8 PTLDEBUG=${PTLDEBUG:--1}
9 LUSTRE=${LUSTRE:-$(dirname $0)/..}
10 . $LUSTRE/tests/test-framework.sh
14 # tool to create lustre filesystem images
15 ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT 32newtarball"
17 # bug number for skipped test: LU-11915
18 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 110"
19 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
22 # bug number for skipped tests: LU-9795 (all below)
23 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 84 86 103"
26 if ! combined_mgs_mds; then
27 # bug number for skipped test: LU-11991 LU-11990
28 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 32a 32b 32c 32d 32e 66"
29 # bug number for skipped test: LU-9897 LU-12032
30 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 84 123F"
34 [ "$SLOW" = "no" ] && EXCEPT_SLOW="45 69 106 111"
38 # use small MDS + OST size to speed formatting time
39 # do not use too small MDSSIZE/OSTSIZE, which affect the default journal size
40 # STORED_MDSSIZE is used in test_18
41 STORED_MDSSIZE=$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
138 for num in $(seq $MDSCOUNT); do
139 start_mdt $num "$@" || return 94
141 for num in $(seq $MDSCOUNT); do
142 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds${num} FULL
147 if ! combined_mgs_mds ; then
155 for num in $(seq $MDSCOUNT); do
156 stop_mdt $num || return 97
161 echo "stop mgs service on `facet_active_host mgs`"
162 # These tests all use non-failover stop
163 stop mgs -f || return 97
167 echo "start ost1 service on `facet_active_host ost1`"
168 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS "$@" || return 95
169 wait_clients_import_state ${CLIENTS:-$HOSTNAME} ost1 FULL
173 echo "stop ost1 service on `facet_active_host ost1`"
174 # These tests all use non-failover stop
175 stop ost1 -f || return 98
179 echo "start ost2 service on `facet_active_host ost2`"
180 start ost2 $(ostdevname 2) $OST_MOUNT_OPTS "$@" || return 92
181 wait_clients_import_state ${CLIENTS:-$HOSTNAME} ost2 FULL
185 echo "stop ost2 service on `facet_active_host ost2`"
186 # These tests all use non-failover stop
187 stop ost2 -f || return 93
194 echo "mount $FSNAME ${mountopt:+with opts $mountopt} on $mountpath....."
195 zconf_mount $HOSTNAME $mountpath $mountopt || return 96
201 echo "umount lustre on $mountpath....."
202 zconf_umount $HOSTNAME $mountpath "$@" || return 97
205 manual_umount_client(){
208 echo "manual umount lustre on ${MOUNT}...."
209 do_facet client "umount ${FORCE} $MOUNT"
215 start_mds || error "MDT start failed"
216 start_ost || error "Unable to start OST1"
217 mount_client $MOUNT || error "client start failed"
218 client_up || error "client_up failed"
227 unload_modules_conf () {
228 if combined_mgs_mds || ! local_mode; then
229 unload_modules || return 1
234 stop_ost || return 202
235 stop_mds || return 201
236 unload_modules_conf || return 203
241 [ "x$1" != "x" ] && force='-f'
242 umount_client $MOUNT $force|| return 200
243 cleanup_nocli || return $?
248 echo "umount $MOUNT2 ..."
249 umount $MOUNT2 || true
250 echo "stopping fs2mds ..."
251 stop fs2mds -f || true
252 echo "stopping fs2ost ..."
253 stop fs2ost -f || true
257 do_facet client "cp /etc/passwd $DIR/a" || return 71
258 do_facet client "rm $DIR/a" || return 72
259 # make sure lustre is actually mounted (touch will block,
260 # but grep won't, so do it after)
261 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
262 echo "setup single mount lustre success"
266 do_facet client "touch $DIR/a" || return 71
267 do_facet client "rm $DIR/a" || return 72
268 do_facet client "touch $DIR2/a" || return 73
269 do_facet client "rm $DIR2/a" || return 74
270 echo "setup double mount lustre success"
274 cat /dev/urandom | tr -dc 'a-zA-Z0-9' | fold -w $1 | head -n 1
277 if [ "$ONLY" == "setup" ]; then
282 if [ "$ONLY" == "cleanup" ]; then
289 #create single point mountpoint
295 check_mount || error "check_mount failed"
296 cleanup || error "cleanup failed with $?"
298 run_test 0 "single mount setup"
301 start_mds || error "MDS start failed"
302 start_ost || error "unable to start OST"
303 echo "start ost second time..."
304 start_ost && error "2nd OST start should fail"
305 mount_client $MOUNT || error "client start failed"
306 check_mount || error "check_mount failed"
307 cleanup || error "cleanup failed with $?"
309 run_test 1 "start up ost twice (should return errors)"
312 start_mds || error "MDT start failed"
313 echo "start mds second time.."
314 start_mds && error "2nd MDT start should fail"
315 start_ost || error "OST start failed"
316 mount_client $MOUNT || error "mount_client failed to start client"
317 check_mount || error "check_mount failed"
318 cleanup || error "cleanup failed with $?"
320 run_test 2 "start up mds twice (should return err)"
324 #mount.lustre returns an error if already in mtab
325 mount_client $MOUNT && error "2nd client mount should fail"
326 check_mount || error "check_mount failed"
327 cleanup || error "cleanup failed with $?"
329 run_test 3 "mount client twice (should return err)"
333 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
334 stop_ost || error "Unable to stop OST1"
335 umount_client $MOUNT -f || error "unmount $MOUNT failed"
338 # ok for ost to fail shutdown
339 if [ 202 -ne $eno ] && [ 0 -ne $eno ]; then
340 error "cleanup failed with $?"
343 run_test 4 "force cleanup ost, then cleanup"
345 test_5a() { # was test_5
347 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
348 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
350 stop_mds || error "Unable to stop MDS"
352 # cleanup may return an error from the failed
353 # disconnects; for now I'll consider this successful
354 # if all the modules have unloaded.
358 echo "killing umount"
359 kill -TERM $UMOUNT_PID
360 echo "waiting for umount to finish"
362 if grep " $MOUNT " /proc/mounts; then
363 echo "test 5: /proc/mounts after failed umount"
367 echo "killing umount"
368 kill -TERM $UMOUNT_PID
369 echo "waiting for umount to finish"
371 grep " $MOUNT " /proc/mounts &&
372 error "/proc/mounts after second umount"
375 # manual_mount_client may fail due to umount succeeding above
377 # stop_mds is a no-op here, and should not fail
378 cleanup_nocli || error "cleanup_nocli failed with $?"
379 # df may have lingering entry
381 # mtab may have lingering entry
385 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
387 grep -q $MOUNT" " /etc/mtab || break
388 echo "Waiting /etc/mtab updated ... "
389 WAIT=$(( WAIT + sleep))
391 [ "$WAIT" -eq "$MAX_WAIT" ] &&
392 error "/etc/mtab is not updated in $WAIT secs"
393 echo "/etc/mtab updated in $WAIT secs"
395 run_test 5a "force cleanup mds, then cleanup"
403 grep " $MOUNT " /etc/mtab &&
404 error false "unexpected entry in mtab before mount" && return 10
406 start_ost || error "OST start failed"
407 if ! combined_mgs_mds ; then
408 trap cleanup_5b EXIT ERR
409 start_mds || error "MDS start failed"
413 mount_client $MOUNT && error "mount_client $MOUNT should fail"
414 grep " $MOUNT " /etc/mtab &&
415 error "$MOUNT entry in mtab after failed mount"
417 # stop_mds is a no-op here, and should not fail
418 cleanup_nocli || error "cleanup_nocli failed with $?"
419 if ! combined_mgs_mds ; then
423 run_test 5b "Try to start a client with no MGS (should return errs)"
426 grep " $MOUNT " /etc/mtab &&
427 error false "unexpected entry in mtab before mount" && return 10
429 start_mds || error "MDS start failed"
430 start_ost || error "OST start failed"
431 local oldfs="${FSNAME}"
432 FSNAME="wrong.${FSNAME}"
433 mount_client $MOUNT || :
435 grep " $MOUNT " /etc/mtab &&
436 error "$MOUNT entry in mtab after failed mount"
438 cleanup_nocli || error "cleanup_nocli failed with $?"
440 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
443 grep " $MOUNT " /etc/mtab &&
444 error "unexpected entry in mtab before mount"
446 start_ost || error "OST start failed"
447 start_mds || error "MDS start failed"
448 stop_ost -f || error "Unable to stop OST1"
449 mount_client $MOUNT || error "mount_client $MOUNT failed"
450 umount_client $MOUNT -f || error "umount_client $MOUNT failed"
451 cleanup_nocli || error "cleanup_nocli failed with $?"
452 ! grep " $MOUNT " /etc/mtab ||
453 error "$MOUNT entry in mtab after unmount"
455 run_test 5d "mount with ost down"
458 grep " $MOUNT " /etc/mtab &&
459 error false "unexpected entry in mtab before mount" && return 10
461 start_mds || error "MDS start failed"
462 start_ost || error "OST start failed"
464 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
465 do_facet client "$LCTL set_param fail_loc=0x80000506"
466 mount_client $MOUNT || echo "mount failed (not fatal)"
467 cleanup || error "cleanup failed with $?"
468 grep " $MOUNT " /etc/mtab &&
469 error "$MOUNT entry in mtab after unmount"
472 run_test 5e "delayed connect, don't crash (bug 10268)"
475 combined_mgs_mds && skip "needs separate mgs and mds"
477 grep " $MOUNT " /etc/mtab &&
478 error false "unexpected entry in mtab before mount" && return 10
481 start_ost || error "OST start failed"
482 mount_client $MOUNT &
484 echo client_mount pid is $pid
488 if ! ps -f -p $pid >/dev/null; then
491 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
492 error "mount returns $rc, expected to hang"
494 cleanup || error "cleanup failed with $?"
499 start_mds || error "start MDS failed"
501 # mount should succeed after start mds
503 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
504 cleanup || error "final call to cleanup failed with rc $?"
506 run_test 5f "mds down, cleanup after failed mount (bug 2712)"
510 [ "$CLIENT_VERSION" -lt $(version_code 2.9.53) ] &&
511 skip "automount of debugfs missing before 2.9.53"
512 umount /sys/kernel/debug
513 $LCTL get_param -n devices | egrep -v "error" && \
514 error "lctl can't access debugfs data"
515 grep " debugfs " /etc/mtab || error "debugfs failed to remount"
517 run_test 5g "handle missing debugfs"
523 #define OBD_FAIL_MDS_FS_SETUP 0x135
524 do_facet mds1 "$LCTL set_param fail_loc=0x80000135"
525 start_mdt 1 && error "start mdt should fail"
526 start_mdt 1 || error "start mdt failed"
527 client_up || error "client_up failed"
530 run_test 5h "start mdt failure at mdt_fs_setup()"
533 (( $MDS1_VERSION >= $(version_code 2.12.54) )) ||
534 skip "Need MDS version at least 2.12.54"
539 #define OBD_FAIL_QUOTA_INIT 0xA05
540 do_facet mds1 "$LCTL set_param fail_loc=0x80000A05"
541 start_mdt 1 && error "start mdt should fail"
542 start_mdt 1 || error "start mdt failed"
543 client_up || error "client_up failed"
546 run_test 5i "start mdt failure at mdt_quota_init()"
551 mount_client $MOUNT || error "mount_client $MOUNT failed"
552 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
553 cleanup || error "cleanup failed with rc $?"
555 run_test 6 "manual umount, then mount again"
560 cleanup_nocli || error "cleanup_nocli failed with $?"
562 run_test 7 "manual umount, then cleanup"
566 mount_client $MOUNT2 || error "mount_client $MOUNT2 failed"
567 check_mount2 || error "check_mount2 failed"
568 umount_client $MOUNT2 || error "umount_client $MOUNT2 failed"
569 cleanup || error "cleanup failed with rc $?"
571 run_test 8 "double mount setup"
574 start_ost || error "OST start failed"
576 do_facet ost1 $LCTL set_param debug=\'inode trace\' ||
577 error "do_facet ost1 set_param inode trace failed."
578 do_facet ost1 $LCTL set_param subsystem_debug=\'mds ost\' ||
579 error "do_facet ost1 set_param debug mds ost failed."
581 CHECK_PTLDEBUG="`do_facet ost1 $LCTL get_param -n debug`"
582 if [ "$CHECK_PTLDEBUG" ] && { \
583 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
584 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
585 echo "lnet.debug success"
587 error "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
589 CHECK_SUBSYS="`do_facet ost1 $LCTL get_param -n subsystem_debug`"
590 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
591 echo "lnet.subsystem_debug success"
593 error "lnet.subsystem_debug: want 'mds ost' got '$CHECK_SUBSYS'"
595 stop_ost || error "Unable to stop OST1"
597 run_test 9 "test ptldebug and subsystem for mkfs"
602 if ! combined_mgs_mds; then
603 files=$(do_facet mgs "find /{proc/fs,sys/fs,sys/kernel/debug}/lustre -type l -exec test ! -e {} \; -print")
604 [ -z $files ] || echo "MGS $files is a broken symlink"
607 files=$(do_facet mds1 "find /{proc/fs,sys/fs,sys/kernel/debug}/lustre -type l -exec test ! -e {} \; -print")
608 [ -z $files ] || echo "MDS $files is a broken symlink"
610 files=$(do_facet ost1 "find /{proc/fs,sys/fs,sys/kernel/debug}/lustre -type l -exec test ! -e {} \; -print")
611 [ -z $files ] || echo "OSS $files is a broken symlink"
613 files=$(do_facet client "find /{proc/fs,sys/fs,sys/kernel/debug}/lustre -type l -exec test ! -e {} \; -print")
614 [ -z $files ] || echo "clients $files is a broken symlink"
616 cleanup || error "cleanup failed with rc $?"
618 run_test 10a "find lctl param broken symlinks"
621 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
622 # But with new MDS stack we don't care about the mode of local objects
623 # anymore, so this test is removed. See bug 22944 for more details.
627 if [ "$mds1_FSTYPE" != ldiskfs ]; then
628 skip "ldiskfs only test"
632 check_mount || error "check_mount failed"
633 cleanup || error "cleanup failed with rc $?"
635 echo "Remove mds config log"
636 if ! combined_mgs_mds ; then
640 do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' \
641 $(mgsdevname) || return \$?" ||
642 error "do_facet mgs failed with $?"
644 if ! combined_mgs_mds ; then
648 start_ost || error "OST start failed"
649 start_mds && error "MDS start succeeded, but should fail"
652 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
655 if [ "$mds1_FSTYPE" != ldiskfs ]; then
656 skip "ldiskfs only test"
659 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
664 # check if current MDSSIZE is large enough
665 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE &&
666 log "use MDSSIZE=$MDSSIZE"
668 # check if the global config has a large enough MDSSIZE
669 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] &&
670 [ $STORED_MDSSIZE -ge $MIN ] &&
671 OK=1 && myMDSSIZE=$STORED_MDSSIZE &&
672 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
674 # check if the block device is large enough
675 is_blkdev $SINGLEMDS $MDSDEV $MIN
676 local large_enough=$?
677 if [ -n "$OK" ]; then
678 [ $large_enough -ne 0 ] && OK=""
680 [ $large_enough -eq 0 ] && OK=1 && myMDSSIZE=$MIN &&
681 log "use device $MDSDEV with MIN=$MIN"
684 # check if a loopback device has enough space for fs metadata (5%)
686 if [ -z "$OK" ]; then
687 local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! \
688 -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
689 awk '($1 != "Filesystem") { print $4 }')
690 ! [ -z "$SPACE" ] && [ $SPACE -gt $((MIN / 20)) ] &&
691 OK=1 && myMDSSIZE=$MIN &&
692 log "use file $MDSDEV with MIN=$MIN"
695 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS"
697 echo "mount mds with large journal..."
699 local OLD_MDSSIZE=$MDSSIZE
703 echo "mount lustre system..."
705 check_mount || error "check_mount failed"
707 echo "check journal size..."
708 local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
709 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
710 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
712 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
715 cleanup || error "cleanup failed with rc $?"
720 run_test 18 "check mkfs creates large journals"
723 start_mds || error "MDS start failed"
724 stop_mds || error "Unable to stop MDS"
726 run_test 19a "start/stop MDS without OSTs"
729 start_ost || error "Unable to start OST1"
730 stop_ost -f || error "Unable to stop OST1"
732 run_test 19b "start/stop OSTs without MDS"
735 # first format the ost/mdt
736 start_mds || error "MDS start failed"
737 start_ost || error "Unable to start OST1"
738 mount_client $MOUNT || error "mount_client $MOUNT failed"
739 check_mount || error "check_mount failed"
740 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed."
741 mount_client $MOUNT remount,ro || error "remount client with ro failed"
742 touch $DIR/$tfile && error "$DIR/$tfile created incorrectly"
743 [ -e $DIR/$tfile ] && error "$DIR/$tfile exists incorrectly"
744 mount_client $MOUNT remount,rw || error "remount client with rw failed"
745 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
746 MCNT=$(grep -c $MOUNT' ' /etc/mtab)
747 [ "$MCNT" -ne 1 ] && error "$MOUNT in /etc/mtab $MCNT times"
749 stop_mds || error "Unable to stop MDS"
750 stop_ost || error "Unable to stop OST1"
752 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
755 start_mds || error "MDS start failed"
756 start_ost || error "unable to start OST1"
757 wait_osc_import_state mds ost FULL
758 stop_ost || error "unable to stop OST1"
759 stop_mds || error "unable to stop MDS"
761 run_test 21a "start mds before ost, stop ost first"
764 start_ost || error "unable to start OST1"
765 start_mds || error "MDS start failed"
766 wait_osc_import_state mds ost FULL
767 stop_mds || error "unable to stop MDS"
768 stop_ost || error "unable to stop OST1"
770 run_test 21b "start ost before mds, stop mds first"
773 start_ost || error "Unable to start OST1"
774 start_mds || error "MDS start failed"
775 start_ost2 || error "Unable to start OST2"
776 wait_osc_import_state mds ost2 FULL
777 stop_ost || error "Unable to stop OST1"
778 stop_ost2 || error "Unable to stop OST2"
779 stop_mds || error "Unable to stop MDS"
780 #writeconf to remove all ost2 traces for subsequent tests
781 writeconf_or_reformat
783 run_test 21c "start mds between two osts, stop mds last"
786 combined_mgs_mds && skip "need separate mgs device"
791 start_mgs || error "unable to start MGS"
792 start_ost || error "unable to start OST1"
793 start_ost2 || error "unable to start OST2"
794 start_mds || error "MDS start failed"
795 wait_osc_import_state mds ost2 FULL
797 local zkeeper=${KEEP_ZPOOL}
798 stack_trap "KEEP_ZPOOL=$zkeeper" EXIT
801 stop_ost || error "Unable to stop OST1"
802 stop_ost2 || error "Unable to stop OST2"
803 stop_mds || error "Unable to stop MDS"
805 #writeconf to remove all ost2 traces for subsequent tests
806 writeconf_or_reformat
807 KEEP_ZPOOL="${zkeeper}"
809 start_mgs || error "unable to start MGS"
811 run_test 21d "start mgs then ost and then mds"
814 MGSNID="$saved_mgsnid"
816 echo "stopping fs2mgs ..."
817 stop $fs2mgs -f || true
820 test_21e() { # LU-5863
821 if [[ -z "$fs3ost_DEV" || -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
822 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
823 skip_env "mixed loopback and real device not working"
826 local fs2mdsdev=$(mdsdevname 1_2)
827 local fs2ostdev=$(ostdevname 1_2)
828 local fs3ostdev=$(ostdevname 2_2)
830 local fs2mdsvdev=$(mdsvdevname 1_2)
831 local fs2ostvdev=$(ostvdevname 1_2)
832 local fs3ostvdev=$(ostvdevname 2_2)
834 # temporarily use fs3ost as fs2mgs
836 local fs2mgsdev=$fs3ostdev
837 local fs2mgsvdev=$fs3ostvdev
839 local fsname=test1234
841 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
842 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
843 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS && trap cleanup_21e EXIT INT ||
844 error "start fs2mgs failed"
846 local saved_mgsnid="$MGSNID"
847 MGSNID=$(do_facet $fs2mgs $LCTL list_nids | xargs | tr ' ' ,)
849 add fs2mds $(mkfs_opts mds1 $fs2mdsdev $fsname) \
850 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
851 add fs2ost $(mkfs_opts ost1 $fs2ostdev $fsname) \
852 --reformat $fs2ostdev $fs2ostvdev || error "add fs2ost failed"
854 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
855 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS || error "start fs2mds failed"
857 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
858 $MOUNT_CMD $MGSNID:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
859 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
863 run_test 21e "separate MGS and MDS"
866 start_mds || error "MDS start failed"
868 echo "Client mount with ost in logs, but none running"
869 start_ost || error "unable to start OST1"
870 # wait until mds connected to ost and open client connection
871 wait_osc_import_state mds ost FULL
872 stop_ost || error "unable to stop OST1"
873 mount_client $MOUNT || error "mount_client $MOUNT failed"
874 # check_mount will block trying to contact ost
875 mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed"
876 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed"
877 umount_client $MOUNT -f
880 echo "Client mount with a running ost"
881 start_ost || error "unable to start OST1"
883 # if gss enabled, wait full time to let connection from
884 # mds to ost be established, due to the mismatch between
885 # initial connect timeout and gss context negotiation timeout.
886 # This perhaps could be remove after AT landed.
887 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
888 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
890 mount_client $MOUNT || error "mount_client $MOUNT failed"
891 wait_osc_import_state mds ost FULL
892 wait_osc_import_ready client ost
893 check_mount || error "check_mount failed"
896 cleanup || error "cleanup failed with rc $?"
898 run_test 22 "start a client before osts (should return errs)"
900 test_23a() { # was test_23
903 stop $SINGLEMDS || error "failed to stop $SINGLEMDS"
904 # force down client so that recovering mds waits for reconnect
905 local running=$(grep -c $MOUNT /proc/mounts) || true
906 if [ $running -ne 0 ]; then
907 echo "Stopping client $MOUNT (opts: -f)"
911 # enter recovery on failed mds
912 local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/})
913 start $SINGLEMDS $MDT_DEV $MDS_MOUNT_OPTS || error "MDS start failed"
914 # try to start a new client
915 mount_client $MOUNT &
917 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
918 MOUNT_LUSTRE_PID=$(ps -ef | grep mount.lustre |
919 grep -v grep | awk '{print $2}')
920 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
922 ps --ppid $MOUNT_LUSTRE_PID
923 echo "waiting for mount to finish"
925 # "ctrl-c" sends SIGINT but it usually (in script) does not work on child process
926 # SIGTERM works but it does not spread to offspring processses
927 kill -s TERM $MOUNT_PID
928 kill -s TERM $MOUNT_LUSTRE_PID
929 # we can not wait $MOUNT_PID because it is not a child of this shell
935 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
937 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
938 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
941 [ -z "$PID1" -a -z "$PID2" ] && break
942 echo "waiting for mount to finish ... "
943 WAIT=$(( WAIT + sleep))
945 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
946 error "MOUNT_PID $MOUNT_PID and "\
947 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
950 cleanup || error "cleanup failed with rc $?"
952 run_test 23a "interrupt client during recovery mount delay"
954 test_23b() { # was test_23
955 start_mds || error "MDS start failed"
956 start_ost || error "Unable to start OST1"
957 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
958 $LCTL set_param fail_loc=0x80000313
960 cleanup || error "cleanup failed with rc $?"
962 run_test 23b "Simulate -EINTR during mount"
965 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
967 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
968 is_blkdev $SINGLEMDS $MDSDEV &&
969 skip_env "mixed loopback and real device not working"
972 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
974 local fs2mdsdev=$(mdsdevname 1_2)
975 local fs2ostdev=$(ostdevname 1_2)
976 local fs2mdsvdev=$(mdsvdevname 1_2)
977 local fs2ostvdev=$(ostvdevname 1_2)
980 # LU-9733 test fsname started with numbers as well
981 local FSNAME2=969362ae
983 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --nomgs --mgsnode=$MGSNID \
984 --fsname=${FSNAME2} --reformat $fs2mdsdev $fs2mdsvdev || exit 10
986 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --fsname=${FSNAME2} \
987 --reformat $fs2ostdev $fs2ostvdev || exit 10
990 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
991 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
992 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
993 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
995 # LU-9733 test fsname started with numbers
996 cl_user=$(do_facet $SINGLEMDS lctl --device $FSNAME2-MDT0000 \
997 changelog_register -n) ||
998 error "register changelog failed"
1000 do_facet $SINGLEMDS lctl --device $FSNAME2-MDT0000 \
1001 changelog_deregister $cl_user ||
1002 error "deregister changelog failed"
1004 check_mount || error "check_mount failed"
1005 # files written on 1 should not show up on 2
1006 cp /etc/passwd $DIR/$tfile
1008 [ -e $MOUNT2/$tfile ] && error "File bleed"
1011 cp /etc/passwd $MOUNT2/$tfile ||
1012 error "cp /etc/passwd $MOUNT2/$tfile failed"
1013 rm $MOUNT2/$tfile || error "remove $MOUNT2/$tfile failed"
1014 # 2 is actually mounted
1015 grep $MOUNT2' ' /proc/mounts > /dev/null || error "$MOUNT2 not mounted"
1017 facet_failover fs2mds
1018 facet_failover fs2ost
1020 umount_client $MOUNT
1021 # the MDS must remain up until last MDT
1023 MDS=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
1024 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
1025 [ -z "$MDS" ] && error "No MDT"
1027 cleanup_nocli || error "cleanup_nocli failed with rc $?"
1029 run_test 24a "Multiple MDTs on a single node"
1032 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1034 skip "needs combined MGT and MDT device"
1036 if [ -z "$fs2mds_DEV" ]; then
1037 local dev=${SINGLEMDS}_dev
1038 local MDSDEV=${!dev}
1039 is_blkdev $SINGLEMDS $MDSDEV &&
1040 skip_env "mixed loopback and real device not working"
1043 local fs2mdsdev=$(mdsdevname 1_2)
1044 local fs2mdsvdev=$(mdsvdevname 1_2)
1046 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --mgs --fsname=${FSNAME}2 \
1047 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
1049 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && {
1051 error "start MDS should fail"
1054 cleanup || error "cleanup failed with rc $?"
1056 run_test 24b "Multiple MGSs on a single node (should return err)"
1060 check_mount || error "check_mount failed"
1061 local MODULES=$($LCTL modules | awk '{ print $2 }')
1062 rmmod $MODULES 2>/dev/null || true
1063 cleanup || error "cleanup failed with $?"
1065 run_test 25 "Verify modules are referenced"
1069 # we need modules before mount for sysctl, so make sure...
1070 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
1071 #define OBD_FAIL_MDS_FS_SETUP 0x135
1072 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000135"
1073 start_mds && error "MDS started but should not have started"
1074 $LCTL get_param -n devices
1075 DEVS=$($LCTL get_param -n devices | egrep -v MG | wc -l)
1076 [ $DEVS -gt 0 ] && error "number of devices is $DEVS, should be zero"
1077 # start mds to drop writeconf setting
1078 start_mds || error "Unable to start MDS"
1079 stop_mds || error "Unable to stop MDS"
1080 unload_modules_conf || error "unload_modules_conf failed with $?"
1082 run_test 26 "MDT startup failure cleans LOV (should return errs)"
1086 start_ost || error "Unable to start OST1"
1087 start_mds || error "Unable to start MDS"
1088 echo "Requeue thread should have started: "
1089 ps -e | grep ll_cfg_requeue
1090 set_persistent_param_and_check ost1 \
1091 "obdfilter.$FSNAME-OST0000.client_cache_seconds" \
1092 "$FSNAME-OST0000.ost.client_cache_seconds"
1093 cleanup_nocli || error "cleanup_nocli failed with rc $?"
1095 run_test 27a "Reacquire MGS lock if OST started first"
1100 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
1101 awk '($3 ~ "mdt" && $4 ~ "MDT0000") { print $4 }')
1103 facet_failover $SINGLEMDS
1104 set_persistent_param_and_check $SINGLEMDS \
1105 "mdt.$device.identity_acquire_expire" \
1106 "$device.mdt.identity_acquire_expire"
1107 set_persistent_param_and_check client \
1108 "mdc.$device-mdc-*.max_rpcs_in_flight" \
1109 "$device.mdc.max_rpcs_in_flight"
1111 cleanup || error "cleanup failed with $?"
1113 run_test 27b "Reacquire MGS lock after failover"
1115 test_28A() { # was test_28
1118 local TEST="llite.$FSNAME-*.max_read_ahead_whole_mb"
1119 local PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
1120 local orig=$($LCTL get_param -n $TEST)
1121 local max=$($LCTL get_param -n \
1122 llite.$FSNAME-*.max_read_ahead_per_file_mb)
1124 orig=${orig%%.[0-9]*}
1126 echo "ORIG:$orig MAX:$max"
1127 [[ $max -le $orig ]] && orig=$((max - 3))
1128 echo "ORIG:$orig MAX:$max"
1130 local final=$((orig + 1))
1132 set_persistent_param_and_check client "$TEST" "$PARAM" $final
1133 final=$((final + 1))
1134 set_persistent_param_and_check client "$TEST" "$PARAM" $final
1135 umount_client $MOUNT || error "umount_client $MOUNT failed"
1136 mount_client $MOUNT || error "mount_client $MOUNT failed"
1138 local result=$($LCTL get_param -n $TEST)
1140 if [ $result -ne $final ]; then
1141 error "New config not seen: wanted $final got $result"
1143 echo "New config success: got $result"
1145 set_persistent_param_and_check client "$TEST" "$PARAM" $orig
1146 cleanup || error "cleanup failed with rc $?"
1148 run_test 28A "permanent parameter setting"
1150 test_28a() { # LU-4221
1151 [[ "$OST1_VERSION" -ge $(version_code 2.5.52) ]] ||
1152 skip "Need OST version at least 2.5.52"
1153 [ "$ost1_FSTYPE" = zfs ] &&
1154 skip "LU-4221: no such proc params for ZFS OSTs"
1161 local device="$FSNAME-OST0000"
1165 # In this test we will set three kinds of proc parameters with
1166 # lctl set_param -P or lctl conf_param:
1167 # 1. non-symlink ones in the OFD
1168 # 2. non-symlink ones in the OSD
1171 # prepare a non-symlink parameter in the OFD
1172 name="client_cache_seconds"
1173 param="$device.ost.$name"
1174 cmd="obdfilter.$device.$name"
1176 # permanently setting the parameter in the OFD
1177 old=$(do_facet ost1 $LCTL get_param -n $cmd)
1179 set_persistent_param_and_check ost1 "$cmd" "$param" $new
1180 set_persistent_param_and_check ost1 "$cmd" "$param" $old
1183 # prepare a non-symlink parameter in the OSD
1185 param="$device.osd.$name"
1186 cmd="osd-*.$device.$name"
1188 # conf_param the parameter in the OSD
1189 old=$(do_facet ost1 $LCTL get_param -n $cmd)
1190 new=$(((old + 1) % 2))
1191 set_persistent_param_and_check ost1 "$cmd" "$param" $new
1192 set_persistent_param_and_check ost1 "$cmd" "$param" $old
1194 cleanup || error "cleanup failed with $?"
1196 run_test 28a "set symlink parameters permanently with lctl"
1199 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >= 2 OSTs"
1200 setup_noconfig > /dev/null 2>&1
1201 start_ost2 || error "Unable to start OST2"
1204 local PARAM="$FSNAME-OST0001.osc.active"
1205 # With lctl set_param -P the value $PROC_ACT will be sent to
1206 # all nodes. The [!M] filter out the ability to set active
1207 # on the MDS servers which is tested with wait_osp_* below.
1208 # For ost_server_uuid that only exist on client so filtering
1210 local PROC_ACT="os[cp].$FSNAME-OST0001-osc-[^M]*.active"
1211 local PROC_UUID="os[cp].$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
1213 ACTV=$($LCTL get_param -n $PROC_ACT)
1215 set_persistent_param_and_check client $PROC_ACT $PARAM $DEAC
1216 # also check ost_server_uuid status
1217 RESULT=$($LCTL get_param -n $PROC_UUID | grep DEACTIV)
1218 if [ -z "$RESULT" ]; then
1219 error "Client not deactivated: $($LCTL get_param \
1222 echo "Live client success: got $RESULT"
1226 wait_osp_active ost ${FSNAME}-OST0001 1 0
1228 # test new client starts deactivated
1229 umount_client $MOUNT || error "umount_client $MOUNT failed"
1230 mount_client $MOUNT || error "mount_client $MOUNT failed"
1232 # the 2nd and 3rd field of ost_server_uuid do not update at the same
1233 # time when using lctl set_param -P
1234 wait_update_facet client \
1235 "$LCTL get_param -n $PROC_UUID | awk '{print \\\$3 }'" \
1237 error "New client start active: $($LCTL get_param -n $PROC_UUID)"
1239 echo "New client success: got '$($LCTL get_param -n $PROC_UUID)'"
1241 # make sure it reactivates
1242 set_persistent_param_and_check client $PROC_ACT $PARAM $ACTV
1244 umount_client $MOUNT
1245 stop_ost2 || error "Unable to stop OST2"
1246 cleanup_nocli || error "cleanup_nocli failed with $?"
1247 #writeconf to remove all ost2 traces for subsequent tests
1248 writeconf_or_reformat
1250 run_test 29 "permanently remove an OST"
1255 echo Big config llog
1256 local path="llite.$FSNAME-*"
1257 local cpath="$FSNAME.llite"
1258 local param="max_read_ahead_whole_mb"
1260 local test="${path}.$param"
1261 local conf="${cpath}.$param"
1263 local orig=$($LCTL get_param -n $test)
1264 local list=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1265 for i in ${list[@]}; do
1266 set_persistent_param_and_check client $test $conf $i
1268 # make sure client restart still works
1269 umount_client $MOUNT
1270 mount_client $MOUNT || error "mount_client $MOUNT failed"
1271 [ "$($LCTL get_param -n $test)" -ne "$i" ] &&
1272 error "Param didn't stick across restart $($test) != $i"
1275 echo Erase parameter setting
1276 if [[ $PERM_CMD == *"set_param -P"* ]]; then
1277 do_facet mgs "$PERM_CMD -d $test" ||
1278 error "Erase param $test failed"
1280 do_facet mgs "$PERM_CMD -d $conf" ||
1281 error "Erase param $conf failed"
1283 umount_client $MOUNT
1284 mount_client $MOUNT || error "mount_client $MOUNT failed"
1285 local final=$($LCTL get_param -n $test)
1286 echo "deleted (default) value=$final, orig=$orig"
1287 orig=${orig%%.[0-9]*}
1288 final=${final%%.[0-9]*}
1289 # assumes this parameter started at the default value
1290 [ "$final" -eq "$orig" ] ||
1291 error "Deleted value=$final -ne orig=$orig"
1293 cleanup || error "cleanup failed with rc $?"
1295 run_test 30a "Big config llog and permanent parameter deletion"
1300 local orignids=$($LCTL get_param -n \
1301 osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1303 local orignidcount=$(echo "$orignids" | wc -w)
1305 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1306 # numerical part of it. Hopefully that's not already a failover address
1308 local OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | \
1310 local ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1311 local NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1312 local NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1313 echo "Using fake nid $NEW"
1315 local TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1316 grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1317 if [[ $PERM_CMD == *"set_param -P"* ]]; then
1318 PARAM="osc.$FSNAME-OST0000-osc-[^M]*.import"
1319 echo "Setting $PARAM from $TEST to $NEW"
1320 do_facet mgs "$PERM_CMD $PARAM='connection=$NEW'" ||
1321 error "$PERM_CMD $PARAM failed"
1323 PARAM="$FSNAME-OST0000.failover.node"
1324 echo "Setting $PARAM from $TEST to $NEW"
1325 do_facet mgs "$PERM_CMD $PARAM='$NEW'" ||
1326 error "$PARAM $PARAM failed"
1328 wait_update_facet client "$TEST" "$NEW" ||
1329 error "check $PARAM failed!"
1331 local NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1333 local NIDCOUNT=$(echo "$NIDS" | wc -w)
1334 echo "should have $((orignidcount + 1)) entries \
1335 in failover nids string, have $NIDCOUNT"
1336 [ $NIDCOUNT -eq $((orignidcount + 1)) ] ||
1337 error "Failover nid not added"
1339 if [[ $PERM_CMD == *"set_param -P"* ]]; then
1340 do_facet mgs "$PERM_CMD -d osc.$FSNAME-OST0000-osc-*.import"
1342 do_facet mgs "$PERM_CMD -d $FSNAME-OST0000.failover.node" ||
1343 error "$PERM_CMD delete failed"
1345 umount_client $MOUNT
1346 mount_client $MOUNT || error "mount_client $MOUNT failed"
1348 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1350 NIDCOUNT=$(echo "$NIDS" | wc -w)
1351 echo "only $orignidcount final entries should remain \
1352 in failover nids string, have $NIDCOUNT"
1353 [ $NIDCOUNT -eq $orignidcount ] || error "Failover nids not removed"
1355 cleanup || error "cleanup failed with rc $?"
1357 run_test 30b "Remove failover nids"
1359 test_31() { # bug 10734
1360 # ipaddr must not exist
1361 $MOUNT_CMD 4.3.2.1@tcp:/lustre $MOUNT || true
1362 cleanup || error "cleanup failed with rc $?"
1364 run_test 31 "Connect to non-existent node (shouldn't crash)"
1368 T32_BLIMIT=40960 # Kbytes
1371 T32_PROLIMIT=$((T32_BLIMIT/10))
1373 # This is not really a test but a tool to create new disk
1374 # image tarballs for the upgrade tests.
1376 # Disk image tarballs should be created on single-node
1377 # clusters by running this test with default configurations
1378 # plus a few mandatory environment settings that are verified
1379 # at the beginning of the test.
