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 $tmp/src/flr_dir
1527 local large_xattr_dir=/mnt/$FSNAME/large_xattr_test_dir
1528 local xattr_file=$large_xattr_dir/large_xattr_file
1530 mkdir $large_xattr_dir
1532 setfattr -n user.fooattr -v $(printf "%c" {1..4096} ) $xattr_file ||
1537 mkdir $tmp/img || return 1
1541 pushd_dir=/mnt/$FSNAME
1542 if [[ $MDSCOUNT -ge 2 ]]; then
1543 pushd_dir=$remote_dir
1544 if [[ "$MDS1_VERSION" -ge $(version_code 2.7.0) ]]; then
1546 BLOCKSIZE=1024 ls -Rni --time-style=+%s >$tmp/img/list2
1552 BLOCKSIZE=1024 ls -Rni --time-style=+%s >$tmp/img/list
1553 find ! -name .lustre -type f -exec sha1sum {} \; |
1554 sort -k 2 >$tmp/img/sha1sums
1556 $LCTL get_param -n version | head -n 1 |
1557 sed -e 's/^lustre: *//' >$tmp/img/commit
1559 [[ "$MDS1_VERSION" -ge $(version_code 2.3.50) ]] ||
1560 $LFS quotaon -ug /mnt/$FSNAME
1561 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1562 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1563 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1564 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1565 | tr -d "*" > $tmp/img/bspace
1566 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1567 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1568 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1569 | tr -d "*" > $tmp/img/ispace
1570 echo $T32_BLIMIT > $tmp/img/blimit
1571 echo $T32_ILIMIT > $tmp/img/ilimit
1573 $MULTIOP /mnt/$FSNAME/orph_file Ouw_c&
1577 debugfs -R "ls /PENDING" ${MDSDEV1:-$MDSDEV}
1578 cp ${MDSDEV1:-$MDSDEV} $tmp/img
1587 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1590 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1591 echo "Data verification failed"
1594 uname -r >$tmp/img/kernel
1595 uname -m >$tmp/img/arch
1597 for ((num=1; num <= $MDSCOUNT; num++)); do
1598 local devname=$(mdsdevname $num)
1602 [[ num -eq 1 ]] && image_name=mdt || image_name=mdt$num
1603 [[ $(facet_fstype $facet) != zfs ]] ||
1604 devname=$(mdsvdevname $num)
1605 dd conv=sparse bs=4k if=$devname of=$tmp/img/$image_name
1608 for ((num=1; num <= $OSTCOUNT; num++)); do
1609 local devname=$(ostdevname $num)
1613 [[ num -eq 1 ]] && image_name=ost || image_name=ost$num
1614 [[ $(facet_fstype $facet) != zfs ]] ||
1615 devname=$(ostdevname $num)
1616 dd conv=sparse bs=4k if=$devname of=$tmp/img/$image_name
1619 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1620 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1623 tar cjvf $dst/disk$version-"$mds1_FSTYPE".tar.bz2 -S *
1628 # run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1631 # The list of applicable tarballs is returned via the caller's
1632 # variable "tarballs".
1635 [ "$CLIENTONLY" ] && skip "Client-only testing"
1637 local node=$(facet_active_host $SINGLEMDS)
1638 local r="do_node $node"
1640 ! $r which "$TUNEFS" && skip_env "tunefs.lustre required on $node"
1642 local IMGTYPE="$mds1_FSTYPE"
1644 [[ -n "$T32IMAGE" ]] && tarballs=$T32IMAGE ||
1645 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 \
1646 -name \'disk*-$IMGTYPE.tar.bz2\')
1648 [ -z "$tarballs" ] && skip "No applicable tarballs found"
1651 t32_test_cleanup() {
1653 local facet=$SINGLEMDS
1656 echo "== cleanup with rc=$rc =="
1657 if $shall_cleanup_lustre; then
1658 umount $tmp/mnt/lustre || rc=$?
1660 if $shall_cleanup_mdt; then
1661 $r $UMOUNT $tmp/mnt/mdt || rc=$?
1663 if $shall_cleanup_mdt1; then
1664 $r $UMOUNT $tmp/mnt/mdt1 || rc=$?
1666 if $shall_cleanup_ost; then
1667 $r $UMOUNT $tmp/mnt/ost || rc=$?
1669 if $shall_cleanup_ost1; then
1670 $r $UMOUNT $tmp/mnt/ost1 || rc=$?
1675 if [[ "$mds1_FSTYPE" == zfs ]]; then
1677 local poolname_list="t32fs-mdt1 t32fs-ost1"
1679 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1681 for poolname in $poolname_list; do
1682 destroy_zpool $facet $poolname
1685 combined_mgs_mds || start_mgs || rc=$?
1689 t32_bits_per_long() {
1691 # Yes, this is not meant to be perfect.
1701 t32_reload_modules() {
1703 local all_removed=false
1706 [ "$mds1_FSTYPE" == zfs ] && do_rpc_nodes $node "service zed stop"
1708 while ((i < 20)); do
1709 echo "Unloading modules on $node: Attempt $i"
1710 do_rpc_nodes $node $LUSTRE_RMMOD "$mds1_FSTYPE" &&
1712 do_rpc_nodes $node check_mem_leak || return 1
1713 if $all_removed; then
1714 do_rpc_nodes $node load_modules
1717 if [ "$mds1_FSTYPE" == zfs ]; then
1718 do_rpc_nodes $node "$ZPOOL status -v"
1723 echo "Unloading modules on $node: Given up"
1727 t32_wait_til_devices_gone() {
1733 echo wait for devices to go
1734 while ((i < 20)); do
1735 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1736 loops=$(do_rpc_nodes $node losetup -a | grep -c t32)
1737 ((devices == 0 && loops == 0)) && return 0
1741 echo "waiting for dev on $node: dev $devices loop $loops given up"
1742 do_rpc_nodes $node "losetup -a"
1743 do_rpc_nodes $node "$LCTL device_list"
1747 t32_verify_quota() {
1754 # LU-2435: if the underlying zfs doesn't support userobj_accounting,
1755 # lustre will estimate the object count usage. This fails quota
1756 # verification in 32b. The object quota usage should be accurate after
1757 # zfs-0.7.0 is released.
1758 [ "$mds1_FSTYPE" == zfs ] && {
1759 local zfs_version=$(do_facet $facet cat /sys/module/zfs/version)
1761 [ $(version_code $zfs_version) -lt $(version_code 0.7.0) ] && {
1762 echo "Skip quota verify for zfs: $zfs_version"
1767 $LFS quota -u $T32_QID -v $mnt
1769 qval=$($LFS quota -v -u $T32_QID $mnt |
1770 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1771 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1773 [ $qval -eq $img_bspace ] || {
1774 echo "bspace, act:$qval, exp:$img_bspace"
1778 qval=$($LFS quota -v -u $T32_QID $mnt |
1779 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1780 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1782 [ $qval -eq $img_ispace ] || {
1783 echo "ispace, act:$qval, exp:$img_ispace"
1787 qval=$($LFS quota -v -u $T32_QID $mnt |
1788 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1789 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1791 [ $qval -eq $img_blimit ] || {
1792 echo "blimit, act:$qval, exp:$img_blimit"
1796 qval=$($LFS quota -v -u $T32_QID $mnt |
1797 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1798 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1800 [ $qval -eq $img_ilimit ] || {
1801 echo "ilimit, act:$qval, exp:$img_ilimit"
1805 set_persistent_param_and_check $facet \
1806 "osd-$mds1_FSTYPE.$fsname-MDT0000.quota_slave.enabled" \
1807 "$fsname.quota.mdt" ug
1809 set_persistent_param_and_check $facet \
1810 "osd-$mds1_FSTYPE.$fsname-OST0000.quota_slave.enabled" \
1811 "$fsname.quota.ost" ug
1814 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1815 bs=1M count=$((img_blimit / 1024)) oflag=sync && {
1816 echo "Write succeed, but expect -EDQUOT"
1819 rm -f $mnt/t32_qf_new
1821 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1823 echo "Create succeed, but expect -EDQUOT"
1826 unlinkmany $mnt/t32_qf_ $img_ilimit
1836 sync_all_data > /dev/null 2>&1 || true
1838 [ "$uuid" != "global" ] || uuid=$mnt
1840 $LFS quota -v "$1" "$2" $mnt |
1841 awk 'BEGIN { num='$spec' } { if ($1 == "'$uuid'") \
1842 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1849 local dne_upgrade=${dne_upgrade:-"no"}
1850 local dom_upgrade=${dom_upgrade:-"no"}
1851 local project_upgrade=${project_upgrade:-"no"}
1852 local ff_convert=${ff_convert:-"no"}
1853 local pfl_upgrade=${pfl_upgrade:-"no"}
1854 local project_quota_upgrade=${project_quota_upgrade:-"no"}
1855 local dom_new_upgrade=${dom_new_upgrade:-"no"}
1856 local flr_upgrade=${flr_upgrade:-"no"}
1857 local shall_cleanup_mdt=false
1858 local shall_cleanup_mdt1=false
1859 local shall_cleanup_ost=false
1860 local shall_cleanup_ost1=false
1861 local shall_cleanup_lustre=false
1862 local mdt2_is_available=false
1863 local ost2_is_available=false
1864 local node=$(facet_active_host $SINGLEMDS)
1865 local r="do_node $node"
1881 local mdt_dev=$tmp/mdt
1882 local mdt2_dev=$tmp/mdt2
1883 local ost_dev=$tmp/ost
1884 local ost2_dev=$tmp/ost2
1888 local pfl_file=$tmp/mnt/lustre/remote_dir/pfl_dir/pfl_file
1889 local flr_file=$tmp/mnt/lustre/remote_dir/flr_dir/flr_file
1890 local dom_file=$tmp/mnt/lustre/remote_dir/dom_dir/dom_file
1891 local quota_dir=$tmp/mnt/lustre/remote_dir/project_quota_dir
1893 combined_mgs_mds || stop_mgs || error "Unable to stop MGS"
1894 trap 'trap - RETURN; t32_test_cleanup' RETURN
1897 nid=$($r $LCTL list_nids | head -1)
1899 mkdir -p $tmp/mnt/lustre || error "mkdir $tmp/mnt/lustre failed"
1900 $r mkdir -p $tmp/mnt/{mdt,mdt1,ost,ost1}
1901 $r tar xjvf $tarball -S -C $tmp || {
1902 error_noexit "Unpacking the disk image tarball"
1905 img_commit=$($r cat $tmp/commit)
1906 img_kernel=$($r cat $tmp/kernel)
1907 img_arch=$($r cat $tmp/arch)
1908 img_bspace=$($r cat $tmp/bspace)
1909 img_ispace=$($r cat $tmp/ispace)
1911 # older images did not have "blimit" and "ilimit" files
1912 # use old values for T32_BLIMIT and T32_ILIMIT
1913 $r test -f $tmp/blimit && img_blimit=$($r cat $tmp/blimit) ||
1915 $r test -f $tmp/ilimit && img_ilimit=$($r cat $tmp/ilimit) ||
1918 echo "Upgrading from $(basename $tarball), created with:"
1919 echo " Commit: $img_commit"
1920 echo " Kernel: $img_kernel"
1921 echo " Arch: $img_arch"
1922 echo "OST version: $(lustre_build_version ost1)"
1924 # The conversion can be made only when both of the following
1925 # conditions are satisfied:
1926 # - ost device img version < 2.3.64
1927 # - ost server version >= 2.5
1928 [ $(version_code $img_commit) -ge $(version_code 2.3.64) -o \
1929 "$OST1_VERSION" -lt $(version_code 2.5.0) ] &&
1932 ! $r test -f $mdt2_dev || mdt2_is_available=true
1933 ! $r test -f $ost2_dev || ost2_is_available=true
1935 if [[ "$mds1_FSTYPE" == zfs ]]; then
1936 echo "== fstype is zfs =="
1939 local poolname_list="t32fs-mdt1 t32fs-ost1"
1941 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1943 for poolname in $poolname_list; do
1945 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
1946 $ZPOOL import -f -d $tmp $poolname"
1949 # upgrade zpool to latest supported features, including
1950 # dnode quota accounting in 0.7.0
1951 $r "$ZPOOL upgrade -a"
1953 mdt_dev=t32fs-mdt1/mdt1
1954 ost_dev=t32fs-ost1/ost1
1955 ! $mdt2_is_available || mdt2_dev=t32fs-mdt2/mdt2
1956 ! $ost2_is_available || ost2_dev=t32fs-ost2/ost2
1957 wait_update_facet $SINGLEMDS "$ZPOOL list |
1958 awk '/^t32fs-mdt1/ { print \\\$1 }'" "t32fs-mdt1" || {
1959 error_noexit "import zfs pool failed"
1962 elif [ "$project_upgrade" != "no" ]; then
1963 echo "== project upgrade =="
1964 ! $r tune2fs -O project $mdt_dev &&
1965 error_noexit "enable project on mdt0 failed" &&
1967 $mdt2_is_available && ! $r tune2fs "-O project" $mdt2_dev &&
1968 error_noexit "enable project on mdt failed" &&
1970 ! $r tune2fs -O project $ost_dev &&
1971 error_noexit "enable project on ost failed" &&
1973 $ost2_is_available && ! $r tune2fs -O project $ost2_dev &&
1974 error_noexit "enable project on ost2 failed" &&
1978 $r $LCTL set_param debug="$PTLDEBUG"
1980 $r $TUNEFS --dryrun $mdt_dev || {
1982 error_noexit "tunefs.lustre before mounting the MDT"
1986 if $mdt2_is_available; then
1987 $r $TUNEFS --dryrun $mdt2_dev || {
1989 error_noexit "tunefs.lustre before mounting the MDT"
1994 if [ "$writeconf" ]; then
1995 echo "== writeconf mdt tunefs and quota =="
1997 if [ "$mds1_FSTYPE" == ldiskfs ]; then
1999 $r $TUNEFS --quota $mdt_dev || {
2001 error_noexit "Enable mdt quota feature"
2005 if [ "$project_quota_upgrade" != "no" ]; then
2006 echo "== enable mdt2 project quota =="
2007 $r $TUNE2FS -O project $mdt_dev || {
2009 error_noexit "tune2fs $mdt_dev failed"
2014 if $mdt2_is_available; then
2015 $r $TUNEFS --quota $mdt2_dev || {
2017 error_noexit "Enable mdt2 quota feature"
2020 if [ "$project_quota_upgrade" != "no" ]; then
2021 echo "== enable mdt2 project quota =="
2022 $r $TUNE2FS -O project $mdt2_dev || {
2025 "tune2fs $mdt2_dev failed"
2032 echo "== replace nids =="
2033 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
2034 [[ "$MGS_VERSION" -ge $(version_code 2.3.59) ]] ||
2035 skip "LU-2200: Cannot run over IB w/o lctl replace_nids "
2036 "(Need MGS version at least 2.3.59)"
2039 local osthost=$(facet_active_host ost1)
2040 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
2043 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2046 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt
2047 $r $LCTL replace_nids $fsname-OST0000 $ostnid
2048 if $ost2_is_available; then
2049 $r $LCTL replace_nids $fsname-OST0001 $ostnid
2051 $r $LCTL replace_nids $fsname-MDT0000 $nid
2052 if $mdt2_is_available; then
2053 $r $LCTL replace_nids $fsname-MDT0001 $nid
2055 $r $UMOUNT $tmp/mnt/mdt
2057 mopts=exclude=$fsname-OST0000
2058 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2063 t32_wait_til_devices_gone $node
2065 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2067 error_noexit "Mounting the MDT"
2070 shall_cleanup_mdt=true
2072 if $mdt2_is_available; then
2073 echo "== mdt2 available =="
2074 mopts=mgsnode=$nid,$mopts
2075 $r $MOUNT_CMD -o $mopts $mdt2_dev $tmp/mnt/mdt1 || {
2077 error_noexit "Mounting the MDT"
2080 shall_cleanup_mdt1=true
2081 echo "== mount new MDT $mdt2_dev =="
2083 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 || {
2084 error_noexit "enable remote dir create failed"
2088 elif [ "$dne_upgrade" != "no" ]; then
2089 local fs2mdsdev=$(mdsdevname 1_2)
2090 local fs2mdsvdev=$(mdsvdevname 1_2)
2092 echo "== mkfs new MDT on ${fs2mdsdev} =="
2093 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2094 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
2097 add $SINGLEMDS $(mds2failover_HOST="" \
2098 mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
2099 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
2100 error_noexit "Mkfs new MDT failed"
2104 [[ "$mds1_FSTYPE" != zfs ]] || import_zpool mds1
2106 $r $TUNEFS --dryrun $fs2mdsdev || {
2107 error_noexit "tunefs.lustre before mounting the MDT"
2111 echo "== mount new MDT....$fs2mdsdev =="
2112 $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
2113 error_noexit "mount mdt1 failed"
2117 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
2118 error_noexit "enable remote dir create failed"
2120 shall_cleanup_mdt1=true
2123 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
2124 error_noexit "Getting MDT UUID"
2127 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
2128 error_noexit "Unexpected MDT UUID: \"$uuid\""
2132 $r $TUNEFS --dryrun $ost_dev || {
2133 error_noexit "tunefs.lustre before mounting the OST"
2137 if $ost2_is_available; then
2138 $r $TUNEFS --dryrun $ost2_dev || {
2139 error_noexit "tunefs.lustre before mounting the OST"
2144 if [ "$writeconf" ]; then
2145 echo "== writeconf and ost quota tunefs =="
2146 mopts=mgsnode=$nid,$writeconf
2147 if [ "$ost1_FSTYPE" == ldiskfs ]; then
2149 $r $TUNEFS --quota $ost_dev || {
2151 error_noexit "Enable ost quota feature"
2155 if [ "$project_quota_upgrade" != "no" ]; then
2156 echo "== enable ost project quota =="
2157 $r $TUNE2FS -O project $ost_dev || {
2159 error_noexit "tune2fs $ost_dev failed"
2164 if $ost2_is_available; then
2165 $r $TUNEFS --quota $ost2_dev || {
2167 error_noexit "Enable ost2 quota feature"
2170 if [ "$project_quota_upgrade" != "no" ]; then
2171 echo "== enable ost2 project quota =="
2172 $r $TUNE2FS -O project $ost2_dev || {
2175 "tune2fs $ost2_dev failed"
2183 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2188 $r $MOUNT_CMD -onomgs -o$mopts $ost_dev $tmp/mnt/ost || {
2189 error_noexit "Mounting the OST"
2193 if $ost2_is_available; then
2194 $r $MOUNT_CMD -onomgs -o$mopts $ost2_dev $tmp/mnt/ost1 || {
2195 error_noexit "Mounting the OST2"
2198 shall_cleanup_ost1=true
2201 shall_cleanup_ost=true
2203 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
2204 error_noexit "Getting OST UUID"
2207 if [ "$uuid" != $fsname-OST0000_UUID ]; then
2208 error_noexit "Unexpected OST UUID: \"$uuid\""
2212 if $ost2_is_available; then
2213 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0001.uuid) ||{
2214 error_noexit "Getting OST1 UUID"
2217 if [ "$uuid" != $fsname-OST0001_UUID ]; then
2218 error_noexit "Unexpected OST1 UUID: \"$uuid\""
2223 if [[ $PERM_CMD =~ "set_param" ]]; then
2224 echo "== perm cmd = '$PERM_CMD' =="
2225 $r $PERM_CMD osc.$fsname-OST0000*.import=connection=$nid || {
2226 error_noexit "Setting OST \"failover.node\""
2229 $r $PERM_CMD mdc.$fsname-MDT0000*.import=connection=$nid || {
2230 error_noexit "Setting MDT \"failover.node\""
2233 $r $PERM_CMD osc.$fsname-OST0000-*.max_dirty_mb=15 || {
2234 error_noexit "Setting \"max_dirty_mb\""
2237 $r $PERM_CMD mdc.$fsname-MDT0000-*.max_rpcs_in_flight=9 || {
2238 error_noexit "Setting \"max_rpcs_in_flight\""
2241 $r $PERM_CMD lov.$fsname-MDT0000-*.stripesize=4M || {
2242 error_noexit "Setting \"lov.stripesize\""
2245 $r $PERM_CMD mdd.$fsname-MDT0000-*.atime_diff=70 || {
2246 error_noexit "Setting \"mdd.atime_diff\""
2250 echo "== perm cmd = '$PERM_CMD' =="
2251 $r $PERM_CMD $fsname-OST0000.failover.node=$nid || {
2252 error_noexit "Setting OST \"failover.node\""
2256 $r $PERM_CMD $fsname-MDT0000.failover.node=$nid || {
2257 error_noexit "Setting MDT \"failover.node\""
2261 $r $PERM_CMD $fsname-OST0000.osc.max_dirty_mb=15 || {
2262 error_noexit "Setting \"max_dirty_mb\""
2265 $r $PERM_CMD $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
2266 error_noexit "Setting \"max_rpcs_in_flight\""
2269 $r $PERM_CMD $fsname-MDT0000.lov.stripesize=4M || {
2270 error_noexit "Setting \"lov.stripesize\""
2273 $r $PERM_CMD $fsname-MDT0000.mdd.atime_diff=70 || {
2274 error_noexit "Setting \"mdd.atime_diff\""
2279 $r $LCTL pool_new $fsname.interop || {
2280 error_noexit "Setting \"interop\""
2284 if [ "$ff_convert" != "no" -a "$ost1_FSTYPE" == ldiskfs ]; then
2285 echo "== ff convert ($ff_convert) =="
2286 $r $LCTL lfsck_start -M $fsname-OST0000 || {
2287 error_noexit "Start OI scrub on OST0"
2291 # The oi_scrub should be on ost1, but for test_32(),
2292 # all on the SINGLEMDS.
2293 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
2294 osd-ldiskfs.$fsname-OST0000.oi_scrub |
2295 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
2296 error_noexit "Failed to get the expected 'completed'"
2300 local UPDATED=$($r $LCTL get_param -n \
2301 osd-ldiskfs.$fsname-OST0000.oi_scrub |
2302 awk '/^updated/ { print $2 }')
2303 [ $UPDATED -ge 1 ] || {
2304 error_noexit "Only $UPDATED objects have been converted"
2309 if [[ "$dne_upgrade" != "no" ]] || $mdt2_is_available; then
2310 echo "== dne upgrate ($dne_upgrade) or mdt2 available =="
2311 if [[ $PERM_CMD =~ "set_param" ]]; then
2312 $r $PERM_CMD mdc.$fsname-MDT0001*.import=connection=$nid || {
2313 error_noexit "Setting MDT1 \"failover.node\""
2317 $r $PERM_CMD mdc.$fsname-MDT0001-*.max_rpcs_in_flight=9 || {
2318 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
2321 $r $PERM_CMD lov.$fsname-MDT0001-*.stripesize=4M || {
2322 error_noexit "Setting MDT1 \"lov.stripesize\""
2326 $r $PERM_CMD $fsname-MDT0001.failover.node=$nid || {
2327 error_noexit "Setting MDT1 \"failover.node\""
2330 $r $PERM_CMD $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
2331 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
2334 $r $PERM_CMD $fsname-MDT0001.lov.stripesize=4M || {
2335 error_noexit "Setting MDT1 \"lov.stripesize\""
2341 if [ "$writeconf" ]; then
2342 echo "== writeconf and client mount =="
2343 $MOUNT_CMD $nid:/$fsname -o user_xattr $tmp/mnt/lustre || {
2344 error_noexit "Mounting the client"
2348 shall_cleanup_lustre=true
2349 $r $LCTL set_param debug="$PTLDEBUG"
2351 t32_verify_quota $SINGLEMDS $fsname $tmp/mnt/lustre || {
2352 error_noexit "verify quota failed"
2356 if $r test -f $tmp/list; then
2357 echo "== list verification =="
2359 # There is not a Test Framework API to copy files to or
2360 # from a remote node.
2362 # LU-2393 - do both sorts on same node to ensure locale
2364 local list_file=$tmp/list
2366 if $mdt2_is_available; then
2367 if [[ -d $tmp/mnt/lustre/striped_dir_old ]] &&
2368 $r test -f $tmp/list2; then
2369 list_file=$tmp/list2
2370 pushd $tmp/mnt/lustre/striped_dir_old
2372 pushd $tmp/mnt/lustre/remote_dir
2375 pushd $tmp/mnt/lustre
2377 $r cat $list_file | sort -k 6 >$tmp/list.orig
2378 BLOCKSIZE=1024 ls -Rni --time-style=+%s | sort -k 6 |
2379 sed 's/\. / /' >$tmp/list || {
2385 # 32-bit and 64-bit clients use different algorithms to
2386 # convert FIDs into inode numbers. Hence, remove the
2387 # inode numbers from the lists, if the original list was
2388 # created on an architecture with different number of
2391 if [ $(t32_bits_per_long $(uname -m)) != \
2392 $(t32_bits_per_long $img_arch) ]; then
2393 echo "Different number of bits per \"long\"" \
2394 "from the disk image"
2395 for list in list.orig list; do
2396 sed -i -e 's/^[0-9]\+[ \t]\+//' \
2400 if ! diff -ub $tmp/list.orig $tmp/list; then
2401 error_noexit "list verification failed"
2405 echo "== list verification skipped =="
2408 #non-dom upgrade to dom
2409 if [ "$dom_upgrade" != "no" ]; then
2410 echo "== dom upgrade =="
2411 $LFS setstripe -E 1M -L mdt -E EOF $tmp/mnt/lustre/dom || {
2412 error_noexit "Verify DoM creation"
2415 [ $($LFS getstripe -L $tmp/mnt/lustre/dom) == "mdt" ] || {
2416 error_noexit "Verify a DoM file"
2419 dd if=/dev/urandom of=$tmp/mnt/lustre/dom bs=4096 \
2420 count=1 conv=fsync || {
2421 error_noexit "Cannot write to DoM file"
2424 [ $(stat -c%s $tmp/mnt/lustre/dom) == 4096 ] || {
2425 error_noexit "DoM: bad size after write"
2428 rm $tmp/mnt/lustre/dom
2430 set_persistent_param_and_check mds \
2431 "lod.*MDT0000*.dom_stripesize" \
2432 "$fsname-MDT0000.lod.dom_stripesize" 0 || {
2433 error_noexit "Changing \"dom_stripesize\""
2439 #$LFS setstripe -E 1M -L mdt -E -1 -S 4M $dom_file
2440 if [ "$dom_new_upgrade" != "no" ]; then
2441 if ! $mdt2_is_available; then
2442 dom_file=$tmp/mnt/lustre/dom_dir/dom_file
2444 echo "== check DoM file can be accessed =="
2445 [ $($LFS getstripe -I1 -L $dom_file) == "mdt" ] || {
2446 error_noexit "Verify a DoM file"
2449 [ $(stat -c%s $dom_file) == $((2 * 1024 * 1024)) ] || {
2450 error_noexit "DoM: bad size after write"
2455 if [ "$flr_upgrade" != "no" ]; then
2456 if ! $mdt2_is_available; then
2457 flr_file=$tmp/mnt/lustre/flr_dir/flr_file
2459 local mirror_count=$($LFS getstripe -N $flr_file)
2460 echo "== check FLR file =="
2461 [ $mirror_count == 2 ] || {
2462 error_noexit "FLR mirror count wrong"
2467 if [ "$dne_upgrade" != "no" ]; then
2468 echo "== dne upgrade striped_dir =="
2469 $LFS mkdir -i 1 -c2 $tmp/mnt/lustre/striped_dir || {
2470 error_noexit "set striped dir failed"
2473 $LFS setdirstripe -D -c2 $tmp/mnt/lustre/striped_dir
2475 pushd $tmp/mnt/lustre
2476 tar -c --exclude=./striped_dir \
2477 --exclude=./striped_dir_old \
2478 --exclude=./remote_dir -f - .|
2479 tar -xvf - -C striped_dir 1>/dev/null || {
2480 error_noexit "cp to striped dir failed"
2486 if [ "$pfl_upgrade" != "no" ]; then
2493 echo "== check PFL file =="
2494 if ! $mdt2_is_available; then
2495 pfl_file=$tmp/mnt/lustre/pfl_dir/pfl_file
2497 comp_cnt=$($LFS getstripe --component-count $pfl_file)
2498 [ $comp_cnt == 2 ] || {
2499 error_noexit "wrong comp_cnt $comp_cnt"
2503 comp_size=$($LFS getstripe -I1 -E $pfl_file)
2504 [ $comp_size == 2097152 ] || {
2505 error_noexit "wrong component size $comp_size"
2509 comp_id=$($LFS getstripe -I1 -i $pfl_file)
2510 [ $comp_id == 0 ] || {
2511 error_noexit "wrong comp id $comp_id"
2515 comp_id=$($LFS getstripe -I2 -i $pfl_file)
2516 [ $comp_id -eq 1 ] || {
2517 error_noexit "wrong comp id $comp_id"
2521 stripe_size=$($LFS getstripe -I1 -S $pfl_file)
2522 [ $stripe_size -eq 1048576 ] || {
2523 error_noexit "wrong stripe size $stripe_size"
2527 stripe_size=$($LFS getstripe -I2 -S $pfl_file)
2528 [ $comp_size -eq 2097152 ] || {
2529 error_noexit "wrong component size $comp_size"
2534 local large_xattr_dir=$tmp/mnt/lustre/large_xattr_test_dir
2536 if [[ -d $large_xattr_dir ]]; then
2537 echo "== check Large EA =="
2538 local xattr_file=$large_xattr_dir/large_xattr_file
2540 xattr_val=$(getfattr --only-values\
2541 -n user.fooattr $xattr_file) || {
2542 error_noexit "Large EA cannot be read"
2546 [[ $xattr_val == $(printf "%c" {1..4096}) ]] || {
2547 error_noexit "Wrong large EA value"
2552 if [ "$project_quota_upgrade" != "no" ]; then
2553 if ! $mdt2_is_available; then
2554 quota_dir=$tmp/mnt/lustre/project_quota_dir
2558 echo "== check Project Quota =="
2559 hardlimit=$(getquota -p $T32_PRJID global 3 \
2561 [ $hardlimit == $T32_PROLIMIT ] || {
2562 error_noexit "wrong hardlimit $hardlimit"
2567 # If it is upgrade from DNE (2.5), then rename the remote dir,
2568 # which is created in 2.5 to striped dir.
2569 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]]; then
2570 echo "== rename remote_dir =="
2571 stripe_index=$($LFS getdirstripe -i \
2572 $tmp/mnt/lustre/remote_dir)
2574 [[ $stripe_index -eq 1 ]] || {
2575 error_noexit "get index \"$stripe_index\"" \
2576 "from remote dir failed"
2579 mv $tmp/mnt/lustre/remote_dir \
2580 $tmp/mnt/lustre/striped_dir/ || {
2581 error_noexit "mv remote dir failed"
2586 # If it is upgraded from DNE (2.7), then move the striped dir
2587 # which was created in 2.7 to the new striped dir.
2588 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]] &&
2589 echo "== rename striped dir =="
2590 [[ -d $tmp/mnt/lustre/striped_dir_old ]]; then
2591 stripe_count=$($LFS getdirstripe -c \
2592 $tmp/mnt/lustre/striped_dir_old)
2593 [[ $stripe_count -eq 2 ]] || {
2594 error_noexit "get count $stripe_count" \
2595 "from striped dir failed"
2598 mv $tmp/mnt/lustre/striped_dir_old \
2599 $tmp/mnt/lustre/striped_dir/ || {
2600 error_noexit "mv striped dir failed"
2606 $r $LCTL set_param -n osd*.*.force_sync=1
2607 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
2608 error_noexit "dd failed"
2611 rm -rf $tmp/mnt/lustre/tmp_file || {
2612 error_noexit "rm failed"
2616 if $r test -f $tmp/sha1sums; then
2617 echo "== checking sha1sums =="
2618 # LU-2393 - do both sorts on same node to ensure locale
2620 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
2621 if [[ "$dne_upgrade" != "no" ]]; then
2622 if [[ -d $tmp/mnt/lustre/striped_dir/remote_dir ]]; then
2623 pushd $tmp/mnt/lustre/striped_dir/remote_dir
2625 pushd $tmp/mnt/lustre/striped_dir
2627 elif [[ "$pfl_upgrade" != "no" ]] ||
2628 [[ "$flr_upgrade" != "no" ]] ||
2629 [[ "$dom_new_upgrade" != "no" ]] ||
2630 [[ "$project_quota_upgrade" != "no" ]] ||
2631 [[ -d $tmp/mnt/lustre/remote_dir ]]; then
2632 pushd $tmp/mnt/lustre/remote_dir
2634 pushd $tmp/mnt/lustre
2637 find ! -path "*remote_dir*" ! -path "*striped_dir*" \
2638 ! -name .lustre -type f -exec sha1sum {} \; |
2639 sort -k 2 >$tmp/sha1sums || {
2641 error_noexit "sha1sum"
2647 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
2648 echo "==** find $save files start **=="
2649 find $save -type f -ls
2650 echo "==** find $save files finish **=="
2651 error_noexit "sha1sum verification failed"
2655 # if upgrade from DNE(2.5), then check remote directory
2656 # if upgrade from DNE(2.7), then check striped directory
2657 if $mdt2_is_available &&
2658 [[ "$dne_upgrade" != "no" ]]; then
2659 echo "== mdt2 upgrade sha1sum check =="
2660 local new_dir="$tmp/mnt/lustre/striped_dir"
2661 local striped_dir_old="$new_dir/striped_dir_old"
2663 local dir_list="$new_dir/remote_dir"
2664 [[ ! -d $triped_dir_old ]] ||
2665 dir_list+=" $striped_dir_old"
2667 for dir in $dir_list; do
2668 echo "== dir = $dir =="
2670 find ! -name .lustre -type f \
2671 -exec sha1sum {} \; |
2672 sort -k 2 >$tmp/sha1sums || {
2675 error_noexit "sha1sum"
2679 if ! diff -ub $tmp/sha1sums.orig \
2682 error_noexit "sha1sum $dir" \
2689 echo "== sha1sum verification skipped =="
2692 # PFL write test after sha1sum check
2693 if [ "$pfl_upgrade" != "no" ]; then
2694 local rw_len=$((3 * 1034 * 1024))
2696 if ! $mdt2_is_available; then
2697 pfl_file=$tmp/mnt/lustre/pfl_dir/pfl_file
2699 small_write $pfl_file $rw_len || {
2700 error_noexit "PFL RW Failed"
2705 if [ "$dne_upgrade" != "no" ]; then
2706 rm -rf $tmp/mnt/lustre/striped_dir || {
2707 error_noexit "remove remote dir failed"
2712 # migrate files/dirs to remote MDT, then move them back
2713 if [ "$MDS1_VERSION" -ge $(version_code 2.7.50) -a \
2714 $dne_upgrade != "no" ]; then
2715 $r $LCTL set_param -n \
2716 mdt.${fsname}*.enable_remote_dir=1 2>/dev/null
2718 echo "== test migration =="
2719 pushd $tmp/mnt/lustre
2720 for dir in $(find ! -name .lustre ! -name . -type d); do
2721 mdt_index=$($LFS getdirstripe -i $dir)
2722 stripe_cnt=$($LFS getdirstripe -c $dir)
2723 if [ $mdt_index = 0 -a $stripe_cnt -le 1 ]; then
2724 $LFS migrate -m 1 $dir || {
2726 error_noexit "migrate MDT1 failed"
2732 for dir in $(find ! -name . ! -name .lustre -type d); do
2733 mdt_index=$($LFS getdirstripe -i $dir)
2734 stripe_cnt=$($LFS getdirstripe -c $dir)
2735 if [ $mdt_index = 1 -a $stripe_cnt -le 1 ]; then
2736 $LFS migrate -m 0 $dir || {
2738 error_noexit "migrate MDT0 failed"
2747 # When adding new data verification tests, please check for
2748 # the presence of the required reference files first, like
2749 # the "sha1sums" and "list" tests above, to avoid the need to
2750 # regenerate every image for each test addition.
