8 PTLDEBUG=${PTLDEBUG:--1}
9 LUSTRE=${LUSTRE:-$(dirname $0)/..}
10 . $LUSTRE/tests/test-framework.sh
14 # tool to create lustre filesystem images
15 ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT 32newtarball"
17 # bug number for skipped test: LU-11915
18 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 110"
19 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
22 # bug number for skipped tests: LU-9795 (all below)
23 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 84 86 103"
26 if ! combined_mgs_mds; then
27 # bug number for skipped test: LU-11991 LU-11990
28 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 32a 32b 32c 32d 32e 66"
29 # bug number for skipped test: LU-9897 LU-12032
30 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 84 123F"
34 [ "$SLOW" = "no" ] && EXCEPT_SLOW="45 69 106 111"
38 # use small MDS + OST size to speed formatting time
39 # do not use too small MDSSIZE/OSTSIZE, which affect the default journal size
40 # STORED_MDSSIZE is used in test_18
41 STORED_MDSSIZE=$MDSSIZE
42 STORED_OSTSIZE=$OSTSIZE
44 [ "$mds1_FSTYPE" = zfs ] && MDSSIZE=400000
46 [ "$ost1_FSTYPE" = zfs ] && OSTSIZE=400000
56 # pass "-E lazy_itable_init" to mke2fs to speed up the formatting time
57 if [[ "$LDISKFS_MKFS_OPTS" != *lazy_itable_init* ]]; then
58 LDISKFS_MKFS_OPTS=$(csa_add "$LDISKFS_MKFS_OPTS" -E lazy_itable_init)
62 require_dsh_mds || exit 0
63 require_dsh_ost || exit 0
68 # The MGS must be started before the OSTs for a new fs, so start
69 # and stop to generate the startup logs.
72 wait_osc_import_state mds ost FULL
77 reformat_and_config() {
79 if ! combined_mgs_mds ; then
85 writeconf_or_reformat() {
86 # There are at most 2 OSTs for write_conf test
87 # who knows if/where $TUNEFS is installed?
88 # Better reformat if it fails...
89 writeconf_all $MDSCOUNT 2 ||
90 { echo "tunefs failed, reformatting instead" &&
91 reformat_and_config && return 0; }
100 echo "start mgs service on $(facet_active_host mgs)"
101 start mgs $(mgsdevname) $MGS_MOUNT_OPTS $@
107 local dev=$(mdsdevname $num)
110 echo "start mds service on `facet_active_host $facet`"
111 start $facet ${dev} $MDS_MOUNT_OPTS $@ || return 94
114 stop_mdt_no_force() {
117 local dev=$(mdsdevname $num)
120 echo "stop mds service on `facet_active_host $facet`"
121 stop $facet || return 97
127 local dev=$(mdsdevname $num)
130 echo "stop mds service on `facet_active_host $facet`"
131 # These tests all use non-failover stop
132 stop $facet -f || return 97
138 for num in $(seq $MDSCOUNT); do
139 start_mdt $num $@ || return 94
141 for num in $(seq $MDSCOUNT); do
142 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds${num} FULL
147 if ! combined_mgs_mds ; then
155 for num in $(seq $MDSCOUNT); do
156 stop_mdt $num || return 97
161 echo "stop mgs service on `facet_active_host mgs`"
162 # These tests all use non-failover stop
163 stop mgs -f || return 97
167 echo "start ost1 service on `facet_active_host ost1`"
168 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS $@ || return 95
169 wait_clients_import_state ${CLIENTS:-$HOSTNAME} ost1 FULL
173 echo "stop ost1 service on `facet_active_host ost1`"
174 # These tests all use non-failover stop
175 stop ost1 -f || return 98
179 echo "start ost2 service on `facet_active_host ost2`"
180 start ost2 $(ostdevname 2) $OST_MOUNT_OPTS $@ || return 92
181 wait_clients_import_state ${CLIENTS:-$HOSTNAME} ost2 FULL
185 echo "stop ost2 service on `facet_active_host ost2`"
186 # These tests all use non-failover stop
187 stop ost2 -f || return 93
194 echo "mount $FSNAME ${mountopt:+with opts $mountopt} on $mountpath....."
195 zconf_mount $HOSTNAME $mountpath $mountopt || return 96
201 echo "umount lustre on $mountpath....."
202 zconf_umount $HOSTNAME $mountpath $@ || return 97
205 manual_umount_client(){
208 echo "manual umount lustre on ${MOUNT}...."
209 do_facet client "umount ${FORCE} $MOUNT"
215 start_mds || error "MDT start failed"
216 start_ost || error "Unable to start OST1"
217 mount_client $MOUNT || error "client start failed"
218 client_up || error "client_up failed"
227 unload_modules_conf () {
228 if combined_mgs_mds || ! local_mode; then
229 unload_modules || return 1
234 stop_ost || return 202
235 stop_mds || return 201
236 unload_modules_conf || return 203
241 [ "x$1" != "x" ] && force='-f'
242 umount_client $MOUNT $force|| return 200
243 cleanup_nocli || return $?
248 echo "umount $MOUNT2 ..."
249 umount $MOUNT2 || true
250 echo "stopping fs2mds ..."
251 stop fs2mds -f || true
252 echo "stopping fs2ost ..."
253 stop fs2ost -f || true
257 do_facet client "cp /etc/passwd $DIR/a" || return 71
258 do_facet client "rm $DIR/a" || return 72
259 # make sure lustre is actually mounted (touch will block,
260 # but grep won't, so do it after)
261 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
262 echo "setup single mount lustre success"
266 do_facet client "touch $DIR/a" || return 71
267 do_facet client "rm $DIR/a" || return 72
268 do_facet client "touch $DIR2/a" || return 73
269 do_facet client "rm $DIR2/a" || return 74
270 echo "setup double mount lustre success"
274 cat /dev/urandom | tr -dc 'a-zA-Z0-9' | fold -w $1 | head -n 1
277 if [ "$ONLY" == "setup" ]; then
282 if [ "$ONLY" == "cleanup" ]; then
289 #create single point mountpoint
295 check_mount || error "check_mount failed"
296 cleanup || error "cleanup failed with $?"
298 run_test 0 "single mount setup"
301 start_mds || error "MDS start failed"
302 start_ost || error "unable to start OST"
303 echo "start ost second time..."
304 start_ost && error "2nd OST start should fail"
305 mount_client $MOUNT || error "client start failed"
306 check_mount || error "check_mount failed"
307 cleanup || error "cleanup failed with $?"
309 run_test 1 "start up ost twice (should return errors)"
312 start_mds || error "MDT start failed"
313 echo "start mds second time.."
314 start_mds && error "2nd MDT start should fail"
315 start_ost || error "OST start failed"
316 mount_client $MOUNT || error "mount_client failed to start client"
317 check_mount || error "check_mount failed"
318 cleanup || error "cleanup failed with $?"
320 run_test 2 "start up mds twice (should return err)"
324 #mount.lustre returns an error if already in mtab
325 mount_client $MOUNT && error "2nd client mount should fail"
326 check_mount || error "check_mount failed"
327 cleanup || error "cleanup failed with $?"
329 run_test 3 "mount client twice (should return err)"
333 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
334 stop_ost || error "Unable to stop OST1"
335 umount_client $MOUNT -f || error “unmount $MOUNT failed”
338 # ok for ost to fail shutdown
339 if [ 202 -ne $eno ] && [ 0 -ne $eno ]; then
340 error "cleanup failed with $?"
343 run_test 4 "force cleanup ost, then cleanup"
345 test_5a() { # was test_5
347 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
348 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
350 stop_mds || error "Unable to stop MDS"
352 # cleanup may return an error from the failed
353 # disconnects; for now I'll consider this successful
354 # if all the modules have unloaded.
358 echo "killing umount"
359 kill -TERM $UMOUNT_PID
360 echo "waiting for umount to finish"
362 if grep " $MOUNT " /proc/mounts; then
363 echo "test 5: /proc/mounts after failed umount"
367 echo "killing umount"
368 kill -TERM $UMOUNT_PID
369 echo "waiting for umount to finish"
371 grep " $MOUNT " /proc/mounts &&
372 error "/proc/mounts after second umount"
375 # manual_mount_client may fail due to umount succeeding above
377 # stop_mds is a no-op here, and should not fail
378 cleanup_nocli || error "cleanup_nocli failed with $?"
379 # df may have lingering entry
381 # mtab may have lingering entry
385 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
387 grep -q $MOUNT" " /etc/mtab || break
388 echo "Waiting /etc/mtab updated ... "
389 WAIT=$(( WAIT + sleep))
391 [ "$WAIT" -eq "$MAX_WAIT" ] &&
392 error "/etc/mtab is not updated in $WAIT secs"
393 echo "/etc/mtab updated in $WAIT secs"
395 run_test 5a "force cleanup mds, then cleanup"
403 grep " $MOUNT " /etc/mtab &&
404 error false "unexpected entry in mtab before mount" && return 10
406 start_ost || error "OST start failed"
407 if ! combined_mgs_mds ; then
408 trap cleanup_5b EXIT ERR
409 start_mds || error "MDS start failed"
413 mount_client $MOUNT && error "mount_client $MOUNT should fail"
414 grep " $MOUNT " /etc/mtab &&
415 error "$MOUNT entry in mtab after failed mount"
417 # stop_mds is a no-op here, and should not fail
418 cleanup_nocli || error "cleanup_nocli failed with $?"
419 if ! combined_mgs_mds ; then
423 run_test 5b "Try to start a client with no MGS (should return errs)"
426 grep " $MOUNT " /etc/mtab &&
427 error false "unexpected entry in mtab before mount" && return 10
429 start_mds || error "MDS start failed"
430 start_ost || error "OST start failed"
431 local oldfs="${FSNAME}"
432 FSNAME="wrong.${FSNAME}"
433 mount_client $MOUNT || :
435 grep " $MOUNT " /etc/mtab &&
436 error "$MOUNT entry in mtab after failed mount"
438 cleanup_nocli || error "cleanup_nocli failed with $?"
440 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
443 grep " $MOUNT " /etc/mtab &&
444 error "unexpected entry in mtab before mount"
446 start_ost || error "OST start failed"
447 start_mds || error "MDS start failed"
448 stop_ost -f || error "Unable to stop OST1"
449 mount_client $MOUNT || error "mount_client $MOUNT failed"
450 umount_client $MOUNT -f || error "umount_client $MOUNT failed"
451 cleanup_nocli || error "cleanup_nocli failed with $?"
452 ! grep " $MOUNT " /etc/mtab ||
453 error "$MOUNT entry in mtab after unmount"
455 run_test 5d "mount with ost down"
458 grep " $MOUNT " /etc/mtab &&
459 error false "unexpected entry in mtab before mount" && return 10
461 start_mds || error "MDS start failed"
462 start_ost || error "OST start failed"
464 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
465 do_facet client "$LCTL set_param fail_loc=0x80000506"
466 mount_client $MOUNT || echo "mount failed (not fatal)"
467 cleanup || error "cleanup failed with $?"
468 grep " $MOUNT " /etc/mtab &&
469 error "$MOUNT entry in mtab after unmount"
472 run_test 5e "delayed connect, don't crash (bug 10268)"
475 combined_mgs_mds && skip "needs separate mgs and mds"
477 grep " $MOUNT " /etc/mtab &&
478 error false "unexpected entry in mtab before mount" && return 10
481 start_ost || error "OST start failed"
482 mount_client $MOUNT &
484 echo client_mount pid is $pid
488 if ! ps -f -p $pid >/dev/null; then
491 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
492 error "mount returns $rc, expected to hang"
494 cleanup || error "cleanup failed with $?"
499 start_mds || error "start MDS failed"
501 # mount should succeed after start mds
503 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
504 cleanup || error "final call to cleanup failed with rc $?"
506 run_test 5f "mds down, cleanup after failed mount (bug 2712)"
510 [ "$CLIENT_VERSION" -lt $(version_code 2.9.53) ] &&
511 skip "automount of debugfs missing before 2.9.53"
512 umount /sys/kernel/debug
513 $LCTL get_param -n devices | egrep -v "error" && \
514 error "lctl can't access debugfs data"
515 grep " debugfs " /etc/mtab || error "debugfs failed to remount"
517 run_test 5g "handle missing debugfs"
523 #define OBD_FAIL_MDS_FS_SETUP 0x135
524 do_facet mds1 "$LCTL set_param fail_loc=0x80000135"
525 start_mdt 1 && error "start mdt should fail"
526 start_mdt 1 || error "start mdt failed"
527 client_up || error "client_up failed"
530 run_test 5h "start mdt failure at mdt_fs_setup()"
533 (( $MDS1_VERSION >= $(version_code 2.12.54) )) ||
534 skip "Need MDS version at least 2.12.54"
539 #define OBD_FAIL_QUOTA_INIT 0xA05
540 do_facet mds1 "$LCTL set_param fail_loc=0x80000A05"
541 start_mdt 1 && error "start mdt should fail"
542 start_mdt 1 || error "start mdt failed"
543 client_up || error "client_up failed"
546 run_test 5i "start mdt failure at mdt_quota_init()"
551 mount_client $MOUNT || error "mount_client $MOUNT failed"
552 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
553 cleanup || error "cleanup failed with rc $?"