1381 test_32newtarball() {
1385 local tmp=$TMP/t32_image_create
1399 if [ $FSNAME != t32fs -o \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o \
1400 $OSTCOUNT -ne 2 -o -z "$OSTDEV1" ]; then
1401 error "Needs FSNAME=t32fs MDSCOUNT=2 " \
1402 "MDSDEV1=<nonexistent_file> " \
1403 "MDSDEV2=<nonexistent_file> " \
1404 "(or MDSDEV, in the case of b1_8) " \
1405 "OSTCOUNT=2 OSTDEV1=<nonexistent_file> " \
1406 "OSTDEV2=<nonexistent_file>"
1410 echo "Found stale $tmp"
1414 mkdir $tmp/src || return 1
1415 tar cf - -C $src . | tar xf - -C $tmp/src
1417 # format ost with comma-separated NIDs to verify LU-4460
1418 local failnid="$(h2nettype 1.2.3.4),$(h2nettype 4.3.2.1)"
1419 MGSNID="$MGSNID,$MGSNID" OSTOPT="--failnode=$failnid" formatall
1423 [[ "$MDS1_VERSION" -ge $(version_code 2.3.50) ]] ||
1424 $LFS quotacheck -ug /mnt/$FSNAME
1425 $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
1428 if [[ $MDSCOUNT -ge 2 ]]; then
1429 remote_dir=/mnt/$FSNAME/remote_dir
1430 $LFS mkdir -i 1 $remote_dir
1431 tar cf - -C $tmp/src . | tar xf - -C $remote_dir
1433 target_dir=$remote_dir
1434 if [[ $MDS1_VERSION -ge $(version_code 2.7.0) ]]; then
1435 striped_dir=/mnt/$FSNAME/striped_dir_old
1436 $LFS mkdir -i 1 -c 2 $striped_dir
1437 tar cf - -C $tmp/src . | tar xf - -C $striped_dir
1440 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1441 target_dir=/mnt/$FSNAME
1444 # add project quota #
1445 [[ $mds1_FSTYPE == "ldiskfs" &&
1446 $MDS1_VERSION -gt $(version_code 2.9.55) ]] ||
1447 [[ $mds1_FSTYPE == "zfs" &&
1448 $MDS1_VERSION -gt $(version_code 2.10.53) ]] && {
1449 pj_quota_dir=$target_dir/project_quota_dir
1450 pj_quota_file_old=$pj_quota_dir/pj_quota_file_old
1452 enable_project_quota
1453 set_mdt_qtype ugp ||
1454 error "enable mdt quota failed"
1456 set_ost_qtype ugp ||
1457 error "enable ost quota failed"
1459 mkdir -p $pj_quota_dir
1460 $LFS setquota -p $T32_PRJID -b 0 -B $T32_PROLIMIT -i 0 \
1461 -I $T32_ILIMIT $pj_quota_dir ||
1462 error "setquota -p $T32_PRJID failed"
1464 $LFS setstripe $pj_quota_file_old -c 1 ||
1465 error "setstripe $pj_quota_file_old failed"
1467 chown $T32_QID:$T32_QID $pj_quota_file_old ||
1468 error "chown $pj_quota_file_old failed"
1470 change_project -p $T32_PRJID $pj_quota_file_old
1472 mkdir -p $tmp/src/project_quota_dir
1473 cp $pj_quota_file_old $tmp/src/project_quota_dir/
1476 #####################
1477 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1479 #if [[ $MDSCOUNT -ge 2 ]]; then
1480 # remote_dir=/mnt/$FSNAME/remote_dir
1481 # $LFS mkdir -i 1 $remote_dir
1482 # tar cf - -C $tmp/src . | tar xf - -C $remote_dir
1484 # if [[ "$MDS1_VERSION" -ge $(version_code 2.7.0) ]]; then
1485 # striped_dir=/mnt/$FSNAME/striped_dir_old
1486 # $LFS mkdir -i 1 -c 2 $striped_dir
1487 # tar cf - -C $tmp/src . | tar xf - -C $striped_dir
1492 if [[ $MDS1_VERSION -ge $(version_code 2.9.51) ]]; then
1493 pfl_dir=$target_dir/pfl_dir
1494 pfl_file=$pfl_dir/pfl_file
1496 $LFS setstripe -E 2M -c 1 -o 0 -E -1 -S 2M -c 1 -o 1 \
1498 error "Create PFL file failed"
1500 dd if=/dev/urandom of=$pfl_file bs=1k count=3k
1501 mkdir -p $tmp/src/pfl_dir
1502 cp $pfl_file $tmp/src/pfl_dir/
1507 if [[ $MDS1_VERSION -ge $(version_code 2.10.56) ]]; then
1508 dom_dir=$target_dir/dom_dir
1509 dom_file=$dom_dir/dom_file
1510 flr_dir=$target_dir/flr_dir
1511 flr_file=$flr_dir/flr_file
1514 $LFS setstripe -E 1M -L mdt -E -1 -S 4M $dom_file
1515 dd if=/dev/urandom of=$dom_file bs=1k count=2k
1516 mkdir -p $tmp/src/dom_dir
1517 cp $dom_file $tmp/src/dom_dir
1520 LFS mirror create -N2 $flr_file
1521 dd if=/dev/urandom of=$flr_file bs=1k count=1
1522 mkdir -p $tmp/src/flr_dir
1523 cp $flr_file $tmp/src/flr_dir
1528 mkdir $tmp/img || return 1
1532 pushd_dir=/mnt/$FSNAME
1533 if [[ $MDSCOUNT -ge 2 ]]; then
1534 pushd_dir=$remote_dir
1535 if [[ "$MDS1_VERSION" -ge $(version_code 2.7.0) ]]; then
1537 ls -Rni --time-style=+%s >$tmp/img/list2
1543 ls -Rni --time-style=+%s >$tmp/img/list
1544 find ! -name .lustre -type f -exec sha1sum {} \; |
1545 sort -k 2 >$tmp/img/sha1sums
1547 $LCTL get_param -n version | head -n 1 |
1548 sed -e 's/^lustre: *//' >$tmp/img/commit
1550 [[ "$MDS1_VERSION" -ge $(version_code 2.3.50) ]] ||
1551 $LFS quotaon -ug /mnt/$FSNAME
1552 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1553 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1554 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1555 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1556 | tr -d "*" > $tmp/img/bspace
1557 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1558 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1559 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1560 | tr -d "*" > $tmp/img/ispace
1561 echo $T32_BLIMIT > $tmp/img/blimit
1562 echo $T32_ILIMIT > $tmp/img/ilimit
1564 $MULTIOP /mnt/$FSNAME/orph_file Ouw_c&
1568 debugfs -R "ls /PENDING" ${MDSDEV1:-$MDSDEV}
1569 cp ${MDSDEV1:-$MDSDEV} $tmp/img
1578 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1581 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1582 echo "Data verification failed"
1585 uname -r >$tmp/img/kernel
1586 uname -m >$tmp/img/arch
1588 for num in $(seq 2 $MDSCOUNT); do
1589 local devname=$(mdsdevname $num)
1591 [[ $(facet_fstype $facet) != zfs ]] ||
1592 devname=$(mdsvdevname $num)
1593 mv $devname $tmp/img
1596 for num in $(seq $OSTCOUNT); do
1597 local devname=$(ostdevname $num)
1599 [[ $(facet_fstype $facet) != zfs ]] ||
1600 devname=$(ostdevname $num)
1601 mv $devname $tmp/img
1604 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1605 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1608 tar cjvf $dst/disk$version-"$mds1_FSTYPE".tar.bz2 -S *
1613 #run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1616 # The list of applicable tarballs is returned via the caller's
1617 # variable "tarballs".
1620 [ "$CLIENTONLY" ] && skip "Client-only testing"
1622 local node=$(facet_active_host $SINGLEMDS)
1623 local r="do_node $node"
1625 ! $r which "$TUNEFS" && skip_env "tunefs.lustre required on $node"
1627 local IMGTYPE="$mds1_FSTYPE"
1629 [[ -n "$T32IMAGE" ]] && tarballs=$T32IMAGE ||
1630 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 \
1631 -name \'disk*-$IMGTYPE.tar.bz2\')
1633 [ -z "$tarballs" ] && skip "No applicable tarballs found"
1636 t32_test_cleanup() {
1638 local facet=$SINGLEMDS
1641 echo "== cleanup with rc=$rc =="
1642 if $shall_cleanup_lustre; then
1643 umount $tmp/mnt/lustre || rc=$?
1645 if $shall_cleanup_mdt; then
1646 $r $UMOUNT $tmp/mnt/mdt || rc=$?
1648 if $shall_cleanup_mdt1; then
1649 $r $UMOUNT $tmp/mnt/mdt1 || rc=$?
1651 if $shall_cleanup_ost; then
1652 $r $UMOUNT $tmp/mnt/ost || rc=$?
1654 if $shall_cleanup_ost1; then
1655 $r $UMOUNT $tmp/mnt/ost1 || rc=$?
1660 if [[ "$mds1_FSTYPE" == zfs ]]; then
1662 local poolname_list="t32fs-mdt1 t32fs-ost1"
1664 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1666 for poolname in $poolname_list; do
1667 destroy_zpool $facet $poolname
1670 combined_mgs_mds || start_mgs || rc=$?
1674 t32_bits_per_long() {
1676 # Yes, this is not meant to be perfect.
1686 t32_reload_modules() {
1688 local all_removed=false
1691 [ "$mds1_FSTYPE" == zfs ] && do_rpc_nodes $node "service zed stop"
1693 while ((i < 20)); do
1694 echo "Unloading modules on $node: Attempt $i"
1695 do_rpc_nodes $node $LUSTRE_RMMOD "$mds1_FSTYPE" &&
1697 do_rpc_nodes $node check_mem_leak || return 1
1698 if $all_removed; then
1699 do_rpc_nodes $node load_modules
1702 if [ "$mds1_FSTYPE" == zfs ]; then
1703 do_rpc_nodes $node "$ZPOOL status -v"
1708 echo "Unloading modules on $node: Given up"
1712 t32_wait_til_devices_gone() {
1718 echo wait for devices to go
1719 while ((i < 20)); do
1720 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1721 loops=$(do_rpc_nodes $node losetup -a | grep -c t32)
1722 ((devices == 0 && loops == 0)) && return 0
1726 echo "waiting for dev on $node: dev $devices loop $loops given up"
1727 do_rpc_nodes $node "losetup -a"
1728 do_rpc_nodes $node "$LCTL device_list"
1732 t32_verify_quota() {
1739 # LU-2435: if the underlying zfs doesn't support userobj_accounting,
1740 # lustre will estimate the object count usage. This fails quota
1741 # verification in 32b. The object quota usage should be accurate after
1742 # zfs-0.7.0 is released.
1743 [ "$mds1_FSTYPE" == zfs ] && {
1744 local zfs_version=$(do_facet $facet cat /sys/module/zfs/version)
1746 [ $(version_code $zfs_version) -lt $(version_code 0.7.0) ] && {
1747 echo "Skip quota verify for zfs: $zfs_version"
1752 $LFS quota -u $T32_QID -v $mnt
1754 qval=$($LFS quota -v -u $T32_QID $mnt |
1755 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1756 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1758 [ $qval -eq $img_bspace ] || {
1759 echo "bspace, act:$qval, exp:$img_bspace"
1763 qval=$($LFS quota -v -u $T32_QID $mnt |
1764 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1765 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1767 [ $qval -eq $img_ispace ] || {
1768 echo "ispace, act:$qval, exp:$img_ispace"
1772 qval=$($LFS quota -v -u $T32_QID $mnt |
1773 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1774 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1776 [ $qval -eq $img_blimit ] || {
1777 echo "blimit, act:$qval, exp:$img_blimit"
1781 qval=$($LFS quota -v -u $T32_QID $mnt |
1782 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1783 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1785 [ $qval -eq $img_ilimit ] || {
1786 echo "ilimit, act:$qval, exp:$img_ilimit"
1790 set_persistent_param_and_check $facet \
1791 "osd-$mds1_FSTYPE.$fsname-MDT0000.quota_slave.enabled" \
1792 "$fsname.quota.mdt" ug
1794 set_persistent_param_and_check $facet \
1795 "osd-$mds1_FSTYPE.$fsname-OST0000.quota_slave.enabled" \
1796 "$fsname.quota.ost" ug
1799 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1800 bs=1M count=$((img_blimit / 1024)) oflag=sync && {
1801 echo "Write succeed, but expect -EDQUOT"
1804 rm -f $mnt/t32_qf_new
1806 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1808 echo "Create succeed, but expect -EDQUOT"
1811 unlinkmany $mnt/t32_qf_ $img_ilimit
1821 sync_all_data > /dev/null 2>&1 || true
1823 [ "$uuid" != "global" ] || uuid=$mnt
1825 $LFS quota -v "$1" "$2" $mnt |
1826 awk 'BEGIN { num='$spec' } { if ($1 == "'$uuid'") \
1827 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1834 local dne_upgrade=${dne_upgrade:-"no"}
1835 local dom_upgrade=${dom_upgrade:-"no"}
1836 local project_upgrade=${project_upgrade:-"no"}
1837 local ff_convert=${ff_convert:-"no"}
1838 local pfl_upgrade=${pfl_upgrade:-"no"}
1839 local project_quota_upgrade=${project_quota_upgrade:-"no"}
1840 local dom_new_upgrade=${dom_new_upgrade:-"no"}
1841 local flr_upgrade=${flr_upgrade:-"no"}
1842 local shall_cleanup_mdt=false
1843 local shall_cleanup_mdt1=false
1844 local shall_cleanup_ost=false
1845 local shall_cleanup_ost1=false
1846 local shall_cleanup_lustre=false
1847 local mdt2_is_available=false
1848 local ost2_is_available=false
1849 local node=$(facet_active_host $SINGLEMDS)
1850 local r="do_node $node"
1866 local mdt_dev=$tmp/mdt
1867 local mdt2_dev=$tmp/mdt2
1868 local ost_dev=$tmp/ost
1869 local ost2_dev=$tmp/ost2
1873 local pfl_file=$tmp/mnt/lustre/remote_dir/pfl_dir/pfl_file
1874 local flr_file=$tmp/mnt/lustre/remote_dir/flr_dir/flr_file
1875 local dom_file=$tmp/mnt/lustre/remote_dir/dom_dir/dom_file
1876 local quota_dir=$tmp/mnt/lustre/remote_dir/project_quota_dir
1878 combined_mgs_mds || stop_mgs || error "Unable to stop MGS"
1879 trap 'trap - RETURN; t32_test_cleanup' RETURN
1882 nid=$($r $LCTL list_nids | head -1)
1884 mkdir -p $tmp/mnt/lustre || error "mkdir $tmp/mnt/lustre failed"
1885 $r mkdir -p $tmp/mnt/{mdt,mdt1,ost,ost1}
1886 $r tar xjvf $tarball -S -C $tmp || {
1887 error_noexit "Unpacking the disk image tarball"
1890 img_commit=$($r cat $tmp/commit)
1891 img_kernel=$($r cat $tmp/kernel)
1892 img_arch=$($r cat $tmp/arch)
1893 img_bspace=$($r cat $tmp/bspace)
1894 img_ispace=$($r cat $tmp/ispace)
1896 # older images did not have "blimit" and "ilimit" files
1897 # use old values for T32_BLIMIT and T32_ILIMIT
1898 $r test -f $tmp/blimit && img_blimit=$($r cat $tmp/blimit) ||
1900 $r test -f $tmp/ilimit && img_ilimit=$($r cat $tmp/ilimit) ||
1903 echo "Upgrading from $(basename $tarball), created with:"
1904 echo " Commit: $img_commit"
1905 echo " Kernel: $img_kernel"
1906 echo " Arch: $img_arch"
1907 echo "OST version: $(lustre_build_version ost1)"
1909 # The conversion can be made only when both of the following
1910 # conditions are satisfied:
1911 # - ost device img version < 2.3.64
1912 # - ost server version >= 2.5
1913 [ $(version_code $img_commit) -ge $(version_code 2.3.64) -o \
1914 "$OST1_VERSION" -lt $(version_code 2.5.0) ] &&
1917 ! $r test -f $mdt2_dev || mdt2_is_available=true
1918 ! $r test -f $ost2_dev || ost2_is_available=true
1920 if [[ "$mds1_FSTYPE" == zfs ]]; then
1921 echo "== fstype is zfs =="
1924 local poolname_list="t32fs-mdt1 t32fs-ost1"
1926 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1928 for poolname in $poolname_list; do
1930 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
1931 $ZPOOL import -f -d $tmp $poolname"
1934 # upgrade zpool to latest supported features, including
1935 # dnode quota accounting in 0.7.0
1936 $r "$ZPOOL upgrade -a"
1938 mdt_dev=t32fs-mdt1/mdt1
1939 ost_dev=t32fs-ost1/ost1
1940 ! $mdt2_is_available || mdt2_dev=t32fs-mdt2/mdt2
1941 ! $ost2_is_available || ost2_dev=t32fs-ost2/ost2
1942 wait_update_facet $SINGLEMDS "$ZPOOL list |
1943 awk '/^t32fs-mdt1/ { print \\\$1 }'" "t32fs-mdt1" || {
1944 error_noexit "import zfs pool failed"
1947 elif [ "$project_upgrade" != "no" ]; then
1948 echo "== project upgrade =="
1949 ! $r tune2fs -O project $mdt_dev &&
1950 error_noexit "enable project on mdt0 failed" &&
1952 $mdt2_is_available && ! $r tune2fs "-O project" $mdt2_dev &&
1953 error_noexit "enable project on mdt failed" &&
1955 ! $r tune2fs -O project $ost_dev &&
1956 error_noexit "enable project on ost failed" &&
1958 $ost2_is_available && ! $r tune2fs -O project $ost2_dev &&
1959 error_noexit "enable project on ost2 failed" &&
1963 $r $LCTL set_param debug="$PTLDEBUG"
1965 $r $TUNEFS --dryrun $mdt_dev || {
1967 error_noexit "tunefs.lustre before mounting the MDT"
1971 if $mdt2_is_available; then
1972 $r $TUNEFS --dryrun $mdt2_dev || {
1974 error_noexit "tunefs.lustre before mounting the MDT"
1979 if [ "$writeconf" ]; then
1980 echo "== writeconf mdt tunefs and quota =="
1982 if [ "$mds1_FSTYPE" == ldiskfs ]; then
1984 $r $TUNEFS --quota $mdt_dev || {
1986 error_noexit "Enable mdt quota feature"
1990 if [ "$project_quota_upgrade" != "no" ]; then
1991 echo "== enable mdt2 project quota =="
1992 $r $TUNE2FS -O project $mdt_dev || {
1994 error_noexit "tune2fs $mdt_dev failed"
1999 if $mdt2_is_available; then
2000 $r $TUNEFS --quota $mdt2_dev || {
2002 error_noexit "Enable mdt2 quota feature"
2005 if [ "$project_quota_upgrade" != "no" ]; then
2006 echo "== enable mdt2 project quota =="
2007 $r $TUNE2FS -O project $mdt2_dev || {
2010 "tune2fs $mdt2_dev failed"
2017 echo "== replace nids =="
2018 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
2019 [[ "$MGS_VERSION" -ge $(version_code 2.3.59) ]] ||
2020 skip "LU-2200: Cannot run over IB w/o lctl replace_nids "
2021 "(Need MGS version at least 2.3.59)"
2024 local osthost=$(facet_active_host ost1)
2025 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
2028 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2031 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt
2032 $r $LCTL replace_nids $fsname-OST0000 $ostnid
2033 if $ost2_is_available; then
2034 $r $LCTL replace_nids $fsname-OST0001 $ostnid
2036 $r $LCTL replace_nids $fsname-MDT0000 $nid
2037 if $mdt2_is_available; then
2038 $r $LCTL replace_nids $fsname-MDT0001 $nid
2040 $r $UMOUNT $tmp/mnt/mdt
2042 mopts=exclude=$fsname-OST0000
2043 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2048 t32_wait_til_devices_gone $node
2050 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2052 error_noexit "Mounting the MDT"
2055 shall_cleanup_mdt=true
2057 if $mdt2_is_available; then
2058 echo "== mdt2 available =="
2059 mopts=mgsnode=$nid,$mopts
2060 $r $MOUNT_CMD -o $mopts $mdt2_dev $tmp/mnt/mdt1 || {
2062 error_noexit "Mounting the MDT"
2065 shall_cleanup_mdt1=true
2066 echo "== mount new MDT $mdt2_dev =="
2068 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 || {
2069 error_noexit "enable remote dir create failed"
2073 elif [ "$dne_upgrade" != "no" ]; then
2074 local fs2mdsdev=$(mdsdevname 1_2)
2075 local fs2mdsvdev=$(mdsvdevname 1_2)
2077 echo "== mkfs new MDT on ${fs2mdsdev} =="
2078 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2079 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
2082 add $SINGLEMDS $(mds2failover_HOST="" \
2083 mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
2084 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
2085 error_noexit "Mkfs new MDT failed"
2089 [[ "$mds1_FSTYPE" != zfs ]] || import_zpool mds1
2091 $r $TUNEFS --dryrun $fs2mdsdev || {
2092 error_noexit "tunefs.lustre before mounting the MDT"
2096 echo "== mount new MDT....$fs2mdsdev =="
2097 $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
2098 error_noexit "mount mdt1 failed"
2102 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
2103 error_noexit "enable remote dir create failed"
2105 shall_cleanup_mdt1=true
2108 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
2109 error_noexit "Getting MDT UUID"
2112 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
2113 error_noexit "Unexpected MDT UUID: \"$uuid\""
2117 $r $TUNEFS --dryrun $ost_dev || {
2118 error_noexit "tunefs.lustre before mounting the OST"
2122 if $ost2_is_available; then
2123 $r $TUNEFS --dryrun $ost2_dev || {
2124 error_noexit "tunefs.lustre before mounting the OST"
2129 if [ "$writeconf" ]; then
2130 echo "== writeconf and ost quota tunefs =="
2131 mopts=mgsnode=$nid,$writeconf
2132 if [ "$ost1_FSTYPE" == ldiskfs ]; then
2134 $r $TUNEFS --quota $ost_dev || {
2136 error_noexit "Enable ost quota feature"
2140 if [ "$project_quota_upgrade" != "no" ]; then
2141 echo "== enable ost project quota =="
2142 $r $TUNE2FS -O project $ost_dev || {
2144 error_noexit "tune2fs $ost_dev failed"
2149 if $ost2_is_available; then
2150 $r $TUNEFS --quota $ost2_dev || {
2152 error_noexit "Enable ost2 quota feature"
2155 if [ "$project_quota_upgrade" != "no" ]; then
2156 echo "== enable ost2 project quota =="
2157 $r $TUNE2FS -O project $ost2_dev || {
2160 "tune2fs $ost2_dev failed"
2168 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2173 $r $MOUNT_CMD -onomgs -o$mopts $ost_dev $tmp/mnt/ost || {
2174 error_noexit "Mounting the OST"
2178 if $ost2_is_available; then
2179 $r $MOUNT_CMD -onomgs -o$mopts $ost2_dev $tmp/mnt/ost1 || {
2180 error_noexit "Mounting the OST2"
2183 shall_cleanup_ost1=true
2186 shall_cleanup_ost=true
2188 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
2189 error_noexit "Getting OST UUID"
2192 if [ "$uuid" != $fsname-OST0000_UUID ]; then
2193 error_noexit "Unexpected OST UUID: \"$uuid\""
2197 if $ost2_is_available; then
2198 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0001.uuid) ||{
2199 error_noexit "Getting OST1 UUID"
2202 if [ "$uuid" != $fsname-OST0001_UUID ]; then
2203 error_noexit "Unexpected OST1 UUID: \"$uuid\""
2208 if [[ $PERM_CMD =~ "set_param" ]]; then
2209 echo "== perm cmd = '$PERM_CMD' =="
2210 $r $PERM_CMD osc.$fsname-OST0000*.import=connection=$nid || {
2211 error_noexit "Setting OST \"failover.node\""
2214 $r $PERM_CMD mdc.$fsname-MDT0000*.import=connection=$nid || {
2215 error_noexit "Setting MDT \"failover.node\""
2218 $r $PERM_CMD osc.$fsname-OST0000-*.max_dirty_mb=15 || {
2219 error_noexit "Setting \"max_dirty_mb\""
2222 $r $PERM_CMD mdc.$fsname-MDT0000-*.max_rpcs_in_flight=9 || {
2223 error_noexit "Setting \"max_rpcs_in_flight\""
2226 $r $PERM_CMD lov.$fsname-MDT0000-*.stripesize=4M || {
2227 error_noexit "Setting \"lov.stripesize\""
2230 $r $PERM_CMD mdd.$fsname-MDT0000-*.atime_diff=70 || {
2231 error_noexit "Setting \"mdd.atime_diff\""
2235 echo "== perm cmd = '$PERM_CMD' =="
2236 $r $PERM_CMD $fsname-OST0000.failover.node=$nid || {
2237 error_noexit "Setting OST \"failover.node\""
2241 $r $PERM_CMD $fsname-MDT0000.failover.node=$nid || {
2242 error_noexit "Setting MDT \"failover.node\""
2246 $r $PERM_CMD $fsname-OST0000.osc.max_dirty_mb=15 || {
2247 error_noexit "Setting \"max_dirty_mb\""
2250 $r $PERM_CMD $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
2251 error_noexit "Setting \"max_rpcs_in_flight\""
2254 $r $PERM_CMD $fsname-MDT0000.lov.stripesize=4M || {
2255 error_noexit "Setting \"lov.stripesize\""
2258 $r $PERM_CMD $fsname-MDT0000.mdd.atime_diff=70 || {
2259 error_noexit "Setting \"mdd.atime_diff\""
2264 $r $LCTL pool_new $fsname.interop || {
2265 error_noexit "Setting \"interop\""
2269 if [ "$ff_convert" != "no" -a "$ost1_FSTYPE" == ldiskfs ]; then
2270 echo "== ff convert ($ff_convert) =="
2271 $r $LCTL lfsck_start -M $fsname-OST0000 || {
2272 error_noexit "Start OI scrub on OST0"
2276 # The oi_scrub should be on ost1, but for test_32(),
2277 # all on the SINGLEMDS.
2278 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
2279 osd-ldiskfs.$fsname-OST0000.oi_scrub |
2280 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
2281 error_noexit "Failed to get the expected 'completed'"
2285 local UPDATED=$($r $LCTL get_param -n \
2286 osd-ldiskfs.$fsname-OST0000.oi_scrub |
2287 awk '/^updated/ { print $2 }')
2288 [ $UPDATED -ge 1 ] || {
2289 error_noexit "Only $UPDATED objects have been converted"
2294 if [[ "$dne_upgrade" != "no" ]] || $mdt2_is_available; then
2295 echo "== dne upgrate ($dne_upgrade) or mdt2 available =="
2296 if [[ $PERM_CMD =~ "set_param" ]]; then
2297 $r $PERM_CMD mdc.$fsname-MDT0001*.import=connection=$nid || {
2298 error_noexit "Setting MDT1 \"failover.node\""
2302 $r $PERM_CMD mdc.$fsname-MDT0001-*.max_rpcs_in_flight=9 || {
2303 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
2306 $r $PERM_CMD lov.$fsname-MDT0001-*.stripesize=4M || {
2307 error_noexit "Setting MDT1 \"lov.stripesize\""
2311 $r $PERM_CMD $fsname-MDT0001.failover.node=$nid || {
2312 error_noexit "Setting MDT1 \"failover.node\""
2315 $r $PERM_CMD $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
2316 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
2319 $r $PERM_CMD $fsname-MDT0001.lov.stripesize=4M || {
2320 error_noexit "Setting MDT1 \"lov.stripesize\""
2326 if [ "$writeconf" ]; then
2327 echo "== writeconf and client mount =="
2328 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
2329 error_noexit "Mounting the client"
2333 shall_cleanup_lustre=true
2334 $r $LCTL set_param debug="$PTLDEBUG"
2336 t32_verify_quota $SINGLEMDS $fsname $tmp/mnt/lustre || {
2337 error_noexit "verify quota failed"
2341 if $r test -f $tmp/list; then
2342 echo "== list verification =="
2344 # There is not a Test Framework API to copy files to or
2345 # from a remote node.
2347 # LU-2393 - do both sorts on same node to ensure locale
2349 local list_file=$tmp/list
2351 if $mdt2_is_available; then
2352 if [[ -d $tmp/mnt/lustre/striped_dir_old ]] &&
2353 $r test -f $tmp/list2; then
2354 list_file=$tmp/list2
2355 pushd $tmp/mnt/lustre/striped_dir_old
2357 pushd $tmp/mnt/lustre/remote_dir
2360 pushd $tmp/mnt/lustre
2362 $r cat $list_file | sort -k 6 >$tmp/list.orig
2363 ls -Rni --time-style=+%s | sort -k 6 |
2364 sed 's/\. / /' >$tmp/list || {
2370 # 32-bit and 64-bit clients use different algorithms to
2371 # convert FIDs into inode numbers. Hence, remove the
2372 # inode numbers from the lists, if the original list was
2373 # created on an architecture with different number of
2376 if [ $(t32_bits_per_long $(uname -m)) != \
2377 $(t32_bits_per_long $img_arch) ]; then
2378 echo "Different number of bits per \"long\"" \
2379 "from the disk image"
2380 for list in list.orig list; do
2381 sed -i -e 's/^[0-9]\+[ \t]\+//' \
2385 if ! diff -ub $tmp/list.orig $tmp/list; then
2386 error_noexit "list verification failed"
2390 echo "== list verification skipped =="
2393 #non-dom upgrade to dom
2394 if [ "$dom_upgrade" != "no" ]; then
2395 echo "== dom upgrade =="
2396 $LFS setstripe -E 1M -L mdt -E EOF $tmp/mnt/lustre/dom || {
2397 error_noexit "Verify DoM creation"
2400 [ $($LFS getstripe -L $tmp/mnt/lustre/dom) == "mdt" ] || {
2401 error_noexit "Verify a DoM file"
2404 dd if=/dev/urandom of=$tmp/mnt/lustre/dom bs=4096 \
2405 count=1 conv=fsync || {
2406 error_noexit "Cannot write to DoM file"
2409 [ $(stat -c%s $tmp/mnt/lustre/dom) == 4096 ] || {
2410 error_noexit "DoM: bad size after write"
2413 rm $tmp/mnt/lustre/dom
2415 set_persistent_param_and_check mds \
2416 "lod.*MDT0000*.dom_stripesize" \
2417 "$fsname-MDT0000.lod.dom_stripesize" 0 || {
2418 error_noexit "Changing \"dom_stripesize\""
2424 #$LFS setstripe -E 1M -L mdt -E -1 -S 4M $dom_file
2425 if [ "$dom_new_upgrade" != "no" ]; then
2426 if ! $mdt2_is_available; then
2427 dom_file=$tmp/mnt/lustre/dom_dir/dom_file
2429 echo "== check DoM file can be accessed =="
2430 [ $($LFS getstripe -I1 -L $dom_file) == "mdt" ] || {
2431 error_noexit "Verify a DoM file"
2434 [ $(stat -c%s $dom_file) == $((2 * 1024 * 1024)) ] || {
2435 error_noexit "DoM: bad size after write"
2440 if [ "$flr_upgrade" != "no" ]; then
2441 if ! $mdt2_is_available; then
2442 flr_file=$tmp/mnt/lustre/flr_dir/flr_file
2444 local mirror_count=$($LFS getstripe -N $flr_file)
2445 echo "== check FLR file =="
2446 [ $mirror_count == 2 ] || {
2447 error_noexit "FLR mirror count wrong"
2452 if [ "$dne_upgrade" != "no" ]; then
2453 echo "== dne upgrade striped_dir =="
2454 $LFS mkdir -i 1 -c2 $tmp/mnt/lustre/striped_dir || {
2455 error_noexit "set striped dir failed"
2458 $LFS setdirstripe -D -c2 $tmp/mnt/lustre/striped_dir
2460 pushd $tmp/mnt/lustre
2461 tar -c --exclude=./striped_dir \
2462 --exclude=./striped_dir_old \
2463 --exclude=./remote_dir -f - .|
2464 tar -xvf - -C striped_dir 1>/dev/null || {
2465 error_noexit "cp to striped dir failed"
2471 if [ "$pfl_upgrade" != "no" ]; then
2478 echo "== check PFL file =="
2479 if ! $mdt2_is_available; then
2480 pfl_file=$tmp/mnt/lustre/pfl_dir/pfl_file
2482 comp_cnt=$($LFS getstripe --component-count $pfl_file)
2483 [ $comp_cnt == 2 ] || {
2484 error_noexit "wrong comp_cnt $comp_cnt"
2488 comp_size=$($LFS getstripe -I1 -E $pfl_file)
2489 [ $comp_size == 2097152 ] || {
2490 error_noexit "wrong component size $comp_size"
2494 comp_id=$($LFS getstripe -I1 -i $pfl_file)
2495 [ $comp_id == 0 ] || {
2496 error_noexit "wrong comp id $comp_id"
2500 comp_id=$($LFS getstripe -I2 -i $pfl_file)
2501 [ $comp_id -eq 1 ] || {
2502 error_noexit "wrong comp id $comp_id"
2506 stripe_size=$($LFS getstripe -I1 -S $pfl_file)
2507 [ $stripe_size -eq 1048576 ] || {
2508 error_noexit "wrong stripe size $stripe_size"
2512 stripe_size=$($LFS getstripe -I2 -S $pfl_file)
2513 [ $comp_size -eq 2097152 ] || {
2514 error_noexit "wrong component size $comp_size"
2519 if [ "$project_quota_upgrade" != "no" ]; then
2520 if ! $mdt2_is_available; then
2521 quota_dir=$tmp/mnt/lustre/project_quota_dir
2525 echo "== check Project Quota =="
2526 hardlimit=$(getquota -p $T32_PRJID global 3 \
2528 [ $hardlimit == $T32_PROLIMIT ] || {
2529 error_noexit "wrong hardlimit $hardlimit"
2534 # If it is upgrade from DNE (2.5), then rename the remote dir,
2535 # which is created in 2.5 to striped dir.
2536 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]]; then
2537 echo "== rename remote_dir =="
2538 stripe_index=$($LFS getdirstripe -i \
2539 $tmp/mnt/lustre/remote_dir)
2541 [[ $stripe_index -eq 1 ]] || {
2542 error_noexit "get index \"$stripe_index\"" \
2543 "from remote dir failed"
2546 mv $tmp/mnt/lustre/remote_dir \
2547 $tmp/mnt/lustre/striped_dir/ || {
2548 error_noexit "mv remote dir failed"
2553 # If it is upgraded from DNE (2.7), then move the striped dir
2554 # which was created in 2.7 to the new striped dir.
2555 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]] &&
2556 echo "== rename striped dir =="
2557 [[ -d $tmp/mnt/lustre/striped_dir_old ]]; then
2558 stripe_count=$($LFS getdirstripe -c \
2559 $tmp/mnt/lustre/striped_dir_old)
2560 [[ $stripe_count -eq 2 ]] || {
2561 error_noexit "get count $stripe_count" \
2562 "from striped dir failed"
2565 mv $tmp/mnt/lustre/striped_dir_old \
2566 $tmp/mnt/lustre/striped_dir/ || {
2567 error_noexit "mv striped dir failed"
2573 $r $LCTL set_param -n osd*.*.force_sync=1
2574 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
2575 error_noexit "dd failed"
2578 rm -rf $tmp/mnt/lustre/tmp_file || {
2579 error_noexit "rm failed"
2583 if $r test -f $tmp/sha1sums; then
2584 echo "== checking sha1sums =="
2585 # LU-2393 - do both sorts on same node to ensure locale
2587 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
2588 if [[ "$dne_upgrade" != "no" ]]; then
2589 if [[ -d $tmp/mnt/lustre/striped_dir/remote_dir ]]; then
2590 pushd $tmp/mnt/lustre/striped_dir/remote_dir
2592 pushd $tmp/mnt/lustre/striped_dir
2594 elif [[ "$pfl_upgrade" != "no" ]] ||
2595 [[ "$flr_upgrade" != "no" ]] ||
2596 [[ "$dom_new_upgrade" != "no" ]] ||
2597 [[ "$project_quota_upgrade" != "no" ]] ||
2598 [[ -d $tmp/mnt/lustre/remote_dir ]]; then
2599 pushd $tmp/mnt/lustre/remote_dir
2601 pushd $tmp/mnt/lustre
2604 find ! -path "*remote_dir*" ! -path "*striped_dir*" \
2605 ! -name .lustre -type f -exec sha1sum {} \; |
2606 sort -k 2 >$tmp/sha1sums || {
2608 error_noexit "sha1sum"
2614 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
2615 echo "==** find $save files start **=="
2616 find $save -type f -ls
2617 echo "==** find $save files finish **=="
2618 error_noexit "sha1sum verification failed"
2622 # if upgrade from DNE(2.5), then check remote directory
2623 # if upgrade from DNE(2.7), then check striped directory
2624 if $mdt2_is_available &&
2625 [[ "$dne_upgrade" != "no" ]]; then
2626 echo "== mdt2 upgrade sha1sum check =="
2627 local new_dir="$tmp/mnt/lustre/striped_dir"
2628 local striped_dir_old="$new_dir/striped_dir_old"
2630 local dir_list="$new_dir/remote_dir"
2631 [[ ! -d $triped_dir_old ]] ||
2632 dir_list+=" $striped_dir_old"
2634 for dir in $dir_list; do
2635 echo "== dir = $dir =="
2637 find ! -name .lustre -type f \
2638 -exec sha1sum {} \; |
2639 sort -k 2 >$tmp/sha1sums || {
2642 error_noexit "sha1sum"
2646 if ! diff -ub $tmp/sha1sums.orig \
2649 error_noexit "sha1sum $dir" \
2656 echo "== sha1sum verification skipped =="
2659 # PFL write test after sha1sum check
2660 if [ "$pfl_upgrade" != "no" ]; then
2661 local rw_len=$((3 * 1034 * 1024))
2663 if ! $mdt2_is_available; then
2664 pfl_file=$tmp/mnt/lustre/pfl_dir/pfl_file
2666 small_write $pfl_file $rw_len || {
2667 error_noexit "PFL RW Failed"
2672 if [ "$dne_upgrade" != "no" ]; then
2673 rm -rf $tmp/mnt/lustre/striped_dir || {
2674 error_noexit "remove remote dir failed"
2679 # migrate files/dirs to remote MDT, then move them back
2680 if [ "$MDS1_VERSION" -ge $(version_code 2.7.50) -a \
2681 $dne_upgrade != "no" ]; then
2682 $r $LCTL set_param -n \
2683 mdt.${fsname}*.enable_remote_dir=1 2>/dev/null
2685 echo "== test migration =="
2686 pushd $tmp/mnt/lustre
2687 for dir in $(find ! -name .lustre ! -name . -type d); do
2688 mdt_index=$($LFS getdirstripe -i $dir)
2689 stripe_cnt=$($LFS getdirstripe -c $dir)
2690 if [ $mdt_index = 0 -a $stripe_cnt -le 1 ]; then
2691 $LFS migrate -m 1 $dir || {
2693 error_noexit "migrate MDT1 failed"
2699 for dir in $(find ! -name . ! -name .lustre -type d); do
2700 mdt_index=$($LFS getdirstripe -i $dir)
2701 stripe_cnt=$($LFS getdirstripe -c $dir)
2702 if [ $mdt_index = 1 -a $stripe_cnt -le 1 ]; then
2703 $LFS migrate -m 0 $dir || {
2705 error_noexit "migrate MDT0 failed"
2714 # When adding new data verification tests, please check for
2715 # the presence of the required reference files first, like
2716 # the "sha1sums" and "list" tests above, to avoid the need to
2717 # regenerate every image for each test addition.