2753 nrpcs_orig=$($LCTL get_param \
2754 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
2755 error_noexit "Getting \"max_rpcs_in_flight\""
2758 nrpcs=$((nrpcs_orig + 5))
2760 set_persistent_param_and_check client \
2761 "mdc.$fsname-MDT0000*.max_rpcs_in_flight" \
2762 "$fsname-MDT0000.mdc.max_rpcs_in_flight" $nrpcs || {
2763 error_noexit "Changing \"max_rpcs_in_flight\""
2767 umount $tmp/mnt/lustre || {
2768 error_noexit "Unmounting the client"
2771 shall_cleanup_lustre=false
2773 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
2774 error_noexit "Mounting the client"
2778 if [ "$mds1_FSTYPE" == ldiskfs -a \
2779 "$project_upgrade" != "no" ]; then
2780 ! $LFS project -d -p 1 $tmp/mnt/lustre/* &&
2781 error_noexit "set project failed" &&
2785 [[ $(do_facet mds1 pgrep orph_.*-MDD | wc -l) == 0 ]] ||
2786 error "MDD orphan cleanup thread not quit"
2788 umount $tmp/mnt/lustre || {
2789 error_noexit "Unmounting the client"
2793 if [[ "$dne_upgrade" != "no" ]] || $mdt2_is_available; then
2794 $r $UMOUNT $tmp/mnt/mdt1 || {
2795 error_noexit "Unmounting the MDT2"
2798 if [[ "$mds1_FSTYPE" == zfs ]]; then
2799 $r "$ZPOOL export t32fs-mdt2"
2801 shall_cleanup_mdt1=false
2804 $r $UMOUNT $tmp/mnt/mdt || {
2805 error_noexit "Unmounting the MDT"
2808 if [[ "$mds1_FSTYPE" == zfs ]]; then
2809 $r "$ZPOOL export t32fs-mdt1"
2811 shall_cleanup_mdt=false
2813 $r $UMOUNT $tmp/mnt/ost || {
2814 error_noexit "Unmounting the OST"
2817 if $ost2_is_available; then
2818 $r $UMOUNT $tmp/mnt/ost1 || {
2819 error_noexit "Unmounting the OST1"
2822 shall_cleanup_ost1=false
2825 if [[ $ost1_FSTYPE == zfs ]]; then
2826 $r "$ZPOOL export t32fs-ost1"
2827 if $ost2_is_available; then
2828 $r "$ZPOOL export t32fs-ost2"
2831 shall_cleanup_ost=false
2833 t32_reload_modules $node || {
2834 error_noexit "Reloading modules"
2838 if [[ "$mds1_FSTYPE" == zfs ]]; then
2839 local poolname=t32fs-mdt1
2841 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
2842 $ZPOOL import -f -d $tmp $poolname"
2844 # upgrade zpool to latest supported features,
2845 # including dnode quota accounting in 0.7.0
2846 $r "$ZPOOL upgrade $poolname"
2849 # mount a second time to make sure we didnt leave upgrade flag on
2850 $r $TUNEFS --dryrun $mdt_dev || {
2852 error_noexit "tunefs.lustre before remounting the MDT"
2856 mopts=exclude=$fsname-OST0000
2857 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2860 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2861 error_noexit "Remounting the MDT"
2864 shall_cleanup_mdt=true
2874 for tarball in $tarballs; do
2875 banner "testing $tarball upgrade"
2876 t32_test $tarball || let "rc += $?"
2880 run_test 32a "Upgrade (not live)"
2888 for tarball in $tarballs; do
2889 banner "testing $tarball upgrade with writeconf"
2890 t32_test $tarball writeconf || let "rc += $?"
2894 run_test 32b "Upgrade with writeconf"
2902 for tarball in $tarballs; do
2903 banner "testing $tarball upgrade with DNE"
2905 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
2909 run_test 32c "dne upgrade test"
2917 for tarball in $tarballs; do
2918 banner "testing $tarball upgrade with ff and project upgrade"
2919 project_upgrade="no"
2920 [[ "$MDS1_VERSION" -ge $(version_code 2.13.54) ]] &&
2921 [[ "$tarball" =~ "disk2_4-ldiskfs" ]] &&
2922 project_upgrade="yes"
2923 project_upgrade=$project_upgrade ff_convert=yes t32_test \
2928 run_test 32d "convert ff and project quota upgrade test"
2931 [[ "$MDS1_VERSION" -ge $(version_code 2.10.56) ]] ||
2932 skip "Need MDS version at least 2.10.56"
2939 for tarball in $tarballs; do
2940 [[ "$tarball" =~ "2_9" ]] ||
2941 { echo "skip $(basename $tarball)"; continue; }
2943 banner "testing $tarball upgrade with DoM"
2944 dom_upgrade=yes t32_test $tarball writeconf || let "rc += $?"
2948 run_test 32e "dom upgrade test"
2951 [[ $MDS1_VERSION -ge $(version_code 2.10.56) ]] ||
2952 skip "Need MDS version at least 2.10.56"
2959 for tarball in $tarballs; do
2960 [[ "$tarball" =~ "2_10" ]] ||
2961 { echo "skip $(basename $tarball)"; continue; }
2962 pfl_upgrade=yes project_quota_upgrade=yes \
2963 t32_test $tarball writeconf || let "rc += $?"
2967 run_test 32f "pfl upgrade test"
2970 [[ $MDS1_VERSION -ge $(version_code 2.10.56) ]] ||
2971 skip "Need MDS version at least 2.10.56"
2978 for tarball in $tarballs; do
2979 [[ $tarball =~ "2_12" ]] ||
2980 { echo "skip $(basename $tarball)"; continue; }
2981 flr_upgrade=yes dom_new_upgrade=yes \
2982 t32_test $tarball writeconf || let "rc += $?"
2986 run_test 32g "flr/dom upgrade test"
2988 test_33a() { # bug 12333, was test_33
2989 local FSNAME2=test-$testnum
2991 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2994 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
2996 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
2997 local dev=${SINGLEMDS}_dev
2998 local MDSDEV=${!dev}
3000 is_blkdev $SINGLEMDS $MDSDEV &&
3001 skip_env "mixed loopback and real device not working"
3004 local fs2mdsdev=$(mdsdevname 1_2)
3005 local fs2ostdev=$(ostdevname 1_2)
3006 local fs2mdsvdev=$(mdsvdevname 1_2)
3007 local fs2ostvdev=$(ostvdevname 1_2)
3009 if [ "$mds1_FSTYPE" == ldiskfs ]; then
3010 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
3013 if combined_mgs_mds; then
3014 local mgs_flag="--mgs"
3017 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --fsname=${FSNAME2} \
3018 --reformat $mgs_flag $mkfsoptions $fs2mdsdev $fs2mdsvdev ||
3020 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
3021 --fsname=${FSNAME2} --index=0x1fff --reformat $fs2ostdev \
3022 $fs2ostvdev || exit 10
3024 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
3025 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
3027 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3028 do_facet mgs "$PERM_CMD timeout=200" ||
3029 error "$PERM_CMD timeout=200 failed"
3031 do_facet mgs "$PERM_CMD $FSNAME2.sys.timeout=200" ||
3032 error "$PERM_CMD $FSNAME2.sys.timeout=200 failed"
3034 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
3035 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
3038 cp /etc/hosts $MOUNT2/ || error "copy /etc/hosts $MOUNT2/ failed"
3039 $LFS getstripe $MOUNT2/hosts ||
3040 error "$LFS getstripe $MOUNT2/hosts failed"
3044 # test lctl del_ost on large index
3045 do_facet mgs "$LCTL del_ost -t ${FSNAME2}-OST1fff" ||
3046 error "del_ost failed with $?"
3047 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
3049 $LFS df | grep -q OST1fff && error "del_ost did not remove OST1fff!"
3054 cleanup_nocli || error "cleanup_nocli failed with $?"
3056 run_test 33a "Mount ost with a large index number"
3058 test_33b() { # was test_34
3061 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
3062 # Drop lock cancelation reply during umount
3063 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
3064 do_facet client $LCTL set_param fail_loc=0x80000304
3065 #lctl set_param debug=-1
3066 umount_client $MOUNT
3067 cleanup || error "cleanup failed with $?"
3069 run_test 33b "Drop cancel during umount"
3073 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
3074 manual_umount_client
3076 do_facet client killall -USR1 multiop
3077 if [ $rc -eq 0 ]; then
3078 error "umount not fail!"
3081 cleanup || error "cleanup failed with rc $?"
3083 run_test 34a "umount with opened file should be fail"
3087 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3088 stop_mds || error "Unable to stop MDS"
3090 manual_umount_client --force || error "mtab after failed umount with $?"
3092 cleanup || error "cleanup failed with $?"
3094 run_test 34b "force umount with failed mds should be normal"
3098 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3099 stop_ost || error "Unable to stop OST1"
3101 manual_umount_client --force || error "mtab after failed umount with $?"
3103 cleanup || error "cleanup failed with $?"
3105 run_test 34c "force umount with failed ost should be normal"
3107 test_35a() { # bug 12459
3110 DBG_SAVE="`$LCTL get_param -n debug`"
3111 $LCTL set_param debug="ha"
3113 log "Set up a fake failnode for the MDS"
3115 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
3116 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
3118 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3119 do_facet mgs "$PERM_CMD \
3120 mdc.*${device}*.import=connection=$(h2nettype $FAKENID)" ||
3121 error "Setting mdc.*${device}*.import=connection=\
3122 $(h2nettype $FAKENID) failed."
3124 do_facet mgs "$PERM_CMD \
3125 ${device}.failover.node=$(h2nettype $FAKENID)" ||
3126 error "Setting ${device}.failover.node=\
3127 $(h2nettype $FAKENID) failed."
3129 log "Wait for RECONNECT_INTERVAL seconds (10s)"
3132 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
3135 log "Stopping the MDT: $device"
3136 stop_mdt 1 || error "MDT0 stop fail"
3138 df $MOUNT > /dev/null 2>&1 &
3140 log "Restarting the MDT: $device"
3141 start_mdt 1 || error "MDT0 start fail"
3142 log "Wait for df ($DFPID) ... "
3145 $LCTL set_param debug="$DBG_SAVE"
3147 # retrieve from the log the first server that the client tried to
3148 # contact after the connection loss
3149 $LCTL dk $TMP/lustre-log-$TESTNAME.log
3150 NEXTCONN=`awk "/${MSG}/ {start = 1;}
3151 /import_select_connection.*$device-mdc.* using connection/ {
3153 if (\\\$NF ~ /$FAKENID/)
3159 }" $TMP/lustre-log-$TESTNAME.log`
3160 [ "$NEXTCONN" != "0" ] &&
3161 error "Tried to connect to ${NEXTCONN} not last active server"
3162 cleanup || error "cleanup failed with $?"
3163 # remove nid settings
3164 writeconf_or_reformat
3166 run_test 35a "Reconnect to the last active server first"
3168 test_35b() { # bug 18674
3169 remote_mds || skip "local MDS"
3173 $LCTL set_param debug="ha"
3175 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
3178 log "Set up a fake failnode for the MDS"
3180 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
3181 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
3183 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3184 do_facet mgs "$PERM_CMD \
3185 mdc.*${device}*.import=connection=$(h2nettype $FAKENID)" ||
3186 error "Set mdc.*${device}*.import=connection=\
3187 $(h2nettype $FAKENID) failed"
3189 do_facet mgs "$PERM_CMD \
3190 ${device}.failover.node=$(h2nettype $FAKENID)" ||
3191 error "Set ${device}.failover.node=\
3192 $(h2nettype $FAKENID) failed"
3195 local at_max_saved=0
3196 # adaptive timeouts may prevent seeing the issue
3197 if at_is_enabled; then
3198 at_max_saved=$(at_max_get mds)
3199 at_max_set 0 mds client
3202 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
3204 log "Injecting EBUSY on MDS"
3205 # Setting OBD_FAIL_MDS_RESEND=0x136
3206 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" ||
3207 error "unable to set param fail_loc=0x80000136"
3209 $LCTL set_param mdc.${FSNAME}*.stats=clear
3211 log "Creating a test file and stat it"
3212 touch $MOUNT/$tdir/$tfile || error "touch $MOUNT/$tdir/$tfile failed"
3213 stat $MOUNT/$tdir/$tfile
3215 log "Stop injecting EBUSY on MDS"
3216 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" ||
3217 error "unable to set param fail_loc=0"
3218 rm -f $MOUNT/$tdir/$tfile || error "remove $MOUNT/$tdir/$tfile failed"
3221 # restore adaptive timeout
3222 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
3224 $LCTL dk $TMP/lustre-log-$TESTNAME.log
3226 CONNCNT=$($LCTL get_param mdc.${FSNAME}*.stats |
3227 awk '/mds_connect/{print $2}')
3229 # retrieve from the log if the client has ever tried to
3230 # contact the fake server after the loss of connection
3231 FAILCONN=`awk "BEGIN {ret = 0;}
3232 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
3234 if (\\\$NF ~ /$FAKENID/) {
3239 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
3241 [ "$FAILCONN" == "0" ] &&
3242 error "The client reconnection has not been triggered"
3243 [ "$FAILCONN" == "2" ] &&
3244 error "Primary server busy, client reconnect to failover failed"
3247 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
3248 # Reconnects are supposed to be rate limited to one every 5s
3249 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] &&
3250 error "Too many reconnects $CONNCNT"
3252 cleanup || error "cleanup failed with $?"
3253 # remove nid settings
3254 writeconf_or_reformat
3256 run_test 35b "Continue reconnection retries, if the active server is busy"
3259 [ $OSTCOUNT -lt 2 ] && skip_env "needs >= 2 OSTs"
3261 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] ||
3265 local FSNAME2=test1234
3266 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3268 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
3270 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
3271 is_blkdev $SINGLEMDS $MDSDEV &&
3272 skip_env "mixed loopback and real device not working"
3275 local fs2mdsdev=$(mdsdevname 1_2)
3276 local fs2ostdev=$(ostdevname 1_2)
3277 local fs3ostdev=$(ostdevname 2_2)
3278 local fs2mdsvdev=$(mdsvdevname 1_2)
3279 local fs2ostvdev=$(ostvdevname 1_2)
3280 local fs3ostvdev=$(ostvdevname 2_2)
3283 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
3284 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
3285 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
3286 # different one than the default value here.
3287 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
3288 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
3289 add fs3ost $(mkfs_opts ost2 ${fs3ostdev}) --mgsnode=$MGSNID \
3290 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
3292 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
3293 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
3294 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
3295 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
3296 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
3298 sleep 5 # until 11778 fixed
3300 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || error "dd failed"
3302 BKTOTAL=$($LCTL get_param -n obdfilter.*.kbytestotal |
3303 awk 'BEGIN{total=0}; {total+=$1}; END{print total}')
3304 BKFREE=$($LCTL get_param -n obdfilter.*.kbytesfree |
3305 awk 'BEGIN{free=0}; {free+=$1}; END{print free}')
3306 BKAVAIL=$($LCTL get_param -n obdfilter.*.kbytesavail |
3307 awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}')
3308 STRING=$(df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}')
3309 DFTOTAL=$(echo $STRING | cut -d, -f1)
3310 DFUSED=$(echo $STRING | cut -d, -f2)
3311 DFAVAIL=$(echo $STRING | cut -d, -f3)
3312 DFFREE=$(($DFTOTAL - $DFUSED))
3314 ALLOWANCE=$((64 * $OSTCOUNT))
3316 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
3317 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
3318 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
3321 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
3322 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
3323 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
3326 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
3327 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
3328 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
3333 stop fs3ost -f || error "unable to stop OST3"
3334 stop fs2ost -f || error "unable to stop OST2"
3335 stop fs2mds -f || error "unable to stop second MDS"
3336 unload_modules_conf || error "unable unload modules"
3339 run_test 36 "df report consistency on OSTs with different block size"
3342 local mntpt=$(facet_mntpt $SINGLEMDS)
3343 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
3344 local mdsdev_sym="$TMP/sym_mdt.img"
3345 local opts=$MDS_MOUNT_OPTS
3348 if [ "$mds1_FSTYPE" != ldiskfs ]; then
3349 skip "ldiskfs only test"
3352 echo "MDS : $mdsdev"
3353 echo "SYMLINK : $mdsdev_sym"
3354 do_facet $SINGLEMDS rm -f $mdsdev_sym
3356 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
3358 echo "mount symlink device - $mdsdev_sym"
3360 if ! do_facet $SINGLEMDS test -b $mdsdev; then
3361 opts=$(csa_add "$opts" -o loop)
3365 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
3366 $mdsdev_sym $mntpt 2>&1)
3369 echo mount_op=$mount_op
3371 do_facet $SINGLEMDS "$UMOUNT $mntpt && rm -f $mdsdev_sym"
3373 if $(echo $mount_op | grep -q "unable to set tunable"); then
3374 error "set tunables failed for symlink device"
3377 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
3379 run_test 37 "verify set tunables works for symlink device"
3381 test_38() { # bug 14222
3382 local mntpt=$(facet_mntpt $SINGLEMDS)
3387 local SRC="/etc /bin"
3388 local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT)
3390 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
3391 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3392 tar cf - $FILES | tar xf - -C $DIR/$tdir ||
3393 error "copying $SRC to $DIR/$tdir"
3395 umount_client $MOUNT || error "umount_client $MOUNT failed"
3396 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3397 stop_mds || error "Unable to stop MDS"
3398 log "delete lov_objid file on MDS"
3400 mount_fstype $SINGLEMDS || error "mount MDS failed (1)"
3402 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid; rm $mntpt/lov_objid"
3404 unmount_fstype $SINGLEMDS || error "umount failed (1)"
3406 # check create in mds_lov_connect
3407 start_mds || error "unable to start MDS"
3408 mount_client $MOUNT || error "mount_client $MOUNT failed"
3410 [ $V ] && log "verifying $DIR/$tdir/$f"
3411 diff -q $f $DIR/$tdir/$f || ERROR=y
3413 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3414 if [ "$ERROR" = "y" ]; then
3415 # check it's updates in sync
3416 umount_client $MOUNT
3418 mount_fstype $SIGNLEMDS
3419 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3420 unmount_fstype $SINGLEMDS
3421 error "old and new files are different after connect" || true
3423 touch $DIR/$tdir/f2 || error "f2 file create failed"
3425 # check it's updates in sync
3426 umount_client $MOUNT || error "second umount_client $MOUNT failed"
3429 mount_fstype $SINGLEMDS || error "mount MDS failed (3)"
3431 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3432 do_facet $SINGLEMDS dd if=/dev/zero of=$mntpt/lov_objid.clear count=8
3434 unmount_fstype $SINGLEMDS || error "umount failed (3)"
3436 start_mds || error "unable to start MDS"
3437 mount_client $MOUNT || error "mount_client $MOUNT failed"
3439 [ $V ] && log "verifying $DIR/$tdir/$f"
3440 diff -q $f $DIR/$tdir/$f || ERROR=y
3442 touch $DIR/$tdir/f3 || error "f3 file create failed"
3443 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3444 umount_client $MOUNT || error "third umount_client $MOUNT failed"
3446 mount_fstype $SINGLEMDS || error "mount MDS failed (4)"
3447 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3448 unmount_fstype $SINGLEMDS || error "umount failed (4)"
3450 [ "$ERROR" = "y" ] &&
3451 error "old and new files are different after sync" || true
3453 log "files compared the same"
3454 cleanup || error "cleanup failed with $?"
3456 run_test 38 "MDS recreates missing lov_objid file from OST data"
3461 cleanup || error "cleanup failed with $?"
3462 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
3463 error "memory leak detected" || true
3465 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
3467 test_40() { # bug 15759
3468 start_ost || error "Unable to start OST1"
3469 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
3470 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
3472 cleanup || error "cleanup failed with rc $?"
3474 run_test 40 "race during service thread startup"
3476 test_41a() { #bug 14134
3477 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3478 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
3479 skip "Loop devices does not work with nosvc option"
3482 combined_mgs_mds || skip "needs combined MGT and MDT device"
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 ost1 $(ostdevname 1) $OST_MOUNT_OPTS
3491 start_mdt 1 -o nomgs,force
3492 mount_client $MOUNT || error "mount_client $MOUNT failed"
3495 echo "blah blah" > $MOUNT/$tfile
3498 umount_client $MOUNT || error "umount_client $MOUNT failed"
3499 stop ost1 -f || error "unable to stop OST1"
3500 stop_mds || error "Unable to stop MDS"
3501 stop_mds || error "Unable to stop MDS on second try"
3503 run_test 41a "mount mds with --nosvc and --nomgs"
3506 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3507 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
3508 skip "Loop devices does not work with nosvc option"
3511 ! combined_mgs_mds && skip "needs combined mgs device"
3515 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3517 start_mdt 1 -o nosvc -n
3518 if [ $MDSCOUNT -ge 2 ]; then
3519 for num in $(seq 2 $MDSCOUNT); do
3520 start_mdt $num || return
3523 start_ost || error "Unable to start OST1"
3524 start_mdt 1 -o nomgs,force
3525 mount_client $MOUNT || error "mount_client $MOUNT failed"
3528 echo "blah blah" > $MOUNT/$tfile
3529 cat $MOUNT/$tfile || error "cat $MOUNT/$tfile failed"
3531 umount_client $MOUNT -f || error "umount_client $MOUNT failed"
3532 stop_ost || error "Unable to stop OST1"
3533 stop_mds || error "Unable to stop MDS"
3534 stop_mds || error "Unable to stop MDS on second try"
3536 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
3539 local oss_list=$(comma_list $(osts_nodes))
3541 [[ "$MDS1_VERSION" -ge $(version_code 2.6.52) ]] ||
3542 [[ "$MDS1_VERSION" -ge $(version_code 2.5.26) &&
3543 "$MDS1_VERSION" -lt $(version_code 2.5.50) ]] ||
3544 [[ "$MDS1_VERSION" -ge $(version_code 2.5.4) &&
3545 "$MDS1_VERSION" -lt $(version_code 2.5.11) ]] ||
3546 skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"
3548 # ensure mds1 ost1 have been created even if running sub-test standalone
3551 cleanup || error "cleanup failed"
3553 # using directly mount command instead of start() function to avoid
3554 # any side effect of // with others/externals tools/features
3555 # ("zpool import", ...)
3557 # MDT concurrent start
3559 LOAD_MODULES_REMOTE=true load_modules
3560 do_facet $SINGLEMDS "lsmod | grep -q libcfs" ||
3561 error "MDT concurrent start: libcfs module not loaded"
3563 local mds1dev=$(mdsdevname 1)
3564 local mds1mnt=$(facet_mntpt mds1)
3565 local mds1opts=$MDS_MOUNT_OPTS
3567 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3568 ! do_facet mds1 test -b $mds1dev; then
3569 mds1opts=$(csa_add "$mds1opts" -o loop)
3571 if [[ "$mds1_FSTYPE" == zfs ]]; then
3572 import_zpool mds1 || return ${PIPESTATUS[0]}
3575 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3576 do_facet mds1 "$LCTL set_param fail_loc=0x80000716"
3578 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts &
3581 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts
3586 do_facet mds1 "$LCTL set_param fail_loc=0x0"
3587 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3588 echo "1st MDT start succeed"
3589 echo "2nd MDT start failed with $rc2"
3590 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3591 echo "1st MDT start failed with $rc"
3592 echo "2nd MDT start succeed"
3595 error "unexpected concurrent MDT mounts result, rc=$rc rc2=$rc2"
3598 if [ $MDSCOUNT -ge 2 ]; then
3599 for num in $(seq 2 $MDSCOUNT); do
3600 start_mdt $num || return
3604 # OST concurrent start
3606 do_rpc_nodes $oss_list "lsmod | grep -q libcfs" ||
3607 error "OST concurrent start: libcfs module not loaded"
3609 local ost1dev=$(ostdevname 1)
3610 local ost1mnt=$(facet_mntpt ost1)
3611 local ost1opts=$OST_MOUNT_OPTS
3613 if [ "$ost1_FSTYPE" == ldiskfs ] &&
3614 ! do_facet ost1 test -b $ost1dev; then
3615 ost1opts=$(csa_add "$ost1opts" -o loop)
3617 if [[ "$ost1_FSTYPE" == zfs ]]; then
3618 import_zpool ost1 || return ${PIPESTATUS[0]}
3621 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3622 do_facet ost1 "$LCTL set_param fail_loc=0x80000716"
3624 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts &
3627 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts
3631 do_facet ost1 "$LCTL set_param fail_loc=0x0"
3632 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3633 echo "1st OST start succeed"
3634 echo "2nd OST start failed with $rc2"
3635 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3636 echo "1st OST start failed with $rc"
3637 echo "2nd OST start succeed"
3641 error "unexpected concurrent OST mounts result, rc=$rc rc2=$rc2"
3647 # verify everything ok
3652 error "MDT(s) start failed"
3660 error "OST(s) start failed"
3668 error "client start failed"
3675 error "client mount failed"
3679 run_test 41c "concurrent mounts of MDT/OST should all fail but one"
3681 test_42() { #bug 14693
3685 check_mount || error "client was not mounted"
3687 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3688 PARAM="llite.$FSNAME-*.some_wrong_param"
3690 PARAM="$FSNAME.llite.some_wrong_param"
3693 do_facet mgs $PERM_CMD $PARAM=10
3694 umount_client $MOUNT ||
3695 error "unmounting client failed with invalid llite param"
3696 mount_client $MOUNT ||
3697 error "mounting client failed with invalid llite param"
3699 do_facet mgs $PERM_CMD $PARAM=20
3700 cleanup || error "stopping $FSNAME failed with invalid sys param"
3702 check_mount || error "client was not mounted with invalid sys param"
3703 cleanup || error "stopping $FSNAME failed with invalid sys param"
3705 run_test 42 "allow client/server mount/unmount with invalid config param"
3708 [[ "$MGS_VERSION" -ge $(version_code 2.5.58) ]] ||
3709 skip "Need MDS version at least 2.5.58"
3710 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
3713 USER1=$(getent passwd | grep :$ID1:$ID1: | cut -d: -f1)
3714 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1"
3717 chmod ugo+x $DIR || error "chmod 0 failed"
3718 set_persistent_param_and_check mds1 \
3719 "mdt.$FSNAME-MDT0000.root_squash" \
3720 "$FSNAME.mdt.root_squash" \
3722 wait_update $HOSTNAME \
3723 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3725 error "check llite root_squash failed!"
3726 set_persistent_param_and_check mds1 \
3727 "mdt.$FSNAME-MDT0000.nosquash_nids" \
3728 "$FSNAME.mdt.nosquash_nids" \
3730 wait_update $HOSTNAME \
3731 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3733 error "check llite nosquash_nids failed!"
3736 # create set of test files
3738 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
3739 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
3740 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
3742 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
3743 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
3745 mkdir_on_mdt0 $DIR/$tdir-rootdir || error "mkdir failed"
3746 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
3747 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
3749 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
3750 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
3751 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
3754 # check root_squash:
3755 # set root squash UID:GID to RUNAS_ID
3756 # root should be able to access only files owned by RUNAS_ID
3758 set_persistent_param_and_check mds1 \
3759 "mdt.$FSNAME-MDT0000.root_squash" \
3760 "$FSNAME.mdt.root_squash" \
3761 "$RUNAS_ID:$RUNAS_ID"
3762 wait_update $HOSTNAME \
3763 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3764 "$RUNAS_ID:$RUNAS_ID" ||
3765 error "check llite root_squash failed!"
3767 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
3768 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3769 error "$ST: root read permission is denied"
3770 echo "$ST: root read permission is granted - ok"
3773 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3774 error "$ST: root write permission is denied"
3775 echo "$ST: root write permission is granted - ok"
3777 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3778 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3779 error "$ST: root read permission is granted"
3780 echo "$ST: root read permission is denied - ok"
3783 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3784 error "$ST: root write permission is granted"
3785 echo "$ST: root write permission is denied - ok"
3787 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3788 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null &&
3789 error "$ST: root unlink permission is granted"
3790 echo "$ST: root unlink permission is denied - ok"
3792 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null &&
3793 error "$ST: root create permission is granted"
3794 echo "$ST: root create permission is denied - ok"
3797 # check root_squash is enforced independently
3798 # of client cache content
3800 # access file by USER1, keep access open
3801 # root should be denied access to user file
3803 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
3807 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
3808 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3809 { kill $pid; error "$ST: root read permission is granted"; }
3810 echo "$ST: root read permission is denied - ok"
3813 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3814 { kill $pid; error "$ST: root write permission is granted"; }
3815 echo "$ST: root write permission is denied - ok"
3821 # check nosquash_nids:
3822 # put client's NID into nosquash_nids list,
3823 # root should be able to access root file after that
3825 local NIDLIST=$($LCTL list_nids all | tr '\n' ' ')
3826 NIDLIST="2@gni $NIDLIST 192.168.0.[2,10]@tcp"
3827 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
3828 set_persistent_param_and_check mds1 \
3829 "mdt.$FSNAME-MDT0000.nosquash_nids" \
3830 "$FSNAME-MDTall.mdt.nosquash_nids" \
3832 wait_update $HOSTNAME \
3833 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3835 error "check llite nosquash_nids failed!"
3837 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3838 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3839 error "$ST: root read permission is denied"
3840 echo "$ST: root read permission is granted - ok"
3843 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3844 error "$ST: root write permission is denied"
3845 echo "$ST: root write permission is granted - ok"
3847 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3848 rm $DIR/$tdir-rootdir/tfile-1 ||
3849 error "$ST: root unlink permission is denied"
3850 echo "$ST: root unlink permission is granted - ok"
3851 touch $DIR/$tdir-rootdir/tfile-2 ||
3852 error "$ST: root create permission is denied"
3853 echo "$ST: root create permission is granted - ok"
3854 cleanup || error "cleanup failed with $?"
3856 run_test 43a "check root_squash and nosquash_nids"
3858 test_43b() { # LU-5690
3859 [[ "$MGS_VERSION" -ge $(version_code 2.7.62) ]] ||
3860 skip "Need MGS version 2.7.62+"
3862 if [[ -z "$fs2mds_DEV" ]]; then
3863 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
3864 skip_env "mixed loopback and real device not working"
3867 local fs2mdsdev=$(mdsdevname 1_2)
3868 local fs2mdsvdev=$(mdsvdevname 1_2)
3870 # temporarily use fs2mds as fs2mgs
3872 local fs2mgsdev=$fs2mdsdev
3873 local fs2mgsvdev=$fs2mdsvdev
3875 local fsname=test1234
3877 load_module llite/lustre
3878 local client_ip=$(host_nids_address $HOSTNAME $NETTYPE)
3879 local host=${client_ip//*./}
3880 local net=${client_ip/%$host/}
3881 local nosquash_nids=$(h2nettype $net[$host,$host,$host])
3883 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
3884 --param mdt.root_squash=$RUNAS_ID:$RUNAS_ID \
3885 --param mdt.nosquash_nids=$nosquash_nids \
3886 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
3887 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS || error "start fs2mgs failed"
3888 stop $fs2mgs -f || error "stop fs2mgs failed"
3889 cleanup || error "cleanup failed with $?"
3891 run_test 43b "parse nosquash_nids with commas in expr_list"
3895 check_mount || error "check_mount"
3896 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
3898 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
3899 for VAL in $UUIDS; do
3900 NID=$(echo $VAL | cut -d= -f1)
3901 CLUUID=$(echo $VAL | cut -d= -f2)
3902 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
3904 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
3905 cleanup || error "cleanup failed with $?"
3907 run_test 44 "mounted client proc entry exists"
3911 check_mount || error "check_mount"
3912 stop_mds || error "Unable to stop MDS"
3916 #define OBD_FAIL_PTLRPC_LONG_REPL_UNLINK 0x50f
3917 do_facet client "$LCTL set_param fail_loc=0x8000050f"
3920 manual_umount_client --force || error "manual_umount_client failed"
3921 do_facet client "$LCTL set_param fail_loc=0x0"
3922 start_mds || error "unable to start MDS"
3923 mount_client $MOUNT || error "mount_client $MOUNT failed"
3924 cleanup || error "cleanup failed with $?"
3926 run_test 45 "long unlink handling in ptlrpcd"
3933 umount_client $MOUNT2 || rc=$?
3934 umount_client $MOUNT || rc=$?
3935 while [ $count -gt 0 ]; do
3936 stop ost${count} -f || rc=$?
3940 cleanup_nocli || rc=$?
3941 #writeconf to remove all ost2 traces for subsequent tests
3942 writeconf_or_reformat
3947 echo "Testing with $OSTCOUNT OSTs"
3949 start_mds || error "unable to start MDS"
3950 #first client should see only one ost
3951 start_ost || error "Unable to start OST1"
3952 wait_osc_import_state mds ost FULL
3954 mount_client $MOUNT || error "mount_client $MOUNT failed"
3955 trap "cleanup_46a $OSTCOUNT" EXIT ERR
3958 for (( i=2; i<=$OSTCOUNT; i++ )); do
3959 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
3960 error "start_ost$i $(ostdevname $i) failed"
3963 # wait until osts in sync
3964 for (( i=2; i<=$OSTCOUNT; i++ )); do
3965 wait_osc_import_state mds ost$i FULL
3966 wait_osc_import_ready client ost$i
3969 #second client see all ost's
3971 mount_client $MOUNT2 || error "mount_client failed"
3972 $LFS setstripe -c -1 $MOUNT2 ||
3973 error "$LFS setstripe -c -1 $MOUNT2 failed"
3974 $LFS getstripe $MOUNT2 || error "$LFS getstripe $MOUNT2 failed"
3976 echo "ok" > $MOUNT2/widestripe
3977 $LFS getstripe $MOUNT2/widestripe ||
3978 error "$LFS getstripe $MOUNT2/widestripe failed"
3979 # fill acl buffer for avoid expand lsm to them
3980 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd |
3982 setfacl -m $acl $MOUNT2/widestripe
3986 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3988 cleanup_46a $OSTCOUNT || error "cleanup_46a failed"
3990 run_test 46a "handle ost additional - wide striped file"
3995 check_mount || error "check_mount failed"
3996 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
4000 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
4001 if echo $ns | grep "MDT[[:digit:]]*"; then
4004 lrs=$(echo $ns | sed 's/.*lru_size=//')
4005 lru_size[count]=$lrs
4010 facet_failover $SINGLEMDS
4011 client_up || error "client_up failed"
4014 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
4015 if echo $ns | grep "MDT[[:digit:]]*"; then
4018 lrs=$(echo $ns | sed 's/.*lru_size=//')
4019 if ! test "$lrs" -eq "${lru_size[count]}"; then
4020 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
4021 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
4026 cleanup || error "cleanup failed with $?"
4028 run_test 47 "server restart does not make client loss lru_resize settings"
4033 # reformat after this test is needed - if the test fails,
4034 # we will have unkillable file at FS
4038 test_48() { # bz-17636 LU-7473
4042 check_mount || error "check_mount failed"
4044 $LFS setstripe -c -1 $MOUNT ||
4045 error "$LFS setstripe -c -1 $MOUNT failed"
4046 $LFS getstripe $MOUNT || error "$LFS getstripe $MOUNT failed"
4048 echo "ok" > $MOUNT/widestripe
4049 $LFS getstripe $MOUNT/widestripe ||
4050 error "$LFS getstripe $MOUNT/widestripe failed"
4052 # In the future, we may introduce more EAs, such as selinux, enlarged
4053 # LOV EA, and so on. These EA will use some EA space that is shared by
4054 # ACL entries. So here we only check some reasonable ACL entries count,
4055 # instead of the max number that is calculated from the max_ea_size.
