8 PTLDEBUG=${PTLDEBUG:--1}
9 LUSTRE=${LUSTRE:-$(dirname $0)/..}
10 . $LUSTRE/tests/test-framework.sh
14 # tool to create lustre filesystem images
15 ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT 32newtarball"
17 # bug number for skipped test: LU-11915
18 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 110"
19 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
22 # bug number for skipped tests: LU-9795 (all below)
23 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 84 86 103"
26 if ! combined_mgs_mds; then
27 # bug number for skipped test: LU-11991 LU-11990
28 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 32a 32b 32c 32d 32e 66"
29 # bug number for skipped test: LU-9897 LU-12032
30 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 84 123F"
34 [ "$SLOW" = "no" ] && EXCEPT_SLOW="45 69 106 111 114"
38 # use small MDS + OST size to speed formatting time
39 # do not use too small MDSSIZE/OSTSIZE, which affect the default journal size
40 # STORED_MDSSIZE is used in test_18
41 STORED_MDSSIZE=$MDSSIZE
42 STORED_OSTSIZE=$OSTSIZE
44 [ "$mds1_FSTYPE" = zfs ] && MDSSIZE=400000
46 [ "$ost1_FSTYPE" = zfs ] && OSTSIZE=400000
56 # pass "-E lazy_itable_init" to mke2fs to speed up the formatting time
57 if [[ "$LDISKFS_MKFS_OPTS" != *lazy_itable_init* ]]; then
58 LDISKFS_MKFS_OPTS=$(csa_add "$LDISKFS_MKFS_OPTS" -E lazy_itable_init)
62 require_dsh_mds || exit 0
63 require_dsh_ost || exit 0
68 # The MGS must be started before the OSTs for a new fs, so start
69 # and stop to generate the startup logs.
72 wait_osc_import_state mds ost FULL
77 reformat_and_config() {
79 if ! combined_mgs_mds ; then
85 writeconf_or_reformat() {
86 # There are at most 2 OSTs for write_conf test
87 # who knows if/where $TUNEFS is installed?
88 # Better reformat if it fails...
89 writeconf_all $MDSCOUNT 2 ||
90 { echo "tunefs failed, reformatting instead" &&
91 reformat_and_config && return 0; }
100 echo "start mgs service on $(facet_active_host mgs)"
101 start mgs $(mgsdevname) $MGS_MOUNT_OPTS "$@"
107 local dev=$(mdsdevname $num)
110 echo "start mds service on `facet_active_host $facet`"
111 start $facet ${dev} $MDS_MOUNT_OPTS "$@" || return 94
114 stop_mdt_no_force() {
117 local dev=$(mdsdevname $num)
120 echo "stop mds service on `facet_active_host $facet`"
121 stop $facet || return 97
127 local dev=$(mdsdevname $num)
130 echo "stop mds service on `facet_active_host $facet`"
131 # These tests all use non-failover stop
132 stop $facet -f || return 97
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 $(version_code 2.7.0) ]]; then
1537 BLOCKSIZE=1024 ls -Rni --time-style=+%s >$tmp/img/list2
1543 BLOCKSIZE=1024 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 BLOCKSIZE=1024 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=0x1fff --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 $MOUNT2/hosts failed"
3011 # test lctl del_ost on large index
3012 do_facet mgs "$LCTL del_ost -t ${FSNAME2}-OST1fff" ||
3013 error "del_ost failed with $?"
3014 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
3016 $LFS df | grep -q OST1fff && error "del_ost did not remove OST1fff!"
3021 cleanup_nocli || error "cleanup_nocli failed with $?"
3023 run_test 33a "Mount ost with a large index number"
3025 test_33b() { # was test_34
3028 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
3029 # Drop lock cancelation reply during umount
3030 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
3031 do_facet client $LCTL set_param fail_loc=0x80000304
3032 #lctl set_param debug=-1
3033 umount_client $MOUNT
3034 cleanup || error "cleanup failed with $?"
3036 run_test 33b "Drop cancel during umount"
3040 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
3041 manual_umount_client
3043 do_facet client killall -USR1 multiop
3044 if [ $rc -eq 0 ]; then
3045 error "umount not fail!"
3048 cleanup || error "cleanup failed with rc $?"
3050 run_test 34a "umount with opened file should be fail"
3054 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3055 stop_mds || error "Unable to stop MDS"
3057 manual_umount_client --force || error "mtab after failed umount with $?"
3059 cleanup || error "cleanup failed with $?"
3061 run_test 34b "force umount with failed mds should be normal"
3065 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3066 stop_ost || error "Unable to stop OST1"
3068 manual_umount_client --force || error "mtab after failed umount with $?"
3070 cleanup || error "cleanup failed with $?"
3072 run_test 34c "force umount with failed ost should be normal"
3074 test_35a() { # bug 12459
3077 DBG_SAVE="`$LCTL get_param -n debug`"
3078 $LCTL set_param debug="ha"
3080 log "Set up a fake failnode for the MDS"
3082 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
3083 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
3085 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3086 do_facet mgs "$PERM_CMD \
3087 mdc.*${device}*.import=connection=$(h2nettype $FAKENID)" ||
3088 error "Setting mdc.*${device}*.import=connection=\
3089 $(h2nettype $FAKENID) failed."
3091 do_facet mgs "$PERM_CMD \
3092 ${device}.failover.node=$(h2nettype $FAKENID)" ||
3093 error "Setting ${device}.failover.node=\
3094 $(h2nettype $FAKENID) failed."
3096 log "Wait for RECONNECT_INTERVAL seconds (10s)"
3099 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
3102 log "Stopping the MDT: $device"
3103 stop_mdt 1 || error "MDT0 stop fail"
3105 df $MOUNT > /dev/null 2>&1 &
3107 log "Restarting the MDT: $device"
3108 start_mdt 1 || error "MDT0 start fail"
3109 log "Wait for df ($DFPID) ... "
3112 $LCTL set_param debug="$DBG_SAVE"
3114 # retrieve from the log the first server that the client tried to
3115 # contact after the connection loss
3116 $LCTL dk $TMP/lustre-log-$TESTNAME.log
3117 NEXTCONN=`awk "/${MSG}/ {start = 1;}
3118 /import_select_connection.*$device-mdc.* using connection/ {
3120 if (\\\$NF ~ /$FAKENID/)
3126 }" $TMP/lustre-log-$TESTNAME.log`
3127 [ "$NEXTCONN" != "0" ] &&
3128 error "Tried to connect to ${NEXTCONN} not last active server"
3129 cleanup || error "cleanup failed with $?"
3130 # remove nid settings
3131 writeconf_or_reformat
3133 run_test 35a "Reconnect to the last active server first"
3135 test_35b() { # bug 18674
3136 remote_mds || skip "local MDS"
3140 $LCTL set_param debug="ha"
3142 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
3145 log "Set up a fake failnode for the MDS"
3147 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
3148 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
3150 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3151 do_facet mgs "$PERM_CMD \
3152 mdc.*${device}*.import=connection=$(h2nettype $FAKENID)" ||
3153 error "Set mdc.*${device}*.import=connection=\
3154 $(h2nettype $FAKENID) failed"
3156 do_facet mgs "$PERM_CMD \
3157 ${device}.failover.node=$(h2nettype $FAKENID)" ||
3158 error "Set ${device}.failover.node=\
3159 $(h2nettype $FAKENID) failed"
3162 local at_max_saved=0
3163 # adaptive timeouts may prevent seeing the issue
3164 if at_is_enabled; then
3165 at_max_saved=$(at_max_get mds)
3166 at_max_set 0 mds client
3169 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
3171 log "Injecting EBUSY on MDS"
3172 # Setting OBD_FAIL_MDS_RESEND=0x136
3173 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" ||
3174 error "unable to set param fail_loc=0x80000136"
3176 $LCTL set_param mdc.${FSNAME}*.stats=clear
3178 log "Creating a test file and stat it"
3179 touch $MOUNT/$tdir/$tfile || error "touch $MOUNT/$tdir/$tfile failed"
3180 stat $MOUNT/$tdir/$tfile
3182 log "Stop injecting EBUSY on MDS"
3183 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" ||
3184 error "unable to set param fail_loc=0"
3185 rm -f $MOUNT/$tdir/$tfile || error "remove $MOUNT/$tdir/$tfile failed"
3188 # restore adaptive timeout
3189 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
3191 $LCTL dk $TMP/lustre-log-$TESTNAME.log
3193 CONNCNT=$($LCTL get_param mdc.${FSNAME}*.stats |
3194 awk '/mds_connect/{print $2}')
3196 # retrieve from the log if the client has ever tried to
3197 # contact the fake server after the loss of connection
3198 FAILCONN=`awk "BEGIN {ret = 0;}
3199 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
3201 if (\\\$NF ~ /$FAKENID/) {
3206 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
3208 [ "$FAILCONN" == "0" ] &&
3209 error "The client reconnection has not been triggered"
3210 [ "$FAILCONN" == "2" ] &&
3211 error "Primary server busy, client reconnect to failover failed"
3214 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
3215 # Reconnects are supposed to be rate limited to one every 5s
3216 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] &&
3217 error "Too many reconnects $CONNCNT"
3219 cleanup || error "cleanup failed with $?"
3220 # remove nid settings
3221 writeconf_or_reformat
3223 run_test 35b "Continue reconnection retries, if the active server is busy"
3226 [ $OSTCOUNT -lt 2 ] && skip_env "needs >= 2 OSTs"
3228 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] ||
3232 local FSNAME2=test1234
3233 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3235 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
3237 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
3238 is_blkdev $SINGLEMDS $MDSDEV &&
3239 skip_env "mixed loopback and real device not working"
3242 local fs2mdsdev=$(mdsdevname 1_2)
3243 local fs2ostdev=$(ostdevname 1_2)
3244 local fs3ostdev=$(ostdevname 2_2)
3245 local fs2mdsvdev=$(mdsvdevname 1_2)
3246 local fs2ostvdev=$(ostvdevname 1_2)
3247 local fs3ostvdev=$(ostvdevname 2_2)
3250 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
3251 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
3252 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
3253 # different one than the default value here.
3254 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
3255 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
3256 add fs3ost $(mkfs_opts ost2 ${fs3ostdev}) --mgsnode=$MGSNID \
3257 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
3259 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
3260 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
3261 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
3262 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
3263 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
3265 sleep 5 # until 11778 fixed
3267 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || error "dd failed"
3269 BKTOTAL=$($LCTL get_param -n obdfilter.*.kbytestotal |
3270 awk 'BEGIN{total=0}; {total+=$1}; END{print total}')
3271 BKFREE=$($LCTL get_param -n obdfilter.*.kbytesfree |
3272 awk 'BEGIN{free=0}; {free+=$1}; END{print free}')
3273 BKAVAIL=$($LCTL get_param -n obdfilter.*.kbytesavail |
3274 awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}')
3275 STRING=$(df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}')
3276 DFTOTAL=$(echo $STRING | cut -d, -f1)
3277 DFUSED=$(echo $STRING | cut -d, -f2)
3278 DFAVAIL=$(echo $STRING | cut -d, -f3)
3279 DFFREE=$(($DFTOTAL - $DFUSED))
3281 ALLOWANCE=$((64 * $OSTCOUNT))
3283 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
3284 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
3285 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
3288 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
3289 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
3290 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
3293 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
3294 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
3295 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
3300 stop fs3ost -f || error "unable to stop OST3"
3301 stop fs2ost -f || error "unable to stop OST2"
3302 stop fs2mds -f || error "unable to stop second MDS"
3303 unload_modules_conf || error "unable unload modules"
3306 run_test 36 "df report consistency on OSTs with different block size"
3309 local mntpt=$(facet_mntpt $SINGLEMDS)
3310 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
3311 local mdsdev_sym="$TMP/sym_mdt.img"
3312 local opts=$MDS_MOUNT_OPTS
3315 if [ "$mds1_FSTYPE" != ldiskfs ]; then
3316 skip "ldiskfs only test"
3319 echo "MDS : $mdsdev"
3320 echo "SYMLINK : $mdsdev_sym"
3321 do_facet $SINGLEMDS rm -f $mdsdev_sym
3323 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
3325 echo "mount symlink device - $mdsdev_sym"
3327 if ! do_facet $SINGLEMDS test -b $mdsdev; then
3328 opts=$(csa_add "$opts" -o loop)
3332 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
3333 $mdsdev_sym $mntpt 2>&1)
3336 echo mount_op=$mount_op
3338 do_facet $SINGLEMDS "$UMOUNT $mntpt && rm -f $mdsdev_sym"
3340 if $(echo $mount_op | grep -q "unable to set tunable"); then
3341 error "set tunables failed for symlink device"
3344 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
3346 run_test 37 "verify set tunables works for symlink device"
3348 test_38() { # bug 14222
3349 local mntpt=$(facet_mntpt $SINGLEMDS)
3354 local SRC="/etc /bin"
3355 local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT)
3357 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
3358 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3359 tar cf - $FILES | tar xf - -C $DIR/$tdir ||
3360 error "copying $SRC to $DIR/$tdir"
3362 umount_client $MOUNT || error "umount_client $MOUNT failed"
3363 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3364 stop_mds || error "Unable to stop MDS"
3365 log "delete lov_objid file on MDS"
3367 mount_fstype $SINGLEMDS || error "mount MDS failed (1)"
3369 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid; rm $mntpt/lov_objid"
3371 unmount_fstype $SINGLEMDS || error "umount failed (1)"
3373 # check create in mds_lov_connect
3374 start_mds || error "unable to start MDS"
3375 mount_client $MOUNT || error "mount_client $MOUNT failed"
3377 [ $V ] && log "verifying $DIR/$tdir/$f"
3378 diff -q $f $DIR/$tdir/$f || ERROR=y
3380 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3381 if [ "$ERROR" = "y" ]; then
3382 # check it's updates in sync
3383 umount_client $MOUNT
3385 mount_fstype $SIGNLEMDS
3386 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3387 unmount_fstype $SINGLEMDS
3388 error "old and new files are different after connect" || true
3390 touch $DIR/$tdir/f2 || error "f2 file create failed"
3392 # check it's updates in sync
3393 umount_client $MOUNT || error "second umount_client $MOUNT failed"
3396 mount_fstype $SINGLEMDS || error "mount MDS failed (3)"
3398 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3399 do_facet $SINGLEMDS dd if=/dev/zero of=$mntpt/lov_objid.clear count=8
3401 unmount_fstype $SINGLEMDS || error "umount failed (3)"
3403 start_mds || error "unable to start MDS"
3404 mount_client $MOUNT || error "mount_client $MOUNT failed"
3406 [ $V ] && log "verifying $DIR/$tdir/$f"
3407 diff -q $f $DIR/$tdir/$f || ERROR=y
3409 touch $DIR/$tdir/f3 || error "f3 file create failed"
3410 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3411 umount_client $MOUNT || error "third umount_client $MOUNT failed"
3413 mount_fstype $SINGLEMDS || error "mount MDS failed (4)"
3414 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3415 unmount_fstype $SINGLEMDS || error "umount failed (4)"
3417 [ "$ERROR" = "y" ] &&
3418 error "old and new files are different after sync" || true
3420 log "files compared the same"
3421 cleanup || error "cleanup failed with $?"
3423 run_test 38 "MDS recreates missing lov_objid file from OST data"
3428 cleanup || error "cleanup failed with $?"
3429 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
3430 error "memory leak detected" || true
3432 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
3434 test_40() { # bug 15759
3435 start_ost || error "Unable to start OST1"
3436 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
3437 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
3439 cleanup || error "cleanup failed with rc $?"
3441 run_test 40 "race during service thread startup"
3443 test_41a() { #bug 14134
3444 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3445 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
3446 skip "Loop devices does not work with nosvc option"
3449 combined_mgs_mds || skip "needs combined MGT and MDT device"
3451 start_mdt 1 -o nosvc -n
3452 if [ $MDSCOUNT -ge 2 ]; then
3453 for num in $(seq 2 $MDSCOUNT); do
3454 start_mdt $num || return
3457 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
3458 start_mdt 1 -o nomgs,force
3459 mount_client $MOUNT || error "mount_client $MOUNT failed"
3462 echo "blah blah" > $MOUNT/$tfile
3465 umount_client $MOUNT || error "umount_client $MOUNT failed"
3466 stop ost1 -f || error "unable to stop OST1"
3467 stop_mds || error "Unable to stop MDS"
3468 stop_mds || error "Unable to stop MDS on second try"
3470 run_test 41a "mount mds with --nosvc and --nomgs"
3473 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3474 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
3475 skip "Loop devices does not work with nosvc option"
3478 ! combined_mgs_mds && skip "needs combined mgs device"
3482 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3484 start_mdt 1 -o nosvc -n
3485 if [ $MDSCOUNT -ge 2 ]; then
3486 for num in $(seq 2 $MDSCOUNT); do
3487 start_mdt $num || return
3490 start_ost || error "Unable to start OST1"
3491 start_mdt 1 -o nomgs,force
3492 mount_client $MOUNT || error "mount_client $MOUNT failed"
3495 echo "blah blah" > $MOUNT/$tfile
3496 cat $MOUNT/$tfile || error "cat $MOUNT/$tfile failed"
3498 umount_client $MOUNT -f || error "umount_client $MOUNT failed"
3499 stop_ost || error "Unable to stop OST1"
3500 stop_mds || error "Unable to stop MDS"
3501 stop_mds || error "Unable to stop MDS on second try"
3503 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
3506 local oss_list=$(comma_list $(osts_nodes))
3508 [[ "$MDS1_VERSION" -ge $(version_code 2.6.52) ]] ||
3509 [[ "$MDS1_VERSION" -ge $(version_code 2.5.26) &&
3510 "$MDS1_VERSION" -lt $(version_code 2.5.50) ]] ||
3511 [[ "$MDS1_VERSION" -ge $(version_code 2.5.4) &&
3512 "$MDS1_VERSION" -lt $(version_code 2.5.11) ]] ||
3513 skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"
3515 # ensure mds1 ost1 have been created even if running sub-test standalone
3518 cleanup || error "cleanup failed"
3520 # using directly mount command instead of start() function to avoid
3521 # any side effect of // with others/externals tools/features
3522 # ("zpool import", ...)
3524 # MDT concurrent start
3526 LOAD_MODULES_REMOTE=true load_modules
3527 do_facet $SINGLEMDS "lsmod | grep -q libcfs" ||
3528 error "MDT concurrent start: libcfs module not loaded"
3530 local mds1dev=$(mdsdevname 1)
3531 local mds1mnt=$(facet_mntpt mds1)
3532 local mds1opts=$MDS_MOUNT_OPTS
3534 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3535 ! do_facet mds1 test -b $mds1dev; then
3536 mds1opts=$(csa_add "$mds1opts" -o loop)
3538 if [[ "$mds1_FSTYPE" == zfs ]]; then
3539 import_zpool mds1 || return ${PIPESTATUS[0]}
3542 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3543 do_facet mds1 "$LCTL set_param fail_loc=0x80000716"
3545 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts &
3548 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts
3553 do_facet mds1 "$LCTL set_param fail_loc=0x0"
3554 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3555 echo "1st MDT start succeed"
3556 echo "2nd MDT start failed with $rc2"
3557 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3558 echo "1st MDT start failed with $rc"
3559 echo "2nd MDT start succeed"
3562 error "unexpected concurrent MDT mounts result, rc=$rc rc2=$rc2"
3565 if [ $MDSCOUNT -ge 2 ]; then
3566 for num in $(seq 2 $MDSCOUNT); do
3567 start_mdt $num || return
3571 # OST concurrent start
3573 do_rpc_nodes $oss_list "lsmod | grep -q libcfs" ||
3574 error "OST concurrent start: libcfs module not loaded"
3576 local ost1dev=$(ostdevname 1)
3577 local ost1mnt=$(facet_mntpt ost1)
3578 local ost1opts=$OST_MOUNT_OPTS
3580 if [ "$ost1_FSTYPE" == ldiskfs ] &&
3581 ! do_facet ost1 test -b $ost1dev; then
3582 ost1opts=$(csa_add "$ost1opts" -o loop)
3584 if [[ "$ost1_FSTYPE" == zfs ]]; then
3585 import_zpool ost1 || return ${PIPESTATUS[0]}
3588 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3589 do_facet ost1 "$LCTL set_param fail_loc=0x80000716"
3591 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts &
3594 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts
3598 do_facet ost1 "$LCTL set_param fail_loc=0x0"
3599 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3600 echo "1st OST start succeed"
3601 echo "2nd OST start failed with $rc2"
3602 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3603 echo "1st OST start failed with $rc"
3604 echo "2nd OST start succeed"
3608 error "unexpected concurrent OST mounts result, rc=$rc rc2=$rc2"
3614 # verify everything ok
3619 error "MDT(s) start failed"
3627 error "OST(s) start failed"
3635 error "client start failed"
3642 error "client mount failed"
3646 run_test 41c "concurrent mounts of MDT/OST should all fail but one"
3648 test_42() { #bug 14693
3652 check_mount || error "client was not mounted"
3654 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3655 PARAM="llite.$FSNAME-*.some_wrong_param"
3657 PARAM="$FSNAME.llite.some_wrong_param"
3660 do_facet mgs $PERM_CMD $PARAM=10
3661 umount_client $MOUNT ||
3662 error "unmounting client failed with invalid llite param"
3663 mount_client $MOUNT ||
3664 error "mounting client failed with invalid llite param"
3666 do_facet mgs $PERM_CMD $PARAM=20
3667 cleanup || error "stopping $FSNAME failed with invalid sys param"
3669 check_mount || error "client was not mounted with invalid sys param"
3670 cleanup || error "stopping $FSNAME failed with invalid sys param"
3672 run_test 42 "allow client/server mount/unmount with invalid config param"
3675 [[ "$MGS_VERSION" -ge $(version_code 2.5.58) ]] ||
3676 skip "Need MDS version at least 2.5.58"
3677 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
3680 USER1=$(getent passwd | grep :$ID1:$ID1: | cut -d: -f1)
3681 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1"
3684 chmod ugo+x $DIR || error "chmod 0 failed"
3685 set_persistent_param_and_check mds1 \
3686 "mdt.$FSNAME-MDT0000.root_squash" \
3687 "$FSNAME.mdt.root_squash" \
3689 wait_update $HOSTNAME \
3690 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3692 error "check llite root_squash failed!"
3693 set_persistent_param_and_check mds1 \
3694 "mdt.$FSNAME-MDT0000.nosquash_nids" \
3695 "$FSNAME.mdt.nosquash_nids" \
3697 wait_update $HOSTNAME \
3698 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3700 error "check llite nosquash_nids failed!"
3703 # create set of test files
3705 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
3706 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
3707 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
3709 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
3710 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
3712 mkdir_on_mdt0 $DIR/$tdir-rootdir || error "mkdir failed"
3713 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
3714 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
3716 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
3717 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
3718 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
3721 # check root_squash:
3722 # set root squash UID:GID to RUNAS_ID
3723 # root should be able to access only files owned by RUNAS_ID
3725 set_persistent_param_and_check mds1 \
3726 "mdt.$FSNAME-MDT0000.root_squash" \
3727 "$FSNAME.mdt.root_squash" \
3728 "$RUNAS_ID:$RUNAS_ID"
3729 wait_update $HOSTNAME \
3730 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3731 "$RUNAS_ID:$RUNAS_ID" ||
3732 error "check llite root_squash failed!"
3734 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
3735 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3736 error "$ST: root read permission is denied"
3737 echo "$ST: root read permission is granted - ok"
3740 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3741 error "$ST: root write permission is denied"
3742 echo "$ST: root write permission is granted - ok"
3744 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3745 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3746 error "$ST: root read permission is granted"
3747 echo "$ST: root read permission is denied - ok"
3750 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3751 error "$ST: root write permission is granted"
3752 echo "$ST: root write permission is denied - ok"
3754 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3755 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null &&
3756 error "$ST: root unlink permission is granted"
3757 echo "$ST: root unlink permission is denied - ok"
3759 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null &&
3760 error "$ST: root create permission is granted"
3761 echo "$ST: root create permission is denied - ok"
3764 # check root_squash is enforced independently
3765 # of client cache content
3767 # access file by USER1, keep access open
3768 # root should be denied access to user file
3770 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
3774 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
3775 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3776 { kill $pid; error "$ST: root read permission is granted"; }
3777 echo "$ST: root read permission is denied - ok"
3780 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3781 { kill $pid; error "$ST: root write permission is granted"; }
3782 echo "$ST: root write permission is denied - ok"
3788 # check nosquash_nids:
3789 # put client's NID into nosquash_nids list,
3790 # root should be able to access root file after that
3792 local NIDLIST=$($LCTL list_nids all | tr '\n' ' ')
3793 NIDLIST="2@gni $NIDLIST 192.168.0.[2,10]@tcp"
3794 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
3795 set_persistent_param_and_check mds1 \
3796 "mdt.$FSNAME-MDT0000.nosquash_nids" \
3797 "$FSNAME-MDTall.mdt.nosquash_nids" \
3799 wait_update $HOSTNAME \
3800 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3802 error "check llite nosquash_nids failed!"
3804 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3805 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3806 error "$ST: root read permission is denied"
3807 echo "$ST: root read permission is granted - ok"
3810 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3811 error "$ST: root write permission is denied"
3812 echo "$ST: root write permission is granted - ok"
3814 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3815 rm $DIR/$tdir-rootdir/tfile-1 ||
3816 error "$ST: root unlink permission is denied"
3817 echo "$ST: root unlink permission is granted - ok"
3818 touch $DIR/$tdir-rootdir/tfile-2 ||
3819 error "$ST: root create permission is denied"
3820 echo "$ST: root create permission is granted - ok"
3821 cleanup || error "cleanup failed with $?"
3823 run_test 43a "check root_squash and nosquash_nids"
3825 test_43b() { # LU-5690
3826 [[ "$MGS_VERSION" -ge $(version_code 2.7.62) ]] ||
3827 skip "Need MGS version 2.7.62+"
3829 if [[ -z "$fs2mds_DEV" ]]; then
3830 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
3831 skip_env "mixed loopback and real device not working"
3834 local fs2mdsdev=$(mdsdevname 1_2)
3835 local fs2mdsvdev=$(mdsvdevname 1_2)
3837 # temporarily use fs2mds as fs2mgs
3839 local fs2mgsdev=$fs2mdsdev
3840 local fs2mgsvdev=$fs2mdsvdev
3842 local fsname=test1234
3844 load_module llite/lustre
3845 local client_ip=$(host_nids_address $HOSTNAME $NETTYPE)
3846 local host=${client_ip//*./}
3847 local net=${client_ip/%$host/}
3848 local nosquash_nids=$(h2nettype $net[$host,$host,$host])
3850 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
3851 --param mdt.root_squash=$RUNAS_ID:$RUNAS_ID \
3852 --param mdt.nosquash_nids=$nosquash_nids \
3853 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
3854 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS || error "start fs2mgs failed"
3855 stop $fs2mgs -f || error "stop fs2mgs failed"
3856 cleanup || error "cleanup failed with $?"
3858 run_test 43b "parse nosquash_nids with commas in expr_list"
3862 check_mount || error "check_mount"
3863 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
3865 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
3866 for VAL in $UUIDS; do
3867 NID=$(echo $VAL | cut -d= -f1)
3868 CLUUID=$(echo $VAL | cut -d= -f2)
3869 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
3871 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
3872 cleanup || error "cleanup failed with $?"
3874 run_test 44 "mounted client proc entry exists"
3878 check_mount || error "check_mount"
3879 stop_mds || error "Unable to stop MDS"
3883 #define OBD_FAIL_PTLRPC_LONG_REPL_UNLINK 0x50f
3884 do_facet client "$LCTL set_param fail_loc=0x8000050f"
3887 manual_umount_client --force || error "manual_umount_client failed"
3888 do_facet client "$LCTL set_param fail_loc=0x0"
3889 start_mds || error "unable to start MDS"
3890 mount_client $MOUNT || error "mount_client $MOUNT failed"
3891 cleanup || error "cleanup failed with $?"
3893 run_test 45 "long unlink handling in ptlrpcd"
3900 umount_client $MOUNT2 || rc=$?
3901 umount_client $MOUNT || rc=$?
3902 while [ $count -gt 0 ]; do
3903 stop ost${count} -f || rc=$?
3907 cleanup_nocli || rc=$?
3908 #writeconf to remove all ost2 traces for subsequent tests
3909 writeconf_or_reformat
3914 echo "Testing with $OSTCOUNT OSTs"
3916 start_mds || error "unable to start MDS"
3917 #first client should see only one ost
3918 start_ost || error "Unable to start OST1"
3919 wait_osc_import_state mds ost FULL
3921 mount_client $MOUNT || error "mount_client $MOUNT failed"
3922 trap "cleanup_46a $OSTCOUNT" EXIT ERR
3925 for (( i=2; i<=$OSTCOUNT; i++ )); do
3926 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
3927 error "start_ost$i $(ostdevname $i) failed"
3930 # wait until osts in sync
3931 for (( i=2; i<=$OSTCOUNT; i++ )); do
3932 wait_osc_import_state mds ost$i FULL
3933 wait_osc_import_ready client ost$i
3936 #second client see all ost's
3938 mount_client $MOUNT2 || error "mount_client failed"
3939 $LFS setstripe -c -1 $MOUNT2 ||
3940 error "$LFS setstripe -c -1 $MOUNT2 failed"
3941 $LFS getstripe $MOUNT2 || error "$LFS getstripe $MOUNT2 failed"
3943 echo "ok" > $MOUNT2/widestripe
3944 $LFS getstripe $MOUNT2/widestripe ||
3945 error "$LFS getstripe $MOUNT2/widestripe failed"
3946 # fill acl buffer for avoid expand lsm to them
3947 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd |
3949 setfacl -m $acl $MOUNT2/widestripe
3953 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3955 cleanup_46a $OSTCOUNT || error "cleanup_46a failed"
3957 run_test 46a "handle ost additional - wide striped file"
3962 check_mount || error "check_mount failed"
3963 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
3967 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3968 if echo $ns | grep "MDT[[:digit:]]*"; then
3971 lrs=$(echo $ns | sed 's/.*lru_size=//')
3972 lru_size[count]=$lrs
3977 facet_failover $SINGLEMDS
3978 client_up || error "client_up failed"
3981 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3982 if echo $ns | grep "MDT[[:digit:]]*"; then
3985 lrs=$(echo $ns | sed 's/.*lru_size=//')
3986 if ! test "$lrs" -eq "${lru_size[count]}"; then
3987 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
3988 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
3993 cleanup || error "cleanup failed with $?"