2720 nrpcs_orig=$($LCTL get_param \
2721 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
2722 error_noexit "Getting \"max_rpcs_in_flight\""
2725 nrpcs=$((nrpcs_orig + 5))
2727 set_persistent_param_and_check client \
2728 "mdc.$fsname-MDT0000*.max_rpcs_in_flight" \
2729 "$fsname-MDT0000.mdc.max_rpcs_in_flight" $nrpcs || {
2730 error_noexit "Changing \"max_rpcs_in_flight\""
2734 umount $tmp/mnt/lustre || {
2735 error_noexit "Unmounting the client"
2738 shall_cleanup_lustre=false
2740 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
2741 error_noexit "Mounting the client"
2745 if [ "$mds1_FSTYPE" == ldiskfs -a \
2746 "$project_upgrade" != "no" ]; then
2747 ! $LFS project -d -p 1 $tmp/mnt/lustre/* &&
2748 error_noexit "set project failed" &&
2752 [[ $(do_facet mds1 pgrep orph_.*-MDD | wc -l) == 0 ]] ||
2753 error "MDD orphan cleanup thread not quit"
2755 umount $tmp/mnt/lustre || {
2756 error_noexit "Unmounting the client"
2760 if [[ "$dne_upgrade" != "no" ]] || $mdt2_is_available; then
2761 $r $UMOUNT $tmp/mnt/mdt1 || {
2762 error_noexit "Unmounting the MDT2"
2765 if [[ "$mds1_FSTYPE" == zfs ]]; then
2766 $r "$ZPOOL export t32fs-mdt2"
2768 shall_cleanup_mdt1=false
2771 $r $UMOUNT $tmp/mnt/mdt || {
2772 error_noexit "Unmounting the MDT"
2775 if [[ "$mds1_FSTYPE" == zfs ]]; then
2776 $r "$ZPOOL export t32fs-mdt1"
2778 shall_cleanup_mdt=false
2780 $r $UMOUNT $tmp/mnt/ost || {
2781 error_noexit "Unmounting the OST"
2784 if $ost2_is_available; then
2785 $r $UMOUNT $tmp/mnt/ost1 || {
2786 error_noexit "Unmounting the OST1"
2789 shall_cleanup_ost1=false
2792 if [[ $ost1_FSTYPE == zfs ]]; then
2793 $r "$ZPOOL export t32fs-ost1"
2794 if $ost2_is_available; then
2795 $r "$ZPOOL export t32fs-ost2"
2798 shall_cleanup_ost=false
2800 t32_reload_modules $node || {
2801 error_noexit "Reloading modules"
2805 if [[ "$mds1_FSTYPE" == zfs ]]; then
2806 local poolname=t32fs-mdt1
2808 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
2809 $ZPOOL import -f -d $tmp $poolname"
2811 # upgrade zpool to latest supported features,
2812 # including dnode quota accounting in 0.7.0
2813 $r "$ZPOOL upgrade $poolname"
2816 # mount a second time to make sure we didnt leave upgrade flag on
2817 $r $TUNEFS --dryrun $mdt_dev || {
2819 error_noexit "tunefs.lustre before remounting the MDT"
2823 mopts=exclude=$fsname-OST0000
2824 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2827 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2828 error_noexit "Remounting the MDT"
2831 shall_cleanup_mdt=true
2841 for tarball in $tarballs; do
2842 banner "testing $tarball upgrade"
2843 t32_test $tarball || let "rc += $?"
2847 run_test 32a "Upgrade (not live)"
2855 for tarball in $tarballs; do
2856 banner "testing $tarball upgrade with writeconf"
2857 t32_test $tarball writeconf || let "rc += $?"
2861 run_test 32b "Upgrade with writeconf"
2869 for tarball in $tarballs; do
2870 banner "testing $tarball upgrade with DNE"
2872 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
2876 run_test 32c "dne upgrade test"
2884 for tarball in $tarballs; do
2885 banner "testing $tarball upgrade with ff and project upgrade"
2886 project_upgrade="no"
2887 [[ "$MDS1_VERSION" -ge $(version_code 2.13.54) ]] &&
2888 [[ "$tarball" =~ "disk2_4-ldiskfs" ]] &&
2889 project_upgrade="yes"
2890 project_upgrade=$project_upgrade ff_convert=yes t32_test \
2895 run_test 32d "convert ff and project quota upgrade test"
2898 [[ "$MDS1_VERSION" -ge $(version_code 2.10.56) ]] ||
2899 skip "Need MDS version at least 2.10.56"
2906 for tarball in $tarballs; do
2907 [[ "$tarball" =~ "2_9" ]] ||
2908 { echo "skip $(basename $tarball)"; continue; }
2910 banner "testing $tarball upgrade with DoM"
2911 dom_upgrade=yes t32_test $tarball writeconf || let "rc += $?"
2915 run_test 32e "dom upgrade test"
2918 [[ $MDS1_VERSION -ge $(version_code 2.10.56) ]] ||
2919 skip "Need MDS version at least 2.10.56"
2926 for tarball in $tarballs; do
2927 [[ "$tarball" =~ "2_10" ]] ||
2928 { echo "skip $(basename $tarball)"; continue; }
2929 pfl_upgrade=yes project_quota_upgrade=yes \
2930 t32_test $tarball writeconf || let "rc += $?"
2934 run_test 32f "pfl upgrade test"
2937 [[ $MDS1_VERSION -ge $(version_code 2.10.56) ]] ||
2938 skip "Need MDS version at least 2.10.56"
2945 for tarball in $tarballs; do
2946 [[ $tarball =~ "2_12" ]] ||
2947 { echo "skip $(basename $tarball)"; continue; }
2948 flr_upgrade=yes dom_new_upgrade=yes \
2949 t32_test $tarball writeconf || let "rc += $?"
2953 run_test 32g "flr/dom upgrade test"
2955 test_33a() { # bug 12333, was test_33
2956 local FSNAME2=test-$testnum
2958 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2961 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
2963 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
2964 local dev=${SINGLEMDS}_dev
2965 local MDSDEV=${!dev}
2967 is_blkdev $SINGLEMDS $MDSDEV &&
2968 skip_env "mixed loopback and real device not working"
2971 local fs2mdsdev=$(mdsdevname 1_2)
2972 local fs2ostdev=$(ostdevname 1_2)
2973 local fs2mdsvdev=$(mdsvdevname 1_2)
2974 local fs2ostvdev=$(ostvdevname 1_2)
2976 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2977 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
2980 if combined_mgs_mds; then
2981 local mgs_flag="--mgs"
2984 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --fsname=${FSNAME2} \
2985 --reformat $mgs_flag $mkfsoptions $fs2mdsdev $fs2mdsvdev ||
2987 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2988 --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \
2989 $fs2ostvdev || exit 10
2991 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
2992 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2994 if [[ $PERM_CMD == *"set_param -P"* ]]; then
2995 do_facet mgs "$PERM_CMD timeout=200" ||
2996 error "$PERM_CMD timeout=200 failed"
2998 do_facet mgs "$PERM_CMD $FSNAME2.sys.timeout=200" ||
2999 error "$PERM_CMD $FSNAME2.sys.timeout=200 failed"
3001 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
3002 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
3005 cp /etc/hosts $MOUNT2/ || error "copy /etc/hosts $MOUNT2/ failed"
3006 $LFS getstripe $MOUNT2/hosts ||
3007 error "$LFS getstripe $MOUNT2/hosts failed"
3012 cleanup_nocli || error "cleanup_nocli failed with $?"
3014 run_test 33a "Mount ost with a large index number"
3016 test_33b() { # was test_34
3019 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
3020 # Drop lock cancelation reply during umount
3021 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
3022 do_facet client $LCTL set_param fail_loc=0x80000304
3023 #lctl set_param debug=-1
3024 umount_client $MOUNT
3025 cleanup || error "cleanup failed with $?"
3027 run_test 33b "Drop cancel during umount"
3031 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
3032 manual_umount_client
3034 do_facet client killall -USR1 multiop
3035 if [ $rc -eq 0 ]; then
3036 error "umount not fail!"
3039 cleanup || error "cleanup failed with rc $?"
3041 run_test 34a "umount with opened file should be fail"
3045 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3046 stop_mds || error "Unable to stop MDS"
3048 manual_umount_client --force || error "mtab after failed umount with $?"
3050 cleanup || error "cleanup failed with $?"
3052 run_test 34b "force umount with failed mds should be normal"
3056 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3057 stop_ost || error "Unable to stop OST1"
3059 manual_umount_client --force || error "mtab after failed umount with $?"
3061 cleanup || error "cleanup failed with $?"
3063 run_test 34c "force umount with failed ost should be normal"
3065 test_35a() { # bug 12459
3068 DBG_SAVE="`$LCTL get_param -n debug`"
3069 $LCTL set_param debug="ha"
3071 log "Set up a fake failnode for the MDS"
3073 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
3074 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
3076 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3077 do_facet mgs "$PERM_CMD \
3078 mdc.*${device}*.import=connection=$(h2nettype $FAKENID)" ||
3079 error "Setting mdc.*${device}*.import=connection=\
3080 $(h2nettype $FAKENID) failed."
3082 do_facet mgs "$PERM_CMD \
3083 ${device}.failover.node=$(h2nettype $FAKENID)" ||
3084 error "Setting ${device}.failover.node=\
3085 $(h2nettype $FAKENID) failed."
3087 log "Wait for RECONNECT_INTERVAL seconds (10s)"
3090 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
3093 log "Stopping the MDT: $device"
3094 stop_mdt 1 || error "MDT0 stop fail"
3096 df $MOUNT > /dev/null 2>&1 &
3098 log "Restarting the MDT: $device"
3099 start_mdt 1 || error "MDT0 start fail"
3100 log "Wait for df ($DFPID) ... "
3103 $LCTL set_param debug="$DBG_SAVE"
3105 # retrieve from the log the first server that the client tried to
3106 # contact after the connection loss
3107 $LCTL dk $TMP/lustre-log-$TESTNAME.log
3108 NEXTCONN=`awk "/${MSG}/ {start = 1;}
3109 /import_select_connection.*$device-mdc.* using connection/ {
3111 if (\\\$NF ~ /$FAKENID/)
3117 }" $TMP/lustre-log-$TESTNAME.log`
3118 [ "$NEXTCONN" != "0" ] &&
3119 error "Tried to connect to ${NEXTCONN} not last active server"
3120 cleanup || error "cleanup failed with $?"
3121 # remove nid settings
3122 writeconf_or_reformat
3124 run_test 35a "Reconnect to the last active server first"
3126 test_35b() { # bug 18674
3127 remote_mds || skip "local MDS"
3131 $LCTL set_param debug="ha"
3133 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
3136 log "Set up a fake failnode for the MDS"
3138 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
3139 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
3141 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3142 do_facet mgs "$PERM_CMD \
3143 mdc.*${device}*.import=connection=$(h2nettype $FAKENID)" ||
3144 error "Set mdc.*${device}*.import=connection=\
3145 $(h2nettype $FAKENID) failed"
3147 do_facet mgs "$PERM_CMD \
3148 ${device}.failover.node=$(h2nettype $FAKENID)" ||
3149 error "Set ${device}.failover.node=\
3150 $(h2nettype $FAKENID) failed"
3153 local at_max_saved=0
3154 # adaptive timeouts may prevent seeing the issue
3155 if at_is_enabled; then
3156 at_max_saved=$(at_max_get mds)
3157 at_max_set 0 mds client
3160 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
3162 log "Injecting EBUSY on MDS"
3163 # Setting OBD_FAIL_MDS_RESEND=0x136
3164 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" ||
3165 error "unable to set param fail_loc=0x80000136"
3167 $LCTL set_param mdc.${FSNAME}*.stats=clear
3169 log "Creating a test file and stat it"
3170 touch $MOUNT/$tdir/$tfile || error "touch $MOUNT/$tdir/$tfile failed"
3171 stat $MOUNT/$tdir/$tfile
3173 log "Stop injecting EBUSY on MDS"
3174 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" ||
3175 error "unable to set param fail_loc=0"
3176 rm -f $MOUNT/$tdir/$tfile || error "remove $MOUNT/$tdir/$tfile failed"
3179 # restore adaptive timeout
3180 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
3182 $LCTL dk $TMP/lustre-log-$TESTNAME.log
3184 CONNCNT=$($LCTL get_param mdc.${FSNAME}*.stats |
3185 awk '/mds_connect/{print $2}')
3187 # retrieve from the log if the client has ever tried to
3188 # contact the fake server after the loss of connection
3189 FAILCONN=`awk "BEGIN {ret = 0;}
3190 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
3192 if (\\\$NF ~ /$FAKENID/) {
3197 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
3199 [ "$FAILCONN" == "0" ] &&
3200 error "The client reconnection has not been triggered"
3201 [ "$FAILCONN" == "2" ] &&
3202 error "Primary server busy, client reconnect to failover failed"
3205 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
3206 # Reconnects are supposed to be rate limited to one every 5s
3207 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] &&
3208 error "Too many reconnects $CONNCNT"
3210 cleanup || error "cleanup failed with $?"
3211 # remove nid settings
3212 writeconf_or_reformat
3214 run_test 35b "Continue reconnection retries, if the active server is busy"
3217 [ $OSTCOUNT -lt 2 ] && skip_env "needs >= 2 OSTs"
3219 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] ||
3223 local FSNAME2=test1234
3224 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3226 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
3228 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
3229 is_blkdev $SINGLEMDS $MDSDEV &&
3230 skip_env "mixed loopback and real device not working"
3233 local fs2mdsdev=$(mdsdevname 1_2)
3234 local fs2ostdev=$(ostdevname 1_2)
3235 local fs3ostdev=$(ostdevname 2_2)
3236 local fs2mdsvdev=$(mdsvdevname 1_2)
3237 local fs2ostvdev=$(ostvdevname 1_2)
3238 local fs3ostvdev=$(ostvdevname 2_2)
3241 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
3242 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
3243 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
3244 # different one than the default value here.
3245 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
3246 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
3247 add fs3ost $(mkfs_opts ost2 ${fs3ostdev}) --mgsnode=$MGSNID \
3248 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
3250 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
3251 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
3252 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
3253 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
3254 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
3256 sleep 5 # until 11778 fixed
3258 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || error "dd failed"
3260 BKTOTAL=$($LCTL get_param -n obdfilter.*.kbytestotal |
3261 awk 'BEGIN{total=0}; {total+=$1}; END{print total}')
3262 BKFREE=$($LCTL get_param -n obdfilter.*.kbytesfree |
3263 awk 'BEGIN{free=0}; {free+=$1}; END{print free}')
3264 BKAVAIL=$($LCTL get_param -n obdfilter.*.kbytesavail |
3265 awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}')
3266 STRING=$(df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}')
3267 DFTOTAL=$(echo $STRING | cut -d, -f1)
3268 DFUSED=$(echo $STRING | cut -d, -f2)
3269 DFAVAIL=$(echo $STRING | cut -d, -f3)
3270 DFFREE=$(($DFTOTAL - $DFUSED))
3272 ALLOWANCE=$((64 * $OSTCOUNT))
3274 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
3275 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
3276 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
3279 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
3280 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
3281 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
3284 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
3285 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
3286 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
3291 stop fs3ost -f || error "unable to stop OST3"
3292 stop fs2ost -f || error "unable to stop OST2"
3293 stop fs2mds -f || error "unable to stop second MDS"
3294 unload_modules_conf || error "unable unload modules"
3297 run_test 36 "df report consistency on OSTs with different block size"
3300 local mntpt=$(facet_mntpt $SINGLEMDS)
3301 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
3302 local mdsdev_sym="$TMP/sym_mdt.img"
3303 local opts=$MDS_MOUNT_OPTS
3306 if [ "$mds1_FSTYPE" != ldiskfs ]; then
3307 skip "ldiskfs only test"
3310 echo "MDS : $mdsdev"
3311 echo "SYMLINK : $mdsdev_sym"
3312 do_facet $SINGLEMDS rm -f $mdsdev_sym
3314 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
3316 echo "mount symlink device - $mdsdev_sym"
3318 if ! do_facet $SINGLEMDS test -b $mdsdev; then
3319 opts=$(csa_add "$opts" -o loop)
3323 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
3324 $mdsdev_sym $mntpt 2>&1)
3327 echo mount_op=$mount_op
3329 do_facet $SINGLEMDS "$UMOUNT $mntpt && rm -f $mdsdev_sym"
3331 if $(echo $mount_op | grep -q "unable to set tunable"); then
3332 error "set tunables failed for symlink device"
3335 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
3337 run_test 37 "verify set tunables works for symlink device"
3339 test_38() { # bug 14222
3340 local mntpt=$(facet_mntpt $SINGLEMDS)
3345 local SRC="/etc /bin"
3346 local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT)
3348 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
3349 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3350 tar cf - $FILES | tar xf - -C $DIR/$tdir ||
3351 error "copying $SRC to $DIR/$tdir"
3353 umount_client $MOUNT || error "umount_client $MOUNT failed"
3354 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3355 stop_mds || error "Unable to stop MDS"
3356 log "delete lov_objid file on MDS"
3358 mount_fstype $SINGLEMDS || error "mount MDS failed (1)"
3360 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid; rm $mntpt/lov_objid"
3362 unmount_fstype $SINGLEMDS || error "umount failed (1)"
3364 # check create in mds_lov_connect
3365 start_mds || error "unable to start MDS"
3366 mount_client $MOUNT || error "mount_client $MOUNT failed"
3368 [ $V ] && log "verifying $DIR/$tdir/$f"
3369 diff -q $f $DIR/$tdir/$f || ERROR=y
3371 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3372 if [ "$ERROR" = "y" ]; then
3373 # check it's updates in sync
3374 umount_client $MOUNT
3376 mount_fstype $SIGNLEMDS
3377 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3378 unmount_fstype $SINGLEMDS
3379 error "old and new files are different after connect" || true
3381 touch $DIR/$tdir/f2 || error "f2 file create failed"
3383 # check it's updates in sync
3384 umount_client $MOUNT || error "second umount_client $MOUNT failed"
3387 mount_fstype $SINGLEMDS || error "mount MDS failed (3)"
3389 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3390 do_facet $SINGLEMDS dd if=/dev/zero of=$mntpt/lov_objid.clear count=8
3392 unmount_fstype $SINGLEMDS || error "umount failed (3)"
3394 start_mds || error "unable to start MDS"
3395 mount_client $MOUNT || error "mount_client $MOUNT failed"
3397 [ $V ] && log "verifying $DIR/$tdir/$f"
3398 diff -q $f $DIR/$tdir/$f || ERROR=y
3400 touch $DIR/$tdir/f3 || error "f3 file create failed"
3401 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3402 umount_client $MOUNT || error "third umount_client $MOUNT failed"
3404 mount_fstype $SINGLEMDS || error "mount MDS failed (4)"
3405 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3406 unmount_fstype $SINGLEMDS || error "umount failed (4)"
3408 [ "$ERROR" = "y" ] &&
3409 error "old and new files are different after sync" || true
3411 log "files compared the same"
3412 cleanup || error "cleanup failed with $?"
3414 run_test 38 "MDS recreates missing lov_objid file from OST data"
3419 cleanup || error "cleanup failed with $?"
3420 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
3421 error "memory leak detected" || true
3423 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
3425 test_40() { # bug 15759
3426 start_ost || error "Unable to start OST1"
3427 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
3428 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
3430 cleanup || error "cleanup failed with rc $?"
3432 run_test 40 "race during service thread startup"
3434 test_41a() { #bug 14134
3435 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3436 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
3437 skip "Loop devices does not work with nosvc option"
3440 combined_mgs_mds || skip "needs combined MGT and MDT device"
3442 start_mdt 1 -o nosvc -n
3443 if [ $MDSCOUNT -ge 2 ]; then
3444 for num in $(seq 2 $MDSCOUNT); do
3445 start_mdt $num || return
3448 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
3449 start_mdt 1 -o nomgs,force
3450 mount_client $MOUNT || error "mount_client $MOUNT failed"
3453 echo "blah blah" > $MOUNT/$tfile
3456 umount_client $MOUNT || error "umount_client $MOUNT failed"
3457 stop ost1 -f || error "unable to stop OST1"
3458 stop_mds || error "Unable to stop MDS"
3459 stop_mds || error "Unable to stop MDS on second try"
3461 run_test 41a "mount mds with --nosvc and --nomgs"
3464 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3465 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
3466 skip "Loop devices does not work with nosvc option"
3469 ! combined_mgs_mds && skip "needs combined mgs device"
3473 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3475 start_mdt 1 -o nosvc -n
3476 if [ $MDSCOUNT -ge 2 ]; then
3477 for num in $(seq 2 $MDSCOUNT); do
3478 start_mdt $num || return
3481 start_ost || error "Unable to start OST1"
3482 start_mdt 1 -o nomgs,force
3483 mount_client $MOUNT || error "mount_client $MOUNT failed"
3486 echo "blah blah" > $MOUNT/$tfile
3487 cat $MOUNT/$tfile || error "cat $MOUNT/$tfile failed"
3489 umount_client $MOUNT -f || error "umount_client $MOUNT failed"
3490 stop_ost || error "Unable to stop OST1"
3491 stop_mds || error "Unable to stop MDS"
3492 stop_mds || error "Unable to stop MDS on second try"
3494 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
3497 local oss_list=$(comma_list $(osts_nodes))
3499 [[ "$MDS1_VERSION" -ge $(version_code 2.6.52) ]] ||
3500 [[ "$MDS1_VERSION" -ge $(version_code 2.5.26) &&
3501 "$MDS1_VERSION" -lt $(version_code 2.5.50) ]] ||
3502 [[ "$MDS1_VERSION" -ge $(version_code 2.5.4) &&
3503 "$MDS1_VERSION" -lt $(version_code 2.5.11) ]] ||
3504 skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"
3506 # ensure mds1 ost1 have been created even if running sub-test standalone
3509 cleanup || error "cleanup failed"
3511 # using directly mount command instead of start() function to avoid
3512 # any side effect of // with others/externals tools/features
3513 # ("zpool import", ...)
3515 # MDT concurrent start
3517 LOAD_MODULES_REMOTE=true load_modules
3518 do_facet $SINGLEMDS "lsmod | grep -q libcfs" ||
3519 error "MDT concurrent start: libcfs module not loaded"
3521 local mds1dev=$(mdsdevname 1)
3522 local mds1mnt=$(facet_mntpt mds1)
3523 local mds1opts=$MDS_MOUNT_OPTS
3525 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3526 ! do_facet mds1 test -b $mds1dev; then
3527 mds1opts=$(csa_add "$mds1opts" -o loop)
3529 if [[ "$mds1_FSTYPE" == zfs ]]; then
3530 import_zpool mds1 || return ${PIPESTATUS[0]}
3533 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3534 do_facet mds1 "$LCTL set_param fail_loc=0x80000716"
3536 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts &
3539 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts
3544 do_facet mds1 "$LCTL set_param fail_loc=0x0"
3545 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3546 echo "1st MDT start succeed"
3547 echo "2nd MDT start failed with $rc2"
3548 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3549 echo "1st MDT start failed with $rc"
3550 echo "2nd MDT start succeed"
3553 error "unexpected concurrent MDT mounts result, rc=$rc rc2=$rc2"
3556 if [ $MDSCOUNT -ge 2 ]; then
3557 for num in $(seq 2 $MDSCOUNT); do
3558 start_mdt $num || return
3562 # OST concurrent start
3564 do_rpc_nodes $oss_list "lsmod | grep -q libcfs" ||
3565 error "OST concurrent start: libcfs module not loaded"
3567 local ost1dev=$(ostdevname 1)
3568 local ost1mnt=$(facet_mntpt ost1)
3569 local ost1opts=$OST_MOUNT_OPTS
3571 if [ "$ost1_FSTYPE" == ldiskfs ] &&
3572 ! do_facet ost1 test -b $ost1dev; then
3573 ost1opts=$(csa_add "$ost1opts" -o loop)
3575 if [[ "$ost1_FSTYPE" == zfs ]]; then
3576 import_zpool ost1 || return ${PIPESTATUS[0]}
3579 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3580 do_facet ost1 "$LCTL set_param fail_loc=0x80000716"
3582 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts &
3585 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts
3589 do_facet ost1 "$LCTL set_param fail_loc=0x0"
3590 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3591 echo "1st OST start succeed"
3592 echo "2nd OST start failed with $rc2"
3593 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3594 echo "1st OST start failed with $rc"
3595 echo "2nd OST start succeed"
3599 error "unexpected concurrent OST mounts result, rc=$rc rc2=$rc2"
3605 # verify everything ok
3610 error "MDT(s) start failed"
3618 error "OST(s) start failed"
3626 error "client start failed"
3633 error "client mount failed"
3637 run_test 41c "concurrent mounts of MDT/OST should all fail but one"
3639 test_42() { #bug 14693
3643 check_mount || error "client was not mounted"
3645 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3646 PARAM="llite.$FSNAME-*.some_wrong_param"
3648 PARAM="$FSNAME.llite.some_wrong_param"
3651 do_facet mgs $PERM_CMD $PARAM=10
3652 umount_client $MOUNT ||
3653 error "unmounting client failed with invalid llite param"
3654 mount_client $MOUNT ||
3655 error "mounting client failed with invalid llite param"
3657 do_facet mgs $PERM_CMD $PARAM=20
3658 cleanup || error "stopping $FSNAME failed with invalid sys param"
3660 check_mount || error "client was not mounted with invalid sys param"
3661 cleanup || error "stopping $FSNAME failed with invalid sys param"
3663 run_test 42 "allow client/server mount/unmount with invalid config param"
3666 [[ "$MGS_VERSION" -ge $(version_code 2.5.58) ]] ||
3667 skip "Need MDS version at least 2.5.58"
3668 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
3671 USER1=$(getent passwd | grep :$ID1:$ID1: | cut -d: -f1)
3672 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1"
3675 chmod ugo+x $DIR || error "chmod 0 failed"
3676 set_persistent_param_and_check mds1 \
3677 "mdt.$FSNAME-MDT0000.root_squash" \
3678 "$FSNAME.mdt.root_squash" \
3680 wait_update $HOSTNAME \
3681 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3683 error "check llite root_squash failed!"
3684 set_persistent_param_and_check mds1 \
3685 "mdt.$FSNAME-MDT0000.nosquash_nids" \
3686 "$FSNAME.mdt.nosquash_nids" \
3688 wait_update $HOSTNAME \
3689 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3691 error "check llite nosquash_nids failed!"
3694 # create set of test files
3696 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
3697 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
3698 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
3700 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
3701 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
3703 mkdir_on_mdt0 $DIR/$tdir-rootdir || error "mkdir failed"
3704 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
3705 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
3707 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
3708 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
3709 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
3712 # check root_squash:
3713 # set root squash UID:GID to RUNAS_ID
3714 # root should be able to access only files owned by RUNAS_ID
3716 set_persistent_param_and_check mds1 \
3717 "mdt.$FSNAME-MDT0000.root_squash" \
3718 "$FSNAME.mdt.root_squash" \
3719 "$RUNAS_ID:$RUNAS_ID"
3720 wait_update $HOSTNAME \
3721 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3722 "$RUNAS_ID:$RUNAS_ID" ||
3723 error "check llite root_squash failed!"
3725 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
3726 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3727 error "$ST: root read permission is denied"
3728 echo "$ST: root read permission is granted - ok"
3731 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3732 error "$ST: root write permission is denied"
3733 echo "$ST: root write permission is granted - ok"
3735 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3736 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3737 error "$ST: root read permission is granted"
3738 echo "$ST: root read permission is denied - ok"
3741 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3742 error "$ST: root write permission is granted"
3743 echo "$ST: root write permission is denied - ok"
3745 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3746 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null &&
3747 error "$ST: root unlink permission is granted"
3748 echo "$ST: root unlink permission is denied - ok"
3750 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null &&
3751 error "$ST: root create permission is granted"
3752 echo "$ST: root create permission is denied - ok"
3755 # check root_squash is enforced independently
3756 # of client cache content
3758 # access file by USER1, keep access open
3759 # root should be denied access to user file
3761 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
3765 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
3766 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3767 { kill $pid; error "$ST: root read permission is granted"; }
3768 echo "$ST: root read permission is denied - ok"
3771 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3772 { kill $pid; error "$ST: root write permission is granted"; }
3773 echo "$ST: root write permission is denied - ok"
3779 # check nosquash_nids:
3780 # put client's NID into nosquash_nids list,
3781 # root should be able to access root file after that
3783 local NIDLIST=$($LCTL list_nids all | tr '\n' ' ')
3784 NIDLIST="2@gni $NIDLIST 192.168.0.[2,10]@tcp"
3785 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
3786 set_persistent_param_and_check mds1 \
3787 "mdt.$FSNAME-MDT0000.nosquash_nids" \
3788 "$FSNAME-MDTall.mdt.nosquash_nids" \
3790 wait_update $HOSTNAME \
3791 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3793 error "check llite nosquash_nids failed!"
3795 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3796 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3797 error "$ST: root read permission is denied"
3798 echo "$ST: root read permission is granted - ok"
3801 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3802 error "$ST: root write permission is denied"
3803 echo "$ST: root write permission is granted - ok"
3805 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3806 rm $DIR/$tdir-rootdir/tfile-1 ||
3807 error "$ST: root unlink permission is denied"
3808 echo "$ST: root unlink permission is granted - ok"
3809 touch $DIR/$tdir-rootdir/tfile-2 ||
3810 error "$ST: root create permission is denied"
3811 echo "$ST: root create permission is granted - ok"
3812 cleanup || error "cleanup failed with $?"
3814 run_test 43a "check root_squash and nosquash_nids"
3816 test_43b() { # LU-5690
3817 [[ "$MGS_VERSION" -ge $(version_code 2.7.62) ]] ||
3818 skip "Need MGS version 2.7.62+"
3820 if [[ -z "$fs2mds_DEV" ]]; then
3821 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
3822 skip_env "mixed loopback and real device not working"
3825 local fs2mdsdev=$(mdsdevname 1_2)
3826 local fs2mdsvdev=$(mdsvdevname 1_2)
3828 # temporarily use fs2mds as fs2mgs
3830 local fs2mgsdev=$fs2mdsdev
3831 local fs2mgsvdev=$fs2mdsvdev
3833 local fsname=test1234
3835 load_module llite/lustre
3836 local client_ip=$(host_nids_address $HOSTNAME $NETTYPE)
3837 local host=${client_ip//*./}
3838 local net=${client_ip/%$host/}
3839 local nosquash_nids=$(h2nettype $net[$host,$host,$host])
3841 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
3842 --param mdt.root_squash=$RUNAS_ID:$RUNAS_ID \
3843 --param mdt.nosquash_nids=$nosquash_nids \
3844 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
3845 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS || error "start fs2mgs failed"
3846 stop $fs2mgs -f || error "stop fs2mgs failed"
3847 cleanup || error "cleanup failed with $?"
3849 run_test 43b "parse nosquash_nids with commas in expr_list"
3853 check_mount || error "check_mount"
3854 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
3856 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
3857 for VAL in $UUIDS; do
3858 NID=$(echo $VAL | cut -d= -f1)
3859 CLUUID=$(echo $VAL | cut -d= -f2)
3860 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
3862 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
3863 cleanup || error "cleanup failed with $?"
3865 run_test 44 "mounted client proc entry exists"
3869 check_mount || error "check_mount"
3870 stop_mds || error "Unable to stop MDS"
3874 #define OBD_FAIL_PTLRPC_LONG_REPL_UNLINK 0x50f
3875 do_facet client "$LCTL set_param fail_loc=0x8000050f"
3878 manual_umount_client --force || error "manual_umount_client failed"
3879 do_facet client "$LCTL set_param fail_loc=0x0"
3880 start_mds || error "unable to start MDS"
3881 mount_client $MOUNT || error "mount_client $MOUNT failed"
3882 cleanup || error "cleanup failed with $?"
3884 run_test 45 "long unlink handling in ptlrpcd"
3891 umount_client $MOUNT2 || rc=$?
3892 umount_client $MOUNT || rc=$?
3893 while [ $count -gt 0 ]; do
3894 stop ost${count} -f || rc=$?
3898 cleanup_nocli || rc=$?
3899 #writeconf to remove all ost2 traces for subsequent tests
3900 writeconf_or_reformat
3905 echo "Testing with $OSTCOUNT OSTs"
3907 start_mds || error "unable to start MDS"
3908 #first client should see only one ost
3909 start_ost || error "Unable to start OST1"
3910 wait_osc_import_state mds ost FULL
3912 mount_client $MOUNT || error "mount_client $MOUNT failed"
3913 trap "cleanup_46a $OSTCOUNT" EXIT ERR
3916 for (( i=2; i<=$OSTCOUNT; i++ )); do
3917 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
3918 error "start_ost$i $(ostdevname $i) failed"
3921 # wait until osts in sync
3922 for (( i=2; i<=$OSTCOUNT; i++ )); do
3923 wait_osc_import_state mds ost$i FULL
3924 wait_osc_import_ready client ost$i
3927 #second client see all ost's
3929 mount_client $MOUNT2 || error "mount_client failed"
3930 $LFS setstripe -c -1 $MOUNT2 ||
3931 error "$LFS setstripe -c -1 $MOUNT2 failed"
3932 $LFS getstripe $MOUNT2 || error "$LFS getstripe $MOUNT2 failed"
3934 echo "ok" > $MOUNT2/widestripe
3935 $LFS getstripe $MOUNT2/widestripe ||
3936 error "$LFS getstripe $MOUNT2/widestripe failed"
3937 # fill acl buffer for avoid expand lsm to them
3938 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd |
3940 setfacl -m $acl $MOUNT2/widestripe
3944 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3946 cleanup_46a $OSTCOUNT || error "cleanup_46a failed"
3948 run_test 46a "handle ost additional - wide striped file"
3953 check_mount || error "check_mount failed"
3954 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
3958 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3959 if echo $ns | grep "MDT[[:digit:]]*"; then
3962 lrs=$(echo $ns | sed 's/.*lru_size=//')
3963 lru_size[count]=$lrs
3968 facet_failover $SINGLEMDS
3969 client_up || error "client_up failed"
3972 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3973 if echo $ns | grep "MDT[[:digit:]]*"; then
3976 lrs=$(echo $ns | sed 's/.*lru_size=//')
3977 if ! test "$lrs" -eq "${lru_size[count]}"; then
3978 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
3979 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
3984 cleanup || error "cleanup failed with $?"
3986 run_test 47 "server restart does not make client loss lru_resize settings"
3991 # reformat after this test is needed - if the test fails,
3992 # we will have unkillable file at FS
3996 test_48() { # bz-17636 LU-7473
4000 check_mount || error "check_mount failed"
4002 $LFS setstripe -c -1 $MOUNT ||
4003 error "$LFS setstripe -c -1 $MOUNT failed"
4004 $LFS getstripe $MOUNT || error "$LFS getstripe $MOUNT failed"
4006 echo "ok" > $MOUNT/widestripe
4007 $LFS getstripe $MOUNT/widestripe ||
4008 error "$LFS getstripe $MOUNT/widestripe failed"
4010 # In the future, we may introduce more EAs, such as selinux, enlarged
4011 # LOV EA, and so on. These EA will use some EA space that is shared by
4012 # ACL entries. So here we only check some reasonable ACL entries count,
4013 # instead of the max number that is calculated from the max_ea_size.