555 run_test 6 "manual umount, then mount again"
560 cleanup_nocli || error "cleanup_nocli failed with $?"
562 run_test 7 "manual umount, then cleanup"
566 mount_client $MOUNT2 || error "mount_client $MOUNT2 failed"
567 check_mount2 || error "check_mount2 failed"
568 umount_client $MOUNT2 || error "umount_client $MOUNT2 failed"
569 cleanup || error "cleanup failed with rc $?"
571 run_test 8 "double mount setup"
574 start_ost || error "OST start failed"
576 do_facet ost1 $LCTL set_param debug=\'inode trace\' ||
577 error "do_facet ost1 set_param inode trace failed."
578 do_facet ost1 $LCTL set_param subsystem_debug=\'mds ost\' ||
579 error "do_facet ost1 set_param debug mds ost failed."
581 CHECK_PTLDEBUG="`do_facet ost1 $LCTL get_param -n debug`"
582 if [ "$CHECK_PTLDEBUG" ] && { \
583 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
584 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
585 echo "lnet.debug success"
587 error "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
589 CHECK_SUBSYS="`do_facet ost1 $LCTL get_param -n subsystem_debug`"
590 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
591 echo "lnet.subsystem_debug success"
593 error "lnet.subsystem_debug: want 'mds ost' got '$CHECK_SUBSYS'"
595 stop_ost || error "Unable to stop OST1"
597 run_test 9 "test ptldebug and subsystem for mkfs"
602 if ! combined_mgs_mds; then
603 files=$(do_facet mgs "find /{proc/fs,sys/fs,sys/kernel/debug}/lustre -type l -exec test ! -e {} \; -print")
604 [ -z $files ] || echo "MGS $files is a broken symlink"
607 files=$(do_facet mds1 "find /{proc/fs,sys/fs,sys/kernel/debug}/lustre -type l -exec test ! -e {} \; -print")
608 [ -z $files ] || echo "MDS $files is a broken symlink"
610 files=$(do_facet ost1 "find /{proc/fs,sys/fs,sys/kernel/debug}/lustre -type l -exec test ! -e {} \; -print")
611 [ -z $files ] || echo "OSS $files is a broken symlink"
613 files=$(do_facet client "find /{proc/fs,sys/fs,sys/kernel/debug}/lustre -type l -exec test ! -e {} \; -print")
614 [ -z $files ] || echo "clients $files is a broken symlink"
616 cleanup || error "cleanup failed with rc $?"
618 run_test 10a "find lctl param broken symlinks"
621 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
622 # But with new MDS stack we don't care about the mode of local objects
623 # anymore, so this test is removed. See bug 22944 for more details.
627 if [ "$mds1_FSTYPE" != ldiskfs ]; then
628 skip "ldiskfs only test"
632 check_mount || error "check_mount failed"
633 cleanup || error "cleanup failed with rc $?"
635 echo "Remove mds config log"
636 if ! combined_mgs_mds ; then
640 do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' \
641 $(mgsdevname) || return \$?" ||
642 error "do_facet mgs failed with $?"
644 if ! combined_mgs_mds ; then
648 start_ost || error "OST start failed"
649 start_mds && error "MDS start succeeded, but should fail"
652 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
655 if [ "$mds1_FSTYPE" != ldiskfs ]; then
656 skip "ldiskfs only test"
659 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
664 # check if current MDSSIZE is large enough
665 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE &&
666 log "use MDSSIZE=$MDSSIZE"
668 # check if the global config has a large enough MDSSIZE
669 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] &&
670 [ $STORED_MDSSIZE -ge $MIN ] &&
671 OK=1 && myMDSSIZE=$STORED_MDSSIZE &&
672 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
674 # check if the block device is large enough
675 is_blkdev $SINGLEMDS $MDSDEV $MIN
676 local large_enough=$?
677 if [ -n "$OK" ]; then
678 [ $large_enough -ne 0 ] && OK=""
680 [ $large_enough -eq 0 ] && OK=1 && myMDSSIZE=$MIN &&
681 log "use device $MDSDEV with MIN=$MIN"
684 # check if a loopback device has enough space for fs metadata (5%)
686 if [ -z "$OK" ]; then
687 local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! \
688 -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
689 awk '($1 != "Filesystem") { print $4 }')
690 ! [ -z "$SPACE" ] && [ $SPACE -gt $((MIN / 20)) ] &&
691 OK=1 && myMDSSIZE=$MIN &&
692 log "use file $MDSDEV with MIN=$MIN"
695 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS"
697 echo "mount mds with large journal..."
699 local OLD_MDSSIZE=$MDSSIZE
703 echo "mount lustre system..."
705 check_mount || error "check_mount failed"
707 echo "check journal size..."
708 local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
709 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
710 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
712 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
715 cleanup || error "cleanup failed with rc $?"
720 run_test 18 "check mkfs creates large journals"
723 start_mds || error "MDS start failed"
724 stop_mds || error "Unable to stop MDS"
726 run_test 19a "start/stop MDS without OSTs"
729 start_ost || error "Unable to start OST1"
730 stop_ost -f || error "Unable to stop OST1"
732 run_test 19b "start/stop OSTs without MDS"
735 # first format the ost/mdt
736 start_mds || error "MDS start failed"
737 start_ost || error "Unable to start OST1"
738 mount_client $MOUNT || error "mount_client $MOUNT failed"
739 check_mount || error "check_mount failed"
740 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed."
741 mount_client $MOUNT remount,ro || error "remount client with ro failed"
742 touch $DIR/$tfile && error "$DIR/$tfile created incorrectly"
743 [ -e $DIR/$tfile ] && error "$DIR/$tfile exists incorrectly"
744 mount_client $MOUNT remount,rw || error "remount client with rw failed"
745 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
746 MCNT=$(grep -c $MOUNT' ' /etc/mtab)
747 [ "$MCNT" -ne 1 ] && error "$MOUNT in /etc/mtab $MCNT times"
749 stop_mds || error "Unable to stop MDS"
750 stop_ost || error "Unable to stop OST1"
752 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
755 start_mds || error "MDS start failed"
756 start_ost || error "unable to start OST1"
757 wait_osc_import_state mds ost FULL
758 stop_ost || error "unable to stop OST1"
759 stop_mds || error "unable to stop MDS"
761 run_test 21a "start mds before ost, stop ost first"
764 start_ost || error "unable to start OST1"
765 start_mds || error "MDS start failed"
766 wait_osc_import_state mds ost FULL
767 stop_mds || error "unable to stop MDS"
768 stop_ost || error "unable to stop OST1"
770 run_test 21b "start ost before mds, stop mds first"
773 start_ost || error "Unable to start OST1"
774 start_mds || error "MDS start failed"
775 start_ost2 || error "Unable to start OST2"
776 wait_osc_import_state mds ost2 FULL
777 stop_ost || error "Unable to stop OST1"
778 stop_ost2 || error "Unable to stop OST2"
779 stop_mds || error "Unable to stop MDS"
780 #writeconf to remove all ost2 traces for subsequent tests
781 writeconf_or_reformat
783 run_test 21c "start mds between two osts, stop mds last"
786 combined_mgs_mds && skip "need separate mgs device"
791 start_mgs || error "unable to start MGS"
792 start_ost || error "unable to start OST1"
793 start_ost2 || error "unable to start OST2"
794 start_mds || error "MDS start failed"
795 wait_osc_import_state mds ost2 FULL
797 local zkeeper=${KEEP_ZPOOL}
798 stack_trap "KEEP_ZPOOL=$zkeeper" EXIT
801 stop_ost || error "Unable to stop OST1"
802 stop_ost2 || error "Unable to stop OST2"
803 stop_mds || error "Unable to stop MDS"
805 #writeconf to remove all ost2 traces for subsequent tests
806 writeconf_or_reformat
807 KEEP_ZPOOL="${zkeeper}"
809 start_mgs || error "unable to start MGS"
811 run_test 21d "start mgs then ost and then mds"
814 MGSNID="$saved_mgsnid"
816 echo "stopping fs2mgs ..."
817 stop $fs2mgs -f || true
820 test_21e() { # LU-5863
821 if [[ -z "$fs3ost_DEV" || -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
822 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
823 skip_env "mixed loopback and real device not working"
826 local fs2mdsdev=$(mdsdevname 1_2)
827 local fs2ostdev=$(ostdevname 1_2)
828 local fs3ostdev=$(ostdevname 2_2)
830 local fs2mdsvdev=$(mdsvdevname 1_2)
831 local fs2ostvdev=$(ostvdevname 1_2)
832 local fs3ostvdev=$(ostvdevname 2_2)
834 # temporarily use fs3ost as fs2mgs
836 local fs2mgsdev=$fs3ostdev
837 local fs2mgsvdev=$fs3ostvdev
839 local fsname=test1234
841 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
842 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
843 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS && trap cleanup_21e EXIT INT ||
844 error "start fs2mgs failed"
846 local saved_mgsnid="$MGSNID"
847 MGSNID=$(do_facet $fs2mgs $LCTL list_nids | xargs | tr ' ' ,)
849 add fs2mds $(mkfs_opts mds1 $fs2mdsdev $fsname) \
850 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
851 add fs2ost $(mkfs_opts ost1 $fs2ostdev $fsname) \
852 --reformat $fs2ostdev $fs2ostvdev || error "add fs2ost failed"
854 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
855 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS || error "start fs2mds failed"
857 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
858 $MOUNT_CMD $MGSNID:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
859 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
863 run_test 21e "separate MGS and MDS"
866 start_mds || error "MDS start failed"
868 echo "Client mount with ost in logs, but none running"
869 start_ost || error "unable to start OST1"
870 # wait until mds connected to ost and open client connection
871 wait_osc_import_state mds ost FULL
872 stop_ost || error "unable to stop OST1"
873 mount_client $MOUNT || error "mount_client $MOUNT failed"
874 # check_mount will block trying to contact ost
875 mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed"
876 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed"
877 umount_client $MOUNT -f
880 echo "Client mount with a running ost"
881 start_ost || error "unable to start OST1"
883 # if gss enabled, wait full time to let connection from
884 # mds to ost be established, due to the mismatch between
885 # initial connect timeout and gss context negotiation timeout.
886 # This perhaps could be remove after AT landed.
887 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
888 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
890 mount_client $MOUNT || error "mount_client $MOUNT failed"
891 wait_osc_import_state mds ost FULL
892 wait_osc_import_ready client ost
893 check_mount || error "check_mount failed"
896 cleanup || error "cleanup failed with rc $?"
898 run_test 22 "start a client before osts (should return errs)"
900 test_23a() { # was test_23
903 stop $SINGLEMDS || error "failed to stop $SINGLEMDS"
904 # force down client so that recovering mds waits for reconnect
905 local running=$(grep -c $MOUNT /proc/mounts) || true
906 if [ $running -ne 0 ]; then
907 echo "Stopping client $MOUNT (opts: -f)"
911 # enter recovery on failed mds
912 local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/})
913 start $SINGLEMDS $MDT_DEV $MDS_MOUNT_OPTS || error "MDS start failed"
914 # try to start a new client
915 mount_client $MOUNT &
917 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
918 MOUNT_LUSTRE_PID=$(ps -ef | grep mount.lustre |
919 grep -v grep | awk '{print $2}')
920 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
922 ps --ppid $MOUNT_LUSTRE_PID
923 echo "waiting for mount to finish"
925 # "ctrl-c" sends SIGINT but it usually (in script) does not work on child process
926 # SIGTERM works but it does not spread to offspring processses
927 kill -s TERM $MOUNT_PID
928 kill -s TERM $MOUNT_LUSTRE_PID
929 # we can not wait $MOUNT_PID because it is not a child of this shell
935 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
937 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
938 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
941 [ -z "$PID1" -a -z "$PID2" ] && break
942 echo "waiting for mount to finish ... "
943 WAIT=$(( WAIT + sleep))
945 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
946 error "MOUNT_PID $MOUNT_PID and "\
947 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
950 cleanup || error "cleanup failed with rc $?"
952 run_test 23a "interrupt client during recovery mount delay"
954 test_23b() { # was test_23
955 start_mds || error "MDS start failed"
956 start_ost || error "Unable to start OST1"
957 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
958 $LCTL set_param fail_loc=0x80000313
960 cleanup || error "cleanup failed with rc $?"