3995 run_test 47 "server restart does not make client loss lru_resize settings"
4000 # reformat after this test is needed - if the test fails,
4001 # we will have unkillable file at FS
4005 test_48() { # bz-17636 LU-7473
4009 check_mount || error "check_mount failed"
4011 $LFS setstripe -c -1 $MOUNT ||
4012 error "$LFS setstripe -c -1 $MOUNT failed"
4013 $LFS getstripe $MOUNT || error "$LFS getstripe $MOUNT failed"
4015 echo "ok" > $MOUNT/widestripe
4016 $LFS getstripe $MOUNT/widestripe ||
4017 error "$LFS getstripe $MOUNT/widestripe failed"
4019 # In the future, we may introduce more EAs, such as selinux, enlarged
4020 # LOV EA, and so on. These EA will use some EA space that is shared by
4021 # ACL entries. So here we only check some reasonable ACL entries count,
4022 # instead of the max number that is calculated from the max_ea_size.
4023 if [ "$MDS1_VERSION" -lt $(version_code 2.8.57) ]; then
4024 count=28 # hard coded of RPC protocol
4025 elif large_xattr_enabled; then
4026 count=4500 # max_num 8187 max_ea_size = 65452
4027 # not create too many (4500) to save test time
4029 count=450 # max_num 497 max_ea_size = 4012
4032 echo "It is expected to hold at least $count ACL entries"
4033 trap cleanup_48 EXIT ERR
4034 for ((i = 0; i < $count; i++)) do
4035 setfacl -m u:$((i + 100)):rw $MOUNT/widestripe ||
4036 error "Fail to setfacl for $MOUNT/widestripe at $i"
4039 cancel_lru_locks mdc
4040 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
4041 local r_count=$(getfacl $MOUNT/widestripe | grep "user:" | wc -l)
4043 count=$((count + 1)) # for the entry "user::rw-"
4045 [ $count -eq $r_count ] ||
4046 error "Expected ACL entries $count, but got $r_count"
4050 run_test 48 "too many acls on file"
4052 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
4053 test_49a() { # bug 17710
4054 local timeout_orig=$TIMEOUT
4055 local ldlm_timeout_orig=$LDLM_TIMEOUT
4056 local LOCAL_TIMEOUT=20
4058 LDLM_TIMEOUT=$LOCAL_TIMEOUT
4059 TIMEOUT=$LOCAL_TIMEOUT
4063 check_mount || error "client mount failed"
4065 echo "check ldlm_timout..."
4066 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
4067 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
4068 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
4070 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
4071 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
4074 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
4075 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
4078 umount_client $MOUNT || error "umount_client $MOUNT failed"
4079 stop_ost || error "problem stopping OSS"
4080 stop_mds || error "problem stopping MDS"
4082 LDLM_TIMEOUT=$ldlm_timeout_orig
4083 TIMEOUT=$timeout_orig
4085 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
4087 test_49b() { # bug 17710
4088 local timeout_orig=$TIMEOUT
4089 local ldlm_timeout_orig=$LDLM_TIMEOUT
4090 local LOCAL_TIMEOUT=20
4092 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
4093 TIMEOUT=$LOCAL_TIMEOUT
4097 check_mount || error "client mount failed"
4099 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
4100 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
4101 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
4103 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
4104 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
4107 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
4108 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
4111 cleanup || error "cleanup failed"
4113 LDLM_TIMEOUT=$ldlm_timeout_orig
4114 TIMEOUT=$timeout_orig
4116 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
4119 # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1
4121 # Test both statfs and lfs df and fail if either one fails
4122 multiop_bg_pause $1 f_
4125 killall -USR1 multiop
4126 [ $RC -ne 0 ] && log "lazystatfs multiop failed"
4127 wait $PID || { RC=$?; log "multiop return error "; }
4129 # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1
4134 if kill -s 0 $PID; then
4137 log "lazystatfs lfs df failed to complete in 5s"
4145 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4146 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4148 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
4150 cleanup || error "cleanup failed with rc $?"
4152 run_test 50a "lazystatfs all servers available"
4156 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4157 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4159 # Wait for client to detect down OST
4160 stop_ost || error "Unable to stop OST1"
4161 wait_osc_import_state client ost DISCONN
4163 log "OSCs should all be DISCONN"
4165 lazystatfs $MOUNT || error "lazystatfs should not return EIO"
4167 umount_client $MOUNT || error "Unable to unmount client"
4168 stop_mds || error "Unable to stop MDS"
4170 run_test 50b "lazystatfs all servers down"
4173 start_mds || error "Unable to start MDS"
4174 start_ost || error "Unable to start OST1"
4175 start_ost2 || error "Unable to start OST2"
4176 mount_client $MOUNT || error "Unable to mount client"
4177 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4178 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4180 # Wait for client to detect down OST
4181 stop_ost || error "Unable to stop OST1"
4182 wait_osc_import_state mds ost DISCONN
4183 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
4185 umount_client $MOUNT || error "Unable to unmount client"
4186 stop_ost2 || error "Unable to stop OST2"
4187 stop_mds || error "Unable to stop MDS"
4188 #writeconf to remove all ost2 traces for subsequent tests
4189 writeconf_or_reformat
4191 run_test 50c "lazystatfs one server down"
4194 start_mds || error "Unable to start MDS"
4195 start_ost || error "Unable to start OST1"
4196 start_ost2 || error "Unable to start OST2"
4197 mount_client $MOUNT || error "Unable to mount client"
4198 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4199 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4201 # Issue the statfs during the window where the client still
4202 # belives the OST to be available but it is in fact down.
4203 # No failure just a statfs which hangs for a timeout interval.
4204 stop_ost || error "Unable to stop OST1"
4205 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
4207 umount_client $MOUNT || error "Unable to unmount client"
4208 stop_ost2 || error "Unable to stop OST2"
4209 stop_mds || error "Unable to stop MDS"
4210 #writeconf to remove all ost2 traces for subsequent tests
4211 writeconf_or_reformat
4213 run_test 50d "lazystatfs client/server conn race"
4220 start_mds || error "Unable to start MDS"
4221 #first client should see only one ost
4222 start_ost || error "Unable to start OST1"
4223 wait_osc_import_state mds ost FULL
4225 # Wait for client to detect down OST
4226 stop_ost || error "Unable to stop OST1"
4227 wait_osc_import_state mds ost DISCONN
4229 mount_client $MOUNT || error "Unable to mount client"
4230 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
4232 multiop_bg_pause $MOUNT _f
4236 if [ $RC1 -ne 0 ]; then
4237 log "multiop failed $RC1"
4240 sleep $(( $TIMEOUT+1 ))
4242 [ $? -ne 0 ] && error "process isn't sleep"
4243 start_ost || error "Unable to start OST1"
4244 wait $pid || error "statfs failed"
4247 umount_client $MOUNT || error "Unable to unmount client"
4248 stop_ost || error "Unable to stop OST1"
4249 stop_mds || error "Unable to stop MDS"
4251 run_test 50e "normal statfs all servers down"
4256 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
4258 start_mds || error "Unable to start mds"
4259 #first client should see only one ost
4260 start_ost || error "Unable to start OST1"
4261 wait_osc_import_state mds ost FULL
4263 start_ost2 || error "Unable to start OST2"
4264 wait_osc_import_state mds ost2 FULL
4266 # Wait for client to detect down OST
4267 stop_ost2 || error "Unable to stop OST2"
4269 wait_osc_import_state mds ost2 DISCONN
4270 mount_client $MOUNT || error "Unable to mount client"
4271 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
4273 multiop_bg_pause $MOUNT _f
4277 if [ $RC1 -ne 0 ]; then
4278 log "lazystatfs multiop failed $RC1"
4281 sleep $(( $TIMEOUT+1 ))
4283 [ $? -ne 0 ] && error "process isn't sleep"
4284 start_ost2 || error "Unable to start OST2"
4285 wait $pid || error "statfs failed"
4286 stop_ost2 || error "Unable to stop OST2"
4289 umount_client $MOUNT -f || error "Unable to unmount client"
4290 stop_ost || error "Unable to stop OST1"
4291 stop_mds || error "Unable to stop MDS"
4292 #writeconf to remove all ost2 traces for subsequent tests
4293 writeconf_or_reformat
4295 run_test 50f "normal statfs one server in down"
4298 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs"
4300 start_ost2 || error "Unable to start OST2"
4301 wait_osc_import_state mds ost2 FULL
4302 wait_osc_import_ready client ost2
4304 if [[ $PERM_CMD == *"set_param -P"* ]]; then
4305 local PARAM="osc.${FSNAME}-OST0001*.active"
4307 local PARAM="${FSNAME}-OST0001.osc.active"
4310 $LFS setstripe -c -1 $DIR/$tfile || error "$LFS setstripe failed"
4311 do_facet mgs $PERM_CMD $PARAM=0 || error "Unable to deactivate OST"
4313 umount_client $MOUNT || error "Unable to unmount client"
4314 mount_client $MOUNT || error "Unable to mount client"
4315 # This df should not cause a panic
4318 do_facet mgs $PERM_CMD $PARAM=1 || error "Unable to activate OST"
4319 rm -f $DIR/$tfile || error "unable to remove file $DIR/$tfile"
4320 umount_client $MOUNT || error "Unable to unmount client"
4321 stop_ost2 || error "Unable to stop OST2"
4322 stop_ost || error "Unable to stop OST1"
4323 stop_mds || error "Unable to stop MDS"
4324 #writeconf to remove all ost2 traces for subsequent tests
4325 writeconf_or_reformat
4327 run_test 50g "deactivated OST should not cause panic"
4331 # prepare MDT/OST, make OSC inactive for OST1
4332 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs"
4334 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
4335 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
4336 error "tunefs OST1 failed"
4337 start_mds || error "Unable to start MDT"
4338 start_ost || error "Unable to start OST1"
4339 start_ost2 || error "Unable to start OST2"
4340 mount_client $MOUNT || error "client start failed"
4342 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4344 # activatate OSC for OST1
4345 set_persistent_param_and_check client \
4346 "osc.${FSNAME}-OST0000-osc-[!M]*.active" \
4347 "${FSNAME}-OST0000.osc.active" 1
4349 mkdir $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
4350 $LFS setstripe -c -1 -i 0 $DIR/$tdir/2 ||
4351 error "$LFS setstripe $DIR/$tdir/2 failed"
4352 sleep 1 && echo "create a file after OST1 is activated"
4353 # doing some io, shouldn't crash
4354 dd if=/dev/zero of=$DIR/$tdir/2/$tfile-io bs=1M count=10
4356 # check OSC import is working
4357 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
4358 error "some OSC imports are still not connected"
4362 umount_client $MOUNT || error "Unable to umount client"
4363 stop_ost2 || error "Unable to stop OST2"
4364 cleanup_nocli || error "cleanup_nocli failed with $?"
4366 run_test 50h "LU-642: activate deactivated OST"
4369 # prepare MDT/OST, make OSC inactive for OST1
4370 [ "$MDSCOUNT" -lt "2" ] && skip "needs >= 2 MDTs"
4373 [ $(facet_fstype mds2) == zfs ] && import_zpool mds2
4374 do_facet mds2 "$TUNEFS --param mdc.active=0 $(mdsdevname 2)" ||
4375 error "tunefs MDT2 failed"
4376 start_mds || error "Unable to start MDT"
4377 start_ost || error "Unable to start OST1"
4378 start_ost2 || error "Unable to start OST2"
4379 mount_client $MOUNT || error "client start failed"
4381 mkdir_on_mdt0 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4383 if [[ $PERM_CMD == *"set_param -P"* ]]; then
4384 $PERM_CMD mdc.${FSNAME}-MDT0001-mdc-*.active=0 &&
4385 error "deactive MDC0 succeeds"
4387 $PERM_CMD ${FSNAME}-MDT0000.mdc.active=0 &&
4388 error "deactive MDC0 succeeds"
4391 # activate MDC for MDT2
4392 set_persistent_param_and_check client \
4393 "mdc.${FSNAME}-MDT0001-mdc-*.active" \
4394 "${FSNAME}-MDT0001.mdc.active" 1
4396 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
4397 if [ "$MDS1_VERSION" -ge $(version_code 2.7.60) ]
4399 wait_dne_interconnect
4401 $LFS mkdir -i1 $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
4403 createmany -o $DIR/$tdir/2/$tfile-%d 1 || error "create files failed"
4405 rm -rf $DIR/$tdir/2 || error "unlink dir failed"
4407 # deactivate MDC for MDT2
4408 set_persistent_param_and_check client \
4409 "mdc.${FSNAME}-MDT0001-mdc-*.active" \
4410 "${FSNAME}-MDT0001.mdc.active" 0
4412 wait_osp_active mds ${FSNAME}-MDT0001 1 0
4414 $LFS mkdir -i1 $DIR/$tdir/2 &&
4415 error "mkdir $DIR/$tdir/2 succeeds after deactive MDT"
4417 $LFS mkdir -i0 -c$MDSCOUNT $DIR/$tdir/striped_dir ||
4418 error "mkdir $DIR/$tdir/striped_dir fails after deactive MDT2"
4420 local stripe_count=$($LFS getdirstripe -c $DIR/$tdir/striped_dir)
4421 [ $stripe_count -eq $((MDSCOUNT - 1)) ] ||
4422 error "wrong $stripe_count != $((MDSCOUNT -1)) for striped_dir"
4425 umount_client $MOUNT || error "Unable to umount client"
4430 run_test 50i "activate deactivated MDT"
4433 local LOCAL_TIMEOUT=20
4437 check_mount || error "check_mount failed"
4439 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
4440 $LFS setstripe -c -1 $MOUNT/$tdir ||
4441 error "$LFS setstripe -c -1 $MOUNT/$tdir failed"
4442 #define OBD_FAIL_MDS_REINT_DELAY 0x142
4443 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x142"
4444 touch $MOUNT/$tdir/$tfile &
4447 start_ost2 || error "Unable to start OST1"
4449 stop_ost2 || error "Unable to stop OST1"
4450 umount_client $MOUNT -f || error “unmount $MOUNT failed”
4451 cleanup_nocli || error “stop server failed”
4452 #writeconf to remove all ost2 traces for subsequent tests
4453 writeconf_or_reformat
4455 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
4464 do_node $node mkdir -p $dest
4465 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
4467 do_node $node 'tar cf - '$*' | tar xf - -C '$dest';
4468 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
4469 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
4471 do_node $node 'getfattr -d -m "[a-z]*\\." '$*' > '$xattrs
4472 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
4482 local backup2=${TMP}/backup2
4484 do_node $node mkdir -p $backup2
4485 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
4487 do_node $node 'tar cf - '$*' | tar xf - -C '$backup2';
4488 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
4489 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
4491 do_node $node "diff -rq $backup $backup2"
4492 [ $? -eq 0 ] || { error "contents differ"; return 3; }
4494 local xattrs2=${TMP}/xattrs2
4495 do_node $node 'getfattr -d -m "[a-z]*\\." '$*' > '$xattrs2
4496 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
4498 do_node $node "diff $xattrs $xattrs2"
4499 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
4501 do_node $node "rm -rf $backup2 $xattrs2"
4502 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
4506 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4507 skip "ldiskfs only test"
4510 start_mds || error "Unable to start MDS"
4511 start_ost || error "Unable to start OST1"
4512 mount_client $MOUNT || error "Unable to mount client"
4515 local ost1mnt=$(facet_mntpt ost1)
4516 local ost1node=$(facet_active_host ost1)
4517 local ost1tmp=$TMP/conf52
4520 mkdir $DIR/$tdir || error "Unable to create $DIR/$tdir"
4521 touch $TMP/modified_first || error "Unable to create temporary file"
4522 local mtime=$(stat -c %Y $TMP/modified_first)
4523 do_node $ost1node "mkdir -p $ost1tmp &&
4524 touch -m -d @$mtime $ost1tmp/modified_first" ||
4525 error "Unable to create temporary file"
4528 $LFS setstripe -c -1 -S 1M $DIR/$tdir || error "$LFS setstripe failed"
4530 for (( i=0; i < nrfiles; i++ )); do
4531 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c ||
4532 error "multiop failed"
4537 # sync all the data and make sure no pending data on the client,
4538 # thus the SOM xattr would not be changed any more.
4539 cancel_lru_locks osc
4542 echo backup files to $TMP/$tdir
4543 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
4544 copy_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
4545 error "Unable to copy files"
4547 umount_client $MOUNT || error "Unable to umount client"
4548 stop_ost || error "Unable to stop ost1"
4550 echo mount ost1 as ldiskfs
4551 do_node $ost1node mkdir -p $ost1mnt || error "Unable to create $ost1mnt"
4552 if ! do_node $ost1node test -b $ost1_dev; then
4555 do_node $ost1node mount -t "$ost1_FSTYPE" $loop $ost1_dev \
4557 error "Unable to mount ost1 as ldiskfs"
4560 echo backup objects to $ost1tmp/objects
4561 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
4562 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
4563 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
4565 error "Unable to copy objects"
4567 # move objects to lost+found
4568 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
4569 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
4571 do_node $ost1node "umount $ost1mnt" ||
4572 error "Unable to umount ost1 as ldiskfs"
4574 start_ost || error "Unable to start OST1"
4575 mount_client $MOUNT || error "Unable to mount client"
4577 local REPAIRED=$(do_node $ost1node "$LCTL get_param \
4578 -n osd-ldiskfs.$FSNAME-OST0000.oi_scrub" |
4579 awk '/^lf_repa[ri]*ed/ { print $2 }')
4580 [ $REPAIRED -gt 0 ] ||
4581 error "Some entry under /lost+found should be repaired"
4584 diff_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
4585 error "Unable to diff files"
4587 rm -rf $TMP/$tdir $TMP/file_xattrs ||
4588 error "Unable to delete temporary files"
4589 do_node $ost1node "rm -rf $ost1tmp" ||
4590 error "Unable to delete temporary files"
4591 cleanup || error "cleanup failed with $?"
4593 run_test 52 "check recovering objects from lost+found"
4595 # Checks threads_min/max/started for some service
4597 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
4598 # parameter pattern prefix like 'ost.*.ost'.
4611 local msg="Insane $modname thread counts"
4612 local ncpts=$(check_cpt_number $facet)
4616 check_mount || return 41
4618 # We need to expand $parampat, but it may match multiple parameters, so
4619 # we'll pick the first one
4620 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
4621 error "Couldn't expand ${parampat}.threads_min parameter name"
4625 # Remove the .threads_min part
4626 paramp=${paramp%.threads_min}
4628 # Check for sanity in defaults
4629 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4631 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4633 tstarted=$(do_facet $facet "$LCTL get_param \
4634 -n ${paramp}.threads_started" || echo 0)
4635 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' ||
4637 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4639 nthrs=$(expr $tmax - $tmin)
4640 if [ $nthrs -lt $ncpts ]; then
4646 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
4647 skip_env "module parameter forced $facet thread count"
4649 # Check that we can change min/max
4650 do_facet $facet "$LCTL set_param \
4651 ${paramp}.threads_min=$((tmin + nthrs))"
4652 do_facet $facet "$LCTL set_param \
4653 ${paramp}.threads_max=$((tmax - nthrs))"
4654 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4656 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4658 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) &&
4659 $tmax2 == ($tmax - $nthrs)))' || return $?
4661 # Check that we can set min/max to the same value
4662 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4664 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$tmin"
4665 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4667 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4669 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
4671 # Check that we can't set max < min
4672 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$((tmin - 1))"
4673 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4675 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4677 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
4679 # We need to ensure that we get the module options desired; to do this
4680 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
4681 LOAD_MODULES_REMOTE=true
4684 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
4686 setmodopts -a $modname "$newvalue" oldvalue
4689 check_mount || return 41
4691 # Restore previous setting of MODOPTS_*
4692 setmodopts $modname "$oldvalue"
4694 # Check that $opts took
4695 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4697 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4699 tstarted=$(do_facet $facet \
4700 "$LCTL get_param -n ${paramp}.threads_started" || echo 0)
4701 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4710 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
4711 cleanup || error "cleanup failed with rc $?"
4713 run_test 53a "check OSS thread count params"
4717 local mds=$(do_facet $SINGLEMDS "$LCTL get_param \
4718 -N mds.*.*.threads_max 2>/dev/null")
4719 if [ -z "$mds" ]; then
4720 #running this on an old MDT
4721 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
4723 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
4725 cleanup || error "cleanup failed with $?"
4727 run_test 53b "check MDS thread count params"
4730 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4731 skip "ldiskfs only test"
4734 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p ||
4735 error "llverdev failed with rc=$?"
4738 run_test 54a "test llverdev and partial verify of device"
4741 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4742 skip "ldiskfs only test"
4746 run_llverfs $MOUNT -p || error "llverfs failed with rc=$?"
4747 cleanup || error "cleanup failed with rc=$?"
4749 run_test 54b "test llverfs and partial verify of filesystem"
4753 local max_ost_index=$1
4755 echo -n $(((max_ost_index + 1) * 8))
4759 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4760 skip "ldiskfs only test"
4763 local mdsdev=$(mdsdevname 1)
4764 local mdsvdev=$(mdsvdevname 1)
4768 if ! combined_mgs_mds; then
4769 stop_mgs || error "stopping MGS service failed"
4770 format_mgs || error "formatting MGT failed"
4772 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
4774 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
4775 --reformat $(ostdevname 1) $(ostvdevname 1)
4781 echo checking size of lov_objid for ost index $i
4782 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" |
4783 grep ^User | awk -F 'Size: ' '{print $2}')
4784 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
4785 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
4787 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
4794 run_test 55 "check lov_objid size"
4797 local mds_journal_size_orig=$MDSJOURNALSIZE
4803 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
4804 $(ostdevname 1) $(ostvdevname 1)
4805 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
4806 $(ostdevname 2) $(ostvdevname 2)
4809 start_ost || error "Unable to start first ost (idx 10000)"
4810 start_ost2 || error "Unable to start second ost (idx 1000)"
4811 mount_client $MOUNT || error "Unable to mount client"
4815 # test instantiating PFL components with sparse index LU-15513
4816 mkdir -p $MOUNT/$tdir
4817 $LFS setstripe -E 4M -c 1 -E 1G -c 4 -S4M -E eof -c -1 $MOUNT/$tdir
4818 dd if=/dev/zero of=$MOUNT/$tdir/$tfile bs=4K count=1 seek=10k ||
4819 error "dd to second component failed"
4821 if [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
4822 [[ "$MDS1_VERSION" -ge $(version_code 2.5.4) &&
4823 "$MDS1_VERSION" -lt $(version_code 2.5.11) ]]; then
4824 wait_osc_import_state mds ost1 FULL
4825 wait_osc_import_state mds ost2 FULL
4826 $LFS setstripe --stripe-count=-1 $DIR/$tfile ||
4827 error "Unable to setstripe $DIR/$tfile"
4828 n=$($LFS getstripe --stripe-count $DIR/$tfile)
4829 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
4834 MDSJOURNALSIZE=$mds_journal_size_orig
4837 run_test 56a "check big OST indexes and out-of-index-order start"
4842 umount_client $MOUNT -f || error "unmount client failed"
4851 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs"
4853 trap cleanup_56b EXIT RETURN ERR
4856 if ! combined_mgs_mds ; then
4861 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) --index=0 --reformat \
4862 $(mdsdevname 1) $(mdsvdevname 1)
4863 add mds2 $(mkfs_opts mds2 $(mdsdevname 2)) --index=1 --reformat \
4864 $(mdsdevname 2) $(mdsvdevname 2)
4865 add mds3 $(mkfs_opts mds3 $(mdsdevname 3)) --index=1000 --reformat \
4866 $(mdsdevname 3) $(mdsvdevname 3)
4870 start_mdt 1 || error "MDT 1 (idx 0) start failed"
4871 start_mdt 2 || error "MDT 2 (idx 1) start failed"
4872 start_mdt 3 || error "MDT 3 (idx 1000) start failed"
4873 start_ost || error "Unable to start first ost"
4874 start_ost2 || error "Unable to start second ost"
4876 do_nodes $(comma_list $(mdts_nodes)) \
4877 "$LCTL set_param mdt.*.enable_remote_dir=1 \
4878 mdt.*.enable_remote_dir_gid=-1"
4880 mount_client $MOUNT || error "Unable to mount client"
4882 $LFS mkdir -c3 $MOUNT/$tdir || error "failed to make testdir"
4884 echo "This is test file 1!" > $MOUNT/$tdir/$tfile.1 ||
4885 error "failed to make test file 1"
4886 echo "This is test file 2!" > $MOUNT/$tdir/$tfile.2 ||
4887 error "failed to make test file 2"
4888 echo "This is test file 1000!" > $MOUNT/$tdir/$tfile.1000 ||
4889 error "failed to make test file 1000"
4891 rm -rf $MOUNT/$tdir || error "failed to remove testdir"
4893 $LFS mkdir -i1000 $MOUNT/$tdir.1000 ||
4894 error "create remote dir at idx 1000 failed"
4897 echo "=== START lfs df OUTPUT ==="
4899 echo "==== END lfs df OUTPUT ===="
4901 mdtcnt=$(echo -e "$output" | grep $FSNAME-MDT | wc -l)
4902 ostcnt=$(echo -e "$output" | grep $FSNAME-OST | wc -l)
4904 echo "lfs df returned mdt count $mdtcnt and ost count $ostcnt"
4905 [ $mdtcnt -eq 3 ] || error "lfs df returned wrong mdt count"
4906 [ $ostcnt -eq 2 ] || error "lfs df returned wrong ost count"
4908 echo "This is test file 1!" > $MOUNT/$tdir.1000/$tfile.1 ||
4909 error "failed to make test file 1"
4910 echo "This is test file 2!" > $MOUNT/$tdir.1000/$tfile.2 ||
4911 error "failed to make test file 2"
4912 echo "This is test file 1000!" > $MOUNT/$tdir.1000/$tfile.1000 ||
4913 error "failed to make test file 1000"
4914 rm -rf $MOUNT/$tdir.1000 || error "failed to remove remote_dir"
4917 echo "=== START lfs mdts OUTPUT ==="
4919 echo "==== END lfs mdts OUTPUT ===="
4921 echo -e "$output" | grep -v "MDTS:" | awk '{print $1}' |
4922 sed 's/://g' > $TMP/mdts-actual.txt
4923 sort $TMP/mdts-actual.txt -o $TMP/mdts-actual.txt
4925 echo -e "0\n1\n1000" > $TMP/mdts-expected.txt
4927 diff $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4930 rm $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4932 [ $result -eq 0 ] || error "target_obd proc file is incorrect!"
