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_ready ${CLIENTS:-$HOSTNAME} ost1
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_ready ${CLIENTS:-$HOSTNAME} ost2
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 $MOUNT2/hosts failed"
3044 if (( "$MGS_VERSION" >= $(version_code 2.15.51) &&
3045 "$MDS1_VERSION" >= $(version_code 2.15.51) )); then
3046 # test lctl del_ost on large index
3047 do_facet mgs "$LCTL del_ost -t ${FSNAME2}-OST1fff" ||
3048 error "del_ost failed with $?"
3049 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 ||
3050 error "$MOUNT_CMD failed"
3052 $LFS df | grep -q OST1fff &&
3053 error "del_ost did not remove OST1fff!"
3059 cleanup_nocli || error "cleanup_nocli failed with $?"
3061 run_test 33a "Mount ost with a large index number"
3063 test_33b() { # was test_34
3066 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
3067 # Drop lock cancelation reply during umount
3068 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
3069 do_facet client $LCTL set_param fail_loc=0x80000304
3070 #lctl set_param debug=-1
3071 umount_client $MOUNT
3072 cleanup || error "cleanup failed with $?"
3074 run_test 33b "Drop cancel during umount"
3078 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
3079 manual_umount_client
3081 do_facet client killall -USR1 multiop
3082 if [ $rc -eq 0 ]; then
3083 error "umount not fail!"
3086 cleanup || error "cleanup failed with rc $?"
3088 run_test 34a "umount with opened file should be fail"
3092 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3093 stop_mds || error "Unable to stop MDS"
3095 manual_umount_client --force || error "mtab after failed umount with $?"
3097 cleanup || error "cleanup failed with $?"
3099 run_test 34b "force umount with failed mds should be normal"
3103 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3104 stop_ost || error "Unable to stop OST1"
3106 manual_umount_client --force || error "mtab after failed umount with $?"
3108 cleanup || error "cleanup failed with $?"
3110 run_test 34c "force umount with failed ost should be normal"
3112 test_35a() { # bug 12459
3115 DBG_SAVE="`$LCTL get_param -n debug`"
3116 $LCTL set_param debug="ha"
3118 log "Set up a fake failnode for the MDS"
3120 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
3121 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
3123 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3124 do_facet mgs "$PERM_CMD \
3125 mdc.*${device}*.import=connection=$(h2nettype $FAKENID)" ||
3126 error "Setting mdc.*${device}*.import=connection=\
3127 $(h2nettype $FAKENID) failed."
3129 do_facet mgs "$PERM_CMD \
3130 ${device}.failover.node=$(h2nettype $FAKENID)" ||
3131 error "Setting ${device}.failover.node=\
3132 $(h2nettype $FAKENID) failed."
3134 log "Wait for RECONNECT_INTERVAL seconds (10s)"
3137 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
3140 log "Stopping the MDT: $device"
3141 stop_mdt 1 || error "MDT0 stop fail"
3143 df $MOUNT > /dev/null 2>&1 &
3145 log "Restarting the MDT: $device"
3146 start_mdt 1 || error "MDT0 start fail"
3147 log "Wait for df ($DFPID) ... "
3150 $LCTL set_param debug="$DBG_SAVE"
3152 # retrieve from the log the first server that the client tried to
3153 # contact after the connection loss
3154 $LCTL dk $TMP/lustre-log-$TESTNAME.log
3155 NEXTCONN=`awk "/${MSG}/ {start = 1;}
3156 /import_select_connection.*$device-mdc.* using connection/ {
3158 if (\\\$NF ~ /$FAKENID/)
3164 }" $TMP/lustre-log-$TESTNAME.log`
3165 [ "$NEXTCONN" != "0" ] &&
3166 error "Tried to connect to ${NEXTCONN} not last active server"
3167 cleanup || error "cleanup failed with $?"
3168 # remove nid settings
3169 writeconf_or_reformat
3171 run_test 35a "Reconnect to the last active server first"
3173 test_35b() { # bug 18674
3174 remote_mds || skip "local MDS"
3178 $LCTL set_param debug="ha"
3180 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
3183 log "Set up a fake failnode for the MDS"
3185 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
3186 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
3188 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3189 do_facet mgs "$PERM_CMD \
3190 mdc.*${device}*.import=connection=$(h2nettype $FAKENID)" ||
3191 error "Set mdc.*${device}*.import=connection=\
3192 $(h2nettype $FAKENID) failed"
3194 do_facet mgs "$PERM_CMD \
3195 ${device}.failover.node=$(h2nettype $FAKENID)" ||
3196 error "Set ${device}.failover.node=\
3197 $(h2nettype $FAKENID) failed"
3200 local at_max_saved=0
3201 # adaptive timeouts may prevent seeing the issue
3202 if at_is_enabled; then
3203 at_max_saved=$(at_max_get mds)
3204 at_max_set 0 mds client
3207 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
3209 log "Injecting EBUSY on MDS"
3210 # Setting OBD_FAIL_MDS_RESEND=0x136
3211 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" ||
3212 error "unable to set param fail_loc=0x80000136"
3214 $LCTL set_param mdc.${FSNAME}*.stats=clear
3216 log "Creating a test file and stat it"
3217 touch $MOUNT/$tdir/$tfile || error "touch $MOUNT/$tdir/$tfile failed"
3218 stat $MOUNT/$tdir/$tfile
3220 log "Stop injecting EBUSY on MDS"
3221 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" ||
3222 error "unable to set param fail_loc=0"
3223 rm -f $MOUNT/$tdir/$tfile || error "remove $MOUNT/$tdir/$tfile failed"
3226 # restore adaptive timeout
3227 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
3229 $LCTL dk $TMP/lustre-log-$TESTNAME.log
3231 CONNCNT=$($LCTL get_param mdc.${FSNAME}*.stats |
3232 awk '/mds_connect/{print $2}')
3234 # retrieve from the log if the client has ever tried to
3235 # contact the fake server after the loss of connection
3236 FAILCONN=`awk "BEGIN {ret = 0;}
3237 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
3239 if (\\\$NF ~ /$FAKENID/) {
3244 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
3246 [ "$FAILCONN" == "0" ] &&
3247 error "The client reconnection has not been triggered"
3248 [ "$FAILCONN" == "2" ] &&
3249 error "Primary server busy, client reconnect to failover failed"
3252 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
3253 # Reconnects are supposed to be rate limited to one every 5s
3254 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] &&
3255 error "Too many reconnects $CONNCNT"
3257 cleanup || error "cleanup failed with $?"
3258 # remove nid settings
3259 writeconf_or_reformat
3261 run_test 35b "Continue reconnection retries, if the active server is busy"
3264 [ $OSTCOUNT -lt 2 ] && skip_env "needs >= 2 OSTs"
3266 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] ||
3270 local FSNAME2=test1234
3271 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3273 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
3275 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
3276 is_blkdev $SINGLEMDS $MDSDEV &&
3277 skip_env "mixed loopback and real device not working"
3280 local fs2mdsdev=$(mdsdevname 1_2)
3281 local fs2ostdev=$(ostdevname 1_2)
3282 local fs3ostdev=$(ostdevname 2_2)
3283 local fs2mdsvdev=$(mdsvdevname 1_2)
3284 local fs2ostvdev=$(ostvdevname 1_2)
3285 local fs3ostvdev=$(ostvdevname 2_2)
3288 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
3289 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
3290 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
3291 # different one than the default value here.
3292 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
3293 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
3294 add fs3ost $(mkfs_opts ost2 ${fs3ostdev}) --mgsnode=$MGSNID \
3295 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
3297 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
3298 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
3299 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
3300 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
3301 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
3303 sleep 5 # until 11778 fixed
3305 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || error "dd failed"
3307 BKTOTAL=$($LCTL get_param -n obdfilter.*.kbytestotal |
3308 awk 'BEGIN{total=0}; {total+=$1}; END{print total}')
3309 BKFREE=$($LCTL get_param -n obdfilter.*.kbytesfree |
3310 awk 'BEGIN{free=0}; {free+=$1}; END{print free}')
3311 BKAVAIL=$($LCTL get_param -n obdfilter.*.kbytesavail |
3312 awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}')
3313 STRING=$(df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}')
3314 DFTOTAL=$(echo $STRING | cut -d, -f1)
3315 DFUSED=$(echo $STRING | cut -d, -f2)
3316 DFAVAIL=$(echo $STRING | cut -d, -f3)
3317 DFFREE=$(($DFTOTAL - $DFUSED))
3319 ALLOWANCE=$((64 * $OSTCOUNT))
3321 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
3322 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
3323 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
3326 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
3327 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
3328 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
3331 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
3332 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
3333 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
3338 stop fs3ost -f || error "unable to stop OST3"
3339 stop fs2ost -f || error "unable to stop OST2"
3340 stop fs2mds -f || error "unable to stop second MDS"
3341 unload_modules_conf || error "unable unload modules"
3344 run_test 36 "df report consistency on OSTs with different block size"
3347 local mntpt=$(facet_mntpt $SINGLEMDS)
3348 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
3349 local mdsdev_sym="$TMP/sym_mdt.img"
3350 local opts=$MDS_MOUNT_OPTS
3353 if [ "$mds1_FSTYPE" != ldiskfs ]; then
3354 skip "ldiskfs only test"
3357 echo "MDS : $mdsdev"
3358 echo "SYMLINK : $mdsdev_sym"
3359 do_facet $SINGLEMDS rm -f $mdsdev_sym
3361 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
3363 echo "mount symlink device - $mdsdev_sym"
3365 if ! do_facet $SINGLEMDS test -b $mdsdev; then
3366 opts=$(csa_add "$opts" -o loop)
3370 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
3371 $mdsdev_sym $mntpt 2>&1)
3374 echo mount_op=$mount_op
3376 do_facet $SINGLEMDS "$UMOUNT $mntpt && rm -f $mdsdev_sym"
3378 if $(echo $mount_op | grep -q "unable to set tunable"); then
3379 error "set tunables failed for symlink device"
3382 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
3384 run_test 37 "verify set tunables works for symlink device"
3386 test_38() { # bug 14222
3387 local mntpt=$(facet_mntpt $SINGLEMDS)
3392 local SRC="/etc /bin"
3393 local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT)
3395 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
3396 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3397 tar cf - $FILES | tar xf - -C $DIR/$tdir ||
3398 error "copying $SRC to $DIR/$tdir"
3400 umount_client $MOUNT || error "umount_client $MOUNT failed"
3401 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3402 stop_mds || error "Unable to stop MDS"
3403 log "delete lov_objid file on MDS"
3405 mount_fstype $SINGLEMDS || error "mount MDS failed (1)"
3407 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid; rm $mntpt/lov_objid"
3409 unmount_fstype $SINGLEMDS || error "umount failed (1)"
3411 # check create in mds_lov_connect
3412 start_mds || error "unable to start MDS"
3413 mount_client $MOUNT || error "mount_client $MOUNT failed"
3415 [ $V ] && log "verifying $DIR/$tdir/$f"
3416 diff -q $f $DIR/$tdir/$f || ERROR=y
3418 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3419 if [ "$ERROR" = "y" ]; then
3420 # check it's updates in sync
3421 umount_client $MOUNT
3423 mount_fstype $SIGNLEMDS
3424 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3425 unmount_fstype $SINGLEMDS
3426 error "old and new files are different after connect" || true
3428 touch $DIR/$tdir/f2 || error "f2 file create failed"
3430 # check it's updates in sync
3431 umount_client $MOUNT || error "second umount_client $MOUNT failed"
3434 mount_fstype $SINGLEMDS || error "mount MDS failed (3)"
3436 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3437 do_facet $SINGLEMDS dd if=/dev/zero of=$mntpt/lov_objid.clear count=8
3439 unmount_fstype $SINGLEMDS || error "umount failed (3)"
3441 start_mds || error "unable to start MDS"
3442 mount_client $MOUNT || error "mount_client $MOUNT failed"
3444 [ $V ] && log "verifying $DIR/$tdir/$f"
3445 diff -q $f $DIR/$tdir/$f || ERROR=y
3447 touch $DIR/$tdir/f3 || error "f3 file create failed"
3448 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3449 umount_client $MOUNT || error "third umount_client $MOUNT failed"
3451 mount_fstype $SINGLEMDS || error "mount MDS failed (4)"
3452 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3453 unmount_fstype $SINGLEMDS || error "umount failed (4)"
3455 [ "$ERROR" = "y" ] &&
3456 error "old and new files are different after sync" || true
3458 log "files compared the same"
3459 cleanup || error "cleanup failed with $?"
3461 run_test 38 "MDS recreates missing lov_objid file from OST data"
3466 cleanup || error "cleanup failed with $?"
3467 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
3468 error "memory leak detected" || true
3470 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
3472 test_40() { # bug 15759
3473 start_ost || error "Unable to start OST1"
3474 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
3475 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
3477 cleanup || error "cleanup failed with rc $?"
3479 run_test 40 "race during service thread startup"
3481 test_41a() { #bug 14134
3482 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3483 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
3484 skip "Loop devices does not work with nosvc option"
3487 combined_mgs_mds || skip "needs combined MGT and MDT device"
3489 start_mdt 1 -o nosvc -n
3490 if [ $MDSCOUNT -ge 2 ]; then
3491 for num in $(seq 2 $MDSCOUNT); do
3492 start_mdt $num || return
3495 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
3496 start_mdt 1 -o nomgs,force
3497 mount_client $MOUNT || error "mount_client $MOUNT failed"
3500 echo "blah blah" > $MOUNT/$tfile
3503 umount_client $MOUNT || error "umount_client $MOUNT failed"
3504 stop ost1 -f || error "unable to stop OST1"
3505 stop_mds || error "Unable to stop MDS"
3506 stop_mds || error "Unable to stop MDS on second try"
3508 run_test 41a "mount mds with --nosvc and --nomgs"
3511 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3512 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
3513 skip "Loop devices does not work with nosvc option"
3516 ! combined_mgs_mds && skip "needs combined mgs device"
3520 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3522 start_mdt 1 -o nosvc -n
3523 if [ $MDSCOUNT -ge 2 ]; then
3524 for num in $(seq 2 $MDSCOUNT); do
3525 start_mdt $num || return
3528 start_ost || error "Unable to start OST1"
3529 start_mdt 1 -o nomgs,force
3530 mount_client $MOUNT || error "mount_client $MOUNT failed"
3533 echo "blah blah" > $MOUNT/$tfile
3534 cat $MOUNT/$tfile || error "cat $MOUNT/$tfile failed"
3536 umount_client $MOUNT -f || error "umount_client $MOUNT failed"
3537 stop_ost || error "Unable to stop OST1"
3538 stop_mds || error "Unable to stop MDS"
3539 stop_mds || error "Unable to stop MDS on second try"
3541 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
3544 local oss_list=$(comma_list $(osts_nodes))
3546 [[ "$MDS1_VERSION" -ge $(version_code 2.6.52) ]] ||
3547 [[ "$MDS1_VERSION" -ge $(version_code 2.5.26) &&
3548 "$MDS1_VERSION" -lt $(version_code 2.5.50) ]] ||
3549 [[ "$MDS1_VERSION" -ge $(version_code 2.5.4) &&
3550 "$MDS1_VERSION" -lt $(version_code 2.5.11) ]] ||
3551 skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"
3553 # ensure mds1 ost1 have been created even if running sub-test standalone
3556 cleanup || error "cleanup failed"
3558 # using directly mount command instead of start() function to avoid
3559 # any side effect of // with others/externals tools/features
3560 # ("zpool import", ...)
3562 # MDT concurrent start
3564 LOAD_MODULES_REMOTE=true load_modules
3565 do_facet $SINGLEMDS "lsmod | grep -q libcfs" ||
3566 error "MDT concurrent start: libcfs module not loaded"
3568 local mds1dev=$(mdsdevname 1)
3569 local mds1mnt=$(facet_mntpt mds1)
3570 local mds1opts=$MDS_MOUNT_OPTS
3572 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3573 ! do_facet mds1 test -b $mds1dev; then
3574 mds1opts=$(csa_add "$mds1opts" -o loop)
3576 if [[ "$mds1_FSTYPE" == zfs ]]; then
3577 import_zpool mds1 || return ${PIPESTATUS[0]}
3580 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3581 do_facet mds1 "$LCTL set_param fail_loc=0x80000716"
3583 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts &
3586 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts
3591 do_facet mds1 "$LCTL set_param fail_loc=0x0"
3592 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3593 echo "1st MDT start succeed"
3594 echo "2nd MDT start failed with $rc2"
3595 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3596 echo "1st MDT start failed with $rc"
3597 echo "2nd MDT start succeed"
3600 error "unexpected concurrent MDT mounts result, rc=$rc rc2=$rc2"
3603 if [ $MDSCOUNT -ge 2 ]; then
3604 for num in $(seq 2 $MDSCOUNT); do
3605 start_mdt $num || return
3609 # OST concurrent start
3611 do_rpc_nodes $oss_list "lsmod | grep -q libcfs" ||
3612 error "OST concurrent start: libcfs module not loaded"
3614 local ost1dev=$(ostdevname 1)
3615 local ost1mnt=$(facet_mntpt ost1)
3616 local ost1opts=$OST_MOUNT_OPTS
3618 if [ "$ost1_FSTYPE" == ldiskfs ] &&
3619 ! do_facet ost1 test -b $ost1dev; then
3620 ost1opts=$(csa_add "$ost1opts" -o loop)
3622 if [[ "$ost1_FSTYPE" == zfs ]]; then
3623 import_zpool ost1 || return ${PIPESTATUS[0]}
3626 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3627 do_facet ost1 "$LCTL set_param fail_loc=0x80000716"
3629 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts &
3632 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts
3636 do_facet ost1 "$LCTL set_param fail_loc=0x0"
3637 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3638 echo "1st OST start succeed"
3639 echo "2nd OST start failed with $rc2"
3640 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3641 echo "1st OST start failed with $rc"
3642 echo "2nd OST start succeed"
3646 error "unexpected concurrent OST mounts result, rc=$rc rc2=$rc2"
3652 # verify everything ok
3657 error "MDT(s) start failed"
3665 error "OST(s) start failed"
3673 error "client start failed"
3680 error "client mount failed"
3684 run_test 41c "concurrent mounts of MDT/OST should all fail but one"
3686 test_42() { #bug 14693
3690 check_mount || error "client was not mounted"
3692 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3693 PARAM="llite.$FSNAME-*.some_wrong_param"
3695 PARAM="$FSNAME.llite.some_wrong_param"
3698 do_facet mgs $PERM_CMD $PARAM=10
3699 umount_client $MOUNT ||
3700 error "unmounting client failed with invalid llite param"
3701 mount_client $MOUNT ||
3702 error "mounting client failed with invalid llite param"
3704 do_facet mgs $PERM_CMD $PARAM=20
3705 cleanup || error "stopping $FSNAME failed with invalid sys param"
3707 check_mount || error "client was not mounted with invalid sys param"
3708 cleanup || error "stopping $FSNAME failed with invalid sys param"
3710 run_test 42 "allow client/server mount/unmount with invalid config param"
3713 [[ "$MGS_VERSION" -ge $(version_code 2.5.58) ]] ||
3714 skip "Need MDS version at least 2.5.58"
3715 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
3718 USER1=$(getent passwd | grep :$ID1:$ID1: | cut -d: -f1)
3719 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1"
3722 chmod ugo+x $DIR || error "chmod 0 failed"
3723 set_persistent_param_and_check mds1 \
3724 "mdt.$FSNAME-MDT0000.root_squash" \
3725 "$FSNAME.mdt.root_squash" \
3727 wait_update $HOSTNAME \
3728 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3730 error "check llite root_squash failed!"
3731 set_persistent_param_and_check mds1 \
3732 "mdt.$FSNAME-MDT0000.nosquash_nids" \
3733 "$FSNAME.mdt.nosquash_nids" \
3735 wait_update $HOSTNAME \
3736 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3738 error "check llite nosquash_nids failed!"
3741 # create set of test files
3743 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
3744 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
3745 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
3747 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
3748 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
3750 mkdir_on_mdt0 $DIR/$tdir-rootdir || error "mkdir failed"
3751 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
3752 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
3754 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
3755 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
3756 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
3759 # check root_squash:
3760 # set root squash UID:GID to RUNAS_ID
3761 # root should be able to access only files owned by RUNAS_ID
3763 set_persistent_param_and_check mds1 \
3764 "mdt.$FSNAME-MDT0000.root_squash" \
3765 "$FSNAME.mdt.root_squash" \
3766 "$RUNAS_ID:$RUNAS_ID"
3767 wait_update $HOSTNAME \
3768 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3769 "$RUNAS_ID:$RUNAS_ID" ||
3770 error "check llite root_squash failed!"
3772 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
3773 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3774 error "$ST: root read permission is denied"
3775 echo "$ST: root read permission is granted - ok"
3778 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3779 error "$ST: root write permission is denied"
3780 echo "$ST: root write permission is granted - ok"
3782 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3783 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3784 error "$ST: root read permission is granted"
3785 echo "$ST: root read permission is denied - ok"
3788 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3789 error "$ST: root write permission is granted"
3790 echo "$ST: root write permission is denied - ok"
3792 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3793 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null &&
3794 error "$ST: root unlink permission is granted"
3795 echo "$ST: root unlink permission is denied - ok"
3797 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null &&
3798 error "$ST: root create permission is granted"
3799 echo "$ST: root create permission is denied - ok"
3802 # check root_squash is enforced independently
3803 # of client cache content
3805 # access file by USER1, keep access open
3806 # root should be denied access to user file
3808 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
3812 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
3813 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3814 { kill $pid; error "$ST: root read permission is granted"; }
3815 echo "$ST: root read permission is denied - ok"
3818 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3819 { kill $pid; error "$ST: root write permission is granted"; }
3820 echo "$ST: root write permission is denied - ok"
3826 # check nosquash_nids:
3827 # put client's NID into nosquash_nids list,
3828 # root should be able to access root file after that
3830 local NIDLIST=$($LCTL list_nids all | tr '\n' ' ')
3831 NIDLIST="2@gni $NIDLIST 192.168.0.[2,10]@tcp"
3832 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
3833 set_persistent_param_and_check mds1 \
3834 "mdt.$FSNAME-MDT0000.nosquash_nids" \
3835 "$FSNAME-MDTall.mdt.nosquash_nids" \
3837 wait_update $HOSTNAME \
3838 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3840 error "check llite nosquash_nids failed!"
3842 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3843 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3844 error "$ST: root read permission is denied"
3845 echo "$ST: root read permission is granted - ok"
3848 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3849 error "$ST: root write permission is denied"
3850 echo "$ST: root write permission is granted - ok"
3852 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3853 rm $DIR/$tdir-rootdir/tfile-1 ||
3854 error "$ST: root unlink permission is denied"
3855 echo "$ST: root unlink permission is granted - ok"
3856 touch $DIR/$tdir-rootdir/tfile-2 ||
3857 error "$ST: root create permission is denied"
3858 echo "$ST: root create permission is granted - ok"
3859 cleanup || error "cleanup failed with $?"
3861 run_test 43a "check root_squash and nosquash_nids"
3863 test_43b() { # LU-5690
3864 [[ "$MGS_VERSION" -ge $(version_code 2.7.62) ]] ||
3865 skip "Need MGS version 2.7.62+"
3867 if [[ -z "$fs2mds_DEV" ]]; then
3868 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
3869 skip_env "mixed loopback and real device not working"
3872 local fs2mdsdev=$(mdsdevname 1_2)
3873 local fs2mdsvdev=$(mdsvdevname 1_2)
3875 # temporarily use fs2mds as fs2mgs
3877 local fs2mgsdev=$fs2mdsdev
3878 local fs2mgsvdev=$fs2mdsvdev
3880 local fsname=test1234
3882 load_module llite/lustre
3883 local client_ip=$(host_nids_address $HOSTNAME $NETTYPE)
3884 local host=${client_ip//*./}
3885 local net=${client_ip/%$host/}
3886 local nosquash_nids=$(h2nettype $net[$host,$host,$host])
3888 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
3889 --param mdt.root_squash=$RUNAS_ID:$RUNAS_ID \
3890 --param mdt.nosquash_nids=$nosquash_nids \
3891 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
3892 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS || error "start fs2mgs failed"
3893 stop $fs2mgs -f || error "stop fs2mgs failed"
3894 cleanup || error "cleanup failed with $?"
3896 run_test 43b "parse nosquash_nids with commas in expr_list"
3900 check_mount || error "check_mount"
3901 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
3903 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
3904 for VAL in $UUIDS; do
3905 NID=$(echo $VAL | cut -d= -f1)
3906 CLUUID=$(echo $VAL | cut -d= -f2)
3907 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
3909 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
3910 cleanup || error "cleanup failed with $?"
3912 run_test 44 "mounted client proc entry exists"
3916 check_mount || error "check_mount"
3917 stop_mds || error "Unable to stop MDS"
3921 #define OBD_FAIL_PTLRPC_LONG_REPL_UNLINK 0x50f
3922 do_facet client "$LCTL set_param fail_loc=0x8000050f"
3925 manual_umount_client --force || error "manual_umount_client failed"
3926 do_facet client "$LCTL set_param fail_loc=0x0"
3927 start_mds || error "unable to start MDS"
3928 mount_client $MOUNT || error "mount_client $MOUNT failed"
3929 cleanup || error "cleanup failed with $?"
3931 run_test 45 "long unlink handling in ptlrpcd"
3938 umount_client $MOUNT2 || rc=$?
3939 umount_client $MOUNT || rc=$?
3940 while [ $count -gt 0 ]; do
3941 stop ost${count} -f || rc=$?
3945 cleanup_nocli || rc=$?
3946 #writeconf to remove all ost2 traces for subsequent tests
3947 writeconf_or_reformat
3952 echo "Testing with $OSTCOUNT OSTs"
3954 start_mds || error "unable to start MDS"
3955 #first client should see only one ost
3956 start_ost || error "Unable to start OST1"
3957 wait_osc_import_state mds ost FULL
3959 mount_client $MOUNT || error "mount_client $MOUNT failed"
3960 trap "cleanup_46a $OSTCOUNT" EXIT ERR
3963 for (( i=2; i<=$OSTCOUNT; i++ )); do
3964 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
3965 error "start_ost$i $(ostdevname $i) failed"
3968 # wait until osts in sync
3969 for (( i=2; i<=$OSTCOUNT; i++ )); do
3970 wait_osc_import_state mds ost$i FULL
3971 wait_osc_import_ready client ost$i
3974 #second client see all ost's
3976 mount_client $MOUNT2 || error "mount_client failed"
3977 $LFS setstripe -c -1 $MOUNT2 ||
3978 error "$LFS setstripe -c -1 $MOUNT2 failed"
3979 $LFS getstripe $MOUNT2 || error "$LFS getstripe $MOUNT2 failed"
3981 echo "ok" > $MOUNT2/widestripe
3982 $LFS getstripe $MOUNT2/widestripe ||
3983 error "$LFS getstripe $MOUNT2/widestripe failed"
3984 # fill acl buffer for avoid expand lsm to them
3985 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd |
3987 setfacl -m $acl $MOUNT2/widestripe
3991 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3993 cleanup_46a $OSTCOUNT || error "cleanup_46a failed"
3995 run_test 46a "handle ost additional - wide striped file"
4000 check_mount || error "check_mount failed"
4001 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
4005 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
4006 if echo $ns | grep "MDT[[:digit:]]*"; then
4009 lrs=$(echo $ns | sed 's/.*lru_size=//')
4010 lru_size[count]=$lrs
4015 facet_failover $SINGLEMDS
4016 client_up || error "client_up failed"
4019 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
4020 if echo $ns | grep "MDT[[:digit:]]*"; then
4023 lrs=$(echo $ns | sed 's/.*lru_size=//')
4024 if ! test "$lrs" -eq "${lru_size[count]}"; then
4025 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
4026 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
4031 cleanup || error "cleanup failed with $?"
4033 run_test 47 "server restart does not make client loss lru_resize settings"
4038 # reformat after this test is needed - if the test fails,
4039 # we will have unkillable file at FS
4043 test_48() { # bz-17636 LU-7473
4047 check_mount || error "check_mount failed"
4049 $LFS setstripe -c -1 $MOUNT ||
4050 error "$LFS setstripe -c -1 $MOUNT failed"
4051 $LFS getstripe $MOUNT || error "$LFS getstripe $MOUNT failed"
4053 echo "ok" > $MOUNT/widestripe
4054 $LFS getstripe $MOUNT/widestripe ||
4055 error "$LFS getstripe $MOUNT/widestripe failed"
4057 # In the future, we may introduce more EAs, such as selinux, enlarged
4058 # LOV EA, and so on. These EA will use some EA space that is shared by
4059 # ACL entries. So here we only check some reasonable ACL entries count,
4060 # instead of the max number that is calculated from the max_ea_size.
4061 if [ "$MDS1_VERSION" -lt $(version_code 2.8.57) ]; then
4062 count=28 # hard coded of RPC protocol
4063 elif large_xattr_enabled; then
4064 count=4500 # max_num 8187 max_ea_size = 65452
4065 # not create too many (4500) to save test time
4067 count=450 # max_num 497 max_ea_size = 4012
4070 echo "It is expected to hold at least $count ACL entries"
4071 trap cleanup_48 EXIT ERR
4072 for ((i = 0; i < $count; i++)) do
4073 setfacl -m u:$((i + 100)):rw $MOUNT/widestripe ||
4074 error "Fail to setfacl for $MOUNT/widestripe at $i"
4077 cancel_lru_locks mdc
4078 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
4079 local r_count=$(getfacl $MOUNT/widestripe | grep "user:" | wc -l)
4081 count=$((count + 1)) # for the entry "user::rw-"
4083 [ $count -eq $r_count ] ||
4084 error "Expected ACL entries $count, but got $r_count"
4088 run_test 48 "too many acls on file"
4090 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
4091 test_49a() { # bug 17710
4092 local timeout_orig=$TIMEOUT
4093 local ldlm_timeout_orig=$LDLM_TIMEOUT
4094 local LOCAL_TIMEOUT=20
4096 LDLM_TIMEOUT=$LOCAL_TIMEOUT
4097 TIMEOUT=$LOCAL_TIMEOUT
4101 check_mount || error "client mount failed"
4103 echo "check ldlm_timout..."