962 run_test 23b "Simulate -EINTR during mount"
965 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
967 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
968 is_blkdev $SINGLEMDS $MDSDEV &&
969 skip_env "mixed loopback and real device not working"
972 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
974 local fs2mdsdev=$(mdsdevname 1_2)
975 local fs2ostdev=$(ostdevname 1_2)
976 local fs2mdsvdev=$(mdsvdevname 1_2)
977 local fs2ostvdev=$(ostvdevname 1_2)
980 # LU-9733 test fsname started with numbers as well
981 local FSNAME2=969362ae
983 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --nomgs --mgsnode=$MGSNID \
984 --fsname=${FSNAME2} --reformat $fs2mdsdev $fs2mdsvdev || exit 10
986 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --fsname=${FSNAME2} \
987 --reformat $fs2ostdev $fs2ostvdev || exit 10
990 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
991 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
992 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
993 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
995 # LU-9733 test fsname started with numbers
996 cl_user=$(do_facet $SINGLEMDS lctl --device $FSNAME2-MDT0000 \
997 changelog_register -n) ||
998 error "register changelog failed"
1000 do_facet $SINGLEMDS lctl --device $FSNAME2-MDT0000 \
1001 changelog_deregister $cl_user ||
1002 error "deregister changelog failed"
1004 check_mount || error "check_mount failed"
1005 # files written on 1 should not show up on 2
1006 cp /etc/passwd $DIR/$tfile
1008 [ -e $MOUNT2/$tfile ] && error "File bleed"
1011 cp /etc/passwd $MOUNT2/$tfile ||
1012 error "cp /etc/passwd $MOUNT2/$tfile failed"
1013 rm $MOUNT2/$tfile || error "remove $MOUNT2/$tfile failed"
1014 # 2 is actually mounted
1015 grep $MOUNT2' ' /proc/mounts > /dev/null || error "$MOUNT2 not mounted"
1017 facet_failover fs2mds
1018 facet_failover fs2ost
1020 umount_client $MOUNT
1021 # the MDS must remain up until last MDT
1023 MDS=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
1024 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
1025 [ -z "$MDS" ] && error "No MDT"
1027 cleanup_nocli || error "cleanup_nocli failed with rc $?"
1029 run_test 24a "Multiple MDTs on a single node"
1032 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1034 skip "needs combined MGT and MDT device"
1036 if [ -z "$fs2mds_DEV" ]; then
1037 local dev=${SINGLEMDS}_dev
1038 local MDSDEV=${!dev}
1039 is_blkdev $SINGLEMDS $MDSDEV &&
1040 skip_env "mixed loopback and real device not working"
1043 local fs2mdsdev=$(mdsdevname 1_2)
1044 local fs2mdsvdev=$(mdsvdevname 1_2)
1046 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --mgs --fsname=${FSNAME}2 \
1047 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
1049 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && {
1051 error "start MDS should fail"
1054 cleanup || error "cleanup failed with rc $?"
1056 run_test 24b "Multiple MGSs on a single node (should return err)"
1060 check_mount || error "check_mount failed"
1061 local MODULES=$($LCTL modules | awk '{ print $2 }')
1062 rmmod $MODULES 2>/dev/null || true
1063 cleanup || error "cleanup failed with $?"
1065 run_test 25 "Verify modules are referenced"
1069 # we need modules before mount for sysctl, so make sure...
1070 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
1071 #define OBD_FAIL_MDS_FS_SETUP 0x135
1072 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000135"
1073 start_mds && error "MDS started but should not have started"
1074 $LCTL get_param -n devices
1075 DEVS=$($LCTL get_param -n devices | egrep -v MG | wc -l)
1076 [ $DEVS -gt 0 ] && error "number of devices is $DEVS, should be zero"
1077 # start mds to drop writeconf setting
1078 start_mds || error "Unable to start MDS"
1079 stop_mds || error "Unable to stop MDS"
1080 unload_modules_conf || error "unload_modules_conf failed with $?"
1082 run_test 26 "MDT startup failure cleans LOV (should return errs)"
1086 start_ost || error "Unable to start OST1"
1087 start_mds || error "Unable to start MDS"
1088 echo "Requeue thread should have started: "
1089 ps -e | grep ll_cfg_requeue
1090 set_persistent_param_and_check ost1 \
1091 "obdfilter.$FSNAME-OST0000.client_cache_seconds" \
1092 "$FSNAME-OST0000.ost.client_cache_seconds"
1093 cleanup_nocli || error "cleanup_nocli failed with rc $?"
1095 run_test 27a "Reacquire MGS lock if OST started first"
1100 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
1101 awk '($3 ~ "mdt" && $4 ~ "MDT0000") { print $4 }')
1103 facet_failover $SINGLEMDS
1104 set_persistent_param_and_check $SINGLEMDS \
1105 "mdt.$device.identity_acquire_expire" \
1106 "$device.mdt.identity_acquire_expire"
1107 set_persistent_param_and_check client \
1108 "mdc.$device-mdc-*.max_rpcs_in_flight" \
1109 "$device.mdc.max_rpcs_in_flight"
1111 cleanup || error "cleanup failed with $?"
1113 run_test 27b "Reacquire MGS lock after failover"
1115 test_28A() { # was test_28
1118 local TEST="llite.$FSNAME-*.max_read_ahead_whole_mb"
1119 local PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
1120 local orig=$($LCTL get_param -n $TEST)
1121 local max=$($LCTL get_param -n \
1122 llite.$FSNAME-*.max_read_ahead_per_file_mb)
1124 orig=${orig%%.[0-9]*}
1126 echo "ORIG:$orig MAX:$max"
1127 [[ $max -le $orig ]] && orig=$((max - 3))
1128 echo "ORIG:$orig MAX:$max"
1130 local final=$((orig + 1))
1132 set_persistent_param_and_check client "$TEST" "$PARAM" $final
1133 final=$((final + 1))
1134 set_persistent_param_and_check client "$TEST" "$PARAM" $final
1135 umount_client $MOUNT || error "umount_client $MOUNT failed"
1136 mount_client $MOUNT || error "mount_client $MOUNT failed"
1138 local result=$($LCTL get_param -n $TEST)
1140 if [ $result -ne $final ]; then
1141 error "New config not seen: wanted $final got $result"
1143 echo "New config success: got $result"
1145 set_persistent_param_and_check client "$TEST" "$PARAM" $orig
1146 cleanup || error "cleanup failed with rc $?"
1148 run_test 28A "permanent parameter setting"
1150 test_28a() { # LU-4221
1151 [[ "$OST1_VERSION" -ge $(version_code 2.5.52) ]] ||
1152 skip "Need OST version at least 2.5.52"
1153 [ "$ost1_FSTYPE" = zfs ] &&
1154 skip "LU-4221: no such proc params for ZFS OSTs"
1161 local device="$FSNAME-OST0000"
1165 # In this test we will set three kinds of proc parameters with
1166 # lctl set_param -P or lctl conf_param:
1167 # 1. non-symlink ones in the OFD
1168 # 2. non-symlink ones in the OSD
1171 # prepare a non-symlink parameter in the OFD
1172 name="client_cache_seconds"
1173 param="$device.ost.$name"
1174 cmd="obdfilter.$device.$name"
1176 # permanently setting the parameter in the OFD
1177 old=$(do_facet ost1 $LCTL get_param -n $cmd)
1179 set_persistent_param_and_check ost1 "$cmd" "$param" $new
1180 set_persistent_param_and_check ost1 "$cmd" "$param" $old
1183 # prepare a non-symlink parameter in the OSD
1185 param="$device.osd.$name"
1186 cmd="osd-*.$device.$name"
1188 # conf_param the parameter in the OSD
1189 old=$(do_facet ost1 $LCTL get_param -n $cmd)
1190 new=$(((old + 1) % 2))
1191 set_persistent_param_and_check ost1 "$cmd" "$param" $new
1192 set_persistent_param_and_check ost1 "$cmd" "$param" $old
1194 cleanup || error "cleanup failed with $?"
1196 run_test 28a "set symlink parameters permanently with lctl"
1199 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >= 2 OSTs"
1200 setup_noconfig > /dev/null 2>&1
1201 start_ost2 || error "Unable to start OST2"
1204 local PARAM="$FSNAME-OST0001.osc.active"
1205 # With lctl set_param -P the value $PROC_ACT will be sent to
1206 # all nodes. The [!M] filter out the ability to set active
1207 # on the MDS servers which is tested with wait_osp_* below.
1208 # For ost_server_uuid that only exist on client so filtering
1210 local PROC_ACT="os[cp].$FSNAME-OST0001-osc-[^M]*.active"
1211 local PROC_UUID="os[cp].$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
1213 ACTV=$($LCTL get_param -n $PROC_ACT)
1215 set_persistent_param_and_check client $PROC_ACT $PARAM $DEAC
1216 # also check ost_server_uuid status
1217 RESULT=$($LCTL get_param -n $PROC_UUID | grep DEACTIV)
1218 if [ -z "$RESULT" ]; then
1219 error "Client not deactivated: $($LCTL get_param \
1222 echo "Live client success: got $RESULT"
1226 wait_osp_active ost ${FSNAME}-OST0001 1 0
1228 # test new client starts deactivated
1229 umount_client $MOUNT || error "umount_client $MOUNT failed"
1230 mount_client $MOUNT || error "mount_client $MOUNT failed"
1232 # the 2nd and 3rd field of ost_server_uuid do not update at the same
1233 # time when using lctl set_param -P
1234 wait_update_facet client \
1235 "$LCTL get_param -n $PROC_UUID | awk '{print \\\$3 }'" \
1237 error "New client start active: $($LCTL get_param -n $PROC_UUID)"
1239 echo "New client success: got '$($LCTL get_param -n $PROC_UUID)'"
1241 # make sure it reactivates
1242 set_persistent_param_and_check client $PROC_ACT $PARAM $ACTV
1244 umount_client $MOUNT
1245 stop_ost2 || error "Unable to stop OST2"
1246 cleanup_nocli || error "cleanup_nocli failed with $?"
1247 #writeconf to remove all ost2 traces for subsequent tests
1248 writeconf_or_reformat
1250 run_test 29 "permanently remove an OST"
1255 echo Big config llog
1256 local path="llite.$FSNAME-*"
1257 local cpath="$FSNAME.llite"
1258 local param="max_read_ahead_whole_mb"
1260 local test="${path}.$param"
1261 local conf="${cpath}.$param"
1263 local orig=$($LCTL get_param -n $test)
1264 local list=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1265 for i in ${list[@]}; do
1266 set_persistent_param_and_check client $test $conf $i
1268 # make sure client restart still works
1269 umount_client $MOUNT
1270 mount_client $MOUNT || error "mount_client $MOUNT failed"
1271 [ "$($LCTL get_param -n $test)" -ne "$i" ] &&
1272 error "Param didn't stick across restart $($test) != $i"
1275 echo Erase parameter setting
1276 if [[ $PERM_CMD == *"set_param -P"* ]]; then
1277 do_facet mgs "$PERM_CMD -d $test" ||
1278 error "Erase param $test failed"
1280 do_facet mgs "$PERM_CMD -d $conf" ||
1281 error "Erase param $conf failed"
1283 umount_client $MOUNT
1284 mount_client $MOUNT || error "mount_client $MOUNT failed"
1285 local final=$($LCTL get_param -n $test)
1286 echo "deleted (default) value=$final, orig=$orig"
1287 orig=${orig%%.[0-9]*}
1288 final=${final%%.[0-9]*}
1289 # assumes this parameter started at the default value
1290 [ "$final" -eq "$orig" ] ||
1291 error "Deleted value=$final -ne orig=$orig"
1293 cleanup || error "cleanup failed with rc $?"
1295 run_test 30a "Big config llog and permanent parameter deletion"
1300 local orignids=$($LCTL get_param -n \
1301 osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1303 local orignidcount=$(echo "$orignids" | wc -w)
1305 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1306 # numerical part of it. Hopefully that's not already a failover address
1308 local OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | \
1310 local ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1311 local NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1312 local NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1313 echo "Using fake nid $NEW"
1315 local TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1316 grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1317 if [[ $PERM_CMD == *"set_param -P"* ]]; then
1318 PARAM="osc.$FSNAME-OST0000-osc-[^M]*.import"
1319 echo "Setting $PARAM from $TEST to $NEW"
1320 do_facet mgs "$PERM_CMD $PARAM='connection=$NEW'" ||
1321 error "$PERM_CMD $PARAM failed"
1323 PARAM="$FSNAME-OST0000.failover.node"
1324 echo "Setting $PARAM from $TEST to $NEW"
1325 do_facet mgs "$PERM_CMD $PARAM='$NEW'" ||
1326 error "$PARAM $PARAM failed"
1328 wait_update_facet client "$TEST" "$NEW" ||
1329 error "check $PARAM failed!"
1331 local NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1333 local NIDCOUNT=$(echo "$NIDS" | wc -w)
1334 echo "should have $((orignidcount + 1)) entries \
1335 in failover nids string, have $NIDCOUNT"
1336 [ $NIDCOUNT -eq $((orignidcount + 1)) ] ||
1337 error "Failover nid not added"
1339 if [[ $PERM_CMD == *"set_param -P"* ]]; then
1340 do_facet mgs "$PERM_CMD -d osc.$FSNAME-OST0000-osc-*.import"
1342 do_facet mgs "$PERM_CMD -d $FSNAME-OST0000.failover.node" ||
1343 error "$PERM_CMD delete failed"
1345 umount_client $MOUNT
1346 mount_client $MOUNT || error "mount_client $MOUNT failed"
1348 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1350 NIDCOUNT=$(echo "$NIDS" | wc -w)
1351 echo "only $orignidcount final entries should remain \
1352 in failover nids string, have $NIDCOUNT"
1353 [ $NIDCOUNT -eq $orignidcount ] || error "Failover nids not removed"
1355 cleanup || error "cleanup failed with rc $?"
