089 umount_client $MOUNT2
7095 run_test 84 "check recovery_hard_time"
7098 [[ "$OST1_VERSION" -ge $(version_code 2.7.55) ]] ||
7099 skip "Need OST version at least 2.7.55"
7100 ##define OBD_FAIL_OSD_OST_EA_FID_SET 0x197
7101 do_facet ost1 "lctl set_param fail_loc=0x197"
7105 run_test 85 "osd_ost init: fail ea_fid_set"
7110 # ost1 has already registered to the MGS before the reformat.
7111 # So after reformatting it with option "-G", it could not be
7112 # mounted to the MGS. Cleanup the system for subsequent tests.
7117 [ "$ost1_FSTYPE" = zfs ] &&
7118 skip "LU-6442: no such mkfs params for ZFS OSTs"
7119 [[ "$MDS1_VERSION" -ge $(version_code 2.7.56) ]] ||
7120 skip "Need server version newer than 2.7.55"
7123 local OLDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
7124 awk '/Flex block group size: / { print $NF; exit; }')
7126 [ "$OLDSIZE" == "$NEWSIZE" ] && skip "$NEWSIZE groups already"
7128 local opts=" -O flex_bg -G $NEWSIZE"
7129 opts=$(OST_FS_MKFS_OPTS+="$opts" mkfs_opts ost1 $(ostdevname 1))
7130 opts+=" --reformat $(ostdevname 1) $(ostvdevname 1)"
7131 echo "params: $opts"
7133 trap cleanup_86 EXIT ERR
7136 add ost1 $opts || error "add ost1 failed with new params"
7138 local FOUNDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
7139 awk '/Flex block group size: / { print $NF; exit; }')
7141 [[ $FOUNDSIZE == $NEWSIZE ]] ||
7142 error "Flex block group size: $FOUNDSIZE, expected: $NEWSIZE"
7146 run_test 86 "Replacing mkfs.lustre -G option"
7148 test_87() { #LU-6544
7149 [[ "$MDS1_VERSION" -ge $(version_code 2.9.51) ]] ||
7150 skip "Need MDS version at least 2.9.51"
7151 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
7152 skip "ldiskfs only test"
7153 [[ $OSTCOUNT -gt 59 ]] &&
7154 skip "Ignore wide striping situation"
7155 [ -n "$FILESET" ] && skip "Not functional for FILESET set"
7157 local mdsdev=$(mdsdevname 1)
7158 local mdsvdev=$(mdsvdevname 1)
7159 local file=$DIR/$tfile
7160 local mntpt=$(facet_mntpt $SINGLEMDS)
7161 local used_xattr_blk=0
7162 local inode_size=${1:-1024}
7164 local xtest="trusted.test"
7168 local stripe_cnt=$(($OSTCOUNT + 2))
7170 #Please see ldiskfs_make_lustre() for MDT inode size calculation
7171 if [ $stripe_cnt -gt 16 ]; then
7174 left_size=$(expr $inode_size - \
7177 32 - 40 \* 3 - 32 \* 3 - $stripe_cnt \* 24 - 16 - 3 - \
7179 24 - 18 - $(expr length $tfile) - 16 - 4)
7180 if [ $left_size -le 0 ]; then
7181 echo "No space($left_size) is expected in inode."
7182 echo "Try 1-byte xattr instead to verify this."
7185 echo "Estimate: at most $left_size-byte space left in inode."
7191 if ! combined_mgs_mds ; then
7195 add mds1 $(mkfs_opts mds1 ${mdsdev}) --stripe-count-hint=$stripe_cnt \
7196 --reformat $mdsdev $mdsvdev || error "add mds1 failed"
7197 start_mdt 1 > /dev/null || error "start mdt1 failed"
7198 for i in $(seq $OSTCOUNT); do
7199 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS > /dev/null ||
7200 error "start ost$i failed"
7202 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
7203 check_mount || error "check client $MOUNT failed"
7206 $LFS setstripe -E 1M -S 1M -c 1 -E 64M -c 1 -E -1 -c -1 $file ||
7207 error "Create file with 3 components failed"
7208 $TRUNCATE $file $((1024*1024*64+1)) || error "truncate file failed"
7209 i=$($LFS getstripe -I3 -c $file) || error "get 3rd stripe count failed"
7210 if [ $i -ne $OSTCOUNT ]; then
7211 left_size=$(expr $left_size + $(expr $OSTCOUNT - $i) \* 24)
7212 echo -n "Since only $i out $OSTCOUNT OSTs are used, "
7213 echo -n "the expected left space is changed to "
7214 echo "$left_size bytes at most."
7216 value=$(generate_string $left_size)
7217 setfattr -n $xtest -v $value $file
7218 orig=$(get_xattr_value $xtest $file)
7219 [[ "$orig" != "$value" ]] && error "$xtest changed"
7221 #Verify if inode has some expected space left
7222 umount $MOUNT > /dev/null || error "umount $MOUNT failed"
7223 stop_mdt 1 > /dev/null || error "stop mdt1 failed"
7224 mount_ldiskfs $SINGLEMDS || error "mount -t ldiskfs $SINGLEMDS failed"
7226 do_facet $SINGLEMDS ls -sal $mntpt/ROOT/$tfile
7227 used_xattr_blk=$(do_facet $SINGLEMDS ls -s $mntpt/ROOT/$tfile |
7229 [[ $used_xattr_blk -eq 0 ]] &&
7230 error "Please check MDS inode size calculation: \
7231 more than $left_size-byte space left in inode."
7232 echo "Verified: at most $left_size-byte space left in inode."
7234 unmount_ldiskfs $SINGLEMDS
7236 for i in $(seq $OSTCOUNT); do
7237 stop ost$i -f || error "stop ost$i failed"
7240 run_test 87 "check if MDT inode can hold EAs with N stripes properly"
7243 [ "$mds1_FSTYPE" == zfs ] &&
7244 skip "LU-6662: no implementation for ZFS"
7248 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
7249 --reformat $(mdsdevname 1) || error "add mds1 failed"
7251 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
7252 grep -e \".*opts:.*errors=remount-ro.*\"" ||
7253 error "default mount options is missing"
7255 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
7256 --mountfsoptions="user_xattr,errors=panic" \
7257 --reformat $(mdsdevname 1) || error "add mds1 failed"
7259 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
7260 grep -e \".*opts:.*errors=panic.*\"" ||
7261 error "user can't override default mount options"
7263 run_test 88 "check the default mount options can be overridden"
7265 test_89() { # LU-7131
7266 [[ "$MDS1_VERSION" -ge $(version_code 2.9.54) ]] ||
7267 skip "Need MDT version at least 2.9.54"
7269 local key=failover.node
7270 local val1=192.0.2.254@tcp0 # Reserved IPs, see RFC 5735
7271 local val2=192.0.2.255@tcp0
7272 local mdsdev=$(mdsdevname 1)
7277 if [[ "$mds1_FSTYPE" == zfs ]]; then
7278 import_zpool mds1 || return ${PIPESTATUS[0]}
7281 # Check that parameters are added correctly
7282 echo "tunefs --param $key=$val1"
7283 do_facet mds1 "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
7284 error "tunefs --param $key=$val1 failed"
7285 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7286 error "tunefs --dryrun failed"
7287 params=${params##*Parameters:}
7288 params=${params%%exiting*}
7289 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val1) = "1" ] ||
7290 error "on-disk parameter not added correctly via tunefs"
7292 # Check that parameters replace existing instances when added
7293 echo "tunefs --param $key=$val2"
7294 do_facet mds1 "$TUNEFS --param $key=$val2 $mdsdev >/dev/null" ||
7295 error "tunefs --param $key=$val2 failed"
7296 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7297 error "tunefs --dryrun failed"
7298 params=${params##*Parameters:}
7299 params=${params%%exiting*}
7300 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "1" ] ||
7301 error "on-disk parameter not replaced via tunefs"
7302 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val2) = "1" ] ||
7303 error "on-disk parameter not replaced correctly via tunefs"
7305 # Check that a parameter is erased properly
7306 echo "tunefs --erase-param $key"
7307 do_facet mds1 "$TUNEFS --erase-param $key $mdsdev >/dev/null" ||
7308 error "tunefs --erase-param $key failed"
7309 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7310 error "tunefs --dryrun failed"
7311 params=${params##*Parameters:}
7312 params=${params%%exiting*}
7313 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "0" ] ||
7314 error "on-disk parameter not erased correctly via tunefs"
7316 # Check that all the parameters are erased
7317 do_facet mds1 "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
7318 error "tunefs --param $key=$val1 failed"
7319 echo "tunefs --erase-params"
7320 do_facet mds1 "$TUNEFS --erase-params $mdsdev >/dev/null" ||
7321 error "tunefs --erase-params failed"
7322 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7323 error "tunefs --dryrun failed"
7324 params=${params##*Parameters:}
7325 params=${params%%exiting*}
7326 params=$(echo $params | tr ' ' '\n')
7328 error "all on-disk parameters not erased correctly via tunefs $params"
7330 # Check the order of options --erase-params and --param
7331 echo "tunefs --param $key=$val1 --erase-params"
7333 "$TUNEFS --param $key=$val1 --erase-params $mdsdev >/dev/null"||
7334 error "tunefs --param $key=$val1 --erase-params failed"
7335 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7336 error "tunefs --dryrun failed"
7337 params=${params##*Parameters:}
7338 params=${params%%exiting*}
7339 [ $(echo $params | tr ' ' '\n') == "$key=$val1" ] ||
7340 error "on-disk param not added correctly with --erase-params"
7344 run_test 89 "check tunefs --param and --erase-param{s} options"
7347 # $2 (optional) value of max_mod_rpcs_in_flight to set
7348 check_max_mod_rpcs_in_flight() {
7356 idx=$(printf "%04x" $($LFS getdirstripe -i $dir))
7357 facet="mds$((0x$idx + 1))"
7359 if [ -z "$mmr" ]; then
7360 # get value of max_mod_rcps_in_flight
7361 mmr=$($LCTL get_param -n \
7362 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) ||
7363 error "Unable to get max_mod_rpcs_in_flight"
7364 echo "max_mod_rcps_in_flight is $mmr"
7366 # set value of max_mod_rpcs_in_flight
7368 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight=$mmr ||
7369 error "Unable to set max_mod_rpcs_in_flight to $mmr"
7370 echo "max_mod_rpcs_in_flight set to $mmr"
7373 # create mmr+1 files
7374 echo "creating $((mmr + 1)) files ..."
7376 for i in $(seq $((mmr + 1))); do
7382 # consumes mmr-1 modify RPC slots
7383 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7384 # drop requests on MDT so that RPC slots are consumed
7385 # during all the request resend interval
7386 do_facet $facet "$LCTL set_param fail_loc=0x159"
7387 echo "launch $((mmr - 1)) chmod in parallel ..."
7388 for i in $(seq $((mmr - 1))); do
7389 chmod 0600 $dir/file-$i &
7393 # send one additional modify RPC
7394 do_facet $facet "$LCTL set_param fail_loc=0"
7395 echo "launch 1 additional chmod in parallel ..."
7396 chmod 0600 $dir/file-$mmr &
7399 # check this additional modify RPC get a modify RPC slot
7400 # and succeed its operation
7401 checkstat -vp 0600 $dir/file-$mmr ||
7402 error "Unable to send $mmr modify RPCs in parallel"
7407 # consumes mmr modify RPC slots
7408 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7409 # drop requests on MDT so that RPC slots are consumed
7410 # during all the request resend interval
7411 do_facet $facet "$LCTL set_param fail_loc=0x159"
7412 echo "launch $mmr chmod in parallel ..."
7413 for i in $(seq $mmr); do
7414 chmod 0666 $dir/file-$i &
7418 # send one additional modify RPC
7419 do_facet $facet "$LCTL set_param fail_loc=0"
7420 echo "launch 1 additional chmod in parallel ..."
7421 chmod 0666 $dir/file-$((mmr + 1)) &
7424 # check this additional modify RPC blocked getting a modify RPC slot
7425 checkstat -vp 0644 $dir/file-$((mmr + 1)) ||
7426 error "Unexpectedly send $(($mmr + 1)) modify RPCs in parallel"
7430 get_mdt_max_mod_rpcs_in_flight_val() {
7434 # It will be enough to get one value from one MDT
7435 max_mrif=$(do_facet $facet \
7436 "$LCTL get_param -n \
7437 mdt.$FSNAME-MDT*.max_mod_rpcs_in_flight | head -n 1")
7438 if [[ $max_mrif ]]; then
7441 max_mrif=$(do_facet $facet \
7442 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
7447 set_mdt_max_mod_rpcs_in_flight() {
7452 lctl_op=$($LCTL get_param \
7453 mdt.*.max_mod_rpcs_in_flight)
7454 if [[ $lctl_op ]]; then
7457 mdt.$FSNAME-MDT*.max_mod_rpcs_in_flight=$max_mrif"
7461 /sys/module/mdt/parameters/max_mod_rpcs_per_client"
7462 echo "the deprecated max_mod_rpcs_per_client \
7463 parameter was involved"
7470 [[ $($LCTL get_param mdc.*.import |
7471 grep "connect_flags:.*multi_mod_rpc") ]] ||
7472 skip "Need MDC with 'multi_mod_rpcs' feature"
7474 # check default value
7475 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
7476 check_max_mod_rpcs_in_flight $DIR/$tdir
7480 run_test 90a "check max_mod_rpcs_in_flight is enforced"
7490 [[ $($LCTL get_param mdc.*.import |
7491 grep "connect_flags:.*multi_mod_rpc") ]] ||
7492 skip "Need MDC with 'multi_mod_rpcs' feature"
7495 # update max_mod_rpcs_in_flight
7496 $LFS mkdir -c1 $DIR/${tdir}1 || error "mkdir $DIR/${tdir}1 failed"
7497 check_max_mod_rpcs_in_flight $DIR/${tdir}1 1
7500 # check client is able to send multiple modify RPCs in paralell
7501 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
7502 grep -c "multi_mod_rpcs")
7503 if [ "$tmp" -ne $MDSCOUNT ]; then
7505 skip "Client not able to send multiple modify RPCs in parallel"
7508 # update max_mod_rpcs_in_flight
7509 $LFS mkdir -c1 $DIR/${tdir}2 || error "mkdir $DIR/${tdir}2 failed"
7510 check_max_mod_rpcs_in_flight $DIR/${tdir}2 5
7513 $LFS mkdir -c1 $DIR/${tdir}3 || error "mkdir $DIR/${tdir}3 failed"
7514 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/${tdir}3))
7515 facet="mds$((0x$idx + 1))"
7517 mmrpc=$(get_mdt_max_mod_rpcs_in_flight_val $facet)
7518 echo "mdt_max_mod_rpcs_in_flight is $mmrpc"
7519 # update max_mod_rpcs_in_flight
7520 umount_client $MOUNT
7521 set_mdt_max_mod_rpcs_in_flight 16 $facet
7523 $LCTL set_param mdc.$FSNAME-MDT$idx-mdc-*.max_rpcs_in_flight=17
7524 check_max_mod_rpcs_in_flight $DIR/${tdir}3 16
7526 # restore MDT max_mod_rpcs_in_flight initial value
7527 set_mdt_max_mod_rpcs_in_flight $mmrpc $facet
7529 rm -rf $DIR/${tdir}?
7532 run_test 90b "check max_mod_rpcs_in_flight is enforced after update"
7535 # get max_rpcs_in_flight value
7536 mrif_90c=$($LCTL get_param -n \
7537 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
7538 echo "max_rpcs_in_flight is $mrif_90c"
7540 # get MDC max_mod_rpcs_in_flight value
7541 mmrif_90c=$($LCTL get_param -n \
7542 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight)
7543 echo "MDC max_mod_rpcs_in_flight is $mmrif_90c"
7545 # get MDT max_mod_rpcs_in_flight value
7546 mmrpc_90c=$(get_mdt_max_mod_rpcs_in_flight_val "mds1")
7547 echo "mdt_max_mod_rpcs_in_flight is $mmrpc_90c"
7550 restore_params_90c() {
7553 # restore max_rpcs_in_flight value
7554 do_facet mgs $LCTL set_param -P \
7555 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$mrif_90c
7557 # restore max_mod_rpcs_in_flight value
7558 do_facet mgs $LCTL set_param -P \
7559 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mmrif_90c
7561 # restore MDT max_mod_rpcs_in_flight value
7562 set_mdt_max_mod_rpcs_in_flight $mmrpc_90c "mds1"
7570 [[ $($LCTL get_param mdc.*.import |
7571 grep "connect_flags:.*multi_mod_rpc") ]] ||
7572 skip "Need MDC with 'multi_mod_rpcs' feature"
7574 # check client is able to send multiple modify RPCs in paralell
7575 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
7576 grep -c "multi_mod_rpcs")
7577 if [ "$tmp" -ne $MDSCOUNT ]; then
7579 skip "Client not able to send multiple modify RPCs in parallel"
7583 stack_trap restore_params_90c
7586 # attempt to set max_mod_rpcs_in_flight to max_rpcs_in_flight value
7587 # prerequisite: set MDT max_mod_rpcs_in_flight to
7588 # max_rpcs_in_flight value
7589 set_mdt_max_mod_rpcs_in_flight $mrif_90c "mds1"
7591 # if max_mod_rpcs_in_flight is set to be equal to or larger than
7592 # max_rpcs_in_flight, then max_rpcs_in_flight will be increased
7593 if [[ "$CLIENT_VERSION" -ge $(version_code 2.13.53) ]]; then
7595 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif_90c ||
7596 error "set max_mod_rpcs_in_flight to $mrif_90c failed"
7598 local new_mrif=$($LCTL get_param -n \
7599 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
7600 ((new_mrif == mrif_90c + 1)) ||
7601 error "max_rpcs_in_flight was not increased"
7604 umount_client $MOUNT
7605 set_mdt_max_mod_rpcs_in_flight $mmrpc_90c "mds1"
7609 # attempt to set max_mod_rpcs_in_flight to MDT max_mod_rpcs_in_flight+1
7610 # prerequisite: set max_rpcs_in_flight to MDT max_mod_rpcs_in_flight+2
7612 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$((mmrpc_90c + 2))
7615 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$((mmrpc_90c + 1)) &&
7616 error "set max_mod_rpcs_in_flight to $((mmrpc_90c + 1)) should fail"
7619 # attempt to set max_mod_rpcs_in_flight permanently
7620 do_facet mgs $LCTL set_param -P \
7621 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$mrif_90c
7623 do_facet mgs $LCTL set_param -P \
7624 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif_90c
7626 remount_client $MOUNT
7628 wait_update_facet --verbose client "$LCTL get_param -n \
7629 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight" \
7630 "$((mrif_90c + 1))" ||
7631 error "expected '$((mrif_90c + 1))' for max_rpcs_in_flight"
7633 wait_update_facet --verbose client "$LCTL get_param -n \
7634 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight" \
7636 error "expected '$mrif_90c' for max_mod_rpcs_in_flight"
7641 run_test 90c "check max_mod_rpcs_in_flight update limits"
7652 [[ $($LCTL get_param mdc.*.import |
7653 grep "connect_flags:.*multi_mod_rpc") ]] ||
7654 skip "Need MDC with 'multi_mod_rpcs' feature"
7656 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
7657 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
7658 facet="mds$((0x$idx + 1))"
7660 # check client version supports multislots
7661 tmp=$($LCTL get_param -N \
7662 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
7663 if [ -z "$tmp" ]; then
7665 skip "Client does not support multiple modify RPCs in flight"
7668 # get current value of max_mod_rcps_in_flight
7669 mmr=$($LCTL get_param -n \
7670 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
7671 echo "max_mod_rcps_in_flight is $mmr"
7674 echo "creating $mmr files ..."