4104 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
4105 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
4106 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
4108 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
4109 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
4112 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
4113 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
4116 umount_client $MOUNT || error "umount_client $MOUNT failed"
4117 stop_ost || error "problem stopping OSS"
4118 stop_mds || error "problem stopping MDS"
4120 LDLM_TIMEOUT=$ldlm_timeout_orig
4121 TIMEOUT=$timeout_orig
4123 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
4125 test_49b() { # bug 17710
4126 local timeout_orig=$TIMEOUT
4127 local ldlm_timeout_orig=$LDLM_TIMEOUT
4128 local LOCAL_TIMEOUT=20
4130 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
4131 TIMEOUT=$LOCAL_TIMEOUT
4135 check_mount || error "client mount failed"
4137 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
4138 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
4139 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
4141 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
4142 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
4145 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
4146 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
4149 cleanup || error "cleanup failed"
4151 LDLM_TIMEOUT=$ldlm_timeout_orig
4152 TIMEOUT=$timeout_orig
4154 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
4157 # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1
4159 # Test both statfs and lfs df and fail if either one fails
4160 multiop_bg_pause $1 f_
4163 killall -USR1 multiop
4164 [ $RC -ne 0 ] && log "lazystatfs multiop failed"
4165 wait $PID || { RC=$?; log "multiop return error "; }
4167 # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1
4172 if kill -s 0 $PID; then
4175 log "lazystatfs lfs df failed to complete in 5s"
4183 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4184 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4186 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
4188 cleanup || error "cleanup failed with rc $?"
4190 run_test 50a "lazystatfs all servers available"
4194 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4195 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4197 # Wait for client to detect down OST
4198 stop_ost || error "Unable to stop OST1"
4199 wait_osc_import_state client ost DISCONN
4201 log "OSCs should all be DISCONN"
4203 lazystatfs $MOUNT || error "lazystatfs should not return EIO"
4205 umount_client $MOUNT || error "Unable to unmount client"
4206 stop_mds || error "Unable to stop MDS"
4208 run_test 50b "lazystatfs all servers down"
4211 start_mds || error "Unable to start MDS"
4212 start_ost || error "Unable to start OST1"
4213 start_ost2 || error "Unable to start OST2"
4214 mount_client $MOUNT || error "Unable to mount client"
4215 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4216 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4218 # Wait for client to detect down OST
4219 stop_ost || error "Unable to stop OST1"
4220 wait_osc_import_state mds ost DISCONN
4221 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
4223 umount_client $MOUNT || error "Unable to unmount client"
4224 stop_ost2 || error "Unable to stop OST2"
4225 stop_mds || error "Unable to stop MDS"
4226 #writeconf to remove all ost2 traces for subsequent tests
4227 writeconf_or_reformat
4229 run_test 50c "lazystatfs one server down"
4232 start_mds || error "Unable to start MDS"
4233 start_ost || error "Unable to start OST1"
4234 start_ost2 || error "Unable to start OST2"
4235 mount_client $MOUNT || error "Unable to mount client"
4236 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4237 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4239 # Issue the statfs during the window where the client still
4240 # belives the OST to be available but it is in fact down.
4241 # No failure just a statfs which hangs for a timeout interval.
4242 stop_ost || error "Unable to stop OST1"
4243 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
4245 umount_client $MOUNT || error "Unable to unmount client"
4246 stop_ost2 || error "Unable to stop OST2"
4247 stop_mds || error "Unable to stop MDS"
4248 #writeconf to remove all ost2 traces for subsequent tests
4249 writeconf_or_reformat
4251 run_test 50d "lazystatfs client/server conn race"
4258 start_mds || error "Unable to start MDS"
4259 #first client should see only one ost
4260 start_ost || error "Unable to start OST1"
4261 wait_osc_import_state mds ost FULL
4263 # Wait for client to detect down OST
4264 stop_ost || error "Unable to stop OST1"
4265 wait_osc_import_state mds ost DISCONN
4267 mount_client $MOUNT || error "Unable to mount client"
4268 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
4270 multiop_bg_pause $MOUNT _f
4274 if [ $RC1 -ne 0 ]; then
4275 log "multiop failed $RC1"
4278 sleep $(( $TIMEOUT+1 ))
4280 [ $? -ne 0 ] && error "process isn't sleep"
4281 start_ost || error "Unable to start OST1"
4282 wait $pid || error "statfs failed"
4285 umount_client $MOUNT || error "Unable to unmount client"
4286 stop_ost || error "Unable to stop OST1"
4287 stop_mds || error "Unable to stop MDS"
4289 run_test 50e "normal statfs all servers down"
4294 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
4296 start_mds || error "Unable to start mds"
4297 #first client should see only one ost
4298 start_ost || error "Unable to start OST1"
4299 wait_osc_import_state mds ost FULL
4301 start_ost2 || error "Unable to start OST2"
4302 wait_osc_import_state mds ost2 FULL
4304 # Wait for client to detect down OST
4305 stop_ost2 || error "Unable to stop OST2"
4307 wait_osc_import_state mds ost2 DISCONN
4308 mount_client $MOUNT || error "Unable to mount client"
4309 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
4311 multiop_bg_pause $MOUNT _f
4315 if [ $RC1 -ne 0 ]; then
4316 log "lazystatfs multiop failed $RC1"
4319 sleep $(( $TIMEOUT+1 ))
4321 [ $? -ne 0 ] && error "process isn't sleep"
4322 start_ost2 || error "Unable to start OST2"
4323 wait $pid || error "statfs failed"
4324 stop_ost2 || error "Unable to stop OST2"
4327 umount_client $MOUNT -f || error "Unable to unmount client"
4328 stop_ost || error "Unable to stop OST1"
4329 stop_mds || error "Unable to stop MDS"
4330 #writeconf to remove all ost2 traces for subsequent tests
4331 writeconf_or_reformat
4333 run_test 50f "normal statfs one server in down"
4336 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs"
4338 start_ost2 || error "Unable to start OST2"
4339 wait_osc_import_state mds ost2 FULL
4340 wait_osc_import_ready client ost2
4342 if [[ $PERM_CMD == *"set_param -P"* ]]; then
4343 local PARAM="osc.${FSNAME}-OST0001*.active"
4345 local PARAM="${FSNAME}-OST0001.osc.active"
4348 $LFS setstripe -c -1 $DIR/$tfile || error "$LFS setstripe failed"
4349 do_facet mgs $PERM_CMD $PARAM=0 || error "Unable to deactivate OST"
4351 umount_client $MOUNT || error "Unable to unmount client"
4352 mount_client $MOUNT || error "Unable to mount client"
4353 # This df should not cause a panic
4356 do_facet mgs $PERM_CMD $PARAM=1 || error "Unable to activate OST"
4357 rm -f $DIR/$tfile || error "unable to remove file $DIR/$tfile"
4358 umount_client $MOUNT || error "Unable to unmount client"
4359 stop_ost2 || error "Unable to stop OST2"
4360 stop_ost || error "Unable to stop OST1"
4361 stop_mds || error "Unable to stop MDS"
4362 #writeconf to remove all ost2 traces for subsequent tests
4363 writeconf_or_reformat
4365 run_test 50g "deactivated OST should not cause panic"
4369 # prepare MDT/OST, make OSC inactive for OST1
4370 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs"
4372 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
4373 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
4374 error "tunefs OST1 failed"
4375 start_mds || error "Unable to start MDT"
4376 start_ost || error "Unable to start OST1"
4377 start_ost2 || error "Unable to start OST2"
4378 mount_client $MOUNT || error "client start failed"
4380 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4382 # activatate OSC for OST1
4383 set_persistent_param_and_check client \
4384 "osc.${FSNAME}-OST0000-osc-[!M]*.active" \
4385 "${FSNAME}-OST0000.osc.active" 1
4387 mkdir $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
4388 $LFS setstripe -c -1 -i 0 $DIR/$tdir/2 ||
4389 error "$LFS setstripe $DIR/$tdir/2 failed"
4390 sleep 1 && echo "create a file after OST1 is activated"
4391 # doing some io, shouldn't crash
4392 dd if=/dev/zero of=$DIR/$tdir/2/$tfile-io bs=1M count=10
4394 # check OSC import is working
4395 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
4396 error "some OSC imports are still not connected"
4400 umount_client $MOUNT || error "Unable to umount client"
4401 stop_ost2 || error "Unable to stop OST2"
4402 cleanup_nocli || error "cleanup_nocli failed with $?"
4404 run_test 50h "LU-642: activate deactivated OST"
4407 # prepare MDT/OST, make OSC inactive for OST1
4408 [ "$MDSCOUNT" -lt "2" ] && skip "needs >= 2 MDTs"
4411 [ $(facet_fstype mds2) == zfs ] && import_zpool mds2
4412 do_facet mds2 "$TUNEFS --param mdc.active=0 $(mdsdevname 2)" ||
4413 error "tunefs MDT2 failed"
4414 start_mds || error "Unable to start MDT"
4415 start_ost || error "Unable to start OST1"
4416 start_ost2 || error "Unable to start OST2"
4417 mount_client $MOUNT || error "client start failed"
4419 mkdir_on_mdt0 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4421 if [[ $PERM_CMD == *"set_param -P"* ]]; then
4422 $PERM_CMD mdc.${FSNAME}-MDT0001-mdc-*.active=0 &&
4423 error "deactive MDC0 succeeds"
4425 $PERM_CMD ${FSNAME}-MDT0000.mdc.active=0 &&
4426 error "deactive MDC0 succeeds"
4429 # activate MDC for MDT2
4430 set_persistent_param_and_check client \
4431 "mdc.${FSNAME}-MDT0001-mdc-*.active" \
4432 "${FSNAME}-MDT0001.mdc.active" 1
4434 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
4435 if [ "$MDS1_VERSION" -ge $(version_code 2.7.60) ]
4437 wait_dne_interconnect
4439 $LFS mkdir -i1 $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
4441 createmany -o $DIR/$tdir/2/$tfile-%d 1 || error "create files failed"
4443 rm -rf $DIR/$tdir/2 || error "unlink dir failed"
4445 # deactivate MDC for MDT2
4446 set_persistent_param_and_check client \
4447 "mdc.${FSNAME}-MDT0001-mdc-*.active" \
4448 "${FSNAME}-MDT0001.mdc.active" 0
4450 wait_osp_active mds ${FSNAME}-MDT0001 1 0
4452 $LFS mkdir -i1 $DIR/$tdir/2 &&
4453 error "mkdir $DIR/$tdir/2 succeeds after deactive MDT"
4455 $LFS mkdir -i0 -c$MDSCOUNT $DIR/$tdir/striped_dir ||
4456 error "mkdir $DIR/$tdir/striped_dir fails after deactive MDT2"
4458 local stripe_count=$($LFS getdirstripe -c $DIR/$tdir/striped_dir)
4459 [ $stripe_count -eq $((MDSCOUNT - 1)) ] ||
4460 error "wrong $stripe_count != $((MDSCOUNT -1)) for striped_dir"
4463 umount_client $MOUNT || error "Unable to umount client"
4468 run_test 50i "activate deactivated MDT"
4471 local LOCAL_TIMEOUT=20
4475 check_mount || error "check_mount failed"
4477 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
4478 $LFS setstripe -c -1 $MOUNT/$tdir ||
4479 error "$LFS setstripe -c -1 $MOUNT/$tdir failed"
4480 #define OBD_FAIL_MDS_REINT_DELAY 0x142
4481 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x142"
4482 touch $MOUNT/$tdir/$tfile &
4485 start_ost2 || error "Unable to start OST1"
4487 stop_ost2 || error "Unable to stop OST1"
4488 umount_client $MOUNT -f || error “unmount $MOUNT failed”
4489 cleanup_nocli || error “stop server failed”
4490 #writeconf to remove all ost2 traces for subsequent tests
4491 writeconf_or_reformat
4493 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
4502 do_node $node mkdir -p $dest
4503 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
4505 do_node $node 'tar cf - '$*' | tar xf - -C '$dest';
4506 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
4507 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
4509 do_node $node 'getfattr -d -m "[a-z]*\\." '$*' > '$xattrs
4510 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
4520 local backup2=${TMP}/backup2
4522 do_node $node mkdir -p $backup2
4523 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
4525 do_node $node 'tar cf - '$*' | tar xf - -C '$backup2';
4526 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
4527 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
4529 do_node $node "diff -rq $backup $backup2"
4530 [ $? -eq 0 ] || { error "contents differ"; return 3; }
4532 local xattrs2=${TMP}/xattrs2
4533 do_node $node 'getfattr -d -m "[a-z]*\\." '$*' > '$xattrs2
4534 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
4536 do_node $node "diff $xattrs $xattrs2"
4537 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
4539 do_node $node "rm -rf $backup2 $xattrs2"
4540 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
4544 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4545 skip "ldiskfs only test"
4548 start_mds || error "Unable to start MDS"
4549 start_ost || error "Unable to start OST1"
4550 mount_client $MOUNT || error "Unable to mount client"
4553 local ost1mnt=$(facet_mntpt ost1)
4554 local ost1node=$(facet_active_host ost1)
4555 local ost1tmp=$TMP/conf52
4558 mkdir $DIR/$tdir || error "Unable to create $DIR/$tdir"
4559 touch $TMP/modified_first || error "Unable to create temporary file"
4560 local mtime=$(stat -c %Y $TMP/modified_first)
4561 do_node $ost1node "mkdir -p $ost1tmp &&
4562 touch -m -d @$mtime $ost1tmp/modified_first" ||
4563 error "Unable to create temporary file"
4566 $LFS setstripe -c -1 -S 1M $DIR/$tdir || error "$LFS setstripe failed"
4568 for (( i=0; i < nrfiles; i++ )); do
4569 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c ||
4570 error "multiop failed"
4575 # sync all the data and make sure no pending data on the client,
4576 # thus the SOM xattr would not be changed any more.
4577 cancel_lru_locks osc
4580 echo backup files to $TMP/$tdir
4581 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
4582 copy_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
4583 error "Unable to copy files"
4585 umount_client $MOUNT || error "Unable to umount client"
4586 stop_ost || error "Unable to stop ost1"
4588 echo mount ost1 as ldiskfs
4589 do_node $ost1node mkdir -p $ost1mnt || error "Unable to create $ost1mnt"
4590 if ! do_node $ost1node test -b $ost1_dev; then
4593 do_node $ost1node mount -t "$ost1_FSTYPE" $loop $ost1_dev \
4595 error "Unable to mount ost1 as ldiskfs"
4598 echo backup objects to $ost1tmp/objects
4599 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
4600 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
4601 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
4603 error "Unable to copy objects"
4605 # move objects to lost+found
4606 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
4607 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
4609 do_node $ost1node "umount $ost1mnt" ||
4610 error "Unable to umount ost1 as ldiskfs"
4612 start_ost || error "Unable to start OST1"
4613 mount_client $MOUNT || error "Unable to mount client"
4615 local REPAIRED=$(do_node $ost1node "$LCTL get_param \
4616 -n osd-ldiskfs.$FSNAME-OST0000.oi_scrub" |
4617 awk '/^lf_repa[ri]*ed/ { print $2 }')
4618 [ $REPAIRED -gt 0 ] ||
4619 error "Some entry under /lost+found should be repaired"
4622 diff_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
4623 error "Unable to diff files"
4625 rm -rf $TMP/$tdir $TMP/file_xattrs ||
4626 error "Unable to delete temporary files"
4627 do_node $ost1node "rm -rf $ost1tmp" ||
4628 error "Unable to delete temporary files"
4629 cleanup || error "cleanup failed with $?"
4631 run_test 52 "check recovering objects from lost+found"
4633 # Checks threads_min/max/started for some service
4635 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
4636 # parameter pattern prefix like 'ost.*.ost'.
4649 local msg="Insane $modname thread counts"
4650 local ncpts=$(check_cpt_number $facet)
4654 check_mount || return 41
4656 # We need to expand $parampat, but it may match multiple parameters, so
4657 # we'll pick the first one
4658 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
4659 error "Couldn't expand ${parampat}.threads_min parameter name"
4663 # Remove the .threads_min part
4664 paramp=${paramp%.threads_min}
4666 # Check for sanity in defaults
4667 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4669 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4671 tstarted=$(do_facet $facet "$LCTL get_param \
4672 -n ${paramp}.threads_started" || echo 0)
4673 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' ||
4675 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4677 nthrs=$(expr $tmax - $tmin)
4678 if [ $nthrs -lt $ncpts ]; then
4684 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
4685 skip_env "module parameter forced $facet thread count"
4687 # Check that we can change min/max
4688 do_facet $facet "$LCTL set_param \
4689 ${paramp}.threads_min=$((tmin + nthrs))"
4690 do_facet $facet "$LCTL set_param \
4691 ${paramp}.threads_max=$((tmax - nthrs))"
4692 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4694 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4696 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) &&
4697 $tmax2 == ($tmax - $nthrs)))' || return $?
4699 # Check that we can set min/max to the same value
4700 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4702 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$tmin"
4703 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4705 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4707 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
4709 # Check that we can't set max < min
4710 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$((tmin - 1))"
4711 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4713 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4715 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
4717 # We need to ensure that we get the module options desired; to do this
4718 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
4719 LOAD_MODULES_REMOTE=true
4722 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
4724 setmodopts -a $modname "$newvalue" oldvalue
4727 check_mount || return 41
4729 # Restore previous setting of MODOPTS_*
4730 setmodopts $modname "$oldvalue"
4732 # Check that $opts took
4733 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4735 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4737 tstarted=$(do_facet $facet \
4738 "$LCTL get_param -n ${paramp}.threads_started" || echo 0)
4739 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4748 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
4749 cleanup || error "cleanup failed with rc $?"
4751 run_test 53a "check OSS thread count params"
4755 local mds=$(do_facet $SINGLEMDS "$LCTL get_param \
4756 -N mds.*.*.threads_max 2>/dev/null")
4757 if [ -z "$mds" ]; then
4758 #running this on an old MDT
4759 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
4761 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
4763 cleanup || error "cleanup failed with $?"
4765 run_test 53b "check MDS thread count params"
4768 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4769 skip "ldiskfs only test"
4772 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p ||
4773 error "llverdev failed with rc=$?"
4776 run_test 54a "test llverdev and partial verify of device"
4779 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4780 skip "ldiskfs only test"
4784 run_llverfs $MOUNT -p || error "llverfs failed with rc=$?"
4785 cleanup || error "cleanup failed with rc=$?"
4787 run_test 54b "test llverfs and partial verify of filesystem"
4791 local max_ost_index=$1
4793 echo -n $(((max_ost_index + 1) * 8))
4797 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4798 skip "ldiskfs only test"
4801 local mdsdev=$(mdsdevname 1)
4802 local mdsvdev=$(mdsvdevname 1)
4806 if ! combined_mgs_mds; then
4807 stop_mgs || error "stopping MGS service failed"
4808 format_mgs || error "formatting MGT failed"
4810 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
4812 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
4813 --reformat $(ostdevname 1) $(ostvdevname 1)
4819 echo checking size of lov_objid for ost index $i
4820 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" |
4821 grep ^User | awk -F 'Size: ' '{print $2}')
4822 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
4823 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
4825 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
4832 run_test 55 "check lov_objid size"
4835 local mds_journal_size_orig=$MDSJOURNALSIZE
4841 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
4842 $(ostdevname 1) $(ostvdevname 1)
4843 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
4844 $(ostdevname 2) $(ostvdevname 2)
4847 start_ost || error "Unable to start first ost (idx 10000)"
4848 start_ost2 || error "Unable to start second ost (idx 1000)"
4849 mount_client $MOUNT || error "Unable to mount client"
4853 # test instantiating PFL components with sparse index LU-15513
4854 mkdir -p $MOUNT/$tdir
4855 $LFS setstripe -E 4M -c 1 -E 1G -c 4 -S4M -E eof -c -1 $MOUNT/$tdir
4856 dd if=/dev/zero of=$MOUNT/$tdir/$tfile bs=4K count=1 seek=10k ||
4857 error "dd to second component failed"
4859 if [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
4860 [[ "$MDS1_VERSION" -ge $(version_code 2.5.4) &&
4861 "$MDS1_VERSION" -lt $(version_code 2.5.11) ]]; then
4862 wait_osc_import_state mds ost1 FULL
4863 wait_osc_import_state mds ost2 FULL
4864 $LFS setstripe --stripe-count=-1 $DIR/$tfile ||
4865 error "Unable to setstripe $DIR/$tfile"
4866 n=$($LFS getstripe --stripe-count $DIR/$tfile)
4867 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
4872 MDSJOURNALSIZE=$mds_journal_size_orig
4875 run_test 56a "check big OST indexes and out-of-index-order start"
4880 umount_client $MOUNT -f || error "unmount client failed"
4889 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs"
4891 trap cleanup_56b EXIT RETURN ERR
4894 if ! combined_mgs_mds ; then
4899 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) --index=0 --reformat \
4900 $(mdsdevname 1) $(mdsvdevname 1)
4901 add mds2 $(mkfs_opts mds2 $(mdsdevname 2)) --index=1 --reformat \
4902 $(mdsdevname 2) $(mdsvdevname 2)
4903 add mds3 $(mkfs_opts mds3 $(mdsdevname 3)) --index=1000 --reformat \
4904 $(mdsdevname 3) $(mdsvdevname 3)
4908 start_mdt 1 || error "MDT 1 (idx 0) start failed"
4909 start_mdt 2 || error "MDT 2 (idx 1) start failed"
4910 start_mdt 3 || error "MDT 3 (idx 1000) start failed"
4911 start_ost || error "Unable to start first ost"
4912 start_ost2 || error "Unable to start second ost"
4914 do_nodes $(comma_list $(mdts_nodes)) \
4915 "$LCTL set_param mdt.*.enable_remote_dir=1 \
4916 mdt.*.enable_remote_dir_gid=-1"
4918 mount_client $MOUNT || error "Unable to mount client"
4920 $LFS mkdir -c3 $MOUNT/$tdir || error "failed to make testdir"
4922 echo "This is test file 1!" > $MOUNT/$tdir/$tfile.1 ||
4923 error "failed to make test file 1"
4924 echo "This is test file 2!" > $MOUNT/$tdir/$tfile.2 ||
4925 error "failed to make test file 2"
4926 echo "This is test file 1000!" > $MOUNT/$tdir/$tfile.1000 ||
4927 error "failed to make test file 1000"
4929 rm -rf $MOUNT/$tdir || error "failed to remove testdir"
4931 $LFS mkdir -i1000 $MOUNT/$tdir.1000 ||
4932 error "create remote dir at idx 1000 failed"
4935 echo "=== START lfs df OUTPUT ==="
4937 echo "==== END lfs df OUTPUT ===="
4939 mdtcnt=$(echo -e "$output" | grep $FSNAME-MDT | wc -l)
4940 ostcnt=$(echo -e "$output" | grep $FSNAME-OST | wc -l)
4942 echo "lfs df returned mdt count $mdtcnt and ost count $ostcnt"
4943 [ $mdtcnt -eq 3 ] || error "lfs df returned wrong mdt count"
4944 [ $ostcnt -eq 2 ] || error "lfs df returned wrong ost count"
4946 echo "This is test file 1!" > $MOUNT/$tdir.1000/$tfile.1 ||
4947 error "failed to make test file 1"
4948 echo "This is test file 2!" > $MOUNT/$tdir.1000/$tfile.2 ||
4949 error "failed to make test file 2"
4950 echo "This is test file 1000!" > $MOUNT/$tdir.1000/$tfile.1000 ||
4951 error "failed to make test file 1000"
4952 rm -rf $MOUNT/$tdir.1000 || error "failed to remove remote_dir"
4955 echo "=== START lfs mdts OUTPUT ==="
4957 echo "==== END lfs mdts OUTPUT ===="
4959 echo -e "$output" | grep -v "MDTS:" | awk '{print $1}' |
4960 sed 's/://g' > $TMP/mdts-actual.txt
4961 sort $TMP/mdts-actual.txt -o $TMP/mdts-actual.txt
4963 echo -e "0\n1\n1000" > $TMP/mdts-expected.txt
4965 diff $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4968 rm $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4970 [ $result -eq 0 ] || error "target_obd proc file is incorrect!"
4972 run_test 56b "test target_obd correctness with nonconsecutive MDTs"
4974 test_57a() { # bug 22656
4975 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4976 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4977 tail -1 | awk '{print $1}')
4978 writeconf_or_reformat
4979 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
4980 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" ||
4981 error "tunefs failed"
4983 start_ost && error "OST registration from failnode should fail"
4986 run_test 57a "initial registration from failnode should fail (should return errs)"
4989 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4990 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4991 tail -1 | awk '{print $1}')
4992 writeconf_or_reformat
4993 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
4994 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" ||
4995 error "tunefs failed"
4997 start_ost || error "OST registration from servicenode should not fail"
5000 run_test 57b "initial registration from servicenode should not fail"
5003 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
5006 test_58() { # bug 22658
5007 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
5009 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5010 createmany -o $DIR/$tdir/$tfile-%d 100
5011 unlinkmany $DIR/$tdir/$tfile-%d 100
5012 stop_mds || error "Unable to stop MDS"
5014 local MNTDIR=$(facet_mntpt $SINGLEMDS)
5015 local devname=$(mdsdevname ${SINGLEMDS//mds/})
5017 # remove all files from the OBJECTS dir
5018 mount_fstype $SINGLEMDS
5020 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
5022 unmount_fstype $SINGLEMDS
5023 # restart MDS with missing llog files
5024 start_mds || error "unable to start MDS"
5025 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0"
5028 run_test 58 "missing llog files must not prevent MDT from mounting"
5031 start_mgsmds >> /dev/null
5032 local C1=$(count_osts)
5033 if [ $C1 -eq 0 ]; then
5034 start_ost >> /dev/null
5038 echo "original ost count: $C1 (expect > 0)"
5039 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
5040 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
5041 local C2=$(count_osts)
5042 echo "after mdt writeconf count: $C2 (expect 0)"
5043 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
5044 echo "OST start without writeconf should fail:"
5045 start_ost >> /dev/null &&
5046 error "OST start without writeconf didn't fail"
5047 echo "OST start with writeconf should succeed:"
5048 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
5049 local C3=$(count_osts)
5050 echo "after ost writeconf count: $C3 (expect 1)"
5051 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
5052 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
5053 local C4=$(count_osts)
5054 echo "after ost2 writeconf count: $C4 (expect 2)"
5055 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
5056 stop_ost2 >> /dev/null
5057 cleanup_nocli >> /dev/null
5058 #writeconf to remove all ost2 traces for subsequent tests
5059 writeconf_or_reformat
5061 run_test 59 "writeconf mount option"
5063 test_60a() { # LU-471
5064 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5065 skip "ldiskfs only test"
5070 for num in $(seq $MDSCOUNT); do
5071 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
5072 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
5073 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
5077 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
5078 [ ${PIPESTATUS[0]} -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
5080 # MDT default has dirdata feature
5081 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
5082 # we disable uninit_bg feature
5083 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
5084 # we set stride extended options
5085 echo $dump | grep stride > /dev/null || error "stride is not set"
5089 run_test 60a "check mkfs.lustre --mkfsoptions -E -O options setting"
5092 [[ "$mds1_FSTYPE" == ldiskfs ]] || skip "ldiskfs only test"
5094 local features=$(do_facet $SINGLEMDS $DUMPE2FS $(mdsdevname 1) |
5096 [ ${PIPESTATUS[0]} -eq 0 ] || error "$DUMPE2FS $(mdsdevname 1) failed"
5099 # ea_inode feature should be enabled by default for MDTs
5100 [[ "$features" =~ "ea_inode" ]] || error "ea_inode is not set"
5101 # large_dir feature should be enabled by default for MDTs
5102 [[ "$features" =~ "large_dir" ]] || error "large_dir is not set"
5104 run_test 60b "check mkfs.lustre MDT default features"
5106 test_61a() { # LU-80
5107 local lxattr=$(large_xattr_enabled)
5109 (( "$MDS1_VERSION" >= $(version_code 2.1.53) )) ||
5110 skip "Need MDS version at least 2.1.53 for large_xattr"
5112 if [[ "$mds1_FSTYPE" == ldiskfs ]] && ! large_xattr_enabled; then
5115 for ((num=1; num <= $MDSCOUNT; num++)); do
5116 do_facet mds${num} $TUNE2FS -O ea_inode \
5117 $(mdsdevname $num) ||
5118 error "tune2fs on mds $num failed"
5122 setup || error "setting up the filesystem failed"
5123 client_up || error "starting client failed"
5125 local file=$DIR/$tfile
5126 touch $file || error "touch $file failed"
5128 local large_value="$(generate_string $(max_xattr_size))"
5129 local small_value="bar"
5131 local name="trusted.big"
5132 log "save large xattr of $(max_xattr_size) bytes on $name on $file"
5133 setfattr -n $name -v $large_value $file ||
5134 error "saving $name on $file failed"
5136 local new_value=$(get_xattr_value $name $file)
5137 [[ "$new_value" != "$large_value" ]] &&
5138 error "$name different after saving"
5140 log "shrink value of $name on $file"
5141 setfattr -n $name -v $small_value $file ||
5142 error "shrinking value of $name on $file failed"
5144 new_value=$(get_xattr_value $name $file)
5145 [[ "$new_value" != "$small_value" ]] &&
5146 error "$name different after shrinking"
5148 log "grow value of $name on $file"
5149 setfattr -n $name -v $large_value $file ||
5150 error "growing value of $name on $file failed"
5152 new_value=$(get_xattr_value $name $file)
5153 [[ "$new_value" != "$large_value" ]] &&
5154 error "$name different after growing"
5156 log "check value of $name on $file after remounting MDS"
5158 new_value=$(get_xattr_value $name $file)
5159 [[ "$new_value" != "$large_value" ]] &&
5160 error "$name different after remounting MDS"
5162 log "remove large xattr $name from $file"
5163 setfattr -x $name $file || error "removing $name from $file failed"
5165 if $lxattr && [ "$mds1_FSTYPE" == ldiskfs ]; then
5166 stopall || error "stopping for e2fsck run"
5167 for num in $(seq $MDSCOUNT); do
5168 run_e2fsck $(facet_active_host mds$num) \
5169 $(mdsdevname $num) "-y" ||
5170 error "e2fsck MDT$num failed"
5172 setup_noconfig || error "remounting the filesystem failed"
5175 # need to delete this file to avoid problems in other tests
5177 cleanup || error "stopping systems failed"
5179 run_test 61a "large xattr"
5181 test_61b() { # LU-80
5182 local lxattr=$(large_xattr_enabled)
5184 (( $MDS1_VERSION >= $(version_code 2.15.51) )) ||
5185 skip "Need MDS version at least 2.15.51 for large_xattr fix"
5187 [[ "$mds1_FSTYPE" == "ldiskfs" ]] || skip "ldiskfs specific bug"
5189 if ! large_xattr_enabled; then
5192 for (( num=1; num <= $MDSCOUNT; num++ )); do
5193 do_facet mds${num} $TUNE2FS -O ea_inode \
5194 $(mdsdevname $num) ||
5195 error "tune2fs on mds $num failed"
5199 setup || error "setting up the filesystem failed"
5200 client_up || error "starting client failed"
5202 local _file=$MOUNT/panda
5203 local large_value="$(generate_string $(max_xattr_size))"
5204 local name="trusted.big"
5206 touch ${_file} || error "touch ${_file} failed"
5207 setfattr -n $name -v $large_value ${_file} ||
5208 error "saving $name on $file failed"
5210 MDT_DEV="${FSNAME}-MDT0000"
5211 MDT_DEVNAME=$(mdsdevname ${SINGLEMDS//mds/})
5213 stopall || error "stopping for e2fsck run"
5216 ino=$(do_facet $SINGLEMDS "$DEBUGFS -R 'stat /ROOT/panda' \
5217 ${MDT_DEVNAME} | grep trusted.big")
5218 ino=$(echo "${ino}" | awk '{print $2;}')
5219 echo "large ea "${ino}
5221 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"ln $ino /lost+found\\\" \
5224 setup_noconfig || error "remounting the filesystem failed"
5226 do_facet $SINGLEMDS $LCTL lfsck_start -M ${MDT_DEV} -t namespace || {
5227 error "can't start lfsck namespace"
5231 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
5232 mdd.${MDT_DEV}.lfsck_namespace |
5233 awk '/^status/ { print \\\$2 }'" "completed" 32 || {
5234 error "(2) unexpected status"
5237 stopall || error "stopping for e2fsck run"
5238 for num in $(seq $MDSCOUNT); do
5239 run_e2fsck $(facet_active_host mds$num) \
5240 $(mdsdevname $num) "-y" ||
5241 error "e2fsck MDT$num failed"
5243 setup_noconfig || error "remounting the filesystem failed"
5245 # need to delete this file to avoid problems in other tests
5247 cleanup || error "stopping systems failed"
5249 run_test 61b "large xattr"
5252 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5253 skip "ldiskfs only test"
5255 [[ "$MDS1_VERSION" -ge $(version_code 2.2.51) ]] ||
5256 skip "Need MDS version at least 2.2.51"
5259 local mdsdev=$(mdsdevname 1)
5260 local ostdev=$(ostdevname 1)
5262 echo "disable journal for mds"
5263 do_facet mds1 $TUNE2FS -O ^has_journal $mdsdev || error "tune2fs failed"
5264 start_mds && error "MDT start should fail"
5265 echo "disable journal for ost"
5266 do_facet ost1 $TUNE2FS -O ^has_journal $ostdev || error "tune2fs failed"
5267 start_ost && error "OST start should fail"
5268 cleanup || error "cleanup failed with rc $?"