1357 run_test 30b "Remove failover nids"
1359 test_31() { # bug 10734
1360 # ipaddr must not exist
1361 $MOUNT_CMD 4.3.2.1@tcp:/lustre $MOUNT || true
1362 cleanup || error "cleanup failed with rc $?"
1364 run_test 31 "Connect to non-existent node (shouldn't crash)"
1368 T32_BLIMIT=40960 # Kbytes
1371 T32_PROLIMIT=$((T32_BLIMIT/10))
1373 # This is not really a test but a tool to create new disk
1374 # image tarballs for the upgrade tests.
1376 # Disk image tarballs should be created on single-node
1377 # clusters by running this test with default configurations
1378 # plus a few mandatory environment settings that are verified
1379 # at the beginning of the test.
1381 test_32newtarball() {
1385 local tmp=$TMP/t32_image_create
1399 if [ $FSNAME != t32fs -o \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o \
1400 $OSTCOUNT -ne 2 -o -z "$OSTDEV1" ]; then
1401 error "Needs FSNAME=t32fs MDSCOUNT=2 " \
1402 "MDSDEV1=<nonexistent_file> " \
1403 "MDSDEV2=<nonexistent_file> " \
1404 "(or MDSDEV, in the case of b1_8) " \
1405 "OSTCOUNT=2 OSTDEV1=<nonexistent_file> " \
1406 "OSTDEV2=<nonexistent_file>"
1410 echo "Found stale $tmp"
1414 mkdir $tmp/src || return 1
1415 tar cf - -C $src . | tar xf - -C $tmp/src
1417 # format ost with comma-separated NIDs to verify LU-4460
1418 local failnid="$(h2nettype 1.2.3.4),$(h2nettype 4.3.2.1)"
1419 MGSNID="$MGSNID,$MGSNID" OSTOPT="--failnode=$failnid" formatall
1423 [[ "$MDS1_VERSION" -ge $(version_code 2.3.50) ]] ||
1424 $LFS quotacheck -ug /mnt/$FSNAME
1425 $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
1428 if [[ $MDSCOUNT -ge 2 ]]; then
1429 remote_dir=/mnt/$FSNAME/remote_dir
1430 $LFS mkdir -i 1 $remote_dir
1431 tar cf - -C $tmp/src . | tar xf - -C $remote_dir
1433 target_dir=$remote_dir
1434 if [[ $MDS1_VERSION -ge $(version_code 2.7.0) ]]; then
1435 striped_dir=/mnt/$FSNAME/striped_dir_old
1436 $LFS mkdir -i 1 -c 2 $striped_dir
1437 tar cf - -C $tmp/src . | tar xf - -C $striped_dir
1440 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1441 target_dir=/mnt/$FSNAME
1444 # add project quota #
1445 [[ $mds1_FSTYPE == "ldiskfs" &&
1446 $MDS1_VERSION -gt $(version_code 2.9.55) ]] ||
1447 [[ $mds1_FSTYPE == "zfs" &&
1448 $MDS1_VERSION -gt $(version_code 2.10.53) ]] && {
1449 pj_quota_dir=$target_dir/project_quota_dir
1450 pj_quota_file_old=$pj_quota_dir/pj_quota_file_old
1452 enable_project_quota
1453 set_mdt_qtype ugp ||
1454 error "enable mdt quota failed"
1456 set_ost_qtype ugp ||
1457 error "enable ost quota failed"
1459 mkdir -p $pj_quota_dir
1460 $LFS setquota -p $T32_PRJID -b 0 -B $T32_PROLIMIT -i 0 \
1461 -I $T32_ILIMIT $pj_quota_dir ||
1462 error "setquota -p $T32_PRJID failed"
1464 $LFS setstripe $pj_quota_file_old -c 1 ||
1465 error "setstripe $pj_quota_file_old failed"
1467 chown $T32_QID:$T32_QID $pj_quota_file_old ||
1468 error "chown $pj_quota_file_old failed"
1470 change_project -p $T32_PRJID $pj_quota_file_old
1472 mkdir -p $tmp/src/project_quota_dir
1473 cp $pj_quota_file_old $tmp/src/project_quota_dir/
1476 #####################
1477 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1479 #if [[ $MDSCOUNT -ge 2 ]]; then
1480 # remote_dir=/mnt/$FSNAME/remote_dir
1481 # $LFS mkdir -i 1 $remote_dir
1482 # tar cf - -C $tmp/src . | tar xf - -C $remote_dir
1484 # if [[ "$MDS1_VERSION" -ge $(version_code 2.7.0) ]]; then
1485 # striped_dir=/mnt/$FSNAME/striped_dir_old
1486 # $LFS mkdir -i 1 -c 2 $striped_dir
1487 # tar cf - -C $tmp/src . | tar xf - -C $striped_dir
1492 if [[ $MDS1_VERSION -ge $(version_code 2.9.51) ]]; then
1493 pfl_dir=$target_dir/pfl_dir
1494 pfl_file=$pfl_dir/pfl_file
1496 $LFS setstripe -E 2M -c 1 -o 0 -E -1 -S 2M -c 1 -o 1 \
1498 error "Create PFL file failed"
1500 dd if=/dev/urandom of=$pfl_file bs=1k count=3k
1501 mkdir -p $tmp/src/pfl_dir
1502 cp $pfl_file $tmp/src/pfl_dir/
1507 if [[ $MDS1_VERSION -ge $(version_code 2.10.56) ]]; then
1508 dom_dir=$target_dir/dom_dir
1509 dom_file=$dom_dir/dom_file
1510 flr_dir=$target_dir/flr_dir
1511 flr_file=$flr_dir/flr_file
1514 $LFS setstripe -E 1M -L mdt -E -1 -S 4M $dom_file
1515 dd if=/dev/urandom of=$dom_file bs=1k count=2k
1516 mkdir -p $tmp/src/dom_dir
1517 cp $dom_file $tmp/src/dom_dir
1520 LFS mirror create -N2 $flr_file
1521 dd if=/dev/urandom of=$flr_file bs=1k count=1
1522 mkdir -p $tmp/src/flr_dir
1523 cp $flr_file $tmp/src/flr_dir
1528 mkdir $tmp/img || return 1
1532 pushd_dir=/mnt/$FSNAME
1533 if [[ $MDSCOUNT -ge 2 ]]; then
1534 pushd_dir=$remote_dir
1535 if [[ "$MDS1_VERSION" -ge $(version_code 2.7.0) ]]; then
1537 ls -Rni --time-style=+%s >$tmp/img/list2
1543 ls -Rni --time-style=+%s >$tmp/img/list
1544 find ! -name .lustre -type f -exec sha1sum {} \; |
1545 sort -k 2 >$tmp/img/sha1sums
1547 $LCTL get_param -n version | head -n 1 |
1548 sed -e 's/^lustre: *//' >$tmp/img/commit
1550 [[ "$MDS1_VERSION" -ge $(version_code 2.3.50) ]] ||
1551 $LFS quotaon -ug /mnt/$FSNAME
1552 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1553 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1554 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1555 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1556 | tr -d "*" > $tmp/img/bspace
1557 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1558 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1559 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1560 | tr -d "*" > $tmp/img/ispace
1561 echo $T32_BLIMIT > $tmp/img/blimit
1562 echo $T32_ILIMIT > $tmp/img/ilimit
1564 $MULTIOP /mnt/$FSNAME/orph_file Ouw_c&
1568 debugfs -R "ls /PENDING" ${MDSDEV1:-$MDSDEV}
1569 cp ${MDSDEV1:-$MDSDEV} $tmp/img
1578 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1581 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1582 echo "Data verification failed"
1585 uname -r >$tmp/img/kernel
1586 uname -m >$tmp/img/arch
1588 for num in $(seq 2 $MDSCOUNT); do
1589 local devname=$(mdsdevname $num)
1591 [[ $(facet_fstype $facet) != zfs ]] ||
1592 devname=$(mdsvdevname $num)
1593 mv $devname $tmp/img
1596 for num in $(seq $OSTCOUNT); do
1597 local devname=$(ostdevname $num)
1599 [[ $(facet_fstype $facet) != zfs ]] ||
1600 devname=$(ostdevname $num)
1601 mv $devname $tmp/img
1604 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1605 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1608 tar cjvf $dst/disk$version-"$mds1_FSTYPE".tar.bz2 -S *
1613 #run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1616 # The list of applicable tarballs is returned via the caller's
1617 # variable "tarballs".
1620 [ "$CLIENTONLY" ] && skip "Client-only testing"
1622 local node=$(facet_active_host $SINGLEMDS)
1623 local r="do_node $node"
1625 ! $r which "$TUNEFS" && skip_env "tunefs.lustre required on $node"
1627 local IMGTYPE="$mds1_FSTYPE"
1629 [[ -n "$T32IMAGE" ]] && tarballs=$T32IMAGE ||
1630 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 \
1631 -name \'disk*-$IMGTYPE.tar.bz2\')
1633 [ -z "$tarballs" ] && skip "No applicable tarballs found"
1636 t32_test_cleanup() {
1638 local facet=$SINGLEMDS
1641 echo "== cleanup with rc=$rc =="
1642 if $shall_cleanup_lustre; then
1643 umount $tmp/mnt/lustre || rc=$?
1645 if $shall_cleanup_mdt; then
1646 $r $UMOUNT $tmp/mnt/mdt || rc=$?
1648 if $shall_cleanup_mdt1; then
1649 $r $UMOUNT $tmp/mnt/mdt1 || rc=$?
1651 if $shall_cleanup_ost; then
1652 $r $UMOUNT $tmp/mnt/ost || rc=$?
1654 if $shall_cleanup_ost1; then
1655 $r $UMOUNT $tmp/mnt/ost1 || rc=$?
1660 if [[ "$mds1_FSTYPE" == zfs ]]; then
1662 local poolname_list="t32fs-mdt1 t32fs-ost1"
1664 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1666 for poolname in $poolname_list; do
1667 destroy_zpool $facet $poolname
1670 combined_mgs_mds || start_mgs || rc=$?
1674 t32_bits_per_long() {
1676 # Yes, this is not meant to be perfect.
1686 t32_reload_modules() {
1688 local all_removed=false
1691 [ "$mds1_FSTYPE" == zfs ] && do_rpc_nodes $node "service zed stop"
1693 while ((i < 20)); do
1694 echo "Unloading modules on $node: Attempt $i"
1695 do_rpc_nodes $node $LUSTRE_RMMOD "$mds1_FSTYPE" &&
1697 do_rpc_nodes $node check_mem_leak || return 1
1698 if $all_removed; then
1699 do_rpc_nodes $node load_modules
1702 if [ "$mds1_FSTYPE" == zfs ]; then
1703 do_rpc_nodes $node "$ZPOOL status -v"
1708 echo "Unloading modules on $node: Given up"
1712 t32_wait_til_devices_gone() {
1718 echo wait for devices to go
1719 while ((i < 20)); do
1720 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1721 loops=$(do_rpc_nodes $node losetup -a | grep -c t32)
1722 ((devices == 0 && loops == 0)) && return 0
1726 echo "waiting for dev on $node: dev $devices loop $loops given up"
1727 do_rpc_nodes $node "losetup -a"
1728 do_rpc_nodes $node "$LCTL device_list"
1732 t32_verify_quota() {
1739 # LU-2435: if the underlying zfs doesn't support userobj_accounting,
1740 # lustre will estimate the object count usage. This fails quota
1741 # verification in 32b. The object quota usage should be accurate after
1742 # zfs-0.7.0 is released.