4934 run_test 56b "test target_obd correctness with nonconsecutive MDTs"
4936 test_57a() { # bug 22656
4937 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4938 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4939 tail -1 | awk '{print $1}')
4940 writeconf_or_reformat
4941 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
4942 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" ||
4943 error "tunefs failed"
4945 start_ost && error "OST registration from failnode should fail"
4948 run_test 57a "initial registration from failnode should fail (should return errs)"
4951 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4952 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4953 tail -1 | awk '{print $1}')
4954 writeconf_or_reformat
4955 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
4956 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" ||
4957 error "tunefs failed"
4959 start_ost || error "OST registration from servicenode should not fail"
4962 run_test 57b "initial registration from servicenode should not fail"
4965 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
4968 test_58() { # bug 22658
4969 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
4971 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4972 createmany -o $DIR/$tdir/$tfile-%d 100
4973 unlinkmany $DIR/$tdir/$tfile-%d 100
4974 stop_mds || error "Unable to stop MDS"
4976 local MNTDIR=$(facet_mntpt $SINGLEMDS)
4977 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4979 # remove all files from the OBJECTS dir
4980 mount_fstype $SINGLEMDS
4982 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
4984 unmount_fstype $SINGLEMDS
4985 # restart MDS with missing llog files
4986 start_mds || error "unable to start MDS"
4987 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0"
4990 run_test 58 "missing llog files must not prevent MDT from mounting"
4993 start_mgsmds >> /dev/null
4994 local C1=$(count_osts)
4995 if [ $C1 -eq 0 ]; then
4996 start_ost >> /dev/null
5000 echo "original ost count: $C1 (expect > 0)"
5001 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
5002 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
5003 local C2=$(count_osts)
5004 echo "after mdt writeconf count: $C2 (expect 0)"
5005 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
5006 echo "OST start without writeconf should fail:"
5007 start_ost >> /dev/null &&
5008 error "OST start without writeconf didn't fail"
5009 echo "OST start with writeconf should succeed:"
5010 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
5011 local C3=$(count_osts)
5012 echo "after ost writeconf count: $C3 (expect 1)"
5013 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
5014 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
5015 local C4=$(count_osts)
5016 echo "after ost2 writeconf count: $C4 (expect 2)"
5017 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
5018 stop_ost2 >> /dev/null
5019 cleanup_nocli >> /dev/null
5020 #writeconf to remove all ost2 traces for subsequent tests
5021 writeconf_or_reformat
5023 run_test 59 "writeconf mount option"
5025 test_60a() { # LU-471
5026 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5027 skip "ldiskfs only test"
5032 for num in $(seq $MDSCOUNT); do
5033 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
5034 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
5035 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
5039 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
5040 [ ${PIPESTATUS[0]} -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
5042 # MDT default has dirdata feature
5043 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
5044 # we disable uninit_bg feature
5045 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
5046 # we set stride extended options
5047 echo $dump | grep stride > /dev/null || error "stride is not set"
5051 run_test 60a "check mkfs.lustre --mkfsoptions -E -O options setting"
5054 [[ "$mds1_FSTYPE" == ldiskfs ]] || skip "ldiskfs only test"
5056 local features=$(do_facet $SINGLEMDS $DUMPE2FS $(mdsdevname 1) |
5058 [ ${PIPESTATUS[0]} -eq 0 ] || error "$DUMPE2FS $(mdsdevname 1) failed"
5061 # ea_inode feature should be enabled by default for MDTs
5062 [[ "$features" =~ "ea_inode" ]] || error "ea_inode is not set"
5063 # large_dir feature should be enabled by default for MDTs
5064 [[ "$features" =~ "large_dir" ]] || error "large_dir is not set"
5066 run_test 60b "check mkfs.lustre MDT default features"
5068 test_61a() { # LU-80
5069 local lxattr=$(large_xattr_enabled)
5071 [ "$MDS1_VERSION" -ge $(version_code 2.1.53) ] ||
5072 skip "Need MDS version at least 2.1.53"
5074 if [ "$mds1_FSTYPE" == ldiskfs ] &&
5075 ! large_xattr_enabled; then
5078 for num in $(seq $MDSCOUNT); do
5079 do_facet mds${num} $TUNE2FS -O ea_inode \
5080 $(mdsdevname $num) ||
5081 error "tune2fs on mds $num failed"
5085 setup || error "setting up the filesystem failed"
5086 client_up || error "starting client failed"
5088 local file=$DIR/$tfile
5089 touch $file || error "touch $file failed"
5091 local large_value="$(generate_string $(max_xattr_size))"
5092 local small_value="bar"
5094 local name="trusted.big"
5095 log "save large xattr of $(max_xattr_size) bytes on $name on $file"
5096 setfattr -n $name -v $large_value $file ||
5097 error "saving $name on $file failed"
5099 local new_value=$(get_xattr_value $name $file)
5100 [[ "$new_value" != "$large_value" ]] &&
5101 error "$name different after saving"
5103 log "shrink value of $name on $file"
5104 setfattr -n $name -v $small_value $file ||
5105 error "shrinking value of $name on $file failed"
5107 new_value=$(get_xattr_value $name $file)
5108 [[ "$new_value" != "$small_value" ]] &&
5109 error "$name different after shrinking"
5111 log "grow value of $name on $file"
5112 setfattr -n $name -v $large_value $file ||
5113 error "growing value of $name on $file failed"
5115 new_value=$(get_xattr_value $name $file)
5116 [[ "$new_value" != "$large_value" ]] &&
5117 error "$name different after growing"
5119 log "check value of $name on $file after remounting MDS"
5121 new_value=$(get_xattr_value $name $file)
5122 [[ "$new_value" != "$large_value" ]] &&
5123 error "$name different after remounting MDS"
5125 log "remove large xattr $name from $file"
5126 setfattr -x $name $file || error "removing $name from $file failed"
5128 if $lxattr && [ "$mds1_FSTYPE" == ldiskfs ]; then
5129 stopall || error "stopping for e2fsck run"
5130 for num in $(seq $MDSCOUNT); do
5131 run_e2fsck $(facet_active_host mds$num) \
5132 $(mdsdevname $num) "-y" ||
5133 error "e2fsck MDT$num failed"
5135 setup_noconfig || error "remounting the filesystem failed"
5138 # need to delete this file to avoid problems in other tests
5140 cleanup || error "stopping systems failed"
5142 run_test 61a "large xattr"
5144 test_61b() { # LU-80
5145 local lxattr=$(large_xattr_enabled)
5147 [ "$MDS1_VERSION" -ge $(version_code 2.1.53) ] ||
5148 skip "Need MDS version at least 2.1.53"
5150 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5151 skip "ldiskfs specific bug"
5154 if ! large_xattr_enabled; then
5157 for num in $(seq $MDSCOUNT); do
5158 do_facet mds${num} $TUNE2FS -O ea_inode \
5159 $(mdsdevname $num) ||
5160 error "tune2fs on mds $num failed"
5164 setup || error "setting up the filesystem failed"
5165 client_up || error "starting client failed"
5167 local _file=$MOUNT/panda
5168 local large_value="$(generate_string $(max_xattr_size))"
5169 local name="trusted.big"
5171 touch ${_file} || error "touch ${_file} failed"
5172 setfattr -n $name -v $large_value ${_file} ||
5173 error "saving $name on $file failed"
5175 MDT_DEV="${FSNAME}-MDT0000"
5176 MDT_DEVNAME=$(mdsdevname ${SINGLEMDS//mds/})
5178 stopall || error "stopping for e2fsck run"
5181 ino=$(do_facet $SINGLEMDS "$DEBUGFS -R 'stat /ROOT/panda' \
5182 ${MDT_DEVNAME} | grep trusted.big")
5183 ino=$(echo "${ino}" | awk '{print $2;}')
5184 echo "large ea "${ino}
5186 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"ln $ino /lost+found\\\" \
5189 setup_noconfig || error "remounting the filesystem failed"
5191 do_facet $SINGLEMDS $LCTL lfsck_start -M ${MDT_DEV} -t namespace || {
5192 error "can't start lfsck namespace"
5196 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
5197 mdd.${MDT_DEV}.lfsck_namespace |
5198 awk '/^status/ { print \\\$2 }'" "completed" 32 || {
5199 error "(2) unexpected status"
5202 stopall || error "stopping for e2fsck run"
5203 for num in $(seq $MDSCOUNT); do
5204 run_e2fsck $(facet_active_host mds$num) \
5205 $(mdsdevname $num) "-y" ||
5206 error "e2fsck MDT$num failed"
5208 setup_noconfig || error "remounting the filesystem failed"
5210 # need to delete this file to avoid problems in other tests
5212 cleanup || error "stopping systems failed"
5214 run_test 61b "large xattr"
5217 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5218 skip "ldiskfs only test"
5220 [[ "$MDS1_VERSION" -ge $(version_code 2.2.51) ]] ||
5221 skip "Need MDS version at least 2.2.51"
5224 local mdsdev=$(mdsdevname 1)
5225 local ostdev=$(ostdevname 1)
5227 echo "disable journal for mds"
5228 do_facet mds1 $TUNE2FS -O ^has_journal $mdsdev || error "tune2fs failed"
5229 start_mds && error "MDT start should fail"
5230 echo "disable journal for ost"
5231 do_facet ost1 $TUNE2FS -O ^has_journal $ostdev || error "tune2fs failed"
5232 start_ost && error "OST start should fail"
5233 cleanup || error "cleanup failed with rc $?"
5236 run_test 62 "start with disabled journal"
5239 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5240 skip "ldiskfs only test"
5243 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ldiskfs
5244 local inode_slab=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
5245 awk '/ldiskfs_inode_cache/ { print $5 / $6 }')
5246 if [ -z "$inode_slab" ]; then
5247 skip "ldiskfs module has not been loaded"
5250 if grep -q "CONFIG_DEBUG_LOCK_ALLOC=y" /boot/config-$(uname -r); then
5251 skip "test is not compatible with CONFIG_DEBUG_LOCK_ALLOC=y"
5254 echo "$inode_slab ldiskfs inodes per page"
5255 [ "${inode_slab%.*}" -ge "3" ] && return 0
5257 # If kmalloc-128 is also 1 per page - this is a debug kernel
5258 # and so this is not an error.
5259 local kmalloc128=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
5260 awk '/^(dma-kmalloc|size)-128 / { print $5 / $6 }')
5261 # 32 128-byte chunks in 4k
5262 [ "${kmalloc128%.*}" -lt "32" ] ||
5263 error "ldiskfs inode too big, only $inode_slab objs/page, " \
5264 "kmalloc128 = $kmalloc128 objs/page"
5266 run_test 63 "Verify each page can at least hold 3 ldiskfs inodes"
5269 start_mds || error "unable to start MDS"
5270 start_ost || error "Unable to start OST1"
5271 start_ost2 || error "Unable to start second ost"
5272 mount_client $MOUNT || error "Unable to mount client"
5273 stop_ost2 || error "Unable to stop second ost"
5276 umount_client $MOUNT -f || error “unmount $MOUNT failed”
5277 cleanup_nocli || error "cleanup_nocli failed with $?"
5278 #writeconf to remove all ost2 traces for subsequent tests
5279 writeconf_or_reformat
5281 run_test 64 "check lfs df --lazy "
5283 test_65() { # LU-2237
5284 # Currently, the test is only valid for ldiskfs backend
5285 [ "$mds1_FSTYPE" != ldiskfs ] &&
5286 skip "ldiskfs only test"
5288 local devname=$(mdsdevname ${SINGLEMDS//mds/})
5289 local brpt=$(facet_mntpt brpt)
5292 if ! do_facet $SINGLEMDS "test -b $devname"; then
5296 stop_mds || error "Unable to stop MDS"
5297 local obj=$(do_facet $SINGLEMDS \
5298 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
5300 if [ -z "$obj" ]; then
5301 # The MDT may be just re-formatted, mount the MDT for the
5302 # first time to guarantee the "last_rcvd" file is there.
5303 start_mds || error "fail to mount the MDS for the first time"
5304 stop_mds || error "Unable to stop MDS"
5307 # remove the "last_rcvd" file
5308 do_facet $SINGLEMDS "mkdir -p $brpt"
5309 do_facet $SINGLEMDS \
5310 "mount -t $mds1_FSTYPE $opts $devname $brpt"
5311 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
5312 do_facet $SINGLEMDS "$UMOUNT $brpt"
5314 # restart MDS, the "last_rcvd" file should be recreated.
5315 start_mds || error "fail to restart the MDS"
5316 stop_mds || error "Unable to stop MDS"
5317 obj=$(do_facet $SINGLEMDS \
5318 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
5319 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
5321 run_test 65 "re-create the lost last_rcvd file when server mount"
5324 [[ "$MGS_VERSION" -ge $(version_code 2.3.59) ]] ||
5325 skip "Need MGS version at least 2.3.59"
5328 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
5329 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
5331 # add EXCLUDE records to config log, they are not to be
5332 # removed by lctl replace_nids
5333 set_conf_param_and_check mds \
5334 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
5335 "$FSNAME-OST0000.osc.active" \
5338 echo "replace_nids should fail if MDS, OSTs and clients are UP"
5339 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5340 error "replace_nids fail"
5342 umount_client $MOUNT || error "unmounting client failed"
5343 echo "replace_nids should fail if MDS and OSTs are UP"
5344 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5345 error "replace_nids fail"
5347 stop_ost || error "Unable to stop OST1"
5348 echo "replace_nids should fail if MDS is UP"
5349 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5350 error "replace_nids fail"
5352 stop_mds || error "stopping mds failed"
5354 if combined_mgs_mds; then
5355 start_mdt 1 "-o nosvc" ||
5356 error "starting mds with nosvc option failed"
5359 echo "command should accept two parameters"
5360 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
5361 error "command should accept two params"
5363 echo "correct device name should be passed"
5364 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
5365 error "wrong devname"
5367 echo "wrong nids list should not destroy the system"
5368 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
5370 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "asdfasdf, asdfadf" &&
5373 echo "replace OST nid"
5374 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
5375 error "replace nids failed"
5377 echo "command should accept two parameters"
5378 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
5379 error "command should accept two params"
5381 echo "wrong nids list should not destroy the system"
5382 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
5385 local FAKE_NIDS="192.168.0.112@tcp1,192.168.0.112@tcp2"
5386 local FAKE_FAILOVER="192.168.0.113@tcp1,192.168.0.113@tcp2"
5387 local NIDS_AND_FAILOVER="$MDS_NID,$FAKE_NIDS:$FAKE_FAILOVER"
5388 echo "set NIDs with failover"
5389 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $NIDS_AND_FAILOVER ||
5390 error "replace nids failed"
5393 echo "replace MDS nid"
5394 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
5395 error "replace nids failed"
5397 if ! combined_mgs_mds ; then
5400 stop_mds || error "Unable to stop MDS"
5403 start_mgsmds || error "start mgsmds failed"
5404 set_conf_param_and_check mds \
5405 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
5406 "$FSNAME-OST0000.osc.active" \
5408 start_ost || error "unable to start OST"
5409 mount_client $MOUNT || error "mount client failed"
5411 check_mount || error "error after nid replace"
5412 cleanup || error "cleanup failed"
5415 run_test 66 "replace nids"
5417 test_67() { #LU-2950
5418 local legacy="$TMP/legacy_lnet_config"
5419 local new="$TMP/new_routes_test"
5420 local out="$TMP/config_out_file"
5421 local verify="$TMP/conv_verify"
5422 local verify_conf="$TMP/conf_verify"
5424 # Create the legacy file that will be run through the
5425 # lustre_routes_conversion script
5426 cat <<- LEGACY_LNET_CONFIG > $legacy
5427 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
5428 tcp2 54 10.1.3.2@tcp;
5429 tcp3 10.3.4.3@tcp:3;
5433 # Create the verification file to verify the output of
5434 # lustre_routes_conversion script against.
5435 cat <<- VERIFY_LNET_CONFIG > $verify
5436 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
5437 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
5438 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
5439 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
5440 tcp4: { gateway: 10.3.3.4@tcp }
5443 # Create the verification file to verify the output of
5444 # lustre_routes_config script against
5445 cat <<- VERIFY_LNET_CONFIG > $verify_conf
5446 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
5447 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
5448 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
5449 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
5450 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
5453 $LUSTRE_ROUTES_CONVERSION $legacy $new > /dev/null
5454 if [ -f $new ]; then
5455 # verify the conversion output
5456 cmp -s $new $verify > /dev/null
5457 if [ $? -eq 1 ]; then
5458 error "routes conversion failed"
5461 lustre_routes_config --dry-run --verbose $new > $out
5462 # check that the script succeeded
5463 cmp -s $out $verify_conf > /dev/null
5464 if [ $? -eq 1 ]; then
5465 error "routes config failed"
5468 error "routes conversion test failed"
5470 # remove generated files
5471 rm -f $new $legacy $verify $verify_conf $out
5473 run_test 67 "test routes conversion and configuration"
5481 [ "$MDS1_VERSION" -ge $(version_code 2.4.53) ] ||
5482 skip "Need MDS version at least 2.4.53"
5484 umount_client $MOUNT || error "umount client failed"
5489 # START-END - the sequences we'll be reserving
5490 START=$(do_facet $SINGLEMDS \
5491 $LCTL get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
5492 END=$((START + (1 << 30)))
5493 do_facet $SINGLEMDS \
5494 $LCTL set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
5496 # reset the sequences MDT0000 has already assigned
5497 do_facet $SINGLEMDS \
5498 $LCTL set_param seq.srv*MDT0000.space=clear
5500 # remount to let the client allocate new sequence
5501 mount_client $MOUNT || error "mount client failed"
5503 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
5504 do_facet $SINGLEMDS \
5505 $LCTL get_param seq.srv*MDT0000.space
5506 $LFS path2fid $DIR/$tfile
5508 local old_ifs="$IFS"
5510 fid=($($LFS path2fid $DIR/$tfile))
5514 if [[ $seq < $END ]]; then
5515 error "used reserved sequence $seq?"
5517 cleanup || error "cleanup failed with $?"
5519 run_test 68 "be able to reserve specific sequences in FLDB"
5521 # Test 69: is about the total number of objects ever created on an OST.
5522 # so that when it is reformatted the normal MDS->OST orphan recovery won't
5523 # just "precreate" the missing objects. In the past it might try to recreate
5524 # millions of objects after an OST was reformatted
5526 [[ "$MDS1_VERSION" -lt $(version_code 2.4.2) ]] &&
5527 skip "Need MDS version at least 2.4.2"
5529 [[ "$MDS1_VERSION" -ge $(version_code 2.4.50) ]] &&
5530 [[ "$MDS1_VERSION" -lt $(version_code 2.5.0) ]] &&
5531 skip "Need MDS version at least 2.5.0"
5534 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5536 # use OST0000 since it probably has the most creations
5537 local OSTNAME=$(ostname_from_index 0)
5538 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
5539 local last_id=$(do_facet mds1 $LCTL get_param -n \
5540 osp.$mdtosc_proc1.prealloc_last_id)
5542 # Want to have OST LAST_ID over 5 * OST_MAX_PRECREATE to
5543 # verify that the LAST_ID recovery is working properly. If
5544 # not, then the OST will refuse to allow the MDS connect
5545 # because the LAST_ID value is too different from the MDS
5546 #define OST_MAX_PRECREATE=20000
5547 local ost_max_pre=20000
5548 local num_create=$(( ost_max_pre * 5 + 1 - last_id))
5550 # If the LAST_ID is already over 5 * OST_MAX_PRECREATE, we don't
5551 # need to create any files. So, skip this section.
5552 if [ $num_create -gt 0 ]; then
5553 # Check the number of inodes available on OST0
5555 local ifree=$($LFS df -i $MOUNT |
5556 awk '/OST0000/ { print $4 }'; exit ${PIPESTATUS[0]})
5557 log "On OST0, $ifree inodes available. Want $num_create. rc=$?"
5559 $LFS setstripe -i 0 $DIR/$tdir ||
5560 error "$LFS setstripe -i 0 $DIR/$tdir failed"
5561 if [ $ifree -lt 10000 ]; then
5562 files=$(( ifree - 50 ))
5567 local j=$((num_create / files + 1))
5568 for i in $(seq 1 $j); do
5569 createmany -o $DIR/$tdir/$tfile-$i- $files ||
5570 error "createmany fail create $files files: $?"
5571 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
5572 error "unlinkmany failed unlink $files files"
5576 # delete all of the files with objects on OST0 so the
5577 # filesystem is not inconsistent later on
5578 $LFS find $MOUNT --ost 0 -print0 | xargs -0 rm
5580 umount_client $MOUNT || error "umount client failed"
5581 stop_ost || error "OST0 stop failure"
5582 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
5583 $(ostdevname 1) $(ostvdevname 1) ||
5584 error "reformat and replace $ostdev failed"
5585 start_ost || error "OST0 restart failure"
5586 wait_osc_import_state mds ost FULL
5588 mount_client $MOUNT || error "mount client failed"
5589 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
5590 local idx=$($LFS getstripe -i $DIR/$tdir/$tfile-last)
5591 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
5593 local iused=$($LFS df -i $MOUNT |
5594 awk '/OST0000/ { print $3 }'; exit ${PIPESTATUS[0]})
5595 log "On OST0, $iused used inodes rc=$?"
5596 [ $iused -ge $((ost_max_pre + 1000)) ] &&
5597 error "OST replacement created too many inodes; $iused"
5598 cleanup || error "cleanup failed with $?"
5600 run_test 69 "replace an OST with the same index"
5603 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5606 cleanup || error "cleanup failed with $?"
5608 start_mdt 1 || error "MDT0 start fail"
5610 start_ost || error "OST0 start fail"
5611 for num in $(seq 2 $MDSCOUNT); do
5612 start_mdt $num || return
5615 mount_client $MOUNT || error "mount client fails"
5617 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
5619 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5620 error "create remote dir fail"
5622 rm -rf $DIR/$tdir || error "delete dir fail"
5623 cleanup || error "cleanup failed with $?"
5625 run_test 70a "start MDT0, then OST, then MDT1"
5628 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5631 start_ost || error "OST0 start fail"
5633 start_mds || error "MDS start fail"
5635 mount_client $MOUNT || error "mount client fails"
5637 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
5639 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5640 error "create remote dir fail"
5642 rm -rf $DIR/$tdir || error "delete dir fail"
5644 cleanup || error "cleanup failed with $?"
5646 run_test 70b "start OST, MDT1, MDT0"
5649 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5652 start_mds || error "MDS start fail"
5653 start_ost || error "OST0 start fail"
5655 mount_client $MOUNT || error "mount client fails"
5656 stop_mdt 1 || error "MDT1 start fail"
5658 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
5659 echo "deactivate $mdc_for_mdt1"
5660 $LCTL --device $mdc_for_mdt1 deactivate ||
5661 error "set $mdc_for_mdt1 deactivate failed"
5663 mkdir $DIR/$tdir && error "mkdir succeed"
5665 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
5666 error "create remote dir succeed"
5668 cleanup || error "cleanup failed with $?"
5670 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
5673 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5676 start_mds || error "MDS start fail"
5677 start_ost || error "OST0 start fail"
5679 mount_client $MOUNT || error "mount client fails"
5681 stop_mdt 2 || error "MDT1 start fail"
5683 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
5685 echo "deactivate $mdc_for_mdt2"
5686 $LCTL --device $mdc_for_mdt2 deactivate ||
5687 error "set $mdc_for_mdt2 deactivate failed"
5689 mkdir $DIR/$tdir || error "mkdir fail"
5690 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
5691 error "create remote dir succeed"
5693 rm -rf $DIR/$tdir || error "delete dir fail"
5695 cleanup || error "cleanup failed with $?"
5697 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
5700 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5702 [ "$MDS1_VERSION" -ge $(version_code 2.7.62) ] ||
5703 skip "Need MDS version at least 2.7.62"
5705 reformat || error "reformat failed with $?"
5709 local mdsdev=$(mdsdevname 1)
5710 local ostdev=$(ostdevname 1)
5711 local mdsvdev=$(mdsvdevname 1)
5712 local ostvdev=$(ostvdevname 1)
5713 local opts_mds="$(mkfs_opts mds1 $mdsdev) --reformat $mdsdev $mdsvdev"
5714 local opts_ost="$(mkfs_opts ost1 $ostdev) --reformat $ostdev $ostvdev"
5716 if ! combined_mgs_mds ; then
5720 add mds1 $opts_mds || error "add mds1 failed"
5721 start_mdt 1 || error "start mdt1 failed"
5722 add ost1 $opts_ost || error "add ost1 failed"
5723 start_ost || error "start ost failed"
5724 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
5726 local soc=$(do_facet mds1 "$LCTL get_param -n \
5727 mdt.*MDT0000.sync_lock_cancel")
5728 [ $soc == "never" ] || error "SoC enabled on single MDS"
5730 for i in $(seq 2 $MDSCOUNT); do
5731 mdsdev=$(mdsdevname $i)
5732 mdsvdev=$(mdsvdevname $i)
5733 opts_mds="$(mkfs_opts mds$i $mdsdev) --reformat $mdsdev \
5735 add mds$i $opts_mds || error "add mds$i failed"
5736 start_mdt $i || error "start mdt$i fail"
5739 wait_dne_interconnect
5741 for i in $(seq $MDSCOUNT); do
5742 soc=$(do_facet mds$i "$LCTL get_param -n \
5743 mdt.*MDT000$((i - 1)).sync_lock_cancel")
5744 [ $soc == "blocking" ] || error "SoC not enabled on DNE"
5747 for i in $(seq 2 $MDSCOUNT); do
5748 stop_mdt $i || error "stop mdt$i fail"
5750 soc=$(do_facet mds1 "$LCTL get_param -n \
5751 mdt.*MDT0000.sync_lock_cancel")
5752 [ $soc == "never" ] || error "SoC enabled on single MDS"
5753 umount_client $MOUNT -f > /dev/null
5755 cleanup || error "cleanup failed with $?"