5271 run_test 62 "start with disabled journal"
5274 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5275 skip "ldiskfs only test"
5278 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ldiskfs
5279 local inode_slab=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
5280 awk '/ldiskfs_inode_cache/ { print $5 / $6 }')
5281 if [ -z "$inode_slab" ]; then
5282 skip "ldiskfs module has not been loaded"
5285 if grep -q "CONFIG_DEBUG_LOCK_ALLOC=y" /boot/config-$(uname -r); then
5286 skip "test is not compatible with CONFIG_DEBUG_LOCK_ALLOC=y"
5289 echo "$inode_slab ldiskfs inodes per page"
5290 [ "${inode_slab%.*}" -ge "3" ] && return 0
5292 # If kmalloc-128 is also 1 per page - this is a debug kernel
5293 # and so this is not an error.
5294 local kmalloc128=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
5295 awk '/^(dma-kmalloc|size)-128 / { print $5 / $6 }')
5296 # 32 128-byte chunks in 4k
5297 [ "${kmalloc128%.*}" -lt "32" ] ||
5298 error "ldiskfs inode too big, only $inode_slab objs/page, " \
5299 "kmalloc128 = $kmalloc128 objs/page"
5301 run_test 63 "Verify each page can at least hold 3 ldiskfs inodes"
5304 start_mds || error "unable to start MDS"
5305 start_ost || error "Unable to start OST1"
5306 start_ost2 || error "Unable to start second ost"
5307 mount_client $MOUNT || error "Unable to mount client"
5308 stop_ost2 || error "Unable to stop second ost"
5311 umount_client $MOUNT -f || error “unmount $MOUNT failed”
5312 cleanup_nocli || error "cleanup_nocli failed with $?"
5313 #writeconf to remove all ost2 traces for subsequent tests
5314 writeconf_or_reformat
5316 run_test 64 "check lfs df --lazy "
5318 test_65() { # LU-2237
5319 # Currently, the test is only valid for ldiskfs backend
5320 [ "$mds1_FSTYPE" != ldiskfs ] &&
5321 skip "ldiskfs only test"
5323 local devname=$(mdsdevname ${SINGLEMDS//mds/})
5324 local brpt=$(facet_mntpt brpt)
5327 if ! do_facet $SINGLEMDS "test -b $devname"; then
5331 stop_mds || error "Unable to stop MDS"
5332 local obj=$(do_facet $SINGLEMDS \
5333 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
5335 if [ -z "$obj" ]; then
5336 # The MDT may be just re-formatted, mount the MDT for the
5337 # first time to guarantee the "last_rcvd" file is there.
5338 start_mds || error "fail to mount the MDS for the first time"
5339 stop_mds || error "Unable to stop MDS"
5342 # remove the "last_rcvd" file
5343 do_facet $SINGLEMDS "mkdir -p $brpt"
5344 do_facet $SINGLEMDS \
5345 "mount -t $mds1_FSTYPE $opts $devname $brpt"
5346 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
5347 do_facet $SINGLEMDS "$UMOUNT $brpt"
5349 # restart MDS, the "last_rcvd" file should be recreated.
5350 start_mds || error "fail to restart the MDS"
5351 stop_mds || error "Unable to stop MDS"
5352 obj=$(do_facet $SINGLEMDS \
5353 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
5354 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
5356 run_test 65 "re-create the lost last_rcvd file when server mount"
5359 [[ "$MGS_VERSION" -ge $(version_code 2.3.59) ]] ||
5360 skip "Need MGS version at least 2.3.59"
5363 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
5364 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
5366 # add EXCLUDE records to config log, they are not to be
5367 # removed by lctl replace_nids
5368 set_conf_param_and_check mds \
5369 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
5370 "$FSNAME-OST0000.osc.active" \
5373 echo "replace_nids should fail if MDS, OSTs and clients are UP"
5374 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5375 error "replace_nids fail"
5377 umount_client $MOUNT || error "unmounting client failed"
5378 echo "replace_nids should fail if MDS and OSTs are UP"
5379 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5380 error "replace_nids fail"
5382 stop_ost || error "Unable to stop OST1"
5383 echo "replace_nids should fail if MDS is UP"
5384 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5385 error "replace_nids fail"
5387 stop_mds || error "stopping mds failed"
5389 if combined_mgs_mds; then
5390 start_mdt 1 "-o nosvc" ||
5391 error "starting mds with nosvc option failed"
5394 echo "command should accept two parameters"
5395 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
5396 error "command should accept two params"
5398 echo "correct device name should be passed"
5399 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
5400 error "wrong devname"
5402 echo "wrong nids list should not destroy the system"
5403 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
5405 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "asdfasdf, asdfadf" &&
5408 echo "replace OST nid"
5409 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
5410 error "replace nids failed"
5412 echo "command should accept two parameters"
5413 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
5414 error "command should accept two params"
5416 echo "wrong nids list should not destroy the system"
5417 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
5420 local FAKE_NIDS="192.168.0.112@tcp1,192.168.0.112@tcp2"
5421 local FAKE_FAILOVER="192.168.0.113@tcp1,192.168.0.113@tcp2"
5422 local NIDS_AND_FAILOVER="$MDS_NID,$FAKE_NIDS:$FAKE_FAILOVER"
5423 echo "set NIDs with failover"
5424 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $NIDS_AND_FAILOVER ||
5425 error "replace nids failed"
5428 echo "replace MDS nid"
5429 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
5430 error "replace nids failed"
5432 if ! combined_mgs_mds ; then
5435 stop_mds || error "Unable to stop MDS"
5438 start_mgsmds || error "start mgsmds failed"
5439 set_conf_param_and_check mds \
5440 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
5441 "$FSNAME-OST0000.osc.active" \
5443 start_ost || error "unable to start OST"
5444 mount_client $MOUNT || error "mount client failed"
5446 check_mount || error "error after nid replace"
5447 cleanup || error "cleanup failed"
5450 run_test 66 "replace nids"
5452 test_67() { #LU-2950
5453 local legacy="$TMP/legacy_lnet_config"
5454 local new="$TMP/new_routes_test"
5455 local out="$TMP/config_out_file"
5456 local verify="$TMP/conv_verify"
5457 local verify_conf="$TMP/conf_verify"
5459 # Create the legacy file that will be run through the
5460 # lustre_routes_conversion script
5461 cat <<- LEGACY_LNET_CONFIG > $legacy
5462 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
5463 tcp2 54 10.1.3.2@tcp;
5464 tcp3 10.3.4.3@tcp:3;
5468 # Create the verification file to verify the output of
5469 # lustre_routes_conversion script against.
5470 cat <<- VERIFY_LNET_CONFIG > $verify
5471 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
5472 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
5473 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
5474 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
5475 tcp4: { gateway: 10.3.3.4@tcp }
5478 # Create the verification file to verify the output of
5479 # lustre_routes_config script against
5480 cat <<- VERIFY_LNET_CONFIG > $verify_conf
5481 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
5482 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
5483 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
5484 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
5485 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
5488 $LUSTRE_ROUTES_CONVERSION $legacy $new > /dev/null
5489 if [ -f $new ]; then
5490 # verify the conversion output
5491 cmp -s $new $verify > /dev/null
5492 if [ $? -eq 1 ]; then
5493 error "routes conversion failed"
5496 lustre_routes_config --dry-run --verbose $new > $out
5497 # check that the script succeeded
5498 cmp -s $out $verify_conf > /dev/null
5499 if [ $? -eq 1 ]; then
5500 error "routes config failed"
5503 error "routes conversion test failed"
5505 # remove generated files
5506 rm -f $new $legacy $verify $verify_conf $out
5508 run_test 67 "test routes conversion and configuration"
5516 [ "$MDS1_VERSION" -ge $(version_code 2.4.53) ] ||
5517 skip "Need MDS version at least 2.4.53"
5519 umount_client $MOUNT || error "umount client failed"
5524 # START-END - the sequences we'll be reserving
5525 START=$(do_facet $SINGLEMDS \
5526 $LCTL get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
5527 END=$((START + (1 << 30)))
5528 do_facet $SINGLEMDS \
5529 $LCTL set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
5531 # reset the sequences MDT0000 has already assigned
5532 do_facet $SINGLEMDS \
5533 $LCTL set_param seq.srv*MDT0000.space=clear
5535 # remount to let the client allocate new sequence
5536 mount_client $MOUNT || error "mount client failed"
5538 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
5539 do_facet $SINGLEMDS \
5540 $LCTL get_param seq.srv*MDT0000.space
5541 $LFS path2fid $DIR/$tfile
5543 local old_ifs="$IFS"
5545 fid=($($LFS path2fid $DIR/$tfile))
5549 if [[ $seq < $END ]]; then
5550 error "used reserved sequence $seq?"
5552 cleanup || error "cleanup failed with $?"
5554 run_test 68 "be able to reserve specific sequences in FLDB"
5556 # Test 69: is about the total number of objects ever created on an OST.
5557 # so that when it is reformatted the normal MDS->OST orphan recovery won't
5558 # just "precreate" the missing objects. In the past it might try to recreate
5559 # millions of objects after an OST was reformatted
5561 [[ "$MDS1_VERSION" -lt $(version_code 2.4.2) ]] &&
5562 skip "Need MDS version at least 2.4.2"
5564 [[ "$MDS1_VERSION" -ge $(version_code 2.4.50) ]] &&
5565 [[ "$MDS1_VERSION" -lt $(version_code 2.5.0) ]] &&
5566 skip "Need MDS version at least 2.5.0"
5569 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5570 do_nodes $(comma_list $(osts_nodes)) $LCTL set_param \
5571 seq.*OST*-super.width=$DATA_SEQ_MAX_WIDTH
5573 # use OST0000 since it probably has the most creations
5574 local OSTNAME=$(ostname_from_index 0)
5575 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
5576 local last_id=$(do_facet mds1 $LCTL get_param -n \
5577 osp.$mdtosc_proc1.prealloc_last_id)
5579 # Want to have OST LAST_ID over 5 * OST_MAX_PRECREATE to
5580 # verify that the LAST_ID recovery is working properly. If
5581 # not, then the OST will refuse to allow the MDS connect
5582 # because the LAST_ID value is too different from the MDS
5583 #define OST_MAX_PRECREATE=20000
5584 local ost_max_pre=20000
5585 local num_create=$(( ost_max_pre * 5 + 1 - last_id))
5587 # If the LAST_ID is already over 5 * OST_MAX_PRECREATE, we don't
5588 # need to create any files. So, skip this section.
5589 if [ $num_create -gt 0 ]; then
5590 # Check the number of inodes available on OST0
5592 local ifree=$($LFS df -i $MOUNT |
5593 awk '/OST0000/ { print $4 }'; exit ${PIPESTATUS[0]})
5594 log "On OST0, $ifree inodes available. Want $num_create. rc=$?"
5596 $LFS setstripe -i 0 $DIR/$tdir ||
5597 error "$LFS setstripe -i 0 $DIR/$tdir failed"
5598 if [ $ifree -lt 10000 ]; then
5599 files=$(( ifree - 50 ))
5604 local j=$((num_create / files + 1))
5605 for i in $(seq 1 $j); do
5606 createmany -o $DIR/$tdir/$tfile-$i- $files ||
5607 error "createmany fail create $files files: $?"
5608 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
5609 error "unlinkmany failed unlink $files files"
5613 # delete all of the files with objects on OST0 so the
5614 # filesystem is not inconsistent later on
5615 $LFS find $MOUNT --ost 0 -print0 | xargs -0 rm
5617 umount_client $MOUNT || error "umount client failed"
5618 stop_ost || error "OST0 stop failure"
5619 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
5620 $(ostdevname 1) $(ostvdevname 1) ||
5621 error "reformat and replace $ostdev failed"
5622 start_ost || error "OST0 restart failure"
5623 wait_osc_import_state mds ost FULL
5625 mount_client $MOUNT || error "mount client failed"
5626 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
5627 local idx=$($LFS getstripe -i $DIR/$tdir/$tfile-last)
5628 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
5630 local iused=$($LFS df -i $MOUNT |
5631 awk '/OST0000/ { print $3 }'; exit ${PIPESTATUS[0]})
5632 log "On OST0, $iused used inodes rc=$?"
5633 [ $iused -ge $((ost_max_pre + 1000)) ] &&
5634 error "OST replacement created too many inodes; $iused"
5635 cleanup || error "cleanup failed with $?"
5637 run_test 69 "replace an OST with the same index"
5640 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5643 cleanup || error "cleanup failed with $?"
5645 start_mdt 1 || error "MDT0 start fail"
5647 start_ost || error "OST0 start fail"
5648 for num in $(seq 2 $MDSCOUNT); do
5649 start_mdt $num || return
5652 mount_client $MOUNT || error "mount client fails"
5654 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
5656 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5657 error "create remote dir fail"
5659 rm -rf $DIR/$tdir || error "delete dir fail"
5660 cleanup || error "cleanup failed with $?"
5662 run_test 70a "start MDT0, then OST, then MDT1"
5665 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5668 start_ost || error "OST0 start fail"
5670 start_mds || error "MDS start fail"
5672 mount_client $MOUNT || error "mount client fails"
5674 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
5676 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5677 error "create remote dir fail"
5679 rm -rf $DIR/$tdir || error "delete dir fail"
5681 cleanup || error "cleanup failed with $?"
5683 run_test 70b "start OST, MDT1, MDT0"
5686 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5689 start_mds || error "MDS start fail"
5690 start_ost || error "OST0 start fail"
5692 mount_client $MOUNT || error "mount client fails"
5693 stop_mdt 1 || error "MDT1 start fail"
5695 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
5696 echo "deactivate $mdc_for_mdt1"
5697 $LCTL --device $mdc_for_mdt1 deactivate ||
5698 error "set $mdc_for_mdt1 deactivate failed"
5700 mkdir $DIR/$tdir && error "mkdir succeed"
5702 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
5703 error "create remote dir succeed"
5705 cleanup || error "cleanup failed with $?"
5707 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
5710 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5713 start_mds || error "MDS start fail"
5714 start_ost || error "OST0 start fail"
5716 mount_client $MOUNT || error "mount client fails"
5718 stop_mdt 2 || error "MDT1 start fail"
5720 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
5722 echo "deactivate $mdc_for_mdt2"
5723 $LCTL --device $mdc_for_mdt2 deactivate ||
5724 error "set $mdc_for_mdt2 deactivate failed"
5726 mkdir $DIR/$tdir || error "mkdir fail"
5727 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
5728 error "create remote dir succeed"
5730 rm -rf $DIR/$tdir || error "delete dir fail"
5732 cleanup || error "cleanup failed with $?"
5734 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
5737 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5739 [ "$MDS1_VERSION" -ge $(version_code 2.7.62) ] ||
5740 skip "Need MDS version at least 2.7.62"
5742 reformat || error "reformat failed with $?"
5746 local mdsdev=$(mdsdevname 1)
5747 local ostdev=$(ostdevname 1)
5748 local mdsvdev=$(mdsvdevname 1)
5749 local ostvdev=$(ostvdevname 1)
5750 local opts_mds="$(mkfs_opts mds1 $mdsdev) --reformat $mdsdev $mdsvdev"
5751 local opts_ost="$(mkfs_opts ost1 $ostdev) --reformat $ostdev $ostvdev"
5753 if ! combined_mgs_mds ; then
5757 add mds1 $opts_mds || error "add mds1 failed"
5758 start_mdt 1 || error "start mdt1 failed"
5759 add ost1 $opts_ost || error "add ost1 failed"
5760 start_ost || error "start ost failed"
5761 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
5763 local soc=$(do_facet mds1 "$LCTL get_param -n \
5764 mdt.*MDT0000.sync_lock_cancel")
5765 [ $soc == "never" ] || error "SoC enabled on single MDS"
5767 for i in $(seq 2 $MDSCOUNT); do
5768 mdsdev=$(mdsdevname $i)
5769 mdsvdev=$(mdsvdevname $i)
5770 opts_mds="$(mkfs_opts mds$i $mdsdev) --reformat $mdsdev \
5772 add mds$i $opts_mds || error "add mds$i failed"
5773 start_mdt $i || error "start mdt$i fail"
5776 wait_dne_interconnect
5778 for i in $(seq $MDSCOUNT); do
5779 soc=$(do_facet mds$i "$LCTL get_param -n \
5780 mdt.*MDT000$((i - 1)).sync_lock_cancel")
5781 [ $soc == "blocking" ] || error "SoC not enabled on DNE"
5784 for i in $(seq 2 $MDSCOUNT); do
5785 stop_mdt $i || error "stop mdt$i fail"
5787 soc=$(do_facet mds1 "$LCTL get_param -n \
5788 mdt.*MDT0000.sync_lock_cancel")
5789 [ $soc == "never" ] || error "SoC enabled on single MDS"
5790 umount_client $MOUNT -f > /dev/null
5792 cleanup || error "cleanup failed with $?"
5794 run_test 70e "Sync-on-Cancel will be enabled by default on DNE"
5797 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5798 if combined_mgs_mds; then
5799 skip "needs separate MGS/MDT"
5803 start_mdt 1 || error "MDT0 start fail"
5804 start_ost || error "OST0 start fail"
5805 for num in $(seq 2 $MDSCOUNT); do
5806 start_mdt $num || return
5809 start_ost2 || error "OST1 start fail"
5811 mount_client $MOUNT || error "mount client fails"
5813 mkdir $DIR/$tdir || error "mkdir fail"
5814 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5815 error "create remote dir succeed"
5817 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5818 rm -rf $DIR/$tdir || error "delete dir fail"
5820 umount_client $MOUNT || error "umount_client failed"
5821 stop_mds || error "MDS stop fail"
5822 stop_ost || error "OST0 stop fail"
5823 stop_ost2 || error "OST1 stop fail"
5825 run_test 71a "start MDT0 OST0, MDT1, OST1"
5828 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5829 if combined_mgs_mds; then
5830 skip "needs separate MGS/MDT"
5834 for num in $(seq 2 $MDSCOUNT); do
5835 start_mdt $num || return
5837 start_ost || error "OST0 start fail"
5838 start_mdt 1 || error "MDT0 start fail"
5839 start_ost2 || error "OST1 start fail"
5841 mount_client $MOUNT || error "mount client fails"
5843 mkdir $DIR/$tdir || error "mkdir fail"
5844 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5845 error "create remote dir succeed"
5847 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5848 rm -rf $DIR/$tdir || error "delete dir fail"
5850 umount_client $MOUNT || error "umount_client failed"
5851 stop_mds || error "MDT0 stop fail"
5852 stop_ost || error "OST0 stop fail"
5853 stop_ost2 || error "OST1 stop fail"
5855 run_test 71b "start MDT1, OST0, MDT0, OST1"
5858 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5859 combined_mgs_mds && skip "needs separate MGS/MDT"
5863 start_ost || error "OST0 start fail"
5864 start_ost2 || error "OST1 start fail"
5865 for num in $(seq 2 $MDSCOUNT); do
5866 start_mdt $num || return
5868 start_mdt 1 || error "MDT0 start fail"
5870 mount_client $MOUNT || error "mount client fails"
5872 mkdir $DIR/$tdir || error "mkdir fail"
5873 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5874 error "create remote dir succeed"
5876 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5877 rm -rf $DIR/$tdir || error "delete dir fail"
5879 umount_client $MOUNT || error "umount_client failed"
5880 stop_mds || error "MDS stop fail"
5881 stop_ost || error "OST0 stop fail"
5882 stop_ost2 || error "OST1 stop fail"
5885 run_test 71c "start OST0, OST1, MDT1, MDT0"
5888 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5889 combined_mgs_mds && skip "needs separate MGS/MDT"
5893 start_ost || error "OST0 start fail"
5894 for num in $(seq 2 $MDSCOUNT); do
5895 start_mdt $num || return
5897 start_mdt 1 || error "MDT0 start fail"
5898 start_ost2 || error "OST1 start fail"
5900 mount_client $MOUNT || error "mount client fails"
5902 mkdir $DIR/$tdir || error "mkdir fail"
5903 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5904 error "create remote dir succeed"
5906 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5907 rm -rf $DIR/$tdir || error "delete dir fail"
5909 umount_client $MOUNT || error "umount_client failed"
5910 stop_mds || error "MDS stop fail"
5911 stop_ost || error "OST0 stop fail"
5912 stop_ost2 || error "OST1 stop fail"
5915 run_test 71d "start OST0, MDT1, MDT0, OST1"
5918 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5919 combined_mgs_mds && skip "needs separate MGS/MDT"
5923 start_ost || error "OST0 start fail"
5924 for num in $(seq 2 $MDSCOUNT); do
5925 start_mdt $num || return
5927 start_ost2 || error "OST1 start fail"
5928 start_mdt 1 || error "MDT0 start fail"
5930 mount_client $MOUNT || error "mount client fails"
5932 mkdir $DIR/$tdir || error "mkdir fail"
5933 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5934 error "create remote dir succeed"
5936 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5937 rm -rf $DIR/$tdir || error "delete dir fail"
5939 umount_client $MOUNT || error "umount_client failed"
5940 stop_mds || error "MDS stop fail"
5941 stop_ost || error "OST0 stop fail"
5942 stop_ost2 || error "OST1 stop fail"
5945 run_test 71e "start OST0, MDT1, OST1, MDT0"
5947 test_72() { #LU-2634
5948 [ "$mds1_FSTYPE" != ldiskfs ] &&
5949 skip "ldiskfs only test"
5951 local mdsdev=$(mdsdevname 1)
5952 local ostdev=$(ostdevname 1)
5953 local cmd="$E2FSCK -fnvd $mdsdev"
5960 if combined_mgs_mds; then
5961 add_options='--reformat'
5963 add_options='--reformat --replace'
5966 #tune MDT with "-O extents"
5968 for num in $(seq $MDSCOUNT); do
5969 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
5970 $add_options $(mdsdevname $num) $(mdsvdevname $num) ||
5971 error "add mds $num failed"
5972 do_facet mds${num} "$TUNE2FS -O extents $(mdsdevname $num)" ||
5973 error "$TUNE2FS failed on mds${num}"
5976 add ost1 $(mkfs_opts ost1 $ostdev) $add_options $ostdev ||
5977 error "add $ostdev failed"
5978 start_mds || error "start mds failed"
5979 start_ost || error "start ost failed"
5980 mount_client $MOUNT || error "mount client failed"
5982 #create some short symlinks
5983 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5984 createmany -o $DIR/$tdir/$tfile-%d $fn
5985 echo "create $fn short symlinks"
5986 for i in $(seq -w 1 $fn); do
5987 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
5992 umount_client $MOUNT || error "umount client failed"
5993 stop_mds || error "stop mds failed"
5994 stop_ost || error "stop ost failed"
5997 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
5999 run_test 72 "test fast symlink with extents flag enabled"
6001 test_73() { #LU-3006
6002 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
6003 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
6004 error "1st tunefs failed"
6005 start_mgsmds || error "start mds failed"
6006 start_ost || error "start ost failed"
6007 mount_client $MOUNT || error "mount client failed"
6008 $LCTL get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
6009 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
6010 umount_client $MOUNT || error "umount client failed"
6014 run_test 73 "failnode to update from mountdata properly"
6016 test_75() { # LU-2374
6017 [[ "$MDS1_VERSION" -lt $(version_code 2.4.1) ]] &&
6018 skip "Need MDS version at least 2.4.1"
6021 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
6022 --replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
6023 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
6024 --replace --reformat $(ostdevname 1) $(ostvdevname 1)"
6026 #check with default parameters
6027 add mds1 $opts_mds || error "add mds1 failed for default params"
6028 add ost1 $opts_ost || error "add ost1 failed for default params"
6030 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
6031 opts_mds=$(echo $opts_mds |
6032 sed -e "s/--index=$index/--index=$index --mdt/")
6033 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
6034 opts_ost=$(echo $opts_ost |
6035 sed -e "s/--index=$index/--index=$index --ost/")
6037 add mds1 $opts_mds || error "add mds1 failed for new params"
6038 add ost1 $opts_ost || error "add ost1 failed for new params"
6043 run_test 75 "The order of --index should be irrelevant"
6046 [[ "$MGS_VERSION" -ge $(version_code 2.4.52) ]] ||
6047 skip "Need MDS version at least 2.4.52"
6050 local MDMB_PARAM="osc.*.max_dirty_mb"
6051 echo "Change MGS params"
6052 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
6054 echo "max_dirty_mb: $MAX_DIRTY_MB"
6055 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB - 10))
6056 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
6057 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
6058 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
6059 head -1" $NEW_MAX_DIRTY_MB
6060 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
6061 echo "$MAX_DIRTY_MB"
6062 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
6063 error "error while apply max_dirty_mb"
6065 echo "Check the value is stored after remount"
6068 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
6069 head -1" $NEW_MAX_DIRTY_MB
6070 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
6071 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
6072 error "max_dirty_mb is not saved after remount"
6074 echo "Change OST params"
6075 CLIENT_PARAM="obdfilter.*.client_cache_count"
6076 local CLIENT_CACHE_COUNT
6077 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
6079 echo "client_cache_count: $CLIENT_CACHE_COUNT"
6080 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
6081 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
6082 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
6083 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
6084 head -1" $NEW_CLIENT_CACHE_COUNT
6085 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
6087 echo "$CLIENT_CACHE_COUNT"
6088 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
6089 error "error while apply client_cache_count"
6091 echo "Check the value is stored after remount"
6094 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
6095 head -1" $NEW_CLIENT_CACHE_COUNT
6096 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
6098 echo "$CLIENT_CACHE_COUNT"
6099 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
6100 error "client_cache_count is not saved after remount"
6103 run_test 76a "set permanent params with lctl across mounts"
6105 test_76b() { # LU-4783
6106 [[ "$MGS_VERSION" -ge $(version_code 2.5.57) ]] ||
6107 skip "Need MGS version at least 2.5.57"
6110 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
6111 error "start params log failed"
6114 run_test 76b "verify params log setup correctly"
6117 [[ "$MGS_VERSION" -ge $(version_code 2.8.54) ]] ||
6118 skip "Need MDS version at least 2.4.52"
6120 local MASK_PARAM="mdd.*.changelog_mask"
6121 echo "Change changelog_mask"
6122 do_facet mgs $LCTL set_param -P $MASK_PARAM=-CLOSE ||
6123 error "Can't change changlog_mask"
6124 wait_update $(facet_host mds) "$LCTL get_param -n $MASK_PARAM |
6127 echo "Check the value is stored after mds remount"
6128 stop_mds || error "Failed to stop MDS"
6129 start_mds || error "Failed to start MDS"
6130 local CHANGELOG_MASK=$(do_facet mgs $LCTL get_param -n $MASK_PARAM)
6131 echo $CHANGELOG_MASK | grep CLOSE > /dev/null &&
6132 error "changelog_mask is not changed"
6136 run_test 76c "verify changelog_mask is applied with lctl set_param -P"
6138 test_76d() { #LU-9399
6141 local xattr_cache="llite.*.xattr_cache"
6142 local cmd="$LCTL get_param -n $xattr_cache | head -1"
6143 local new=$((($(eval $cmd) + 1) % 2))
6145 echo "lctl set_param -P llite.*.xattr_cache=$new"
6146 do_facet mgs $LCTL set_param -P $xattr_cache=$new ||
6147 error "Can't change xattr_cache"
6148 wait_update $HOSTNAME "$cmd" "$new"
6150 echo "Check $xattr_cache on client $MOUNT"
6151 umount_client $MOUNT || error "umount $MOUNT failed"
6152 mount_client $MOUNT || error "mount $MOUNT failed"
6153 [ $(eval $cmd) -eq $new ] ||
6154 error "$xattr_cache != $new on client $MOUNT"
6156 echo "Check $xattr_cache on the new client $MOUNT2"
6157 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
6158 [ $(eval $cmd) -eq $new ] ||
6159 error "$xattr_cache != $new on client $MOUNT2"
6160 umount_client $MOUNT2 || error "umount $MOUNT2 failed"
6164 run_test 76d "verify llite.*.xattr_cache can be set by 'lctl set_param -P' correctly"
6166 test_77() { # LU-3445
6167 [[ "$MDS1_VERSION" -ge $(version_code 2.8.55) ]] ||
6168 skip "Need MDS version 2.8.55+ "
6170 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
6171 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
6172 skip_env "mixed loopback and real device not working"
6175 local fs2mdsdev=$(mdsdevname 1_2)
6176 local fs2ostdev=$(ostdevname 1_2)
6177 local fs2mdsvdev=$(mdsvdevname 1_2)
6178 local fs2ostvdev=$(ostvdevname 1_2)
6179 local fsname=test1234
6181 local failnid="$(h2nettype 1.2.3.4),$(h2nettype 4.3.2.1)"
6183 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
6185 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
6186 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
6187 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
6188 error "start fs2mds failed"
6190 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
6191 mgsnid="0.0.0.0@tcp,$mgsnid,$mgsnid:$mgsnid"
6193 add fs2ost --mgsnode=$mgsnid $(mkfs_opts ost1 $fs2ostdev) \
6194 --failnode=$failnid --fsname=$fsname \
6195 --reformat $fs2ostdev $fs2ostvdev ||
6196 error "add fs2ost failed"
6197 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
6199 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
6200 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
6201 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
6204 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
6207 [[ "$mds1_FSTYPE" != ldiskfs ||
6208 "$ost1_FSTYPE" != ldiskfs ]] &&
6209 skip "ldiskfs only test"
6211 # reformat the Lustre filesystem with a smaller size
6212 local saved_MDSCOUNT=$MDSCOUNT
6213 local saved_MDSSIZE=$MDSSIZE
6214 local saved_OSTCOUNT=$OSTCOUNT
6215 local saved_OSTSIZE=$OSTSIZE
6218 MDSSIZE=$((MDSSIZE - 20000))
6219 OSTSIZE=$((OSTSIZE - 20000))
6220 reformat || error "(1) reformat Lustre filesystem failed"
6221 MDSSIZE=$saved_MDSSIZE
6222 OSTSIZE=$saved_OSTSIZE
6224 # mount the Lustre filesystem
6225 setup_noconfig || error "(2) setup Lustre filesystem failed"
6228 log "create test files"
6233 mkdir $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
6235 for i in $(seq $num_files); do
6236 file=$MOUNT/$tdir/$tfile-$i
6237 dd if=/dev/urandom of=$file count=1 bs=1M || {
6238 $LCTL get_param osc.*.cur*grant*
6239 $LFS df; $LFS df -i;
6240 # stop creating files if there is no more space
6241 if [ ! -e $file ]; then
6242 num_files=$((i - 1))
6246 $LFS getstripe -v $file
6247 local ost_idx=$(LFS getstripe -i $file)
6248 do_facet ost$((ost_idx + 1)) \
6249 $LCTL get_param obdfilter.*.*grant*
6250 error "(4) create $file failed"
6254 # unmount the Lustre filesystem
6255 cleanup || error "(5) cleanup Lustre filesystem failed"
6257 # run e2fsck on the MDT and OST devices
6258 local mds_host=$(facet_active_host $SINGLEMDS)
6259 local ost_host=$(facet_active_host ost1)
6260 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
6261 local ost_dev=$(ostdevname 1)
6263 run_e2fsck $mds_host $mds_dev "-y"
6264 run_e2fsck $ost_host $ost_dev "-y"
6266 # get the original block count of the MDT and OST filesystems
6267 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
6268 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
6270 # expand the MDT and OST filesystems to the device size
6271 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
6272 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
6274 # run e2fsck on the MDT and OST devices again
6275 run_e2fsck $mds_host $mds_dev "-y"
6276 run_e2fsck $ost_host $ost_dev "-y"
6278 # mount the Lustre filesystem
6282 log "check files after expanding the MDT and OST filesystems"
6283 for i in $(seq $num_files); do
6284 file=$MOUNT/$tdir/$tfile-$i
6285 $CHECKSTAT -t file -s 1048576 $file ||
6286 error "(6) checkstat $file failed"
6290 log "create more files after expanding the MDT and OST filesystems"
6291 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
6292 file=$MOUNT/$tdir/$tfile-$i
6293 dd if=/dev/urandom of=$file count=1 bs=1M ||
6294 error "(7) create $file failed"
6297 # unmount the Lustre filesystem
6298 cleanup || error "(8) cleanup Lustre filesystem failed"
6300 # run e2fsck on the MDT and OST devices
6301 run_e2fsck $mds_host $mds_dev "-y"
6302 run_e2fsck $ost_host $ost_dev "-y"
6304 # get the maximum block count of the MDT and OST filesystems
6305 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
6306 local ost_max_blks=$(get_block_count ost1 $ost_dev)
6308 # get the minimum block count of the MDT and OST filesystems
6309 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
6310 grep minimum | sed -e 's/^.*filesystem: //g')
6311 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
6312 grep minimum | sed -e 's/^.*filesystem: //g')
6314 # shrink the MDT and OST filesystems to a smaller size
6318 if [[ $mds_max_blks -gt $mds_min_blks &&
6319 $mds_max_blks -gt $mds_orig_blks ]]; then
6320 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
6321 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
6322 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
6323 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
6324 error "shrink $SINGLEMDS to $new_blks failed"
6328 if [[ $ost_max_blks -gt $ost_min_blks &&
6329 $ost_max_blks -gt $ost_orig_blks ]]; then
6330 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
6331 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
6332 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
6333 run_resize2fs ost1 $ost_dev $new_blks ||
6334 error "shrink ost1 to $new_blks failed"
6338 # check whether the MDT or OST filesystem was shrunk or not
6340 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
6341 reformat_and_config ||
6342 error "(10) reformat Lustre filesystem failed"
6346 # run e2fsck on the MDT and OST devices again
6347 run_e2fsck $mds_host $mds_dev "-y"
6348 run_e2fsck $ost_host $ost_dev "-y"
6350 # mount the Lustre filesystem again
6354 log "check files after shrinking the MDT and OST filesystems"
6355 for i in $(seq $((num_files + 10))); do
6356 file=$MOUNT/$tdir/$tfile-$i
6357 $CHECKSTAT -t file -s 1048576 $file ||
6358 error "(11) checkstat $file failed"
6361 # unmount and reformat the Lustre filesystem
6362 cleanup || error "(12) cleanup Lustre filesystem failed"
6363 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
6365 MDSCOUNT=$saved_MDSCOUNT
6366 OSTCOUNT=$saved_OSTCOUNT
6367 reformat_and_config || error "(14) reformat Lustre filesystem failed"
6369 run_test 78 "run resize2fs on MDT and OST filesystems"
6371 test_79() { # LU-4227
6372 [[ "$MDS1_VERSION" -ge $(version_code 2.5.59) ]] ||
6373 skip "Need MDS version at least 2.5.59"
6375 local mdsdev1=$(mdsdevname 1)
6376 local mdsvdev1=$(mdsvdevname 1)
6377 local mdsdev2=$(mdsdevname 2)
6378 local mdsvdev2=$(mdsvdevname 2)
6379 local ostdev1=$(ostdevname 1)
6380 local ostvdev1=$(ostvdevname 1)
6381 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
6382 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
6383 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
6386 # remove --mgs/--mgsnode from mkfs.lustre options
6387 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
6389 mgsnode_opt=$(echo $opts_mds2 |
6390 awk '{ for ( i = 1; i < NF; i++ )
6391 if ( $i ~ "--mgsnode" ) { print $i; break } }')
6392 [ -n "$mgsnode_opt" ] &&
6393 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
6395 mgsnode_opt=$(echo $opts_ost1 |
6396 awk '{ for ( i = 1; i < NF; i++ )
6397 if ( $i ~ "--mgsnode" ) { print $i; break } }')
6398 [ -n "$mgsnode_opt" ] &&
6399 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
6401 # -MGS, format a mdt without --mgs option
6402 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
6403 error "Must specify --mgs when formatting mdt combined with mgs"
6405 # +MGS, format a mdt/ost without --mgsnode option
6406 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
6407 > /dev/null || error "start mds1 failed"
6408 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
6409 error "Must specify --mgsnode when formatting a mdt"
6410 add ost1 $opts_ost1 $ostdev1 $ostvdev2 &&
6411 error "Must specify --mgsnode when formatting an ost"
6415 run_test 79 "format MDT/OST without mgs option (should return errors)"
6418 start_mds || error "Failed to start MDT"
6419 start_ost || error "Failed to start OST1"
6420 uuid=$(do_facet ost1 $LCTL get_param -n mgc.*.uuid)
6421 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
6422 do_facet ost1 "$LCTL set_param fail_val=10 fail_loc=0x906"
6423 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x906"
6424 do_facet mgs "$LCTL set_param -n mgs/MGS/evict_client $uuid"
6426 start_ost2 || error "Failed to start OST2"
6428 do_facet ost1 "$LCTL set_param fail_loc=0"
6433 run_test 80 "mgc import reconnect race"
6435 #Save the original values of $OSTCOUNT and $OSTINDEX$i.