4056 if [ "$MDS1_VERSION" -lt $(version_code 2.8.57) ]; then
4057 count=28 # hard coded of RPC protocol
4058 elif large_xattr_enabled; then
4059 count=4500 # max_num 8187 max_ea_size = 65452
4060 # not create too many (4500) to save test time
4062 count=450 # max_num 497 max_ea_size = 4012
4065 echo "It is expected to hold at least $count ACL entries"
4066 trap cleanup_48 EXIT ERR
4067 for ((i = 0; i < $count; i++)) do
4068 setfacl -m u:$((i + 100)):rw $MOUNT/widestripe ||
4069 error "Fail to setfacl for $MOUNT/widestripe at $i"
4072 cancel_lru_locks mdc
4073 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
4074 local r_count=$(getfacl $MOUNT/widestripe | grep "user:" | wc -l)
4076 count=$((count + 1)) # for the entry "user::rw-"
4078 [ $count -eq $r_count ] ||
4079 error "Expected ACL entries $count, but got $r_count"
4083 run_test 48 "too many acls on file"
4085 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
4086 test_49a() { # bug 17710
4087 local timeout_orig=$TIMEOUT
4088 local ldlm_timeout_orig=$LDLM_TIMEOUT
4089 local LOCAL_TIMEOUT=20
4091 LDLM_TIMEOUT=$LOCAL_TIMEOUT
4092 TIMEOUT=$LOCAL_TIMEOUT
4096 check_mount || error "client mount failed"
4098 echo "check ldlm_timout..."
4099 local LDLM_MDS="$(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 $((LOCAL_TIMEOUT / 3)) ]; then
4108 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
4111 umount_client $MOUNT || error "umount_client $MOUNT failed"
4112 stop_ost || error "problem stopping OSS"
4113 stop_mds || error "problem stopping MDS"
4115 LDLM_TIMEOUT=$ldlm_timeout_orig
4116 TIMEOUT=$timeout_orig
4118 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
4120 test_49b() { # bug 17710
4121 local timeout_orig=$TIMEOUT
4122 local ldlm_timeout_orig=$LDLM_TIMEOUT
4123 local LOCAL_TIMEOUT=20
4125 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
4126 TIMEOUT=$LOCAL_TIMEOUT
4130 check_mount || error "client mount failed"
4132 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
4133 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
4134 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
4136 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
4137 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
4140 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
4141 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
4144 cleanup || error "cleanup failed"
4146 LDLM_TIMEOUT=$ldlm_timeout_orig
4147 TIMEOUT=$timeout_orig
4149 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
4152 # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1
4154 # Test both statfs and lfs df and fail if either one fails
4155 multiop_bg_pause $1 f_
4158 killall -USR1 multiop
4159 [ $RC -ne 0 ] && log "lazystatfs multiop failed"
4160 wait $PID || { RC=$?; log "multiop return error "; }
4162 # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1
4167 if kill -s 0 $PID; then
4170 log "lazystatfs lfs df failed to complete in 5s"
4178 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4179 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4181 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
4183 cleanup || error "cleanup failed with rc $?"
4185 run_test 50a "lazystatfs all servers available"
4189 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4190 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4192 # Wait for client to detect down OST
4193 stop_ost || error "Unable to stop OST1"
4194 wait_osc_import_state client ost DISCONN
4196 log "OSCs should all be DISCONN"
4198 lazystatfs $MOUNT || error "lazystatfs should not return EIO"
4200 umount_client $MOUNT || error "Unable to unmount client"
4201 stop_mds || error "Unable to stop MDS"
4203 run_test 50b "lazystatfs all servers down"
4206 start_mds || error "Unable to start MDS"
4207 start_ost || error "Unable to start OST1"
4208 start_ost2 || error "Unable to start OST2"
4209 mount_client $MOUNT || error "Unable to mount client"
4210 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4211 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4213 # Wait for client to detect down OST
4214 stop_ost || error "Unable to stop OST1"
4215 wait_osc_import_state mds ost DISCONN
4216 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
4218 umount_client $MOUNT || error "Unable to unmount client"
4219 stop_ost2 || error "Unable to stop OST2"
4220 stop_mds || error "Unable to stop MDS"
4221 #writeconf to remove all ost2 traces for subsequent tests
4222 writeconf_or_reformat
4224 run_test 50c "lazystatfs one server down"
4227 start_mds || error "Unable to start MDS"
4228 start_ost || error "Unable to start OST1"
4229 start_ost2 || error "Unable to start OST2"
4230 mount_client $MOUNT || error "Unable to mount client"
4231 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4232 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4234 # Issue the statfs during the window where the client still
4235 # belives the OST to be available but it is in fact down.
4236 # No failure just a statfs which hangs for a timeout interval.
4237 stop_ost || error "Unable to stop OST1"
4238 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
4240 umount_client $MOUNT || error "Unable to unmount client"
4241 stop_ost2 || error "Unable to stop OST2"
4242 stop_mds || error "Unable to stop MDS"
4243 #writeconf to remove all ost2 traces for subsequent tests
4244 writeconf_or_reformat
4246 run_test 50d "lazystatfs client/server conn race"
4253 start_mds || error "Unable to start MDS"
4254 #first client should see only one ost
4255 start_ost || error "Unable to start OST1"
4256 wait_osc_import_state mds ost FULL
4258 # Wait for client to detect down OST
4259 stop_ost || error "Unable to stop OST1"
4260 wait_osc_import_state mds ost DISCONN
4262 mount_client $MOUNT || error "Unable to mount client"
4263 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
4265 multiop_bg_pause $MOUNT _f
4269 if [ $RC1 -ne 0 ]; then
4270 log "multiop failed $RC1"
4273 sleep $(( $TIMEOUT+1 ))
4275 [ $? -ne 0 ] && error "process isn't sleep"
4276 start_ost || error "Unable to start OST1"
4277 wait $pid || error "statfs failed"
4280 umount_client $MOUNT || error "Unable to unmount client"
4281 stop_ost || error "Unable to stop OST1"
4282 stop_mds || error "Unable to stop MDS"
4284 run_test 50e "normal statfs all servers down"
4289 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
4291 start_mds || error "Unable to start mds"
4292 #first client should see only one ost
4293 start_ost || error "Unable to start OST1"
4294 wait_osc_import_state mds ost FULL
4296 start_ost2 || error "Unable to start OST2"
4297 wait_osc_import_state mds ost2 FULL
4299 # Wait for client to detect down OST
4300 stop_ost2 || error "Unable to stop OST2"
4302 wait_osc_import_state mds ost2 DISCONN
4303 mount_client $MOUNT || error "Unable to mount client"
4304 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
4306 multiop_bg_pause $MOUNT _f
4310 if [ $RC1 -ne 0 ]; then
4311 log "lazystatfs multiop failed $RC1"
4314 sleep $(( $TIMEOUT+1 ))
4316 [ $? -ne 0 ] && error "process isn't sleep"
4317 start_ost2 || error "Unable to start OST2"
4318 wait $pid || error "statfs failed"
4319 stop_ost2 || error "Unable to stop OST2"
4322 umount_client $MOUNT -f || error "Unable to unmount client"
4323 stop_ost || error "Unable to stop OST1"
4324 stop_mds || error "Unable to stop MDS"
4325 #writeconf to remove all ost2 traces for subsequent tests
4326 writeconf_or_reformat
4328 run_test 50f "normal statfs one server in down"
4331 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs"
4333 start_ost2 || error "Unable to start OST2"
4334 wait_osc_import_state mds ost2 FULL
4335 wait_osc_import_ready client ost2
4337 if [[ $PERM_CMD == *"set_param -P"* ]]; then
4338 local PARAM="osc.${FSNAME}-OST0001*.active"
4340 local PARAM="${FSNAME}-OST0001.osc.active"
4343 $LFS setstripe -c -1 $DIR/$tfile || error "$LFS setstripe failed"
4344 do_facet mgs $PERM_CMD $PARAM=0 || error "Unable to deactivate OST"
4346 umount_client $MOUNT || error "Unable to unmount client"
4347 mount_client $MOUNT || error "Unable to mount client"
4348 # This df should not cause a panic
4351 do_facet mgs $PERM_CMD $PARAM=1 || error "Unable to activate OST"
4352 rm -f $DIR/$tfile || error "unable to remove file $DIR/$tfile"
4353 umount_client $MOUNT || error "Unable to unmount client"
4354 stop_ost2 || error "Unable to stop OST2"
4355 stop_ost || error "Unable to stop OST1"
4356 stop_mds || error "Unable to stop MDS"
4357 #writeconf to remove all ost2 traces for subsequent tests
4358 writeconf_or_reformat
4360 run_test 50g "deactivated OST should not cause panic"
4364 # prepare MDT/OST, make OSC inactive for OST1
4365 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs"
4367 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
4368 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
4369 error "tunefs OST1 failed"
4370 start_mds || error "Unable to start MDT"
4371 start_ost || error "Unable to start OST1"
4372 start_ost2 || error "Unable to start OST2"
4373 mount_client $MOUNT || error "client start failed"
4375 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4377 # activatate OSC for OST1
4378 set_persistent_param_and_check client \
4379 "osc.${FSNAME}-OST0000-osc-[!M]*.active" \
4380 "${FSNAME}-OST0000.osc.active" 1
4382 mkdir $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
4383 $LFS setstripe -c -1 -i 0 $DIR/$tdir/2 ||
4384 error "$LFS setstripe $DIR/$tdir/2 failed"
4385 sleep 1 && echo "create a file after OST1 is activated"
4386 # doing some io, shouldn't crash
4387 dd if=/dev/zero of=$DIR/$tdir/2/$tfile-io bs=1M count=10
4389 # check OSC import is working
4390 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
4391 error "some OSC imports are still not connected"
4395 umount_client $MOUNT || error "Unable to umount client"
4396 stop_ost2 || error "Unable to stop OST2"
4397 cleanup_nocli || error "cleanup_nocli failed with $?"
4399 run_test 50h "LU-642: activate deactivated OST"
4402 # prepare MDT/OST, make OSC inactive for OST1
4403 [ "$MDSCOUNT" -lt "2" ] && skip "needs >= 2 MDTs"
4406 [ $(facet_fstype mds2) == zfs ] && import_zpool mds2
4407 do_facet mds2 "$TUNEFS --param mdc.active=0 $(mdsdevname 2)" ||
4408 error "tunefs MDT2 failed"
4409 start_mds || error "Unable to start MDT"
4410 start_ost || error "Unable to start OST1"
4411 start_ost2 || error "Unable to start OST2"
4412 mount_client $MOUNT || error "client start failed"
4414 mkdir_on_mdt0 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4416 if [[ $PERM_CMD == *"set_param -P"* ]]; then
4417 $PERM_CMD mdc.${FSNAME}-MDT0001-mdc-*.active=0 &&
4418 error "deactive MDC0 succeeds"
4420 $PERM_CMD ${FSNAME}-MDT0000.mdc.active=0 &&
4421 error "deactive MDC0 succeeds"
4424 # activate MDC for MDT2
4425 set_persistent_param_and_check client \
4426 "mdc.${FSNAME}-MDT0001-mdc-*.active" \
4427 "${FSNAME}-MDT0001.mdc.active" 1
4429 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
4430 if [ "$MDS1_VERSION" -ge $(version_code 2.7.60) ]
4432 wait_dne_interconnect
4434 $LFS mkdir -i1 $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
4436 createmany -o $DIR/$tdir/2/$tfile-%d 1 || error "create files failed"
4438 rm -rf $DIR/$tdir/2 || error "unlink dir failed"
4440 # deactivate MDC for MDT2
4441 set_persistent_param_and_check client \
4442 "mdc.${FSNAME}-MDT0001-mdc-*.active" \
4443 "${FSNAME}-MDT0001.mdc.active" 0
4445 wait_osp_active mds ${FSNAME}-MDT0001 1 0
4447 $LFS mkdir -i1 $DIR/$tdir/2 &&
4448 error "mkdir $DIR/$tdir/2 succeeds after deactive MDT"
4450 $LFS mkdir -i0 -c$MDSCOUNT $DIR/$tdir/striped_dir ||
4451 error "mkdir $DIR/$tdir/striped_dir fails after deactive MDT2"
4453 local stripe_count=$($LFS getdirstripe -c $DIR/$tdir/striped_dir)
4454 [ $stripe_count -eq $((MDSCOUNT - 1)) ] ||
4455 error "wrong $stripe_count != $((MDSCOUNT -1)) for striped_dir"
4458 umount_client $MOUNT || error "Unable to umount client"
4463 run_test 50i "activate deactivated MDT"
4466 local LOCAL_TIMEOUT=20
4470 check_mount || error "check_mount failed"
4472 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
4473 $LFS setstripe -c -1 $MOUNT/$tdir ||
4474 error "$LFS setstripe -c -1 $MOUNT/$tdir failed"
4475 #define OBD_FAIL_MDS_REINT_DELAY 0x142
4476 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x142"
4477 touch $MOUNT/$tdir/$tfile &
4480 start_ost2 || error "Unable to start OST1"
4482 stop_ost2 || error "Unable to stop OST1"
4483 umount_client $MOUNT -f || error “unmount $MOUNT failed”
4484 cleanup_nocli || error “stop server failed”
4485 #writeconf to remove all ost2 traces for subsequent tests
4486 writeconf_or_reformat
4488 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
4497 do_node $node mkdir -p $dest
4498 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
4500 do_node $node 'tar cf - '$*' | tar xf - -C '$dest';
4501 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
4502 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
4504 do_node $node 'getfattr -d -m "[a-z]*\\." '$*' > '$xattrs
4505 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
4515 local backup2=${TMP}/backup2
4517 do_node $node mkdir -p $backup2
4518 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
4520 do_node $node 'tar cf - '$*' | tar xf - -C '$backup2';
4521 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
4522 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
4524 do_node $node "diff -rq $backup $backup2"
4525 [ $? -eq 0 ] || { error "contents differ"; return 3; }
4527 local xattrs2=${TMP}/xattrs2
4528 do_node $node 'getfattr -d -m "[a-z]*\\." '$*' > '$xattrs2
4529 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
4531 do_node $node "diff $xattrs $xattrs2"
4532 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
4534 do_node $node "rm -rf $backup2 $xattrs2"
4535 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
4539 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4540 skip "ldiskfs only test"
4543 start_mds || error "Unable to start MDS"
4544 start_ost || error "Unable to start OST1"
4545 mount_client $MOUNT || error "Unable to mount client"
4548 local ost1mnt=$(facet_mntpt ost1)
4549 local ost1node=$(facet_active_host ost1)
4550 local ost1tmp=$TMP/conf52
4553 mkdir $DIR/$tdir || error "Unable to create $DIR/$tdir"
4554 touch $TMP/modified_first || error "Unable to create temporary file"
4555 local mtime=$(stat -c %Y $TMP/modified_first)
4556 do_node $ost1node "mkdir -p $ost1tmp &&
4557 touch -m -d @$mtime $ost1tmp/modified_first" ||
4558 error "Unable to create temporary file"
4561 $LFS setstripe -c -1 -S 1M $DIR/$tdir || error "$LFS setstripe failed"
4563 for (( i=0; i < nrfiles; i++ )); do
4564 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c ||
4565 error "multiop failed"
4570 # sync all the data and make sure no pending data on the client,
4571 # thus the SOM xattr would not be changed any more.
4572 cancel_lru_locks osc
4575 echo backup files to $TMP/$tdir
4576 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
4577 copy_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
4578 error "Unable to copy files"
4580 umount_client $MOUNT || error "Unable to umount client"
4581 stop_ost || error "Unable to stop ost1"
4583 echo mount ost1 as ldiskfs
4584 do_node $ost1node mkdir -p $ost1mnt || error "Unable to create $ost1mnt"
4585 if ! do_node $ost1node test -b $ost1_dev; then
4588 do_node $ost1node mount -t "$ost1_FSTYPE" $loop $ost1_dev \
4590 error "Unable to mount ost1 as ldiskfs"
4593 echo backup objects to $ost1tmp/objects
4594 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
4595 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
4596 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
4598 error "Unable to copy objects"
4600 # move objects to lost+found
4601 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
4602 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
4604 do_node $ost1node "umount $ost1mnt" ||
4605 error "Unable to umount ost1 as ldiskfs"
4607 start_ost || error "Unable to start OST1"
4608 mount_client $MOUNT || error "Unable to mount client"
4610 local REPAIRED=$(do_node $ost1node "$LCTL get_param \
4611 -n osd-ldiskfs.$FSNAME-OST0000.oi_scrub" |
4612 awk '/^lf_repa[ri]*ed/ { print $2 }')
4613 [ $REPAIRED -gt 0 ] ||
4614 error "Some entry under /lost+found should be repaired"
4617 diff_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
4618 error "Unable to diff files"
4620 rm -rf $TMP/$tdir $TMP/file_xattrs ||
4621 error "Unable to delete temporary files"
4622 do_node $ost1node "rm -rf $ost1tmp" ||
4623 error "Unable to delete temporary files"
4624 cleanup || error "cleanup failed with $?"
4626 run_test 52 "check recovering objects from lost+found"
4628 # Checks threads_min/max/started for some service
4630 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
4631 # parameter pattern prefix like 'ost.*.ost'.
4644 local msg="Insane $modname thread counts"
4645 local ncpts=$(check_cpt_number $facet)
4649 check_mount || return 41
4651 # We need to expand $parampat, but it may match multiple parameters, so
4652 # we'll pick the first one
4653 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
4654 error "Couldn't expand ${parampat}.threads_min parameter name"
4658 # Remove the .threads_min part
4659 paramp=${paramp%.threads_min}
4661 # Check for sanity in defaults
4662 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4664 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4666 tstarted=$(do_facet $facet "$LCTL get_param \
4667 -n ${paramp}.threads_started" || echo 0)
4668 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' ||
4670 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4672 nthrs=$(expr $tmax - $tmin)
4673 if [ $nthrs -lt $ncpts ]; then
4679 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
4680 skip_env "module parameter forced $facet thread count"
4682 # Check that we can change min/max
4683 do_facet $facet "$LCTL set_param \
4684 ${paramp}.threads_min=$((tmin + nthrs))"
4685 do_facet $facet "$LCTL set_param \
4686 ${paramp}.threads_max=$((tmax - nthrs))"
4687 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4689 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4691 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) &&
4692 $tmax2 == ($tmax - $nthrs)))' || return $?
4694 # Check that we can set min/max to the same value
4695 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4697 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$tmin"
4698 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4700 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4702 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
4704 # Check that we can't set max < min
4705 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$((tmin - 1))"
4706 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4708 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4710 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
4712 # We need to ensure that we get the module options desired; to do this
4713 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
4714 LOAD_MODULES_REMOTE=true
4717 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
4719 setmodopts -a $modname "$newvalue" oldvalue
4722 check_mount || return 41
4724 # Restore previous setting of MODOPTS_*
4725 setmodopts $modname "$oldvalue"
4727 # Check that $opts took
4728 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4730 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4732 tstarted=$(do_facet $facet \
4733 "$LCTL get_param -n ${paramp}.threads_started" || echo 0)
4734 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4743 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
4744 cleanup || error "cleanup failed with rc $?"
4746 run_test 53a "check OSS thread count params"
4750 local mds=$(do_facet $SINGLEMDS "$LCTL get_param \
4751 -N mds.*.*.threads_max 2>/dev/null")
4752 if [ -z "$mds" ]; then
4753 #running this on an old MDT
4754 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
4756 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
4758 cleanup || error "cleanup failed with $?"
4760 run_test 53b "check MDS thread count params"
4763 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4764 skip "ldiskfs only test"
4767 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p ||
4768 error "llverdev failed with rc=$?"
4771 run_test 54a "test llverdev and partial verify of device"
4774 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4775 skip "ldiskfs only test"
4779 run_llverfs $MOUNT -p || error "llverfs failed with rc=$?"
4780 cleanup || error "cleanup failed with rc=$?"
4782 run_test 54b "test llverfs and partial verify of filesystem"
4786 local max_ost_index=$1
4788 echo -n $(((max_ost_index + 1) * 8))
4792 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4793 skip "ldiskfs only test"
4796 local mdsdev=$(mdsdevname 1)
4797 local mdsvdev=$(mdsvdevname 1)
4801 if ! combined_mgs_mds; then
4802 stop_mgs || error "stopping MGS service failed"
4803 format_mgs || error "formatting MGT failed"
4805 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
4807 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
4808 --reformat $(ostdevname 1) $(ostvdevname 1)
4814 echo checking size of lov_objid for ost index $i
4815 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" |
4816 grep ^User | awk -F 'Size: ' '{print $2}')
4817 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
4818 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
4820 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
4827 run_test 55 "check lov_objid size"
4830 local mds_journal_size_orig=$MDSJOURNALSIZE
4836 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
4837 $(ostdevname 1) $(ostvdevname 1)
4838 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
4839 $(ostdevname 2) $(ostvdevname 2)
4842 start_ost || error "Unable to start first ost (idx 10000)"
4843 start_ost2 || error "Unable to start second ost (idx 1000)"
4844 mount_client $MOUNT || error "Unable to mount client"
4848 # test instantiating PFL components with sparse index LU-15513
4849 mkdir -p $MOUNT/$tdir
4850 $LFS setstripe -E 4M -c 1 -E 1G -c 4 -S4M -E eof -c -1 $MOUNT/$tdir
4851 dd if=/dev/zero of=$MOUNT/$tdir/$tfile bs=4K count=1 seek=10k ||
4852 error "dd to second component failed"
4854 if [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
4855 [[ "$MDS1_VERSION" -ge $(version_code 2.5.4) &&
4856 "$MDS1_VERSION" -lt $(version_code 2.5.11) ]]; then
4857 wait_osc_import_state mds ost1 FULL
4858 wait_osc_import_state mds ost2 FULL
4859 $LFS setstripe --stripe-count=-1 $DIR/$tfile ||
4860 error "Unable to setstripe $DIR/$tfile"
4861 n=$($LFS getstripe --stripe-count $DIR/$tfile)
4862 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
4867 MDSJOURNALSIZE=$mds_journal_size_orig
4870 run_test 56a "check big OST indexes and out-of-index-order start"
4875 umount_client $MOUNT -f || error "unmount client failed"
4884 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs"
4886 trap cleanup_56b EXIT RETURN ERR
4889 if ! combined_mgs_mds ; then
4894 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) --index=0 --reformat \
4895 $(mdsdevname 1) $(mdsvdevname 1)
4896 add mds2 $(mkfs_opts mds2 $(mdsdevname 2)) --index=1 --reformat \
4897 $(mdsdevname 2) $(mdsvdevname 2)
4898 add mds3 $(mkfs_opts mds3 $(mdsdevname 3)) --index=1000 --reformat \
4899 $(mdsdevname 3) $(mdsvdevname 3)
4903 start_mdt 1 || error "MDT 1 (idx 0) start failed"
4904 start_mdt 2 || error "MDT 2 (idx 1) start failed"
4905 start_mdt 3 || error "MDT 3 (idx 1000) start failed"
4906 start_ost || error "Unable to start first ost"
4907 start_ost2 || error "Unable to start second ost"
4909 do_nodes $(comma_list $(mdts_nodes)) \
4910 "$LCTL set_param mdt.*.enable_remote_dir=1 \
4911 mdt.*.enable_remote_dir_gid=-1"
4913 mount_client $MOUNT || error "Unable to mount client"
4915 $LFS mkdir -c3 $MOUNT/$tdir || error "failed to make testdir"
4917 echo "This is test file 1!" > $MOUNT/$tdir/$tfile.1 ||
4918 error "failed to make test file 1"
4919 echo "This is test file 2!" > $MOUNT/$tdir/$tfile.2 ||
4920 error "failed to make test file 2"
4921 echo "This is test file 1000!" > $MOUNT/$tdir/$tfile.1000 ||
4922 error "failed to make test file 1000"
4924 rm -rf $MOUNT/$tdir || error "failed to remove testdir"
4926 $LFS mkdir -i1000 $MOUNT/$tdir.1000 ||
4927 error "create remote dir at idx 1000 failed"
4930 echo "=== START lfs df OUTPUT ==="
4932 echo "==== END lfs df OUTPUT ===="
4934 mdtcnt=$(echo -e "$output" | grep $FSNAME-MDT | wc -l)
4935 ostcnt=$(echo -e "$output" | grep $FSNAME-OST | wc -l)
4937 echo "lfs df returned mdt count $mdtcnt and ost count $ostcnt"
4938 [ $mdtcnt -eq 3 ] || error "lfs df returned wrong mdt count"
4939 [ $ostcnt -eq 2 ] || error "lfs df returned wrong ost count"
4941 echo "This is test file 1!" > $MOUNT/$tdir.1000/$tfile.1 ||
4942 error "failed to make test file 1"
4943 echo "This is test file 2!" > $MOUNT/$tdir.1000/$tfile.2 ||
4944 error "failed to make test file 2"
4945 echo "This is test file 1000!" > $MOUNT/$tdir.1000/$tfile.1000 ||
4946 error "failed to make test file 1000"
4947 rm -rf $MOUNT/$tdir.1000 || error "failed to remove remote_dir"
4950 echo "=== START lfs mdts OUTPUT ==="
4952 echo "==== END lfs mdts OUTPUT ===="
4954 echo -e "$output" | grep -v "MDTS:" | awk '{print $1}' |
4955 sed 's/://g' > $TMP/mdts-actual.txt
4956 sort $TMP/mdts-actual.txt -o $TMP/mdts-actual.txt
4958 echo -e "0\n1\n1000" > $TMP/mdts-expected.txt
4960 diff $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4963 rm $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4965 [ $result -eq 0 ] || error "target_obd proc file is incorrect!"
4967 run_test 56b "test target_obd correctness with nonconsecutive MDTs"
4969 test_57a() { # bug 22656
4970 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4971 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4972 tail -1 | awk '{print $1}')
4973 writeconf_or_reformat
4974 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
4975 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" ||
4976 error "tunefs failed"
4978 start_ost && error "OST registration from failnode should fail"
4981 run_test 57a "initial registration from failnode should fail (should return errs)"
4984 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4985 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4986 tail -1 | awk '{print $1}')
4987 writeconf_or_reformat
4988 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
4989 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" ||
4990 error "tunefs failed"
4992 start_ost || error "OST registration from servicenode should not fail"
4995 run_test 57b "initial registration from servicenode should not fail"
4998 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
5001 test_58() { # bug 22658
5002 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
5004 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5005 createmany -o $DIR/$tdir/$tfile-%d 100
5006 unlinkmany $DIR/$tdir/$tfile-%d 100
5007 stop_mds || error "Unable to stop MDS"
5009 local MNTDIR=$(facet_mntpt $SINGLEMDS)
5010 local devname=$(mdsdevname ${SINGLEMDS//mds/})
5012 # remove all files from the OBJECTS dir
5013 mount_fstype $SINGLEMDS
5015 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
5017 unmount_fstype $SINGLEMDS
5018 # restart MDS with missing llog files
5019 start_mds || error "unable to start MDS"
5020 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0"
5023 run_test 58 "missing llog files must not prevent MDT from mounting"
5026 start_mgsmds >> /dev/null
5027 local C1=$(count_osts)
5028 if [ $C1 -eq 0 ]; then
5029 start_ost >> /dev/null
5033 echo "original ost count: $C1 (expect > 0)"
5034 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
5035 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
5036 local C2=$(count_osts)
5037 echo "after mdt writeconf count: $C2 (expect 0)"
5038 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
5039 echo "OST start without writeconf should fail:"
5040 start_ost >> /dev/null &&
5041 error "OST start without writeconf didn't fail"
5042 echo "OST start with writeconf should succeed:"
5043 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
5044 local C3=$(count_osts)
5045 echo "after ost writeconf count: $C3 (expect 1)"
5046 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
5047 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
5048 local C4=$(count_osts)
5049 echo "after ost2 writeconf count: $C4 (expect 2)"
5050 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
5051 stop_ost2 >> /dev/null
5052 cleanup_nocli >> /dev/null
5053 #writeconf to remove all ost2 traces for subsequent tests
5054 writeconf_or_reformat
5056 run_test 59 "writeconf mount option"
5058 test_60a() { # LU-471
5059 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5060 skip "ldiskfs only test"
5065 for num in $(seq $MDSCOUNT); do
5066 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
5067 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
5068 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
5072 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
5073 [ ${PIPESTATUS[0]} -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
5075 # MDT default has dirdata feature
5076 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
5077 # we disable uninit_bg feature
5078 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
5079 # we set stride extended options
5080 echo $dump | grep stride > /dev/null || error "stride is not set"
5084 run_test 60a "check mkfs.lustre --mkfsoptions -E -O options setting"
5087 [[ "$mds1_FSTYPE" == ldiskfs ]] || skip "ldiskfs only test"
5089 local features=$(do_facet $SINGLEMDS $DUMPE2FS $(mdsdevname 1) |
5091 [ ${PIPESTATUS[0]} -eq 0 ] || error "$DUMPE2FS $(mdsdevname 1) failed"
5094 # ea_inode feature should be enabled by default for MDTs
5095 [[ "$features" =~ "ea_inode" ]] || error "ea_inode is not set"
5096 # large_dir feature should be enabled by default for MDTs
5097 [[ "$features" =~ "large_dir" ]] || error "large_dir is not set"
5099 run_test 60b "check mkfs.lustre MDT default features"
5101 test_61a() { # LU-80
5102 local lxattr=$(large_xattr_enabled)
5104 (( "$MDS1_VERSION" >= $(version_code 2.1.53) )) ||
5105 skip "Need MDS version at least 2.1.53 for large_xattr"
5107 if [[ "$mds1_FSTYPE" == ldiskfs ]] && ! large_xattr_enabled; then
5110 for ((num=1; num <= $MDSCOUNT; num++)); do
5111 do_facet mds${num} $TUNE2FS -O ea_inode \
5112 $(mdsdevname $num) ||
5113 error "tune2fs on mds $num failed"
5117 setup || error "setting up the filesystem failed"
5118 client_up || error "starting client failed"
5120 local file=$DIR/$tfile
5121 touch $file || error "touch $file failed"
5123 local large_value="$(generate_string $(max_xattr_size))"
5124 local small_value="bar"
5126 local name="trusted.big"
5127 log "save large xattr of $(max_xattr_size) bytes on $name on $file"
5128 setfattr -n $name -v $large_value $file ||
5129 error "saving $name on $file failed"
5131 local new_value=$(get_xattr_value $name $file)
5132 [[ "$new_value" != "$large_value" ]] &&
5133 error "$name different after saving"
5135 log "shrink value of $name on $file"
5136 setfattr -n $name -v $small_value $file ||
5137 error "shrinking value of $name on $file failed"
5139 new_value=$(get_xattr_value $name $file)
5140 [[ "$new_value" != "$small_value" ]] &&
5141 error "$name different after shrinking"
5143 log "grow value of $name on $file"
5144 setfattr -n $name -v $large_value $file ||
5145 error "growing value of $name on $file failed"
5147 new_value=$(get_xattr_value $name $file)
5148 [[ "$new_value" != "$large_value" ]] &&
5149 error "$name different after growing"
5151 log "check value of $name on $file after remounting MDS"
5153 new_value=$(get_xattr_value $name $file)
5154 [[ "$new_value" != "$large_value" ]] &&
5155 error "$name different after remounting MDS"
5157 log "remove large xattr $name from $file"
5158 setfattr -x $name $file || error "removing $name from $file failed"
5160 if $lxattr && [ "$mds1_FSTYPE" == ldiskfs ]; then
5161 stopall || error "stopping for e2fsck run"
5162 for num in $(seq $MDSCOUNT); do
5163 run_e2fsck $(facet_active_host mds$num) \
5164 $(mdsdevname $num) "-y" ||
5165 error "e2fsck MDT$num failed"
5167 setup_noconfig || error "remounting the filesystem failed"
5170 # need to delete this file to avoid problems in other tests
5172 cleanup || error "stopping systems failed"
5174 run_test 61a "large xattr"
5176 test_61b() { # LU-80
5177 local lxattr=$(large_xattr_enabled)
5179 (( $MDS1_VERSION >= $(version_code 2.15.51) )) ||
5180 skip "Need MDS version at least 2.15.51 for large_xattr fix"
5182 [[ "$mds1_FSTYPE" == "ldiskfs" ]] || skip "ldiskfs specific bug"
5184 if ! large_xattr_enabled; then
5187 for (( num=1; num <= $MDSCOUNT; num++ )); do
5188 do_facet mds${num} $TUNE2FS -O ea_inode \
5189 $(mdsdevname $num) ||
5190 error "tune2fs on mds $num failed"
5194 setup || error "setting up the filesystem failed"
5195 client_up || error "starting client failed"
5197 local _file=$MOUNT/panda
5198 local large_value="$(generate_string $(max_xattr_size))"
5199 local name="trusted.big"
5201 touch ${_file} || error "touch ${_file} failed"
5202 setfattr -n $name -v $large_value ${_file} ||
5203 error "saving $name on $file failed"
5205 MDT_DEV="${FSNAME}-MDT0000"
5206 MDT_DEVNAME=$(mdsdevname ${SINGLEMDS//mds/})
5208 stopall || error "stopping for e2fsck run"
5211 ino=$(do_facet $SINGLEMDS "$DEBUGFS -R 'stat /ROOT/panda' \
5212 ${MDT_DEVNAME} | grep trusted.big")
5213 ino=$(echo "${ino}" | awk '{print $2;}')
5214 echo "large ea "${ino}
5216 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"ln $ino /lost+found\\\" \
5219 setup_noconfig || error "remounting the filesystem failed"
5221 do_facet $SINGLEMDS $LCTL lfsck_start -M ${MDT_DEV} -t namespace || {
5222 error "can't start lfsck namespace"
5226 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
5227 mdd.${MDT_DEV}.lfsck_namespace |
5228 awk '/^status/ { print \\\$2 }'" "completed" 32 || {
5229 error "(2) unexpected status"
5232 stopall || error "stopping for e2fsck run"
5233 for num in $(seq $MDSCOUNT); do
5234 run_e2fsck $(facet_active_host mds$num) \
5235 $(mdsdevname $num) "-y" ||
5236 error "e2fsck MDT$num failed"
5238 setup_noconfig || error "remounting the filesystem failed"
5240 # need to delete this file to avoid problems in other tests
5242 cleanup || error "stopping systems failed"
5244 run_test 61b "large xattr"
5247 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5248 skip "ldiskfs only test"
5250 [[ "$MDS1_VERSION" -ge $(version_code 2.2.51) ]] ||
5251 skip "Need MDS version at least 2.2.51"
5254 local mdsdev=$(mdsdevname 1)
5255 local ostdev=$(ostdevname 1)
5257 echo "disable journal for mds"
5258 do_facet mds1 $TUNE2FS -O ^has_journal $mdsdev || error "tune2fs failed"
5259 start_mds && error "MDT start should fail"
5260 echo "disable journal for ost"
5261 do_facet ost1 $TUNE2FS -O ^has_journal $ostdev || error "tune2fs failed"
5262 start_ost && error "OST start should fail"
5263 cleanup || error "cleanup failed with rc $?"
5266 run_test 62 "start with disabled journal"
5269 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5270 skip "ldiskfs only test"
5273 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ldiskfs
5274 local inode_slab=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
5275 awk '/ldiskfs_inode_cache/ { print $5 / $6 }')
5276 if [ -z "$inode_slab" ]; then
5277 skip "ldiskfs module has not been loaded"
5280 if grep -q "CONFIG_DEBUG_LOCK_ALLOC=y" /boot/config-$(uname -r); then
5281 skip "test is not compatible with CONFIG_DEBUG_LOCK_ALLOC=y"
5284 echo "$inode_slab ldiskfs inodes per page"
5285 [ "${inode_slab%.*}" -ge "3" ] && return 0
5287 # If kmalloc-128 is also 1 per page - this is a debug kernel
5288 # and so this is not an error.