5757 run_test 70e "Sync-on-Cancel will be enabled by default on DNE"
5760 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5761 if combined_mgs_mds; then
5762 skip "needs separate MGS/MDT"
5766 start_mdt 1 || error "MDT0 start fail"
5767 start_ost || error "OST0 start fail"
5768 for num in $(seq 2 $MDSCOUNT); do
5769 start_mdt $num || return
5772 start_ost2 || error "OST1 start fail"
5774 mount_client $MOUNT || error "mount client fails"
5776 mkdir $DIR/$tdir || error "mkdir fail"
5777 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5778 error "create remote dir succeed"
5780 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5781 rm -rf $DIR/$tdir || error "delete dir fail"
5783 umount_client $MOUNT || error "umount_client failed"
5784 stop_mds || error "MDS stop fail"
5785 stop_ost || error "OST0 stop fail"
5786 stop_ost2 || error "OST1 stop fail"
5788 run_test 71a "start MDT0 OST0, MDT1, OST1"
5791 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5792 if combined_mgs_mds; then
5793 skip "needs separate MGS/MDT"
5797 for num in $(seq 2 $MDSCOUNT); do
5798 start_mdt $num || return
5800 start_ost || error "OST0 start fail"
5801 start_mdt 1 || error "MDT0 start fail"
5802 start_ost2 || error "OST1 start fail"
5804 mount_client $MOUNT || error "mount client fails"
5806 mkdir $DIR/$tdir || error "mkdir fail"
5807 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5808 error "create remote dir succeed"
5810 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5811 rm -rf $DIR/$tdir || error "delete dir fail"
5813 umount_client $MOUNT || error "umount_client failed"
5814 stop_mds || error "MDT0 stop fail"
5815 stop_ost || error "OST0 stop fail"
5816 stop_ost2 || error "OST1 stop fail"
5818 run_test 71b "start MDT1, OST0, MDT0, OST1"
5821 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5822 combined_mgs_mds && skip "needs separate MGS/MDT"
5826 start_ost || error "OST0 start fail"
5827 start_ost2 || error "OST1 start fail"
5828 for num in $(seq 2 $MDSCOUNT); do
5829 start_mdt $num || return
5831 start_mdt 1 || error "MDT0 start fail"
5833 mount_client $MOUNT || error "mount client fails"
5835 mkdir $DIR/$tdir || error "mkdir fail"
5836 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5837 error "create remote dir succeed"
5839 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5840 rm -rf $DIR/$tdir || error "delete dir fail"
5842 umount_client $MOUNT || error "umount_client failed"
5843 stop_mds || error "MDS stop fail"
5844 stop_ost || error "OST0 stop fail"
5845 stop_ost2 || error "OST1 stop fail"
5848 run_test 71c "start OST0, OST1, MDT1, MDT0"
5851 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5852 combined_mgs_mds && skip "needs separate MGS/MDT"
5856 start_ost || error "OST0 start fail"
5857 for num in $(seq 2 $MDSCOUNT); do
5858 start_mdt $num || return
5860 start_mdt 1 || error "MDT0 start fail"
5861 start_ost2 || error "OST1 start fail"
5863 mount_client $MOUNT || error "mount client fails"
5865 mkdir $DIR/$tdir || error "mkdir fail"
5866 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5867 error "create remote dir succeed"
5869 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5870 rm -rf $DIR/$tdir || error "delete dir fail"
5872 umount_client $MOUNT || error "umount_client failed"
5873 stop_mds || error "MDS stop fail"
5874 stop_ost || error "OST0 stop fail"
5875 stop_ost2 || error "OST1 stop fail"
5878 run_test 71d "start OST0, MDT1, MDT0, OST1"
5881 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5882 combined_mgs_mds && skip "needs separate MGS/MDT"
5886 start_ost || error "OST0 start fail"
5887 for num in $(seq 2 $MDSCOUNT); do
5888 start_mdt $num || return
5890 start_ost2 || error "OST1 start fail"
5891 start_mdt 1 || error "MDT0 start fail"
5893 mount_client $MOUNT || error "mount client fails"
5895 mkdir $DIR/$tdir || error "mkdir fail"
5896 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5897 error "create remote dir succeed"
5899 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5900 rm -rf $DIR/$tdir || error "delete dir fail"
5902 umount_client $MOUNT || error "umount_client failed"
5903 stop_mds || error "MDS stop fail"
5904 stop_ost || error "OST0 stop fail"
5905 stop_ost2 || error "OST1 stop fail"
5908 run_test 71e "start OST0, MDT1, OST1, MDT0"
5910 test_72() { #LU-2634
5911 [ "$mds1_FSTYPE" != ldiskfs ] &&
5912 skip "ldiskfs only test"
5914 local mdsdev=$(mdsdevname 1)
5915 local ostdev=$(ostdevname 1)
5916 local cmd="$E2FSCK -fnvd $mdsdev"
5923 if combined_mgs_mds; then
5924 add_options='--reformat'
5926 add_options='--reformat --replace'
5929 #tune MDT with "-O extents"
5931 for num in $(seq $MDSCOUNT); do
5932 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
5933 $add_options $(mdsdevname $num) $(mdsvdevname $num) ||
5934 error "add mds $num failed"
5935 do_facet mds${num} "$TUNE2FS -O extents $(mdsdevname $num)" ||
5936 error "$TUNE2FS failed on mds${num}"
5939 add ost1 $(mkfs_opts ost1 $ostdev) $add_options $ostdev ||
5940 error "add $ostdev failed"
5941 start_mds || error "start mds failed"
5942 start_ost || error "start ost failed"
5943 mount_client $MOUNT || error "mount client failed"
5945 #create some short symlinks
5946 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5947 createmany -o $DIR/$tdir/$tfile-%d $fn
5948 echo "create $fn short symlinks"
5949 for i in $(seq -w 1 $fn); do
5950 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
5955 umount_client $MOUNT || error "umount client failed"
5956 stop_mds || error "stop mds failed"
5957 stop_ost || error "stop ost failed"
5960 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
5962 run_test 72 "test fast symlink with extents flag enabled"
5964 test_73() { #LU-3006
5965 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
5966 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
5967 error "1st tunefs failed"
5968 start_mgsmds || error "start mds failed"
5969 start_ost || error "start ost failed"
5970 mount_client $MOUNT || error "mount client failed"
5971 $LCTL get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
5972 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
5973 umount_client $MOUNT || error "umount client failed"
5977 run_test 73 "failnode to update from mountdata properly"
5979 test_75() { # LU-2374
5980 [[ "$MDS1_VERSION" -lt $(version_code 2.4.1) ]] &&
5981 skip "Need MDS version at least 2.4.1"
5984 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
5985 --replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
5986 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
5987 --replace --reformat $(ostdevname 1) $(ostvdevname 1)"
5989 #check with default parameters
5990 add mds1 $opts_mds || error "add mds1 failed for default params"
5991 add ost1 $opts_ost || error "add ost1 failed for default params"
5993 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
5994 opts_mds=$(echo $opts_mds |
5995 sed -e "s/--index=$index/--index=$index --mdt/")
5996 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
5997 opts_ost=$(echo $opts_ost |
5998 sed -e "s/--index=$index/--index=$index --ost/")
6000 add mds1 $opts_mds || error "add mds1 failed for new params"
6001 add ost1 $opts_ost || error "add ost1 failed for new params"
6006 run_test 75 "The order of --index should be irrelevant"
6009 [[ "$MGS_VERSION" -ge $(version_code 2.4.52) ]] ||
6010 skip "Need MDS version at least 2.4.52"
6013 local MDMB_PARAM="osc.*.max_dirty_mb"
6014 echo "Change MGS params"
6015 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
6017 echo "max_dirty_mb: $MAX_DIRTY_MB"
6018 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB - 10))
6019 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
6020 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
6021 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
6022 head -1" $NEW_MAX_DIRTY_MB
6023 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
6024 echo "$MAX_DIRTY_MB"
6025 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
6026 error "error while apply max_dirty_mb"
6028 echo "Check the value is stored after remount"
6031 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
6032 head -1" $NEW_MAX_DIRTY_MB
6033 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
6034 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
6035 error "max_dirty_mb is not saved after remount"
6037 echo "Change OST params"
6038 CLIENT_PARAM="obdfilter.*.client_cache_count"
6039 local CLIENT_CACHE_COUNT
6040 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
6042 echo "client_cache_count: $CLIENT_CACHE_COUNT"
6043 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
6044 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
6045 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
6046 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
6047 head -1" $NEW_CLIENT_CACHE_COUNT
6048 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
6050 echo "$CLIENT_CACHE_COUNT"
6051 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
6052 error "error while apply client_cache_count"
6054 echo "Check the value is stored after remount"
6057 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
6058 head -1" $NEW_CLIENT_CACHE_COUNT
6059 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
6061 echo "$CLIENT_CACHE_COUNT"
6062 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
6063 error "client_cache_count is not saved after remount"
6066 run_test 76a "set permanent params with lctl across mounts"
6068 test_76b() { # LU-4783
6069 [[ "$MGS_VERSION" -ge $(version_code 2.5.57) ]] ||
6070 skip "Need MGS version at least 2.5.57"
6073 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
6074 error "start params log failed"
6077 run_test 76b "verify params log setup correctly"
6080 [[ "$MGS_VERSION" -ge $(version_code 2.8.54) ]] ||
6081 skip "Need MDS version at least 2.4.52"
6083 local MASK_PARAM="mdd.*.changelog_mask"
6084 echo "Change changelog_mask"
6085 do_facet mgs $LCTL set_param -P $MASK_PARAM=-CLOSE ||
6086 error "Can't change changlog_mask"
6087 wait_update $(facet_host mds) "$LCTL get_param -n $MASK_PARAM |
6090 echo "Check the value is stored after mds remount"
6091 stop_mds || error "Failed to stop MDS"
6092 start_mds || error "Failed to start MDS"
6093 local CHANGELOG_MASK=$(do_facet mgs $LCTL get_param -n $MASK_PARAM)
6094 echo $CHANGELOG_MASK | grep CLOSE > /dev/null &&
6095 error "changelog_mask is not changed"
6099 run_test 76c "verify changelog_mask is applied with lctl set_param -P"
6101 test_76d() { #LU-9399
6104 local xattr_cache="llite.*.xattr_cache"
6105 local cmd="$LCTL get_param -n $xattr_cache | head -1"
6106 local new=$((($(eval $cmd) + 1) % 2))
6108 echo "lctl set_param -P llite.*.xattr_cache=$new"
6109 do_facet mgs $LCTL set_param -P $xattr_cache=$new ||
6110 error "Can't change xattr_cache"
6111 wait_update $HOSTNAME "$cmd" "$new"
6113 echo "Check $xattr_cache on client $MOUNT"
6114 umount_client $MOUNT || error "umount $MOUNT failed"
6115 mount_client $MOUNT || error "mount $MOUNT failed"
6116 [ $(eval $cmd) -eq $new ] ||
6117 error "$xattr_cache != $new on client $MOUNT"
6119 echo "Check $xattr_cache on the new client $MOUNT2"
6120 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
6121 [ $(eval $cmd) -eq $new ] ||
6122 error "$xattr_cache != $new on client $MOUNT2"
6123 umount_client $MOUNT2 || error "umount $MOUNT2 failed"
6127 run_test 76d "verify llite.*.xattr_cache can be set by 'lctl set_param -P' correctly"
6129 test_77() { # LU-3445
6130 [[ "$MDS1_VERSION" -ge $(version_code 2.8.55) ]] ||
6131 skip "Need MDS version 2.8.55+ "
6133 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
6134 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
6135 skip_env "mixed loopback and real device not working"
6138 local fs2mdsdev=$(mdsdevname 1_2)
6139 local fs2ostdev=$(ostdevname 1_2)
6140 local fs2mdsvdev=$(mdsvdevname 1_2)
6141 local fs2ostvdev=$(ostvdevname 1_2)
6142 local fsname=test1234
6144 local failnid="$(h2nettype 1.2.3.4),$(h2nettype 4.3.2.1)"
6146 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
6148 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
6149 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
6150 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
6151 error "start fs2mds failed"
6153 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
6154 mgsnid="0.0.0.0@tcp,$mgsnid,$mgsnid:$mgsnid"
6156 add fs2ost --mgsnode=$mgsnid $(mkfs_opts ost1 $fs2ostdev) \
6157 --failnode=$failnid --fsname=$fsname \
6158 --reformat $fs2ostdev $fs2ostvdev ||
6159 error "add fs2ost failed"
6160 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
6162 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
6163 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
6164 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
6167 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
6170 [[ "$mds1_FSTYPE" != ldiskfs ||
6171 "$ost1_FSTYPE" != ldiskfs ]] &&
6172 skip "ldiskfs only test"
6174 # reformat the Lustre filesystem with a smaller size
6175 local saved_MDSCOUNT=$MDSCOUNT
6176 local saved_MDSSIZE=$MDSSIZE
6177 local saved_OSTCOUNT=$OSTCOUNT
6178 local saved_OSTSIZE=$OSTSIZE
6181 MDSSIZE=$((MDSSIZE - 20000))
6182 OSTSIZE=$((OSTSIZE - 20000))
6183 reformat || error "(1) reformat Lustre filesystem failed"
6184 MDSSIZE=$saved_MDSSIZE
6185 OSTSIZE=$saved_OSTSIZE
6187 # mount the Lustre filesystem
6188 setup_noconfig || error "(2) setup Lustre filesystem failed"
6191 log "create test files"
6196 mkdir $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
6198 for i in $(seq $num_files); do
6199 file=$MOUNT/$tdir/$tfile-$i
6200 dd if=/dev/urandom of=$file count=1 bs=1M || {
6201 $LCTL get_param osc.*.cur*grant*
6202 $LFS df; $LFS df -i;
6203 # stop creating files if there is no more space
6204 if [ ! -e $file ]; then
6205 num_files=$((i - 1))
6209 $LFS getstripe -v $file
6210 local ost_idx=$(LFS getstripe -i $file)
6211 do_facet ost$((ost_idx + 1)) \
6212 $LCTL get_param obdfilter.*.*grant*
6213 error "(4) create $file failed"
6217 # unmount the Lustre filesystem
6218 cleanup || error "(5) cleanup Lustre filesystem failed"
6220 # run e2fsck on the MDT and OST devices
6221 local mds_host=$(facet_active_host $SINGLEMDS)
6222 local ost_host=$(facet_active_host ost1)
6223 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
6224 local ost_dev=$(ostdevname 1)
6226 run_e2fsck $mds_host $mds_dev "-y"
6227 run_e2fsck $ost_host $ost_dev "-y"
6229 # get the original block count of the MDT and OST filesystems
6230 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
6231 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
6233 # expand the MDT and OST filesystems to the device size
6234 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
6235 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
6237 # run e2fsck on the MDT and OST devices again
6238 run_e2fsck $mds_host $mds_dev "-y"
6239 run_e2fsck $ost_host $ost_dev "-y"
6241 # mount the Lustre filesystem
6245 log "check files after expanding the MDT and OST filesystems"
6246 for i in $(seq $num_files); do
6247 file=$MOUNT/$tdir/$tfile-$i
6248 $CHECKSTAT -t file -s 1048576 $file ||
6249 error "(6) checkstat $file failed"
6253 log "create more files after expanding the MDT and OST filesystems"
6254 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
6255 file=$MOUNT/$tdir/$tfile-$i
6256 dd if=/dev/urandom of=$file count=1 bs=1M ||
6257 error "(7) create $file failed"
6260 # unmount the Lustre filesystem
6261 cleanup || error "(8) cleanup Lustre filesystem failed"
6263 # run e2fsck on the MDT and OST devices
6264 run_e2fsck $mds_host $mds_dev "-y"
6265 run_e2fsck $ost_host $ost_dev "-y"
6267 # get the maximum block count of the MDT and OST filesystems
6268 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
6269 local ost_max_blks=$(get_block_count ost1 $ost_dev)
6271 # get the minimum block count of the MDT and OST filesystems
6272 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
6273 grep minimum | sed -e 's/^.*filesystem: //g')
6274 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
6275 grep minimum | sed -e 's/^.*filesystem: //g')
6277 # shrink the MDT and OST filesystems to a smaller size
6281 if [[ $mds_max_blks -gt $mds_min_blks &&
6282 $mds_max_blks -gt $mds_orig_blks ]]; then
6283 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
6284 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
6285 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
6286 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
6287 error "shrink $SINGLEMDS to $new_blks failed"
6291 if [[ $ost_max_blks -gt $ost_min_blks &&
6292 $ost_max_blks -gt $ost_orig_blks ]]; then
6293 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
6294 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
6295 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
6296 run_resize2fs ost1 $ost_dev $new_blks ||
6297 error "shrink ost1 to $new_blks failed"
6301 # check whether the MDT or OST filesystem was shrunk or not
6303 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
6304 reformat_and_config ||
6305 error "(10) reformat Lustre filesystem failed"
6309 # run e2fsck on the MDT and OST devices again
6310 run_e2fsck $mds_host $mds_dev "-y"
6311 run_e2fsck $ost_host $ost_dev "-y"
6313 # mount the Lustre filesystem again
6317 log "check files after shrinking the MDT and OST filesystems"
6318 for i in $(seq $((num_files + 10))); do
6319 file=$MOUNT/$tdir/$tfile-$i
6320 $CHECKSTAT -t file -s 1048576 $file ||
6321 error "(11) checkstat $file failed"
6324 # unmount and reformat the Lustre filesystem
6325 cleanup || error "(12) cleanup Lustre filesystem failed"
6326 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
6328 MDSCOUNT=$saved_MDSCOUNT
6329 OSTCOUNT=$saved_OSTCOUNT
6330 reformat_and_config || error "(14) reformat Lustre filesystem failed"
6332 run_test 78 "run resize2fs on MDT and OST filesystems"
6334 test_79() { # LU-4227
6335 [[ "$MDS1_VERSION" -ge $(version_code 2.5.59) ]] ||
6336 skip "Need MDS version at least 2.5.59"
6338 local mdsdev1=$(mdsdevname 1)
6339 local mdsvdev1=$(mdsvdevname 1)
6340 local mdsdev2=$(mdsdevname 2)
6341 local mdsvdev2=$(mdsvdevname 2)
6342 local ostdev1=$(ostdevname 1)
6343 local ostvdev1=$(ostvdevname 1)
6344 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
6345 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
6346 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
6349 # remove --mgs/--mgsnode from mkfs.lustre options
6350 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
6352 mgsnode_opt=$(echo $opts_mds2 |
6353 awk '{ for ( i = 1; i < NF; i++ )
6354 if ( $i ~ "--mgsnode" ) { print $i; break } }')
6355 [ -n "$mgsnode_opt" ] &&
6356 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
6358 mgsnode_opt=$(echo $opts_ost1 |
6359 awk '{ for ( i = 1; i < NF; i++ )
6360 if ( $i ~ "--mgsnode" ) { print $i; break } }')
6361 [ -n "$mgsnode_opt" ] &&
6362 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
6364 # -MGS, format a mdt without --mgs option
6365 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
6366 error "Must specify --mgs when formatting mdt combined with mgs"
6368 # +MGS, format a mdt/ost without --mgsnode option
6369 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
6370 > /dev/null || error "start mds1 failed"
6371 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
6372 error "Must specify --mgsnode when formatting a mdt"
6373 add ost1 $opts_ost1 $ostdev1 $ostvdev2 &&
6374 error "Must specify --mgsnode when formatting an ost"
6378 run_test 79 "format MDT/OST without mgs option (should return errors)"
6381 start_mds || error "Failed to start MDT"
6382 start_ost || error "Failed to start OST1"
6383 uuid=$(do_facet ost1 $LCTL get_param -n mgc.*.uuid)
6384 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
6385 do_facet ost1 "$LCTL set_param fail_val=10 fail_loc=0x906"
6386 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x906"
6387 do_facet mgs "$LCTL set_param -n mgs/MGS/evict_client $uuid"
6389 start_ost2 || error "Failed to start OST2"
6391 do_facet ost1 "$LCTL set_param fail_loc=0"
6396 run_test 80 "mgc import reconnect race"
6398 #Save the original values of $OSTCOUNT and $OSTINDEX$i.
6400 local new_ostcount=$1
6401 saved_ostcount=$OSTCOUNT
6402 OSTCOUNT=$new_ostcount
6406 for i in $(seq $OSTCOUNT); do
6408 eval saved_ostindex$i=${!index}
6413 # Restore the original values of $OSTCOUNT and $OSTINDEX$i.
6414 restore_ostindex() {
6419 for i in $(seq $OSTCOUNT); do
6420 index=saved_ostindex$i
6421 eval OSTINDEX$i=${!index}
6423 OSTCOUNT=$saved_ostcount
6426 if ! combined_mgs_mds ; then
6431 # The main purpose of this test is to ensure the OST_INDEX_LIST functions as
6432 # expected. This test uses OST_INDEX_LIST to format OSTs with a randomly
6433 # assigned index and ensures we can mount such a formatted file system
6434 test_81() { # LU-4665
6435 [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
6436 skip "Need MDS version at least 2.6.54"
6437 [[ $OSTCOUNT -ge 3 ]] || skip_env "needs >= 3 OSTs"
6441 # Each time RANDOM is referenced, a random integer between 0 and 32767
6444 local saved_ostindex1=$OSTINDEX1
6445 for i in 65535 $((RANDOM + 65536)); do
6446 echo -e "\nFormat ost1 with --index=$i, should fail"
6448 if add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat \
6449 $(ostdevname 1) $(ostvdevname 1); then
6450 OSTINDEX1=$saved_ostindex1
6451 error "format ost1 with --index=$i should fail"
6454 OSTINDEX1=$saved_ostindex1
6458 # Format OSTs with random sparse indices.
6459 trap "restore_ostindex" EXIT
6460 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices"
6461 OST_INDEX_LIST=[0,$((RANDOM * 2 % 65533 + 1)),65534] formatall
6463 # Setup and check Lustre filesystem.
6464 start_mgsmds || error "start_mgsmds failed"
6465 for i in $(seq $OSTCOUNT); do
6466 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6467 error "start ost$i failed"
6470 mount_client $MOUNT || error "mount client $MOUNT failed"
6471 check_mount || error "check client $MOUNT failed"
6474 local max_easize=$($LCTL get_param -n llite.*.max_easize)
6475 # 65452 is XATTR_SIZE_MAX less ldiskfs ea overhead
6476 if large_xattr_enabled; then
6477 [[ $max_easize -ge 65452 ]] ||
6478 error "max_easize is $max_easize, should be at least 65452 bytes"
6481 # 4012 is 4096 less ldiskfs ea overhead
6482 [[ $max_easize -ge 4012 ]] ||
6483 error "max_easize is $max_easize, should be at least 4012 bytes"
6488 run_test 81 "sparse OST indexing"
6490 # Here we exercise the stripe placement functionality on a file system that
6491 # has formatted the OST with a random index. With the file system the following
6492 # functionality is tested:
6494 # 1. Creating a new file with a specific stripe layout.
6496 # 2. Modifiy a existing empty file with a specific stripe layout.
6498 # 3. Ensure we fail to set the stripe layout of a file that already has one.
6500 # 4. If ost-index is defined we need to ensure it is the first entry in the
6501 # ost index list returned by lfs getstripe.
6503 # 5. Lastly ensure this functionality fails with directories.
6504 test_82a() { # LU-4665
6505 [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
6506 skip "Need MDS version at least 2.6.54"
6507 [[ $OSTCOUNT -ge 3 ]] || skip_env "needs >= 3 OSTs"
6513 # Format OSTs with random sparse indices.
6517 local LOV_V1_INSANE_STRIPE_COUNT=65532
6518 for i in $(seq $OSTCOUNT); do
6519 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
6520 ost_indices+=" $index"
6522 ost_indices=$(comma_list $ost_indices)
6524 stack_trap "restore_ostindex" EXIT
6525 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
6526 OST_INDEX_LIST=[$ost_indices] formatall
6528 # Setup Lustre filesystem.
6529 start_mgsmds || error "start_mgsmds failed"
6530 for i in $(seq $OSTCOUNT); do
6531 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6532 error "start ost$i failed"
6535 # Collect debug information - start of test
6536 do_nodes $(comma_list $(mdts_nodes)) \
6537 $LCTL get_param osp.*.prealloc_*_id
6539 mount_client $MOUNT || error "mount client $MOUNT failed"
6543 check_lfs_df_ret_val $? || error "$LFS df $MOUNT failed"
6544 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6546 stack_trap "do_nodes $(comma_list $(mdts_nodes)) \
6547 $LCTL get_param osp.*.prealloc_*_id" EXIT
6549 # 1. If the file does not exist, new file will be created
6550 # with specified OSTs.
6551 local file=$DIR/$tdir/$tfile-1
6552 local cmd="$LFS setstripe -o $ost_indices $file"
6554 eval $cmd || error "$cmd failed"
6555 check_stripe_count $file $OSTCOUNT
6556 check_obdidx $file $ost_indices
6557 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6558 error "write $file failed"
6560 # 2. If the file already exists and is an empty file, the file
6561 # will be attached with specified layout.
6562 file=$DIR/$tdir/$tfile-2
6563 mcreate $file || error "mcreate $file failed"
6564 cmd="$LFS setstripe -o $ost_indices $file"
6566 eval $cmd || error "$cmd failed"
6567 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6568 error "write $file failed"
6569 check_stripe_count $file $OSTCOUNT
6570 check_obdidx $file $ost_indices
6572 # 3. If the file already has a valid layout attached, the command
6573 # should fail with EBUSY.
6575 eval $cmd && error "stripe is already set on $file, $cmd should fail"
6577 # 4. If [--stripe-index|-i <start_ost_idx>] is used, the index must
6578 # be in the OST indices list.
6579 local start_ost_idx=${ost_indices##*,}
6580 file=$DIR/$tdir/$tfile-3
6581 cmd="$LFS setstripe -o $ost_indices -i $start_ost_idx $file"
6583 eval $cmd || error "$cmd failed"
6584 check_stripe_count $file $OSTCOUNT
6585 check_obdidx $file $ost_indices
6586 check_start_ost_idx $file $start_ost_idx
6588 file=$DIR/$tdir/$tfile-4
6589 cmd="$LFS setstripe"
6590 cmd+=" -o $(exclude_items_from_list $ost_indices $start_ost_idx)"
6591 cmd+=" -i $start_ost_idx $file"
6593 eval $cmd && error "index $start_ost_idx should be in $ost_indices"
6595 # 5. Specifying OST indices for directory should succeed.
6596 local dir=$DIR/$tdir/$tdir
6597 mkdir $dir || error "mkdir $dir failed"
6598 cmd="$LFS setstripe -o $ost_indices $dir"
6599 if [ "$MDS1_VERSION" -gt $(version_code 2.11.53) ] &&
6600 [ "$CLIENT_VERSION" -gt $(version_code 2.11.53) ]; then
6602 eval $cmd || error "unable to specify OST indices on directory"
6604 echo "need MDS+client version at least 2.11.53"
6609 run_test 82a "specify OSTs for file (succeed) or directory (succeed)"
6614 # Remove OSTs from a pool and destroy the pool.
6615 destroy_pool $ost_pool || true
6620 # Test 82b is run to ensure that if the user supplies a pool with a specific
6621 # stripe layout that it behaves proprerly. It should fail in the case that
6622 # the supplied OST index list points to OSTs not contained in the user
6624 test_82b() { # LU-4665
6625 [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
6626 skip "Need MDS version at least 2.6.54"
6627 [[ $OSTCOUNT -ge 4 ]] || skip_env "needs >= 4 OSTs"
6633 # Format OSTs with random sparse indices.
6637 local LOV_V1_INSANE_STRIPE_COUNT=65532
6638 for i in $(seq $OSTCOUNT); do
6639 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
6640 ost_indices+=" $index"
6642 ost_indices=$(comma_list $ost_indices)
6644 trap "restore_ostindex" EXIT
6645 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
6646 OST_INDEX_LIST=[$ost_indices] formatall
6648 # Setup Lustre filesystem.
6649 start_mgsmds || error "start_mgsmds failed"
6650 for i in $(seq $OSTCOUNT); do
6651 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6652 error "start ost$i failed"
6655 mount_client $MOUNT || error "mount client $MOUNT failed"
6659 check_lfs_df_ret_val $? || error "$LFS df $MOUNT failed"
6660 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6662 # Create a new pool and add OSTs into it.
6663 local ost_pool=$FSNAME.$TESTNAME
6664 create_pool $ost_pool || error "create OST pool $ost_pool failed"
6667 trap "cleanup_82b" EXIT
6669 local ost_idx_in_list=${ost_indices##*,}
6670 local ost_idx_in_pool=$(exclude_items_from_list $ost_indices \
6673 local ost_targets="$FSNAME-OST["
6674 for i in ${ost_idx_in_pool//,/ }; do
6675 ost_targets=$ost_targets$(printf "%04x," $i)
6677 ost_targets="${ost_targets%,}]"
6679 local ost_targets_uuid=$(for i in ${ost_idx_in_pool//,/ }; \
6680 do printf "$FSNAME-OST%04x_UUID\n" $i; done |
6681 sort -u | tr '\n' ' ')
6683 local cmd="$LCTL pool_add $ost_pool $ost_targets"
6684 do_facet mgs $cmd || error "$cmd failed"
6685 wait_update $HOSTNAME "$LCTL get_param -n lov.$FSNAME-*.pools.$TESTNAME|
6686 sort -u | tr '\n' ' ' " "$ost_targets_uuid" ||
6687 error "wait_update $ost_pool failed"
6688 wait_update_facet $SINGLEMDS "$LCTL pool_list $ost_pool | wc -l" 4 ||
6689 error "wait_update pool_list $ost_pool failed"
6691 # If [--pool|-p <pool_name>] is set with [--ost-list|-o <ost_indices>],
6692 # then the OSTs must be the members of the pool.
6693 local file=$DIR/$tdir/$tfile
6694 cmd="$LFS setstripe -p $ost_pool -o $ost_idx_in_list $file"
6696 eval $cmd && error "OST with index $ost_idx_in_list should be" \
6697 "in OST pool $ost_pool"
6699 # Only select OST $ost_idx_in_list from $ost_pool for file.
6700 ost_idx_in_list=${ost_idx_in_pool#*,}
6701 cmd="$LFS setstripe -p $ost_pool -o $ost_idx_in_list $file"
6703 eval $cmd || error "$cmd failed"
6704 cmd="$LFS getstripe $file"
6706 eval $cmd || error "$cmd failed"
6707 check_stripe_count $file 2
6708 check_obdidx $file $ost_idx_in_list
6709 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6710 error "write $file failed"
6714 run_test 82b "specify OSTs for file with --pool and --ost-list options"
6717 [[ "$OST1_VERSION" -ge $(version_code 2.6.91) ]] ||
6718 skip "Need OST version at least 2.6.91"
6719 if [ "$ost1_FSTYPE" != ldiskfs ]; then
6720 skip "ldiskfs only test"
6729 ostmnt=$(facet_mntpt ost1)
6731 # Mount the OST as an ldiskfs filesystem.
6732 log "mount the OST $dev as a $ost1_FSTYPE filesystem"
6733 add ost1 $(mkfs_opts ost1 $dev) $FSTYPE_OPT \
6734 --reformat $dev > /dev/null ||
6735 error "format ost1 error"
6737 if ! test -b $dev; then
6738 mnt_opts=$(csa_add "$OST_MOUNT_FS_OPTS" -o loop)
6740 echo "mnt_opts $mnt_opts"
6741 do_facet ost1 mount -t "$ost1_FSTYPE" $dev \
6743 # Run llverfs on the mounted ldiskfs filesystem.
6744 # It is needed to get ENOSPACE.
6745 log "run llverfs in partial mode on the OST $ost1_FSTYPE $ostmnt"
6746 do_rpc_nodes $(facet_host ost1) run_llverfs $ostmnt -vpl \
6747 "no" || error "run_llverfs error on $ost1_FSTYPE"
6750 log "unmount the OST $dev"
6753 # Delete file IO_scrub. Later osd_scrub_setup will try to
6754 # create "IO_scrub" but will get ENOSPACE.
6756 echo "start ost1 service on `facet_active_host ost1`"
6757 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
6760 err=$(do_facet ost1 dmesg | grep "VFS: Busy inodes after unmount of")
6761 echo "string err $err"
6762 [ -z "$err" ] || error $err
6765 run_test 83 "ENOSPACE on OST doesn't cause message VFS: \
6766 Busy inodes after unmount ..."
6769 local facet=$SINGLEMDS
6770 local num=$(echo $facet | tr -d "mds")
6771 local dev=$(mdsdevname $num)
6772 local time_min=$(recovery_time_min)
6773 local recovery_duration
6774 local completed_clients
6775 local correct_clients
6778 echo "start mds service on $(facet_active_host $facet)"
6780 "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
6781 error "start MDS failed"
6783 start_ost || error "start OST0000 failed"
6784 start_ost2 || error "start OST0001 failed"
6786 echo "recovery_time=$time_min, timeout=$TIMEOUT, wrap_up=$wrap_up"
6788 mount_client $MOUNT1 || error "mount $MOUNT1 failed"
6789 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
6790 # make sure new superblock labels are sync'd before disabling writes
6794 replay_barrier $SINGLEMDS
6795 createmany -o $DIR1/$tfile-%d 1000
6797 # We need to catch the end of recovery window to extend it.
6798 # Skip 5 requests and add delay to request handling.
6799 #define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP
6800 do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5"
6802 facet_failover --fsck $SINGLEMDS || error "failover: $?"
6805 echo "recovery status"
6806 do_facet $SINGLEMDS \
6807 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status"
6809 recovery_duration=$(do_facet $SINGLEMDS \
6810 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
6811 awk '/recovery_duration/ { print $2 }')
6812 (( $recovery_duration > $time_min + $wrap_up )) &&
6813 error "recovery_duration > recovery_time_hard + wrap up"
6814 completed_clients=$(do_facet $SINGLEMDS \
6815 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
6816 awk '/completed_clients/ { print $2 }')
6818 correct_clients="$MDSCOUNT/$((MDSCOUNT+1))"
6819 [ "$completed_clients" = "${correct_clients}" ] ||
6820 error "$completed_clients != $correct_clients"
6822 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
6823 umount_client $MOUNT1
6824 umount_client $MOUNT2
6830 run_test 84 "check recovery_hard_time"
6833 [[ "$OST1_VERSION" -ge $(version_code 2.7.55) ]] ||
6834 skip "Need OST version at least 2.7.55"
6835 ##define OBD_FAIL_OSD_OST_EA_FID_SET 0x197
6836 do_facet ost1 "lctl set_param fail_loc=0x197"
6840 run_test 85 "osd_ost init: fail ea_fid_set"
6845 # ost1 has already registered to the MGS before the reformat.
6846 # So after reformatting it with option "-G", it could not be
6847 # mounted to the MGS. Cleanup the system for subsequent tests.
6852 [ "$ost1_FSTYPE" = zfs ] &&
6853 skip "LU-6442: no such mkfs params for ZFS OSTs"
6854 [[ "$MDS1_VERSION" -ge $(version_code 2.7.56) ]] ||
6855 skip "Need server version newer than 2.7.55"
6858 local OLDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6859 awk '/Flex block group size: / { print $NF; exit; }')
6861 [ "$OLDSIZE" == "$NEWSIZE" ] && skip "$NEWSIZE groups already"
6863 local opts=" -O flex_bg -G $NEWSIZE"
6864 opts=$(OST_FS_MKFS_OPTS+="$opts" mkfs_opts ost1 $(ostdevname 1))
6865 opts+=" --reformat $(ostdevname 1) $(ostvdevname 1)"
6866 echo "params: $opts"
6868 trap cleanup_86 EXIT ERR
6871 add ost1 $opts || error "add ost1 failed with new params"
6873 local FOUNDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6874 awk '/Flex block group size: / { print $NF; exit; }')
6876 [[ $FOUNDSIZE == $NEWSIZE ]] ||
6877 error "Flex block group size: $FOUNDSIZE, expected: $NEWSIZE"
6881 run_test 86 "Replacing mkfs.lustre -G option"
6883 test_87() { #LU-6544
6884 [[ "$MDS1_VERSION" -ge $(version_code 2.9.51) ]] ||
6885 skip "Need MDS version at least 2.9.51"
6886 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
6887 skip "ldiskfs only test"
6888 [[ $OSTCOUNT -gt 59 ]] &&
6889 skip "Ignore wide striping situation"
6890 [ -n "$FILESET" ] && skip "Not functional for FILESET set"
6892 local mdsdev=$(mdsdevname 1)
6893 local mdsvdev=$(mdsvdevname 1)
6894 local file=$DIR/$tfile
6895 local mntpt=$(facet_mntpt $SINGLEMDS)
6896 local used_xattr_blk=0
6897 local inode_size=${1:-1024}
6899 local xtest="trusted.test"
6903 local stripe_cnt=$(($OSTCOUNT + 2))
6905 #Please see ldiskfs_make_lustre() for MDT inode size calculation
6906 if [ $stripe_cnt -gt 16 ]; then
6909 left_size=$(expr $inode_size - \
6912 32 - 40 \* 3 - 32 \* 3 - $stripe_cnt \* 24 - 16 - 3 - \
6914 24 - 18 - $(expr length $tfile) - 16 - 4)
6915 if [ $left_size -le 0 ]; then
6916 echo "No space($left_size) is expected in inode."
6917 echo "Try 1-byte xattr instead to verify this."
6920 echo "Estimate: at most $left_size-byte space left in inode."
6926 if ! combined_mgs_mds ; then
6930 add mds1 $(mkfs_opts mds1 ${mdsdev}) --stripe-count-hint=$stripe_cnt \
6931 --reformat $mdsdev $mdsvdev || error "add mds1 failed"
6932 start_mdt 1 > /dev/null || error "start mdt1 failed"
6933 for i in $(seq $OSTCOUNT); do
6934 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS > /dev/null ||
6935 error "start ost$i failed"
6937 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
6938 check_mount || error "check client $MOUNT failed"
6941 $LFS setstripe -E 1M -S 1M -c 1 -E 64M -c 1 -E -1 -c -1 $file ||
6942 error "Create file with 3 components failed"
6943 $TRUNCATE $file $((1024*1024*64+1)) || error "truncate file failed"
6944 i=$($LFS getstripe -I3 -c $file) || error "get 3rd stripe count failed"
6945 if [ $i -ne $OSTCOUNT ]; then
6946 left_size=$(expr $left_size + $(expr $OSTCOUNT - $i) \* 24)
6947 echo -n "Since only $i out $OSTCOUNT OSTs are used, "
6948 echo -n "the expected left space is changed to "
6949 echo "$left_size bytes at most."
6951 value=$(generate_string $left_size)
6952 setfattr -n $xtest -v $value $file
6953 orig=$(get_xattr_value $xtest $file)
6954 [[ "$orig" != "$value" ]] && error "$xtest changed"
6956 #Verify if inode has some expected space left
6957 umount $MOUNT > /dev/null || error "umount $MOUNT failed"
6958 stop_mdt 1 > /dev/null || error "stop mdt1 failed"
6959 mount_ldiskfs $SINGLEMDS || error "mount -t ldiskfs $SINGLEMDS failed"
6961 do_facet $SINGLEMDS ls -sal $mntpt/ROOT/$tfile
6962 used_xattr_blk=$(do_facet $SINGLEMDS ls -s $mntpt/ROOT/$tfile |
6964 [[ $used_xattr_blk -eq 0 ]] &&
6965 error "Please check MDS inode size calculation: \
6966 more than $left_size-byte space left in inode."