4014 if [ "$MDS1_VERSION" -lt $(version_code 2.8.57) ]; then
4015 count=28 # hard coded of RPC protocol
4016 elif large_xattr_enabled; then
4017 count=4500 # max_num 8187 max_ea_size = 65452
4018 # not create too many (4500) to save test time
4020 count=450 # max_num 497 max_ea_size = 4012
4023 echo "It is expected to hold at least $count ACL entries"
4024 trap cleanup_48 EXIT ERR
4025 for ((i = 0; i < $count; i++)) do
4026 setfacl -m u:$((i + 100)):rw $MOUNT/widestripe ||
4027 error "Fail to setfacl for $MOUNT/widestripe at $i"
4030 cancel_lru_locks mdc
4031 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
4032 local r_count=$(getfacl $MOUNT/widestripe | grep "user:" | wc -l)
4034 count=$((count + 1)) # for the entry "user::rw-"
4036 [ $count -eq $r_count ] ||
4037 error "Expected ACL entries $count, but got $r_count"
4041 run_test 48 "too many acls on file"
4043 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
4044 test_49a() { # bug 17710
4045 local timeout_orig=$TIMEOUT
4046 local ldlm_timeout_orig=$LDLM_TIMEOUT
4047 local LOCAL_TIMEOUT=20
4049 LDLM_TIMEOUT=$LOCAL_TIMEOUT
4050 TIMEOUT=$LOCAL_TIMEOUT
4054 check_mount || error "client mount failed"
4056 echo "check ldlm_timout..."
4057 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
4058 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
4059 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
4061 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
4062 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
4065 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
4066 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
4069 umount_client $MOUNT || error "umount_client $MOUNT failed"
4070 stop_ost || error "problem stopping OSS"
4071 stop_mds || error "problem stopping MDS"
4073 LDLM_TIMEOUT=$ldlm_timeout_orig
4074 TIMEOUT=$timeout_orig
4076 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
4078 test_49b() { # bug 17710
4079 local timeout_orig=$TIMEOUT
4080 local ldlm_timeout_orig=$LDLM_TIMEOUT
4081 local LOCAL_TIMEOUT=20
4083 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
4084 TIMEOUT=$LOCAL_TIMEOUT
4088 check_mount || error "client mount failed"
4090 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
4091 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
4092 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
4094 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
4095 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
4098 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
4099 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
4102 cleanup || error "cleanup failed"
4104 LDLM_TIMEOUT=$ldlm_timeout_orig
4105 TIMEOUT=$timeout_orig
4107 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
4110 # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1
4112 # Test both statfs and lfs df and fail if either one fails
4113 multiop_bg_pause $1 f_
4116 killall -USR1 multiop
4117 [ $RC -ne 0 ] && log "lazystatfs multiop failed"
4118 wait $PID || { RC=$?; log "multiop return error "; }
4120 # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1
4125 if kill -s 0 $PID; then
4128 log "lazystatfs lfs df failed to complete in 5s"
4136 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4137 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4139 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
4141 cleanup || error "cleanup failed with rc $?"
4143 run_test 50a "lazystatfs all servers available"
4147 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4148 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4150 # Wait for client to detect down OST
4151 stop_ost || error "Unable to stop OST1"
4152 wait_osc_import_state client ost DISCONN
4154 log "OSCs should all be DISCONN"
4156 lazystatfs $MOUNT || error "lazystatfs should not return EIO"
4158 umount_client $MOUNT || error "Unable to unmount client"
4159 stop_mds || error "Unable to stop MDS"
4161 run_test 50b "lazystatfs all servers down"
4164 start_mds || error "Unable to start MDS"
4165 start_ost || error "Unable to start OST1"
4166 start_ost2 || error "Unable to start OST2"
4167 mount_client $MOUNT || error "Unable to mount client"
4168 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4169 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4171 # Wait for client to detect down OST
4172 stop_ost || error "Unable to stop OST1"
4173 wait_osc_import_state mds ost DISCONN
4174 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
4176 umount_client $MOUNT || error "Unable to unmount client"
4177 stop_ost2 || error "Unable to stop OST2"
4178 stop_mds || error "Unable to stop MDS"
4179 #writeconf to remove all ost2 traces for subsequent tests
4180 writeconf_or_reformat
4182 run_test 50c "lazystatfs one server down"
4185 start_mds || error "Unable to start MDS"
4186 start_ost || error "Unable to start OST1"
4187 start_ost2 || error "Unable to start OST2"
4188 mount_client $MOUNT || error "Unable to mount client"
4189 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4190 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4192 # Issue the statfs during the window where the client still
4193 # belives the OST to be available but it is in fact down.
4194 # No failure just a statfs which hangs for a timeout interval.
4195 stop_ost || error "Unable to stop OST1"
4196 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
4198 umount_client $MOUNT || error "Unable to unmount client"
4199 stop_ost2 || error "Unable to stop OST2"
4200 stop_mds || error "Unable to stop MDS"
4201 #writeconf to remove all ost2 traces for subsequent tests
4202 writeconf_or_reformat
4204 run_test 50d "lazystatfs client/server conn race"
4211 start_mds || error "Unable to start MDS"
4212 #first client should see only one ost
4213 start_ost || error "Unable to start OST1"
4214 wait_osc_import_state mds ost FULL
4216 # Wait for client to detect down OST
4217 stop_ost || error "Unable to stop OST1"
4218 wait_osc_import_state mds ost DISCONN
4220 mount_client $MOUNT || error "Unable to mount client"
4221 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
4223 multiop_bg_pause $MOUNT _f
4227 if [ $RC1 -ne 0 ]; then
4228 log "multiop failed $RC1"
4231 sleep $(( $TIMEOUT+1 ))
4233 [ $? -ne 0 ] && error "process isn't sleep"
4234 start_ost || error "Unable to start OST1"
4235 wait $pid || error "statfs failed"
4238 umount_client $MOUNT || error "Unable to unmount client"
4239 stop_ost || error "Unable to stop OST1"
4240 stop_mds || error "Unable to stop MDS"
4242 run_test 50e "normal statfs all servers down"
4247 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
4249 start_mds || error "Unable to start mds"
4250 #first client should see only one ost
4251 start_ost || error "Unable to start OST1"
4252 wait_osc_import_state mds ost FULL
4254 start_ost2 || error "Unable to start OST2"
4255 wait_osc_import_state mds ost2 FULL
4257 # Wait for client to detect down OST
4258 stop_ost2 || error "Unable to stop OST2"
4260 wait_osc_import_state mds ost2 DISCONN
4261 mount_client $MOUNT || error "Unable to mount client"
4262 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
4264 multiop_bg_pause $MOUNT _f
4268 if [ $RC1 -ne 0 ]; then
4269 log "lazystatfs multiop failed $RC1"
4272 sleep $(( $TIMEOUT+1 ))
4274 [ $? -ne 0 ] && error "process isn't sleep"
4275 start_ost2 || error "Unable to start OST2"
4276 wait $pid || error "statfs failed"
4277 stop_ost2 || error "Unable to stop OST2"
4280 umount_client $MOUNT -f || error "Unable to unmount client"
4281 stop_ost || error "Unable to stop OST1"
4282 stop_mds || error "Unable to stop MDS"
4283 #writeconf to remove all ost2 traces for subsequent tests
4284 writeconf_or_reformat
4286 run_test 50f "normal statfs one server in down"
4289 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs"
4291 start_ost2 || error "Unable to start OST2"
4292 wait_osc_import_state mds ost2 FULL
4293 wait_osc_import_ready client ost2
4295 if [[ $PERM_CMD == *"set_param -P"* ]]; then
4296 local PARAM="osc.${FSNAME}-OST0001*.active"
4298 local PARAM="${FSNAME}-OST0001.osc.active"
4301 $LFS setstripe -c -1 $DIR/$tfile || error "$LFS setstripe failed"
4302 do_facet mgs $PERM_CMD $PARAM=0 || error "Unable to deactivate OST"
4304 umount_client $MOUNT || error "Unable to unmount client"
4305 mount_client $MOUNT || error "Unable to mount client"
4306 # This df should not cause a panic
4309 do_facet mgs $PERM_CMD $PARAM=1 || error "Unable to activate OST"
4310 rm -f $DIR/$tfile || error "unable to remove file $DIR/$tfile"
4311 umount_client $MOUNT || error "Unable to unmount client"
4312 stop_ost2 || error "Unable to stop OST2"
4313 stop_ost || error "Unable to stop OST1"
4314 stop_mds || error "Unable to stop MDS"
4315 #writeconf to remove all ost2 traces for subsequent tests
4316 writeconf_or_reformat
4318 run_test 50g "deactivated OST should not cause panic"
4322 # prepare MDT/OST, make OSC inactive for OST1
4323 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs"
4325 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
4326 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
4327 error "tunefs OST1 failed"
4328 start_mds || error "Unable to start MDT"
4329 start_ost || error "Unable to start OST1"
4330 start_ost2 || error "Unable to start OST2"
4331 mount_client $MOUNT || error "client start failed"
4333 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4335 # activatate OSC for OST1
4336 set_persistent_param_and_check client \
4337 "osc.${FSNAME}-OST0000-osc-[!M]*.active" \
4338 "${FSNAME}-OST0000.osc.active" 1
4340 mkdir $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
4341 $LFS setstripe -c -1 -i 0 $DIR/$tdir/2 ||
4342 error "$LFS setstripe $DIR/$tdir/2 failed"
4343 sleep 1 && echo "create a file after OST1 is activated"
4344 # doing some io, shouldn't crash
4345 dd if=/dev/zero of=$DIR/$tdir/2/$tfile-io bs=1M count=10
4347 # check OSC import is working
4348 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
4349 error "some OSC imports are still not connected"
4353 umount_client $MOUNT || error "Unable to umount client"
4354 stop_ost2 || error "Unable to stop OST2"
4355 cleanup_nocli || error "cleanup_nocli failed with $?"
4357 run_test 50h "LU-642: activate deactivated OST"
4360 # prepare MDT/OST, make OSC inactive for OST1
4361 [ "$MDSCOUNT" -lt "2" ] && skip "needs >= 2 MDTs"
4364 [ $(facet_fstype mds2) == zfs ] && import_zpool mds2
4365 do_facet mds2 "$TUNEFS --param mdc.active=0 $(mdsdevname 2)" ||
4366 error "tunefs MDT2 failed"
4367 start_mds || error "Unable to start MDT"
4368 start_ost || error "Unable to start OST1"
4369 start_ost2 || error "Unable to start OST2"
4370 mount_client $MOUNT || error "client start failed"
4372 mkdir_on_mdt0 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4374 if [[ $PERM_CMD == *"set_param -P"* ]]; then
4375 $PERM_CMD mdc.${FSNAME}-MDT0001-mdc-*.active=0 &&
4376 error "deactive MDC0 succeeds"
4378 $PERM_CMD ${FSNAME}-MDT0000.mdc.active=0 &&
4379 error "deactive MDC0 succeeds"
4382 # activate MDC for MDT2
4383 set_persistent_param_and_check client \
4384 "mdc.${FSNAME}-MDT0001-mdc-*.active" \
4385 "${FSNAME}-MDT0001.mdc.active" 1
4387 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
4388 if [ "$MDS1_VERSION" -ge $(version_code 2.7.60) ]
4390 wait_dne_interconnect
4392 $LFS mkdir -i1 $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
4394 createmany -o $DIR/$tdir/2/$tfile-%d 1 || error "create files failed"
4396 rm -rf $DIR/$tdir/2 || error "unlink dir failed"
4398 # deactivate MDC for MDT2
4399 set_persistent_param_and_check client \
4400 "mdc.${FSNAME}-MDT0001-mdc-*.active" \
4401 "${FSNAME}-MDT0001.mdc.active" 0
4403 wait_osp_active mds ${FSNAME}-MDT0001 1 0
4405 $LFS mkdir -i1 $DIR/$tdir/2 &&
4406 error "mkdir $DIR/$tdir/2 succeeds after deactive MDT"
4408 $LFS mkdir -i0 -c$MDSCOUNT $DIR/$tdir/striped_dir ||
4409 error "mkdir $DIR/$tdir/striped_dir fails after deactive MDT2"
4411 local stripe_count=$($LFS getdirstripe -c $DIR/$tdir/striped_dir)
4412 [ $stripe_count -eq $((MDSCOUNT - 1)) ] ||
4413 error "wrong $stripe_count != $((MDSCOUNT -1)) for striped_dir"
4416 umount_client $MOUNT || error "Unable to umount client"
4421 run_test 50i "activate deactivated MDT"
4424 local LOCAL_TIMEOUT=20
4428 check_mount || error "check_mount failed"
4430 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
4431 $LFS setstripe -c -1 $MOUNT/$tdir ||
4432 error "$LFS setstripe -c -1 $MOUNT/$tdir failed"
4433 #define OBD_FAIL_MDS_REINT_DELAY 0x142
4434 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x142"
4435 touch $MOUNT/$tdir/$tfile &
4438 start_ost2 || error "Unable to start OST1"
4440 stop_ost2 || error "Unable to stop OST1"
4441 umount_client $MOUNT -f || error “unmount $MOUNT failed”
4442 cleanup_nocli || error “stop server failed”
4443 #writeconf to remove all ost2 traces for subsequent tests
4444 writeconf_or_reformat
4446 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
4455 do_node $node mkdir -p $dest
4456 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
4458 do_node $node 'tar cf - '$*' | tar xf - -C '$dest';
4459 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
4460 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
4462 do_node $node 'getfattr -d -m "[a-z]*\\." '$*' > '$xattrs
4463 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
4473 local backup2=${TMP}/backup2
4475 do_node $node mkdir -p $backup2
4476 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
4478 do_node $node 'tar cf - '$*' | tar xf - -C '$backup2';
4479 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
4480 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
4482 do_node $node "diff -rq $backup $backup2"
4483 [ $? -eq 0 ] || { error "contents differ"; return 3; }
4485 local xattrs2=${TMP}/xattrs2
4486 do_node $node 'getfattr -d -m "[a-z]*\\." '$*' > '$xattrs2
4487 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
4489 do_node $node "diff $xattrs $xattrs2"
4490 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
4492 do_node $node "rm -rf $backup2 $xattrs2"
4493 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
4497 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4498 skip "ldiskfs only test"
4501 start_mds || error "Unable to start MDS"
4502 start_ost || error "Unable to start OST1"
4503 mount_client $MOUNT || error "Unable to mount client"
4506 local ost1mnt=$(facet_mntpt ost1)
4507 local ost1node=$(facet_active_host ost1)
4508 local ost1tmp=$TMP/conf52
4511 mkdir $DIR/$tdir || error "Unable to create $DIR/$tdir"
4512 touch $TMP/modified_first || error "Unable to create temporary file"
4513 local mtime=$(stat -c %Y $TMP/modified_first)
4514 do_node $ost1node "mkdir -p $ost1tmp &&
4515 touch -m -d @$mtime $ost1tmp/modified_first" ||
4516 error "Unable to create temporary file"
4519 $LFS setstripe -c -1 -S 1M $DIR/$tdir || error "$LFS setstripe failed"
4521 for (( i=0; i < nrfiles; i++ )); do
4522 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c ||
4523 error "multiop failed"
4528 # sync all the data and make sure no pending data on the client,
4529 # thus the SOM xattr would not be changed any more.
4530 cancel_lru_locks osc
4533 echo backup files to $TMP/$tdir
4534 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
4535 copy_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
4536 error "Unable to copy files"
4538 umount_client $MOUNT || error "Unable to umount client"
4539 stop_ost || error "Unable to stop ost1"
4541 echo mount ost1 as ldiskfs
4542 do_node $ost1node mkdir -p $ost1mnt || error "Unable to create $ost1mnt"
4543 if ! do_node $ost1node test -b $ost1_dev; then
4546 do_node $ost1node mount -t "$ost1_FSTYPE" $loop $ost1_dev \
4548 error "Unable to mount ost1 as ldiskfs"
4551 echo backup objects to $ost1tmp/objects
4552 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
4553 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
4554 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
4556 error "Unable to copy objects"
4558 # move objects to lost+found
4559 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
4560 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
4562 do_node $ost1node "umount $ost1mnt" ||
4563 error "Unable to umount ost1 as ldiskfs"
4565 start_ost || error "Unable to start OST1"
4566 mount_client $MOUNT || error "Unable to mount client"
4568 local REPAIRED=$(do_node $ost1node "$LCTL get_param \
4569 -n osd-ldiskfs.$FSNAME-OST0000.oi_scrub" |
4570 awk '/^lf_repa[ri]*ed/ { print $2 }')
4571 [ $REPAIRED -gt 0 ] ||
4572 error "Some entry under /lost+found should be repaired"
4575 diff_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
4576 error "Unable to diff files"
4578 rm -rf $TMP/$tdir $TMP/file_xattrs ||
4579 error "Unable to delete temporary files"
4580 do_node $ost1node "rm -rf $ost1tmp" ||
4581 error "Unable to delete temporary files"
4582 cleanup || error "cleanup failed with $?"
4584 run_test 52 "check recovering objects from lost+found"
4586 # Checks threads_min/max/started for some service
4588 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
4589 # parameter pattern prefix like 'ost.*.ost'.
4602 local msg="Insane $modname thread counts"
4603 local ncpts=$(check_cpt_number $facet)
4607 check_mount || return 41
4609 # We need to expand $parampat, but it may match multiple parameters, so
4610 # we'll pick the first one
4611 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
4612 error "Couldn't expand ${parampat}.threads_min parameter name"
4616 # Remove the .threads_min part
4617 paramp=${paramp%.threads_min}
4619 # Check for sanity in defaults
4620 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4622 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4624 tstarted=$(do_facet $facet "$LCTL get_param \
4625 -n ${paramp}.threads_started" || echo 0)
4626 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' ||
4628 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4630 nthrs=$(expr $tmax - $tmin)
4631 if [ $nthrs -lt $ncpts ]; then
4637 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
4638 skip_env "module parameter forced $facet thread count"
4640 # Check that we can change min/max
4641 do_facet $facet "$LCTL set_param \
4642 ${paramp}.threads_min=$((tmin + nthrs))"
4643 do_facet $facet "$LCTL set_param \
4644 ${paramp}.threads_max=$((tmax - nthrs))"
4645 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4647 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4649 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) &&
4650 $tmax2 == ($tmax - $nthrs)))' || return $?
4652 # Check that we can set min/max to the same value
4653 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4655 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$tmin"
4656 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4658 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4660 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
4662 # Check that we can't set max < min
4663 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$((tmin - 1))"
4664 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4666 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4668 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
4670 # We need to ensure that we get the module options desired; to do this
4671 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
4672 LOAD_MODULES_REMOTE=true
4675 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
4677 setmodopts -a $modname "$newvalue" oldvalue
4680 check_mount || return 41
4682 # Restore previous setting of MODOPTS_*
4683 setmodopts $modname "$oldvalue"
4685 # Check that $opts took
4686 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4688 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4690 tstarted=$(do_facet $facet \
4691 "$LCTL get_param -n ${paramp}.threads_started" || echo 0)
4692 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4701 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
4702 cleanup || error "cleanup failed with rc $?"
4704 run_test 53a "check OSS thread count params"
4708 local mds=$(do_facet $SINGLEMDS "$LCTL get_param \
4709 -N mds.*.*.threads_max 2>/dev/null")
4710 if [ -z "$mds" ]; then
4711 #running this on an old MDT
4712 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
4714 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
4716 cleanup || error "cleanup failed with $?"
4718 run_test 53b "check MDS thread count params"
4721 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4722 skip "ldiskfs only test"
4725 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p ||
4726 error "llverdev failed with rc=$?"
4729 run_test 54a "test llverdev and partial verify of device"
4732 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4733 skip "ldiskfs only test"
4737 run_llverfs $MOUNT -p || error "llverfs failed with rc=$?"
4738 cleanup || error "cleanup failed with rc=$?"
4740 run_test 54b "test llverfs and partial verify of filesystem"
4744 local max_ost_index=$1
4746 echo -n $(((max_ost_index + 1) * 8))
4750 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4751 skip "ldiskfs only test"
4754 local mdsdev=$(mdsdevname 1)
4755 local mdsvdev=$(mdsvdevname 1)
4759 if ! combined_mgs_mds; then
4760 stop_mgs || error "stopping MGS service failed"
4761 format_mgs || error "formatting MGT failed"
4763 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
4765 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
4766 --reformat $(ostdevname 1) $(ostvdevname 1)
4772 echo checking size of lov_objid for ost index $i
4773 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" |
4774 grep ^User | awk -F 'Size: ' '{print $2}')
4775 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
4776 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
4778 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
4785 run_test 55 "check lov_objid size"
4788 local mds_journal_size_orig=$MDSJOURNALSIZE
4794 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
4795 $(ostdevname 1) $(ostvdevname 1)
4796 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
4797 $(ostdevname 2) $(ostvdevname 2)
4800 start_ost || error "Unable to start first ost (idx 10000)"
4801 start_ost2 || error "Unable to start second ost (idx 1000)"
4802 mount_client $MOUNT || error "Unable to mount client"
4806 # test instantiating PFL components with sparse index LU-15513
4807 mkdir -p $MOUNT/$tdir
4808 $LFS setstripe -E 4M -c 1 -E 1G -c 4 -S4M -E eof -c -1 $MOUNT/$tdir
4809 dd if=/dev/zero of=$MOUNT/$tdir/$tfile bs=4K count=1 seek=10k ||
4810 error "dd to second component failed"
4812 if [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
4813 [[ "$MDS1_VERSION" -ge $(version_code 2.5.4) &&
4814 "$MDS1_VERSION" -lt $(version_code 2.5.11) ]]; then
4815 wait_osc_import_state mds ost1 FULL
4816 wait_osc_import_state mds ost2 FULL
4817 $LFS setstripe --stripe-count=-1 $DIR/$tfile ||
4818 error "Unable to setstripe $DIR/$tfile"
4819 n=$($LFS getstripe --stripe-count $DIR/$tfile)
4820 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
4825 MDSJOURNALSIZE=$mds_journal_size_orig
4828 run_test 56a "check big OST indexes and out-of-index-order start"
4833 umount_client $MOUNT -f || error "unmount client failed"
4842 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs"
4844 trap cleanup_56b EXIT RETURN ERR
4847 if ! combined_mgs_mds ; then
4852 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) --index=0 --reformat \
4853 $(mdsdevname 1) $(mdsvdevname 1)
4854 add mds2 $(mkfs_opts mds2 $(mdsdevname 2)) --index=1 --reformat \
4855 $(mdsdevname 2) $(mdsvdevname 2)
4856 add mds3 $(mkfs_opts mds3 $(mdsdevname 3)) --index=1000 --reformat \
4857 $(mdsdevname 3) $(mdsvdevname 3)
4861 start_mdt 1 || error "MDT 1 (idx 0) start failed"
4862 start_mdt 2 || error "MDT 2 (idx 1) start failed"
4863 start_mdt 3 || error "MDT 3 (idx 1000) start failed"
4864 start_ost || error "Unable to start first ost"
4865 start_ost2 || error "Unable to start second ost"
4867 do_nodes $(comma_list $(mdts_nodes)) \
4868 "$LCTL set_param mdt.*.enable_remote_dir=1 \
4869 mdt.*.enable_remote_dir_gid=-1"
4871 mount_client $MOUNT || error "Unable to mount client"
4873 $LFS mkdir -c3 $MOUNT/$tdir || error "failed to make testdir"
4875 echo "This is test file 1!" > $MOUNT/$tdir/$tfile.1 ||
4876 error "failed to make test file 1"
4877 echo "This is test file 2!" > $MOUNT/$tdir/$tfile.2 ||
4878 error "failed to make test file 2"
4879 echo "This is test file 1000!" > $MOUNT/$tdir/$tfile.1000 ||
4880 error "failed to make test file 1000"
4882 rm -rf $MOUNT/$tdir || error "failed to remove testdir"
4884 $LFS mkdir -i1000 $MOUNT/$tdir.1000 ||
4885 error "create remote dir at idx 1000 failed"
4888 echo "=== START lfs df OUTPUT ==="
4890 echo "==== END lfs df OUTPUT ===="
4892 mdtcnt=$(echo -e "$output" | grep $FSNAME-MDT | wc -l)
4893 ostcnt=$(echo -e "$output" | grep $FSNAME-OST | wc -l)
4895 echo "lfs df returned mdt count $mdtcnt and ost count $ostcnt"
4896 [ $mdtcnt -eq 3 ] || error "lfs df returned wrong mdt count"
4897 [ $ostcnt -eq 2 ] || error "lfs df returned wrong ost count"
4899 echo "This is test file 1!" > $MOUNT/$tdir.1000/$tfile.1 ||
4900 error "failed to make test file 1"
4901 echo "This is test file 2!" > $MOUNT/$tdir.1000/$tfile.2 ||
4902 error "failed to make test file 2"
4903 echo "This is test file 1000!" > $MOUNT/$tdir.1000/$tfile.1000 ||
4904 error "failed to make test file 1000"
4905 rm -rf $MOUNT/$tdir.1000 || error "failed to remove remote_dir"
4908 echo "=== START lfs mdts OUTPUT ==="
4910 echo "==== END lfs mdts OUTPUT ===="
4912 echo -e "$output" | grep -v "MDTS:" | awk '{print $1}' |
4913 sed 's/://g' > $TMP/mdts-actual.txt
4914 sort $TMP/mdts-actual.txt -o $TMP/mdts-actual.txt
4916 echo -e "0\n1\n1000" > $TMP/mdts-expected.txt
4918 diff $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4921 rm $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4923 [ $result -eq 0 ] || error "target_obd proc file is incorrect!"
4925 run_test 56b "test target_obd correctness with nonconsecutive MDTs"
4927 test_57a() { # bug 22656
4928 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4929 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4930 tail -1 | awk '{print $1}')
4931 writeconf_or_reformat
4932 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
4933 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" ||
4934 error "tunefs failed"
4936 start_ost && error "OST registration from failnode should fail"
4939 run_test 57a "initial registration from failnode should fail (should return errs)"
4942 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4943 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4944 tail -1 | awk '{print $1}')
4945 writeconf_or_reformat
4946 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
4947 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" ||
4948 error "tunefs failed"
4950 start_ost || error "OST registration from servicenode should not fail"
4953 run_test 57b "initial registration from servicenode should not fail"
4956 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
4959 test_58() { # bug 22658
4960 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
4962 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4963 createmany -o $DIR/$tdir/$tfile-%d 100
4964 unlinkmany $DIR/$tdir/$tfile-%d 100
4965 stop_mds || error "Unable to stop MDS"
4967 local MNTDIR=$(facet_mntpt $SINGLEMDS)
4968 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4970 # remove all files from the OBJECTS dir
4971 mount_fstype $SINGLEMDS
4973 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
4975 unmount_fstype $SINGLEMDS
4976 # restart MDS with missing llog files
4977 start_mds || error "unable to start MDS"
4978 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0"
4981 run_test 58 "missing llog files must not prevent MDT from mounting"
4984 start_mgsmds >> /dev/null
4985 local C1=$(count_osts)
4986 if [ $C1 -eq 0 ]; then
4987 start_ost >> /dev/null
4991 echo "original ost count: $C1 (expect > 0)"
4992 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
4993 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
4994 local C2=$(count_osts)
4995 echo "after mdt writeconf count: $C2 (expect 0)"
4996 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
4997 echo "OST start without writeconf should fail:"
4998 start_ost >> /dev/null &&
4999 error "OST start without writeconf didn't fail"
5000 echo "OST start with writeconf should succeed:"
5001 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
5002 local C3=$(count_osts)
5003 echo "after ost writeconf count: $C3 (expect 1)"
5004 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
5005 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
5006 local C4=$(count_osts)
5007 echo "after ost2 writeconf count: $C4 (expect 2)"
5008 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
5009 stop_ost2 >> /dev/null
5010 cleanup_nocli >> /dev/null
5011 #writeconf to remove all ost2 traces for subsequent tests
5012 writeconf_or_reformat
5014 run_test 59 "writeconf mount option"
5016 test_60() { # LU-471
5017 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5018 skip "ldiskfs only test"
5023 for num in $(seq $MDSCOUNT); do
5024 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
5025 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
5026 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
5030 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
5031 [ ${PIPESTATUS[0]} -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
5033 # MDT default has dirdata feature
5034 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
5035 # we disable uninit_bg feature
5036 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
5037 # we set stride extended options
5038 echo $dump | grep stride > /dev/null || error "stride is not set"
5042 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
5045 local lxattr=$(large_xattr_enabled)
5047 [ "$MDS1_VERSION" -ge $(version_code 2.1.53) ] ||
5048 skip "Need MDS version at least 2.1.53"
5050 if [ "$mds1_FSTYPE" == ldiskfs ] &&
5051 ! large_xattr_enabled; then
5054 for num in $(seq $MDSCOUNT); do
5055 do_facet mds${num} $TUNE2FS -O ea_inode \
5056 $(mdsdevname $num) ||
5057 error "tune2fs on mds $num failed"
5061 setup || error "setting up the filesystem failed"
5062 client_up || error "starting client failed"
5064 local file=$DIR/$tfile
5065 touch $file || error "touch $file failed"
5067 local large_value="$(generate_string $(max_xattr_size))"
5068 local small_value="bar"
5070 local name="trusted.big"
5071 log "save large xattr of $(max_xattr_size) bytes on $name on $file"
5072 setfattr -n $name -v $large_value $file ||
5073 error "saving $name on $file failed"
5075 local new_value=$(get_xattr_value $name $file)
5076 [[ "$new_value" != "$large_value" ]] &&
5077 error "$name different after saving"
5079 log "shrink value of $name on $file"
5080 setfattr -n $name -v $small_value $file ||
5081 error "shrinking value of $name on $file failed"
5083 new_value=$(get_xattr_value $name $file)
5084 [[ "$new_value" != "$small_value" ]] &&
5085 error "$name different after shrinking"
5087 log "grow value of $name on $file"
5088 setfattr -n $name -v $large_value $file ||
5089 error "growing value of $name on $file failed"
5091 new_value=$(get_xattr_value $name $file)
5092 [[ "$new_value" != "$large_value" ]] &&
5093 error "$name different after growing"
5095 log "check value of $name on $file after remounting MDS"
5097 new_value=$(get_xattr_value $name $file)
5098 [[ "$new_value" != "$large_value" ]] &&
5099 error "$name different after remounting MDS"
5101 log "remove large xattr $name from $file"
5102 setfattr -x $name $file || error "removing $name from $file failed"
5104 if $lxattr && [ "$mds1_FSTYPE" == ldiskfs ]; then
5105 stopall || error "stopping for e2fsck run"
5106 for num in $(seq $MDSCOUNT); do
5107 run_e2fsck $(facet_active_host mds$num) \
5108 $(mdsdevname $num) "-y" ||
5109 error "e2fsck MDT$num failed"
5111 setup_noconfig || error "remounting the filesystem failed"
5114 # need to delete this file to avoid problems in other tests
5116 cleanup || error "stopping systems failed"
5118 run_test 61 "large xattr"
5121 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5122 skip "ldiskfs only test"
5124 [[ "$MDS1_VERSION" -ge $(version_code 2.2.51) ]] ||
5125 skip "Need MDS version at least 2.2.51"
5128 local mdsdev=$(mdsdevname 1)
5129 local ostdev=$(ostdevname 1)
5131 echo "disable journal for mds"
5132 do_facet mds1 $TUNE2FS -O ^has_journal $mdsdev || error "tune2fs failed"
5133 start_mds && error "MDT start should fail"
5134 echo "disable journal for ost"
5135 do_facet ost1 $TUNE2FS -O ^has_journal $ostdev || error "tune2fs failed"
5136 start_ost && error "OST start should fail"
5137 cleanup || error "cleanup failed with rc $?"
5140 run_test 62 "start with disabled journal"
5143 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5144 skip "ldiskfs only test"
5147 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ldiskfs
5148 local inode_slab=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
5149 awk '/ldiskfs_inode_cache/ { print $5 / $6 }')
5150 if [ -z "$inode_slab" ]; then
5151 skip "ldiskfs module has not been loaded"
5154 if grep -q "CONFIG_DEBUG_LOCK_ALLOC=y" /boot/config-$(uname -r); then
5155 skip "test is not compatible with CONFIG_DEBUG_LOCK_ALLOC=y"
5158 echo "$inode_slab ldiskfs inodes per page"
5159 [ "${inode_slab%.*}" -ge "3" ] && return 0
5161 # If kmalloc-128 is also 1 per page - this is a debug kernel
5162 # and so this is not an error.
5163 local kmalloc128=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
5164 awk '/^(dma-kmalloc|size)-128 / { print $5 / $6 }')
5165 # 32 128-byte chunks in 4k
5166 [ "${kmalloc128%.*}" -lt "32" ] ||
5167 error "ldiskfs inode too big, only $inode_slab objs/page, " \
5168 "kmalloc128 = $kmalloc128 objs/page"
5170 run_test 63 "Verify each page can at least hold 3 ldiskfs inodes"
5173 start_mds || error "unable to start MDS"
5174 start_ost || error "Unable to start OST1"
5175 start_ost2 || error "Unable to start second ost"
5176 mount_client $MOUNT || error "Unable to mount client"
5177 stop_ost2 || error "Unable to stop second ost"
5180 umount_client $MOUNT -f || error “unmount $MOUNT failed”
5181 cleanup_nocli || error "cleanup_nocli failed with $?"
5182 #writeconf to remove all ost2 traces for subsequent tests
5183 writeconf_or_reformat
5185 run_test 64 "check lfs df --lazy "
5187 test_65() { # LU-2237
5188 # Currently, the test is only valid for ldiskfs backend
5189 [ "$mds1_FSTYPE" != ldiskfs ] &&
5190 skip "ldiskfs only test"
5192 local devname=$(mdsdevname ${SINGLEMDS//mds/})
5193 local brpt=$(facet_mntpt brpt)
5196 if ! do_facet $SINGLEMDS "test -b $devname"; then
5200 stop_mds || error "Unable to stop MDS"
5201 local obj=$(do_facet $SINGLEMDS \
5202 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
5204 if [ -z "$obj" ]; then
5205 # The MDT may be just re-formatted, mount the MDT for the
5206 # first time to guarantee the "last_rcvd" file is there.
5207 start_mds || error "fail to mount the MDS for the first time"
5208 stop_mds || error "Unable to stop MDS"
5211 # remove the "last_rcvd" file
5212 do_facet $SINGLEMDS "mkdir -p $brpt"
5213 do_facet $SINGLEMDS \
5214 "mount -t $mds1_FSTYPE $opts $devname $brpt"
5215 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
5216 do_facet $SINGLEMDS "$UMOUNT $brpt"
5218 # restart MDS, the "last_rcvd" file should be recreated.
5219 start_mds || error "fail to restart the MDS"
5220 stop_mds || error "Unable to stop MDS"
5221 obj=$(do_facet $SINGLEMDS \
5222 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
5223 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
5225 run_test 65 "re-create the lost last_rcvd file when server mount"
5228 [[ "$MGS_VERSION" -ge $(version_code 2.3.59) ]] ||
5229 skip "Need MGS version at least 2.3.59"
5232 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
5233 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
5235 # add EXCLUDE records to config log, they are not to be
5236 # removed by lctl replace_nids
5237 set_conf_param_and_check mds \
5238 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
5239 "$FSNAME-OST0000.osc.active" \
5242 echo "replace_nids should fail if MDS, OSTs and clients are UP"
5243 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5244 error "replace_nids fail"
5246 umount_client $MOUNT || error "unmounting client failed"
5247 echo "replace_nids should fail if MDS and OSTs are UP"
5248 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5249 error "replace_nids fail"
5251 stop_ost || error "Unable to stop OST1"
5252 echo "replace_nids should fail if MDS is UP"
5253 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5254 error "replace_nids fail"
5256 stop_mds || error "stopping mds failed"
5258 if combined_mgs_mds; then
5259 start_mdt 1 "-o nosvc" ||
5260 error "starting mds with nosvc option failed"
5263 echo "command should accept two parameters"
5264 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
5265 error "command should accept two params"
5267 echo "correct device name should be passed"
5268 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
5269 error "wrong devname"
5271 echo "wrong nids list should not destroy the system"
5272 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
5274 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "asdfasdf, asdfadf" &&
5277 echo "replace OST nid"
5278 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
5279 error "replace nids failed"
5281 echo "command should accept two parameters"
5282 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
5283 error "command should accept two params"
5285 echo "wrong nids list should not destroy the system"
5286 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
5289 local FAKE_NIDS="192.168.0.112@tcp1,192.168.0.112@tcp2"
5290 local FAKE_FAILOVER="192.168.0.113@tcp1,192.168.0.113@tcp2"
5291 local NIDS_AND_FAILOVER="$MDS_NID,$FAKE_NIDS:$FAKE_FAILOVER"
5292 echo "set NIDs with failover"
5293 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $NIDS_AND_FAILOVER ||
5294 error "replace nids failed"
5297 echo "replace MDS nid"
5298 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
5299 error "replace nids failed"
5301 if ! combined_mgs_mds ; then
5304 stop_mds || error "Unable to stop MDS"
5307 start_mgsmds || error "start mgsmds failed"
5308 set_conf_param_and_check mds \
5309 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
5310 "$FSNAME-OST0000.osc.active" \
5312 start_ost || error "unable to start OST"
5313 mount_client $MOUNT || error "mount client failed"
5315 check_mount || error "error after nid replace"
5316 cleanup || error "cleanup failed"
5319 run_test 66 "replace nids"
5321 test_67() { #LU-2950
5322 local legacy="$TMP/legacy_lnet_config"
5323 local new="$TMP/new_routes_test"
5324 local out="$TMP/config_out_file"
5325 local verify="$TMP/conv_verify"
5326 local verify_conf="$TMP/conf_verify"
5328 # Create the legacy file that will be run through the
5329 # lustre_routes_conversion script
5330 cat <<- LEGACY_LNET_CONFIG > $legacy
5331 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
5332 tcp2 54 10.1.3.2@tcp;
5333 tcp3 10.3.4.3@tcp:3;
5337 # Create the verification file to verify the output of
5338 # lustre_routes_conversion script against.