7676 for i in $(seq $mmr); do
7677 touch $DIR/$tdir/file-$i
7680 # prepare for close RPC
7681 multiop_bg_pause $DIR/$tdir/file-close O_c
7684 # consumes mmr modify RPC slots
7685 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7686 # drop requests on MDT so that RPC slots are consumed
7687 # during all the request resend interval
7688 do_facet $facet "$LCTL set_param fail_loc=0x159"
7689 echo "launch $mmr chmod in parallel ..."
7690 for i in $(seq $mmr); do
7691 chmod 0600 $DIR/$tdir/file-$i &
7694 # send one additional close RPC
7695 do_facet $facet "$LCTL set_param fail_loc=0"
7696 echo "launch 1 additional close in parallel ..."
7698 cancel_lru_locks mdc
7701 # check this additional close RPC get a modify RPC slot
7702 # and multiop process completed
7703 [ -d /proc/$pid ] &&
7704 error "Unable to send the additional close RPC in parallel"
7709 run_test 90d "check one close RPC is allowed above max_mod_rpcs_in_flight"
7711 check_uuid_on_ost() {
7713 do_facet ost1 "$LCTL get_param obdfilter.${FSNAME}*.exports.'$nid'.uuid"
7716 check_uuid_on_mdt() {
7718 do_facet $SINGLEMDS "$LCTL get_param mdt.${mds1_svc}*.exports.'$nid'.uuid"
7726 [[ "$OST1_VERSION" -ge $(version_code 2.7.63) ]] ||
7727 skip "Need OST version at least 2.7.63"
7728 [[ "$MDS1_VERSION" -ge $(version_code 2.7.63) ]] ||
7729 skip "Need MDT version at least 2.7.63"
7731 start_mds || error "MDS start failed"
7732 start_ost || error "unable to start OST"
7733 mount_client $MOUNT || error "client start failed"
7734 check_mount || error "check_mount failed"
7737 nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g")
7741 uuid=$(get_client_uuid $MOUNT)
7743 echo "list nids on mdt:"
7744 do_facet $SINGLEMDS "$LCTL list_param mdt.${FSNAME}*.exports.*"
7745 echo "uuid from $nid:"
7746 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
7748 found=$(check_uuid_on_mdt $nid | grep $uuid)
7749 [ -z "$found" ] && error "can't find $uuid $nid on MDT"
7750 found=$(check_uuid_on_ost $nid | grep $uuid)
7751 [ -z "$found" ] && error "can't find $uuid $nid on OST"
7753 # umount the client so it won't reconnect
7754 manual_umount_client --force || error "failed to umount $?"
7755 # shouldn't disappear on MDS after forced umount
7756 found=$(check_uuid_on_mdt $nid | grep $uuid)
7757 [ -z "$found" ] && error "can't find $uuid $nid"
7760 do_facet $SINGLEMDS \
7761 "$LCTL set_param -n mdt.${mds1_svc}.evict_client nid:$nid"
7762 sleep 1 # eviction above is async, give it some time to proceed
7764 found=$(check_uuid_on_mdt $nid | grep $uuid)
7765 [ -n "$found" ] && error "found $uuid $nid on MDT"
7766 found=$(check_uuid_on_ost $nid | grep $uuid)
7767 [ -n "$found" ] && error "found $uuid $nid on OST"
7769 # check it didn't reconnect (being umounted)
7770 sleep $((TIMEOUT+1))
7771 found=$(check_uuid_on_mdt $nid | grep $uuid)
7772 [ -n "$found" ] && error "found $uuid $nid on MDT"
7773 found=$(check_uuid_on_ost $nid | grep $uuid)
7774 [ -n "$found" ] && error "found $uuid $nid on OST"
7778 run_test 91 "evict-by-nid support"
7780 generate_ldev_conf() {
7781 # generate an ldev.conf file
7782 local ldevconfpath=$1
7784 local fsldevformat=""
7787 local facets="mgs,$(get_facets OST),$(get_facets MDS)"
7788 for facet in ${facets//,/ }; do
7790 fstype=$(facet_fstype $facet)
7791 [ "$fstype" = zfs ] && fsldevformat="$fstype:"
7792 local host=$(facet_host $facet)
7794 local varfo=${facet}failover_HOST
7795 if [ $facet == mgs ] && combined_mgs_mds; then
7796 varfo=mds1failover_HOST
7798 [ -n "${!varfo}" ] && fo=${!varfo}
7799 local type=$(echo $facet | tr -d "[:digit:]" | \
7800 tr "[:lower:]" "[:upper:]" | sed s/MDS/MDT/ )
7802 [ ${facet} == mgs ] ||
7803 num=$(facet_number $facet)
7804 printf "%s\t%s\t%s-%s%04x\t%s%s\n" \
7811 $(facet_device $facet) >> $ldevconfpath
7814 echo "----- $ldevconfpath -----"
7816 echo "--- END $ldevconfpath ---"
7821 # generate a nids file (mapping between hostname to nid)
7822 # looks like we only have the MGS nid available to us
7823 # so just echo that to a file
7825 echo -e "${mgs_HOST}\t${MGSNID}" > $nidspath
7827 echo "----- $nidspath -----"
7829 echo "--- END $nidspath ---"
7832 compare_ldev_output() {
7833 local ldev_output=$1
7834 local expected_output=$2
7836 sort $expected_output -o $expected_output
7837 sort $ldev_output -o $ldev_output
7839 echo "-- START OF LDEV OUTPUT --"
7841 echo "--- END OF LDEV OUTPUT ---"
7843 echo "-- START OF EXPECTED OUTPUT --"
7844 cat $expected_output
7845 echo "--- END OF EXPECTED OUTPUT ---"
7847 diff $expected_output $ldev_output
7852 if [ -z "$LDEV" ]; then
7853 error "ldev is missing!"
7856 local LDEVCONFPATH=$TMP/ldev.conf
7857 local NIDSPATH=$TMP/nids
7859 echo "Host is $(hostname)"
7861 generate_ldev_conf $LDEVCONFPATH
7862 generate_nids $NIDSPATH
7864 # echo the mgs nid and compare it to environment variable MGSNID
7865 local facets="$(get_facets OST),$(get_facets MDS),mgs"
7866 for facet in ${facets//,/ }; do
7867 local host=$(facet_host $facet)
7868 local output=$($LDEV -c $LDEVCONFPATH -H $host -n $NIDSPATH echo %m)
7870 echo "-- START OF LDEV OUTPUT --"
7872 echo "--- END OF LDEV OUTPUT ---"
7875 error "ldev failed to execute!"
7877 # need to process multiple lines because of
7878 # several targets on host
7879 echo -e $output | awk '{ print $2 }' | while read -r line ; do
7880 [ "$line" = "$MGSNID" ] ||
7881 error "ldev failed mgs nid '$line', \
7885 rm -f $LDEVCONFPATH $NIDSPATH
7887 run_test 92 "ldev returns MGS NID correctly in command substitution"
7890 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs"
7893 #start mgs or mgs/mdt0
7894 if ! combined_mgs_mds ; then
7901 start_ost || error "OST0 start fail"
7903 #define OBD_FAIL_MGS_WRITE_TARGET_DELAY 0x90e
7904 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x8000090e"
7905 for num in $(seq 2 $MDSCOUNT); do
7909 mount_client $MOUNT || error "mount client fails"
7910 wait_osc_import_state mds ost FULL
7911 wait_osc_import_ready client ost
7912 check_mount || error "check_mount failed"
7914 cleanup || error "cleanup failed with $?"
7916 run_test 93 "register mulitple MDT at the same time"
7919 if [ -z "$LDEV" ]; then
7920 error "ldev is missing!"
7923 local LDEVCONFPATH=$TMP/ldev.conf
7924 local NIDSPATH=$TMP/nids
7926 generate_ldev_conf $LDEVCONFPATH
7927 generate_nids $NIDSPATH
7929 local LDEV_OUTPUT=$TMP/ldev-output.txt
7930 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME > $LDEV_OUTPUT
7932 # ldev failed, error
7933 if [ $? -ne 0 ]; then
7934 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7935 error "ldev failed to execute!"
7939 local EXPECTED_OUTPUT=$TMP/ldev-expected.txt
7941 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
7943 for num in $(seq $MDSCOUNT); do
7944 printf "%s-MDT%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7947 for num in $(seq $OSTCOUNT); do
7948 printf "%s-OST%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7951 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7953 if [ $? -ne 0 ]; then
7954 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7955 error "ldev failed to produce the correct hostlist!"
7958 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7960 run_test 94 "ldev outputs correct labels for file system name query"
7963 if [ -z "$LDEV" ]; then
7964 error "ldev is missing!"
7967 local LDEVCONFPATH=$TMP/ldev.conf
7968 local NIDSPATH=$TMP/nids
7970 generate_ldev_conf $LDEVCONFPATH
7971 generate_nids $NIDSPATH
7975 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME &>/dev/null
7976 if [ $? -ne 0 ]; then
7977 rm $LDEVCONFPATH $NIDSPATH
7978 error "ldev label filtering w/ -F failed!"
7982 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -l &>/dev/null
7983 if [ $? -ne 0 ]; then
7984 rm $LDEVCONFPATH $NIDSPATH
7985 error "ldev label filtering w/ -l failed!"
7989 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f &>/dev/null
7990 if [ $? -ne 0 ]; then
7991 rm $LDEVCONFPATH $NIDSPATH
7992 error "ldev label filtering w/ -f failed!"
7996 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a &>/dev/null
7997 if [ $? -ne 0 ]; then
7998 rm $LDEVCONFPATH $NIDSPATH
7999 error "ldev label filtering w/ -a failed!"
8004 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -F $FSNAME &>/dev/null
8005 if [ $? -eq 0 ]; then
8006 rm $LDEVCONFPATH $NIDSPATH
8007 error "ldev label filtering w/ -a and -F incorrectly succeeded"
8011 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -f &>/dev/null
8012 if [ $? -eq 0 ]; then
8013 rm $LDEVCONFPATH $NIDSPATH
8014 error "ldev label filtering w/ -a and -f incorrectly succeeded"
8018 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -l &>/dev/null
8019 if [ $? -eq 0 ]; then
8020 rm $LDEVCONFPATH $NIDSPATH
8021 error "ldev label filtering w/ -a and -l incorrectly succeeded"
8025 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f -l &>/dev/null
8026 if [ $? -eq 0 ]; then
8027 rm $LDEVCONFPATH $NIDSPATH
8028 error "ldev label filtering w/ -f and -l incorrectly succeeded"
8032 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -l &>/dev/null
8033 if [ $? -eq 0 ]; then
8034 rm $LDEVCONFPATH $NIDSPATH
8035 error "ldev label filtering w/ -F and -l incorrectly succeeded"
8038 # file sys & foreign
8039 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -f &>/dev/null
8040 if [ $? -eq 0 ]; then
8041 rm $LDEVCONFPATH $NIDSPATH
8042 error "ldev label filtering w/ -F and -f incorrectly succeeded"
8045 rm $LDEVCONFPATH $NIDSPATH
8047 run_test 95 "ldev should only allow one label filter"
8050 if [ -z "$LDEV" ]; then
8051 error "ldev is missing!"
8054 local LDEVCONFPATH=$TMP/ldev.conf
8055 local NIDSPATH=$TMP/nids
8057 generate_ldev_conf $LDEVCONFPATH
8058 generate_nids $NIDSPATH
8060 local LDEV_OUTPUT=$TMP/ldev-output.txt
8061 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -H $mgs_HOST \
8063 awk '{print $2}' > $LDEV_OUTPUT
8065 # ldev failed, error
8066 if [ $? -ne 0 ]; then
8067 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
8068 error "ldev failed to execute!"
8072 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
8074 echo "$mgs_HOST-$(facet_fstype mgs)" > $EXPECTED_OUTPUT
8076 local facets="$(get_facets OST),$(get_facets MDS)"
8077 for facet in ${facets//,/ }; do
8078 local host=$(facet_host $facet)
8079 [ "$mgs_HOST" == "$host" ] &&
8080 echo "$host-$(facet_fstype $facet)" \
8084 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
8086 if [ $? -ne 0 ]; then
8087 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8088 error "ldev failed to produce the correct output!"
8091 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8093 run_test 96 "ldev returns hostname and backend fs correctly in command sub"
8096 if [ -z "$LDEV" ]; then
8097 error "ldev is missing!"
8100 local LDEVCONFPATH=$TMP/ldev.conf
8101 local NIDSPATH=$TMP/nids
8103 generate_ldev_conf $LDEVCONFPATH
8104 generate_nids $NIDSPATH
8106 local LDEV_OUTPUT=$TMP/ldev-output.txt
8107 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
8109 echo -e "\nMDT role"
8110 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mdt > $LDEV_OUTPUT
8112 if [ $? -ne 0 ]; then
8113 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
8114 error "ldev failed to execute for mdt role!"
8117 for num in $(seq $MDSCOUNT); do
8118 printf "%s-MDT%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
8121 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
8123 if [ $? -ne 0 ]; then
8124 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8125 error "ldev failed to produce the correct output for mdt role!"
8128 echo -e "\nOST role"
8129 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R ost > $LDEV_OUTPUT
8131 if [ $? -ne 0 ]; then
8132 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
8133 error "ldev failed to execute for ost role!"
8137 for num in $(seq $OSTCOUNT); do
8138 printf "%s-OST%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
8141 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
8143 if [ $? -ne 0 ]; then
8144 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8145 error "ldev failed to produce the correct output for ost role!"
8148 echo -e "\nMGS role"
8149 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mgs > $LDEV_OUTPUT
8151 if [ $? -ne 0 ]; then
8152 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
8153 error "ldev failed to execute for mgs role!"
8156 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
8158 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
8160 if [ $? -ne 0 ]; then
8161 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8162 error "ldev failed to produce the correct output for mgs role!"
8165 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8167 run_test 97 "ldev returns correct ouput when querying based on role"
8174 check_mount || error "mount failed"
8175 mountopt="user_xattr"
8176 for ((x = 1; x <= PAGE_SIZE/11; x++)); do
8177 mountopt="$mountopt,user_xattr"
8179 mount_client $MOUNT remount,$mountopt 2>&1 | grep "too long" ||
8180 error "Buffer overflow check failed"
8181 cleanup || error "cleanup failed"
8183 run_test 98 "Buffer-overflow check while parsing mount_opts"
8187 [[ "$ost1_FSTYPE" != ldiskfs ]] &&
8188 skip "ldiskfs only test"
8189 [[ "$OST1_VERSION" -ge $(version_code 2.8.57) ]] ||
8190 skip "Need OST version at least 2.8.57"
8192 local ost_opts="$(mkfs_opts ost1 $(ostdevname 1)) \
8193 --reformat $(ostdevname 1) $(ostvdevname 1)"
8194 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" &&
8195 skip "meta_bg already set"
8198 if [[ ${!opts} != *mkfsoptions* ]]; then
8199 eval opts=\"${!opts} \
8200 --mkfsoptions='\\\"-O ^resize_inode,meta_bg\\\"'\"
8202 local val=${!opts//--mkfsoptions=\\\"/ \
8203 --mkfsoptions=\\\"-O ^resize_inode,meta_bg }
8207 echo "params: $opts"
8210 add ost1 $opts || error "add ost1 failed with new params"
8212 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" ||
8213 error "meta_bg is not set"
8215 run_test 99 "Adding meta_bg option"
8220 start_mgsmds || error "MDS start failed"
8221 start_ost || error "unable to start OST"
8222 mount_client $MOUNT || error "client start failed"
8223 check_mount || error "check_mount failed"
8232 do_facet mgs 'lshowmount -v' | awk 'BEGIN {NR == 0; rc=1} /MGS:/ {rc=0}
8233 END {exit rc}' || error "lshowmount have no output MGS"
8235 do_facet mds1 'lshowmount -v' | awk 'BEGIN {NR == 2; rc=1} /-MDT0000:/
8236 {rc=0} END {exit rc}' || error "lshowmount have no output MDT0"
8238 do_facet ost1 'lshowmount -v' | awk 'BEGIN {NR == 4; rc=1} /-OST0000:/
8239 {rc=0} END {exit rc}' || error "lshowmount have no output OST0"
8241 cleanup || error "cleanup failed with $?"