6967 echo "Verified: at most $left_size-byte space left in inode."
6969 unmount_ldiskfs $SINGLEMDS
6971 for i in $(seq $OSTCOUNT); do
6972 stop ost$i -f || error "stop ost$i failed"
6975 run_test 87 "check if MDT inode can hold EAs with N stripes properly"
6978 [ "$mds1_FSTYPE" == zfs ] &&
6979 skip "LU-6662: no implementation for ZFS"
6983 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
6984 --reformat $(mdsdevname 1) || error "add mds1 failed"
6986 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
6987 grep -e \".*opts:.*errors=remount-ro.*\"" ||
6988 error "default mount options is missing"
6990 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
6991 --mountfsoptions="user_xattr,errors=panic" \
6992 --reformat $(mdsdevname 1) || error "add mds1 failed"
6994 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
6995 grep -e \".*opts:.*errors=panic.*\"" ||
6996 error "user can't override default mount options"
6998 run_test 88 "check the default mount options can be overridden"
7000 test_89() { # LU-7131
7001 [[ "$MDS1_VERSION" -ge $(version_code 2.9.54) ]] ||
7002 skip "Need MDT version at least 2.9.54"
7004 local key=failover.node
7005 local val1=192.0.2.254@tcp0 # Reserved IPs, see RFC 5735
7006 local val2=192.0.2.255@tcp0
7007 local mdsdev=$(mdsdevname 1)
7012 if [[ "$mds1_FSTYPE" == zfs ]]; then
7013 import_zpool mds1 || return ${PIPESTATUS[0]}
7016 # Check that parameters are added correctly
7017 echo "tunefs --param $key=$val1"
7018 do_facet mds1 "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
7019 error "tunefs --param $key=$val1 failed"
7020 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7021 error "tunefs --dryrun failed"
7022 params=${params##*Parameters:}
7023 params=${params%%exiting*}
7024 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val1) = "1" ] ||
7025 error "on-disk parameter not added correctly via tunefs"
7027 # Check that parameters replace existing instances when added
7028 echo "tunefs --param $key=$val2"
7029 do_facet mds1 "$TUNEFS --param $key=$val2 $mdsdev >/dev/null" ||
7030 error "tunefs --param $key=$val2 failed"
7031 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7032 error "tunefs --dryrun failed"
7033 params=${params##*Parameters:}
7034 params=${params%%exiting*}
7035 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "1" ] ||
7036 error "on-disk parameter not replaced via tunefs"
7037 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val2) = "1" ] ||
7038 error "on-disk parameter not replaced correctly via tunefs"
7040 # Check that a parameter is erased properly
7041 echo "tunefs --erase-param $key"
7042 do_facet mds1 "$TUNEFS --erase-param $key $mdsdev >/dev/null" ||
7043 error "tunefs --erase-param $key failed"
7044 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7045 error "tunefs --dryrun failed"
7046 params=${params##*Parameters:}
7047 params=${params%%exiting*}
7048 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "0" ] ||
7049 error "on-disk parameter not erased correctly via tunefs"
7051 # Check that all the parameters are erased
7052 do_facet mds1 "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
7053 error "tunefs --param $key=$val1 failed"
7054 echo "tunefs --erase-params"
7055 do_facet mds1 "$TUNEFS --erase-params $mdsdev >/dev/null" ||
7056 error "tunefs --erase-params failed"
7057 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7058 error "tunefs --dryrun failed"
7059 params=${params##*Parameters:}
7060 params=${params%%exiting*}
7061 params=$(echo $params | tr ' ' '\n')
7063 error "all on-disk parameters not erased correctly via tunefs $params"
7065 # Check the order of options --erase-params and --param
7066 echo "tunefs --param $key=$val1 --erase-params"
7068 "$TUNEFS --param $key=$val1 --erase-params $mdsdev >/dev/null"||
7069 error "tunefs --param $key=$val1 --erase-params failed"
7070 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7071 error "tunefs --dryrun failed"
7072 params=${params##*Parameters:}
7073 params=${params%%exiting*}
7074 [ $(echo $params | tr ' ' '\n') == "$key=$val1" ] ||
7075 error "on-disk param not added correctly with --erase-params"
7079 run_test 89 "check tunefs --param and --erase-param{s} options"
7082 # $2 (optional) value of max_mod_rpcs_in_flight to set
7083 check_max_mod_rpcs_in_flight() {
7091 idx=$(printf "%04x" $($LFS getdirstripe -i $dir))
7092 facet="mds$((0x$idx + 1))"
7094 if [ -z "$mmr" ]; then
7095 # get value of max_mod_rcps_in_flight
7096 mmr=$($LCTL get_param -n \
7097 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) ||
7098 error "Unable to get max_mod_rpcs_in_flight"
7099 echo "max_mod_rcps_in_flight is $mmr"
7101 # set value of max_mod_rpcs_in_flight
7103 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight=$mmr ||
7104 error "Unable to set max_mod_rpcs_in_flight to $mmr"
7105 echo "max_mod_rpcs_in_flight set to $mmr"
7108 # create mmr+1 files
7109 echo "creating $((mmr + 1)) files ..."
7111 for i in $(seq $((mmr + 1))); do
7117 # consumes mmr-1 modify RPC slots
7118 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7119 # drop requests on MDT so that RPC slots are consumed
7120 # during all the request resend interval
7121 do_facet $facet "$LCTL set_param fail_loc=0x159"
7122 echo "launch $((mmr - 1)) chmod in parallel ..."
7123 for i in $(seq $((mmr - 1))); do
7124 chmod 0600 $dir/file-$i &
7128 # send one additional modify RPC
7129 do_facet $facet "$LCTL set_param fail_loc=0"
7130 echo "launch 1 additional chmod in parallel ..."
7131 chmod 0600 $dir/file-$mmr &
7134 # check this additional modify RPC get a modify RPC slot
7135 # and succeed its operation
7136 checkstat -vp 0600 $dir/file-$mmr ||
7137 error "Unable to send $mmr modify RPCs in parallel"
7142 # consumes mmr modify RPC slots
7143 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7144 # drop requests on MDT so that RPC slots are consumed
7145 # during all the request resend interval
7146 do_facet $facet "$LCTL set_param fail_loc=0x159"
7147 echo "launch $mmr chmod in parallel ..."
7148 for i in $(seq $mmr); do
7149 chmod 0666 $dir/file-$i &
7153 # send one additional modify RPC
7154 do_facet $facet "$LCTL set_param fail_loc=0"
7155 echo "launch 1 additional chmod in parallel ..."
7156 chmod 0666 $dir/file-$((mmr + 1)) &
7159 # check this additional modify RPC blocked getting a modify RPC slot
7160 checkstat -vp 0644 $dir/file-$((mmr + 1)) ||
7161 error "Unexpectedly send $(($mmr + 1)) modify RPCs in parallel"
7168 [[ $($LCTL get_param mdc.*.import |
7169 grep "connect_flags:.*multi_mod_rpc") ]] ||
7170 skip "Need MDC with 'multi_mod_rpcs' feature"
7172 # check default value
7173 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
7174 check_max_mod_rpcs_in_flight $DIR/$tdir
7178 run_test 90a "check max_mod_rpcs_in_flight is enforced"
7188 [[ $($LCTL get_param mdc.*.import |
7189 grep "connect_flags:.*multi_mod_rpc") ]] ||
7190 skip "Need MDC with 'multi_mod_rpcs' feature"
7193 # update max_mod_rpcs_in_flight
7194 $LFS mkdir -c1 $DIR/${tdir}1 || error "mkdir $DIR/${tdir}1 failed"
7195 check_max_mod_rpcs_in_flight $DIR/${tdir}1 1
7198 # check client is able to send multiple modify RPCs in paralell
7199 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
7200 grep -c "multi_mod_rpcs")
7201 if [ "$tmp" -ne $MDSCOUNT ]; then
7203 skip "Client not able to send multiple modify RPCs in parallel"
7206 # update max_mod_rpcs_in_flight
7207 $LFS mkdir -c1 $DIR/${tdir}2 || error "mkdir $DIR/${tdir}2 failed"
7208 check_max_mod_rpcs_in_flight $DIR/${tdir}2 5
7211 $LFS mkdir -c1 $DIR/${tdir}3 || error "mkdir $DIR/${tdir}3 failed"
7212 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/${tdir}3))
7213 facet="mds$((0x$idx + 1))"
7215 # save MDT max_mod_rpcs_per_client
7216 mmrpc=$(do_facet $facet \
7217 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
7219 # update max_mod_rpcs_in_flight
7220 umount_client $MOUNT
7222 "echo 16 > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
7224 $LCTL set_param mdc.$FSNAME-MDT$idx-mdc-*.max_rpcs_in_flight=17
7225 check_max_mod_rpcs_in_flight $DIR/${tdir}3 16
7227 # restore MDT max_mod_rpcs_per_client initial value
7229 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
7231 rm -rf $DIR/${tdir}?
7234 run_test 90b "check max_mod_rpcs_in_flight is enforced after update"
7237 # get max_rpcs_in_flight value
7238 mrif_90c=$($LCTL get_param -n \
7239 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
7240 echo "max_rpcs_in_flight is $mrif_90c"
7242 # get max_mod_rpcs_in_flight value
7243 mmrif_90c=$($LCTL get_param -n \
7244 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight)
7245 echo "max_mod_rpcs_in_flight is $mmrif_90c"
7247 # get MDT max_mod_rpcs_per_client value
7248 mmrpc_90c=$(do_facet mds1 \
7249 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
7250 echo "max_mod_rpcs_per_client is $mmrpc_90c"
7253 restore_params_90c() {
7256 # restore max_rpcs_in_flight value
7257 do_facet mgs $LCTL set_param -P \
7258 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$mrif_90c
7260 # restore max_mod_rpcs_in_flight value
7261 do_facet mgs $LCTL set_param -P \
7262 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mmrif_90c
7264 # restore MDT max_mod_rpcs_per_client value
7265 do_facet mds1 "echo $mmrpc_90c > \
7266 /sys/module/mdt/parameters/max_mod_rpcs_per_client"
7274 [[ $($LCTL get_param mdc.*.import |
7275 grep "connect_flags:.*multi_mod_rpc") ]] ||
7276 skip "Need MDC with 'multi_mod_rpcs' feature"
7278 # check client is able to send multiple modify RPCs in paralell
7279 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
7280 grep -c "multi_mod_rpcs")
7281 if [ "$tmp" -ne $MDSCOUNT ]; then
7283 skip "Client not able to send multiple modify RPCs in parallel"
7287 stack_trap restore_params_90c
7290 # attempt to set max_mod_rpcs_in_flight to max_rpcs_in_flight value
7291 # prerequisite: set max_mod_rpcs_per_client to max_rpcs_in_flight value
7292 do_facet mds1 "echo $mrif_90c > \
7293 /sys/module/mdt/parameters/max_mod_rpcs_per_client"
7295 # if max_mod_rpcs_in_flight is set to be equal to or larger than
7296 # max_rpcs_in_flight, then max_rpcs_in_flight will be increased
7297 if [[ "$CLIENT_VERSION" -ge $(version_code 2.13.53) ]]; then
7299 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif_90c ||
7300 error "set max_mod_rpcs_in_flight to $mrif_90c failed"
7302 local new_mrif=$($LCTL get_param -n \
7303 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
7304 ((new_mrif == mrif_90c + 1)) ||
7305 error "max_rpcs_in_flight was not increased"
7308 umount_client $MOUNT
7309 do_facet mds1 "echo $mmrpc_90c > \
7310 /sys/module/mdt/parameters/max_mod_rpcs_per_client"
7314 # attempt to set max_mod_rpcs_in_flight to max_mod_rpcs_per_client+1
7315 # prerequisite: set max_rpcs_in_flight to max_mod_rpcs_per_client+2
7317 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$((mmrpc_90c + 2))
7320 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$((mmrpc_90c + 1)) &&
7321 error "set max_mod_rpcs_in_flight to $((mmrpc_90c + 1)) should fail"
7324 # attempt to set max_mod_rpcs_in_flight permanently
7325 do_facet mgs $LCTL set_param -P \
7326 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$mrif_90c
7328 do_facet mgs $LCTL set_param -P \
7329 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif_90c
7331 remount_client $MOUNT
7333 wait_update_facet --verbose client "$LCTL get_param -n \
7334 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight" \
7335 "$((mrif_90c + 1))" ||
7336 error "expected '$((mrif_90c + 1))' for max_rpcs_in_flight"
7338 wait_update_facet --verbose client "$LCTL get_param -n \
7339 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight" \
7341 error "expected '$mrif_90c' for max_mod_rpcs_in_flight"
7346 run_test 90c "check max_mod_rpcs_in_flight update limits"
7357 [[ $($LCTL get_param mdc.*.import |
7358 grep "connect_flags:.*multi_mod_rpc") ]] ||
7359 skip "Need MDC with 'multi_mod_rpcs' feature"
7361 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
7362 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
7363 facet="mds$((0x$idx + 1))"
7365 # check client version supports multislots
7366 tmp=$($LCTL get_param -N \
7367 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
7368 if [ -z "$tmp" ]; then
7370 skip "Client does not support multiple modify RPCs in flight"
7373 # get current value of max_mod_rcps_in_flight
7374 mmr=$($LCTL get_param -n \
7375 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
7376 echo "max_mod_rcps_in_flight is $mmr"
7379 echo "creating $mmr files ..."
7381 for i in $(seq $mmr); do
7382 touch $DIR/$tdir/file-$i
7385 # prepare for close RPC
7386 multiop_bg_pause $DIR/$tdir/file-close O_c
7389 # consumes mmr modify RPC slots
7390 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7391 # drop requests on MDT so that RPC slots are consumed
7392 # during all the request resend interval
7393 do_facet $facet "$LCTL set_param fail_loc=0x159"
7394 echo "launch $mmr chmod in parallel ..."
7395 for i in $(seq $mmr); do
7396 chmod 0600 $DIR/$tdir/file-$i &
7399 # send one additional close RPC
7400 do_facet $facet "$LCTL set_param fail_loc=0"
7401 echo "launch 1 additional close in parallel ..."
7403 cancel_lru_locks mdc
7406 # check this additional close RPC get a modify RPC slot
7407 # and multiop process completed
7408 [ -d /proc/$pid ] &&
7409 error "Unable to send the additional close RPC in parallel"
7414 run_test 90d "check one close RPC is allowed above max_mod_rpcs_in_flight"
7416 check_uuid_on_ost() {
7418 do_facet ost1 "$LCTL get_param obdfilter.${FSNAME}*.exports.'$nid'.uuid"
7421 check_uuid_on_mdt() {
7423 do_facet $SINGLEMDS "$LCTL get_param mdt.${mds1_svc}*.exports.'$nid'.uuid"
7431 [[ "$OST1_VERSION" -ge $(version_code 2.7.63) ]] ||
7432 skip "Need OST version at least 2.7.63"
7433 [[ "$MDS1_VERSION" -ge $(version_code 2.7.63) ]] ||
7434 skip "Need MDT version at least 2.7.63"
7436 start_mds || error "MDS start failed"
7437 start_ost || error "unable to start OST"
7438 mount_client $MOUNT || error "client start failed"
7439 check_mount || error "check_mount failed"
7442 nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g")
7446 uuid=$(get_client_uuid $MOUNT)
7448 echo "list nids on mdt:"
7449 do_facet $SINGLEMDS "$LCTL list_param mdt.${FSNAME}*.exports.*"
7450 echo "uuid from $nid:"
7451 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
7453 found=$(check_uuid_on_mdt $nid | grep $uuid)
7454 [ -z "$found" ] && error "can't find $uuid $nid on MDT"
7455 found=$(check_uuid_on_ost $nid | grep $uuid)
7456 [ -z "$found" ] && error "can't find $uuid $nid on OST"
7458 # umount the client so it won't reconnect
7459 manual_umount_client --force || error "failed to umount $?"
7460 # shouldn't disappear on MDS after forced umount
7461 found=$(check_uuid_on_mdt $nid | grep $uuid)
7462 [ -z "$found" ] && error "can't find $uuid $nid"
7465 do_facet $SINGLEMDS \
7466 "$LCTL set_param -n mdt.${mds1_svc}.evict_client nid:$nid"
7467 sleep 1 # eviction above is async, give it some time to proceed
7469 found=$(check_uuid_on_mdt $nid | grep $uuid)
7470 [ -n "$found" ] && error "found $uuid $nid on MDT"
7471 found=$(check_uuid_on_ost $nid | grep $uuid)
7472 [ -n "$found" ] && error "found $uuid $nid on OST"
7474 # check it didn't reconnect (being umounted)
7475 sleep $((TIMEOUT+1))
7476 found=$(check_uuid_on_mdt $nid | grep $uuid)
7477 [ -n "$found" ] && error "found $uuid $nid on MDT"
7478 found=$(check_uuid_on_ost $nid | grep $uuid)
7479 [ -n "$found" ] && error "found $uuid $nid on OST"
7483 run_test 91 "evict-by-nid support"
7485 generate_ldev_conf() {
7486 # generate an ldev.conf file
7487 local ldevconfpath=$1
7489 local fsldevformat=""
7492 local facets="mgs,$(get_facets OST),$(get_facets MDS)"
7493 for facet in ${facets//,/ }; do
7495 fstype=$(facet_fstype $facet)
7496 [ "$fstype" = zfs ] && fsldevformat="$fstype:"
7497 local host=$(facet_host $facet)
7499 local varfo=${facet}failover_HOST
7500 if [ $facet == mgs ] && combined_mgs_mds; then
7501 varfo=mds1failover_HOST
7503 [ -n "${!varfo}" ] && fo=${!varfo}
7504 local type=$(echo $facet | tr -d "[:digit:]" | \
7505 tr "[:lower:]" "[:upper:]" | sed s/MDS/MDT/ )
7507 [ ${facet} == mgs ] ||
7508 num=$(facet_number $facet)
7509 printf "%s\t%s\t%s-%s%04x\t%s%s\n" \
7516 $(facet_device $facet) >> $ldevconfpath
7519 echo "----- $ldevconfpath -----"
7521 echo "--- END $ldevconfpath ---"
7526 # generate a nids file (mapping between hostname to nid)
7527 # looks like we only have the MGS nid available to us
7528 # so just echo that to a file
7530 echo -e "${mgs_HOST}\t${MGSNID}" > $nidspath
7532 echo "----- $nidspath -----"
7534 echo "--- END $nidspath ---"
7537 compare_ldev_output() {
7538 local ldev_output=$1
7539 local expected_output=$2
7541 sort $expected_output -o $expected_output
7542 sort $ldev_output -o $ldev_output
7544 echo "-- START OF LDEV OUTPUT --"
7546 echo "--- END OF LDEV OUTPUT ---"
7548 echo "-- START OF EXPECTED OUTPUT --"
7549 cat $expected_output
7550 echo "--- END OF EXPECTED OUTPUT ---"
7552 diff $expected_output $ldev_output
7557 if [ -z "$LDEV" ]; then
7558 error "ldev is missing!"
7561 local LDEVCONFPATH=$TMP/ldev.conf
7562 local NIDSPATH=$TMP/nids
7564 echo "Host is $(hostname)"
7566 generate_ldev_conf $LDEVCONFPATH
7567 generate_nids $NIDSPATH
7569 # echo the mgs nid and compare it to environment variable MGSNID
7570 local facets="$(get_facets OST),$(get_facets MDS),mgs"
7571 for facet in ${facets//,/ }; do
7572 local host=$(facet_host $facet)
7573 local output=$($LDEV -c $LDEVCONFPATH -H $host -n $NIDSPATH echo %m)
7575 echo "-- START OF LDEV OUTPUT --"
7577 echo "--- END OF LDEV OUTPUT ---"
7580 error "ldev failed to execute!"
7582 # need to process multiple lines because of
7583 # several targets on host
7584 echo -e $output | awk '{ print $2 }' | while read -r line ; do
7585 [ "$line" = "$MGSNID" ] ||
7586 error "ldev failed mgs nid '$line', \
7590 rm -f $LDEVCONFPATH $NIDSPATH
7592 run_test 92 "ldev returns MGS NID correctly in command substitution"
7595 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs"
7598 #start mgs or mgs/mdt0
7599 if ! combined_mgs_mds ; then
7606 start_ost || error "OST0 start fail"
7608 #define OBD_FAIL_MGS_WRITE_TARGET_DELAY 0x90e
7609 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x8000090e"
7610 for num in $(seq 2 $MDSCOUNT); do
7614 mount_client $MOUNT || error "mount client fails"
7615 wait_osc_import_state mds ost FULL
7616 wait_osc_import_ready client ost
7617 check_mount || error "check_mount failed"
7619 cleanup || error "cleanup failed with $?"
7621 run_test 93 "register mulitple MDT at the same time"
7624 if [ -z "$LDEV" ]; then
7625 error "ldev is missing!"
7628 local LDEVCONFPATH=$TMP/ldev.conf
7629 local NIDSPATH=$TMP/nids
7631 generate_ldev_conf $LDEVCONFPATH
7632 generate_nids $NIDSPATH
7634 local LDEV_OUTPUT=$TMP/ldev-output.txt
7635 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME > $LDEV_OUTPUT
7637 # ldev failed, error
7638 if [ $? -ne 0 ]; then
7639 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7640 error "ldev failed to execute!"
7644 local EXPECTED_OUTPUT=$TMP/ldev-expected.txt
7646 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
7648 for num in $(seq $MDSCOUNT); do
7649 printf "%s-MDT%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7652 for num in $(seq $OSTCOUNT); do
7653 printf "%s-OST%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7656 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7658 if [ $? -ne 0 ]; then
7659 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7660 error "ldev failed to produce the correct hostlist!"
7663 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7665 run_test 94 "ldev outputs correct labels for file system name query"
7668 if [ -z "$LDEV" ]; then
7669 error "ldev is missing!"
7672 local LDEVCONFPATH=$TMP/ldev.conf
7673 local NIDSPATH=$TMP/nids
7675 generate_ldev_conf $LDEVCONFPATH
7676 generate_nids $NIDSPATH
7680 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME &>/dev/null
7681 if [ $? -ne 0 ]; then
7682 rm $LDEVCONFPATH $NIDSPATH
7683 error "ldev label filtering w/ -F failed!"
7687 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -l &>/dev/null
7688 if [ $? -ne 0 ]; then
7689 rm $LDEVCONFPATH $NIDSPATH
7690 error "ldev label filtering w/ -l failed!"
7694 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f &>/dev/null
7695 if [ $? -ne 0 ]; then
7696 rm $LDEVCONFPATH $NIDSPATH
7697 error "ldev label filtering w/ -f failed!"
7701 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a &>/dev/null
7702 if [ $? -ne 0 ]; then
7703 rm $LDEVCONFPATH $NIDSPATH
7704 error "ldev label filtering w/ -a failed!"
7709 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -F $FSNAME &>/dev/null
7710 if [ $? -eq 0 ]; then
7711 rm $LDEVCONFPATH $NIDSPATH
7712 error "ldev label filtering w/ -a and -F incorrectly succeeded"
7716 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -f &>/dev/null
7717 if [ $? -eq 0 ]; then
7718 rm $LDEVCONFPATH $NIDSPATH
7719 error "ldev label filtering w/ -a and -f incorrectly succeeded"
7723 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -l &>/dev/null
7724 if [ $? -eq 0 ]; then
7725 rm $LDEVCONFPATH $NIDSPATH
7726 error "ldev label filtering w/ -a and -l incorrectly succeeded"
7730 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f -l &>/dev/null
7731 if [ $? -eq 0 ]; then
7732 rm $LDEVCONFPATH $NIDSPATH
7733 error "ldev label filtering w/ -f and -l incorrectly succeeded"
7737 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -l &>/dev/null
7738 if [ $? -eq 0 ]; then
7739 rm $LDEVCONFPATH $NIDSPATH
7740 error "ldev label filtering w/ -F and -l incorrectly succeeded"
7743 # file sys & foreign
7744 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -f &>/dev/null
7745 if [ $? -eq 0 ]; then
7746 rm $LDEVCONFPATH $NIDSPATH
7747 error "ldev label filtering w/ -F and -f incorrectly succeeded"
7750 rm $LDEVCONFPATH $NIDSPATH
7752 run_test 95 "ldev should only allow one label filter"
7755 if [ -z "$LDEV" ]; then
7756 error "ldev is missing!"
7759 local LDEVCONFPATH=$TMP/ldev.conf
7760 local NIDSPATH=$TMP/nids
7762 generate_ldev_conf $LDEVCONFPATH
7763 generate_nids $NIDSPATH
7765 local LDEV_OUTPUT=$TMP/ldev-output.txt
7766 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -H $mgs_HOST \
7768 awk '{print $2}' > $LDEV_OUTPUT
7770 # ldev failed, error
7771 if [ $? -ne 0 ]; then
7772 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7773 error "ldev failed to execute!"
7777 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
7779 echo "$mgs_HOST-$(facet_fstype mgs)" > $EXPECTED_OUTPUT
7781 local facets="$(get_facets OST),$(get_facets MDS)"
7782 for facet in ${facets//,/ }; do
7783 local host=$(facet_host $facet)
7784 [ "$mgs_HOST" == "$host" ] &&
7785 echo "$host-$(facet_fstype $facet)" \
7789 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7791 if [ $? -ne 0 ]; then
7792 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7793 error "ldev failed to produce the correct output!"
7796 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7798 run_test 96 "ldev returns hostname and backend fs correctly in command sub"
7801 if [ -z "$LDEV" ]; then
7802 error "ldev is missing!"
7805 local LDEVCONFPATH=$TMP/ldev.conf
7806 local NIDSPATH=$TMP/nids
7808 generate_ldev_conf $LDEVCONFPATH
7809 generate_nids $NIDSPATH
7811 local LDEV_OUTPUT=$TMP/ldev-output.txt
7812 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
7814 echo -e "\nMDT role"
7815 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mdt > $LDEV_OUTPUT
7817 if [ $? -ne 0 ]; then
7818 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7819 error "ldev failed to execute for mdt role!"
7822 for num in $(seq $MDSCOUNT); do
7823 printf "%s-MDT%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7826 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7828 if [ $? -ne 0 ]; then
7829 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7830 error "ldev failed to produce the correct output for mdt role!"
7833 echo -e "\nOST role"
7834 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R ost > $LDEV_OUTPUT
7836 if [ $? -ne 0 ]; then
7837 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
7838 error "ldev failed to execute for ost role!"
7842 for num in $(seq $OSTCOUNT); do
7843 printf "%s-OST%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7846 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7848 if [ $? -ne 0 ]; then
7849 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7850 error "ldev failed to produce the correct output for ost role!"
7853 echo -e "\nMGS role"
7854 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mgs > $LDEV_OUTPUT
7856 if [ $? -ne 0 ]; then
7857 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
7858 error "ldev failed to execute for mgs role!"
7861 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
7863 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7865 if [ $? -ne 0 ]; then
7866 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7867 error "ldev failed to produce the correct output for mgs role!"
7870 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7872 run_test 97 "ldev returns correct ouput when querying based on role"
7877 local temp=$MDS_MOUNT_OPTS
7880 check_mount || error "mount failed"
7881 mountopt="user_xattr"
7882 for ((x = 1; x <= 400; x++)); do
7883 mountopt="$mountopt,user_xattr"
7885 mount_client $MOUNT remount,$mountopt 2>&1 | grep "too long" ||
7886 error "Buffer overflow check failed"
7887 cleanup || error "cleanup failed"
7889 run_test 98 "Buffer-overflow check while parsing mount_opts"
7893 [[ "$ost1_FSTYPE" != ldiskfs ]] &&
7894 skip "ldiskfs only test"
7895 [[ "$OST1_VERSION" -ge $(version_code 2.8.57) ]] ||
7896 skip "Need OST version at least 2.8.57"
7898 local ost_opts="$(mkfs_opts ost1 $(ostdevname 1)) \
7899 --reformat $(ostdevname 1) $(ostvdevname 1)"
7900 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" &&
7901 skip "meta_bg already set"
7904 if [[ ${!opts} != *mkfsoptions* ]]; then
7905 eval opts=\"${!opts} \
7906 --mkfsoptions='\\\"-O ^resize_inode,meta_bg\\\"'\"
7908 local val=${!opts//--mkfsoptions=\\\"/ \
7909 --mkfsoptions=\\\"-O ^resize_inode,meta_bg }
7913 echo "params: $opts"
7916 add ost1 $opts || error "add ost1 failed with new params"
7918 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" ||
7919 error "meta_bg is not set"
7921 run_test 99 "Adding meta_bg option"
7926 start_mgsmds || error "MDS start failed"
7927 start_ost || error "unable to start OST"
7928 mount_client $MOUNT || error "client start failed"
7929 check_mount || error "check_mount failed"
7938 do_facet mgs 'lshowmount -v' | awk 'BEGIN {NR == 0; rc=1} /MGS:/ {rc=0}
7939 END {exit rc}' || error "lshowmount have no output MGS"
7941 do_facet mds1 'lshowmount -v' | awk 'BEGIN {NR == 2; rc=1} /-MDT0000:/
7942 {rc=0} END {exit rc}' || error "lshowmount have no output MDT0"
7944 do_facet ost1 'lshowmount -v' | awk 'BEGIN {NR == 4; rc=1} /-OST0000:/
7945 {rc=0} END {exit rc}' || error "lshowmount have no output OST0"
7947 cleanup || error "cleanup failed with $?"
7949 run_test 100 "check lshowmount lists MGS, MDT, OST and 0@lo"
7952 local createmany_pid
7953 local dev=$FSNAME-OST0000-osc-MDT0000
7957 createmany -o $DIR1/$tdir/$tfile-%d 50000 &
7959 # MDT->OST reconnection causes MDT<->OST last_id synchornisation
7960 # via osp_precreate_cleanup_orphans.