5339 cat <<- VERIFY_LNET_CONFIG > $verify
5340 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
5341 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
5342 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
5343 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
5344 tcp4: { gateway: 10.3.3.4@tcp }
5347 # Create the verification file to verify the output of
5348 # lustre_routes_config script against
5349 cat <<- VERIFY_LNET_CONFIG > $verify_conf
5350 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
5351 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
5352 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
5353 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
5354 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
5357 $LUSTRE_ROUTES_CONVERSION $legacy $new > /dev/null
5358 if [ -f $new ]; then
5359 # verify the conversion output
5360 cmp -s $new $verify > /dev/null
5361 if [ $? -eq 1 ]; then
5362 error "routes conversion failed"
5365 lustre_routes_config --dry-run --verbose $new > $out
5366 # check that the script succeeded
5367 cmp -s $out $verify_conf > /dev/null
5368 if [ $? -eq 1 ]; then
5369 error "routes config failed"
5372 error "routes conversion test failed"
5374 # remove generated files
5375 rm -f $new $legacy $verify $verify_conf $out
5377 run_test 67 "test routes conversion and configuration"
5385 [ "$MDS1_VERSION" -ge $(version_code 2.4.53) ] ||
5386 skip "Need MDS version at least 2.4.53"
5388 umount_client $MOUNT || error "umount client failed"
5393 # START-END - the sequences we'll be reserving
5394 START=$(do_facet $SINGLEMDS \
5395 $LCTL get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
5396 END=$((START + (1 << 30)))
5397 do_facet $SINGLEMDS \
5398 $LCTL set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
5400 # reset the sequences MDT0000 has already assigned
5401 do_facet $SINGLEMDS \
5402 $LCTL set_param seq.srv*MDT0000.space=clear
5404 # remount to let the client allocate new sequence
5405 mount_client $MOUNT || error "mount client failed"
5407 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
5408 do_facet $SINGLEMDS \
5409 $LCTL get_param seq.srv*MDT0000.space
5410 $LFS path2fid $DIR/$tfile
5412 local old_ifs="$IFS"
5414 fid=($($LFS path2fid $DIR/$tfile))
5418 if [[ $seq < $END ]]; then
5419 error "used reserved sequence $seq?"
5421 cleanup || error "cleanup failed with $?"
5423 run_test 68 "be able to reserve specific sequences in FLDB"
5425 # Test 69: is about the total number of objects ever created on an OST.
5426 # so that when it is reformatted the normal MDS->OST orphan recovery won't
5427 # just "precreate" the missing objects. In the past it might try to recreate
5428 # millions of objects after an OST was reformatted
5430 [[ "$MDS1_VERSION" -lt $(version_code 2.4.2) ]] &&
5431 skip "Need MDS version at least 2.4.2"
5433 [[ "$MDS1_VERSION" -ge $(version_code 2.4.50) ]] &&
5434 [[ "$MDS1_VERSION" -lt $(version_code 2.5.0) ]] &&
5435 skip "Need MDS version at least 2.5.0"
5438 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5440 # use OST0000 since it probably has the most creations
5441 local OSTNAME=$(ostname_from_index 0)
5442 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
5443 local last_id=$(do_facet mds1 $LCTL get_param -n \
5444 osp.$mdtosc_proc1.prealloc_last_id)
5446 # Want to have OST LAST_ID over 5 * OST_MAX_PRECREATE to
5447 # verify that the LAST_ID recovery is working properly. If
5448 # not, then the OST will refuse to allow the MDS connect
5449 # because the LAST_ID value is too different from the MDS
5450 #define OST_MAX_PRECREATE=20000
5451 local ost_max_pre=20000
5452 local num_create=$(( ost_max_pre * 5 + 1 - last_id))
5454 # If the LAST_ID is already over 5 * OST_MAX_PRECREATE, we don't
5455 # need to create any files. So, skip this section.
5456 if [ $num_create -gt 0 ]; then
5457 # Check the number of inodes available on OST0
5459 local ifree=$($LFS df -i $MOUNT |
5460 awk '/OST0000/ { print $4 }'; exit ${PIPESTATUS[0]})
5461 log "On OST0, $ifree inodes available. Want $num_create. rc=$?"
5463 $LFS setstripe -i 0 $DIR/$tdir ||
5464 error "$LFS setstripe -i 0 $DIR/$tdir failed"
5465 if [ $ifree -lt 10000 ]; then
5466 files=$(( ifree - 50 ))
5471 local j=$((num_create / files + 1))
5472 for i in $(seq 1 $j); do
5473 createmany -o $DIR/$tdir/$tfile-$i- $files ||
5474 error "createmany fail create $files files: $?"
5475 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
5476 error "unlinkmany failed unlink $files files"
5480 # delete all of the files with objects on OST0 so the
5481 # filesystem is not inconsistent later on
5482 $LFS find $MOUNT --ost 0 -print0 | xargs -0 rm
5484 umount_client $MOUNT || error "umount client failed"
5485 stop_ost || error "OST0 stop failure"
5486 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
5487 $(ostdevname 1) $(ostvdevname 1) ||
5488 error "reformat and replace $ostdev failed"
5489 start_ost || error "OST0 restart failure"
5490 wait_osc_import_state mds ost FULL
5492 mount_client $MOUNT || error "mount client failed"
5493 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
5494 local idx=$($LFS getstripe -i $DIR/$tdir/$tfile-last)
5495 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
5497 local iused=$($LFS df -i $MOUNT |
5498 awk '/OST0000/ { print $3 }'; exit ${PIPESTATUS[0]})
5499 log "On OST0, $iused used inodes rc=$?"
5500 [ $iused -ge $((ost_max_pre/2 + 1000)) ] &&
5501 error "OST replacement created too many inodes; $iused"
5502 cleanup || error "cleanup failed with $?"
5504 run_test 69 "replace an OST with the same index"
5507 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5510 cleanup || error "cleanup failed with $?"
5512 start_mdt 1 || error "MDT0 start fail"
5514 start_ost || error "OST0 start fail"
5515 for num in $(seq 2 $MDSCOUNT); do
5516 start_mdt $num || return
5519 mount_client $MOUNT || error "mount client fails"
5521 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
5523 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5524 error "create remote dir fail"
5526 rm -rf $DIR/$tdir || error "delete dir fail"
5527 cleanup || error "cleanup failed with $?"
5529 run_test 70a "start MDT0, then OST, then MDT1"
5532 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5535 start_ost || error "OST0 start fail"
5537 start_mds || error "MDS start fail"
5539 mount_client $MOUNT || error "mount client fails"
5541 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
5543 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5544 error "create remote dir fail"
5546 rm -rf $DIR/$tdir || error "delete dir fail"
5548 cleanup || error "cleanup failed with $?"
5550 run_test 70b "start OST, MDT1, MDT0"
5553 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5556 start_mds || error "MDS start fail"
5557 start_ost || error "OST0 start fail"
5559 mount_client $MOUNT || error "mount client fails"
5560 stop_mdt 1 || error "MDT1 start fail"
5562 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
5563 echo "deactivate $mdc_for_mdt1"
5564 $LCTL --device $mdc_for_mdt1 deactivate ||
5565 error "set $mdc_for_mdt1 deactivate failed"
5567 mkdir $DIR/$tdir && error "mkdir succeed"
5569 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
5570 error "create remote dir succeed"
5572 cleanup || error "cleanup failed with $?"
5574 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
5577 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5580 start_mds || error "MDS start fail"
5581 start_ost || error "OST0 start fail"
5583 mount_client $MOUNT || error "mount client fails"
5585 stop_mdt 2 || error "MDT1 start fail"
5587 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
5589 echo "deactivate $mdc_for_mdt2"
5590 $LCTL --device $mdc_for_mdt2 deactivate ||
5591 error "set $mdc_for_mdt2 deactivate failed"
5593 mkdir $DIR/$tdir || error "mkdir fail"
5594 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
5595 error "create remote dir succeed"
5597 rm -rf $DIR/$tdir || error "delete dir fail"
5599 cleanup || error "cleanup failed with $?"
5601 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
5604 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5606 [ "$MDS1_VERSION" -ge $(version_code 2.7.62) ] ||
5607 skip "Need MDS version at least 2.7.62"
5609 reformat || error "reformat failed with $?"
5613 local mdsdev=$(mdsdevname 1)
5614 local ostdev=$(ostdevname 1)
5615 local mdsvdev=$(mdsvdevname 1)
5616 local ostvdev=$(ostvdevname 1)
5617 local opts_mds="$(mkfs_opts mds1 $mdsdev) --reformat $mdsdev $mdsvdev"
5618 local opts_ost="$(mkfs_opts ost1 $ostdev) --reformat $ostdev $ostvdev"
5620 if ! combined_mgs_mds ; then
5624 add mds1 $opts_mds || error "add mds1 failed"
5625 start_mdt 1 || error "start mdt1 failed"
5626 add ost1 $opts_ost || error "add ost1 failed"
5627 start_ost || error "start ost failed"
5628 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
5630 local soc=$(do_facet mds1 "$LCTL get_param -n \
5631 mdt.*MDT0000.sync_lock_cancel")
5632 [ $soc == "never" ] || error "SoC enabled on single MDS"
5634 for i in $(seq 2 $MDSCOUNT); do
5635 mdsdev=$(mdsdevname $i)
5636 mdsvdev=$(mdsvdevname $i)
5637 opts_mds="$(mkfs_opts mds$i $mdsdev) --reformat $mdsdev \
5639 add mds$i $opts_mds || error "add mds$i failed"
5640 start_mdt $i || error "start mdt$i fail"
5643 wait_dne_interconnect
5645 for i in $(seq $MDSCOUNT); do
5646 soc=$(do_facet mds$i "$LCTL get_param -n \
5647 mdt.*MDT000$((i - 1)).sync_lock_cancel")
5648 [ $soc == "blocking" ] || error "SoC not enabled on DNE"
5651 for i in $(seq 2 $MDSCOUNT); do
5652 stop_mdt $i || error "stop mdt$i fail"
5654 soc=$(do_facet mds1 "$LCTL get_param -n \
5655 mdt.*MDT0000.sync_lock_cancel")
5656 [ $soc == "never" ] || error "SoC enabled on single MDS"
5657 umount_client $MOUNT -f > /dev/null
5659 cleanup || error "cleanup failed with $?"
5661 run_test 70e "Sync-on-Cancel will be enabled by default on DNE"
5664 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5665 if combined_mgs_mds; then
5666 skip "needs separate MGS/MDT"
5670 start_mdt 1 || error "MDT0 start fail"
5671 start_ost || error "OST0 start fail"
5672 for num in $(seq 2 $MDSCOUNT); do
5673 start_mdt $num || return
5676 start_ost2 || error "OST1 start fail"
5678 mount_client $MOUNT || error "mount client fails"
5680 mkdir $DIR/$tdir || error "mkdir fail"
5681 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5682 error "create remote dir succeed"
5684 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5685 rm -rf $DIR/$tdir || error "delete dir fail"
5687 umount_client $MOUNT || error "umount_client failed"
5688 stop_mds || error "MDS stop fail"
5689 stop_ost || error "OST0 stop fail"
5690 stop_ost2 || error "OST1 stop fail"
5692 run_test 71a "start MDT0 OST0, MDT1, OST1"
5695 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5696 if combined_mgs_mds; then
5697 skip "needs separate MGS/MDT"
5701 for num in $(seq 2 $MDSCOUNT); do
5702 start_mdt $num || return
5704 start_ost || error "OST0 start fail"
5705 start_mdt 1 || error "MDT0 start fail"
5706 start_ost2 || error "OST1 start fail"
5708 mount_client $MOUNT || error "mount client fails"
5710 mkdir $DIR/$tdir || error "mkdir fail"
5711 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5712 error "create remote dir succeed"
5714 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5715 rm -rf $DIR/$tdir || error "delete dir fail"
5717 umount_client $MOUNT || error "umount_client failed"
5718 stop_mds || error "MDT0 stop fail"
5719 stop_ost || error "OST0 stop fail"
5720 stop_ost2 || error "OST1 stop fail"
5722 run_test 71b "start MDT1, OST0, MDT0, OST1"
5725 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5726 combined_mgs_mds && skip "needs separate MGS/MDT"
5730 start_ost || error "OST0 start fail"
5731 start_ost2 || error "OST1 start fail"
5732 for num in $(seq 2 $MDSCOUNT); do
5733 start_mdt $num || return
5735 start_mdt 1 || error "MDT0 start fail"
5737 mount_client $MOUNT || error "mount client fails"
5739 mkdir $DIR/$tdir || error "mkdir fail"
5740 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5741 error "create remote dir succeed"
5743 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5744 rm -rf $DIR/$tdir || error "delete dir fail"
5746 umount_client $MOUNT || error "umount_client failed"
5747 stop_mds || error "MDS stop fail"
5748 stop_ost || error "OST0 stop fail"
5749 stop_ost2 || error "OST1 stop fail"
5752 run_test 71c "start OST0, OST1, MDT1, MDT0"
5755 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5756 combined_mgs_mds && skip "needs separate MGS/MDT"
5760 start_ost || error "OST0 start fail"
5761 for num in $(seq 2 $MDSCOUNT); do
5762 start_mdt $num || return
5764 start_mdt 1 || error "MDT0 start fail"
5765 start_ost2 || error "OST1 start fail"
5767 mount_client $MOUNT || error "mount client fails"
5769 mkdir $DIR/$tdir || error "mkdir fail"
5770 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5771 error "create remote dir succeed"
5773 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5774 rm -rf $DIR/$tdir || error "delete dir fail"
5776 umount_client $MOUNT || error "umount_client failed"
5777 stop_mds || error "MDS stop fail"
5778 stop_ost || error "OST0 stop fail"
5779 stop_ost2 || error "OST1 stop fail"
5782 run_test 71d "start OST0, MDT1, MDT0, OST1"
5785 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5786 combined_mgs_mds && skip "needs separate MGS/MDT"
5790 start_ost || error "OST0 start fail"
5791 for num in $(seq 2 $MDSCOUNT); do
5792 start_mdt $num || return
5794 start_ost2 || error "OST1 start fail"
5795 start_mdt 1 || error "MDT0 start fail"
5797 mount_client $MOUNT || error "mount client fails"
5799 mkdir $DIR/$tdir || error "mkdir fail"
5800 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5801 error "create remote dir succeed"
5803 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5804 rm -rf $DIR/$tdir || error "delete dir fail"
5806 umount_client $MOUNT || error "umount_client failed"
5807 stop_mds || error "MDS stop fail"
5808 stop_ost || error "OST0 stop fail"
5809 stop_ost2 || error "OST1 stop fail"
5812 run_test 71e "start OST0, MDT1, OST1, MDT0"
5814 test_72() { #LU-2634
5815 [ "$mds1_FSTYPE" != ldiskfs ] &&
5816 skip "ldiskfs only test"
5818 local mdsdev=$(mdsdevname 1)
5819 local ostdev=$(ostdevname 1)
5820 local cmd="$E2FSCK -fnvd $mdsdev"
5827 if combined_mgs_mds; then
5828 add_options='--reformat'
5830 add_options='--reformat --replace'
5833 #tune MDT with "-O extents"
5835 for num in $(seq $MDSCOUNT); do
5836 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
5837 $add_options $(mdsdevname $num) $(mdsvdevname $num) ||
5838 error "add mds $num failed"
5839 do_facet mds${num} "$TUNE2FS -O extents $(mdsdevname $num)" ||
5840 error "$TUNE2FS failed on mds${num}"
5843 add ost1 $(mkfs_opts ost1 $ostdev) $add_options $ostdev ||
5844 error "add $ostdev failed"
5845 start_mds || error "start mds failed"
5846 start_ost || error "start ost failed"
5847 mount_client $MOUNT || error "mount client failed"
5849 #create some short symlinks
5850 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5851 createmany -o $DIR/$tdir/$tfile-%d $fn
5852 echo "create $fn short symlinks"
5853 for i in $(seq -w 1 $fn); do
5854 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
5859 umount_client $MOUNT || error "umount client failed"
5860 stop_mds || error "stop mds failed"
5861 stop_ost || error "stop ost failed"
5864 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
5866 run_test 72 "test fast symlink with extents flag enabled"
5868 test_73() { #LU-3006
5869 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
5870 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
5871 error "1st tunefs failed"
5872 start_mgsmds || error "start mds failed"
5873 start_ost || error "start ost failed"
5874 mount_client $MOUNT || error "mount client failed"
5875 $LCTL get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
5876 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
5877 umount_client $MOUNT || error "umount client failed"
5881 run_test 73 "failnode to update from mountdata properly"
5883 test_75() { # LU-2374
5884 [[ "$MDS1_VERSION" -lt $(version_code 2.4.1) ]] &&
5885 skip "Need MDS version at least 2.4.1"
5888 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
5889 --replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
5890 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
5891 --replace --reformat $(ostdevname 1) $(ostvdevname 1)"
5893 #check with default parameters
5894 add mds1 $opts_mds || error "add mds1 failed for default params"
5895 add ost1 $opts_ost || error "add ost1 failed for default params"
5897 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
5898 opts_mds=$(echo $opts_mds |
5899 sed -e "s/--index=$index/--index=$index --mdt/")
5900 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
5901 opts_ost=$(echo $opts_ost |
5902 sed -e "s/--index=$index/--index=$index --ost/")
5904 add mds1 $opts_mds || error "add mds1 failed for new params"
5905 add ost1 $opts_ost || error "add ost1 failed for new params"
5910 run_test 75 "The order of --index should be irrelevant"
5913 [[ "$MGS_VERSION" -ge $(version_code 2.4.52) ]] ||
5914 skip "Need MDS version at least 2.4.52"
5917 local MDMB_PARAM="osc.*.max_dirty_mb"
5918 echo "Change MGS params"
5919 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
5921 echo "max_dirty_mb: $MAX_DIRTY_MB"
5922 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB - 10))
5923 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
5924 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
5925 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5926 head -1" $NEW_MAX_DIRTY_MB
5927 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5928 echo "$MAX_DIRTY_MB"
5929 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5930 error "error while apply max_dirty_mb"
5932 echo "Check the value is stored after remount"
5935 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5936 head -1" $NEW_MAX_DIRTY_MB
5937 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5938 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5939 error "max_dirty_mb is not saved after remount"
5941 echo "Change OST params"
5942 CLIENT_PARAM="obdfilter.*.client_cache_count"
5943 local CLIENT_CACHE_COUNT
5944 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5946 echo "client_cache_count: $CLIENT_CACHE_COUNT"
5947 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
5948 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
5949 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
5950 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5951 head -1" $NEW_CLIENT_CACHE_COUNT
5952 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5954 echo "$CLIENT_CACHE_COUNT"
5955 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5956 error "error while apply client_cache_count"
5958 echo "Check the value is stored after remount"
5961 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5962 head -1" $NEW_CLIENT_CACHE_COUNT
5963 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5965 echo "$CLIENT_CACHE_COUNT"
5966 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5967 error "client_cache_count is not saved after remount"
5970 run_test 76a "set permanent params with lctl across mounts"
5972 test_76b() { # LU-4783
5973 [[ "$MGS_VERSION" -ge $(version_code 2.5.57) ]] ||
5974 skip "Need MGS version at least 2.5.57"
5977 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
5978 error "start params log failed"
5981 run_test 76b "verify params log setup correctly"
5984 [[ "$MGS_VERSION" -ge $(version_code 2.8.54) ]] ||
5985 skip "Need MDS version at least 2.4.52"
5987 local MASK_PARAM="mdd.*.changelog_mask"
5988 echo "Change changelog_mask"
5989 do_facet mgs $LCTL set_param -P $MASK_PARAM=-CLOSE ||
5990 error "Can't change changlog_mask"
5991 wait_update $(facet_host mds) "$LCTL get_param -n $MASK_PARAM |
5994 echo "Check the value is stored after mds remount"
5995 stop_mds || error "Failed to stop MDS"
5996 start_mds || error "Failed to start MDS"
5997 local CHANGELOG_MASK=$(do_facet mgs $LCTL get_param -n $MASK_PARAM)
5998 echo $CHANGELOG_MASK | grep CLOSE > /dev/null &&
5999 error "changelog_mask is not changed"
6003 run_test 76c "verify changelog_mask is applied with lctl set_param -P"
6005 test_76d() { #LU-9399
6008 local xattr_cache="llite.*.xattr_cache"
6009 local cmd="$LCTL get_param -n $xattr_cache | head -1"
6010 local new=$((($(eval $cmd) + 1) % 2))
6012 echo "lctl set_param -P llite.*.xattr_cache=$new"
6013 do_facet mgs $LCTL set_param -P $xattr_cache=$new ||
6014 error "Can't change xattr_cache"
6015 wait_update $HOSTNAME "$cmd" "$new"
6017 echo "Check $xattr_cache on client $MOUNT"
6018 umount_client $MOUNT || error "umount $MOUNT failed"
6019 mount_client $MOUNT || error "mount $MOUNT failed"
6020 [ $(eval $cmd) -eq $new ] ||
6021 error "$xattr_cache != $new on client $MOUNT"
6023 echo "Check $xattr_cache on the new client $MOUNT2"
6024 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
6025 [ $(eval $cmd) -eq $new ] ||
6026 error "$xattr_cache != $new on client $MOUNT2"
6027 umount_client $MOUNT2 || error "umount $MOUNT2 failed"
6031 run_test 76d "verify llite.*.xattr_cache can be set by 'lctl set_param -P' correctly"
6033 test_77() { # LU-3445
6034 [[ "$MDS1_VERSION" -ge $(version_code 2.8.55) ]] ||
6035 skip "Need MDS version 2.8.55+ "
6037 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
6038 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
6039 skip_env "mixed loopback and real device not working"
6042 local fs2mdsdev=$(mdsdevname 1_2)
6043 local fs2ostdev=$(ostdevname 1_2)
6044 local fs2mdsvdev=$(mdsvdevname 1_2)
6045 local fs2ostvdev=$(ostvdevname 1_2)
6046 local fsname=test1234
6048 local failnid="$(h2nettype 1.2.3.4),$(h2nettype 4.3.2.1)"
6050 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
6052 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
6053 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
6054 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
6055 error "start fs2mds failed"
6057 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
6058 mgsnid="0.0.0.0@tcp,$mgsnid,$mgsnid:$mgsnid"
6060 add fs2ost --mgsnode=$mgsnid $(mkfs_opts ost1 $fs2ostdev) \
6061 --failnode=$failnid --fsname=$fsname \
6062 --reformat $fs2ostdev $fs2ostvdev ||
6063 error "add fs2ost failed"
6064 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
6066 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
6067 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
6068 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
6071 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
6074 [[ "$mds1_FSTYPE" != ldiskfs ||
6075 "$ost1_FSTYPE" != ldiskfs ]] &&
6076 skip "ldiskfs only test"
6078 # reformat the Lustre filesystem with a smaller size
6079 local saved_MDSCOUNT=$MDSCOUNT
6080 local saved_MDSSIZE=$MDSSIZE
6081 local saved_OSTCOUNT=$OSTCOUNT
6082 local saved_OSTSIZE=$OSTSIZE
6085 MDSSIZE=$((MDSSIZE - 20000))
6086 OSTSIZE=$((OSTSIZE - 20000))
6087 reformat || error "(1) reformat Lustre filesystem failed"
6088 MDSSIZE=$saved_MDSSIZE
6089 OSTSIZE=$saved_OSTSIZE
6091 # mount the Lustre filesystem
6092 setup_noconfig || error "(2) setup Lustre filesystem failed"
6095 log "create test files"
6100 mkdir $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
6102 for i in $(seq $num_files); do
6103 file=$MOUNT/$tdir/$tfile-$i
6104 dd if=/dev/urandom of=$file count=1 bs=1M || {
6105 $LCTL get_param osc.*.cur*grant*
6106 $LFS df; $LFS df -i;
6107 # stop creating files if there is no more space
6108 if [ ! -e $file ]; then
6109 num_files=$((i - 1))
6113 $LFS getstripe -v $file
6114 local ost_idx=$(LFS getstripe -i $file)
6115 do_facet ost$((ost_idx + 1)) \
6116 $LCTL get_param obdfilter.*.*grant*
6117 error "(4) create $file failed"
6121 # unmount the Lustre filesystem
6122 cleanup || error "(5) cleanup Lustre filesystem failed"
6124 # run e2fsck on the MDT and OST devices
6125 local mds_host=$(facet_active_host $SINGLEMDS)
6126 local ost_host=$(facet_active_host ost1)
6127 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
6128 local ost_dev=$(ostdevname 1)
6130 run_e2fsck $mds_host $mds_dev "-y"
6131 run_e2fsck $ost_host $ost_dev "-y"
6133 # get the original block count of the MDT and OST filesystems
6134 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
6135 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
6137 # expand the MDT and OST filesystems to the device size
6138 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
6139 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
6141 # run e2fsck on the MDT and OST devices again
6142 run_e2fsck $mds_host $mds_dev "-y"
6143 run_e2fsck $ost_host $ost_dev "-y"
6145 # mount the Lustre filesystem
6149 log "check files after expanding the MDT and OST filesystems"
6150 for i in $(seq $num_files); do
6151 file=$MOUNT/$tdir/$tfile-$i
6152 $CHECKSTAT -t file -s 1048576 $file ||
6153 error "(6) checkstat $file failed"
6157 log "create more files after expanding the MDT and OST filesystems"
6158 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
6159 file=$MOUNT/$tdir/$tfile-$i
6160 dd if=/dev/urandom of=$file count=1 bs=1M ||
6161 error "(7) create $file failed"
6164 # unmount the Lustre filesystem
6165 cleanup || error "(8) cleanup Lustre filesystem failed"
6167 # run e2fsck on the MDT and OST devices
6168 run_e2fsck $mds_host $mds_dev "-y"
6169 run_e2fsck $ost_host $ost_dev "-y"
6171 # get the maximum block count of the MDT and OST filesystems
6172 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
6173 local ost_max_blks=$(get_block_count ost1 $ost_dev)
6175 # get the minimum block count of the MDT and OST filesystems
6176 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
6177 grep minimum | sed -e 's/^.*filesystem: //g')
6178 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
6179 grep minimum | sed -e 's/^.*filesystem: //g')
6181 # shrink the MDT and OST filesystems to a smaller size
6185 if [[ $mds_max_blks -gt $mds_min_blks &&
6186 $mds_max_blks -gt $mds_orig_blks ]]; then
6187 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
6188 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
6189 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
6190 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
6191 error "shrink $SINGLEMDS to $new_blks failed"
6195 if [[ $ost_max_blks -gt $ost_min_blks &&
6196 $ost_max_blks -gt $ost_orig_blks ]]; then
6197 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
6198 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
6199 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
6200 run_resize2fs ost1 $ost_dev $new_blks ||
6201 error "shrink ost1 to $new_blks failed"
6205 # check whether the MDT or OST filesystem was shrunk or not
6207 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
6208 reformat_and_config ||
6209 error "(10) reformat Lustre filesystem failed"
6213 # run e2fsck on the MDT and OST devices again
6214 run_e2fsck $mds_host $mds_dev "-y"
6215 run_e2fsck $ost_host $ost_dev "-y"
6217 # mount the Lustre filesystem again
6221 log "check files after shrinking the MDT and OST filesystems"
6222 for i in $(seq $((num_files + 10))); do
6223 file=$MOUNT/$tdir/$tfile-$i
6224 $CHECKSTAT -t file -s 1048576 $file ||
6225 error "(11) checkstat $file failed"
6228 # unmount and reformat the Lustre filesystem
6229 cleanup || error "(12) cleanup Lustre filesystem failed"
6230 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
6232 MDSCOUNT=$saved_MDSCOUNT
6233 OSTCOUNT=$saved_OSTCOUNT
6234 reformat_and_config || error "(14) reformat Lustre filesystem failed"
6236 run_test 78 "run resize2fs on MDT and OST filesystems"
6238 test_79() { # LU-4227
6239 [[ "$MDS1_VERSION" -ge $(version_code 2.5.59) ]] ||
6240 skip "Need MDS version at least 2.5.59"
6242 local mdsdev1=$(mdsdevname 1)
6243 local mdsvdev1=$(mdsvdevname 1)
6244 local mdsdev2=$(mdsdevname 2)
6245 local mdsvdev2=$(mdsvdevname 2)
6246 local ostdev1=$(ostdevname 1)
6247 local ostvdev1=$(ostvdevname 1)
6248 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
6249 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
6250 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
6253 # remove --mgs/--mgsnode from mkfs.lustre options
6254 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
6256 mgsnode_opt=$(echo $opts_mds2 |
6257 awk '{ for ( i = 1; i < NF; i++ )
6258 if ( $i ~ "--mgsnode" ) { print $i; break } }')
6259 [ -n "$mgsnode_opt" ] &&
6260 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
6262 mgsnode_opt=$(echo $opts_ost1 |
6263 awk '{ for ( i = 1; i < NF; i++ )
6264 if ( $i ~ "--mgsnode" ) { print $i; break } }')
6265 [ -n "$mgsnode_opt" ] &&
6266 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
6268 # -MGS, format a mdt without --mgs option
6269 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
6270 error "Must specify --mgs when formatting mdt combined with mgs"
6272 # +MGS, format a mdt/ost without --mgsnode option
6273 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
6274 > /dev/null || error "start mds1 failed"
6275 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
6276 error "Must specify --mgsnode when formatting a mdt"
6277 add ost1 $opts_ost1 $ostdev1 $ostvdev2 &&
6278 error "Must specify --mgsnode when formatting an ost"
6282 run_test 79 "format MDT/OST without mgs option (should return errors)"
6285 start_mds || error "Failed to start MDT"
6286 start_ost || error "Failed to start OST1"
6287 uuid=$(do_facet ost1 $LCTL get_param -n mgc.*.uuid)
6288 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
6289 do_facet ost1 "$LCTL set_param fail_val=10 fail_loc=0x906"
6290 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x906"
6291 do_facet mgs "$LCTL set_param -n mgs/MGS/evict_client $uuid"
6293 start_ost2 || error "Failed to start OST2"
6295 do_facet ost1 "$LCTL set_param fail_loc=0"
6300 run_test 80 "mgc import reconnect race"
6302 #Save the original values of $OSTCOUNT and $OSTINDEX$i.
6304 local new_ostcount=$1
6305 saved_ostcount=$OSTCOUNT
6306 OSTCOUNT=$new_ostcount
6310 for i in $(seq $OSTCOUNT); do
6312 eval saved_ostindex$i=${!index}
6317 # Restore the original values of $OSTCOUNT and $OSTINDEX$i.
6318 restore_ostindex() {
6323 for i in $(seq $OSTCOUNT); do
6324 index=saved_ostindex$i
6325 eval OSTINDEX$i=${!index}
6327 OSTCOUNT=$saved_ostcount
6330 if ! combined_mgs_mds ; then
6335 # The main purpose of this test is to ensure the OST_INDEX_LIST functions as
6336 # expected. This test uses OST_INDEX_LIST to format OSTs with a randomly
6337 # assigned index and ensures we can mount such a formatted file system
6338 test_81() { # LU-4665
6339 [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
6340 skip "Need MDS version at least 2.6.54"
6341 [[ $OSTCOUNT -ge 3 ]] || skip_env "needs >= 3 OSTs"
6345 # Each time RANDOM is referenced, a random integer between 0 and 32767
6348 local saved_ostindex1=$OSTINDEX1
6349 for i in 65535 $((RANDOM + 65536)); do
6350 echo -e "\nFormat ost1 with --index=$i, should fail"
6352 if add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat \
6353 $(ostdevname 1) $(ostvdevname 1); then
6354 OSTINDEX1=$saved_ostindex1
6355 error "format ost1 with --index=$i should fail"
6358 OSTINDEX1=$saved_ostindex1
6362 # Format OSTs with random sparse indices.
6363 trap "restore_ostindex" EXIT
6364 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices"
6365 OST_INDEX_LIST=[0,$((RANDOM * 2 % 65533 + 1)),65534] formatall
6367 # Setup and check Lustre filesystem.
6368 start_mgsmds || error "start_mgsmds failed"
6369 for i in $(seq $OSTCOUNT); do
6370 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6371 error "start ost$i failed"
6374 mount_client $MOUNT || error "mount client $MOUNT failed"
6375 check_mount || error "check client $MOUNT failed"
6378 local max_easize=$($LCTL get_param -n llite.*.max_easize)
6379 # 65452 is XATTR_SIZE_MAX less ldiskfs ea overhead
6380 if large_xattr_enabled; then
6381 [[ $max_easize -ge 65452 ]] ||
6382 error "max_easize is $max_easize, should be at least 65452 bytes"
6385 # 4012 is 4096 less ldiskfs ea overhead
6386 [[ $max_easize -ge 4012 ]] ||
6387 error "max_easize is $max_easize, should be at least 4012 bytes"
6392 run_test 81 "sparse OST indexing"
6394 # Here we exercise the stripe placement functionality on a file system that
6395 # has formatted the OST with a random index. With the file system the following
6396 # functionality is tested:
6398 # 1. Creating a new file with a specific stripe layout.
6400 # 2. Modifiy a existing empty file with a specific stripe layout.
6402 # 3. Ensure we fail to set the stripe layout of a file that already has one.
6404 # 4. If ost-index is defined we need to ensure it is the first entry in the
6405 # ost index list returned by lfs getstripe.
6407 # 5. Lastly ensure this functionality fails with directories.
6408 test_82a() { # LU-4665
6409 [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
6410 skip "Need MDS version at least 2.6.54"
6411 [[ $OSTCOUNT -ge 3 ]] || skip_env "needs >= 3 OSTs"
6417 # Format OSTs with random sparse indices.
6421 local LOV_V1_INSANE_STRIPE_COUNT=65532
6422 for i in $(seq $OSTCOUNT); do
6423 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
6424 ost_indices+=" $index"
6426 ost_indices=$(comma_list $ost_indices)
6428 stack_trap "restore_ostindex" EXIT
6429 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
6430 OST_INDEX_LIST=[$ost_indices] formatall
6432 # Setup Lustre filesystem.
6433 start_mgsmds || error "start_mgsmds failed"
6434 for i in $(seq $OSTCOUNT); do
6435 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6436 error "start ost$i failed"
6439 # Collect debug information - start of test
6440 do_nodes $(comma_list $(mdts_nodes)) \
6441 $LCTL get_param osp.*.prealloc_*_id
6443 mount_client $MOUNT || error "mount client $MOUNT failed"
6447 check_lfs_df_ret_val $? || error "$LFS df $MOUNT failed"
6448 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6450 stack_trap "do_nodes $(comma_list $(mdts_nodes)) \
6451 $LCTL get_param osp.*.prealloc_*_id" EXIT
6453 # 1. If the file does not exist, new file will be created
6454 # with specified OSTs.
6455 local file=$DIR/$tdir/$tfile-1
6456 local cmd="$LFS setstripe -o $ost_indices $file"
6458 eval $cmd || error "$cmd failed"
6459 check_stripe_count $file $OSTCOUNT
6460 check_obdidx $file $ost_indices
6461 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6462 error "write $file failed"
6464 # 2. If the file already exists and is an empty file, the file
6465 # will be attached with specified layout.
6466 file=$DIR/$tdir/$tfile-2
6467 mcreate $file || error "mcreate $file failed"
6468 cmd="$LFS setstripe -o $ost_indices $file"
6470 eval $cmd || error "$cmd failed"
6471 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6472 error "write $file failed"
6473 check_stripe_count $file $OSTCOUNT
6474 check_obdidx $file $ost_indices
6476 # 3. If the file already has a valid layout attached, the command
6477 # should fail with EBUSY.
6479 eval $cmd && error "stripe is already set on $file, $cmd should fail"
6481 # 4. If [--stripe-index|-i <start_ost_idx>] is used, the index must
6482 # be in the OST indices list.
6483 local start_ost_idx=${ost_indices##*,}
6484 file=$DIR/$tdir/$tfile-3
6485 cmd="$LFS setstripe -o $ost_indices -i $start_ost_idx $file"
6487 eval $cmd || error "$cmd failed"
6488 check_stripe_count $file $OSTCOUNT
6489 check_obdidx $file $ost_indices
6490 check_start_ost_idx $file $start_ost_idx
6492 file=$DIR/$tdir/$tfile-4
6493 cmd="$LFS setstripe"
6494 cmd+=" -o $(exclude_items_from_list $ost_indices $start_ost_idx)"
6495 cmd+=" -i $start_ost_idx $file"
6497 eval $cmd && error "index $start_ost_idx should be in $ost_indices"
6499 # 5. Specifying OST indices for directory should succeed.
6500 local dir=$DIR/$tdir/$tdir
6501 mkdir $dir || error "mkdir $dir failed"
6502 cmd="$LFS setstripe -o $ost_indices $dir"
6503 if [ "$MDS1_VERSION" -gt $(version_code 2.11.53) ] &&
6504 [ "$CLIENT_VERSION" -gt $(version_code 2.11.53) ]; then
6506 eval $cmd || error "unable to specify OST indices on directory"
6508 echo "need MDS+client version at least 2.11.53"
6513 run_test 82a "specify OSTs for file (succeed) or directory (succeed)"
6518 # Remove OSTs from a pool and destroy the pool.
6519 destroy_pool $ost_pool || true
6524 # Test 82b is run to ensure that if the user supplies a pool with a specific
6525 # stripe layout that it behaves proprerly. It should fail in the case that
6526 # the supplied OST index list points to OSTs not contained in the user
6528 test_82b() { # LU-4665
6529 [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
6530 skip "Need MDS version at least 2.6.54"
6531 [[ $OSTCOUNT -ge 4 ]] || skip_env "needs >= 4 OSTs"
6537 # Format OSTs with random sparse indices.
6541 local LOV_V1_INSANE_STRIPE_COUNT=65532
6542 for i in $(seq $OSTCOUNT); do
6543 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
6544 ost_indices+=" $index"
6546 ost_indices=$(comma_list $ost_indices)
6548 trap "restore_ostindex" EXIT
6549 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
6550 OST_INDEX_LIST=[$ost_indices] formatall
6552 # Setup Lustre filesystem.
6553 start_mgsmds || error "start_mgsmds failed"
6554 for i in $(seq $OSTCOUNT); do
6555 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6556 error "start ost$i failed"
6559 mount_client $MOUNT || error "mount client $MOUNT failed"
6563 check_lfs_df_ret_val $? || error "$LFS df $MOUNT failed"
6564 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6566 # Create a new pool and add OSTs into it.