1743 [ "$mds1_FSTYPE" == zfs ] && {
1744 local zfs_version=$(do_facet $facet cat /sys/module/zfs/version)
1746 [ $(version_code $zfs_version) -lt $(version_code 0.7.0) ] && {
1747 echo "Skip quota verify for zfs: $zfs_version"
1752 $LFS quota -u $T32_QID -v $mnt
1754 qval=$($LFS quota -v -u $T32_QID $mnt |
1755 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1756 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1758 [ $qval -eq $img_bspace ] || {
1759 echo "bspace, act:$qval, exp:$img_bspace"
1763 qval=$($LFS quota -v -u $T32_QID $mnt |
1764 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1765 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1767 [ $qval -eq $img_ispace ] || {
1768 echo "ispace, act:$qval, exp:$img_ispace"
1772 qval=$($LFS quota -v -u $T32_QID $mnt |
1773 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1774 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1776 [ $qval -eq $img_blimit ] || {
1777 echo "blimit, act:$qval, exp:$img_blimit"
1781 qval=$($LFS quota -v -u $T32_QID $mnt |
1782 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1783 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1785 [ $qval -eq $img_ilimit ] || {
1786 echo "ilimit, act:$qval, exp:$img_ilimit"
1790 set_persistent_param_and_check $facet \
1791 "osd-$mds1_FSTYPE.$fsname-MDT0000.quota_slave.enabled" \
1792 "$fsname.quota.mdt" ug
1794 set_persistent_param_and_check $facet \
1795 "osd-$mds1_FSTYPE.$fsname-OST0000.quota_slave.enabled" \
1796 "$fsname.quota.ost" ug
1799 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1800 bs=1M count=$((img_blimit / 1024)) oflag=sync && {
1801 echo "Write succeed, but expect -EDQUOT"
1804 rm -f $mnt/t32_qf_new
1806 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1808 echo "Create succeed, but expect -EDQUOT"
1811 unlinkmany $mnt/t32_qf_ $img_ilimit
1821 sync_all_data > /dev/null 2>&1 || true
1823 [ "$uuid" != "global" ] || uuid=$mnt
1825 $LFS quota -v "$1" "$2" $mnt |
1826 awk 'BEGIN { num='$spec' } { if ($1 == "'$uuid'") \
1827 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1834 local dne_upgrade=${dne_upgrade:-"no"}
1835 local dom_upgrade=${dom_upgrade:-"no"}
1836 local project_upgrade=${project_upgrade:-"no"}
1837 local ff_convert=${ff_convert:-"no"}
1838 local pfl_upgrade=${pfl_upgrade:-"no"}
1839 local project_quota_upgrade=${project_quota_upgrade:-"no"}
1840 local dom_new_upgrade=${dom_new_upgrade:-"no"}
1841 local flr_upgrade=${flr_upgrade:-"no"}
1842 local shall_cleanup_mdt=false
1843 local shall_cleanup_mdt1=false
1844 local shall_cleanup_ost=false
1845 local shall_cleanup_ost1=false
1846 local shall_cleanup_lustre=false
1847 local mdt2_is_available=false
1848 local ost2_is_available=false
1849 local node=$(facet_active_host $SINGLEMDS)
1850 local r="do_node $node"
1866 local mdt_dev=$tmp/mdt
1867 local mdt2_dev=$tmp/mdt2
1868 local ost_dev=$tmp/ost
1869 local ost2_dev=$tmp/ost2
1873 local pfl_file=$tmp/mnt/lustre/remote_dir/pfl_dir/pfl_file
1874 local flr_file=$tmp/mnt/lustre/remote_dir/flr_dir/flr_file
1875 local dom_file=$tmp/mnt/lustre/remote_dir/dom_dir/dom_file
1876 local quota_dir=$tmp/mnt/lustre/remote_dir/project_quota_dir
1878 combined_mgs_mds || stop_mgs || error "Unable to stop MGS"
1879 trap 'trap - RETURN; t32_test_cleanup' RETURN
1882 nid=$($r $LCTL list_nids | head -1)
1884 mkdir -p $tmp/mnt/lustre || error "mkdir $tmp/mnt/lustre failed"
1885 $r mkdir -p $tmp/mnt/{mdt,mdt1,ost,ost1}
1886 $r tar xjvf $tarball -S -C $tmp || {
1887 error_noexit "Unpacking the disk image tarball"
1890 img_commit=$($r cat $tmp/commit)
1891 img_kernel=$($r cat $tmp/kernel)
1892 img_arch=$($r cat $tmp/arch)
1893 img_bspace=$($r cat $tmp/bspace)
1894 img_ispace=$($r cat $tmp/ispace)
1896 # older images did not have "blimit" and "ilimit" files
1897 # use old values for T32_BLIMIT and T32_ILIMIT
1898 $r test -f $tmp/blimit && img_blimit=$($r cat $tmp/blimit) ||
1900 $r test -f $tmp/ilimit && img_ilimit=$($r cat $tmp/ilimit) ||
1903 echo "Upgrading from $(basename $tarball), created with:"
1904 echo " Commit: $img_commit"
1905 echo " Kernel: $img_kernel"
1906 echo " Arch: $img_arch"
1907 echo "OST version: $(lustre_build_version ost1)"
1909 # The conversion can be made only when both of the following
1910 # conditions are satisfied:
1911 # - ost device img version < 2.3.64
1912 # - ost server version >= 2.5
1913 [ $(version_code $img_commit) -ge $(version_code 2.3.64) -o \
1914 "$OST1_VERSION" -lt $(version_code 2.5.0) ] &&
1917 ! $r test -f $mdt2_dev || mdt2_is_available=true
1918 ! $r test -f $ost2_dev || ost2_is_available=true
1920 if [[ "$mds1_FSTYPE" == zfs ]]; then
1921 echo "== fstype is zfs =="
1924 local poolname_list="t32fs-mdt1 t32fs-ost1"
1926 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1928 for poolname in $poolname_list; do
1930 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
1931 $ZPOOL import -f -d $tmp $poolname"
1934 # upgrade zpool to latest supported features, including
1935 # dnode quota accounting in 0.7.0
1936 $r "$ZPOOL upgrade -a"
1938 mdt_dev=t32fs-mdt1/mdt1
1939 ost_dev=t32fs-ost1/ost1
1940 ! $mdt2_is_available || mdt2_dev=t32fs-mdt2/mdt2
1941 ! $ost2_is_available || ost2_dev=t32fs-ost2/ost2
1942 wait_update_facet $SINGLEMDS "$ZPOOL list |
1943 awk '/^t32fs-mdt1/ { print \\\$1 }'" "t32fs-mdt1" || {
1944 error_noexit "import zfs pool failed"
1947 elif [ "$project_upgrade" != "no" ]; then
1948 echo "== project upgrade =="
1949 ! $r tune2fs -O project $mdt_dev &&
1950 error_noexit "enable project on mdt0 failed" &&
1952 $mdt2_is_available && ! $r tune2fs "-O project" $mdt2_dev &&
1953 error_noexit "enable project on mdt failed" &&
1955 ! $r tune2fs -O project $ost_dev &&
1956 error_noexit "enable project on ost failed" &&
1958 $ost2_is_available && ! $r tune2fs -O project $ost2_dev &&
1959 error_noexit "enable project on ost2 failed" &&
1963 $r $LCTL set_param debug="$PTLDEBUG"
1965 $r $TUNEFS --dryrun $mdt_dev || {
1967 error_noexit "tunefs.lustre before mounting the MDT"
1971 if $mdt2_is_available; then
1972 $r $TUNEFS --dryrun $mdt2_dev || {
1974 error_noexit "tunefs.lustre before mounting the MDT"
1979 if [ "$writeconf" ]; then
1980 echo "== writeconf mdt tunefs and quota =="
1982 if [ "$mds1_FSTYPE" == ldiskfs ]; then
1984 $r $TUNEFS --quota $mdt_dev || {
1986 error_noexit "Enable mdt quota feature"
1990 if [ "$project_quota_upgrade" != "no" ]; then
1991 echo "== enable mdt2 project quota =="
1992 $r $TUNE2FS -O project $mdt_dev || {
1994 error_noexit "tune2fs $mdt_dev failed"
1999 if $mdt2_is_available; then
2000 $r $TUNEFS --quota $mdt2_dev || {
2002 error_noexit "Enable mdt2 quota feature"
2005 if [ "$project_quota_upgrade" != "no" ]; then
2006 echo "== enable mdt2 project quota =="
2007 $r $TUNE2FS -O project $mdt2_dev || {
2010 "tune2fs $mdt2_dev failed"
2017 echo "== replace nids =="
2018 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
2019 [[ "$MGS_VERSION" -ge $(version_code 2.3.59) ]] ||
2020 skip "LU-2200: Cannot run over IB w/o lctl replace_nids "
2021 "(Need MGS version at least 2.3.59)"
2024 local osthost=$(facet_active_host ost1)
2025 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
2028 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2031 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt
2032 $r $LCTL replace_nids $fsname-OST0000 $ostnid
2033 if $ost2_is_available; then
2034 $r $LCTL replace_nids $fsname-OST0001 $ostnid
2036 $r $LCTL replace_nids $fsname-MDT0000 $nid
2037 if $mdt2_is_available; then
2038 $r $LCTL replace_nids $fsname-MDT0001 $nid
2040 $r $UMOUNT $tmp/mnt/mdt
2042 mopts=exclude=$fsname-OST0000
2043 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2048 t32_wait_til_devices_gone $node
2050 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2052 error_noexit "Mounting the MDT"
2055 shall_cleanup_mdt=true
2057 if $mdt2_is_available; then
2058 echo "== mdt2 available =="
2059 mopts=mgsnode=$nid,$mopts
2060 $r $MOUNT_CMD -o $mopts $mdt2_dev $tmp/mnt/mdt1 || {
2062 error_noexit "Mounting the MDT"
2065 shall_cleanup_mdt1=true
2066 echo "== mount new MDT $mdt2_dev =="
2068 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 || {
2069 error_noexit "enable remote dir create failed"
2073 elif [ "$dne_upgrade" != "no" ]; then
2074 local fs2mdsdev=$(mdsdevname 1_2)
2075 local fs2mdsvdev=$(mdsvdevname 1_2)
2077 echo "== mkfs new MDT on ${fs2mdsdev} =="
2078 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2079 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
2082 add $SINGLEMDS $(mds2failover_HOST="" \
2083 mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
2084 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
2085 error_noexit "Mkfs new MDT failed"
2089 [[ "$mds1_FSTYPE" != zfs ]] || import_zpool mds1
2091 $r $TUNEFS --dryrun $fs2mdsdev || {
2092 error_noexit "tunefs.lustre before mounting the MDT"
2096 echo "== mount new MDT....$fs2mdsdev =="
2097 $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
2098 error_noexit "mount mdt1 failed"
2102 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
2103 error_noexit "enable remote dir create failed"
2105 shall_cleanup_mdt1=true
2108 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
2109 error_noexit "Getting MDT UUID"
2112 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
2113 error_noexit "Unexpected MDT UUID: \"$uuid\""
2117 $r $TUNEFS --dryrun $ost_dev || {
2118 error_noexit "tunefs.lustre before mounting the OST"
2122 if $ost2_is_available; then
2123 $r $TUNEFS --dryrun $ost2_dev || {
2124 error_noexit "tunefs.lustre before mounting the OST"
2129 if [ "$writeconf" ]; then
2130 echo "== writeconf and ost quota tunefs =="
2131 mopts=mgsnode=$nid,$writeconf
2132 if [ "$ost1_FSTYPE" == ldiskfs ]; then
2134 $r $TUNEFS --quota $ost_dev || {
2136 error_noexit "Enable ost quota feature"
2140 if [ "$project_quota_upgrade" != "no" ]; then
2141 echo "== enable ost project quota =="
2142 $r $TUNE2FS -O project $ost_dev || {
2144 error_noexit "tune2fs $ost_dev failed"
2149 if $ost2_is_available; then
2150 $r $TUNEFS --quota $ost2_dev || {
2152 error_noexit "Enable ost2 quota feature"
2155 if [ "$project_quota_upgrade" != "no" ]; then
2156 echo "== enable ost2 project quota =="
2157 $r $TUNE2FS -O project $ost2_dev || {
2160 "tune2fs $ost2_dev failed"
2168 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2173 $r $MOUNT_CMD -onomgs -o$mopts $ost_dev $tmp/mnt/ost || {
2174 error_noexit "Mounting the OST"
2178 if $ost2_is_available; then
2179 $r $MOUNT_CMD -onomgs -o$mopts $ost2_dev $tmp/mnt/ost1 || {
2180 error_noexit "Mounting the OST2"
2183 shall_cleanup_ost1=true
2186 shall_cleanup_ost=true
2188 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
2189 error_noexit "Getting OST UUID"
2192 if [ "$uuid" != $fsname-OST0000_UUID ]; then
2193 error_noexit "Unexpected OST UUID: \"$uuid\""
2197 if $ost2_is_available; then
2198 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0001.uuid) ||{
2199 error_noexit "Getting OST1 UUID"
2202 if [ "$uuid" != $fsname-OST0001_UUID ]; then
2203 error_noexit "Unexpected OST1 UUID: \"$uuid\""
2208 if [[ $PERM_CMD =~ "set_param" ]]; then
2209 echo "== perm cmd = '$PERM_CMD' =="
2210 $r $PERM_CMD osc.$fsname-OST0000*.import=connection=$nid || {
2211 error_noexit "Setting OST \"failover.node\""
2214 $r $PERM_CMD mdc.$fsname-MDT0000*.import=connection=$nid || {
2215 error_noexit "Setting MDT \"failover.node\""
2218 $r $PERM_CMD osc.$fsname-OST0000-*.max_dirty_mb=15 || {
2219 error_noexit "Setting \"max_dirty_mb\""
2222 $r $PERM_CMD mdc.$fsname-MDT0000-*.max_rpcs_in_flight=9 || {
2223 error_noexit "Setting \"max_rpcs_in_flight\""
2226 $r $PERM_CMD lov.$fsname-MDT0000-*.stripesize=4M || {
2227 error_noexit "Setting \"lov.stripesize\""
2230 $r $PERM_CMD mdd.$fsname-MDT0000-*.atime_diff=70 || {
2231 error_noexit "Setting \"mdd.atime_diff\""
2235 echo "== perm cmd = '$PERM_CMD' =="
2236 $r $PERM_CMD $fsname-OST0000.failover.node=$nid || {
2237 error_noexit "Setting OST \"failover.node\""
2241 $r $PERM_CMD $fsname-MDT0000.failover.node=$nid || {
2242 error_noexit "Setting MDT \"failover.node\""
2246 $r $PERM_CMD $fsname-OST0000.osc.max_dirty_mb=15 || {
2247 error_noexit "Setting \"max_dirty_mb\""
2250 $r $PERM_CMD $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
2251 error_noexit "Setting \"max_rpcs_in_flight\""
2254 $r $PERM_CMD $fsname-MDT0000.lov.stripesize=4M || {
2255 error_noexit "Setting \"lov.stripesize\""
2258 $r $PERM_CMD $fsname-MDT0000.mdd.atime_diff=70 || {
2259 error_noexit "Setting \"mdd.atime_diff\""
2264 $r $LCTL pool_new $fsname.interop || {
2265 error_noexit "Setting \"interop\""
2269 if [ "$ff_convert" != "no" -a "$ost1_FSTYPE" == ldiskfs ]; then
2270 echo "== ff convert ($ff_convert) =="
2271 $r $LCTL lfsck_start -M $fsname-OST0000 || {
2272 error_noexit "Start OI scrub on OST0"
2276 # The oi_scrub should be on ost1, but for test_32(),
2277 # all on the SINGLEMDS.