7961 for ((i = 0; i < 100; i++)); do
7962 for ((k = 0; k < 10; k++)); do
7963 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;" \
7964 "$LCTL --device $dev activate"
7967 ls -asl $MOUNT | grep '???' &&
7968 { kill -9 $createmany_pid &>/dev/null;
7969 error "File has no object on OST"; }
7971 kill -s 0 $createmany_pid || break
7973 wait $createmany_pid
7975 unlinkmany $DIR1/$tdir/$tfile-%d 50000
7978 run_test 101a "Race MDT->OST reconnection with create"
7981 local dev=$FSNAME-OST0000-osc-MDT0000
7982 local dir=$DIR1/$tdir
7986 $LFS setstripe -c 1 -i 0 $dir
7987 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;"
7988 #define OBD_FAIL_OSP_CON_EVENT_DELAY 0x2107
7989 do_facet mds1 "$LCTL set_param fail_loc=0x80002107 fail_val=20"
7990 do_facet $SINGLEMDS "$LCTL --device $dev activate;"
7995 wait_osc_import_state client ost1 FULL
7996 touch $dir/$tfile || error "Can't create file"
8000 run_test 101b "Race events DISCONNECT and ACTIVE in osp"
8003 [[ "$MDS1_VERSION" -gt $(version_code 2.9.53) ]] ||
8004 skip "Need server version greater than 2.9.53"
8005 [[ “$(mdsdevname 1)” != “$(mgsdevname)” ]] &&
8006 [[ “$(facet_host mds1)” = “$(facet_host mgs)” ]] &&
8007 skip "MGS must be on different node or combined"
8009 cleanup || error "cleanup failed with $?"
8011 local mds1dev=$(mdsdevname 1)
8012 local mds1mnt=$(facet_mntpt mds1)
8013 local mds1opts=$MDS_MOUNT_OPTS
8015 if [ "$mds1_FSTYPE" == ldiskfs ] &&
8016 ! do_facet mds1 test -b $mds1dev; then
8017 mds1opts=$(csa_add "$mds1opts" -o loop)
8019 if [[ "$mds1_FSTYPE" == zfs ]]; then
8020 import_zpool mds1 || return ${PIPESTATUS[0]}
8023 # unload all and only load libcfs to allow fail_loc setting
8024 do_facet mds1 $LUSTRE_RMMOD || error "unable to unload modules"
8025 do_facet mds1 modprobe libcfs || error "libcfs not loaded"
8026 do_facet mds1 lsmod \| grep libcfs || error "libcfs not loaded"
8028 #define OBD_FAIL_OBDCLASS_MODULE_LOAD 0x60a
8029 do_facet mds1 "$LCTL set_param fail_loc=0x8000060a"
8031 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts &&
8032 error "mdt start must fail"
8033 do_facet mds1 lsmod \| grep obdclass && error "obdclass must not load"
8035 do_facet mds1 "$LCTL set_param fail_loc=0x0"
8037 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts ||
8038 error "mdt start must not fail"
8040 cleanup || error "cleanup failed with $?"
8042 run_test 102 "obdclass module cleanup upon error"
8047 echo "rename $FSNAME to $newname"
8049 if ! combined_mgs_mds ; then
8050 local dev=$(mgsdevname)
8053 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
8054 error "(7) Fail to rename MGS"
8055 if [ "$(facet_fstype mgs)" = "zfs" ]; then
8056 reimport_zpool mgs $newname-mgs
8060 for num in $(seq $MDSCOUNT); do
8061 local dev=$(mdsdevname $num)
8063 do_facet mds${num} \
8064 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
8065 error "(8) Fail to rename MDT $num"
8066 if [ "$(facet_fstype mds${num})" = "zfs" ]; then
8067 reimport_zpool mds${num} $newname-mdt${num}
8071 for num in $(seq $OSTCOUNT); do
8072 local dev=$(ostdevname $num)
8074 do_facet ost${num} \
8075 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
8076 error "(9) Fail to rename OST $num"
8077 if [ "$(facet_fstype ost${num})" = "zfs" ]; then
8078 reimport_zpool ost${num} $newname-ost${num}
8083 test_103_set_pool() {
8087 do_facet mgs $LCTL pool_add $FSNAME.$pname ${FSNAME}-$ost_x ||
8088 error "Fail to add $ost_x to $FSNAME.$pname"
8089 wait_update $HOSTNAME \
8090 "lctl get_param -n lov.$FSNAME-clilov-*.pools.$pname |
8091 grep $ost_x" "$FSNAME-${ost_x}_UUID" ||
8092 error "$ost_x is NOT in pool $FSNAME.$pname"
8095 test_103_check_pool() {
8096 local save_fsname=$1
8099 stat $DIR/$tdir/test-framework.sh ||
8100 error "($errno) Fail to stat"
8101 do_facet mgs $LCTL pool_list $FSNAME.pool1 ||
8102 error "($errno) Fail to list $FSNAME.pool1"
8103 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname ||
8104 error "($errno) Fail to list $FSNAME.$save_fsname"
8105 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname |
8106 grep ${FSNAME}-OST0000 ||
8107 error "($errno) List $FSNAME.$save_fsname is invalid"
8109 local pname=$($LFS getstripe --pool $DIR/$tdir/d0)
8110 [ "$pname" = "$save_fsname" ] ||
8111 error "($errno) Unexpected pool name $pname"
8115 check_mount_and_prep
8117 mkdir $DIR/$tdir || error "(1) Fail to mkdir $DIR/$tdir"
8118 cp $LUSTRE/tests/test-framework.sh $DIR/$tdir ||
8119 error "(2) Fail to copy test-framework.sh"
8121 do_facet mgs $LCTL pool_new $FSNAME.pool1 ||
8122 error "(3) Fail to create $FSNAME.pool1"
8123 # name the pool name as the fsname
8124 do_facet mgs $LCTL pool_new $FSNAME.$FSNAME ||
8125 error "(4) Fail to create $FSNAME.$FSNAME"
8127 test_103_set_pool $FSNAME OST0000
8129 $LFS setstripe -p $FSNAME $DIR/$tdir/d0 ||
8130 error "(6) Fail to setstripe on $DIR/$tdir/d0"
8135 test_renamefs mylustre
8137 local save_fsname=$FSNAME
8141 test_103_check_pool $save_fsname 7
8143 if [ $OSTCOUNT -ge 2 ]; then
8144 test_103_set_pool $save_fsname OST0001
8147 $LFS setstripe -p $save_fsname $DIR/$tdir/f0 ||
8148 error "(16) Fail to setstripe on $DIR/$tdir/f0"
8157 test_103_check_pool $save_fsname 17
8161 test_renamefs $save_fsname
8167 run_test 103 "rename filesystem name"
8169 test_104a() { # LU-6952
8170 local mds_mountopts=$MDS_MOUNT_OPTS
8171 local ost_mountopts=$OST_MOUNT_OPTS
8172 local mds_mountfsopts=$MDS_MOUNT_FS_OPTS
8173 local lctl_ver=$(do_facet $SINGLEMDS $LCTL --version |
8176 [[ $(version_code $lctl_ver) -lt $(version_code 2.9.55) ]] &&
8177 skip "this test needs utils above 2.9.55"
8179 # specify "acl" in mount options used by mkfs.lustre
8180 if [ -z "$MDS_MOUNT_FS_OPTS" ]; then
8181 MDS_MOUNT_FS_OPTS="acl,user_xattr"
8184 MDS_MOUNT_FS_OPTS="${MDS_MOUNT_FS_OPTS},acl,user_xattr"
8187 echo "mountfsopt: $MDS_MOUNT_FS_OPTS"
8189 #reformat/remount the MDT to apply the MDT_MOUNT_FS_OPT options
8191 if ! combined_mgs_mds ; then
8195 if [ -z "$MDS_MOUNT_OPTS" ]; then
8196 MDS_MOUNT_OPTS="-o noacl"
8198 MDS_MOUNT_OPTS="${MDS_MOUNT_OPTS},noacl"
8201 for num in $(seq $MDSCOUNT); do
8202 start mds$num $(mdsdevname $num) $MDS_MOUNT_OPTS ||
8203 error "Failed to start MDS"
8206 for num in $(seq $OSTCOUNT); do
8207 start ost$num $(ostdevname $num) $OST_MOUNT_OPTS ||
8208 error "Failed to start OST"
8212 setfacl -m "d:$RUNAS_ID:rwx" $MOUNT &&
8213 error "ACL is applied when FS is mounted with noacl."
8215 MDS_MOUNT_OPTS=$mds_mountopts
8216 OST_MOUNT_OPTS=$ost_mountopts
8217 MDS_MOUNT_FS_OPTS=$mds_mountfsopts
8219 run_test 104a "Make sure user defined options are reflected in mount"
8221 test_104b() { # LU-12859
8222 mount_client $MOUNT3 flock,localflock
8223 stack_trap "umount_client $MOUNT3" EXIT
8224 mount | grep "$MOUNT3 .*,flock" && error "flock is still set"
8225 mount | grep "$MOUNT3 .*,localflock" || error "localflock is not set"
8226 umount_client $MOUNT3
8227 mount_client $MOUNT3 localflock,flock
8228 mount | grep "$MOUNT3 .*,localflock" && error "localflock is still set"
8229 mount | grep "$MOUNT3 .*,flock" || error "flock is not set"
8230 umount_client $MOUNT3
8231 mount_client $MOUNT3 localflock,flock,noflock
8232 flock_is_enabled $MOUNT3 && error "some flock is still enabled" || true
8234 run_test 104b "Mount uses last flock argument"
8236 error_and_umount() {
8247 mount --bind $DIR $TMP/$tdir || error "mount bind mnt pt failed"
8248 rm -f $TMP/$tdir/$tfile
8249 rm -f $TMP/$tdir/${tfile}1
8251 # Files should not be created in ro bind mount point
8252 # remounting from rw to ro
8253 mount -o remount,ro $TMP/$tdir ||
8254 error_and_umount "readonly remount of bind mnt pt failed"
8255 touch $TMP/$tdir/$tfile &&
8256 error_and_umount "touch succeeds on ro bind mnt pt"
8257 [ -e $TMP/$tdir/$tfile ] &&
8258 error_and_umount "file created on ro bind mnt pt"
8260 # Files should be created in rw bind mount point
8261 # remounting from ro to rw
8262 mount -o remount,rw $TMP/$tdir ||
8263 error_and_umount "read-write remount of bind mnt pt failed"
8264 touch $TMP/$tdir/${tfile}1 ||
8265 error_and_umount "touch fails on rw bind mnt pt"
8266 [ -e $TMP/$tdir/${tfile}1 ] ||
8267 error_and_umount "file not created on rw bind mnt pt"
8268 umount $TMP/$tdir || error "umount of bind mnt pt failed"
8270 cleanup || error "cleanup failed with $?"
8272 run_test 105 "check file creation for ro and rw bind mnt pt"
8279 mkdir -p $DIR/$tdir || error "create $tdir failed"
8280 lfs setstripe -c 1 -i 0 $DIR/$tdir
8281 #define OBD_FAIL_CAT_RECORDS 0x1312
8282 do_facet mds1 $LCTL set_param fail_loc=0x1312 fail_val=$repeat
8284 for ((i = 1; i <= $repeat; i++)); do
8286 #one full plain llog
8287 createmany -o $DIR/$tdir/f- 64768
8289 createmany -u $DIR/$tdir/f- 64768
8291 wait_delete_completed $((TIMEOUT * 7))
8292 #ASSERTION osp_sync_thread() ( thread->t_flags != SVC_RUNNING ) failed
8293 #shows that osp code is buggy
8294 do_facet mds1 $LCTL set_param fail_loc=0 fail_val=0
8298 run_test 106 "check osp llog processing when catalog is wrapped"
8301 [[ "$MDS1_VERSION" -ge $(version_code 2.10.50) ]] ||
8302 skip "Need MDS version > 2.10.50"
8305 start_mgsmds || error "start_mgsmds failed"
8306 start_ost || error "unable to start OST"
8308 # add unknown configuration parameter.
8309 if [[ $PERM_CMD == *"set_param -P"* ]]; then
8310 cmd="$PERM_CMD ost.$FSNAME-OST0000*.unknown_param"
8312 cmd="$PERM_CMD $FSNAME-OST0000*.ost.unknown_param"
8314 do_facet mgs "$cmd=50"
8315 cleanup_nocli || error "cleanup_nocli failed with $?"
8318 # unknown param should be ignored while mounting.
8319 start_ost || error "unable to start OST after unknown param set"
8321 cleanup || error "cleanup failed with $?"
8323 run_test 107 "Unknown config param should not fail target mounting"
8328 $rcmd rm -rf $tmp > /dev/null 2>&1
8329 $rcmd mkdir -p $tmp/{mnt,images} || error "failed to mkdir remotely"
8331 for facet in $facets; do
8332 [ "$mds1_FSTYPE" = zfs ] &&
8333 $rcmd $ZPOOL -f export lustre-$facet > /dev/null 2>&1
8334 $rcmd mkdir $tmp/mnt/$facet ||
8335 error "failed to mkdir $tmp/mnt/$facet"
8336 $rcmd dd if=/dev/zero of=$tmp/images/$facet \
8337 seek=199 bs=1M count=1 ||
8338 error "failed to create $tmp/images/$facet"
8347 local facet=${role}$((idx + 1))
8349 [ $# -eq 5 ] && pool=$5
8351 do_facet $SINGLEMDS $MKFS --fsname=lustre --$mgs \
8352 --$role --index=$idx --replace --backfstype=$bkfs \
8353 --device-size=200000 --reformat $pool $tmp/images/$facet ||
8354 error "failed to mkfs for $facet"
8358 echo "mounting client..."
8359 mount -t lustre ${nid}:/lustre $MOUNT ||
8360 error "failed to mount lustre"
8363 ls -l $MOUNT/local_dir || error "failed to list"
8365 echo "check truncate && write"
8366 echo "dummmmmmmmmmmmm" > $MOUNT/remote_dir/fsx.c ||
8367 error "failed to tuncate & write"
8370 touch $MOUNT/foooo ||
8371 error "failed to create"
8373 echo "check read && write && append"
8374 sha1sum $MOUNT/conf-sanity.sh |
8375 awk '{ print $1 }' > $MOUNT/checksum.new ||
8376 error "failed to read(1)"
8377 sha1sum $MOUNT/remote_dir/unlinkmany.c |
8378 awk '{ print $1 }' >> $MOUNT/checksum.new ||
8379 error "failed to read(2)"
8380 sha1sum $MOUNT/striped_dir/lockahead_test.o |
8381 awk '{ print $1 }' >> $MOUNT/checksum.new ||
8382 error "failed to read(3)"
8385 diff $MOUNT/checksum.new $MOUNT/checksum.src ||
8386 error "failed to verify data"
8396 umount -f $MOUNT || error "failed to umount client"
8397 for facet in $facets; do
8398 $rcmd umount -f $tmp/mnt/$facet ||
8399 error "failed to umount $facet"
8400 if [ "$mds1_FSTYPE" = zfs ]; then
8401 $rcmd $ZPOOL export -f lustre-$facet ||
8402 error "failed to export lustre-$facet"
8406 $rcmd rm -rf $tmp || error "failed to rm the dir $tmp"
8410 [ "$CLIENTONLY" ] && skip "Client-only testing"
8411 [ "$mds1_FSTYPE" != zfs ] && skip "zfs only test"
8412 [ "$MDS1_VERSION" -lt $(version_code 2.10.58) ] &&
8413 skip "Need server version at least 2.10.58"
8418 local tmp=$TMP/$tdir
8419 local rcmd="do_facet $SINGLEMDS"
8420 local facets="mdt1 mdt2 ost1 ost2"
8421 local nid=$($rcmd $LCTL list_nids | head -1)
8424 trap t_108_cleanup EXIT ERR
8427 t_108_mkfs mdt 0 zfs mgs lustre-mdt1/mdt1
8428 t_108_mkfs mdt 1 zfs mgsnode=$nid lustre-mdt2/mdt2
8429 t_108_mkfs ost 0 zfs mgsnode=$nid lustre-ost1/ost1
8430 t_108_mkfs ost 1 zfs mgsnode=$nid lustre-ost2/ost2
8432 for facet in $facets; do
8433 $rcmd zfs set mountpoint=$tmp/mnt/$facet canmount=on \
8434 lustre-$facet/$facet ||
8435 error "failed to zfs set for $facet (1)"
8436 $rcmd zfs mount lustre-$facet/$facet ||
8437 error "failed to local mount $facet"
8438 $rcmd tar jxf $LUSTRE/tests/ldiskfs_${facet}_2_11.tar.bz2 \
8439 --xattrs --xattrs-include="trusted.*" \
8440 -C $tmp/mnt/$facet/ > /dev/null 2>&1 ||
8441 error "failed to untar image for $facet"
8442 $rcmd "cd $tmp/mnt/$facet && rm -rf oi.* OI_* lfsck_* LFSCK" ||
8443 error "failed to cleanup for $facet"
8444 $rcmd zfs umount lustre-$facet/$facet ||
8445 error "failed to local umount $facet"
8446 $rcmd zfs set canmount=off lustre-$facet/$facet ||
8447 error "failed to zfs set $facet (2)"
8450 echo "changing server nid..."
8451 $rcmd mount -t lustre -o nosvc lustre-mdt1/mdt1 $tmp/mnt/mdt1
8452 $rcmd lctl replace_nids $FSNAME-MDT0000 $nid
8453 $rcmd lctl replace_nids $FSNAME-MDT0001 $nid
8454 $rcmd lctl replace_nids $FSNAME-OST0000 $nid
8455 $rcmd lctl replace_nids $FSNAME-OST0001 $nid
8456 $rcmd umount $tmp/mnt/mdt1
8458 for facet in $facets; do
8459 echo "mounting $facet from backup..."
8460 $rcmd mount -t lustre -o abort_recov lustre-$facet/$facet \
8461 $tmp/mnt/$facet || error "failed to mount $facet"
8464 # ZFS backend can detect migration and trigger OI scrub automatically
8465 # sleep 3 seconds for scrub done
8471 run_test 108a "migrate from ldiskfs to ZFS"
8474 [ "$CLIENTONLY" ] && skip "Client-only testing"
8475 [ "$mds1_FSTYPE" != ldiskfs ] && skip "ldiskfs only test"
8476 [ "$MDS1_VERSION" -lt $(version_code 2.10.58) ] &&
8477 skip "Need server version at least 2.10.58"
8482 local tmp=$TMP/$tdir
8483 local rcmd="do_facet $SINGLEMDS"
8484 local facets="mdt1 mdt2 ost1 ost2"
8485 local scrub_list="MDT0000 MDT0001 OST0000 OST0001"
8486 local nid=$($rcmd $LCTL list_nids | head -1)
8489 trap t_108_cleanup EXIT ERR
8492 t_108_mkfs mdt 0 ldiskfs mgs
8493 t_108_mkfs mdt 1 ldiskfs mgsnode=$nid
8494 t_108_mkfs ost 0 ldiskfs mgsnode=$nid
8495 t_108_mkfs ost 1 ldiskfs mgsnode=$nid
8497 for facet in $facets; do
8498 $rcmd mount -t ldiskfs -o loop $tmp/images/$facet \
8500 error "failed to local mount $facet"
8502 $rcmd tar jxf $LUSTRE/tests/zfs_${facet}_2_11.tar.bz2 \
8503 --xattrs --xattrs-include="*.*" \
8504 -C $tmp/mnt/$facet/ > /dev/null 2>&1 ||
8505 error "failed to untar image for $facet"
8506 $rcmd "cd $tmp/mnt/$facet && rm -rf oi.* OI_* lfsck_* LFSCK" ||
8507 error "failed to cleanup for $facet"
8508 $rcmd umount $tmp/mnt/$facet ||
8509 error "failed to local umount $facet"
8512 echo "changing server nid..."
8513 $rcmd mount -t lustre -o nosvc,loop $tmp/images/mdt1 $tmp/mnt/mdt1
8514 $rcmd lctl replace_nids $FSNAME-MDT0000 $nid
8515 $rcmd lctl replace_nids $FSNAME-MDT0001 $nid
8516 $rcmd lctl replace_nids $FSNAME-OST0000 $nid
8517 $rcmd lctl replace_nids $FSNAME-OST0001 $nid
8518 $rcmd umount $tmp/mnt/mdt1
8520 for facet in $facets; do
8521 echo "mounting $facet from backup..."
8522 $rcmd mount -t lustre -o loop,abort_recov $tmp/images/$facet \
8523 $tmp/mnt/$facet || error "failed to mount $facet"
8526 for facet in $scrub_list; do
8527 $rcmd $LCTL lfsck_start -M $FSNAME-$facet -t scrub ||
8528 error "failed to start OI scrub on $facet"
8531 # sleep 3 seconds for scrub done
8537 run_test 108b "migrate from ZFS to ldiskfs"
8540 # set number of permanent parameters
8542 test_109_set_params() {
8545 set_persistent_param_and_check mds \
8546 "mdd.$fsname-MDT0000.atime_diff" \
8547 "$fsname-MDT0000.mdd.atime_diff" \
8549 set_persistent_param_and_check mds \
8550 "mdd.$fsname-MDT0000.atime_diff" \
8551 "$fsname-MDT0000.mdd.atime_diff" \
8553 set_persistent_param_and_check client \
8554 "llite.$fsname*.max_read_ahead_mb" \
8555 "$fsname.llite.max_read_ahead_mb" \
8557 set_persistent_param_and_check client \
8558 "llite.$fsname*.max_read_ahead_mb" \
8559 "$fsname.llite.max_read_ahead_mb" \
8561 create_pool $fsname.pool1 || error "create pool failed"
8562 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
8563 error "pool_add failed"
8564 do_facet mgs $LCTL pool_remove $fsname.pool1 OST0000 ||
8565 error "pool_remove failed"
8566 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
8567 error "pool_add failed"
8571 # check permanent parameters
8573 test_109_test_params() {
8576 local atime_diff=$(do_facet mds $LCTL \
8577 get_param -n mdd.$fsname-MDT0000.atime_diff)
8578 [ $atime_diff == 63 ] || error "wrong mdd parameter after clear_conf"
8579 local max_read_ahead_mb=$(do_facet client $LCTL \
8580 get_param -n llite.$fsname*.max_read_ahead_mb)
8581 [ $max_read_ahead_mb == 64 ] ||
8582 error "wrong llite parameter after clear_conf"
8583 local ost_in_pool=$(do_facet mds $LCTL pool_list $fsname.pool1 |
8584 grep -v "^Pool:" | sed 's/_UUID//')
8585 [ $ost_in_pool = "$fsname-OST0000" ] ||
8586 error "wrong pool after clear_conf"
8590 # run lctl clear_conf, store CONFIGS before and after that
8592 test_109_clear_conf()
8594 local clear_conf_arg=$1
8597 if ! combined_mgs_mds ; then
8598 mgsdev=$(mgsdevname)
8599 stop_mgs || error "stop_mgs failed"
8600 start_mgs "-o nosvc" || error "start_mgs nosvc failed"
8602 mgsdev=$(mdsdevname 1)
8603 start_mdt 1 "-o nosvc" || error "start_mdt 1 nosvc failed"
8606 do_facet mgs "rm -rf $TMP/${tdir}/conf1; mkdir -p $TMP/${tdir}/conf1;" \
8607 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf1\\\" \
8611 # the command being tested
8613 do_facet mgs $LCTL clear_conf $clear_conf_arg ||
8614 error "clear_conf failed"
8615 if ! combined_mgs_mds ; then
8616 stop_mgs || error "stop_mgs failed"
8618 stop_mdt 1 || error "stop_mdt 1 failed"
8621 do_facet mgs "rm -rf $TMP/${tdir}/conf2; mkdir -p $TMP/${tdir}/conf2;" \
8622 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf2\\\" \
8626 test_109_file_shortened() {
8628 local sizes=($(do_facet mgs "stat -c %s " \
8629 "$TMP/${tdir}/conf1/CONFIGS/$file" \
8630 "$TMP/${tdir}/conf2/CONFIGS/$file"))
8631 [ ${sizes[1]} -lt ${sizes[0]} ] && return 0
8637 [ $MDS1_VERSION -lt $(version_code 2.10.59) ] &&
8638 skip "Needs MDS version 2.10.59 or later."
8639 [ "$(facet_fstype mgs)" == zfs ] &&
8640 skip "LU-8727: no implementation for ZFS"
8645 client_up || error "client_up failed"
8648 # set number of permanent parameters
8650 test_109_set_params $FSNAME
8652 umount_client $MOUNT || error "umount_client failed"
8653 stop_ost || error "stop_ost failed"
8654 stop_mds || error "stop_mds failed"
8656 test_109_clear_conf $FSNAME
8658 # make sure that all configs are cleared
8660 test_109_file_shortened $FSNAME-MDT0000 ||
8661 error "failed to clear MDT0000 config"
8662 test_109_file_shortened $FSNAME-client ||
8663 error "failed to clear client config"
8668 # check that configurations are intact
8670 test_109_test_params $FSNAME
8675 destroy_test_pools || error "destroy test pools failed"
8679 run_test 109a "test lctl clear_conf fsname"
8683 [ $MDS1_VERSION -lt $(version_code 2.10.59) ] &&
8684 skip "Needs MDS version 2.10.59 or later."
8685 [ "$(facet_fstype mgs)" == zfs ] &&
8686 skip "LU-8727: no implementation for ZFS"
8691 client_up || error "client_up failed"
8694 # set number of permanent parameters
8696 test_109_set_params $FSNAME
8698 umount_client $MOUNT || error "umount_client failed"
8699 stop_ost || error "stop_ost failed"
8700 stop_mds || error "stop_mds failed"
8702 test_109_clear_conf $FSNAME-MDT0000
8704 # make sure that only one config is cleared
8706 test_109_file_shortened $FSNAME-MDT0000 ||
8707 error "failed to clear MDT0000 config"
8708 test_109_file_shortened $FSNAME-client &&
8709 error "failed to clear client config"
8713 # check that configurations are intact
8715 test_109_test_params $FSNAME
8720 destroy_test_pools || error "destroy test pools failed"
8724 run_test 109b "test lctl clear_conf one config"
8728 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
8729 skip "Only applicable to ldiskfs-based MDTs"
8731 do_facet $SINGLEMDS $DEBUGFS -w -R supported_features |grep large_dir ||
8732 skip "large_dir option is not supported on MDS"
8733 do_facet ost1 $DEBUGFS -w -R supported_features | grep large_dir ||
8734 skip "large_dir option is not supported on OSS"
8736 stopall # stop all targets before modifying the target counts
8737 local old_mdscount=$MDSCOUNT
8738 local old_ostcount=$OSTCOUNT
8740 stack_trap "MDSCOUNT=$old_mdscount OSTCOUNT=$old_ostcount; reformat"
8744 # ext4_dir_entry_2 struct size:264
8745 # dx_root struct size:8
8746 # dx_node struct size:8
8747 # dx_entry struct size:8
8748 # For 1024 bytes block size.
8749 # First level directory entries: 126
8750 # Second level directory entries: 127
8751 # Entries in leaf: 3
8752 # For 2 levels limit: 48006
8753 # For 3 levels limit : 6096762
8754 # Create 80000 files to safely exceed 2-level htree limit.
8755 CONF_SANITY_110_LINKS=${CONF_SANITY_110_LINKS:-80000}
8757 # can fit at most 3 filenames per 1KB leaf block, but each
8758 # leaf/index block will only be 3/4 full before split at each level
8759 (( MDSSIZE < CONF_SANITY_110_LINKS / 3 * 4/3 * 4/3 )) &&
8760 CONF_SANITY_110_LINKS=$((MDSSIZE * 3 * 3/4 * 3/4))
8762 combined_mgs_mds || replace=" --replace "
8763 local opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
8764 $replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
8765 if [[ $opts != *large_dir* ]]; then
8766 if [[ $opts != *mkfsoptions* ]]; then
8767 opts+=" --mkfsoptions=\\\"-O large_dir -b 1024 -i 65536\\\""
8769 opts="${opts//--mkfsoptions=\\\"/ \
8770 --mkfsoptions=\\\"-O large_dir -b 1024 -i 65536 }"
8773 echo "MDT params: $opts"
8775 combined_mgs_mds || start_mgs
8776 add mds1 $opts || error "add mds1 failed with new params"
8777 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
8779 opts="$(mkfs_opts ost1 $(ostdevname 1)) \
8780 $replace --reformat $(ostdevname 1) $(ostvdevname 1)"
8782 if [[ $opts != *large_dir* ]]; then
8783 if [[ $opts != *mkfsoptions* ]]; then
8784 opts+=" --mkfsoptions=\\\"-O large_dir\\\" "
8786 opts="${opts//--mkfsoptions=\\\"/ \
8787 --mkfsoptions=\\\"-O large_dir }"
8790 echo "OST params: $opts"
8791 add ost1 $opts || error "add ost1 failed with new params"
8792 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
8794 MOUNT_2=yes mountcli || error "mount clients failed"
8796 mkdir -v $DIR/$tdir || error "cannot create $DIR/$tdir"
8797 local pids count=0 group=0
8799 echo "creating $CONF_SANITY_110_LINKS in total"
8800 while (( count < CONF_SANITY_110_LINKS )); do
8801 local len=$((253 - $(wc -c <<<"$tfile-$group-40000-")))
8802 local dir=DIR$((group % 2 + 1))
8803 local target=${!dir}/$tdir/$tfile-$group
8804 local long=$target-$(generate_name $len)-
8805 local create=$((CONF_SANITY_110_LINKS - count))
8807 (( create > 40000 )) && create=40000
8808 touch $target || error "creating $target failed"
8809 echo "creating $create hard links to $target"
8810 createmany -l $target $long $create &
8813 count=$((count + create))
8814 group=$((group + 1))
8816 echo "waiting for PIDs$pids to complete"
8817 wait $pids || error "createmany failed after $group groups"
8819 umount_client $MOUNT2 -f
8822 run_e2fsck $(facet_active_host mds1) $(mdsdevname 1) -n
8823 MDSCOUNT=$old_mdscount
8824 OSTCOUNT=$old_ostcount
8826 run_test 110 "Adding large_dir with 3-level htree"
8829 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
8830 skip "Only applicable to ldiskfs-based MDTs"
8832 is_dm_flakey_dev $SINGLEMDS $(mdsdevname 1) &&
8833 skip "This test can not be executed on flakey dev"
8835 do_facet $SINGLEMDS $DEBUGFS -w -R supported_features |grep large_dir ||
8836 skip "large_dir option is not supported on MDS"
8838 do_facet ost1 $DEBUGFS -w -R supported_features | grep large_dir ||
8839 skip "large_dir option is not supported on OSS"
8841 # cleanup before changing target counts
8843 local old_mdscount=$MDSCOUNT
8844 local old_ostcount=$OSTCOUNT
8845 local old_mdssize=$MDSSIZE
8847 stack_trap "MDSSIZE=$old_mdssize MDSCOUNT=$old_mdscount \
8848 OSTCOUNT=$old_ostcount; reformat"
8851 (( MDSSIZE < 2400000 )) && MDSSIZE=2400000 # need at least 2.4GB
8853 local mdsdev=$(mdsdevname 1)
8854 combined_mgs_mds || replace=" --replace "
8855 local opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
8856 $replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
8857 if [[ $opts != *large_dir* ]]; then
8858 if [[ $opts != *mkfsoptions* ]]; then
8859 opts+=" --mkfsoptions=\\\"-O large_dir -i 1048576 \\\" "
8861 opts="${opts//--mkfsoptions=\\\"/ \
8862 --mkfsoptions=\\\"-O large_dir -i 1048576 }"
8865 echo "MDT params: $opts"
8867 combined_mgs_mds || start_mgs
8868 __touch_device mds 1
8869 add mds1 $opts || error "add mds1 failed with new params"
8870 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
8872 opts="$(mkfs_opts ost1 $(ostdevname 1)) \
8873 $replace --reformat $(ostdevname 1) $(ostvdevname 1)"
8874 if [[ $opts != *large_dir* ]]; then
8875 if [[ $opts != *mkfsoptions* ]]; then
8876 opts+=" --mkfsoptions=\\\"-O large_dir \\\""
8878 opts="${opts//--mkfsoptions=\\\"/ --mkfsoptions=\\\"-O large_dir }"
8881 echo "OST params: $opts"
8882 __touch_device ost 1
8883 add ost1 $opts || error "add ost1 failed with new params"
8884 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
8886 MOUNT_2=yes mountcli
8887 mkdir $DIR/$tdir || error "cannot create $DIR/$tdir"
8889 lfs df -i $DIR/$tdir
8893 local start=$SECONDS
8895 local dirmax=$((2 << 30))
8901 while (( !needskip & dirsize < dirmax )); do
8904 for cli in ${CLIENTS//,/ }; do
8905 local len=$((253 - $(wc -c <<<"$cli-$group-60000-")))
8906 local target=$cli-$group
8907 local long=$DIR/$tdir/$target-$(generate_name $len)-
8909 RPWD=$DIR/$tdir do_node $cli touch $target ||
8910 error "creating $target failed"
8911 echo "creating 60000 hardlinks to $target"
8912 RPWD=$DIR/$tdir do_node $cli createmany -l $target $long 60000 &
8915 group=$((group + 1))
8917 long=$DIR2/$tdir/$target-$(generate_name $len)-
8919 RPWD=$DIR2/$tdir do_node $cli touch $target ||
8920 error "creating $target failed"
8921 echo "creating 60000 hardlinks to $target"
8922 RPWD=$DIR2/$tdir do_node $cli createmany -l $target $long 60000 &
8925 group=$((group + 1))
8927 echo "waiting for PIDs$pids to complete"
8928 wait $pids || error "createmany failed after $group groups"
8929 dirsize=$(stat -c %s $DIR/$tdir)
8930 taken=$((SECONDS - start))
8931 rate=$((dirsize / taken))
8932 left=$(((dirmax - dirsize) / rate))
8933 num=$((group * 60000))
8934 echo "estimate ${left}s left after $num files / ${taken}s"
8935 # if the estimated time remaining is too large (it may change
8936 # over time as the create rate is not constant) then exit
8937 # without declaring a failure.
