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"
3458 log "files compared the same"
3459 cleanup || error "cleanup failed with $?"
3461 run_test 38 "MDS recreates missing lov_objid file from OST data"
3464 [[ -n "$(type -p perl)" ]] || skip_env "need perl for leak_finder.pl"
3468 cleanup || error "cleanup failed with $?"
3469 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
3470 error "memory leak detected" || true
3472 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
3474 test_40() { # bug 15759
3475 start_ost || error "Unable to start OST1"
3476 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
3477 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
3479 cleanup || error "cleanup failed with rc $?"
3481 run_test 40 "race during service thread startup"
3483 test_41a() { #bug 14134
3484 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3485 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
3486 skip "Loop devices does not work with nosvc option"
3489 combined_mgs_mds || skip "needs combined MGT and MDT device"
3491 start_mdt 1 -o nosvc -n
3492 if [ $MDSCOUNT -ge 2 ]; then
3493 for num in $(seq 2 $MDSCOUNT); do
3494 start_mdt $num || return
3497 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
3498 start_mdt 1 -o nomgs,force
3499 mount_client $MOUNT || error "mount_client $MOUNT failed"
3502 echo "blah blah" > $MOUNT/$tfile
3505 umount_client $MOUNT || error "umount_client $MOUNT failed"
3506 stop ost1 -f || error "unable to stop OST1"
3507 stop_mds || error "Unable to stop MDS"
3508 stop_mds || error "Unable to stop MDS on second try"
3510 run_test 41a "mount mds with --nosvc and --nomgs"
3513 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3514 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
3515 skip "Loop devices does not work with nosvc option"
3518 ! combined_mgs_mds && skip "needs combined mgs device"
3522 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3524 start_mdt 1 -o nosvc -n
3525 if [ $MDSCOUNT -ge 2 ]; then
3526 for num in $(seq 2 $MDSCOUNT); do
3527 start_mdt $num || return
3530 start_ost || error "Unable to start OST1"
3531 start_mdt 1 -o nomgs,force
3532 mount_client $MOUNT || error "mount_client $MOUNT failed"
3535 echo "blah blah" > $MOUNT/$tfile
3536 cat $MOUNT/$tfile || error "cat $MOUNT/$tfile failed"
3538 umount_client $MOUNT -f || error "umount_client $MOUNT failed"
3539 stop_ost || error "Unable to stop OST1"
3540 stop_mds || error "Unable to stop MDS"
3541 stop_mds || error "Unable to stop MDS on second try"
3543 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
3546 local oss_list=$(comma_list $(osts_nodes))
3548 [[ "$MDS1_VERSION" -ge $(version_code 2.6.52) ]] ||
3549 [[ "$MDS1_VERSION" -ge $(version_code 2.5.26) &&
3550 "$MDS1_VERSION" -lt $(version_code 2.5.50) ]] ||
3551 [[ "$MDS1_VERSION" -ge $(version_code 2.5.4) &&
3552 "$MDS1_VERSION" -lt $(version_code 2.5.11) ]] ||
3553 skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"
3555 # ensure mds1 ost1 have been created even if running sub-test standalone
3558 cleanup || error "cleanup failed"
3560 # using directly mount command instead of start() function to avoid
3561 # any side effect of // with others/externals tools/features
3562 # ("zpool import", ...)
3564 # MDT concurrent start
3566 LOAD_MODULES_REMOTE=true load_modules
3567 do_facet $SINGLEMDS "lsmod | grep -q libcfs" ||
3568 error "MDT concurrent start: libcfs module not loaded"
3570 local mds1dev=$(mdsdevname 1)
3571 local mds1mnt=$(facet_mntpt mds1)
3572 local mds1opts=$MDS_MOUNT_OPTS
3574 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3575 ! do_facet mds1 test -b $mds1dev; then
3576 mds1opts=$(csa_add "$mds1opts" -o loop)
3578 if [[ "$mds1_FSTYPE" == zfs ]]; then
3579 import_zpool mds1 || return ${PIPESTATUS[0]}
3582 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3583 do_facet mds1 "$LCTL set_param fail_loc=0x80000716"
3585 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts &
3588 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts
3593 do_facet mds1 "$LCTL set_param fail_loc=0x0"
3594 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3595 echo "1st MDT start succeed"
3596 echo "2nd MDT start failed with $rc2"
3597 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3598 echo "1st MDT start failed with $rc"
3599 echo "2nd MDT start succeed"
3602 error "unexpected concurrent MDT mounts result, rc=$rc rc2=$rc2"
3605 if [ $MDSCOUNT -ge 2 ]; then
3606 for num in $(seq 2 $MDSCOUNT); do
3607 start_mdt $num || return
3611 # OST concurrent start
3613 do_rpc_nodes $oss_list "lsmod | grep -q libcfs" ||
3614 error "OST concurrent start: libcfs module not loaded"
3616 local ost1dev=$(ostdevname 1)
3617 local ost1mnt=$(facet_mntpt ost1)
3618 local ost1opts=$OST_MOUNT_OPTS
3620 if [ "$ost1_FSTYPE" == ldiskfs ] &&
3621 ! do_facet ost1 test -b $ost1dev; then
3622 ost1opts=$(csa_add "$ost1opts" -o loop)
3624 if [[ "$ost1_FSTYPE" == zfs ]]; then
3625 import_zpool ost1 || return ${PIPESTATUS[0]}
3628 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3629 do_facet ost1 "$LCTL set_param fail_loc=0x80000716"
3631 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts &
3634 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts
3638 do_facet ost1 "$LCTL set_param fail_loc=0x0"
3639 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3640 echo "1st OST start succeed"
3641 echo "2nd OST start failed with $rc2"
3642 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3643 echo "1st OST start failed with $rc"
3644 echo "2nd OST start succeed"
3648 error "unexpected concurrent OST mounts result, rc=$rc rc2=$rc2"
3654 # verify everything ok
3659 error "MDT(s) start failed"
3667 error "OST(s) start failed"
3675 error "client start failed"
3682 error "client mount failed"
3686 run_test 41c "concurrent mounts of MDT/OST should all fail but one"
3688 test_42() { #bug 14693
3692 check_mount || error "client was not mounted"
3694 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3695 PARAM="llite.$FSNAME-*.some_wrong_param"
3697 PARAM="$FSNAME.llite.some_wrong_param"
3700 do_facet mgs $PERM_CMD $PARAM=10
3701 umount_client $MOUNT ||
3702 error "unmounting client failed with invalid llite param"
3703 mount_client $MOUNT ||
3704 error "mounting client failed with invalid llite param"
3706 do_facet mgs $PERM_CMD $PARAM=20
3707 cleanup || error "stopping $FSNAME failed with invalid sys param"
3709 check_mount || error "client was not mounted with invalid sys param"
3710 cleanup || error "stopping $FSNAME failed with invalid sys param"
3712 run_test 42 "allow client/server mount/unmount with invalid config param"
3715 [[ "$MGS_VERSION" -ge $(version_code 2.5.58) ]] ||
3716 skip "Need MDS version at least 2.5.58"
3717 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
3720 USER1=$(getent passwd | grep :$ID1:$ID1: | cut -d: -f1)
3721 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1"
3724 chmod ugo+x $DIR || error "chmod 0 failed"
3725 set_persistent_param_and_check mds1 \
3726 "mdt.$FSNAME-MDT0000.root_squash" \
3727 "$FSNAME.mdt.root_squash" \
3729 wait_update $HOSTNAME \
3730 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3732 error "check llite root_squash failed!"
3733 set_persistent_param_and_check mds1 \
3734 "mdt.$FSNAME-MDT0000.nosquash_nids" \
3735 "$FSNAME.mdt.nosquash_nids" \
3737 wait_update $HOSTNAME \
3738 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3740 error "check llite nosquash_nids failed!"
3743 # create set of test files
3745 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
3746 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
3747 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
3749 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
3750 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
3752 mkdir_on_mdt0 $DIR/$tdir-rootdir || error "mkdir failed"
3753 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
3754 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
3756 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
3757 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
3758 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
3761 # check root_squash:
3762 # set root squash UID:GID to RUNAS_ID
3763 # root should be able to access only files owned by RUNAS_ID
3765 set_persistent_param_and_check mds1 \
3766 "mdt.$FSNAME-MDT0000.root_squash" \
3767 "$FSNAME.mdt.root_squash" \
3768 "$RUNAS_ID:$RUNAS_ID"
3769 wait_update $HOSTNAME \
3770 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3771 "$RUNAS_ID:$RUNAS_ID" ||
3772 error "check llite root_squash failed!"
3774 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
3775 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3776 error "$ST: root read permission is denied"
3777 echo "$ST: root read permission is granted - ok"
3780 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3781 error "$ST: root write permission is denied"
3782 echo "$ST: root write permission is granted - ok"
3784 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3785 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3786 error "$ST: root read permission is granted"
3787 echo "$ST: root read permission is denied - ok"
3790 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3791 error "$ST: root write permission is granted"
3792 echo "$ST: root write permission is denied - ok"
3794 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3795 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null &&
3796 error "$ST: root unlink permission is granted"
3797 echo "$ST: root unlink permission is denied - ok"
3799 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null &&
3800 error "$ST: root create permission is granted"
3801 echo "$ST: root create permission is denied - ok"
3804 # check root_squash is enforced independently
3805 # of client cache content
3807 # access file by USER1, keep access open
3808 # root should be denied access to user file
3810 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
3814 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
3815 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3816 { kill $pid; error "$ST: root read permission is granted"; }
3817 echo "$ST: root read permission is denied - ok"
3820 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3821 { kill $pid; error "$ST: root write permission is granted"; }
3822 echo "$ST: root write permission is denied - ok"
3828 # check nosquash_nids:
3829 # put client's NID into nosquash_nids list,
3830 # root should be able to access root file after that
3832 local NIDLIST=$($LCTL list_nids all | tr '\n' ' ')
3833 NIDLIST="2@gni $NIDLIST 192.168.0.[2,10]@tcp"
3834 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
3835 set_persistent_param_and_check mds1 \
3836 "mdt.$FSNAME-MDT0000.nosquash_nids" \
3837 "$FSNAME-MDTall.mdt.nosquash_nids" \
3839 wait_update $HOSTNAME \
3840 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3842 error "check llite nosquash_nids failed!"
3844 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3845 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3846 error "$ST: root read permission is denied"
3847 echo "$ST: root read permission is granted - ok"
3850 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3851 error "$ST: root write permission is denied"
3852 echo "$ST: root write permission is granted - ok"
3854 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3855 rm $DIR/$tdir-rootdir/tfile-1 ||
3856 error "$ST: root unlink permission is denied"
3857 echo "$ST: root unlink permission is granted - ok"
3858 touch $DIR/$tdir-rootdir/tfile-2 ||
3859 error "$ST: root create permission is denied"
3860 echo "$ST: root create permission is granted - ok"
3861 cleanup || error "cleanup failed with $?"
3863 run_test 43a "check root_squash and nosquash_nids"
3865 test_43b() { # LU-5690
3866 [[ "$MGS_VERSION" -ge $(version_code 2.7.62) ]] ||
3867 skip "Need MGS version 2.7.62+"
3869 if [[ -z "$fs2mds_DEV" ]]; then
3870 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
3871 skip_env "mixed loopback and real device not working"
3874 local fs2mdsdev=$(mdsdevname 1_2)
3875 local fs2mdsvdev=$(mdsvdevname 1_2)
3877 # temporarily use fs2mds as fs2mgs
3879 local fs2mgsdev=$fs2mdsdev
3880 local fs2mgsvdev=$fs2mdsvdev
3882 local fsname=test1234
3884 load_module llite/lustre
3885 local client_ip=$(host_nids_address $HOSTNAME $NETTYPE)
3886 local host=${client_ip//*./}
3887 local net=${client_ip/%$host/}
3888 local nosquash_nids=$(h2nettype $net[$host,$host,$host])
3890 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
3891 --param mdt.root_squash=$RUNAS_ID:$RUNAS_ID \
3892 --param mdt.nosquash_nids=$nosquash_nids \
3893 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
3894 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS || error "start fs2mgs failed"
3895 stop $fs2mgs -f || error "stop fs2mgs failed"
3896 cleanup || error "cleanup failed with $?"
3898 run_test 43b "parse nosquash_nids with commas in expr_list"
3902 check_mount || error "check_mount"
3903 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
3905 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
3906 for VAL in $UUIDS; do
3907 NID=$(echo $VAL | cut -d= -f1)
3908 CLUUID=$(echo $VAL | cut -d= -f2)
3909 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
3911 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
3912 cleanup || error "cleanup failed with $?"
3914 run_test 44 "mounted client proc entry exists"
3918 check_mount || error "check_mount"
3919 stop_mds || error "Unable to stop MDS"
3923 #define OBD_FAIL_PTLRPC_LONG_REPL_UNLINK 0x50f
3924 do_facet client "$LCTL set_param fail_loc=0x8000050f"
3927 manual_umount_client --force || error "manual_umount_client failed"
3928 do_facet client "$LCTL set_param fail_loc=0x0"
3929 start_mds || error "unable to start MDS"
3930 mount_client $MOUNT || error "mount_client $MOUNT failed"
3931 cleanup || error "cleanup failed with $?"
3933 run_test 45 "long unlink handling in ptlrpcd"
3940 umount_client $MOUNT2 || rc=$?
3941 umount_client $MOUNT || rc=$?
3942 while [ $count -gt 0 ]; do
3943 stop ost${count} -f || rc=$?
3947 cleanup_nocli || rc=$?
3948 #writeconf to remove all ost2 traces for subsequent tests
3949 writeconf_or_reformat
3954 echo "Testing with $OSTCOUNT OSTs"
3956 start_mds || error "unable to start MDS"
3957 #first client should see only one ost
3958 start_ost || error "Unable to start OST1"
3959 wait_osc_import_state mds ost FULL
3961 mount_client $MOUNT || error "mount_client $MOUNT failed"
3962 trap "cleanup_46a $OSTCOUNT" EXIT ERR
3965 for (( i=2; i<=$OSTCOUNT; i++ )); do
3966 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
3967 error "start_ost$i $(ostdevname $i) failed"
3970 # wait until osts in sync
3971 for (( i=2; i<=$OSTCOUNT; i++ )); do
3972 wait_osc_import_state mds ost$i FULL
3973 wait_osc_import_ready client ost$i
3976 #second client see all ost's
3978 mount_client $MOUNT2 || error "mount_client failed"
3979 $LFS setstripe -c -1 $MOUNT2 ||
3980 error "$LFS setstripe -c -1 $MOUNT2 failed"
3981 $LFS getstripe $MOUNT2 || error "$LFS getstripe $MOUNT2 failed"
3983 echo "ok" > $MOUNT2/widestripe
3984 $LFS getstripe $MOUNT2/widestripe ||
3985 error "$LFS getstripe $MOUNT2/widestripe failed"
3986 # fill acl buffer for avoid expand lsm to them
3987 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd |
3989 setfacl -m $acl $MOUNT2/widestripe
3993 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3995 cleanup_46a $OSTCOUNT || error "cleanup_46a failed"
3997 run_test 46a "handle ost additional - wide striped file"
4002 check_mount || error "check_mount failed"
4003 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
4007 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
4008 if echo $ns | grep "MDT[[:digit:]]*"; then
4011 lrs=$(echo $ns | sed 's/.*lru_size=//')
4012 lru_size[count]=$lrs
4017 facet_failover $SINGLEMDS
4018 client_up || error "client_up failed"
4021 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
4022 if echo $ns | grep "MDT[[:digit:]]*"; then
4025 lrs=$(echo $ns | sed 's/.*lru_size=//')
4026 if ! test "$lrs" -eq "${lru_size[count]}"; then
4027 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
4028 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
4033 cleanup || error "cleanup failed with $?"
4035 run_test 47 "server restart does not make client loss lru_resize settings"
4040 # reformat after this test is needed - if the test fails,
4041 # we will have unkillable file at FS
4045 test_48() { # bz-17636 LU-7473
4049 check_mount || error "check_mount failed"
4051 $LFS setstripe -c -1 $MOUNT ||
4052 error "$LFS setstripe -c -1 $MOUNT failed"
4053 $LFS getstripe $MOUNT || error "$LFS getstripe $MOUNT failed"
4055 echo "ok" > $MOUNT/widestripe
4056 $LFS getstripe $MOUNT/widestripe ||
4057 error "$LFS getstripe $MOUNT/widestripe failed"
4059 # In the future, we may introduce more EAs, such as selinux, enlarged
4060 # LOV EA, and so on. These EA will use some EA space that is shared by
4061 # ACL entries. So here we only check some reasonable ACL entries count,
4062 # instead of the max number that is calculated from the max_ea_size.
4063 if [ "$MDS1_VERSION" -lt $(version_code 2.8.57) ]; then
4064 count=28 # hard coded of RPC protocol
4065 elif large_xattr_enabled; then
4066 count=4500 # max_num 8187 max_ea_size = 65452
4067 # not create too many (4500) to save test time
4069 count=450 # max_num 497 max_ea_size = 4012
4072 echo "It is expected to hold at least $count ACL entries"
4073 trap cleanup_48 EXIT ERR
4074 for ((i = 0; i < $count; i++)) do
4075 setfacl -m u:$((i + 100)):rw $MOUNT/widestripe ||
4076 error "Fail to setfacl for $MOUNT/widestripe at $i"
4079 cancel_lru_locks mdc
4080 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
4081 local r_count=$(getfacl $MOUNT/widestripe | grep "user:" | wc -l)
4083 count=$((count + 1)) # for the entry "user::rw-"
4085 [ $count -eq $r_count ] ||
4086 error "Expected ACL entries $count, but got $r_count"
4090 run_test 48 "too many acls on file"
4092 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
4093 test_49a() { # bug 17710
4094 local timeout_orig=$TIMEOUT
4095 local ldlm_timeout_orig=$LDLM_TIMEOUT
4096 local LOCAL_TIMEOUT=20
4098 LDLM_TIMEOUT=$LOCAL_TIMEOUT
4099 TIMEOUT=$LOCAL_TIMEOUT
4103 check_mount || error "client mount failed"
4105 echo "check ldlm_timout..."
4106 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
4107 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
4108 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
4110 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
4111 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
4114 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
4115 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
4118 umount_client $MOUNT || error "umount_client $MOUNT failed"
4119 stop_ost || error "problem stopping OSS"
4120 stop_mds || error "problem stopping MDS"
4122 LDLM_TIMEOUT=$ldlm_timeout_orig
4123 TIMEOUT=$timeout_orig
4125 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
4127 test_49b() { # bug 17710
4128 local timeout_orig=$TIMEOUT
4129 local ldlm_timeout_orig=$LDLM_TIMEOUT
4130 local LOCAL_TIMEOUT=20
4132 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
4133 TIMEOUT=$LOCAL_TIMEOUT
4137 check_mount || error "client mount failed"
4139 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
4140 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
4141 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
4143 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
4144 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
4147 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
4148 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
4151 cleanup || error "cleanup failed"
4153 LDLM_TIMEOUT=$ldlm_timeout_orig
4154 TIMEOUT=$timeout_orig
4156 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
4159 # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1
4161 # Test both statfs and lfs df and fail if either one fails
4162 multiop_bg_pause $1 f_
4165 killall -USR1 multiop
4166 [ $RC -ne 0 ] && log "lazystatfs multiop failed"
4167 wait $PID || { RC=$?; log "multiop return error "; }
4169 # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1
4174 if kill -s 0 $PID; then
4177 log "lazystatfs lfs df failed to complete in 5s"
4185 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4186 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4188 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
4190 cleanup || error "cleanup failed with rc $?"
4192 run_test 50a "lazystatfs all servers available"
4196 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4197 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4199 # Wait for client to detect down OST
4200 stop_ost || error "Unable to stop OST1"
4201 wait_osc_import_state client ost DISCONN
4203 log "OSCs should all be DISCONN"
4205 lazystatfs $MOUNT || error "lazystatfs should not return EIO"
4207 umount_client $MOUNT || error "Unable to unmount client"
4208 stop_mds || error "Unable to stop MDS"
4210 run_test 50b "lazystatfs all servers down"
4213 start_mds || error "Unable to start MDS"
4214 start_ost || error "Unable to start OST1"
4215 start_ost2 || error "Unable to start OST2"
4216 mount_client $MOUNT || error "Unable to mount client"
4217 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4218 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4220 # Wait for client to detect down OST
4221 stop_ost || error "Unable to stop OST1"
4222 wait_osc_import_state mds ost DISCONN
4223 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
4225 umount_client $MOUNT || error "Unable to unmount client"
4226 stop_ost2 || error "Unable to stop OST2"
4227 stop_mds || error "Unable to stop MDS"
4228 #writeconf to remove all ost2 traces for subsequent tests
4229 writeconf_or_reformat
4231 run_test 50c "lazystatfs one server down"
4234 start_mds || error "Unable to start MDS"
4235 start_ost || error "Unable to start OST1"
4236 start_ost2 || error "Unable to start OST2"
4237 mount_client $MOUNT || error "Unable to mount client"
4238 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4239 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4241 # Issue the statfs during the window where the client still
4242 # belives the OST to be available but it is in fact down.
4243 # No failure just a statfs which hangs for a timeout interval.
4244 stop_ost || error "Unable to stop OST1"
4245 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
4247 umount_client $MOUNT || error "Unable to unmount client"
4248 stop_ost2 || error "Unable to stop OST2"
4249 stop_mds || error "Unable to stop MDS"
4250 #writeconf to remove all ost2 traces for subsequent tests
4251 writeconf_or_reformat
4253 run_test 50d "lazystatfs client/server conn race"
4260 start_mds || error "Unable to start MDS"
4261 #first client should see only one ost
4262 start_ost || error "Unable to start OST1"
4263 wait_osc_import_state mds ost FULL
4265 # Wait for client to detect down OST
4266 stop_ost || error "Unable to stop OST1"
4267 wait_osc_import_state mds ost DISCONN
4269 mount_client $MOUNT || error "Unable to mount client"
4270 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
4272 multiop_bg_pause $MOUNT _f
4276 if [ $RC1 -ne 0 ]; then
4277 log "multiop failed $RC1"
4280 sleep $(( $TIMEOUT+1 ))
4282 [ $? -ne 0 ] && error "process isn't sleep"
4283 start_ost || error "Unable to start OST1"
4284 wait $pid || error "statfs failed"
4287 umount_client $MOUNT || error "Unable to unmount client"
4288 stop_ost || error "Unable to stop OST1"
4289 stop_mds || error "Unable to stop MDS"
4291 run_test 50e "normal statfs all servers down"
4296 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
4298 start_mds || error "Unable to start mds"
4299 #first client should see only one ost
4300 start_ost || error "Unable to start OST1"
4301 wait_osc_import_state mds ost FULL
4303 start_ost2 || error "Unable to start OST2"
4304 wait_osc_import_state mds ost2 FULL
4306 # Wait for client to detect down OST
4307 stop_ost2 || error "Unable to stop OST2"
4309 wait_osc_import_state mds ost2 DISCONN
4310 mount_client $MOUNT || error "Unable to mount client"
4311 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
4313 multiop_bg_pause $MOUNT _f
4317 if [ $RC1 -ne 0 ]; then
4318 log "lazystatfs multiop failed $RC1"
4321 sleep $(( $TIMEOUT+1 ))
4323 [ $? -ne 0 ] && error "process isn't sleep"
4324 start_ost2 || error "Unable to start OST2"
4325 wait $pid || error "statfs failed"
4326 stop_ost2 || error "Unable to stop OST2"
4329 umount_client $MOUNT -f || error "Unable to unmount client"
4330 stop_ost || error "Unable to stop OST1"
4331 stop_mds || error "Unable to stop MDS"
4332 #writeconf to remove all ost2 traces for subsequent tests
4333 writeconf_or_reformat
4335 run_test 50f "normal statfs one server in down"
4338 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs"
4340 start_ost2 || error "Unable to start OST2"
4341 wait_osc_import_state mds ost2 FULL
4342 wait_osc_import_ready client ost2
4344 if [[ $PERM_CMD == *"set_param -P"* ]]; then
4345 local PARAM="osc.${FSNAME}-OST0001*.active"
4347 local PARAM="${FSNAME}-OST0001.osc.active"
4350 $LFS setstripe -c -1 $DIR/$tfile || error "$LFS setstripe failed"
4351 do_facet mgs $PERM_CMD $PARAM=0 || error "Unable to deactivate OST"
4353 umount_client $MOUNT || error "Unable to unmount client"
4354 mount_client $MOUNT || error "Unable to mount client"
4355 # This df should not cause a panic
4358 do_facet mgs $PERM_CMD $PARAM=1 || error "Unable to activate OST"
4359 rm -f $DIR/$tfile || error "unable to remove file $DIR/$tfile"
4360 umount_client $MOUNT || error "Unable to unmount client"
4361 stop_ost2 || error "Unable to stop OST2"
4362 stop_ost || error "Unable to stop OST1"
4363 stop_mds || error "Unable to stop MDS"
4364 #writeconf to remove all ost2 traces for subsequent tests
4365 writeconf_or_reformat
4367 run_test 50g "deactivated OST should not cause panic"
4371 # prepare MDT/OST, make OSC inactive for OST1
4372 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs"
4374 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
4375 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
4376 error "tunefs OST1 failed"
4377 start_mds || error "Unable to start MDT"
4378 start_ost || error "Unable to start OST1"
4379 start_ost2 || error "Unable to start OST2"
4380 mount_client $MOUNT || error "client start failed"
4382 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4384 # activatate OSC for OST1
4385 set_persistent_param_and_check client \
4386 "osc.${FSNAME}-OST0000-osc-[!M]*.active" \
4387 "${FSNAME}-OST0000.osc.active" 1
4389 mkdir $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
4390 $LFS setstripe -c -1 -i 0 $DIR/$tdir/2 ||
4391 error "$LFS setstripe $DIR/$tdir/2 failed"
4392 sleep 1 && echo "create a file after OST1 is activated"
4393 # doing some io, shouldn't crash
4394 dd if=/dev/zero of=$DIR/$tdir/2/$tfile-io bs=1M count=10
4396 # check OSC import is working
4397 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
4398 error "some OSC imports are still not connected"
4402 umount_client $MOUNT || error "Unable to umount client"
4403 stop_ost2 || error "Unable to stop OST2"
4404 cleanup_nocli || error "cleanup_nocli failed with $?"
4406 run_test 50h "LU-642: activate deactivated OST"
4409 # prepare MDT/OST, make OSC inactive for OST1
4410 [ "$MDSCOUNT" -lt "2" ] && skip "needs >= 2 MDTs"
4413 [ $(facet_fstype mds2) == zfs ] && import_zpool mds2
4414 do_facet mds2 "$TUNEFS --param mdc.active=0 $(mdsdevname 2)" ||
4415 error "tunefs MDT2 failed"
4416 start_mds || error "Unable to start MDT"
4417 start_ost || error "Unable to start OST1"
4418 start_ost2 || error "Unable to start OST2"
4419 mount_client $MOUNT || error "client start failed"
4421 mkdir_on_mdt0 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4423 if [[ $PERM_CMD == *"set_param -P"* ]]; then
4424 $PERM_CMD mdc.${FSNAME}-MDT0001-mdc-*.active=0 &&
4425 error "deactive MDC0 succeeds"
4427 $PERM_CMD ${FSNAME}-MDT0000.mdc.active=0 &&
4428 error "deactive MDC0 succeeds"
4431 # activate MDC for MDT2
4432 set_persistent_param_and_check client \
4433 "mdc.${FSNAME}-MDT0001-mdc-*.active" \
4434 "${FSNAME}-MDT0001.mdc.active" 1
4436 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
4437 if [ "$MDS1_VERSION" -ge $(version_code 2.7.60) ]
4439 wait_dne_interconnect
4441 $LFS mkdir -i1 $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
4443 createmany -o $DIR/$tdir/2/$tfile-%d 1 || error "create files failed"
4445 rm -rf $DIR/$tdir/2 || error "unlink dir failed"
4447 # deactivate MDC for MDT2
4448 set_persistent_param_and_check client \
4449 "mdc.${FSNAME}-MDT0001-mdc-*.active" \
4450 "${FSNAME}-MDT0001.mdc.active" 0
4452 wait_osp_active mds ${FSNAME}-MDT0001 1 0
4454 $LFS mkdir -i1 $DIR/$tdir/2 &&
4455 error "mkdir $DIR/$tdir/2 succeeds after deactive MDT"
4457 $LFS mkdir -i0 -c$MDSCOUNT $DIR/$tdir/striped_dir ||
4458 error "mkdir $DIR/$tdir/striped_dir fails after deactive MDT2"
4460 local stripe_count=$($LFS getdirstripe -c $DIR/$tdir/striped_dir)
4461 [ $stripe_count -eq $((MDSCOUNT - 1)) ] ||
4462 error "wrong $stripe_count != $((MDSCOUNT -1)) for striped_dir"
4465 umount_client $MOUNT || error "Unable to umount client"
4470 run_test 50i "activate deactivated MDT"
4473 local LOCAL_TIMEOUT=20
4477 check_mount || error "check_mount failed"
4479 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
4480 $LFS setstripe -c -1 $MOUNT/$tdir ||
4481 error "$LFS setstripe -c -1 $MOUNT/$tdir failed"
4482 #define OBD_FAIL_MDS_REINT_DELAY 0x142
4483 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x142"
4484 touch $MOUNT/$tdir/$tfile &
4487 start_ost2 || error "Unable to start OST1"
4489 stop_ost2 || error "Unable to stop OST1"
4490 umount_client $MOUNT -f || error “unmount $MOUNT failed”
4491 cleanup_nocli || error “stop server failed”
4492 #writeconf to remove all ost2 traces for subsequent tests
4493 writeconf_or_reformat
4495 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
4504 do_node $node mkdir -p $dest
4505 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
4507 do_node $node 'tar cf - '$*' | tar xf - -C '$dest';
4508 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
4509 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
4511 do_node $node 'getfattr -d -m "[a-z]*\\." '$*' > '$xattrs
4512 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
4522 local backup2=${TMP}/backup2
4524 do_node $node mkdir -p $backup2
4525 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
4527 do_node $node 'tar cf - '$*' | tar xf - -C '$backup2';
4528 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
4529 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
4531 do_node $node "diff -rq $backup $backup2"
4532 [ $? -eq 0 ] || { error "contents differ"; return 3; }
4534 local xattrs2=${TMP}/xattrs2
4535 do_node $node 'getfattr -d -m "[a-z]*\\." '$*' > '$xattrs2
4536 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
4538 do_node $node "diff $xattrs $xattrs2"
4539 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
4541 do_node $node "rm -rf $backup2 $xattrs2"
4542 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
4546 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4547 skip "ldiskfs only test"
4550 start_mds || error "Unable to start MDS"
4551 start_ost || error "Unable to start OST1"
4552 mount_client $MOUNT || error "Unable to mount client"
4555 local ost1mnt=$(facet_mntpt ost1)
4556 local ost1node=$(facet_active_host ost1)
4557 local ost1tmp=$TMP/conf52
4560 mkdir $DIR/$tdir || error "Unable to create $DIR/$tdir"
4561 touch $TMP/modified_first || error "Unable to create temporary file"
4562 local mtime=$(stat -c %Y $TMP/modified_first)
4563 do_node $ost1node "mkdir -p $ost1tmp &&
4564 touch -m -d @$mtime $ost1tmp/modified_first" ||
4565 error "Unable to create temporary file"
4568 $LFS setstripe -c -1 -S 1M $DIR/$tdir || error "$LFS setstripe failed"
4570 for (( i=0; i < nrfiles; i++ )); do
4571 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c ||
4572 error "multiop failed"
4577 # sync all the data and make sure no pending data on the client,
4578 # thus the SOM xattr would not be changed any more.