6567 local ost_pool=$FSNAME.$TESTNAME
6568 create_pool $ost_pool || error "create OST pool $ost_pool failed"
6571 trap "cleanup_82b" EXIT
6573 local ost_idx_in_list=${ost_indices##*,}
6574 local ost_idx_in_pool=$(exclude_items_from_list $ost_indices \
6577 local ost_targets="$FSNAME-OST["
6578 for i in ${ost_idx_in_pool//,/ }; do
6579 ost_targets=$ost_targets$(printf "%04x," $i)
6581 ost_targets="${ost_targets%,}]"
6583 local ost_targets_uuid=$(for i in ${ost_idx_in_pool//,/ }; \
6584 do printf "$FSNAME-OST%04x_UUID\n" $i; done |
6585 sort -u | tr '\n' ' ')
6587 local cmd="$LCTL pool_add $ost_pool $ost_targets"
6588 do_facet mgs $cmd || error "$cmd failed"
6589 wait_update $HOSTNAME "$LCTL get_param -n lov.$FSNAME-*.pools.$TESTNAME|
6590 sort -u | tr '\n' ' ' " "$ost_targets_uuid" ||
6591 error "wait_update $ost_pool failed"
6592 wait_update_facet $SINGLEMDS "$LCTL pool_list $ost_pool | wc -l" 4 ||
6593 error "wait_update pool_list $ost_pool failed"
6595 # If [--pool|-p <pool_name>] is set with [--ost-list|-o <ost_indices>],
6596 # then the OSTs must be the members of the pool.
6597 local file=$DIR/$tdir/$tfile
6598 cmd="$LFS setstripe -p $ost_pool -o $ost_idx_in_list $file"
6600 eval $cmd && error "OST with index $ost_idx_in_list should be" \
6601 "in OST pool $ost_pool"
6603 # Only select OST $ost_idx_in_list from $ost_pool for file.
6604 ost_idx_in_list=${ost_idx_in_pool#*,}
6605 cmd="$LFS setstripe -p $ost_pool -o $ost_idx_in_list $file"
6607 eval $cmd || error "$cmd failed"
6608 cmd="$LFS getstripe $file"
6610 eval $cmd || error "$cmd failed"
6611 check_stripe_count $file 2
6612 check_obdidx $file $ost_idx_in_list
6613 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6614 error "write $file failed"
6618 run_test 82b "specify OSTs for file with --pool and --ost-list options"
6621 [[ "$OST1_VERSION" -ge $(version_code 2.6.91) ]] ||
6622 skip "Need OST version at least 2.6.91"
6623 if [ "$ost1_FSTYPE" != ldiskfs ]; then
6624 skip "ldiskfs only test"
6633 ostmnt=$(facet_mntpt ost1)
6635 # Mount the OST as an ldiskfs filesystem.
6636 log "mount the OST $dev as a $ost1_FSTYPE filesystem"
6637 add ost1 $(mkfs_opts ost1 $dev) $FSTYPE_OPT \
6638 --reformat $dev > /dev/null ||
6639 error "format ost1 error"
6641 if ! test -b $dev; then
6642 mnt_opts=$(csa_add "$OST_MOUNT_FS_OPTS" -o loop)
6644 echo "mnt_opts $mnt_opts"
6645 do_facet ost1 mount -t "$ost1_FSTYPE" $dev \
6647 # Run llverfs on the mounted ldiskfs filesystem.
6648 # It is needed to get ENOSPACE.
6649 log "run llverfs in partial mode on the OST $ost1_FSTYPE $ostmnt"
6650 do_rpc_nodes $(facet_host ost1) run_llverfs $ostmnt -vpl \
6651 "no" || error "run_llverfs error on $ost1_FSTYPE"
6654 log "unmount the OST $dev"
6657 # Delete file IO_scrub. Later osd_scrub_setup will try to
6658 # create "IO_scrub" but will get ENOSPACE.
6660 echo "start ost1 service on `facet_active_host ost1`"
6661 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
6664 err=$(do_facet ost1 dmesg | grep "VFS: Busy inodes after unmount of")
6665 echo "string err $err"
6666 [ -z "$err" ] || error $err
6669 run_test 83 "ENOSPACE on OST doesn't cause message VFS: \
6670 Busy inodes after unmount ..."
6673 local facet=$SINGLEMDS
6674 local num=$(echo $facet | tr -d "mds")
6675 local dev=$(mdsdevname $num)
6676 local time_min=$(recovery_time_min)
6677 local recovery_duration
6678 local completed_clients
6679 local correct_clients
6682 echo "start mds service on $(facet_active_host $facet)"
6684 "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
6685 error "start MDS failed"
6687 start_ost || error "start OST0000 failed"
6688 start_ost2 || error "start OST0001 failed"
6690 echo "recovery_time=$time_min, timeout=$TIMEOUT, wrap_up=$wrap_up"
6692 mount_client $MOUNT1 || error "mount $MOUNT1 failed"
6693 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
6694 # make sure new superblock labels are sync'd before disabling writes
6698 replay_barrier $SINGLEMDS
6699 createmany -o $DIR1/$tfile-%d 1000
6701 # We need to catch the end of recovery window to extend it.
6702 # Skip 5 requests and add delay to request handling.
6703 #define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP
6704 do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5"
6706 facet_failover --fsck $SINGLEMDS || error "failover: $?"
6709 echo "recovery status"
6710 do_facet $SINGLEMDS \
6711 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status"
6713 recovery_duration=$(do_facet $SINGLEMDS \
6714 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
6715 awk '/recovery_duration/ { print $2 }')
6716 (( $recovery_duration > $time_min + $wrap_up )) &&
6717 error "recovery_duration > recovery_time_hard + wrap up"
6718 completed_clients=$(do_facet $SINGLEMDS \
6719 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
6720 awk '/completed_clients/ { print $2 }')
6722 correct_clients="$MDSCOUNT/$((MDSCOUNT+1))"
6723 [ "$completed_clients" = "${correct_clients}" ] ||
6724 error "$completed_clients != $correct_clients"
6726 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
6727 umount_client $MOUNT1
6728 umount_client $MOUNT2
6734 run_test 84 "check recovery_hard_time"
6737 [[ "$OST1_VERSION" -ge $(version_code 2.7.55) ]] ||
6738 skip "Need OST version at least 2.7.55"
6739 ##define OBD_FAIL_OSD_OST_EA_FID_SET 0x197
6740 do_facet ost1 "lctl set_param fail_loc=0x197"
6744 run_test 85 "osd_ost init: fail ea_fid_set"
6749 # ost1 has already registered to the MGS before the reformat.
6750 # So after reformatting it with option "-G", it could not be
6751 # mounted to the MGS. Cleanup the system for subsequent tests.
6756 [ "$ost1_FSTYPE" = zfs ] &&
6757 skip "LU-6442: no such mkfs params for ZFS OSTs"
6758 [[ "$MDS1_VERSION" -ge $(version_code 2.7.56) ]] ||
6759 skip "Need server version newer than 2.7.55"
6762 local OLDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6763 awk '/Flex block group size: / { print $NF; exit; }')
6765 [ "$OLDSIZE" == "$NEWSIZE" ] && skip "$NEWSIZE groups already"
6767 local opts=" -O flex_bg -G $NEWSIZE"
6768 opts=$(OST_FS_MKFS_OPTS+="$opts" mkfs_opts ost1 $(ostdevname 1))
6769 opts+=" --reformat $(ostdevname 1) $(ostvdevname 1)"
6770 echo "params: $opts"
6772 trap cleanup_86 EXIT ERR
6775 add ost1 $opts || error "add ost1 failed with new params"
6777 local FOUNDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6778 awk '/Flex block group size: / { print $NF; exit; }')
6780 [[ $FOUNDSIZE == $NEWSIZE ]] ||
6781 error "Flex block group size: $FOUNDSIZE, expected: $NEWSIZE"
6785 run_test 86 "Replacing mkfs.lustre -G option"
6787 test_87() { #LU-6544
6788 [[ "$MDS1_VERSION" -ge $(version_code 2.9.51) ]] ||
6789 skip "Need MDS version at least 2.9.51"
6790 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
6791 skip "ldiskfs only test"
6792 [[ $OSTCOUNT -gt 59 ]] &&
6793 skip "Ignore wide striping situation"
6794 [ -n "$FILESET" ] && skip "Not functional for FILESET set"
6796 local mdsdev=$(mdsdevname 1)
6797 local mdsvdev=$(mdsvdevname 1)
6798 local file=$DIR/$tfile
6799 local mntpt=$(facet_mntpt $SINGLEMDS)
6800 local used_xattr_blk=0
6801 local inode_size=${1:-1024}
6803 local xtest="trusted.test"
6807 local stripe_cnt=$(($OSTCOUNT + 2))
6809 #Please see ldiskfs_make_lustre() for MDT inode size calculation
6810 if [ $stripe_cnt -gt 16 ]; then
6813 left_size=$(expr $inode_size - \
6816 32 - 40 \* 3 - 32 \* 3 - $stripe_cnt \* 24 - 16 - 3 - \
6818 24 - 18 - $(expr length $tfile) - 16 - 4)
6819 if [ $left_size -le 0 ]; then
6820 echo "No space($left_size) is expected in inode."
6821 echo "Try 1-byte xattr instead to verify this."
6824 echo "Estimate: at most $left_size-byte space left in inode."
6830 if ! combined_mgs_mds ; then
6834 add mds1 $(mkfs_opts mds1 ${mdsdev}) --stripe-count-hint=$stripe_cnt \
6835 --reformat $mdsdev $mdsvdev || error "add mds1 failed"
6836 start_mdt 1 > /dev/null || error "start mdt1 failed"
6837 for i in $(seq $OSTCOUNT); do
6838 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS > /dev/null ||
6839 error "start ost$i failed"
6841 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
6842 check_mount || error "check client $MOUNT failed"
6845 $LFS setstripe -E 1M -S 1M -c 1 -E 64M -c 1 -E -1 -c -1 $file ||
6846 error "Create file with 3 components failed"
6847 $TRUNCATE $file $((1024*1024*64+1)) || error "truncate file failed"
6848 i=$($LFS getstripe -I3 -c $file) || error "get 3rd stripe count failed"
6849 if [ $i -ne $OSTCOUNT ]; then
6850 left_size=$(expr $left_size + $(expr $OSTCOUNT - $i) \* 24)
6851 echo -n "Since only $i out $OSTCOUNT OSTs are used, "
6852 echo -n "the expected left space is changed to "
6853 echo "$left_size bytes at most."
6855 value=$(generate_string $left_size)
6856 setfattr -n $xtest -v $value $file
6857 orig=$(get_xattr_value $xtest $file)
6858 [[ "$orig" != "$value" ]] && error "$xtest changed"
6860 #Verify if inode has some expected space left
6861 umount $MOUNT > /dev/null || error "umount $MOUNT failed"
6862 stop_mdt 1 > /dev/null || error "stop mdt1 failed"
6863 mount_ldiskfs $SINGLEMDS || error "mount -t ldiskfs $SINGLEMDS failed"
6865 do_facet $SINGLEMDS ls -sal $mntpt/ROOT/$tfile
6866 used_xattr_blk=$(do_facet $SINGLEMDS ls -s $mntpt/ROOT/$tfile |
6868 [[ $used_xattr_blk -eq 0 ]] &&
6869 error "Please check MDS inode size calculation: \
6870 more than $left_size-byte space left in inode."
6871 echo "Verified: at most $left_size-byte space left in inode."
6873 unmount_ldiskfs $SINGLEMDS
6875 for i in $(seq $OSTCOUNT); do
6876 stop ost$i -f || error "stop ost$i failed"
6879 run_test 87 "check if MDT inode can hold EAs with N stripes properly"
6882 [ "$mds1_FSTYPE" == zfs ] &&
6883 skip "LU-6662: no implementation for ZFS"
6887 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
6888 --reformat $(mdsdevname 1) || error "add mds1 failed"
6890 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
6891 grep -e \".*opts:.*errors=remount-ro.*\"" ||
6892 error "default mount options is missing"
6894 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
6895 --mountfsoptions="user_xattr,errors=panic" \
6896 --reformat $(mdsdevname 1) || error "add mds1 failed"
6898 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
6899 grep -e \".*opts:.*errors=panic.*\"" ||
6900 error "user can't override default mount options"
6902 run_test 88 "check the default mount options can be overridden"
6904 test_89() { # LU-7131
6905 [[ "$MDS1_VERSION" -ge $(version_code 2.9.54) ]] ||
6906 skip "Need MDT version at least 2.9.54"
6908 local key=failover.node
6909 local val1=192.0.2.254@tcp0 # Reserved IPs, see RFC 5735
6910 local val2=192.0.2.255@tcp0
6911 local mdsdev=$(mdsdevname 1)
6916 if [[ "$mds1_FSTYPE" == zfs ]]; then
6917 import_zpool mds1 || return ${PIPESTATUS[0]}
6920 # Check that parameters are added correctly
6921 echo "tunefs --param $key=$val1"
6922 do_facet mds1 "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
6923 error "tunefs --param $key=$val1 failed"
6924 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
6925 error "tunefs --dryrun failed"
6926 params=${params##*Parameters:}
6927 params=${params%%exiting*}
6928 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val1) = "1" ] ||
6929 error "on-disk parameter not added correctly via tunefs"
6931 # Check that parameters replace existing instances when added
6932 echo "tunefs --param $key=$val2"
6933 do_facet mds1 "$TUNEFS --param $key=$val2 $mdsdev >/dev/null" ||
6934 error "tunefs --param $key=$val2 failed"
6935 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
6936 error "tunefs --dryrun failed"
6937 params=${params##*Parameters:}
6938 params=${params%%exiting*}
6939 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "1" ] ||
6940 error "on-disk parameter not replaced via tunefs"
6941 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val2) = "1" ] ||
6942 error "on-disk parameter not replaced correctly via tunefs"
6944 # Check that a parameter is erased properly
6945 echo "tunefs --erase-param $key"
6946 do_facet mds1 "$TUNEFS --erase-param $key $mdsdev >/dev/null" ||
6947 error "tunefs --erase-param $key failed"
6948 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
6949 error "tunefs --dryrun failed"
6950 params=${params##*Parameters:}
6951 params=${params%%exiting*}
6952 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "0" ] ||
6953 error "on-disk parameter not erased correctly via tunefs"
6955 # Check that all the parameters are erased
6956 do_facet mds1 "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
6957 error "tunefs --param $key=$val1 failed"
6958 echo "tunefs --erase-params"
6959 do_facet mds1 "$TUNEFS --erase-params $mdsdev >/dev/null" ||
6960 error "tunefs --erase-params failed"
6961 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
6962 error "tunefs --dryrun failed"
6963 params=${params##*Parameters:}
6964 params=${params%%exiting*}
6965 params=$(echo $params | tr ' ' '\n')
6967 error "all on-disk parameters not erased correctly via tunefs $params"
6969 # Check the order of options --erase-params and --param
6970 echo "tunefs --param $key=$val1 --erase-params"
6972 "$TUNEFS --param $key=$val1 --erase-params $mdsdev >/dev/null"||
6973 error "tunefs --param $key=$val1 --erase-params failed"
6974 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
6975 error "tunefs --dryrun failed"
6976 params=${params##*Parameters:}
6977 params=${params%%exiting*}
6978 [ $(echo $params | tr ' ' '\n') == "$key=$val1" ] ||
6979 error "on-disk param not added correctly with --erase-params"
6983 run_test 89 "check tunefs --param and --erase-param{s} options"
6986 # $2 (optional) value of max_mod_rpcs_in_flight to set
6987 check_max_mod_rpcs_in_flight() {
6995 idx=$(printf "%04x" $($LFS getdirstripe -i $dir))
6996 facet="mds$((0x$idx + 1))"
6998 if [ -z "$mmr" ]; then
6999 # get value of max_mod_rcps_in_flight
7000 mmr=$($LCTL get_param -n \
7001 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) ||
7002 error "Unable to get max_mod_rpcs_in_flight"
7003 echo "max_mod_rcps_in_flight is $mmr"
7005 # set value of max_mod_rpcs_in_flight
7007 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight=$mmr ||
7008 error "Unable to set max_mod_rpcs_in_flight to $mmr"
7009 echo "max_mod_rpcs_in_flight set to $mmr"
7012 # create mmr+1 files
7013 echo "creating $((mmr + 1)) files ..."
7015 for i in $(seq $((mmr + 1))); do
7021 # consumes mmr-1 modify RPC slots
7022 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7023 # drop requests on MDT so that RPC slots are consumed
7024 # during all the request resend interval
7025 do_facet $facet "$LCTL set_param fail_loc=0x159"
7026 echo "launch $((mmr - 1)) chmod in parallel ..."
7027 for i in $(seq $((mmr - 1))); do
7028 chmod 0600 $dir/file-$i &
7032 # send one additional modify RPC
7033 do_facet $facet "$LCTL set_param fail_loc=0"
7034 echo "launch 1 additional chmod in parallel ..."
7035 chmod 0600 $dir/file-$mmr &
7038 # check this additional modify RPC get a modify RPC slot
7039 # and succeed its operation
7040 checkstat -vp 0600 $dir/file-$mmr ||
7041 error "Unable to send $mmr modify RPCs in parallel"
7046 # consumes mmr modify RPC slots
7047 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7048 # drop requests on MDT so that RPC slots are consumed
7049 # during all the request resend interval
7050 do_facet $facet "$LCTL set_param fail_loc=0x159"
7051 echo "launch $mmr chmod in parallel ..."
7052 for i in $(seq $mmr); do
7053 chmod 0666 $dir/file-$i &
7057 # send one additional modify RPC
7058 do_facet $facet "$LCTL set_param fail_loc=0"
7059 echo "launch 1 additional chmod in parallel ..."
7060 chmod 0666 $dir/file-$((mmr + 1)) &
7063 # check this additional modify RPC blocked getting a modify RPC slot
7064 checkstat -vp 0644 $dir/file-$((mmr + 1)) ||
7065 error "Unexpectedly send $(($mmr + 1)) modify RPCs in parallel"
7072 [[ $($LCTL get_param mdc.*.import |
7073 grep "connect_flags:.*multi_mod_rpc") ]] ||
7074 skip "Need MDC with 'multi_mod_rpcs' feature"
7076 # check default value
7077 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
7078 check_max_mod_rpcs_in_flight $DIR/$tdir
7082 run_test 90a "check max_mod_rpcs_in_flight is enforced"
7092 [[ $($LCTL get_param mdc.*.import |
7093 grep "connect_flags:.*multi_mod_rpc") ]] ||
7094 skip "Need MDC with 'multi_mod_rpcs' feature"
7097 # update max_mod_rpcs_in_flight
7098 $LFS mkdir -c1 $DIR/${tdir}1 || error "mkdir $DIR/${tdir}1 failed"
7099 check_max_mod_rpcs_in_flight $DIR/${tdir}1 1
7102 # check client is able to send multiple modify RPCs in paralell
7103 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
7104 grep -c "multi_mod_rpcs")
7105 if [ "$tmp" -ne $MDSCOUNT ]; then
7107 skip "Client not able to send multiple modify RPCs in parallel"
7110 # update max_mod_rpcs_in_flight
7111 $LFS mkdir -c1 $DIR/${tdir}2 || error "mkdir $DIR/${tdir}2 failed"
7112 check_max_mod_rpcs_in_flight $DIR/${tdir}2 5
7115 $LFS mkdir -c1 $DIR/${tdir}3 || error "mkdir $DIR/${tdir}3 failed"
7116 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/${tdir}3))
7117 facet="mds$((0x$idx + 1))"
7119 # save MDT max_mod_rpcs_per_client
7120 mmrpc=$(do_facet $facet \
7121 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
7123 # update max_mod_rpcs_in_flight
7124 umount_client $MOUNT
7126 "echo 16 > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
7128 $LCTL set_param mdc.$FSNAME-MDT$idx-mdc-*.max_rpcs_in_flight=17
7129 check_max_mod_rpcs_in_flight $DIR/${tdir}3 16
7131 # restore MDT max_mod_rpcs_per_client initial value
7133 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
7135 rm -rf $DIR/${tdir}?
7138 run_test 90b "check max_mod_rpcs_in_flight is enforced after update"
7141 # get max_rpcs_in_flight value
7142 mrif_90c=$($LCTL get_param -n \
7143 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
7144 echo "max_rpcs_in_flight is $mrif_90c"
7146 # get max_mod_rpcs_in_flight value
7147 mmrif_90c=$($LCTL get_param -n \
7148 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight)
7149 echo "max_mod_rpcs_in_flight is $mmrif_90c"
7151 # get MDT max_mod_rpcs_per_client value
7152 mmrpc_90c=$(do_facet mds1 \
7153 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
7154 echo "max_mod_rpcs_per_client is $mmrpc_90c"
7157 restore_params_90c() {
7160 # restore max_rpcs_in_flight value
7161 do_facet mgs $LCTL set_param -P \
7162 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$mrif_90c
7164 # restore max_mod_rpcs_in_flight value
7165 do_facet mgs $LCTL set_param -P \
7166 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mmrif_90c
7168 # restore MDT max_mod_rpcs_per_client value
7169 do_facet mds1 "echo $mmrpc_90c > \
7170 /sys/module/mdt/parameters/max_mod_rpcs_per_client"
7178 [[ $($LCTL get_param mdc.*.import |
7179 grep "connect_flags:.*multi_mod_rpc") ]] ||
7180 skip "Need MDC with 'multi_mod_rpcs' feature"
7182 # check client is able to send multiple modify RPCs in paralell
7183 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
7184 grep -c "multi_mod_rpcs")
7185 if [ "$tmp" -ne $MDSCOUNT ]; then
7187 skip "Client not able to send multiple modify RPCs in parallel"
7191 stack_trap restore_params_90c
7194 # attempt to set max_mod_rpcs_in_flight to max_rpcs_in_flight value
7195 # prerequisite: set max_mod_rpcs_per_client to max_rpcs_in_flight value
7196 do_facet mds1 "echo $mrif_90c > \
7197 /sys/module/mdt/parameters/max_mod_rpcs_per_client"
7199 # if max_mod_rpcs_in_flight is set to be equal to or larger than
7200 # max_rpcs_in_flight, then max_rpcs_in_flight will be increased
7201 if [[ "$CLIENT_VERSION" -ge $(version_code 2.13.53) ]]; then
7203 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif_90c ||
7204 error "set max_mod_rpcs_in_flight to $mrif_90c failed"
7206 local new_mrif=$($LCTL get_param -n \
7207 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
7208 ((new_mrif == mrif_90c + 1)) ||
7209 error "max_rpcs_in_flight was not increased"
7212 umount_client $MOUNT
7213 do_facet mds1 "echo $mmrpc_90c > \
7214 /sys/module/mdt/parameters/max_mod_rpcs_per_client"
7218 # attempt to set max_mod_rpcs_in_flight to max_mod_rpcs_per_client+1
7219 # prerequisite: set max_rpcs_in_flight to max_mod_rpcs_per_client+2
7221 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$((mmrpc_90c + 2))
7224 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$((mmrpc_90c + 1)) &&
7225 error "set max_mod_rpcs_in_flight to $((mmrpc_90c + 1)) should fail"
7228 # attempt to set max_mod_rpcs_in_flight permanently
7229 do_facet mgs $LCTL set_param -P \
7230 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$mrif_90c
7232 do_facet mgs $LCTL set_param -P \
7233 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif_90c
7235 remount_client $MOUNT
7237 wait_update_facet --verbose client "$LCTL get_param -n \
7238 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight" \
7239 "$((mrif_90c + 1))" ||
7240 error "expected '$((mrif_90c + 1))' for max_rpcs_in_flight"
7242 wait_update_facet --verbose client "$LCTL get_param -n \
7243 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight" \
7245 error "expected '$mrif_90c' for max_mod_rpcs_in_flight"
7250 run_test 90c "check max_mod_rpcs_in_flight update limits"
7261 [[ $($LCTL get_param mdc.*.import |
7262 grep "connect_flags:.*multi_mod_rpc") ]] ||
7263 skip "Need MDC with 'multi_mod_rpcs' feature"
7265 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
7266 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
7267 facet="mds$((0x$idx + 1))"
7269 # check client version supports multislots
7270 tmp=$($LCTL get_param -N \
7271 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
7272 if [ -z "$tmp" ]; then
7274 skip "Client does not support multiple modify RPCs in flight"
7277 # get current value of max_mod_rcps_in_flight
7278 mmr=$($LCTL get_param -n \
7279 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
7280 echo "max_mod_rcps_in_flight is $mmr"
7283 echo "creating $mmr files ..."
7285 for i in $(seq $mmr); do
7286 touch $DIR/$tdir/file-$i
7289 # prepare for close RPC
7290 multiop_bg_pause $DIR/$tdir/file-close O_c
7293 # consumes mmr modify RPC slots
7294 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7295 # drop requests on MDT so that RPC slots are consumed
7296 # during all the request resend interval
7297 do_facet $facet "$LCTL set_param fail_loc=0x159"
7298 echo "launch $mmr chmod in parallel ..."
7299 for i in $(seq $mmr); do
7300 chmod 0600 $DIR/$tdir/file-$i &
7303 # send one additional close RPC
7304 do_facet $facet "$LCTL set_param fail_loc=0"
7305 echo "launch 1 additional close in parallel ..."
7307 cancel_lru_locks mdc
7310 # check this additional close RPC get a modify RPC slot
7311 # and multiop process completed
7312 [ -d /proc/$pid ] &&
7313 error "Unable to send the additional close RPC in parallel"
7318 run_test 90d "check one close RPC is allowed above max_mod_rpcs_in_flight"
7320 check_uuid_on_ost() {
7322 do_facet ost1 "$LCTL get_param obdfilter.${FSNAME}*.exports.'$nid'.uuid"
7325 check_uuid_on_mdt() {
7327 do_facet $SINGLEMDS "$LCTL get_param mdt.${mds1_svc}*.exports.'$nid'.uuid"
7335 [[ "$OST1_VERSION" -ge $(version_code 2.7.63) ]] ||
7336 skip "Need OST version at least 2.7.63"
7337 [[ "$MDS1_VERSION" -ge $(version_code 2.7.63) ]] ||
7338 skip "Need MDT version at least 2.7.63"
7340 start_mds || error "MDS start failed"
7341 start_ost || error "unable to start OST"
7342 mount_client $MOUNT || error "client start failed"
7343 check_mount || error "check_mount failed"
7346 nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g")
7350 uuid=$(get_client_uuid $MOUNT)
7352 echo "list nids on mdt:"
7353 do_facet $SINGLEMDS "$LCTL list_param mdt.${FSNAME}*.exports.*"
7354 echo "uuid from $nid:"
7355 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
7357 found=$(check_uuid_on_mdt $nid | grep $uuid)
7358 [ -z "$found" ] && error "can't find $uuid $nid on MDT"
7359 found=$(check_uuid_on_ost $nid | grep $uuid)
7360 [ -z "$found" ] && error "can't find $uuid $nid on OST"
7362 # umount the client so it won't reconnect
7363 manual_umount_client --force || error "failed to umount $?"
7364 # shouldn't disappear on MDS after forced umount
7365 found=$(check_uuid_on_mdt $nid | grep $uuid)
7366 [ -z "$found" ] && error "can't find $uuid $nid"
7369 do_facet $SINGLEMDS \
7370 "$LCTL set_param -n mdt.${mds1_svc}.evict_client nid:$nid"
7371 sleep 1 # eviction above is async, give it some time to proceed
7373 found=$(check_uuid_on_mdt $nid | grep $uuid)
7374 [ -n "$found" ] && error "found $uuid $nid on MDT"
7375 found=$(check_uuid_on_ost $nid | grep $uuid)
7376 [ -n "$found" ] && error "found $uuid $nid on OST"
7378 # check it didn't reconnect (being umounted)
7379 sleep $((TIMEOUT+1))
7380 found=$(check_uuid_on_mdt $nid | grep $uuid)
7381 [ -n "$found" ] && error "found $uuid $nid on MDT"
7382 found=$(check_uuid_on_ost $nid | grep $uuid)
7383 [ -n "$found" ] && error "found $uuid $nid on OST"
7387 run_test 91 "evict-by-nid support"
7389 generate_ldev_conf() {
7390 # generate an ldev.conf file
7391 local ldevconfpath=$1
7393 local fsldevformat=""
7396 local facets="mgs,$(get_facets OST),$(get_facets MDS)"
7397 for facet in ${facets//,/ }; do
7399 fstype=$(facet_fstype $facet)
7400 [ "$fstype" = zfs ] && fsldevformat="$fstype:"
7401 local host=$(facet_host $facet)
7403 local varfo=${facet}failover_HOST
7404 if [ $facet == mgs ] && combined_mgs_mds; then
7405 varfo=mds1failover_HOST
7407 [ -n "${!varfo}" ] && fo=${!varfo}
7408 local type=$(echo $facet | tr -d "[:digit:]" | \
7409 tr "[:lower:]" "[:upper:]" | sed s/MDS/MDT/ )
7411 [ ${facet} == mgs ] ||
7412 num=$(facet_number $facet)
7413 printf "%s\t%s\t%s-%s%04x\t%s%s\n" \
7420 $(facet_device $facet) >> $ldevconfpath
7423 echo "----- $ldevconfpath -----"
7425 echo "--- END $ldevconfpath ---"
7430 # generate a nids file (mapping between hostname to nid)
7431 # looks like we only have the MGS nid available to us
7432 # so just echo that to a file
7434 echo -e "${mgs_HOST}\t${MGSNID}" > $nidspath
7436 echo "----- $nidspath -----"
7438 echo "--- END $nidspath ---"
7441 compare_ldev_output() {
7442 local ldev_output=$1
7443 local expected_output=$2
7445 sort $expected_output -o $expected_output
7446 sort $ldev_output -o $ldev_output
7448 echo "-- START OF LDEV OUTPUT --"
7450 echo "--- END OF LDEV OUTPUT ---"
7452 echo "-- START OF EXPECTED OUTPUT --"
7453 cat $expected_output
7454 echo "--- END OF EXPECTED OUTPUT ---"
7456 diff $expected_output $ldev_output
7461 if [ -z "$LDEV" ]; then
7462 error "ldev is missing!"
7465 local LDEVCONFPATH=$TMP/ldev.conf
7466 local NIDSPATH=$TMP/nids
7468 echo "Host is $(hostname)"
7470 generate_ldev_conf $LDEVCONFPATH
7471 generate_nids $NIDSPATH
7473 # echo the mgs nid and compare it to environment variable MGSNID
7474 local facets="$(get_facets OST),$(get_facets MDS),mgs"
7475 for facet in ${facets//,/ }; do
7476 local host=$(facet_host $facet)
7477 local output=$($LDEV -c $LDEVCONFPATH -H $host -n $NIDSPATH echo %m)
7479 echo "-- START OF LDEV OUTPUT --"
7481 echo "--- END OF LDEV OUTPUT ---"
7484 error "ldev failed to execute!"
7486 # need to process multiple lines because of
7487 # several targets on host
7488 echo -e $output | awk '{ print $2 }' | while read -r line ; do
7489 [ "$line" = "$MGSNID" ] ||
7490 error "ldev failed mgs nid '$line', \
7494 rm -f $LDEVCONFPATH $NIDSPATH
7496 run_test 92 "ldev returns MGS NID correctly in command substitution"
7499 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs"
7502 #start mgs or mgs/mdt0
7503 if ! combined_mgs_mds ; then
7510 start_ost || error "OST0 start fail"
7512 #define OBD_FAIL_MGS_WRITE_TARGET_DELAY 0x90e
7513 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x8000090e"
7514 for num in $(seq 2 $MDSCOUNT); do
7518 mount_client $MOUNT || error "mount client fails"
7519 wait_osc_import_state mds ost FULL
7520 wait_osc_import_ready client ost
7521 check_mount || error "check_mount failed"
7523 cleanup || error "cleanup failed with $?"
7525 run_test 93 "register mulitple MDT at the same time"
7528 if [ -z "$LDEV" ]; then
7529 error "ldev is missing!"
7532 local LDEVCONFPATH=$TMP/ldev.conf
7533 local NIDSPATH=$TMP/nids
7535 generate_ldev_conf $LDEVCONFPATH
7536 generate_nids $NIDSPATH
7538 local LDEV_OUTPUT=$TMP/ldev-output.txt
7539 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME > $LDEV_OUTPUT
7541 # ldev failed, error
7542 if [ $? -ne 0 ]; then
7543 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7544 error "ldev failed to execute!"
7548 local EXPECTED_OUTPUT=$TMP/ldev-expected.txt
7550 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
7552 for num in $(seq $MDSCOUNT); do
7553 printf "%s-MDT%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7556 for num in $(seq $OSTCOUNT); do
7557 printf "%s-OST%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7560 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7562 if [ $? -ne 0 ]; then
7563 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7564 error "ldev failed to produce the correct hostlist!"
7567 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7569 run_test 94 "ldev outputs correct labels for file system name query"
7572 if [ -z "$LDEV" ]; then
7573 error "ldev is missing!"
7576 local LDEVCONFPATH=$TMP/ldev.conf
7577 local NIDSPATH=$TMP/nids
7579 generate_ldev_conf $LDEVCONFPATH
7580 generate_nids $NIDSPATH
7584 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME &>/dev/null
7585 if [ $? -ne 0 ]; then
7586 rm $LDEVCONFPATH $NIDSPATH
7587 error "ldev label filtering w/ -F failed!"
7591 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -l &>/dev/null
7592 if [ $? -ne 0 ]; then
7593 rm $LDEVCONFPATH $NIDSPATH
7594 error "ldev label filtering w/ -l failed!"
7598 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f &>/dev/null
7599 if [ $? -ne 0 ]; then
7600 rm $LDEVCONFPATH $NIDSPATH
7601 error "ldev label filtering w/ -f failed!"
7605 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a &>/dev/null
7606 if [ $? -ne 0 ]; then
7607 rm $LDEVCONFPATH $NIDSPATH
7608 error "ldev label filtering w/ -a failed!"
7613 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -F $FSNAME &>/dev/null
7614 if [ $? -eq 0 ]; then
7615 rm $LDEVCONFPATH $NIDSPATH
7616 error "ldev label filtering w/ -a and -F incorrectly succeeded"
7620 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -f &>/dev/null
7621 if [ $? -eq 0 ]; then
7622 rm $LDEVCONFPATH $NIDSPATH
7623 error "ldev label filtering w/ -a and -f incorrectly succeeded"
7627 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -l &>/dev/null
7628 if [ $? -eq 0 ]; then
7629 rm $LDEVCONFPATH $NIDSPATH
7630 error "ldev label filtering w/ -a and -l incorrectly succeeded"
7634 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f -l &>/dev/null
7635 if [ $? -eq 0 ]; then
7636 rm $LDEVCONFPATH $NIDSPATH
7637 error "ldev label filtering w/ -f and -l incorrectly succeeded"
7641 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -l &>/dev/null
7642 if [ $? -eq 0 ]; then
7643 rm $LDEVCONFPATH $NIDSPATH
7644 error "ldev label filtering w/ -F and -l incorrectly succeeded"
7647 # file sys & foreign
7648 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -f &>/dev/null
7649 if [ $? -eq 0 ]; then
7650 rm $LDEVCONFPATH $NIDSPATH
7651 error "ldev label filtering w/ -F and -f incorrectly succeeded"
7654 rm $LDEVCONFPATH $NIDSPATH
7656 run_test 95 "ldev should only allow one label filter"
7659 if [ -z "$LDEV" ]; then
7660 error "ldev is missing!"
7663 local LDEVCONFPATH=$TMP/ldev.conf
7664 local NIDSPATH=$TMP/nids
7666 generate_ldev_conf $LDEVCONFPATH
7667 generate_nids $NIDSPATH
7669 local LDEV_OUTPUT=$TMP/ldev-output.txt
7670 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -H $mgs_HOST \
7672 awk '{print $2}' > $LDEV_OUTPUT
7674 # ldev failed, error
7675 if [ $? -ne 0 ]; then
7676 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7677 error "ldev failed to execute!"
7681 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
7683 echo "$mgs_HOST-$(facet_fstype mgs)" > $EXPECTED_OUTPUT
7685 local facets="$(get_facets OST),$(get_facets MDS)"
7686 for facet in ${facets//,/ }; do
7687 local host=$(facet_host $facet)
7688 [ "$mgs_HOST" == "$host" ] &&
7689 echo "$host-$(facet_fstype $facet)" \
7693 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7695 if [ $? -ne 0 ]; then
7696 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7697 error "ldev failed to produce the correct output!"
7700 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7702 run_test 96 "ldev returns hostname and backend fs correctly in command sub"
7705 if [ -z "$LDEV" ]; then
7706 error "ldev is missing!"
7709 local LDEVCONFPATH=$TMP/ldev.conf
7710 local NIDSPATH=$TMP/nids
7712 generate_ldev_conf $LDEVCONFPATH
7713 generate_nids $NIDSPATH
7715 local LDEV_OUTPUT=$TMP/ldev-output.txt
7716 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
7718 echo -e "\nMDT role"
7719 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mdt > $LDEV_OUTPUT
7721 if [ $? -ne 0 ]; then
7722 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7723 error "ldev failed to execute for mdt role!"
7726 for num in $(seq $MDSCOUNT); do
7727 printf "%s-MDT%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7730 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7732 if [ $? -ne 0 ]; then
7733 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7734 error "ldev failed to produce the correct output for mdt role!"
7737 echo -e "\nOST role"
7738 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R ost > $LDEV_OUTPUT
7740 if [ $? -ne 0 ]; then
7741 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
7742 error "ldev failed to execute for ost role!"
7746 for num in $(seq $OSTCOUNT); do
7747 printf "%s-OST%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7750 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7752 if [ $? -ne 0 ]; then
7753 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7754 error "ldev failed to produce the correct output for ost role!"