6437 local new_ostcount=$1
6438 saved_ostcount=$OSTCOUNT
6439 OSTCOUNT=$new_ostcount
6443 for i in $(seq $OSTCOUNT); do
6445 eval saved_ostindex$i=${!index}
6450 # Restore the original values of $OSTCOUNT and $OSTINDEX$i.
6451 restore_ostindex() {
6456 for i in $(seq $OSTCOUNT); do
6457 index=saved_ostindex$i
6458 eval OSTINDEX$i=${!index}
6460 OSTCOUNT=$saved_ostcount
6463 if ! combined_mgs_mds ; then
6468 # The main purpose of this test is to ensure the OST_INDEX_LIST functions as
6469 # expected. This test uses OST_INDEX_LIST to format OSTs with a randomly
6470 # assigned index and ensures we can mount such a formatted file system
6471 test_81() { # LU-4665
6472 [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
6473 skip "Need MDS version at least 2.6.54"
6474 [[ $OSTCOUNT -ge 3 ]] || skip_env "needs >= 3 OSTs"
6478 # Each time RANDOM is referenced, a random integer between 0 and 32767
6481 local saved_ostindex1=$OSTINDEX1
6482 for i in 65535 $((RANDOM + 65536)); do
6483 echo -e "\nFormat ost1 with --index=$i, should fail"
6485 if add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat \
6486 $(ostdevname 1) $(ostvdevname 1); then
6487 OSTINDEX1=$saved_ostindex1
6488 error "format ost1 with --index=$i should fail"
6491 OSTINDEX1=$saved_ostindex1
6495 # Format OSTs with random sparse indices.
6496 trap "restore_ostindex" EXIT
6497 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices"
6498 OST_INDEX_LIST=[0,$((RANDOM * 2 % 65533 + 1)),65534] formatall
6500 # Setup and check Lustre filesystem.
6501 start_mgsmds || error "start_mgsmds failed"
6502 for i in $(seq $OSTCOUNT); do
6503 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6504 error "start ost$i failed"
6507 mount_client $MOUNT || error "mount client $MOUNT failed"
6508 check_mount || error "check client $MOUNT failed"
6511 local max_easize=$($LCTL get_param -n llite.*.max_easize)
6512 # 65452 is XATTR_SIZE_MAX less ldiskfs ea overhead
6513 if large_xattr_enabled; then
6514 [[ $max_easize -ge 65452 ]] ||
6515 error "max_easize is $max_easize, should be at least 65452 bytes"
6518 # 4012 is 4096 less ldiskfs ea overhead
6519 [[ $max_easize -ge 4012 ]] ||
6520 error "max_easize is $max_easize, should be at least 4012 bytes"
6525 run_test 81 "sparse OST indexing"
6527 # Here we exercise the stripe placement functionality on a file system that
6528 # has formatted the OST with a random index. With the file system the following
6529 # functionality is tested:
6531 # 1. Creating a new file with a specific stripe layout.
6533 # 2. Modifiy a existing empty file with a specific stripe layout.
6535 # 3. Ensure we fail to set the stripe layout of a file that already has one.
6537 # 4. If ost-index is defined we need to ensure it is the first entry in the
6538 # ost index list returned by lfs getstripe.
6540 # 5. Lastly ensure this functionality fails with directories.
6541 test_82a() { # LU-4665
6542 [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
6543 skip "Need MDS version at least 2.6.54"
6544 [[ $OSTCOUNT -ge 3 ]] || skip_env "needs >= 3 OSTs"
6550 # Format OSTs with random sparse indices.
6554 local LOV_V1_INSANE_STRIPE_COUNT=65532
6555 for i in $(seq $OSTCOUNT); do
6556 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
6557 ost_indices+=" $index"
6559 ost_indices=$(comma_list $ost_indices)
6561 stack_trap "restore_ostindex" EXIT
6562 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
6563 OST_INDEX_LIST=[$ost_indices] formatall
6565 # Setup Lustre filesystem.
6566 start_mgsmds || error "start_mgsmds failed"
6567 for i in $(seq $OSTCOUNT); do
6568 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6569 error "start ost$i failed"
6572 # Collect debug information - start of test
6573 do_nodes $(comma_list $(mdts_nodes)) \
6574 $LCTL get_param osp.*.prealloc_*_id
6576 mount_client $MOUNT || error "mount client $MOUNT failed"
6580 check_lfs_df_ret_val $? || error "$LFS df $MOUNT failed"
6581 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6583 stack_trap "do_nodes $(comma_list $(mdts_nodes)) \
6584 $LCTL get_param osp.*.prealloc_*_id" EXIT
6586 # 1. If the file does not exist, new file will be created
6587 # with specified OSTs.
6588 local file=$DIR/$tdir/$tfile-1
6589 local cmd="$LFS setstripe -o $ost_indices $file"
6591 eval $cmd || error "$cmd failed"
6592 check_stripe_count $file $OSTCOUNT
6593 check_obdidx $file $ost_indices
6594 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6595 error "write $file failed"
6597 # 2. If the file already exists and is an empty file, the file
6598 # will be attached with specified layout.
6599 file=$DIR/$tdir/$tfile-2
6600 mcreate $file || error "mcreate $file failed"
6601 cmd="$LFS setstripe -o $ost_indices $file"
6603 eval $cmd || error "$cmd failed"
6604 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6605 error "write $file failed"
6606 check_stripe_count $file $OSTCOUNT
6607 check_obdidx $file $ost_indices
6609 # 3. If the file already has a valid layout attached, the command
6610 # should fail with EBUSY.
6612 eval $cmd && error "stripe is already set on $file, $cmd should fail"
6614 # 4. If [--stripe-index|-i <start_ost_idx>] is used, the index must
6615 # be in the OST indices list.
6616 local start_ost_idx=${ost_indices##*,}
6617 file=$DIR/$tdir/$tfile-3
6618 cmd="$LFS setstripe -o $ost_indices -i $start_ost_idx $file"
6620 eval $cmd || error "$cmd failed"
6621 check_stripe_count $file $OSTCOUNT
6622 check_obdidx $file $ost_indices
6623 check_start_ost_idx $file $start_ost_idx
6625 file=$DIR/$tdir/$tfile-4
6626 cmd="$LFS setstripe"
6627 cmd+=" -o $(exclude_items_from_list $ost_indices $start_ost_idx)"
6628 cmd+=" -i $start_ost_idx $file"
6630 eval $cmd && error "index $start_ost_idx should be in $ost_indices"
6632 # 5. Specifying OST indices for directory should succeed.
6633 local dir=$DIR/$tdir/$tdir
6634 mkdir $dir || error "mkdir $dir failed"
6635 cmd="$LFS setstripe -o $ost_indices $dir"
6636 if [ "$MDS1_VERSION" -gt $(version_code 2.11.53) ] &&
6637 [ "$CLIENT_VERSION" -gt $(version_code 2.11.53) ]; then
6639 eval $cmd || error "unable to specify OST indices on directory"
6641 echo "need MDS+client version at least 2.11.53"
6646 run_test 82a "specify OSTs for file (succeed) or directory (succeed)"
6651 # Remove OSTs from a pool and destroy the pool.
6652 destroy_pool $ost_pool || true
6657 # Test 82b is run to ensure that if the user supplies a pool with a specific
6658 # stripe layout that it behaves proprerly. It should fail in the case that
6659 # the supplied OST index list points to OSTs not contained in the user
6661 test_82b() { # LU-4665
6662 [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
6663 skip "Need MDS version at least 2.6.54"
6664 [[ $OSTCOUNT -ge 4 ]] || skip_env "needs >= 4 OSTs"
6670 # Format OSTs with random sparse indices.
6674 local LOV_V1_INSANE_STRIPE_COUNT=65532
6675 for i in $(seq $OSTCOUNT); do
6676 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
6677 ost_indices+=" $index"
6679 ost_indices=$(comma_list $ost_indices)
6681 trap "restore_ostindex" EXIT
6682 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
6683 OST_INDEX_LIST=[$ost_indices] formatall
6685 # Setup Lustre filesystem.
6686 start_mgsmds || error "start_mgsmds failed"
6687 for i in $(seq $OSTCOUNT); do
6688 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6689 error "start ost$i failed"
6692 mount_client $MOUNT || error "mount client $MOUNT failed"
6696 check_lfs_df_ret_val $? || error "$LFS df $MOUNT failed"
6697 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6699 # Create a new pool and add OSTs into it.
6700 local ost_pool=$FSNAME.$TESTNAME
6701 create_pool $ost_pool || error "create OST pool $ost_pool failed"
6704 trap "cleanup_82b" EXIT
6706 local ost_idx_in_list=${ost_indices##*,}
6707 local ost_idx_in_pool=$(exclude_items_from_list $ost_indices \
6710 local ost_targets="$FSNAME-OST["
6711 for i in ${ost_idx_in_pool//,/ }; do
6712 ost_targets=$ost_targets$(printf "%04x," $i)
6714 ost_targets="${ost_targets%,}]"
6716 local ost_targets_uuid=$(for i in ${ost_idx_in_pool//,/ }; \
6717 do printf "$FSNAME-OST%04x_UUID\n" $i; done |
6718 sort -u | tr '\n' ' ')
6720 local cmd="$LCTL pool_add $ost_pool $ost_targets"
6721 do_facet mgs $cmd || error "$cmd failed"
6722 wait_update $HOSTNAME "$LCTL get_param -n lov.$FSNAME-*.pools.$TESTNAME|
6723 sort -u | tr '\n' ' ' " "$ost_targets_uuid" ||
6724 error "wait_update $ost_pool failed"
6725 wait_update_facet $SINGLEMDS "$LCTL pool_list $ost_pool | wc -l" 4 ||
6726 error "wait_update pool_list $ost_pool failed"
6728 # If [--pool|-p <pool_name>] is set with [--ost-list|-o <ost_indices>],
6729 # then the OSTs must be the members of the pool.
6730 local file=$DIR/$tdir/$tfile
6731 cmd="$LFS setstripe -p $ost_pool -o $ost_idx_in_list $file"
6733 eval $cmd && error "OST with index $ost_idx_in_list should be" \
6734 "in OST pool $ost_pool"
6736 # Only select OST $ost_idx_in_list from $ost_pool for file.
6737 ost_idx_in_list=${ost_idx_in_pool#*,}
6738 cmd="$LFS setstripe -p $ost_pool -o $ost_idx_in_list $file"
6740 eval $cmd || error "$cmd failed"
6741 cmd="$LFS getstripe $file"
6743 eval $cmd || error "$cmd failed"
6744 check_stripe_count $file 2
6745 check_obdidx $file $ost_idx_in_list
6746 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6747 error "write $file failed"
6751 run_test 82b "specify OSTs for file with --pool and --ost-list options"
6754 [[ "$OST1_VERSION" -ge $(version_code 2.6.91) ]] ||
6755 skip "Need OST version at least 2.6.91"
6756 if [ "$ost1_FSTYPE" != ldiskfs ]; then
6757 skip "ldiskfs only test"
6766 ostmnt=$(facet_mntpt ost1)
6768 # Mount the OST as an ldiskfs filesystem.
6769 log "mount the OST $dev as a $ost1_FSTYPE filesystem"
6770 add ost1 $(mkfs_opts ost1 $dev) $FSTYPE_OPT \
6771 --reformat $dev > /dev/null ||
6772 error "format ost1 error"
6774 if ! test -b $dev; then
6775 mnt_opts=$(csa_add "$OST_MOUNT_FS_OPTS" -o loop)
6777 echo "mnt_opts $mnt_opts"
6778 do_facet ost1 mount -t "$ost1_FSTYPE" $dev \
6780 # Run llverfs on the mounted ldiskfs filesystem.
6781 # It is needed to get ENOSPACE.
6782 log "run llverfs in partial mode on the OST $ost1_FSTYPE $ostmnt"
6783 do_rpc_nodes $(facet_host ost1) run_llverfs $ostmnt -vpl \
6784 "no" || error "run_llverfs error on $ost1_FSTYPE"
6787 log "unmount the OST $dev"
6790 # Delete file IO_scrub. Later osd_scrub_setup will try to
6791 # create "IO_scrub" but will get ENOSPACE.
6793 echo "start ost1 service on `facet_active_host ost1`"
6794 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
6797 err=$(do_facet ost1 dmesg | grep "VFS: Busy inodes after unmount of")
6798 echo "string err $err"
6799 [ -z "$err" ] || error $err
6802 run_test 83 "ENOSPACE on OST doesn't cause message VFS: \
6803 Busy inodes after unmount ..."
6806 local facet=$SINGLEMDS
6807 local num=$(echo $facet | tr -d "mds")
6808 local dev=$(mdsdevname $num)
6809 local time_min=$(recovery_time_min)
6810 local recovery_duration
6811 local completed_clients
6812 local correct_clients
6815 echo "start mds service on $(facet_active_host $facet)"
6817 "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
6818 error "start MDS failed"
6820 start_ost || error "start OST0000 failed"
6821 start_ost2 || error "start OST0001 failed"
6823 echo "recovery_time=$time_min, timeout=$TIMEOUT, wrap_up=$wrap_up"
6825 mount_client $MOUNT1 || error "mount $MOUNT1 failed"
6826 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
6827 # make sure new superblock labels are sync'd before disabling writes
6831 replay_barrier $SINGLEMDS
6832 createmany -o $DIR1/$tfile-%d 1000
6834 # We need to catch the end of recovery window to extend it.
6835 # Skip 5 requests and add delay to request handling.
6836 #define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP
6837 do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5"
6839 facet_failover --fsck $SINGLEMDS || error "failover: $?"
6842 echo "recovery status"
6843 do_facet $SINGLEMDS \
6844 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status"
6846 recovery_duration=$(do_facet $SINGLEMDS \
6847 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
6848 awk '/recovery_duration/ { print $2 }')
6849 (( $recovery_duration > $time_min + $wrap_up )) &&
6850 error "recovery_duration > recovery_time_hard + wrap up"
6851 completed_clients=$(do_facet $SINGLEMDS \
6852 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
6853 awk '/completed_clients/ { print $2 }')
6855 correct_clients="$MDSCOUNT/$((MDSCOUNT+1))"
6856 [ "$completed_clients" = "${correct_clients}" ] ||
6857 error "$completed_clients != $correct_clients"
6859 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
6860 umount_client $MOUNT1
6861 umount_client $MOUNT2
6867 run_test 84 "check recovery_hard_time"
6870 [[ "$OST1_VERSION" -ge $(version_code 2.7.55) ]] ||
6871 skip "Need OST version at least 2.7.55"
6872 ##define OBD_FAIL_OSD_OST_EA_FID_SET 0x197
6873 do_facet ost1 "lctl set_param fail_loc=0x197"
6877 run_test 85 "osd_ost init: fail ea_fid_set"
6882 # ost1 has already registered to the MGS before the reformat.
6883 # So after reformatting it with option "-G", it could not be
6884 # mounted to the MGS. Cleanup the system for subsequent tests.
6889 [ "$ost1_FSTYPE" = zfs ] &&
6890 skip "LU-6442: no such mkfs params for ZFS OSTs"
6891 [[ "$MDS1_VERSION" -ge $(version_code 2.7.56) ]] ||
6892 skip "Need server version newer than 2.7.55"
6895 local OLDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6896 awk '/Flex block group size: / { print $NF; exit; }')
6898 [ "$OLDSIZE" == "$NEWSIZE" ] && skip "$NEWSIZE groups already"
6900 local opts=" -O flex_bg -G $NEWSIZE"
6901 opts=$(OST_FS_MKFS_OPTS+="$opts" mkfs_opts ost1 $(ostdevname 1))
6902 opts+=" --reformat $(ostdevname 1) $(ostvdevname 1)"
6903 echo "params: $opts"
6905 trap cleanup_86 EXIT ERR
6908 add ost1 $opts || error "add ost1 failed with new params"
6910 local FOUNDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6911 awk '/Flex block group size: / { print $NF; exit; }')
6913 [[ $FOUNDSIZE == $NEWSIZE ]] ||
6914 error "Flex block group size: $FOUNDSIZE, expected: $NEWSIZE"
6918 run_test 86 "Replacing mkfs.lustre -G option"
6920 test_87() { #LU-6544
6921 [[ "$MDS1_VERSION" -ge $(version_code 2.9.51) ]] ||
6922 skip "Need MDS version at least 2.9.51"
6923 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
6924 skip "ldiskfs only test"
6925 [[ $OSTCOUNT -gt 59 ]] &&
6926 skip "Ignore wide striping situation"
6927 [ -n "$FILESET" ] && skip "Not functional for FILESET set"
6929 local mdsdev=$(mdsdevname 1)
6930 local mdsvdev=$(mdsvdevname 1)
6931 local file=$DIR/$tfile
6932 local mntpt=$(facet_mntpt $SINGLEMDS)
6933 local used_xattr_blk=0
6934 local inode_size=${1:-1024}
6936 local xtest="trusted.test"
6940 local stripe_cnt=$(($OSTCOUNT + 2))
6942 #Please see ldiskfs_make_lustre() for MDT inode size calculation
6943 if [ $stripe_cnt -gt 16 ]; then
6946 left_size=$(expr $inode_size - \
6949 32 - 40 \* 3 - 32 \* 3 - $stripe_cnt \* 24 - 16 - 3 - \
6951 24 - 18 - $(expr length $tfile) - 16 - 4)
6952 if [ $left_size -le 0 ]; then
6953 echo "No space($left_size) is expected in inode."
6954 echo "Try 1-byte xattr instead to verify this."
6957 echo "Estimate: at most $left_size-byte space left in inode."
6963 if ! combined_mgs_mds ; then
6967 add mds1 $(mkfs_opts mds1 ${mdsdev}) --stripe-count-hint=$stripe_cnt \
6968 --reformat $mdsdev $mdsvdev || error "add mds1 failed"
6969 start_mdt 1 > /dev/null || error "start mdt1 failed"
6970 for i in $(seq $OSTCOUNT); do
6971 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS > /dev/null ||
6972 error "start ost$i failed"
6974 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
6975 check_mount || error "check client $MOUNT failed"
6978 $LFS setstripe -E 1M -S 1M -c 1 -E 64M -c 1 -E -1 -c -1 $file ||
6979 error "Create file with 3 components failed"
6980 $TRUNCATE $file $((1024*1024*64+1)) || error "truncate file failed"
6981 i=$($LFS getstripe -I3 -c $file) || error "get 3rd stripe count failed"
6982 if [ $i -ne $OSTCOUNT ]; then
6983 left_size=$(expr $left_size + $(expr $OSTCOUNT - $i) \* 24)
6984 echo -n "Since only $i out $OSTCOUNT OSTs are used, "
6985 echo -n "the expected left space is changed to "
6986 echo "$left_size bytes at most."
6988 value=$(generate_string $left_size)
6989 setfattr -n $xtest -v $value $file
6990 orig=$(get_xattr_value $xtest $file)
6991 [[ "$orig" != "$value" ]] && error "$xtest changed"
6993 #Verify if inode has some expected space left
6994 umount $MOUNT > /dev/null || error "umount $MOUNT failed"
6995 stop_mdt 1 > /dev/null || error "stop mdt1 failed"
6996 mount_ldiskfs $SINGLEMDS || error "mount -t ldiskfs $SINGLEMDS failed"
6998 do_facet $SINGLEMDS ls -sal $mntpt/ROOT/$tfile
6999 used_xattr_blk=$(do_facet $SINGLEMDS ls -s $mntpt/ROOT/$tfile |
7001 [[ $used_xattr_blk -eq 0 ]] &&
7002 error "Please check MDS inode size calculation: \
7003 more than $left_size-byte space left in inode."
7004 echo "Verified: at most $left_size-byte space left in inode."