5289 local kmalloc128=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
5290 awk '/^(dma-kmalloc|size)-128 / { print $5 / $6 }')
5291 # 32 128-byte chunks in 4k
5292 [ "${kmalloc128%.*}" -lt "32" ] ||
5293 error "ldiskfs inode too big, only $inode_slab objs/page, " \
5294 "kmalloc128 = $kmalloc128 objs/page"
5296 run_test 63 "Verify each page can at least hold 3 ldiskfs inodes"
5299 start_mds || error "unable to start MDS"
5300 start_ost || error "Unable to start OST1"
5301 start_ost2 || error "Unable to start second ost"
5302 mount_client $MOUNT || error "Unable to mount client"
5303 stop_ost2 || error "Unable to stop second ost"
5306 umount_client $MOUNT -f || error “unmount $MOUNT failed”
5307 cleanup_nocli || error "cleanup_nocli failed with $?"
5308 #writeconf to remove all ost2 traces for subsequent tests
5309 writeconf_or_reformat
5311 run_test 64 "check lfs df --lazy "
5313 test_65() { # LU-2237
5314 # Currently, the test is only valid for ldiskfs backend
5315 [ "$mds1_FSTYPE" != ldiskfs ] &&
5316 skip "ldiskfs only test"
5318 local devname=$(mdsdevname ${SINGLEMDS//mds/})
5319 local brpt=$(facet_mntpt brpt)
5322 if ! do_facet $SINGLEMDS "test -b $devname"; then
5326 stop_mds || error "Unable to stop MDS"
5327 local obj=$(do_facet $SINGLEMDS \
5328 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
5330 if [ -z "$obj" ]; then
5331 # The MDT may be just re-formatted, mount the MDT for the
5332 # first time to guarantee the "last_rcvd" file is there.
5333 start_mds || error "fail to mount the MDS for the first time"
5334 stop_mds || error "Unable to stop MDS"
5337 # remove the "last_rcvd" file
5338 do_facet $SINGLEMDS "mkdir -p $brpt"
5339 do_facet $SINGLEMDS \
5340 "mount -t $mds1_FSTYPE $opts $devname $brpt"
5341 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
5342 do_facet $SINGLEMDS "$UMOUNT $brpt"
5344 # restart MDS, the "last_rcvd" file should be recreated.
5345 start_mds || error "fail to restart the MDS"
5346 stop_mds || error "Unable to stop MDS"
5347 obj=$(do_facet $SINGLEMDS \
5348 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
5349 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
5351 run_test 65 "re-create the lost last_rcvd file when server mount"
5354 [[ "$MGS_VERSION" -ge $(version_code 2.3.59) ]] ||
5355 skip "Need MGS version at least 2.3.59"
5358 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
5359 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
5361 # add EXCLUDE records to config log, they are not to be
5362 # removed by lctl replace_nids
5363 set_conf_param_and_check mds \
5364 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
5365 "$FSNAME-OST0000.osc.active" \
5368 echo "replace_nids should fail if MDS, OSTs and clients are UP"
5369 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5370 error "replace_nids fail"
5372 umount_client $MOUNT || error "unmounting client failed"
5373 echo "replace_nids should fail if MDS and OSTs are UP"
5374 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5375 error "replace_nids fail"
5377 stop_ost || error "Unable to stop OST1"
5378 echo "replace_nids should fail if MDS is UP"
5379 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5380 error "replace_nids fail"
5382 stop_mds || error "stopping mds failed"
5384 if combined_mgs_mds; then
5385 start_mdt 1 "-o nosvc" ||
5386 error "starting mds with nosvc option failed"
5389 echo "command should accept two parameters"
5390 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
5391 error "command should accept two params"
5393 echo "correct device name should be passed"
5394 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
5395 error "wrong devname"
5397 echo "wrong nids list should not destroy the system"
5398 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
5400 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "asdfasdf, asdfadf" &&
5403 echo "replace OST nid"
5404 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
5405 error "replace nids failed"
5407 echo "command should accept two parameters"
5408 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
5409 error "command should accept two params"
5411 echo "wrong nids list should not destroy the system"
5412 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
5415 local FAKE_NIDS="192.168.0.112@tcp1,192.168.0.112@tcp2"
5416 local FAKE_FAILOVER="192.168.0.113@tcp1,192.168.0.113@tcp2"
5417 local NIDS_AND_FAILOVER="$MDS_NID,$FAKE_NIDS:$FAKE_FAILOVER"
5418 echo "set NIDs with failover"
5419 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $NIDS_AND_FAILOVER ||
5420 error "replace nids failed"
5423 echo "replace MDS nid"
5424 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
5425 error "replace nids failed"
5427 if ! combined_mgs_mds ; then
5430 stop_mds || error "Unable to stop MDS"
5433 start_mgsmds || error "start mgsmds failed"
5434 set_conf_param_and_check mds \
5435 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
5436 "$FSNAME-OST0000.osc.active" \
5438 start_ost || error "unable to start OST"
5439 mount_client $MOUNT || error "mount client failed"
5441 check_mount || error "error after nid replace"
5442 cleanup || error "cleanup failed"
5445 run_test 66 "replace nids"
5447 test_67() { #LU-2950
5448 local legacy="$TMP/legacy_lnet_config"
5449 local new="$TMP/new_routes_test"
5450 local out="$TMP/config_out_file"
5451 local verify="$TMP/conv_verify"
5452 local verify_conf="$TMP/conf_verify"
5454 # Create the legacy file that will be run through the
5455 # lustre_routes_conversion script
5456 cat <<- LEGACY_LNET_CONFIG > $legacy
5457 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
5458 tcp2 54 10.1.3.2@tcp;
5459 tcp3 10.3.4.3@tcp:3;
5463 # Create the verification file to verify the output of
5464 # lustre_routes_conversion script against.
5465 cat <<- VERIFY_LNET_CONFIG > $verify
5466 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
5467 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
5468 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
5469 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
5470 tcp4: { gateway: 10.3.3.4@tcp }
5473 # Create the verification file to verify the output of
5474 # lustre_routes_config script against
5475 cat <<- VERIFY_LNET_CONFIG > $verify_conf
5476 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
5477 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
5478 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
5479 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
5480 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
5483 $LUSTRE_ROUTES_CONVERSION $legacy $new > /dev/null
5484 if [ -f $new ]; then
5485 # verify the conversion output
5486 cmp -s $new $verify > /dev/null
5487 if [ $? -eq 1 ]; then
5488 error "routes conversion failed"
5491 lustre_routes_config --dry-run --verbose $new > $out
5492 # check that the script succeeded
5493 cmp -s $out $verify_conf > /dev/null
5494 if [ $? -eq 1 ]; then
5495 error "routes config failed"
5498 error "routes conversion test failed"
5500 # remove generated files
5501 rm -f $new $legacy $verify $verify_conf $out
5503 run_test 67 "test routes conversion and configuration"
5511 [ "$MDS1_VERSION" -ge $(version_code 2.4.53) ] ||
5512 skip "Need MDS version at least 2.4.53"
5514 umount_client $MOUNT || error "umount client failed"
5519 # START-END - the sequences we'll be reserving
5520 START=$(do_facet $SINGLEMDS \
5521 $LCTL get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
5522 END=$((START + (1 << 30)))
5523 do_facet $SINGLEMDS \
5524 $LCTL set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
5526 # reset the sequences MDT0000 has already assigned
5527 do_facet $SINGLEMDS \
5528 $LCTL set_param seq.srv*MDT0000.space=clear
5530 # remount to let the client allocate new sequence
5531 mount_client $MOUNT || error "mount client failed"
5533 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
5534 do_facet $SINGLEMDS \
5535 $LCTL get_param seq.srv*MDT0000.space
5536 $LFS path2fid $DIR/$tfile
5538 local old_ifs="$IFS"
5540 fid=($($LFS path2fid $DIR/$tfile))
5544 if [[ $seq < $END ]]; then
5545 error "used reserved sequence $seq?"
5547 cleanup || error "cleanup failed with $?"
5549 run_test 68 "be able to reserve specific sequences in FLDB"
5551 # Test 69: is about the total number of objects ever created on an OST.
5552 # so that when it is reformatted the normal MDS->OST orphan recovery won't
5553 # just "precreate" the missing objects. In the past it might try to recreate
5554 # millions of objects after an OST was reformatted
5556 [[ "$MDS1_VERSION" -lt $(version_code 2.4.2) ]] &&
5557 skip "Need MDS version at least 2.4.2"
5559 [[ "$MDS1_VERSION" -ge $(version_code 2.4.50) ]] &&
5560 [[ "$MDS1_VERSION" -lt $(version_code 2.5.0) ]] &&
5561 skip "Need MDS version at least 2.5.0"
5564 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5566 # use OST0000 since it probably has the most creations
5567 local OSTNAME=$(ostname_from_index 0)
5568 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
5569 local last_id=$(do_facet mds1 $LCTL get_param -n \
5570 osp.$mdtosc_proc1.prealloc_last_id)
5572 # Want to have OST LAST_ID over 5 * OST_MAX_PRECREATE to
5573 # verify that the LAST_ID recovery is working properly. If
5574 # not, then the OST will refuse to allow the MDS connect
5575 # because the LAST_ID value is too different from the MDS
5576 #define OST_MAX_PRECREATE=20000
5577 local ost_max_pre=20000
5578 local num_create=$(( ost_max_pre * 5 + 1 - last_id))
5580 # If the LAST_ID is already over 5 * OST_MAX_PRECREATE, we don't
5581 # need to create any files. So, skip this section.
5582 if [ $num_create -gt 0 ]; then
5583 # Check the number of inodes available on OST0
5585 local ifree=$($LFS df -i $MOUNT |
5586 awk '/OST0000/ { print $4 }'; exit ${PIPESTATUS[0]})
5587 log "On OST0, $ifree inodes available. Want $num_create. rc=$?"
5589 $LFS setstripe -i 0 $DIR/$tdir ||
5590 error "$LFS setstripe -i 0 $DIR/$tdir failed"
5591 if [ $ifree -lt 10000 ]; then
5592 files=$(( ifree - 50 ))
5597 local j=$((num_create / files + 1))
5598 for i in $(seq 1 $j); do
5599 createmany -o $DIR/$tdir/$tfile-$i- $files ||
5600 error "createmany fail create $files files: $?"
5601 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
5602 error "unlinkmany failed unlink $files files"
5606 # delete all of the files with objects on OST0 so the
5607 # filesystem is not inconsistent later on
5608 $LFS find $MOUNT --ost 0 -print0 | xargs -0 rm
5610 umount_client $MOUNT || error "umount client failed"
5611 stop_ost || error "OST0 stop failure"
5612 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
5613 $(ostdevname 1) $(ostvdevname 1) ||
5614 error "reformat and replace $ostdev failed"
5615 start_ost || error "OST0 restart failure"
5616 wait_osc_import_state mds ost FULL
5618 mount_client $MOUNT || error "mount client failed"
5619 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
5620 local idx=$($LFS getstripe -i $DIR/$tdir/$tfile-last)
5621 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
5623 local iused=$($LFS df -i $MOUNT |
5624 awk '/OST0000/ { print $3 }'; exit ${PIPESTATUS[0]})
5625 log "On OST0, $iused used inodes rc=$?"
5626 [ $iused -ge $((ost_max_pre + 1000)) ] &&
5627 error "OST replacement created too many inodes; $iused"
5628 cleanup || error "cleanup failed with $?"
5630 run_test 69 "replace an OST with the same index"
5633 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5636 cleanup || error "cleanup failed with $?"
5638 start_mdt 1 || error "MDT0 start fail"
5640 start_ost || error "OST0 start fail"
5641 for num in $(seq 2 $MDSCOUNT); do
5642 start_mdt $num || return
5645 mount_client $MOUNT || error "mount client fails"
5647 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
5649 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5650 error "create remote dir fail"
5652 rm -rf $DIR/$tdir || error "delete dir fail"
5653 cleanup || error "cleanup failed with $?"
5655 run_test 70a "start MDT0, then OST, then MDT1"
5658 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5661 start_ost || error "OST0 start fail"
5663 start_mds || error "MDS start fail"
5665 mount_client $MOUNT || error "mount client fails"
5667 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
5669 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5670 error "create remote dir fail"
5672 rm -rf $DIR/$tdir || error "delete dir fail"
5674 cleanup || error "cleanup failed with $?"
5676 run_test 70b "start OST, MDT1, MDT0"
5679 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5682 start_mds || error "MDS start fail"
5683 start_ost || error "OST0 start fail"
5685 mount_client $MOUNT || error "mount client fails"
5686 stop_mdt 1 || error "MDT1 start fail"
5688 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
5689 echo "deactivate $mdc_for_mdt1"
5690 $LCTL --device $mdc_for_mdt1 deactivate ||
5691 error "set $mdc_for_mdt1 deactivate failed"
5693 mkdir $DIR/$tdir && error "mkdir succeed"
5695 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
5696 error "create remote dir succeed"
5698 cleanup || error "cleanup failed with $?"
5700 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
5703 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5706 start_mds || error "MDS start fail"
5707 start_ost || error "OST0 start fail"
5709 mount_client $MOUNT || error "mount client fails"
5711 stop_mdt 2 || error "MDT1 start fail"
5713 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
5715 echo "deactivate $mdc_for_mdt2"
5716 $LCTL --device $mdc_for_mdt2 deactivate ||
5717 error "set $mdc_for_mdt2 deactivate failed"
5719 mkdir $DIR/$tdir || error "mkdir fail"
5720 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
5721 error "create remote dir succeed"
5723 rm -rf $DIR/$tdir || error "delete dir fail"
5725 cleanup || error "cleanup failed with $?"
5727 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
5730 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5732 [ "$MDS1_VERSION" -ge $(version_code 2.7.62) ] ||
5733 skip "Need MDS version at least 2.7.62"
5735 reformat || error "reformat failed with $?"
5739 local mdsdev=$(mdsdevname 1)
5740 local ostdev=$(ostdevname 1)
5741 local mdsvdev=$(mdsvdevname 1)
5742 local ostvdev=$(ostvdevname 1)
5743 local opts_mds="$(mkfs_opts mds1 $mdsdev) --reformat $mdsdev $mdsvdev"
5744 local opts_ost="$(mkfs_opts ost1 $ostdev) --reformat $ostdev $ostvdev"
5746 if ! combined_mgs_mds ; then
5750 add mds1 $opts_mds || error "add mds1 failed"
5751 start_mdt 1 || error "start mdt1 failed"
5752 add ost1 $opts_ost || error "add ost1 failed"
5753 start_ost || error "start ost failed"
5754 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
5756 local soc=$(do_facet mds1 "$LCTL get_param -n \
5757 mdt.*MDT0000.sync_lock_cancel")
5758 [ $soc == "never" ] || error "SoC enabled on single MDS"
5760 for i in $(seq 2 $MDSCOUNT); do
5761 mdsdev=$(mdsdevname $i)
5762 mdsvdev=$(mdsvdevname $i)
5763 opts_mds="$(mkfs_opts mds$i $mdsdev) --reformat $mdsdev \
5765 add mds$i $opts_mds || error "add mds$i failed"
5766 start_mdt $i || error "start mdt$i fail"
5769 wait_dne_interconnect
5771 for i in $(seq $MDSCOUNT); do
5772 soc=$(do_facet mds$i "$LCTL get_param -n \
5773 mdt.*MDT000$((i - 1)).sync_lock_cancel")
5774 [ $soc == "blocking" ] || error "SoC not enabled on DNE"
5777 for i in $(seq 2 $MDSCOUNT); do
5778 stop_mdt $i || error "stop mdt$i fail"
5780 soc=$(do_facet mds1 "$LCTL get_param -n \
5781 mdt.*MDT0000.sync_lock_cancel")
5782 [ $soc == "never" ] || error "SoC enabled on single MDS"
5783 umount_client $MOUNT -f > /dev/null
5785 cleanup || error "cleanup failed with $?"
5787 run_test 70e "Sync-on-Cancel will be enabled by default on DNE"
5790 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5791 if combined_mgs_mds; then
5792 skip "needs separate MGS/MDT"
5796 start_mdt 1 || error "MDT0 start fail"
5797 start_ost || error "OST0 start fail"
5798 for num in $(seq 2 $MDSCOUNT); do
5799 start_mdt $num || return
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 "MDS stop fail"
5815 stop_ost || error "OST0 stop fail"
5816 stop_ost2 || error "OST1 stop fail"
5818 run_test 71a "start MDT0 OST0, MDT1, OST1"
5821 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5822 if combined_mgs_mds; then
5823 skip "needs separate MGS/MDT"
5827 for num in $(seq 2 $MDSCOUNT); do
5828 start_mdt $num || return
5830 start_ost || error "OST0 start fail"
5831 start_mdt 1 || error "MDT0 start fail"
5832 start_ost2 || error "OST1 start fail"
5834 mount_client $MOUNT || error "mount client fails"
5836 mkdir $DIR/$tdir || error "mkdir fail"
5837 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5838 error "create remote dir succeed"
5840 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5841 rm -rf $DIR/$tdir || error "delete dir fail"
5843 umount_client $MOUNT || error "umount_client failed"
5844 stop_mds || error "MDT0 stop fail"
5845 stop_ost || error "OST0 stop fail"
5846 stop_ost2 || error "OST1 stop fail"
5848 run_test 71b "start MDT1, OST0, MDT0, OST1"
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 start_ost2 || error "OST1 start fail"
5858 for num in $(seq 2 $MDSCOUNT); do
5859 start_mdt $num || return
5861 start_mdt 1 || error "MDT0 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 71c "start OST0, OST1, MDT1, MDT0"
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_mdt 1 || error "MDT0 start fail"
5891 start_ost2 || error "OST1 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 71d "start OST0, MDT1, MDT0, OST1"
5911 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5912 combined_mgs_mds && skip "needs separate MGS/MDT"
5916 start_ost || error "OST0 start fail"
5917 for num in $(seq 2 $MDSCOUNT); do
5918 start_mdt $num || return
5920 start_ost2 || error "OST1 start fail"
5921 start_mdt 1 || error "MDT0 start fail"
5923 mount_client $MOUNT || error "mount client fails"
5925 mkdir $DIR/$tdir || error "mkdir fail"
5926 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5927 error "create remote dir succeed"
5929 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5930 rm -rf $DIR/$tdir || error "delete dir fail"
5932 umount_client $MOUNT || error "umount_client failed"
5933 stop_mds || error "MDS stop fail"
5934 stop_ost || error "OST0 stop fail"
5935 stop_ost2 || error "OST1 stop fail"
5938 run_test 71e "start OST0, MDT1, OST1, MDT0"
5940 test_72() { #LU-2634
5941 [ "$mds1_FSTYPE" != ldiskfs ] &&
5942 skip "ldiskfs only test"
5944 local mdsdev=$(mdsdevname 1)
5945 local ostdev=$(ostdevname 1)
5946 local cmd="$E2FSCK -fnvd $mdsdev"
5953 if combined_mgs_mds; then
5954 add_options='--reformat'
5956 add_options='--reformat --replace'
5959 #tune MDT with "-O extents"
5961 for num in $(seq $MDSCOUNT); do
5962 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
5963 $add_options $(mdsdevname $num) $(mdsvdevname $num) ||
5964 error "add mds $num failed"
5965 do_facet mds${num} "$TUNE2FS -O extents $(mdsdevname $num)" ||
5966 error "$TUNE2FS failed on mds${num}"
5969 add ost1 $(mkfs_opts ost1 $ostdev) $add_options $ostdev ||
5970 error "add $ostdev failed"
5971 start_mds || error "start mds failed"
5972 start_ost || error "start ost failed"
5973 mount_client $MOUNT || error "mount client failed"
5975 #create some short symlinks
5976 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5977 createmany -o $DIR/$tdir/$tfile-%d $fn
5978 echo "create $fn short symlinks"
5979 for i in $(seq -w 1 $fn); do
5980 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
5985 umount_client $MOUNT || error "umount client failed"
5986 stop_mds || error "stop mds failed"
5987 stop_ost || error "stop ost failed"
5990 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
5992 run_test 72 "test fast symlink with extents flag enabled"
5994 test_73() { #LU-3006
5995 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
5996 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
5997 error "1st tunefs failed"
5998 start_mgsmds || error "start mds failed"
5999 start_ost || error "start ost failed"
6000 mount_client $MOUNT || error "mount client failed"
6001 $LCTL get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
6002 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
6003 umount_client $MOUNT || error "umount client failed"
6007 run_test 73 "failnode to update from mountdata properly"
6009 test_75() { # LU-2374
6010 [[ "$MDS1_VERSION" -lt $(version_code 2.4.1) ]] &&
6011 skip "Need MDS version at least 2.4.1"
6014 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
6015 --replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
6016 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
6017 --replace --reformat $(ostdevname 1) $(ostvdevname 1)"
6019 #check with default parameters
6020 add mds1 $opts_mds || error "add mds1 failed for default params"
6021 add ost1 $opts_ost || error "add ost1 failed for default params"
6023 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
6024 opts_mds=$(echo $opts_mds |
6025 sed -e "s/--index=$index/--index=$index --mdt/")
6026 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
6027 opts_ost=$(echo $opts_ost |
6028 sed -e "s/--index=$index/--index=$index --ost/")
6030 add mds1 $opts_mds || error "add mds1 failed for new params"
6031 add ost1 $opts_ost || error "add ost1 failed for new params"
6036 run_test 75 "The order of --index should be irrelevant"
6039 [[ "$MGS_VERSION" -ge $(version_code 2.4.52) ]] ||
6040 skip "Need MDS version at least 2.4.52"
6043 local MDMB_PARAM="osc.*.max_dirty_mb"
6044 echo "Change MGS params"
6045 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
6047 echo "max_dirty_mb: $MAX_DIRTY_MB"
6048 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB - 10))
6049 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
6050 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
6051 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
6052 head -1" $NEW_MAX_DIRTY_MB
6053 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
6054 echo "$MAX_DIRTY_MB"
6055 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
6056 error "error while apply max_dirty_mb"
6058 echo "Check the value is stored after remount"
6061 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
6062 head -1" $NEW_MAX_DIRTY_MB
6063 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
6064 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
6065 error "max_dirty_mb is not saved after remount"
6067 echo "Change OST params"
6068 CLIENT_PARAM="obdfilter.*.client_cache_count"
6069 local CLIENT_CACHE_COUNT
6070 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
6072 echo "client_cache_count: $CLIENT_CACHE_COUNT"
6073 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
6074 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
6075 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
6076 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
6077 head -1" $NEW_CLIENT_CACHE_COUNT
6078 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
6080 echo "$CLIENT_CACHE_COUNT"
6081 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
6082 error "error while apply client_cache_count"
6084 echo "Check the value is stored after remount"
6087 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
6088 head -1" $NEW_CLIENT_CACHE_COUNT
6089 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
6091 echo "$CLIENT_CACHE_COUNT"
6092 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
6093 error "client_cache_count is not saved after remount"
6096 run_test 76a "set permanent params with lctl across mounts"
6098 test_76b() { # LU-4783
6099 [[ "$MGS_VERSION" -ge $(version_code 2.5.57) ]] ||
6100 skip "Need MGS version at least 2.5.57"
6103 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
6104 error "start params log failed"
6107 run_test 76b "verify params log setup correctly"
6110 [[ "$MGS_VERSION" -ge $(version_code 2.8.54) ]] ||
6111 skip "Need MDS version at least 2.4.52"
6113 local MASK_PARAM="mdd.*.changelog_mask"
6114 echo "Change changelog_mask"
6115 do_facet mgs $LCTL set_param -P $MASK_PARAM=-CLOSE ||
6116 error "Can't change changlog_mask"
6117 wait_update $(facet_host mds) "$LCTL get_param -n $MASK_PARAM |
6120 echo "Check the value is stored after mds remount"
6121 stop_mds || error "Failed to stop MDS"
6122 start_mds || error "Failed to start MDS"
6123 local CHANGELOG_MASK=$(do_facet mgs $LCTL get_param -n $MASK_PARAM)
6124 echo $CHANGELOG_MASK | grep CLOSE > /dev/null &&
6125 error "changelog_mask is not changed"
6129 run_test 76c "verify changelog_mask is applied with lctl set_param -P"
6131 test_76d() { #LU-9399
6134 local xattr_cache="llite.*.xattr_cache"
6135 local cmd="$LCTL get_param -n $xattr_cache | head -1"
6136 local new=$((($(eval $cmd) + 1) % 2))
6138 echo "lctl set_param -P llite.*.xattr_cache=$new"
6139 do_facet mgs $LCTL set_param -P $xattr_cache=$new ||
6140 error "Can't change xattr_cache"
6141 wait_update $HOSTNAME "$cmd" "$new"
6143 echo "Check $xattr_cache on client $MOUNT"
6144 umount_client $MOUNT || error "umount $MOUNT failed"
6145 mount_client $MOUNT || error "mount $MOUNT failed"
6146 [ $(eval $cmd) -eq $new ] ||
6147 error "$xattr_cache != $new on client $MOUNT"
6149 echo "Check $xattr_cache on the new client $MOUNT2"
6150 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
6151 [ $(eval $cmd) -eq $new ] ||
6152 error "$xattr_cache != $new on client $MOUNT2"
6153 umount_client $MOUNT2 || error "umount $MOUNT2 failed"
6157 run_test 76d "verify llite.*.xattr_cache can be set by 'lctl set_param -P' correctly"
6159 test_77() { # LU-3445
6160 [[ "$MDS1_VERSION" -ge $(version_code 2.8.55) ]] ||
6161 skip "Need MDS version 2.8.55+ "
6163 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
6164 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
6165 skip_env "mixed loopback and real device not working"
6168 local fs2mdsdev=$(mdsdevname 1_2)
6169 local fs2ostdev=$(ostdevname 1_2)
6170 local fs2mdsvdev=$(mdsvdevname 1_2)
6171 local fs2ostvdev=$(ostvdevname 1_2)
6172 local fsname=test1234
6174 local failnid="$(h2nettype 1.2.3.4),$(h2nettype 4.3.2.1)"
6176 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
6178 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
6179 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
6180 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
6181 error "start fs2mds failed"
6183 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
6184 mgsnid="0.0.0.0@tcp,$mgsnid,$mgsnid:$mgsnid"
6186 add fs2ost --mgsnode=$mgsnid $(mkfs_opts ost1 $fs2ostdev) \
6187 --failnode=$failnid --fsname=$fsname \
6188 --reformat $fs2ostdev $fs2ostvdev ||
6189 error "add fs2ost failed"
6190 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
6192 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
6193 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
6194 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
6197 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
6200 [[ "$mds1_FSTYPE" != ldiskfs ||
6201 "$ost1_FSTYPE" != ldiskfs ]] &&
6202 skip "ldiskfs only test"
6204 # reformat the Lustre filesystem with a smaller size
6205 local saved_MDSCOUNT=$MDSCOUNT
6206 local saved_MDSSIZE=$MDSSIZE
6207 local saved_OSTCOUNT=$OSTCOUNT
6208 local saved_OSTSIZE=$OSTSIZE
6211 MDSSIZE=$((MDSSIZE - 20000))
6212 OSTSIZE=$((OSTSIZE - 20000))
6213 reformat || error "(1) reformat Lustre filesystem failed"
6214 MDSSIZE=$saved_MDSSIZE
6215 OSTSIZE=$saved_OSTSIZE
6217 # mount the Lustre filesystem
6218 setup_noconfig || error "(2) setup Lustre filesystem failed"
6221 log "create test files"
6226 mkdir $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
6228 for i in $(seq $num_files); do
6229 file=$MOUNT/$tdir/$tfile-$i
6230 dd if=/dev/urandom of=$file count=1 bs=1M || {
6231 $LCTL get_param osc.*.cur*grant*
6232 $LFS df; $LFS df -i;
6233 # stop creating files if there is no more space
6234 if [ ! -e $file ]; then
6235 num_files=$((i - 1))
6239 $LFS getstripe -v $file
6240 local ost_idx=$(LFS getstripe -i $file)
6241 do_facet ost$((ost_idx + 1)) \
6242 $LCTL get_param obdfilter.*.*grant*
6243 error "(4) create $file failed"
6247 # unmount the Lustre filesystem
6248 cleanup || error "(5) cleanup Lustre filesystem failed"
6250 # run e2fsck on the MDT and OST devices
6251 local mds_host=$(facet_active_host $SINGLEMDS)
6252 local ost_host=$(facet_active_host ost1)
6253 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
6254 local ost_dev=$(ostdevname 1)
6256 run_e2fsck $mds_host $mds_dev "-y"
6257 run_e2fsck $ost_host $ost_dev "-y"
6259 # get the original block count of the MDT and OST filesystems
6260 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
6261 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
6263 # expand the MDT and OST filesystems to the device size
6264 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
6265 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
6267 # run e2fsck on the MDT and OST devices again
6268 run_e2fsck $mds_host $mds_dev "-y"
6269 run_e2fsck $ost_host $ost_dev "-y"
6271 # mount the Lustre filesystem
6275 log "check files after expanding the MDT and OST filesystems"
6276 for i in $(seq $num_files); do
6277 file=$MOUNT/$tdir/$tfile-$i
6278 $CHECKSTAT -t file -s 1048576 $file ||
6279 error "(6) checkstat $file failed"
6283 log "create more files after expanding the MDT and OST filesystems"
6284 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
6285 file=$MOUNT/$tdir/$tfile-$i
6286 dd if=/dev/urandom of=$file count=1 bs=1M ||
6287 error "(7) create $file failed"
6290 # unmount the Lustre filesystem
6291 cleanup || error "(8) cleanup Lustre filesystem failed"
6293 # run e2fsck on the MDT and OST devices
6294 run_e2fsck $mds_host $mds_dev "-y"
6295 run_e2fsck $ost_host $ost_dev "-y"
6297 # get the maximum block count of the MDT and OST filesystems
6298 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
6299 local ost_max_blks=$(get_block_count ost1 $ost_dev)
6301 # get the minimum block count of the MDT and OST filesystems
6302 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
6303 grep minimum | sed -e 's/^.*filesystem: //g')
6304 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
6305 grep minimum | sed -e 's/^.*filesystem: //g')
6307 # shrink the MDT and OST filesystems to a smaller size
6311 if [[ $mds_max_blks -gt $mds_min_blks &&
6312 $mds_max_blks -gt $mds_orig_blks ]]; then
6313 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
6314 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
6315 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
6316 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
6317 error "shrink $SINGLEMDS to $new_blks failed"
6321 if [[ $ost_max_blks -gt $ost_min_blks &&
6322 $ost_max_blks -gt $ost_orig_blks ]]; then
6323 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
6324 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
6325 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
6326 run_resize2fs ost1 $ost_dev $new_blks ||
6327 error "shrink ost1 to $new_blks failed"
6331 # check whether the MDT or OST filesystem was shrunk or not
6333 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
6334 reformat_and_config ||
6335 error "(10) reformat Lustre filesystem failed"
6339 # run e2fsck on the MDT and OST devices again
6340 run_e2fsck $mds_host $mds_dev "-y"
6341 run_e2fsck $ost_host $ost_dev "-y"
6343 # mount the Lustre filesystem again
6347 log "check files after shrinking the MDT and OST filesystems"
6348 for i in $(seq $((num_files + 10))); do
6349 file=$MOUNT/$tdir/$tfile-$i
6350 $CHECKSTAT -t file -s 1048576 $file ||
6351 error "(11) checkstat $file failed"
6354 # unmount and reformat the Lustre filesystem
6355 cleanup || error "(12) cleanup Lustre filesystem failed"
6356 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
6358 MDSCOUNT=$saved_MDSCOUNT
6359 OSTCOUNT=$saved_OSTCOUNT
6360 reformat_and_config || error "(14) reformat Lustre filesystem failed"
6362 run_test 78 "run resize2fs on MDT and OST filesystems"
6364 test_79() { # LU-4227
6365 [[ "$MDS1_VERSION" -ge $(version_code 2.5.59) ]] ||
6366 skip "Need MDS version at least 2.5.59"
6368 local mdsdev1=$(mdsdevname 1)
6369 local mdsvdev1=$(mdsvdevname 1)
6370 local mdsdev2=$(mdsdevname 2)
6371 local mdsvdev2=$(mdsvdevname 2)
6372 local ostdev1=$(ostdevname 1)
6373 local ostvdev1=$(ostvdevname 1)
6374 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
6375 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
6376 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
6379 # remove --mgs/--mgsnode from mkfs.lustre options
6380 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
6382 mgsnode_opt=$(echo $opts_mds2 |
6383 awk '{ for ( i = 1; i < NF; i++ )
6384 if ( $i ~ "--mgsnode" ) { print $i; break } }')
6385 [ -n "$mgsnode_opt" ] &&
6386 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
6388 mgsnode_opt=$(echo $opts_ost1 |
6389 awk '{ for ( i = 1; i < NF; i++ )
6390 if ( $i ~ "--mgsnode" ) { print $i; break } }')
6391 [ -n "$mgsnode_opt" ] &&
6392 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
6394 # -MGS, format a mdt without --mgs option
6395 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
6396 error "Must specify --mgs when formatting mdt combined with mgs"
6398 # +MGS, format a mdt/ost without --mgsnode option
6399 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
6400 > /dev/null || error "start mds1 failed"
6401 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
6402 error "Must specify --mgsnode when formatting a mdt"
6403 add ost1 $opts_ost1 $ostdev1 $ostvdev2 &&
6404 error "Must specify --mgsnode when formatting an ost"
6408 run_test 79 "format MDT/OST without mgs option (should return errors)"
6411 start_mds || error "Failed to start MDT"
6412 start_ost || error "Failed to start OST1"
6413 uuid=$(do_facet ost1 $LCTL get_param -n mgc.*.uuid)
6414 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
6415 do_facet ost1 "$LCTL set_param fail_val=10 fail_loc=0x906"
6416 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x906"
6417 do_facet mgs "$LCTL set_param -n mgs/MGS/evict_client $uuid"
6419 start_ost2 || error "Failed to start OST2"
6421 do_facet ost1 "$LCTL set_param fail_loc=0"
6426 run_test 80 "mgc import reconnect race"
6428 #Save the original values of $OSTCOUNT and $OSTINDEX$i.
6430 local new_ostcount=$1
6431 saved_ostcount=$OSTCOUNT
6432 OSTCOUNT=$new_ostcount
6436 for i in $(seq $OSTCOUNT); do
6438 eval saved_ostindex$i=${!index}
6443 # Restore the original values of $OSTCOUNT and $OSTINDEX$i.
6444 restore_ostindex() {
6449 for i in $(seq $OSTCOUNT); do
6450 index=saved_ostindex$i
6451 eval OSTINDEX$i=${!index}
6453 OSTCOUNT=$saved_ostcount
6456 if ! combined_mgs_mds ; then
6461 # The main purpose of this test is to ensure the OST_INDEX_LIST functions as
6462 # expected. This test uses OST_INDEX_LIST to format OSTs with a randomly
6463 # assigned index and ensures we can mount such a formatted file system
6464 test_81() { # LU-4665
6465 [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
6466 skip "Need MDS version at least 2.6.54"
6467 [[ $OSTCOUNT -ge 3 ]] || skip_env "needs >= 3 OSTs"
6471 # Each time RANDOM is referenced, a random integer between 0 and 32767
6474 local saved_ostindex1=$OSTINDEX1
6475 for i in 65535 $((RANDOM + 65536)); do
6476 echo -e "\nFormat ost1 with --index=$i, should fail"
6478 if add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat \
6479 $(ostdevname 1) $(ostvdevname 1); then
6480 OSTINDEX1=$saved_ostindex1
6481 error "format ost1 with --index=$i should fail"
6484 OSTINDEX1=$saved_ostindex1
6488 # Format OSTs with random sparse indices.