8243 run_test 100 "check lshowmount lists MGS, MDT, OST and 0@lo"
8246 local createmany_pid
8247 local dev=$FSNAME-OST0000-osc-MDT0000
8251 do_nodes $(comma_list $(osts_nodes)) $LCTL set_param \
8252 seq.*OST*-super.width=$DATA_SEQ_MAX_WIDTH
8253 createmany -o $DIR1/$tdir/$tfile-%d 50000 &
8255 # MDT->OST reconnection causes MDT<->OST last_id synchornisation
8256 # via osp_precreate_cleanup_orphans.
8257 for ((i = 0; i < 100; i++)); do
8258 for ((k = 0; k < 10; k++)); do
8259 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;" \
8260 "$LCTL --device $dev activate"
8263 ls -asl $MOUNT | grep '???' &&
8264 { kill -9 $createmany_pid &>/dev/null;
8265 error "File has no object on OST"; }
8267 kill -s 0 $createmany_pid || break
8269 wait $createmany_pid
8271 unlinkmany $DIR1/$tdir/$tfile-%d 50000
8274 run_test 101a "Race MDT->OST reconnection with create"
8277 local dev=$FSNAME-OST0000-osc-MDT0000
8278 local dir=$DIR1/$tdir
8282 $LFS setstripe -c 1 -i 0 $dir
8283 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;"
8284 #define OBD_FAIL_OSP_CON_EVENT_DELAY 0x2107
8285 do_facet mds1 "$LCTL set_param fail_loc=0x80002107 fail_val=20"
8286 do_facet $SINGLEMDS "$LCTL --device $dev activate;"
8291 wait_osc_import_ready client ost1
8292 touch $dir/$tfile || error "Can't create file"
8296 run_test 101b "Race events DISCONNECT and ACTIVE in osp"
8299 [[ "$MDS1_VERSION" -gt $(version_code 2.9.53) ]] ||
8300 skip "Need server version greater than 2.9.53"
8301 [[ “$(mdsdevname 1)” != “$(mgsdevname)” ]] &&
8302 [[ “$(facet_host mds1)” = “$(facet_host mgs)” ]] &&
8303 skip "MGS must be on different node or combined"
8305 cleanup || error "cleanup failed with $?"
8307 local mds1dev=$(mdsdevname 1)
8308 local mds1mnt=$(facet_mntpt mds1)
8309 local mds1opts=$MDS_MOUNT_OPTS
8311 if [ "$mds1_FSTYPE" == ldiskfs ] &&
8312 ! do_facet mds1 test -b $mds1dev; then
8313 mds1opts=$(csa_add "$mds1opts" -o loop)
8315 if [[ "$mds1_FSTYPE" == zfs ]]; then
8316 import_zpool mds1 || return ${PIPESTATUS[0]}
8319 # unload all and only load libcfs to allow fail_loc setting
8320 do_facet mds1 $LUSTRE_RMMOD || error "unable to unload modules"
8321 do_facet mds1 modprobe libcfs || error "libcfs not loaded"
8322 do_facet mds1 lsmod \| grep libcfs || error "libcfs not loaded"
8324 #define OBD_FAIL_OBDCLASS_MODULE_LOAD 0x60a
8325 do_facet mds1 "$LCTL set_param fail_loc=0x8000060a"
8327 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts &&
8328 error "mdt start must fail"
8329 do_facet mds1 lsmod \| grep obdclass && error "obdclass must not load"
8331 do_facet mds1 "$LCTL set_param fail_loc=0x0"
8333 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts ||
8334 error "mdt start must not fail"
8336 cleanup || error "cleanup failed with $?"
8338 run_test 102 "obdclass module cleanup upon error"
8343 echo "rename $FSNAME to $newname"
8345 if ! combined_mgs_mds ; then
8346 local dev=$(mgsdevname)
8349 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
8350 error "(7) Fail to rename MGS"
8351 if [ "$(facet_fstype mgs)" = "zfs" ]; then
8352 reimport_zpool mgs $newname-mgs
8356 for num in $(seq $MDSCOUNT); do
8357 local dev=$(mdsdevname $num)
8359 do_facet mds${num} \
8360 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
8361 error "(8) Fail to rename MDT $num"
8362 if [ "$(facet_fstype mds${num})" = "zfs" ]; then
8363 reimport_zpool mds${num} $newname-mdt${num}
8367 for num in $(seq $OSTCOUNT); do
8368 local dev=$(ostdevname $num)
8370 do_facet ost${num} \
8371 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
8372 error "(9) Fail to rename OST $num"
8373 if [ "$(facet_fstype ost${num})" = "zfs" ]; then
8374 reimport_zpool ost${num} $newname-ost${num}
8379 test_103_set_pool() {
8383 do_facet mgs $LCTL pool_add $FSNAME.$pname ${FSNAME}-$ost_x ||
8384 error "Fail to add $ost_x to $FSNAME.$pname"
8385 wait_update $HOSTNAME \
8386 "lctl get_param -n lov.$FSNAME-clilov-*.pools.$pname |
8387 grep $ost_x" "$FSNAME-${ost_x}_UUID" ||
8388 error "$ost_x is NOT in pool $FSNAME.$pname"
8391 test_103_check_pool() {
8392 local save_fsname=$1
8395 stat $DIR/$tdir/test-framework.sh ||
8396 error "($errno) Fail to stat"
8397 do_facet mgs $LCTL pool_list $FSNAME.pool1 ||
8398 error "($errno) Fail to list $FSNAME.pool1"
8399 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname ||
8400 error "($errno) Fail to list $FSNAME.$save_fsname"
8401 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname |
8402 grep ${FSNAME}-OST0000 ||
8403 error "($errno) List $FSNAME.$save_fsname is invalid"
8405 local pname=$($LFS getstripe --pool $DIR/$tdir/d0)
8406 [ "$pname" = "$save_fsname" ] ||
8407 error "($errno) Unexpected pool name $pname"
8411 check_mount_and_prep
8413 mkdir $DIR/$tdir || error "(1) Fail to mkdir $DIR/$tdir"
8414 cp $LUSTRE/tests/test-framework.sh $DIR/$tdir ||
8415 error "(2) Fail to copy test-framework.sh"
8417 do_facet mgs $LCTL pool_new $FSNAME.pool1 ||
8418 error "(3) Fail to create $FSNAME.pool1"
8419 # name the pool name as the fsname
8420 do_facet mgs $LCTL pool_new $FSNAME.$FSNAME ||
8421 error "(4) Fail to create $FSNAME.$FSNAME"
8423 test_103_set_pool $FSNAME OST0000
8425 $LFS setstripe -p $FSNAME $DIR/$tdir/d0 ||
8426 error "(6) Fail to setstripe on $DIR/$tdir/d0"
8431 test_renamefs mylustre
8433 local save_fsname=$FSNAME
8437 test_103_check_pool $save_fsname 7
8439 if [ $OSTCOUNT -ge 2 ]; then
8440 test_103_set_pool $save_fsname OST0001
8443 $LFS setstripe -p $save_fsname $DIR/$tdir/f0 ||
8444 error "(16) Fail to setstripe on $DIR/$tdir/f0"
8453 test_103_check_pool $save_fsname 17
8457 test_renamefs $save_fsname
8463 run_test 103 "rename filesystem name"
8465 test_104a() { # LU-6952
8466 local mds_mountopts=$MDS_MOUNT_OPTS
8467 local ost_mountopts=$OST_MOUNT_OPTS
8468 local mds_mountfsopts=$MDS_MOUNT_FS_OPTS
8469 local lctl_ver=$(do_facet $SINGLEMDS $LCTL --version |
8472 [[ $(version_code $lctl_ver) -lt $(version_code 2.9.55) ]] &&
8473 skip "this test needs utils above 2.9.55"
8475 # specify "acl" in mount options used by mkfs.lustre
8476 if [ -z "$MDS_MOUNT_FS_OPTS" ]; then
8477 MDS_MOUNT_FS_OPTS="acl,user_xattr"
8480 MDS_MOUNT_FS_OPTS="${MDS_MOUNT_FS_OPTS},acl,user_xattr"
8483 echo "mountfsopt: $MDS_MOUNT_FS_OPTS"
8485 #reformat/remount the MDT to apply the MDT_MOUNT_FS_OPT options
8487 if ! combined_mgs_mds ; then
8491 if [ -z "$MDS_MOUNT_OPTS" ]; then
8492 MDS_MOUNT_OPTS="-o noacl"
8494 MDS_MOUNT_OPTS="${MDS_MOUNT_OPTS},noacl"
8497 for num in $(seq $MDSCOUNT); do
8498 start mds$num $(mdsdevname $num) $MDS_MOUNT_OPTS ||
8499 error "Failed to start MDS"
8502 for num in $(seq $OSTCOUNT); do
8503 start ost$num $(ostdevname $num) $OST_MOUNT_OPTS ||
8504 error "Failed to start OST"
8508 setfacl -m "d:$RUNAS_ID:rwx" $MOUNT &&
8509 error "ACL is applied when FS is mounted with noacl."
8511 MDS_MOUNT_OPTS=$mds_mountopts
8512 OST_MOUNT_OPTS=$ost_mountopts
8513 MDS_MOUNT_FS_OPTS=$mds_mountfsopts
8515 run_test 104a "Make sure user defined options are reflected in mount"
8517 test_104b() { # LU-12859
8518 mount_client $MOUNT3 flock,localflock
8519 stack_trap "umount_client $MOUNT3" EXIT
8520 mount | grep "$MOUNT3 .*,flock" && error "flock is still set"
8521 mount | grep "$MOUNT3 .*,localflock" || error "localflock is not set"
8522 umount_client $MOUNT3
8523 mount_client $MOUNT3 localflock,flock
8524 mount | grep "$MOUNT3 .*,localflock" && error "localflock is still set"
8525 mount | grep "$MOUNT3 .*,flock" || error "flock is not set"
8526 umount_client $MOUNT3
8527 mount_client $MOUNT3 localflock,flock,noflock
8528 flock_is_enabled $MOUNT3 && error "some flock is still enabled" || true
8530 run_test 104b "Mount uses last flock argument"
8532 error_and_umount() {
8543 mount --bind $DIR $TMP/$tdir || error "mount bind mnt pt failed"
8544 rm -f $TMP/$tdir/$tfile
8545 rm -f $TMP/$tdir/${tfile}1
8547 # Files should not be created in ro bind mount point
8548 # remounting from rw to ro
8549 mount -o remount,ro $TMP/$tdir ||
8550 error_and_umount "readonly remount of bind mnt pt failed"
8551 touch $TMP/$tdir/$tfile &&
8552 error_and_umount "touch succeeds on ro bind mnt pt"
8553 [ -e $TMP/$tdir/$tfile ] &&
8554 error_and_umount "file created on ro bind mnt pt"
8556 # Files should be created in rw bind mount point
8557 # remounting from ro to rw
8558 mount -o remount,rw $TMP/$tdir ||
8559 error_and_umount "read-write remount of bind mnt pt failed"
8560 touch $TMP/$tdir/${tfile}1 ||
8561 error_and_umount "touch fails on rw bind mnt pt"
8562 [ -e $TMP/$tdir/${tfile}1 ] ||
8563 error_and_umount "file not created on rw bind mnt pt"
8564 umount $TMP/$tdir || error "umount of bind mnt pt failed"
8566 cleanup || error "cleanup failed with $?"
8568 run_test 105 "check file creation for ro and rw bind mnt pt"
8572 local creates=64768 # one full plain llog
8574 # ensure there are enough inodes in the filesystem
8575 (( OSTSIZE < (creates + 1024) * 8)) && OSTSIZE=$(((creates + 1024) * 8))
8579 mkdir -p $DIR/$tdir || error "create $tdir failed"
8580 do_nodes $(comma_list $(osts_nodes)) $LCTL set_param \
8581 seq.*OST*-super.width=$DATA_SEQ_MAX_WIDTH
8582 lfs setstripe -c 1 -i 0 $DIR/$tdir
8583 #define OBD_FAIL_CAT_RECORDS 0x1312
8584 do_facet mds1 $LCTL set_param fail_loc=0x1312 fail_val=$repeat
8586 for ((i = 1; i <= $repeat; i++)); do
8587 createmany -o $DIR/$tdir/f- $creates || lfs df -i
8588 createmany -u $DIR/$tdir/f- $creates
8589 wait_delete_completed $((TIMEOUT * 7))
8591 #ASSERTION osp_sync_thread() ( thread->t_flags != SVC_RUNNING ) failed
8592 #shows that osp code is buggy
8593 do_facet mds1 $LCTL set_param fail_loc=0 fail_val=0
8597 run_test 106 "check osp llog processing when catalog is wrapped"
8600 [[ "$MDS1_VERSION" -ge $(version_code 2.10.50) ]] ||
8601 skip "Need MDS version > 2.10.50"
8604 start_mgsmds || error "start_mgsmds failed"
8605 start_ost || error "unable to start OST"
8607 # add unknown configuration parameter.
8608 if [[ $PERM_CMD == *"set_param -P"* ]]; then
8609 cmd="$PERM_CMD ost.$FSNAME-OST0000*.unknown_param"
8611 cmd="$PERM_CMD $FSNAME-OST0000*.ost.unknown_param"
8613 do_facet mgs "$cmd=50"
8614 cleanup_nocli || error "cleanup_nocli failed with $?"
8617 # unknown param should be ignored while mounting.
8618 start_ost || error "unable to start OST after unknown param set"
8620 cleanup || error "cleanup failed with $?"
8622 run_test 107 "Unknown config param should not fail target mounting"
8627 $rcmd rm -rf $tmp > /dev/null 2>&1
8628 $rcmd mkdir -p $tmp/{mnt,images} || error "failed to mkdir remotely"
8630 for facet in $facets; do
8631 [ "$mds1_FSTYPE" = zfs ] &&
8632 $rcmd $ZPOOL -f export lustre-$facet > /dev/null 2>&1
8633 $rcmd mkdir $tmp/mnt/$facet ||
8634 error "failed to mkdir $tmp/mnt/$facet"
8635 $rcmd dd if=/dev/zero of=$tmp/images/$facet \
8636 seek=199 bs=1M count=1 ||
8637 error "failed to create $tmp/images/$facet"
8646 local facet=${role}$((idx + 1))
8648 [ $# -eq 5 ] && pool=$5
8650 do_facet $SINGLEMDS $MKFS --fsname=lustre --$mgs \
8651 --$role --index=$idx --replace --backfstype=$bkfs \
8652 --device-size=200000 --reformat $pool $tmp/images/$facet ||
8653 error "failed to mkfs for $facet"
8657 echo "mounting client..."
8658 mount -t lustre ${nid}:/lustre $MOUNT ||
8659 error "failed to mount lustre"
8662 ls -l $MOUNT/local_dir || error "failed to list"
8664 echo "check truncate && write"
8665 echo "dummmmmmmmmmmmm" > $MOUNT/remote_dir/fsx.c ||
8666 error "failed to tuncate & write"
8669 touch $MOUNT/foooo ||
8670 error "failed to create"
8672 echo "check read && write && append"
8673 sha1sum $MOUNT/conf-sanity.sh |
8674 awk '{ print $1 }' > $MOUNT/checksum.new ||
8675 error "failed to read(1)"
8676 sha1sum $MOUNT/remote_dir/unlinkmany.c |
8677 awk '{ print $1 }' >> $MOUNT/checksum.new ||
8678 error "failed to read(2)"
8679 sha1sum $MOUNT/striped_dir/lockahead_test.o |
8680 awk '{ print $1 }' >> $MOUNT/checksum.new ||
8681 error "failed to read(3)"
8684 diff $MOUNT/checksum.new $MOUNT/checksum.src ||
8685 error "failed to verify data"
8695 umount -f $MOUNT || error "failed to umount client"
8696 for facet in $facets; do
8697 $rcmd umount -f $tmp/mnt/$facet ||
8698 error "failed to umount $facet"
8699 if [ "$mds1_FSTYPE" = zfs ]; then
8700 $rcmd $ZPOOL export -f lustre-$facet ||
8701 error "failed to export lustre-$facet"
8705 $rcmd rm -rf $tmp || error "failed to rm the dir $tmp"
8709 [ "$CLIENTONLY" ] && skip "Client-only testing"
8710 [ "$mds1_FSTYPE" != zfs ] && skip "zfs only test"
8711 [ "$MDS1_VERSION" -lt $(version_code 2.10.58) ] &&
8712 skip "Need server version at least 2.10.58"
8717 local tmp=$TMP/$tdir
8718 local rcmd="do_facet $SINGLEMDS"
8719 local facets="mdt1 mdt2 ost1 ost2"
8720 local nid=$($rcmd $LCTL list_nids | head -1)
8723 trap t_108_cleanup EXIT ERR
8726 t_108_mkfs mdt 0 zfs mgs lustre-mdt1/mdt1
8727 t_108_mkfs mdt 1 zfs mgsnode=$nid lustre-mdt2/mdt2
8728 t_108_mkfs ost 0 zfs mgsnode=$nid lustre-ost1/ost1
8729 t_108_mkfs ost 1 zfs mgsnode=$nid lustre-ost2/ost2
8731 for facet in $facets; do
8732 $rcmd zfs set mountpoint=$tmp/mnt/$facet canmount=on \
8733 lustre-$facet/$facet ||
8734 error "failed to zfs set for $facet (1)"
8735 $rcmd zfs mount lustre-$facet/$facet ||
8736 error "failed to local mount $facet"
8737 $rcmd tar jxf $LUSTRE/tests/ldiskfs_${facet}_2_11.tar.bz2 \
8738 --xattrs --xattrs-include="trusted.*" \
8739 -C $tmp/mnt/$facet/ > /dev/null 2>&1 ||
8740 error "failed to untar image for $facet"
8741 $rcmd "cd $tmp/mnt/$facet && rm -rf oi.* OI_* lfsck_* LFSCK" ||
8742 error "failed to cleanup for $facet"
8743 $rcmd zfs umount lustre-$facet/$facet ||
8744 error "failed to local umount $facet"
8745 $rcmd zfs set canmount=off lustre-$facet/$facet ||
8746 error "failed to zfs set $facet (2)"
8749 echo "changing server nid..."