7006 unmount_ldiskfs $SINGLEMDS
7008 for i in $(seq $OSTCOUNT); do
7009 stop ost$i -f || error "stop ost$i failed"
7012 run_test 87 "check if MDT inode can hold EAs with N stripes properly"
7015 [ "$mds1_FSTYPE" == zfs ] &&
7016 skip "LU-6662: no implementation for ZFS"
7020 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
7021 --reformat $(mdsdevname 1) || error "add mds1 failed"
7023 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
7024 grep -e \".*opts:.*errors=remount-ro.*\"" ||
7025 error "default mount options is missing"
7027 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
7028 --mountfsoptions="user_xattr,errors=panic" \
7029 --reformat $(mdsdevname 1) || error "add mds1 failed"
7031 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
7032 grep -e \".*opts:.*errors=panic.*\"" ||
7033 error "user can't override default mount options"
7035 run_test 88 "check the default mount options can be overridden"
7037 test_89() { # LU-7131
7038 [[ "$MDS1_VERSION" -ge $(version_code 2.9.54) ]] ||
7039 skip "Need MDT version at least 2.9.54"
7041 local key=failover.node
7042 local val1=192.0.2.254@tcp0 # Reserved IPs, see RFC 5735
7043 local val2=192.0.2.255@tcp0
7044 local mdsdev=$(mdsdevname 1)
7049 if [[ "$mds1_FSTYPE" == zfs ]]; then
7050 import_zpool mds1 || return ${PIPESTATUS[0]}
7053 # Check that parameters are added correctly
7054 echo "tunefs --param $key=$val1"
7055 do_facet mds1 "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
7056 error "tunefs --param $key=$val1 failed"
7057 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7058 error "tunefs --dryrun failed"
7059 params=${params##*Parameters:}
7060 params=${params%%exiting*}
7061 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val1) = "1" ] ||
7062 error "on-disk parameter not added correctly via tunefs"
7064 # Check that parameters replace existing instances when added
7065 echo "tunefs --param $key=$val2"
7066 do_facet mds1 "$TUNEFS --param $key=$val2 $mdsdev >/dev/null" ||
7067 error "tunefs --param $key=$val2 failed"
7068 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7069 error "tunefs --dryrun failed"
7070 params=${params##*Parameters:}
7071 params=${params%%exiting*}
7072 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "1" ] ||
7073 error "on-disk parameter not replaced via tunefs"
7074 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val2) = "1" ] ||
7075 error "on-disk parameter not replaced correctly via tunefs"
7077 # Check that a parameter is erased properly
7078 echo "tunefs --erase-param $key"
7079 do_facet mds1 "$TUNEFS --erase-param $key $mdsdev >/dev/null" ||
7080 error "tunefs --erase-param $key failed"
7081 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7082 error "tunefs --dryrun failed"
7083 params=${params##*Parameters:}
7084 params=${params%%exiting*}
7085 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "0" ] ||
7086 error "on-disk parameter not erased correctly via tunefs"
7088 # Check that all the parameters are erased
7089 do_facet mds1 "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
7090 error "tunefs --param $key=$val1 failed"
7091 echo "tunefs --erase-params"
7092 do_facet mds1 "$TUNEFS --erase-params $mdsdev >/dev/null" ||
7093 error "tunefs --erase-params failed"
7094 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7095 error "tunefs --dryrun failed"
7096 params=${params##*Parameters:}
7097 params=${params%%exiting*}
7098 params=$(echo $params | tr ' ' '\n')
7100 error "all on-disk parameters not erased correctly via tunefs $params"
7102 # Check the order of options --erase-params and --param
7103 echo "tunefs --param $key=$val1 --erase-params"
7105 "$TUNEFS --param $key=$val1 --erase-params $mdsdev >/dev/null"||
7106 error "tunefs --param $key=$val1 --erase-params failed"
7107 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7108 error "tunefs --dryrun failed"
7109 params=${params##*Parameters:}
7110 params=${params%%exiting*}
7111 [ $(echo $params | tr ' ' '\n') == "$key=$val1" ] ||
7112 error "on-disk param not added correctly with --erase-params"
7116 run_test 89 "check tunefs --param and --erase-param{s} options"
7119 # $2 (optional) value of max_mod_rpcs_in_flight to set
7120 check_max_mod_rpcs_in_flight() {
7128 idx=$(printf "%04x" $($LFS getdirstripe -i $dir))
7129 facet="mds$((0x$idx + 1))"
7131 if [ -z "$mmr" ]; then
7132 # get value of max_mod_rcps_in_flight
7133 mmr=$($LCTL get_param -n \
7134 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) ||
7135 error "Unable to get max_mod_rpcs_in_flight"
7136 echo "max_mod_rcps_in_flight is $mmr"
7138 # set value of max_mod_rpcs_in_flight
7140 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight=$mmr ||
7141 error "Unable to set max_mod_rpcs_in_flight to $mmr"
7142 echo "max_mod_rpcs_in_flight set to $mmr"
7145 # create mmr+1 files
7146 echo "creating $((mmr + 1)) files ..."
7148 for i in $(seq $((mmr + 1))); do
7154 # consumes mmr-1 modify RPC slots
7155 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7156 # drop requests on MDT so that RPC slots are consumed
7157 # during all the request resend interval
7158 do_facet $facet "$LCTL set_param fail_loc=0x159"
7159 echo "launch $((mmr - 1)) chmod in parallel ..."
7160 for i in $(seq $((mmr - 1))); do
7161 chmod 0600 $dir/file-$i &
7165 # send one additional modify RPC
7166 do_facet $facet "$LCTL set_param fail_loc=0"
7167 echo "launch 1 additional chmod in parallel ..."
7168 chmod 0600 $dir/file-$mmr &
7171 # check this additional modify RPC get a modify RPC slot
7172 # and succeed its operation
7173 checkstat -vp 0600 $dir/file-$mmr ||
7174 error "Unable to send $mmr modify RPCs in parallel"
7179 # consumes mmr modify RPC slots
7180 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7181 # drop requests on MDT so that RPC slots are consumed
7182 # during all the request resend interval
7183 do_facet $facet "$LCTL set_param fail_loc=0x159"
7184 echo "launch $mmr chmod in parallel ..."
7185 for i in $(seq $mmr); do
7186 chmod 0666 $dir/file-$i &
7190 # send one additional modify RPC
7191 do_facet $facet "$LCTL set_param fail_loc=0"
7192 echo "launch 1 additional chmod in parallel ..."
7193 chmod 0666 $dir/file-$((mmr + 1)) &
7196 # check this additional modify RPC blocked getting a modify RPC slot
7197 checkstat -vp 0644 $dir/file-$((mmr + 1)) ||
7198 error "Unexpectedly send $(($mmr + 1)) modify RPCs in parallel"
7202 get_mdt_max_mod_rpcs_in_flight_val() {
7206 # It will be enough to get one value from one MDT
7207 max_mrif=$(do_facet $facet \
7208 "$LCTL get_param -n \
7209 mdt.$FSNAME-MDT*.max_mod_rpcs_in_flight | head -n 1")
7210 if [[ $max_mrif ]]; then
7213 max_mrif=$(do_facet $facet \
7214 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
7219 set_mdt_max_mod_rpcs_in_flight() {
7224 lctl_op=$($LCTL get_param \
7225 mdt.*.max_mod_rpcs_in_flight)
7226 if [[ $lctl_op ]]; then
7229 mdt.$FSNAME-MDT*.max_mod_rpcs_in_flight=$max_mrif"
7233 /sys/module/mdt/parameters/max_mod_rpcs_per_client"
7234 echo "the deprecated max_mod_rpcs_per_client \
7235 parameter was involved"
7242 [[ $($LCTL get_param mdc.*.import |
7243 grep "connect_flags:.*multi_mod_rpc") ]] ||
7244 skip "Need MDC with 'multi_mod_rpcs' feature"
7246 # check default value
7247 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
7248 check_max_mod_rpcs_in_flight $DIR/$tdir
7252 run_test 90a "check max_mod_rpcs_in_flight is enforced"
7262 [[ $($LCTL get_param mdc.*.import |
7263 grep "connect_flags:.*multi_mod_rpc") ]] ||
7264 skip "Need MDC with 'multi_mod_rpcs' feature"
7267 # update max_mod_rpcs_in_flight
7268 $LFS mkdir -c1 $DIR/${tdir}1 || error "mkdir $DIR/${tdir}1 failed"
7269 check_max_mod_rpcs_in_flight $DIR/${tdir}1 1
7272 # check client is able to send multiple modify RPCs in paralell
7273 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
7274 grep -c "multi_mod_rpcs")
7275 if [ "$tmp" -ne $MDSCOUNT ]; then
7277 skip "Client not able to send multiple modify RPCs in parallel"
7280 # update max_mod_rpcs_in_flight
7281 $LFS mkdir -c1 $DIR/${tdir}2 || error "mkdir $DIR/${tdir}2 failed"
7282 check_max_mod_rpcs_in_flight $DIR/${tdir}2 5
7285 $LFS mkdir -c1 $DIR/${tdir}3 || error "mkdir $DIR/${tdir}3 failed"
7286 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/${tdir}3))
7287 facet="mds$((0x$idx + 1))"
7289 mmrpc=$(get_mdt_max_mod_rpcs_in_flight_val $facet)
7290 echo "mdt_max_mod_rpcs_in_flight is $mmrpc"
7291 # update max_mod_rpcs_in_flight
7292 umount_client $MOUNT
7293 set_mdt_max_mod_rpcs_in_flight 16 $facet
7295 $LCTL set_param mdc.$FSNAME-MDT$idx-mdc-*.max_rpcs_in_flight=17
7296 check_max_mod_rpcs_in_flight $DIR/${tdir}3 16
7298 # restore MDT max_mod_rpcs_in_flight initial value
7299 set_mdt_max_mod_rpcs_in_flight $mmrpc $facet
7301 rm -rf $DIR/${tdir}?
7304 run_test 90b "check max_mod_rpcs_in_flight is enforced after update"
7307 # get max_rpcs_in_flight value
7308 mrif_90c=$($LCTL get_param -n \
7309 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
7310 echo "max_rpcs_in_flight is $mrif_90c"
7312 # get MDC max_mod_rpcs_in_flight value
7313 mmrif_90c=$($LCTL get_param -n \
7314 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight)
7315 echo "MDC max_mod_rpcs_in_flight is $mmrif_90c"
7317 # get MDT max_mod_rpcs_in_flight value
7318 mmrpc_90c=$(get_mdt_max_mod_rpcs_in_flight_val "mds1")
7319 echo "mdt_max_mod_rpcs_in_flight is $mmrpc_90c"
7322 restore_params_90c() {
7325 # restore max_rpcs_in_flight value
7326 do_facet mgs $LCTL set_param -P \
7327 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$mrif_90c
7329 # restore max_mod_rpcs_in_flight value
7330 do_facet mgs $LCTL set_param -P \
7331 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mmrif_90c
7333 # restore MDT max_mod_rpcs_in_flight value
7334 set_mdt_max_mod_rpcs_in_flight $mmrpc_90c "mds1"
7342 [[ $($LCTL get_param mdc.*.import |
7343 grep "connect_flags:.*multi_mod_rpc") ]] ||
7344 skip "Need MDC with 'multi_mod_rpcs' feature"
7346 # check client is able to send multiple modify RPCs in paralell
7347 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
7348 grep -c "multi_mod_rpcs")
7349 if [ "$tmp" -ne $MDSCOUNT ]; then
7351 skip "Client not able to send multiple modify RPCs in parallel"
7355 stack_trap restore_params_90c
7358 # attempt to set max_mod_rpcs_in_flight to max_rpcs_in_flight value
7359 # prerequisite: set MDT max_mod_rpcs_in_flight to
7360 # max_rpcs_in_flight value
7361 set_mdt_max_mod_rpcs_in_flight $mrif_90c "mds1"
7363 # if max_mod_rpcs_in_flight is set to be equal to or larger than
7364 # max_rpcs_in_flight, then max_rpcs_in_flight will be increased
7365 if [[ "$CLIENT_VERSION" -ge $(version_code 2.13.53) ]]; then
7367 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif_90c ||
7368 error "set max_mod_rpcs_in_flight to $mrif_90c failed"
7370 local new_mrif=$($LCTL get_param -n \
7371 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
7372 ((new_mrif == mrif_90c + 1)) ||
7373 error "max_rpcs_in_flight was not increased"
7376 umount_client $MOUNT
7377 set_mdt_max_mod_rpcs_in_flight $mmrpc_90c "mds1"
7381 # attempt to set max_mod_rpcs_in_flight to MDT max_mod_rpcs_in_flight+1
7382 # prerequisite: set max_rpcs_in_flight to MDT max_mod_rpcs_in_flight+2
7384 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$((mmrpc_90c + 2))
7387 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$((mmrpc_90c + 1)) &&
7388 error "set max_mod_rpcs_in_flight to $((mmrpc_90c + 1)) should fail"
7391 # attempt to set max_mod_rpcs_in_flight permanently
7392 do_facet mgs $LCTL set_param -P \
7393 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$mrif_90c
7395 do_facet mgs $LCTL set_param -P \
7396 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif_90c
7398 remount_client $MOUNT
7400 wait_update_facet --verbose client "$LCTL get_param -n \
7401 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight" \
7402 "$((mrif_90c + 1))" ||
7403 error "expected '$((mrif_90c + 1))' for max_rpcs_in_flight"
7405 wait_update_facet --verbose client "$LCTL get_param -n \
7406 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight" \
7408 error "expected '$mrif_90c' for max_mod_rpcs_in_flight"
7413 run_test 90c "check max_mod_rpcs_in_flight update limits"
7424 [[ $($LCTL get_param mdc.*.import |
7425 grep "connect_flags:.*multi_mod_rpc") ]] ||
7426 skip "Need MDC with 'multi_mod_rpcs' feature"
7428 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
7429 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
7430 facet="mds$((0x$idx + 1))"
7432 # check client version supports multislots
7433 tmp=$($LCTL get_param -N \
7434 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
7435 if [ -z "$tmp" ]; then
7437 skip "Client does not support multiple modify RPCs in flight"
7440 # get current value of max_mod_rcps_in_flight
7441 mmr=$($LCTL get_param -n \
7442 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
7443 echo "max_mod_rcps_in_flight is $mmr"
7446 echo "creating $mmr files ..."
7448 for i in $(seq $mmr); do
7449 touch $DIR/$tdir/file-$i
7452 # prepare for close RPC
7453 multiop_bg_pause $DIR/$tdir/file-close O_c
7456 # consumes mmr modify RPC slots
7457 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7458 # drop requests on MDT so that RPC slots are consumed
7459 # during all the request resend interval
7460 do_facet $facet "$LCTL set_param fail_loc=0x159"
7461 echo "launch $mmr chmod in parallel ..."
7462 for i in $(seq $mmr); do
7463 chmod 0600 $DIR/$tdir/file-$i &
7466 # send one additional close RPC
7467 do_facet $facet "$LCTL set_param fail_loc=0"
7468 echo "launch 1 additional close in parallel ..."
7470 cancel_lru_locks mdc
7473 # check this additional close RPC get a modify RPC slot
7474 # and multiop process completed
7475 [ -d /proc/$pid ] &&
7476 error "Unable to send the additional close RPC in parallel"
7481 run_test 90d "check one close RPC is allowed above max_mod_rpcs_in_flight"
7483 check_uuid_on_ost() {
7485 do_facet ost1 "$LCTL get_param obdfilter.${FSNAME}*.exports.'$nid'.uuid"
7488 check_uuid_on_mdt() {
7490 do_facet $SINGLEMDS "$LCTL get_param mdt.${mds1_svc}*.exports.'$nid'.uuid"
7498 [[ "$OST1_VERSION" -ge $(version_code 2.7.63) ]] ||
7499 skip "Need OST version at least 2.7.63"
7500 [[ "$MDS1_VERSION" -ge $(version_code 2.7.63) ]] ||
7501 skip "Need MDT version at least 2.7.63"
7503 start_mds || error "MDS start failed"
7504 start_ost || error "unable to start OST"
7505 mount_client $MOUNT || error "client start failed"
7506 check_mount || error "check_mount failed"
7509 nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g")
7513 uuid=$(get_client_uuid $MOUNT)
7515 echo "list nids on mdt:"
7516 do_facet $SINGLEMDS "$LCTL list_param mdt.${FSNAME}*.exports.*"
7517 echo "uuid from $nid:"
7518 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
7520 found=$(check_uuid_on_mdt $nid | grep $uuid)
7521 [ -z "$found" ] && error "can't find $uuid $nid on MDT"
7522 found=$(check_uuid_on_ost $nid | grep $uuid)
7523 [ -z "$found" ] && error "can't find $uuid $nid on OST"
7525 # umount the client so it won't reconnect
7526 manual_umount_client --force || error "failed to umount $?"
7527 # shouldn't disappear on MDS after forced umount
7528 found=$(check_uuid_on_mdt $nid | grep $uuid)
7529 [ -z "$found" ] && error "can't find $uuid $nid"
7532 do_facet $SINGLEMDS \
7533 "$LCTL set_param -n mdt.${mds1_svc}.evict_client nid:$nid"
7534 sleep 1 # eviction above is async, give it some time to proceed
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"
7541 # check it didn't reconnect (being umounted)
7542 sleep $((TIMEOUT+1))
7543 found=$(check_uuid_on_mdt $nid | grep $uuid)
7544 [ -n "$found" ] && error "found $uuid $nid on MDT"
7545 found=$(check_uuid_on_ost $nid | grep $uuid)
7546 [ -n "$found" ] && error "found $uuid $nid on OST"
7550 run_test 91 "evict-by-nid support"
7552 generate_ldev_conf() {
7553 # generate an ldev.conf file
7554 local ldevconfpath=$1
7556 local fsldevformat=""
7559 local facets="mgs,$(get_facets OST),$(get_facets MDS)"
7560 for facet in ${facets//,/ }; do
7562 fstype=$(facet_fstype $facet)
7563 [ "$fstype" = zfs ] && fsldevformat="$fstype:"
7564 local host=$(facet_host $facet)
7566 local varfo=${facet}failover_HOST
7567 if [ $facet == mgs ] && combined_mgs_mds; then
7568 varfo=mds1failover_HOST
7570 [ -n "${!varfo}" ] && fo=${!varfo}
7571 local type=$(echo $facet | tr -d "[:digit:]" | \
7572 tr "[:lower:]" "[:upper:]" | sed s/MDS/MDT/ )
7574 [ ${facet} == mgs ] ||
7575 num=$(facet_number $facet)
7576 printf "%s\t%s\t%s-%s%04x\t%s%s\n" \
7583 $(facet_device $facet) >> $ldevconfpath
7586 echo "----- $ldevconfpath -----"
7588 echo "--- END $ldevconfpath ---"
7593 # generate a nids file (mapping between hostname to nid)
7594 # looks like we only have the MGS nid available to us
7595 # so just echo that to a file
7597 echo -e "${mgs_HOST}\t${MGSNID}" > $nidspath
7599 echo "----- $nidspath -----"
7601 echo "--- END $nidspath ---"
7604 compare_ldev_output() {
7605 local ldev_output=$1
7606 local expected_output=$2
7608 sort $expected_output -o $expected_output
7609 sort $ldev_output -o $ldev_output
7611 echo "-- START OF LDEV OUTPUT --"
7613 echo "--- END OF LDEV OUTPUT ---"
7615 echo "-- START OF EXPECTED OUTPUT --"
7616 cat $expected_output
7617 echo "--- END OF EXPECTED OUTPUT ---"
7619 diff $expected_output $ldev_output
7624 if [ -z "$LDEV" ]; then
7625 error "ldev is missing!"
7628 local LDEVCONFPATH=$TMP/ldev.conf
7629 local NIDSPATH=$TMP/nids
7631 echo "Host is $(hostname)"
7633 generate_ldev_conf $LDEVCONFPATH
7634 generate_nids $NIDSPATH
7636 # echo the mgs nid and compare it to environment variable MGSNID
7637 local facets="$(get_facets OST),$(get_facets MDS),mgs"
7638 for facet in ${facets//,/ }; do
7639 local host=$(facet_host $facet)
7640 local output=$($LDEV -c $LDEVCONFPATH -H $host -n $NIDSPATH echo %m)
7642 echo "-- START OF LDEV OUTPUT --"
7644 echo "--- END OF LDEV OUTPUT ---"
7647 error "ldev failed to execute!"
7649 # need to process multiple lines because of
7650 # several targets on host
7651 echo -e $output | awk '{ print $2 }' | while read -r line ; do
7652 [ "$line" = "$MGSNID" ] ||
7653 error "ldev failed mgs nid '$line', \
7657 rm -f $LDEVCONFPATH $NIDSPATH
7659 run_test 92 "ldev returns MGS NID correctly in command substitution"
7662 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs"
7665 #start mgs or mgs/mdt0
7666 if ! combined_mgs_mds ; then
7673 start_ost || error "OST0 start fail"
7675 #define OBD_FAIL_MGS_WRITE_TARGET_DELAY 0x90e
7676 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x8000090e"
7677 for num in $(seq 2 $MDSCOUNT); do
7681 mount_client $MOUNT || error "mount client fails"
7682 wait_osc_import_state mds ost FULL
7683 wait_osc_import_ready client ost
7684 check_mount || error "check_mount failed"
7686 cleanup || error "cleanup failed with $?"
7688 run_test 93 "register mulitple MDT at the same time"
7691 if [ -z "$LDEV" ]; then
7692 error "ldev is missing!"
7695 local LDEVCONFPATH=$TMP/ldev.conf
7696 local NIDSPATH=$TMP/nids
7698 generate_ldev_conf $LDEVCONFPATH
7699 generate_nids $NIDSPATH
7701 local LDEV_OUTPUT=$TMP/ldev-output.txt
7702 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME > $LDEV_OUTPUT
7704 # ldev failed, error
7705 if [ $? -ne 0 ]; then
7706 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7707 error "ldev failed to execute!"
7711 local EXPECTED_OUTPUT=$TMP/ldev-expected.txt
7713 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
7715 for num in $(seq $MDSCOUNT); do
7716 printf "%s-MDT%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7719 for num in $(seq $OSTCOUNT); do
7720 printf "%s-OST%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7723 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7725 if [ $? -ne 0 ]; then
7726 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7727 error "ldev failed to produce the correct hostlist!"
7730 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7732 run_test 94 "ldev outputs correct labels for file system name query"
7735 if [ -z "$LDEV" ]; then
7736 error "ldev is missing!"
7739 local LDEVCONFPATH=$TMP/ldev.conf
7740 local NIDSPATH=$TMP/nids
7742 generate_ldev_conf $LDEVCONFPATH
7743 generate_nids $NIDSPATH
7747 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME &>/dev/null
7748 if [ $? -ne 0 ]; then
7749 rm $LDEVCONFPATH $NIDSPATH
7750 error "ldev label filtering w/ -F failed!"
7754 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -l &>/dev/null
7755 if [ $? -ne 0 ]; then
7756 rm $LDEVCONFPATH $NIDSPATH
7757 error "ldev label filtering w/ -l failed!"
7761 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f &>/dev/null
7762 if [ $? -ne 0 ]; then
7763 rm $LDEVCONFPATH $NIDSPATH
7764 error "ldev label filtering w/ -f failed!"
7768 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a &>/dev/null
7769 if [ $? -ne 0 ]; then
7770 rm $LDEVCONFPATH $NIDSPATH
7771 error "ldev label filtering w/ -a failed!"
7776 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -F $FSNAME &>/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 -f &>/dev/null
7784 if [ $? -eq 0 ]; then
7785 rm $LDEVCONFPATH $NIDSPATH
7786 error "ldev label filtering w/ -a and -f incorrectly succeeded"
7790 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -l &>/dev/null
7791 if [ $? -eq 0 ]; then
7792 rm $LDEVCONFPATH $NIDSPATH
7793 error "ldev label filtering w/ -a and -l incorrectly succeeded"
7797 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f -l &>/dev/null
7798 if [ $? -eq 0 ]; then
7799 rm $LDEVCONFPATH $NIDSPATH
7800 error "ldev label filtering w/ -f and -l incorrectly succeeded"
7804 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -l &>/dev/null
7805 if [ $? -eq 0 ]; then
7806 rm $LDEVCONFPATH $NIDSPATH
7807 error "ldev label filtering w/ -F and -l incorrectly succeeded"
7810 # file sys & foreign
7811 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -f &>/dev/null
7812 if [ $? -eq 0 ]; then
7813 rm $LDEVCONFPATH $NIDSPATH
7814 error "ldev label filtering w/ -F and -f incorrectly succeeded"
7817 rm $LDEVCONFPATH $NIDSPATH
7819 run_test 95 "ldev should only allow one label filter"
7822 if [ -z "$LDEV" ]; then
7823 error "ldev is missing!"
7826 local LDEVCONFPATH=$TMP/ldev.conf
7827 local NIDSPATH=$TMP/nids
7829 generate_ldev_conf $LDEVCONFPATH
7830 generate_nids $NIDSPATH
7832 local LDEV_OUTPUT=$TMP/ldev-output.txt
7833 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -H $mgs_HOST \
7835 awk '{print $2}' > $LDEV_OUTPUT
7837 # ldev failed, error
7838 if [ $? -ne 0 ]; then
7839 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7840 error "ldev failed to execute!"
7844 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
7846 echo "$mgs_HOST-$(facet_fstype mgs)" > $EXPECTED_OUTPUT
7848 local facets="$(get_facets OST),$(get_facets MDS)"
7849 for facet in ${facets//,/ }; do
7850 local host=$(facet_host $facet)
7851 [ "$mgs_HOST" == "$host" ] &&
7852 echo "$host-$(facet_fstype $facet)" \
7856 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7858 if [ $? -ne 0 ]; then
7859 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7860 error "ldev failed to produce the correct output!"
7863 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7865 run_test 96 "ldev returns hostname and backend fs correctly in command sub"
7868 if [ -z "$LDEV" ]; then
7869 error "ldev is missing!"
7872 local LDEVCONFPATH=$TMP/ldev.conf
7873 local NIDSPATH=$TMP/nids
7875 generate_ldev_conf $LDEVCONFPATH
7876 generate_nids $NIDSPATH
7878 local LDEV_OUTPUT=$TMP/ldev-output.txt
7879 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
7881 echo -e "\nMDT role"
7882 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mdt > $LDEV_OUTPUT
7884 if [ $? -ne 0 ]; then
7885 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7886 error "ldev failed to execute for mdt role!"
7889 for num in $(seq $MDSCOUNT); do
7890 printf "%s-MDT%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7893 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7895 if [ $? -ne 0 ]; then
7896 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7897 error "ldev failed to produce the correct output for mdt role!"
7900 echo -e "\nOST role"
7901 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R ost > $LDEV_OUTPUT
7903 if [ $? -ne 0 ]; then
7904 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
7905 error "ldev failed to execute for ost role!"
7909 for num in $(seq $OSTCOUNT); do
7910 printf "%s-OST%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7913 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7915 if [ $? -ne 0 ]; then
7916 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7917 error "ldev failed to produce the correct output for ost role!"
7920 echo -e "\nMGS role"
7921 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mgs > $LDEV_OUTPUT
7923 if [ $? -ne 0 ]; then
7924 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
7925 error "ldev failed to execute for mgs role!"
7928 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
7930 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7932 if [ $? -ne 0 ]; then
7933 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7934 error "ldev failed to produce the correct output for mgs role!"
7937 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7939 run_test 97 "ldev returns correct ouput when querying based on role"
7944 local temp=$MDS_MOUNT_OPTS
7947 check_mount || error "mount failed"
7948 mountopt="user_xattr"
7949 for ((x = 1; x <= PAGE_SIZE/11; x++)); do
7950 mountopt="$mountopt,user_xattr"
7952 mount_client $MOUNT remount,$mountopt 2>&1 | grep "too long" ||
7953 error "Buffer overflow check failed"
7954 cleanup || error "cleanup failed"
7956 run_test 98 "Buffer-overflow check while parsing mount_opts"
7960 [[ "$ost1_FSTYPE" != ldiskfs ]] &&
7961 skip "ldiskfs only test"
7962 [[ "$OST1_VERSION" -ge $(version_code 2.8.57) ]] ||
7963 skip "Need OST version at least 2.8.57"
7965 local ost_opts="$(mkfs_opts ost1 $(ostdevname 1)) \
7966 --reformat $(ostdevname 1) $(ostvdevname 1)"
7967 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" &&
7968 skip "meta_bg already set"
7971 if [[ ${!opts} != *mkfsoptions* ]]; then
7972 eval opts=\"${!opts} \
7973 --mkfsoptions='\\\"-O ^resize_inode,meta_bg\\\"'\"
7975 local val=${!opts//--mkfsoptions=\\\"/ \
7976 --mkfsoptions=\\\"-O ^resize_inode,meta_bg }
7980 echo "params: $opts"
7983 add ost1 $opts || error "add ost1 failed with new params"
7985 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" ||
7986 error "meta_bg is not set"
7988 run_test 99 "Adding meta_bg option"
7993 start_mgsmds || error "MDS start failed"
7994 start_ost || error "unable to start OST"
7995 mount_client $MOUNT || error "client start failed"
7996 check_mount || error "check_mount failed"
8005 do_facet mgs 'lshowmount -v' | awk 'BEGIN {NR == 0; rc=1} /MGS:/ {rc=0}
8006 END {exit rc}' || error "lshowmount have no output MGS"
8008 do_facet mds1 'lshowmount -v' | awk 'BEGIN {NR == 2; rc=1} /-MDT0000:/
8009 {rc=0} END {exit rc}' || error "lshowmount have no output MDT0"
8011 do_facet ost1 'lshowmount -v' | awk 'BEGIN {NR == 4; rc=1} /-OST0000:/
8012 {rc=0} END {exit rc}' || error "lshowmount have no output OST0"
8014 cleanup || error "cleanup failed with $?"
8016 run_test 100 "check lshowmount lists MGS, MDT, OST and 0@lo"
8019 local createmany_pid
8020 local dev=$FSNAME-OST0000-osc-MDT0000
8024 do_nodes $(comma_list $(osts_nodes)) $LCTL set_param \
8025 seq.*OST*-super.width=$DATA_SEQ_MAX_WIDTH
8026 createmany -o $DIR1/$tdir/$tfile-%d 50000 &
8028 # MDT->OST reconnection causes MDT<->OST last_id synchornisation
8029 # via osp_precreate_cleanup_orphans.
8030 for ((i = 0; i < 100; i++)); do
8031 for ((k = 0; k < 10; k++)); do
8032 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;" \
8033 "$LCTL --device $dev activate"
8036 ls -asl $MOUNT | grep '???' &&
8037 { kill -9 $createmany_pid &>/dev/null;
8038 error "File has no object on OST"; }
8040 kill -s 0 $createmany_pid || break
8042 wait $createmany_pid
8044 unlinkmany $DIR1/$tdir/$tfile-%d 50000
8047 run_test 101a "Race MDT->OST reconnection with create"
8050 local dev=$FSNAME-OST0000-osc-MDT0000
8051 local dir=$DIR1/$tdir
8055 $LFS setstripe -c 1 -i 0 $dir
8056 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;"
8057 #define OBD_FAIL_OSP_CON_EVENT_DELAY 0x2107
8058 do_facet mds1 "$LCTL set_param fail_loc=0x80002107 fail_val=20"
8059 do_facet $SINGLEMDS "$LCTL --device $dev activate;"
8064 wait_osc_import_ready client ost1
8065 touch $dir/$tfile || error "Can't create file"
8069 run_test 101b "Race events DISCONNECT and ACTIVE in osp"
8072 [[ "$MDS1_VERSION" -gt $(version_code 2.9.53) ]] ||
8073 skip "Need server version greater than 2.9.53"
8074 [[ “$(mdsdevname 1)” != “$(mgsdevname)” ]] &&
8075 [[ “$(facet_host mds1)” = “$(facet_host mgs)” ]] &&
8076 skip "MGS must be on different node or combined"
8078 cleanup || error "cleanup failed with $?"
8080 local mds1dev=$(mdsdevname 1)
8081 local mds1mnt=$(facet_mntpt mds1)
8082 local mds1opts=$MDS_MOUNT_OPTS
8084 if [ "$mds1_FSTYPE" == ldiskfs ] &&
8085 ! do_facet mds1 test -b $mds1dev; then
8086 mds1opts=$(csa_add "$mds1opts" -o loop)
8088 if [[ "$mds1_FSTYPE" == zfs ]]; then
8089 import_zpool mds1 || return ${PIPESTATUS[0]}
8092 # unload all and only load libcfs to allow fail_loc setting
8093 do_facet mds1 $LUSTRE_RMMOD || error "unable to unload modules"
8094 do_facet mds1 modprobe libcfs || error "libcfs not loaded"
8095 do_facet mds1 lsmod \| grep libcfs || error "libcfs not loaded"
8097 #define OBD_FAIL_OBDCLASS_MODULE_LOAD 0x60a
8098 do_facet mds1 "$LCTL set_param fail_loc=0x8000060a"
8100 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts &&
8101 error "mdt start must fail"
8102 do_facet mds1 lsmod \| grep obdclass && error "obdclass must not load"
8104 do_facet mds1 "$LCTL set_param fail_loc=0x0"
8106 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts ||
8107 error "mdt start must not fail"
8109 cleanup || error "cleanup failed with $?"