6489 trap "restore_ostindex" EXIT
6490 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices"
6491 OST_INDEX_LIST=[0,$((RANDOM * 2 % 65533 + 1)),65534] formatall
6493 # Setup and check Lustre filesystem.
6494 start_mgsmds || error "start_mgsmds failed"
6495 for i in $(seq $OSTCOUNT); do
6496 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6497 error "start ost$i failed"
6500 mount_client $MOUNT || error "mount client $MOUNT failed"
6501 check_mount || error "check client $MOUNT failed"
6504 local max_easize=$($LCTL get_param -n llite.*.max_easize)
6505 # 65452 is XATTR_SIZE_MAX less ldiskfs ea overhead
6506 if large_xattr_enabled; then
6507 [[ $max_easize -ge 65452 ]] ||
6508 error "max_easize is $max_easize, should be at least 65452 bytes"
6511 # 4012 is 4096 less ldiskfs ea overhead
6512 [[ $max_easize -ge 4012 ]] ||
6513 error "max_easize is $max_easize, should be at least 4012 bytes"
6518 run_test 81 "sparse OST indexing"
6520 # Here we exercise the stripe placement functionality on a file system that
6521 # has formatted the OST with a random index. With the file system the following
6522 # functionality is tested:
6524 # 1. Creating a new file with a specific stripe layout.
6526 # 2. Modifiy a existing empty file with a specific stripe layout.
6528 # 3. Ensure we fail to set the stripe layout of a file that already has one.
6530 # 4. If ost-index is defined we need to ensure it is the first entry in the
6531 # ost index list returned by lfs getstripe.
6533 # 5. Lastly ensure this functionality fails with directories.
6534 test_82a() { # LU-4665
6535 [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
6536 skip "Need MDS version at least 2.6.54"
6537 [[ $OSTCOUNT -ge 3 ]] || skip_env "needs >= 3 OSTs"
6543 # Format OSTs with random sparse indices.
6547 local LOV_V1_INSANE_STRIPE_COUNT=65532
6548 for i in $(seq $OSTCOUNT); do
6549 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
6550 ost_indices+=" $index"
6552 ost_indices=$(comma_list $ost_indices)
6554 stack_trap "restore_ostindex" EXIT
6555 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
6556 OST_INDEX_LIST=[$ost_indices] formatall
6558 # Setup Lustre filesystem.
6559 start_mgsmds || error "start_mgsmds failed"
6560 for i in $(seq $OSTCOUNT); do
6561 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6562 error "start ost$i failed"
6565 # Collect debug information - start of test
6566 do_nodes $(comma_list $(mdts_nodes)) \
6567 $LCTL get_param osp.*.prealloc_*_id
6569 mount_client $MOUNT || error "mount client $MOUNT failed"
6573 check_lfs_df_ret_val $? || error "$LFS df $MOUNT failed"
6574 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6576 stack_trap "do_nodes $(comma_list $(mdts_nodes)) \
6577 $LCTL get_param osp.*.prealloc_*_id" EXIT
6579 # 1. If the file does not exist, new file will be created
6580 # with specified OSTs.
6581 local file=$DIR/$tdir/$tfile-1
6582 local cmd="$LFS setstripe -o $ost_indices $file"
6584 eval $cmd || error "$cmd failed"
6585 check_stripe_count $file $OSTCOUNT
6586 check_obdidx $file $ost_indices
6587 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6588 error "write $file failed"
6590 # 2. If the file already exists and is an empty file, the file
6591 # will be attached with specified layout.
6592 file=$DIR/$tdir/$tfile-2
6593 mcreate $file || error "mcreate $file failed"
6594 cmd="$LFS setstripe -o $ost_indices $file"
6596 eval $cmd || error "$cmd failed"
6597 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6598 error "write $file failed"
6599 check_stripe_count $file $OSTCOUNT
6600 check_obdidx $file $ost_indices
6602 # 3. If the file already has a valid layout attached, the command
6603 # should fail with EBUSY.
6605 eval $cmd && error "stripe is already set on $file, $cmd should fail"
6607 # 4. If [--stripe-index|-i <start_ost_idx>] is used, the index must
6608 # be in the OST indices list.
6609 local start_ost_idx=${ost_indices##*,}
6610 file=$DIR/$tdir/$tfile-3
6611 cmd="$LFS setstripe -o $ost_indices -i $start_ost_idx $file"
6613 eval $cmd || error "$cmd failed"
6614 check_stripe_count $file $OSTCOUNT
6615 check_obdidx $file $ost_indices
6616 check_start_ost_idx $file $start_ost_idx
6618 file=$DIR/$tdir/$tfile-4
6619 cmd="$LFS setstripe"
6620 cmd+=" -o $(exclude_items_from_list $ost_indices $start_ost_idx)"
6621 cmd+=" -i $start_ost_idx $file"
6623 eval $cmd && error "index $start_ost_idx should be in $ost_indices"
6625 # 5. Specifying OST indices for directory should succeed.
6626 local dir=$DIR/$tdir/$tdir
6627 mkdir $dir || error "mkdir $dir failed"
6628 cmd="$LFS setstripe -o $ost_indices $dir"
6629 if [ "$MDS1_VERSION" -gt $(version_code 2.11.53) ] &&
6630 [ "$CLIENT_VERSION" -gt $(version_code 2.11.53) ]; then
6632 eval $cmd || error "unable to specify OST indices on directory"
6634 echo "need MDS+client version at least 2.11.53"
6639 run_test 82a "specify OSTs for file (succeed) or directory (succeed)"
6644 # Remove OSTs from a pool and destroy the pool.
6645 destroy_pool $ost_pool || true
6650 # Test 82b is run to ensure that if the user supplies a pool with a specific
6651 # stripe layout that it behaves proprerly. It should fail in the case that
6652 # the supplied OST index list points to OSTs not contained in the user
6654 test_82b() { # LU-4665
6655 [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
6656 skip "Need MDS version at least 2.6.54"
6657 [[ $OSTCOUNT -ge 4 ]] || skip_env "needs >= 4 OSTs"
6663 # Format OSTs with random sparse indices.
6667 local LOV_V1_INSANE_STRIPE_COUNT=65532
6668 for i in $(seq $OSTCOUNT); do
6669 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
6670 ost_indices+=" $index"
6672 ost_indices=$(comma_list $ost_indices)
6674 trap "restore_ostindex" EXIT
6675 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
6676 OST_INDEX_LIST=[$ost_indices] formatall
6678 # Setup Lustre filesystem.
6679 start_mgsmds || error "start_mgsmds failed"
6680 for i in $(seq $OSTCOUNT); do
6681 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6682 error "start ost$i failed"
6685 mount_client $MOUNT || error "mount client $MOUNT failed"
6689 check_lfs_df_ret_val $? || error "$LFS df $MOUNT failed"
6690 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6692 # Create a new pool and add OSTs into it.
6693 local ost_pool=$FSNAME.$TESTNAME
6694 create_pool $ost_pool || error "create OST pool $ost_pool failed"
6697 trap "cleanup_82b" EXIT
6699 local ost_idx_in_list=${ost_indices##*,}
6700 local ost_idx_in_pool=$(exclude_items_from_list $ost_indices \
6703 local ost_targets="$FSNAME-OST["
6704 for i in ${ost_idx_in_pool//,/ }; do
6705 ost_targets=$ost_targets$(printf "%04x," $i)
6707 ost_targets="${ost_targets%,}]"
6709 local ost_targets_uuid=$(for i in ${ost_idx_in_pool//,/ }; \
6710 do printf "$FSNAME-OST%04x_UUID\n" $i; done |
6711 sort -u | tr '\n' ' ')
6713 local cmd="$LCTL pool_add $ost_pool $ost_targets"
6714 do_facet mgs $cmd || error "$cmd failed"
6715 wait_update $HOSTNAME "$LCTL get_param -n lov.$FSNAME-*.pools.$TESTNAME|
6716 sort -u | tr '\n' ' ' " "$ost_targets_uuid" ||
6717 error "wait_update $ost_pool failed"
6718 wait_update_facet $SINGLEMDS "$LCTL pool_list $ost_pool | wc -l" 4 ||
6719 error "wait_update pool_list $ost_pool failed"
6721 # If [--pool|-p <pool_name>] is set with [--ost-list|-o <ost_indices>],
6722 # then the OSTs must be the members of the pool.
6723 local file=$DIR/$tdir/$tfile
6724 cmd="$LFS setstripe -p $ost_pool -o $ost_idx_in_list $file"
6726 eval $cmd && error "OST with index $ost_idx_in_list should be" \
6727 "in OST pool $ost_pool"
6729 # Only select OST $ost_idx_in_list from $ost_pool for file.
6730 ost_idx_in_list=${ost_idx_in_pool#*,}
6731 cmd="$LFS setstripe -p $ost_pool -o $ost_idx_in_list $file"
6733 eval $cmd || error "$cmd failed"
6734 cmd="$LFS getstripe $file"
6736 eval $cmd || error "$cmd failed"
6737 check_stripe_count $file 2
6738 check_obdidx $file $ost_idx_in_list
6739 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6740 error "write $file failed"
6744 run_test 82b "specify OSTs for file with --pool and --ost-list options"
6747 [[ "$OST1_VERSION" -ge $(version_code 2.6.91) ]] ||
6748 skip "Need OST version at least 2.6.91"
6749 if [ "$ost1_FSTYPE" != ldiskfs ]; then
6750 skip "ldiskfs only test"
6759 ostmnt=$(facet_mntpt ost1)
6761 # Mount the OST as an ldiskfs filesystem.
6762 log "mount the OST $dev as a $ost1_FSTYPE filesystem"
6763 add ost1 $(mkfs_opts ost1 $dev) $FSTYPE_OPT \
6764 --reformat $dev > /dev/null ||
6765 error "format ost1 error"
6767 if ! test -b $dev; then
6768 mnt_opts=$(csa_add "$OST_MOUNT_FS_OPTS" -o loop)
6770 echo "mnt_opts $mnt_opts"
6771 do_facet ost1 mount -t "$ost1_FSTYPE" $dev \
6773 # Run llverfs on the mounted ldiskfs filesystem.
6774 # It is needed to get ENOSPACE.
6775 log "run llverfs in partial mode on the OST $ost1_FSTYPE $ostmnt"
6776 do_rpc_nodes $(facet_host ost1) run_llverfs $ostmnt -vpl \
6777 "no" || error "run_llverfs error on $ost1_FSTYPE"
6780 log "unmount the OST $dev"
6783 # Delete file IO_scrub. Later osd_scrub_setup will try to
6784 # create "IO_scrub" but will get ENOSPACE.
6786 echo "start ost1 service on `facet_active_host ost1`"
6787 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
6790 err=$(do_facet ost1 dmesg | grep "VFS: Busy inodes after unmount of")
6791 echo "string err $err"
6792 [ -z "$err" ] || error $err
6795 run_test 83 "ENOSPACE on OST doesn't cause message VFS: \
6796 Busy inodes after unmount ..."
6799 local facet=$SINGLEMDS
6800 local num=$(echo $facet | tr -d "mds")
6801 local dev=$(mdsdevname $num)
6802 local time_min=$(recovery_time_min)
6803 local recovery_duration
6804 local completed_clients
6805 local correct_clients
6808 echo "start mds service on $(facet_active_host $facet)"
6810 "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
6811 error "start MDS failed"
6813 start_ost || error "start OST0000 failed"
6814 start_ost2 || error "start OST0001 failed"
6816 echo "recovery_time=$time_min, timeout=$TIMEOUT, wrap_up=$wrap_up"
6818 mount_client $MOUNT1 || error "mount $MOUNT1 failed"
6819 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
6820 # make sure new superblock labels are sync'd before disabling writes
6824 replay_barrier $SINGLEMDS
6825 createmany -o $DIR1/$tfile-%d 1000
6827 # We need to catch the end of recovery window to extend it.
6828 # Skip 5 requests and add delay to request handling.
6829 #define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP
6830 do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5"
6832 facet_failover --fsck $SINGLEMDS || error "failover: $?"
6835 echo "recovery status"
6836 do_facet $SINGLEMDS \
6837 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status"
6839 recovery_duration=$(do_facet $SINGLEMDS \
6840 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
6841 awk '/recovery_duration/ { print $2 }')
6842 (( $recovery_duration > $time_min + $wrap_up )) &&
6843 error "recovery_duration > recovery_time_hard + wrap up"
6844 completed_clients=$(do_facet $SINGLEMDS \
6845 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
6846 awk '/completed_clients/ { print $2 }')
6848 correct_clients="$MDSCOUNT/$((MDSCOUNT+1))"
6849 [ "$completed_clients" = "${correct_clients}" ] ||
6850 error "$completed_clients != $correct_clients"
6852 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
6853 umount_client $MOUNT1
6854 umount_client $MOUNT2
6860 run_test 84 "check recovery_hard_time"
6863 [[ "$OST1_VERSION" -ge $(version_code 2.7.55) ]] ||
6864 skip "Need OST version at least 2.7.55"
6865 ##define OBD_FAIL_OSD_OST_EA_FID_SET 0x197
6866 do_facet ost1 "lctl set_param fail_loc=0x197"
6870 run_test 85 "osd_ost init: fail ea_fid_set"
6875 # ost1 has already registered to the MGS before the reformat.
6876 # So after reformatting it with option "-G", it could not be
6877 # mounted to the MGS. Cleanup the system for subsequent tests.
6882 [ "$ost1_FSTYPE" = zfs ] &&
6883 skip "LU-6442: no such mkfs params for ZFS OSTs"
6884 [[ "$MDS1_VERSION" -ge $(version_code 2.7.56) ]] ||
6885 skip "Need server version newer than 2.7.55"
6888 local OLDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6889 awk '/Flex block group size: / { print $NF; exit; }')
6891 [ "$OLDSIZE" == "$NEWSIZE" ] && skip "$NEWSIZE groups already"
6893 local opts=" -O flex_bg -G $NEWSIZE"
6894 opts=$(OST_FS_MKFS_OPTS+="$opts" mkfs_opts ost1 $(ostdevname 1))
6895 opts+=" --reformat $(ostdevname 1) $(ostvdevname 1)"
6896 echo "params: $opts"
6898 trap cleanup_86 EXIT ERR
6901 add ost1 $opts || error "add ost1 failed with new params"
6903 local FOUNDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6904 awk '/Flex block group size: / { print $NF; exit; }')
6906 [[ $FOUNDSIZE == $NEWSIZE ]] ||
6907 error "Flex block group size: $FOUNDSIZE, expected: $NEWSIZE"
6911 run_test 86 "Replacing mkfs.lustre -G option"
6913 test_87() { #LU-6544
6914 [[ "$MDS1_VERSION" -ge $(version_code 2.9.51) ]] ||
6915 skip "Need MDS version at least 2.9.51"
6916 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
6917 skip "ldiskfs only test"
6918 [[ $OSTCOUNT -gt 59 ]] &&
6919 skip "Ignore wide striping situation"
6920 [ -n "$FILESET" ] && skip "Not functional for FILESET set"
6922 local mdsdev=$(mdsdevname 1)
6923 local mdsvdev=$(mdsvdevname 1)
6924 local file=$DIR/$tfile
6925 local mntpt=$(facet_mntpt $SINGLEMDS)
6926 local used_xattr_blk=0
6927 local inode_size=${1:-1024}
6929 local xtest="trusted.test"
6933 local stripe_cnt=$(($OSTCOUNT + 2))
6935 #Please see ldiskfs_make_lustre() for MDT inode size calculation
6936 if [ $stripe_cnt -gt 16 ]; then
6939 left_size=$(expr $inode_size - \
6942 32 - 40 \* 3 - 32 \* 3 - $stripe_cnt \* 24 - 16 - 3 - \
6944 24 - 18 - $(expr length $tfile) - 16 - 4)
6945 if [ $left_size -le 0 ]; then
6946 echo "No space($left_size) is expected in inode."
6947 echo "Try 1-byte xattr instead to verify this."
6950 echo "Estimate: at most $left_size-byte space left in inode."
6956 if ! combined_mgs_mds ; then
6960 add mds1 $(mkfs_opts mds1 ${mdsdev}) --stripe-count-hint=$stripe_cnt \
6961 --reformat $mdsdev $mdsvdev || error "add mds1 failed"
6962 start_mdt 1 > /dev/null || error "start mdt1 failed"
6963 for i in $(seq $OSTCOUNT); do
6964 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS > /dev/null ||
6965 error "start ost$i failed"
6967 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
6968 check_mount || error "check client $MOUNT failed"
6971 $LFS setstripe -E 1M -S 1M -c 1 -E 64M -c 1 -E -1 -c -1 $file ||
6972 error "Create file with 3 components failed"
6973 $TRUNCATE $file $((1024*1024*64+1)) || error "truncate file failed"
6974 i=$($LFS getstripe -I3 -c $file) || error "get 3rd stripe count failed"
6975 if [ $i -ne $OSTCOUNT ]; then
6976 left_size=$(expr $left_size + $(expr $OSTCOUNT - $i) \* 24)
6977 echo -n "Since only $i out $OSTCOUNT OSTs are used, "
6978 echo -n "the expected left space is changed to "
6979 echo "$left_size bytes at most."
6981 value=$(generate_string $left_size)
6982 setfattr -n $xtest -v $value $file
6983 orig=$(get_xattr_value $xtest $file)
6984 [[ "$orig" != "$value" ]] && error "$xtest changed"
6986 #Verify if inode has some expected space left
6987 umount $MOUNT > /dev/null || error "umount $MOUNT failed"
6988 stop_mdt 1 > /dev/null || error "stop mdt1 failed"
6989 mount_ldiskfs $SINGLEMDS || error "mount -t ldiskfs $SINGLEMDS failed"
6991 do_facet $SINGLEMDS ls -sal $mntpt/ROOT/$tfile
6992 used_xattr_blk=$(do_facet $SINGLEMDS ls -s $mntpt/ROOT/$tfile |
6994 [[ $used_xattr_blk -eq 0 ]] &&
6995 error "Please check MDS inode size calculation: \
6996 more than $left_size-byte space left in inode."
6997 echo "Verified: at most $left_size-byte space left in inode."
6999 unmount_ldiskfs $SINGLEMDS
7001 for i in $(seq $OSTCOUNT); do
7002 stop ost$i -f || error "stop ost$i failed"
7005 run_test 87 "check if MDT inode can hold EAs with N stripes properly"
7008 [ "$mds1_FSTYPE" == zfs ] &&
7009 skip "LU-6662: no implementation for ZFS"
7013 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
7014 --reformat $(mdsdevname 1) || error "add mds1 failed"
7016 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
7017 grep -e \".*opts:.*errors=remount-ro.*\"" ||
7018 error "default mount options is missing"
7020 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
7021 --mountfsoptions="user_xattr,errors=panic" \
7022 --reformat $(mdsdevname 1) || error "add mds1 failed"
7024 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
7025 grep -e \".*opts:.*errors=panic.*\"" ||
7026 error "user can't override default mount options"
7028 run_test 88 "check the default mount options can be overridden"
7030 test_89() { # LU-7131
7031 [[ "$MDS1_VERSION" -ge $(version_code 2.9.54) ]] ||
7032 skip "Need MDT version at least 2.9.54"
7034 local key=failover.node
7035 local val1=192.0.2.254@tcp0 # Reserved IPs, see RFC 5735
7036 local val2=192.0.2.255@tcp0
7037 local mdsdev=$(mdsdevname 1)
7042 if [[ "$mds1_FSTYPE" == zfs ]]; then
7043 import_zpool mds1 || return ${PIPESTATUS[0]}
7046 # Check that parameters are added correctly
7047 echo "tunefs --param $key=$val1"
7048 do_facet mds1 "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
7049 error "tunefs --param $key=$val1 failed"
7050 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7051 error "tunefs --dryrun failed"
7052 params=${params##*Parameters:}
7053 params=${params%%exiting*}
7054 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val1) = "1" ] ||
7055 error "on-disk parameter not added correctly via tunefs"
7057 # Check that parameters replace existing instances when added
7058 echo "tunefs --param $key=$val2"
7059 do_facet mds1 "$TUNEFS --param $key=$val2 $mdsdev >/dev/null" ||
7060 error "tunefs --param $key=$val2 failed"
7061 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7062 error "tunefs --dryrun failed"
7063 params=${params##*Parameters:}
7064 params=${params%%exiting*}
7065 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "1" ] ||
7066 error "on-disk parameter not replaced via tunefs"
7067 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val2) = "1" ] ||
7068 error "on-disk parameter not replaced correctly via tunefs"
7070 # Check that a parameter is erased properly
7071 echo "tunefs --erase-param $key"
7072 do_facet mds1 "$TUNEFS --erase-param $key $mdsdev >/dev/null" ||
7073 error "tunefs --erase-param $key failed"
7074 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7075 error "tunefs --dryrun failed"
7076 params=${params##*Parameters:}
7077 params=${params%%exiting*}
7078 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "0" ] ||
7079 error "on-disk parameter not erased correctly via tunefs"
7081 # Check that all the parameters are erased
7082 do_facet mds1 "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
7083 error "tunefs --param $key=$val1 failed"
7084 echo "tunefs --erase-params"
7085 do_facet mds1 "$TUNEFS --erase-params $mdsdev >/dev/null" ||
7086 error "tunefs --erase-params failed"
7087 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7088 error "tunefs --dryrun failed"
7089 params=${params##*Parameters:}
7090 params=${params%%exiting*}
7091 params=$(echo $params | tr ' ' '\n')
7093 error "all on-disk parameters not erased correctly via tunefs $params"
7095 # Check the order of options --erase-params and --param
7096 echo "tunefs --param $key=$val1 --erase-params"
7098 "$TUNEFS --param $key=$val1 --erase-params $mdsdev >/dev/null"||
7099 error "tunefs --param $key=$val1 --erase-params failed"
7100 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7101 error "tunefs --dryrun failed"
7102 params=${params##*Parameters:}
7103 params=${params%%exiting*}
7104 [ $(echo $params | tr ' ' '\n') == "$key=$val1" ] ||
7105 error "on-disk param not added correctly with --erase-params"
7109 run_test 89 "check tunefs --param and --erase-param{s} options"
7112 # $2 (optional) value of max_mod_rpcs_in_flight to set
7113 check_max_mod_rpcs_in_flight() {
7121 idx=$(printf "%04x" $($LFS getdirstripe -i $dir))
7122 facet="mds$((0x$idx + 1))"
7124 if [ -z "$mmr" ]; then
7125 # get value of max_mod_rcps_in_flight
7126 mmr=$($LCTL get_param -n \
7127 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) ||
7128 error "Unable to get max_mod_rpcs_in_flight"
7129 echo "max_mod_rcps_in_flight is $mmr"
7131 # set value of max_mod_rpcs_in_flight
7133 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight=$mmr ||
7134 error "Unable to set max_mod_rpcs_in_flight to $mmr"
7135 echo "max_mod_rpcs_in_flight set to $mmr"
7138 # create mmr+1 files
7139 echo "creating $((mmr + 1)) files ..."
7141 for i in $(seq $((mmr + 1))); do
7147 # consumes mmr-1 modify RPC slots
7148 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7149 # drop requests on MDT so that RPC slots are consumed
7150 # during all the request resend interval
7151 do_facet $facet "$LCTL set_param fail_loc=0x159"
7152 echo "launch $((mmr - 1)) chmod in parallel ..."
7153 for i in $(seq $((mmr - 1))); do
7154 chmod 0600 $dir/file-$i &
7158 # send one additional modify RPC
7159 do_facet $facet "$LCTL set_param fail_loc=0"
7160 echo "launch 1 additional chmod in parallel ..."
7161 chmod 0600 $dir/file-$mmr &
7164 # check this additional modify RPC get a modify RPC slot
7165 # and succeed its operation
7166 checkstat -vp 0600 $dir/file-$mmr ||
7167 error "Unable to send $mmr modify RPCs in parallel"
7172 # consumes mmr modify RPC slots
7173 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7174 # drop requests on MDT so that RPC slots are consumed
7175 # during all the request resend interval
7176 do_facet $facet "$LCTL set_param fail_loc=0x159"
7177 echo "launch $mmr chmod in parallel ..."
7178 for i in $(seq $mmr); do
7179 chmod 0666 $dir/file-$i &
7183 # send one additional modify RPC
7184 do_facet $facet "$LCTL set_param fail_loc=0"
7185 echo "launch 1 additional chmod in parallel ..."
7186 chmod 0666 $dir/file-$((mmr + 1)) &
7189 # check this additional modify RPC blocked getting a modify RPC slot
7190 checkstat -vp 0644 $dir/file-$((mmr + 1)) ||
7191 error "Unexpectedly send $(($mmr + 1)) modify RPCs in parallel"
7195 get_mdt_max_mod_rpcs_in_flight_val() {
7199 # It will be enough to get one value from one MDT
7200 max_mrif=$(do_facet $facet \
7201 "$LCTL get_param -n \
7202 mdt.$FSNAME-MDT*.max_mod_rpcs_in_flight | head -n 1")
7203 if [[ $max_mrif ]]; then
7206 max_mrif=$(do_facet $facet \
7207 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
7212 set_mdt_max_mod_rpcs_in_flight() {
7217 lctl_op=$($LCTL get_param \
7218 mdt.*.max_mod_rpcs_in_flight)
7219 if [[ $lctl_op ]]; then
7222 mdt.$FSNAME-MDT*.max_mod_rpcs_in_flight=$max_mrif"
7226 /sys/module/mdt/parameters/max_mod_rpcs_per_client"
7227 echo "the deprecated max_mod_rpcs_per_client \
7228 parameter was involved"
7235 [[ $($LCTL get_param mdc.*.import |
7236 grep "connect_flags:.*multi_mod_rpc") ]] ||
7237 skip "Need MDC with 'multi_mod_rpcs' feature"
7239 # check default value
7240 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
7241 check_max_mod_rpcs_in_flight $DIR/$tdir
7245 run_test 90a "check max_mod_rpcs_in_flight is enforced"
7255 [[ $($LCTL get_param mdc.*.import |
7256 grep "connect_flags:.*multi_mod_rpc") ]] ||
7257 skip "Need MDC with 'multi_mod_rpcs' feature"
7260 # update max_mod_rpcs_in_flight
7261 $LFS mkdir -c1 $DIR/${tdir}1 || error "mkdir $DIR/${tdir}1 failed"
7262 check_max_mod_rpcs_in_flight $DIR/${tdir}1 1
7265 # check client is able to send multiple modify RPCs in paralell
7266 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
7267 grep -c "multi_mod_rpcs")
7268 if [ "$tmp" -ne $MDSCOUNT ]; then
7270 skip "Client not able to send multiple modify RPCs in parallel"
7273 # update max_mod_rpcs_in_flight
7274 $LFS mkdir -c1 $DIR/${tdir}2 || error "mkdir $DIR/${tdir}2 failed"
7275 check_max_mod_rpcs_in_flight $DIR/${tdir}2 5
7278 $LFS mkdir -c1 $DIR/${tdir}3 || error "mkdir $DIR/${tdir}3 failed"
7279 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/${tdir}3))
7280 facet="mds$((0x$idx + 1))"
7282 mmrpc=$(get_mdt_max_mod_rpcs_in_flight_val $facet)
7283 echo "mdt_max_mod_rpcs_in_flight is $mmrpc"
7284 # update max_mod_rpcs_in_flight
7285 umount_client $MOUNT
7286 set_mdt_max_mod_rpcs_in_flight 16 $facet
7288 $LCTL set_param mdc.$FSNAME-MDT$idx-mdc-*.max_rpcs_in_flight=17
7289 check_max_mod_rpcs_in_flight $DIR/${tdir}3 16
7291 # restore MDT max_mod_rpcs_in_flight initial value
7292 set_mdt_max_mod_rpcs_in_flight $mmrpc $facet
7294 rm -rf $DIR/${tdir}?
7297 run_test 90b "check max_mod_rpcs_in_flight is enforced after update"
7300 # get max_rpcs_in_flight value
7301 mrif_90c=$($LCTL get_param -n \
7302 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
7303 echo "max_rpcs_in_flight is $mrif_90c"
7305 # get MDC max_mod_rpcs_in_flight value
7306 mmrif_90c=$($LCTL get_param -n \
7307 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight)
7308 echo "MDC max_mod_rpcs_in_flight is $mmrif_90c"
7310 # get MDT max_mod_rpcs_in_flight value
7311 mmrpc_90c=$(get_mdt_max_mod_rpcs_in_flight_val "mds1")
7312 echo "mdt_max_mod_rpcs_in_flight is $mmrpc_90c"
7315 restore_params_90c() {
7318 # restore max_rpcs_in_flight value
7319 do_facet mgs $LCTL set_param -P \
7320 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$mrif_90c
7322 # restore max_mod_rpcs_in_flight value
7323 do_facet mgs $LCTL set_param -P \
7324 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mmrif_90c
7326 # restore MDT max_mod_rpcs_in_flight value
7327 set_mdt_max_mod_rpcs_in_flight $mmrpc_90c "mds1"
7335 [[ $($LCTL get_param mdc.*.import |
7336 grep "connect_flags:.*multi_mod_rpc") ]] ||
7337 skip "Need MDC with 'multi_mod_rpcs' feature"
7339 # check client is able to send multiple modify RPCs in paralell
7340 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
7341 grep -c "multi_mod_rpcs")
7342 if [ "$tmp" -ne $MDSCOUNT ]; then
7344 skip "Client not able to send multiple modify RPCs in parallel"
7348 stack_trap restore_params_90c
7351 # attempt to set max_mod_rpcs_in_flight to max_rpcs_in_flight value
7352 # prerequisite: set MDT max_mod_rpcs_in_flight to
7353 # max_rpcs_in_flight value
7354 set_mdt_max_mod_rpcs_in_flight $mrif_90c "mds1"
7356 # if max_mod_rpcs_in_flight is set to be equal to or larger than
7357 # max_rpcs_in_flight, then max_rpcs_in_flight will be increased
7358 if [[ "$CLIENT_VERSION" -ge $(version_code 2.13.53) ]]; then
7360 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif_90c ||
7361 error "set max_mod_rpcs_in_flight to $mrif_90c failed"
7363 local new_mrif=$($LCTL get_param -n \
7364 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
7365 ((new_mrif == mrif_90c + 1)) ||
7366 error "max_rpcs_in_flight was not increased"
7369 umount_client $MOUNT
7370 set_mdt_max_mod_rpcs_in_flight $mmrpc_90c "mds1"
7374 # attempt to set max_mod_rpcs_in_flight to MDT max_mod_rpcs_in_flight+1
7375 # prerequisite: set max_rpcs_in_flight to MDT max_mod_rpcs_in_flight+2
7377 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$((mmrpc_90c + 2))
7380 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$((mmrpc_90c + 1)) &&
7381 error "set max_mod_rpcs_in_flight to $((mmrpc_90c + 1)) should fail"
7384 # attempt to set max_mod_rpcs_in_flight permanently
7385 do_facet mgs $LCTL set_param -P \
7386 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$mrif_90c
7388 do_facet mgs $LCTL set_param -P \
7389 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif_90c
7391 remount_client $MOUNT
7393 wait_update_facet --verbose client "$LCTL get_param -n \
7394 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight" \
7395 "$((mrif_90c + 1))" ||
7396 error "expected '$((mrif_90c + 1))' for max_rpcs_in_flight"
7398 wait_update_facet --verbose client "$LCTL get_param -n \
7399 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight" \
7401 error "expected '$mrif_90c' for max_mod_rpcs_in_flight"
7406 run_test 90c "check max_mod_rpcs_in_flight update limits"
7417 [[ $($LCTL get_param mdc.*.import |
7418 grep "connect_flags:.*multi_mod_rpc") ]] ||
7419 skip "Need MDC with 'multi_mod_rpcs' feature"
7421 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
7422 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
7423 facet="mds$((0x$idx + 1))"
7425 # check client version supports multislots
7426 tmp=$($LCTL get_param -N \
7427 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
7428 if [ -z "$tmp" ]; then
7430 skip "Client does not support multiple modify RPCs in flight"
7433 # get current value of max_mod_rcps_in_flight
7434 mmr=$($LCTL get_param -n \
7435 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
7436 echo "max_mod_rcps_in_flight is $mmr"
7439 echo "creating $mmr files ..."
7441 for i in $(seq $mmr); do
7442 touch $DIR/$tdir/file-$i
7445 # prepare for close RPC
7446 multiop_bg_pause $DIR/$tdir/file-close O_c
7449 # consumes mmr modify RPC slots
7450 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7451 # drop requests on MDT so that RPC slots are consumed
7452 # during all the request resend interval
7453 do_facet $facet "$LCTL set_param fail_loc=0x159"
7454 echo "launch $mmr chmod in parallel ..."
7455 for i in $(seq $mmr); do
7456 chmod 0600 $DIR/$tdir/file-$i &
7459 # send one additional close RPC
7460 do_facet $facet "$LCTL set_param fail_loc=0"
7461 echo "launch 1 additional close in parallel ..."
7463 cancel_lru_locks mdc
7466 # check this additional close RPC get a modify RPC slot
7467 # and multiop process completed
7468 [ -d /proc/$pid ] &&
7469 error "Unable to send the additional close RPC in parallel"
7474 run_test 90d "check one close RPC is allowed above max_mod_rpcs_in_flight"
7476 check_uuid_on_ost() {
7478 do_facet ost1 "$LCTL get_param obdfilter.${FSNAME}*.exports.'$nid'.uuid"
7481 check_uuid_on_mdt() {
7483 do_facet $SINGLEMDS "$LCTL get_param mdt.${mds1_svc}*.exports.'$nid'.uuid"
7491 [[ "$OST1_VERSION" -ge $(version_code 2.7.63) ]] ||
7492 skip "Need OST version at least 2.7.63"
7493 [[ "$MDS1_VERSION" -ge $(version_code 2.7.63) ]] ||
7494 skip "Need MDT version at least 2.7.63"
7496 start_mds || error "MDS start failed"
7497 start_ost || error "unable to start OST"
7498 mount_client $MOUNT || error "client start failed"
7499 check_mount || error "check_mount failed"
7502 nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g")
7506 uuid=$(get_client_uuid $MOUNT)
7508 echo "list nids on mdt:"
7509 do_facet $SINGLEMDS "$LCTL list_param mdt.${FSNAME}*.exports.*"
7510 echo "uuid from $nid:"
7511 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
7513 found=$(check_uuid_on_mdt $nid | grep $uuid)
7514 [ -z "$found" ] && error "can't find $uuid $nid on MDT"
7515 found=$(check_uuid_on_ost $nid | grep $uuid)
7516 [ -z "$found" ] && error "can't find $uuid $nid on OST"
7518 # umount the client so it won't reconnect
7519 manual_umount_client --force || error "failed to umount $?"