8750 $rcmd mount -t lustre -o nosvc lustre-mdt1/mdt1 $tmp/mnt/mdt1
8751 $rcmd lctl replace_nids $FSNAME-MDT0000 $nid
8752 $rcmd lctl replace_nids $FSNAME-MDT0001 $nid
8753 $rcmd lctl replace_nids $FSNAME-OST0000 $nid
8754 $rcmd lctl replace_nids $FSNAME-OST0001 $nid
8755 $rcmd umount $tmp/mnt/mdt1
8757 for facet in $facets; do
8758 echo "mounting $facet from backup..."
8759 $rcmd mount -t lustre -o abort_recov lustre-$facet/$facet \
8760 $tmp/mnt/$facet || error "failed to mount $facet"
8763 # ZFS backend can detect migration and trigger OI scrub automatically
8764 # sleep 3 seconds for scrub done
8770 run_test 108a "migrate from ldiskfs to ZFS"
8773 [ "$CLIENTONLY" ] && skip "Client-only testing"
8774 [ "$mds1_FSTYPE" != ldiskfs ] && skip "ldiskfs only test"
8775 [ "$MDS1_VERSION" -lt $(version_code 2.10.58) ] &&
8776 skip "Need server version at least 2.10.58"
8781 local tmp=$TMP/$tdir
8782 local rcmd="do_facet $SINGLEMDS"
8783 local facets="mdt1 mdt2 ost1 ost2"
8784 local scrub_list="MDT0000 MDT0001 OST0000 OST0001"
8785 local nid=$($rcmd $LCTL list_nids | head -1)
8788 trap t_108_cleanup EXIT ERR
8791 t_108_mkfs mdt 0 ldiskfs mgs
8792 t_108_mkfs mdt 1 ldiskfs mgsnode=$nid
8793 t_108_mkfs ost 0 ldiskfs mgsnode=$nid
8794 t_108_mkfs ost 1 ldiskfs mgsnode=$nid
8796 for facet in $facets; do
8797 $rcmd mount -t ldiskfs -o loop $tmp/images/$facet \
8799 error "failed to local mount $facet"
8801 $rcmd tar jxf $LUSTRE/tests/zfs_${facet}_2_11.tar.bz2 \
8802 --xattrs --xattrs-include="*.*" \
8803 -C $tmp/mnt/$facet/ > /dev/null 2>&1 ||
8804 error "failed to untar image for $facet"
8805 $rcmd "cd $tmp/mnt/$facet && rm -rf oi.* OI_* lfsck_* LFSCK" ||
8806 error "failed to cleanup for $facet"
8807 $rcmd umount $tmp/mnt/$facet ||
8808 error "failed to local umount $facet"
8811 echo "changing server nid..."
8812 $rcmd mount -t lustre -o nosvc,loop $tmp/images/mdt1 $tmp/mnt/mdt1
8813 $rcmd lctl replace_nids $FSNAME-MDT0000 $nid
8814 $rcmd lctl replace_nids $FSNAME-MDT0001 $nid
8815 $rcmd lctl replace_nids $FSNAME-OST0000 $nid
8816 $rcmd lctl replace_nids $FSNAME-OST0001 $nid
8817 $rcmd umount $tmp/mnt/mdt1
8819 for facet in $facets; do
8820 echo "mounting $facet from backup..."
8821 $rcmd mount -t lustre -o loop,abort_recov $tmp/images/$facet \
8822 $tmp/mnt/$facet || error "failed to mount $facet"
8825 for facet in $scrub_list; do
8826 $rcmd $LCTL lfsck_start -M $FSNAME-$facet -t scrub ||
8827 error "failed to start OI scrub on $facet"
8830 # sleep 3 seconds for scrub done
8836 run_test 108b "migrate from ZFS to ldiskfs"
8839 # set number of permanent parameters
8841 test_109_set_params() {
8844 set_persistent_param_and_check mds \
8845 "mdd.$fsname-MDT0000.atime_diff" \
8846 "$fsname-MDT0000.mdd.atime_diff" \
8848 set_persistent_param_and_check mds \
8849 "mdd.$fsname-MDT0000.atime_diff" \
8850 "$fsname-MDT0000.mdd.atime_diff" \
8852 set_persistent_param_and_check client \
8853 "llite.$fsname*.max_read_ahead_mb" \
8854 "$fsname.llite.max_read_ahead_mb" \
8856 set_persistent_param_and_check client \
8857 "llite.$fsname*.max_read_ahead_mb" \
8858 "$fsname.llite.max_read_ahead_mb" \
8860 create_pool $fsname.pool1 || error "create pool failed"
8861 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
8862 error "pool_add failed"
8863 do_facet mgs $LCTL pool_remove $fsname.pool1 OST0000 ||
8864 error "pool_remove failed"
8865 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
8866 error "pool_add failed"
8870 # check permanent parameters
8872 test_109_test_params() {
8875 local atime_diff=$(do_facet mds $LCTL \
8876 get_param -n mdd.$fsname-MDT0000.atime_diff)
8877 [ $atime_diff == 63 ] || error "wrong mdd parameter after clear_conf"
8878 local max_read_ahead_mb=$(do_facet client $LCTL \
8879 get_param -n llite.$fsname*.max_read_ahead_mb)
8880 [ $max_read_ahead_mb == 64 ] ||
8881 error "wrong llite parameter after clear_conf"
8882 local ost_in_pool=$(do_facet mds $LCTL pool_list $fsname.pool1 |
8883 grep -v "^Pool:" | sed 's/_UUID//')
8884 [ $ost_in_pool = "$fsname-OST0000" ] ||
8885 error "wrong pool after clear_conf"
8889 # run lctl clear_conf, store CONFIGS before and after that
8891 test_109_clear_conf()
8893 local clear_conf_arg=$1
8896 if ! combined_mgs_mds ; then
8897 mgsdev=$(mgsdevname)
8898 stop_mgs || error "stop_mgs failed"
8899 start_mgs "-o nosvc" || error "start_mgs nosvc failed"
8901 mgsdev=$(mdsdevname 1)
8902 start_mdt 1 "-o nosvc" || error "start_mdt 1 nosvc failed"
8905 do_facet mgs "rm -rf $TMP/${tdir}/conf1; mkdir -p $TMP/${tdir}/conf1;" \
8906 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf1\\\" \
8910 # the command being tested
8912 do_facet mgs $LCTL clear_conf $clear_conf_arg ||
8913 error "clear_conf failed"
8914 if ! combined_mgs_mds ; then
8915 stop_mgs || error "stop_mgs failed"
8917 stop_mdt 1 || error "stop_mdt 1 failed"
8920 do_facet mgs "rm -rf $TMP/${tdir}/conf2; mkdir -p $TMP/${tdir}/conf2;" \
8921 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf2\\\" \
8925 test_109_file_shortened() {
8927 local sizes=($(do_facet mgs "stat -c %s " \
8928 "$TMP/${tdir}/conf1/CONFIGS/$file" \
8929 "$TMP/${tdir}/conf2/CONFIGS/$file"))
8930 [ ${sizes[1]} -lt ${sizes[0]} ] && return 0
8936 [ $MDS1_VERSION -lt $(version_code 2.10.59) ] &&
8937 skip "Needs MDS version 2.10.59 or later."
8938 [ "$(facet_fstype mgs)" == zfs ] &&
8939 skip "LU-8727: no implementation for ZFS"
8944 client_up || error "client_up failed"
8947 # set number of permanent parameters
8949 test_109_set_params $FSNAME
8951 umount_client $MOUNT || error "umount_client failed"
8952 stop_ost || error "stop_ost failed"
8953 stop_mds || error "stop_mds failed"
8955 test_109_clear_conf $FSNAME
8957 # make sure that all configs are cleared
8959 test_109_file_shortened $FSNAME-MDT0000 ||
8960 error "failed to clear MDT0000 config"
8961 test_109_file_shortened $FSNAME-client ||
8962 error "failed to clear client config"
8967 # check that configurations are intact
8969 test_109_test_params $FSNAME
8974 destroy_test_pools || error "destroy test pools failed"
8978 run_test 109a "test lctl clear_conf fsname"
8982 [ $MDS1_VERSION -lt $(version_code 2.10.59) ] &&
8983 skip "Needs MDS version 2.10.59 or later."
8984 [ "$(facet_fstype mgs)" == zfs ] &&
8985 skip "LU-8727: no implementation for ZFS"
8990 client_up || error "client_up failed"
8993 # set number of permanent parameters
8995 test_109_set_params $FSNAME
8997 umount_client $MOUNT || error "umount_client failed"
8998 stop_ost || error "stop_ost failed"
8999 stop_mds || error "stop_mds failed"
9001 test_109_clear_conf $FSNAME-MDT0000
9003 # make sure that only one config is cleared
9005 test_109_file_shortened $FSNAME-MDT0000 ||
9006 error "failed to clear MDT0000 config"
9007 test_109_file_shortened $FSNAME-client &&
9008 error "failed to clear client config"
9012 # check that configurations are intact
9014 test_109_test_params $FSNAME
9019 destroy_test_pools || error "destroy test pools failed"
9023 run_test 109b "test lctl clear_conf one config"
9027 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
9028 skip "Only applicable to ldiskfs-based MDTs"
9030 do_facet $SINGLEMDS $DEBUGFS -w -R supported_features |grep large_dir ||
9031 skip "large_dir option is not supported on MDS"
9032 do_facet ost1 $DEBUGFS -w -R supported_features | grep large_dir ||
9033 skip "large_dir option is not supported on OSS"
9035 stopall # stop all targets before modifying the target counts
9036 local old_mdscount=$MDSCOUNT
9037 local old_ostcount=$OSTCOUNT
9039 stack_trap "MDSCOUNT=$old_mdscount OSTCOUNT=$old_ostcount; reformat"
9043 # ext4_dir_entry_2 struct size:264
9044 # dx_root struct size:8
9045 # dx_node struct size:8
9046 # dx_entry struct size:8
9047 # For 1024 bytes block size.
9048 # First level directory entries: 126
9049 # Second level directory entries: 127
9050 # Entries in leaf: 3
9051 # For 2 levels limit: 48006
9052 # For 3 levels limit : 6096762
9053 # Create 80000 files to safely exceed 2-level htree limit.
9054 CONF_SANITY_110_LINKS=${CONF_SANITY_110_LINKS:-80000}
9056 # can fit at most 3 filenames per 1KB leaf block, but each
9057 # leaf/index block will only be 3/4 full before split at each level
9058 (( MDSSIZE < CONF_SANITY_110_LINKS / 3 * 4/3 * 4/3 )) &&
9059 CONF_SANITY_110_LINKS=$((MDSSIZE * 3 * 3/4 * 3/4))
9061 combined_mgs_mds || replace=" --replace "
9062 local opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
9063 $replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
9064 if [[ $opts != *large_dir* ]]; then
9065 if [[ $opts != *mkfsoptions* ]]; then
9066 opts+=" --mkfsoptions=\\\"-O large_dir -b 1024 -i 65536\\\""
9068 opts="${opts//--mkfsoptions=\\\"/ \
9069 --mkfsoptions=\\\"-O large_dir -b 1024 -i 65536 }"
9072 echo "MDT params: $opts"
9074 combined_mgs_mds || start_mgs
9075 add mds1 $opts || error "add mds1 failed with new params"
9076 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
9078 opts="$(mkfs_opts ost1 $(ostdevname 1)) \
9079 $replace --reformat $(ostdevname 1) $(ostvdevname 1)"
9081 if [[ $opts != *large_dir* ]]; then
9082 if [[ $opts != *mkfsoptions* ]]; then
9083 opts+=" --mkfsoptions=\\\"-O large_dir\\\" "
9085 opts="${opts//--mkfsoptions=\\\"/ \
9086 --mkfsoptions=\\\"-O large_dir }"
9089 echo "OST params: $opts"
9090 add ost1 $opts || error "add ost1 failed with new params"
9091 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
9093 MOUNT_2=yes mountcli || error "mount clients failed"
9095 mkdir -v $DIR/$tdir || error "cannot create $DIR/$tdir"
9096 local pids count=0 group=0
9098 echo "creating $CONF_SANITY_110_LINKS in total"
9099 while (( count < CONF_SANITY_110_LINKS )); do
9100 local len=$((253 - $(wc -c <<<"$tfile-$group-40000-")))
9101 local dir=DIR$((group % 2 + 1))
9102 local target=${!dir}/$tdir/$tfile-$group
9103 local long=$target-$(generate_name $len)-
9104 local create=$((CONF_SANITY_110_LINKS - count))
9106 (( create > 40000 )) && create=40000
9107 touch $target || error "creating $target failed"
9108 echo "creating $create hard links to $target"
9109 createmany -l $target $long $create &
9112 count=$((count + create))
9113 group=$((group + 1))
9115 echo "waiting for PIDs$pids to complete"
9116 wait $pids || error "createmany failed after $group groups"
9118 umount_client $MOUNT2 -f
9121 run_e2fsck $(facet_active_host mds1) $(mdsdevname 1) -n
9122 MDSCOUNT=$old_mdscount
9123 OSTCOUNT=$old_ostcount
9125 run_test 110 "Adding large_dir with 3-level htree"
9128 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
9129 skip "Only applicable to ldiskfs-based MDTs"
9131 is_dm_flakey_dev $SINGLEMDS $(mdsdevname 1) &&
9132 skip "This test can not be executed on flakey dev"
9134 do_facet $SINGLEMDS $DEBUGFS -w -R supported_features |grep large_dir ||
9135 skip "large_dir option is not supported on MDS"
9137 do_facet ost1 $DEBUGFS -w -R supported_features | grep large_dir ||
9138 skip "large_dir option is not supported on OSS"
9140 # cleanup before changing target counts
9142 local old_mdscount=$MDSCOUNT
9143 local old_ostcount=$OSTCOUNT
9144 local old_mdssize=$MDSSIZE
9146 stack_trap "MDSSIZE=$old_mdssize MDSCOUNT=$old_mdscount \
9147 OSTCOUNT=$old_ostcount; reformat"
9150 (( MDSSIZE < 2400000 )) && MDSSIZE=2400000 # need at least 2.4GB
9152 local mdsdev=$(mdsdevname 1)
9153 combined_mgs_mds || replace=" --replace "
9154 local opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
9155 $replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
9156 if [[ $opts != *large_dir* ]]; then
9157 if [[ $opts != *mkfsoptions* ]]; then
9158 opts+=" --mkfsoptions=\\\"-O large_dir -i 1048576 \\\" "
9160 opts="${opts//--mkfsoptions=\\\"/ \
9161 --mkfsoptions=\\\"-O large_dir -i 1048576 }"
9164 echo "MDT params: $opts"
9166 combined_mgs_mds || start_mgs
9167 __touch_device mds 1
9168 add mds1 $opts || error "add mds1 failed with new params"
9169 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
9171 opts="$(mkfs_opts ost1 $(ostdevname 1)) \
9172 $replace --reformat $(ostdevname 1) $(ostvdevname 1)"
9173 if [[ $opts != *large_dir* ]]; then
9174 if [[ $opts != *mkfsoptions* ]]; then
9175 opts+=" --mkfsoptions=\\\"-O large_dir \\\""
9177 opts="${opts//--mkfsoptions=\\\"/ --mkfsoptions=\\\"-O large_dir }"
9180 echo "OST params: $opts"
9181 __touch_device ost 1
9182 add ost1 $opts || error "add ost1 failed with new params"
9183 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
9185 MOUNT_2=yes mountcli
9186 mkdir $DIR/$tdir || error "cannot create $DIR/$tdir"
9188 lfs df -i $DIR/$tdir
9192 local start=$SECONDS
9194 local dirmax=$((2 << 30))
9200 while (( !needskip & dirsize < dirmax )); do
9203 for cli in ${CLIENTS//,/ }; do
9204 local len=$((253 - $(wc -c <<<"$cli-$group-60000-")))
9205 local target=$cli-$group
9206 local long=$DIR/$tdir/$target-$(generate_name $len)-
9208 RPWD=$DIR/$tdir do_node $cli touch $target ||
9209 error "creating $target failed"
9210 echo "creating 60000 hardlinks to $target"
9211 RPWD=$DIR/$tdir do_node $cli createmany -l $target $long 60000 &
9214 group=$((group + 1))
9216 long=$DIR2/$tdir/$target-$(generate_name $len)-
9218 RPWD=$DIR2/$tdir do_node $cli touch $target ||
9219 error "creating $target failed"
9220 echo "creating 60000 hardlinks to $target"
9221 RPWD=$DIR2/$tdir do_node $cli createmany -l $target $long 60000 &
9224 group=$((group + 1))
9226 echo "waiting for PIDs$pids to complete"
9227 wait $pids || error "createmany failed after $group groups"
9228 dirsize=$(stat -c %s $DIR/$tdir)
9229 taken=$((SECONDS - start))
9230 rate=$((dirsize / taken))
9231 left=$(((dirmax - dirsize) / rate))
9232 num=$((group * 60000))
9233 echo "estimate ${left}s left after $num files / ${taken}s"
9234 # if the estimated time remaining is too large (it may change
9235 # over time as the create rate is not constant) then exit
9236 # without declaring a failure.