8111 run_test 102 "obdclass module cleanup upon error"
8116 echo "rename $FSNAME to $newname"
8118 if ! combined_mgs_mds ; then
8119 local dev=$(mgsdevname)
8122 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
8123 error "(7) Fail to rename MGS"
8124 if [ "$(facet_fstype mgs)" = "zfs" ]; then
8125 reimport_zpool mgs $newname-mgs
8129 for num in $(seq $MDSCOUNT); do
8130 local dev=$(mdsdevname $num)
8132 do_facet mds${num} \
8133 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
8134 error "(8) Fail to rename MDT $num"
8135 if [ "$(facet_fstype mds${num})" = "zfs" ]; then
8136 reimport_zpool mds${num} $newname-mdt${num}
8140 for num in $(seq $OSTCOUNT); do
8141 local dev=$(ostdevname $num)
8143 do_facet ost${num} \
8144 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
8145 error "(9) Fail to rename OST $num"
8146 if [ "$(facet_fstype ost${num})" = "zfs" ]; then
8147 reimport_zpool ost${num} $newname-ost${num}
8152 test_103_set_pool() {
8156 do_facet mgs $LCTL pool_add $FSNAME.$pname ${FSNAME}-$ost_x ||
8157 error "Fail to add $ost_x to $FSNAME.$pname"
8158 wait_update $HOSTNAME \
8159 "lctl get_param -n lov.$FSNAME-clilov-*.pools.$pname |
8160 grep $ost_x" "$FSNAME-${ost_x}_UUID" ||
8161 error "$ost_x is NOT in pool $FSNAME.$pname"
8164 test_103_check_pool() {
8165 local save_fsname=$1
8168 stat $DIR/$tdir/test-framework.sh ||
8169 error "($errno) Fail to stat"
8170 do_facet mgs $LCTL pool_list $FSNAME.pool1 ||
8171 error "($errno) Fail to list $FSNAME.pool1"
8172 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname ||
8173 error "($errno) Fail to list $FSNAME.$save_fsname"
8174 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname |
8175 grep ${FSNAME}-OST0000 ||
8176 error "($errno) List $FSNAME.$save_fsname is invalid"
8178 local pname=$($LFS getstripe --pool $DIR/$tdir/d0)
8179 [ "$pname" = "$save_fsname" ] ||
8180 error "($errno) Unexpected pool name $pname"
8184 check_mount_and_prep
8186 mkdir $DIR/$tdir || error "(1) Fail to mkdir $DIR/$tdir"
8187 cp $LUSTRE/tests/test-framework.sh $DIR/$tdir ||
8188 error "(2) Fail to copy test-framework.sh"
8190 do_facet mgs $LCTL pool_new $FSNAME.pool1 ||
8191 error "(3) Fail to create $FSNAME.pool1"
8192 # name the pool name as the fsname
8193 do_facet mgs $LCTL pool_new $FSNAME.$FSNAME ||
8194 error "(4) Fail to create $FSNAME.$FSNAME"
8196 test_103_set_pool $FSNAME OST0000
8198 $LFS setstripe -p $FSNAME $DIR/$tdir/d0 ||
8199 error "(6) Fail to setstripe on $DIR/$tdir/d0"
8204 test_renamefs mylustre
8206 local save_fsname=$FSNAME
8210 test_103_check_pool $save_fsname 7
8212 if [ $OSTCOUNT -ge 2 ]; then
8213 test_103_set_pool $save_fsname OST0001
8216 $LFS setstripe -p $save_fsname $DIR/$tdir/f0 ||
8217 error "(16) Fail to setstripe on $DIR/$tdir/f0"
8226 test_103_check_pool $save_fsname 17
8230 test_renamefs $save_fsname
8236 run_test 103 "rename filesystem name"
8238 test_104a() { # LU-6952
8239 local mds_mountopts=$MDS_MOUNT_OPTS
8240 local ost_mountopts=$OST_MOUNT_OPTS
8241 local mds_mountfsopts=$MDS_MOUNT_FS_OPTS
8242 local lctl_ver=$(do_facet $SINGLEMDS $LCTL --version |
8245 [[ $(version_code $lctl_ver) -lt $(version_code 2.9.55) ]] &&
8246 skip "this test needs utils above 2.9.55"
8248 # specify "acl" in mount options used by mkfs.lustre
8249 if [ -z "$MDS_MOUNT_FS_OPTS" ]; then
8250 MDS_MOUNT_FS_OPTS="acl,user_xattr"
8253 MDS_MOUNT_FS_OPTS="${MDS_MOUNT_FS_OPTS},acl,user_xattr"
8256 echo "mountfsopt: $MDS_MOUNT_FS_OPTS"
8258 #reformat/remount the MDT to apply the MDT_MOUNT_FS_OPT options
8260 if ! combined_mgs_mds ; then
8264 if [ -z "$MDS_MOUNT_OPTS" ]; then
8265 MDS_MOUNT_OPTS="-o noacl"
8267 MDS_MOUNT_OPTS="${MDS_MOUNT_OPTS},noacl"
8270 for num in $(seq $MDSCOUNT); do
8271 start mds$num $(mdsdevname $num) $MDS_MOUNT_OPTS ||
8272 error "Failed to start MDS"
8275 for num in $(seq $OSTCOUNT); do
8276 start ost$num $(ostdevname $num) $OST_MOUNT_OPTS ||
8277 error "Failed to start OST"
8281 setfacl -m "d:$RUNAS_ID:rwx" $MOUNT &&
8282 error "ACL is applied when FS is mounted with noacl."
8284 MDS_MOUNT_OPTS=$mds_mountopts
8285 OST_MOUNT_OPTS=$ost_mountopts
8286 MDS_MOUNT_FS_OPTS=$mds_mountfsopts
8288 run_test 104a "Make sure user defined options are reflected in mount"
8290 test_104b() { # LU-12859
8291 mount_client $MOUNT3 flock,localflock
8292 stack_trap "umount_client $MOUNT3" EXIT
8293 mount | grep "$MOUNT3 .*,flock" && error "flock is still set"
8294 mount | grep "$MOUNT3 .*,localflock" || error "localflock is not set"
8295 umount_client $MOUNT3
8296 mount_client $MOUNT3 localflock,flock
8297 mount | grep "$MOUNT3 .*,localflock" && error "localflock is still set"
8298 mount | grep "$MOUNT3 .*,flock" || error "flock is not set"
8299 umount_client $MOUNT3
8300 mount_client $MOUNT3 localflock,flock,noflock
8301 flock_is_enabled $MOUNT3 && error "some flock is still enabled" || true
8303 run_test 104b "Mount uses last flock argument"
8305 error_and_umount() {
8316 mount --bind $DIR $TMP/$tdir || error "mount bind mnt pt failed"
8317 rm -f $TMP/$tdir/$tfile
8318 rm -f $TMP/$tdir/${tfile}1
8320 # Files should not be created in ro bind mount point
8321 # remounting from rw to ro
8322 mount -o remount,ro $TMP/$tdir ||
8323 error_and_umount "readonly remount of bind mnt pt failed"
8324 touch $TMP/$tdir/$tfile &&
8325 error_and_umount "touch succeeds on ro bind mnt pt"
8326 [ -e $TMP/$tdir/$tfile ] &&
8327 error_and_umount "file created on ro bind mnt pt"
8329 # Files should be created in rw bind mount point
8330 # remounting from ro to rw
8331 mount -o remount,rw $TMP/$tdir ||
8332 error_and_umount "read-write remount of bind mnt pt failed"
8333 touch $TMP/$tdir/${tfile}1 ||
8334 error_and_umount "touch fails on rw bind mnt pt"
8335 [ -e $TMP/$tdir/${tfile}1 ] ||
8336 error_and_umount "file not created on rw bind mnt pt"
8337 umount $TMP/$tdir || error "umount of bind mnt pt failed"
8339 cleanup || error "cleanup failed with $?"
8341 run_test 105 "check file creation for ro and rw bind mnt pt"
8348 mkdir -p $DIR/$tdir || error "create $tdir failed"
8349 do_nodes $(comma_list $(osts_nodes)) $LCTL set_param \
8350 seq.*OST*-super.width=$DATA_SEQ_MAX_WIDTH
8351 lfs setstripe -c 1 -i 0 $DIR/$tdir
8352 #define OBD_FAIL_CAT_RECORDS 0x1312
8353 do_facet mds1 $LCTL set_param fail_loc=0x1312 fail_val=$repeat
8355 for ((i = 1; i <= $repeat; i++)); do
8357 #one full plain llog
8358 createmany -o $DIR/$tdir/f- 64768
8360 createmany -u $DIR/$tdir/f- 64768
8362 wait_delete_completed $((TIMEOUT * 7))
8363 #ASSERTION osp_sync_thread() ( thread->t_flags != SVC_RUNNING ) failed
8364 #shows that osp code is buggy
8365 do_facet mds1 $LCTL set_param fail_loc=0 fail_val=0
8369 run_test 106 "check osp llog processing when catalog is wrapped"
8372 [[ "$MDS1_VERSION" -ge $(version_code 2.10.50) ]] ||
8373 skip "Need MDS version > 2.10.50"
8376 start_mgsmds || error "start_mgsmds failed"
8377 start_ost || error "unable to start OST"
8379 # add unknown configuration parameter.
8380 if [[ $PERM_CMD == *"set_param -P"* ]]; then
8381 cmd="$PERM_CMD ost.$FSNAME-OST0000*.unknown_param"
8383 cmd="$PERM_CMD $FSNAME-OST0000*.ost.unknown_param"
8385 do_facet mgs "$cmd=50"
8386 cleanup_nocli || error "cleanup_nocli failed with $?"
8389 # unknown param should be ignored while mounting.
8390 start_ost || error "unable to start OST after unknown param set"
8392 cleanup || error "cleanup failed with $?"
8394 run_test 107 "Unknown config param should not fail target mounting"
8399 $rcmd rm -rf $tmp > /dev/null 2>&1
8400 $rcmd mkdir -p $tmp/{mnt,images} || error "failed to mkdir remotely"
8402 for facet in $facets; do
8403 [ "$mds1_FSTYPE" = zfs ] &&
8404 $rcmd $ZPOOL -f export lustre-$facet > /dev/null 2>&1
8405 $rcmd mkdir $tmp/mnt/$facet ||
8406 error "failed to mkdir $tmp/mnt/$facet"
8407 $rcmd dd if=/dev/zero of=$tmp/images/$facet \
8408 seek=199 bs=1M count=1 ||
8409 error "failed to create $tmp/images/$facet"
8418 local facet=${role}$((idx + 1))
8420 [ $# -eq 5 ] && pool=$5
8422 do_facet $SINGLEMDS $MKFS --fsname=lustre --$mgs \
8423 --$role --index=$idx --replace --backfstype=$bkfs \
8424 --device-size=200000 --reformat $pool $tmp/images/$facet ||
8425 error "failed to mkfs for $facet"
8429 echo "mounting client..."
8430 mount -t lustre ${nid}:/lustre $MOUNT ||
8431 error "failed to mount lustre"
8434 ls -l $MOUNT/local_dir || error "failed to list"
8436 echo "check truncate && write"
8437 echo "dummmmmmmmmmmmm" > $MOUNT/remote_dir/fsx.c ||
8438 error "failed to tuncate & write"
8441 touch $MOUNT/foooo ||
8442 error "failed to create"
8444 echo "check read && write && append"
8445 sha1sum $MOUNT/conf-sanity.sh |
8446 awk '{ print $1 }' > $MOUNT/checksum.new ||
8447 error "failed to read(1)"
8448 sha1sum $MOUNT/remote_dir/unlinkmany.c |
8449 awk '{ print $1 }' >> $MOUNT/checksum.new ||
8450 error "failed to read(2)"
8451 sha1sum $MOUNT/striped_dir/lockahead_test.o |
8452 awk '{ print $1 }' >> $MOUNT/checksum.new ||
8453 error "failed to read(3)"
8456 diff $MOUNT/checksum.new $MOUNT/checksum.src ||
8457 error "failed to verify data"
8467 umount -f $MOUNT || error "failed to umount client"
8468 for facet in $facets; do
8469 $rcmd umount -f $tmp/mnt/$facet ||
8470 error "failed to umount $facet"
8471 if [ "$mds1_FSTYPE" = zfs ]; then
8472 $rcmd $ZPOOL export -f lustre-$facet ||
8473 error "failed to export lustre-$facet"
8477 $rcmd rm -rf $tmp || error "failed to rm the dir $tmp"
8481 [ "$CLIENTONLY" ] && skip "Client-only testing"
8482 [ "$mds1_FSTYPE" != zfs ] && skip "zfs only test"
8483 [ "$MDS1_VERSION" -lt $(version_code 2.10.58) ] &&
8484 skip "Need server version at least 2.10.58"
8489 local tmp=$TMP/$tdir
8490 local rcmd="do_facet $SINGLEMDS"
8491 local facets="mdt1 mdt2 ost1 ost2"
8492 local nid=$($rcmd $LCTL list_nids | head -1)
8495 trap t_108_cleanup EXIT ERR
8498 t_108_mkfs mdt 0 zfs mgs lustre-mdt1/mdt1
8499 t_108_mkfs mdt 1 zfs mgsnode=$nid lustre-mdt2/mdt2
8500 t_108_mkfs ost 0 zfs mgsnode=$nid lustre-ost1/ost1
8501 t_108_mkfs ost 1 zfs mgsnode=$nid lustre-ost2/ost2
8503 for facet in $facets; do
8504 $rcmd zfs set mountpoint=$tmp/mnt/$facet canmount=on \
8505 lustre-$facet/$facet ||
8506 error "failed to zfs set for $facet (1)"
8507 $rcmd zfs mount lustre-$facet/$facet ||
8508 error "failed to local mount $facet"
8509 $rcmd tar jxf $LUSTRE/tests/ldiskfs_${facet}_2_11.tar.bz2 \
8510 --xattrs --xattrs-include="trusted.*" \
8511 -C $tmp/mnt/$facet/ > /dev/null 2>&1 ||
8512 error "failed to untar image for $facet"
8513 $rcmd "cd $tmp/mnt/$facet && rm -rf oi.* OI_* lfsck_* LFSCK" ||
8514 error "failed to cleanup for $facet"
8515 $rcmd zfs umount lustre-$facet/$facet ||
8516 error "failed to local umount $facet"
8517 $rcmd zfs set canmount=off lustre-$facet/$facet ||
8518 error "failed to zfs set $facet (2)"
8521 echo "changing server nid..."
8522 $rcmd mount -t lustre -o nosvc lustre-mdt1/mdt1 $tmp/mnt/mdt1
8523 $rcmd lctl replace_nids $FSNAME-MDT0000 $nid
8524 $rcmd lctl replace_nids $FSNAME-MDT0001 $nid
8525 $rcmd lctl replace_nids $FSNAME-OST0000 $nid
8526 $rcmd lctl replace_nids $FSNAME-OST0001 $nid
8527 $rcmd umount $tmp/mnt/mdt1
8529 for facet in $facets; do
8530 echo "mounting $facet from backup..."
8531 $rcmd mount -t lustre -o abort_recov lustre-$facet/$facet \
8532 $tmp/mnt/$facet || error "failed to mount $facet"
8535 # ZFS backend can detect migration and trigger OI scrub automatically
8536 # sleep 3 seconds for scrub done
8542 run_test 108a "migrate from ldiskfs to ZFS"
8545 [ "$CLIENTONLY" ] && skip "Client-only testing"
8546 [ "$mds1_FSTYPE" != ldiskfs ] && skip "ldiskfs only test"
8547 [ "$MDS1_VERSION" -lt $(version_code 2.10.58) ] &&
8548 skip "Need server version at least 2.10.58"
8553 local tmp=$TMP/$tdir
8554 local rcmd="do_facet $SINGLEMDS"
8555 local facets="mdt1 mdt2 ost1 ost2"
8556 local scrub_list="MDT0000 MDT0001 OST0000 OST0001"
8557 local nid=$($rcmd $LCTL list_nids | head -1)
8560 trap t_108_cleanup EXIT ERR
8563 t_108_mkfs mdt 0 ldiskfs mgs
8564 t_108_mkfs mdt 1 ldiskfs mgsnode=$nid
8565 t_108_mkfs ost 0 ldiskfs mgsnode=$nid
8566 t_108_mkfs ost 1 ldiskfs mgsnode=$nid
8568 for facet in $facets; do
8569 $rcmd mount -t ldiskfs -o loop $tmp/images/$facet \
8571 error "failed to local mount $facet"
8573 $rcmd tar jxf $LUSTRE/tests/zfs_${facet}_2_11.tar.bz2 \
8574 --xattrs --xattrs-include="*.*" \
8575 -C $tmp/mnt/$facet/ > /dev/null 2>&1 ||
8576 error "failed to untar image for $facet"
8577 $rcmd "cd $tmp/mnt/$facet && rm -rf oi.* OI_* lfsck_* LFSCK" ||
8578 error "failed to cleanup for $facet"
8579 $rcmd umount $tmp/mnt/$facet ||
8580 error "failed to local umount $facet"
8583 echo "changing server nid..."
8584 $rcmd mount -t lustre -o nosvc,loop $tmp/images/mdt1 $tmp/mnt/mdt1
8585 $rcmd lctl replace_nids $FSNAME-MDT0000 $nid
8586 $rcmd lctl replace_nids $FSNAME-MDT0001 $nid
8587 $rcmd lctl replace_nids $FSNAME-OST0000 $nid
8588 $rcmd lctl replace_nids $FSNAME-OST0001 $nid
8589 $rcmd umount $tmp/mnt/mdt1
8591 for facet in $facets; do
8592 echo "mounting $facet from backup..."
8593 $rcmd mount -t lustre -o loop,abort_recov $tmp/images/$facet \
8594 $tmp/mnt/$facet || error "failed to mount $facet"
8597 for facet in $scrub_list; do
8598 $rcmd $LCTL lfsck_start -M $FSNAME-$facet -t scrub ||
8599 error "failed to start OI scrub on $facet"
8602 # sleep 3 seconds for scrub done
8608 run_test 108b "migrate from ZFS to ldiskfs"
8611 # set number of permanent parameters
8613 test_109_set_params() {
8616 set_persistent_param_and_check mds \
8617 "mdd.$fsname-MDT0000.atime_diff" \
8618 "$fsname-MDT0000.mdd.atime_diff" \
8620 set_persistent_param_and_check mds \
8621 "mdd.$fsname-MDT0000.atime_diff" \
8622 "$fsname-MDT0000.mdd.atime_diff" \
8624 set_persistent_param_and_check client \
8625 "llite.$fsname*.max_read_ahead_mb" \
8626 "$fsname.llite.max_read_ahead_mb" \
8628 set_persistent_param_and_check client \
8629 "llite.$fsname*.max_read_ahead_mb" \
8630 "$fsname.llite.max_read_ahead_mb" \
8632 create_pool $fsname.pool1 || error "create pool failed"
8633 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
8634 error "pool_add failed"
8635 do_facet mgs $LCTL pool_remove $fsname.pool1 OST0000 ||
8636 error "pool_remove failed"
8637 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
8638 error "pool_add failed"
8642 # check permanent parameters
8644 test_109_test_params() {
8647 local atime_diff=$(do_facet mds $LCTL \
8648 get_param -n mdd.$fsname-MDT0000.atime_diff)
8649 [ $atime_diff == 63 ] || error "wrong mdd parameter after clear_conf"
8650 local max_read_ahead_mb=$(do_facet client $LCTL \
8651 get_param -n llite.$fsname*.max_read_ahead_mb)
8652 [ $max_read_ahead_mb == 64 ] ||
8653 error "wrong llite parameter after clear_conf"
8654 local ost_in_pool=$(do_facet mds $LCTL pool_list $fsname.pool1 |
8655 grep -v "^Pool:" | sed 's/_UUID//')
8656 [ $ost_in_pool = "$fsname-OST0000" ] ||
8657 error "wrong pool after clear_conf"
8661 # run lctl clear_conf, store CONFIGS before and after that
8663 test_109_clear_conf()
8665 local clear_conf_arg=$1
8668 if ! combined_mgs_mds ; then
8669 mgsdev=$(mgsdevname)
8670 stop_mgs || error "stop_mgs failed"
8671 start_mgs "-o nosvc" || error "start_mgs nosvc failed"
8673 mgsdev=$(mdsdevname 1)
8674 start_mdt 1 "-o nosvc" || error "start_mdt 1 nosvc failed"
8677 do_facet mgs "rm -rf $TMP/${tdir}/conf1; mkdir -p $TMP/${tdir}/conf1;" \
8678 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf1\\\" \
8682 # the command being tested
8684 do_facet mgs $LCTL clear_conf $clear_conf_arg ||
8685 error "clear_conf failed"
8686 if ! combined_mgs_mds ; then
8687 stop_mgs || error "stop_mgs failed"
8689 stop_mdt 1 || error "stop_mdt 1 failed"
8692 do_facet mgs "rm -rf $TMP/${tdir}/conf2; mkdir -p $TMP/${tdir}/conf2;" \
8693 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf2\\\" \
8697 test_109_file_shortened() {
8699 local sizes=($(do_facet mgs "stat -c %s " \
8700 "$TMP/${tdir}/conf1/CONFIGS/$file" \
8701 "$TMP/${tdir}/conf2/CONFIGS/$file"))
8702 [ ${sizes[1]} -lt ${sizes[0]} ] && return 0
8708 [ $MDS1_VERSION -lt $(version_code 2.10.59) ] &&
8709 skip "Needs MDS version 2.10.59 or later."
8710 [ "$(facet_fstype mgs)" == zfs ] &&
8711 skip "LU-8727: no implementation for ZFS"
8716 client_up || error "client_up failed"
8719 # set number of permanent parameters
8721 test_109_set_params $FSNAME
8723 umount_client $MOUNT || error "umount_client failed"
8724 stop_ost || error "stop_ost failed"
8725 stop_mds || error "stop_mds failed"
8727 test_109_clear_conf $FSNAME
8729 # make sure that all configs are cleared
8731 test_109_file_shortened $FSNAME-MDT0000 ||
8732 error "failed to clear MDT0000 config"
8733 test_109_file_shortened $FSNAME-client ||
8734 error "failed to clear client config"
8739 # check that configurations are intact
8741 test_109_test_params $FSNAME
8746 destroy_test_pools || error "destroy test pools failed"
8750 run_test 109a "test lctl clear_conf fsname"
8754 [ $MDS1_VERSION -lt $(version_code 2.10.59) ] &&
8755 skip "Needs MDS version 2.10.59 or later."
8756 [ "$(facet_fstype mgs)" == zfs ] &&
8757 skip "LU-8727: no implementation for ZFS"
8762 client_up || error "client_up failed"
8765 # set number of permanent parameters
8767 test_109_set_params $FSNAME
8769 umount_client $MOUNT || error "umount_client failed"
8770 stop_ost || error "stop_ost failed"
8771 stop_mds || error "stop_mds failed"
8773 test_109_clear_conf $FSNAME-MDT0000
8775 # make sure that only one config is cleared
8777 test_109_file_shortened $FSNAME-MDT0000 ||
8778 error "failed to clear MDT0000 config"
8779 test_109_file_shortened $FSNAME-client &&
8780 error "failed to clear client config"
8784 # check that configurations are intact
8786 test_109_test_params $FSNAME
8791 destroy_test_pools || error "destroy test pools failed"
8795 run_test 109b "test lctl clear_conf one config"
8799 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
8800 skip "Only applicable to ldiskfs-based MDTs"
8802 do_facet $SINGLEMDS $DEBUGFS -w -R supported_features |grep large_dir ||
8803 skip "large_dir option is not supported on MDS"
8804 do_facet ost1 $DEBUGFS -w -R supported_features | grep large_dir ||
8805 skip "large_dir option is not supported on OSS"
8807 stopall # stop all targets before modifying the target counts
8808 local old_mdscount=$MDSCOUNT
8809 local old_ostcount=$OSTCOUNT
8811 stack_trap "MDSCOUNT=$old_mdscount OSTCOUNT=$old_ostcount; reformat"
8815 # ext4_dir_entry_2 struct size:264
8816 # dx_root struct size:8
8817 # dx_node struct size:8
8818 # dx_entry struct size:8
8819 # For 1024 bytes block size.
8820 # First level directory entries: 126
8821 # Second level directory entries: 127
8822 # Entries in leaf: 3
8823 # For 2 levels limit: 48006
8824 # For 3 levels limit : 6096762
8825 # Create 80000 files to safely exceed 2-level htree limit.
8826 CONF_SANITY_110_LINKS=${CONF_SANITY_110_LINKS:-80000}
8828 # can fit at most 3 filenames per 1KB leaf block, but each
8829 # leaf/index block will only be 3/4 full before split at each level
8830 (( MDSSIZE < CONF_SANITY_110_LINKS / 3 * 4/3 * 4/3 )) &&
8831 CONF_SANITY_110_LINKS=$((MDSSIZE * 3 * 3/4 * 3/4))
8833 combined_mgs_mds || replace=" --replace "
8834 local opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
8835 $replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
8836 if [[ $opts != *large_dir* ]]; then
8837 if [[ $opts != *mkfsoptions* ]]; then
8838 opts+=" --mkfsoptions=\\\"-O large_dir -b 1024 -i 65536\\\""
8840 opts="${opts//--mkfsoptions=\\\"/ \
8841 --mkfsoptions=\\\"-O large_dir -b 1024 -i 65536 }"
8844 echo "MDT params: $opts"
8846 combined_mgs_mds || start_mgs
8847 add mds1 $opts || error "add mds1 failed with new params"
8848 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
8850 opts="$(mkfs_opts ost1 $(ostdevname 1)) \
8851 $replace --reformat $(ostdevname 1) $(ostvdevname 1)"
8853 if [[ $opts != *large_dir* ]]; then
8854 if [[ $opts != *mkfsoptions* ]]; then
8855 opts+=" --mkfsoptions=\\\"-O large_dir\\\" "
8857 opts="${opts//--mkfsoptions=\\\"/ \
8858 --mkfsoptions=\\\"-O large_dir }"
8861 echo "OST params: $opts"
8862 add ost1 $opts || error "add ost1 failed with new params"
8863 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
8865 MOUNT_2=yes mountcli || error "mount clients failed"
8867 mkdir -v $DIR/$tdir || error "cannot create $DIR/$tdir"
8868 local pids count=0 group=0
8870 echo "creating $CONF_SANITY_110_LINKS in total"
8871 while (( count < CONF_SANITY_110_LINKS )); do
8872 local len=$((253 - $(wc -c <<<"$tfile-$group-40000-")))
8873 local dir=DIR$((group % 2 + 1))
8874 local target=${!dir}/$tdir/$tfile-$group
8875 local long=$target-$(generate_name $len)-
8876 local create=$((CONF_SANITY_110_LINKS - count))
8878 (( create > 40000 )) && create=40000
8879 touch $target || error "creating $target failed"
8880 echo "creating $create hard links to $target"
8881 createmany -l $target $long $create &
8884 count=$((count + create))
8885 group=$((group + 1))
8887 echo "waiting for PIDs$pids to complete"
8888 wait $pids || error "createmany failed after $group groups"
8890 umount_client $MOUNT2 -f
8893 run_e2fsck $(facet_active_host mds1) $(mdsdevname 1) -n
8894 MDSCOUNT=$old_mdscount
8895 OSTCOUNT=$old_ostcount
8897 run_test 110 "Adding large_dir with 3-level htree"
8900 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
8901 skip "Only applicable to ldiskfs-based MDTs"
8903 is_dm_flakey_dev $SINGLEMDS $(mdsdevname 1) &&
8904 skip "This test can not be executed on flakey dev"
8906 do_facet $SINGLEMDS $DEBUGFS -w -R supported_features |grep large_dir ||
8907 skip "large_dir option is not supported on MDS"
8909 do_facet ost1 $DEBUGFS -w -R supported_features | grep large_dir ||
8910 skip "large_dir option is not supported on OSS"
8912 # cleanup before changing target counts
8914 local old_mdscount=$MDSCOUNT
8915 local old_ostcount=$OSTCOUNT
8916 local old_mdssize=$MDSSIZE
8918 stack_trap "MDSSIZE=$old_mdssize MDSCOUNT=$old_mdscount \
8919 OSTCOUNT=$old_ostcount; reformat"
8922 (( MDSSIZE < 2400000 )) && MDSSIZE=2400000 # need at least 2.4GB
8924 local mdsdev=$(mdsdevname 1)
8925 combined_mgs_mds || replace=" --replace "
8926 local opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
8927 $replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
8928 if [[ $opts != *large_dir* ]]; then
8929 if [[ $opts != *mkfsoptions* ]]; then
8930 opts+=" --mkfsoptions=\\\"-O large_dir -i 1048576 \\\" "
8932 opts="${opts//--mkfsoptions=\\\"/ \
8933 --mkfsoptions=\\\"-O large_dir -i 1048576 }"
8936 echo "MDT params: $opts"
8938 combined_mgs_mds || start_mgs
8939 __touch_device mds 1
8940 add mds1 $opts || error "add mds1 failed with new params"
8941 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
8943 opts="$(mkfs_opts ost1 $(ostdevname 1)) \
8944 $replace --reformat $(ostdevname 1) $(ostvdevname 1)"
8945 if [[ $opts != *large_dir* ]]; then
8946 if [[ $opts != *mkfsoptions* ]]; then
8947 opts+=" --mkfsoptions=\\\"-O large_dir \\\""
8949 opts="${opts//--mkfsoptions=\\\"/ --mkfsoptions=\\\"-O large_dir }"
8952 echo "OST params: $opts"
8953 __touch_device ost 1
8954 add ost1 $opts || error "add ost1 failed with new params"
8955 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
8957 MOUNT_2=yes mountcli
8958 mkdir $DIR/$tdir || error "cannot create $DIR/$tdir"
8960 lfs df -i $DIR/$tdir
8964 local start=$SECONDS
8966 local dirmax=$((2 << 30))
8972 while (( !needskip & dirsize < dirmax )); do
8975 for cli in ${CLIENTS//,/ }; do
8976 local len=$((253 - $(wc -c <<<"$cli-$group-60000-")))
8977 local target=$cli-$group
8978 local long=$DIR/$tdir/$target-$(generate_name $len)-
8980 RPWD=$DIR/$tdir do_node $cli touch $target ||
8981 error "creating $target failed"
8982 echo "creating 60000 hardlinks to $target"
8983 RPWD=$DIR/$tdir do_node $cli createmany -l $target $long 60000 &
8986 group=$((group + 1))
8988 long=$DIR2/$tdir/$target-$(generate_name $len)-
8990 RPWD=$DIR2/$tdir do_node $cli touch $target ||
8991 error "creating $target failed"
8992 echo "creating 60000 hardlinks to $target"
8993 RPWD=$DIR2/$tdir do_node $cli createmany -l $target $long 60000 &
8996 group=$((group + 1))
8998 echo "waiting for PIDs$pids to complete"
8999 wait $pids || error "createmany failed after $group groups"
9000 dirsize=$(stat -c %s $DIR/$tdir)
9001 taken=$((SECONDS - start))
9002 rate=$((dirsize / taken))
9003 left=$(((dirmax - dirsize) / rate))
9004 num=$((group * 60000))
9005 echo "estimate ${left}s left after $num files / ${taken}s"
9006 # if the estimated time remaining is too large (it may change
9007 # over time as the create rate is not constant) then exit
9008 # without declaring a failure.