7757 echo -e "\nMGS role"
7758 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mgs > $LDEV_OUTPUT
7760 if [ $? -ne 0 ]; then
7761 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
7762 error "ldev failed to execute for mgs role!"
7765 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
7767 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7769 if [ $? -ne 0 ]; then
7770 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7771 error "ldev failed to produce the correct output for mgs role!"
7774 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7776 run_test 97 "ldev returns correct ouput when querying based on role"
7781 local temp=$MDS_MOUNT_OPTS
7784 check_mount || error "mount failed"
7785 mountopt="user_xattr"
7786 for ((x = 1; x <= 400; x++)); do
7787 mountopt="$mountopt,user_xattr"
7789 mount_client $MOUNT remount,$mountopt 2>&1 | grep "too long" ||
7790 error "Buffer overflow check failed"
7791 cleanup || error "cleanup failed"
7793 run_test 98 "Buffer-overflow check while parsing mount_opts"
7797 [[ "$ost1_FSTYPE" != ldiskfs ]] &&
7798 skip "ldiskfs only test"
7799 [[ "$OST1_VERSION" -ge $(version_code 2.8.57) ]] ||
7800 skip "Need OST version at least 2.8.57"
7802 local ost_opts="$(mkfs_opts ost1 $(ostdevname 1)) \
7803 --reformat $(ostdevname 1) $(ostvdevname 1)"
7804 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" &&
7805 skip "meta_bg already set"
7808 if [[ ${!opts} != *mkfsoptions* ]]; then
7809 eval opts=\"${!opts} \
7810 --mkfsoptions='\\\"-O ^resize_inode,meta_bg\\\"'\"
7812 local val=${!opts//--mkfsoptions=\\\"/ \
7813 --mkfsoptions=\\\"-O ^resize_inode,meta_bg }
7817 echo "params: $opts"
7820 add ost1 $opts || error "add ost1 failed with new params"
7822 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" ||
7823 error "meta_bg is not set"
7825 run_test 99 "Adding meta_bg option"
7830 start_mgsmds || error "MDS start failed"
7831 start_ost || error "unable to start OST"
7832 mount_client $MOUNT || error "client start failed"
7833 check_mount || error "check_mount failed"
7842 do_facet mgs 'lshowmount -v' | awk 'BEGIN {NR == 0; rc=1} /MGS:/ {rc=0}
7843 END {exit rc}' || error "lshowmount have no output MGS"
7845 do_facet mds1 'lshowmount -v' | awk 'BEGIN {NR == 2; rc=1} /-MDT0000:/
7846 {rc=0} END {exit rc}' || error "lshowmount have no output MDT0"
7848 do_facet ost1 'lshowmount -v' | awk 'BEGIN {NR == 4; rc=1} /-OST0000:/
7849 {rc=0} END {exit rc}' || error "lshowmount have no output OST0"
7851 cleanup || error "cleanup failed with $?"
7853 run_test 100 "check lshowmount lists MGS, MDT, OST and 0@lo"
7856 local createmany_pid
7857 local dev=$FSNAME-OST0000-osc-MDT0000
7861 createmany -o $DIR1/$tdir/$tfile-%d 50000 &
7863 # MDT->OST reconnection causes MDT<->OST last_id synchornisation
7864 # via osp_precreate_cleanup_orphans.
7865 for ((i = 0; i < 100; i++)); do
7866 for ((k = 0; k < 10; k++)); do
7867 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;" \
7868 "$LCTL --device $dev activate"
7871 ls -asl $MOUNT | grep '???' &&
7872 { kill -9 $createmany_pid &>/dev/null;
7873 error "File has no object on OST"; }
7875 kill -s 0 $createmany_pid || break
7877 wait $createmany_pid
7879 unlinkmany $DIR1/$tdir/$tfile-%d 50000
7882 run_test 101a "Race MDT->OST reconnection with create"
7885 local dev=$FSNAME-OST0000-osc-MDT0000
7886 local dir=$DIR1/$tdir
7890 $LFS setstripe -c 1 -i 0 $dir
7891 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;"
7892 #define OBD_FAIL_OSP_CON_EVENT_DELAY 0x2107
7893 do_facet mds1 "$LCTL set_param fail_loc=0x80002107 fail_val=20"
7894 do_facet $SINGLEMDS "$LCTL --device $dev activate;"
7899 wait_osc_import_state client ost1 FULL
7900 touch $dir/$tfile || error "Can't create file"
7904 run_test 101b "Race events DISCONNECT and ACTIVE in osp"
7907 [[ "$MDS1_VERSION" -gt $(version_code 2.9.53) ]] ||
7908 skip "Need server version greater than 2.9.53"
7909 [[ “$(mdsdevname 1)” != “$(mgsdevname)” ]] &&
7910 [[ “$(facet_host mds1)” = “$(facet_host mgs)” ]] &&
7911 skip "MGS must be on different node or combined"
7913 cleanup || error "cleanup failed with $?"
7915 local mds1dev=$(mdsdevname 1)
7916 local mds1mnt=$(facet_mntpt mds1)
7917 local mds1opts=$MDS_MOUNT_OPTS
7919 if [ "$mds1_FSTYPE" == ldiskfs ] &&
7920 ! do_facet mds1 test -b $mds1dev; then
7921 mds1opts=$(csa_add "$mds1opts" -o loop)
7923 if [[ "$mds1_FSTYPE" == zfs ]]; then
7924 import_zpool mds1 || return ${PIPESTATUS[0]}
7927 # unload all and only load libcfs to allow fail_loc setting
7928 do_facet mds1 $LUSTRE_RMMOD || error "unable to unload modules"
7929 do_facet mds1 modprobe libcfs || error "libcfs not loaded"
7930 do_facet mds1 lsmod \| grep libcfs || error "libcfs not loaded"
7932 #define OBD_FAIL_OBDCLASS_MODULE_LOAD 0x60a
7933 do_facet mds1 "$LCTL set_param fail_loc=0x8000060a"
7935 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts &&
7936 error "mdt start must fail"
7937 do_facet mds1 lsmod \| grep obdclass && error "obdclass must not load"
7939 do_facet mds1 "$LCTL set_param fail_loc=0x0"
7941 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts ||
7942 error "mdt start must not fail"
7944 cleanup || error "cleanup failed with $?"
7946 run_test 102 "obdclass module cleanup upon error"
7951 echo "rename $FSNAME to $newname"
7953 if ! combined_mgs_mds ; then
7954 local dev=$(mgsdevname)
7957 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
7958 error "(7) Fail to rename MGS"
7959 if [ "$(facet_fstype mgs)" = "zfs" ]; then
7960 reimport_zpool mgs $newname-mgs
7964 for num in $(seq $MDSCOUNT); do
7965 local dev=$(mdsdevname $num)
7967 do_facet mds${num} \
7968 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
7969 error "(8) Fail to rename MDT $num"
7970 if [ "$(facet_fstype mds${num})" = "zfs" ]; then
7971 reimport_zpool mds${num} $newname-mdt${num}
7975 for num in $(seq $OSTCOUNT); do
7976 local dev=$(ostdevname $num)
7978 do_facet ost${num} \
7979 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
7980 error "(9) Fail to rename OST $num"
7981 if [ "$(facet_fstype ost${num})" = "zfs" ]; then
7982 reimport_zpool ost${num} $newname-ost${num}
7987 test_103_set_pool() {
7991 do_facet mgs $LCTL pool_add $FSNAME.$pname ${FSNAME}-$ost_x ||
7992 error "Fail to add $ost_x to $FSNAME.$pname"
7993 wait_update $HOSTNAME \
7994 "lctl get_param -n lov.$FSNAME-clilov-*.pools.$pname |
7995 grep $ost_x" "$FSNAME-${ost_x}_UUID" ||
7996 error "$ost_x is NOT in pool $FSNAME.$pname"
7999 test_103_check_pool() {
8000 local save_fsname=$1
8003 stat $DIR/$tdir/test-framework.sh ||
8004 error "($errno) Fail to stat"
8005 do_facet mgs $LCTL pool_list $FSNAME.pool1 ||
8006 error "($errno) Fail to list $FSNAME.pool1"
8007 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname ||
8008 error "($errno) Fail to list $FSNAME.$save_fsname"
8009 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname |
8010 grep ${FSNAME}-OST0000 ||
8011 error "($errno) List $FSNAME.$save_fsname is invalid"
8013 local pname=$($LFS getstripe --pool $DIR/$tdir/d0)
8014 [ "$pname" = "$save_fsname" ] ||
8015 error "($errno) Unexpected pool name $pname"
8019 check_mount_and_prep
8021 mkdir $DIR/$tdir || error "(1) Fail to mkdir $DIR/$tdir"
8022 cp $LUSTRE/tests/test-framework.sh $DIR/$tdir ||
8023 error "(2) Fail to copy test-framework.sh"
8025 do_facet mgs $LCTL pool_new $FSNAME.pool1 ||
8026 error "(3) Fail to create $FSNAME.pool1"
8027 # name the pool name as the fsname
8028 do_facet mgs $LCTL pool_new $FSNAME.$FSNAME ||
8029 error "(4) Fail to create $FSNAME.$FSNAME"
8031 test_103_set_pool $FSNAME OST0000
8033 $LFS setstripe -p $FSNAME $DIR/$tdir/d0 ||
8034 error "(6) Fail to setstripe on $DIR/$tdir/d0"
8039 test_renamefs mylustre
8041 local save_fsname=$FSNAME
8045 test_103_check_pool $save_fsname 7
8047 if [ $OSTCOUNT -ge 2 ]; then
8048 test_103_set_pool $save_fsname OST0001
8051 $LFS setstripe -p $save_fsname $DIR/$tdir/f0 ||
8052 error "(16) Fail to setstripe on $DIR/$tdir/f0"
8061 test_103_check_pool $save_fsname 17
8065 test_renamefs $save_fsname
8071 run_test 103 "rename filesystem name"
8073 test_104a() { # LU-6952
8074 local mds_mountopts=$MDS_MOUNT_OPTS
8075 local ost_mountopts=$OST_MOUNT_OPTS
8076 local mds_mountfsopts=$MDS_MOUNT_FS_OPTS
8077 local lctl_ver=$(do_facet $SINGLEMDS $LCTL --version |
8080 [[ $(version_code $lctl_ver) -lt $(version_code 2.9.55) ]] &&
8081 skip "this test needs utils above 2.9.55"
8083 # specify "acl" in mount options used by mkfs.lustre
8084 if [ -z "$MDS_MOUNT_FS_OPTS" ]; then
8085 MDS_MOUNT_FS_OPTS="acl,user_xattr"
8088 MDS_MOUNT_FS_OPTS="${MDS_MOUNT_FS_OPTS},acl,user_xattr"
8091 echo "mountfsopt: $MDS_MOUNT_FS_OPTS"
8093 #reformat/remount the MDT to apply the MDT_MOUNT_FS_OPT options
8095 if ! combined_mgs_mds ; then
8099 if [ -z "$MDS_MOUNT_OPTS" ]; then
8100 MDS_MOUNT_OPTS="-o noacl"
8102 MDS_MOUNT_OPTS="${MDS_MOUNT_OPTS},noacl"
8105 for num in $(seq $MDSCOUNT); do
8106 start mds$num $(mdsdevname $num) $MDS_MOUNT_OPTS ||
8107 error "Failed to start MDS"
8110 for num in $(seq $OSTCOUNT); do
8111 start ost$num $(ostdevname $num) $OST_MOUNT_OPTS ||
8112 error "Failed to start OST"
8116 setfacl -m "d:$RUNAS_ID:rwx" $MOUNT &&
8117 error "ACL is applied when FS is mounted with noacl."
8119 MDS_MOUNT_OPTS=$mds_mountopts
8120 OST_MOUNT_OPTS=$ost_mountopts
8121 MDS_MOUNT_FS_OPTS=$mds_mountfsopts
8123 run_test 104a "Make sure user defined options are reflected in mount"
8125 test_104b() { # LU-12859
8126 mount_client $MOUNT3 flock,localflock
8127 stack_trap "umount_client $MOUNT3" EXIT
8128 mount | grep "$MOUNT3 .*,flock" && error "flock is still set"
8129 mount | grep "$MOUNT3 .*,localflock" || error "localflock is not set"
8130 umount_client $MOUNT3
8131 mount_client $MOUNT3 localflock,flock
8132 mount | grep "$MOUNT3 .*,localflock" && error "localflock is still set"
8133 mount | grep "$MOUNT3 .*,flock" || error "flock is not set"
8134 umount_client $MOUNT3
8135 mount_client $MOUNT3 localflock,flock,noflock
8136 flock_is_enabled $MOUNT3 && error "some flock is still enabled" || true
8138 run_test 104b "Mount uses last flock argument"
8140 error_and_umount() {
8151 mount --bind $DIR $TMP/$tdir || error "mount bind mnt pt failed"
8152 rm -f $TMP/$tdir/$tfile
8153 rm -f $TMP/$tdir/${tfile}1
8155 # Files should not be created in ro bind mount point
8156 # remounting from rw to ro
8157 mount -o remount,ro $TMP/$tdir ||
8158 error_and_umount "readonly remount of bind mnt pt failed"
8159 touch $TMP/$tdir/$tfile &&
8160 error_and_umount "touch succeeds on ro bind mnt pt"
8161 [ -e $TMP/$tdir/$tfile ] &&
8162 error_and_umount "file created on ro bind mnt pt"
8164 # Files should be created in rw bind mount point
8165 # remounting from ro to rw
8166 mount -o remount,rw $TMP/$tdir ||
8167 error_and_umount "read-write remount of bind mnt pt failed"
8168 touch $TMP/$tdir/${tfile}1 ||
8169 error_and_umount "touch fails on rw bind mnt pt"
8170 [ -e $TMP/$tdir/${tfile}1 ] ||
8171 error_and_umount "file not created on rw bind mnt pt"
8172 umount $TMP/$tdir || error "umount of bind mnt pt failed"
8174 cleanup || error "cleanup failed with $?"
8176 run_test 105 "check file creation for ro and rw bind mnt pt"
8183 mkdir -p $DIR/$tdir || error "create $tdir failed"
8184 lfs setstripe -c 1 -i 0 $DIR/$tdir
8185 #define OBD_FAIL_CAT_RECORDS 0x1312
8186 do_facet mds1 $LCTL set_param fail_loc=0x1312 fail_val=$repeat
8188 for ((i = 1; i <= $repeat; i++)); do
8190 #one full plain llog
8191 createmany -o $DIR/$tdir/f- 64768
8193 createmany -u $DIR/$tdir/f- 64768
8195 wait_delete_completed $((TIMEOUT * 7))
8196 #ASSERTION osp_sync_thread() ( thread->t_flags != SVC_RUNNING ) failed
8197 #shows that osp code is buggy
8198 do_facet mds1 $LCTL set_param fail_loc=0 fail_val=0
8202 run_test 106 "check osp llog processing when catalog is wrapped"
8205 [[ "$MDS1_VERSION" -ge $(version_code 2.10.50) ]] ||
8206 skip "Need MDS version > 2.10.50"
8209 start_mgsmds || error "start_mgsmds failed"
8210 start_ost || error "unable to start OST"
8212 # add unknown configuration parameter.
8213 if [[ $PERM_CMD == *"set_param -P"* ]]; then
8214 cmd="$PERM_CMD ost.$FSNAME-OST0000*.unknown_param"
8216 cmd="$PERM_CMD $FSNAME-OST0000*.ost.unknown_param"
8218 do_facet mgs "$cmd=50"
8219 cleanup_nocli || error "cleanup_nocli failed with $?"
8222 # unknown param should be ignored while mounting.
8223 start_ost || error "unable to start OST after unknown param set"
8225 cleanup || error "cleanup failed with $?"
8227 run_test 107 "Unknown config param should not fail target mounting"
8232 $rcmd rm -rf $tmp > /dev/null 2>&1
8233 $rcmd mkdir -p $tmp/{mnt,images} || error "failed to mkdir remotely"
8235 for facet in $facets; do
8236 [ "$mds1_FSTYPE" = zfs ] &&
8237 $rcmd $ZPOOL -f export lustre-$facet > /dev/null 2>&1
8238 $rcmd mkdir $tmp/mnt/$facet ||
8239 error "failed to mkdir $tmp/mnt/$facet"
8240 $rcmd dd if=/dev/zero of=$tmp/images/$facet \
8241 seek=199 bs=1M count=1 ||
8242 error "failed to create $tmp/images/$facet"
8251 local facet=${role}$((idx + 1))
8253 [ $# -eq 5 ] && pool=$5
8255 do_facet $SINGLEMDS $MKFS --fsname=lustre --$mgs \
8256 --$role --index=$idx --replace --backfstype=$bkfs \
8257 --device-size=200000 --reformat $pool $tmp/images/$facet ||
8258 error "failed to mkfs for $facet"
8262 echo "mounting client..."
8263 mount -t lustre ${nid}:/lustre $MOUNT ||
8264 error "failed to mount lustre"
8267 ls -l $MOUNT/local_dir || error "failed to list"
8269 echo "check truncate && write"
8270 echo "dummmmmmmmmmmmm" > $MOUNT/remote_dir/fsx.c ||
8271 error "failed to tuncate & write"
8274 touch $MOUNT/foooo ||
8275 error "failed to create"
8277 echo "check read && write && append"
8278 sha1sum $MOUNT/conf-sanity.sh |
8279 awk '{ print $1 }' > $MOUNT/checksum.new ||
8280 error "failed to read(1)"
8281 sha1sum $MOUNT/remote_dir/unlinkmany.c |
8282 awk '{ print $1 }' >> $MOUNT/checksum.new ||
8283 error "failed to read(2)"
8284 sha1sum $MOUNT/striped_dir/lockahead_test.o |
8285 awk '{ print $1 }' >> $MOUNT/checksum.new ||
8286 error "failed to read(3)"
8289 diff $MOUNT/checksum.new $MOUNT/checksum.src ||
8290 error "failed to verify data"
8300 umount -f $MOUNT || error "failed to umount client"
8301 for facet in $facets; do
8302 $rcmd umount -f $tmp/mnt/$facet ||
8303 error "failed to umount $facet"
8304 if [ "$mds1_FSTYPE" = zfs ]; then
8305 $rcmd $ZPOOL export -f lustre-$facet ||
8306 error "failed to export lustre-$facet"
8310 $rcmd rm -rf $tmp || error "failed to rm the dir $tmp"
8314 [ "$CLIENTONLY" ] && skip "Client-only testing"
8315 [ "$mds1_FSTYPE" != zfs ] && skip "zfs only test"
8316 [ "$MDS1_VERSION" -lt $(version_code 2.10.58) ] &&
8317 skip "Need server version at least 2.10.58"
8322 local tmp=$TMP/$tdir
8323 local rcmd="do_facet $SINGLEMDS"
8324 local facets="mdt1 mdt2 ost1 ost2"
8325 local nid=$($rcmd $LCTL list_nids | head -1)
8328 trap t_108_cleanup EXIT ERR
8331 t_108_mkfs mdt 0 zfs mgs lustre-mdt1/mdt1
8332 t_108_mkfs mdt 1 zfs mgsnode=$nid lustre-mdt2/mdt2
8333 t_108_mkfs ost 0 zfs mgsnode=$nid lustre-ost1/ost1
8334 t_108_mkfs ost 1 zfs mgsnode=$nid lustre-ost2/ost2
8336 for facet in $facets; do
8337 $rcmd zfs set mountpoint=$tmp/mnt/$facet canmount=on \
8338 lustre-$facet/$facet ||
8339 error "failed to zfs set for $facet (1)"
8340 $rcmd zfs mount lustre-$facet/$facet ||
8341 error "failed to local mount $facet"
8342 $rcmd tar jxf $LUSTRE/tests/ldiskfs_${facet}_2_11.tar.bz2 \
8343 --xattrs --xattrs-include="trusted.*" \
8344 -C $tmp/mnt/$facet/ > /dev/null 2>&1 ||
8345 error "failed to untar image for $facet"
8346 $rcmd "cd $tmp/mnt/$facet && rm -rf oi.* OI_* lfsck_* LFSCK" ||
8347 error "failed to cleanup for $facet"
8348 $rcmd zfs umount lustre-$facet/$facet ||
8349 error "failed to local umount $facet"
8350 $rcmd zfs set canmount=off lustre-$facet/$facet ||
8351 error "failed to zfs set $facet (2)"
8354 echo "changing server nid..."
8355 $rcmd mount -t lustre -o nosvc lustre-mdt1/mdt1 $tmp/mnt/mdt1
8356 $rcmd lctl replace_nids $FSNAME-MDT0000 $nid
8357 $rcmd lctl replace_nids $FSNAME-MDT0001 $nid
8358 $rcmd lctl replace_nids $FSNAME-OST0000 $nid
8359 $rcmd lctl replace_nids $FSNAME-OST0001 $nid
8360 $rcmd umount $tmp/mnt/mdt1
8362 for facet in $facets; do
8363 echo "mounting $facet from backup..."
8364 $rcmd mount -t lustre -o abort_recov lustre-$facet/$facet \
8365 $tmp/mnt/$facet || error "failed to mount $facet"
8368 # ZFS backend can detect migration and trigger OI scrub automatically
8369 # sleep 3 seconds for scrub done
8375 run_test 108a "migrate from ldiskfs to ZFS"
8378 [ "$CLIENTONLY" ] && skip "Client-only testing"
8379 [ "$mds1_FSTYPE" != ldiskfs ] && skip "ldiskfs only test"
8380 [ "$MDS1_VERSION" -lt $(version_code 2.10.58) ] &&
8381 skip "Need server version at least 2.10.58"
8386 local tmp=$TMP/$tdir
8387 local rcmd="do_facet $SINGLEMDS"
8388 local facets="mdt1 mdt2 ost1 ost2"
8389 local scrub_list="MDT0000 MDT0001 OST0000 OST0001"
8390 local nid=$($rcmd $LCTL list_nids | head -1)
8393 trap t_108_cleanup EXIT ERR
8396 t_108_mkfs mdt 0 ldiskfs mgs
8397 t_108_mkfs mdt 1 ldiskfs mgsnode=$nid
8398 t_108_mkfs ost 0 ldiskfs mgsnode=$nid
8399 t_108_mkfs ost 1 ldiskfs mgsnode=$nid
8401 for facet in $facets; do
8402 $rcmd mount -t ldiskfs -o loop $tmp/images/$facet \
8404 error "failed to local mount $facet"
8406 $rcmd tar jxf $LUSTRE/tests/zfs_${facet}_2_11.tar.bz2 \
8407 --xattrs --xattrs-include="*.*" \
8408 -C $tmp/mnt/$facet/ > /dev/null 2>&1 ||
8409 error "failed to untar image for $facet"
8410 $rcmd "cd $tmp/mnt/$facet && rm -rf oi.* OI_* lfsck_* LFSCK" ||
8411 error "failed to cleanup for $facet"
8412 $rcmd umount $tmp/mnt/$facet ||
8413 error "failed to local umount $facet"
8416 echo "changing server nid..."
8417 $rcmd mount -t lustre -o nosvc,loop $tmp/images/mdt1 $tmp/mnt/mdt1
8418 $rcmd lctl replace_nids $FSNAME-MDT0000 $nid
8419 $rcmd lctl replace_nids $FSNAME-MDT0001 $nid
8420 $rcmd lctl replace_nids $FSNAME-OST0000 $nid
8421 $rcmd lctl replace_nids $FSNAME-OST0001 $nid
8422 $rcmd umount $tmp/mnt/mdt1
8424 for facet in $facets; do
8425 echo "mounting $facet from backup..."
8426 $rcmd mount -t lustre -o loop,abort_recov $tmp/images/$facet \
8427 $tmp/mnt/$facet || error "failed to mount $facet"
8430 for facet in $scrub_list; do
8431 $rcmd $LCTL lfsck_start -M $FSNAME-$facet -t scrub ||
8432 error "failed to start OI scrub on $facet"
8435 # sleep 3 seconds for scrub done
8441 run_test 108b "migrate from ZFS to ldiskfs"
8444 # set number of permanent parameters
8446 test_109_set_params() {
8449 set_persistent_param_and_check mds \
8450 "mdd.$fsname-MDT0000.atime_diff" \
8451 "$fsname-MDT0000.mdd.atime_diff" \
8453 set_persistent_param_and_check mds \
8454 "mdd.$fsname-MDT0000.atime_diff" \
8455 "$fsname-MDT0000.mdd.atime_diff" \
8457 set_persistent_param_and_check client \
8458 "llite.$fsname*.max_read_ahead_mb" \
8459 "$fsname.llite.max_read_ahead_mb" \
8461 set_persistent_param_and_check client \
8462 "llite.$fsname*.max_read_ahead_mb" \
8463 "$fsname.llite.max_read_ahead_mb" \
8465 create_pool $fsname.pool1 || error "create pool failed"
8466 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
8467 error "pool_add failed"
8468 do_facet mgs $LCTL pool_remove $fsname.pool1 OST0000 ||
8469 error "pool_remove failed"
8470 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
8471 error "pool_add failed"
8475 # check permanent parameters
8477 test_109_test_params() {
8480 local atime_diff=$(do_facet mds $LCTL \
8481 get_param -n mdd.$fsname-MDT0000.atime_diff)
8482 [ $atime_diff == 63 ] || error "wrong mdd parameter after clear_conf"
8483 local max_read_ahead_mb=$(do_facet client $LCTL \
8484 get_param -n llite.$fsname*.max_read_ahead_mb)
8485 [ $max_read_ahead_mb == 64 ] ||
8486 error "wrong llite parameter after clear_conf"
8487 local ost_in_pool=$(do_facet mds $LCTL pool_list $fsname.pool1 |
8488 grep -v "^Pool:" | sed 's/_UUID//')
8489 [ $ost_in_pool = "$fsname-OST0000" ] ||
8490 error "wrong pool after clear_conf"
8494 # run lctl clear_conf, store CONFIGS before and after that
8496 test_109_clear_conf()
8498 local clear_conf_arg=$1
8501 if ! combined_mgs_mds ; then
8502 mgsdev=$(mgsdevname)
8503 stop_mgs || error "stop_mgs failed"
8504 start_mgs "-o nosvc" || error "start_mgs nosvc failed"
8506 mgsdev=$(mdsdevname 1)
8507 start_mdt 1 "-o nosvc" || error "start_mdt 1 nosvc failed"
8510 do_facet mgs "rm -rf $TMP/${tdir}/conf1; mkdir -p $TMP/${tdir}/conf1;" \
8511 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf1\\\" \
8515 # the command being tested
8517 do_facet mgs $LCTL clear_conf $clear_conf_arg ||
8518 error "clear_conf failed"
8519 if ! combined_mgs_mds ; then
8520 stop_mgs || error "stop_mgs failed"
8522 stop_mdt 1 || error "stop_mdt 1 failed"
8525 do_facet mgs "rm -rf $TMP/${tdir}/conf2; mkdir -p $TMP/${tdir}/conf2;" \
8526 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf2\\\" \
8530 test_109_file_shortened() {
8532 local sizes=($(do_facet mgs "stat -c %s " \
8533 "$TMP/${tdir}/conf1/CONFIGS/$file" \
8534 "$TMP/${tdir}/conf2/CONFIGS/$file"))
8535 [ ${sizes[1]} -lt ${sizes[0]} ] && return 0
8541 [ $MDS1_VERSION -lt $(version_code 2.10.59) ] &&
8542 skip "Needs MDS version 2.10.59 or later."
8543 [ "$(facet_fstype mgs)" == zfs ] &&
8544 skip "LU-8727: no implementation for ZFS"
8549 client_up || error "client_up failed"
8552 # set number of permanent parameters
8554 test_109_set_params $FSNAME
8556 umount_client $MOUNT || error "umount_client failed"
8557 stop_ost || error "stop_ost failed"
8558 stop_mds || error "stop_mds failed"
8560 test_109_clear_conf $FSNAME
8562 # make sure that all configs are cleared
8564 test_109_file_shortened $FSNAME-MDT0000 ||
8565 error "failed to clear MDT0000 config"
8566 test_109_file_shortened $FSNAME-client ||
8567 error "failed to clear client config"
8572 # check that configurations are intact
8574 test_109_test_params $FSNAME
8579 destroy_test_pools || error "destroy test pools failed"
8583 run_test 109a "test lctl clear_conf fsname"
8587 [ $MDS1_VERSION -lt $(version_code 2.10.59) ] &&
8588 skip "Needs MDS version 2.10.59 or later."
8589 [ "$(facet_fstype mgs)" == zfs ] &&
8590 skip "LU-8727: no implementation for ZFS"
8595 client_up || error "client_up failed"
8598 # set number of permanent parameters
8600 test_109_set_params $FSNAME
8602 umount_client $MOUNT || error "umount_client failed"
8603 stop_ost || error "stop_ost failed"
8604 stop_mds || error "stop_mds failed"
8606 test_109_clear_conf $FSNAME-MDT0000
8608 # make sure that only one config is cleared
8610 test_109_file_shortened $FSNAME-MDT0000 ||
8611 error "failed to clear MDT0000 config"
8612 test_109_file_shortened $FSNAME-client &&
8613 error "failed to clear client config"
8617 # check that configurations are intact
8619 test_109_test_params $FSNAME
8624 destroy_test_pools || error "destroy test pools failed"
8628 run_test 109b "test lctl clear_conf one config"
8632 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
8633 skip "Only applicable to ldiskfs-based MDTs"
8635 do_facet $SINGLEMDS $DEBUGFS -w -R supported_features |grep large_dir ||
8636 skip "large_dir option is not supported on MDS"
8637 do_facet ost1 $DEBUGFS -w -R supported_features | grep large_dir ||
8638 skip "large_dir option is not supported on OSS"
8640 stopall # stop all targets before modifying the target counts
8641 local old_mdscount=$MDSCOUNT
8642 local old_ostcount=$OSTCOUNT
8644 stack_trap "MDSCOUNT=$old_mdscount OSTCOUNT=$old_ostcount" EXIT
8648 # ext4_dir_entry_2 struct size:264
8649 # dx_root struct size:8
8650 # dx_node struct size:8
8651 # dx_entry struct size:8
8652 # For 1024 bytes block size.
8653 # First level directory entries: 126
8654 # Second level directory entries: 127
8655 # Entries in leaf: 3
8656 # For 2 levels limit: 48006
8657 # For 3 levels limit : 6096762
8658 # Create 80000 files to safely exceed 2-level htree limit.
8659 CONF_SANITY_110_LINKS=${CONF_SANITY_110_LINKS:-80000}
8661 # can fit at most 3 filenames per 1KB leaf block, but each
8662 # leaf/index block will only be 3/4 full before split at each level
8663 (( MDSSIZE < CONF_SANITY_110_LINKS / 3 * 4/3 * 4/3 )) &&
8664 CONF_SANITY_110_LINKS=$((MDSSIZE * 3 * 3/4 * 3/4))
8666 combined_mgs_mds || replace=" --replace "
8667 local opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
8668 $replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
8669 if [[ $opts != *large_dir* ]]; then
8670 if [[ $opts != *mkfsoptions* ]]; then
8671 opts+=" --mkfsoptions=\\\"-O large_dir -b 1024 -i 65536\\\""
8673 opts="${opts//--mkfsoptions=\\\"/ \
8674 --mkfsoptions=\\\"-O large_dir -b 1024 -i 65536 }"
8677 echo "MDT params: $opts"
8679 combined_mgs_mds || start_mgs
8680 add mds1 $opts || error "add mds1 failed with new params"
8681 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
8683 opts="$(mkfs_opts ost1 $(ostdevname 1)) \
8684 $replace --reformat $(ostdevname 1) $(ostvdevname 1)"
8686 if [[ $opts != *large_dir* ]]; then
8687 if [[ $opts != *mkfsoptions* ]]; then
8688 opts+=" --mkfsoptions=\\\"-O large_dir\\\" "
8690 opts="${opts//--mkfsoptions=\\\"/ \
8691 --mkfsoptions=\\\"-O large_dir }"
8694 echo "OST params: $opts"
8695 add ost1 $opts || error "add ost1 failed with new params"
8696 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
8698 MOUNT_2=yes mountcli || error "mount clients failed"
8700 mkdir -v $DIR/$tdir || error "cannot create $DIR/$tdir"
8701 local pids count=0 group=0
8703 echo "creating $CONF_SANITY_110_LINKS in total"
8704 while (( count < CONF_SANITY_110_LINKS )); do
8705 local len=$((253 - $(wc -c <<<"$tfile-$group-40000-")))
8706 local dir=DIR$((group % 2 + 1))
8707 local target=${!dir}/$tdir/$tfile-$group
8708 local long=$target-$(generate_name $len)-
8709 local create=$((CONF_SANITY_110_LINKS - count))
8711 (( create > 40000 )) && create=40000
8712 touch $target || error "creating $target failed"
8713 echo "creating $create hard links to $target"
8714 createmany -l $target $long $create &
8717 count=$((count + create))
8718 group=$((group + 1))
8720 echo "waiting for PIDs$pids to complete"
8721 wait $pids || error "createmany failed after $group groups"
8723 umount_client $MOUNT2 -f
8726 run_e2fsck $(facet_active_host mds1) $(mdsdevname 1) -n
8727 MDSCOUNT=$old_mdscount
8728 OSTCOUNT=$old_ostcount
8730 run_test 110 "Adding large_dir with 3-level htree"
8733 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
8734 skip "Only applicable to ldiskfs-based MDTs"
8736 is_dm_flakey_dev $SINGLEMDS $(mdsdevname 1) &&
8737 skip "This test can not be executed on flakey dev"
8739 do_facet $SINGLEMDS $DEBUGFS -w -R supported_features |grep large_dir ||
8740 skip "large_dir option is not supported on MDS"
8742 do_facet ost1 $DEBUGFS -w -R supported_features | grep large_dir ||
8743 skip "large_dir option is not supported on OSS"
8745 # cleanup before changing target counts
8747 local old_mdscount=$MDSCOUNT
8748 local old_ostcount=$OSTCOUNT
8749 local old_mdssize=$MDSSIZE
8751 stack_trap "MDSSIZE=$MDSSIZE MDSCOUNT=$MDSCOUNT OSTCOUNT=$OSTCOUNT" EXIT
8754 (( MDSSIZE < 2400000 )) && MDSSIZE=2400000 # need at least 2.4GB
8756 local mdsdev=$(mdsdevname 1)
8757 combined_mgs_mds || replace=" --replace "
8758 local opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
8759 $replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
8760 if [[ $opts != *large_dir* ]]; then
8761 if [[ $opts != *mkfsoptions* ]]; then
8762 opts+=" --mkfsoptions=\\\"-O large_dir -i 1048576 \\\" "
8764 opts="${opts//--mkfsoptions=\\\"/ \
8765 --mkfsoptions=\\\"-O large_dir -i 1048576 }"
8768 echo "MDT params: $opts"
8770 combined_mgs_mds || start_mgs
8771 __touch_device mds 1
8772 add mds1 $opts || error "add mds1 failed with new params"
8773 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
8775 opts="$(mkfs_opts ost1 $(ostdevname 1)) \
8776 $replace --reformat $(ostdevname 1) $(ostvdevname 1)"
8777 if [[ $opts != *large_dir* ]]; then
8778 if [[ $opts != *mkfsoptions* ]]; then
8779 opts+=" --mkfsoptions=\\\"-O large_dir \\\""
8781 opts="${opts//--mkfsoptions=\\\"/ --mkfsoptions=\\\"-O large_dir }"
8784 echo "OST params: $opts"
8785 __touch_device ost 1
8786 add ost1 $opts || error "add ost1 failed with new params"
8787 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
8789 MOUNT_2=yes mountcli
8790 mkdir $DIR/$tdir || error "cannot create $DIR/$tdir"
8792 lfs df -i $DIR/$tdir
8796 local start=$SECONDS
8798 local dirmax=$((2 << 30))
8804 while (( !needskip & dirsize < dirmax )); do
8807 for cli in ${CLIENTS//,/ }; do
8808 local len=$((253 - $(wc -c <<<"$cli-$group-60000-")))
8809 local target=$cli-$group
8810 local long=$DIR/$tdir/$target-$(generate_name $len)-
8812 RPWD=$DIR/$tdir do_node $cli touch $target ||
8813 error "creating $target failed"
8814 echo "creating 60000 hardlinks to $target"
8815 RPWD=$DIR/$tdir do_node $cli createmany -l $target $long 60000 &
8818 group=$((group + 1))
8820 long=$DIR2/$tdir/$target-$(generate_name $len)-
8822 RPWD=$DIR2/$tdir do_node $cli touch $target ||
8823 error "creating $target failed"
8824 echo "creating 60000 hardlinks to $target"
8825 RPWD=$DIR2/$tdir do_node $cli createmany -l $target $long 60000 &
8828 group=$((group + 1))
8830 echo "waiting for PIDs$pids to complete"
8831 wait $pids || error "createmany failed after $group groups"
8832 dirsize=$(stat -c %s $DIR/$tdir)
8833 taken=$((SECONDS - start))
8834 rate=$((dirsize / taken))
8835 left=$(((dirmax - dirsize) / rate))
8836 num=$((group * 60000))
8837 echo "estimate ${left}s left after $num files / ${taken}s"
8838 # if the estimated time remaining is too large (it may change
8839 # over time as the create rate is not constant) then exit
8840 # without declaring a failure.