9237 (( left > 1200 )) && needskip=1
9240 umount_client $MOUNT2 -f
9243 ! (( $needskip )) ||
9244 echo "ETA ${left}s after $num files / ${taken}s is too long"
9246 run_e2fsck $(facet_active_host mds1) $(mdsdevname 1) -n
9247 MDSCOUNT=$old_mdscount
9248 OSTCOUNT=$old_ostcount
9249 MDSSIZE=$old_mdssize
9251 run_test 111 "Adding large_dir with over 2GB directory"
9254 local param="no_create"
9256 (( $OST1_VERSION > $(version_code 2.14.0) )) ||
9257 skip "need OSS at least 2.14.0"
9258 (( $OST1_VERSION >= $(version_code 2.15.56.125) )) ||
9259 param="no_precreate"
9261 start_mds || error "MDS start failed"
9262 start_ost || error "OSS start failed"
9263 echo "start ost2 service on $(facet_active_host ost2)"
9264 start ost2 $(ostdevname 2) $(csa_add "$OST_MOUNT_OPTS" -o $param) ||
9265 error "start ost2 facet failed"
9266 local val=$(do_facet ost2 \
9267 "$LCTL get_param -n obdfilter.$FSNAME-OST0001*.$param")
9268 (( $val == 1 )) || error "obdfilter.$FSNAME-OST0001*.$param=$val"
9270 mount_client $MOUNT || error "mount client failed"
9271 wait_osc_import_state mds1 ost1 FULL
9272 wait_osc_import_ready client ost1
9273 wait_osc_import_ready client ost2
9275 $LFS setstripe -i 0 $DIR/$tfile.0 ||
9276 error "problem creating $tfile.0 on OST0000"
9277 $LFS setstripe -i 1 $DIR/$tfile.1 && $LFS getstripe $DIR/$tfile.1 &&
9278 (( $($LFS getstripe -i $DIR/$tfile.1) == 1 )) &&
9279 error "allowed to create $tfile.1 on OST0001"
9281 $LFS df -v $MOUNT | grep -q "OST:1.*N" ||
9282 error "NOCREATE not in 'lfs df'"
9284 do_facet ost2 $LCTL set_param obdfilter.$FSNAME-OST0001*.$param=0
9286 $LFS setstripe -i 1 $DIR/$tfile.2 ||
9287 error "failed to create $tfile.2 on ost1 facet"
9288 # files not cleaned with ONLY_REPEAT because of client unmount below
9289 rm -f $DIR/$tfile.[012]
9290 stop_ost2 || error "stop ost2 facet failed"
9293 run_test 112a "mount OST with no_create option"
9296 (( MDSCOUNT >= 2 )) || skip "need at least 2 MDTs"
9297 (( $MDS1_VERSION >= $(version_code 2.15.56.125) )) ||
9298 skip "need MDS >= 2.15.56.125"
9299 local mdsnum=$MDSCOUNT
9300 local facet=mds$mdsnum
9301 local mdtidx=$((mdsnum - 1))
9302 local mdtname=$FSNAME-MDT$(printf %04x $mdtidx)
9304 start_mds --mdscount $((mdsnum - 1)) || error "MDS start failed"
9305 start_mdt $mdsnum -o no_create || error "start $facet failed"
9306 local val=$(do_facet $facet \
9307 "$LCTL get_param -n mdt.$mdtname*.no_create")
9308 (( $val == 1 )) || error "mdt.$mdtname*.no_create=$val"
9309 start_ost || error "ost1 start failed"
9310 start_ost2 || error "ost1 start failed"
9312 mount_client $MOUNT || error "mount client failed"
9313 wait_osc_import_ready $facet ost2
9316 $LFS df -v $MOUNT | grep -q "MDT:$mdtidx.*N" ||
9317 error "NOCREATE not in 'lfs df'"
9319 $LFS mkdir -i $mdtidx $DIR/$tdir ||
9320 $LFS setdirstripe -D -c 1 -i -1 --max-inherit-rr 2 $DIR/$tdir ||
9321 error "error creating $tdir on $mdtname"
9322 stack_trap "rm -rf $DIR/$tdir"
9324 mkdir $DIR/$tdir/d1.{1..100} || error "mkdir $tdir/d1.{1..100} failed"
9325 $LFS getdirstripe -i $DIR/$tdir/d1.* | sort | uniq -c
9326 do_facet $facet $LCTL set_param mdt.$mdtname*.no_create=0
9327 # allow one initial create for delayed statfs on client
9328 (( $($LFS getdirstripe -i $DIR/$tdir/d1.* | grep -c $mdtidx) < 2 )) ||
9329 error "allowed create on $mdtname"
9332 mkdir $DIR/$tdir/d2.{1..100} || error "mkdir $tdir/d2.{1..100} failed"
9333 $LFS getdirstripe -i $DIR/$tdir/d2.{1..100} | sort | uniq -c
9334 (( $($LFS getdirstripe -i $DIR/$tdir/d2.* | grep -c $mdtidx) > 10 )) ||
9335 error "no create on $mdtname"
9336 # files not cleaned with ONLY_REPEAT because of client unmount below
9338 stop_ost2 || error "ost1 start failed"
9341 run_test 112b "mount MDT with no_create option"
9344 SAVE_MGS_MOUNT_OPTS=$MGS_MOUNT_OPTS
9345 SAVE_MDS_MOUNT_OPTS=$MDS_MOUNT_OPTS
9346 SAVE_OST_MOUNT_OPTS=$OST_MOUNT_OPTS
9352 MGS_MOUNT_OPTS=$SAVE_MGS_MOUNT_OPTS
9353 MDS_MOUNT_OPTS=$SAVE_MDS_MOUNT_OPTS
9354 OST_MOUNT_OPTS=$SAVE_OST_MOUNT_OPTS
9355 # Revert old mount options back
9357 # Subsequent following test requires
9358 # conf-sanity to be in stopall state.
9359 # Force 'stopall' so others following
9364 # Error out with mount info
9366 local server_nodes=$(comma_list $(mdts_nodes) $(osts_nodes))
9369 echo "--Client Mount Info--"
9370 mount | grep -i lustre
9371 echo "--Server Mount Info--"
9372 do_nodes $server_nodes mount | grep -i lustre
9378 local ost_version="2.15.51" # Minimum version required
9380 (( OST1_VERSION >= $(version_code $ost_version) )) ||
9381 skip "Need server version at least $ost_version"
9383 stack_trap cleanup_113 EXIT
9385 # Reset before starting
9389 # Verify MDS's should start with "rw"
9390 do_facet $SINGLEMDS mount | grep "lustre.*rw,.*MDT" ||
9391 error_113 "$SINGLEMDS should be read-write"
9393 # Verify OST's should start with "rw"
9394 for (( i=1; i <= OSTCOUNT; i++ )); do
9395 do_facet ost$i mount | grep "lustre.*rw,.*OST" ||
9396 error_113 "ost$i should be read-write"
9399 # rdonly_dev does not currently work for ldiskfs
9400 # We skip the rdonly_dev check until then.
9401 if [[ $ost1_FSTYPE == ldiskfs ]]; then
9402 echo "Shadow Mountpoint correctly reports rw for ldiskfs"
9407 # Only ZFS specific tests below.
9410 # Must stop all (server+client) and restart to verify new
9414 # add rdonly_dev to mount option
9415 MGS_MOUNT_OPTS=$(csa_add "$MGS_MOUNT_OPTS" -o rdonly_dev)
9416 MDS_MOUNT_OPTS=$(csa_add "$MDS_MOUNT_OPTS" -o rdonly_dev)
9417 OST_MOUNT_OPTS=$(csa_add "$OST_MOUNT_OPTS" -o rdonly_dev)
9419 # Only restart server(mds/ost). Sufficient for test
9420 setupall server_only || error "Fail to start servers"
9422 # Verify MDS's should be "ro"
9423 do_facet $SINGLEMDS mount | grep "lustre.*ro,.*MDT.*rdonly_dev" ||
9424 error_113 "$SINGLEMDS should be read-only"
9426 # Verify OST's should be "ro"
9427 for (( i=1; i <= OSTCOUNT; i++ )); do
9428 do_facet ost$i mount | grep "lustre.*ro,.*OST.*rdonly_dev" ||
9429 error_113 "ost$i should be read-only"
9432 run_test 113 "Shadow mountpoint correctly report ro/rw for mounts"
9435 # Purpose: To verify dynamic thread (OSS) creation.
9436 # (This was sanity/115)
9439 [ $PARALLEL == "yes" ] && skip "skip parallel run"
9440 remote_ost_nodsh && skip "remote OST with nodsh"
9442 # Lustre does not stop service threads once they are started.
9443 # Reset number of running threads to default.
9448 local save_params="$TMP/sanity-$TESTNAME.parameters"
9450 # Get ll_ost_io count before I/O
9451 OSTIO_pre=$(do_facet ost1 \
9452 "$LCTL get_param ost.OSS.ost_io.threads_started | cut -d= -f2")
9453 # Exit if lustre is not running (ll_ost_io not running).
9454 [ -z "$OSTIO_pre" ] && error "no OSS threads"
9456 echo "Starting with $OSTIO_pre threads"
9457 local thread_max=$((OSTIO_pre * 2))
9458 local rpc_in_flight=$((thread_max * 2))
9460 # this is limited to OSC_MAX_RIF_MAX (256)
9461 [ $rpc_in_flight -gt 256 ] && rpc_in_flight=256
9462 thread_max=$((rpc_in_flight / 2))
9463 [ $thread_max -le $OSTIO_pre ] && skip "Too many ost_io threads" &&
9466 # Number of I/O Process proposed to be started.
9468 local facets=$(get_facets OST)
9470 save_lustre_params client "osc.*OST*.max_rpcs_in_flight" > $save_params
9471 save_lustre_params $facets "ost.OSS.ost_io.threads_max" >> $save_params
9473 # Set in_flight to $rpc_in_flight
9474 $LCTL set_param osc.*OST*.max_rpcs_in_flight=$rpc_in_flight ||
9475 error "Failed to set max_rpcs_in_flight to $rpc_in_flight"
9476 nfiles=${rpc_in_flight}
9477 # Set ost thread_max to $thread_max
9478 do_facet ost1 "$LCTL set_param ost.OSS.ost_io.threads_max=$thread_max"
9480 # 5 Minutes should be sufficient for max number of OSS
9481 # threads(thread_max) to be created.
9485 local wtl=${WTL:-"$LUSTRE/tests/write_time_limit"}
9487 test_mkdir $DIR/$tdir
9488 for ((i = 1; i <= nfiles; i++)); do
9489 local file=$DIR/$tdir/${tfile}-$i
9491 $LFS setstripe -c -1 -i 0 $file
9492 ($wtl $file $timeout)&
9495 # I/O Started - Wait for thread_started to reach thread_max or report
9496 # error if thread_started is more than thread_max.
9497 echo "Waiting for thread_started to reach thread_max"
9498 local thread_started=0
9499 local end_time=$((SECONDS + timeout))
9501 while [ $SECONDS -le $end_time ] ; do
9503 # Get ost i/o thread_started count.
9504 thread_started=$(do_facet ost1 \
9506 ost.OSS.ost_io.threads_started | cut -d= -f2")
9507 # Break out if thread_started is equal/greater than thread_max
9508 if (( $thread_started >= $thread_max )); then
9509 echo ll_ost_io thread_started $thread_started, \
9510 equal/greater than thread_max $thread_max
9516 # Cleanup - We have the numbers, Kill i/o jobs if running.
9517 jobcount=($(jobs -p))
9519 for ((i=0; i < ${#jobcount[*]}; i++)); do
9520 kill -9 ${jobcount[$i]}
9521 if [ $? -ne 0 ] ; then
9522 echo "warning: cannot kill WTL pid ${jobcount[$i]}"
9526 # Cleanup files left by WTL binary.
9527 for ((i = 1; i <= nfiles; i++)); do
9528 local file=$DIR/$tdir/${tfile}-$i
9531 if [ $? -ne 0 ] ; then
9532 echo "Warning: Failed to delete file $file"
9536 restore_lustre_params <$save_params
9537 rm -f $save_params || echo "Warning: delete file '$save_params' failed"
9539 # Error out if no new thread has started or Thread started is greater
9541 if (( $thread_started <= $OSTIO_pre ||
9542 $thread_started > $thread_max )); then
9543 error "ll_ost_io: thread_started $thread_started" \
9544 "OSTIO_pre $OSTIO_pre, thread_max $thread_max." \
9545 "No new thread started or thread started greater " \
9549 run_test 114 "verify dynamic thread creation===================="
9555 do_facet mds1 rm -f $1
9559 if [ "$mds1_FSTYPE" != ldiskfs ]; then
9560 skip "Only applicable to ldiskfs-based MDTs"
9562 [ -n "$FILESET" ] && skip "Not functional for FILESET set"
9564 local dbfs_ver=$(do_facet mds1 $DEBUGFS -V 2>&1)
9566 echo "debugfs version: $dbfs_ver"
9567 echo "$dbfs_ver" | egrep -w "1.44.3.wc1|1.44.5.wc1|1.45.2.wc1" &&
9568 skip_env "This version of debugfs doesn't show inode number"
9570 local IMAGESIZE=$((3072 << 30)) # 3072 GiB
9580 do_facet mds1 "hostname; $LCTL dl; mount"
9582 do_facet ost1 "hostname; $LCTL dl; mount"
9583 # We need MDT size 3072GB, because it is smallest
9584 # partition that can store 2B inodes
9585 do_facet mds1 "mkdir -p $TMP/$tdir"
9586 local mdsimgname=$TMP/$tdir/lustre-mdt
9588 do_facet mds1 "rm -f $mdsimgname"
9589 do_facet mds1 "touch $mdsimgname"
9590 stack_trap "cleanup_115 $mdsimgname" EXIT
9591 do_facet mds1 "$TRUNCATE $mdsimgname $IMAGESIZE" ||
9592 skip "Backend FS doesn't support sparse files"
9593 local mdsdev=$(do_facet mds1 "losetup -f")
9595 do_facet mds1 "losetup $mdsdev $mdsimgname"
9597 local mds_opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
9598 --mkfsoptions='-O ea_inode,^resize_inode,meta_bg \
9599 -N 2247484000 -E lazy_itable_init' --device-size=$IMAGESIZE"
9600 add mds1 $mds_opts --mgs --reformat $mdsdev || {
9601 do_facet $SINGLEMDS \
9602 "losetup -d $mdsdev && rm -f $mdsimgname"
9603 skip_env "format large MDT failed"
9606 local ostdev=$(ostdevname 1)
9608 local opts="$(mkfs_opts ost1 $ostdev) \
9609 --reformat $ostdev $ostdev"
9610 add ost1 $opts || error "add ost1 failed with new params"
9611 start mds1 $mdsdev $MDS_MOUNT_OPTS || error "start MDS failed"
9612 start_ost || error "start OSS failed"
9613 mount_client $MOUNT || error "mount client failed"
9615 mkdir -p $DIR/$tdir || error "mkdir $DIR/$tdir fail"
9616 local goal="/sys/fs/ldiskfs/$(basename $mdsdev)/inode_goal"
9618 # 2147483648 is 0x80000000
9619 do_facet mds1 "echo 2147483648 >> $goal; grep . $goal"
9620 touch $DIR/$tdir/$tfile
9622 # attrs from 1 to 16 go to block, 17th - to inode
9625 for i in {1..17}; do
9626 local nm="trusted.ea$i"
9627 setfattr -n $nm -v $(printf "xattr%0250d" $i) $DIR/$tdir/$tfile
9630 do_facet mds1 "$DEBUGFS -c -R 'stat ROOT/$tdir/$tfile' $mdsdev"
9632 # inode <2147483649> trusted.ea16 (255)
9633 local inode_num=$(do_facet mds1 \
9634 "$DEBUGFS -c -R 'stat ROOT/$tdir/$tfile' $mdsdev" |
9635 awk '/ea17/ { print $2 }' |
9636 sed -e 's/>//' -e 's/<//' -e 's/\"//')
9637 echo "inode num: $inode_num"
9638 [ $inode_num -ge 2147483648 ] || error "inode $inode_num too small"
9639 do_facet mds1 "losetup -d $mdsdev"
9641 run_test 115 "Access large xattr with inodes number over 2TB"
9644 [ "$mds1_FSTYPE" != ldiskfs ] && skip "ldiskfs only test"
9645 [ "$MDS1_VERSION" -lt $(version_code 2.10.59) ] &&
9646 skip "Need server version at least 2.10.59"
9647 do_facet $SINGLEMDS which mkfs.xfs ||
9648 skip_env "No mkfs.xfs installed"
9653 local tmpmnt=$TMP/$tdir
9654 local mdtimg=$tfile-mdt0
9656 do_facet $SINGLEMDS mkdir -p $tmpmnt
9657 stack_trap "do_facet $SINGLEMDS rmdir $tmpmnt" EXIT
9659 do_facet $SINGLEMDS touch $TMP/$mdtimg
9660 stack_trap "do_facet $SINGLEMDS rm -f $TMP/$mdtimg" EXIT
9661 do_facet $SINGLEMDS mkfs -t xfs -d file,size=1t,name=$TMP/$mdtimg ||
9662 error "mkfs temporary xfs image"
9664 do_facet $SINGLEMDS mount $TMP/$mdtimg $tmpmnt ||
9665 error "mount temporary xfs image"
9666 stack_trap "do_facet $SINGLEMDS umount $tmpmnt" EXIT
9667 local old_mdssize=$MDSSIZE
9668 local old_mdsisize=$MDSISIZE
9670 MDSSIZE=$((17 * 1024 * 1024 * 1024)) # 17T MDT
9671 MDSISIZE=$((16 << 20))
9672 local opts17t="$(mkfs_opts $SINGLEMDS)"
9674 MDSSIZE=$old_mdssize
9675 MDSISIZE=$old_mdsisize
9677 do_facet $SINGLEMDS $MKFS \
9678 ${opts17t/-E /-E lazy_itable_init,lazy_journal_init,} \
9680 error "failed to mkfs for $tmpmnt/$mdtimg"
9682 do_facet $SINGLEMDS $TUNE2FS -l $tmpmnt/$mdtimg |
9683 grep -qw 'features.*extent' || error "extent should be enabled"
9686 run_test 116 "big size MDT support"
9689 # Call setup only if LustreFS is not mounted
9690 check_mount || setup
9692 do_facet ost1 "$LCTL set_param ost.OSS.ost_io.nrs_policies=fifo"
9693 do_facet ost1 "$LCTL get_param -n ost.OSS.ost_io.nrs_tbf_rule" &&
9694 error "get_param should fail"
9695 cleanup || error "cleanup failed with rc $?"