8938 (( left > 1200 )) && needskip=1
8941 umount_client $MOUNT2 -f
8944 ! (( $needskip )) ||
8945 echo "ETA ${left}s after $num files / ${taken}s is too long"
8947 run_e2fsck $(facet_active_host mds1) $(mdsdevname 1) -n
8948 MDSCOUNT=$old_mdscount
8949 OSTCOUNT=$old_ostcount
8950 MDSSIZE=$old_mdssize
8952 run_test 111 "Adding large_dir with over 2GB directory"
8955 start_mds || error "MDS start failed"
8956 start_ost || error "OSS start failed"
8957 echo "start ost2 service on $(facet_active_host ost2)"
8958 start ost2 $(ostdevname 2) $(csa_add "$OST_MOUNT_OPTS" -o no_precreate) ||
8959 error "start ost2 facet failed"
8960 local val=$(do_facet ost2 \
8961 "$LCTL get_param -n obdfilter.$FSNAME-OST0001*.no_precreate")
8962 (( $val == 1 )) || error "obdfilter.$FSNAME-OST0001*.no_precreate=$val"
8964 mount_client $MOUNT || error "mount client failed"
8965 wait_osc_import_state mds1 ost1 FULL
8966 wait_osc_import_state client ost1 FULL
8967 wait_osc_import_state client ost2 FULL
8969 $LFS setstripe -i 0 $DIR/$tfile.0 ||
8970 error "problem creating $tfile.0 on OST0000"
8971 $LFS setstripe -i 1 $DIR/$tfile.1 && $LFS getstripe $DIR/$tfile.1 &&
8972 (( $($LFS getstripe -i $DIR/$tfile.1) == 1 )) &&
8973 error "allowed to create $tfile.1 on OST0001"
8974 do_facet ost2 $LCTL set_param obdfilter.*.no_precreate=0
8976 $LFS setstripe -i 1 $DIR/$tfile.2 ||
8977 error "failed to create $tfile.2 on ost1 facet"
8978 # files not cleaned with ONLY_REPEAT because of client unmount below
8979 rm -f $DIR/$tfile.[012]
8980 stop_ost2 || error "stop ost2 facet failed"
8983 run_test 112 "mount OST with nocreate option"
8986 SAVE_MGS_MOUNT_OPTS=$MGS_MOUNT_OPTS
8987 SAVE_MDS_MOUNT_OPTS=$MDS_MOUNT_OPTS
8988 SAVE_OST_MOUNT_OPTS=$OST_MOUNT_OPTS
8994 MGS_MOUNT_OPTS=$SAVE_MGS_MOUNT_OPTS
8995 MDS_MOUNT_OPTS=$SAVE_MDS_MOUNT_OPTS
8996 OST_MOUNT_OPTS=$SAVE_OST_MOUNT_OPTS
8997 # Revert old mount options back
8999 # Subsequent following test requires
9000 # conf-sanity to be in stopall state.
9001 # Force 'stopall' so others following
9006 # Error out with mount info
9008 local server_nodes=$(comma_list $(mdts_nodes) $(osts_nodes))
9011 echo "--Client Mount Info--"
9012 mount | grep -i lustre
9013 echo "--Server Mount Info--"
9014 do_nodes $server_nodes mount | grep -i lustre
9020 local ost_version="2.15.51" # Minimum version required
9022 (( OST1_VERSION >= $(version_code $ost_version) )) ||
9023 skip "Need server version at least $ost_version"
9025 stack_trap cleanup_113 EXIT
9027 # Reset before starting
9031 # Verify MDS's should start with "rw"
9032 do_facet $SINGLEMDS mount | grep "lustre.*rw,.*MDT" ||
9033 error_113 "$SINGLEMDS should be read-write"
9035 # Verify OST's should start with "rw"
9036 for (( i=1; i <= OSTCOUNT; i++ )); do
9037 do_facet ost$i mount | grep "lustre.*rw,.*OST" ||
9038 error_113 "ost$i should be read-write"
9041 # rdonly_dev does not currently work for ldiskfs
9042 # We skip the rdonly_dev check until then.
9043 if [[ $ost1_FSTYPE == ldiskfs ]]; then
9044 echo "Shadow Mountpoint correctly reports rw for ldiskfs"
9049 # Only ZFS specific tests below.
9052 # Must stop all (server+client) and restart to verify new
9056 # add rdonly_dev to mount option
9057 MGS_MOUNT_OPTS=$(csa_add "$MGS_MOUNT_OPTS" -o rdonly_dev)
9058 MDS_MOUNT_OPTS=$(csa_add "$MDS_MOUNT_OPTS" -o rdonly_dev)
9059 OST_MOUNT_OPTS=$(csa_add "$OST_MOUNT_OPTS" -o rdonly_dev)
9061 # Only restart server(mds/ost). Sufficient for test
9062 setupall server_only || error "Fail to start servers"
9064 # Verify MDS's should be "ro"
9065 do_facet $SINGLEMDS mount | grep "lustre.*ro,.*MDT.*rdonly_dev" ||
9066 error_113 "$SINGLEMDS should be read-only"
9068 # Verify OST's should be "ro"
9069 for (( i=1; i <= OSTCOUNT; i++ )); do
9070 do_facet ost$i mount | grep "lustre.*ro,.*OST.*rdonly_dev" ||
9071 error_113 "ost$i should be read-only"
9074 run_test 113 "Shadow mountpoint correctly report ro/rw for mounts"
9077 # Purpose: To verify dynamic thread (OSS) creation.
9078 # (This was sanity/115)
9081 [ $PARALLEL == "yes" ] && skip "skip parallel run"
9082 remote_ost_nodsh && skip "remote OST with nodsh"
9084 # Lustre does not stop service threads once they are started.
9085 # Reset number of running threads to default.
9090 local save_params="$TMP/sanity-$TESTNAME.parameters"
9092 # Get ll_ost_io count before I/O
9093 OSTIO_pre=$(do_facet ost1 \
9094 "$LCTL get_param ost.OSS.ost_io.threads_started | cut -d= -f2")
9095 # Exit if lustre is not running (ll_ost_io not running).
9096 [ -z "$OSTIO_pre" ] && error "no OSS threads"
9098 echo "Starting with $OSTIO_pre threads"
9099 local thread_max=$((OSTIO_pre * 2))
9100 local rpc_in_flight=$((thread_max * 2))
9102 # this is limited to OSC_MAX_RIF_MAX (256)
9103 [ $rpc_in_flight -gt 256 ] && rpc_in_flight=256
9104 thread_max=$((rpc_in_flight / 2))
9105 [ $thread_max -le $OSTIO_pre ] && skip "Too many ost_io threads" &&
9108 # Number of I/O Process proposed to be started.
9110 local facets=$(get_facets OST)
9112 save_lustre_params client "osc.*OST*.max_rpcs_in_flight" > $save_params
9113 save_lustre_params $facets "ost.OSS.ost_io.threads_max" >> $save_params
9115 # Set in_flight to $rpc_in_flight
9116 $LCTL set_param osc.*OST*.max_rpcs_in_flight=$rpc_in_flight ||
9117 error "Failed to set max_rpcs_in_flight to $rpc_in_flight"
9118 nfiles=${rpc_in_flight}
9119 # Set ost thread_max to $thread_max
9120 do_facet ost1 "$LCTL set_param ost.OSS.ost_io.threads_max=$thread_max"
9122 # 5 Minutes should be sufficient for max number of OSS
9123 # threads(thread_max) to be created.
9127 local wtl=${WTL:-"$LUSTRE/tests/write_time_limit"}
9129 test_mkdir $DIR/$tdir
9130 for ((i = 1; i <= nfiles; i++)); do
9131 local file=$DIR/$tdir/${tfile}-$i
9133 $LFS setstripe -c -1 -i 0 $file
9134 ($wtl $file $timeout)&
9137 # I/O Started - Wait for thread_started to reach thread_max or report
9138 # error if thread_started is more than thread_max.
9139 echo "Waiting for thread_started to reach thread_max"
9140 local thread_started=0
9141 local end_time=$((SECONDS + timeout))
9143 while [ $SECONDS -le $end_time ] ; do
9145 # Get ost i/o thread_started count.
9146 thread_started=$(do_facet ost1 \
9148 ost.OSS.ost_io.threads_started | cut -d= -f2")
9149 # Break out if thread_started is equal/greater than thread_max
9150 if (( $thread_started >= $thread_max )); then
9151 echo ll_ost_io thread_started $thread_started, \
9152 equal/greater than thread_max $thread_max
9158 # Cleanup - We have the numbers, Kill i/o jobs if running.
9159 jobcount=($(jobs -p))
9161 for ((i=0; i < ${#jobcount[*]}; i++)); do
9162 kill -9 ${jobcount[$i]}
9163 if [ $? -ne 0 ] ; then
9164 echo "warning: cannot kill WTL pid ${jobcount[$i]}"
9168 # Cleanup files left by WTL binary.
9169 for ((i = 1; i <= nfiles; i++)); do
9170 local file=$DIR/$tdir/${tfile}-$i
9173 if [ $? -ne 0 ] ; then
9174 echo "Warning: Failed to delete file $file"
9178 restore_lustre_params <$save_params
9179 rm -f $save_params || echo "Warning: delete file '$save_params' failed"
9181 # Error out if no new thread has started or Thread started is greater
9183 if (( $thread_started <= $OSTIO_pre ||
9184 $thread_started > $thread_max )); then
9185 error "ll_ost_io: thread_started $thread_started" \
9186 "OSTIO_pre $OSTIO_pre, thread_max $thread_max." \
9187 "No new thread started or thread started greater " \
9191 run_test 114 "verify dynamic thread creation===================="
9197 do_facet mds1 rm -f $1
9201 if [ "$mds1_FSTYPE" != ldiskfs ]; then
9202 skip "Only applicable to ldiskfs-based MDTs"
9204 [ -n "$FILESET" ] && skip "Not functional for FILESET set"
9206 local dbfs_ver=$(do_facet mds1 $DEBUGFS -V 2>&1)
9208 echo "debugfs version: $dbfs_ver"
9209 echo "$dbfs_ver" | egrep -w "1.44.3.wc1|1.44.5.wc1|1.45.2.wc1" &&
9210 skip_env "This version of debugfs doesn't show inode number"
9212 local IMAGESIZE=$((3072 << 30)) # 3072 GiB
9222 do_facet mds1 "hostname; $LCTL dl; mount"
9224 do_facet ost1 "hostname; $LCTL dl; mount"
9225 # We need MDT size 3072GB, because it is smallest
9226 # partition that can store 2B inodes
9227 do_facet mds1 "mkdir -p $TMP/$tdir"
9228 local mdsimgname=$TMP/$tdir/lustre-mdt
9230 do_facet mds1 "rm -f $mdsimgname"
9231 do_facet mds1 "touch $mdsimgname"
9232 stack_trap "cleanup_115 $mdsimgname" EXIT
9233 do_facet mds1 "$TRUNCATE $mdsimgname $IMAGESIZE" ||
9234 skip "Backend FS doesn't support sparse files"
9235 local mdsdev=$(do_facet mds1 "losetup -f")
9237 do_facet mds1 "losetup $mdsdev $mdsimgname"
9239 local mds_opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
9240 --mkfsoptions='-O ea_inode,^resize_inode,meta_bg \
9241 -N 2247484000 -E lazy_itable_init' --device-size=$IMAGESIZE"
9242 add mds1 $mds_opts --mgs --reformat $mdsdev ||
9243 skip_env "format large MDT failed"
9245 local ostdev=$(ostdevname 1)
9247 local opts="$(mkfs_opts ost1 $ostdev) \
9248 --reformat $ostdev $ostdev"
9249 add ost1 $opts || error "add ost1 failed with new params"
9250 start mds1 $mdsdev $MDS_MOUNT_OPTS || error "start MDS failed"
9251 start_ost || error "start OSS failed"
9252 mount_client $MOUNT || error "mount client failed"
9254 mkdir -p $DIR/$tdir || error "mkdir $DIR/$tdir fail"
9255 local goal="/sys/fs/ldiskfs/$(basename $mdsdev)/inode_goal"
9257 # 2147483648 is 0x80000000
9258 do_facet mds1 "echo 2147483648 >> $goal; grep . $goal"
9259 touch $DIR/$tdir/$tfile
9261 # attrs from 1 to 16 go to block, 17th - to inode
9264 for i in {1..17}; do
9265 local nm="trusted.ea$i"
9266 setfattr -n $nm -v $(printf "xattr%0250d" $i) $DIR/$tdir/$tfile
9269 do_facet mds1 "$DEBUGFS -c -R 'stat ROOT/$tdir/$tfile' $mdsdev"
9271 # inode <2147483649> trusted.ea16 (255)
9272 local inode_num=$(do_facet mds1 \
9273 "$DEBUGFS -c -R 'stat ROOT/$tdir/$tfile' $mdsdev" |
9274 awk '/ea17/ { print $2 }' |
9275 sed -e 's/>//' -e 's/<//' -e 's/\"//')
9276 echo "inode num: $inode_num"
9277 [ $inode_num -ge 2147483648 ] || error "inode $inode_num too small"
9278 do_facet mds1 "losetup -d $mdsdev"
9280 run_test 115 "Access large xattr with inodes number over 2TB"
9283 [ "$mds1_FSTYPE" != ldiskfs ] && skip "ldiskfs only test"
9284 [ "$MDS1_VERSION" -lt $(version_code 2.10.59) ] &&
9285 skip "Need server version at least 2.10.59"
9286 do_facet $SINGLEMDS which mkfs.xfs ||
9287 skip_env "No mkfs.xfs installed"
9292 local tmpmnt=$TMP/$tdir
9293 local mdtimg=$tfile-mdt0
9295 do_facet $SINGLEMDS mkdir -p $tmpmnt
9296 stack_trap "do_facet $SINGLEMDS rmdir $tmpmnt" EXIT
9298 do_facet $SINGLEMDS touch $TMP/$mdtimg
9299 stack_trap "do_facet $SINGLEMDS rm -f $TMP/$mdtimg" EXIT
9300 do_facet $SINGLEMDS mkfs -t xfs -d file,size=1t,name=$TMP/$mdtimg ||
9301 error "mkfs temporary xfs image"
9303 do_facet $SINGLEMDS mount $TMP/$mdtimg $tmpmnt ||
9304 error "mount temporary xfs image"
9305 stack_trap "do_facet $SINGLEMDS umount $tmpmnt" EXIT
9306 local old_mdssize=$MDSSIZE
9307 local old_mdsisize=$MDSISIZE
9309 MDSSIZE=$((17 * 1024 * 1024 * 1024)) # 17T MDT
9310 MDSISIZE=$((16 << 20))
9311 local opts17t="$(mkfs_opts $SINGLEMDS)"
9313 MDSSIZE=$old_mdssize
9314 MDSISIZE=$old_mdsisize
9316 do_facet $SINGLEMDS $MKFS \
9317 ${opts17t/-E /-E lazy_itable_init,lazy_journal_init,} \
9319 error "failed to mkfs for $tmpmnt/$mdtimg"
9321 do_facet $SINGLEMDS $TUNE2FS -l $tmpmnt/$mdtimg |
9322 grep -qw 'features.*extent' || error "extent should be enabled"
9325 run_test 116 "big size MDT support"
9328 # Call setup only if LustreFS is not mounted
9329 check_mount || setup
9331 do_facet ost1 "$LCTL set_param ost.OSS.ost_io.nrs_policies=fifo"
9332 do_facet ost1 "$LCTL get_param -n ost.OSS.ost_io.nrs_tbf_rule" &&
9333 error "get_param should fail"
9334 cleanup || error "cleanup failed with rc $?"
9336 run_test 117 "lctl get_param return errors properly"
9342 [[ "$MDSCOUNT" -ge 2 ]] || skip "Need more at least 2 MDTs"
9344 had_config=$(do_facet mds1 "$LCTL get_param debug | grep config")
9345 do_facet mds1 "$LCTL set_param debug=+config"
9346 do_facet mds1 "$LCTL clear"
9349 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)" &>/dev/null
9350 # mount after writeconf will make "add osp" added to mdt0 config:
9351 # 53 (224)marker 60 (flags=0x01, v2.5.1.0) lustre-MDT0001 'add osp'
9352 # 54 (080)add_uuid nid=... 0: 1:...
9353 # 55 (144)attach 0:lustre-MDT0001-osp-MDT0000 1:osp 2:...
9354 # 56 (144)setup 0:lustre-MDT0001-osp-MDT0000 1:... 2:...
9355 # 57 (136)modify_mdc_tgts add 0:lustre-MDT0000-mdtlov 1:... 2:1 3:1
9356 # duplicate modify_mdc_tgts caused crashes
9359 # using larger debug_mb size to avoid lctl dk log truncation
9360 size_mb=$((DEBUG_SIZE_SAVED * 4))
9363 # though config processing stops after failed attach and setup
9364 # it will proceed after the failed command after each writeconf
9365 # this is the original scenario of the issue
9366 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)" &>/dev/null
9367 do_facet mds1 "$LCTL set_param debug_mb=$size_mb"
9370 wait_update_facet_cond mds1 \
9371 "$LCTL dk | grep -c Processed.log.$FSNAME-MDT0000" \
9376 [[ -z "$had_config" ]] && do_facet mds1 lctl set_param debug=-config
9380 run_test 119 "writeconf on slave mdt shouldn't duplicate mdc/osp and crash"
9382 test_120() { # LU-11130
9383 [ "$MDSCOUNT" -lt 2 ] && skip "mdt count < 2"
9384 [ "$mds1_FSTYPE" != ldiskfs ] &&
9385 skip "ldiskfs only test"
9386 [ "$MDS1_VERSION" -lt $(version_code 2.11.56) ] &&
9387 skip "Need DNE2 capable MD target with LU-11130 fix"
9391 local mds1host=$(facet_active_host mds1)
9392 local mds1dev=$(mdsdevname 1)
9394 $LFS mkdir -i 1 $DIR/$tdir
9395 $LFS mkdir -i 0 $DIR/$tdir/mds1dir
9397 ln -s foo $DIR/$tdir/bar
9398 mv $DIR/$tdir/bar $DIR/$tdir/mds1dir/bar2 ||
9399 error "cross-target rename failed"
9403 run_e2fsck $mds1host $mds1dev "-n"
9405 run_test 120 "cross-target rename should not create bad symlinks"
9409 start_mgsmds || error "MGS MDS Start failed"
9411 stop_mds || error "Stopping MDSes failed"
9415 stop_mds || error "Stopping MDSes failed"
9417 run_test 121 "failover MGS"
9420 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
9421 [[ "$OST1_VERSION" -ge $(version_code 2.11.53) ]] ||
9422 skip "Need OST version at least 2.11.53"
9425 LOAD_MODULES_REMOTE=true load_modules
9426 #define OBD_FAIL_OFD_SET_OID 0x1e0
9427 do_facet ost1 $LCTL set_param fail_loc=0x00001e0
9430 $LFS mkdir -i1 -c1 $DIR/$tdir
9431 $LFS setstripe -i0 -c1 $DIR/$tdir
9432 do_facet ost1 $LCTL set_param fail_loc=0
9433 createmany -o $DIR/$tdir/file_ 1000 ||
9434 error "Fail to create a new sequence"
9438 run_test 122a "Check OST sequence update"
9441 [[ "$OST1_VERSION" -ge $(version_code 2.11.53) ]] ||
9442 skip "Need OST version at least 2.11.53"
9446 LOAD_MODULES_REMOTE=true load_modules
9447 #define OBD_FAIL_OFD_SET_OID 0x1e0
9448 do_facet ost1 $LCTL set_param fail_loc=0x00001e0
9450 stack_trap cleanup EXIT
9452 do_facet ost1 $LCTL set_param obdfilter.*.precreate_batch=256
9453 $LFS mkdir -i0 -c1 $DIR/$tdir || error "failed to create directory"
9454 $LFS setstripe -i0 -c1 $DIR/$tdir || error "failed to setstripe"
9455 do_facet ost1 $LCTL set_param fail_loc=0
9456 # overflow IDIF 32bit and create > OST_MAX_PRECREATE*5
9457 # so a new wrong sequence would differ from an original with error
9458 #define OST_MAX_PRECREATE 20000
9459 local ost_max_precreate=20100
9460 local num_create=$(( ost_max_precreate * 5 ))
9462 # Check the number of inodes available on OST0
9464 local ifree=$($LFS df -i $MOUNT |
9465 awk '/OST0000/ { print $4 }'; exit ${PIPESTATUS[0]})
9467 log "On OST0, $ifree inodes available. Want $num_create. rc=$?"
9469 if [ $ifree -lt 10000 ]; then
9470 files=$(( ifree - 50 ))
9475 local j=$((num_create / files + 1))
9477 for i in $(seq 1 $j); do
9478 createmany -o $DIR/$tdir/$tfile-$i- $files ||
9479 error "createmany fail create $files files: $?"
9480 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
9481 error "unlinkmany failed unlink $files files"
9484 touch $DIR/$tdir/$tfile
9486 #we need a write req during recovery for ofd_seq_load
9488 dd if=/dev/urandom of=$DIR/$tdir/$tfile bs=1024k count=1 oflag=sync ||
9489 error "failed to write file"
9491 # OBD_FAIL_OST_CREATE_NET 0x204
9492 do_facet ost1 $LCTL set_param fail_loc=0x80000204
9494 createmany -o $DIR/$tdir/file_ 100
9497 err=$(do_facet ost1 dmesg | tac | sed "/Recovery over/,$ d" |
9498 grep "OST replaced or reformatted")
9499 [ -z "$err" ] || error $err
9501 run_test 122b "Check OST sequence wouldn't change when IDIF 32bit overflows"
9504 remote_mgs_nodsh && skip "remote MGS with nodsh"
9505 [ -d $MOUNT/.lustre ] || setup
9507 # test old logid format until removal from llog_ioctl.c::str2logid()
9508 if [ $MGS_VERSION -lt $(version_code 3.1.53) ]; then
9509 do_facet mgs $LCTL dl | grep MGS
9510 do_facet mgs "$LCTL --device %MGS llog_print \
9511 \\\\\\\$$FSNAME-client 1 10" ||
9512 error "old llog_print failed"
9515 # test new logid format
9516 if [ $MGS_VERSION -ge $(version_code 2.9.53) ]; then
9517 do_facet mgs "$LCTL --device MGS llog_print $FSNAME-client" ||
9518 error "new llog_print failed"
9521 run_test 123aa "llog_print works with FIDs and simple names"
9524 remote_mgs_nodsh && skip "remote MGS with nodsh"
9525 [[ $MGS_VERSION -gt $(version_code 2.11.51) ]] ||
9526 skip "Need server with working llog_print support"
9528 [ -d $MOUNT/.lustre ] || setup
9534 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9535 mgs_arg="--device MGS"
9537 orig_val=$(do_facet mgs $LCTL get_param jobid_name)
9538 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME"
9540 yaml=$(do_facet mgs $LCTL $mgs_arg llog_print params |
9541 grep jobid_name | tail -n 1)
9543 local param=$(awk '{ print $10 }' <<< "$yaml")
9544 local val=$(awk '{ print $12 }' <<< "$yaml")
9545 #return to the default
9546 do_facet mgs $LCTL set_param -P jobid_name=$orig_val
9547 [ $val = "TESTNAME" ] || error "bad value: $val"
9548 [ $param = "jobid_name," ] || error "Bad param: $param"
9550 run_test 123ab "llog_print params output values from set_param -P"
9552 test_123ac() { # LU-11566
9553 remote_mgs_nodsh && skip "remote MGS with nodsh"
9554 do_facet mgs "$LCTL help llog_print" 2>&1 | grep -q -- --start ||
9555 skip "Need 'lctl llog_print --start' on MGS"
9561 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9562 mgs_arg="--device MGS"
9564 [ -d $MOUNT/.lustre ] || setup
9566 # - { index: 10, event: add_uuid, nid: 192.168.20.1@tcp(0x20000c0a81401,
9567 # node: 192.168.20.1@tcp }
9568 do_facet mgs $LCTL $mgs_arg \
9569 llog_print --start $start --end $end $FSNAME-client | tr -d , |
9570 while read DASH BRACE INDEX idx EVENT BLAH BLAH BLAH; do
9571 (( idx >= start )) || error "llog_print index $idx < $start"
9572 (( idx <= end )) || error "llog_print index $idx > $end"
9575 run_test 123ac "llog_print with --start and --end"
9577 test_123ad() { # LU-11566
9578 remote_mgs_nodsh && skip "remote MGS with nodsh"
9579 # older versions of lctl may not print all records properly
9580 do_facet mgs "$LCTL help llog_print" 2>&1 | grep -q -- --start ||
9581 skip "Need 'lctl llog_print --start' on MGS"
9583 [ -d $MOUNT/.lustre ] || setup
9585 # append a new record, to avoid issues if last record was cancelled
9586 local old=$($LCTL get_param -n osc.*-OST0000-*.max_dirty_mb | head -1)
9587 do_facet mgs $LCTL conf_param $FSNAME-OST0000.osc.max_dirty_mb=$old
9589 # logid: [0x3:0xa:0x0]:0
9593 local num=$(do_facet mgs $LCTL --device MGS llog_info $FSNAME-client |
9594 awk '/last_index:/ { print $2 - 1 }')
9596 # - { index: 71, event: set_timeout, num: 0x14, param: sys.timeout=20 }
9597 local last=$(do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
9598 tail -1 | awk '{ print $4 }' | tr -d , )
9599 (( last == num )) || error "llog_print only showed $last/$num records"
9601 run_test 123ad "llog_print shows all records"
9603 test_123ae() { # LU-11566
9612 remote_mgs_nodsh && skip "remote MGS with nodsh"
9613 [ -d $MOUNT/.lustre ] || setupall
9615 max=$($LCTL get_param -n osc.*-OST0000-*.max_dirty_mb | head -1)
9616 pgs=$($LCTL get_param -n osc.*-OST0000-*.max_pages_per_rpc | head -1)
9617 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9618 mgs_arg="--device MGS"
9620 if do_facet mgs "$LCTL help llog_cancel" 2>&1| grep -q -- --log_id; then
9621 # save one set_param -P record in case none exist
9623 do_facet mgs $LCTL set_param -P osc.*.max_pages_per_rpc=$pgs
9624 stack_trap "do_facet mgs $LCTL set_param -P -d \
9625 osc.*.max_pages_per_rpc"
9628 orig=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9629 tail -1 | awk '{ print $4 }' | tr -d , )
9630 do_facet mgs $LCTL set_param -P osc.*.max_dirty_mb=$max
9631 do_facet mgs $LCTL $mgs_arg llog_print $log | tail -1 |
9632 grep "parameter: osc.*.max_dirty_mb" ||
9633 error "new set_param -P wasn't stored in params log"
9635 # - { index: 71, event: set_param, device: general,
9636 # param: osc.*.max_dirty_mb, value: 256 }
9637 id=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9638 tail -1 | awk '{ print $4 }' | tr -d , )
9639 do_facet mgs $LCTL $mgs_arg llog_cancel $log --log_idx=$id
9640 local new=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9641 tail -1 | awk '{ print $4 }' | tr -d , )
9642 (( new == orig )) ||
9643 error "new llog_cancel now $new, not at $orig records"
9646 # test old positional parameters for a while still
9647 if [ "$MGS_VERSION" -le $(version_code 3.1.53) ]; then
9650 do_facet mgs $LCTL conf_param \
9651 $FSNAME-OST0000.osc.max_pages_per_rpc=$pgs
9652 stack_trap "do_facet mgs $LCTL conf_param -d \
9653 $FSNAME-OST0000.osc.max_pages_per_rpc"
9655 orig=$(do_facet mgs $LCTL --device MGS llog_print $log |
9656 tail -1 | awk '{ print $4 }' | tr -d , )
9657 do_facet mgs $LCTL conf_param $FSNAME-OST0000.osc.max_dirty_mb=$max
9658 do_facet mgs $LCTL --device MGS llog_print $log |
9659 tail -1 | grep "parameter: osc.max_dirty_mb" ||
9660 error "old conf_param wasn't stored in params log"
9661 do_facet mgs $LCTL --device MGS llog_print $log
9662 # - { index: 71, event: conf_param, device: testfs-OST0000-osc,
9663 # param: osc.max_dirty_mb=256 }
9664 id=$(do_facet mgs $LCTL --device MGS llog_print $log |
9665 tail -1 | awk '{ print $4 }' | tr -d , )
9666 do_facet mgs $LCTL --device MGS llog_cancel $log $id
9667 do_facet mgs $LCTL --device MGS llog_print $log
9668 new=$(do_facet mgs $LCTL --device MGS llog_print $log |
9669 tail -1 | awk '{ print $4 }' | tr -d , )
9670 (( new == orig )) ||
9671 error "old llog_cancel now $new, not at $orig records"
9674 run_test 123ae "llog_cancel can cancel requested record"
9676 test_123af() { #LU-13609
9677 [ "$MGS_VERSION" -ge $(version_code 2.13.54) -a \
9678 "$MDS1_VERSION" -ge $(version_code 2.13.54) ] ||
9679 skip "Need both MGS and MDS version at least 2.13.54"
9681 [ -d $MOUNT/.lustre ] || setupall
9682 stack_trap "do_facet mds1 $LCTL set_param fail_loc=0" EXIT
9692 for device in "MGS" "$FSNAME-MDT0000"; do
9693 cmd="--device $device llog_catlist"
9694 echo "lctl $cmd ..."