7520 # shouldn't disappear on MDS after forced umount
7521 found=$(check_uuid_on_mdt $nid | grep $uuid)
7522 [ -z "$found" ] && error "can't find $uuid $nid"
7525 do_facet $SINGLEMDS \
7526 "$LCTL set_param -n mdt.${mds1_svc}.evict_client nid:$nid"
7527 sleep 1 # eviction above is async, give it some time to proceed
7529 found=$(check_uuid_on_mdt $nid | grep $uuid)
7530 [ -n "$found" ] && error "found $uuid $nid on MDT"
7531 found=$(check_uuid_on_ost $nid | grep $uuid)
7532 [ -n "$found" ] && error "found $uuid $nid on OST"
7534 # check it didn't reconnect (being umounted)
7535 sleep $((TIMEOUT+1))
7536 found=$(check_uuid_on_mdt $nid | grep $uuid)
7537 [ -n "$found" ] && error "found $uuid $nid on MDT"
7538 found=$(check_uuid_on_ost $nid | grep $uuid)
7539 [ -n "$found" ] && error "found $uuid $nid on OST"
7543 run_test 91 "evict-by-nid support"
7545 generate_ldev_conf() {
7546 # generate an ldev.conf file
7547 local ldevconfpath=$1
7549 local fsldevformat=""
7552 local facets="mgs,$(get_facets OST),$(get_facets MDS)"
7553 for facet in ${facets//,/ }; do
7555 fstype=$(facet_fstype $facet)
7556 [ "$fstype" = zfs ] && fsldevformat="$fstype:"
7557 local host=$(facet_host $facet)
7559 local varfo=${facet}failover_HOST
7560 if [ $facet == mgs ] && combined_mgs_mds; then
7561 varfo=mds1failover_HOST
7563 [ -n "${!varfo}" ] && fo=${!varfo}
7564 local type=$(echo $facet | tr -d "[:digit:]" | \
7565 tr "[:lower:]" "[:upper:]" | sed s/MDS/MDT/ )
7567 [ ${facet} == mgs ] ||
7568 num=$(facet_number $facet)
7569 printf "%s\t%s\t%s-%s%04x\t%s%s\n" \
7576 $(facet_device $facet) >> $ldevconfpath
7579 echo "----- $ldevconfpath -----"
7581 echo "--- END $ldevconfpath ---"
7586 # generate a nids file (mapping between hostname to nid)
7587 # looks like we only have the MGS nid available to us
7588 # so just echo that to a file
7590 echo -e "${mgs_HOST}\t${MGSNID}" > $nidspath
7592 echo "----- $nidspath -----"
7594 echo "--- END $nidspath ---"
7597 compare_ldev_output() {
7598 local ldev_output=$1
7599 local expected_output=$2
7601 sort $expected_output -o $expected_output
7602 sort $ldev_output -o $ldev_output
7604 echo "-- START OF LDEV OUTPUT --"
7606 echo "--- END OF LDEV OUTPUT ---"
7608 echo "-- START OF EXPECTED OUTPUT --"
7609 cat $expected_output
7610 echo "--- END OF EXPECTED OUTPUT ---"
7612 diff $expected_output $ldev_output
7617 if [ -z "$LDEV" ]; then
7618 error "ldev is missing!"
7621 local LDEVCONFPATH=$TMP/ldev.conf
7622 local NIDSPATH=$TMP/nids
7624 echo "Host is $(hostname)"
7626 generate_ldev_conf $LDEVCONFPATH
7627 generate_nids $NIDSPATH
7629 # echo the mgs nid and compare it to environment variable MGSNID
7630 local facets="$(get_facets OST),$(get_facets MDS),mgs"
7631 for facet in ${facets//,/ }; do
7632 local host=$(facet_host $facet)
7633 local output=$($LDEV -c $LDEVCONFPATH -H $host -n $NIDSPATH echo %m)
7635 echo "-- START OF LDEV OUTPUT --"
7637 echo "--- END OF LDEV OUTPUT ---"
7640 error "ldev failed to execute!"
7642 # need to process multiple lines because of
7643 # several targets on host
7644 echo -e $output | awk '{ print $2 }' | while read -r line ; do
7645 [ "$line" = "$MGSNID" ] ||
7646 error "ldev failed mgs nid '$line', \
7650 rm -f $LDEVCONFPATH $NIDSPATH
7652 run_test 92 "ldev returns MGS NID correctly in command substitution"
7655 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs"
7658 #start mgs or mgs/mdt0
7659 if ! combined_mgs_mds ; then
7666 start_ost || error "OST0 start fail"
7668 #define OBD_FAIL_MGS_WRITE_TARGET_DELAY 0x90e
7669 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x8000090e"
7670 for num in $(seq 2 $MDSCOUNT); do
7674 mount_client $MOUNT || error "mount client fails"
7675 wait_osc_import_state mds ost FULL
7676 wait_osc_import_ready client ost
7677 check_mount || error "check_mount failed"
7679 cleanup || error "cleanup failed with $?"
7681 run_test 93 "register mulitple MDT at the same time"
7684 if [ -z "$LDEV" ]; then
7685 error "ldev is missing!"
7688 local LDEVCONFPATH=$TMP/ldev.conf
7689 local NIDSPATH=$TMP/nids
7691 generate_ldev_conf $LDEVCONFPATH
7692 generate_nids $NIDSPATH
7694 local LDEV_OUTPUT=$TMP/ldev-output.txt
7695 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME > $LDEV_OUTPUT
7697 # ldev failed, error
7698 if [ $? -ne 0 ]; then
7699 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7700 error "ldev failed to execute!"
7704 local EXPECTED_OUTPUT=$TMP/ldev-expected.txt
7706 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
7708 for num in $(seq $MDSCOUNT); do
7709 printf "%s-MDT%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7712 for num in $(seq $OSTCOUNT); do
7713 printf "%s-OST%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7716 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7718 if [ $? -ne 0 ]; then
7719 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7720 error "ldev failed to produce the correct hostlist!"
7723 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7725 run_test 94 "ldev outputs correct labels for file system name query"
7728 if [ -z "$LDEV" ]; then
7729 error "ldev is missing!"
7732 local LDEVCONFPATH=$TMP/ldev.conf
7733 local NIDSPATH=$TMP/nids
7735 generate_ldev_conf $LDEVCONFPATH
7736 generate_nids $NIDSPATH
7740 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME &>/dev/null
7741 if [ $? -ne 0 ]; then
7742 rm $LDEVCONFPATH $NIDSPATH
7743 error "ldev label filtering w/ -F failed!"
7747 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -l &>/dev/null
7748 if [ $? -ne 0 ]; then
7749 rm $LDEVCONFPATH $NIDSPATH
7750 error "ldev label filtering w/ -l failed!"
7754 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f &>/dev/null
7755 if [ $? -ne 0 ]; then
7756 rm $LDEVCONFPATH $NIDSPATH
7757 error "ldev label filtering w/ -f failed!"
7761 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a &>/dev/null
7762 if [ $? -ne 0 ]; then
7763 rm $LDEVCONFPATH $NIDSPATH
7764 error "ldev label filtering w/ -a failed!"
7769 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -F $FSNAME &>/dev/null
7770 if [ $? -eq 0 ]; then
7771 rm $LDEVCONFPATH $NIDSPATH
7772 error "ldev label filtering w/ -a and -F incorrectly succeeded"
7776 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -f &>/dev/null
7777 if [ $? -eq 0 ]; then
7778 rm $LDEVCONFPATH $NIDSPATH
7779 error "ldev label filtering w/ -a and -f incorrectly succeeded"
7783 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -l &>/dev/null
7784 if [ $? -eq 0 ]; then
7785 rm $LDEVCONFPATH $NIDSPATH
7786 error "ldev label filtering w/ -a and -l incorrectly succeeded"
7790 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f -l &>/dev/null
7791 if [ $? -eq 0 ]; then
7792 rm $LDEVCONFPATH $NIDSPATH
7793 error "ldev label filtering w/ -f and -l incorrectly succeeded"
7797 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -l &>/dev/null
7798 if [ $? -eq 0 ]; then
7799 rm $LDEVCONFPATH $NIDSPATH
7800 error "ldev label filtering w/ -F and -l incorrectly succeeded"
7803 # file sys & foreign
7804 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -f &>/dev/null
7805 if [ $? -eq 0 ]; then
7806 rm $LDEVCONFPATH $NIDSPATH
7807 error "ldev label filtering w/ -F and -f incorrectly succeeded"
7810 rm $LDEVCONFPATH $NIDSPATH
7812 run_test 95 "ldev should only allow one label filter"
7815 if [ -z "$LDEV" ]; then
7816 error "ldev is missing!"
7819 local LDEVCONFPATH=$TMP/ldev.conf
7820 local NIDSPATH=$TMP/nids
7822 generate_ldev_conf $LDEVCONFPATH
7823 generate_nids $NIDSPATH
7825 local LDEV_OUTPUT=$TMP/ldev-output.txt
7826 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -H $mgs_HOST \
7828 awk '{print $2}' > $LDEV_OUTPUT
7830 # ldev failed, error
7831 if [ $? -ne 0 ]; then
7832 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7833 error "ldev failed to execute!"
7837 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
7839 echo "$mgs_HOST-$(facet_fstype mgs)" > $EXPECTED_OUTPUT
7841 local facets="$(get_facets OST),$(get_facets MDS)"
7842 for facet in ${facets//,/ }; do
7843 local host=$(facet_host $facet)
7844 [ "$mgs_HOST" == "$host" ] &&
7845 echo "$host-$(facet_fstype $facet)" \
7849 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7851 if [ $? -ne 0 ]; then
7852 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7853 error "ldev failed to produce the correct output!"
7856 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7858 run_test 96 "ldev returns hostname and backend fs correctly in command sub"
7861 if [ -z "$LDEV" ]; then
7862 error "ldev is missing!"
7865 local LDEVCONFPATH=$TMP/ldev.conf
7866 local NIDSPATH=$TMP/nids
7868 generate_ldev_conf $LDEVCONFPATH
7869 generate_nids $NIDSPATH
7871 local LDEV_OUTPUT=$TMP/ldev-output.txt
7872 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
7874 echo -e "\nMDT role"
7875 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mdt > $LDEV_OUTPUT
7877 if [ $? -ne 0 ]; then
7878 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7879 error "ldev failed to execute for mdt role!"
7882 for num in $(seq $MDSCOUNT); do
7883 printf "%s-MDT%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7886 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7888 if [ $? -ne 0 ]; then
7889 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7890 error "ldev failed to produce the correct output for mdt role!"
7893 echo -e "\nOST role"
7894 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R ost > $LDEV_OUTPUT
7896 if [ $? -ne 0 ]; then
7897 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
7898 error "ldev failed to execute for ost role!"
7902 for num in $(seq $OSTCOUNT); do
7903 printf "%s-OST%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7906 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7908 if [ $? -ne 0 ]; then
7909 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7910 error "ldev failed to produce the correct output for ost role!"
7913 echo -e "\nMGS role"
7914 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mgs > $LDEV_OUTPUT
7916 if [ $? -ne 0 ]; then
7917 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
7918 error "ldev failed to execute for mgs role!"
7921 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
7923 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7925 if [ $? -ne 0 ]; then
7926 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7927 error "ldev failed to produce the correct output for mgs role!"
7930 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7932 run_test 97 "ldev returns correct ouput when querying based on role"
7937 local temp=$MDS_MOUNT_OPTS
7940 check_mount || error "mount failed"
7941 mountopt="user_xattr"
7942 for ((x = 1; x <= 400; x++)); do
7943 mountopt="$mountopt,user_xattr"
7945 mount_client $MOUNT remount,$mountopt 2>&1 | grep "too long" ||
7946 error "Buffer overflow check failed"
7947 cleanup || error "cleanup failed"
7949 run_test 98 "Buffer-overflow check while parsing mount_opts"
7953 [[ "$ost1_FSTYPE" != ldiskfs ]] &&
7954 skip "ldiskfs only test"
7955 [[ "$OST1_VERSION" -ge $(version_code 2.8.57) ]] ||
7956 skip "Need OST version at least 2.8.57"
7958 local ost_opts="$(mkfs_opts ost1 $(ostdevname 1)) \
7959 --reformat $(ostdevname 1) $(ostvdevname 1)"
7960 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" &&
7961 skip "meta_bg already set"
7964 if [[ ${!opts} != *mkfsoptions* ]]; then
7965 eval opts=\"${!opts} \
7966 --mkfsoptions='\\\"-O ^resize_inode,meta_bg\\\"'\"
7968 local val=${!opts//--mkfsoptions=\\\"/ \
7969 --mkfsoptions=\\\"-O ^resize_inode,meta_bg }
7973 echo "params: $opts"
7976 add ost1 $opts || error "add ost1 failed with new params"
7978 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" ||
7979 error "meta_bg is not set"
7981 run_test 99 "Adding meta_bg option"
7986 start_mgsmds || error "MDS start failed"
7987 start_ost || error "unable to start OST"
7988 mount_client $MOUNT || error "client start failed"
7989 check_mount || error "check_mount failed"
7998 do_facet mgs 'lshowmount -v' | awk 'BEGIN {NR == 0; rc=1} /MGS:/ {rc=0}
7999 END {exit rc}' || error "lshowmount have no output MGS"
8001 do_facet mds1 'lshowmount -v' | awk 'BEGIN {NR == 2; rc=1} /-MDT0000:/
8002 {rc=0} END {exit rc}' || error "lshowmount have no output MDT0"
8004 do_facet ost1 'lshowmount -v' | awk 'BEGIN {NR == 4; rc=1} /-OST0000:/
8005 {rc=0} END {exit rc}' || error "lshowmount have no output OST0"
8007 cleanup || error "cleanup failed with $?"
8009 run_test 100 "check lshowmount lists MGS, MDT, OST and 0@lo"
8012 local createmany_pid
8013 local dev=$FSNAME-OST0000-osc-MDT0000
8017 createmany -o $DIR1/$tdir/$tfile-%d 50000 &
8019 # MDT->OST reconnection causes MDT<->OST last_id synchornisation
8020 # via osp_precreate_cleanup_orphans.
8021 for ((i = 0; i < 100; i++)); do
8022 for ((k = 0; k < 10; k++)); do
8023 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;" \
8024 "$LCTL --device $dev activate"
8027 ls -asl $MOUNT | grep '???' &&
8028 { kill -9 $createmany_pid &>/dev/null;
8029 error "File has no object on OST"; }
8031 kill -s 0 $createmany_pid || break
8033 wait $createmany_pid
8035 unlinkmany $DIR1/$tdir/$tfile-%d 50000
8038 run_test 101a "Race MDT->OST reconnection with create"
8041 local dev=$FSNAME-OST0000-osc-MDT0000
8042 local dir=$DIR1/$tdir
8046 $LFS setstripe -c 1 -i 0 $dir
8047 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;"
8048 #define OBD_FAIL_OSP_CON_EVENT_DELAY 0x2107
8049 do_facet mds1 "$LCTL set_param fail_loc=0x80002107 fail_val=20"
8050 do_facet $SINGLEMDS "$LCTL --device $dev activate;"
8055 wait_osc_import_state client ost1 FULL
8056 touch $dir/$tfile || error "Can't create file"
8060 run_test 101b "Race events DISCONNECT and ACTIVE in osp"
8063 [[ "$MDS1_VERSION" -gt $(version_code 2.9.53) ]] ||
8064 skip "Need server version greater than 2.9.53"
8065 [[ “$(mdsdevname 1)” != “$(mgsdevname)” ]] &&
8066 [[ “$(facet_host mds1)” = “$(facet_host mgs)” ]] &&
8067 skip "MGS must be on different node or combined"
8069 cleanup || error "cleanup failed with $?"
8071 local mds1dev=$(mdsdevname 1)
8072 local mds1mnt=$(facet_mntpt mds1)
8073 local mds1opts=$MDS_MOUNT_OPTS
8075 if [ "$mds1_FSTYPE" == ldiskfs ] &&
8076 ! do_facet mds1 test -b $mds1dev; then
8077 mds1opts=$(csa_add "$mds1opts" -o loop)
8079 if [[ "$mds1_FSTYPE" == zfs ]]; then
8080 import_zpool mds1 || return ${PIPESTATUS[0]}
8083 # unload all and only load libcfs to allow fail_loc setting
8084 do_facet mds1 $LUSTRE_RMMOD || error "unable to unload modules"
8085 do_facet mds1 modprobe libcfs || error "libcfs not loaded"
8086 do_facet mds1 lsmod \| grep libcfs || error "libcfs not loaded"
8088 #define OBD_FAIL_OBDCLASS_MODULE_LOAD 0x60a
8089 do_facet mds1 "$LCTL set_param fail_loc=0x8000060a"
8091 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts &&
8092 error "mdt start must fail"
8093 do_facet mds1 lsmod \| grep obdclass && error "obdclass must not load"
8095 do_facet mds1 "$LCTL set_param fail_loc=0x0"
8097 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts ||
8098 error "mdt start must not fail"
8100 cleanup || error "cleanup failed with $?"
8102 run_test 102 "obdclass module cleanup upon error"
8107 echo "rename $FSNAME to $newname"
8109 if ! combined_mgs_mds ; then
8110 local dev=$(mgsdevname)
8113 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
8114 error "(7) Fail to rename MGS"
8115 if [ "$(facet_fstype mgs)" = "zfs" ]; then
8116 reimport_zpool mgs $newname-mgs
8120 for num in $(seq $MDSCOUNT); do
8121 local dev=$(mdsdevname $num)
8123 do_facet mds${num} \
8124 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
8125 error "(8) Fail to rename MDT $num"
8126 if [ "$(facet_fstype mds${num})" = "zfs" ]; then
8127 reimport_zpool mds${num} $newname-mdt${num}
8131 for num in $(seq $OSTCOUNT); do
8132 local dev=$(ostdevname $num)
8134 do_facet ost${num} \
8135 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
8136 error "(9) Fail to rename OST $num"
8137 if [ "$(facet_fstype ost${num})" = "zfs" ]; then
8138 reimport_zpool ost${num} $newname-ost${num}
8143 test_103_set_pool() {
8147 do_facet mgs $LCTL pool_add $FSNAME.$pname ${FSNAME}-$ost_x ||
8148 error "Fail to add $ost_x to $FSNAME.$pname"
8149 wait_update $HOSTNAME \
8150 "lctl get_param -n lov.$FSNAME-clilov-*.pools.$pname |
8151 grep $ost_x" "$FSNAME-${ost_x}_UUID" ||
8152 error "$ost_x is NOT in pool $FSNAME.$pname"
8155 test_103_check_pool() {
8156 local save_fsname=$1
8159 stat $DIR/$tdir/test-framework.sh ||
8160 error "($errno) Fail to stat"
8161 do_facet mgs $LCTL pool_list $FSNAME.pool1 ||
8162 error "($errno) Fail to list $FSNAME.pool1"
8163 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname ||
8164 error "($errno) Fail to list $FSNAME.$save_fsname"
8165 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname |
8166 grep ${FSNAME}-OST0000 ||
8167 error "($errno) List $FSNAME.$save_fsname is invalid"
8169 local pname=$($LFS getstripe --pool $DIR/$tdir/d0)
8170 [ "$pname" = "$save_fsname" ] ||
8171 error "($errno) Unexpected pool name $pname"
8175 check_mount_and_prep
8177 mkdir $DIR/$tdir || error "(1) Fail to mkdir $DIR/$tdir"
8178 cp $LUSTRE/tests/test-framework.sh $DIR/$tdir ||
8179 error "(2) Fail to copy test-framework.sh"
8181 do_facet mgs $LCTL pool_new $FSNAME.pool1 ||
8182 error "(3) Fail to create $FSNAME.pool1"
8183 # name the pool name as the fsname
8184 do_facet mgs $LCTL pool_new $FSNAME.$FSNAME ||
8185 error "(4) Fail to create $FSNAME.$FSNAME"
8187 test_103_set_pool $FSNAME OST0000
8189 $LFS setstripe -p $FSNAME $DIR/$tdir/d0 ||
8190 error "(6) Fail to setstripe on $DIR/$tdir/d0"
8195 test_renamefs mylustre
8197 local save_fsname=$FSNAME
8201 test_103_check_pool $save_fsname 7
8203 if [ $OSTCOUNT -ge 2 ]; then
8204 test_103_set_pool $save_fsname OST0001
8207 $LFS setstripe -p $save_fsname $DIR/$tdir/f0 ||
8208 error "(16) Fail to setstripe on $DIR/$tdir/f0"
8217 test_103_check_pool $save_fsname 17
8221 test_renamefs $save_fsname
8227 run_test 103 "rename filesystem name"
8229 test_104a() { # LU-6952
8230 local mds_mountopts=$MDS_MOUNT_OPTS
8231 local ost_mountopts=$OST_MOUNT_OPTS
8232 local mds_mountfsopts=$MDS_MOUNT_FS_OPTS
8233 local lctl_ver=$(do_facet $SINGLEMDS $LCTL --version |
8236 [[ $(version_code $lctl_ver) -lt $(version_code 2.9.55) ]] &&
8237 skip "this test needs utils above 2.9.55"
8239 # specify "acl" in mount options used by mkfs.lustre
8240 if [ -z "$MDS_MOUNT_FS_OPTS" ]; then
8241 MDS_MOUNT_FS_OPTS="acl,user_xattr"
8244 MDS_MOUNT_FS_OPTS="${MDS_MOUNT_FS_OPTS},acl,user_xattr"
8247 echo "mountfsopt: $MDS_MOUNT_FS_OPTS"
8249 #reformat/remount the MDT to apply the MDT_MOUNT_FS_OPT options
8251 if ! combined_mgs_mds ; then
8255 if [ -z "$MDS_MOUNT_OPTS" ]; then
8256 MDS_MOUNT_OPTS="-o noacl"
8258 MDS_MOUNT_OPTS="${MDS_MOUNT_OPTS},noacl"
8261 for num in $(seq $MDSCOUNT); do
8262 start mds$num $(mdsdevname $num) $MDS_MOUNT_OPTS ||
8263 error "Failed to start MDS"
8266 for num in $(seq $OSTCOUNT); do
8267 start ost$num $(ostdevname $num) $OST_MOUNT_OPTS ||
8268 error "Failed to start OST"
8272 setfacl -m "d:$RUNAS_ID:rwx" $MOUNT &&
8273 error "ACL is applied when FS is mounted with noacl."
8275 MDS_MOUNT_OPTS=$mds_mountopts
8276 OST_MOUNT_OPTS=$ost_mountopts
8277 MDS_MOUNT_FS_OPTS=$mds_mountfsopts
8279 run_test 104a "Make sure user defined options are reflected in mount"
8281 test_104b() { # LU-12859
8282 mount_client $MOUNT3 flock,localflock
8283 stack_trap "umount_client $MOUNT3" EXIT
8284 mount | grep "$MOUNT3 .*,flock" && error "flock is still set"
8285 mount | grep "$MOUNT3 .*,localflock" || error "localflock is not set"
8286 umount_client $MOUNT3
8287 mount_client $MOUNT3 localflock,flock
8288 mount | grep "$MOUNT3 .*,localflock" && error "localflock is still set"
8289 mount | grep "$MOUNT3 .*,flock" || error "flock is not set"
8290 umount_client $MOUNT3
8291 mount_client $MOUNT3 localflock,flock,noflock
8292 flock_is_enabled $MOUNT3 && error "some flock is still enabled" || true
8294 run_test 104b "Mount uses last flock argument"
8296 error_and_umount() {
8307 mount --bind $DIR $TMP/$tdir || error "mount bind mnt pt failed"
8308 rm -f $TMP/$tdir/$tfile
8309 rm -f $TMP/$tdir/${tfile}1
8311 # Files should not be created in ro bind mount point
8312 # remounting from rw to ro
8313 mount -o remount,ro $TMP/$tdir ||
8314 error_and_umount "readonly remount of bind mnt pt failed"
8315 touch $TMP/$tdir/$tfile &&
8316 error_and_umount "touch succeeds on ro bind mnt pt"
8317 [ -e $TMP/$tdir/$tfile ] &&
8318 error_and_umount "file created on ro bind mnt pt"
8320 # Files should be created in rw bind mount point
8321 # remounting from ro to rw
8322 mount -o remount,rw $TMP/$tdir ||
8323 error_and_umount "read-write remount of bind mnt pt failed"
8324 touch $TMP/$tdir/${tfile}1 ||
8325 error_and_umount "touch fails on rw bind mnt pt"
8326 [ -e $TMP/$tdir/${tfile}1 ] ||
8327 error_and_umount "file not created on rw bind mnt pt"
8328 umount $TMP/$tdir || error "umount of bind mnt pt failed"
8330 cleanup || error "cleanup failed with $?"
8332 run_test 105 "check file creation for ro and rw bind mnt pt"
8339 mkdir -p $DIR/$tdir || error "create $tdir failed"
8340 lfs setstripe -c 1 -i 0 $DIR/$tdir
8341 #define OBD_FAIL_CAT_RECORDS 0x1312
8342 do_facet mds1 $LCTL set_param fail_loc=0x1312 fail_val=$repeat
8344 for ((i = 1; i <= $repeat; i++)); do
8346 #one full plain llog
8347 createmany -o $DIR/$tdir/f- 64768
8349 createmany -u $DIR/$tdir/f- 64768
8351 wait_delete_completed $((TIMEOUT * 7))
8352 #ASSERTION osp_sync_thread() ( thread->t_flags != SVC_RUNNING ) failed
8353 #shows that osp code is buggy
8354 do_facet mds1 $LCTL set_param fail_loc=0 fail_val=0
8358 run_test 106 "check osp llog processing when catalog is wrapped"
8361 [[ "$MDS1_VERSION" -ge $(version_code 2.10.50) ]] ||
8362 skip "Need MDS version > 2.10.50"
8365 start_mgsmds || error "start_mgsmds failed"
8366 start_ost || error "unable to start OST"
8368 # add unknown configuration parameter.
8369 if [[ $PERM_CMD == *"set_param -P"* ]]; then
8370 cmd="$PERM_CMD ost.$FSNAME-OST0000*.unknown_param"
8372 cmd="$PERM_CMD $FSNAME-OST0000*.ost.unknown_param"
8374 do_facet mgs "$cmd=50"
8375 cleanup_nocli || error "cleanup_nocli failed with $?"
8378 # unknown param should be ignored while mounting.
8379 start_ost || error "unable to start OST after unknown param set"
8381 cleanup || error "cleanup failed with $?"
8383 run_test 107 "Unknown config param should not fail target mounting"
8388 $rcmd rm -rf $tmp > /dev/null 2>&1
8389 $rcmd mkdir -p $tmp/{mnt,images} || error "failed to mkdir remotely"
8391 for facet in $facets; do
8392 [ "$mds1_FSTYPE" = zfs ] &&
8393 $rcmd $ZPOOL -f export lustre-$facet > /dev/null 2>&1
8394 $rcmd mkdir $tmp/mnt/$facet ||
8395 error "failed to mkdir $tmp/mnt/$facet"
8396 $rcmd dd if=/dev/zero of=$tmp/images/$facet \
8397 seek=199 bs=1M count=1 ||
8398 error "failed to create $tmp/images/$facet"
8407 local facet=${role}$((idx + 1))
8409 [ $# -eq 5 ] && pool=$5
8411 do_facet $SINGLEMDS $MKFS --fsname=lustre --$mgs \
8412 --$role --index=$idx --replace --backfstype=$bkfs \
8413 --device-size=200000 --reformat $pool $tmp/images/$facet ||
8414 error "failed to mkfs for $facet"
8418 echo "mounting client..."
8419 mount -t lustre ${nid}:/lustre $MOUNT ||
8420 error "failed to mount lustre"
8423 ls -l $MOUNT/local_dir || error "failed to list"
8425 echo "check truncate && write"
8426 echo "dummmmmmmmmmmmm" > $MOUNT/remote_dir/fsx.c ||
8427 error "failed to tuncate & write"
8430 touch $MOUNT/foooo ||
8431 error "failed to create"
8433 echo "check read && write && append"
8434 sha1sum $MOUNT/conf-sanity.sh |
8435 awk '{ print $1 }' > $MOUNT/checksum.new ||
8436 error "failed to read(1)"
8437 sha1sum $MOUNT/remote_dir/unlinkmany.c |
8438 awk '{ print $1 }' >> $MOUNT/checksum.new ||
8439 error "failed to read(2)"
8440 sha1sum $MOUNT/striped_dir/lockahead_test.o |
8441 awk '{ print $1 }' >> $MOUNT/checksum.new ||
8442 error "failed to read(3)"
8445 diff $MOUNT/checksum.new $MOUNT/checksum.src ||
8446 error "failed to verify data"
8456 umount -f $MOUNT || error "failed to umount client"
8457 for facet in $facets; do
8458 $rcmd umount -f $tmp/mnt/$facet ||
8459 error "failed to umount $facet"
8460 if [ "$mds1_FSTYPE" = zfs ]; then
8461 $rcmd $ZPOOL export -f lustre-$facet ||
8462 error "failed to export lustre-$facet"
8466 $rcmd rm -rf $tmp || error "failed to rm the dir $tmp"
8470 [ "$CLIENTONLY" ] && skip "Client-only testing"
8471 [ "$mds1_FSTYPE" != zfs ] && skip "zfs only test"
8472 [ "$MDS1_VERSION" -lt $(version_code 2.10.58) ] &&
8473 skip "Need server version at least 2.10.58"
8478 local tmp=$TMP/$tdir
8479 local rcmd="do_facet $SINGLEMDS"
8480 local facets="mdt1 mdt2 ost1 ost2"
8481 local nid=$($rcmd $LCTL list_nids | head -1)
8484 trap t_108_cleanup EXIT ERR
8487 t_108_mkfs mdt 0 zfs mgs lustre-mdt1/mdt1
8488 t_108_mkfs mdt 1 zfs mgsnode=$nid lustre-mdt2/mdt2
8489 t_108_mkfs ost 0 zfs mgsnode=$nid lustre-ost1/ost1
8490 t_108_mkfs ost 1 zfs mgsnode=$nid lustre-ost2/ost2
8492 for facet in $facets; do
8493 $rcmd zfs set mountpoint=$tmp/mnt/$facet canmount=on \
8494 lustre-$facet/$facet ||
8495 error "failed to zfs set for $facet (1)"
8496 $rcmd zfs mount lustre-$facet/$facet ||
8497 error "failed to local mount $facet"
8498 $rcmd tar jxf $LUSTRE/tests/ldiskfs_${facet}_2_11.tar.bz2 \
8499 --xattrs --xattrs-include="trusted.*" \
8500 -C $tmp/mnt/$facet/ > /dev/null 2>&1 ||
8501 error "failed to untar image for $facet"
8502 $rcmd "cd $tmp/mnt/$facet && rm -rf oi.* OI_* lfsck_* LFSCK" ||
8503 error "failed to cleanup for $facet"
8504 $rcmd zfs umount lustre-$facet/$facet ||
8505 error "failed to local umount $facet"
8506 $rcmd zfs set canmount=off lustre-$facet/$facet ||
8507 error "failed to zfs set $facet (2)"
8510 echo "changing server nid..."
8511 $rcmd mount -t lustre -o nosvc lustre-mdt1/mdt1 $tmp/mnt/mdt1
8512 $rcmd lctl replace_nids $FSNAME-MDT0000 $nid
8513 $rcmd lctl replace_nids $FSNAME-MDT0001 $nid
8514 $rcmd lctl replace_nids $FSNAME-OST0000 $nid
8515 $rcmd lctl replace_nids $FSNAME-OST0001 $nid
8516 $rcmd umount $tmp/mnt/mdt1
8518 for facet in $facets; do
8519 echo "mounting $facet from backup..."
8520 $rcmd mount -t lustre -o abort_recov lustre-$facet/$facet \
8521 $tmp/mnt/$facet || error "failed to mount $facet"
8524 # ZFS backend can detect migration and trigger OI scrub automatically
8525 # sleep 3 seconds for scrub done
8531 run_test 108a "migrate from ldiskfs to ZFS"
8534 [ "$CLIENTONLY" ] && skip "Client-only testing"
8535 [ "$mds1_FSTYPE" != ldiskfs ] && skip "ldiskfs only test"
8536 [ "$MDS1_VERSION" -lt $(version_code 2.10.58) ] &&
8537 skip "Need server version at least 2.10.58"
8542 local tmp=$TMP/$tdir
8543 local rcmd="do_facet $SINGLEMDS"
8544 local facets="mdt1 mdt2 ost1 ost2"
8545 local scrub_list="MDT0000 MDT0001 OST0000 OST0001"
8546 local nid=$($rcmd $LCTL list_nids | head -1)
8549 trap t_108_cleanup EXIT ERR
8552 t_108_mkfs mdt 0 ldiskfs mgs
8553 t_108_mkfs mdt 1 ldiskfs mgsnode=$nid
8554 t_108_mkfs ost 0 ldiskfs mgsnode=$nid
8555 t_108_mkfs ost 1 ldiskfs mgsnode=$nid
8557 for facet in $facets; do
8558 $rcmd mount -t ldiskfs -o loop $tmp/images/$facet \
8560 error "failed to local mount $facet"
8562 $rcmd tar jxf $LUSTRE/tests/zfs_${facet}_2_11.tar.bz2 \
8563 --xattrs --xattrs-include="*.*" \
8564 -C $tmp/mnt/$facet/ > /dev/null 2>&1 ||
8565 error "failed to untar image for $facet"
8566 $rcmd "cd $tmp/mnt/$facet && rm -rf oi.* OI_* lfsck_* LFSCK" ||
8567 error "failed to cleanup for $facet"
8568 $rcmd umount $tmp/mnt/$facet ||
8569 error "failed to local umount $facet"
8572 echo "changing server nid..."
8573 $rcmd mount -t lustre -o nosvc,loop $tmp/images/mdt1 $tmp/mnt/mdt1
8574 $rcmd lctl replace_nids $FSNAME-MDT0000 $nid
8575 $rcmd lctl replace_nids $FSNAME-MDT0001 $nid
8576 $rcmd lctl replace_nids $FSNAME-OST0000 $nid
8577 $rcmd lctl replace_nids $FSNAME-OST0001 $nid
8578 $rcmd umount $tmp/mnt/mdt1
8580 for facet in $facets; do
8581 echo "mounting $facet from backup..."
8582 $rcmd mount -t lustre -o loop,abort_recov $tmp/images/$facet \
8583 $tmp/mnt/$facet || error "failed to mount $facet"
8586 for facet in $scrub_list; do
8587 $rcmd $LCTL lfsck_start -M $FSNAME-$facet -t scrub ||
8588 error "failed to start OI scrub on $facet"
8591 # sleep 3 seconds for scrub done
8597 run_test 108b "migrate from ZFS to ldiskfs"
8600 # set number of permanent parameters
8602 test_109_set_params() {
8605 set_persistent_param_and_check mds \
8606 "mdd.$fsname-MDT0000.atime_diff" \
8607 "$fsname-MDT0000.mdd.atime_diff" \
8609 set_persistent_param_and_check mds \
8610 "mdd.$fsname-MDT0000.atime_diff" \
8611 "$fsname-MDT0000.mdd.atime_diff" \
8613 set_persistent_param_and_check client \
8614 "llite.$fsname*.max_read_ahead_mb" \
8615 "$fsname.llite.max_read_ahead_mb" \
8617 set_persistent_param_and_check client \
8618 "llite.$fsname*.max_read_ahead_mb" \
8619 "$fsname.llite.max_read_ahead_mb" \
8621 create_pool $fsname.pool1 || error "create pool failed"
8622 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
8623 error "pool_add failed"
8624 do_facet mgs $LCTL pool_remove $fsname.pool1 OST0000 ||
8625 error "pool_remove failed"
8626 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
8627 error "pool_add failed"
8631 # check permanent parameters
8633 test_109_test_params() {
8636 local atime_diff=$(do_facet mds $LCTL \
8637 get_param -n mdd.$fsname-MDT0000.atime_diff)
8638 [ $atime_diff == 63 ] || error "wrong mdd parameter after clear_conf"
8639 local max_read_ahead_mb=$(do_facet client $LCTL \
8640 get_param -n llite.$fsname*.max_read_ahead_mb)
8641 [ $max_read_ahead_mb == 64 ] ||
8642 error "wrong llite parameter after clear_conf"
8643 local ost_in_pool=$(do_facet mds $LCTL pool_list $fsname.pool1 |
8644 grep -v "^Pool:" | sed 's/_UUID//')
8645 [ $ost_in_pool = "$fsname-OST0000" ] ||
8646 error "wrong pool after clear_conf"
8650 # run lctl clear_conf, store CONFIGS before and after that
8652 test_109_clear_conf()
8654 local clear_conf_arg=$1
8657 if ! combined_mgs_mds ; then
8658 mgsdev=$(mgsdevname)
8659 stop_mgs || error "stop_mgs failed"
8660 start_mgs "-o nosvc" || error "start_mgs nosvc failed"
8662 mgsdev=$(mdsdevname 1)
8663 start_mdt 1 "-o nosvc" || error "start_mdt 1 nosvc failed"
8666 do_facet mgs "rm -rf $TMP/${tdir}/conf1; mkdir -p $TMP/${tdir}/conf1;" \
8667 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf1\\\" \
8671 # the command being tested
8673 do_facet mgs $LCTL clear_conf $clear_conf_arg ||
8674 error "clear_conf failed"
8675 if ! combined_mgs_mds ; then
8676 stop_mgs || error "stop_mgs failed"
8678 stop_mdt 1 || error "stop_mdt 1 failed"
8681 do_facet mgs "rm -rf $TMP/${tdir}/conf2; mkdir -p $TMP/${tdir}/conf2;" \
8682 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf2\\\" \
8686 test_109_file_shortened() {
8688 local sizes=($(do_facet mgs "stat -c %s " \
8689 "$TMP/${tdir}/conf1/CONFIGS/$file" \
8690 "$TMP/${tdir}/conf2/CONFIGS/$file"))
8691 [ ${sizes[1]} -lt ${sizes[0]} ] && return 0
8697 [ $MDS1_VERSION -lt $(version_code 2.10.59) ] &&
8698 skip "Needs MDS version 2.10.59 or later."