2278 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
2279 osd-ldiskfs.$fsname-OST0000.oi_scrub |
2280 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
2281 error_noexit "Failed to get the expected 'completed'"
2285 local UPDATED=$($r $LCTL get_param -n \
2286 osd-ldiskfs.$fsname-OST0000.oi_scrub |
2287 awk '/^updated/ { print $2 }')
2288 [ $UPDATED -ge 1 ] || {
2289 error_noexit "Only $UPDATED objects have been converted"
2294 if [[ "$dne_upgrade" != "no" ]] || $mdt2_is_available; then
2295 echo "== dne upgrate ($dne_upgrade) or mdt2 available =="
2296 if [[ $PERM_CMD =~ "set_param" ]]; then
2297 $r $PERM_CMD mdc.$fsname-MDT0001*.import=connection=$nid || {
2298 error_noexit "Setting MDT1 \"failover.node\""
2302 $r $PERM_CMD mdc.$fsname-MDT0001-*.max_rpcs_in_flight=9 || {
2303 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
2306 $r $PERM_CMD lov.$fsname-MDT0001-*.stripesize=4M || {
2307 error_noexit "Setting MDT1 \"lov.stripesize\""
2311 $r $PERM_CMD $fsname-MDT0001.failover.node=$nid || {
2312 error_noexit "Setting MDT1 \"failover.node\""
2315 $r $PERM_CMD $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
2316 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
2319 $r $PERM_CMD $fsname-MDT0001.lov.stripesize=4M || {
2320 error_noexit "Setting MDT1 \"lov.stripesize\""
2326 if [ "$writeconf" ]; then
2327 echo "== writeconf and client mount =="
2328 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
2329 error_noexit "Mounting the client"
2333 shall_cleanup_lustre=true
2334 $r $LCTL set_param debug="$PTLDEBUG"
2336 t32_verify_quota $SINGLEMDS $fsname $tmp/mnt/lustre || {
2337 error_noexit "verify quota failed"
2341 if $r test -f $tmp/list; then
2342 echo "== list verification =="
2344 # There is not a Test Framework API to copy files to or
2345 # from a remote node.
2347 # LU-2393 - do both sorts on same node to ensure locale
2349 local list_file=$tmp/list
2351 if $mdt2_is_available; then
2352 if [[ -d $tmp/mnt/lustre/striped_dir_old ]] &&
2353 $r test -f $tmp/list2; then
2354 list_file=$tmp/list2
2355 pushd $tmp/mnt/lustre/striped_dir_old
2357 pushd $tmp/mnt/lustre/remote_dir
2360 pushd $tmp/mnt/lustre
2362 $r cat $list_file | sort -k 6 >$tmp/list.orig
2363 ls -Rni --time-style=+%s | sort -k 6 |
2364 sed 's/\. / /' >$tmp/list || {
2370 # 32-bit and 64-bit clients use different algorithms to
2371 # convert FIDs into inode numbers. Hence, remove the
2372 # inode numbers from the lists, if the original list was
2373 # created on an architecture with different number of
2376 if [ $(t32_bits_per_long $(uname -m)) != \
2377 $(t32_bits_per_long $img_arch) ]; then
2378 echo "Different number of bits per \"long\"" \
2379 "from the disk image"
2380 for list in list.orig list; do
2381 sed -i -e 's/^[0-9]\+[ \t]\+//' \
2385 if ! diff -ub $tmp/list.orig $tmp/list; then
2386 error_noexit "list verification failed"
2390 echo "== list verification skipped =="
2393 #non-dom upgrade to dom
2394 if [ "$dom_upgrade" != "no" ]; then
2395 echo "== dom upgrade =="
2396 $LFS setstripe -E 1M -L mdt -E EOF $tmp/mnt/lustre/dom || {
2397 error_noexit "Verify DoM creation"
2400 [ $($LFS getstripe -L $tmp/mnt/lustre/dom) == "mdt" ] || {
2401 error_noexit "Verify a DoM file"
2404 dd if=/dev/urandom of=$tmp/mnt/lustre/dom bs=4096 \
2405 count=1 conv=fsync || {
2406 error_noexit "Cannot write to DoM file"
2409 [ $(stat -c%s $tmp/mnt/lustre/dom) == 4096 ] || {
2410 error_noexit "DoM: bad size after write"
2413 rm $tmp/mnt/lustre/dom
2415 set_persistent_param_and_check mds \
2416 "lod.*MDT0000*.dom_stripesize" \
2417 "$fsname-MDT0000.lod.dom_stripesize" 0 || {
2418 error_noexit "Changing \"dom_stripesize\""
2424 #$LFS setstripe -E 1M -L mdt -E -1 -S 4M $dom_file
2425 if [ "$dom_new_upgrade" != "no" ]; then
2426 if ! $mdt2_is_available; then
2427 dom_file=$tmp/mnt/lustre/dom_dir/dom_file
2429 echo "== check DoM file can be accessed =="
2430 [ $($LFS getstripe -I1 -L $dom_file) == "mdt" ] || {
2431 error_noexit "Verify a DoM file"
2434 [ $(stat -c%s $dom_file) == $((2 * 1024 * 1024)) ] || {
2435 error_noexit "DoM: bad size after write"
2440 if [ "$flr_upgrade" != "no" ]; then
2441 if ! $mdt2_is_available; then
2442 flr_file=$tmp/mnt/lustre/flr_dir/flr_file
2444 local mirror_count=$($LFS getstripe -N $flr_file)
2445 echo "== check FLR file =="
2446 [ $mirror_count == 2 ] || {
2447 error_noexit "FLR mirror count wrong"
2452 if [ "$dne_upgrade" != "no" ]; then
2453 echo "== dne upgrade striped_dir =="
2454 $LFS mkdir -i 1 -c2 $tmp/mnt/lustre/striped_dir || {
2455 error_noexit "set striped dir failed"
2458 $LFS setdirstripe -D -c2 $tmp/mnt/lustre/striped_dir
2460 pushd $tmp/mnt/lustre
2461 tar -c --exclude=./striped_dir \
2462 --exclude=./striped_dir_old \
2463 --exclude=./remote_dir -f - .|
2464 tar -xvf - -C striped_dir 1>/dev/null || {
2465 error_noexit "cp to striped dir failed"
2471 if [ "$pfl_upgrade" != "no" ]; then
2478 echo "== check PFL file =="
2479 if ! $mdt2_is_available; then
2480 pfl_file=$tmp/mnt/lustre/pfl_dir/pfl_file
2482 comp_cnt=$($LFS getstripe --component-count $pfl_file)
2483 [ $comp_cnt == 2 ] || {
2484 error_noexit "wrong comp_cnt $comp_cnt"
2488 comp_size=$($LFS getstripe -I1 -E $pfl_file)
2489 [ $comp_size == 2097152 ] || {
2490 error_noexit "wrong component size $comp_size"
2494 comp_id=$($LFS getstripe -I1 -i $pfl_file)
2495 [ $comp_id == 0 ] || {
2496 error_noexit "wrong comp id $comp_id"
2500 comp_id=$($LFS getstripe -I2 -i $pfl_file)
2501 [ $comp_id -eq 1 ] || {
2502 error_noexit "wrong comp id $comp_id"
2506 stripe_size=$($LFS getstripe -I1 -S $pfl_file)
2507 [ $stripe_size -eq 1048576 ] || {
2508 error_noexit "wrong stripe size $stripe_size"
2512 stripe_size=$($LFS getstripe -I2 -S $pfl_file)
2513 [ $comp_size -eq 2097152 ] || {
2514 error_noexit "wrong component size $comp_size"
2519 if [ "$project_quota_upgrade" != "no" ]; then
2520 if ! $mdt2_is_available; then
2521 quota_dir=$tmp/mnt/lustre/project_quota_dir
2525 echo "== check Project Quota =="
2526 hardlimit=$(getquota -p $T32_PRJID global 3 \
2528 [ $hardlimit == $T32_PROLIMIT ] || {
2529 error_noexit "wrong hardlimit $hardlimit"
2534 # If it is upgrade from DNE (2.5), then rename the remote dir,
2535 # which is created in 2.5 to striped dir.
2536 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]]; then
2537 echo "== rename remote_dir =="
2538 stripe_index=$($LFS getdirstripe -i \
2539 $tmp/mnt/lustre/remote_dir)
2541 [[ $stripe_index -eq 1 ]] || {
2542 error_noexit "get index \"$stripe_index\"" \
2543 "from remote dir failed"
2546 mv $tmp/mnt/lustre/remote_dir \
2547 $tmp/mnt/lustre/striped_dir/ || {
2548 error_noexit "mv remote dir failed"
2553 # If it is upgraded from DNE (2.7), then move the striped dir
2554 # which was created in 2.7 to the new striped dir.
2555 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]] &&
2556 echo "== rename striped dir =="
2557 [[ -d $tmp/mnt/lustre/striped_dir_old ]]; then
2558 stripe_count=$($LFS getdirstripe -c \
2559 $tmp/mnt/lustre/striped_dir_old)
2560 [[ $stripe_count -eq 2 ]] || {
2561 error_noexit "get count $stripe_count" \
2562 "from striped dir failed"
2565 mv $tmp/mnt/lustre/striped_dir_old \
2566 $tmp/mnt/lustre/striped_dir/ || {
2567 error_noexit "mv striped dir failed"
2573 $r $LCTL set_param -n osd*.*.force_sync=1
2574 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
2575 error_noexit "dd failed"
2578 rm -rf $tmp/mnt/lustre/tmp_file || {
2579 error_noexit "rm failed"
2583 if $r test -f $tmp/sha1sums; then
2584 echo "== checking sha1sums =="
2585 # LU-2393 - do both sorts on same node to ensure locale
2587 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
2588 if [[ "$dne_upgrade" != "no" ]]; then
2589 if [[ -d $tmp/mnt/lustre/striped_dir/remote_dir ]]; then
2590 pushd $tmp/mnt/lustre/striped_dir/remote_dir
2592 pushd $tmp/mnt/lustre/striped_dir
2594 elif [[ "$pfl_upgrade" != "no" ]] ||
2595 [[ "$flr_upgrade" != "no" ]] ||
2596 [[ "$dom_new_upgrade" != "no" ]] ||
2597 [[ "$project_quota_upgrade" != "no" ]] ||
2598 [[ -d $tmp/mnt/lustre/remote_dir ]]; then
2599 pushd $tmp/mnt/lustre/remote_dir
2601 pushd $tmp/mnt/lustre
2604 find ! -path "*remote_dir*" ! -path "*striped_dir*" \
2605 ! -name .lustre -type f -exec sha1sum {} \; |
2606 sort -k 2 >$tmp/sha1sums || {
2608 error_noexit "sha1sum"
2614 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
2615 echo "==** find $save files start **=="
2616 find $save -type f -ls
2617 echo "==** find $save files finish **=="
2618 error_noexit "sha1sum verification failed"
2622 # if upgrade from DNE(2.5), then check remote directory
2623 # if upgrade from DNE(2.7), then check striped directory
2624 if $mdt2_is_available &&
2625 [[ "$dne_upgrade" != "no" ]]; then
2626 echo "== mdt2 upgrade sha1sum check =="
2627 local new_dir="$tmp/mnt/lustre/striped_dir"
2628 local striped_dir_old="$new_dir/striped_dir_old"
2630 local dir_list="$new_dir/remote_dir"
2631 [[ ! -d $triped_dir_old ]] ||
2632 dir_list+=" $striped_dir_old"
2634 for dir in $dir_list; do
2635 echo "== dir = $dir =="
2637 find ! -name .lustre -type f \
2638 -exec sha1sum {} \; |
2639 sort -k 2 >$tmp/sha1sums || {
2642 error_noexit "sha1sum"
2646 if ! diff -ub $tmp/sha1sums.orig \
2649 error_noexit "sha1sum $dir" \
2656 echo "== sha1sum verification skipped =="
2659 # PFL write test after sha1sum check
2660 if [ "$pfl_upgrade" != "no" ]; then
2661 local rw_len=$((3 * 1034 * 1024))
2662 if ! $mdt2_is_available; then
2663 pfl_file=$tmp/mnt/lustre/pfl_dir/pfl_file
2665 small_write $pfl_file $rw_len || {
2666 error_noexit "PFL RW Failed"
2671 if [ "$dne_upgrade" != "no" ]; then
2672 rm -rf $tmp/mnt/lustre/striped_dir || {
2673 error_noexit "remove remote dir failed"
2678 # migrate files/dirs to remote MDT, then move them back
2679 if [ "$MDS1_VERSION" -ge $(version_code 2.7.50) -a \
2680 $dne_upgrade != "no" ]; then
2681 $r $LCTL set_param -n \
2682 mdt.${fsname}*.enable_remote_dir=1 2>/dev/null
2684 echo "== test migration =="
2685 pushd $tmp/mnt/lustre
2686 for dir in $(find ! -name .lustre ! -name . -type d); do
2687 mdt_index=$($LFS getdirstripe -i $dir)
2688 stripe_cnt=$($LFS getdirstripe -c $dir)
2689 if [ $mdt_index = 0 -a $stripe_cnt -le 1 ]; then
2690 $LFS migrate -m 1 $dir || {
2692 error_noexit "migrate MDT1 failed"
2698 for dir in $(find ! -name . ! -name .lustre -type d); do
2699 mdt_index=$($LFS getdirstripe -i $dir)
2700 stripe_cnt=$($LFS getdirstripe -c $dir)
2701 if [ $mdt_index = 1 -a $stripe_cnt -le 1 ]; then
2702 $LFS migrate -m 0 $dir || {
2704 error_noexit "migrate MDT0 failed"
2713 # When adding new data verification tests, please check for
2714 # the presence of the required reference files first, like
2715 # the "sha1sums" and "list" tests above, to avoid the need to
2716 # regenerate every image for each test addition.