8841 (( left > 1200 )) && needskip=1
8844 umount_client $MOUNT2 -f
8847 (( $needskip )) && skip "ETA ${left}s after $num files / ${taken}s is too long"
8849 run_e2fsck $(facet_active_host mds1) $(mdsdevname 1) -n
8850 MDSCOUNT=$old_mdscount
8851 OSTCOUNT=$old_ostcount
8852 MDSSIZE=$old_mdssize
8854 run_test 111 "Adding large_dir with over 2GB directory"
8857 start_mds || error "MDS start failed"
8858 start_ost || error "OSS start failed"
8859 echo "start ost2 service on $(facet_active_host ost2)"
8860 start ost2 $(ostdevname 2) $(csa_add "$OST_MOUNT_OPTS" -o no_precreate) ||
8861 error "start ost2 facet failed"
8862 local val=$(do_facet ost2 \
8863 "$LCTL get_param -n obdfilter.$FSNAME-OST0001*.no_precreate")
8864 (( $val == 1 )) || error "obdfilter.$FSNAME-OST0001*.no_precreate=$val"
8866 mount_client $MOUNT || error "mount client failed"
8867 wait_osc_import_state mds1 ost1 FULL
8868 wait_osc_import_state client ost1 FULL
8869 wait_osc_import_state client ost2 FULL
8871 $LFS setstripe -i 0 $DIR/$tfile.0 ||
8872 error "problem creating $tfile.0 on OST0000"
8873 $LFS setstripe -i 1 $DIR/$tfile.1 && $LFS getstripe $DIR/$tfile.1 &&
8874 (( $($LFS getstripe -i $DIR/$tfile.1) == 1 )) &&
8875 error "allowed to create $tfile.1 on OST0001"
8876 do_facet ost2 $LCTL set_param obdfilter.*.no_precreate=0
8878 $LFS setstripe -i 1 $DIR/$tfile.2 ||
8879 error "failed to create $tfile.2 on ost1 facet"
8880 # files not cleaned with ONLY_REPEAT because of client unmount below
8881 rm -f $DIR/$tfile.[012]
8882 stop_ost2 || error "stop ost2 facet failed"
8885 run_test 112 "mount OST with nocreate option"
8891 rm -f $TMP/$tdir/lustre-mdt
8895 if [ "$mds1_FSTYPE" != ldiskfs ]; then
8896 skip "Only applicable to ldiskfs-based MDTs"
8898 [ -n "$FILESET" ] && skip "Not functional for FILESET set"
8900 local dbfs_ver=$(do_facet $SINGLEMDS $DEBUGFS -V 2>&1)
8902 echo "debugfs version: $dbfs_ver"
8903 echo "$dbfs_ver" | egrep -w "1.44.3.wc1|1.44.5.wc1|1.45.2.wc1" &&
8904 skip_env "This version of debugfs doesn't show inode number"
8906 IMAGESIZE=$((3072 << 30)) # 3072 GiB
8910 local saved_flakey=${FLAKEY}
8912 stack_trap "FLAKEY=$saved_flakey" EXIT
8919 do_facet mds1 "hostname; lctl dl; mount"
8921 do_facet ost1 "hostname; lctl dl; mount"
8922 # We need MDT size 3072GB, because it is smallest
8923 # partition that can store 2B inodes
8924 do_facet $SINGLEMDS "mkdir -p $TMP/$tdir"
8925 local mdsimgname=$TMP/$tdir/lustre-mdt
8927 do_facet $SINGLEMDS "rm -f $mdsimgname"
8928 do_facet $SINGLEMDS "touch $mdsimgname"
8929 trap cleanup_115 RETURN EXIT
8930 do_facet $SINGLEMDS "$TRUNCATE $mdsimgname $IMAGESIZE" ||
8931 skip "Backend FS doesn't support sparse files"
8932 local mdsdev=$(do_facet $SINGLEMDS "losetup -f")
8934 do_facet $SINGLEMDS "losetup $mdsdev $mdsimgname"
8936 local mds_opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
8937 --mkfsoptions='-O ea_inode,^resize_inode,meta_bg \
8938 -N 2247484000 -E lazy_itable_init' --device-size=$IMAGESIZE"
8939 add mds1 $mds_opts --mgs --reformat $mdsdev ||
8940 skip_env "format large MDT failed"
8941 opts="$(mkfs_opts ost1 $(ostdevname 1)) \
8942 $replace --reformat $(ostdevname 1) $(ostvdevname 1)"
8943 add ost1 $opts || error "add ost1 failed with new params"
8944 start $SINGLEMDS $mdsdev $MDS_MOUNT_OPTS || error "start MDS failed"
8945 start_ost || error "start OSS failed"
8946 mount_client $MOUNT || error "mount client failed"
8948 mkdir -p $DIR/$tdir || error "mkdir $DIR/$tdir fail"
8949 goal="/sys/fs/ldiskfs/$(basename $mdsdev)/inode_goal"
8951 # 2147483648 is 0x80000000
8952 do_facet $SINGLEMDS "echo 2147483648 >> $goal; grep . $goal"
8953 touch $DIR/$tdir/$tfile
8955 # attrs from 1 to 16 go to block, 17th - to inode
8956 for i in {1..17}; do
8957 local nm="trusted.ea$i"
8959 setfattr -n $nm -v $(printf "xattr%0250d" $i) $DIR/$tdir/$tfile
8962 do_facet $SINGLEMDS $DEBUGFS -c -R "stat ROOT/$tdir/$tfile" $mdsdev
8964 # inode <2147483649> trusted.ea16 (255)
8965 local inode_num=$(do_facet $SINGLEMDS \
8966 "$DEBUGFS -c -R 'stat ROOT/$tdir/$tfile' $mdsdev" |
8967 awk '/ea17/ { print $2 }' |
8968 sed -e 's/>//' -e 's/<//' -e 's/\"//')
8969 echo "inode num: $inode_num"
8970 [ $inode_num -ge 2147483648 ] || error "inode $inode_num too small"
8971 do_facet $SINGLEMDS "losetup -d $mdsdev"
8974 run_test 115 "Access large xattr with inodes number over 2TB"
8977 [ "$mds1_FSTYPE" != ldiskfs ] && skip "ldiskfs only test"
8978 [ "$MDS1_VERSION" -lt $(version_code 2.10.59) ] &&
8979 skip "Need server version at least 2.10.59"
8980 do_facet $SINGLEMDS which mkfs.xfs ||
8981 skip_env "No mkfs.xfs installed"
8986 local tmpmnt=$TMP/$tdir
8987 local mdtimg=$tfile-mdt0
8989 do_facet $SINGLEMDS mkdir -p $tmpmnt
8990 stack_trap "do_facet $SINGLEMDS rmdir $tmpmnt" EXIT
8992 do_facet $SINGLEMDS touch $TMP/$mdtimg
8993 stack_trap "do_facet $SINGLEMDS rm -f $TMP/$mdtimg" EXIT
8994 do_facet $SINGLEMDS mkfs -t xfs -d file,size=1t,name=$TMP/$mdtimg ||
8995 error "mkfs temporary xfs image"
8997 do_facet $SINGLEMDS mount $TMP/$mdtimg $tmpmnt ||
8998 error "mount temporary xfs image"
8999 stack_trap "do_facet $SINGLEMDS umount $tmpmnt" EXIT
9000 local old_mdssize=$MDSSIZE
9001 local old_mdsisize=$MDSISIZE
9003 MDSSIZE=$((17 * 1024 * 1024 * 1024)) # 17T MDT
9004 MDSISIZE=$((16 << 20))
9005 local opts17t="$(mkfs_opts $SINGLEMDS)"
9007 MDSSIZE=$old_mdssize
9008 MDSISIZE=$old_mdsisize
9010 do_facet $SINGLEMDS $MKFS \
9011 ${opts17t/-E /-E lazy_itable_init,lazy_journal_init,} \
9013 error "failed to mkfs for $tmpmnt/$mdtimg"
9015 do_facet $SINGLEMDS $TUNE2FS -l $tmpmnt/$mdtimg |
9016 grep -qw 'features.*extent' || error "extent should be enabled"
9019 run_test 116 "big size MDT support"
9023 do_facet ost1 "$LCTL set_param ost.OSS.ost_io.nrs_policies=fifo"
9024 do_facet ost1 "$LCTL get_param -n ost.OSS.ost_io.nrs_tbf_rule" &&
9025 error "get_param should fail"
9026 cleanup || error "cleanup failed with rc $?"
9028 run_test 117 "lctl get_param return errors properly"
9034 [[ "$MDSCOUNT" -ge 2 ]] || skip "Need more at least 2 MDTs"
9036 had_config=$(do_facet mds1 "$LCTL get_param debug | grep config")
9037 do_facet mds1 "$LCTL set_param debug=+config"
9038 do_facet mds1 "$LCTL clear"
9041 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)" &>/dev/null
9042 # mount after writeconf will make "add osp" added to mdt0 config:
9043 # 53 (224)marker 60 (flags=0x01, v2.5.1.0) lustre-MDT0001 'add osp'
9044 # 54 (080)add_uuid nid=... 0: 1:...
9045 # 55 (144)attach 0:lustre-MDT0001-osp-MDT0000 1:osp 2:...
9046 # 56 (144)setup 0:lustre-MDT0001-osp-MDT0000 1:... 2:...
9047 # 57 (136)modify_mdc_tgts add 0:lustre-MDT0000-mdtlov 1:... 2:1 3:1
9048 # duplicate modify_mdc_tgts caused crashes
9051 # using larger debug_mb size to avoid lctl dk log truncation
9052 size_mb=$((DEBUG_SIZE_SAVED * 4))
9055 # though config processing stops after failed attach and setup
9056 # it will proceed after the failed command after each writeconf
9057 # this is the original scenario of the issue
9058 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)" &>/dev/null
9059 do_facet mds1 "$LCTL set_param debug_mb=$size_mb"
9062 wait_update_facet_cond mds1 \
9063 "$LCTL dk | grep -c Processed.log.$FSNAME-MDT0000" \
9068 [[ -z "$had_config" ]] && do_facet mds1 lctl set_param debug=-config
9072 run_test 119 "writeconf on slave mdt shouldn't duplicate mdc/osp and crash"
9074 test_120() { # LU-11130
9075 [ "$MDSCOUNT" -lt 2 ] && skip "mdt count < 2"
9076 [ "$mds1_FSTYPE" != ldiskfs ] &&
9077 skip "ldiskfs only test"
9078 [ "$MDS1_VERSION" -lt $(version_code 2.11.56) ] &&
9079 skip "Need DNE2 capable MD target with LU-11130 fix"
9083 local mds1host=$(facet_active_host mds1)
9084 local mds1dev=$(mdsdevname 1)
9086 $LFS mkdir -i 1 $DIR/$tdir
9087 $LFS mkdir -i 0 $DIR/$tdir/mds1dir
9089 ln -s foo $DIR/$tdir/bar
9090 mv $DIR/$tdir/bar $DIR/$tdir/mds1dir/bar2 ||
9091 error "cross-target rename failed"
9095 run_e2fsck $mds1host $mds1dev "-n"
9097 run_test 120 "cross-target rename should not create bad symlinks"
9101 start_mgsmds || error "MGS MDS Start failed"
9103 stop_mds || error "Stopping MDSes failed"
9107 stop_mds || error "Stopping MDSes failed"
9109 run_test 121 "failover MGS"
9112 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
9113 [[ "$OST1_VERSION" -ge $(version_code 2.11.53) ]] ||
9114 skip "Need OST version at least 2.11.53"
9117 LOAD_MODULES_REMOTE=true load_modules
9118 #define OBD_FAIL_OFD_SET_OID 0x1e0
9119 do_facet ost1 $LCTL set_param fail_loc=0x00001e0
9122 $LFS mkdir -i1 -c1 $DIR/$tdir
9123 $LFS setstripe -i0 -c1 $DIR/$tdir
9124 do_facet ost1 $LCTL set_param fail_loc=0
9125 createmany -o $DIR/$tdir/file_ 1000 ||
9126 error "Fail to create a new sequence"
9130 run_test 122a "Check OST sequence update"
9133 [[ "$OST1_VERSION" -ge $(version_code 2.11.53) ]] ||
9134 skip "Need OST version at least 2.11.53"
9138 LOAD_MODULES_REMOTE=true load_modules
9139 #define OBD_FAIL_OFD_SET_OID 0x1e0
9140 do_facet ost1 $LCTL set_param fail_loc=0x00001e0
9142 stack_trap cleanup EXIT
9144 do_facet ost1 $LCTL set_param obdfilter.*.precreate_batch=256
9145 $LFS mkdir -i0 -c1 $DIR/$tdir || error "failed to create directory"
9146 $LFS setstripe -i0 -c1 $DIR/$tdir || error "failed to setstripe"
9147 do_facet ost1 $LCTL set_param fail_loc=0
9148 # overflow IDIF 32bit and create > OST_MAX_PRECREATE*5
9149 # so a new wrong sequence would differ from an original with error
9150 #define OST_MAX_PRECREATE 20000
9151 local ost_max_precreate=20100
9152 local num_create=$(( ost_max_precreate * 5 ))
9154 # Check the number of inodes available on OST0
9156 local ifree=$($LFS df -i $MOUNT |
9157 awk '/OST0000/ { print $4 }'; exit ${PIPESTATUS[0]})
9159 log "On OST0, $ifree inodes available. Want $num_create. rc=$?"
9161 if [ $ifree -lt 10000 ]; then
9162 files=$(( ifree - 50 ))
9167 local j=$((num_create / files + 1))
9169 for i in $(seq 1 $j); do
9170 createmany -o $DIR/$tdir/$tfile-$i- $files ||
9171 error "createmany fail create $files files: $?"
9172 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
9173 error "unlinkmany failed unlink $files files"
9176 touch $DIR/$tdir/$tfile
9178 #we need a write req during recovery for ofd_seq_load
9180 dd if=/dev/urandom of=$DIR/$tdir/$tfile bs=1024k count=1 oflag=sync ||
9181 error "failed to write file"
9183 # OBD_FAIL_OST_CREATE_NET 0x204
9184 do_facet ost1 $LCTL set_param fail_loc=0x80000204
9186 createmany -o $DIR/$tdir/file_ 100
9189 err=$(do_facet ost1 dmesg | tac | sed "/Recovery over/,$ d" |
9190 grep "OST replaced or reformatted")
9191 [ -z "$err" ] || error $err
9193 run_test 122b "Check OST sequence wouldn't change when IDIF 32bit overflows"
9196 remote_mgs_nodsh && skip "remote MGS with nodsh"
9197 [ -d $MOUNT/.lustre ] || setup
9199 # test old logid format until removal from llog_ioctl.c::str2logid()
9200 if [ $MGS_VERSION -lt $(version_code 3.1.53) ]; then
9201 do_facet mgs $LCTL dl | grep MGS
9202 do_facet mgs "$LCTL --device %MGS llog_print \
9203 \\\\\\\$$FSNAME-client 1 10" ||
9204 error "old llog_print failed"
9207 # test new logid format
9208 if [ $MGS_VERSION -ge $(version_code 2.9.53) ]; then
9209 do_facet mgs "$LCTL --device MGS llog_print $FSNAME-client" ||
9210 error "new llog_print failed"
9213 run_test 123aa "llog_print works with FIDs and simple names"
9216 remote_mgs_nodsh && skip "remote MGS with nodsh"
9217 [[ $MGS_VERSION -gt $(version_code 2.11.51) ]] ||
9218 skip "Need server with working llog_print support"
9220 [ -d $MOUNT/.lustre ] || setup
9226 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9227 mgs_arg="--device MGS"
9229 orig_val=$(do_facet mgs $LCTL get_param jobid_name)
9230 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME"
9232 yaml=$(do_facet mgs $LCTL $mgs_arg llog_print params |
9233 grep jobid_name | tail -n 1)
9235 local param=$(awk '{ print $10 }' <<< "$yaml")
9236 local val=$(awk '{ print $12 }' <<< "$yaml")
9237 #return to the default
9238 do_facet mgs $LCTL set_param -P jobid_name=$orig_val
9239 [ $val = "TESTNAME" ] || error "bad value: $val"
9240 [ $param = "jobid_name," ] || error "Bad param: $param"
9242 run_test 123ab "llog_print params output values from set_param -P"
9244 test_123ac() { # LU-11566
9245 remote_mgs_nodsh && skip "remote MGS with nodsh"
9246 do_facet mgs "$LCTL help llog_print" 2>&1 | grep -q -- --start ||
9247 skip "Need 'lctl llog_print --start' on MGS"
9253 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9254 mgs_arg="--device MGS"
9256 [ -d $MOUNT/.lustre ] || setup
9258 # - { index: 10, event: add_uuid, nid: 192.168.20.1@tcp(0x20000c0a81401,
9259 # node: 192.168.20.1@tcp }
9260 do_facet mgs $LCTL $mgs_arg \
9261 llog_print --start $start --end $end $FSNAME-client | tr -d , |
9262 while read DASH BRACE INDEX idx EVENT BLAH BLAH BLAH; do
9263 (( idx >= start )) || error "llog_print index $idx < $start"
9264 (( idx <= end )) || error "llog_print index $idx > $end"
9267 run_test 123ac "llog_print with --start and --end"
9269 test_123ad() { # LU-11566
9270 remote_mgs_nodsh && skip "remote MGS with nodsh"
9271 # older versions of lctl may not print all records properly
9272 do_facet mgs "$LCTL help llog_print" 2>&1 | grep -q -- --start ||
9273 skip "Need 'lctl llog_print --start' on MGS"
9275 [ -d $MOUNT/.lustre ] || setup
9277 # append a new record, to avoid issues if last record was cancelled
9278 local old=$($LCTL get_param -n osc.*-OST0000-*.max_dirty_mb | head -1)
9279 do_facet mgs $LCTL conf_param $FSNAME-OST0000.osc.max_dirty_mb=$old
9281 # logid: [0x3:0xa:0x0]:0
9285 local num=$(do_facet mgs $LCTL --device MGS llog_info $FSNAME-client |
9286 awk '/last_index:/ { print $2 - 1 }')
9288 # - { index: 71, event: set_timeout, num: 0x14, param: sys.timeout=20 }
9289 local last=$(do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
9290 tail -1 | awk '{ print $4 }' | tr -d , )
9291 (( last == num )) || error "llog_print only showed $last/$num records"
9293 run_test 123ad "llog_print shows all records"
9295 test_123ae() { # LU-11566
9304 remote_mgs_nodsh && skip "remote MGS with nodsh"
9305 [ -d $MOUNT/.lustre ] || setupall
9307 max=$($LCTL get_param -n osc.*-OST0000-*.max_dirty_mb | head -1)
9308 pgs=$($LCTL get_param -n osc.*-OST0000-*.max_pages_per_rpc | head -1)
9309 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9310 mgs_arg="--device MGS"
9312 if do_facet mgs "$LCTL help llog_cancel" 2>&1| grep -q -- --log_id; then
9313 # save one set_param -P record in case none exist
9315 do_facet mgs $LCTL set_param -P osc.*.max_pages_per_rpc=$pgs
9316 stack_trap "do_facet mgs $LCTL set_param -P -d \
9317 osc.*.max_pages_per_rpc"
9320 orig=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9321 tail -1 | awk '{ print $4 }' | tr -d , )
9322 do_facet mgs $LCTL set_param -P osc.*.max_dirty_mb=$max
9323 do_facet mgs $LCTL $mgs_arg llog_print $log | tail -1 |
9324 grep "parameter: osc.*.max_dirty_mb" ||
9325 error "new set_param -P wasn't stored in params log"
9327 # - { index: 71, event: set_param, device: general,
9328 # param: osc.*.max_dirty_mb, value: 256 }
9329 id=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9330 tail -1 | awk '{ print $4 }' | tr -d , )
9331 do_facet mgs $LCTL $mgs_arg llog_cancel $log --log_idx=$id
9332 local new=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9333 tail -1 | awk '{ print $4 }' | tr -d , )
9334 (( new == orig )) ||
9335 error "new llog_cancel now $new, not at $orig records"
9338 # test old positional parameters for a while still
9339 if [ "$MGS_VERSION" -le $(version_code 3.1.53) ]; then
9342 do_facet mgs $LCTL conf_param \
9343 $FSNAME-OST0000.osc.max_pages_per_rpc=$pgs
9344 stack_trap "do_facet mgs $LCTL conf_param -d \
9345 $FSNAME-OST0000.osc.max_pages_per_rpc"
9347 orig=$(do_facet mgs $LCTL --device MGS llog_print $log |
9348 tail -1 | awk '{ print $4 }' | tr -d , )
9349 do_facet mgs $LCTL conf_param $FSNAME-OST0000.osc.max_dirty_mb=$max
9350 do_facet mgs $LCTL --device MGS llog_print $log |
9351 tail -1 | grep "parameter: osc.max_dirty_mb" ||
9352 error "old conf_param wasn't stored in params log"
9353 do_facet mgs $LCTL --device MGS llog_print $log
9354 # - { index: 71, event: conf_param, device: testfs-OST0000-osc,
9355 # param: osc.max_dirty_mb=256 }
9356 id=$(do_facet mgs $LCTL --device MGS llog_print $log |
9357 tail -1 | awk '{ print $4 }' | tr -d , )
9358 do_facet mgs $LCTL --device MGS llog_cancel $log $id
9359 do_facet mgs $LCTL --device MGS llog_print $log
9360 new=$(do_facet mgs $LCTL --device MGS llog_print $log |
9361 tail -1 | awk '{ print $4 }' | tr -d , )
9362 (( new == orig )) ||
9363 error "old llog_cancel now $new, not at $orig records"
9366 run_test 123ae "llog_cancel can cancel requested record"
9368 test_123af() { #LU-13609
9369 [ "$MGS_VERSION" -ge $(version_code 2.13.54) -a \
9370 "$MDS1_VERSION" -ge $(version_code 2.13.54) ] ||
9371 skip "Need both MGS and MDS version at least 2.13.54"
9373 [ -d $MOUNT/.lustre ] || setupall
9374 stack_trap "do_facet mds1 $LCTL set_param fail_loc=0" EXIT
9384 for device in "MGS" "$FSNAME-MDT0000"; do
9385 cmd="--device $device llog_catlist"
9386 echo "lctl $cmd ..."
9387 if [ "$device" = "MGS" ]; then
9392 orig_clist=($(do_facet $facet $LCTL $cmd | awk '{ print $2 }'))
9393 orig_count=${#orig_clist[@]}
9394 echo "orig_clist: ${orig_clist[*]}"
9396 #define OBD_FAIL_CATLIST 0x131b
9397 #fetch to llog records from the second one
9398 do_facet $facet $LCTL set_param fail_loc=0x131b fail_val=2
9400 new_clist=($(do_facet $facet $LCTL $cmd | awk '{ print $2 }'))
9401 new_count=${#new_clist[@]}
9402 echo "new_clist: ${new_clist[*]}"
9404 [ $new_count -eq $((orig_count - 1)) ] ||
9405 error "$new_count != $orig_count - 1"
9406 for i in $(seq 0 $new_count); do
9408 [ "${orig_clist[$j]}" = "${new_clist[$i]}" ] ||
9409 error "${orig_clist[$j]} != ${new_clist[$i]}"
9411 do_facet mds1 $LCTL set_param fail_loc=0
9415 run_test 123af "llog_catlist can show all config files correctly"
9417 test_123ag() { # LU-15142
9421 remote_mgs_nodsh && skip "remote MGS with nodsh"
9422 (( $MGS_VERSION >= $(version_code 2.14.55) )) ||
9423 skip "Need server version least 2.14.55"
9425 [ -d $MOUNT/.lustre ] || setup
9427 orig_val=$(do_facet mgs $LCTL get_param jobid_name)
9428 stack_trap "do_facet mgs $LCTL set_param -P jobid_name=$orig_val"
9430 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME1"
9431 do_facet mgs $LCTL set_param -P -d jobid_name
9432 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
9434 (( rec == 0 )) || error "parameter was not deleted, check #1"
9435 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME1"
9436 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
9438 (( rec == 1)) || error "parameter is not set"
9439 # usage with ordinary set_param format works too
9440 do_facet mgs $LCTL set_param -P -d jobid_name="ANY"
9441 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
9443 (( rec == 0 )) || error "parameter was not deleted, check #2"
9445 run_test 123ag "llog_print skips values deleted by set_param -P -d"
9448 remote_mgs_nodsh && skip "remote MGS with nodsh"
9450 [ -d $MOUNT/.lustre ] || setup
9451 local yaml_file="$TMP/$tfile.yaml"
9453 do_facet mgs rm "$yaml_file"
9454 local cfgfiles=$(do_facet mgs "lctl --device MGS llog_catlist" |
9455 sed 's/config_log://')
9457 # set jobid_var to a different value for test
9458 local orig_val=$(do_facet mgs $LCTL get_param jobid_var)
9460 do_facet mgs $LCTL set_param -P jobid_var="TESTNAME"
9462 for i in $cfgfiles params; do
9463 do_facet mgs "lctl --device MGS llog_print ${i} >> $yaml_file"
9466 echo "Unmounting FS"
9473 # Reapply the config from before
9474 echo "Setting configuration parameters"
9475 do_facet mgs "lctl set_param -F $yaml_file"
9477 local set_val=$(do_facet mgs $LCTL get_param jobid_var)
9479 do_facet mgs $LCTL set_param -P $orig_val
9481 [ $set_val == "jobid_var=TESTNAME" ] ||
9482 error "$set_val is not TESTNAME"
9484 do_facet mgs rm "$yaml_file"
9487 run_test 123F "clear and reset all parameters using set_param -F"
9491 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
9492 [ -z $mds2failover_HOST ] && skip "needs MDT failover setup"
9495 do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
9496 grep 1.2.3.4@tcp && error "Should not be fake nid"
9497 do_facet mgs $LCTL conf_param $FSNAME-MDT0001.failover.node=1.2.3.4@tcp\
9498 || error "Set params error"
9499 do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
9500 grep 1.2.3.4@tcp || error "Fake nid should be added"
9504 if combined_mgs_mds; then
9505 start_mdt 1 "-o nosvc" ||
9506 error "starting mds with nosvc option failed"
9508 local nid=$(do_facet mds2 $LCTL list_nids | head -1)
9509 local failover_nid=$(do_node $mds2failover_HOST $LCTL list_nids | head -1)
9511 do_facet mgs $LCTL replace_nids $FSNAME-MDT0001 $nid:$failover_nid ||
9512 error "replace_nids execution error"
9514 if combined_mgs_mds; then
9520 echo "lfs setdirstripe"
9521 $LFS setdirstripe -i 1 $MOUNT/$tdir || error "setdirstirpe error"
9524 run_test 124 "check failover after replace_nids"
9526 get_max_sectors_kb() {
9529 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
9530 local max_sectors_path="/sys/block/${dev_base}/queue/max_sectors_kb"
9532 do_facet ${facet} "[[ -e ${max_sectors_path} ]] &&
9533 cat ${max_sectors_path}"
9536 get_max_hw_sectors_kb() {
9539 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
9540 local max_hw_path="/sys/block/${dev_base}/queue/max_hw_sectors_kb"
9542 do_facet ${facet} "[[ -e ${max_hw_path} ]] && cat ${max_hw_path}"
9545 set_max_sectors_kb() {
9549 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
9550 local max_sectors_path="/sys/block/${dev_base}/queue/max_sectors_kb"
9552 do_facet ${facet} "[[ -e ${max_sectors_path} ]] &&
9553 echo ${value} > ${max_sectors_path}"
9556 [[ $rc -ne 0 ]] && echo "Failed to set ${max_sectors_path} to ${value}"
9561 # Return 0 if all slave devices have max_sectors_kb == max_hw_sectors_kb
9562 # Otherwise return > 0
9563 check_slaves_max_sectors_kb()
9567 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
9568 local slaves_dir=/sys/block/${dev_base}/slaves
9569 local slave_devices=$(do_facet ${facet} "ls ${slaves_dir} 2>/dev/null")
9570 [[ -z ${slave_devices} ]] && return 0
9572 local slave max_sectors new_max_sectors max_hw_sectors path
9575 for slave in ${slave_devices}; do
9576 path="/dev/${slave}"
9577 ! is_blkdev ${facet} ${path} && continue
9578 max_sectors=$(get_max_sectors_kb ${facet} ${path})
9579 max_hw_sectors=$(get_max_hw_sectors_kb ${facet} ${path})
9580 new_max_sectors=${max_hw_sectors}
9581 [[ ${new_max_sectors} -gt ${RQ_SIZE_LIMIT} ]] &&
9582 new_max_sectors=${RQ_SIZE_LIMIT}
9584 if [[ ${max_sectors} -ne ${new_max_sectors} ]]; then
9585 echo "${path} ${max_sectors} ${new_max_sectors}"
9588 check_slaves_max_sectors_kb ${facet} ${path}
9597 local facet_list="mgs mds1 ost1"
9598 combined_mgs_mds && facet_list="mgs ost1"
9601 for facet in ${facet_list}; do
9602 [[ $(facet_fstype ${facet}) != ldiskfs ]] &&
9603 skip "ldiskfs only test" &&
9605 ! is_blkdev ${facet} $(facet_device ${facet}) &&
9606 skip "requires all real devices" &&
9611 # We don't increase IO request size limit past 16MB. See comments in
9612 # lustre/utils/libmount_utils_ldiskfs.c:tune_max_sectors_kb()
9613 RQ_SIZE_LIMIT=$((16 * 1024))
9614 local device old_max_sectors new_max_sectors max_hw_sectors
9615 for facet in ${facet_list}; do
9616 device=$(facet_device ${facet})
9617 old_max_sectors=$(get_max_sectors_kb ${facet} ${device})
9618 max_hw_sectors=$(get_max_hw_sectors_kb ${facet} ${device})
9620 # The expected value after l_tunedisk is executed
9621 new_max_sectors=$old_max_sectors
9622 [[ ${new_max_sectors_kb} -gt ${RQ_SIZE_LIMIT} ]] &&
9623 new_max_sectors_kb=${RQ_SIZE_LIMIT}
9625 # Ensure the current value of max_sectors_kb does not equal
9626 # max_hw_sectors_kb, so we can tell whether l_tunedisk did
9628 set_max_sectors_kb ${facet} ${device} $((new_max_sectors - 1))
9630 # Value before l_tunedisk
9631 local pre_max_sectors=$(get_max_sectors_kb ${facet} ${device})
9632 if [[ ${pre_max_sectors} -ne $((new_max_sectors - 1)) ]]; then
9633 echo "unable to satsify test pre-condition:"
9634 echo "${pre_max_sectors} != $((new_max_sectors - 1))"
9639 echo "Before: ${facet} ${device} ${pre_max_sectors} ${max_hw_sectors}"
9641 do_facet ${facet} "l_tunedisk ${device}"
9643 # Value after l_tunedisk
9644 local post_max_sectors=$(get_max_sectors_kb ${facet} ${device})
9646 echo "After: ${facet} ${device} ${post_max_sectors} ${max_hw_sectors}"
9648 if [[ ${facet} != ost1 ]]; then
9649 if [[ ${post_max_sectors} -ne ${pre_max_sectors} ]]; then
9650 echo "l_tunedisk modified max_sectors_kb of ${facet}"
9654 set_max_sectors_kb ${facet} ${device} ${old_max_sectors}
9656 if [[ ${post_max_sectors} -eq ${pre_max_sectors} ]]; then
9657 echo "l_tunedisk failed to modify max_sectors_kb of ${facet}"
9661 check_slaves_max_sectors_kb ${facet} ${device} ||
9668 run_test 125 "check l_tunedisk only tunes OSTs and their slave devices"
9671 [[ "$MDS1_VERSION" -ge $(version_code 2.13.52) ]] ||
9672 skip "Need MDS version at least 2.13.52"
9675 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ../libcfs/libcfs/libcfs
9676 #define OBD_FAIL_OBD_SETUP 0x60d
9677 do_facet mds1 $LCTL set_param fail_loc=0x60d
9678 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_modules &
9679 for i in {1..40}; do
9680 do_facet mds1 lsmod | grep -q osd_$mds1_FSTYPE && break
9683 clear_failloc $SINGLEMDS 20 &
9684 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
9686 run_test 126 "mount in parallel shouldn't cause a crash"
9689 [[ "$ost1_FSTYPE" == ldiskfs ]] || skip "ldiskfs only test"
9693 zconf_umount_clients $RCLIENTS $MOUNT
9695 wait_osp_active ost ${FSNAME}-OST0000 0 1
9696 local osc_tgt="$FSNAME-OST0000-osc-$($LFS getname -i $DIR)"
9697 local avail1=($($LCTL get_param -n osc.${osc_tgt}.kbytesavail))
9699 $LFS setstripe -i 0 $DIR/$tfile || error "failed creating $DIR/$tfile"
9700 dd if=/dev/zero of=$DIR/$tfile bs=1M oflag=direct || true
9702 local avail2=($($LCTL get_param -n osc.${osc_tgt}.kbytesavail))
9704 if ((avail2 * 100 / avail1 > 1)); then
9707 error "more than 1% space left: before=$avail1 after=$avail2"
9710 local mbs=$(($(stat -c %s $DIR/$tfile) / (1024 * 1024)))
9712 dd if=/dev/zero of=$DIR/$tfile bs=1M count=$mbs conv=notrunc \
9713 oflag=direct || error "overwrite failed"
9715 run_test 127 "direct io overwrite on full ost"
9719 combined_mgs_mds && skip "need separate mgs device"
9720 [ "$ost2_FSTYPE" == zfs ] && import_zpool ost2
9723 # Try to apply nolocallogs to the virgin OST. Should fail.
9724 do_facet ost2 "$TUNEFS --nolocallogs $(ostdevname 2)" &&
9725 error "nolocallogs should not be allowed on the virgin target"
9730 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
9731 # Start OST without MGS (local configs)
9732 do_facet ost1 "$TUNEFS --dryrun $(ostdevname 1)"
9733 start_ost || error "unable to start OST1"
9734 stop_ost || error "Unable to stop OST1"
9736 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
9737 # Do not allow reading local configs, should fail
9738 do_facet ost1 "$TUNEFS --nolocallogs $(ostdevname 1)" ||
9739 error "Can not set nolocallogs"
9740 start_ost && error "OST1 started, but should fail"
9742 # Connect to MGS successfully, reset nolocallogs flag
9743 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
9744 start_mgs || error "unable to start MGS"
9745 start_ost || error "unable to start OST1"
9747 do_facet ost1 "$TUNEFS --dryrun $(ostdevname 1)" | grep "nolocallogs" &&
9748 error "nolocallogs expected to be reset"
9750 stop_ost || error "Unable to stop OST1"
9752 run_test 128 "Force using remote logs with --nolocallogs"
9757 start_mds || error "MDS start failed"
9759 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &&
9760 error "start ost1 should fail" || true
9761 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &&
9762 error "second start ost1 should fail" || true
9763 do_facet ost1 "$TUNEFS --writeconf $(ostdevname 1)"
9764 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS ||
9765 error "start ost1 failed"
9769 run_test 129 "attempt to connect an OST with the same index should fail"
9773 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
9775 stop_mdt 2 || error "mdt2 stop failed"
9776 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)"
9777 start_mdt 2 || error "mdt2 start failed"
9778 do_facet mds2 "$LCTL dl" | grep MDT0001-osp-MDT0001 &&
9779 error "Illegal OSP device created" || true
9781 run_test 130 "re-register an MDT after writeconf"
9784 [ "$mds1_FSTYPE" == "ldiskfs" ] || skip "ldiskfs only test"
9785 do_facet mds1 $DEBUGFS -R features $(mdsdevname 1) |
9786 grep -q project || skip "skip project quota not supported"
9791 test_mkdir -c $MDSCOUNT -p $DIR/$tdir
9792 $LFS project -p 1000 $DIR/$tdir || error "set dir project id failed"
9793 createmany -o $DIR/$tdir/f 512
9794 for ((i = 0; i < 512; ++i)); do
9795 $LFS project -p $i $DIR/$tdir/f${i} ||
9796 error "set f${i} project id failed"
9799 test_mkdir -c $MDSCOUNT -p $DIR/$tdir.inherit
9800 $LFS project -p 1001 -s $DIR/$tdir.inherit
9801 createmany -o $DIR/$tdir.inherit/f 128
9802 (( $($LFS project $DIR/$tdir.inherit/f* |
9803 awk '$1 == 1001 { print }' | wc -l) == 128 )) ||
9804 error "files did not inherit projid 1001"
9808 for i in $(seq $MDSCOUNT); do
9809 mds_backup_restore mds$i ||
9810 error "Backup/restore on mds$i failed"
9815 projid=($($LFS project -d $DIR/$tdir))
9816 [ ${projid[0]} == "1000" ] ||
9817 error "projid expected 1000 not ${projid[0]}"
9818 for ((i = 0; i < 512; ++i)); do
9819 projid=($($LFS project $DIR/$tdir/f${i}))
9820 [ ${projid[0]} == "$i" ] ||
9821 error "projid expected $i not ${projid[0]}"
9824 (( $($LFS project $DIR/$tdir.inherit/f* |
9825 awk '$1 == 1001 { print }' | wc -l) == 128 )) ||
9826 error "restore did not copy projid 1001"
9828 run_test 131 "MDT backup restore with project ID"
9835 combined_mgs_mds || start_mgs || error "unable to start MGS"
9836 start_mdt 1 || error "unable to start mdt1"
9838 err_cnt=$(do_facet mds1 dmesg | grep -c "cannot take the layout locks")
9839 stop_mdt 1 || error "stop mdt1 failed"
9841 [ "$mds1_FSTYPE" == zfs ] && import_zpool mds1
9842 do_facet mds1 $TUNEFS --param mdt.hsm_control=enabled $(mdsdevname 1) ||
9843 error "tunefs failed"
9844 start_mdt 1 || error "cannot start mdt1"
9846 err_cnt2=$(do_facet mds1 dmesg | grep -c "cannot take the layout locks")
9847 [ $err_cnt -eq $err_cnt2 ] || error "Can not take the layout lock"
9848 stop_mdt 1 || error "stop mdt1 failed"
9850 run_test 132 "hsm_actions processed after failover"
9852 if ! combined_mgs_mds ; then
9858 # restore the values of MDSSIZE and OSTSIZE
9859 MDSSIZE=$STORED_MDSSIZE
9860 OSTSIZE=$STORED_OSTSIZE
9864 check_and_cleanup_lustre