8699 [ "$(facet_fstype mgs)" == zfs ] &&
8700 skip "LU-8727: no implementation for ZFS"
8705 client_up || error "client_up failed"
8708 # set number of permanent parameters
8710 test_109_set_params $FSNAME
8712 umount_client $MOUNT || error "umount_client failed"
8713 stop_ost || error "stop_ost failed"
8714 stop_mds || error "stop_mds failed"
8716 test_109_clear_conf $FSNAME
8718 # make sure that all configs are cleared
8720 test_109_file_shortened $FSNAME-MDT0000 ||
8721 error "failed to clear MDT0000 config"
8722 test_109_file_shortened $FSNAME-client ||
8723 error "failed to clear client config"
8728 # check that configurations are intact
8730 test_109_test_params $FSNAME
8735 destroy_test_pools || error "destroy test pools failed"
8739 run_test 109a "test lctl clear_conf fsname"
8743 [ $MDS1_VERSION -lt $(version_code 2.10.59) ] &&
8744 skip "Needs MDS version 2.10.59 or later."
8745 [ "$(facet_fstype mgs)" == zfs ] &&
8746 skip "LU-8727: no implementation for ZFS"
8751 client_up || error "client_up failed"
8754 # set number of permanent parameters
8756 test_109_set_params $FSNAME
8758 umount_client $MOUNT || error "umount_client failed"
8759 stop_ost || error "stop_ost failed"
8760 stop_mds || error "stop_mds failed"
8762 test_109_clear_conf $FSNAME-MDT0000
8764 # make sure that only one config is cleared
8766 test_109_file_shortened $FSNAME-MDT0000 ||
8767 error "failed to clear MDT0000 config"
8768 test_109_file_shortened $FSNAME-client &&
8769 error "failed to clear client config"
8773 # check that configurations are intact
8775 test_109_test_params $FSNAME
8780 destroy_test_pools || error "destroy test pools failed"
8784 run_test 109b "test lctl clear_conf one config"
8788 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
8789 skip "Only applicable to ldiskfs-based MDTs"
8791 do_facet $SINGLEMDS $DEBUGFS -w -R supported_features |grep large_dir ||
8792 skip "large_dir option is not supported on MDS"
8793 do_facet ost1 $DEBUGFS -w -R supported_features | grep large_dir ||
8794 skip "large_dir option is not supported on OSS"
8796 stopall # stop all targets before modifying the target counts
8797 local old_mdscount=$MDSCOUNT
8798 local old_ostcount=$OSTCOUNT
8800 stack_trap "MDSCOUNT=$old_mdscount OSTCOUNT=$old_ostcount; reformat"
8804 # ext4_dir_entry_2 struct size:264
8805 # dx_root struct size:8
8806 # dx_node struct size:8
8807 # dx_entry struct size:8
8808 # For 1024 bytes block size.
8809 # First level directory entries: 126
8810 # Second level directory entries: 127
8811 # Entries in leaf: 3
8812 # For 2 levels limit: 48006
8813 # For 3 levels limit : 6096762
8814 # Create 80000 files to safely exceed 2-level htree limit.
8815 CONF_SANITY_110_LINKS=${CONF_SANITY_110_LINKS:-80000}
8817 # can fit at most 3 filenames per 1KB leaf block, but each
8818 # leaf/index block will only be 3/4 full before split at each level
8819 (( MDSSIZE < CONF_SANITY_110_LINKS / 3 * 4/3 * 4/3 )) &&
8820 CONF_SANITY_110_LINKS=$((MDSSIZE * 3 * 3/4 * 3/4))
8822 combined_mgs_mds || replace=" --replace "
8823 local opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
8824 $replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
8825 if [[ $opts != *large_dir* ]]; then
8826 if [[ $opts != *mkfsoptions* ]]; then
8827 opts+=" --mkfsoptions=\\\"-O large_dir -b 1024 -i 65536\\\""
8829 opts="${opts//--mkfsoptions=\\\"/ \
8830 --mkfsoptions=\\\"-O large_dir -b 1024 -i 65536 }"
8833 echo "MDT params: $opts"
8835 combined_mgs_mds || start_mgs
8836 add mds1 $opts || error "add mds1 failed with new params"
8837 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
8839 opts="$(mkfs_opts ost1 $(ostdevname 1)) \
8840 $replace --reformat $(ostdevname 1) $(ostvdevname 1)"
8842 if [[ $opts != *large_dir* ]]; then
8843 if [[ $opts != *mkfsoptions* ]]; then
8844 opts+=" --mkfsoptions=\\\"-O large_dir\\\" "
8846 opts="${opts//--mkfsoptions=\\\"/ \
8847 --mkfsoptions=\\\"-O large_dir }"
8850 echo "OST params: $opts"
8851 add ost1 $opts || error "add ost1 failed with new params"
8852 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
8854 MOUNT_2=yes mountcli || error "mount clients failed"
8856 mkdir -v $DIR/$tdir || error "cannot create $DIR/$tdir"
8857 local pids count=0 group=0
8859 echo "creating $CONF_SANITY_110_LINKS in total"
8860 while (( count < CONF_SANITY_110_LINKS )); do
8861 local len=$((253 - $(wc -c <<<"$tfile-$group-40000-")))
8862 local dir=DIR$((group % 2 + 1))
8863 local target=${!dir}/$tdir/$tfile-$group
8864 local long=$target-$(generate_name $len)-
8865 local create=$((CONF_SANITY_110_LINKS - count))
8867 (( create > 40000 )) && create=40000
8868 touch $target || error "creating $target failed"
8869 echo "creating $create hard links to $target"
8870 createmany -l $target $long $create &
8873 count=$((count + create))
8874 group=$((group + 1))
8876 echo "waiting for PIDs$pids to complete"
8877 wait $pids || error "createmany failed after $group groups"
8879 umount_client $MOUNT2 -f
8882 run_e2fsck $(facet_active_host mds1) $(mdsdevname 1) -n
8883 MDSCOUNT=$old_mdscount
8884 OSTCOUNT=$old_ostcount
8886 run_test 110 "Adding large_dir with 3-level htree"
8889 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
8890 skip "Only applicable to ldiskfs-based MDTs"
8892 is_dm_flakey_dev $SINGLEMDS $(mdsdevname 1) &&
8893 skip "This test can not be executed on flakey dev"
8895 do_facet $SINGLEMDS $DEBUGFS -w -R supported_features |grep large_dir ||
8896 skip "large_dir option is not supported on MDS"
8898 do_facet ost1 $DEBUGFS -w -R supported_features | grep large_dir ||
8899 skip "large_dir option is not supported on OSS"
8901 # cleanup before changing target counts
8903 local old_mdscount=$MDSCOUNT
8904 local old_ostcount=$OSTCOUNT
8905 local old_mdssize=$MDSSIZE
8907 stack_trap "MDSSIZE=$old_mdssize MDSCOUNT=$old_mdscount \
8908 OSTCOUNT=$old_ostcount; reformat"
8911 (( MDSSIZE < 2400000 )) && MDSSIZE=2400000 # need at least 2.4GB
8913 local mdsdev=$(mdsdevname 1)
8914 combined_mgs_mds || replace=" --replace "
8915 local opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
8916 $replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
8917 if [[ $opts != *large_dir* ]]; then
8918 if [[ $opts != *mkfsoptions* ]]; then
8919 opts+=" --mkfsoptions=\\\"-O large_dir -i 1048576 \\\" "
8921 opts="${opts//--mkfsoptions=\\\"/ \
8922 --mkfsoptions=\\\"-O large_dir -i 1048576 }"
8925 echo "MDT params: $opts"
8927 combined_mgs_mds || start_mgs
8928 __touch_device mds 1
8929 add mds1 $opts || error "add mds1 failed with new params"
8930 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
8932 opts="$(mkfs_opts ost1 $(ostdevname 1)) \
8933 $replace --reformat $(ostdevname 1) $(ostvdevname 1)"
8934 if [[ $opts != *large_dir* ]]; then
8935 if [[ $opts != *mkfsoptions* ]]; then
8936 opts+=" --mkfsoptions=\\\"-O large_dir \\\""
8938 opts="${opts//--mkfsoptions=\\\"/ --mkfsoptions=\\\"-O large_dir }"
8941 echo "OST params: $opts"
8942 __touch_device ost 1
8943 add ost1 $opts || error "add ost1 failed with new params"
8944 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
8946 MOUNT_2=yes mountcli
8947 mkdir $DIR/$tdir || error "cannot create $DIR/$tdir"
8949 lfs df -i $DIR/$tdir
8953 local start=$SECONDS
8955 local dirmax=$((2 << 30))
8961 while (( !needskip & dirsize < dirmax )); do
8964 for cli in ${CLIENTS//,/ }; do
8965 local len=$((253 - $(wc -c <<<"$cli-$group-60000-")))
8966 local target=$cli-$group
8967 local long=$DIR/$tdir/$target-$(generate_name $len)-
8969 RPWD=$DIR/$tdir do_node $cli touch $target ||
8970 error "creating $target failed"
8971 echo "creating 60000 hardlinks to $target"
8972 RPWD=$DIR/$tdir do_node $cli createmany -l $target $long 60000 &
8975 group=$((group + 1))
8977 long=$DIR2/$tdir/$target-$(generate_name $len)-
8979 RPWD=$DIR2/$tdir do_node $cli touch $target ||
8980 error "creating $target failed"
8981 echo "creating 60000 hardlinks to $target"
8982 RPWD=$DIR2/$tdir do_node $cli createmany -l $target $long 60000 &
8985 group=$((group + 1))
8987 echo "waiting for PIDs$pids to complete"
8988 wait $pids || error "createmany failed after $group groups"
8989 dirsize=$(stat -c %s $DIR/$tdir)
8990 taken=$((SECONDS - start))
8991 rate=$((dirsize / taken))
8992 left=$(((dirmax - dirsize) / rate))
8993 num=$((group * 60000))
8994 echo "estimate ${left}s left after $num files / ${taken}s"
8995 # if the estimated time remaining is too large (it may change
8996 # over time as the create rate is not constant) then exit
8997 # without declaring a failure.
8998 (( left > 1200 )) && needskip=1
9001 umount_client $MOUNT2 -f
9004 ! (( $needskip )) ||
9005 echo "ETA ${left}s after $num files / ${taken}s is too long"
9007 run_e2fsck $(facet_active_host mds1) $(mdsdevname 1) -n
9008 MDSCOUNT=$old_mdscount
9009 OSTCOUNT=$old_ostcount
9010 MDSSIZE=$old_mdssize
9012 run_test 111 "Adding large_dir with over 2GB directory"
9015 start_mds || error "MDS start failed"
9016 start_ost || error "OSS start failed"
9017 echo "start ost2 service on $(facet_active_host ost2)"
9018 start ost2 $(ostdevname 2) $(csa_add "$OST_MOUNT_OPTS" -o no_precreate) ||
9019 error "start ost2 facet failed"
9020 local val=$(do_facet ost2 \
9021 "$LCTL get_param -n obdfilter.$FSNAME-OST0001*.no_precreate")
9022 (( $val == 1 )) || error "obdfilter.$FSNAME-OST0001*.no_precreate=$val"
9024 mount_client $MOUNT || error "mount client failed"
9025 wait_osc_import_state mds1 ost1 FULL
9026 wait_osc_import_state client ost1 FULL
9027 wait_osc_import_state client ost2 FULL
9029 $LFS setstripe -i 0 $DIR/$tfile.0 ||
9030 error "problem creating $tfile.0 on OST0000"
9031 $LFS setstripe -i 1 $DIR/$tfile.1 && $LFS getstripe $DIR/$tfile.1 &&
9032 (( $($LFS getstripe -i $DIR/$tfile.1) == 1 )) &&
9033 error "allowed to create $tfile.1 on OST0001"
9034 do_facet ost2 $LCTL set_param obdfilter.*.no_precreate=0
9036 $LFS setstripe -i 1 $DIR/$tfile.2 ||
9037 error "failed to create $tfile.2 on ost1 facet"
9038 # files not cleaned with ONLY_REPEAT because of client unmount below
9039 rm -f $DIR/$tfile.[012]
9040 stop_ost2 || error "stop ost2 facet failed"
9043 run_test 112 "mount OST with nocreate option"
9046 SAVE_MGS_MOUNT_OPTS=$MGS_MOUNT_OPTS
9047 SAVE_MDS_MOUNT_OPTS=$MDS_MOUNT_OPTS
9048 SAVE_OST_MOUNT_OPTS=$OST_MOUNT_OPTS
9054 MGS_MOUNT_OPTS=$SAVE_MGS_MOUNT_OPTS
9055 MDS_MOUNT_OPTS=$SAVE_MDS_MOUNT_OPTS
9056 OST_MOUNT_OPTS=$SAVE_OST_MOUNT_OPTS
9057 # Revert old mount options back
9059 # Subsequent following test requires
9060 # conf-sanity to be in stopall state.
9061 # Force 'stopall' so others following
9066 # Error out with mount info
9068 local server_nodes=$(comma_list $(mdts_nodes) $(osts_nodes))
9071 echo "--Client Mount Info--"
9072 mount | grep -i lustre
9073 echo "--Server Mount Info--"
9074 do_nodes $server_nodes mount | grep -i lustre
9080 local ost_version="2.15.51" # Minimum version required
9082 (( OST1_VERSION >= $(version_code $ost_version) )) ||
9083 skip "Need server version at least $ost_version"
9085 stack_trap cleanup_113 EXIT
9087 # Reset before starting
9091 # Verify MDS's should start with "rw"
9092 do_facet $SINGLEMDS mount | grep "lustre.*rw,.*MDT" ||
9093 error_113 "$SINGLEMDS should be read-write"
9095 # Verify OST's should start with "rw"
9096 for (( i=1; i <= OSTCOUNT; i++ )); do
9097 do_facet ost$i mount | grep "lustre.*rw,.*OST" ||
9098 error_113 "ost$i should be read-write"
9101 # rdonly_dev does not currently work for ldiskfs
9102 # We skip the rdonly_dev check until then.
9103 if [[ $ost1_FSTYPE == ldiskfs ]]; then
9104 echo "Shadow Mountpoint correctly reports rw for ldiskfs"
9109 # Only ZFS specific tests below.
9112 # Must stop all (server+client) and restart to verify new
9116 # add rdonly_dev to mount option
9117 MGS_MOUNT_OPTS=$(csa_add "$MGS_MOUNT_OPTS" -o rdonly_dev)
9118 MDS_MOUNT_OPTS=$(csa_add "$MDS_MOUNT_OPTS" -o rdonly_dev)
9119 OST_MOUNT_OPTS=$(csa_add "$OST_MOUNT_OPTS" -o rdonly_dev)
9121 # Only restart server(mds/ost). Sufficient for test
9122 setupall server_only || error "Fail to start servers"
9124 # Verify MDS's should be "ro"
9125 do_facet $SINGLEMDS mount | grep "lustre.*ro,.*MDT.*rdonly_dev" ||
9126 error_113 "$SINGLEMDS should be read-only"
9128 # Verify OST's should be "ro"
9129 for (( i=1; i <= OSTCOUNT; i++ )); do
9130 do_facet ost$i mount | grep "lustre.*ro,.*OST.*rdonly_dev" ||
9131 error_113 "ost$i should be read-only"
9134 run_test 113 "Shadow mountpoint correctly report ro/rw for mounts"
9137 # Purpose: To verify dynamic thread (OSS) creation.
9138 # (This was sanity/115)
9141 [ $PARALLEL == "yes" ] && skip "skip parallel run"
9142 remote_ost_nodsh && skip "remote OST with nodsh"
9144 # Lustre does not stop service threads once they are started.
9145 # Reset number of running threads to default.
9150 local save_params="$TMP/sanity-$TESTNAME.parameters"
9152 # Get ll_ost_io count before I/O
9153 OSTIO_pre=$(do_facet ost1 \
9154 "$LCTL get_param ost.OSS.ost_io.threads_started | cut -d= -f2")
9155 # Exit if lustre is not running (ll_ost_io not running).
9156 [ -z "$OSTIO_pre" ] && error "no OSS threads"
9158 echo "Starting with $OSTIO_pre threads"
9159 local thread_max=$((OSTIO_pre * 2))
9160 local rpc_in_flight=$((thread_max * 2))
9162 # this is limited to OSC_MAX_RIF_MAX (256)
9163 [ $rpc_in_flight -gt 256 ] && rpc_in_flight=256
9164 thread_max=$((rpc_in_flight / 2))
9165 [ $thread_max -le $OSTIO_pre ] && skip "Too many ost_io threads" &&
9168 # Number of I/O Process proposed to be started.
9170 local facets=$(get_facets OST)
9172 save_lustre_params client "osc.*OST*.max_rpcs_in_flight" > $save_params
9173 save_lustre_params $facets "ost.OSS.ost_io.threads_max" >> $save_params
9175 # Set in_flight to $rpc_in_flight
9176 $LCTL set_param osc.*OST*.max_rpcs_in_flight=$rpc_in_flight ||
9177 error "Failed to set max_rpcs_in_flight to $rpc_in_flight"
9178 nfiles=${rpc_in_flight}
9179 # Set ost thread_max to $thread_max
9180 do_facet ost1 "$LCTL set_param ost.OSS.ost_io.threads_max=$thread_max"
9182 # 5 Minutes should be sufficient for max number of OSS
9183 # threads(thread_max) to be created.
9187 local wtl=${WTL:-"$LUSTRE/tests/write_time_limit"}
9189 test_mkdir $DIR/$tdir
9190 for ((i = 1; i <= nfiles; i++)); do
9191 local file=$DIR/$tdir/${tfile}-$i
9193 $LFS setstripe -c -1 -i 0 $file
9194 ($wtl $file $timeout)&
9197 # I/O Started - Wait for thread_started to reach thread_max or report
9198 # error if thread_started is more than thread_max.
9199 echo "Waiting for thread_started to reach thread_max"
9200 local thread_started=0
9201 local end_time=$((SECONDS + timeout))
9203 while [ $SECONDS -le $end_time ] ; do
9205 # Get ost i/o thread_started count.
9206 thread_started=$(do_facet ost1 \
9208 ost.OSS.ost_io.threads_started | cut -d= -f2")
9209 # Break out if thread_started is equal/greater than thread_max
9210 if (( $thread_started >= $thread_max )); then
9211 echo ll_ost_io thread_started $thread_started, \
9212 equal/greater than thread_max $thread_max
9218 # Cleanup - We have the numbers, Kill i/o jobs if running.
9219 jobcount=($(jobs -p))
9221 for ((i=0; i < ${#jobcount[*]}; i++)); do
9222 kill -9 ${jobcount[$i]}
9223 if [ $? -ne 0 ] ; then
9224 echo "warning: cannot kill WTL pid ${jobcount[$i]}"
9228 # Cleanup files left by WTL binary.
9229 for ((i = 1; i <= nfiles; i++)); do
9230 local file=$DIR/$tdir/${tfile}-$i
9233 if [ $? -ne 0 ] ; then
9234 echo "Warning: Failed to delete file $file"
9238 restore_lustre_params <$save_params
9239 rm -f $save_params || echo "Warning: delete file '$save_params' failed"
9241 # Error out if no new thread has started or Thread started is greater
9243 if (( $thread_started <= $OSTIO_pre ||
9244 $thread_started > $thread_max )); then
9245 error "ll_ost_io: thread_started $thread_started" \
9246 "OSTIO_pre $OSTIO_pre, thread_max $thread_max." \
9247 "No new thread started or thread started greater " \
9251 run_test 114 "verify dynamic thread creation===================="
9257 do_facet mds1 rm -f $1
9261 if [ "$mds1_FSTYPE" != ldiskfs ]; then
9262 skip "Only applicable to ldiskfs-based MDTs"
9264 [ -n "$FILESET" ] && skip "Not functional for FILESET set"
9266 local dbfs_ver=$(do_facet mds1 $DEBUGFS -V 2>&1)
9268 echo "debugfs version: $dbfs_ver"
9269 echo "$dbfs_ver" | egrep -w "1.44.3.wc1|1.44.5.wc1|1.45.2.wc1" &&
9270 skip_env "This version of debugfs doesn't show inode number"
9272 local IMAGESIZE=$((3072 << 30)) # 3072 GiB
9282 do_facet mds1 "hostname; $LCTL dl; mount"
9284 do_facet ost1 "hostname; $LCTL dl; mount"
9285 # We need MDT size 3072GB, because it is smallest
9286 # partition that can store 2B inodes
9287 do_facet mds1 "mkdir -p $TMP/$tdir"
9288 local mdsimgname=$TMP/$tdir/lustre-mdt
9290 do_facet mds1 "rm -f $mdsimgname"
9291 do_facet mds1 "touch $mdsimgname"
9292 stack_trap "cleanup_115 $mdsimgname" EXIT
9293 do_facet mds1 "$TRUNCATE $mdsimgname $IMAGESIZE" ||
9294 skip "Backend FS doesn't support sparse files"
9295 local mdsdev=$(do_facet mds1 "losetup -f")
9297 do_facet mds1 "losetup $mdsdev $mdsimgname"
9299 local mds_opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
9300 --mkfsoptions='-O ea_inode,^resize_inode,meta_bg \
9301 -N 2247484000 -E lazy_itable_init' --device-size=$IMAGESIZE"
9302 add mds1 $mds_opts --mgs --reformat $mdsdev ||
9303 skip_env "format large MDT failed"
9305 local ostdev=$(ostdevname 1)
9307 local opts="$(mkfs_opts ost1 $ostdev) \
9308 --reformat $ostdev $ostdev"
9309 add ost1 $opts || error "add ost1 failed with new params"
9310 start mds1 $mdsdev $MDS_MOUNT_OPTS || error "start MDS failed"
9311 start_ost || error "start OSS failed"
9312 mount_client $MOUNT || error "mount client failed"
9314 mkdir -p $DIR/$tdir || error "mkdir $DIR/$tdir fail"
9315 local goal="/sys/fs/ldiskfs/$(basename $mdsdev)/inode_goal"
9317 # 2147483648 is 0x80000000
9318 do_facet mds1 "echo 2147483648 >> $goal; grep . $goal"
9319 touch $DIR/$tdir/$tfile
9321 # attrs from 1 to 16 go to block, 17th - to inode
9324 for i in {1..17}; do
9325 local nm="trusted.ea$i"
9326 setfattr -n $nm -v $(printf "xattr%0250d" $i) $DIR/$tdir/$tfile
9329 do_facet mds1 "$DEBUGFS -c -R 'stat ROOT/$tdir/$tfile' $mdsdev"
9331 # inode <2147483649> trusted.ea16 (255)
9332 local inode_num=$(do_facet mds1 \
9333 "$DEBUGFS -c -R 'stat ROOT/$tdir/$tfile' $mdsdev" |
9334 awk '/ea17/ { print $2 }' |
9335 sed -e 's/>//' -e 's/<//' -e 's/\"//')
9336 echo "inode num: $inode_num"
9337 [ $inode_num -ge 2147483648 ] || error "inode $inode_num too small"
9338 do_facet mds1 "losetup -d $mdsdev"
9340 run_test 115 "Access large xattr with inodes number over 2TB"
9343 [ "$mds1_FSTYPE" != ldiskfs ] && skip "ldiskfs only test"
9344 [ "$MDS1_VERSION" -lt $(version_code 2.10.59) ] &&
9345 skip "Need server version at least 2.10.59"
9346 do_facet $SINGLEMDS which mkfs.xfs ||
9347 skip_env "No mkfs.xfs installed"
9352 local tmpmnt=$TMP/$tdir
9353 local mdtimg=$tfile-mdt0
9355 do_facet $SINGLEMDS mkdir -p $tmpmnt
9356 stack_trap "do_facet $SINGLEMDS rmdir $tmpmnt" EXIT
9358 do_facet $SINGLEMDS touch $TMP/$mdtimg
9359 stack_trap "do_facet $SINGLEMDS rm -f $TMP/$mdtimg" EXIT
9360 do_facet $SINGLEMDS mkfs -t xfs -d file,size=1t,name=$TMP/$mdtimg ||
9361 error "mkfs temporary xfs image"
9363 do_facet $SINGLEMDS mount $TMP/$mdtimg $tmpmnt ||
9364 error "mount temporary xfs image"
9365 stack_trap "do_facet $SINGLEMDS umount $tmpmnt" EXIT
9366 local old_mdssize=$MDSSIZE
9367 local old_mdsisize=$MDSISIZE
9369 MDSSIZE=$((17 * 1024 * 1024 * 1024)) # 17T MDT
9370 MDSISIZE=$((16 << 20))
9371 local opts17t="$(mkfs_opts $SINGLEMDS)"
9373 MDSSIZE=$old_mdssize
9374 MDSISIZE=$old_mdsisize
9376 do_facet $SINGLEMDS $MKFS \
9377 ${opts17t/-E /-E lazy_itable_init,lazy_journal_init,} \
9379 error "failed to mkfs for $tmpmnt/$mdtimg"
9381 do_facet $SINGLEMDS $TUNE2FS -l $tmpmnt/$mdtimg |
9382 grep -qw 'features.*extent' || error "extent should be enabled"
9385 run_test 116 "big size MDT support"
9388 # Call setup only if LustreFS is not mounted
9389 check_mount || setup
9391 do_facet ost1 "$LCTL set_param ost.OSS.ost_io.nrs_policies=fifo"
9392 do_facet ost1 "$LCTL get_param -n ost.OSS.ost_io.nrs_tbf_rule" &&
9393 error "get_param should fail"
9394 cleanup || error "cleanup failed with rc $?"
9396 run_test 117 "lctl get_param return errors properly"
9402 [[ "$MDSCOUNT" -ge 2 ]] || skip "Need more at least 2 MDTs"
9404 had_config=$(do_facet mds1 "$LCTL get_param debug | grep config")
9405 do_facet mds1 "$LCTL set_param debug=+config"
9406 do_facet mds1 "$LCTL clear"
9409 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)" &>/dev/null
9410 # mount after writeconf will make "add osp" added to mdt0 config:
9411 # 53 (224)marker 60 (flags=0x01, v2.5.1.0) lustre-MDT0001 'add osp'
9412 # 54 (080)add_uuid nid=... 0: 1:...
9413 # 55 (144)attach 0:lustre-MDT0001-osp-MDT0000 1:osp 2:...
9414 # 56 (144)setup 0:lustre-MDT0001-osp-MDT0000 1:... 2:...
9415 # 57 (136)modify_mdc_tgts add 0:lustre-MDT0000-mdtlov 1:... 2:1 3:1
9416 # duplicate modify_mdc_tgts caused crashes
9419 # using larger debug_mb size to avoid lctl dk log truncation
9420 size_mb=$((DEBUG_SIZE_SAVED * 4))
9423 # though config processing stops after failed attach and setup
9424 # it will proceed after the failed command after each writeconf
9425 # this is the original scenario of the issue
9426 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)" &>/dev/null
9427 do_facet mds1 "$LCTL set_param debug_mb=$size_mb"
9430 wait_update_facet_cond mds1 \
9431 "$LCTL dk | grep -c Processed.log.$FSNAME-MDT0000" \
9436 [[ -z "$had_config" ]] && do_facet mds1 lctl set_param debug=-config
9440 run_test 119 "writeconf on slave mdt shouldn't duplicate mdc/osp and crash"
9442 test_120() { # LU-11130
9443 [ "$MDSCOUNT" -lt 2 ] && skip "mdt count < 2"
9444 [ "$mds1_FSTYPE" != ldiskfs ] &&
9445 skip "ldiskfs only test"
9446 [ "$MDS1_VERSION" -lt $(version_code 2.11.56) ] &&
9447 skip "Need DNE2 capable MD target with LU-11130 fix"
9451 local mds1host=$(facet_active_host mds1)
9452 local mds1dev=$(mdsdevname 1)
9454 $LFS mkdir -i 1 $DIR/$tdir
9455 $LFS mkdir -i 0 $DIR/$tdir/mds1dir
9457 ln -s foo $DIR/$tdir/bar
9458 mv $DIR/$tdir/bar $DIR/$tdir/mds1dir/bar2 ||
9459 error "cross-target rename failed"
9463 run_e2fsck $mds1host $mds1dev "-n"
9465 run_test 120 "cross-target rename should not create bad symlinks"
9469 start_mgsmds || error "MGS MDS Start failed"
9471 stop_mds || error "Stopping MDSes failed"
9475 stop_mds || error "Stopping MDSes failed"
9477 run_test 121 "failover MGS"
9480 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
9481 [[ "$OST1_VERSION" -ge $(version_code 2.11.53) ]] ||
9482 skip "Need OST version at least 2.11.53"
9485 LOAD_MODULES_REMOTE=true load_modules
9486 #define OBD_FAIL_OFD_SET_OID 0x1e0
9487 do_facet ost1 $LCTL set_param fail_loc=0x00001e0
9490 $LFS mkdir -i1 -c1 $DIR/$tdir
9491 $LFS setstripe -i0 -c1 $DIR/$tdir
9492 do_facet ost1 $LCTL set_param fail_loc=0
9493 createmany -o $DIR/$tdir/file_ 1000 ||
9494 error "Fail to create a new sequence"
9498 run_test 122a "Check OST sequence update"
9501 remote_mgs_nodsh && skip "remote MGS with nodsh"
9502 [ -d $MOUNT/.lustre ] || setup
9504 # test old logid format until removal from llog_ioctl.c::str2logid()
9505 if [ $MGS_VERSION -lt $(version_code 3.1.53) ]; then
9506 do_facet mgs $LCTL dl | grep MGS
9507 do_facet mgs "$LCTL --device %MGS llog_print \
9508 \\\\\\\$$FSNAME-client 1 10" ||
9509 error "old llog_print failed"
9512 # test new logid format
9513 if [ $MGS_VERSION -ge $(version_code 2.9.53) ]; then
9514 do_facet mgs "$LCTL --device MGS llog_print $FSNAME-client" ||
9515 error "new llog_print failed"
9518 run_test 123aa "llog_print works with FIDs and simple names"
9521 remote_mgs_nodsh && skip "remote MGS with nodsh"
9522 [[ $MGS_VERSION -gt $(version_code 2.11.51) ]] ||
9523 skip "Need server with working llog_print support"
9525 [ -d $MOUNT/.lustre ] || setup
9531 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9532 mgs_arg="--device MGS"
9534 orig_val=$(do_facet mgs $LCTL get_param jobid_name)
9535 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME"
9537 yaml=$(do_facet mgs $LCTL $mgs_arg llog_print params |
9538 grep jobid_name | tail -n 1)
9540 local param=$(awk '{ print $10 }' <<< "$yaml")
9541 local val=$(awk '{ print $12 }' <<< "$yaml")
9542 #return to the default
9543 do_facet mgs $LCTL set_param -P jobid_name=$orig_val
9544 [ $val = "TESTNAME" ] || error "bad value: $val"
9545 [ $param = "jobid_name," ] || error "Bad param: $param"
9547 run_test 123ab "llog_print params output values from set_param -P"
9549 test_123ac() { # LU-11566
9550 remote_mgs_nodsh && skip "remote MGS with nodsh"
9551 do_facet mgs "$LCTL help llog_print" 2>&1 | grep -q -- --start ||
9552 skip "Need 'lctl llog_print --start' on MGS"
9558 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9559 mgs_arg="--device MGS"
9561 [ -d $MOUNT/.lustre ] || setup
9563 # - { index: 10, event: add_uuid, nid: 192.168.20.1@tcp(0x20000c0a81401,
9564 # node: 192.168.20.1@tcp }
9565 do_facet mgs $LCTL $mgs_arg \
9566 llog_print --start $start --end $end $FSNAME-client | tr -d , |
9567 while read DASH BRACE INDEX idx EVENT BLAH BLAH BLAH; do
9568 (( idx >= start )) || error "llog_print index $idx < $start"
9569 (( idx <= end )) || error "llog_print index $idx > $end"
9572 run_test 123ac "llog_print with --start and --end"
9574 test_123ad() { # LU-11566
9575 remote_mgs_nodsh && skip "remote MGS with nodsh"
9576 # older versions of lctl may not print all records properly
9577 do_facet mgs "$LCTL help llog_print" 2>&1 | grep -q -- --start ||
9578 skip "Need 'lctl llog_print --start' on MGS"
9580 [ -d $MOUNT/.lustre ] || setup
9582 # append a new record, to avoid issues if last record was cancelled
9583 local old=$($LCTL get_param -n osc.*-OST0000-*.max_dirty_mb | head -1)
9584 do_facet mgs $LCTL conf_param $FSNAME-OST0000.osc.max_dirty_mb=$old
9586 # logid: [0x3:0xa:0x0]:0
9590 local num=$(do_facet mgs $LCTL --device MGS llog_info $FSNAME-client |
9591 awk '/last_index:/ { print $2 - 1 }')
9593 # - { index: 71, event: set_timeout, num: 0x14, param: sys.timeout=20 }
9594 local last=$(do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
9595 tail -1 | awk '{ print $4 }' | tr -d , )
9596 (( last == num )) || error "llog_print only showed $last/$num records"
9598 run_test 123ad "llog_print shows all records"
9600 test_123ae() { # LU-11566
9609 remote_mgs_nodsh && skip "remote MGS with nodsh"
9610 [ -d $MOUNT/.lustre ] || setupall
9612 max=$($LCTL get_param -n osc.*-OST0000-*.max_dirty_mb | head -1)
9613 pgs=$($LCTL get_param -n osc.*-OST0000-*.max_pages_per_rpc | head -1)
9614 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9615 mgs_arg="--device MGS"
9617 if do_facet mgs "$LCTL help llog_cancel" 2>&1| grep -q -- --log_id; then
9618 # save one set_param -P record in case none exist
9620 do_facet mgs $LCTL set_param -P osc.*.max_pages_per_rpc=$pgs
9621 stack_trap "do_facet mgs $LCTL set_param -P -d \
9622 osc.*.max_pages_per_rpc"
9625 orig=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9626 tail -1 | awk '{ print $4 }' | tr -d , )
9627 do_facet mgs $LCTL set_param -P osc.*.max_dirty_mb=$max
9628 do_facet mgs $LCTL $mgs_arg llog_print $log | tail -1 |
9629 grep "parameter: osc.*.max_dirty_mb" ||
9630 error "new set_param -P wasn't stored in params log"
9632 # - { index: 71, event: set_param, device: general,
9633 # param: osc.*.max_dirty_mb, value: 256 }
9634 id=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9635 tail -1 | awk '{ print $4 }' | tr -d , )
9636 do_facet mgs $LCTL $mgs_arg llog_cancel $log --log_idx=$id
9637 local new=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9638 tail -1 | awk '{ print $4 }' | tr -d , )
9639 (( new == orig )) ||
9640 error "new llog_cancel now $new, not at $orig records"
9643 # test old positional parameters for a while still
9644 if [ "$MGS_VERSION" -le $(version_code 3.1.53) ]; then
9647 do_facet mgs $LCTL conf_param \
9648 $FSNAME-OST0000.osc.max_pages_per_rpc=$pgs
9649 stack_trap "do_facet mgs $LCTL conf_param -d \
9650 $FSNAME-OST0000.osc.max_pages_per_rpc"
9652 orig=$(do_facet mgs $LCTL --device MGS llog_print $log |
9653 tail -1 | awk '{ print $4 }' | tr -d , )
9654 do_facet mgs $LCTL conf_param $FSNAME-OST0000.osc.max_dirty_mb=$max
9655 do_facet mgs $LCTL --device MGS llog_print $log |
9656 tail -1 | grep "parameter: osc.max_dirty_mb" ||
9657 error "old conf_param wasn't stored in params log"
9658 do_facet mgs $LCTL --device MGS llog_print $log
9659 # - { index: 71, event: conf_param, device: testfs-OST0000-osc,
9660 # param: osc.max_dirty_mb=256 }
9661 id=$(do_facet mgs $LCTL --device MGS llog_print $log |
9662 tail -1 | awk '{ print $4 }' | tr -d , )
9663 do_facet mgs $LCTL --device MGS llog_cancel $log $id
9664 do_facet mgs $LCTL --device MGS llog_print $log
9665 new=$(do_facet mgs $LCTL --device MGS llog_print $log |
9666 tail -1 | awk '{ print $4 }' | tr -d , )
9667 (( new == orig )) ||
9668 error "old llog_cancel now $new, not at $orig records"
9671 run_test 123ae "llog_cancel can cancel requested record"
9673 test_123af() { #LU-13609
9674 [ "$MGS_VERSION" -ge $(version_code 2.13.54) -a \
9675 "$MDS1_VERSION" -ge $(version_code 2.13.54) ] ||
9676 skip "Need both MGS and MDS version at least 2.13.54"
9678 [ -d $MOUNT/.lustre ] || setupall
9679 stack_trap "do_facet mds1 $LCTL set_param fail_loc=0" EXIT
9689 for device in "MGS" "$FSNAME-MDT0000"; do
9690 cmd="--device $device llog_catlist"
9691 echo "lctl $cmd ..."