2719 nrpcs_orig=$($LCTL get_param \
2720 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
2721 error_noexit "Getting \"max_rpcs_in_flight\""
2724 nrpcs=$((nrpcs_orig + 5))
2726 set_persistent_param_and_check client \
2727 "mdc.$fsname-MDT0000*.max_rpcs_in_flight" \
2728 "$fsname-MDT0000.mdc.max_rpcs_in_flight" $nrpcs || {
2729 error_noexit "Changing \"max_rpcs_in_flight\""
2733 umount $tmp/mnt/lustre || {
2734 error_noexit "Unmounting the client"
2737 shall_cleanup_lustre=false
2739 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
2740 error_noexit "Mounting the client"
2744 if [ "$mds1_FSTYPE" == ldiskfs -a \
2745 "$project_upgrade" != "no" ]; then
2746 ! $LFS project -d -p 1 $tmp/mnt/lustre/* &&
2747 error_noexit "set project failed" &&
2751 [[ $(do_facet mds1 pgrep orph_.*-MDD | wc -l) == 0 ]] ||
2752 error "MDD orphan cleanup thread not quit"
2754 umount $tmp/mnt/lustre || {
2755 error_noexit "Unmounting the client"
2759 if [[ "$dne_upgrade" != "no" ]] || $mdt2_is_available; then
2760 $r $UMOUNT $tmp/mnt/mdt1 || {
2761 error_noexit "Unmounting the MDT2"
2764 if [[ "$mds1_FSTYPE" == zfs ]]; then
2765 $r "$ZPOOL export t32fs-mdt2"
2767 shall_cleanup_mdt1=false
2770 $r $UMOUNT $tmp/mnt/mdt || {
2771 error_noexit "Unmounting the MDT"
2774 if [[ "$mds1_FSTYPE" == zfs ]]; then
2775 $r "$ZPOOL export t32fs-mdt1"
2777 shall_cleanup_mdt=false
2779 $r $UMOUNT $tmp/mnt/ost || {
2780 error_noexit "Unmounting the OST"
2783 if $ost2_is_available; then
2784 $r $UMOUNT $tmp/mnt/ost1 || {
2785 error_noexit "Unmounting the OST1"
2788 shall_cleanup_ost1=false
2791 if [[ $ost1_FSTYPE == zfs ]]; then
2792 $r "$ZPOOL export t32fs-ost1"
2793 if $ost2_is_available; then
2794 $r "$ZPOOL export t32fs-ost2"
2797 shall_cleanup_ost=false
2799 t32_reload_modules $node || {
2800 error_noexit "Reloading modules"
2804 if [[ "$mds1_FSTYPE" == zfs ]]; then
2805 local poolname=t32fs-mdt1
2807 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
2808 $ZPOOL import -f -d $tmp $poolname"
2810 # upgrade zpool to latest supported features,
2811 # including dnode quota accounting in 0.7.0
2812 $r "$ZPOOL upgrade $poolname"
2815 # mount a second time to make sure we didnt leave upgrade flag on
2816 $r $TUNEFS --dryrun $mdt_dev || {
2818 error_noexit "tunefs.lustre before remounting the MDT"
2822 mopts=exclude=$fsname-OST0000
2823 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2826 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2827 error_noexit "Remounting the MDT"
2830 shall_cleanup_mdt=true
2840 for tarball in $tarballs; do
2841 banner "testing $tarball upgrade"
2842 t32_test $tarball || let "rc += $?"
2846 run_test 32a "Upgrade (not live)"
2854 for tarball in $tarballs; do
2855 banner "testing $tarball upgrade with writeconf"
2856 t32_test $tarball writeconf || let "rc += $?"
2860 run_test 32b "Upgrade with writeconf"
2868 for tarball in $tarballs; do
2869 banner "testing $tarball upgrade with DNE"
2871 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
2875 run_test 32c "dne upgrade test"
2883 for tarball in $tarballs; do
2884 banner "testing $tarball upgrade with ff and project upgrade"
2885 project_upgrade="no"
2886 [[ "$MDS1_VERSION" -ge $(version_code 2.13.54) ]] &&
2887 [[ "$tarball" =~ "disk2_4-ldiskfs" ]] &&
2888 project_upgrade="yes"
2889 project_upgrade=$project_upgrade ff_convert=yes t32_test \
2894 run_test 32d "convert ff and project quota upgrade test"
2897 [[ "$MDS1_VERSION" -ge $(version_code 2.10.56) ]] ||
2898 skip "Need MDS version at least 2.10.56"
2905 for tarball in $tarballs; do
2906 [[ "$tarball" =~ "2_9" ]] ||
2907 { echo "skip $(basename $tarball)"; continue; }
2909 banner "testing $tarball upgrade with DoM"
2910 dom_upgrade=yes t32_test $tarball writeconf || let "rc += $?"
2914 run_test 32e "dom upgrade test"
2917 [[ $MDS1_VERSION -ge $(version_code 2.10.56) ]] ||
2918 skip "Need MDS version at least 2.10.56"
2925 for tarball in $tarballs; do
2926 [[ "$tarball" =~ "2_10" ]] ||
2927 { echo "skip $(basename $tarball)"; continue; }
2928 pfl_upgrade=yes project_quota_upgrade=yes \
2929 t32_test $tarball writeconf || let "rc += $?"
2933 run_test 32f "pfl upgrade test"
2936 [[ $MDS1_VERSION -ge $(version_code 2.10.56) ]] ||
2937 skip "Need MDS version at least 2.10.56"
2944 for tarball in $tarballs; do
2945 [[ $tarball =~ "2_12" ]] ||
2946 { echo "skip $(basename $tarball)"; continue; }
2947 flr_upgrade=yes dom_new_upgrade=yes \
2948 t32_test $tarball writeconf || let "rc += $?"
2952 run_test 32g "flr/dom upgrade test"
2954 test_33a() { # bug 12333, was test_33
2955 local FSNAME2=test-$testnum
2956 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2959 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
2961 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
2962 local dev=${SINGLEMDS}_dev
2963 local MDSDEV=${!dev}
2964 is_blkdev $SINGLEMDS $MDSDEV &&
2965 skip_env "mixed loopback and real device not working"
2968 local fs2mdsdev=$(mdsdevname 1_2)
2969 local fs2ostdev=$(ostdevname 1_2)
2970 local fs2mdsvdev=$(mdsvdevname 1_2)
2971 local fs2ostvdev=$(ostvdevname 1_2)
2973 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2974 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
2977 if combined_mgs_mds; then
2978 local mgs_flag="--mgs"
2981 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --fsname=${FSNAME2} \
2982 --reformat $mgs_flag $mkfsoptions $fs2mdsdev $fs2mdsvdev ||
2984 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2985 --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \
2986 $fs2ostvdev || exit 10
2988 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
2989 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2991 if [[ $PERM_CMD == *"set_param -P"* ]]; then
2992 do_facet mgs "$PERM_CMD timeout=200" ||
2993 error "$PERM_CMD timeout=200 failed"
2995 do_facet mgs "$PERM_CMD $FSNAME2.sys.timeout=200" ||
2996 error "$PERM_CMD $FSNAME2.sys.timeout=200 failed"
2998 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2999 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
3002 cp /etc/hosts $MOUNT2/ || error "copy /etc/hosts $MOUNT2/ failed"
3003 $LFS getstripe $MOUNT2/hosts ||
3004 error "$LFS getstripe $MOUNT2/hosts failed"
3009 cleanup_nocli || error "cleanup_nocli failed with $?"
3011 run_test 33a "Mount ost with a large index number"
3013 test_33b() { # was test_34
3016 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
3017 # Drop lock cancelation reply during umount
3018 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
3019 do_facet client $LCTL set_param fail_loc=0x80000304
3020 #lctl set_param debug=-1
3021 umount_client $MOUNT
3022 cleanup || error "cleanup failed with $?"
3024 run_test 33b "Drop cancel during umount"
3028 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
3029 manual_umount_client
3031 do_facet client killall -USR1 multiop
3032 if [ $rc -eq 0 ]; then
3033 error "umount not fail!"
3036 cleanup || error "cleanup failed with rc $?"
3038 run_test 34a "umount with opened file should be fail"
3042 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3043 stop_mds || error "Unable to stop MDS"
3045 manual_umount_client --force || error "mtab after failed umount with $?"
3047 cleanup || error "cleanup failed with $?"
3049 run_test 34b "force umount with failed mds should be normal"
3053 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3054 stop_ost || error "Unable to stop OST1"
3056 manual_umount_client --force || error "mtab after failed umount with $?"
3058 cleanup || error "cleanup failed with $?"
3060 run_test 34c "force umount with failed ost should be normal"
3062 test_35a() { # bug 12459
3065 DBG_SAVE="`$LCTL get_param -n debug`"
3066 $LCTL set_param debug="ha"
3068 log "Set up a fake failnode for the MDS"
3070 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
3071 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
3073 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3074 do_facet mgs "$PERM_CMD \
3075 mdc.*${device}*.import=connection=$(h2nettype $FAKENID)" ||
3076 error "Setting mdc.*${device}*.import=connection=\
3077 $(h2nettype $FAKENID) failed."
3079 do_facet mgs "$PERM_CMD \
3080 ${device}.failover.node=$(h2nettype $FAKENID)" ||
3081 error "Setting ${device}.failover.node=\
3082 $(h2nettype $FAKENID) failed."
3084 log "Wait for RECONNECT_INTERVAL seconds (10s)"
3087 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
3090 log "Stopping the MDT: $device"
3091 stop_mdt 1 || error "MDT0 stop fail"
3093 df $MOUNT > /dev/null 2>&1 &
3095 log "Restarting the MDT: $device"
3096 start_mdt 1 || error "MDT0 start fail"
3097 log "Wait for df ($DFPID) ... "
3100 $LCTL set_param debug="$DBG_SAVE"
3102 # retrieve from the log the first server that the client tried to
3103 # contact after the connection loss
3104 $LCTL dk $TMP/lustre-log-$TESTNAME.log
3105 NEXTCONN=`awk "/${MSG}/ {start = 1;}
3106 /import_select_connection.*$device-mdc.* using connection/ {
3108 if (\\\$NF ~ /$FAKENID/)
3114 }" $TMP/lustre-log-$TESTNAME.log`
3115 [ "$NEXTCONN" != "0" ] &&
3116 error "Tried to connect to ${NEXTCONN} not last active server"
3117 cleanup || error "cleanup failed with $?"