4579 cancel_lru_locks osc
4582 echo backup files to $TMP/$tdir
4583 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
4584 copy_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
4585 error "Unable to copy files"
4587 umount_client $MOUNT || error "Unable to umount client"
4588 stop_ost || error "Unable to stop ost1"
4590 echo mount ost1 as ldiskfs
4591 do_node $ost1node mkdir -p $ost1mnt || error "Unable to create $ost1mnt"
4592 if ! do_node $ost1node test -b $ost1_dev; then
4595 do_node $ost1node mount -t "$ost1_FSTYPE" $loop $ost1_dev \
4597 error "Unable to mount ost1 as ldiskfs"
4600 echo backup objects to $ost1tmp/objects
4601 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
4602 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
4603 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
4605 error "Unable to copy objects"
4607 # move objects to lost+found
4608 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
4609 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
4611 do_node $ost1node "umount $ost1mnt" ||
4612 error "Unable to umount ost1 as ldiskfs"
4614 start_ost || error "Unable to start OST1"
4615 mount_client $MOUNT || error "Unable to mount client"
4617 local REPAIRED=$(do_node $ost1node "$LCTL get_param \
4618 -n osd-ldiskfs.$FSNAME-OST0000.oi_scrub" |
4619 awk '/^lf_repa[ri]*ed/ { print $2 }')
4620 [ $REPAIRED -gt 0 ] ||
4621 error "Some entry under /lost+found should be repaired"
4624 diff_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
4625 error "Unable to diff files"
4627 rm -rf $TMP/$tdir $TMP/file_xattrs ||
4628 error "Unable to delete temporary files"
4629 do_node $ost1node "rm -rf $ost1tmp" ||
4630 error "Unable to delete temporary files"
4631 cleanup || error "cleanup failed with $?"
4633 run_test 52 "check recovering objects from lost+found"
4635 # Checks threads_min/max/started for some service
4637 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
4638 # parameter pattern prefix like 'ost.*.ost'.
4651 local msg="Insane $modname thread counts"
4652 local ncpts=$(check_cpt_number $facet)
4656 check_mount || return 41
4658 # We need to expand $parampat, but it may match multiple parameters, so
4659 # we'll pick the first one
4660 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
4661 error "Couldn't expand ${parampat}.threads_min parameter name"
4665 # Remove the .threads_min part
4666 paramp=${paramp%.threads_min}
4668 # Check for sanity in defaults
4669 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4671 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4673 tstarted=$(do_facet $facet "$LCTL get_param \
4674 -n ${paramp}.threads_started" || echo 0)
4675 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' ||
4677 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4679 nthrs=$(expr $tmax - $tmin)
4680 if [ $nthrs -lt $ncpts ]; then
4686 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
4687 skip_env "module parameter forced $facet thread count"
4689 # Check that we can change min/max
4690 do_facet $facet "$LCTL set_param \
4691 ${paramp}.threads_min=$((tmin + nthrs))"
4692 do_facet $facet "$LCTL set_param \
4693 ${paramp}.threads_max=$((tmax - nthrs))"
4694 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4696 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4698 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) &&
4699 $tmax2 == ($tmax - $nthrs)))' || return $?
4701 # Check that we can set min/max to the same value
4702 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4704 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$tmin"
4705 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4707 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4709 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
4711 # Check that we can't set max < min
4712 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$((tmin - 1))"
4713 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4715 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4717 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
4719 # We need to ensure that we get the module options desired; to do this
4720 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
4721 LOAD_MODULES_REMOTE=true
4724 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
4726 setmodopts -a $modname "$newvalue" oldvalue
4729 check_mount || return 41
4731 # Restore previous setting of MODOPTS_*
4732 setmodopts $modname "$oldvalue"
4734 # Check that $opts took
4735 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4737 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4739 tstarted=$(do_facet $facet \
4740 "$LCTL get_param -n ${paramp}.threads_started" || echo 0)
4741 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4750 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
4751 cleanup || error "cleanup failed with rc $?"
4753 run_test 53a "check OSS thread count params"
4757 local mds=$(do_facet $SINGLEMDS "$LCTL get_param \
4758 -N mds.*.*.threads_max 2>/dev/null")
4759 if [ -z "$mds" ]; then
4760 #running this on an old MDT
4761 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
4763 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
4765 cleanup || error "cleanup failed with $?"
4767 run_test 53b "check MDS thread count params"
4770 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4771 skip "ldiskfs only test"
4774 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p ||
4775 error "llverdev failed with rc=$?"
4778 run_test 54a "test llverdev and partial verify of device"
4781 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4782 skip "ldiskfs only test"
4786 run_llverfs $MOUNT -p || error "llverfs failed with rc=$?"
4787 cleanup || error "cleanup failed with rc=$?"
4789 run_test 54b "test llverfs and partial verify of filesystem"
4793 local max_ost_index=$1
4795 echo -n $(((max_ost_index + 1) * 8))
4799 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4800 skip "ldiskfs only test"
4803 local mdsdev=$(mdsdevname 1)
4804 local mdsvdev=$(mdsvdevname 1)
4808 if ! combined_mgs_mds; then
4809 stop_mgs || error "stopping MGS service failed"
4810 format_mgs || error "formatting MGT failed"
4812 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
4814 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
4815 --reformat $(ostdevname 1) $(ostvdevname 1)
4821 echo checking size of lov_objid for ost index $i
4822 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" |
4823 grep ^User | awk -F 'Size: ' '{print $2}')
4824 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
4825 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
4827 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
4834 run_test 55 "check lov_objid size"
4837 local mds_journal_size_orig=$MDSJOURNALSIZE
4843 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
4844 $(ostdevname 1) $(ostvdevname 1)
4845 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
4846 $(ostdevname 2) $(ostvdevname 2)
4849 start_ost || error "Unable to start first ost (idx 10000)"
4850 start_ost2 || error "Unable to start second ost (idx 1000)"
4851 mount_client $MOUNT || error "Unable to mount client"
4855 # test instantiating PFL components with sparse index LU-15513
4856 mkdir -p $MOUNT/$tdir
4857 $LFS setstripe -E 4M -c 1 -E 1G -c 4 -S4M -E eof -c -1 $MOUNT/$tdir
4858 dd if=/dev/zero of=$MOUNT/$tdir/$tfile bs=4K count=1 seek=10k ||
4859 error "dd to second component failed"
4861 if [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
4862 [[ "$MDS1_VERSION" -ge $(version_code 2.5.4) &&
4863 "$MDS1_VERSION" -lt $(version_code 2.5.11) ]]; then
4864 wait_osc_import_state mds ost1 FULL
4865 wait_osc_import_state mds ost2 FULL
4866 $LFS setstripe --stripe-count=-1 $DIR/$tfile ||
4867 error "Unable to setstripe $DIR/$tfile"
4868 n=$($LFS getstripe --stripe-count $DIR/$tfile)
4869 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
4874 MDSJOURNALSIZE=$mds_journal_size_orig
4877 run_test 56a "check big OST indexes and out-of-index-order start"
4882 umount_client $MOUNT -f || error "unmount client failed"
4891 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs"
4893 trap cleanup_56b EXIT RETURN ERR
4896 if ! combined_mgs_mds ; then
4901 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) --index=0 --reformat \
4902 $(mdsdevname 1) $(mdsvdevname 1)
4903 add mds2 $(mkfs_opts mds2 $(mdsdevname 2)) --index=1 --reformat \
4904 $(mdsdevname 2) $(mdsvdevname 2)
4905 add mds3 $(mkfs_opts mds3 $(mdsdevname 3)) --index=1000 --reformat \
4906 $(mdsdevname 3) $(mdsvdevname 3)
4910 start_mdt 1 || error "MDT 1 (idx 0) start failed"
4911 start_mdt 2 || error "MDT 2 (idx 1) start failed"
4912 start_mdt 3 || error "MDT 3 (idx 1000) start failed"
4913 start_ost || error "Unable to start first ost"
4914 start_ost2 || error "Unable to start second ost"
4916 do_nodes $(comma_list $(mdts_nodes)) \
4917 "$LCTL set_param mdt.*.enable_remote_dir=1 \
4918 mdt.*.enable_remote_dir_gid=-1"
4920 mount_client $MOUNT || error "Unable to mount client"
4922 $LFS mkdir -c3 $MOUNT/$tdir || error "failed to make testdir"
4924 echo "This is test file 1!" > $MOUNT/$tdir/$tfile.1 ||
4925 error "failed to make test file 1"
4926 echo "This is test file 2!" > $MOUNT/$tdir/$tfile.2 ||
4927 error "failed to make test file 2"
4928 echo "This is test file 1000!" > $MOUNT/$tdir/$tfile.1000 ||
4929 error "failed to make test file 1000"
4931 rm -rf $MOUNT/$tdir || error "failed to remove testdir"
4933 $LFS mkdir -i1000 $MOUNT/$tdir.1000 ||
4934 error "create remote dir at idx 1000 failed"
4937 echo "=== START lfs df OUTPUT ==="
4939 echo "==== END lfs df OUTPUT ===="
4941 mdtcnt=$(echo -e "$output" | grep $FSNAME-MDT | wc -l)
4942 ostcnt=$(echo -e "$output" | grep $FSNAME-OST | wc -l)
4944 echo "lfs df returned mdt count $mdtcnt and ost count $ostcnt"
4945 [ $mdtcnt -eq 3 ] || error "lfs df returned wrong mdt count"
4946 [ $ostcnt -eq 2 ] || error "lfs df returned wrong ost count"
4948 echo "This is test file 1!" > $MOUNT/$tdir.1000/$tfile.1 ||
4949 error "failed to make test file 1"
4950 echo "This is test file 2!" > $MOUNT/$tdir.1000/$tfile.2 ||
4951 error "failed to make test file 2"
4952 echo "This is test file 1000!" > $MOUNT/$tdir.1000/$tfile.1000 ||
4953 error "failed to make test file 1000"
4954 rm -rf $MOUNT/$tdir.1000 || error "failed to remove remote_dir"
4957 echo "=== START lfs mdts OUTPUT ==="
4959 echo "==== END lfs mdts OUTPUT ===="
4961 echo -e "$output" | grep -v "MDTS:" | awk '{print $1}' |
4962 sed 's/://g' > $TMP/mdts-actual.txt
4963 sort $TMP/mdts-actual.txt -o $TMP/mdts-actual.txt
4965 echo -e "0\n1\n1000" > $TMP/mdts-expected.txt
4967 diff $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4970 rm $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4972 [ $result -eq 0 ] || error "target_obd proc file is incorrect!"
4974 run_test 56b "test target_obd correctness with nonconsecutive MDTs"
4976 test_57a() { # bug 22656
4977 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4978 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4979 tail -1 | awk '{print $1}')
4980 writeconf_or_reformat
4981 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
4982 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" ||
4983 error "tunefs failed"
4985 start_ost && error "OST registration from failnode should fail"
4988 run_test 57a "initial registration from failnode should fail (should return errs)"
4991 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4992 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4993 tail -1 | awk '{print $1}')
4994 writeconf_or_reformat
4995 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
4996 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" ||
4997 error "tunefs failed"
4999 start_ost || error "OST registration from servicenode should not fail"
5002 run_test 57b "initial registration from servicenode should not fail"
5005 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
5008 test_58() { # bug 22658
5009 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
5011 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5012 createmany -o $DIR/$tdir/$tfile-%d 100
5013 unlinkmany $DIR/$tdir/$tfile-%d 100
5014 stop_mds || error "Unable to stop MDS"
5016 local MNTDIR=$(facet_mntpt $SINGLEMDS)
5017 local devname=$(mdsdevname ${SINGLEMDS//mds/})
5019 # remove all files from the OBJECTS dir
5020 mount_fstype $SINGLEMDS
5022 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
5024 unmount_fstype $SINGLEMDS
5025 # restart MDS with missing llog files
5026 start_mds || error "unable to start MDS"
5027 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0"
5030 run_test 58 "missing llog files must not prevent MDT from mounting"
5033 start_mgsmds >> /dev/null
5034 local C1=$(count_osts)
5035 if [ $C1 -eq 0 ]; then
5036 start_ost >> /dev/null
5040 echo "original ost count: $C1 (expect > 0)"
5041 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
5042 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
5043 local C2=$(count_osts)
5044 echo "after mdt writeconf count: $C2 (expect 0)"
5045 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
5046 echo "OST start without writeconf should fail:"
5047 start_ost >> /dev/null &&
5048 error "OST start without writeconf didn't fail"
5049 echo "OST start with writeconf should succeed:"
5050 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
5051 local C3=$(count_osts)
5052 echo "after ost writeconf count: $C3 (expect 1)"
5053 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
5054 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
5055 local C4=$(count_osts)
5056 echo "after ost2 writeconf count: $C4 (expect 2)"
5057 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
5058 stop_ost2 >> /dev/null
5059 cleanup_nocli >> /dev/null
5060 #writeconf to remove all ost2 traces for subsequent tests
5061 writeconf_or_reformat
5063 run_test 59 "writeconf mount option"
5065 test_60a() { # LU-471
5066 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5067 skip "ldiskfs only test"
5072 for num in $(seq $MDSCOUNT); do
5073 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
5074 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
5075 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
5079 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
5080 [ ${PIPESTATUS[0]} -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
5082 # MDT default has dirdata feature
5083 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
5084 # we disable uninit_bg feature
5085 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
5086 # we set stride extended options
5087 echo $dump | grep stride > /dev/null || error "stride is not set"
5091 run_test 60a "check mkfs.lustre --mkfsoptions -E -O options setting"
5094 [[ "$mds1_FSTYPE" == ldiskfs ]] || skip "ldiskfs only test"
5096 local features=$(do_facet $SINGLEMDS $DUMPE2FS $(mdsdevname 1) |
5098 [ ${PIPESTATUS[0]} -eq 0 ] || error "$DUMPE2FS $(mdsdevname 1) failed"
5101 # ea_inode feature should be enabled by default for MDTs
5102 [[ "$features" =~ "ea_inode" ]] || error "ea_inode is not set"
5103 # large_dir feature should be enabled by default for MDTs
5104 [[ "$features" =~ "large_dir" ]] || error "large_dir is not set"
5106 run_test 60b "check mkfs.lustre MDT default features"
5108 test_61a() { # LU-80
5109 local lxattr=$(large_xattr_enabled)
5111 (( "$MDS1_VERSION" >= $(version_code 2.1.53) )) ||
5112 skip "Need MDS version at least 2.1.53 for large_xattr"
5114 if [[ "$mds1_FSTYPE" == ldiskfs ]] && ! large_xattr_enabled; then
5117 for ((num=1; num <= $MDSCOUNT; num++)); do
5118 do_facet mds${num} $TUNE2FS -O ea_inode \
5119 $(mdsdevname $num) ||
5120 error "tune2fs on mds $num failed"
5124 setup || error "setting up the filesystem failed"
5125 client_up || error "starting client failed"
5127 local file=$DIR/$tfile
5128 touch $file || error "touch $file failed"
5130 local large_value="$(generate_string $(max_xattr_size))"
5131 local small_value="bar"
5133 local name="trusted.big"
5134 log "save large xattr of $(max_xattr_size) bytes on $name on $file"
5135 setfattr -n $name -v $large_value $file ||
5136 error "saving $name on $file failed"
5138 local new_value=$(get_xattr_value $name $file)
5139 [[ "$new_value" != "$large_value" ]] &&
5140 error "$name different after saving"
5142 log "shrink value of $name on $file"
5143 setfattr -n $name -v $small_value $file ||
5144 error "shrinking value of $name on $file failed"
5146 new_value=$(get_xattr_value $name $file)
5147 [[ "$new_value" != "$small_value" ]] &&
5148 error "$name different after shrinking"
5150 log "grow value of $name on $file"
5151 setfattr -n $name -v $large_value $file ||
5152 error "growing value of $name on $file failed"
5154 new_value=$(get_xattr_value $name $file)
5155 [[ "$new_value" != "$large_value" ]] &&
5156 error "$name different after growing"
5158 log "check value of $name on $file after remounting MDS"
5160 new_value=$(get_xattr_value $name $file)
5161 [[ "$new_value" != "$large_value" ]] &&
5162 error "$name different after remounting MDS"
5164 log "remove large xattr $name from $file"
5165 setfattr -x $name $file || error "removing $name from $file failed"
5167 if $lxattr && [ "$mds1_FSTYPE" == ldiskfs ]; then
5168 stopall || error "stopping for e2fsck run"
5169 for num in $(seq $MDSCOUNT); do
5170 run_e2fsck $(facet_active_host mds$num) \
5171 $(mdsdevname $num) "-y" ||
5172 error "e2fsck MDT$num failed"
5174 setup_noconfig || error "remounting the filesystem failed"
5177 # need to delete this file to avoid problems in other tests
5179 cleanup || error "stopping systems failed"
5181 run_test 61a "large xattr"
5183 test_61b() { # LU-80
5184 local lxattr=$(large_xattr_enabled)
5186 (( $MDS1_VERSION >= $(version_code 2.15.51) )) ||
5187 skip "Need MDS version at least 2.15.51 for large_xattr fix"
5189 [[ "$mds1_FSTYPE" == "ldiskfs" ]] || skip "ldiskfs specific bug"
5191 if ! large_xattr_enabled; then
5194 for (( num=1; num <= $MDSCOUNT; num++ )); do
5195 do_facet mds${num} $TUNE2FS -O ea_inode \
5196 $(mdsdevname $num) ||
5197 error "tune2fs on mds $num failed"
5201 setup || error "setting up the filesystem failed"
5202 client_up || error "starting client failed"
5204 local _file=$MOUNT/panda
5205 local large_value="$(generate_string $(max_xattr_size))"
5206 local name="trusted.big"
5208 touch ${_file} || error "touch ${_file} failed"
5209 setfattr -n $name -v $large_value ${_file} ||
5210 error "saving $name on $file failed"
5212 MDT_DEV="${FSNAME}-MDT0000"
5213 MDT_DEVNAME=$(mdsdevname ${SINGLEMDS//mds/})
5215 stopall || error "stopping for e2fsck run"
5218 ino=$(do_facet $SINGLEMDS "$DEBUGFS -R 'stat /ROOT/panda' \
5219 ${MDT_DEVNAME} | grep trusted.big")
5220 ino=$(echo "${ino}" | awk '{print $2;}')
5221 echo "large ea "${ino}
5223 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"ln $ino /lost+found\\\" \
5226 setup_noconfig || error "remounting the filesystem failed"
5228 do_facet $SINGLEMDS $LCTL lfsck_start -M ${MDT_DEV} -t namespace || {
5229 error "can't start lfsck namespace"
5233 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
5234 mdd.${MDT_DEV}.lfsck_namespace |
5235 awk '/^status/ { print \\\$2 }'" "completed" 32 || {
5236 error "(2) unexpected status"
5239 stopall || error "stopping for e2fsck run"
5240 for num in $(seq $MDSCOUNT); do
5241 run_e2fsck $(facet_active_host mds$num) \
5242 $(mdsdevname $num) "-y" ||
5243 error "e2fsck MDT$num failed"
5245 setup_noconfig || error "remounting the filesystem failed"
5247 # need to delete this file to avoid problems in other tests
5249 cleanup || error "stopping systems failed"
5251 run_test 61b "large xattr"
5254 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5255 skip "ldiskfs only test"
5257 [[ "$MDS1_VERSION" -ge $(version_code 2.2.51) ]] ||
5258 skip "Need MDS version at least 2.2.51"
5261 local mdsdev=$(mdsdevname 1)
5262 local ostdev=$(ostdevname 1)
5264 echo "disable journal for mds"
5265 do_facet mds1 $TUNE2FS -O ^has_journal $mdsdev || error "tune2fs failed"
5266 start_mds && error "MDT start should fail"
5267 echo "disable journal for ost"
5268 do_facet ost1 $TUNE2FS -O ^has_journal $ostdev || error "tune2fs failed"
5269 start_ost && error "OST start should fail"
5270 cleanup || error "cleanup failed with rc $?"
5273 run_test 62 "start with disabled journal"
5276 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5277 skip "ldiskfs only test"
5280 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ldiskfs
5281 local inode_slab=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
5282 awk '/ldiskfs_inode_cache/ { print $5 / $6 }')
5283 if [ -z "$inode_slab" ]; then
5284 skip "ldiskfs module has not been loaded"
5287 if grep -q "CONFIG_DEBUG_LOCK_ALLOC=y" /boot/config-$(uname -r); then
5288 skip "test is not compatible with CONFIG_DEBUG_LOCK_ALLOC=y"
5291 echo "$inode_slab ldiskfs inodes per page"
5292 [ "${inode_slab%.*}" -ge "3" ] && return 0
5294 # If kmalloc-128 is also 1 per page - this is a debug kernel
5295 # and so this is not an error.
5296 local kmalloc128=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
5297 awk '/^(dma-kmalloc|size)-128 / { print $5 / $6 }')
5298 # 32 128-byte chunks in 4k
5299 [ "${kmalloc128%.*}" -lt "32" ] ||
5300 error "ldiskfs inode too big, only $inode_slab objs/page, " \
5301 "kmalloc128 = $kmalloc128 objs/page"
5303 run_test 63 "Verify each page can at least hold 3 ldiskfs inodes"
5306 start_mds || error "unable to start MDS"
5307 start_ost || error "Unable to start OST1"
5308 start_ost2 || error "Unable to start second ost"
5309 mount_client $MOUNT || error "Unable to mount client"
5310 stop_ost2 || error "Unable to stop second ost"
5313 umount_client $MOUNT -f || error “unmount $MOUNT failed”
5314 cleanup_nocli || error "cleanup_nocli failed with $?"
5315 #writeconf to remove all ost2 traces for subsequent tests
5316 writeconf_or_reformat
5318 run_test 64 "check lfs df --lazy "
5320 test_65() { # LU-2237
5321 # Currently, the test is only valid for ldiskfs backend
5322 [ "$mds1_FSTYPE" != ldiskfs ] &&
5323 skip "ldiskfs only test"
5325 local devname=$(mdsdevname ${SINGLEMDS//mds/})
5326 local brpt=$(facet_mntpt brpt)
5329 if ! do_facet $SINGLEMDS "test -b $devname"; then
5333 stop_mds || error "Unable to stop MDS"
5334 local obj=$(do_facet $SINGLEMDS \
5335 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
5337 if [ -z "$obj" ]; then
5338 # The MDT may be just re-formatted, mount the MDT for the
5339 # first time to guarantee the "last_rcvd" file is there.
5340 start_mds || error "fail to mount the MDS for the first time"
5341 stop_mds || error "Unable to stop MDS"
5344 # remove the "last_rcvd" file
5345 do_facet $SINGLEMDS "mkdir -p $brpt"
5346 do_facet $SINGLEMDS \
5347 "mount -t $mds1_FSTYPE $opts $devname $brpt"
5348 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
5349 do_facet $SINGLEMDS "$UMOUNT $brpt"
5351 # restart MDS, the "last_rcvd" file should be recreated.
5352 start_mds || error "fail to restart the MDS"
5353 stop_mds || error "Unable to stop MDS"
5354 obj=$(do_facet $SINGLEMDS \
5355 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
5356 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
5358 run_test 65 "re-create the lost last_rcvd file when server mount"
5361 [[ "$MGS_VERSION" -ge $(version_code 2.3.59) ]] ||
5362 skip "Need MGS version at least 2.3.59"
5365 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
5366 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
5368 # add EXCLUDE records to config log, they are not to be
5369 # removed by lctl replace_nids
5370 set_conf_param_and_check mds \
5371 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
5372 "$FSNAME-OST0000.osc.active" \
5375 echo "replace_nids should fail if MDS, OSTs and clients are UP"
5376 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5377 error "replace_nids fail"
5379 umount_client $MOUNT || error "unmounting client failed"
5380 echo "replace_nids should fail if MDS and OSTs are UP"
5381 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5382 error "replace_nids fail"
5384 stop_ost || error "Unable to stop OST1"
5385 echo "replace_nids should fail if MDS is UP"
5386 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5387 error "replace_nids fail"
5389 stop_mds || error "stopping mds failed"
5391 if combined_mgs_mds; then
5392 start_mdt 1 "-o nosvc" ||
5393 error "starting mds with nosvc option failed"
5396 echo "command should accept two parameters"
5397 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
5398 error "command should accept two params"
5400 echo "correct device name should be passed"
5401 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
5402 error "wrong devname"
5404 echo "wrong nids list should not destroy the system"
5405 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
5407 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "asdfasdf, asdfadf" &&
5410 echo "replace OST nid"
5411 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
5412 error "replace nids failed"
5414 echo "command should accept two parameters"
5415 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
5416 error "command should accept two params"
5418 echo "wrong nids list should not destroy the system"
5419 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
5422 local FAKE_NIDS="192.168.0.112@tcp1,192.168.0.112@tcp2"
5423 local FAKE_FAILOVER="192.168.0.113@tcp1,192.168.0.113@tcp2"
5424 local NIDS_AND_FAILOVER="$MDS_NID,$FAKE_NIDS:$FAKE_FAILOVER"
5425 echo "set NIDs with failover"
5426 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $NIDS_AND_FAILOVER ||
5427 error "replace nids failed"
5430 echo "replace MDS nid"
5431 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
5432 error "replace nids failed"
5434 if ! combined_mgs_mds ; then
5437 stop_mds || error "Unable to stop MDS"
5440 start_mgsmds || error "start mgsmds failed"
5441 set_conf_param_and_check mds \
5442 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
5443 "$FSNAME-OST0000.osc.active" \
5445 start_ost || error "unable to start OST"
5446 mount_client $MOUNT || error "mount client failed"
5448 check_mount || error "error after nid replace"
5449 cleanup || error "cleanup failed"
5452 run_test 66 "replace nids"
5454 test_67() { #LU-2950
5455 local legacy="$TMP/legacy_lnet_config"
5456 local new="$TMP/new_routes_test"
5457 local out="$TMP/config_out_file"
5458 local verify="$TMP/conv_verify"
5459 local verify_conf="$TMP/conf_verify"
5461 # Create the legacy file that will be run through the
5462 # lustre_routes_conversion script
5463 cat <<- LEGACY_LNET_CONFIG > $legacy
5464 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
5465 tcp2 54 10.1.3.2@tcp;
5466 tcp3 10.3.4.3@tcp:3;
5470 # Create the verification file to verify the output of
5471 # lustre_routes_conversion script against.