9009 (( left > 1200 )) && needskip=1
9012 umount_client $MOUNT2 -f
9015 ! (( $needskip )) ||
9016 echo "ETA ${left}s after $num files / ${taken}s is too long"
9018 run_e2fsck $(facet_active_host mds1) $(mdsdevname 1) -n
9019 MDSCOUNT=$old_mdscount
9020 OSTCOUNT=$old_ostcount
9021 MDSSIZE=$old_mdssize
9023 run_test 111 "Adding large_dir with over 2GB directory"
9026 start_mds || error "MDS start failed"
9027 start_ost || error "OSS start failed"
9028 echo "start ost2 service on $(facet_active_host ost2)"
9029 start ost2 $(ostdevname 2) $(csa_add "$OST_MOUNT_OPTS" -o no_precreate) ||
9030 error "start ost2 facet failed"
9031 local val=$(do_facet ost2 \
9032 "$LCTL get_param -n obdfilter.$FSNAME-OST0001*.no_precreate")
9033 (( $val == 1 )) || error "obdfilter.$FSNAME-OST0001*.no_precreate=$val"
9035 mount_client $MOUNT || error "mount client failed"
9036 wait_osc_import_state mds1 ost1 FULL
9037 wait_osc_import_ready client ost1
9038 wait_osc_import_ready client ost2
9040 $LFS setstripe -i 0 $DIR/$tfile.0 ||
9041 error "problem creating $tfile.0 on OST0000"
9042 $LFS setstripe -i 1 $DIR/$tfile.1 && $LFS getstripe $DIR/$tfile.1 &&
9043 (( $($LFS getstripe -i $DIR/$tfile.1) == 1 )) &&
9044 error "allowed to create $tfile.1 on OST0001"
9045 do_facet ost2 $LCTL set_param obdfilter.*.no_precreate=0
9047 $LFS setstripe -i 1 $DIR/$tfile.2 ||
9048 error "failed to create $tfile.2 on ost1 facet"
9049 # files not cleaned with ONLY_REPEAT because of client unmount below
9050 rm -f $DIR/$tfile.[012]
9051 stop_ost2 || error "stop ost2 facet failed"
9054 run_test 112 "mount OST with nocreate option"
9057 SAVE_MGS_MOUNT_OPTS=$MGS_MOUNT_OPTS
9058 SAVE_MDS_MOUNT_OPTS=$MDS_MOUNT_OPTS
9059 SAVE_OST_MOUNT_OPTS=$OST_MOUNT_OPTS
9065 MGS_MOUNT_OPTS=$SAVE_MGS_MOUNT_OPTS
9066 MDS_MOUNT_OPTS=$SAVE_MDS_MOUNT_OPTS
9067 OST_MOUNT_OPTS=$SAVE_OST_MOUNT_OPTS
9068 # Revert old mount options back
9070 # Subsequent following test requires
9071 # conf-sanity to be in stopall state.
9072 # Force 'stopall' so others following
9077 # Error out with mount info
9079 local server_nodes=$(comma_list $(mdts_nodes) $(osts_nodes))
9082 echo "--Client Mount Info--"
9083 mount | grep -i lustre
9084 echo "--Server Mount Info--"
9085 do_nodes $server_nodes mount | grep -i lustre
9091 local ost_version="2.15.51" # Minimum version required
9093 (( OST1_VERSION >= $(version_code $ost_version) )) ||
9094 skip "Need server version at least $ost_version"
9096 stack_trap cleanup_113 EXIT
9098 # Reset before starting
9102 # Verify MDS's should start with "rw"
9103 do_facet $SINGLEMDS mount | grep "lustre.*rw,.*MDT" ||
9104 error_113 "$SINGLEMDS should be read-write"
9106 # Verify OST's should start with "rw"
9107 for (( i=1; i <= OSTCOUNT; i++ )); do
9108 do_facet ost$i mount | grep "lustre.*rw,.*OST" ||
9109 error_113 "ost$i should be read-write"
9112 # rdonly_dev does not currently work for ldiskfs
9113 # We skip the rdonly_dev check until then.
9114 if [[ $ost1_FSTYPE == ldiskfs ]]; then
9115 echo "Shadow Mountpoint correctly reports rw for ldiskfs"
9120 # Only ZFS specific tests below.
9123 # Must stop all (server+client) and restart to verify new
9127 # add rdonly_dev to mount option
9128 MGS_MOUNT_OPTS=$(csa_add "$MGS_MOUNT_OPTS" -o rdonly_dev)
9129 MDS_MOUNT_OPTS=$(csa_add "$MDS_MOUNT_OPTS" -o rdonly_dev)
9130 OST_MOUNT_OPTS=$(csa_add "$OST_MOUNT_OPTS" -o rdonly_dev)
9132 # Only restart server(mds/ost). Sufficient for test
9133 setupall server_only || error "Fail to start servers"
9135 # Verify MDS's should be "ro"
9136 do_facet $SINGLEMDS mount | grep "lustre.*ro,.*MDT.*rdonly_dev" ||
9137 error_113 "$SINGLEMDS should be read-only"
9139 # Verify OST's should be "ro"
9140 for (( i=1; i <= OSTCOUNT; i++ )); do
9141 do_facet ost$i mount | grep "lustre.*ro,.*OST.*rdonly_dev" ||
9142 error_113 "ost$i should be read-only"
9145 run_test 113 "Shadow mountpoint correctly report ro/rw for mounts"
9148 # Purpose: To verify dynamic thread (OSS) creation.
9149 # (This was sanity/115)
9152 [ $PARALLEL == "yes" ] && skip "skip parallel run"
9153 remote_ost_nodsh && skip "remote OST with nodsh"
9155 # Lustre does not stop service threads once they are started.
9156 # Reset number of running threads to default.
9161 local save_params="$TMP/sanity-$TESTNAME.parameters"
9163 # Get ll_ost_io count before I/O
9164 OSTIO_pre=$(do_facet ost1 \
9165 "$LCTL get_param ost.OSS.ost_io.threads_started | cut -d= -f2")
9166 # Exit if lustre is not running (ll_ost_io not running).
9167 [ -z "$OSTIO_pre" ] && error "no OSS threads"
9169 echo "Starting with $OSTIO_pre threads"
9170 local thread_max=$((OSTIO_pre * 2))
9171 local rpc_in_flight=$((thread_max * 2))
9173 # this is limited to OSC_MAX_RIF_MAX (256)
9174 [ $rpc_in_flight -gt 256 ] && rpc_in_flight=256
9175 thread_max=$((rpc_in_flight / 2))
9176 [ $thread_max -le $OSTIO_pre ] && skip "Too many ost_io threads" &&
9179 # Number of I/O Process proposed to be started.
9181 local facets=$(get_facets OST)
9183 save_lustre_params client "osc.*OST*.max_rpcs_in_flight" > $save_params
9184 save_lustre_params $facets "ost.OSS.ost_io.threads_max" >> $save_params
9186 # Set in_flight to $rpc_in_flight
9187 $LCTL set_param osc.*OST*.max_rpcs_in_flight=$rpc_in_flight ||
9188 error "Failed to set max_rpcs_in_flight to $rpc_in_flight"
9189 nfiles=${rpc_in_flight}
9190 # Set ost thread_max to $thread_max
9191 do_facet ost1 "$LCTL set_param ost.OSS.ost_io.threads_max=$thread_max"
9193 # 5 Minutes should be sufficient for max number of OSS
9194 # threads(thread_max) to be created.
9198 local wtl=${WTL:-"$LUSTRE/tests/write_time_limit"}
9200 test_mkdir $DIR/$tdir
9201 for ((i = 1; i <= nfiles; i++)); do
9202 local file=$DIR/$tdir/${tfile}-$i
9204 $LFS setstripe -c -1 -i 0 $file
9205 ($wtl $file $timeout)&
9208 # I/O Started - Wait for thread_started to reach thread_max or report
9209 # error if thread_started is more than thread_max.
9210 echo "Waiting for thread_started to reach thread_max"
9211 local thread_started=0
9212 local end_time=$((SECONDS + timeout))
9214 while [ $SECONDS -le $end_time ] ; do
9216 # Get ost i/o thread_started count.
9217 thread_started=$(do_facet ost1 \
9219 ost.OSS.ost_io.threads_started | cut -d= -f2")
9220 # Break out if thread_started is equal/greater than thread_max
9221 if (( $thread_started >= $thread_max )); then
9222 echo ll_ost_io thread_started $thread_started, \
9223 equal/greater than thread_max $thread_max
9229 # Cleanup - We have the numbers, Kill i/o jobs if running.
9230 jobcount=($(jobs -p))
9232 for ((i=0; i < ${#jobcount[*]}; i++)); do
9233 kill -9 ${jobcount[$i]}
9234 if [ $? -ne 0 ] ; then
9235 echo "warning: cannot kill WTL pid ${jobcount[$i]}"
9239 # Cleanup files left by WTL binary.
9240 for ((i = 1; i <= nfiles; i++)); do
9241 local file=$DIR/$tdir/${tfile}-$i
9244 if [ $? -ne 0 ] ; then
9245 echo "Warning: Failed to delete file $file"
9249 restore_lustre_params <$save_params
9250 rm -f $save_params || echo "Warning: delete file '$save_params' failed"
9252 # Error out if no new thread has started or Thread started is greater
9254 if (( $thread_started <= $OSTIO_pre ||
9255 $thread_started > $thread_max )); then
9256 error "ll_ost_io: thread_started $thread_started" \
9257 "OSTIO_pre $OSTIO_pre, thread_max $thread_max." \
9258 "No new thread started or thread started greater " \
9262 run_test 114 "verify dynamic thread creation===================="
9268 do_facet mds1 rm -f $1
9272 if [ "$mds1_FSTYPE" != ldiskfs ]; then
9273 skip "Only applicable to ldiskfs-based MDTs"
9275 [ -n "$FILESET" ] && skip "Not functional for FILESET set"
9277 local dbfs_ver=$(do_facet mds1 $DEBUGFS -V 2>&1)
9279 echo "debugfs version: $dbfs_ver"
9280 echo "$dbfs_ver" | egrep -w "1.44.3.wc1|1.44.5.wc1|1.45.2.wc1" &&
9281 skip_env "This version of debugfs doesn't show inode number"
9283 local IMAGESIZE=$((3072 << 30)) # 3072 GiB
9293 do_facet mds1 "hostname; $LCTL dl; mount"
9295 do_facet ost1 "hostname; $LCTL dl; mount"
9296 # We need MDT size 3072GB, because it is smallest
9297 # partition that can store 2B inodes
9298 do_facet mds1 "mkdir -p $TMP/$tdir"
9299 local mdsimgname=$TMP/$tdir/lustre-mdt
9301 do_facet mds1 "rm -f $mdsimgname"
9302 do_facet mds1 "touch $mdsimgname"
9303 stack_trap "cleanup_115 $mdsimgname" EXIT
9304 do_facet mds1 "$TRUNCATE $mdsimgname $IMAGESIZE" ||
9305 skip "Backend FS doesn't support sparse files"
9306 local mdsdev=$(do_facet mds1 "losetup -f")
9308 do_facet mds1 "losetup $mdsdev $mdsimgname"
9310 local mds_opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
9311 --mkfsoptions='-O ea_inode,^resize_inode,meta_bg \
9312 -N 2247484000 -E lazy_itable_init' --device-size=$IMAGESIZE"
9313 add mds1 $mds_opts --mgs --reformat $mdsdev ||
9314 skip_env "format large MDT failed"
9316 local ostdev=$(ostdevname 1)
9318 local opts="$(mkfs_opts ost1 $ostdev) \
9319 --reformat $ostdev $ostdev"
9320 add ost1 $opts || error "add ost1 failed with new params"
9321 start mds1 $mdsdev $MDS_MOUNT_OPTS || error "start MDS failed"
9322 start_ost || error "start OSS failed"
9323 mount_client $MOUNT || error "mount client failed"
9325 mkdir -p $DIR/$tdir || error "mkdir $DIR/$tdir fail"
9326 local goal="/sys/fs/ldiskfs/$(basename $mdsdev)/inode_goal"
9328 # 2147483648 is 0x80000000
9329 do_facet mds1 "echo 2147483648 >> $goal; grep . $goal"
9330 touch $DIR/$tdir/$tfile
9332 # attrs from 1 to 16 go to block, 17th - to inode
9335 for i in {1..17}; do
9336 local nm="trusted.ea$i"
9337 setfattr -n $nm -v $(printf "xattr%0250d" $i) $DIR/$tdir/$tfile
9340 do_facet mds1 "$DEBUGFS -c -R 'stat ROOT/$tdir/$tfile' $mdsdev"
9342 # inode <2147483649> trusted.ea16 (255)
9343 local inode_num=$(do_facet mds1 \
9344 "$DEBUGFS -c -R 'stat ROOT/$tdir/$tfile' $mdsdev" |
9345 awk '/ea17/ { print $2 }' |
9346 sed -e 's/>//' -e 's/<//' -e 's/\"//')
9347 echo "inode num: $inode_num"
9348 [ $inode_num -ge 2147483648 ] || error "inode $inode_num too small"
9349 do_facet mds1 "losetup -d $mdsdev"
9351 run_test 115 "Access large xattr with inodes number over 2TB"
9354 [ "$mds1_FSTYPE" != ldiskfs ] && skip "ldiskfs only test"
9355 [ "$MDS1_VERSION" -lt $(version_code 2.10.59) ] &&
9356 skip "Need server version at least 2.10.59"
9357 do_facet $SINGLEMDS which mkfs.xfs ||
9358 skip_env "No mkfs.xfs installed"
9363 local tmpmnt=$TMP/$tdir
9364 local mdtimg=$tfile-mdt0
9366 do_facet $SINGLEMDS mkdir -p $tmpmnt
9367 stack_trap "do_facet $SINGLEMDS rmdir $tmpmnt" EXIT
9369 do_facet $SINGLEMDS touch $TMP/$mdtimg
9370 stack_trap "do_facet $SINGLEMDS rm -f $TMP/$mdtimg" EXIT
9371 do_facet $SINGLEMDS mkfs -t xfs -d file,size=1t,name=$TMP/$mdtimg ||
9372 error "mkfs temporary xfs image"
9374 do_facet $SINGLEMDS mount $TMP/$mdtimg $tmpmnt ||
9375 error "mount temporary xfs image"
9376 stack_trap "do_facet $SINGLEMDS umount $tmpmnt" EXIT
9377 local old_mdssize=$MDSSIZE
9378 local old_mdsisize=$MDSISIZE
9380 MDSSIZE=$((17 * 1024 * 1024 * 1024)) # 17T MDT
9381 MDSISIZE=$((16 << 20))
9382 local opts17t="$(mkfs_opts $SINGLEMDS)"
9384 MDSSIZE=$old_mdssize
9385 MDSISIZE=$old_mdsisize
9387 do_facet $SINGLEMDS $MKFS \
9388 ${opts17t/-E /-E lazy_itable_init,lazy_journal_init,} \
9390 error "failed to mkfs for $tmpmnt/$mdtimg"
9392 do_facet $SINGLEMDS $TUNE2FS -l $tmpmnt/$mdtimg |
9393 grep -qw 'features.*extent' || error "extent should be enabled"
9396 run_test 116 "big size MDT support"
9399 # Call setup only if LustreFS is not mounted
9400 check_mount || setup
9402 do_facet ost1 "$LCTL set_param ost.OSS.ost_io.nrs_policies=fifo"
9403 do_facet ost1 "$LCTL get_param -n ost.OSS.ost_io.nrs_tbf_rule" &&
9404 error "get_param should fail"
9405 cleanup || error "cleanup failed with rc $?"
9407 run_test 117 "lctl get_param return errors properly"
9413 [[ "$MDSCOUNT" -ge 2 ]] || skip "Need more at least 2 MDTs"
9415 had_config=$(do_facet mds1 "$LCTL get_param debug | grep config")
9416 do_facet mds1 "$LCTL set_param debug=+config"
9417 do_facet mds1 "$LCTL clear"
9420 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)" &>/dev/null
9421 # mount after writeconf will make "add osp" added to mdt0 config:
9422 # 53 (224)marker 60 (flags=0x01, v2.5.1.0) lustre-MDT0001 'add osp'
9423 # 54 (080)add_uuid nid=... 0: 1:...
9424 # 55 (144)attach 0:lustre-MDT0001-osp-MDT0000 1:osp 2:...
9425 # 56 (144)setup 0:lustre-MDT0001-osp-MDT0000 1:... 2:...
9426 # 57 (136)modify_mdc_tgts add 0:lustre-MDT0000-mdtlov 1:... 2:1 3:1
9427 # duplicate modify_mdc_tgts caused crashes
9430 # using larger debug_mb size to avoid lctl dk log truncation
9431 size_mb=$((DEBUG_SIZE_SAVED * 4))
9434 # though config processing stops after failed attach and setup
9435 # it will proceed after the failed command after each writeconf
9436 # this is the original scenario of the issue
9437 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)" &>/dev/null
9438 do_facet mds1 "$LCTL set_param debug_mb=$size_mb"
9441 wait_update_facet_cond mds1 \
9442 "$LCTL dk | grep -c Processed.log.$FSNAME-MDT0000" \
9447 [[ -z "$had_config" ]] && do_facet mds1 lctl set_param debug=-config
9451 run_test 119 "writeconf on slave mdt shouldn't duplicate mdc/osp and crash"
9453 test_120() { # LU-11130
9454 [ "$MDSCOUNT" -lt 2 ] && skip "mdt count < 2"
9455 [ "$mds1_FSTYPE" != ldiskfs ] &&
9456 skip "ldiskfs only test"
9457 [ "$MDS1_VERSION" -lt $(version_code 2.11.56) ] &&
9458 skip "Need DNE2 capable MD target with LU-11130 fix"
9462 local mds1host=$(facet_active_host mds1)
9463 local mds1dev=$(mdsdevname 1)
9465 $LFS mkdir -i 1 $DIR/$tdir
9466 $LFS mkdir -i 0 $DIR/$tdir/mds1dir
9468 ln -s foo $DIR/$tdir/bar
9469 mv $DIR/$tdir/bar $DIR/$tdir/mds1dir/bar2 ||
9470 error "cross-target rename failed"
9474 run_e2fsck $mds1host $mds1dev "-n"
9476 run_test 120 "cross-target rename should not create bad symlinks"
9480 start_mgsmds || error "MGS MDS Start failed"
9482 stop_mds || error "Stopping MDSes failed"
9486 stop_mds || error "Stopping MDSes failed"
9488 run_test 121 "failover MGS"
9491 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
9492 [[ "$OST1_VERSION" -ge $(version_code 2.11.53) ]] ||
9493 skip "Need OST version at least 2.11.53"
9496 LOAD_MODULES_REMOTE=true load_modules
9497 #define OBD_FAIL_OFD_SET_OID 0x1e0
9498 do_facet ost1 $LCTL set_param fail_loc=0x00001e0
9501 $LFS mkdir -i1 -c1 $DIR/$tdir
9502 $LFS setstripe -i0 -c1 $DIR/$tdir
9503 do_facet ost1 $LCTL set_param fail_loc=0
9504 createmany -o $DIR/$tdir/file_ 1000 ||
9505 error "Fail to create a new sequence"
9509 run_test 122a "Check OST sequence update"
9512 remote_mgs_nodsh && skip "remote MGS with nodsh"
9513 [ -d $MOUNT/.lustre ] || setup
9515 # test old logid format until removal from llog_ioctl.c::str2logid()
9516 if [ $MGS_VERSION -lt $(version_code 3.1.53) ]; then
9517 do_facet mgs $LCTL dl | grep MGS
9518 do_facet mgs "$LCTL --device %MGS llog_print \
9519 \\\\\\\$$FSNAME-client 1 10" ||
9520 error "old llog_print failed"
9523 # test new logid format
9524 if [ $MGS_VERSION -ge $(version_code 2.9.53) ]; then
9525 do_facet mgs "$LCTL --device MGS llog_print $FSNAME-client" ||
9526 error "new llog_print failed"
9529 run_test 123aa "llog_print works with FIDs and simple names"
9532 remote_mgs_nodsh && skip "remote MGS with nodsh"
9533 [[ $MGS_VERSION -gt $(version_code 2.11.51) ]] ||
9534 skip "Need server with working llog_print support"
9536 [ -d $MOUNT/.lustre ] || setup
9542 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9543 mgs_arg="--device MGS"
9545 orig_val=$(do_facet mgs $LCTL get_param jobid_name)
9546 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME"
9548 yaml=$(do_facet mgs $LCTL $mgs_arg llog_print params |
9549 grep jobid_name | tail -n 1)
9551 local param=$(awk '{ print $10 }' <<< "$yaml")
9552 local val=$(awk '{ print $12 }' <<< "$yaml")
9553 #return to the default
9554 do_facet mgs $LCTL set_param -P jobid_name=$orig_val
9555 [ $val = "TESTNAME" ] || error "bad value: $val"
9556 [ $param = "jobid_name," ] || error "Bad param: $param"
9558 run_test 123ab "llog_print params output values from set_param -P"
9560 test_123ac() { # LU-11566
9561 remote_mgs_nodsh && skip "remote MGS with nodsh"
9562 do_facet mgs "$LCTL help llog_print" 2>&1 | grep -q -- --start ||
9563 skip "Need 'lctl llog_print --start' on MGS"
9569 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9570 mgs_arg="--device MGS"
9572 [ -d $MOUNT/.lustre ] || setup
9574 # - { index: 10, event: add_uuid, nid: 192.168.20.1@tcp(0x20000c0a81401,
9575 # node: 192.168.20.1@tcp }
9576 do_facet mgs $LCTL $mgs_arg \
9577 llog_print --start $start --end $end $FSNAME-client | tr -d , |
9578 while read DASH BRACE INDEX idx EVENT BLAH BLAH BLAH; do
9579 (( idx >= start )) || error "llog_print index $idx < $start"
9580 (( idx <= end )) || error "llog_print index $idx > $end"
9583 run_test 123ac "llog_print with --start and --end"
9585 test_123ad() { # LU-11566
9586 remote_mgs_nodsh && skip "remote MGS with nodsh"
9587 # older versions of lctl may not print all records properly
9588 do_facet mgs "$LCTL help llog_print" 2>&1 | grep -q -- --start ||
9589 skip "Need 'lctl llog_print --start' on MGS"
9591 [ -d $MOUNT/.lustre ] || setup
9593 # append a new record, to avoid issues if last record was cancelled
9594 local old=$($LCTL get_param -n osc.*-OST0000-*.max_dirty_mb | head -1)
9595 do_facet mgs $LCTL conf_param $FSNAME-OST0000.osc.max_dirty_mb=$old
9597 # logid: [0x3:0xa:0x0]:0
9601 local num=$(do_facet mgs $LCTL --device MGS llog_info $FSNAME-client |
9602 awk '/last_index:/ { print $2 - 1 }')
9604 # - { index: 71, event: set_timeout, num: 0x14, param: sys.timeout=20 }
9605 local last=$(do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
9606 tail -1 | awk '{ print $4 }' | tr -d , )
9607 (( last == num )) || error "llog_print only showed $last/$num records"
9609 run_test 123ad "llog_print shows all records"
9611 test_123ae() { # LU-11566
9620 remote_mgs_nodsh && skip "remote MGS with nodsh"
9621 [ -d $MOUNT/.lustre ] || setupall
9623 max=$($LCTL get_param -n osc.*-OST0000-*.max_dirty_mb | head -1)
9624 pgs=$($LCTL get_param -n osc.*-OST0000-*.max_pages_per_rpc | head -1)
9625 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9626 mgs_arg="--device MGS"
9628 if do_facet mgs "$LCTL help llog_cancel" 2>&1| grep -q -- --log_id; then
9629 # save one set_param -P record in case none exist
9631 do_facet mgs $LCTL set_param -P osc.*.max_pages_per_rpc=$pgs
9632 stack_trap "do_facet mgs $LCTL set_param -P -d \
9633 osc.*.max_pages_per_rpc"
9636 orig=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9637 tail -1 | awk '{ print $4 }' | tr -d , )
9638 do_facet mgs $LCTL set_param -P osc.*.max_dirty_mb=$max
9639 do_facet mgs $LCTL $mgs_arg llog_print $log | tail -1 |
9640 grep "parameter: osc.*.max_dirty_mb" ||
9641 error "new set_param -P wasn't stored in params log"
9643 # - { index: 71, event: set_param, device: general,
9644 # param: osc.*.max_dirty_mb, value: 256 }
9645 id=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9646 tail -1 | awk '{ print $4 }' | tr -d , )
9647 do_facet mgs $LCTL $mgs_arg llog_cancel $log --log_idx=$id
9648 local new=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9649 tail -1 | awk '{ print $4 }' | tr -d , )
9650 (( new == orig )) ||
9651 error "new llog_cancel now $new, not at $orig records"
9654 # test old positional parameters for a while still
9655 if [ "$MGS_VERSION" -le $(version_code 3.1.53) ]; then
9658 do_facet mgs $LCTL conf_param \
9659 $FSNAME-OST0000.osc.max_pages_per_rpc=$pgs
9660 stack_trap "do_facet mgs $LCTL conf_param -d \
9661 $FSNAME-OST0000.osc.max_pages_per_rpc"
9663 orig=$(do_facet mgs $LCTL --device MGS llog_print $log |
9664 tail -1 | awk '{ print $4 }' | tr -d , )
9665 do_facet mgs $LCTL conf_param $FSNAME-OST0000.osc.max_dirty_mb=$max
9666 do_facet mgs $LCTL --device MGS llog_print $log |
9667 tail -1 | grep "parameter: osc.max_dirty_mb" ||
9668 error "old conf_param wasn't stored in params log"
9669 do_facet mgs $LCTL --device MGS llog_print $log
9670 # - { index: 71, event: conf_param, device: testfs-OST0000-osc,
9671 # param: osc.max_dirty_mb=256 }
9672 id=$(do_facet mgs $LCTL --device MGS llog_print $log |
9673 tail -1 | awk '{ print $4 }' | tr -d , )
9674 do_facet mgs $LCTL --device MGS llog_cancel $log $id
9675 do_facet mgs $LCTL --device MGS llog_print $log
9676 new=$(do_facet mgs $LCTL --device MGS llog_print $log |
9677 tail -1 | awk '{ print $4 }' | tr -d , )
9678 (( new == orig )) ||
9679 error "old llog_cancel now $new, not at $orig records"
9682 run_test 123ae "llog_cancel can cancel requested record"
9684 test_123af() { #LU-13609
9685 [ "$MGS_VERSION" -ge $(version_code 2.13.54) -a \
9686 "$MDS1_VERSION" -ge $(version_code 2.13.54) ] ||
9687 skip "Need both MGS and MDS version at least 2.13.54"
9689 [ -d $MOUNT/.lustre ] || setupall
9690 stack_trap "do_facet mds1 $LCTL set_param fail_loc=0" EXIT
9700 for device in "MGS" "$FSNAME-MDT0000"; do
9701 cmd="--device $device llog_catlist"
9702 echo "lctl $cmd ..."
9703 if [ "$device" = "MGS" ]; then
9708 orig_clist=($(do_facet $facet $LCTL $cmd | awk '{ print $2 }'))
9709 orig_count=${#orig_clist[@]}
9710 echo "orig_clist: ${orig_clist[*]}"
9712 #define OBD_FAIL_CATLIST 0x131b
9713 #fetch to llog records from the second one
9714 do_facet $facet $LCTL set_param fail_loc=0x131b fail_val=2
9716 new_clist=($(do_facet $facet $LCTL $cmd | awk '{ print $2 }'))
9717 new_count=${#new_clist[@]}
9718 echo "new_clist: ${new_clist[*]}"
9720 [ $new_count -eq $((orig_count - 1)) ] ||
9721 error "$new_count != $orig_count - 1"
9722 for i in $(seq 0 $new_count); do
9724 [ "${orig_clist[$j]}" = "${new_clist[$i]}" ] ||
9725 error "${orig_clist[$j]} != ${new_clist[$i]}"
9727 do_facet mds1 $LCTL set_param fail_loc=0
9731 run_test 123af "llog_catlist can show all config files correctly"
9733 test_123ag() { # LU-15142
9737 remote_mgs_nodsh && skip "remote MGS with nodsh"
9738 (( $MGS_VERSION >= $(version_code 2.14.55) )) ||
9739 skip "Need server version least 2.14.55"
9741 [ -d $MOUNT/.lustre ] || setup
9743 orig_val=$(do_facet mgs $LCTL get_param jobid_name)
9744 stack_trap "do_facet mgs $LCTL set_param -P jobid_name=$orig_val"
9746 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME1"
9747 do_facet mgs $LCTL set_param -P -d jobid_name
9748 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
9750 (( rec == 0 )) || error "parameter was not deleted, check #1"
9751 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME1"
9752 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
9754 (( rec == 1)) || error "parameter is not set"
9755 # usage with ordinary set_param format works too
9756 do_facet mgs $LCTL set_param -P -d jobid_name="ANY"
9757 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
9759 (( rec == 0 )) || error "parameter was not deleted, check #2"
9761 run_test 123ag "llog_print skips values deleted by set_param -P -d"
9763 test_123ah() { #LU-7668 del_ost
9764 (( "$MGS_VERSION" >= $(version_code 2.15.51) &&
9765 "$MDS1_VERSION" >= $(version_code 2.15.51) )) ||
9766 skip "Need MGS/MDS version >= 2.15.51 for del_ost"
9768 [ -d $MOUNT/.lustre ] || setupall
9769 stack_trap "do_facet mds1 $LCTL set_param fail_loc=0" EXIT
9771 local cmd="--device MGS llog_print"
9773 cli_llogcnt_orig=$(do_facet mgs $LCTL $cmd $FSNAME-client |
9774 grep -c $FSNAME-OST0000)
9775 mds1_llogcnt_orig=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 |
9776 grep -c $FSNAME-OST0000)
9778 [ $cli_llogcnt_orig -gt 0 ] ||
9779 error "$FSNAME-OST0000 not found (client)"
9780 [ $mds1_llogcnt_orig -gt 0 ] || error "$FSNAME-OST0000 not found (MDT)"
9782 # -n/--dryrun should NOT modify catalog
9783 do_facet mgs "$LCTL del_ost -n -t $FSNAME-OST0000" ||
9784 error "del_ost --dryrun failed with $?"