9697 run_test 117 "lctl get_param return errors properly"
9703 [[ "$MDSCOUNT" -ge 2 ]] || skip "Need more at least 2 MDTs"
9705 had_config=$(do_facet mds1 "$LCTL get_param debug | grep config")
9706 do_facet mds1 "$LCTL set_param debug=+config"
9707 do_facet mds1 "$LCTL clear"
9710 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)" &>/dev/null
9711 # mount after writeconf will make "add osp" added to mdt0 config:
9712 # 53 (224)marker 60 (flags=0x01, v2.5.1.0) lustre-MDT0001 'add osp'
9713 # 54 (080)add_uuid nid=... 0: 1:...
9714 # 55 (144)attach 0:lustre-MDT0001-osp-MDT0000 1:osp 2:...
9715 # 56 (144)setup 0:lustre-MDT0001-osp-MDT0000 1:... 2:...
9716 # 57 (136)modify_mdc_tgts add 0:lustre-MDT0000-mdtlov 1:... 2:1 3:1
9717 # duplicate modify_mdc_tgts caused crashes
9720 # using larger debug_mb size to avoid lctl dk log truncation
9721 size_mb=$((DEBUG_SIZE_SAVED * 4))
9724 # though config processing stops after failed attach and setup
9725 # it will proceed after the failed command after each writeconf
9726 # this is the original scenario of the issue
9727 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)" &>/dev/null
9728 do_facet mds1 "$LCTL set_param debug_mb=$size_mb"
9731 wait_update_facet_cond mds1 \
9732 "$LCTL dk | grep -c Processed.log.$FSNAME-MDT0000" \
9737 [[ -z "$had_config" ]] && do_facet mds1 lctl set_param debug=-config
9741 run_test 119 "writeconf on slave mdt shouldn't duplicate mdc/osp and crash"
9743 test_120() { # LU-11130
9744 [ "$MDSCOUNT" -lt 2 ] && skip "mdt count < 2"
9745 [ "$mds1_FSTYPE" != ldiskfs ] &&
9746 skip "ldiskfs only test"
9747 [ "$MDS1_VERSION" -lt $(version_code 2.11.56) ] &&
9748 skip "Need DNE2 capable MD target with LU-11130 fix"
9752 local mds1host=$(facet_active_host mds1)
9753 local mds1dev=$(mdsdevname 1)
9755 $LFS mkdir -i 1 $DIR/$tdir
9756 $LFS mkdir -i 0 $DIR/$tdir/mds1dir
9758 ln -s foo $DIR/$tdir/bar
9759 mv $DIR/$tdir/bar $DIR/$tdir/mds1dir/bar2 ||
9760 error "cross-target rename failed"
9764 run_e2fsck $mds1host $mds1dev "-n"
9766 run_test 120 "cross-target rename should not create bad symlinks"
9770 start_mgsmds || error "MGS MDS Start failed"
9772 stop_mds || error "Stopping MDSes failed"
9776 stop_mds || error "Stopping MDSes failed"
9778 run_test 121 "failover MGS"
9781 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
9782 [[ "$OST1_VERSION" -ge $(version_code 2.11.53) ]] ||
9783 skip "Need OST version at least 2.11.53"
9786 LOAD_MODULES_REMOTE=true load_modules
9787 #define OBD_FAIL_OFD_SET_OID 0x1e0
9788 do_facet ost1 $LCTL set_param fail_loc=0x00001e0
9791 $LFS mkdir -i1 -c1 $DIR/$tdir
9792 $LFS setstripe -i0 -c1 $DIR/$tdir
9793 do_facet ost1 $LCTL set_param fail_loc=0
9794 createmany -o $DIR/$tdir/file_ 1000 ||
9795 error "Fail to create a new sequence"
9799 run_test 122a "Check OST sequence update"
9802 remote_mgs_nodsh && skip "remote MGS with nodsh"
9803 [ -d $MOUNT/.lustre ] || setup
9805 # test old logid format until removal from llog_ioctl.c::str2logid()
9806 if (( $MGS_VERSION < $(version_code 2.99.53) )); then
9807 do_facet mgs $LCTL dl | grep MGS
9808 do_facet mgs "$LCTL --device %MGS llog_print \
9809 \\\\\\\$$FSNAME-client 1 10" ||
9810 error "old llog_print failed"
9813 # test new logid format
9814 if [ $MGS_VERSION -ge $(version_code 2.9.53) ]; then
9815 do_facet mgs "$LCTL --device MGS llog_print $FSNAME-client" ||
9816 error "new llog_print failed"
9819 run_test 123aa "llog_print works with FIDs and simple names"
9822 remote_mgs_nodsh && skip "remote MGS with nodsh"
9823 [[ $MGS_VERSION -gt $(version_code 2.11.51) ]] ||
9824 skip "Need server with working llog_print support"
9826 [ -d $MOUNT/.lustre ] || setup
9832 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9833 mgs_arg="--device MGS"
9835 orig_val=$(do_facet mgs $LCTL get_param jobid_name)
9836 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME"
9838 yaml=$(do_facet mgs $LCTL $mgs_arg llog_print params |
9839 grep jobid_name | tail -n 1)
9841 local param=$(awk '{ print $10 }' <<< "$yaml")
9842 local val=$(awk '{ print $12 }' <<< "$yaml")
9843 #return to the default
9844 do_facet mgs $LCTL set_param -P jobid_name=$orig_val
9845 [ $val = "TESTNAME" ] || error "bad value: $val"
9846 [ $param = "jobid_name," ] || error "Bad param: $param"
9848 run_test 123ab "llog_print params output values from set_param -P"
9850 test_123ac() { # LU-11566
9851 remote_mgs_nodsh && skip "remote MGS with nodsh"
9852 do_facet mgs "$LCTL help llog_print" 2>&1 | grep -q -- --start ||
9853 skip "Need 'lctl llog_print --start' on MGS"
9859 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9860 mgs_arg="--device MGS"
9862 [ -d $MOUNT/.lustre ] || setup
9864 # - { index: 10, event: add_uuid, nid: 192.168.20.1@tcp(0x20000c0a81401,
9865 # node: 192.168.20.1@tcp }
9866 do_facet mgs $LCTL $mgs_arg \
9867 llog_print --start $start --end $end $FSNAME-client | tr -d , |
9868 while read DASH BRACE INDEX idx EVENT BLAH BLAH BLAH; do
9869 (( idx >= start )) || error "llog_print index $idx < $start"
9870 (( idx <= end )) || error "llog_print index $idx > $end"
9873 run_test 123ac "llog_print with --start and --end"
9875 test_123ad() { # LU-11566
9876 remote_mgs_nodsh && skip "remote MGS with nodsh"
9877 # older versions of lctl may not print all records properly
9878 do_facet mgs "$LCTL help llog_print" 2>&1 | grep -q -- --start ||
9879 skip "Need 'lctl llog_print --start' on MGS"
9881 [ -d $MOUNT/.lustre ] || setup
9883 # append a new record, to avoid issues if last record was cancelled
9884 local old=$($LCTL get_param -n osc.*-OST0000-*.max_dirty_mb | head -1)
9885 do_facet mgs $LCTL conf_param $FSNAME-OST0000.osc.max_dirty_mb=$old
9887 # logid: [0x3:0xa:0x0]:0
9891 local num=$(do_facet mgs $LCTL --device MGS llog_info $FSNAME-client |
9892 awk '/last_index:/ { print $2 - 1 }')
9894 # - { index: 71, event: set_timeout, num: 0x14, param: sys.timeout=20 }
9895 local last=$(do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
9896 tail -1 | awk '{ print $4 }' | tr -d , )
9897 (( last == num )) || error "llog_print only showed $last/$num records"
9899 run_test 123ad "llog_print shows all records"
9901 test_123ae() { # LU-11566
9909 remote_mgs_nodsh && skip "remote MGS with nodsh"
9910 [ -d $MOUNT/.lustre ] || setupall
9912 max=$($LCTL get_param -n osc.*-OST0000-*.max_dirty_mb | head -1)
9913 pgs=$($LCTL get_param -n osc.*-OST0000-*.max_pages_per_rpc | head -1)
9914 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9915 mgs_arg="--device MGS"
9917 if do_facet mgs "$LCTL help llog_cancel" 2>&1| grep -q -- --log_id; then
9918 # save one set_param -P record in case none exist
9920 do_facet mgs $LCTL set_param -P osc.*.max_pages_per_rpc=$pgs
9921 stack_trap "do_facet mgs $LCTL set_param -P -d \
9922 osc.*.max_pages_per_rpc"
9925 orig=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9926 tail -1 | awk '{ print $4 }' | tr -d , )
9927 do_facet mgs $LCTL set_param -P osc.*.max_dirty_mb=$max
9928 do_facet mgs $LCTL $mgs_arg llog_print $log | tail -1 |
9929 grep "parameter: osc.*.max_dirty_mb" ||
9930 error "new set_param -P wasn't stored in params log"
9932 # - { index: 71, event: set_param, device: general,
9933 # param: osc.*.max_dirty_mb, value: 256 }
9934 id=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9935 tail -1 | awk '{ print $4 }' | tr -d , )
9936 do_facet mgs $LCTL $mgs_arg llog_cancel $log --log_idx=$id
9937 local new=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9938 tail -1 | awk '{ print $4 }' | tr -d , )
9939 (( new == orig )) ||
9940 error "new llog_cancel now $new, not at $orig records"
9943 # test old positional parameters for a while still
9944 if [ "$MGS_VERSION" -le $(version_code 3.1.53) ]; then
9947 do_facet mgs $LCTL conf_param \
9948 $FSNAME-OST0000.osc.max_pages_per_rpc=$pgs
9949 stack_trap "do_facet mgs $LCTL conf_param -d \
9950 $FSNAME-OST0000.osc.max_pages_per_rpc"
9952 orig=$(do_facet mgs $LCTL --device MGS llog_print $log |
9953 tail -1 | awk '{ print $4 }' | tr -d , )
9954 do_facet mgs $LCTL conf_param $FSNAME-OST0000.osc.max_dirty_mb=$max
9955 do_facet mgs $LCTL --device MGS llog_print $log |
9956 tail -1 | grep "parameter: osc.max_dirty_mb" ||
9957 error "old conf_param wasn't stored in params log"
9958 do_facet mgs $LCTL --device MGS llog_print $log
9959 # - { index: 71, event: conf_param, device: testfs-OST0000-osc,
9960 # param: osc.max_dirty_mb=256 }
9961 id=$(do_facet mgs $LCTL --device MGS llog_print $log |
9962 tail -1 | awk '{ print $4 }' | tr -d , )
9963 do_facet mgs $LCTL --device MGS llog_cancel $log $id
9964 do_facet mgs $LCTL --device MGS llog_print $log
9965 new=$(do_facet mgs $LCTL --device MGS llog_print $log |
9966 tail -1 | awk '{ print $4 }' | tr -d , )
9967 (( new == orig )) ||
9968 error "old llog_cancel now $new, not at $orig records"
9971 run_test 123ae "llog_cancel can cancel requested record"
9973 test_123af() { #LU-13609
9974 [ "$MGS_VERSION" -ge $(version_code 2.13.54) -a \
9975 "$MDS1_VERSION" -ge $(version_code 2.13.54) ] ||
9976 skip "Need both MGS and MDS version at least 2.13.54"
9978 [ -d $MOUNT/.lustre ] || setupall
9979 stack_trap "do_facet mds1 $LCTL set_param fail_loc=0" EXIT
9989 for device in "MGS" "$FSNAME-MDT0000"; do
9990 cmd="--device $device llog_catlist"
9991 echo "lctl $cmd ..."
9992 if [ "$device" = "MGS" ]; then
9997 orig_clist=($(do_facet $facet $LCTL $cmd | awk '{ print $2 }'))
9998 orig_count=${#orig_clist[@]}
9999 echo "orig_clist: ${orig_clist[*]}"
10001 #define OBD_FAIL_CATLIST 0x131b
10002 #fetch to llog records from the second one
10003 do_facet $facet $LCTL set_param fail_loc=0x131b fail_val=2
10005 new_clist=($(do_facet $facet $LCTL $cmd | awk '{ print $2 }'))
10006 new_count=${#new_clist[@]}
10007 echo "new_clist: ${new_clist[*]}"
10009 [ $new_count -eq $((orig_count - 1)) ] ||
10010 error "$new_count != $orig_count - 1"
10011 for i in $(seq 0 $new_count); do
10013 [ "${orig_clist[$j]}" = "${new_clist[$i]}" ] ||
10014 error "${orig_clist[$j]} != ${new_clist[$i]}"
10016 do_facet mds1 $LCTL set_param fail_loc=0
10020 run_test 123af "llog_catlist can show all config files correctly"
10022 test_123ag() { # LU-15142
10026 remote_mgs_nodsh && skip "remote MGS with nodsh"
10027 (( $MGS_VERSION >= $(version_code 2.14.55) )) ||
10028 skip "Need server version least 2.14.55"
10030 [ -d $MOUNT/.lustre ] || setup
10032 orig_val=$(do_facet mgs $LCTL get_param jobid_name)
10033 stack_trap "do_facet mgs $LCTL set_param -P jobid_name=$orig_val"
10035 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME1"
10036 do_facet mgs $LCTL set_param -P -d jobid_name
10037 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
10038 grep -c jobid_name)
10039 (( rec == 0 )) || error "parameter was not deleted, check #1"
10040 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME1"
10041 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
10042 grep -c jobid_name)
10043 (( rec == 1)) || error "parameter is not set"
10044 # usage with ordinary set_param format works too
10045 do_facet mgs $LCTL set_param -P -d jobid_name="ANY"
10046 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
10047 grep -c jobid_name)
10048 (( rec == 0 )) || error "parameter was not deleted, check #2"
10050 run_test 123ag "llog_print skips values deleted by set_param -P -d"
10052 test_123ah() { #LU-7668 del_ost
10053 (( "$MGS_VERSION" >= $(version_code 2.15.51) &&
10054 "$MDS1_VERSION" >= $(version_code 2.15.51) )) ||
10055 skip "Need MGS/MDS version >= 2.15.51 for del_ost"
10057 [ -d $MOUNT/.lustre ] || setupall
10058 stack_trap "do_facet mds1 $LCTL set_param fail_loc=0" EXIT
10060 local cmd="--device MGS llog_print"
10062 cli_llogcnt_orig=$(do_facet mgs $LCTL $cmd $FSNAME-client |
10063 grep -c $FSNAME-OST0000)
10064 mds1_llogcnt_orig=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 |
10065 grep -c $FSNAME-OST0000)
10067 [ $cli_llogcnt_orig -gt 0 ] ||
10068 error "$FSNAME-OST0000 not found (client)"
10069 [ $mds1_llogcnt_orig -gt 0 ] || error "$FSNAME-OST0000 not found (MDT)"
10071 # -n/--dryrun should NOT modify catalog
10072 do_facet mgs "$LCTL del_ost -n -t $FSNAME-OST0000" ||
10073 error "del_ost --dryrun failed with $?"
10075 local cli_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-client |
10076 grep -c $FSNAME-OST0000)
10077 local mds1_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 |
10078 grep -c $FSNAME-OST0000)
10080 [ $cli_llogcnt -eq $cli_llogcnt_orig ] ||
10081 error "--dryrun error: $cli_llogcnt != $cli_llogcnt_orig"
10082 [ $mds1_llogcnt -eq $mds1_llogcnt_orig ] ||
10083 error "--dryrun error: $mds1_llogcnt != $mds1_llogcnt_orig"
10086 do_facet mgs "$LCTL del_ost --target $FSNAME-OST0000" ||
10087 error "del_ost failed with $?"
10089 local cli_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-client |
10090 grep -c $FSNAME-OST0000)
10091 local mds1_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 |
10092 grep -c $FSNAME-OST0000)
10094 # every catalog entry for OST0000 should have been cancelled
10095 [ $cli_llogcnt -eq 0 ] || error "$FSNAME-OST0000 not cancelled (cli)"
10096 [ $mds1_llogcnt -eq 0 ] || error "$FSNAME-OST0000 not cancelled (MDT)"
10098 umount_client $MOUNT
10099 mount_client $MOUNT
10101 $LFS df | grep -q OST0000 && error "del_ost did not remove OST0000!"