5472 cat <<- VERIFY_LNET_CONFIG > $verify
5473 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
5474 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
5475 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
5476 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
5477 tcp4: { gateway: 10.3.3.4@tcp }
5480 # Create the verification file to verify the output of
5481 # lustre_routes_config script against
5482 cat <<- VERIFY_LNET_CONFIG > $verify_conf
5483 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
5484 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
5485 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
5486 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
5487 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
5490 $LUSTRE_ROUTES_CONVERSION $legacy $new > /dev/null
5491 if [ -f $new ]; then
5492 # verify the conversion output
5493 cmp -s $new $verify > /dev/null
5494 if [ $? -eq 1 ]; then
5495 error "routes conversion failed"
5498 lustre_routes_config --dry-run --verbose $new > $out
5499 # check that the script succeeded
5500 cmp -s $out $verify_conf > /dev/null
5501 if [ $? -eq 1 ]; then
5502 error "routes config failed"
5505 error "routes conversion test failed"
5507 # remove generated files
5508 rm -f $new $legacy $verify $verify_conf $out
5510 run_test 67 "test routes conversion and configuration"
5518 [ "$MDS1_VERSION" -ge $(version_code 2.4.53) ] ||
5519 skip "Need MDS version at least 2.4.53"
5521 umount_client $MOUNT || error "umount client failed"
5526 # START-END - the sequences we'll be reserving
5527 START=$(do_facet $SINGLEMDS \
5528 $LCTL get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
5529 END=$((START + (1 << 30)))
5530 do_facet $SINGLEMDS \
5531 $LCTL set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
5533 # reset the sequences MDT0000 has already assigned
5534 do_facet $SINGLEMDS \
5535 $LCTL set_param seq.srv*MDT0000.space=clear
5537 # remount to let the client allocate new sequence
5538 mount_client $MOUNT || error "mount client failed"
5540 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
5541 do_facet $SINGLEMDS \
5542 $LCTL get_param seq.srv*MDT0000.space
5543 $LFS path2fid $DIR/$tfile
5545 local old_ifs="$IFS"
5547 fid=($($LFS path2fid $DIR/$tfile))
5551 if [[ $seq < $END ]]; then
5552 error "used reserved sequence $seq?"
5554 cleanup || error "cleanup failed with $?"
5556 run_test 68 "be able to reserve specific sequences in FLDB"
5558 # Test 69: is about the total number of objects ever created on an OST.
5559 # so that when it is reformatted the normal MDS->OST orphan recovery won't
5560 # just "precreate" the missing objects. In the past it might try to recreate
5561 # millions of objects after an OST was reformatted
5563 [[ "$MDS1_VERSION" -lt $(version_code 2.4.2) ]] &&
5564 skip "Need MDS version at least 2.4.2"
5566 [[ "$MDS1_VERSION" -ge $(version_code 2.4.50) ]] &&
5567 [[ "$MDS1_VERSION" -lt $(version_code 2.5.0) ]] &&
5568 skip "Need MDS version at least 2.5.0"
5571 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5572 do_nodes $(comma_list $(osts_nodes)) $LCTL set_param \
5573 seq.*OST*-super.width=$DATA_SEQ_MAX_WIDTH
5575 # use OST0000 since it probably has the most creations
5576 local OSTNAME=$(ostname_from_index 0)
5577 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
5578 local last_id=$(do_facet mds1 $LCTL get_param -n \
5579 osp.$mdtosc_proc1.prealloc_last_id)
5581 # Want to have OST LAST_ID over 5 * OST_MAX_PRECREATE to
5582 # verify that the LAST_ID recovery is working properly. If
5583 # not, then the OST will refuse to allow the MDS connect
5584 # because the LAST_ID value is too different from the MDS
5585 #define OST_MAX_PRECREATE=20000
5586 local ost_max_pre=20000
5587 local num_create=$(( ost_max_pre * 5 + 1 - last_id))
5589 # If the LAST_ID is already over 5 * OST_MAX_PRECREATE, we don't
5590 # need to create any files. So, skip this section.
5591 if [ $num_create -gt 0 ]; then
5592 # Check the number of inodes available on OST0
5594 local ifree=$($LFS df -i $MOUNT |
5595 awk '/OST0000/ { print $4 }'; exit ${PIPESTATUS[0]})
5596 log "On OST0, $ifree inodes available. Want $num_create. rc=$?"
5598 $LFS setstripe -i 0 $DIR/$tdir ||
5599 error "$LFS setstripe -i 0 $DIR/$tdir failed"
5600 if [ $ifree -lt 10000 ]; then
5601 files=$(( ifree - 50 ))
5606 local j=$((num_create / files + 1))
5607 for i in $(seq 1 $j); do
5608 createmany -o $DIR/$tdir/$tfile-$i- $files ||
5609 error "createmany fail create $files files: $?"
5610 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
5611 error "unlinkmany failed unlink $files files"
5615 # delete all of the files with objects on OST0 so the
5616 # filesystem is not inconsistent later on
5617 $LFS find $MOUNT --ost 0 -print0 | xargs -0 rm
5619 umount_client $MOUNT || error "umount client failed"
5620 stop_ost || error "OST0 stop failure"
5621 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
5622 $(ostdevname 1) $(ostvdevname 1) ||
5623 error "reformat and replace $ostdev failed"
5624 start_ost || error "OST0 restart failure"
5625 wait_osc_import_state mds ost FULL
5627 mount_client $MOUNT || error "mount client failed"
5628 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
5629 local idx=$($LFS getstripe -i $DIR/$tdir/$tfile-last)
5630 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
5632 local iused=$($LFS df -i $MOUNT |
5633 awk '/OST0000/ { print $3 }'; exit ${PIPESTATUS[0]})
5634 log "On OST0, $iused used inodes rc=$?"
5635 [ $iused -ge $((ost_max_pre + 1000)) ] &&
5636 error "OST replacement created too many inodes; $iused"
5637 cleanup || error "cleanup failed with $?"
5639 run_test 69 "replace an OST with the same index"
5642 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5645 cleanup || error "cleanup failed with $?"
5647 start_mdt 1 || error "MDT0 start fail"
5649 start_ost || error "OST0 start fail"
5650 for num in $(seq 2 $MDSCOUNT); do
5651 start_mdt $num || return
5654 mount_client $MOUNT || error "mount client fails"
5656 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
5658 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5659 error "create remote dir fail"
5661 rm -rf $DIR/$tdir || error "delete dir fail"
5662 cleanup || error "cleanup failed with $?"
5664 run_test 70a "start MDT0, then OST, then MDT1"
5667 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5670 start_ost || error "OST0 start fail"
5672 start_mds || error "MDS start fail"
5674 mount_client $MOUNT || error "mount client fails"
5676 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
5678 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5679 error "create remote dir fail"
5681 rm -rf $DIR/$tdir || error "delete dir fail"
5683 cleanup || error "cleanup failed with $?"
5685 run_test 70b "start OST, MDT1, MDT0"
5688 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5691 start_mds || error "MDS start fail"
5692 start_ost || error "OST0 start fail"
5694 mount_client $MOUNT || error "mount client fails"
5695 stop_mdt 1 || error "MDT1 start fail"
5697 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
5698 echo "deactivate $mdc_for_mdt1"
5699 $LCTL --device $mdc_for_mdt1 deactivate ||
5700 error "set $mdc_for_mdt1 deactivate failed"
5702 mkdir $DIR/$tdir && error "mkdir succeed"
5704 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
5705 error "create remote dir succeed"
5707 cleanup || error "cleanup failed with $?"
5709 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
5712 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5715 start_mds || error "MDS start fail"
5716 start_ost || error "OST0 start fail"
5718 mount_client $MOUNT || error "mount client fails"
5720 stop_mdt 2 || error "MDT1 start fail"
5722 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
5724 echo "deactivate $mdc_for_mdt2"
5725 $LCTL --device $mdc_for_mdt2 deactivate ||
5726 error "set $mdc_for_mdt2 deactivate failed"
5728 mkdir $DIR/$tdir || error "mkdir fail"
5729 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
5730 error "create remote dir succeed"
5732 rm -rf $DIR/$tdir || error "delete dir fail"
5734 cleanup || error "cleanup failed with $?"
5736 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
5739 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5741 [ "$MDS1_VERSION" -ge $(version_code 2.7.62) ] ||
5742 skip "Need MDS version at least 2.7.62"
5744 reformat || error "reformat failed with $?"
5748 local mdsdev=$(mdsdevname 1)
5749 local ostdev=$(ostdevname 1)
5750 local mdsvdev=$(mdsvdevname 1)
5751 local ostvdev=$(ostvdevname 1)
5752 local opts_mds="$(mkfs_opts mds1 $mdsdev) --reformat $mdsdev $mdsvdev"
5753 local opts_ost="$(mkfs_opts ost1 $ostdev) --reformat $ostdev $ostvdev"
5755 if ! combined_mgs_mds ; then
5759 add mds1 $opts_mds || error "add mds1 failed"
5760 start_mdt 1 || error "start mdt1 failed"
5761 add ost1 $opts_ost || error "add ost1 failed"
5762 start_ost || error "start ost failed"
5763 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
5765 local soc=$(do_facet mds1 "$LCTL get_param -n \
5766 mdt.*MDT0000.sync_lock_cancel")
5767 [ $soc == "never" ] || error "SoC enabled on single MDS"
5769 for i in $(seq 2 $MDSCOUNT); do
5770 mdsdev=$(mdsdevname $i)
5771 mdsvdev=$(mdsvdevname $i)
5772 opts_mds="$(mkfs_opts mds$i $mdsdev) --reformat $mdsdev \
5774 add mds$i $opts_mds || error "add mds$i failed"
5775 start_mdt $i || error "start mdt$i fail"
5778 wait_dne_interconnect
5780 for i in $(seq $MDSCOUNT); do
5781 soc=$(do_facet mds$i "$LCTL get_param -n \
5782 mdt.*MDT000$((i - 1)).sync_lock_cancel")
5783 [ $soc == "blocking" ] || error "SoC not enabled on DNE"
5786 for i in $(seq 2 $MDSCOUNT); do
5787 stop_mdt $i || error "stop mdt$i fail"
5789 soc=$(do_facet mds1 "$LCTL get_param -n \
5790 mdt.*MDT0000.sync_lock_cancel")
5791 [ $soc == "never" ] || error "SoC enabled on single MDS"
5792 umount_client $MOUNT -f > /dev/null
5794 cleanup || error "cleanup failed with $?"
5796 run_test 70e "Sync-on-Cancel will be enabled by default on DNE"
5799 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5800 if combined_mgs_mds; then
5801 skip "needs separate MGS/MDT"
5805 start_mdt 1 || error "MDT0 start fail"
5806 start_ost || error "OST0 start fail"
5807 for num in $(seq 2 $MDSCOUNT); do
5808 start_mdt $num || return
5811 start_ost2 || error "OST1 start fail"
5813 mount_client $MOUNT || error "mount client fails"
5815 mkdir $DIR/$tdir || error "mkdir fail"
5816 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5817 error "create remote dir succeed"
5819 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5820 rm -rf $DIR/$tdir || error "delete dir fail"
5822 umount_client $MOUNT || error "umount_client failed"
5823 stop_mds || error "MDS stop fail"
5824 stop_ost || error "OST0 stop fail"
5825 stop_ost2 || error "OST1 stop fail"
5827 run_test 71a "start MDT0 OST0, MDT1, OST1"
5830 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5831 if combined_mgs_mds; then
5832 skip "needs separate MGS/MDT"
5836 for num in $(seq 2 $MDSCOUNT); do
5837 start_mdt $num || return
5839 start_ost || error "OST0 start fail"
5840 start_mdt 1 || error "MDT0 start fail"
5841 start_ost2 || error "OST1 start fail"
5843 mount_client $MOUNT || error "mount client fails"
5845 mkdir $DIR/$tdir || error "mkdir fail"
5846 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5847 error "create remote dir succeed"
5849 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5850 rm -rf $DIR/$tdir || error "delete dir fail"
5852 umount_client $MOUNT || error "umount_client failed"
5853 stop_mds || error "MDT0 stop fail"
5854 stop_ost || error "OST0 stop fail"
5855 stop_ost2 || error "OST1 stop fail"
5857 run_test 71b "start MDT1, OST0, MDT0, OST1"
5860 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5861 combined_mgs_mds && skip "needs separate MGS/MDT"
5865 start_ost || error "OST0 start fail"
5866 start_ost2 || error "OST1 start fail"
5867 for num in $(seq 2 $MDSCOUNT); do
5868 start_mdt $num || return
5870 start_mdt 1 || error "MDT0 start fail"
5872 mount_client $MOUNT || error "mount client fails"
5874 mkdir $DIR/$tdir || error "mkdir fail"
5875 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5876 error "create remote dir succeed"
5878 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5879 rm -rf $DIR/$tdir || error "delete dir fail"
5881 umount_client $MOUNT || error "umount_client failed"
5882 stop_mds || error "MDS stop fail"
5883 stop_ost || error "OST0 stop fail"
5884 stop_ost2 || error "OST1 stop fail"
5887 run_test 71c "start OST0, OST1, MDT1, MDT0"
5890 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5891 combined_mgs_mds && skip "needs separate MGS/MDT"
5895 start_ost || error "OST0 start fail"
5896 for num in $(seq 2 $MDSCOUNT); do
5897 start_mdt $num || return
5899 start_mdt 1 || error "MDT0 start fail"
5900 start_ost2 || error "OST1 start fail"
5902 mount_client $MOUNT || error "mount client fails"
5904 mkdir $DIR/$tdir || error "mkdir fail"
5905 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5906 error "create remote dir succeed"
5908 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5909 rm -rf $DIR/$tdir || error "delete dir fail"
5911 umount_client $MOUNT || error "umount_client failed"
5912 stop_mds || error "MDS stop fail"
5913 stop_ost || error "OST0 stop fail"
5914 stop_ost2 || error "OST1 stop fail"
5917 run_test 71d "start OST0, MDT1, MDT0, OST1"
5920 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5921 combined_mgs_mds && skip "needs separate MGS/MDT"
5925 start_ost || error "OST0 start fail"
5926 for num in $(seq 2 $MDSCOUNT); do
5927 start_mdt $num || return
5929 start_ost2 || error "OST1 start fail"
5930 start_mdt 1 || error "MDT0 start fail"
5932 mount_client $MOUNT || error "mount client fails"
5934 mkdir $DIR/$tdir || error "mkdir fail"
5935 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5936 error "create remote dir succeed"
5938 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5939 rm -rf $DIR/$tdir || error "delete dir fail"
5941 umount_client $MOUNT || error "umount_client failed"
5942 stop_mds || error "MDS stop fail"
5943 stop_ost || error "OST0 stop fail"
5944 stop_ost2 || error "OST1 stop fail"
5947 run_test 71e "start OST0, MDT1, OST1, MDT0"
5949 test_72() { #LU-2634
5950 [ "$mds1_FSTYPE" != ldiskfs ] &&
5951 skip "ldiskfs only test"
5953 local mdsdev=$(mdsdevname 1)
5954 local ostdev=$(ostdevname 1)
5955 local cmd="$E2FSCK -fnvd $mdsdev"
5962 if combined_mgs_mds; then
5963 add_options='--reformat'
5965 add_options='--reformat --replace'
5968 #tune MDT with "-O extents"
5970 for num in $(seq $MDSCOUNT); do
5971 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
5972 $add_options $(mdsdevname $num) $(mdsvdevname $num) ||
5973 error "add mds $num failed"
5974 do_facet mds${num} "$TUNE2FS -O extents $(mdsdevname $num)" ||
5975 error "$TUNE2FS failed on mds${num}"
5978 add ost1 $(mkfs_opts ost1 $ostdev) $add_options $ostdev ||
5979 error "add $ostdev failed"
5980 start_mds || error "start mds failed"
5981 start_ost || error "start ost failed"
5982 mount_client $MOUNT || error "mount client failed"
5984 #create some short symlinks
5985 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5986 createmany -o $DIR/$tdir/$tfile-%d $fn
5987 echo "create $fn short symlinks"
5988 for i in $(seq -w 1 $fn); do
5989 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
5994 umount_client $MOUNT || error "umount client failed"
5995 stop_mds || error "stop mds failed"
5996 stop_ost || error "stop ost failed"
5999 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
6001 run_test 72 "test fast symlink with extents flag enabled"
6003 test_73() { #LU-3006
6004 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
6005 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
6006 error "1st tunefs failed"
6007 start_mgsmds || error "start mds failed"
6008 start_ost || error "start ost failed"
6009 mount_client $MOUNT || error "mount client failed"
6010 $LCTL get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
6011 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
6012 umount_client $MOUNT || error "umount client failed"
6016 run_test 73 "failnode to update from mountdata properly"
6018 test_75() { # LU-2374
6019 [[ "$MDS1_VERSION" -lt $(version_code 2.4.1) ]] &&
6020 skip "Need MDS version at least 2.4.1"
6023 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
6024 --replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
6025 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
6026 --replace --reformat $(ostdevname 1) $(ostvdevname 1)"
6028 #check with default parameters
6029 add mds1 $opts_mds || error "add mds1 failed for default params"
6030 add ost1 $opts_ost || error "add ost1 failed for default params"
6032 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
6033 opts_mds=$(echo $opts_mds |
6034 sed -e "s/--index=$index/--index=$index --mdt/")
6035 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
6036 opts_ost=$(echo $opts_ost |
6037 sed -e "s/--index=$index/--index=$index --ost/")
6039 add mds1 $opts_mds || error "add mds1 failed for new params"
6040 add ost1 $opts_ost || error "add ost1 failed for new params"
6045 run_test 75 "The order of --index should be irrelevant"
6048 [[ "$MGS_VERSION" -ge $(version_code 2.4.52) ]] ||
6049 skip "Need MDS version at least 2.4.52"
6052 local MDMB_PARAM="osc.*.max_dirty_mb"
6053 echo "Change MGS params"
6054 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
6056 echo "max_dirty_mb: $MAX_DIRTY_MB"
6057 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB - 10))
6058 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
6059 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
6060 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
6061 head -1" $NEW_MAX_DIRTY_MB
6062 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
6063 echo "$MAX_DIRTY_MB"
6064 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
6065 error "error while apply max_dirty_mb"
6067 echo "Check the value is stored after remount"
6070 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
6071 head -1" $NEW_MAX_DIRTY_MB
6072 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
6073 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
6074 error "max_dirty_mb is not saved after remount"
6076 echo "Change OST params"
6077 CLIENT_PARAM="obdfilter.*.client_cache_count"
6078 local CLIENT_CACHE_COUNT
6079 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
6081 echo "client_cache_count: $CLIENT_CACHE_COUNT"
6082 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
6083 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
6084 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
6085 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
6086 head -1" $NEW_CLIENT_CACHE_COUNT
6087 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
6089 echo "$CLIENT_CACHE_COUNT"
6090 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
6091 error "error while apply client_cache_count"
6093 echo "Check the value is stored after remount"
6096 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
6097 head -1" $NEW_CLIENT_CACHE_COUNT
6098 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
6100 echo "$CLIENT_CACHE_COUNT"
6101 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
6102 error "client_cache_count is not saved after remount"
6105 run_test 76a "set permanent params with lctl across mounts"
6107 test_76b() { # LU-4783
6108 [[ "$MGS_VERSION" -ge $(version_code 2.5.57) ]] ||
6109 skip "Need MGS version at least 2.5.57"
6112 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
6113 error "start params log failed"
6116 run_test 76b "verify params log setup correctly"
6119 [[ "$MGS_VERSION" -ge $(version_code 2.8.54) ]] ||
6120 skip "Need MDS version at least 2.4.52"
6122 local MASK_PARAM="mdd.*.changelog_mask"
6123 echo "Change changelog_mask"
6124 do_facet mgs $LCTL set_param -P $MASK_PARAM=-CLOSE ||
6125 error "Can't change changlog_mask"
6126 wait_update $(facet_host mds) "$LCTL get_param -n $MASK_PARAM |
6129 echo "Check the value is stored after mds remount"
6130 stop_mds || error "Failed to stop MDS"
6131 start_mds || error "Failed to start MDS"
6132 local CHANGELOG_MASK=$(do_facet mgs $LCTL get_param -n $MASK_PARAM)
6133 echo $CHANGELOG_MASK | grep CLOSE > /dev/null &&
6134 error "changelog_mask is not changed"
6138 run_test 76c "verify changelog_mask is applied with lctl set_param -P"
6140 test_76d() { #LU-9399
6143 local xattr_cache="llite.*.xattr_cache"
6144 local cmd="$LCTL get_param -n $xattr_cache | head -1"
6145 local new=$((($(eval $cmd) + 1) % 2))
6147 echo "lctl set_param -P llite.*.xattr_cache=$new"
6148 do_facet mgs $LCTL set_param -P $xattr_cache=$new ||
6149 error "Can't change xattr_cache"
6150 wait_update $HOSTNAME "$cmd" "$new"
6152 echo "Check $xattr_cache on client $MOUNT"
6153 umount_client $MOUNT || error "umount $MOUNT failed"
6154 mount_client $MOUNT || error "mount $MOUNT failed"
6155 [ $(eval $cmd) -eq $new ] ||
6156 error "$xattr_cache != $new on client $MOUNT"
6158 echo "Check $xattr_cache on the new client $MOUNT2"
6159 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
6160 [ $(eval $cmd) -eq $new ] ||
6161 error "$xattr_cache != $new on client $MOUNT2"
6162 umount_client $MOUNT2 || error "umount $MOUNT2 failed"
6166 run_test 76d "verify llite.*.xattr_cache can be set by 'lctl set_param -P' correctly"
6168 test_77() { # LU-3445
6169 [[ "$MDS1_VERSION" -ge $(version_code 2.8.55) ]] ||
6170 skip "Need MDS version 2.8.55+ "
6172 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
6173 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
6174 skip_env "mixed loopback and real device not working"
6177 local fs2mdsdev=$(mdsdevname 1_2)
6178 local fs2ostdev=$(ostdevname 1_2)
6179 local fs2mdsvdev=$(mdsvdevname 1_2)
6180 local fs2ostvdev=$(ostvdevname 1_2)
6181 local fsname=test1234
6183 local failnid="$(h2nettype 1.2.3.4),$(h2nettype 4.3.2.1)"
6185 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
6187 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
6188 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
6189 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
6190 error "start fs2mds failed"
6192 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
6193 mgsnid="0.0.0.0@tcp,$mgsnid,$mgsnid:$mgsnid"
6195 add fs2ost --mgsnode=$mgsnid $(mkfs_opts ost1 $fs2ostdev) \
6196 --failnode=$failnid --fsname=$fsname \
6197 --reformat $fs2ostdev $fs2ostvdev ||
6198 error "add fs2ost failed"
6199 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
6201 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
6202 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
6203 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
6206 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
6209 [[ "$mds1_FSTYPE" != ldiskfs ||
6210 "$ost1_FSTYPE" != ldiskfs ]] &&
6211 skip "ldiskfs only test"
6213 # reformat the Lustre filesystem with a smaller size
6214 local saved_MDSCOUNT=$MDSCOUNT
6215 local saved_MDSSIZE=$MDSSIZE
6216 local saved_OSTCOUNT=$OSTCOUNT
6217 local saved_OSTSIZE=$OSTSIZE
6220 MDSSIZE=$((MDSSIZE - 20000))
6221 OSTSIZE=$((OSTSIZE - 20000))
6222 reformat || error "(1) reformat Lustre filesystem failed"
6223 MDSSIZE=$saved_MDSSIZE
6224 OSTSIZE=$saved_OSTSIZE
6226 # mount the Lustre filesystem
6227 setup_noconfig || error "(2) setup Lustre filesystem failed"
6230 log "create test files"
6235 mkdir $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
6237 for i in $(seq $num_files); do
6238 file=$MOUNT/$tdir/$tfile-$i
6239 dd if=/dev/urandom of=$file count=1 bs=1M || {
6240 $LCTL get_param osc.*.cur*grant*
6241 $LFS df; $LFS df -i;
6242 # stop creating files if there is no more space
6243 if [ ! -e $file ]; then
6244 num_files=$((i - 1))
6248 $LFS getstripe -v $file
6249 local ost_idx=$(LFS getstripe -i $file)
6250 do_facet ost$((ost_idx + 1)) \
6251 $LCTL get_param obdfilter.*.*grant*
6252 error "(4) create $file failed"
6256 # unmount the Lustre filesystem
6257 cleanup || error "(5) cleanup Lustre filesystem failed"
6259 # run e2fsck on the MDT and OST devices
6260 local mds_host=$(facet_active_host $SINGLEMDS)
6261 local ost_host=$(facet_active_host ost1)
6262 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
6263 local ost_dev=$(ostdevname 1)
6265 run_e2fsck $mds_host $mds_dev "-y"
6266 run_e2fsck $ost_host $ost_dev "-y"
6268 # get the original block count of the MDT and OST filesystems
6269 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
6270 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
6272 # expand the MDT and OST filesystems to the device size
6273 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
6274 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
6276 # run e2fsck on the MDT and OST devices again
6277 run_e2fsck $mds_host $mds_dev "-y"
6278 run_e2fsck $ost_host $ost_dev "-y"
6280 # mount the Lustre filesystem
6284 log "check files after expanding the MDT and OST filesystems"
6285 for i in $(seq $num_files); do
6286 file=$MOUNT/$tdir/$tfile-$i
6287 $CHECKSTAT -t file -s 1048576 $file ||
6288 error "(6) checkstat $file failed"
6292 log "create more files after expanding the MDT and OST filesystems"
6293 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
6294 file=$MOUNT/$tdir/$tfile-$i
6295 dd if=/dev/urandom of=$file count=1 bs=1M ||
6296 error "(7) create $file failed"
6299 # unmount the Lustre filesystem
6300 cleanup || error "(8) cleanup Lustre filesystem failed"
6302 # run e2fsck on the MDT and OST devices
6303 run_e2fsck $mds_host $mds_dev "-y"
6304 run_e2fsck $ost_host $ost_dev "-y"
6306 # get the maximum block count of the MDT and OST filesystems
6307 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
6308 local ost_max_blks=$(get_block_count ost1 $ost_dev)
6310 # get the minimum block count of the MDT and OST filesystems
6311 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
6312 grep minimum | sed -e 's/^.*filesystem: //g')
6313 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
6314 grep minimum | sed -e 's/^.*filesystem: //g')
6316 # shrink the MDT and OST filesystems to a smaller size
6320 if [[ $mds_max_blks -gt $mds_min_blks &&
6321 $mds_max_blks -gt $mds_orig_blks ]]; then
6322 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
6323 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
6324 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
6325 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
6326 error "shrink $SINGLEMDS to $new_blks failed"
6330 if [[ $ost_max_blks -gt $ost_min_blks &&
6331 $ost_max_blks -gt $ost_orig_blks ]]; then
6332 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
6333 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
6334 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
6335 run_resize2fs ost1 $ost_dev $new_blks ||
6336 error "shrink ost1 to $new_blks failed"
6340 # check whether the MDT or OST filesystem was shrunk or not
6342 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
6343 reformat_and_config ||
6344 error "(10) reformat Lustre filesystem failed"
6348 # run e2fsck on the MDT and OST devices again
6349 run_e2fsck $mds_host $mds_dev "-y"
6350 run_e2fsck $ost_host $ost_dev "-y"
6352 # mount the Lustre filesystem again
6356 log "check files after shrinking the MDT and OST filesystems"
6357 for i in $(seq $((num_files + 10))); do
6358 file=$MOUNT/$tdir/$tfile-$i
6359 $CHECKSTAT -t file -s 1048576 $file ||
6360 error "(11) checkstat $file failed"
6363 # unmount and reformat the Lustre filesystem
6364 cleanup || error "(12) cleanup Lustre filesystem failed"
6365 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
6367 MDSCOUNT=$saved_MDSCOUNT
6368 OSTCOUNT=$saved_OSTCOUNT
6369 reformat_and_config || error "(14) reformat Lustre filesystem failed"
6371 run_test 78 "run resize2fs on MDT and OST filesystems"
6373 test_79() { # LU-4227
6374 [[ "$MDS1_VERSION" -ge $(version_code 2.5.59) ]] ||
6375 skip "Need MDS version at least 2.5.59"
6377 local mdsdev1=$(mdsdevname 1)
6378 local mdsvdev1=$(mdsvdevname 1)
6379 local mdsdev2=$(mdsdevname 2)
6380 local mdsvdev2=$(mdsvdevname 2)
6381 local ostdev1=$(ostdevname 1)
6382 local ostvdev1=$(ostvdevname 1)
6383 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
6384 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
6385 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
6388 # remove --mgs/--mgsnode from mkfs.lustre options
6389 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
6391 mgsnode_opt=$(echo $opts_mds2 |
6392 awk '{ for ( i = 1; i < NF; i++ )
6393 if ( $i ~ "--mgsnode" ) { print $i; break } }')
6394 [ -n "$mgsnode_opt" ] &&
6395 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
6397 mgsnode_opt=$(echo $opts_ost1 |
6398 awk '{ for ( i = 1; i < NF; i++ )
6399 if ( $i ~ "--mgsnode" ) { print $i; break } }')
6400 [ -n "$mgsnode_opt" ] &&
6401 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
6403 # -MGS, format a mdt without --mgs option
6404 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
6405 error "Must specify --mgs when formatting mdt combined with mgs"
6407 # +MGS, format a mdt/ost without --mgsnode option
6408 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
6409 > /dev/null || error "start mds1 failed"
6410 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
6411 error "Must specify --mgsnode when formatting a mdt"
6412 add ost1 $opts_ost1 $ostdev1 $ostvdev2 &&
6413 error "Must specify --mgsnode when formatting an ost"
6417 run_test 79 "format MDT/OST without mgs option (should return errors)"
6420 start_mds || error "Failed to start MDT"
6421 start_ost || error "Failed to start OST1"
6422 uuid=$(do_facet ost1 $LCTL get_param -n mgc.*.uuid)
6423 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
6424 do_facet ost1 "$LCTL set_param fail_val=10 fail_loc=0x906"
6425 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x906"
6426 do_facet mgs "$LCTL set_param -n mgs/MGS/evict_client $uuid"
6428 start_ost2 || error "Failed to start OST2"
6430 do_facet ost1 "$LCTL set_param fail_loc=0"
6435 run_test 80 "mgc import reconnect race"
6437 #Save the original values of $OSTCOUNT and $OSTINDEX$i.