9786 local cli_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-client |
9787 grep -c $FSNAME-OST0000)
9788 local mds1_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 |
9789 grep -c $FSNAME-OST0000)
9791 [ $cli_llogcnt -eq $cli_llogcnt_orig ] ||
9792 error "--dryrun error: $cli_llogcnt != $cli_llogcnt_orig"
9793 [ $mds1_llogcnt -eq $mds1_llogcnt_orig ] ||
9794 error "--dryrun error: $mds1_llogcnt != $mds1_llogcnt_orig"
9797 do_facet mgs "$LCTL del_ost --target $FSNAME-OST0000" ||
9798 error "del_ost failed with $?"
9800 local cli_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-client |
9801 grep -c $FSNAME-OST0000)
9802 local mds1_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 |
9803 grep -c $FSNAME-OST0000)
9805 # every catalog entry for OST0000 should have been cancelled
9806 [ $cli_llogcnt -eq 0 ] || error "$FSNAME-OST0000 not cancelled (cli)"
9807 [ $mds1_llogcnt -eq 0 ] || error "$FSNAME-OST0000 not cancelled (MDT)"
9809 umount_client $MOUNT
9812 $LFS df | grep -q OST0000 && error "del_ost did not remove OST0000!"
9816 run_test 123ah "del_ost cancels config log entries correctly"
9821 echo "cleanup test 123ai"
9823 # cancel last timeout record
9824 do_facet mgs "$LCTL set_param -P -d timeout"
9826 # restore timeout value
9827 do_nodes $(comma_list $(all_nodes)) "$LCTL set_param timeout=$timeout"
9830 test_123ai() { #LU-16167
9836 remote_mgs_nodsh && skip "remote MGS with nodsh"
9837 (( MDS1_VERSION >= $(version_code 2.15.51) )) ||
9838 skip "Need MDS version at least 2.15.51"
9840 [ -d $MOUNT/.lustre ] || setup
9842 old_count=$(do_facet mgs "$LCTL --device MGS llog_print -r params" |
9843 grep -c "parameter: timeout")
9845 old_timeout=$($LCTL get_param -n timeout)
9846 stack_trap "cleanup_123ai $old_timeout" EXIT
9848 # add and cancel params (2 * MAX_IOC_BUFLEN / 128 + 1 = 129)
9849 for i in {1..129}; do
9850 do_facet mgs "$LCTL set_param -P timeout=$i" ||
9851 error "fail to set timeout=$i on MGS"
9854 count=$(do_facet mgs "$LCTL --device MGS llog_print -r params" |
9855 grep -c "parameter: timeout")
9857 ((count - old_count == 129)) ||
9858 error "number timeout record missmatch ($count - $old_count != 129)"
9860 do_facet mgs "$LCTL --device MGS llog_print params" |
9861 tail -1 | grep "timeout, value: 129" ||
9862 error "llog_print could not display the last record (timeout=129)"
9865 run_test 123ai "llog_print display all non skipped records"
9868 remote_mgs_nodsh && skip "remote MGS with nodsh"
9870 [ -d $MOUNT/.lustre ] || setup
9871 local yaml_file="$TMP/$tfile.yaml"
9873 do_facet mgs rm "$yaml_file"
9874 local cfgfiles=$(do_facet mgs "lctl --device MGS llog_catlist" |
9875 sed 's/config_log://')
9877 # set jobid_var to a different value for test
9878 local orig_val=$(do_facet mgs $LCTL get_param jobid_var)
9880 do_facet mgs $LCTL set_param -P jobid_var="TESTNAME"
9882 for i in $cfgfiles params; do
9883 do_facet mgs "lctl --device MGS llog_print ${i} >> $yaml_file"
9886 echo "Unmounting FS"
9893 # Reapply the config from before
9894 echo "Setting configuration parameters"
9895 do_facet mgs "lctl set_param -F $yaml_file"
9897 local set_val=$(do_facet mgs $LCTL get_param jobid_var)
9899 do_facet mgs $LCTL set_param -P $orig_val
9901 [ $set_val == "jobid_var=TESTNAME" ] ||
9902 error "$set_val is not TESTNAME"
9904 do_facet mgs rm "$yaml_file"
9907 run_test 123F "clear and reset all parameters using set_param -F"
9911 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
9912 [ -z $mds2failover_HOST ] && skip "needs MDT failover setup"
9915 do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
9916 grep 1.2.3.4@tcp && error "Should not be fake nid"
9917 do_facet mgs $LCTL conf_param $FSNAME-MDT0001.failover.node=1.2.3.4@tcp\
9918 || error "Set params error"
9919 do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
9920 grep 1.2.3.4@tcp || error "Fake nid should be added"
9924 if combined_mgs_mds; then
9925 start_mdt 1 "-o nosvc" ||
9926 error "starting mds with nosvc option failed"
9928 local nid=$(do_facet mds2 $LCTL list_nids | head -1)
9929 local failover_nid=$(do_node $mds2failover_HOST $LCTL list_nids | head -1)
9931 do_facet mgs $LCTL replace_nids $FSNAME-MDT0001 $nid:$failover_nid ||
9932 error "replace_nids execution error"
9934 if combined_mgs_mds; then
9940 echo "lfs setdirstripe"
9941 $LFS setdirstripe -i 1 $MOUNT/$tdir || error "setdirstirpe error"
9944 run_test 124 "check failover after replace_nids"
9946 get_max_sectors_kb() {
9949 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
9950 local max_sectors_path="/sys/block/${dev_base}/queue/max_sectors_kb"
9952 do_facet ${facet} "[[ -e ${max_sectors_path} ]] &&
9953 cat ${max_sectors_path}"
9956 get_max_hw_sectors_kb() {
9959 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
9960 local max_hw_path="/sys/block/${dev_base}/queue/max_hw_sectors_kb"
9962 do_facet ${facet} "[[ -e ${max_hw_path} ]] && cat ${max_hw_path}"
9965 set_max_sectors_kb() {
9969 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
9970 local max_sectors_path="/sys/block/${dev_base}/queue/max_sectors_kb"
9972 do_facet ${facet} "[[ -e ${max_sectors_path} ]] &&
9973 echo ${value} > ${max_sectors_path}"
9976 [[ $rc -ne 0 ]] && echo "Failed to set ${max_sectors_path} to ${value}"
9981 # Return 0 if all slave devices have max_sectors_kb == max_hw_sectors_kb
9982 # Otherwise return > 0
9983 check_slaves_max_sectors_kb()
9987 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
9988 local slaves_dir=/sys/block/${dev_base}/slaves
9989 local slave_devices=$(do_facet ${facet} "ls ${slaves_dir} 2>/dev/null")
9990 [[ -z ${slave_devices} ]] && return 0
9992 local slave max_sectors new_max_sectors max_hw_sectors path
9995 for slave in ${slave_devices}; do
9996 path="/dev/${slave}"
9997 ! is_blkdev ${facet} ${path} && continue
9998 max_sectors=$(get_max_sectors_kb ${facet} ${path})
9999 max_hw_sectors=$(get_max_hw_sectors_kb ${facet} ${path})
10000 new_max_sectors=${max_hw_sectors}
10001 [[ ${new_max_sectors} -gt ${RQ_SIZE_LIMIT} ]] &&
10002 new_max_sectors=${RQ_SIZE_LIMIT}
10004 if [[ ${max_sectors} -ne ${new_max_sectors} ]]; then
10005 echo "${path} ${max_sectors} ${new_max_sectors}"
10008 check_slaves_max_sectors_kb ${facet} ${path}
10017 local facet_list="mgs mds1 ost1"
10018 combined_mgs_mds && facet_list="mgs ost1"
10021 for facet in ${facet_list}; do
10022 [[ $(facet_fstype ${facet}) != ldiskfs ]] &&
10023 skip "ldiskfs only test" &&
10025 ! is_blkdev ${facet} $(facet_device ${facet}) &&
10026 skip "requires all real devices" &&
10031 # We don't increase IO request size limit past 16MB. See comments in
10032 # lustre/utils/libmount_utils_ldiskfs.c:tune_max_sectors_kb()
10033 RQ_SIZE_LIMIT=$((16 * 1024))
10034 local device old_max_sectors new_max_sectors max_hw_sectors
10035 for facet in ${facet_list}; do
10036 device=$(facet_device ${facet})
10037 old_max_sectors=$(get_max_sectors_kb ${facet} ${device})
10038 max_hw_sectors=$(get_max_hw_sectors_kb ${facet} ${device})
10040 # The expected value after l_tunedisk is executed
10041 new_max_sectors=$old_max_sectors
10042 [[ ${new_max_sectors_kb} -gt ${RQ_SIZE_LIMIT} ]] &&
10043 new_max_sectors_kb=${RQ_SIZE_LIMIT}
10045 # Ensure the current value of max_sectors_kb does not equal
10046 # max_hw_sectors_kb, so we can tell whether l_tunedisk did
10048 set_max_sectors_kb ${facet} ${device} $((new_max_sectors - 1))
10050 # Value before l_tunedisk
10051 local pre_max_sectors=$(get_max_sectors_kb ${facet} ${device})
10052 if [[ ${pre_max_sectors} -ne $((new_max_sectors - 1)) ]]; then
10053 echo "unable to satsify test pre-condition:"
10054 echo "${pre_max_sectors} != $((new_max_sectors - 1))"
10059 echo "Before: ${facet} ${device} ${pre_max_sectors} ${max_hw_sectors}"
10061 do_facet ${facet} "l_tunedisk ${device}"
10063 # Value after l_tunedisk
10064 local post_max_sectors=$(get_max_sectors_kb ${facet} ${device})
10066 echo "After: ${facet} ${device} ${post_max_sectors} ${max_hw_sectors}"
10068 if [[ ${facet} != ost1 ]]; then
10069 if [[ ${post_max_sectors} -ne ${pre_max_sectors} ]]; then
10070 echo "l_tunedisk modified max_sectors_kb of ${facet}"
10074 set_max_sectors_kb ${facet} ${device} ${old_max_sectors}
10076 if [[ ${post_max_sectors} -eq ${pre_max_sectors} ]]; then
10077 echo "l_tunedisk failed to modify max_sectors_kb of ${facet}"
10081 check_slaves_max_sectors_kb ${facet} ${device} ||
10088 run_test 125 "check l_tunedisk only tunes OSTs and their slave devices"
10091 [[ "$MDS1_VERSION" -ge $(version_code 2.13.52) ]] ||
10092 skip "Need MDS version at least 2.13.52"
10095 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ../libcfs/libcfs/libcfs
10096 #define OBD_FAIL_OBD_SETUP 0x60d
10097 do_facet mds1 $LCTL set_param fail_loc=0x60d
10098 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_modules &
10099 for i in {1..40}; do
10100 do_facet mds1 lsmod | grep -q osd_$mds1_FSTYPE && break
10103 clear_failloc $SINGLEMDS 20 &
10104 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
10106 run_test 126 "mount in parallel shouldn't cause a crash"
10109 [[ "$ost1_FSTYPE" == ldiskfs ]] || skip "ldiskfs only test"
10113 zconf_umount_clients $RCLIENTS $MOUNT
10115 wait_osp_active ost ${FSNAME}-OST0000 0 1
10116 local osc_tgt="$FSNAME-OST0000-osc-$($LFS getname -i $DIR)"
10117 local avail1=($($LCTL get_param -n osc.${osc_tgt}.kbytesavail))
10119 $LFS setstripe -i 0 $DIR/$tfile || error "failed creating $DIR/$tfile"
10120 dd if=/dev/zero of=$DIR/$tfile bs=1M oflag=direct || true
10122 local avail2=($($LCTL get_param -n osc.${osc_tgt}.kbytesavail))
10124 if ((avail2 * 100 / avail1 > 1)); then
10127 error "more than 1% space left: before=$avail1 after=$avail2"
10130 local mbs=$(($(stat -c %s $DIR/$tfile) / (1024 * 1024)))
10132 dd if=/dev/zero of=$DIR/$tfile bs=1M count=$mbs conv=notrunc \
10133 oflag=direct || error "overwrite failed"
10135 run_test 127 "direct io overwrite on full ost"
10139 combined_mgs_mds && skip "need separate mgs device"
10140 [ "$ost2_FSTYPE" == zfs ] && import_zpool ost2
10143 # Try to apply nolocallogs to the virgin OST. Should fail.
10144 do_facet ost2 "$TUNEFS --nolocallogs $(ostdevname 2)" &&
10145 error "nolocallogs should not be allowed on the virgin target"
10150 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
10151 # Start OST without MGS (local configs)
10152 do_facet ost1 "$TUNEFS --dryrun $(ostdevname 1)"
10153 start_ost || error "unable to start OST1"
10154 stop_ost || error "Unable to stop OST1"
10156 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
10157 # Do not allow reading local configs, should fail
10158 do_facet ost1 "$TUNEFS --nolocallogs $(ostdevname 1)" ||
10159 error "Can not set nolocallogs"
10160 start_ost && error "OST1 started, but should fail"
10162 # Connect to MGS successfully, reset nolocallogs flag
10163 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
10164 start_mgs || error "unable to start MGS"
10165 start_ost || error "unable to start OST1"
10167 do_facet ost1 "$TUNEFS --dryrun $(ostdevname 1)" | grep "nolocallogs" &&
10168 error "nolocallogs expected to be reset"
10170 stop_ost || error "Unable to stop OST1"
10172 run_test 128 "Force using remote logs with --nolocallogs"
10176 (( MDS1_VERSION >= $(version_code 2.14.57) )) ||
10177 skip "Need MDS version at least 2.14.57"
10179 start_mgsmds || error "MGS/MDS start failed"
10181 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &&
10182 error "start ost1 should fail" || true
10183 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &&
10184 error "second start ost1 should fail" || true
10185 do_facet ost1 "$TUNEFS --writeconf $(ostdevname 1)"
10186 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS ||
10187 error "start ost1 failed"
10191 run_test 129 "attempt to connect an OST with the same index should fail"
10195 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
10197 stop_mdt 2 || error "mdt2 stop failed"
10198 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)"
10199 start_mdt 2 || error "mdt2 start failed"
10200 do_facet mds2 "$LCTL dl" | grep MDT0001-osp-MDT0001 &&
10201 error "Illegal OSP device created" || true
10203 run_test 130 "re-register an MDT after writeconf"
10206 [ "$mds1_FSTYPE" == "ldiskfs" ] || skip "ldiskfs only test"
10207 (( $MDS1_VERSION >= $(version_code 2.14.56.35) )) ||
10208 skip "Need MDS version at least 2.14.56.35"
10209 do_facet mds1 $DEBUGFS -R features $(mdsdevname 1) |
10210 grep -q project || skip "skip project quota not supported"
10215 test_mkdir -c $MDSCOUNT -p $DIR/$tdir
10216 $LFS project -p 1000 $DIR/$tdir || error "set dir project id failed"
10217 createmany -o $DIR/$tdir/f 512
10218 for ((i = 0; i < 512; ++i)); do
10219 $LFS project -p $i $DIR/$tdir/f${i} ||
10220 error "set f${i} project id failed"
10223 test_mkdir -c $MDSCOUNT -p $DIR/$tdir.inherit
10224 $LFS project -p 1001 -s $DIR/$tdir.inherit
10225 createmany -o $DIR/$tdir.inherit/f 128
10226 (( $($LFS project $DIR/$tdir.inherit/f* |
10227 awk '$1 == 1001 { print }' | wc -l) == 128 )) ||
10228 error "files did not inherit projid 1001"
10232 for i in $(seq $MDSCOUNT); do
10233 mds_backup_restore mds$i ||
10234 error "Backup/restore on mds$i failed"
10239 projid=($($LFS project -d $DIR/$tdir))
10240 [ ${projid[0]} == "1000" ] ||
10241 error "projid expected 1000 not ${projid[0]}"
10242 for ((i = 0; i < 512; ++i)); do
10243 projid=($($LFS project $DIR/$tdir/f${i}))
10244 [ ${projid[0]} == "$i" ] ||
10245 error "projid expected $i not ${projid[0]}"
10248 (( $($LFS project $DIR/$tdir.inherit/f* |
10249 awk '$1 == 1001 { print }' | wc -l) == 128 )) ||
10250 error "restore did not copy projid 1001"
10252 run_test 131 "MDT backup restore with project ID"
10258 (( MDS1_VERSION >= $(version_code 2.14.57) )) ||
10259 skip "Need MDS version at least 2.14.57"
10261 combined_mgs_mds || start_mgs || error "unable to start MGS"
10262 start_mdt 1 || error "unable to start mdt1"
10264 err_cnt=$(do_facet mds1 dmesg | grep -c "cannot take the layout locks")
10265 stop_mdt 1 || error "stop mdt1 failed"
10267 [ "$mds1_FSTYPE" == zfs ] && import_zpool mds1
10268 do_facet mds1 $TUNEFS --param mdt.hsm_control=enabled $(mdsdevname 1) ||
10269 error "tunefs failed"
10270 start_mdt 1 || error "cannot start mdt1"
10272 err_cnt2=$(do_facet mds1 dmesg | grep -c "cannot take the layout locks")
10273 [ $err_cnt -eq $err_cnt2 ] || error "Can not take the layout lock"
10274 stop_mdt 1 || error "stop mdt1 failed"
10276 run_test 132 "hsm_actions processed after failover"
10278 # This test verifies we do RR allocation within a pool even if there is a
10279 # significant imbalance vs an OST outside the pool
10281 [[ $OSTCOUNT -lt 4 ]] && skip_env "needs >= 4 OSTs"
10282 [[ "$OST1_VERSION" -ge $(version_code 2.15.51) ]] ||
10283 skip "Need OST version at least 2.15.51"
10284 # This is the easiest way to ensure OSTs start out balanced
10285 reformat_and_config
10288 check_set_fallocate_or_skip
10290 local testfile=$DIR/$tdir/$tfile
10291 local pool="testpool"
10292 local ostrange=$((OSTCOUNT - 1))
10293 # Select all but the last OST to add to the pool
10294 local poolostrange=$((OSTCOUNT - 2))
10302 [ $MINV -eq 0 ] && error "no free space in OST$MINI"
10303 [ $MAXV -gt $((2 * $MINV)) ] &&
10304 error "OSTs badly unbalanced after reformat"
10306 create_pool $FSNAME.$pool || error "failed to create a pool"
10307 do_facet mgs $LCTL pool_add $FSNAME.$pool OST[0-$poolostrange] ||
10308 error "failed to add OST[0-$poolostrange] to the pool"
10310 test_mkdir -p $DIR/$tdir || error "failed to mkdir $DIR/$tdir"
10311 # Consume space on the OSTs in the pool so they are unbalanced with the
10312 # OST outside of the pool
10313 # fill each OST 90% with fallocate so they are widely
10315 local size=$(((MINV * 9 / 10) * 1024))
10316 for ((i = 0; i <= poolostrange; i++)); do
10317 $LFS setstripe -c 1 -i $i $testfile$i ||
10318 error "failed to setstripe $testfile$i"
10319 fallocate -l $size $testfile$i || error "fallocate failed"
10325 # Create files in the pool now that there is an imbalance
10326 filecount=$(((OSTCOUNT - 1) * filenum))
10327 for ((i = 0; i < filecount; i++)); do
10328 $LFS setstripe -p $pool $testfile-$i ||
10329 error "failed to setstripe -p $pool $testfile-$i"
10331 $LFS getstripe -i $testfile-* > /tmp/$tfile.log
10332 # Count the number of files with a stripe on each OST to verify the
10333 # pool allocated with round-robin
10334 for ((i = 0; i <= poolostrange; i++)); do
10335 stripecount=$(grep -c $i /tmp/$tfile.log)
10336 # Allow a little leeway
10337 if (( stripecount < filenum - 1 ||
10338 stripecount > filenum + 1 )); then
10339 cat /tmp/$tfile.log
10340 error "$stripecount != $filenum files on OST$i"
10344 # Create files across the system now that there is an imbalance
10345 filecount=$((OSTCOUNT * filenum))
10346 for ((i = 1; i < filecount; i++)); do
10347 $LFS setstripe $testfile-$i.2 ||
10348 error "failed to setstripe $testilfe-$i.2"
10350 $LFS getstripe -i $testfile-*.2 > /tmp/$tfile.log
10352 # Count the number of files with a stripe on each OST to verify the
10353 # files are *NOT* allocated with round-robin
10354 for ((i = 0; i <= ostrange; i++)); do
10355 stripecount=$(grep -c $i /tmp/$tfile.log)
10356 if [[ $stripecount -ne $filenum ]]; then
10358 echo "QOS: $stripecount != $filenum files on OST$i"
10361 if [ -z "$qos_used" ]; then
10362 error "QOS not used on imbalanced OSTs!"
10365 rm -rf /tmp/$tfile.log $DIR/$tdir
10366 do_facet mgs $LCTL pool_remove $FSNAME.$pool OST[0-$poolostrange] ||
10367 "failed to remove OST[0-$poolostrange] from the pool"
10368 do_facet mgs $LCTL pool_destroy $FSNAME.$pool ||
10369 error "failed to destroy pool"
10371 run_test 133 "stripe QOS: free space balance in a pool"
10374 [ "$mds1_FSTYPE" == "ldiskfs" ] || skip "ldiskfs only test"
10377 local mdt_dev=$(facet_device mds1)
10378 local tmp_dir=$TMP/$tdir
10379 local out=$tmp_dir/check_iam.txt
10380 local CHECK_IAM=${CHECK_IAM:-$(do_facet mds1 "which check_iam 2> /dev/null || true")}
10382 [[ -n "$CHECK_IAM" ]] || skip "check_iam not found"
10385 for ((i=0; i<64; i++)); do
10387 #cmd introduce a random corruption to IAM file
10388 local cmd="dd if=/dev/urandom of=$tmp_dir/$f bs=2 conv=notrunc count=1 seek=$((RANDOM % 36))"
10389 do_facet mds1 "mkdir -p $tmp_dir; \
10390 $DEBUGFS -c -R 'dump $f $tmp_dir/$f' $mdt_dev 2>&1; \
10391 $CHECK_IAM -v $tmp_dir/$f 2>&1; \
10392 echo $cmd; eval $cmd 2>/dev/null;
10393 $CHECK_IAM -v $tmp_dir/$f 2>&1; echo \\\$?" >> $out
10398 stack_trap "rm -rf $tmp_dir && do_facet mds1 rm -rf $tmp_dir" EXIT
10400 rc=$(grep -c "fault\|except" $out)
10401 (( rc == 0 )) || { cat $out &&
10402 error "check_iam failed with fault or exception $rc"; }
10404 rc=$(grep -c "^255" $out)
10405 error=$(grep -c "FINISHED WITH ERRORS" $out)
10407 (( rc == error )) || { cat $out &&
10408 error "check_iam errcode does not fit with errors $rc $error"; }
10410 run_test 134 "check_iam works without faults"
10412 cleanup_test_135(){
10415 printf "\nCleanup test_135\n" >&2
10416 do_facet mds1 "$LCTL set_param -n $oldgc"
10417 rm -rf $DIR/$tdir &> /dev/null
10421 __test_135_file_thread() {
10423 local init_time=$(awk '{print $1}' /proc/uptime)
10424 local awkcmd="/crosses index zero/ {if (\$1 > $init_time) exit(1);}"
10426 #Generate a full plain llogs
10427 while dmesg | sed -r 's/(\[|\])//g' | awk "$awkcmd" ; do
10428 createmany -o $DIR/$tdir/f 4500 >&2
10429 createmany -u $DIR/$tdir/f 4500 >&2
10433 __test_135_reader() {
10436 local firstidx=$(changelog_user_rec mds1 $cl_user)
10437 local oldidx=$firstidx
10442 while read -t10 -u$fd newidx other; do
10443 (( (newidx - oldidx) == 1 )) ||
10444 error "changelog jump detected (last: $oldidx, current: $newidx)"
10446 if (( (newidx - firstidx + 1) % 13000 == 0 )); then
10447 [[ $pid -eq 0 ]] ||
10448 wait $pid || error "changelog_clear failed"
10449 changelog_clear $((newidx - 1)) mds1 >&2 & pid=$!
10454 [[ $pid -eq 0 ]] ||
10455 wait $pid || error "changelog_clear failed"
10461 (( MDS1_VERSION >= $(version_code 2.15.52) )) ||
10462 skip "need MDS version at least 2.15.52"
10464 local service=$(facet_svc mds1)
10466 local lastread lastidx
10467 local files_pid reader_pid
10472 # Need to reformat because we are changing llog catalog sizes to 5.
10473 # Otherwise, processing could fail with existing catalogs (last_idx>5).
10477 # Disable changelog garbage colector
10478 local oldgc=$(do_facet mds1 "$LCTL get_param mdd.${service}.changelog_gc")
10479 do_facet mds1 "$LCTL set_param -n mdd.${service}.changelog_gc=0"
10480 stack_trap "cleanup_test_135 $oldgc" EXIT INT
10482 # change the changelog_catalog size to 5 entries for everybody
10483 #define OBD_FAIL_CAT_RECORDS 0x1312
10484 do_node $(comma_list $(all_nodes)) $LCTL set_param fail_loc=0x1312 fail_val=5
10486 # disable console ratelimit
10487 local rl=$(cat /sys/module/libcfs/parameters/libcfs_console_ratelimit)
10488 echo 0 > /sys/module/libcfs/parameters/libcfs_console_ratelimit
10489 stack_trap "echo $rl > /sys/module/libcfs/parameters/libcfs_console_ratelimit" EXIT
10491 test_mkdir -c 1 -i 0 $DIR/$tdir || error "Failed to create directory"
10492 changelog_chmask "ALL" || error "changelog_chmask failed"
10493 changelog_register || error "changelog_register failed"
10495 cl_user="${CL_USERS[mds1]%% *}"
10496 changelog_users mds1 | grep -q $cl_user ||
10497 error "User $cl_user not found in changelog_users"
10499 # Start reader thread
10500 coproc $LFS changelog --follow $service
10503 stack_trap "echo kill changelog reader; kill $reader_pid" EXIT
10505 echo -e "\nWrap arround changelog catalog"
10507 # Start file writer thread
10508 __test_135_file_thread "$service" & files_pid=$!
10509 stack_trap "(pkill -P$files_pid; kill $files_pid) &> /dev/null || true" EXIT
10511 # Check changelog entries
10512 lastread=$(__test_135_reader $fd $cl_user) || exit $?
10513 ! kill -0 $files_pid 2>/dev/null ||
10514 error "creation thread is running. Is changelog reader stuck?"
10516 lastidx=$(changelog_users mds1 | awk '/current_index/ {print $NF}' )
10517 [[ "$lastread" -eq "$lastidx" ]] ||
10518 error "invalid changelog lastidx (read: $lastread, mds: $lastidx)"
10520 run_test 135 "check the behavior when changelog is wrapped around"
10523 # (This was sanity/802a)
10525 saved_MGS_MOUNT_OPTS=$MGS_MOUNT_OPTS
10526 saved_MDS_MOUNT_OPTS=$MDS_MOUNT_OPTS
10527 saved_OST_MOUNT_OPTS=$OST_MOUNT_OPTS
10528 saved_MOUNT_OPTS=$MOUNT_OPTS
10532 MGS_MOUNT_OPTS=$saved_MGS_MOUNT_OPTS
10533 MDS_MOUNT_OPTS=$saved_MDS_MOUNT_OPTS
10534 OST_MOUNT_OPTS=$saved_OST_MOUNT_OPTS
10535 MOUNT_OPTS=$saved_MOUNT_OPTS
10540 [[ $mds1_FSTYPE = zfs ]] || skip "ZFS specific test"
10541 [[ $MDS1_VERSION -lt $(version_code 2.9.55) ]] ||
10542 [[ $OST1_VERSION -lt $(version_code 2.9.55) ]] &&
10543 skip "Need server version at least 2.9.55"
10545 [[ $ENABLE_QUOTA ]] && skip "Quota enabled for read-only test"
10547 # Reset before starting
10551 mkdir $DIR/$tdir || error "(1) fail to mkdir"
10553 cp $LUSTRE/tests/test-framework.sh $DIR/$tdir/ ||
10554 error "(2) Fail to copy"
10556 stack_trap cleanup_802a EXIT
10558 # sync by force before remount as readonly
10559 sync; sync_all_data; sleep 3; sync_all_data
10563 MGS_MOUNT_OPTS=$(csa_add "$MGS_MOUNT_OPTS" -o rdonly_dev)
10564 MDS_MOUNT_OPTS=$(csa_add "$MDS_MOUNT_OPTS" -o rdonly_dev)
10565 OST_MOUNT_OPTS=$(csa_add "$OST_MOUNT_OPTS" -o rdonly_dev)
10567 echo "Mount the server as read only"
10568 setupall server_only || error "(3) Fail to start servers"
10570 echo "Mount client without ro should fail"
10571 mount_client $MOUNT &&
10572 error "(4) Mount client without 'ro' should fail"
10574 echo "Mount client with ro should succeed"
10575 MOUNT_OPTS=$(csa_add "$MOUNT_OPTS" -o ro)
10576 mount_client $MOUNT ||
10577 error "(5) Mount client with 'ro' should succeed"
10579 echo "Modify should be refused"
10580 touch $DIR/$tdir/guard && error "(6) Touch should fail under ro mode"
10582 echo "Read should be allowed"
10583 diff $LUSTRE/tests/test-framework.sh $DIR/$tdir/test-framework.sh ||
10584 error "(7) Read should succeed under ro mode"
10586 run_test 802a "simulate readonly device"
10588 if ! combined_mgs_mds ; then
10594 # restore the values of MDSSIZE and OSTSIZE
10595 MDSSIZE=$STORED_MDSSIZE
10596 OSTSIZE=$STORED_OSTSIZE
10600 check_and_cleanup_lustre