10103 reformat_and_config
10105 run_test 123ah "del_ost cancels config log entries correctly"
10110 echo "cleanup test 123ai"
10112 # cancel last timeout record
10113 do_facet mgs "$LCTL set_param -P -d timeout"
10115 # restore timeout value
10116 do_nodes $(comma_list $(all_nodes)) "$LCTL set_param timeout=$timeout"
10119 test_123ai() { #LU-16167
10125 remote_mgs_nodsh && skip "remote MGS with nodsh"
10126 (( MDS1_VERSION >= $(version_code 2.15.51) )) ||
10127 skip "Need MDS version at least 2.15.51"
10129 [ -d $MOUNT/.lustre ] || setup
10131 old_count=$(do_facet mgs "$LCTL --device MGS llog_print -r params" |
10132 grep -c "parameter: timeout")
10134 old_timeout=$($LCTL get_param -n timeout)
10135 stack_trap "cleanup_123ai $old_timeout" EXIT
10137 # add and cancel params (2 * MAX_IOC_BUFLEN / 128 + 1 = 129)
10138 for i in {1..129}; do
10139 do_facet mgs "$LCTL set_param -P timeout=$i" ||
10140 error "fail to set timeout=$i on MGS"
10143 count=$(do_facet mgs "$LCTL --device MGS llog_print -r params" |
10144 grep -c "parameter: timeout")
10146 ((count - old_count == 129)) ||
10147 error "number timeout record missmatch ($count - $old_count != 129)"
10149 do_facet mgs "$LCTL --device MGS llog_print params" |
10150 tail -1 | grep "timeout, value: 129" ||
10151 error "llog_print could not display the last record (timeout=129)"
10154 run_test 123ai "llog_print display all non skipped records"
10157 remote_mgs_nodsh && skip "remote MGS with nodsh"
10159 [ -d $MOUNT/.lustre ] || setup
10160 yaml_file="$TMP/$tfile.yaml"
10162 do_facet mgs rm "$yaml_file"
10163 cfgfiles=$(do_facet mgs "lctl --device MGS llog_catlist" |
10164 sed 's/config_log://')
10166 # set jobid_var to a different value for test
10167 orig_val=$(do_facet mgs $LCTL get_param jobid_var)
10169 do_facet mgs $LCTL set_param -P jobid_var="TESTNAME"
10171 for i in $cfgfiles params; do
10172 do_facet mgs "lctl --device MGS llog_print ${i} >> $yaml_file"
10175 echo "Unmounting FS"
10183 test_123_restore() {
10184 set_val=$(do_facet mgs $LCTL get_param jobid_var)
10186 do_facet mgs $LCTL set_param -P $orig_val
10188 [ $set_val == "jobid_var=TESTNAME" ] ||
10189 error "$set_val is not TESTNAME"
10191 do_facet mgs rm "$yaml_file"
10203 # Reapply the config from before
10204 echo "Setting configuration parameters"
10205 do_facet mgs "lctl set_param -F $yaml_file"
10209 run_test 123F "clear and reset all parameters using set_param -F"
10219 # Reapply the config from before
10220 echo "Setting configuration parameters"
10221 do_facet mgs "lctl apply_yaml $yaml_file"
10225 run_test 123G "clear and reset all parameters using apply_yaml"
10230 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
10231 [ -z $mds2failover_HOST ] && skip "needs MDT failover setup"
10234 do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
10235 grep 1.2.3.4@tcp && error "Should not be fake nid"
10236 do_facet mgs $LCTL conf_param $FSNAME-MDT0001.failover.node=1.2.3.4@tcp\
10237 || error "Set params error"
10238 do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
10239 grep 1.2.3.4@tcp || error "Fake nid should be added"
10243 if combined_mgs_mds; then
10244 start_mdt 1 "-o nosvc" ||
10245 error "starting mds with nosvc option failed"
10247 local nid=$(do_facet mds2 $LCTL list_nids | head -1)
10248 local failover_nid=$(do_node $mds2failover_HOST $LCTL list_nids | head -1)
10250 do_facet mgs $LCTL replace_nids $FSNAME-MDT0001 $nid:$failover_nid ||
10251 error "replace_nids execution error"
10253 if combined_mgs_mds; then
10259 echo "lfs setdirstripe"
10260 $LFS setdirstripe -i 1 $MOUNT/$tdir || error "setdirstirpe error"
10263 run_test 124 "check failover after replace_nids"
10265 get_max_sectors_kb() {
10268 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
10269 local max_sectors_path="/sys/block/${dev_base}/queue/max_sectors_kb"
10271 do_facet ${facet} "[[ -e ${max_sectors_path} ]] &&
10272 cat ${max_sectors_path}"
10275 get_max_hw_sectors_kb() {
10278 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
10279 local max_hw_path="/sys/block/${dev_base}/queue/max_hw_sectors_kb"
10281 do_facet ${facet} "[[ -e ${max_hw_path} ]] && cat ${max_hw_path}"
10284 set_max_sectors_kb() {
10288 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
10289 local max_sectors_path="/sys/block/${dev_base}/queue/max_sectors_kb"
10291 do_facet ${facet} "[[ -e ${max_sectors_path} ]] &&
10292 echo ${value} > ${max_sectors_path}"
10295 [[ $rc -ne 0 ]] && echo "Failed to set ${max_sectors_path} to ${value}"
10300 # Return 0 if all slave devices have max_sectors_kb == max_hw_sectors_kb
10301 # Otherwise return > 0
10302 check_slaves_max_sectors_kb()
10306 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
10307 local slaves_dir=/sys/block/${dev_base}/slaves
10308 local slave_devices=$(do_facet ${facet} "ls ${slaves_dir} 2>/dev/null")
10309 [[ -z ${slave_devices} ]] && return 0
10311 local slave max_sectors new_max_sectors max_hw_sectors path
10314 for slave in ${slave_devices}; do
10315 path="/dev/${slave}"
10316 ! is_blkdev ${facet} ${path} && continue
10317 max_sectors=$(get_max_sectors_kb ${facet} ${path})
10318 max_hw_sectors=$(get_max_hw_sectors_kb ${facet} ${path})
10319 new_max_sectors=${max_hw_sectors}
10320 [[ ${new_max_sectors} -gt ${RQ_SIZE_LIMIT} ]] &&
10321 new_max_sectors=${RQ_SIZE_LIMIT}
10323 if [[ ${max_sectors} -ne ${new_max_sectors} ]]; then
10324 echo "${path} ${max_sectors} ${new_max_sectors}"
10327 check_slaves_max_sectors_kb ${facet} ${path}
10336 local facet_list="mgs mds1 ost1"
10337 combined_mgs_mds && facet_list="mgs ost1"
10340 for facet in ${facet_list}; do
10341 [[ $(facet_fstype ${facet}) != ldiskfs ]] &&
10342 skip "ldiskfs only test" &&
10344 ! is_blkdev ${facet} $(facet_device ${facet}) &&
10345 skip "requires all real devices" &&
10350 # We don't increase IO request size limit past 16MB. See comments in
10351 # lustre/utils/libmount_utils_ldiskfs.c:tune_max_sectors_kb()
10352 RQ_SIZE_LIMIT=$((16 * 1024))
10353 local device old_max_sectors new_max_sectors max_hw_sectors
10354 for facet in ${facet_list}; do
10355 device=$(facet_device ${facet})
10356 old_max_sectors=$(get_max_sectors_kb ${facet} ${device})
10357 max_hw_sectors=$(get_max_hw_sectors_kb ${facet} ${device})
10359 # The expected value after l_tunedisk is executed
10360 new_max_sectors=$old_max_sectors
10361 [[ ${new_max_sectors_kb} -gt ${RQ_SIZE_LIMIT} ]] &&
10362 new_max_sectors_kb=${RQ_SIZE_LIMIT}
10364 # Ensure the current value of max_sectors_kb does not equal
10365 # max_hw_sectors_kb, so we can tell whether l_tunedisk did
10367 set_max_sectors_kb ${facet} ${device} $((new_max_sectors - 1))
10369 # Value before l_tunedisk
10370 local pre_max_sectors=$(get_max_sectors_kb ${facet} ${device})
10371 if [[ ${pre_max_sectors} -ne $((new_max_sectors - 1)) ]]; then
10372 echo "unable to satsify test pre-condition:"
10373 echo "${pre_max_sectors} != $((new_max_sectors - 1))"
10378 echo "Before: ${facet} ${device} ${pre_max_sectors} ${max_hw_sectors}"
10380 do_facet ${facet} "l_tunedisk ${device}"
10382 # Value after l_tunedisk
10383 local post_max_sectors=$(get_max_sectors_kb ${facet} ${device})
10385 echo "After: ${facet} ${device} ${post_max_sectors} ${max_hw_sectors}"
10387 if [[ ${facet} != ost1 ]]; then
10388 if [[ ${post_max_sectors} -ne ${pre_max_sectors} ]]; then
10389 echo "l_tunedisk modified max_sectors_kb of ${facet}"
10393 set_max_sectors_kb ${facet} ${device} ${old_max_sectors}
10395 if [[ ${post_max_sectors} -eq ${pre_max_sectors} ]]; then
10396 echo "l_tunedisk failed to modify max_sectors_kb of ${facet}"
10400 check_slaves_max_sectors_kb ${facet} ${device} ||
10407 run_test 125 "check l_tunedisk only tunes OSTs and their slave devices"
10410 [[ "$MDS1_VERSION" -ge $(version_code 2.13.52) ]] ||
10411 skip "Need MDS version at least 2.13.52"
10414 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ../libcfs/libcfs/libcfs
10415 #define OBD_FAIL_OBD_SETUP 0x60d
10416 do_facet mds1 $LCTL set_param fail_loc=0x60d
10417 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_modules &
10418 for i in {1..40}; do
10419 do_facet mds1 lsmod | grep -q osd_$mds1_FSTYPE && break
10422 clear_failloc $SINGLEMDS 20 &
10423 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
10425 run_test 126 "mount in parallel shouldn't cause a crash"
10428 [[ "$ost1_FSTYPE" == ldiskfs ]] || skip "ldiskfs only test"
10432 zconf_umount_clients $RCLIENTS $MOUNT
10434 wait_osp_active ost ${FSNAME}-OST0000 0 1
10435 local osc_tgt="$FSNAME-OST0000-osc-$($LFS getname -i $DIR)"
10436 local avail1=($($LCTL get_param -n osc.${osc_tgt}.kbytesavail))
10438 $LFS setstripe -i 0 $DIR/$tfile || error "failed creating $DIR/$tfile"
10439 dd if=/dev/zero of=$DIR/$tfile bs=1M oflag=direct || true
10441 local avail2=($($LCTL get_param -n osc.${osc_tgt}.kbytesavail))
10443 if ((avail2 * 100 / avail1 > 1)); then
10446 error "more than 1% space left: before=$avail1 after=$avail2"
10449 local mbs=$(($(stat -c %s $DIR/$tfile) / (1024 * 1024)))
10451 dd if=/dev/zero of=$DIR/$tfile bs=1M count=$mbs conv=notrunc \
10452 oflag=direct || error "overwrite failed"
10454 run_test 127 "direct io overwrite on full ost"
10458 combined_mgs_mds && skip "need separate mgs device"
10459 [ "$ost2_FSTYPE" == zfs ] && import_zpool ost2
10462 # Try to apply nolocallogs to the virgin OST. Should fail.
10463 do_facet ost2 "$TUNEFS --nolocallogs $(ostdevname 2)" &&
10464 error "nolocallogs should not be allowed on the virgin target"
10469 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
10470 # Start OST without MGS (local configs)
10471 do_facet ost1 "$TUNEFS --dryrun $(ostdevname 1)"
10472 start_ost || error "unable to start OST1"
10473 stop_ost || error "Unable to stop OST1"
10475 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
10476 # Do not allow reading local configs, should fail
10477 do_facet ost1 "$TUNEFS --nolocallogs $(ostdevname 1)" ||
10478 error "Can not set nolocallogs"
10479 start_ost && error "OST1 started, but should fail"
10481 # Connect to MGS successfully, reset nolocallogs flag
10482 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
10483 start_mgs || error "unable to start MGS"
10484 start_ost || error "unable to start OST1"
10486 do_facet ost1 "$TUNEFS --dryrun $(ostdevname 1)" | grep "nolocallogs" &&
10487 error "nolocallogs expected to be reset"
10489 stop_ost || error "Unable to stop OST1"
10491 run_test 128 "Force using remote logs with --nolocallogs"
10495 (( MDS1_VERSION >= $(version_code 2.14.57) )) ||
10496 skip "Need MDS version at least 2.14.57"
10498 start_mgsmds || error "MGS/MDS start failed"
10500 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &&
10501 error "start ost1 should fail" || true
10502 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &&
10503 error "second start ost1 should fail" || true
10504 do_facet ost1 "$TUNEFS --writeconf $(ostdevname 1)"
10505 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS ||
10506 error "start ost1 failed"
10510 run_test 129 "attempt to connect an OST with the same index should fail"
10514 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
10516 stop_mdt 2 || error "mdt2 stop failed"
10517 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)"
10518 start_mdt 2 || error "mdt2 start failed"
10519 do_facet mds2 "$LCTL dl" | grep MDT0001-osp-MDT0001 &&
10520 error "Illegal OSP device created" || true
10522 run_test 130 "re-register an MDT after writeconf"
10525 [ "$mds1_FSTYPE" == "ldiskfs" ] || skip "ldiskfs only test"
10526 (( $MDS1_VERSION >= $(version_code 2.14.56.35) )) ||
10527 skip "Need MDS version at least 2.14.56.35"
10528 do_facet mds1 $DEBUGFS -R features $(mdsdevname 1) |
10529 grep -q project || skip "skip project quota not supported"
10534 test_mkdir -c $MDSCOUNT -p $DIR/$tdir
10535 $LFS project -p 1000 $DIR/$tdir || error "set dir project id failed"
10536 createmany -o $DIR/$tdir/f 512
10537 for ((i = 0; i < 512; ++i)); do
10538 $LFS project -p $i $DIR/$tdir/f${i} ||
10539 error "set f${i} project id failed"
10542 test_mkdir -c $MDSCOUNT -p $DIR/$tdir.inherit
10543 $LFS project -p 1001 -s $DIR/$tdir.inherit
10544 createmany -o $DIR/$tdir.inherit/f 128
10545 (( $($LFS project $DIR/$tdir.inherit/f* |
10546 awk '$1 == 1001 { print }' | wc -l) == 128 )) ||
10547 error "files did not inherit projid 1001"
10551 for i in $(seq $MDSCOUNT); do
10552 mds_backup_restore mds$i ||
10553 error "Backup/restore on mds$i failed"
10558 projid=($($LFS project -d $DIR/$tdir))
10559 [ ${projid[0]} == "1000" ] ||
10560 error "projid expected 1000 not ${projid[0]}"
10561 for ((i = 0; i < 512; ++i)); do
10562 projid=($($LFS project $DIR/$tdir/f${i}))
10563 [ ${projid[0]} == "$i" ] ||
10564 error "projid expected $i not ${projid[0]}"
10567 (( $($LFS project $DIR/$tdir.inherit/f* |
10568 awk '$1 == 1001 { print }' | wc -l) == 128 )) ||
10569 error "restore did not copy projid 1001"
10571 run_test 131 "MDT backup restore with project ID"
10577 (( MDS1_VERSION >= $(version_code 2.14.57) )) ||
10578 skip "Need MDS version at least 2.14.57"
10580 combined_mgs_mds || start_mgs || error "unable to start MGS"
10581 start_mdt 1 || error "unable to start mdt1"
10583 err_cnt=$(do_facet mds1 dmesg | grep -c "cannot take the layout locks")
10584 stop_mdt 1 || error "stop mdt1 failed"
10586 [ "$mds1_FSTYPE" == zfs ] && import_zpool mds1
10587 do_facet mds1 $TUNEFS --param mdt.hsm_control=enabled $(mdsdevname 1) ||
10588 error "tunefs failed"
10589 start_mdt 1 || error "cannot start mdt1"
10591 err_cnt2=$(do_facet mds1 dmesg | grep -c "cannot take the layout locks")
10592 [ $err_cnt -eq $err_cnt2 ] || error "Can not take the layout lock"
10593 stop_mdt 1 || error "stop mdt1 failed"
10595 run_test 132 "hsm_actions processed after failover"
10597 # This test verifies we do RR allocation within a pool even if there is a
10598 # significant imbalance vs an OST outside the pool
10600 [[ $OSTCOUNT -lt 4 ]] && skip_env "needs >= 4 OSTs"
10601 [[ "$OST1_VERSION" -ge $(version_code 2.15.51) ]] ||
10602 skip "Need OST version at least 2.15.51"
10603 # This is the easiest way to ensure OSTs start out balanced
10604 reformat_and_config
10607 check_set_fallocate_or_skip
10609 local testfile=$DIR/$tdir/$tfile
10610 local pool="testpool"
10611 local ostrange=$((OSTCOUNT - 1))
10612 # Select all but the last OST to add to the pool
10613 local poolostrange=$((OSTCOUNT - 2))
10621 [ $MINV -eq 0 ] && error "no free space in OST$MINI"
10622 [ $MAXV -gt $((2 * $MINV)) ] &&
10623 error "OSTs badly unbalanced after reformat"
10625 create_pool $FSNAME.$pool || error "failed to create a pool"
10626 do_facet mgs $LCTL pool_add $FSNAME.$pool OST[0-$poolostrange] ||
10627 error "failed to add OST[0-$poolostrange] to the pool"
10629 test_mkdir -p $DIR/$tdir || error "failed to mkdir $DIR/$tdir"
10630 # Consume space on the OSTs in the pool so they are unbalanced with the
10631 # OST outside of the pool
10632 # fill each OST 90% with fallocate so they are widely
10634 local size=$(((MINV * 9 / 10) * 1024))
10635 for ((i = 0; i <= poolostrange; i++)); do
10636 $LFS setstripe -c 1 -i $i $testfile$i ||
10637 error "failed to setstripe $testfile$i"
10638 fallocate -l $size $testfile$i || error "fallocate failed"
10644 # Create files in the pool now that there is an imbalance
10645 filecount=$(((OSTCOUNT - 1) * filenum))
10646 for ((i = 0; i < filecount; i++)); do
10647 $LFS setstripe -p $pool $testfile-$i ||
10648 error "failed to setstripe -p $pool $testfile-$i"
10650 $LFS getstripe -i $testfile-* > /tmp/$tfile.log
10651 # Count the number of files with a stripe on each OST to verify the
10652 # pool allocated with round-robin
10653 for ((i = 0; i <= poolostrange; i++)); do
10654 stripecount=$(grep -c $i /tmp/$tfile.log)
10655 # Allow a little leeway
10656 if (( stripecount < filenum - 1 ||
10657 stripecount > filenum + 1 )); then
10658 cat /tmp/$tfile.log
10659 error "$stripecount != $filenum files on OST$i"
10663 # Create files across the system now that there is an imbalance
10664 filecount=$((OSTCOUNT * filenum))
10665 for ((i = 1; i < filecount; i++)); do
10666 $LFS setstripe $testfile-$i.2 ||
10667 error "failed to setstripe $testilfe-$i.2"
10669 $LFS getstripe -i $testfile-*.2 > /tmp/$tfile.log
10671 # Count the number of files with a stripe on each OST to verify the
10672 # files are *NOT* allocated with round-robin
10673 for ((i = 0; i <= ostrange; i++)); do
10674 stripecount=$(grep -c $i /tmp/$tfile.log)
10675 if [[ $stripecount -ne $filenum ]]; then
10677 echo "QOS: $stripecount != $filenum files on OST$i"
10680 if [ -z "$qos_used" ]; then
10681 error "QOS not used on imbalanced OSTs!"