6439 local new_ostcount=$1
6440 saved_ostcount=$OSTCOUNT
6441 OSTCOUNT=$new_ostcount
6445 for i in $(seq $OSTCOUNT); do
6447 eval saved_ostindex$i=${!index}
6452 # Restore the original values of $OSTCOUNT and $OSTINDEX$i.
6453 restore_ostindex() {
6458 for i in $(seq $OSTCOUNT); do
6459 index=saved_ostindex$i
6460 eval OSTINDEX$i=${!index}
6462 OSTCOUNT=$saved_ostcount
6465 if ! combined_mgs_mds ; then
6470 # The main purpose of this test is to ensure the OST_INDEX_LIST functions as
6471 # expected. This test uses OST_INDEX_LIST to format OSTs with a randomly
6472 # assigned index and ensures we can mount such a formatted file system
6473 test_81() { # LU-4665
6474 [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
6475 skip "Need MDS version at least 2.6.54"
6476 [[ $OSTCOUNT -ge 3 ]] || skip_env "needs >= 3 OSTs"
6480 # Each time RANDOM is referenced, a random integer between 0 and 32767
6483 local saved_ostindex1=$OSTINDEX1
6484 for i in 65535 $((RANDOM + 65536)); do
6485 echo -e "\nFormat ost1 with --index=$i, should fail"
6487 if add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat \
6488 $(ostdevname 1) $(ostvdevname 1); then
6489 OSTINDEX1=$saved_ostindex1
6490 error "format ost1 with --index=$i should fail"
6493 OSTINDEX1=$saved_ostindex1
6497 # Format OSTs with random sparse indices.
6498 trap "restore_ostindex" EXIT
6499 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices"
6500 OST_INDEX_LIST=[0,$((RANDOM * 2 % 65533 + 1)),65534] formatall
6502 # Setup and check Lustre filesystem.
6503 start_mgsmds || error "start_mgsmds failed"
6504 for i in $(seq $OSTCOUNT); do
6505 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6506 error "start ost$i failed"
6509 mount_client $MOUNT || error "mount client $MOUNT failed"
6510 check_mount || error "check client $MOUNT failed"
6513 local max_easize=$($LCTL get_param -n llite.*.max_easize)
6514 # 65452 is XATTR_SIZE_MAX less ldiskfs ea overhead
6515 if large_xattr_enabled; then
6516 [[ $max_easize -ge 65452 ]] ||
6517 error "max_easize is $max_easize, should be at least 65452 bytes"
6520 # 4012 is 4096 less ldiskfs ea overhead
6521 [[ $max_easize -ge 4012 ]] ||
6522 error "max_easize is $max_easize, should be at least 4012 bytes"
6527 run_test 81 "sparse OST indexing"
6529 # Here we exercise the stripe placement functionality on a file system that
6530 # has formatted the OST with a random index. With the file system the following
6531 # functionality is tested:
6533 # 1. Creating a new file with a specific stripe layout.
6535 # 2. Modifiy a existing empty file with a specific stripe layout.
6537 # 3. Ensure we fail to set the stripe layout of a file that already has one.
6539 # 4. If ost-index is defined we need to ensure it is the first entry in the
6540 # ost index list returned by lfs getstripe.
6542 # 5. Lastly ensure this functionality fails with directories.
6543 test_82a() { # LU-4665
6544 [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
6545 skip "Need MDS version at least 2.6.54"
6546 [[ $OSTCOUNT -ge 3 ]] || skip_env "needs >= 3 OSTs"
6552 # Format OSTs with random sparse indices.
6556 local LOV_V1_INSANE_STRIPE_COUNT=65532
6557 for i in $(seq $OSTCOUNT); do
6558 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
6559 ost_indices+=" $index"
6561 ost_indices=$(comma_list $ost_indices)
6563 stack_trap "restore_ostindex" EXIT
6564 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
6565 OST_INDEX_LIST=[$ost_indices] formatall
6567 # Setup Lustre filesystem.
6568 start_mgsmds || error "start_mgsmds failed"
6569 for i in $(seq $OSTCOUNT); do
6570 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6571 error "start ost$i failed"
6574 # Collect debug information - start of test
6575 do_nodes $(comma_list $(mdts_nodes)) \
6576 $LCTL get_param osp.*.prealloc_*_id
6578 mount_client $MOUNT || error "mount client $MOUNT failed"
6582 check_lfs_df_ret_val $? || error "$LFS df $MOUNT failed"
6583 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6585 stack_trap "do_nodes $(comma_list $(mdts_nodes)) \
6586 $LCTL get_param osp.*.prealloc_*_id" EXIT
6588 # 1. If the file does not exist, new file will be created
6589 # with specified OSTs.
6590 local file=$DIR/$tdir/$tfile-1
6591 local cmd="$LFS setstripe -o $ost_indices $file"
6593 eval $cmd || error "$cmd failed"
6594 check_stripe_count $file $OSTCOUNT
6595 check_obdidx $file $ost_indices
6596 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6597 error "write $file failed"
6599 # 2. If the file already exists and is an empty file, the file
6600 # will be attached with specified layout.
6601 file=$DIR/$tdir/$tfile-2
6602 mcreate $file || error "mcreate $file failed"
6603 cmd="$LFS setstripe -o $ost_indices $file"
6605 eval $cmd || error "$cmd failed"
6606 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6607 error "write $file failed"
6608 check_stripe_count $file $OSTCOUNT
6609 check_obdidx $file $ost_indices
6611 # 3. If the file already has a valid layout attached, the command
6612 # should fail with EBUSY.
6614 eval $cmd && error "stripe is already set on $file, $cmd should fail"
6616 # 4. If [--stripe-index|-i <start_ost_idx>] is used, the index must
6617 # be in the OST indices list.
6618 local start_ost_idx=${ost_indices##*,}
6619 file=$DIR/$tdir/$tfile-3
6620 cmd="$LFS setstripe -o $ost_indices -i $start_ost_idx $file"
6622 eval $cmd || error "$cmd failed"
6623 check_stripe_count $file $OSTCOUNT
6624 check_obdidx $file $ost_indices
6625 check_start_ost_idx $file $start_ost_idx
6627 file=$DIR/$tdir/$tfile-4
6628 cmd="$LFS setstripe"
6629 cmd+=" -o $(exclude_items_from_list $ost_indices $start_ost_idx)"
6630 cmd+=" -i $start_ost_idx $file"
6632 eval $cmd && error "index $start_ost_idx should be in $ost_indices"
6634 # 5. Specifying OST indices for directory should succeed.
6635 local dir=$DIR/$tdir/$tdir
6636 mkdir $dir || error "mkdir $dir failed"
6637 cmd="$LFS setstripe -o $ost_indices $dir"
6638 if [ "$MDS1_VERSION" -gt $(version_code 2.11.53) ] &&
6639 [ "$CLIENT_VERSION" -gt $(version_code 2.11.53) ]; then
6641 eval $cmd || error "unable to specify OST indices on directory"
6643 echo "need MDS+client version at least 2.11.53"
6648 run_test 82a "specify OSTs for file (succeed) or directory (succeed)"
6653 # Remove OSTs from a pool and destroy the pool.
6654 destroy_pool $ost_pool || true
6659 # Test 82b is run to ensure that if the user supplies a pool with a specific
6660 # stripe layout that it behaves proprerly. It should fail in the case that
6661 # the supplied OST index list points to OSTs not contained in the user
6663 test_82b() { # LU-4665
6664 [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
6665 skip "Need MDS version at least 2.6.54"
6666 [[ $OSTCOUNT -ge 4 ]] || skip_env "needs >= 4 OSTs"
6672 # Format OSTs with random sparse indices.
6676 local LOV_V1_INSANE_STRIPE_COUNT=65532
6677 for i in $(seq $OSTCOUNT); do
6678 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
6679 ost_indices+=" $index"
6681 ost_indices=$(comma_list $ost_indices)
6683 trap "restore_ostindex" EXIT
6684 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
6685 OST_INDEX_LIST=[$ost_indices] formatall
6687 # Setup Lustre filesystem.
6688 start_mgsmds || error "start_mgsmds failed"
6689 for i in $(seq $OSTCOUNT); do
6690 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6691 error "start ost$i failed"
6694 mount_client $MOUNT || error "mount client $MOUNT failed"
6698 check_lfs_df_ret_val $? || error "$LFS df $MOUNT failed"
6699 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6701 # Create a new pool and add OSTs into it.
6702 local ost_pool=$FSNAME.$TESTNAME
6703 create_pool $ost_pool || error "create OST pool $ost_pool failed"
6706 trap "cleanup_82b" EXIT
6708 local ost_idx_in_list=${ost_indices##*,}
6709 local ost_idx_in_pool=$(exclude_items_from_list $ost_indices \
6712 local ost_targets="$FSNAME-OST["
6713 for i in ${ost_idx_in_pool//,/ }; do
6714 ost_targets=$ost_targets$(printf "%04x," $i)
6716 ost_targets="${ost_targets%,}]"
6718 local ost_targets_uuid=$(for i in ${ost_idx_in_pool//,/ }; \
6719 do printf "$FSNAME-OST%04x_UUID\n" $i; done |
6720 sort -u | tr '\n' ' ')
6722 local cmd="$LCTL pool_add $ost_pool $ost_targets"
6723 do_facet mgs $cmd || error "$cmd failed"
6724 wait_update $HOSTNAME "$LCTL get_param -n lov.$FSNAME-*.pools.$TESTNAME|
6725 sort -u | tr '\n' ' ' " "$ost_targets_uuid" ||
6726 error "wait_update $ost_pool failed"
6727 wait_update_facet $SINGLEMDS "$LCTL pool_list $ost_pool | wc -l" 4 ||
6728 error "wait_update pool_list $ost_pool failed"
6730 # If [--pool|-p <pool_name>] is set with [--ost-list|-o <ost_indices>],
6731 # then the OSTs must be the members of the pool.
6732 local file=$DIR/$tdir/$tfile
6733 cmd="$LFS setstripe -p $ost_pool -o $ost_idx_in_list $file"
6735 eval $cmd && error "OST with index $ost_idx_in_list should be" \
6736 "in OST pool $ost_pool"
6738 # Only select OST $ost_idx_in_list from $ost_pool for file.
6739 ost_idx_in_list=${ost_idx_in_pool#*,}
6740 cmd="$LFS setstripe -p $ost_pool -o $ost_idx_in_list $file"
6742 eval $cmd || error "$cmd failed"
6743 cmd="$LFS getstripe $file"
6745 eval $cmd || error "$cmd failed"
6746 check_stripe_count $file 2
6747 check_obdidx $file $ost_idx_in_list
6748 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6749 error "write $file failed"
6753 run_test 82b "specify OSTs for file with --pool and --ost-list options"
6756 [[ "$OST1_VERSION" -ge $(version_code 2.6.91) ]] ||
6757 skip "Need OST version at least 2.6.91"
6758 if [ "$ost1_FSTYPE" != ldiskfs ]; then
6759 skip "ldiskfs only test"
6768 ostmnt=$(facet_mntpt ost1)
6770 # Mount the OST as an ldiskfs filesystem.
6771 log "mount the OST $dev as a $ost1_FSTYPE filesystem"
6772 add ost1 $(mkfs_opts ost1 $dev) $FSTYPE_OPT \
6773 --reformat $dev > /dev/null ||
6774 error "format ost1 error"
6776 if ! test -b $dev; then
6777 mnt_opts=$(csa_add "$OST_MOUNT_FS_OPTS" -o loop)
6779 echo "mnt_opts $mnt_opts"
6780 do_facet ost1 mount -t "$ost1_FSTYPE" $dev \
6782 # Run llverfs on the mounted ldiskfs filesystem.
6783 # It is needed to get ENOSPACE.
6784 log "run llverfs in partial mode on the OST $ost1_FSTYPE $ostmnt"
6785 do_rpc_nodes $(facet_host ost1) run_llverfs $ostmnt -vpl \
6786 "no" || error "run_llverfs error on $ost1_FSTYPE"
6789 log "unmount the OST $dev"
6792 # Delete file IO_scrub. Later osd_scrub_setup will try to
6793 # create "IO_scrub" but will get ENOSPACE.
6795 echo "start ost1 service on `facet_active_host ost1`"
6796 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
6799 err=$(do_facet ost1 dmesg | grep "VFS: Busy inodes after unmount of")
6800 echo "string err $err"
6801 [ -z "$err" ] || error $err
6804 run_test 83 "ENOSPACE on OST doesn't cause message VFS: \
6805 Busy inodes after unmount ..."
6808 local facet=$SINGLEMDS
6809 local num=$(echo $facet | tr -d "mds")
6810 local dev=$(mdsdevname $num)
6811 local time_min=$(recovery_time_min)
6812 local recovery_duration
6813 local completed_clients
6814 local correct_clients
6817 echo "start mds service on $(facet_active_host $facet)"
6819 "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
6820 error "start MDS failed"
6822 start_ost || error "start OST0000 failed"
6823 wait_osc_import_state mds ost1 FULL
6824 start_ost2 || error "start OST0001 failed"
6825 wait_osc_import_state mds ost2 FULL
6827 echo "recovery_time=$time_min, timeout=$TIMEOUT, wrap_up=$wrap_up"
6829 mount_client $MOUNT1 || error "mount $MOUNT1 failed"
6830 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
6831 # make sure new superblock labels are sync'd before disabling writes
6835 replay_barrier $SINGLEMDS
6836 createmany -o $DIR1/$tfile-%d 1000
6838 # We need to catch the end of recovery window to extend it.
6839 # Skip 5 requests and add delay to request handling.
6840 #define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP
6841 do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5"
6843 facet_failover --fsck $SINGLEMDS || error "failover: $?"
6846 echo "recovery status"
6847 do_facet $SINGLEMDS \
6848 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status"
6850 recovery_duration=$(do_facet $SINGLEMDS \
6851 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
6852 awk '/recovery_duration/ { print $2 }')
6853 (( $recovery_duration > $time_min + $wrap_up )) &&
6854 error "recovery_duration > recovery_time_hard + wrap up"
6855 completed_clients=$(do_facet $SINGLEMDS \
6856 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
6857 awk '/completed_clients/ { print $2 }')
6859 correct_clients="$MDSCOUNT/$((MDSCOUNT+1))"
6860 [ "$completed_clients" = "${correct_clients}" ] ||
6861 error "$completed_clients != $correct_clients"
6863 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
6864 umount_client $MOUNT1
6865 umount_client $MOUNT2
6871 run_test 84 "check recovery_hard_time"
6874 [[ "$OST1_VERSION" -ge $(version_code 2.7.55) ]] ||
6875 skip "Need OST version at least 2.7.55"
6876 ##define OBD_FAIL_OSD_OST_EA_FID_SET 0x197
6877 do_facet ost1 "lctl set_param fail_loc=0x197"
6881 run_test 85 "osd_ost init: fail ea_fid_set"
6886 # ost1 has already registered to the MGS before the reformat.
6887 # So after reformatting it with option "-G", it could not be
6888 # mounted to the MGS. Cleanup the system for subsequent tests.
6893 [ "$ost1_FSTYPE" = zfs ] &&
6894 skip "LU-6442: no such mkfs params for ZFS OSTs"
6895 [[ "$MDS1_VERSION" -ge $(version_code 2.7.56) ]] ||
6896 skip "Need server version newer than 2.7.55"
6899 local OLDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6900 awk '/Flex block group size: / { print $NF; exit; }')
6902 [ "$OLDSIZE" == "$NEWSIZE" ] && skip "$NEWSIZE groups already"
6904 local opts=" -O flex_bg -G $NEWSIZE"
6905 opts=$(OST_FS_MKFS_OPTS+="$opts" mkfs_opts ost1 $(ostdevname 1))
6906 opts+=" --reformat $(ostdevname 1) $(ostvdevname 1)"
6907 echo "params: $opts"
6909 trap cleanup_86 EXIT ERR
6912 add ost1 $opts || error "add ost1 failed with new params"
6914 local FOUNDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6915 awk '/Flex block group size: / { print $NF; exit; }')
6917 [[ $FOUNDSIZE == $NEWSIZE ]] ||
6918 error "Flex block group size: $FOUNDSIZE, expected: $NEWSIZE"
6922 run_test 86 "Replacing mkfs.lustre -G option"
6924 test_87() { #LU-6544
6925 [[ "$MDS1_VERSION" -ge $(version_code 2.9.51) ]] ||
6926 skip "Need MDS version at least 2.9.51"
6927 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
6928 skip "ldiskfs only test"
6929 [[ $OSTCOUNT -gt 59 ]] &&
6930 skip "Ignore wide striping situation"
6931 [ -n "$FILESET" ] && skip "Not functional for FILESET set"
6933 local mdsdev=$(mdsdevname 1)
6934 local mdsvdev=$(mdsvdevname 1)
6935 local file=$DIR/$tfile
6936 local mntpt=$(facet_mntpt $SINGLEMDS)
6937 local used_xattr_blk=0
6938 local inode_size=${1:-1024}
6940 local xtest="trusted.test"
6944 local stripe_cnt=$(($OSTCOUNT + 2))
6946 #Please see ldiskfs_make_lustre() for MDT inode size calculation
6947 if [ $stripe_cnt -gt 16 ]; then
6950 left_size=$(expr $inode_size - \
6953 32 - 40 \* 3 - 32 \* 3 - $stripe_cnt \* 24 - 16 - 3 - \
6955 24 - 18 - $(expr length $tfile) - 16 - 4)
6956 if [ $left_size -le 0 ]; then
6957 echo "No space($left_size) is expected in inode."
6958 echo "Try 1-byte xattr instead to verify this."
6961 echo "Estimate: at most $left_size-byte space left in inode."
6967 if ! combined_mgs_mds ; then
6971 add mds1 $(mkfs_opts mds1 ${mdsdev}) --stripe-count-hint=$stripe_cnt \
6972 --reformat $mdsdev $mdsvdev || error "add mds1 failed"
6973 start_mdt 1 > /dev/null || error "start mdt1 failed"
6974 for i in $(seq $OSTCOUNT); do
6975 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS > /dev/null ||
6976 error "start ost$i failed"
6978 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
6979 check_mount || error "check client $MOUNT failed"
6982 $LFS setstripe -E 1M -S 1M -c 1 -E 64M -c 1 -E -1 -c -1 $file ||
6983 error "Create file with 3 components failed"
6984 $TRUNCATE $file $((1024*1024*64+1)) || error "truncate file failed"
6985 i=$($LFS getstripe -I3 -c $file) || error "get 3rd stripe count failed"
6986 if [ $i -ne $OSTCOUNT ]; then
6987 left_size=$(expr $left_size + $(expr $OSTCOUNT - $i) \* 24)
6988 echo -n "Since only $i out $OSTCOUNT OSTs are used, "
6989 echo -n "the expected left space is changed to "
6990 echo "$left_size bytes at most."
6992 value=$(generate_string $left_size)
6993 setfattr -n $xtest -v $value $file
6994 orig=$(get_xattr_value $xtest $file)
6995 [[ "$orig" != "$value" ]] && error "$xtest changed"
6997 #Verify if inode has some expected space left
6998 umount $MOUNT > /dev/null || error "umount $MOUNT failed"
6999 stop_mdt 1 > /dev/null || error "stop mdt1 failed"
7000 mount_ldiskfs $SINGLEMDS || error "mount -t ldiskfs $SINGLEMDS failed"
7002 do_facet $SINGLEMDS ls -sal $mntpt/ROOT/$tfile
7003 used_xattr_blk=$(do_facet $SINGLEMDS ls -s $mntpt/ROOT/$tfile |
7005 [[ $used_xattr_blk -eq 0 ]] &&
7006 error "Please check MDS inode size calculation: \
7007 more than $left_size-byte space left in inode."
7008 echo "Verified: at most $left_size-byte space left in inode."
7010 unmount_ldiskfs $SINGLEMDS
7012 for i in $(seq $OSTCOUNT); do
7013 stop ost$i -f || error "stop ost$i failed"
7016 run_test 87 "check if MDT inode can hold EAs with N stripes properly"
7019 [ "$mds1_FSTYPE" == zfs ] &&
7020 skip "LU-6662: no implementation for ZFS"
7024 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
7025 --reformat $(mdsdevname 1) || error "add mds1 failed"
7027 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
7028 grep -e \".*opts:.*errors=remount-ro.*\"" ||
7029 error "default mount options is missing"
7031 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
7032 --mountfsoptions="user_xattr,errors=panic" \
7033 --reformat $(mdsdevname 1) || error "add mds1 failed"
7035 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
7036 grep -e \".*opts:.*errors=panic.*\"" ||
7037 error "user can't override default mount options"
7039 run_test 88 "check the default mount options can be overridden"
7041 test_89() { # LU-7131
7042 [[ "$MDS1_VERSION" -ge $(version_code 2.9.54) ]] ||
7043 skip "Need MDT version at least 2.9.54"
7045 local key=failover.node
7046 local val1=192.0.2.254@tcp0 # Reserved IPs, see RFC 5735
7047 local val2=192.0.2.255@tcp0
7048 local mdsdev=$(mdsdevname 1)
7053 if [[ "$mds1_FSTYPE" == zfs ]]; then
7054 import_zpool mds1 || return ${PIPESTATUS[0]}
7057 # Check that parameters are added correctly
7058 echo "tunefs --param $key=$val1"
7059 do_facet mds1 "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
7060 error "tunefs --param $key=$val1 failed"
7061 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7062 error "tunefs --dryrun failed"
7063 params=${params##*Parameters:}
7064 params=${params%%exiting*}
7065 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val1) = "1" ] ||
7066 error "on-disk parameter not added correctly via tunefs"
7068 # Check that parameters replace existing instances when added
7069 echo "tunefs --param $key=$val2"
7070 do_facet mds1 "$TUNEFS --param $key=$val2 $mdsdev >/dev/null" ||
7071 error "tunefs --param $key=$val2 failed"
7072 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7073 error "tunefs --dryrun failed"
7074 params=${params##*Parameters:}
7075 params=${params%%exiting*}
7076 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "1" ] ||
7077 error "on-disk parameter not replaced via tunefs"
7078 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val2) = "1" ] ||
7079 error "on-disk parameter not replaced correctly via tunefs"
7081 # Check that a parameter is erased properly
7082 echo "tunefs --erase-param $key"
7083 do_facet mds1 "$TUNEFS --erase-param $key $mdsdev >/dev/null" ||
7084 error "tunefs --erase-param $key failed"
7085 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7086 error "tunefs --dryrun failed"
7087 params=${params##*Parameters:}
7088 params=${params%%exiting*}
7089 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "0" ] ||
7090 error "on-disk parameter not erased correctly via tunefs"
7092 # Check that all the parameters are erased
7093 do_facet mds1 "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
7094 error "tunefs --param $key=$val1 failed"
7095 echo "tunefs --erase-params"
7096 do_facet mds1 "$TUNEFS --erase-params $mdsdev >/dev/null" ||
7097 error "tunefs --erase-params failed"
7098 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7099 error "tunefs --dryrun failed"
7100 params=${params##*Parameters:}
7101 params=${params%%exiting*}
7102 params=$(echo $params | tr ' ' '\n')
7104 error "all on-disk parameters not erased correctly via tunefs $params"
7106 # Check the order of options --erase-params and --param
7107 echo "tunefs --param $key=$val1 --erase-params"
7109 "$TUNEFS --param $key=$val1 --erase-params $mdsdev >/dev/null"||
7110 error "tunefs --param $key=$val1 --erase-params failed"
7111 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7112 error "tunefs --dryrun failed"
7113 params=${params##*Parameters:}
7114 params=${params%%exiting*}
7115 [ $(echo $params | tr ' ' '\n') == "$key=$val1" ] ||
7116 error "on-disk param not added correctly with --erase-params"
7120 run_test 89 "check tunefs --param and --erase-param{s} options"
7123 # $2 (optional) value of max_mod_rpcs_in_flight to set
7124 check_max_mod_rpcs_in_flight() {
7132 idx=$(printf "%04x" $($LFS getdirstripe -i $dir))
7133 facet="mds$((0x$idx + 1))"
7135 if [ -z "$mmr" ]; then
7136 # get value of max_mod_rcps_in_flight
7137 mmr=$($LCTL get_param -n \
7138 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) ||
7139 error "Unable to get max_mod_rpcs_in_flight"
7140 echo "max_mod_rcps_in_flight is $mmr"
7142 # set value of max_mod_rpcs_in_flight
7144 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight=$mmr ||
7145 error "Unable to set max_mod_rpcs_in_flight to $mmr"
7146 echo "max_mod_rpcs_in_flight set to $mmr"
7149 # create mmr+1 files
7150 echo "creating $((mmr + 1)) files ..."
7152 for i in $(seq $((mmr + 1))); do
7158 # consumes mmr-1 modify RPC slots
7159 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7160 # drop requests on MDT so that RPC slots are consumed
7161 # during all the request resend interval
7162 do_facet $facet "$LCTL set_param fail_loc=0x159"
7163 echo "launch $((mmr - 1)) chmod in parallel ..."
7164 for i in $(seq $((mmr - 1))); do
7165 chmod 0600 $dir/file-$i &
7169 # send one additional modify RPC
7170 do_facet $facet "$LCTL set_param fail_loc=0"
7171 echo "launch 1 additional chmod in parallel ..."
7172 chmod 0600 $dir/file-$mmr &
7175 # check this additional modify RPC get a modify RPC slot
7176 # and succeed its operation
7177 checkstat -vp 0600 $dir/file-$mmr ||
7178 error "Unable to send $mmr modify RPCs in parallel"
7183 # consumes mmr modify RPC slots
7184 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7185 # drop requests on MDT so that RPC slots are consumed
7186 # during all the request resend interval
7187 do_facet $facet "$LCTL set_param fail_loc=0x159"
7188 echo "launch $mmr chmod in parallel ..."
7189 for i in $(seq $mmr); do
7190 chmod 0666 $dir/file-$i &
7194 # send one additional modify RPC
7195 do_facet $facet "$LCTL set_param fail_loc=0"
7196 echo "launch 1 additional chmod in parallel ..."
7197 chmod 0666 $dir/file-$((mmr + 1)) &
7200 # check this additional modify RPC blocked getting a modify RPC slot
7201 checkstat -vp 0644 $dir/file-$((mmr + 1)) ||
7202 error "Unexpectedly send $(($mmr + 1)) modify RPCs in parallel"
7206 get_mdt_max_mod_rpcs_in_flight_val() {
7210 # It will be enough to get one value from one MDT
7211 max_mrif=$(do_facet $facet \
7212 "$LCTL get_param -n \
7213 mdt.$FSNAME-MDT*.max_mod_rpcs_in_flight | head -n 1")
7214 if [[ $max_mrif ]]; then
7217 max_mrif=$(do_facet $facet \
7218 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
7223 set_mdt_max_mod_rpcs_in_flight() {
7228 lctl_op=$($LCTL get_param \
7229 mdt.*.max_mod_rpcs_in_flight)
7230 if [[ $lctl_op ]]; then
7233 mdt.$FSNAME-MDT*.max_mod_rpcs_in_flight=$max_mrif"
7237 /sys/module/mdt/parameters/max_mod_rpcs_per_client"
7238 echo "the deprecated max_mod_rpcs_per_client \
7239 parameter was involved"
7246 [[ $($LCTL get_param mdc.*.import |
7247 grep "connect_flags:.*multi_mod_rpc") ]] ||
7248 skip "Need MDC with 'multi_mod_rpcs' feature"
7250 # check default value
7251 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
7252 check_max_mod_rpcs_in_flight $DIR/$tdir
7256 run_test 90a "check max_mod_rpcs_in_flight is enforced"
7266 [[ $($LCTL get_param mdc.*.import |
7267 grep "connect_flags:.*multi_mod_rpc") ]] ||
7268 skip "Need MDC with 'multi_mod_rpcs' feature"
7271 # update max_mod_rpcs_in_flight
7272 $LFS mkdir -c1 $DIR/${tdir}1 || error "mkdir $DIR/${tdir}1 failed"
7273 check_max_mod_rpcs_in_flight $DIR/${tdir}1 1
7276 # check client is able to send multiple modify RPCs in paralell
7277 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
7278 grep -c "multi_mod_rpcs")
7279 if [ "$tmp" -ne $MDSCOUNT ]; then
7281 skip "Client not able to send multiple modify RPCs in parallel"
7284 # update max_mod_rpcs_in_flight
7285 $LFS mkdir -c1 $DIR/${tdir}2 || error "mkdir $DIR/${tdir}2 failed"
7286 check_max_mod_rpcs_in_flight $DIR/${tdir}2 5
7289 $LFS mkdir -c1 $DIR/${tdir}3 || error "mkdir $DIR/${tdir}3 failed"
7290 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/${tdir}3))
7291 facet="mds$((0x$idx + 1))"
7293 mmrpc=$(get_mdt_max_mod_rpcs_in_flight_val $facet)
7294 echo "mdt_max_mod_rpcs_in_flight is $mmrpc"
7295 # update max_mod_rpcs_in_flight
7296 umount_client $MOUNT
7297 set_mdt_max_mod_rpcs_in_flight 16 $facet
7299 $LCTL set_param mdc.$FSNAME-MDT$idx-mdc-*.max_rpcs_in_flight=17
7300 check_max_mod_rpcs_in_flight $DIR/${tdir}3 16
7302 # restore MDT max_mod_rpcs_in_flight initial value
7303 set_mdt_max_mod_rpcs_in_flight $mmrpc $facet
7305 rm -rf $DIR/${tdir}?