9695 if [ "$device" = "MGS" ]; then
9700 orig_clist=($(do_facet $facet $LCTL $cmd | awk '{ print $2 }'))
9701 orig_count=${#orig_clist[@]}
9702 echo "orig_clist: ${orig_clist[*]}"
9704 #define OBD_FAIL_CATLIST 0x131b
9705 #fetch to llog records from the second one
9706 do_facet $facet $LCTL set_param fail_loc=0x131b fail_val=2
9708 new_clist=($(do_facet $facet $LCTL $cmd | awk '{ print $2 }'))
9709 new_count=${#new_clist[@]}
9710 echo "new_clist: ${new_clist[*]}"
9712 [ $new_count -eq $((orig_count - 1)) ] ||
9713 error "$new_count != $orig_count - 1"
9714 for i in $(seq 0 $new_count); do
9716 [ "${orig_clist[$j]}" = "${new_clist[$i]}" ] ||
9717 error "${orig_clist[$j]} != ${new_clist[$i]}"
9719 do_facet mds1 $LCTL set_param fail_loc=0
9723 run_test 123af "llog_catlist can show all config files correctly"
9725 test_123ag() { # LU-15142
9729 remote_mgs_nodsh && skip "remote MGS with nodsh"
9730 (( $MGS_VERSION >= $(version_code 2.14.55) )) ||
9731 skip "Need server version least 2.14.55"
9733 [ -d $MOUNT/.lustre ] || setup
9735 orig_val=$(do_facet mgs $LCTL get_param jobid_name)
9736 stack_trap "do_facet mgs $LCTL set_param -P jobid_name=$orig_val"
9738 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME1"
9739 do_facet mgs $LCTL set_param -P -d jobid_name
9740 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
9742 (( rec == 0 )) || error "parameter was not deleted, check #1"
9743 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME1"
9744 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
9746 (( rec == 1)) || error "parameter is not set"
9747 # usage with ordinary set_param format works too
9748 do_facet mgs $LCTL set_param -P -d jobid_name="ANY"
9749 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
9751 (( rec == 0 )) || error "parameter was not deleted, check #2"
9753 run_test 123ag "llog_print skips values deleted by set_param -P -d"
9755 test_123ah() { #LU-7668 del_ost
9756 [ "$MGS_VERSION" -ge $(version_code 2.15.50) -a \
9757 "$MDS1_VERSION" -ge $(version_code 2.15.50) ] ||
9758 skip "Need both MGS and MDS version at least 2.15.50"
9760 [ -d $MOUNT/.lustre ] || setupall
9761 stack_trap "do_facet mds1 $LCTL set_param fail_loc=0" EXIT
9763 local cmd="--device MGS llog_print"
9765 cli_llogcnt_orig=$(do_facet mgs $LCTL $cmd $FSNAME-client |
9766 grep -c $FSNAME-OST0000)
9767 mds1_llogcnt_orig=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 |
9768 grep -c $FSNAME-OST0000)
9770 [ $cli_llogcnt_orig -gt 0 ] ||
9771 error "$FSNAME-OST0000 not found (client)"
9772 [ $mds1_llogcnt_orig -gt 0 ] || error "$FSNAME-OST0000 not found (MDT)"
9774 # -n/--dryrun should NOT modify catalog
9775 do_facet mgs "$LCTL del_ost -n -t $FSNAME-OST0000" ||
9776 error "del_ost --dryrun failed with $?"
9778 local cli_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-client |
9779 grep -c $FSNAME-OST0000)
9780 local mds1_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 |
9781 grep -c $FSNAME-OST0000)
9783 [ $cli_llogcnt -eq $cli_llogcnt_orig ] ||
9784 error "--dryrun error: $cli_llogcnt != $cli_llogcnt_orig"
9785 [ $mds1_llogcnt -eq $mds1_llogcnt_orig ] ||
9786 error "--dryrun error: $mds1_llogcnt != $mds1_llogcnt_orig"
9789 do_facet mgs "$LCTL del_ost --target $FSNAME-OST0000" ||
9790 error "del_ost failed with $?"
9792 local cli_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-client |
9793 grep -c $FSNAME-OST0000)
9794 local mds1_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 |
9795 grep -c $FSNAME-OST0000)
9797 # every catalog entry for OST0000 should have been cancelled
9798 [ $cli_llogcnt -eq 0 ] || error "$FSNAME-OST0000 not cancelled (cli)"
9799 [ $mds1_llogcnt -eq 0 ] || error "$FSNAME-OST0000 not cancelled (MDT)"
9801 umount_client $MOUNT
9804 $LFS df | grep -q OST0000 && error "del_ost did not remove OST0000!"
9808 run_test 123ah "del_ost cancels config log entries correctly"
9813 echo "cleanup test 123ai"
9815 # cancel last timeout record
9816 do_facet mgs "$LCTL set_param -P -d timeout"
9818 # restore timeout value
9819 do_nodes $(comma_list $(all_nodes)) "$LCTL set_param timeout=$timeout"
9822 test_123ai() { #LU-16167
9828 remote_mgs_nodsh && skip "remote MGS with nodsh"
9829 (( MDS1_VERSION >= $(version_code 2.15.51) )) ||
9830 skip "Need MDS version at least 2.15.51"
9832 [ -d $MOUNT/.lustre ] || setup
9834 old_count=$(do_facet mgs "$LCTL --device MGS llog_print -r params" |
9835 grep -c "parameter: timeout")
9837 old_timeout=$($LCTL get_param -n timeout)
9838 stack_trap "cleanup_123ai $old_timeout" EXIT
9840 # add and cancel params (2 * MAX_IOC_BUFLEN / 128 + 1 = 129)
9841 for i in {1..129}; do
9842 do_facet mgs "$LCTL set_param -P timeout=$i" ||
9843 error "fail to set timeout=$i on MGS"
9846 count=$(do_facet mgs "$LCTL --device MGS llog_print -r params" |
9847 grep -c "parameter: timeout")
9849 ((count - old_count == 129)) ||
9850 error "number timeout record missmatch ($count - $old_count != 129)"
9852 do_facet mgs "$LCTL --device MGS llog_print params" |
9853 tail -1 | grep "timeout, value: 129" ||
9854 error "llog_print could not display the last record (timeout=129)"
9857 run_test 123ai "llog_print display all non skipped records"
9860 remote_mgs_nodsh && skip "remote MGS with nodsh"
9862 [ -d $MOUNT/.lustre ] || setup
9863 local yaml_file="$TMP/$tfile.yaml"
9865 do_facet mgs rm "$yaml_file"
9866 local cfgfiles=$(do_facet mgs "lctl --device MGS llog_catlist" |
9867 sed 's/config_log://')
9869 # set jobid_var to a different value for test
9870 local orig_val=$(do_facet mgs $LCTL get_param jobid_var)
9872 do_facet mgs $LCTL set_param -P jobid_var="TESTNAME"
9874 for i in $cfgfiles params; do
9875 do_facet mgs "lctl --device MGS llog_print ${i} >> $yaml_file"
9878 echo "Unmounting FS"
9885 # Reapply the config from before
9886 echo "Setting configuration parameters"
9887 do_facet mgs "lctl set_param -F $yaml_file"
9889 local set_val=$(do_facet mgs $LCTL get_param jobid_var)
9891 do_facet mgs $LCTL set_param -P $orig_val
9893 [ $set_val == "jobid_var=TESTNAME" ] ||
9894 error "$set_val is not TESTNAME"
9896 do_facet mgs rm "$yaml_file"
9899 run_test 123F "clear and reset all parameters using set_param -F"
9903 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
9904 [ -z $mds2failover_HOST ] && skip "needs MDT failover setup"
9907 do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
9908 grep 1.2.3.4@tcp && error "Should not be fake nid"
9909 do_facet mgs $LCTL conf_param $FSNAME-MDT0001.failover.node=1.2.3.4@tcp\
9910 || error "Set params error"
9911 do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
9912 grep 1.2.3.4@tcp || error "Fake nid should be added"
9916 if combined_mgs_mds; then
9917 start_mdt 1 "-o nosvc" ||
9918 error "starting mds with nosvc option failed"
9920 local nid=$(do_facet mds2 $LCTL list_nids | head -1)
9921 local failover_nid=$(do_node $mds2failover_HOST $LCTL list_nids | head -1)
9923 do_facet mgs $LCTL replace_nids $FSNAME-MDT0001 $nid:$failover_nid ||
9924 error "replace_nids execution error"
9926 if combined_mgs_mds; then
9932 echo "lfs setdirstripe"
9933 $LFS setdirstripe -i 1 $MOUNT/$tdir || error "setdirstirpe error"
9936 run_test 124 "check failover after replace_nids"
9938 get_max_sectors_kb() {
9941 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
9942 local max_sectors_path="/sys/block/${dev_base}/queue/max_sectors_kb"
9944 do_facet ${facet} "[[ -e ${max_sectors_path} ]] &&
9945 cat ${max_sectors_path}"
9948 get_max_hw_sectors_kb() {
9951 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
9952 local max_hw_path="/sys/block/${dev_base}/queue/max_hw_sectors_kb"
9954 do_facet ${facet} "[[ -e ${max_hw_path} ]] && cat ${max_hw_path}"
9957 set_max_sectors_kb() {
9961 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
9962 local max_sectors_path="/sys/block/${dev_base}/queue/max_sectors_kb"
9964 do_facet ${facet} "[[ -e ${max_sectors_path} ]] &&
9965 echo ${value} > ${max_sectors_path}"
9968 [[ $rc -ne 0 ]] && echo "Failed to set ${max_sectors_path} to ${value}"
9973 # Return 0 if all slave devices have max_sectors_kb == max_hw_sectors_kb
9974 # Otherwise return > 0
9975 check_slaves_max_sectors_kb()
9979 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
9980 local slaves_dir=/sys/block/${dev_base}/slaves
9981 local slave_devices=$(do_facet ${facet} "ls ${slaves_dir} 2>/dev/null")
9982 [[ -z ${slave_devices} ]] && return 0
9984 local slave max_sectors new_max_sectors max_hw_sectors path
9987 for slave in ${slave_devices}; do
9988 path="/dev/${slave}"
9989 ! is_blkdev ${facet} ${path} && continue
9990 max_sectors=$(get_max_sectors_kb ${facet} ${path})
9991 max_hw_sectors=$(get_max_hw_sectors_kb ${facet} ${path})
9992 new_max_sectors=${max_hw_sectors}
9993 [[ ${new_max_sectors} -gt ${RQ_SIZE_LIMIT} ]] &&
9994 new_max_sectors=${RQ_SIZE_LIMIT}
9996 if [[ ${max_sectors} -ne ${new_max_sectors} ]]; then
9997 echo "${path} ${max_sectors} ${new_max_sectors}"
10000 check_slaves_max_sectors_kb ${facet} ${path}
10009 local facet_list="mgs mds1 ost1"
10010 combined_mgs_mds && facet_list="mgs ost1"
10013 for facet in ${facet_list}; do
10014 [[ $(facet_fstype ${facet}) != ldiskfs ]] &&
10015 skip "ldiskfs only test" &&
10017 ! is_blkdev ${facet} $(facet_device ${facet}) &&
10018 skip "requires all real devices" &&
10023 # We don't increase IO request size limit past 16MB. See comments in
10024 # lustre/utils/libmount_utils_ldiskfs.c:tune_max_sectors_kb()
10025 RQ_SIZE_LIMIT=$((16 * 1024))
10026 local device old_max_sectors new_max_sectors max_hw_sectors
10027 for facet in ${facet_list}; do
10028 device=$(facet_device ${facet})
10029 old_max_sectors=$(get_max_sectors_kb ${facet} ${device})
10030 max_hw_sectors=$(get_max_hw_sectors_kb ${facet} ${device})
10032 # The expected value after l_tunedisk is executed
10033 new_max_sectors=$old_max_sectors
10034 [[ ${new_max_sectors_kb} -gt ${RQ_SIZE_LIMIT} ]] &&
10035 new_max_sectors_kb=${RQ_SIZE_LIMIT}
10037 # Ensure the current value of max_sectors_kb does not equal
10038 # max_hw_sectors_kb, so we can tell whether l_tunedisk did
10040 set_max_sectors_kb ${facet} ${device} $((new_max_sectors - 1))
10042 # Value before l_tunedisk
10043 local pre_max_sectors=$(get_max_sectors_kb ${facet} ${device})
10044 if [[ ${pre_max_sectors} -ne $((new_max_sectors - 1)) ]]; then
10045 echo "unable to satsify test pre-condition:"
10046 echo "${pre_max_sectors} != $((new_max_sectors - 1))"
10051 echo "Before: ${facet} ${device} ${pre_max_sectors} ${max_hw_sectors}"
10053 do_facet ${facet} "l_tunedisk ${device}"
10055 # Value after l_tunedisk
10056 local post_max_sectors=$(get_max_sectors_kb ${facet} ${device})
10058 echo "After: ${facet} ${device} ${post_max_sectors} ${max_hw_sectors}"
10060 if [[ ${facet} != ost1 ]]; then
10061 if [[ ${post_max_sectors} -ne ${pre_max_sectors} ]]; then
10062 echo "l_tunedisk modified max_sectors_kb of ${facet}"
10066 set_max_sectors_kb ${facet} ${device} ${old_max_sectors}
10068 if [[ ${post_max_sectors} -eq ${pre_max_sectors} ]]; then
10069 echo "l_tunedisk failed to modify max_sectors_kb of ${facet}"
10073 check_slaves_max_sectors_kb ${facet} ${device} ||
10080 run_test 125 "check l_tunedisk only tunes OSTs and their slave devices"
10083 [[ "$MDS1_VERSION" -ge $(version_code 2.13.52) ]] ||
10084 skip "Need MDS version at least 2.13.52"
10087 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ../libcfs/libcfs/libcfs
10088 #define OBD_FAIL_OBD_SETUP 0x60d
10089 do_facet mds1 $LCTL set_param fail_loc=0x60d
10090 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_modules &
10091 for i in {1..40}; do
10092 do_facet mds1 lsmod | grep -q osd_$mds1_FSTYPE && break
10095 clear_failloc $SINGLEMDS 20 &
10096 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
10098 run_test 126 "mount in parallel shouldn't cause a crash"
10101 [[ "$ost1_FSTYPE" == ldiskfs ]] || skip "ldiskfs only test"
10105 zconf_umount_clients $RCLIENTS $MOUNT
10107 wait_osp_active ost ${FSNAME}-OST0000 0 1
10108 local osc_tgt="$FSNAME-OST0000-osc-$($LFS getname -i $DIR)"
10109 local avail1=($($LCTL get_param -n osc.${osc_tgt}.kbytesavail))
10111 $LFS setstripe -i 0 $DIR/$tfile || error "failed creating $DIR/$tfile"
10112 dd if=/dev/zero of=$DIR/$tfile bs=1M oflag=direct || true
10114 local avail2=($($LCTL get_param -n osc.${osc_tgt}.kbytesavail))
10116 if ((avail2 * 100 / avail1 > 1)); then
10119 error "more than 1% space left: before=$avail1 after=$avail2"
10122 local mbs=$(($(stat -c %s $DIR/$tfile) / (1024 * 1024)))
10124 dd if=/dev/zero of=$DIR/$tfile bs=1M count=$mbs conv=notrunc \
10125 oflag=direct || error "overwrite failed"
10127 run_test 127 "direct io overwrite on full ost"
10131 combined_mgs_mds && skip "need separate mgs device"
10132 [ "$ost2_FSTYPE" == zfs ] && import_zpool ost2
10135 # Try to apply nolocallogs to the virgin OST. Should fail.
10136 do_facet ost2 "$TUNEFS --nolocallogs $(ostdevname 2)" &&
10137 error "nolocallogs should not be allowed on the virgin target"
10142 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
10143 # Start OST without MGS (local configs)
10144 do_facet ost1 "$TUNEFS --dryrun $(ostdevname 1)"
10145 start_ost || error "unable to start OST1"
10146 stop_ost || error "Unable to stop OST1"
10148 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
10149 # Do not allow reading local configs, should fail
10150 do_facet ost1 "$TUNEFS --nolocallogs $(ostdevname 1)" ||
10151 error "Can not set nolocallogs"
10152 start_ost && error "OST1 started, but should fail"
10154 # Connect to MGS successfully, reset nolocallogs flag
10155 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
10156 start_mgs || error "unable to start MGS"
10157 start_ost || error "unable to start OST1"
10159 do_facet ost1 "$TUNEFS --dryrun $(ostdevname 1)" | grep "nolocallogs" &&
10160 error "nolocallogs expected to be reset"
10162 stop_ost || error "Unable to stop OST1"
10164 run_test 128 "Force using remote logs with --nolocallogs"
10169 start_mgsmds || error "MGS/MDS start failed"
10171 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &&
10172 error "start ost1 should fail" || true
10173 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &&
10174 error "second start ost1 should fail" || true
10175 do_facet ost1 "$TUNEFS --writeconf $(ostdevname 1)"
10176 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS ||
10177 error "start ost1 failed"
10181 run_test 129 "attempt to connect an OST with the same index should fail"
10185 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
10187 stop_mdt 2 || error "mdt2 stop failed"
10188 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)"
10189 start_mdt 2 || error "mdt2 start failed"
10190 do_facet mds2 "$LCTL dl" | grep MDT0001-osp-MDT0001 &&
10191 error "Illegal OSP device created" || true
10193 run_test 130 "re-register an MDT after writeconf"
10196 [ "$mds1_FSTYPE" == "ldiskfs" ] || skip "ldiskfs only test"
10197 (( $MDS1_VERSION >= $(version_code 2.14.56.35) )) ||
10198 skip "Need MDS version at least 2.14.56.35"
10199 do_facet mds1 $DEBUGFS -R features $(mdsdevname 1) |
10200 grep -q project || skip "skip project quota not supported"
10205 test_mkdir -c $MDSCOUNT -p $DIR/$tdir
10206 $LFS project -p 1000 $DIR/$tdir || error "set dir project id failed"
10207 createmany -o $DIR/$tdir/f 512
10208 for ((i = 0; i < 512; ++i)); do
10209 $LFS project -p $i $DIR/$tdir/f${i} ||
10210 error "set f${i} project id failed"
10213 test_mkdir -c $MDSCOUNT -p $DIR/$tdir.inherit
10214 $LFS project -p 1001 -s $DIR/$tdir.inherit
10215 createmany -o $DIR/$tdir.inherit/f 128
10216 (( $($LFS project $DIR/$tdir.inherit/f* |
10217 awk '$1 == 1001 { print }' | wc -l) == 128 )) ||
10218 error "files did not inherit projid 1001"
10222 for i in $(seq $MDSCOUNT); do
10223 mds_backup_restore mds$i ||
10224 error "Backup/restore on mds$i failed"
10229 projid=($($LFS project -d $DIR/$tdir))
10230 [ ${projid[0]} == "1000" ] ||
10231 error "projid expected 1000 not ${projid[0]}"
10232 for ((i = 0; i < 512; ++i)); do
10233 projid=($($LFS project $DIR/$tdir/f${i}))
10234 [ ${projid[0]} == "$i" ] ||
10235 error "projid expected $i not ${projid[0]}"
10238 (( $($LFS project $DIR/$tdir.inherit/f* |
10239 awk '$1 == 1001 { print }' | wc -l) == 128 )) ||
10240 error "restore did not copy projid 1001"
10242 run_test 131 "MDT backup restore with project ID"
10249 combined_mgs_mds || start_mgs || error "unable to start MGS"
10250 start_mdt 1 || error "unable to start mdt1"
10252 err_cnt=$(do_facet mds1 dmesg | grep -c "cannot take the layout locks")
10253 stop_mdt 1 || error "stop mdt1 failed"
10255 [ "$mds1_FSTYPE" == zfs ] && import_zpool mds1
10256 do_facet mds1 $TUNEFS --param mdt.hsm_control=enabled $(mdsdevname 1) ||
10257 error "tunefs failed"
10258 start_mdt 1 || error "cannot start mdt1"
10260 err_cnt2=$(do_facet mds1 dmesg | grep -c "cannot take the layout locks")
10261 [ $err_cnt -eq $err_cnt2 ] || error "Can not take the layout lock"
10262 stop_mdt 1 || error "stop mdt1 failed"
10264 run_test 132 "hsm_actions processed after failover"
10266 # This test verifies we do RR allocation within a pool even if there is a
10267 # significant imbalance vs an OST outside the pool
10269 [[ $OSTCOUNT -lt 4 ]] && skip_env "needs >= 4 OSTs"
10270 [[ "$OST1_VERSION" -ge $(version_code 2.15.51) ]] ||
10271 skip "Need OST version at least 2.15.51"
10272 # This is the easiest way to ensure OSTs start out balanced
10273 reformat_and_config
10276 check_set_fallocate_or_skip
10278 local testfile=$DIR/$tdir/$tfile
10279 local pool="testpool"
10280 local ostrange=$((OSTCOUNT - 1))
10281 # Select all but the last OST to add to the pool
10282 local poolostrange=$((OSTCOUNT - 2))
10290 [ $MINV -eq 0 ] && error "no free space in OST$MINI"
10291 [ $MAXV -gt $((2 * $MINV)) ] &&
10292 error "OSTs badly unbalanced after reformat"
10294 create_pool $FSNAME.$pool || error "failed to create a pool"
10295 do_facet mgs $LCTL pool_add $FSNAME.$pool OST[0-$poolostrange] ||
10296 error "failed to add OST[0-$poolostrange] to the pool"
10298 test_mkdir -p $DIR/$tdir || error "failed to mkdir $DIR/$tdir"
10299 # Consume space on the OSTs in the pool so they are unbalanced with the
10300 # OST outside of the pool
10301 # fill each OST 90% with fallocate so they are widely
10303 local size=$(((MINV * 9 / 10) * 1024))
10304 for ((i = 0; i <= poolostrange; i++)); do
10305 $LFS setstripe -c 1 -i $i $testfile$i ||
10306 error "failed to setstripe $testfile$i"
10307 fallocate -l $size $testfile$i || error "fallocate failed"
10313 # Create files in the pool now that there is an imbalance
10314 filecount=$(((OSTCOUNT - 1) * filenum))
10315 for ((i = 0; i < filecount; i++)); do
10316 $LFS setstripe -p $pool $testfile-$i ||
10317 error "failed to setstripe -p $pool $testfile-$i"
10319 $LFS getstripe -i $testfile-* > /tmp/$tfile.log
10320 # Count the number of files with a stripe on each OST to verify the
10321 # pool allocated with round-robin
10322 for ((i = 0; i <= poolostrange; i++)); do
10323 stripecount=$(grep -c $i /tmp/$tfile.log)
10324 # Allow a little leeway
10325 if (( stripecount < filenum - 1 ||
10326 stripecount > filenum + 1 )); then
10327 cat /tmp/$tfile.log
10328 error "$stripecount != $filenum files on OST$i"
10332 # Create files across the system now that there is an imbalance
10333 filecount=$((OSTCOUNT * filenum))
10334 for ((i = 1; i < filecount; i++)); do
10335 $LFS setstripe $testfile-$i.2 ||
10336 error "failed to setstripe $testilfe-$i.2"
10338 $LFS getstripe -i $testfile-*.2 > /tmp/$tfile.log
10340 # Count the number of files with a stripe on each OST to verify the
10341 # files are *NOT* allocated with round-robin
10342 for ((i = 0; i <= ostrange; i++)); do
10343 stripecount=$(grep -c $i /tmp/$tfile.log)
10344 if [[ $stripecount -ne $filenum ]]; then
10346 echo "QOS: $stripecount != $filenum files on OST$i"
10349 if [ -z "$qos_used" ]; then
10350 error "QOS not used on imbalanced OSTs!"
10353 rm -rf /tmp/$tfile.log $DIR/$tdir
10354 do_facet mgs $LCTL pool_remove $FSNAME.$pool OST[0-$poolostrange] ||
10355 "failed to remove OST[0-$poolostrange] from the pool"
10356 do_facet mgs $LCTL pool_destroy $FSNAME.$pool ||
10357 error "failed to destroy pool"
10359 run_test 133 "stripe QOS: free space balance in a pool"
10362 [ "$mds1_FSTYPE" == "ldiskfs" ] || skip "ldiskfs only test"
10365 local mdt_dev=$(facet_device mds1)
10366 local tmp_dir=$TMP/$tdir
10367 local out=$tmp_dir/check_iam.txt
10368 local CHECK_IAM=${CHECK_IAM:-$(do_facet mds1 "which check_iam 2> /dev/null || true")}
10370 [[ -n "$CHECK_IAM" ]] || skip "check_iam not found"
10373 for ((i=0; i<64; i++)); do
10375 #cmd introduce a random corruption to IAM file
10376 local cmd="dd if=/dev/urandom of=$tmp_dir/$f bs=2 conv=notrunc count=1 seek=$((RANDOM % 36))"
10377 do_facet mds1 "mkdir -p $tmp_dir; \
10378 $DEBUGFS -c -R 'dump $f $tmp_dir/$f' $mdt_dev 2>&1; \
10379 $CHECK_IAM -v $tmp_dir/$f 2>&1; \
10380 echo $cmd; eval $cmd 2>/dev/null;
10381 $CHECK_IAM -v $tmp_dir/$f 2>&1; echo \\\$?" >> $out
10386 stack_trap "rm -rf $tmp_dir && do_facet mds1 rm -rf $tmp_dir" EXIT
10388 rc=$(grep -c "fault\|except" $out)
10389 (( rc == 0 )) || { cat $out &&
10390 error "check_iam failed with fault or exception $rc"; }
10392 rc=$(grep -c "^255" $out)
10393 error=$(grep -c "FINISHED WITH ERRORS" $out)
10395 (( rc == error )) || { cat $out &&
10396 error "check_iam errcode does not fit with errors $rc $error"; }
10398 run_test 134 "check_iam works without faults"
10401 # (This was sanity/802a)
10403 saved_MGS_MOUNT_OPTS=$MGS_MOUNT_OPTS
10404 saved_MDS_MOUNT_OPTS=$MDS_MOUNT_OPTS
10405 saved_OST_MOUNT_OPTS=$OST_MOUNT_OPTS
10406 saved_MOUNT_OPTS=$MOUNT_OPTS
10410 MGS_MOUNT_OPTS=$saved_MGS_MOUNT_OPTS
10411 MDS_MOUNT_OPTS=$saved_MDS_MOUNT_OPTS
10412 OST_MOUNT_OPTS=$saved_OST_MOUNT_OPTS
10413 MOUNT_OPTS=$saved_MOUNT_OPTS
10418 [[ $mds1_FSTYPE = zfs ]] || skip "ZFS specific test"
10419 [[ $MDS1_VERSION -lt $(version_code 2.9.55) ]] ||
10420 [[ $OST1_VERSION -lt $(version_code 2.9.55) ]] &&
10421 skip "Need server version at least 2.9.55"
10423 [[ $ENABLE_QUOTA ]] && skip "Quota enabled for read-only test"
10425 # Reset before starting
10429 mkdir $DIR/$tdir || error "(1) fail to mkdir"
10431 cp $LUSTRE/tests/test-framework.sh $DIR/$tdir/ ||
10432 error "(2) Fail to copy"
10434 stack_trap cleanup_802a EXIT
10436 # sync by force before remount as readonly
10437 sync; sync_all_data; sleep 3; sync_all_data
10441 MGS_MOUNT_OPTS=$(csa_add "$MGS_MOUNT_OPTS" -o rdonly_dev)
10442 MDS_MOUNT_OPTS=$(csa_add "$MDS_MOUNT_OPTS" -o rdonly_dev)
10443 OST_MOUNT_OPTS=$(csa_add "$OST_MOUNT_OPTS" -o rdonly_dev)
10445 echo "Mount the server as read only"
10446 setupall server_only || error "(3) Fail to start servers"
10448 echo "Mount client without ro should fail"
10449 mount_client $MOUNT &&
10450 error "(4) Mount client without 'ro' should fail"
10452 echo "Mount client with ro should succeed"
10453 MOUNT_OPTS=$(csa_add "$MOUNT_OPTS" -o ro)
10454 mount_client $MOUNT ||
10455 error "(5) Mount client with 'ro' should succeed"
10457 echo "Modify should be refused"
10458 touch $DIR/$tdir/guard && error "(6) Touch should fail under ro mode"
10460 echo "Read should be allowed"
10461 diff $LUSTRE/tests/test-framework.sh $DIR/$tdir/test-framework.sh ||
10462 error "(7) Read should succeed under ro mode"
10464 run_test 802a "simulate readonly device"
10466 if ! combined_mgs_mds ; then
10472 # restore the values of MDSSIZE and OSTSIZE
10473 MDSSIZE=$STORED_MDSSIZE
10474 OSTSIZE=$STORED_OSTSIZE
10478 check_and_cleanup_lustre