10684 rm -rf /tmp/$tfile.log $DIR/$tdir
10685 do_facet mgs $LCTL pool_remove $FSNAME.$pool OST[0-$poolostrange] ||
10686 "failed to remove OST[0-$poolostrange] from the pool"
10687 do_facet mgs $LCTL pool_destroy $FSNAME.$pool ||
10688 error "failed to destroy pool"
10690 run_test 133 "stripe QOS: free space balance in a pool"
10693 [ "$mds1_FSTYPE" == "ldiskfs" ] || skip "ldiskfs only test"
10696 local mdt_dev=$(facet_device mds1)
10697 local tmp_dir=$TMP/$tdir
10698 local out=$tmp_dir/check_iam.txt
10699 local CHECK_IAM=${CHECK_IAM:-$(do_facet mds1 "which check_iam 2> /dev/null || true")}
10701 [[ -n "$CHECK_IAM" ]] || skip "check_iam not found"
10704 for ((i=0; i<64; i++)); do
10706 #cmd introduce a random corruption to IAM file
10707 local cmd="dd if=/dev/urandom of=$tmp_dir/$f bs=2 conv=notrunc count=1 seek=$((RANDOM % 36))"
10708 do_facet mds1 "mkdir -p $tmp_dir; \
10709 $DEBUGFS -c -R 'dump $f $tmp_dir/$f' $mdt_dev 2>&1; \
10710 $CHECK_IAM -v $tmp_dir/$f 2>&1; \
10711 echo $cmd; eval $cmd 2>/dev/null;
10712 $CHECK_IAM -v $tmp_dir/$f 2>&1; echo \\\$?" >> $out 2>&1
10717 stack_trap "rm -rf $tmp_dir && do_facet mds1 rm -rf $tmp_dir" EXIT
10719 rc=$(grep -c "fault\|except" $out)
10720 (( rc == 0 )) || { cat $out &&
10721 error "check_iam failed with fault or exception $rc"; }
10723 rc=$(grep -c "^255" $out)
10724 errors=$(grep -c "FINISHED WITH ERRORS" $out)
10726 (( rc == errors )) || { cat $out &&
10727 error "check_iam errcode does not fit with errors $rc $errors"; }
10729 run_test 134 "check_iam works without faults"
10731 cleanup_test_135(){
10734 printf "\nCleanup test_135\n" >&2
10735 do_facet mds1 "$LCTL set_param -n $oldgc"
10736 rm -rf $DIR/$tdir &> /dev/null
10740 __test_135_file_thread() {
10742 local init_time=$(awk '{print $1}' /proc/uptime)
10743 local awkcmd="/crosses index zero/ {if (\$1 > $init_time) exit(1);}"
10745 #Generate a full plain llogs
10746 while dmesg | sed -r 's/(\[|\])//g' | awk "$awkcmd" ; do
10747 createmany -o $DIR/$tdir/f 4500 >&2
10748 createmany -u $DIR/$tdir/f 4500 >&2
10752 __test_135_reader() {
10755 local firstidx=$(changelog_user_rec mds1 $cl_user)
10756 local oldidx=$firstidx
10761 while read -t10 -u$fd newidx other; do
10762 (( (newidx - oldidx) == 1 )) ||
10763 error "changelog jump detected (last: $oldidx, current: $newidx)"
10765 if (( (newidx - firstidx + 1) % 13000 == 0 )); then
10766 [[ $pid -eq 0 ]] ||
10767 wait $pid || error "changelog_clear failed"
10768 changelog_clear $((newidx - 1)) mds1 >&2 & pid=$!
10773 [[ $pid -eq 0 ]] ||
10774 wait $pid || error "changelog_clear failed"
10780 (( MDS1_VERSION >= $(version_code 2.15.52) )) ||
10781 skip "need MDS version at least 2.15.52"
10783 local service=$(facet_svc mds1)
10785 local lastread lastidx
10786 local files_pid reader_pid
10791 # Need to reformat because we are changing llog catalog sizes to 5.
10792 # Otherwise, processing could fail with existing catalogs (last_idx>5).
10796 # Disable changelog garbage colector
10797 local oldgc=$(do_facet mds1 "$LCTL get_param mdd.${service}.changelog_gc")
10798 do_facet mds1 "$LCTL set_param -n mdd.${service}.changelog_gc=0"
10799 stack_trap "cleanup_test_135 $oldgc" EXIT INT
10801 # change the changelog_catalog size to 5 entries for everybody
10802 #define OBD_FAIL_CAT_RECORDS 0x1312
10803 do_node $(comma_list $(all_nodes)) $LCTL set_param fail_loc=0x1312 fail_val=5
10805 # disable console ratelimit
10806 local rl=$(cat /sys/module/libcfs/parameters/libcfs_console_ratelimit)
10807 echo 0 > /sys/module/libcfs/parameters/libcfs_console_ratelimit
10808 stack_trap "echo $rl > /sys/module/libcfs/parameters/libcfs_console_ratelimit" EXIT
10810 test_mkdir -c 1 -i 0 $DIR/$tdir || error "Failed to create directory"
10811 changelog_chmask "ALL" || error "changelog_chmask failed"
10812 changelog_register || error "changelog_register failed"
10814 cl_user="${CL_USERS[mds1]%% *}"
10815 changelog_users mds1 | grep -q $cl_user ||
10816 error "User $cl_user not found in changelog_users"
10818 # Start reader thread
10819 coproc $LFS changelog --follow $service
10822 stack_trap "echo kill changelog reader; kill $reader_pid" EXIT
10824 echo -e "\nWrap arround changelog catalog"
10826 # Start file writer thread
10827 __test_135_file_thread "$service" & files_pid=$!
10828 stack_trap "(pkill -P$files_pid; kill $files_pid) &> /dev/null || true" EXIT
10830 # Check changelog entries
10831 lastread=$(__test_135_reader $fd $cl_user) || exit $?
10832 ! kill -0 $files_pid 2>/dev/null ||
10833 error "creation thread is running. Is changelog reader stuck?"
10835 lastidx=$(changelog_users mds1 | awk '/current_index/ {print $NF}' )
10836 [[ "$lastread" -eq "$lastidx" ]] ||
10837 error "invalid changelog lastidx (read: $lastread, mds: $lastidx)"
10839 run_test 135 "check the behavior when changelog is wrapped around"
10842 do_facet mds2 "$LCTL --device ec cleanup" || true
10843 do_facet mds2 "$LCTL --device ec detach" || true
10846 reformat_and_config
10850 (( MDSCOUNT >= 2 )) || skip "needs >= 2 MDTs"
10855 do_facet mds2 "$LCTL attach echo_client ec ec_uuid" ||
10856 error "echo attach fail"
10858 stack_trap cleanup_136 EXIT
10860 do_facet mds2 "$LCTL --device ec setup lustre-MDT0001 mdt" &&
10861 error "attach to MDT should fail!"
10863 do_facet mds2 "$LCTL --device ec setup lustre-MDT0001 mdd" ||
10864 error "attach to MDD should OK"
10865 do_facet mds2 "$LCTL --device ec test_mkdir /tt" &&
10866 error "mkdir test should fail with remote object"
10870 run_test 136 "don't panic with bad obdecho setup"
10873 (( MDS1_VERSION >= $(version_code 2.15.55) )) ||
10874 skip "need MDS version at least 2.15.55"
10875 (( MDSCOUNT >= 2 )) || skip "needs >= 2 MDTs"
10877 RM_UPDATELOG=$(do_facet mds2 "which remove_updatelog 2> /dev/null")
10878 RM_UPDATELOG=${RM_UPDATELOG:-"$LUSTRE/scripts/remove_updatelog"}
10880 [ -f "$RM_UPDATELOG" ] ||
10881 skip_env "remove_updatelog is not found on mds2"
10883 local mntpt=$(facet_mntpt mds2)
10886 test_mkdir -c $MDSCOUNT -p $DIR/$tdir ||
10887 error "mkdir $DIR/$tdir failed"
10888 mkdir $DIR/$tdir/d{1..256}
10893 mount_fstype mds2 || error "mount as fstype mds2 failed"
10894 do_facet mds2 $RM_UPDATELOG -n $mntpt
10895 MDTS=($(do_facet mds2 $RM_UPDATELOG -n $mntpt |
10896 grep -o "Processing MDT[0-9]*" | awk -F'MDT' '{print $2}'))
10897 (( ${#MDTS[@]} == MDSCOUNT )) ||
10898 error "Processed ${#MDTS[@]} from $MDSCOUNT"
10900 do_facet mds2 $RM_UPDATELOG -n -m 1,0 $mntpt
10901 MDTS=($(do_facet mds2 $RM_UPDATELOG -n -m 1,0 $mntpt |
10902 grep -o "Processing MDT[0-9]*" | awk -F'MDT' '{print $2}'))
10903 (( ${#MDTS[@]} == 2 )) ||
10904 error "Processed ${#MDTS[@]} instead of 2"
10905 (( ${MDTS[0]} == 1 && ${MDTS[1]} == 0 )) ||
10906 error "Processed: ${MDTS[*]}, expected: 1 0"
10908 do_facet mds2 $RM_UPDATELOG -m 0 $mntpt
10909 unmount_fstype mds2
10910 start_mdt 2 || error "mds2 start fail"
10911 start_mdt 1 || error "mds1 start fail"
10912 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds1 FULL
10913 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
10915 rm -rf $DIR/$tdir || error "Can't remove $tdir"
10917 reformat_and_config
10919 run_test 140 "remove_updatelog script actions"
10924 local max_cached_mb=$($LCTL get_param llite.*.max_cached_mb |
10925 awk '/^max_cached_mb/ { print $2 }')
10926 stack_trap "$LCTL set_param -n llite.*.max_cached_mb=$max_cached_mb"
10928 $LCTL set_param llite.*.max_cached_mb='100%'
10930 local new_max_cached_mb=$($LCTL get_param llite.*.max_cached_mb |
10931 awk '/^max_cached_mb/ { print $2 }')
10932 local total_ram_mb=$(free -m | grep 'Mem:' | awk '{print $2}')
10934 $LCTL get_param llite.*.max_cached_mb
10935 echo "total ram mb: $total_ram_mb"
10936 (( new_max_cached_mb == total_ram_mb )) ||
10937 error "setting cache to 100% not equal to total RAM"
10939 $LCTL set_param llite.*.max_cached_mb='50%'
10940 new_max_cached_mb=$($LCTL get_param llite.*.max_cached_mb |
10941 awk '/^max_cached_mb/ { print $2 }')
10943 $LCTL get_param llite.*.max_cached_mb
10944 (( new_max_cached_mb == $((total_ram_mb / 2)) )) ||
10945 error "setting cache to 50% not equal to 50% of RAM"
10947 $LCTL set_param llite.*.max_cached_mb='105%' &&
10948 error "should not be able to set insane value"
10950 $LCTL set_param llite.*.max_cached_mb='0%'
10951 new_max_cached_mb=$($LCTL get_param llite.*.max_cached_mb |
10952 awk '/^max_cached_mb/ { print $2 }')
10953 # Minimum cache size is 64 MiB
10954 $LCTL get_param llite.*.max_cached_mb
10955 (( new_max_cached_mb == 64 )) ||
10956 error "setting cache to 0% != minimum cache size"
10958 run_test 150 "test setting max_cached_mb to a %"
10961 (( MDS1_VERSION >= $(version_code 2.15.58) )) ||
10962 skip "need MDS version at least 2.15.58"
10963 [[ "$ost1_FSTYPE" == ldiskfs ]] || skip "ldiskfs only test"
10966 if ! combined_mgs_mds ; then
10970 echo "Damage ost1 local config log"
10971 do_facet ost1 "$DEBUGFS -w -R 'punch CONFIGS/$FSNAME-OST0000 0 1' \
10972 $(ostdevname 1) || return \$?" ||
10973 error "do_facet ost1 failed with $?"
10975 # expect OST to fail mount with no MGS and bad local config
10976 start_ost && error "OST start should fail"
10978 if ! combined_mgs_mds ; then
10981 start_mds || error "MDS start failed"
10982 # now it should start with MGS config
10983 start_ost || error "OST start failed"
10984 reformat_and_config
10986 run_test 151 "damaged local config doesn't prevent mounting"
10989 (( MDS1_VERSION >= $(version_code 2.15.59.53) )) ||
10990 skip "need MDS >= 2.15.59.53 for sequence allocation retry"
10991 (( MDSCOUNT >= 2 )) || skip "needs >= 2 MDTs"
10992 local tf=$DIR/$tdir/$tfile
10993 local nost=$((OSTCOUNT+1))
10994 local nostdevname=$(ostdevname $nost)
10997 test_mkdir -i 1 -c1 $DIR/$tdir || error "can't mkdir"
11000 add ost$nost $(mkfs_opts ost1 $nostdevname) --index=$nost \
11001 --reformat $nostdevname $(ostvdevname $nost)
11002 [[ -d "$nostdevname" ]] || stack_trap "do_facet mds1 rm -f $nostdevname"
11004 #define OBD_FAIL_OPS_FAIL_SEQ_ALLOC 0x2109
11005 do_facet mds1 $LCTL set_param fail_loc=0x80002109 fail_val=2
11006 echo "START OST$nost"
11007 stack_trap "stop ost$nost"
11008 start ost$nost $nostdevname $OST_MOUNT_OPTS &
11012 $LFS setstripe -c -1 $tf &
11016 log "STOP OST$nost"
11017 # probably mount hasn't completed yet, so stop races with it
11020 jobs -pr | grep -E "^$PID\$" && sleep 0.5 && continue
11025 do_facet mds1 $LCTL set_param fail_loc=0
11026 log "START OST$nost again"
11027 start ost$nost $nostdevname $OST_MOUNT_OPTS ||
11028 error "can't start ost$nost"
11030 $LFS setstripe -c -1 $tf-2 || error "can't touch $tf-2"
11031 $LFS getstripe -v $tf-2
11032 local stripes=$($LFS getstripe -c $tf-2)
11033 (( stripes == $nost )) || error "$tf-2 $stripes != $nost"
11035 run_test 152 "seq allocation error in OSP"
11038 reformat_and_config
11040 start_mds || error "MDS start failed"
11041 start_ost || error "OST start failed"
11043 local nid=$($LCTL list_nids | grep ${NETTYPE} | head -n1)
11044 local net=${nid#*@}
11045 local MGS_NID=$(do_facet mgs $LCTL list_nids | head -1)
11046 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
11047 local FAKE_PNID="192.168.252.112@${net}"
11048 local FAKE_NIDS="${FAKE_PNID},${FAKE_PNID}2"
11049 local FAKE_FAILOVER="10.252.252.113@${net},10.252.252.113@${net}2"
11050 local NIDS_AND_FAILOVER="$FAKE_NIDS:$FAKE_FAILOVER:$OST1_NID:$MGS_NID"
11055 mount -t lustre $NIDS_AND_FAILOVER:/lustre $MOUNT &
11057 while (( period < 30 )); do
11058 [[ -n "$(ps -p $pid -o pid=)" ]] || break
11059 echo "waiting for mount ..."
11061 period=$((period + 5))
11063 $LCTL get_param mgc.MGC${FAKE_PNID}.import | grep "uptodate:"
11064 check_mount || error "check_mount failed"
11066 cleanup || error "cleanup failed with rc $?"
11068 run_test 153a "bypass invalid NIDs quickly"
11071 # (This was sanity/802a)
11073 saved_MGS_MOUNT_OPTS=$MGS_MOUNT_OPTS
11074 saved_MDS_MOUNT_OPTS=$MDS_MOUNT_OPTS
11075 saved_OST_MOUNT_OPTS=$OST_MOUNT_OPTS
11076 saved_MOUNT_OPTS=$MOUNT_OPTS
11080 MGS_MOUNT_OPTS=$saved_MGS_MOUNT_OPTS
11081 MDS_MOUNT_OPTS=$saved_MDS_MOUNT_OPTS
11082 OST_MOUNT_OPTS=$saved_OST_MOUNT_OPTS
11083 MOUNT_OPTS=$saved_MOUNT_OPTS
11088 [[ $mds1_FSTYPE = zfs ]] || skip "ZFS specific test"
11089 [[ $MDS1_VERSION -lt $(version_code 2.9.55) ]] ||
11090 [[ $OST1_VERSION -lt $(version_code 2.9.55) ]] &&
11091 skip "Need server version at least 2.9.55"
11093 [[ $ENABLE_QUOTA ]] && skip "Quota enabled for read-only test"
11095 # Reset before starting
11099 mkdir $DIR/$tdir || error "(1) fail to mkdir"
11101 cp $LUSTRE/tests/test-framework.sh $DIR/$tdir/ ||
11102 error "(2) Fail to copy"
11104 stack_trap cleanup_802a EXIT
11106 # sync by force before remount as readonly
11107 sync; sync_all_data; sleep 3; sync_all_data
11111 MGS_MOUNT_OPTS=$(csa_add "$MGS_MOUNT_OPTS" -o rdonly_dev)
11112 MDS_MOUNT_OPTS=$(csa_add "$MDS_MOUNT_OPTS" -o rdonly_dev)
11113 OST_MOUNT_OPTS=$(csa_add "$OST_MOUNT_OPTS" -o rdonly_dev)
11115 echo "Mount the server as read only"
11116 setupall server_only || error "(3) Fail to start servers"
11118 echo "Mount client without ro should fail"
11119 mount_client $MOUNT &&
11120 error "(4) Mount client without 'ro' should fail"
11122 echo "Mount client with ro should succeed"
11123 MOUNT_OPTS=$(csa_add "$MOUNT_OPTS" -o ro)
11124 mount_client $MOUNT ||
11125 error "(5) Mount client with 'ro' should succeed"
11127 echo "Modify should be refused"
11128 touch $DIR/$tdir/guard && error "(6) Touch should fail under ro mode"
11130 echo "Read should be allowed"
11131 diff $LUSTRE/tests/test-framework.sh $DIR/$tdir/test-framework.sh ||
11132 error "(7) Read should succeed under ro mode"
11134 run_test 802a "simulate readonly device"
11136 if ! combined_mgs_mds ; then
11142 # restore the values of MDSSIZE and OSTSIZE
11143 MDSSIZE=$STORED_MDSSIZE
11144 OSTSIZE=$STORED_OSTSIZE
11147 complete_test $SECONDS
11148 check_and_cleanup_lustre