7308 run_test 90b "check max_mod_rpcs_in_flight is enforced after update"
7311 # get max_rpcs_in_flight value
7312 mrif_90c=$($LCTL get_param -n \
7313 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
7314 echo "max_rpcs_in_flight is $mrif_90c"
7316 # get MDC max_mod_rpcs_in_flight value
7317 mmrif_90c=$($LCTL get_param -n \
7318 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight)
7319 echo "MDC max_mod_rpcs_in_flight is $mmrif_90c"
7321 # get MDT max_mod_rpcs_in_flight value
7322 mmrpc_90c=$(get_mdt_max_mod_rpcs_in_flight_val "mds1")
7323 echo "mdt_max_mod_rpcs_in_flight is $mmrpc_90c"
7326 restore_params_90c() {
7329 # restore max_rpcs_in_flight value
7330 do_facet mgs $LCTL set_param -P \
7331 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$mrif_90c
7333 # restore max_mod_rpcs_in_flight value
7334 do_facet mgs $LCTL set_param -P \
7335 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mmrif_90c
7337 # restore MDT max_mod_rpcs_in_flight value
7338 set_mdt_max_mod_rpcs_in_flight $mmrpc_90c "mds1"
7346 [[ $($LCTL get_param mdc.*.import |
7347 grep "connect_flags:.*multi_mod_rpc") ]] ||
7348 skip "Need MDC with 'multi_mod_rpcs' feature"
7350 # check client is able to send multiple modify RPCs in paralell
7351 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
7352 grep -c "multi_mod_rpcs")
7353 if [ "$tmp" -ne $MDSCOUNT ]; then
7355 skip "Client not able to send multiple modify RPCs in parallel"
7359 stack_trap restore_params_90c
7362 # attempt to set max_mod_rpcs_in_flight to max_rpcs_in_flight value
7363 # prerequisite: set MDT max_mod_rpcs_in_flight to
7364 # max_rpcs_in_flight value
7365 set_mdt_max_mod_rpcs_in_flight $mrif_90c "mds1"
7367 # if max_mod_rpcs_in_flight is set to be equal to or larger than
7368 # max_rpcs_in_flight, then max_rpcs_in_flight will be increased
7369 if [[ "$CLIENT_VERSION" -ge $(version_code 2.13.53) ]]; then
7371 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif_90c ||
7372 error "set max_mod_rpcs_in_flight to $mrif_90c failed"
7374 local new_mrif=$($LCTL get_param -n \
7375 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
7376 ((new_mrif == mrif_90c + 1)) ||
7377 error "max_rpcs_in_flight was not increased"
7380 umount_client $MOUNT
7381 set_mdt_max_mod_rpcs_in_flight $mmrpc_90c "mds1"
7385 # attempt to set max_mod_rpcs_in_flight to MDT max_mod_rpcs_in_flight+1
7386 # prerequisite: set max_rpcs_in_flight to MDT max_mod_rpcs_in_flight+2
7388 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$((mmrpc_90c + 2))
7391 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$((mmrpc_90c + 1)) &&
7392 error "set max_mod_rpcs_in_flight to $((mmrpc_90c + 1)) should fail"
7395 # attempt to set max_mod_rpcs_in_flight permanently
7396 do_facet mgs $LCTL set_param -P \
7397 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$mrif_90c
7399 do_facet mgs $LCTL set_param -P \
7400 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif_90c
7402 remount_client $MOUNT
7404 wait_update_facet --verbose client "$LCTL get_param -n \
7405 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight" \
7406 "$((mrif_90c + 1))" ||
7407 error "expected '$((mrif_90c + 1))' for max_rpcs_in_flight"
7409 wait_update_facet --verbose client "$LCTL get_param -n \
7410 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight" \
7412 error "expected '$mrif_90c' for max_mod_rpcs_in_flight"
7417 run_test 90c "check max_mod_rpcs_in_flight update limits"
7428 [[ $($LCTL get_param mdc.*.import |
7429 grep "connect_flags:.*multi_mod_rpc") ]] ||
7430 skip "Need MDC with 'multi_mod_rpcs' feature"
7432 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
7433 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
7434 facet="mds$((0x$idx + 1))"
7436 # check client version supports multislots
7437 tmp=$($LCTL get_param -N \
7438 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
7439 if [ -z "$tmp" ]; then
7441 skip "Client does not support multiple modify RPCs in flight"
7444 # get current value of max_mod_rcps_in_flight
7445 mmr=$($LCTL get_param -n \
7446 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
7447 echo "max_mod_rcps_in_flight is $mmr"
7450 echo "creating $mmr files ..."
7452 for i in $(seq $mmr); do
7453 touch $DIR/$tdir/file-$i
7456 # prepare for close RPC
7457 multiop_bg_pause $DIR/$tdir/file-close O_c
7460 # consumes mmr modify RPC slots
7461 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7462 # drop requests on MDT so that RPC slots are consumed
7463 # during all the request resend interval
7464 do_facet $facet "$LCTL set_param fail_loc=0x159"
7465 echo "launch $mmr chmod in parallel ..."
7466 for i in $(seq $mmr); do
7467 chmod 0600 $DIR/$tdir/file-$i &
7470 # send one additional close RPC
7471 do_facet $facet "$LCTL set_param fail_loc=0"
7472 echo "launch 1 additional close in parallel ..."
7474 cancel_lru_locks mdc
7477 # check this additional close RPC get a modify RPC slot
7478 # and multiop process completed
7479 [ -d /proc/$pid ] &&
7480 error "Unable to send the additional close RPC in parallel"
7485 run_test 90d "check one close RPC is allowed above max_mod_rpcs_in_flight"
7487 check_uuid_on_ost() {
7489 do_facet ost1 "$LCTL get_param obdfilter.${FSNAME}*.exports.'$nid'.uuid"
7492 check_uuid_on_mdt() {
7494 do_facet $SINGLEMDS "$LCTL get_param mdt.${mds1_svc}*.exports.'$nid'.uuid"
7502 [[ "$OST1_VERSION" -ge $(version_code 2.7.63) ]] ||
7503 skip "Need OST version at least 2.7.63"
7504 [[ "$MDS1_VERSION" -ge $(version_code 2.7.63) ]] ||
7505 skip "Need MDT version at least 2.7.63"
7507 start_mds || error "MDS start failed"
7508 start_ost || error "unable to start OST"
7509 mount_client $MOUNT || error "client start failed"
7510 check_mount || error "check_mount failed"
7513 nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g")
7517 uuid=$(get_client_uuid $MOUNT)
7519 echo "list nids on mdt:"
7520 do_facet $SINGLEMDS "$LCTL list_param mdt.${FSNAME}*.exports.*"
7521 echo "uuid from $nid:"
7522 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
7524 found=$(check_uuid_on_mdt $nid | grep $uuid)
7525 [ -z "$found" ] && error "can't find $uuid $nid on MDT"
7526 found=$(check_uuid_on_ost $nid | grep $uuid)
7527 [ -z "$found" ] && error "can't find $uuid $nid on OST"
7529 # umount the client so it won't reconnect
7530 manual_umount_client --force || error "failed to umount $?"
7531 # shouldn't disappear on MDS after forced umount
7532 found=$(check_uuid_on_mdt $nid | grep $uuid)
7533 [ -z "$found" ] && error "can't find $uuid $nid"
7536 do_facet $SINGLEMDS \
7537 "$LCTL set_param -n mdt.${mds1_svc}.evict_client nid:$nid"
7538 sleep 1 # eviction above is async, give it some time to proceed
7540 found=$(check_uuid_on_mdt $nid | grep $uuid)
7541 [ -n "$found" ] && error "found $uuid $nid on MDT"
7542 found=$(check_uuid_on_ost $nid | grep $uuid)
7543 [ -n "$found" ] && error "found $uuid $nid on OST"
7545 # check it didn't reconnect (being umounted)
7546 sleep $((TIMEOUT+1))
7547 found=$(check_uuid_on_mdt $nid | grep $uuid)
7548 [ -n "$found" ] && error "found $uuid $nid on MDT"
7549 found=$(check_uuid_on_ost $nid | grep $uuid)
7550 [ -n "$found" ] && error "found $uuid $nid on OST"
7554 run_test 91 "evict-by-nid support"
7556 generate_ldev_conf() {
7557 # generate an ldev.conf file
7558 local ldevconfpath=$1
7560 local fsldevformat=""
7563 local facets="mgs,$(get_facets OST),$(get_facets MDS)"
7564 for facet in ${facets//,/ }; do
7566 fstype=$(facet_fstype $facet)
7567 [ "$fstype" = zfs ] && fsldevformat="$fstype:"
7568 local host=$(facet_host $facet)
7570 local varfo=${facet}failover_HOST
7571 if [ $facet == mgs ] && combined_mgs_mds; then
7572 varfo=mds1failover_HOST
7574 [ -n "${!varfo}" ] && fo=${!varfo}
7575 local type=$(echo $facet | tr -d "[:digit:]" | \
7576 tr "[:lower:]" "[:upper:]" | sed s/MDS/MDT/ )
7578 [ ${facet} == mgs ] ||
7579 num=$(facet_number $facet)
7580 printf "%s\t%s\t%s-%s%04x\t%s%s\n" \
7587 $(facet_device $facet) >> $ldevconfpath
7590 echo "----- $ldevconfpath -----"
7592 echo "--- END $ldevconfpath ---"
7597 # generate a nids file (mapping between hostname to nid)
7598 # looks like we only have the MGS nid available to us
7599 # so just echo that to a file
7601 echo -e "${mgs_HOST}\t${MGSNID}" > $nidspath
7603 echo "----- $nidspath -----"
7605 echo "--- END $nidspath ---"
7608 compare_ldev_output() {
7609 local ldev_output=$1
7610 local expected_output=$2
7612 sort $expected_output -o $expected_output
7613 sort $ldev_output -o $ldev_output
7615 echo "-- START OF LDEV OUTPUT --"
7617 echo "--- END OF LDEV OUTPUT ---"
7619 echo "-- START OF EXPECTED OUTPUT --"
7620 cat $expected_output
7621 echo "--- END OF EXPECTED OUTPUT ---"
7623 diff $expected_output $ldev_output
7628 if [ -z "$LDEV" ]; then
7629 error "ldev is missing!"
7632 local LDEVCONFPATH=$TMP/ldev.conf
7633 local NIDSPATH=$TMP/nids
7635 echo "Host is $(hostname)"
7637 generate_ldev_conf $LDEVCONFPATH
7638 generate_nids $NIDSPATH
7640 # echo the mgs nid and compare it to environment variable MGSNID
7641 local facets="$(get_facets OST),$(get_facets MDS),mgs"
7642 for facet in ${facets//,/ }; do
7643 local host=$(facet_host $facet)
7644 local output=$($LDEV -c $LDEVCONFPATH -H $host -n $NIDSPATH echo %m)
7646 echo "-- START OF LDEV OUTPUT --"
7648 echo "--- END OF LDEV OUTPUT ---"
7651 error "ldev failed to execute!"
7653 # need to process multiple lines because of
7654 # several targets on host
7655 echo -e $output | awk '{ print $2 }' | while read -r line ; do
7656 [ "$line" = "$MGSNID" ] ||
7657 error "ldev failed mgs nid '$line', \
7661 rm -f $LDEVCONFPATH $NIDSPATH
7663 run_test 92 "ldev returns MGS NID correctly in command substitution"
7666 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs"
7669 #start mgs or mgs/mdt0
7670 if ! combined_mgs_mds ; then
7677 start_ost || error "OST0 start fail"
7679 #define OBD_FAIL_MGS_WRITE_TARGET_DELAY 0x90e
7680 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x8000090e"
7681 for num in $(seq 2 $MDSCOUNT); do
7685 mount_client $MOUNT || error "mount client fails"
7686 wait_osc_import_state mds ost FULL
7687 wait_osc_import_ready client ost
7688 check_mount || error "check_mount failed"
7690 cleanup || error "cleanup failed with $?"
7692 run_test 93 "register mulitple MDT at the same time"
7695 if [ -z "$LDEV" ]; then
7696 error "ldev is missing!"
7699 local LDEVCONFPATH=$TMP/ldev.conf
7700 local NIDSPATH=$TMP/nids
7702 generate_ldev_conf $LDEVCONFPATH
7703 generate_nids $NIDSPATH
7705 local LDEV_OUTPUT=$TMP/ldev-output.txt
7706 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME > $LDEV_OUTPUT
7708 # ldev failed, error
7709 if [ $? -ne 0 ]; then
7710 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7711 error "ldev failed to execute!"
7715 local EXPECTED_OUTPUT=$TMP/ldev-expected.txt
7717 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
7719 for num in $(seq $MDSCOUNT); do
7720 printf "%s-MDT%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7723 for num in $(seq $OSTCOUNT); do
7724 printf "%s-OST%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7727 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7729 if [ $? -ne 0 ]; then
7730 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7731 error "ldev failed to produce the correct hostlist!"
7734 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7736 run_test 94 "ldev outputs correct labels for file system name query"
7739 if [ -z "$LDEV" ]; then
7740 error "ldev is missing!"
7743 local LDEVCONFPATH=$TMP/ldev.conf
7744 local NIDSPATH=$TMP/nids
7746 generate_ldev_conf $LDEVCONFPATH
7747 generate_nids $NIDSPATH
7751 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME &>/dev/null
7752 if [ $? -ne 0 ]; then
7753 rm $LDEVCONFPATH $NIDSPATH
7754 error "ldev label filtering w/ -F failed!"
7758 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -l &>/dev/null
7759 if [ $? -ne 0 ]; then
7760 rm $LDEVCONFPATH $NIDSPATH
7761 error "ldev label filtering w/ -l failed!"
7765 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f &>/dev/null
7766 if [ $? -ne 0 ]; then
7767 rm $LDEVCONFPATH $NIDSPATH
7768 error "ldev label filtering w/ -f failed!"
7772 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a &>/dev/null
7773 if [ $? -ne 0 ]; then
7774 rm $LDEVCONFPATH $NIDSPATH
7775 error "ldev label filtering w/ -a failed!"
7780 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -F $FSNAME &>/dev/null
7781 if [ $? -eq 0 ]; then
7782 rm $LDEVCONFPATH $NIDSPATH
7783 error "ldev label filtering w/ -a and -F incorrectly succeeded"
7787 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -f &>/dev/null
7788 if [ $? -eq 0 ]; then
7789 rm $LDEVCONFPATH $NIDSPATH
7790 error "ldev label filtering w/ -a and -f incorrectly succeeded"
7794 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -l &>/dev/null
7795 if [ $? -eq 0 ]; then
7796 rm $LDEVCONFPATH $NIDSPATH
7797 error "ldev label filtering w/ -a and -l incorrectly succeeded"
7801 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f -l &>/dev/null
7802 if [ $? -eq 0 ]; then
7803 rm $LDEVCONFPATH $NIDSPATH
7804 error "ldev label filtering w/ -f and -l incorrectly succeeded"
7808 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -l &>/dev/null
7809 if [ $? -eq 0 ]; then
7810 rm $LDEVCONFPATH $NIDSPATH
7811 error "ldev label filtering w/ -F and -l incorrectly succeeded"
7814 # file sys & foreign
7815 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -f &>/dev/null
7816 if [ $? -eq 0 ]; then
7817 rm $LDEVCONFPATH $NIDSPATH
7818 error "ldev label filtering w/ -F and -f incorrectly succeeded"
7821 rm $LDEVCONFPATH $NIDSPATH
7823 run_test 95 "ldev should only allow one label filter"
7826 if [ -z "$LDEV" ]; then
7827 error "ldev is missing!"
7830 local LDEVCONFPATH=$TMP/ldev.conf
7831 local NIDSPATH=$TMP/nids
7833 generate_ldev_conf $LDEVCONFPATH
7834 generate_nids $NIDSPATH
7836 local LDEV_OUTPUT=$TMP/ldev-output.txt
7837 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -H $mgs_HOST \
7839 awk '{print $2}' > $LDEV_OUTPUT
7841 # ldev failed, error
7842 if [ $? -ne 0 ]; then
7843 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7844 error "ldev failed to execute!"
7848 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
7850 echo "$mgs_HOST-$(facet_fstype mgs)" > $EXPECTED_OUTPUT
7852 local facets="$(get_facets OST),$(get_facets MDS)"
7853 for facet in ${facets//,/ }; do
7854 local host=$(facet_host $facet)
7855 [ "$mgs_HOST" == "$host" ] &&
7856 echo "$host-$(facet_fstype $facet)" \
7860 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7862 if [ $? -ne 0 ]; then
7863 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7864 error "ldev failed to produce the correct output!"
7867 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7869 run_test 96 "ldev returns hostname and backend fs correctly in command sub"
7872 if [ -z "$LDEV" ]; then
7873 error "ldev is missing!"
7876 local LDEVCONFPATH=$TMP/ldev.conf
7877 local NIDSPATH=$TMP/nids
7879 generate_ldev_conf $LDEVCONFPATH
7880 generate_nids $NIDSPATH
7882 local LDEV_OUTPUT=$TMP/ldev-output.txt
7883 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
7885 echo -e "\nMDT role"
7886 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mdt > $LDEV_OUTPUT
7888 if [ $? -ne 0 ]; then
7889 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7890 error "ldev failed to execute for mdt role!"
7893 for num in $(seq $MDSCOUNT); do
7894 printf "%s-MDT%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7897 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7899 if [ $? -ne 0 ]; then
7900 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7901 error "ldev failed to produce the correct output for mdt role!"
7904 echo -e "\nOST role"
7905 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R ost > $LDEV_OUTPUT
7907 if [ $? -ne 0 ]; then
7908 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
7909 error "ldev failed to execute for ost role!"
7913 for num in $(seq $OSTCOUNT); do
7914 printf "%s-OST%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7917 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7919 if [ $? -ne 0 ]; then
7920 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7921 error "ldev failed to produce the correct output for ost role!"
7924 echo -e "\nMGS role"
7925 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mgs > $LDEV_OUTPUT
7927 if [ $? -ne 0 ]; then
7928 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
7929 error "ldev failed to execute for mgs role!"
7932 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
7934 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7936 if [ $? -ne 0 ]; then
7937 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7938 error "ldev failed to produce the correct output for mgs role!"
7941 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7943 run_test 97 "ldev returns correct ouput when querying based on role"
7948 local temp=$MDS_MOUNT_OPTS
7951 check_mount || error "mount failed"
7952 mountopt="user_xattr"
7953 for ((x = 1; x <= PAGE_SIZE/11; x++)); do
7954 mountopt="$mountopt,user_xattr"
7956 mount_client $MOUNT remount,$mountopt 2>&1 | grep "too long" ||
7957 error "Buffer overflow check failed"
7958 cleanup || error "cleanup failed"
7960 run_test 98 "Buffer-overflow check while parsing mount_opts"
7964 [[ "$ost1_FSTYPE" != ldiskfs ]] &&
7965 skip "ldiskfs only test"
7966 [[ "$OST1_VERSION" -ge $(version_code 2.8.57) ]] ||
7967 skip "Need OST version at least 2.8.57"
7969 local ost_opts="$(mkfs_opts ost1 $(ostdevname 1)) \
7970 --reformat $(ostdevname 1) $(ostvdevname 1)"
7971 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" &&
7972 skip "meta_bg already set"
7975 if [[ ${!opts} != *mkfsoptions* ]]; then
7976 eval opts=\"${!opts} \
7977 --mkfsoptions='\\\"-O ^resize_inode,meta_bg\\\"'\"
7979 local val=${!opts//--mkfsoptions=\\\"/ \
7980 --mkfsoptions=\\\"-O ^resize_inode,meta_bg }
7984 echo "params: $opts"
7987 add ost1 $opts || error "add ost1 failed with new params"
7989 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" ||
7990 error "meta_bg is not set"
7992 run_test 99 "Adding meta_bg option"
7997 start_mgsmds || error "MDS start failed"
7998 start_ost || error "unable to start OST"
7999 mount_client $MOUNT || error "client start failed"
8000 check_mount || error "check_mount failed"
8009 do_facet mgs 'lshowmount -v' | awk 'BEGIN {NR == 0; rc=1} /MGS:/ {rc=0}
8010 END {exit rc}' || error "lshowmount have no output MGS"
8012 do_facet mds1 'lshowmount -v' | awk 'BEGIN {NR == 2; rc=1} /-MDT0000:/
8013 {rc=0} END {exit rc}' || error "lshowmount have no output MDT0"
8015 do_facet ost1 'lshowmount -v' | awk 'BEGIN {NR == 4; rc=1} /-OST0000:/
8016 {rc=0} END {exit rc}' || error "lshowmount have no output OST0"
8018 cleanup || error "cleanup failed with $?"
8020 run_test 100 "check lshowmount lists MGS, MDT, OST and 0@lo"
8023 local createmany_pid
8024 local dev=$FSNAME-OST0000-osc-MDT0000
8028 do_nodes $(comma_list $(osts_nodes)) $LCTL set_param \
8029 seq.*OST*-super.width=$DATA_SEQ_MAX_WIDTH
8030 createmany -o $DIR1/$tdir/$tfile-%d 50000 &
8032 # MDT->OST reconnection causes MDT<->OST last_id synchornisation
8033 # via osp_precreate_cleanup_orphans.
8034 for ((i = 0; i < 100; i++)); do
8035 for ((k = 0; k < 10; k++)); do
8036 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;" \
8037 "$LCTL --device $dev activate"
8040 ls -asl $MOUNT | grep '???' &&
8041 { kill -9 $createmany_pid &>/dev/null;
8042 error "File has no object on OST"; }
8044 kill -s 0 $createmany_pid || break
8046 wait $createmany_pid
8048 unlinkmany $DIR1/$tdir/$tfile-%d 50000
8051 run_test 101a "Race MDT->OST reconnection with create"
8054 local dev=$FSNAME-OST0000-osc-MDT0000
8055 local dir=$DIR1/$tdir
8059 $LFS setstripe -c 1 -i 0 $dir
8060 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;"
8061 #define OBD_FAIL_OSP_CON_EVENT_DELAY 0x2107
8062 do_facet mds1 "$LCTL set_param fail_loc=0x80002107 fail_val=20"
8063 do_facet $SINGLEMDS "$LCTL --device $dev activate;"
8068 wait_osc_import_ready client ost1
8069 touch $dir/$tfile || error "Can't create file"
8073 run_test 101b "Race events DISCONNECT and ACTIVE in osp"
8076 [[ "$MDS1_VERSION" -gt $(version_code 2.9.53) ]] ||
8077 skip "Need server version greater than 2.9.53"
8078 [[ “$(mdsdevname 1)” != “$(mgsdevname)” ]] &&
8079 [[ “$(facet_host mds1)” = “$(facet_host mgs)” ]] &&
8080 skip "MGS must be on different node or combined"
8082 cleanup || error "cleanup failed with $?"
8084 local mds1dev=$(mdsdevname 1)
8085 local mds1mnt=$(facet_mntpt mds1)
8086 local mds1opts=$MDS_MOUNT_OPTS
8088 if [ "$mds1_FSTYPE" == ldiskfs ] &&
8089 ! do_facet mds1 test -b $mds1dev; then
8090 mds1opts=$(csa_add "$mds1opts" -o loop)
8092 if [[ "$mds1_FSTYPE" == zfs ]]; then
8093 import_zpool mds1 || return ${PIPESTATUS[0]}
8096 # unload all and only load libcfs to allow fail_loc setting
8097 do_facet mds1 $LUSTRE_RMMOD || error "unable to unload modules"
8098 do_facet mds1 modprobe libcfs || error "libcfs not loaded"
8099 do_facet mds1 lsmod \| grep libcfs || error "libcfs not loaded"
8101 #define OBD_FAIL_OBDCLASS_MODULE_LOAD 0x60a
8102 do_facet mds1 "$LCTL set_param fail_loc=0x8000060a"
8104 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts &&
8105 error "mdt start must fail"
8106 do_facet mds1 lsmod \| grep obdclass && error "obdclass must not load"
8108 do_facet mds1 "$LCTL set_param fail_loc=0x0"
8110 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts ||
8111 error "mdt start must not fail"
8113 cleanup || error "cleanup failed with $?"
8115 run_test 102 "obdclass module cleanup upon error"
8120 echo "rename $FSNAME to $newname"
8122 if ! combined_mgs_mds ; then
8123 local dev=$(mgsdevname)
8126 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
8127 error "(7) Fail to rename MGS"
8128 if [ "$(facet_fstype mgs)" = "zfs" ]; then
8129 reimport_zpool mgs $newname-mgs
8133 for num in $(seq $MDSCOUNT); do
8134 local dev=$(mdsdevname $num)
8136 do_facet mds${num} \
8137 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
8138 error "(8) Fail to rename MDT $num"
8139 if [ "$(facet_fstype mds${num})" = "zfs" ]; then
8140 reimport_zpool mds${num} $newname-mdt${num}
8144 for num in $(seq $OSTCOUNT); do
8145 local dev=$(ostdevname $num)
8147 do_facet ost${num} \
8148 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
8149 error "(9) Fail to rename OST $num"
8150 if [ "$(facet_fstype ost${num})" = "zfs" ]; then
8151 reimport_zpool ost${num} $newname-ost${num}
8156 test_103_set_pool() {
8160 do_facet mgs $LCTL pool_add $FSNAME.$pname ${FSNAME}-$ost_x ||
8161 error "Fail to add $ost_x to $FSNAME.$pname"
8162 wait_update $HOSTNAME \
8163 "lctl get_param -n lov.$FSNAME-clilov-*.pools.$pname |
8164 grep $ost_x" "$FSNAME-${ost_x}_UUID" ||
8165 error "$ost_x is NOT in pool $FSNAME.$pname"
8168 test_103_check_pool() {
8169 local save_fsname=$1
8172 stat $DIR/$tdir/test-framework.sh ||
8173 error "($errno) Fail to stat"
8174 do_facet mgs $LCTL pool_list $FSNAME.pool1 ||
8175 error "($errno) Fail to list $FSNAME.pool1"
8176 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname ||
8177 error "($errno) Fail to list $FSNAME.$save_fsname"
8178 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname |
8179 grep ${FSNAME}-OST0000 ||
8180 error "($errno) List $FSNAME.$save_fsname is invalid"
8182 local pname=$($LFS getstripe --pool $DIR/$tdir/d0)
8183 [ "$pname" = "$save_fsname" ] ||
8184 error "($errno) Unexpected pool name $pname"
8188 check_mount_and_prep
8190 mkdir $DIR/$tdir || error "(1) Fail to mkdir $DIR/$tdir"
8191 cp $LUSTRE/tests/test-framework.sh $DIR/$tdir ||
8192 error "(2) Fail to copy test-framework.sh"
8194 do_facet mgs $LCTL pool_new $FSNAME.pool1 ||
8195 error "(3) Fail to create $FSNAME.pool1"
8196 # name the pool name as the fsname
8197 do_facet mgs $LCTL pool_new $FSNAME.$FSNAME ||
8198 error "(4) Fail to create $FSNAME.$FSNAME"
8200 test_103_set_pool $FSNAME OST0000
8202 $LFS setstripe -p $FSNAME $DIR/$tdir/d0 ||
8203 error "(6) Fail to setstripe on $DIR/$tdir/d0"
8208 test_renamefs mylustre
8210 local save_fsname=$FSNAME
8214 test_103_check_pool $save_fsname 7
8216 if [ $OSTCOUNT -ge 2 ]; then
8217 test_103_set_pool $save_fsname OST0001
8220 $LFS setstripe -p $save_fsname $DIR/$tdir/f0 ||
8221 error "(16) Fail to setstripe on $DIR/$tdir/f0"
8230 test_103_check_pool $save_fsname 17
8234 test_renamefs $save_fsname
8240 run_test 103 "rename filesystem name"
8242 test_104a() { # LU-6952
8243 local mds_mountopts=$MDS_MOUNT_OPTS
8244 local ost_mountopts=$OST_MOUNT_OPTS
8245 local mds_mountfsopts=$MDS_MOUNT_FS_OPTS
8246 local lctl_ver=$(do_facet $SINGLEMDS $LCTL --version |
8249 [[ $(version_code $lctl_ver) -lt $(version_code 2.9.55) ]] &&
8250 skip "this test needs utils above 2.9.55"
8252 # specify "acl" in mount options used by mkfs.lustre
8253 if [ -z "$MDS_MOUNT_FS_OPTS" ]; then
8254 MDS_MOUNT_FS_OPTS="acl,user_xattr"
8257 MDS_MOUNT_FS_OPTS="${MDS_MOUNT_FS_OPTS},acl,user_xattr"
8260 echo "mountfsopt: $MDS_MOUNT_FS_OPTS"
8262 #reformat/remount the MDT to apply the MDT_MOUNT_FS_OPT options
8264 if ! combined_mgs_mds ; then
8268 if [ -z "$MDS_MOUNT_OPTS" ]; then
8269 MDS_MOUNT_OPTS="-o noacl"
8271 MDS_MOUNT_OPTS="${MDS_MOUNT_OPTS},noacl"
8274 for num in $(seq $MDSCOUNT); do
8275 start mds$num $(mdsdevname $num) $MDS_MOUNT_OPTS ||
8276 error "Failed to start MDS"
8279 for num in $(seq $OSTCOUNT); do
8280 start ost$num $(ostdevname $num) $OST_MOUNT_OPTS ||
8281 error "Failed to start OST"
8285 setfacl -m "d:$RUNAS_ID:rwx" $MOUNT &&
8286 error "ACL is applied when FS is mounted with noacl."