9692 if [ "$device" = "MGS" ]; then
9697 orig_clist=($(do_facet $facet $LCTL $cmd | awk '{ print $2 }'))
9698 orig_count=${#orig_clist[@]}
9699 echo "orig_clist: ${orig_clist[*]}"
9701 #define OBD_FAIL_CATLIST 0x131b
9702 #fetch to llog records from the second one
9703 do_facet $facet $LCTL set_param fail_loc=0x131b fail_val=2
9705 new_clist=($(do_facet $facet $LCTL $cmd | awk '{ print $2 }'))
9706 new_count=${#new_clist[@]}
9707 echo "new_clist: ${new_clist[*]}"
9709 [ $new_count -eq $((orig_count - 1)) ] ||
9710 error "$new_count != $orig_count - 1"
9711 for i in $(seq 0 $new_count); do
9713 [ "${orig_clist[$j]}" = "${new_clist[$i]}" ] ||
9714 error "${orig_clist[$j]} != ${new_clist[$i]}"
9716 do_facet mds1 $LCTL set_param fail_loc=0
9720 run_test 123af "llog_catlist can show all config files correctly"
9722 test_123ag() { # LU-15142
9726 remote_mgs_nodsh && skip "remote MGS with nodsh"
9727 (( $MGS_VERSION >= $(version_code 2.14.55) )) ||
9728 skip "Need server version least 2.14.55"
9730 [ -d $MOUNT/.lustre ] || setup
9732 orig_val=$(do_facet mgs $LCTL get_param jobid_name)
9733 stack_trap "do_facet mgs $LCTL set_param -P jobid_name=$orig_val"
9735 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME1"
9736 do_facet mgs $LCTL set_param -P -d jobid_name
9737 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
9739 (( rec == 0 )) || error "parameter was not deleted, check #1"
9740 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME1"
9741 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
9743 (( rec == 1)) || error "parameter is not set"
9744 # usage with ordinary set_param format works too
9745 do_facet mgs $LCTL set_param -P -d jobid_name="ANY"
9746 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
9748 (( rec == 0 )) || error "parameter was not deleted, check #2"
9750 run_test 123ag "llog_print skips values deleted by set_param -P -d"
9752 test_123ah() { #LU-7668 del_ost
9753 [ "$MGS_VERSION" -ge $(version_code 2.15.50) -a \
9754 "$MDS1_VERSION" -ge $(version_code 2.15.50) ] ||
9755 skip "Need both MGS and MDS version at least 2.15.50"
9757 [ -d $MOUNT/.lustre ] || setupall
9758 stack_trap "do_facet mds1 $LCTL set_param fail_loc=0" EXIT
9760 local cmd="--device MGS llog_print"
9762 cli_llogcnt_orig=$(do_facet mgs $LCTL $cmd $FSNAME-client |
9763 grep -c $FSNAME-OST0000)
9764 mds1_llogcnt_orig=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 |
9765 grep -c $FSNAME-OST0000)
9767 [ $cli_llogcnt_orig -gt 0 ] ||
9768 error "$FSNAME-OST0000 not found (client)"
9769 [ $mds1_llogcnt_orig -gt 0 ] || error "$FSNAME-OST0000 not found (MDT)"
9771 # -n/--dryrun should NOT modify catalog
9772 do_facet mgs "$LCTL del_ost -n -t $FSNAME-OST0000" ||
9773 error "del_ost --dryrun failed with $?"
9775 local cli_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-client |
9776 grep -c $FSNAME-OST0000)
9777 local mds1_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 |
9778 grep -c $FSNAME-OST0000)
9780 [ $cli_llogcnt -eq $cli_llogcnt_orig ] ||
9781 error "--dryrun error: $cli_llogcnt != $cli_llogcnt_orig"
9782 [ $mds1_llogcnt -eq $mds1_llogcnt_orig ] ||
9783 error "--dryrun error: $mds1_llogcnt != $mds1_llogcnt_orig"
9786 do_facet mgs "$LCTL del_ost --target $FSNAME-OST0000" ||
9787 error "del_ost failed with $?"
9789 local cli_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-client |
9790 grep -c $FSNAME-OST0000)
9791 local mds1_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 |
9792 grep -c $FSNAME-OST0000)
9794 # every catalog entry for OST0000 should have been cancelled
9795 [ $cli_llogcnt -eq 0 ] || error "$FSNAME-OST0000 not cancelled (cli)"
9796 [ $mds1_llogcnt -eq 0 ] || error "$FSNAME-OST0000 not cancelled (MDT)"
9798 umount_client $MOUNT
9801 $LFS df | grep -q OST0000 && error "del_ost did not remove OST0000!"
9805 run_test 123ah "del_ost cancels config log entries correctly"
9810 echo "cleanup test 123ai"
9812 # cancel last timeout record
9813 do_facet mgs "$LCTL set_param -P -d timeout"
9815 # restore timeout value
9816 do_nodes $(comma_list $(all_nodes)) "$LCTL set_param timeout=$timeout"
9819 test_123ai() { #LU-16167
9825 remote_mgs_nodsh && skip "remote MGS with nodsh"
9826 (( MDS1_VERSION >= $(version_code 2.15.51) )) ||
9827 skip "Need MDS version at least 2.15.51"
9829 [ -d $MOUNT/.lustre ] || setup
9831 old_count=$(do_facet mgs "$LCTL --device MGS llog_print -r params" |
9832 grep -c "parameter: timeout")
9834 old_timeout=$($LCTL get_param -n timeout)
9835 stack_trap "cleanup_123ai $old_timeout" EXIT
9837 # add and cancel params (2 * MAX_IOC_BUFLEN / 128 + 1 = 129)
9838 for i in {1..129}; do
9839 do_facet mgs "$LCTL set_param -P timeout=$i" ||
9840 error "fail to set timeout=$i on MGS"
9843 count=$(do_facet mgs "$LCTL --device MGS llog_print -r params" |
9844 grep -c "parameter: timeout")
9846 ((count - old_count == 129)) ||
9847 error "number timeout record missmatch ($count - $old_count != 129)"
9849 do_facet mgs "$LCTL --device MGS llog_print params" |
9850 tail -1 | grep "timeout, value: 129" ||
9851 error "llog_print could not display the last record (timeout=129)"
9854 run_test 123ai "llog_print display all non skipped records"
9857 remote_mgs_nodsh && skip "remote MGS with nodsh"
9859 [ -d $MOUNT/.lustre ] || setup
9860 local yaml_file="$TMP/$tfile.yaml"
9862 do_facet mgs rm "$yaml_file"
9863 local cfgfiles=$(do_facet mgs "lctl --device MGS llog_catlist" |
9864 sed 's/config_log://')
9866 # set jobid_var to a different value for test
9867 local orig_val=$(do_facet mgs $LCTL get_param jobid_var)
9869 do_facet mgs $LCTL set_param -P jobid_var="TESTNAME"
9871 for i in $cfgfiles params; do
9872 do_facet mgs "lctl --device MGS llog_print ${i} >> $yaml_file"
9875 echo "Unmounting FS"
9882 # Reapply the config from before
9883 echo "Setting configuration parameters"
9884 do_facet mgs "lctl set_param -F $yaml_file"
9886 local set_val=$(do_facet mgs $LCTL get_param jobid_var)
9888 do_facet mgs $LCTL set_param -P $orig_val
9890 [ $set_val == "jobid_var=TESTNAME" ] ||
9891 error "$set_val is not TESTNAME"
9893 do_facet mgs rm "$yaml_file"
9896 run_test 123F "clear and reset all parameters using set_param -F"
9900 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
9901 [ -z $mds2failover_HOST ] && skip "needs MDT failover setup"
9904 do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
9905 grep 1.2.3.4@tcp && error "Should not be fake nid"
9906 do_facet mgs $LCTL conf_param $FSNAME-MDT0001.failover.node=1.2.3.4@tcp\
9907 || error "Set params error"
9908 do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
9909 grep 1.2.3.4@tcp || error "Fake nid should be added"
9913 if combined_mgs_mds; then
9914 start_mdt 1 "-o nosvc" ||
9915 error "starting mds with nosvc option failed"
9917 local nid=$(do_facet mds2 $LCTL list_nids | head -1)
9918 local failover_nid=$(do_node $mds2failover_HOST $LCTL list_nids | head -1)
9920 do_facet mgs $LCTL replace_nids $FSNAME-MDT0001 $nid:$failover_nid ||
9921 error "replace_nids execution error"
9923 if combined_mgs_mds; then
9929 echo "lfs setdirstripe"
9930 $LFS setdirstripe -i 1 $MOUNT/$tdir || error "setdirstirpe error"
9933 run_test 124 "check failover after replace_nids"
9935 get_max_sectors_kb() {
9938 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
9939 local max_sectors_path="/sys/block/${dev_base}/queue/max_sectors_kb"
9941 do_facet ${facet} "[[ -e ${max_sectors_path} ]] &&
9942 cat ${max_sectors_path}"
9945 get_max_hw_sectors_kb() {
9948 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
9949 local max_hw_path="/sys/block/${dev_base}/queue/max_hw_sectors_kb"
9951 do_facet ${facet} "[[ -e ${max_hw_path} ]] && cat ${max_hw_path}"
9954 set_max_sectors_kb() {
9958 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
9959 local max_sectors_path="/sys/block/${dev_base}/queue/max_sectors_kb"
9961 do_facet ${facet} "[[ -e ${max_sectors_path} ]] &&
9962 echo ${value} > ${max_sectors_path}"
9965 [[ $rc -ne 0 ]] && echo "Failed to set ${max_sectors_path} to ${value}"
9970 # Return 0 if all slave devices have max_sectors_kb == max_hw_sectors_kb
9971 # Otherwise return > 0
9972 check_slaves_max_sectors_kb()
9976 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
9977 local slaves_dir=/sys/block/${dev_base}/slaves
9978 local slave_devices=$(do_facet ${facet} "ls ${slaves_dir} 2>/dev/null")
9979 [[ -z ${slave_devices} ]] && return 0
9981 local slave max_sectors new_max_sectors max_hw_sectors path
9984 for slave in ${slave_devices}; do
9985 path="/dev/${slave}"
9986 ! is_blkdev ${facet} ${path} && continue
9987 max_sectors=$(get_max_sectors_kb ${facet} ${path})
9988 max_hw_sectors=$(get_max_hw_sectors_kb ${facet} ${path})
9989 new_max_sectors=${max_hw_sectors}
9990 [[ ${new_max_sectors} -gt ${RQ_SIZE_LIMIT} ]] &&
9991 new_max_sectors=${RQ_SIZE_LIMIT}
9993 if [[ ${max_sectors} -ne ${new_max_sectors} ]]; then
9994 echo "${path} ${max_sectors} ${new_max_sectors}"
9997 check_slaves_max_sectors_kb ${facet} ${path}
10006 local facet_list="mgs mds1 ost1"
10007 combined_mgs_mds && facet_list="mgs ost1"
10010 for facet in ${facet_list}; do
10011 [[ $(facet_fstype ${facet}) != ldiskfs ]] &&
10012 skip "ldiskfs only test" &&
10014 ! is_blkdev ${facet} $(facet_device ${facet}) &&
10015 skip "requires all real devices" &&
10020 # We don't increase IO request size limit past 16MB. See comments in
10021 # lustre/utils/libmount_utils_ldiskfs.c:tune_max_sectors_kb()
10022 RQ_SIZE_LIMIT=$((16 * 1024))
10023 local device old_max_sectors new_max_sectors max_hw_sectors
10024 for facet in ${facet_list}; do
10025 device=$(facet_device ${facet})
10026 old_max_sectors=$(get_max_sectors_kb ${facet} ${device})
10027 max_hw_sectors=$(get_max_hw_sectors_kb ${facet} ${device})
10029 # The expected value after l_tunedisk is executed
10030 new_max_sectors=$old_max_sectors
10031 [[ ${new_max_sectors_kb} -gt ${RQ_SIZE_LIMIT} ]] &&
10032 new_max_sectors_kb=${RQ_SIZE_LIMIT}
10034 # Ensure the current value of max_sectors_kb does not equal
10035 # max_hw_sectors_kb, so we can tell whether l_tunedisk did
10037 set_max_sectors_kb ${facet} ${device} $((new_max_sectors - 1))
10039 # Value before l_tunedisk
10040 local pre_max_sectors=$(get_max_sectors_kb ${facet} ${device})
10041 if [[ ${pre_max_sectors} -ne $((new_max_sectors - 1)) ]]; then
10042 echo "unable to satsify test pre-condition:"
10043 echo "${pre_max_sectors} != $((new_max_sectors - 1))"
10048 echo "Before: ${facet} ${device} ${pre_max_sectors} ${max_hw_sectors}"
10050 do_facet ${facet} "l_tunedisk ${device}"
10052 # Value after l_tunedisk
10053 local post_max_sectors=$(get_max_sectors_kb ${facet} ${device})
10055 echo "After: ${facet} ${device} ${post_max_sectors} ${max_hw_sectors}"
10057 if [[ ${facet} != ost1 ]]; then
10058 if [[ ${post_max_sectors} -ne ${pre_max_sectors} ]]; then
10059 echo "l_tunedisk modified max_sectors_kb of ${facet}"
10063 set_max_sectors_kb ${facet} ${device} ${old_max_sectors}
10065 if [[ ${post_max_sectors} -eq ${pre_max_sectors} ]]; then
10066 echo "l_tunedisk failed to modify max_sectors_kb of ${facet}"
10070 check_slaves_max_sectors_kb ${facet} ${device} ||
10077 run_test 125 "check l_tunedisk only tunes OSTs and their slave devices"
10080 [[ "$MDS1_VERSION" -ge $(version_code 2.13.52) ]] ||
10081 skip "Need MDS version at least 2.13.52"
10084 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ../libcfs/libcfs/libcfs
10085 #define OBD_FAIL_OBD_SETUP 0x60d
10086 do_facet mds1 $LCTL set_param fail_loc=0x60d
10087 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_modules &
10088 for i in {1..40}; do
10089 do_facet mds1 lsmod | grep -q osd_$mds1_FSTYPE && break
10092 clear_failloc $SINGLEMDS 20 &
10093 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
10095 run_test 126 "mount in parallel shouldn't cause a crash"
10098 [[ "$ost1_FSTYPE" == ldiskfs ]] || skip "ldiskfs only test"
10102 zconf_umount_clients $RCLIENTS $MOUNT
10104 wait_osp_active ost ${FSNAME}-OST0000 0 1
10105 local osc_tgt="$FSNAME-OST0000-osc-$($LFS getname -i $DIR)"
10106 local avail1=($($LCTL get_param -n osc.${osc_tgt}.kbytesavail))
10108 $LFS setstripe -i 0 $DIR/$tfile || error "failed creating $DIR/$tfile"
10109 dd if=/dev/zero of=$DIR/$tfile bs=1M oflag=direct || true
10111 local avail2=($($LCTL get_param -n osc.${osc_tgt}.kbytesavail))
10113 if ((avail2 * 100 / avail1 > 1)); then
10116 error "more than 1% space left: before=$avail1 after=$avail2"
10119 local mbs=$(($(stat -c %s $DIR/$tfile) / (1024 * 1024)))
10121 dd if=/dev/zero of=$DIR/$tfile bs=1M count=$mbs conv=notrunc \
10122 oflag=direct || error "overwrite failed"
10124 run_test 127 "direct io overwrite on full ost"
10128 combined_mgs_mds && skip "need separate mgs device"
10129 [ "$ost2_FSTYPE" == zfs ] && import_zpool ost2
10132 # Try to apply nolocallogs to the virgin OST. Should fail.
10133 do_facet ost2 "$TUNEFS --nolocallogs $(ostdevname 2)" &&
10134 error "nolocallogs should not be allowed on the virgin target"
10139 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
10140 # Start OST without MGS (local configs)
10141 do_facet ost1 "$TUNEFS --dryrun $(ostdevname 1)"
10142 start_ost || error "unable to start OST1"
10143 stop_ost || error "Unable to stop OST1"
10145 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
10146 # Do not allow reading local configs, should fail
10147 do_facet ost1 "$TUNEFS --nolocallogs $(ostdevname 1)" ||
10148 error "Can not set nolocallogs"
10149 start_ost && error "OST1 started, but should fail"
10151 # Connect to MGS successfully, reset nolocallogs flag
10152 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
10153 start_mgs || error "unable to start MGS"
10154 start_ost || error "unable to start OST1"
10156 do_facet ost1 "$TUNEFS --dryrun $(ostdevname 1)" | grep "nolocallogs" &&
10157 error "nolocallogs expected to be reset"
10159 stop_ost || error "Unable to stop OST1"
10161 run_test 128 "Force using remote logs with --nolocallogs"
10165 (( MDS1_VERSION >= $(version_code 2.14.57) )) ||
10166 skip "Need MDS version at least 2.14.57"
10168 start_mgsmds || error "MGS/MDS start failed"
10170 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &&
10171 error "start ost1 should fail" || true
10172 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &&
10173 error "second start ost1 should fail" || true
10174 do_facet ost1 "$TUNEFS --writeconf $(ostdevname 1)"
10175 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS ||
10176 error "start ost1 failed"
10180 run_test 129 "attempt to connect an OST with the same index should fail"
10184 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
10186 stop_mdt 2 || error "mdt2 stop failed"
10187 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)"
10188 start_mdt 2 || error "mdt2 start failed"
10189 do_facet mds2 "$LCTL dl" | grep MDT0001-osp-MDT0001 &&
10190 error "Illegal OSP device created" || true
10192 run_test 130 "re-register an MDT after writeconf"
10195 [ "$mds1_FSTYPE" == "ldiskfs" ] || skip "ldiskfs only test"
10196 (( $MDS1_VERSION >= $(version_code 2.14.56.35) )) ||
10197 skip "Need MDS version at least 2.14.56.35"
10198 do_facet mds1 $DEBUGFS -R features $(mdsdevname 1) |
10199 grep -q project || skip "skip project quota not supported"
10204 test_mkdir -c $MDSCOUNT -p $DIR/$tdir
10205 $LFS project -p 1000 $DIR/$tdir || error "set dir project id failed"
10206 createmany -o $DIR/$tdir/f 512
10207 for ((i = 0; i < 512; ++i)); do
10208 $LFS project -p $i $DIR/$tdir/f${i} ||
10209 error "set f${i} project id failed"
10212 test_mkdir -c $MDSCOUNT -p $DIR/$tdir.inherit
10213 $LFS project -p 1001 -s $DIR/$tdir.inherit
10214 createmany -o $DIR/$tdir.inherit/f 128
10215 (( $($LFS project $DIR/$tdir.inherit/f* |
10216 awk '$1 == 1001 { print }' | wc -l) == 128 )) ||
10217 error "files did not inherit projid 1001"
10221 for i in $(seq $MDSCOUNT); do
10222 mds_backup_restore mds$i ||
10223 error "Backup/restore on mds$i failed"
10228 projid=($($LFS project -d $DIR/$tdir))
10229 [ ${projid[0]} == "1000" ] ||
10230 error "projid expected 1000 not ${projid[0]}"
10231 for ((i = 0; i < 512; ++i)); do
10232 projid=($($LFS project $DIR/$tdir/f${i}))
10233 [ ${projid[0]} == "$i" ] ||
10234 error "projid expected $i not ${projid[0]}"
10237 (( $($LFS project $DIR/$tdir.inherit/f* |
10238 awk '$1 == 1001 { print }' | wc -l) == 128 )) ||
10239 error "restore did not copy projid 1001"
10241 run_test 131 "MDT backup restore with project ID"
10247 (( MDS1_VERSION >= $(version_code 2.14.57) )) ||
10248 skip "Need MDS version at least 2.14.57"
10250 combined_mgs_mds || start_mgs || error "unable to start MGS"
10251 start_mdt 1 || error "unable to start mdt1"
10253 err_cnt=$(do_facet mds1 dmesg | grep -c "cannot take the layout locks")
10254 stop_mdt 1 || error "stop mdt1 failed"
10256 [ "$mds1_FSTYPE" == zfs ] && import_zpool mds1
10257 do_facet mds1 $TUNEFS --param mdt.hsm_control=enabled $(mdsdevname 1) ||
10258 error "tunefs failed"
10259 start_mdt 1 || error "cannot start mdt1"
10261 err_cnt2=$(do_facet mds1 dmesg | grep -c "cannot take the layout locks")
10262 [ $err_cnt -eq $err_cnt2 ] || error "Can not take the layout lock"
10263 stop_mdt 1 || error "stop mdt1 failed"
10265 run_test 132 "hsm_actions processed after failover"
10267 # This test verifies we do RR allocation within a pool even if there is a
10268 # significant imbalance vs an OST outside the pool
10270 [[ $OSTCOUNT -lt 4 ]] && skip_env "needs >= 4 OSTs"
10271 [[ "$OST1_VERSION" -ge $(version_code 2.15.51) ]] ||
10272 skip "Need OST version at least 2.15.51"
10273 # This is the easiest way to ensure OSTs start out balanced
10274 reformat_and_config
10277 check_set_fallocate_or_skip
10279 local testfile=$DIR/$tdir/$tfile
10280 local pool="testpool"
10281 local ostrange=$((OSTCOUNT - 1))
10282 # Select all but the last OST to add to the pool
10283 local poolostrange=$((OSTCOUNT - 2))
10291 [ $MINV -eq 0 ] && error "no free space in OST$MINI"
10292 [ $MAXV -gt $((2 * $MINV)) ] &&
10293 error "OSTs badly unbalanced after reformat"
10295 create_pool $FSNAME.$pool || error "failed to create a pool"
10296 do_facet mgs $LCTL pool_add $FSNAME.$pool OST[0-$poolostrange] ||
10297 error "failed to add OST[0-$poolostrange] to the pool"
10299 test_mkdir -p $DIR/$tdir || error "failed to mkdir $DIR/$tdir"
10300 # Consume space on the OSTs in the pool so they are unbalanced with the
10301 # OST outside of the pool
10302 # fill each OST 90% with fallocate so they are widely
10304 local size=$(((MINV * 9 / 10) * 1024))
10305 for ((i = 0; i <= poolostrange; i++)); do
10306 $LFS setstripe -c 1 -i $i $testfile$i ||
10307 error "failed to setstripe $testfile$i"
10308 fallocate -l $size $testfile$i || error "fallocate failed"
10314 # Create files in the pool now that there is an imbalance
10315 filecount=$(((OSTCOUNT - 1) * filenum))
10316 for ((i = 0; i < filecount; i++)); do
10317 $LFS setstripe -p $pool $testfile-$i ||
10318 error "failed to setstripe -p $pool $testfile-$i"
10320 $LFS getstripe -i $testfile-* > /tmp/$tfile.log
10321 # Count the number of files with a stripe on each OST to verify the
10322 # pool allocated with round-robin
10323 for ((i = 0; i <= poolostrange; i++)); do
10324 stripecount=$(grep -c $i /tmp/$tfile.log)
10325 # Allow a little leeway
10326 if (( stripecount < filenum - 1 ||
10327 stripecount > filenum + 1 )); then
10328 cat /tmp/$tfile.log
10329 error "$stripecount != $filenum files on OST$i"
10333 # Create files across the system now that there is an imbalance
10334 filecount=$((OSTCOUNT * filenum))
10335 for ((i = 1; i < filecount; i++)); do
10336 $LFS setstripe $testfile-$i.2 ||
10337 error "failed to setstripe $testilfe-$i.2"
10339 $LFS getstripe -i $testfile-*.2 > /tmp/$tfile.log
10341 # Count the number of files with a stripe on each OST to verify the
10342 # files are *NOT* allocated with round-robin
10343 for ((i = 0; i <= ostrange; i++)); do
10344 stripecount=$(grep -c $i /tmp/$tfile.log)
10345 if [[ $stripecount -ne $filenum ]]; then
10347 echo "QOS: $stripecount != $filenum files on OST$i"
10350 if [ -z "$qos_used" ]; then
10351 error "QOS not used on imbalanced OSTs!"
10354 rm -rf /tmp/$tfile.log $DIR/$tdir
10355 do_facet mgs $LCTL pool_remove $FSNAME.$pool OST[0-$poolostrange] ||
10356 "failed to remove OST[0-$poolostrange] from the pool"
10357 do_facet mgs $LCTL pool_destroy $FSNAME.$pool ||
10358 error "failed to destroy pool"
10360 run_test 133 "stripe QOS: free space balance in a pool"
10363 [ "$mds1_FSTYPE" == "ldiskfs" ] || skip "ldiskfs only test"
10366 local mdt_dev=$(facet_device mds1)
10367 local tmp_dir=$TMP/$tdir
10368 local out=$tmp_dir/check_iam.txt
10369 local CHECK_IAM=${CHECK_IAM:-$(do_facet mds1 "which check_iam 2> /dev/null || true")}
10371 [[ -n "$CHECK_IAM" ]] || skip "check_iam not found"
10374 for ((i=0; i<64; i++)); do
10376 #cmd introduce a random corruption to IAM file
10377 local cmd="dd if=/dev/urandom of=$tmp_dir/$f bs=2 conv=notrunc count=1 seek=$((RANDOM % 36))"
10378 do_facet mds1 "mkdir -p $tmp_dir; \
10379 $DEBUGFS -c -R 'dump $f $tmp_dir/$f' $mdt_dev 2>&1; \
10380 $CHECK_IAM -v $tmp_dir/$f 2>&1; \
10381 echo $cmd; eval $cmd 2>/dev/null;
10382 $CHECK_IAM -v $tmp_dir/$f 2>&1; echo \\\$?" >> $out
10387 stack_trap "rm -rf $tmp_dir && do_facet mds1 rm -rf $tmp_dir" EXIT
10389 rc=$(grep -c "fault\|except" $out)
10390 (( rc == 0 )) || { cat $out &&
10391 error "check_iam failed with fault or exception $rc"; }
10393 rc=$(grep -c "^255" $out)
10394 error=$(grep -c "FINISHED WITH ERRORS" $out)
10396 (( rc == error )) || { cat $out &&
10397 error "check_iam errcode does not fit with errors $rc $error"; }
10399 run_test 134 "check_iam works without faults"
10401 cleanup_test_135(){
10404 printf "\nCleanup test_135\n" >&2
10405 do_facet mds1 "$LCTL set_param -n $oldgc"
10406 rm -rf $DIR/$tdir &> /dev/null
10410 __test_135_file_thread() {
10412 local init_time=$(awk '{print $1}' /proc/uptime)
10413 local awkcmd="/crosses index zero/ {if (\$1 > $init_time) exit(1);}"
10415 #Generate a full plain llogs
10416 while dmesg | sed -r 's/(\[|\])//g' | awk "$awkcmd" ; do
10417 createmany -o $DIR/$tdir/f 4500 >&2
10418 createmany -u $DIR/$tdir/f 4500 >&2
10422 __test_135_reader() {
10425 local firstidx=$(changelog_user_rec mds1 $cl_user)
10426 local oldidx=$firstidx
10431 while read -t10 -u$fd newidx other; do
10432 (( (newidx - oldidx) == 1 )) ||
10433 error "changelog jump detected (last: $oldidx, current: $newidx)"
10435 if (( (newidx - firstidx + 1) % 13000 == 0 )); then
10436 [[ $pid -eq 0 ]] ||
10437 wait $pid || error "changelog_clear failed"
10438 changelog_clear $((newidx - 1)) mds1 >&2 & pid=$!
10443 [[ $pid -eq 0 ]] ||
10444 wait $pid || error "changelog_clear failed"
10450 (( MDS1_VERSION >= $(version_code 2.15.52) )) ||
10451 skip "need MDS version at least 2.15.52"
10453 local service=$(facet_svc mds1)
10455 local lastread lastidx
10456 local files_pid reader_pid
10461 # Need to reformat because we are changing llog catalog sizes to 5.
10462 # Otherwise, processing could fail with existing catalogs (last_idx>5).
10466 # Disable changelog garbage colector
10467 local oldgc=$(do_facet mds1 "$LCTL get_param mdd.${service}.changelog_gc")
10468 do_facet mds1 "$LCTL set_param -n mdd.${service}.changelog_gc=0"
10469 stack_trap "cleanup_test_135 $oldgc" EXIT INT
10471 # change the changelog_catalog size to 5 entries for everybody
10472 #define OBD_FAIL_CAT_RECORDS 0x1312
10473 do_node $(comma_list $(all_nodes)) $LCTL set_param fail_loc=0x1312 fail_val=5
10475 # disable console ratelimit
10476 local rl=$(cat /sys/module/libcfs/parameters/libcfs_console_ratelimit)
10477 echo 0 > /sys/module/libcfs/parameters/libcfs_console_ratelimit
10478 stack_trap "echo $rl > /sys/module/libcfs/parameters/libcfs_console_ratelimit" EXIT
10480 test_mkdir -c 1 -i 0 $DIR/$tdir || error "Failed to create directory"
10481 changelog_chmask "ALL" || error "changelog_chmask failed"
10482 changelog_register || error "changelog_register failed"
10484 cl_user="${CL_USERS[mds1]%% *}"
10485 changelog_users mds1 | grep -q $cl_user ||
10486 error "User $cl_user not found in changelog_users"
10488 # Start reader thread
10489 coproc $LFS changelog --follow $service
10492 stack_trap "echo kill changelog reader; kill $reader_pid" EXIT
10494 echo -e "\nWrap arround changelog catalog"
10496 # Start file writer thread
10497 __test_135_file_thread "$service" & files_pid=$!
10498 stack_trap "(pkill -P$files_pid; kill $files_pid) &> /dev/null || true" EXIT
10500 # Check changelog entries
10501 lastread=$(__test_135_reader $fd $cl_user) || exit $?
10502 ! kill -0 $files_pid 2>/dev/null ||
10503 error "creation thread is running. Is changelog reader stuck?"
10505 lastidx=$(changelog_users mds1 | awk '/current_index/ {print $NF}' )
10506 [[ "$lastread" -eq "$lastidx" ]] ||
10507 error "invalid changelog lastidx (read: $lastread, mds: $lastidx)"
10509 run_test 135 "check the behavior when changelog is wrapped around"
10512 # (This was sanity/802a)
10514 saved_MGS_MOUNT_OPTS=$MGS_MOUNT_OPTS
10515 saved_MDS_MOUNT_OPTS=$MDS_MOUNT_OPTS
10516 saved_OST_MOUNT_OPTS=$OST_MOUNT_OPTS
10517 saved_MOUNT_OPTS=$MOUNT_OPTS
10521 MGS_MOUNT_OPTS=$saved_MGS_MOUNT_OPTS
10522 MDS_MOUNT_OPTS=$saved_MDS_MOUNT_OPTS
10523 OST_MOUNT_OPTS=$saved_OST_MOUNT_OPTS
10524 MOUNT_OPTS=$saved_MOUNT_OPTS
10529 [[ $mds1_FSTYPE = zfs ]] || skip "ZFS specific test"
10530 [[ $MDS1_VERSION -lt $(version_code 2.9.55) ]] ||
10531 [[ $OST1_VERSION -lt $(version_code 2.9.55) ]] &&
10532 skip "Need server version at least 2.9.55"
10534 [[ $ENABLE_QUOTA ]] && skip "Quota enabled for read-only test"
10536 # Reset before starting
10540 mkdir $DIR/$tdir || error "(1) fail to mkdir"
10542 cp $LUSTRE/tests/test-framework.sh $DIR/$tdir/ ||
10543 error "(2) Fail to copy"
10545 stack_trap cleanup_802a EXIT
10547 # sync by force before remount as readonly
10548 sync; sync_all_data; sleep 3; sync_all_data
10552 MGS_MOUNT_OPTS=$(csa_add "$MGS_MOUNT_OPTS" -o rdonly_dev)
10553 MDS_MOUNT_OPTS=$(csa_add "$MDS_MOUNT_OPTS" -o rdonly_dev)
10554 OST_MOUNT_OPTS=$(csa_add "$OST_MOUNT_OPTS" -o rdonly_dev)
10556 echo "Mount the server as read only"
10557 setupall server_only || error "(3) Fail to start servers"
10559 echo "Mount client without ro should fail"
10560 mount_client $MOUNT &&
10561 error "(4) Mount client without 'ro' should fail"
10563 echo "Mount client with ro should succeed"
10564 MOUNT_OPTS=$(csa_add "$MOUNT_OPTS" -o ro)
10565 mount_client $MOUNT ||
10566 error "(5) Mount client with 'ro' should succeed"
10568 echo "Modify should be refused"
10569 touch $DIR/$tdir/guard && error "(6) Touch should fail under ro mode"
10571 echo "Read should be allowed"
10572 diff $LUSTRE/tests/test-framework.sh $DIR/$tdir/test-framework.sh ||
10573 error "(7) Read should succeed under ro mode"
10575 run_test 802a "simulate readonly device"
10577 if ! combined_mgs_mds ; then
10583 # restore the values of MDSSIZE and OSTSIZE
10584 MDSSIZE=$STORED_MDSSIZE
10585 OSTSIZE=$STORED_OSTSIZE
10589 check_and_cleanup_lustre