8288 MDS_MOUNT_OPTS=$mds_mountopts
8289 OST_MOUNT_OPTS=$ost_mountopts
8290 MDS_MOUNT_FS_OPTS=$mds_mountfsopts
8292 run_test 104a "Make sure user defined options are reflected in mount"
8294 test_104b() { # LU-12859
8295 mount_client $MOUNT3 flock,localflock
8296 stack_trap "umount_client $MOUNT3" EXIT
8297 mount | grep "$MOUNT3 .*,flock" && error "flock is still set"
8298 mount | grep "$MOUNT3 .*,localflock" || error "localflock is not set"
8299 umount_client $MOUNT3
8300 mount_client $MOUNT3 localflock,flock
8301 mount | grep "$MOUNT3 .*,localflock" && error "localflock is still set"
8302 mount | grep "$MOUNT3 .*,flock" || error "flock is not set"
8303 umount_client $MOUNT3
8304 mount_client $MOUNT3 localflock,flock,noflock
8305 flock_is_enabled $MOUNT3 && error "some flock is still enabled" || true
8307 run_test 104b "Mount uses last flock argument"
8309 error_and_umount() {
8320 mount --bind $DIR $TMP/$tdir || error "mount bind mnt pt failed"
8321 rm -f $TMP/$tdir/$tfile
8322 rm -f $TMP/$tdir/${tfile}1
8324 # Files should not be created in ro bind mount point
8325 # remounting from rw to ro
8326 mount -o remount,ro $TMP/$tdir ||
8327 error_and_umount "readonly remount of bind mnt pt failed"
8328 touch $TMP/$tdir/$tfile &&
8329 error_and_umount "touch succeeds on ro bind mnt pt"
8330 [ -e $TMP/$tdir/$tfile ] &&
8331 error_and_umount "file created on ro bind mnt pt"
8333 # Files should be created in rw bind mount point
8334 # remounting from ro to rw
8335 mount -o remount,rw $TMP/$tdir ||
8336 error_and_umount "read-write remount of bind mnt pt failed"
8337 touch $TMP/$tdir/${tfile}1 ||
8338 error_and_umount "touch fails on rw bind mnt pt"
8339 [ -e $TMP/$tdir/${tfile}1 ] ||
8340 error_and_umount "file not created on rw bind mnt pt"
8341 umount $TMP/$tdir || error "umount of bind mnt pt failed"
8343 cleanup || error "cleanup failed with $?"
8345 run_test 105 "check file creation for ro and rw bind mnt pt"
8352 mkdir -p $DIR/$tdir || error "create $tdir failed"
8353 do_nodes $(comma_list $(osts_nodes)) $LCTL set_param \
8354 seq.*OST*-super.width=$DATA_SEQ_MAX_WIDTH
8355 lfs setstripe -c 1 -i 0 $DIR/$tdir
8356 #define OBD_FAIL_CAT_RECORDS 0x1312
8357 do_facet mds1 $LCTL set_param fail_loc=0x1312 fail_val=$repeat
8359 for ((i = 1; i <= $repeat; i++)); do
8361 #one full plain llog
8362 createmany -o $DIR/$tdir/f- 64768
8364 createmany -u $DIR/$tdir/f- 64768
8366 wait_delete_completed $((TIMEOUT * 7))
8367 #ASSERTION osp_sync_thread() ( thread->t_flags != SVC_RUNNING ) failed
8368 #shows that osp code is buggy
8369 do_facet mds1 $LCTL set_param fail_loc=0 fail_val=0
8373 run_test 106 "check osp llog processing when catalog is wrapped"
8376 [[ "$MDS1_VERSION" -ge $(version_code 2.10.50) ]] ||
8377 skip "Need MDS version > 2.10.50"
8380 start_mgsmds || error "start_mgsmds failed"
8381 start_ost || error "unable to start OST"
8383 # add unknown configuration parameter.
8384 if [[ $PERM_CMD == *"set_param -P"* ]]; then
8385 cmd="$PERM_CMD ost.$FSNAME-OST0000*.unknown_param"
8387 cmd="$PERM_CMD $FSNAME-OST0000*.ost.unknown_param"
8389 do_facet mgs "$cmd=50"
8390 cleanup_nocli || error "cleanup_nocli failed with $?"
8393 # unknown param should be ignored while mounting.
8394 start_ost || error "unable to start OST after unknown param set"
8396 cleanup || error "cleanup failed with $?"
8398 run_test 107 "Unknown config param should not fail target mounting"
8403 $rcmd rm -rf $tmp > /dev/null 2>&1
8404 $rcmd mkdir -p $tmp/{mnt,images} || error "failed to mkdir remotely"
8406 for facet in $facets; do
8407 [ "$mds1_FSTYPE" = zfs ] &&
8408 $rcmd $ZPOOL -f export lustre-$facet > /dev/null 2>&1
8409 $rcmd mkdir $tmp/mnt/$facet ||
8410 error "failed to mkdir $tmp/mnt/$facet"
8411 $rcmd dd if=/dev/zero of=$tmp/images/$facet \
8412 seek=199 bs=1M count=1 ||
8413 error "failed to create $tmp/images/$facet"
8422 local facet=${role}$((idx + 1))
8424 [ $# -eq 5 ] && pool=$5
8426 do_facet $SINGLEMDS $MKFS --fsname=lustre --$mgs \
8427 --$role --index=$idx --replace --backfstype=$bkfs \
8428 --device-size=200000 --reformat $pool $tmp/images/$facet ||
8429 error "failed to mkfs for $facet"
8433 echo "mounting client..."
8434 mount -t lustre ${nid}:/lustre $MOUNT ||
8435 error "failed to mount lustre"
8438 ls -l $MOUNT/local_dir || error "failed to list"
8440 echo "check truncate && write"
8441 echo "dummmmmmmmmmmmm" > $MOUNT/remote_dir/fsx.c ||
8442 error "failed to tuncate & write"
8445 touch $MOUNT/foooo ||
8446 error "failed to create"
8448 echo "check read && write && append"
8449 sha1sum $MOUNT/conf-sanity.sh |
8450 awk '{ print $1 }' > $MOUNT/checksum.new ||
8451 error "failed to read(1)"
8452 sha1sum $MOUNT/remote_dir/unlinkmany.c |
8453 awk '{ print $1 }' >> $MOUNT/checksum.new ||
8454 error "failed to read(2)"
8455 sha1sum $MOUNT/striped_dir/lockahead_test.o |
8456 awk '{ print $1 }' >> $MOUNT/checksum.new ||
8457 error "failed to read(3)"
8460 diff $MOUNT/checksum.new $MOUNT/checksum.src ||
8461 error "failed to verify data"
8471 umount -f $MOUNT || error "failed to umount client"
8472 for facet in $facets; do
8473 $rcmd umount -f $tmp/mnt/$facet ||
8474 error "failed to umount $facet"
8475 if [ "$mds1_FSTYPE" = zfs ]; then
8476 $rcmd $ZPOOL export -f lustre-$facet ||
8477 error "failed to export lustre-$facet"
8481 $rcmd rm -rf $tmp || error "failed to rm the dir $tmp"
8485 [ "$CLIENTONLY" ] && skip "Client-only testing"
8486 [ "$mds1_FSTYPE" != zfs ] && skip "zfs only test"
8487 [ "$MDS1_VERSION" -lt $(version_code 2.10.58) ] &&
8488 skip "Need server version at least 2.10.58"
8493 local tmp=$TMP/$tdir
8494 local rcmd="do_facet $SINGLEMDS"
8495 local facets="mdt1 mdt2 ost1 ost2"
8496 local nid=$($rcmd $LCTL list_nids | head -1)
8499 trap t_108_cleanup EXIT ERR
8502 t_108_mkfs mdt 0 zfs mgs lustre-mdt1/mdt1
8503 t_108_mkfs mdt 1 zfs mgsnode=$nid lustre-mdt2/mdt2
8504 t_108_mkfs ost 0 zfs mgsnode=$nid lustre-ost1/ost1
8505 t_108_mkfs ost 1 zfs mgsnode=$nid lustre-ost2/ost2
8507 for facet in $facets; do
8508 $rcmd zfs set mountpoint=$tmp/mnt/$facet canmount=on \
8509 lustre-$facet/$facet ||
8510 error "failed to zfs set for $facet (1)"
8511 $rcmd zfs mount lustre-$facet/$facet ||
8512 error "failed to local mount $facet"
8513 $rcmd tar jxf $LUSTRE/tests/ldiskfs_${facet}_2_11.tar.bz2 \
8514 --xattrs --xattrs-include="trusted.*" \
8515 -C $tmp/mnt/$facet/ > /dev/null 2>&1 ||
8516 error "failed to untar image for $facet"
8517 $rcmd "cd $tmp/mnt/$facet && rm -rf oi.* OI_* lfsck_* LFSCK" ||
8518 error "failed to cleanup for $facet"
8519 $rcmd zfs umount lustre-$facet/$facet ||
8520 error "failed to local umount $facet"
8521 $rcmd zfs set canmount=off lustre-$facet/$facet ||
8522 error "failed to zfs set $facet (2)"
8525 echo "changing server nid..."
8526 $rcmd mount -t lustre -o nosvc lustre-mdt1/mdt1 $tmp/mnt/mdt1
8527 $rcmd lctl replace_nids $FSNAME-MDT0000 $nid
8528 $rcmd lctl replace_nids $FSNAME-MDT0001 $nid
8529 $rcmd lctl replace_nids $FSNAME-OST0000 $nid
8530 $rcmd lctl replace_nids $FSNAME-OST0001 $nid
8531 $rcmd umount $tmp/mnt/mdt1
8533 for facet in $facets; do
8534 echo "mounting $facet from backup..."
8535 $rcmd mount -t lustre -o abort_recov lustre-$facet/$facet \
8536 $tmp/mnt/$facet || error "failed to mount $facet"
8539 # ZFS backend can detect migration and trigger OI scrub automatically
8540 # sleep 3 seconds for scrub done
8546 run_test 108a "migrate from ldiskfs to ZFS"
8549 [ "$CLIENTONLY" ] && skip "Client-only testing"
8550 [ "$mds1_FSTYPE" != ldiskfs ] && skip "ldiskfs only test"
8551 [ "$MDS1_VERSION" -lt $(version_code 2.10.58) ] &&
8552 skip "Need server version at least 2.10.58"
8557 local tmp=$TMP/$tdir
8558 local rcmd="do_facet $SINGLEMDS"
8559 local facets="mdt1 mdt2 ost1 ost2"
8560 local scrub_list="MDT0000 MDT0001 OST0000 OST0001"
8561 local nid=$($rcmd $LCTL list_nids | head -1)
8564 trap t_108_cleanup EXIT ERR
8567 t_108_mkfs mdt 0 ldiskfs mgs
8568 t_108_mkfs mdt 1 ldiskfs mgsnode=$nid
8569 t_108_mkfs ost 0 ldiskfs mgsnode=$nid
8570 t_108_mkfs ost 1 ldiskfs mgsnode=$nid
8572 for facet in $facets; do
8573 $rcmd mount -t ldiskfs -o loop $tmp/images/$facet \
8575 error "failed to local mount $facet"
8577 $rcmd tar jxf $LUSTRE/tests/zfs_${facet}_2_11.tar.bz2 \
8578 --xattrs --xattrs-include="*.*" \
8579 -C $tmp/mnt/$facet/ > /dev/null 2>&1 ||
8580 error "failed to untar image for $facet"
8581 $rcmd "cd $tmp/mnt/$facet && rm -rf oi.* OI_* lfsck_* LFSCK" ||
8582 error "failed to cleanup for $facet"
8583 $rcmd umount $tmp/mnt/$facet ||
8584 error "failed to local umount $facet"
8587 echo "changing server nid..."
8588 $rcmd mount -t lustre -o nosvc,loop $tmp/images/mdt1 $tmp/mnt/mdt1
8589 $rcmd lctl replace_nids $FSNAME-MDT0000 $nid
8590 $rcmd lctl replace_nids $FSNAME-MDT0001 $nid
8591 $rcmd lctl replace_nids $FSNAME-OST0000 $nid
8592 $rcmd lctl replace_nids $FSNAME-OST0001 $nid
8593 $rcmd umount $tmp/mnt/mdt1
8595 for facet in $facets; do
8596 echo "mounting $facet from backup..."
8597 $rcmd mount -t lustre -o loop,abort_recov $tmp/images/$facet \
8598 $tmp/mnt/$facet || error "failed to mount $facet"
8601 for facet in $scrub_list; do
8602 $rcmd $LCTL lfsck_start -M $FSNAME-$facet -t scrub ||
8603 error "failed to start OI scrub on $facet"
8606 # sleep 3 seconds for scrub done
8612 run_test 108b "migrate from ZFS to ldiskfs"
8615 # set number of permanent parameters
8617 test_109_set_params() {
8620 set_persistent_param_and_check mds \
8621 "mdd.$fsname-MDT0000.atime_diff" \
8622 "$fsname-MDT0000.mdd.atime_diff" \
8624 set_persistent_param_and_check mds \
8625 "mdd.$fsname-MDT0000.atime_diff" \
8626 "$fsname-MDT0000.mdd.atime_diff" \
8628 set_persistent_param_and_check client \
8629 "llite.$fsname*.max_read_ahead_mb" \
8630 "$fsname.llite.max_read_ahead_mb" \
8632 set_persistent_param_and_check client \
8633 "llite.$fsname*.max_read_ahead_mb" \
8634 "$fsname.llite.max_read_ahead_mb" \
8636 create_pool $fsname.pool1 || error "create pool failed"
8637 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
8638 error "pool_add failed"
8639 do_facet mgs $LCTL pool_remove $fsname.pool1 OST0000 ||
8640 error "pool_remove failed"
8641 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
8642 error "pool_add failed"
8646 # check permanent parameters
8648 test_109_test_params() {
8651 local atime_diff=$(do_facet mds $LCTL \
8652 get_param -n mdd.$fsname-MDT0000.atime_diff)
8653 [ $atime_diff == 63 ] || error "wrong mdd parameter after clear_conf"
8654 local max_read_ahead_mb=$(do_facet client $LCTL \
8655 get_param -n llite.$fsname*.max_read_ahead_mb)
8656 [ $max_read_ahead_mb == 64 ] ||
8657 error "wrong llite parameter after clear_conf"
8658 local ost_in_pool=$(do_facet mds $LCTL pool_list $fsname.pool1 |
8659 grep -v "^Pool:" | sed 's/_UUID//')
8660 [ $ost_in_pool = "$fsname-OST0000" ] ||
8661 error "wrong pool after clear_conf"
8665 # run lctl clear_conf, store CONFIGS before and after that
8667 test_109_clear_conf()
8669 local clear_conf_arg=$1
8672 if ! combined_mgs_mds ; then
8673 mgsdev=$(mgsdevname)
8674 stop_mgs || error "stop_mgs failed"
8675 start_mgs "-o nosvc" || error "start_mgs nosvc failed"
8677 mgsdev=$(mdsdevname 1)
8678 start_mdt 1 "-o nosvc" || error "start_mdt 1 nosvc failed"
8681 do_facet mgs "rm -rf $TMP/${tdir}/conf1; mkdir -p $TMP/${tdir}/conf1;" \
8682 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf1\\\" \
8686 # the command being tested
8688 do_facet mgs $LCTL clear_conf $clear_conf_arg ||
8689 error "clear_conf failed"
8690 if ! combined_mgs_mds ; then
8691 stop_mgs || error "stop_mgs failed"
8693 stop_mdt 1 || error "stop_mdt 1 failed"
8696 do_facet mgs "rm -rf $TMP/${tdir}/conf2; mkdir -p $TMP/${tdir}/conf2;" \
8697 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf2\\\" \
8701 test_109_file_shortened() {
8703 local sizes=($(do_facet mgs "stat -c %s " \
8704 "$TMP/${tdir}/conf1/CONFIGS/$file" \
8705 "$TMP/${tdir}/conf2/CONFIGS/$file"))
8706 [ ${sizes[1]} -lt ${sizes[0]} ] && return 0
8712 [ $MDS1_VERSION -lt $(version_code 2.10.59) ] &&
8713 skip "Needs MDS version 2.10.59 or later."
8714 [ "$(facet_fstype mgs)" == zfs ] &&
8715 skip "LU-8727: no implementation for ZFS"
8720 client_up || error "client_up failed"
8723 # set number of permanent parameters
8725 test_109_set_params $FSNAME
8727 umount_client $MOUNT || error "umount_client failed"
8728 stop_ost || error "stop_ost failed"
8729 stop_mds || error "stop_mds failed"
8731 test_109_clear_conf $FSNAME
8733 # make sure that all configs are cleared
8735 test_109_file_shortened $FSNAME-MDT0000 ||
8736 error "failed to clear MDT0000 config"
8737 test_109_file_shortened $FSNAME-client ||
8738 error "failed to clear client config"
8743 # check that configurations are intact
8745 test_109_test_params $FSNAME
8750 destroy_test_pools || error "destroy test pools failed"
8754 run_test 109a "test lctl clear_conf fsname"
8758 [ $MDS1_VERSION -lt $(version_code 2.10.59) ] &&
8759 skip "Needs MDS version 2.10.59 or later."
8760 [ "$(facet_fstype mgs)" == zfs ] &&
8761 skip "LU-8727: no implementation for ZFS"
8766 client_up || error "client_up failed"
8769 # set number of permanent parameters
8771 test_109_set_params $FSNAME
8773 umount_client $MOUNT || error "umount_client failed"
8774 stop_ost || error "stop_ost failed"
8775 stop_mds || error "stop_mds failed"
8777 test_109_clear_conf $FSNAME-MDT0000
8779 # make sure that only one config is cleared
8781 test_109_file_shortened $FSNAME-MDT0000 ||
8782 error "failed to clear MDT0000 config"
8783 test_109_file_shortened $FSNAME-client &&
8784 error "failed to clear client config"
8788 # check that configurations are intact
8790 test_109_test_params $FSNAME
8795 destroy_test_pools || error "destroy test pools failed"
8799 run_test 109b "test lctl clear_conf one config"
8803 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
8804 skip "Only applicable to ldiskfs-based MDTs"
8806 do_facet $SINGLEMDS $DEBUGFS -w -R supported_features |grep large_dir ||
8807 skip "large_dir option is not supported on MDS"
8808 do_facet ost1 $DEBUGFS -w -R supported_features | grep large_dir ||
8809 skip "large_dir option is not supported on OSS"
8811 stopall # stop all targets before modifying the target counts
8812 local old_mdscount=$MDSCOUNT
8813 local old_ostcount=$OSTCOUNT
8815 stack_trap "MDSCOUNT=$old_mdscount OSTCOUNT=$old_ostcount; reformat"
8819 # ext4_dir_entry_2 struct size:264
8820 # dx_root struct size:8
8821 # dx_node struct size:8
8822 # dx_entry struct size:8
8823 # For 1024 bytes block size.
8824 # First level directory entries: 126
8825 # Second level directory entries: 127
8826 # Entries in leaf: 3
8827 # For 2 levels limit: 48006
8828 # For 3 levels limit : 6096762
8829 # Create 80000 files to safely exceed 2-level htree limit.
8830 CONF_SANITY_110_LINKS=${CONF_SANITY_110_LINKS:-80000}
8832 # can fit at most 3 filenames per 1KB leaf block, but each
8833 # leaf/index block will only be 3/4 full before split at each level
8834 (( MDSSIZE < CONF_SANITY_110_LINKS / 3 * 4/3 * 4/3 )) &&
8835 CONF_SANITY_110_LINKS=$((MDSSIZE * 3 * 3/4 * 3/4))
8837 combined_mgs_mds || replace=" --replace "
8838 local opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
8839 $replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
8840 if [[ $opts != *large_dir* ]]; then
8841 if [[ $opts != *mkfsoptions* ]]; then
8842 opts+=" --mkfsoptions=\\\"-O large_dir -b 1024 -i 65536\\\""
8844 opts="${opts//--mkfsoptions=\\\"/ \
8845 --mkfsoptions=\\\"-O large_dir -b 1024 -i 65536 }"
8848 echo "MDT params: $opts"
8850 combined_mgs_mds || start_mgs
8851 add mds1 $opts || error "add mds1 failed with new params"
8852 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
8854 opts="$(mkfs_opts ost1 $(ostdevname 1)) \
8855 $replace --reformat $(ostdevname 1) $(ostvdevname 1)"
8857 if [[ $opts != *large_dir* ]]; then
8858 if [[ $opts != *mkfsoptions* ]]; then
8859 opts+=" --mkfsoptions=\\\"-O large_dir\\\" "
8861 opts="${opts//--mkfsoptions=\\\"/ \
8862 --mkfsoptions=\\\"-O large_dir }"
8865 echo "OST params: $opts"
8866 add ost1 $opts || error "add ost1 failed with new params"
8867 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
8869 MOUNT_2=yes mountcli || error "mount clients failed"
8871 mkdir -v $DIR/$tdir || error "cannot create $DIR/$tdir"
8872 local pids count=0 group=0
8874 echo "creating $CONF_SANITY_110_LINKS in total"
8875 while (( count < CONF_SANITY_110_LINKS )); do
8876 local len=$((253 - $(wc -c <<<"$tfile-$group-40000-")))
8877 local dir=DIR$((group % 2 + 1))
8878 local target=${!dir}/$tdir/$tfile-$group
8879 local long=$target-$(generate_name $len)-
8880 local create=$((CONF_SANITY_110_LINKS - count))
8882 (( create > 40000 )) && create=40000
8883 touch $target || error "creating $target failed"
8884 echo "creating $create hard links to $target"
8885 createmany -l $target $long $create &
8888 count=$((count + create))
8889 group=$((group + 1))
8891 echo "waiting for PIDs$pids to complete"
8892 wait $pids || error "createmany failed after $group groups"
8894 umount_client $MOUNT2 -f
8897 run_e2fsck $(facet_active_host mds1) $(mdsdevname 1) -n
8898 MDSCOUNT=$old_mdscount
8899 OSTCOUNT=$old_ostcount
8901 run_test 110 "Adding large_dir with 3-level htree"
8904 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
8905 skip "Only applicable to ldiskfs-based MDTs"
8907 is_dm_flakey_dev $SINGLEMDS $(mdsdevname 1) &&
8908 skip "This test can not be executed on flakey dev"
8910 do_facet $SINGLEMDS $DEBUGFS -w -R supported_features |grep large_dir ||
8911 skip "large_dir option is not supported on MDS"
8913 do_facet ost1 $DEBUGFS -w -R supported_features | grep large_dir ||
8914 skip "large_dir option is not supported on OSS"
8916 # cleanup before changing target counts
8918 local old_mdscount=$MDSCOUNT
8919 local old_ostcount=$OSTCOUNT
8920 local old_mdssize=$MDSSIZE
8922 stack_trap "MDSSIZE=$old_mdssize MDSCOUNT=$old_mdscount \
8923 OSTCOUNT=$old_ostcount; reformat"
8926 (( MDSSIZE < 2400000 )) && MDSSIZE=2400000 # need at least 2.4GB
8928 local mdsdev=$(mdsdevname 1)
8929 combined_mgs_mds || replace=" --replace "
8930 local opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
8931 $replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
8932 if [[ $opts != *large_dir* ]]; then
8933 if [[ $opts != *mkfsoptions* ]]; then
8934 opts+=" --mkfsoptions=\\\"-O large_dir -i 1048576 \\\" "
8936 opts="${opts//--mkfsoptions=\\\"/ \
8937 --mkfsoptions=\\\"-O large_dir -i 1048576 }"
8940 echo "MDT params: $opts"
8942 combined_mgs_mds || start_mgs
8943 __touch_device mds 1
8944 add mds1 $opts || error "add mds1 failed with new params"
8945 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
8947 opts="$(mkfs_opts ost1 $(ostdevname 1)) \
8948 $replace --reformat $(ostdevname 1) $(ostvdevname 1)"
8949 if [[ $opts != *large_dir* ]]; then
8950 if [[ $opts != *mkfsoptions* ]]; then
8951 opts+=" --mkfsoptions=\\\"-O large_dir \\\""
8953 opts="${opts//--mkfsoptions=\\\"/ --mkfsoptions=\\\"-O large_dir }"
8956 echo "OST params: $opts"
8957 __touch_device ost 1
8958 add ost1 $opts || error "add ost1 failed with new params"
8959 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
8961 MOUNT_2=yes mountcli
8962 mkdir $DIR/$tdir || error "cannot create $DIR/$tdir"
8964 lfs df -i $DIR/$tdir
8968 local start=$SECONDS
8970 local dirmax=$((2 << 30))
8976 while (( !needskip & dirsize < dirmax )); do
8979 for cli in ${CLIENTS//,/ }; do
8980 local len=$((253 - $(wc -c <<<"$cli-$group-60000-")))
8981 local target=$cli-$group
8982 local long=$DIR/$tdir/$target-$(generate_name $len)-
8984 RPWD=$DIR/$tdir do_node $cli touch $target ||
8985 error "creating $target failed"
8986 echo "creating 60000 hardlinks to $target"
8987 RPWD=$DIR/$tdir do_node $cli createmany -l $target $long 60000 &
8990 group=$((group + 1))
8992 long=$DIR2/$tdir/$target-$(generate_name $len)-
8994 RPWD=$DIR2/$tdir do_node $cli touch $target ||
8995 error "creating $target failed"
8996 echo "creating 60000 hardlinks to $target"
8997 RPWD=$DIR2/$tdir do_node $cli createmany -l $target $long 60000 &
9000 group=$((group + 1))
9002 echo "waiting for PIDs$pids to complete"
9003 wait $pids || error "createmany failed after $group groups"
9004 dirsize=$(stat -c %s $DIR/$tdir)
9005 taken=$((SECONDS - start))
9006 rate=$((dirsize / taken))
9007 left=$(((dirmax - dirsize) / rate))
9008 num=$((group * 60000))
9009 echo "estimate ${left}s left after $num files / ${taken}s"
9010 # if the estimated time remaining is too large (it may change
9011 # over time as the create rate is not constant) then exit
9012 # without declaring a failure.
9013 (( left > 1200 )) && needskip=1
9016 umount_client $MOUNT2 -f
9019 ! (( $needskip )) ||
9020 echo "ETA ${left}s after $num files / ${taken}s is too long"
9022 run_e2fsck $(facet_active_host mds1) $(mdsdevname 1) -n
9023 MDSCOUNT=$old_mdscount
9024 OSTCOUNT=$old_ostcount
9025 MDSSIZE=$old_mdssize
9027 run_test 111 "Adding large_dir with over 2GB directory"
9030 start_mds || error "MDS start failed"
9031 start_ost || error "OSS start failed"
9032 echo "start ost2 service on $(facet_active_host ost2)"
9033 start ost2 $(ostdevname 2) $(csa_add "$OST_MOUNT_OPTS" -o no_precreate) ||
9034 error "start ost2 facet failed"
9035 local val=$(do_facet ost2 \
9036 "$LCTL get_param -n obdfilter.$FSNAME-OST0001*.no_precreate")
9037 (( $val == 1 )) || error "obdfilter.$FSNAME-OST0001*.no_precreate=$val"
9039 mount_client $MOUNT || error "mount client failed"
9040 wait_osc_import_state mds1 ost1 FULL
9041 wait_osc_import_ready client ost1
9042 wait_osc_import_ready client ost2
9044 $LFS setstripe -i 0 $DIR/$tfile.0 ||
9045 error "problem creating $tfile.0 on OST0000"
9046 $LFS setstripe -i 1 $DIR/$tfile.1 && $LFS getstripe $DIR/$tfile.1 &&
9047 (( $($LFS getstripe -i $DIR/$tfile.1) == 1 )) &&
9048 error "allowed to create $tfile.1 on OST0001"
9049 do_facet ost2 $LCTL set_param obdfilter.*.no_precreate=0
9051 $LFS setstripe -i 1 $DIR/$tfile.2 ||
9052 error "failed to create $tfile.2 on ost1 facet"
9053 # files not cleaned with ONLY_REPEAT because of client unmount below
9054 rm -f $DIR/$tfile.[012]
9055 stop_ost2 || error "stop ost2 facet failed"
9058 run_test 112 "mount OST with nocreate option"
9061 SAVE_MGS_MOUNT_OPTS=$MGS_MOUNT_OPTS
9062 SAVE_MDS_MOUNT_OPTS=$MDS_MOUNT_OPTS
9063 SAVE_OST_MOUNT_OPTS=$OST_MOUNT_OPTS
9069 MGS_MOUNT_OPTS=$SAVE_MGS_MOUNT_OPTS
9070 MDS_MOUNT_OPTS=$SAVE_MDS_MOUNT_OPTS
9071 OST_MOUNT_OPTS=$SAVE_OST_MOUNT_OPTS
9072 # Revert old mount options back
9074 # Subsequent following test requires
9075 # conf-sanity to be in stopall state.
9076 # Force 'stopall' so others following
9081 # Error out with mount info
9083 local server_nodes=$(comma_list $(mdts_nodes) $(osts_nodes))
9086 echo "--Client Mount Info--"
9087 mount | grep -i lustre
9088 echo "--Server Mount Info--"
9089 do_nodes $server_nodes mount | grep -i lustre
9095 local ost_version="2.15.51" # Minimum version required
9097 (( OST1_VERSION >= $(version_code $ost_version) )) ||
9098 skip "Need server version at least $ost_version"
9100 stack_trap cleanup_113 EXIT
9102 # Reset before starting
9106 # Verify MDS's should start with "rw"
9107 do_facet $SINGLEMDS mount | grep "lustre.*rw,.*MDT" ||
9108 error_113 "$SINGLEMDS should be read-write"
9110 # Verify OST's should start with "rw"
9111 for (( i=1; i <= OSTCOUNT; i++ )); do
9112 do_facet ost$i mount | grep "lustre.*rw,.*OST" ||
9113 error_113 "ost$i should be read-write"
9116 # rdonly_dev does not currently work for ldiskfs
9117 # We skip the rdonly_dev check until then.
9118 if [[ $ost1_FSTYPE == ldiskfs ]]; then
9119 echo "Shadow Mountpoint correctly reports rw for ldiskfs"
9124 # Only ZFS specific tests below.
9127 # Must stop all (server+client) and restart to verify new
9131 # add rdonly_dev to mount option
9132 MGS_MOUNT_OPTS=$(csa_add "$MGS_MOUNT_OPTS" -o rdonly_dev)
9133 MDS_MOUNT_OPTS=$(csa_add "$MDS_MOUNT_OPTS" -o rdonly_dev)
9134 OST_MOUNT_OPTS=$(csa_add "$OST_MOUNT_OPTS" -o rdonly_dev)
9136 # Only restart server(mds/ost). Sufficient for test
9137 setupall server_only || error "Fail to start servers"
9139 # Verify MDS's should be "ro"
9140 do_facet $SINGLEMDS mount | grep "lustre.*ro,.*MDT.*rdonly_dev" ||
9141 error_113 "$SINGLEMDS should be read-only"
9143 # Verify OST's should be "ro"
9144 for (( i=1; i <= OSTCOUNT; i++ )); do
9145 do_facet ost$i mount | grep "lustre.*ro,.*OST.*rdonly_dev" ||
9146 error_113 "ost$i should be read-only"
9149 run_test 113 "Shadow mountpoint correctly report ro/rw for mounts"
9152 # Purpose: To verify dynamic thread (OSS) creation.
9153 # (This was sanity/115)
9156 [ $PARALLEL == "yes" ] && skip "skip parallel run"
9157 remote_ost_nodsh && skip "remote OST with nodsh"
9159 # Lustre does not stop service threads once they are started.
9160 # Reset number of running threads to default.
9165 local save_params="$TMP/sanity-$TESTNAME.parameters"
9167 # Get ll_ost_io count before I/O
9168 OSTIO_pre=$(do_facet ost1 \
9169 "$LCTL get_param ost.OSS.ost_io.threads_started | cut -d= -f2")
9170 # Exit if lustre is not running (ll_ost_io not running).
9171 [ -z "$OSTIO_pre" ] && error "no OSS threads"
9173 echo "Starting with $OSTIO_pre threads"
9174 local thread_max=$((OSTIO_pre * 2))
9175 local rpc_in_flight=$((thread_max * 2))
9177 # this is limited to OSC_MAX_RIF_MAX (256)
9178 [ $rpc_in_flight -gt 256 ] && rpc_in_flight=256
9179 thread_max=$((rpc_in_flight / 2))
9180 [ $thread_max -le $OSTIO_pre ] && skip "Too many ost_io threads" &&
9183 # Number of I/O Process proposed to be started.
9185 local facets=$(get_facets OST)
9187 save_lustre_params client "osc.*OST*.max_rpcs_in_flight" > $save_params
9188 save_lustre_params $facets "ost.OSS.ost_io.threads_max" >> $save_params
9190 # Set in_flight to $rpc_in_flight
9191 $LCTL set_param osc.*OST*.max_rpcs_in_flight=$rpc_in_flight ||
9192 error "Failed to set max_rpcs_in_flight to $rpc_in_flight"
9193 nfiles=${rpc_in_flight}
9194 # Set ost thread_max to $thread_max
9195 do_facet ost1 "$LCTL set_param ost.OSS.ost_io.threads_max=$thread_max"
9197 # 5 Minutes should be sufficient for max number of OSS
9198 # threads(thread_max) to be created.
9202 local wtl=${WTL:-"$LUSTRE/tests/write_time_limit"}
9204 test_mkdir $DIR/$tdir
9205 for ((i = 1; i <= nfiles; i++)); do
9206 local file=$DIR/$tdir/${tfile}-$i
9208 $LFS setstripe -c -1 -i 0 $file
9209 ($wtl $file $timeout)&
9212 # I/O Started - Wait for thread_started to reach thread_max or report
9213 # error if thread_started is more than thread_max.
9214 echo "Waiting for thread_started to reach thread_max"
9215 local thread_started=0
9216 local end_time=$((SECONDS + timeout))
9218 while [ $SECONDS -le $end_time ] ; do
9220 # Get ost i/o thread_started count.
9221 thread_started=$(do_facet ost1 \
9223 ost.OSS.ost_io.threads_started | cut -d= -f2")
9224 # Break out if thread_started is equal/greater than thread_max
9225 if (( $thread_started >= $thread_max )); then
9226 echo ll_ost_io thread_started $thread_started, \
9227 equal/greater than thread_max $thread_max
9233 # Cleanup - We have the numbers, Kill i/o jobs if running.
9234 jobcount=($(jobs -p))
9236 for ((i=0; i < ${#jobcount[*]}; i++)); do
9237 kill -9 ${jobcount[$i]}
9238 if [ $? -ne 0 ] ; then
9239 echo "warning: cannot kill WTL pid ${jobcount[$i]}"
9243 # Cleanup files left by WTL binary.
9244 for ((i = 1; i <= nfiles; i++)); do
9245 local file=$DIR/$tdir/${tfile}-$i
9248 if [ $? -ne 0 ] ; then
9249 echo "Warning: Failed to delete file $file"
9253 restore_lustre_params <$save_params
9254 rm -f $save_params || echo "Warning: delete file '$save_params' failed"
9256 # Error out if no new thread has started or Thread started is greater
9258 if (( $thread_started <= $OSTIO_pre ||
9259 $thread_started > $thread_max )); then
9260 error "ll_ost_io: thread_started $thread_started" \
9261 "OSTIO_pre $OSTIO_pre, thread_max $thread_max." \
9262 "No new thread started or thread started greater " \
9266 run_test 114 "verify dynamic thread creation===================="
9272 do_facet mds1 rm -f $1
9276 if [ "$mds1_FSTYPE" != ldiskfs ]; then
9277 skip "Only applicable to ldiskfs-based MDTs"
9279 [ -n "$FILESET" ] && skip "Not functional for FILESET set"
9281 local dbfs_ver=$(do_facet mds1 $DEBUGFS -V 2>&1)
9283 echo "debugfs version: $dbfs_ver"
9284 echo "$dbfs_ver" | egrep -w "1.44.3.wc1|1.44.5.wc1|1.45.2.wc1" &&
9285 skip_env "This version of debugfs doesn't show inode number"
9287 local IMAGESIZE=$((3072 << 30)) # 3072 GiB
9297 do_facet mds1 "hostname; $LCTL dl; mount"
9299 do_facet ost1 "hostname; $LCTL dl; mount"
9300 # We need MDT size 3072GB, because it is smallest
9301 # partition that can store 2B inodes
9302 do_facet mds1 "mkdir -p $TMP/$tdir"
9303 local mdsimgname=$TMP/$tdir/lustre-mdt
9305 do_facet mds1 "rm -f $mdsimgname"
9306 do_facet mds1 "touch $mdsimgname"
9307 stack_trap "cleanup_115 $mdsimgname" EXIT
9308 do_facet mds1 "$TRUNCATE $mdsimgname $IMAGESIZE" ||
9309 skip "Backend FS doesn't support sparse files"
9310 local mdsdev=$(do_facet mds1 "losetup -f")
9312 do_facet mds1 "losetup $mdsdev $mdsimgname"
9314 local mds_opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
9315 --mkfsoptions='-O ea_inode,^resize_inode,meta_bg \
9316 -N 2247484000 -E lazy_itable_init' --device-size=$IMAGESIZE"
9317 add mds1 $mds_opts --mgs --reformat $mdsdev ||
9318 skip_env "format large MDT failed"
9320 local ostdev=$(ostdevname 1)
9322 local opts="$(mkfs_opts ost1 $ostdev) \
9323 --reformat $ostdev $ostdev"
9324 add ost1 $opts || error "add ost1 failed with new params"
9325 start mds1 $mdsdev $MDS_MOUNT_OPTS || error "start MDS failed"
9326 start_ost || error "start OSS failed"
9327 mount_client $MOUNT || error "mount client failed"
9329 mkdir -p $DIR/$tdir || error "mkdir $DIR/$tdir fail"
9330 local goal="/sys/fs/ldiskfs/$(basename $mdsdev)/inode_goal"
9332 # 2147483648 is 0x80000000
9333 do_facet mds1 "echo 2147483648 >> $goal; grep . $goal"
9334 touch $DIR/$tdir/$tfile
9336 # attrs from 1 to 16 go to block, 17th - to inode
9339 for i in {1..17}; do
9340 local nm="trusted.ea$i"
9341 setfattr -n $nm -v $(printf "xattr%0250d" $i) $DIR/$tdir/$tfile
9344 do_facet mds1 "$DEBUGFS -c -R 'stat ROOT/$tdir/$tfile' $mdsdev"
9346 # inode <2147483649> trusted.ea16 (255)
9347 local inode_num=$(do_facet mds1 \
9348 "$DEBUGFS -c -R 'stat ROOT/$tdir/$tfile' $mdsdev" |
9349 awk '/ea17/ { print $2 }' |
9350 sed -e 's/>//' -e 's/<//' -e 's/\"//')
9351 echo "inode num: $inode_num"
9352 [ $inode_num -ge 2147483648 ] || error "inode $inode_num too small"
9353 do_facet mds1 "losetup -d $mdsdev"
9355 run_test 115 "Access large xattr with inodes number over 2TB"
9358 [ "$mds1_FSTYPE" != ldiskfs ] && skip "ldiskfs only test"
9359 [ "$MDS1_VERSION" -lt $(version_code 2.10.59) ] &&
9360 skip "Need server version at least 2.10.59"
9361 do_facet $SINGLEMDS which mkfs.xfs ||
9362 skip_env "No mkfs.xfs installed"
9367 local tmpmnt=$TMP/$tdir
9368 local mdtimg=$tfile-mdt0
9370 do_facet $SINGLEMDS mkdir -p $tmpmnt
9371 stack_trap "do_facet $SINGLEMDS rmdir $tmpmnt" EXIT
9373 do_facet $SINGLEMDS touch $TMP/$mdtimg
9374 stack_trap "do_facet $SINGLEMDS rm -f $TMP/$mdtimg" EXIT
9375 do_facet $SINGLEMDS mkfs -t xfs -d file,size=1t,name=$TMP/$mdtimg ||
9376 error "mkfs temporary xfs image"
9378 do_facet $SINGLEMDS mount $TMP/$mdtimg $tmpmnt ||
9379 error "mount temporary xfs image"
9380 stack_trap "do_facet $SINGLEMDS umount $tmpmnt" EXIT
9381 local old_mdssize=$MDSSIZE
9382 local old_mdsisize=$MDSISIZE
9384 MDSSIZE=$((17 * 1024 * 1024 * 1024)) # 17T MDT
9385 MDSISIZE=$((16 << 20))
9386 local opts17t="$(mkfs_opts $SINGLEMDS)"
9388 MDSSIZE=$old_mdssize
9389 MDSISIZE=$old_mdsisize
9391 do_facet $SINGLEMDS $MKFS \
9392 ${opts17t/-E /-E lazy_itable_init,lazy_journal_init,} \
9394 error "failed to mkfs for $tmpmnt/$mdtimg"
9396 do_facet $SINGLEMDS $TUNE2FS -l $tmpmnt/$mdtimg |
9397 grep -qw 'features.*extent' || error "extent should be enabled"
9400 run_test 116 "big size MDT support"
9403 # Call setup only if LustreFS is not mounted
9404 check_mount || setup
9406 do_facet ost1 "$LCTL set_param ost.OSS.ost_io.nrs_policies=fifo"
9407 do_facet ost1 "$LCTL get_param -n ost.OSS.ost_io.nrs_tbf_rule" &&
9408 error "get_param should fail"
9409 cleanup || error "cleanup failed with rc $?"
9411 run_test 117 "lctl get_param return errors properly"
9417 [[ "$MDSCOUNT" -ge 2 ]] || skip "Need more at least 2 MDTs"
9419 had_config=$(do_facet mds1 "$LCTL get_param debug | grep config")
9420 do_facet mds1 "$LCTL set_param debug=+config"
9421 do_facet mds1 "$LCTL clear"
9424 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)" &>/dev/null
9425 # mount after writeconf will make "add osp" added to mdt0 config:
9426 # 53 (224)marker 60 (flags=0x01, v2.5.1.0) lustre-MDT0001 'add osp'
9427 # 54 (080)add_uuid nid=... 0: 1:...
9428 # 55 (144)attach 0:lustre-MDT0001-osp-MDT0000 1:osp 2:...
9429 # 56 (144)setup 0:lustre-MDT0001-osp-MDT0000 1:... 2:...
9430 # 57 (136)modify_mdc_tgts add 0:lustre-MDT0000-mdtlov 1:... 2:1 3:1
9431 # duplicate modify_mdc_tgts caused crashes
9434 # using larger debug_mb size to avoid lctl dk log truncation
9435 size_mb=$((DEBUG_SIZE_SAVED * 4))
9438 # though config processing stops after failed attach and setup
9439 # it will proceed after the failed command after each writeconf
9440 # this is the original scenario of the issue
9441 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)" &>/dev/null
9442 do_facet mds1 "$LCTL set_param debug_mb=$size_mb"
9445 wait_update_facet_cond mds1 \
9446 "$LCTL dk | grep -c Processed.log.$FSNAME-MDT0000" \
9451 [[ -z "$had_config" ]] && do_facet mds1 lctl set_param debug=-config
9455 run_test 119 "writeconf on slave mdt shouldn't duplicate mdc/osp and crash"
9457 test_120() { # LU-11130
9458 [ "$MDSCOUNT" -lt 2 ] && skip "mdt count < 2"
9459 [ "$mds1_FSTYPE" != ldiskfs ] &&
9460 skip "ldiskfs only test"
9461 [ "$MDS1_VERSION" -lt $(version_code 2.11.56) ] &&
9462 skip "Need DNE2 capable MD target with LU-11130 fix"
9466 local mds1host=$(facet_active_host mds1)
9467 local mds1dev=$(mdsdevname 1)
9469 $LFS mkdir -i 1 $DIR/$tdir
9470 $LFS mkdir -i 0 $DIR/$tdir/mds1dir
9472 ln -s foo $DIR/$tdir/bar
9473 mv $DIR/$tdir/bar $DIR/$tdir/mds1dir/bar2 ||
9474 error "cross-target rename failed"
9478 run_e2fsck $mds1host $mds1dev "-n"
9480 run_test 120 "cross-target rename should not create bad symlinks"
9484 start_mgsmds || error "MGS MDS Start failed"
9486 stop_mds || error "Stopping MDSes failed"
9490 stop_mds || error "Stopping MDSes failed"
9492 run_test 121 "failover MGS"
9495 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
9496 [[ "$OST1_VERSION" -ge $(version_code 2.11.53) ]] ||
9497 skip "Need OST version at least 2.11.53"
9500 LOAD_MODULES_REMOTE=true load_modules
9501 #define OBD_FAIL_OFD_SET_OID 0x1e0
9502 do_facet ost1 $LCTL set_param fail_loc=0x00001e0
9505 $LFS mkdir -i1 -c1 $DIR/$tdir
9506 $LFS setstripe -i0 -c1 $DIR/$tdir
9507 do_facet ost1 $LCTL set_param fail_loc=0
9508 createmany -o $DIR/$tdir/file_ 1000 ||
9509 error "Fail to create a new sequence"
9513 run_test 122a "Check OST sequence update"
9516 remote_mgs_nodsh && skip "remote MGS with nodsh"
9517 [ -d $MOUNT/.lustre ] || setup
9519 # test old logid format until removal from llog_ioctl.c::str2logid()
9520 if (( $MGS_VERSION < $(version_code 2.99.53) )); then
9521 do_facet mgs $LCTL dl | grep MGS
9522 do_facet mgs "$LCTL --device %MGS llog_print \
9523 \\\\\\\$$FSNAME-client 1 10" ||
9524 error "old llog_print failed"
9527 # test new logid format
9528 if [ $MGS_VERSION -ge $(version_code 2.9.53) ]; then
9529 do_facet mgs "$LCTL --device MGS llog_print $FSNAME-client" ||
9530 error "new llog_print failed"
9533 run_test 123aa "llog_print works with FIDs and simple names"
9536 remote_mgs_nodsh && skip "remote MGS with nodsh"
9537 [[ $MGS_VERSION -gt $(version_code 2.11.51) ]] ||
9538 skip "Need server with working llog_print support"
9540 [ -d $MOUNT/.lustre ] || setup
9546 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9547 mgs_arg="--device MGS"
9549 orig_val=$(do_facet mgs $LCTL get_param jobid_name)
9550 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME"
9552 yaml=$(do_facet mgs $LCTL $mgs_arg llog_print params |
9553 grep jobid_name | tail -n 1)
9555 local param=$(awk '{ print $10 }' <<< "$yaml")
9556 local val=$(awk '{ print $12 }' <<< "$yaml")
9557 #return to the default
9558 do_facet mgs $LCTL set_param -P jobid_name=$orig_val
9559 [ $val = "TESTNAME" ] || error "bad value: $val"
9560 [ $param = "jobid_name," ] || error "Bad param: $param"
9562 run_test 123ab "llog_print params output values from set_param -P"
9564 test_123ac() { # LU-11566
9565 remote_mgs_nodsh && skip "remote MGS with nodsh"
9566 do_facet mgs "$LCTL help llog_print" 2>&1 | grep -q -- --start ||
9567 skip "Need 'lctl llog_print --start' on MGS"
9573 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9574 mgs_arg="--device MGS"
9576 [ -d $MOUNT/.lustre ] || setup
9578 # - { index: 10, event: add_uuid, nid: 192.168.20.1@tcp(0x20000c0a81401,
9579 # node: 192.168.20.1@tcp }
9580 do_facet mgs $LCTL $mgs_arg \
9581 llog_print --start $start --end $end $FSNAME-client | tr -d , |
9582 while read DASH BRACE INDEX idx EVENT BLAH BLAH BLAH; do
9583 (( idx >= start )) || error "llog_print index $idx < $start"
9584 (( idx <= end )) || error "llog_print index $idx > $end"
9587 run_test 123ac "llog_print with --start and --end"
9589 test_123ad() { # LU-11566
9590 remote_mgs_nodsh && skip "remote MGS with nodsh"
9591 # older versions of lctl may not print all records properly
9592 do_facet mgs "$LCTL help llog_print" 2>&1 | grep -q -- --start ||
9593 skip "Need 'lctl llog_print --start' on MGS"
9595 [ -d $MOUNT/.lustre ] || setup
9597 # append a new record, to avoid issues if last record was cancelled
9598 local old=$($LCTL get_param -n osc.*-OST0000-*.max_dirty_mb | head -1)
9599 do_facet mgs $LCTL conf_param $FSNAME-OST0000.osc.max_dirty_mb=$old
9601 # logid: [0x3:0xa:0x0]:0
9605 local num=$(do_facet mgs $LCTL --device MGS llog_info $FSNAME-client |
9606 awk '/last_index:/ { print $2 - 1 }')
9608 # - { index: 71, event: set_timeout, num: 0x14, param: sys.timeout=20 }
9609 local last=$(do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
9610 tail -1 | awk '{ print $4 }' | tr -d , )
9611 (( last == num )) || error "llog_print only showed $last/$num records"
9613 run_test 123ad "llog_print shows all records"
9615 test_123ae() { # LU-11566
9624 remote_mgs_nodsh && skip "remote MGS with nodsh"
9625 [ -d $MOUNT/.lustre ] || setupall
9627 max=$($LCTL get_param -n osc.*-OST0000-*.max_dirty_mb | head -1)
9628 pgs=$($LCTL get_param -n osc.*-OST0000-*.max_pages_per_rpc | head -1)
9629 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9630 mgs_arg="--device MGS"
9632 if do_facet mgs "$LCTL help llog_cancel" 2>&1| grep -q -- --log_id; then
9633 # save one set_param -P record in case none exist
9635 do_facet mgs $LCTL set_param -P osc.*.max_pages_per_rpc=$pgs
9636 stack_trap "do_facet mgs $LCTL set_param -P -d \
9637 osc.*.max_pages_per_rpc"
9640 orig=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9641 tail -1 | awk '{ print $4 }' | tr -d , )
9642 do_facet mgs $LCTL set_param -P osc.*.max_dirty_mb=$max
9643 do_facet mgs $LCTL $mgs_arg llog_print $log | tail -1 |
9644 grep "parameter: osc.*.max_dirty_mb" ||
9645 error "new set_param -P wasn't stored in params log"
9647 # - { index: 71, event: set_param, device: general,
9648 # param: osc.*.max_dirty_mb, value: 256 }
9649 id=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9650 tail -1 | awk '{ print $4 }' | tr -d , )
9651 do_facet mgs $LCTL $mgs_arg llog_cancel $log --log_idx=$id
9652 local new=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9653 tail -1 | awk '{ print $4 }' | tr -d , )
9654 (( new == orig )) ||
9655 error "new llog_cancel now $new, not at $orig records"
9658 # test old positional parameters for a while still
9659 if [ "$MGS_VERSION" -le $(version_code 3.1.53) ]; then
9662 do_facet mgs $LCTL conf_param \
9663 $FSNAME-OST0000.osc.max_pages_per_rpc=$pgs
9664 stack_trap "do_facet mgs $LCTL conf_param -d \
9665 $FSNAME-OST0000.osc.max_pages_per_rpc"
9667 orig=$(do_facet mgs $LCTL --device MGS llog_print $log |
9668 tail -1 | awk '{ print $4 }' | tr -d , )
9669 do_facet mgs $LCTL conf_param $FSNAME-OST0000.osc.max_dirty_mb=$max
9670 do_facet mgs $LCTL --device MGS llog_print $log |
9671 tail -1 | grep "parameter: osc.max_dirty_mb" ||
9672 error "old conf_param wasn't stored in params log"
9673 do_facet mgs $LCTL --device MGS llog_print $log
9674 # - { index: 71, event: conf_param, device: testfs-OST0000-osc,
9675 # param: osc.max_dirty_mb=256 }
9676 id=$(do_facet mgs $LCTL --device MGS llog_print $log |
9677 tail -1 | awk '{ print $4 }' | tr -d , )
9678 do_facet mgs $LCTL --device MGS llog_cancel $log $id
9679 do_facet mgs $LCTL --device MGS llog_print $log
9680 new=$(do_facet mgs $LCTL --device MGS llog_print $log |
9681 tail -1 | awk '{ print $4 }' | tr -d , )
9682 (( new == orig )) ||
9683 error "old llog_cancel now $new, not at $orig records"
9686 run_test 123ae "llog_cancel can cancel requested record"
9688 test_123af() { #LU-13609
9689 [ "$MGS_VERSION" -ge $(version_code 2.13.54) -a \
9690 "$MDS1_VERSION" -ge $(version_code 2.13.54) ] ||
9691 skip "Need both MGS and MDS version at least 2.13.54"
9693 [ -d $MOUNT/.lustre ] || setupall
9694 stack_trap "do_facet mds1 $LCTL set_param fail_loc=0" EXIT
9704 for device in "MGS" "$FSNAME-MDT0000"; do
9705 cmd="--device $device llog_catlist"
9706 echo "lctl $cmd ..."
9707 if [ "$device" = "MGS" ]; then
9712 orig_clist=($(do_facet $facet $LCTL $cmd | awk '{ print $2 }'))
9713 orig_count=${#orig_clist[@]}
9714 echo "orig_clist: ${orig_clist[*]}"
9716 #define OBD_FAIL_CATLIST 0x131b
9717 #fetch to llog records from the second one
9718 do_facet $facet $LCTL set_param fail_loc=0x131b fail_val=2
9720 new_clist=($(do_facet $facet $LCTL $cmd | awk '{ print $2 }'))
9721 new_count=${#new_clist[@]}
9722 echo "new_clist: ${new_clist[*]}"
9724 [ $new_count -eq $((orig_count - 1)) ] ||
9725 error "$new_count != $orig_count - 1"
9726 for i in $(seq 0 $new_count); do
9728 [ "${orig_clist[$j]}" = "${new_clist[$i]}" ] ||
9729 error "${orig_clist[$j]} != ${new_clist[$i]}"
9731 do_facet mds1 $LCTL set_param fail_loc=0
9735 run_test 123af "llog_catlist can show all config files correctly"
9737 test_123ag() { # LU-15142
9741 remote_mgs_nodsh && skip "remote MGS with nodsh"
9742 (( $MGS_VERSION >= $(version_code 2.14.55) )) ||
9743 skip "Need server version least 2.14.55"
9745 [ -d $MOUNT/.lustre ] || setup
9747 orig_val=$(do_facet mgs $LCTL get_param jobid_name)
9748 stack_trap "do_facet mgs $LCTL set_param -P jobid_name=$orig_val"
9750 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME1"
9751 do_facet mgs $LCTL set_param -P -d jobid_name
9752 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
9754 (( rec == 0 )) || error "parameter was not deleted, check #1"
9755 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME1"
9756 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
9758 (( rec == 1)) || error "parameter is not set"
9759 # usage with ordinary set_param format works too
9760 do_facet mgs $LCTL set_param -P -d jobid_name="ANY"
9761 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
9763 (( rec == 0 )) || error "parameter was not deleted, check #2"
9765 run_test 123ag "llog_print skips values deleted by set_param -P -d"
9767 test_123ah() { #LU-7668 del_ost
9768 (( "$MGS_VERSION" >= $(version_code 2.15.51) &&
9769 "$MDS1_VERSION" >= $(version_code 2.15.51) )) ||
9770 skip "Need MGS/MDS version >= 2.15.51 for del_ost"
9772 [ -d $MOUNT/.lustre ] || setupall
9773 stack_trap "do_facet mds1 $LCTL set_param fail_loc=0" EXIT
9775 local cmd="--device MGS llog_print"
9777 cli_llogcnt_orig=$(do_facet mgs $LCTL $cmd $FSNAME-client |
9778 grep -c $FSNAME-OST0000)
9779 mds1_llogcnt_orig=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 |
9780 grep -c $FSNAME-OST0000)
9782 [ $cli_llogcnt_orig -gt 0 ] ||
9783 error "$FSNAME-OST0000 not found (client)"
9784 [ $mds1_llogcnt_orig -gt 0 ] || error "$FSNAME-OST0000 not found (MDT)"
9786 # -n/--dryrun should NOT modify catalog
9787 do_facet mgs "$LCTL del_ost -n -t $FSNAME-OST0000" ||
9788 error "del_ost --dryrun failed with $?"
9790 local cli_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-client |
9791 grep -c $FSNAME-OST0000)
9792 local mds1_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 |
9793 grep -c $FSNAME-OST0000)
9795 [ $cli_llogcnt -eq $cli_llogcnt_orig ] ||
9796 error "--dryrun error: $cli_llogcnt != $cli_llogcnt_orig"
9797 [ $mds1_llogcnt -eq $mds1_llogcnt_orig ] ||
9798 error "--dryrun error: $mds1_llogcnt != $mds1_llogcnt_orig"
9801 do_facet mgs "$LCTL del_ost --target $FSNAME-OST0000" ||
9802 error "del_ost failed with $?"
9804 local cli_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-client |
9805 grep -c $FSNAME-OST0000)
9806 local mds1_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 |
9807 grep -c $FSNAME-OST0000)
9809 # every catalog entry for OST0000 should have been cancelled
9810 [ $cli_llogcnt -eq 0 ] || error "$FSNAME-OST0000 not cancelled (cli)"
9811 [ $mds1_llogcnt -eq 0 ] || error "$FSNAME-OST0000 not cancelled (MDT)"
9813 umount_client $MOUNT
9816 $LFS df | grep -q OST0000 && error "del_ost did not remove OST0000!"
9820 run_test 123ah "del_ost cancels config log entries correctly"
9825 echo "cleanup test 123ai"
9827 # cancel last timeout record
9828 do_facet mgs "$LCTL set_param -P -d timeout"
9830 # restore timeout value
9831 do_nodes $(comma_list $(all_nodes)) "$LCTL set_param timeout=$timeout"
9834 test_123ai() { #LU-16167
9840 remote_mgs_nodsh && skip "remote MGS with nodsh"
9841 (( MDS1_VERSION >= $(version_code 2.15.51) )) ||
9842 skip "Need MDS version at least 2.15.51"
9844 [ -d $MOUNT/.lustre ] || setup
9846 old_count=$(do_facet mgs "$LCTL --device MGS llog_print -r params" |
9847 grep -c "parameter: timeout")
9849 old_timeout=$($LCTL get_param -n timeout)
9850 stack_trap "cleanup_123ai $old_timeout" EXIT
9852 # add and cancel params (2 * MAX_IOC_BUFLEN / 128 + 1 = 129)
9853 for i in {1..129}; do
9854 do_facet mgs "$LCTL set_param -P timeout=$i" ||
9855 error "fail to set timeout=$i on MGS"
9858 count=$(do_facet mgs "$LCTL --device MGS llog_print -r params" |
9859 grep -c "parameter: timeout")
9861 ((count - old_count == 129)) ||
9862 error "number timeout record missmatch ($count - $old_count != 129)"
9864 do_facet mgs "$LCTL --device MGS llog_print params" |
9865 tail -1 | grep "timeout, value: 129" ||
9866 error "llog_print could not display the last record (timeout=129)"
9869 run_test 123ai "llog_print display all non skipped records"
9872 remote_mgs_nodsh && skip "remote MGS with nodsh"
9874 [ -d $MOUNT/.lustre ] || setup
9875 local yaml_file="$TMP/$tfile.yaml"
9877 do_facet mgs rm "$yaml_file"
9878 local cfgfiles=$(do_facet mgs "lctl --device MGS llog_catlist" |
9879 sed 's/config_log://')
9881 # set jobid_var to a different value for test
9882 local orig_val=$(do_facet mgs $LCTL get_param jobid_var)
9884 do_facet mgs $LCTL set_param -P jobid_var="TESTNAME"
9886 for i in $cfgfiles params; do
9887 do_facet mgs "lctl --device MGS llog_print ${i} >> $yaml_file"
9890 echo "Unmounting FS"
9897 # Reapply the config from before
9898 echo "Setting configuration parameters"
9899 do_facet mgs "lctl set_param -F $yaml_file"
9901 local set_val=$(do_facet mgs $LCTL get_param jobid_var)
9903 do_facet mgs $LCTL set_param -P $orig_val
9905 [ $set_val == "jobid_var=TESTNAME" ] ||
9906 error "$set_val is not TESTNAME"
9908 do_facet mgs rm "$yaml_file"
9911 run_test 123F "clear and reset all parameters using set_param -F"
9915 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
9916 [ -z $mds2failover_HOST ] && skip "needs MDT failover setup"
9919 do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
9920 grep 1.2.3.4@tcp && error "Should not be fake nid"
9921 do_facet mgs $LCTL conf_param $FSNAME-MDT0001.failover.node=1.2.3.4@tcp\
9922 || error "Set params error"
9923 do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
9924 grep 1.2.3.4@tcp || error "Fake nid should be added"
9928 if combined_mgs_mds; then
9929 start_mdt 1 "-o nosvc" ||
9930 error "starting mds with nosvc option failed"
9932 local nid=$(do_facet mds2 $LCTL list_nids | head -1)
9933 local failover_nid=$(do_node $mds2failover_HOST $LCTL list_nids | head -1)
9935 do_facet mgs $LCTL replace_nids $FSNAME-MDT0001 $nid:$failover_nid ||
9936 error "replace_nids execution error"
9938 if combined_mgs_mds; then
9944 echo "lfs setdirstripe"
9945 $LFS setdirstripe -i 1 $MOUNT/$tdir || error "setdirstirpe error"
9948 run_test 124 "check failover after replace_nids"
9950 get_max_sectors_kb() {
9953 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
9954 local max_sectors_path="/sys/block/${dev_base}/queue/max_sectors_kb"
9956 do_facet ${facet} "[[ -e ${max_sectors_path} ]] &&
9957 cat ${max_sectors_path}"
9960 get_max_hw_sectors_kb() {
9963 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
9964 local max_hw_path="/sys/block/${dev_base}/queue/max_hw_sectors_kb"
9966 do_facet ${facet} "[[ -e ${max_hw_path} ]] && cat ${max_hw_path}"
9969 set_max_sectors_kb() {
9973 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
9974 local max_sectors_path="/sys/block/${dev_base}/queue/max_sectors_kb"
9976 do_facet ${facet} "[[ -e ${max_sectors_path} ]] &&
9977 echo ${value} > ${max_sectors_path}"
9980 [[ $rc -ne 0 ]] && echo "Failed to set ${max_sectors_path} to ${value}"
9985 # Return 0 if all slave devices have max_sectors_kb == max_hw_sectors_kb
9986 # Otherwise return > 0
9987 check_slaves_max_sectors_kb()
9991 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
9992 local slaves_dir=/sys/block/${dev_base}/slaves
9993 local slave_devices=$(do_facet ${facet} "ls ${slaves_dir} 2>/dev/null")
9994 [[ -z ${slave_devices} ]] && return 0
9996 local slave max_sectors new_max_sectors max_hw_sectors path
9999 for slave in ${slave_devices}; do
10000 path="/dev/${slave}"
10001 ! is_blkdev ${facet} ${path} && continue
10002 max_sectors=$(get_max_sectors_kb ${facet} ${path})
10003 max_hw_sectors=$(get_max_hw_sectors_kb ${facet} ${path})
10004 new_max_sectors=${max_hw_sectors}
10005 [[ ${new_max_sectors} -gt ${RQ_SIZE_LIMIT} ]] &&
10006 new_max_sectors=${RQ_SIZE_LIMIT}
10008 if [[ ${max_sectors} -ne ${new_max_sectors} ]]; then
10009 echo "${path} ${max_sectors} ${new_max_sectors}"
10012 check_slaves_max_sectors_kb ${facet} ${path}
10021 local facet_list="mgs mds1 ost1"
10022 combined_mgs_mds && facet_list="mgs ost1"
10025 for facet in ${facet_list}; do
10026 [[ $(facet_fstype ${facet}) != ldiskfs ]] &&
10027 skip "ldiskfs only test" &&
10029 ! is_blkdev ${facet} $(facet_device ${facet}) &&
10030 skip "requires all real devices" &&
10035 # We don't increase IO request size limit past 16MB. See comments in
10036 # lustre/utils/libmount_utils_ldiskfs.c:tune_max_sectors_kb()
10037 RQ_SIZE_LIMIT=$((16 * 1024))
10038 local device old_max_sectors new_max_sectors max_hw_sectors
10039 for facet in ${facet_list}; do
10040 device=$(facet_device ${facet})
10041 old_max_sectors=$(get_max_sectors_kb ${facet} ${device})
10042 max_hw_sectors=$(get_max_hw_sectors_kb ${facet} ${device})
10044 # The expected value after l_tunedisk is executed
10045 new_max_sectors=$old_max_sectors
10046 [[ ${new_max_sectors_kb} -gt ${RQ_SIZE_LIMIT} ]] &&
10047 new_max_sectors_kb=${RQ_SIZE_LIMIT}
10049 # Ensure the current value of max_sectors_kb does not equal
10050 # max_hw_sectors_kb, so we can tell whether l_tunedisk did
10052 set_max_sectors_kb ${facet} ${device} $((new_max_sectors - 1))
10054 # Value before l_tunedisk
10055 local pre_max_sectors=$(get_max_sectors_kb ${facet} ${device})
10056 if [[ ${pre_max_sectors} -ne $((new_max_sectors - 1)) ]]; then
10057 echo "unable to satsify test pre-condition:"
10058 echo "${pre_max_sectors} != $((new_max_sectors - 1))"
10063 echo "Before: ${facet} ${device} ${pre_max_sectors} ${max_hw_sectors}"
10065 do_facet ${facet} "l_tunedisk ${device}"
10067 # Value after l_tunedisk
10068 local post_max_sectors=$(get_max_sectors_kb ${facet} ${device})
10070 echo "After: ${facet} ${device} ${post_max_sectors} ${max_hw_sectors}"
10072 if [[ ${facet} != ost1 ]]; then
10073 if [[ ${post_max_sectors} -ne ${pre_max_sectors} ]]; then
10074 echo "l_tunedisk modified max_sectors_kb of ${facet}"
10078 set_max_sectors_kb ${facet} ${device} ${old_max_sectors}
10080 if [[ ${post_max_sectors} -eq ${pre_max_sectors} ]]; then
10081 echo "l_tunedisk failed to modify max_sectors_kb of ${facet}"
10085 check_slaves_max_sectors_kb ${facet} ${device} ||
10092 run_test 125 "check l_tunedisk only tunes OSTs and their slave devices"
10095 [[ "$MDS1_VERSION" -ge $(version_code 2.13.52) ]] ||
10096 skip "Need MDS version at least 2.13.52"
10099 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ../libcfs/libcfs/libcfs
10100 #define OBD_FAIL_OBD_SETUP 0x60d
10101 do_facet mds1 $LCTL set_param fail_loc=0x60d
10102 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_modules &
10103 for i in {1..40}; do
10104 do_facet mds1 lsmod | grep -q osd_$mds1_FSTYPE && break
10107 clear_failloc $SINGLEMDS 20 &
10108 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
10110 run_test 126 "mount in parallel shouldn't cause a crash"
10113 [[ "$ost1_FSTYPE" == ldiskfs ]] || skip "ldiskfs only test"
10117 zconf_umount_clients $RCLIENTS $MOUNT
10119 wait_osp_active ost ${FSNAME}-OST0000 0 1
10120 local osc_tgt="$FSNAME-OST0000-osc-$($LFS getname -i $DIR)"
10121 local avail1=($($LCTL get_param -n osc.${osc_tgt}.kbytesavail))
10123 $LFS setstripe -i 0 $DIR/$tfile || error "failed creating $DIR/$tfile"
10124 dd if=/dev/zero of=$DIR/$tfile bs=1M oflag=direct || true
10126 local avail2=($($LCTL get_param -n osc.${osc_tgt}.kbytesavail))
10128 if ((avail2 * 100 / avail1 > 1)); then
10131 error "more than 1% space left: before=$avail1 after=$avail2"
10134 local mbs=$(($(stat -c %s $DIR/$tfile) / (1024 * 1024)))
10136 dd if=/dev/zero of=$DIR/$tfile bs=1M count=$mbs conv=notrunc \
10137 oflag=direct || error "overwrite failed"
10139 run_test 127 "direct io overwrite on full ost"
10143 combined_mgs_mds && skip "need separate mgs device"
10144 [ "$ost2_FSTYPE" == zfs ] && import_zpool ost2
10147 # Try to apply nolocallogs to the virgin OST. Should fail.
10148 do_facet ost2 "$TUNEFS --nolocallogs $(ostdevname 2)" &&
10149 error "nolocallogs should not be allowed on the virgin target"
10154 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
10155 # Start OST without MGS (local configs)
10156 do_facet ost1 "$TUNEFS --dryrun $(ostdevname 1)"
10157 start_ost || error "unable to start OST1"
10158 stop_ost || error "Unable to stop OST1"
10160 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
10161 # Do not allow reading local configs, should fail
10162 do_facet ost1 "$TUNEFS --nolocallogs $(ostdevname 1)" ||
10163 error "Can not set nolocallogs"
10164 start_ost && error "OST1 started, but should fail"
10166 # Connect to MGS successfully, reset nolocallogs flag
10167 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
10168 start_mgs || error "unable to start MGS"
10169 start_ost || error "unable to start OST1"
10171 do_facet ost1 "$TUNEFS --dryrun $(ostdevname 1)" | grep "nolocallogs" &&
10172 error "nolocallogs expected to be reset"
10174 stop_ost || error "Unable to stop OST1"
10176 run_test 128 "Force using remote logs with --nolocallogs"
10180 (( MDS1_VERSION >= $(version_code 2.14.57) )) ||
10181 skip "Need MDS version at least 2.14.57"
10183 start_mgsmds || error "MGS/MDS start failed"
10185 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &&
10186 error "start ost1 should fail" || true
10187 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &&
10188 error "second start ost1 should fail" || true
10189 do_facet ost1 "$TUNEFS --writeconf $(ostdevname 1)"
10190 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS ||
10191 error "start ost1 failed"
10195 run_test 129 "attempt to connect an OST with the same index should fail"
10199 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
10201 stop_mdt 2 || error "mdt2 stop failed"
10202 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)"
10203 start_mdt 2 || error "mdt2 start failed"
10204 do_facet mds2 "$LCTL dl" | grep MDT0001-osp-MDT0001 &&
10205 error "Illegal OSP device created" || true
10207 run_test 130 "re-register an MDT after writeconf"
10210 [ "$mds1_FSTYPE" == "ldiskfs" ] || skip "ldiskfs only test"
10211 (( $MDS1_VERSION >= $(version_code 2.14.56.35) )) ||
10212 skip "Need MDS version at least 2.14.56.35"
10213 do_facet mds1 $DEBUGFS -R features $(mdsdevname 1) |
10214 grep -q project || skip "skip project quota not supported"
10219 test_mkdir -c $MDSCOUNT -p $DIR/$tdir
10220 $LFS project -p 1000 $DIR/$tdir || error "set dir project id failed"
10221 createmany -o $DIR/$tdir/f 512
10222 for ((i = 0; i < 512; ++i)); do
10223 $LFS project -p $i $DIR/$tdir/f${i} ||
10224 error "set f${i} project id failed"
10227 test_mkdir -c $MDSCOUNT -p $DIR/$tdir.inherit
10228 $LFS project -p 1001 -s $DIR/$tdir.inherit
10229 createmany -o $DIR/$tdir.inherit/f 128
10230 (( $($LFS project $DIR/$tdir.inherit/f* |
10231 awk '$1 == 1001 { print }' | wc -l) == 128 )) ||
10232 error "files did not inherit projid 1001"
10236 for i in $(seq $MDSCOUNT); do
10237 mds_backup_restore mds$i ||
10238 error "Backup/restore on mds$i failed"
10243 projid=($($LFS project -d $DIR/$tdir))
10244 [ ${projid[0]} == "1000" ] ||
10245 error "projid expected 1000 not ${projid[0]}"
10246 for ((i = 0; i < 512; ++i)); do
10247 projid=($($LFS project $DIR/$tdir/f${i}))
10248 [ ${projid[0]} == "$i" ] ||
10249 error "projid expected $i not ${projid[0]}"
10252 (( $($LFS project $DIR/$tdir.inherit/f* |
10253 awk '$1 == 1001 { print }' | wc -l) == 128 )) ||
10254 error "restore did not copy projid 1001"
10256 run_test 131 "MDT backup restore with project ID"
10262 (( MDS1_VERSION >= $(version_code 2.14.57) )) ||
10263 skip "Need MDS version at least 2.14.57"
10265 combined_mgs_mds || start_mgs || error "unable to start MGS"
10266 start_mdt 1 || error "unable to start mdt1"
10268 err_cnt=$(do_facet mds1 dmesg | grep -c "cannot take the layout locks")
10269 stop_mdt 1 || error "stop mdt1 failed"
10271 [ "$mds1_FSTYPE" == zfs ] && import_zpool mds1
10272 do_facet mds1 $TUNEFS --param mdt.hsm_control=enabled $(mdsdevname 1) ||
10273 error "tunefs failed"
10274 start_mdt 1 || error "cannot start mdt1"
10276 err_cnt2=$(do_facet mds1 dmesg | grep -c "cannot take the layout locks")
10277 [ $err_cnt -eq $err_cnt2 ] || error "Can not take the layout lock"
10278 stop_mdt 1 || error "stop mdt1 failed"
10280 run_test 132 "hsm_actions processed after failover"
10282 # This test verifies we do RR allocation within a pool even if there is a
10283 # significant imbalance vs an OST outside the pool
10285 [[ $OSTCOUNT -lt 4 ]] && skip_env "needs >= 4 OSTs"
10286 [[ "$OST1_VERSION" -ge $(version_code 2.15.51) ]] ||
10287 skip "Need OST version at least 2.15.51"
10288 # This is the easiest way to ensure OSTs start out balanced
10289 reformat_and_config
10292 check_set_fallocate_or_skip
10294 local testfile=$DIR/$tdir/$tfile
10295 local pool="testpool"
10296 local ostrange=$((OSTCOUNT - 1))
10297 # Select all but the last OST to add to the pool
10298 local poolostrange=$((OSTCOUNT - 2))
10306 [ $MINV -eq 0 ] && error "no free space in OST$MINI"
10307 [ $MAXV -gt $((2 * $MINV)) ] &&
10308 error "OSTs badly unbalanced after reformat"
10310 create_pool $FSNAME.$pool || error "failed to create a pool"
10311 do_facet mgs $LCTL pool_add $FSNAME.$pool OST[0-$poolostrange] ||
10312 error "failed to add OST[0-$poolostrange] to the pool"
10314 test_mkdir -p $DIR/$tdir || error "failed to mkdir $DIR/$tdir"
10315 # Consume space on the OSTs in the pool so they are unbalanced with the
10316 # OST outside of the pool
10317 # fill each OST 90% with fallocate so they are widely
10319 local size=$(((MINV * 9 / 10) * 1024))
10320 for ((i = 0; i <= poolostrange; i++)); do
10321 $LFS setstripe -c 1 -i $i $testfile$i ||
10322 error "failed to setstripe $testfile$i"
10323 fallocate -l $size $testfile$i || error "fallocate failed"
10329 # Create files in the pool now that there is an imbalance
10330 filecount=$(((OSTCOUNT - 1) * filenum))
10331 for ((i = 0; i < filecount; i++)); do
10332 $LFS setstripe -p $pool $testfile-$i ||
10333 error "failed to setstripe -p $pool $testfile-$i"
10335 $LFS getstripe -i $testfile-* > /tmp/$tfile.log
10336 # Count the number of files with a stripe on each OST to verify the
10337 # pool allocated with round-robin
10338 for ((i = 0; i <= poolostrange; i++)); do
10339 stripecount=$(grep -c $i /tmp/$tfile.log)
10340 # Allow a little leeway
10341 if (( stripecount < filenum - 1 ||
10342 stripecount > filenum + 1 )); then
10343 cat /tmp/$tfile.log
10344 error "$stripecount != $filenum files on OST$i"
10348 # Create files across the system now that there is an imbalance
10349 filecount=$((OSTCOUNT * filenum))
10350 for ((i = 1; i < filecount; i++)); do
10351 $LFS setstripe $testfile-$i.2 ||
10352 error "failed to setstripe $testilfe-$i.2"
10354 $LFS getstripe -i $testfile-*.2 > /tmp/$tfile.log
10356 # Count the number of files with a stripe on each OST to verify the
10357 # files are *NOT* allocated with round-robin
10358 for ((i = 0; i <= ostrange; i++)); do
10359 stripecount=$(grep -c $i /tmp/$tfile.log)
10360 if [[ $stripecount -ne $filenum ]]; then
10362 echo "QOS: $stripecount != $filenum files on OST$i"
10365 if [ -z "$qos_used" ]; then
10366 error "QOS not used on imbalanced OSTs!"
10369 rm -rf /tmp/$tfile.log $DIR/$tdir
10370 do_facet mgs $LCTL pool_remove $FSNAME.$pool OST[0-$poolostrange] ||
10371 "failed to remove OST[0-$poolostrange] from the pool"
10372 do_facet mgs $LCTL pool_destroy $FSNAME.$pool ||
10373 error "failed to destroy pool"
10375 run_test 133 "stripe QOS: free space balance in a pool"
10378 [ "$mds1_FSTYPE" == "ldiskfs" ] || skip "ldiskfs only test"
10381 local mdt_dev=$(facet_device mds1)
10382 local tmp_dir=$TMP/$tdir
10383 local out=$tmp_dir/check_iam.txt
10384 local CHECK_IAM=${CHECK_IAM:-$(do_facet mds1 "which check_iam 2> /dev/null || true")}
10386 [[ -n "$CHECK_IAM" ]] || skip "check_iam not found"
10389 for ((i=0; i<64; i++)); do
10391 #cmd introduce a random corruption to IAM file
10392 local cmd="dd if=/dev/urandom of=$tmp_dir/$f bs=2 conv=notrunc count=1 seek=$((RANDOM % 36))"
10393 do_facet mds1 "mkdir -p $tmp_dir; \
10394 $DEBUGFS -c -R 'dump $f $tmp_dir/$f' $mdt_dev 2>&1; \
10395 $CHECK_IAM -v $tmp_dir/$f 2>&1; \
10396 echo $cmd; eval $cmd 2>/dev/null;
10397 $CHECK_IAM -v $tmp_dir/$f 2>&1; echo \\\$?" >> $out
10402 stack_trap "rm -rf $tmp_dir && do_facet mds1 rm -rf $tmp_dir" EXIT
10404 rc=$(grep -c "fault\|except" $out)
10405 (( rc == 0 )) || { cat $out &&
10406 error "check_iam failed with fault or exception $rc"; }
10408 rc=$(grep -c "^255" $out)
10409 error=$(grep -c "FINISHED WITH ERRORS" $out)
10411 (( rc == error )) || { cat $out &&
10412 error "check_iam errcode does not fit with errors $rc $error"; }
10414 run_test 134 "check_iam works without faults"
10416 cleanup_test_135(){
10419 printf "\nCleanup test_135\n" >&2
10420 do_facet mds1 "$LCTL set_param -n $oldgc"
10421 rm -rf $DIR/$tdir &> /dev/null
10425 __test_135_file_thread() {
10427 local init_time=$(awk '{print $1}' /proc/uptime)
10428 local awkcmd="/crosses index zero/ {if (\$1 > $init_time) exit(1);}"
10430 #Generate a full plain llogs
10431 while dmesg | sed -r 's/(\[|\])//g' | awk "$awkcmd" ; do
10432 createmany -o $DIR/$tdir/f 4500 >&2
10433 createmany -u $DIR/$tdir/f 4500 >&2
10437 __test_135_reader() {
10440 local firstidx=$(changelog_user_rec mds1 $cl_user)
10441 local oldidx=$firstidx
10446 while read -t10 -u$fd newidx other; do
10447 (( (newidx - oldidx) == 1 )) ||
10448 error "changelog jump detected (last: $oldidx, current: $newidx)"
10450 if (( (newidx - firstidx + 1) % 13000 == 0 )); then
10451 [[ $pid -eq 0 ]] ||
10452 wait $pid || error "changelog_clear failed"
10453 changelog_clear $((newidx - 1)) mds1 >&2 & pid=$!
10458 [[ $pid -eq 0 ]] ||
10459 wait $pid || error "changelog_clear failed"
10465 (( MDS1_VERSION >= $(version_code 2.15.52) )) ||
10466 skip "need MDS version at least 2.15.52"
10468 local service=$(facet_svc mds1)
10470 local lastread lastidx
10471 local files_pid reader_pid
10476 # Need to reformat because we are changing llog catalog sizes to 5.
10477 # Otherwise, processing could fail with existing catalogs (last_idx>5).
10481 # Disable changelog garbage colector
10482 local oldgc=$(do_facet mds1 "$LCTL get_param mdd.${service}.changelog_gc")
10483 do_facet mds1 "$LCTL set_param -n mdd.${service}.changelog_gc=0"
10484 stack_trap "cleanup_test_135 $oldgc" EXIT INT
10486 # change the changelog_catalog size to 5 entries for everybody
10487 #define OBD_FAIL_CAT_RECORDS 0x1312
10488 do_node $(comma_list $(all_nodes)) $LCTL set_param fail_loc=0x1312 fail_val=5
10490 # disable console ratelimit
10491 local rl=$(cat /sys/module/libcfs/parameters/libcfs_console_ratelimit)
10492 echo 0 > /sys/module/libcfs/parameters/libcfs_console_ratelimit
10493 stack_trap "echo $rl > /sys/module/libcfs/parameters/libcfs_console_ratelimit" EXIT
10495 test_mkdir -c 1 -i 0 $DIR/$tdir || error "Failed to create directory"
10496 changelog_chmask "ALL" || error "changelog_chmask failed"
10497 changelog_register || error "changelog_register failed"
10499 cl_user="${CL_USERS[mds1]%% *}"
10500 changelog_users mds1 | grep -q $cl_user ||
10501 error "User $cl_user not found in changelog_users"
10503 # Start reader thread
10504 coproc $LFS changelog --follow $service
10507 stack_trap "echo kill changelog reader; kill $reader_pid" EXIT
10509 echo -e "\nWrap arround changelog catalog"
10511 # Start file writer thread
10512 __test_135_file_thread "$service" & files_pid=$!
10513 stack_trap "(pkill -P$files_pid; kill $files_pid) &> /dev/null || true" EXIT
10515 # Check changelog entries
10516 lastread=$(__test_135_reader $fd $cl_user) || exit $?
10517 ! kill -0 $files_pid 2>/dev/null ||
10518 error "creation thread is running. Is changelog reader stuck?"
10520 lastidx=$(changelog_users mds1 | awk '/current_index/ {print $NF}' )
10521 [[ "$lastread" -eq "$lastidx" ]] ||
10522 error "invalid changelog lastidx (read: $lastread, mds: $lastidx)"
10524 run_test 135 "check the behavior when changelog is wrapped around"
10527 # (This was sanity/802a)
10529 saved_MGS_MOUNT_OPTS=$MGS_MOUNT_OPTS
10530 saved_MDS_MOUNT_OPTS=$MDS_MOUNT_OPTS
10531 saved_OST_MOUNT_OPTS=$OST_MOUNT_OPTS
10532 saved_MOUNT_OPTS=$MOUNT_OPTS
10536 MGS_MOUNT_OPTS=$saved_MGS_MOUNT_OPTS
10537 MDS_MOUNT_OPTS=$saved_MDS_MOUNT_OPTS
10538 OST_MOUNT_OPTS=$saved_OST_MOUNT_OPTS
10539 MOUNT_OPTS=$saved_MOUNT_OPTS
10544 [[ $mds1_FSTYPE = zfs ]] || skip "ZFS specific test"
10545 [[ $MDS1_VERSION -lt $(version_code 2.9.55) ]] ||
10546 [[ $OST1_VERSION -lt $(version_code 2.9.55) ]] &&
10547 skip "Need server version at least 2.9.55"
10549 [[ $ENABLE_QUOTA ]] && skip "Quota enabled for read-only test"
10551 # Reset before starting
10555 mkdir $DIR/$tdir || error "(1) fail to mkdir"
10557 cp $LUSTRE/tests/test-framework.sh $DIR/$tdir/ ||
10558 error "(2) Fail to copy"
10560 stack_trap cleanup_802a EXIT
10562 # sync by force before remount as readonly
10563 sync; sync_all_data; sleep 3; sync_all_data
10567 MGS_MOUNT_OPTS=$(csa_add "$MGS_MOUNT_OPTS" -o rdonly_dev)
10568 MDS_MOUNT_OPTS=$(csa_add "$MDS_MOUNT_OPTS" -o rdonly_dev)
10569 OST_MOUNT_OPTS=$(csa_add "$OST_MOUNT_OPTS" -o rdonly_dev)
10571 echo "Mount the server as read only"
10572 setupall server_only || error "(3) Fail to start servers"
10574 echo "Mount client without ro should fail"
10575 mount_client $MOUNT &&
10576 error "(4) Mount client without 'ro' should fail"
10578 echo "Mount client with ro should succeed"
10579 MOUNT_OPTS=$(csa_add "$MOUNT_OPTS" -o ro)
10580 mount_client $MOUNT ||
10581 error "(5) Mount client with 'ro' should succeed"
10583 echo "Modify should be refused"
10584 touch $DIR/$tdir/guard && error "(6) Touch should fail under ro mode"
10586 echo "Read should be allowed"
10587 diff $LUSTRE/tests/test-framework.sh $DIR/$tdir/test-framework.sh ||
10588 error "(7) Read should succeed under ro mode"
10590 run_test 802a "simulate readonly device"
10592 if ! combined_mgs_mds ; then
10598 # restore the values of MDSSIZE and OSTSIZE
10599 MDSSIZE=$STORED_MDSSIZE
10600 OSTSIZE=$STORED_OSTSIZE
10604 check_and_cleanup_lustre