3118 # remove nid settings
3119 writeconf_or_reformat
3121 run_test 35a "Reconnect to the last active server first"
3123 test_35b() { # bug 18674
3124 remote_mds || skip "local MDS"
3128 $LCTL set_param debug="ha"
3130 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
3133 log "Set up a fake failnode for the MDS"
3135 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
3136 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
3138 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3139 do_facet mgs "$PERM_CMD \
3140 mdc.*${device}*.import=connection=$(h2nettype $FAKENID)" ||
3141 error "Set mdc.*${device}*.import=connection=\
3142 $(h2nettype $FAKENID) failed"
3144 do_facet mgs "$PERM_CMD \
3145 ${device}.failover.node=$(h2nettype $FAKENID)" ||
3146 error "Set ${device}.failover.node=\
3147 $(h2nettype $FAKENID) failed"
3150 local at_max_saved=0
3151 # adaptive timeouts may prevent seeing the issue
3152 if at_is_enabled; then
3153 at_max_saved=$(at_max_get mds)
3154 at_max_set 0 mds client
3157 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
3159 log "Injecting EBUSY on MDS"
3160 # Setting OBD_FAIL_MDS_RESEND=0x136
3161 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" ||
3162 error "unable to set param fail_loc=0x80000136"
3164 $LCTL set_param mdc.${FSNAME}*.stats=clear
3166 log "Creating a test file and stat it"
3167 touch $MOUNT/$tdir/$tfile || error "touch $MOUNT/$tdir/$tfile failed"
3168 stat $MOUNT/$tdir/$tfile
3170 log "Stop injecting EBUSY on MDS"
3171 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" ||
3172 error "unable to set param fail_loc=0"
3173 rm -f $MOUNT/$tdir/$tfile || error "remove $MOUNT/$tdir/$tfile failed"
3176 # restore adaptive timeout
3177 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
3179 $LCTL dk $TMP/lustre-log-$TESTNAME.log
3181 CONNCNT=$($LCTL get_param mdc.${FSNAME}*.stats |
3182 awk '/mds_connect/{print $2}')
3184 # retrieve from the log if the client has ever tried to
3185 # contact the fake server after the loss of connection
3186 FAILCONN=`awk "BEGIN {ret = 0;}
3187 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
3189 if (\\\$NF ~ /$FAKENID/) {
3194 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
3196 [ "$FAILCONN" == "0" ] &&
3197 error "The client reconnection has not been triggered"
3198 [ "$FAILCONN" == "2" ] &&
3199 error "Primary server busy, client reconnect to failover failed"
3202 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
3203 # Reconnects are supposed to be rate limited to one every 5s
3204 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] &&
3205 error "Too many reconnects $CONNCNT"
3207 cleanup || error "cleanup failed with $?"
3208 # remove nid settings
3209 writeconf_or_reformat
3211 run_test 35b "Continue reconnection retries, if the active server is busy"
3214 [ $OSTCOUNT -lt 2 ] && skip_env "needs >= 2 OSTs"
3216 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] ||
3220 local FSNAME2=test1234
3221 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3223 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
3225 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
3226 is_blkdev $SINGLEMDS $MDSDEV &&
3227 skip_env "mixed loopback and real device not working"
3230 local fs2mdsdev=$(mdsdevname 1_2)
3231 local fs2ostdev=$(ostdevname 1_2)
3232 local fs3ostdev=$(ostdevname 2_2)
3233 local fs2mdsvdev=$(mdsvdevname 1_2)
3234 local fs2ostvdev=$(ostvdevname 1_2)
3235 local fs3ostvdev=$(ostvdevname 2_2)
3238 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
3239 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
3240 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
3241 # different one than the default value here.
3242 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
3243 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
3244 add fs3ost $(mkfs_opts ost2 ${fs3ostdev}) --mgsnode=$MGSNID \
3245 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
3247 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
3248 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
3249 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
3250 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
3251 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
3253 sleep 5 # until 11778 fixed
3255 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || error "dd failed"
3257 BKTOTAL=$($LCTL get_param -n obdfilter.*.kbytestotal |
3258 awk 'BEGIN{total=0}; {total+=$1}; END{print total}')
3259 BKFREE=$($LCTL get_param -n obdfilter.*.kbytesfree |
3260 awk 'BEGIN{free=0}; {free+=$1}; END{print free}')
3261 BKAVAIL=$($LCTL get_param -n obdfilter.*.kbytesavail |
3262 awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}')
3263 STRING=$(df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}')
3264 DFTOTAL=$(echo $STRING | cut -d, -f1)
3265 DFUSED=$(echo $STRING | cut -d, -f2)
3266 DFAVAIL=$(echo $STRING | cut -d, -f3)
3267 DFFREE=$(($DFTOTAL - $DFUSED))
3269 ALLOWANCE=$((64 * $OSTCOUNT))
3271 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
3272 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
3273 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
3276 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
3277 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
3278 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
3281 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
3282 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
3283 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
3288 stop fs3ost -f || error "unable to stop OST3"
3289 stop fs2ost -f || error "unable to stop OST2"
3290 stop fs2mds -f || error "unable to stop second MDS"
3291 unload_modules_conf || error "unable unload modules"
3294 run_test 36 "df report consistency on OSTs with different block size"
3297 local mntpt=$(facet_mntpt $SINGLEMDS)
3298 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
3299 local mdsdev_sym="$TMP/sym_mdt.img"
3300 local opts=$MDS_MOUNT_OPTS
3303 if [ "$mds1_FSTYPE" != ldiskfs ]; then
3304 skip "ldiskfs only test"
3307 echo "MDS : $mdsdev"
3308 echo "SYMLINK : $mdsdev_sym"
3309 do_facet $SINGLEMDS rm -f $mdsdev_sym
3311 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
3313 echo "mount symlink device - $mdsdev_sym"
3315 if ! do_facet $SINGLEMDS test -b $mdsdev; then
3316 opts=$(csa_add "$opts" -o loop)
3320 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
3321 $mdsdev_sym $mntpt 2>&1)
3324 echo mount_op=$mount_op
3326 do_facet $SINGLEMDS "$UMOUNT $mntpt && rm -f $mdsdev_sym"
3328 if $(echo $mount_op | grep -q "unable to set tunable"); then
3329 error "set tunables failed for symlink device"
3332 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
3334 run_test 37 "verify set tunables works for symlink device"
3336 test_38() { # bug 14222
3337 local mntpt=$(facet_mntpt $SINGLEMDS)
3342 local SRC="/etc /bin"
3343 local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT)
3344 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
3345 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3346 tar cf - $FILES | tar xf - -C $DIR/$tdir ||
3347 error "copying $SRC to $DIR/$tdir"
3349 umount_client $MOUNT || error "umount_client $MOUNT failed"
3350 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3351 stop_mds || error "Unable to stop MDS"
3352 log "delete lov_objid file on MDS"
3354 mount_fstype $SINGLEMDS || error "mount MDS failed (1)"
3356 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid; rm $mntpt/lov_objid"
3358 unmount_fstype $SINGLEMDS || error "umount failed (1)"
3360 # check create in mds_lov_connect
3361 start_mds || error "unable to start MDS"
3362 mount_client $MOUNT || error "mount_client $MOUNT failed"
3364 [ $V ] && log "verifying $DIR/$tdir/$f"
3365 diff -q $f $DIR/$tdir/$f || ERROR=y
3367 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3368 if [ "$ERROR" = "y" ]; then
3369 # check it's updates in sync
3370 umount_client $MOUNT
3372 mount_fstype $SIGNLEMDS
3373 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3374 unmount_fstype $SINGLEMDS
3375 error "old and new files are different after connect" || true
3377 touch $DIR/$tdir/f2 || error "f2 file create failed"
3379 # check it's updates in sync
3380 umount_client $MOUNT || error "second umount_client $MOUNT failed"
3383 mount_fstype $SINGLEMDS || error "mount MDS failed (3)"
3385 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3386 do_facet $SINGLEMDS dd if=/dev/zero of=$mntpt/lov_objid.clear count=8
3388 unmount_fstype $SINGLEMDS || error "umount failed (3)"
3390 start_mds || error "unable to start MDS"
3391 mount_client $MOUNT || error "mount_client $MOUNT failed"
3393 [ $V ] && log "verifying $DIR/$tdir/$f"
3394 diff -q $f $DIR/$tdir/$f || ERROR=y
3396 touch $DIR/$tdir/f3 || error "f3 file create failed"
3397 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3398 umount_client $MOUNT || error "third umount_client $MOUNT failed"
3400 mount_fstype $SINGLEMDS || error "mount MDS failed (4)"
3401 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3402 unmount_fstype $SINGLEMDS || error "umount failed (4)"
3404 [ "$ERROR" = "y" ] &&
3405 error "old and new files are different after sync" || true
3407 log "files compared the same"
3408 cleanup || error "cleanup failed with $?"
3410 run_test 38 "MDS recreates missing lov_objid file from OST data"
3415 cleanup || error "cleanup failed with $?"
3416 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
3417 error "memory leak detected" || true
3419 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
3421 test_40() { # bug 15759
3422 start_ost || error "Unable to start OST1"
3423 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
3424 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
3426 cleanup || error "cleanup failed with rc $?"
3428 run_test 40 "race during service thread startup"
3430 test_41a() { #bug 14134
3431 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3432 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
3433 skip "Loop devices does not work with nosvc option"
3436 combined_mgs_mds || skip "needs combined MGT and MDT device"
3438 start_mdt 1 -o nosvc -n
3439 if [ $MDSCOUNT -ge 2 ]; then
3440 for num in $(seq 2 $MDSCOUNT); do
3441 start_mdt $num || return
3444 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
3445 start_mdt 1 -o nomgs,force
3446 mount_client $MOUNT || error "mount_client $MOUNT failed"
3449 echo "blah blah" > $MOUNT/$tfile
3452 umount_client $MOUNT || error "umount_client $MOUNT failed"
3453 stop ost1 -f || error "unable to stop OST1"
3454 stop_mds || error "Unable to stop MDS"
3455 stop_mds || error "Unable to stop MDS on second try"
3457 run_test 41a "mount mds with --nosvc and --nomgs"
3460 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3461 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
3462 skip "Loop devices does not work with nosvc option"
3465 ! combined_mgs_mds && skip "needs combined mgs device"
3469 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3471 start_mdt 1 -o nosvc -n
3472 if [ $MDSCOUNT -ge 2 ]; then
3473 for num in $(seq 2 $MDSCOUNT); do
3474 start_mdt $num || return
3477 start_ost || error "Unable to start OST1"
3478 start_mdt 1 -o nomgs,force
3479 mount_client $MOUNT || error "mount_client $MOUNT failed"
3482 echo "blah blah" > $MOUNT/$tfile
3483 cat $MOUNT/$tfile || error "cat $MOUNT/$tfile failed"
3485 umount_client $MOUNT -f || error "umount_client $MOUNT failed"
3486 stop_ost || error "Unable to stop OST1"
3487 stop_mds || error "Unable to stop MDS"
3488 stop_mds || error "Unable to stop MDS on second try"
3490 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
3493 local oss_list=$(comma_list $(osts_nodes))
3495 [[ "$MDS1_VERSION" -ge $(version_code 2.6.52) ]] ||
3496 [[ "$MDS1_VERSION" -ge $(version_code 2.5.26) &&
3497 "$MDS1_VERSION" -lt $(version_code 2.5.50) ]] ||
3498 [[ "$MDS1_VERSION" -ge $(version_code 2.5.4) &&
3499 "$MDS1_VERSION" -lt $(version_code 2.5.11) ]] ||
3500 skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"
3502 # ensure mds1 ost1 have been created even if running sub-test standalone
3505 cleanup || error "cleanup failed"
3507 # using directly mount command instead of start() function to avoid
3508 # any side effect of // with others/externals tools/features
3509 # ("zpool import", ...)
3511 # MDT concurrent start
3513 LOAD_MODULES_REMOTE=true load_modules
3514 do_facet $SINGLEMDS "lsmod | grep -q libcfs" ||
3515 error "MDT concurrent start: libcfs module not loaded"
3517 local mds1dev=$(mdsdevname 1)
3518 local mds1mnt=$(facet_mntpt mds1)
3519 local mds1opts=$MDS_MOUNT_OPTS
3521 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3522 ! do_facet mds1 test -b $mds1dev; then
3523 mds1opts=$(csa_add "$mds1opts" -o loop)
3525 if [[ "$mds1_FSTYPE" == zfs ]]; then
3526 import_zpool mds1 || return ${PIPESTATUS[0]}
3529 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3530 do_facet mds1 "$LCTL set_param fail_loc=0x80000716"
3532 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts &
3535 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts
3539 do_facet mds1 "$LCTL set_param fail_loc=0x0"
3540 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3541 echo "1st MDT start succeed"
3542 echo "2nd MDT start failed with $rc2"
3543 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3544 echo "1st MDT start failed with $rc"
3545 echo "2nd MDT start succeed"
3548 error "unexpected concurrent MDT mounts result, rc=$rc rc2=$rc2"
3551 if [ $MDSCOUNT -ge 2 ]; then
3552 for num in $(seq 2 $MDSCOUNT); do
3553 start_mdt $num || return
3557 # OST concurrent start
3559 do_rpc_nodes $oss_list "lsmod | grep -q libcfs" ||
3560 error "OST concurrent start: libcfs module not loaded"
3562 local ost1dev=$(ostdevname 1)
3563 local ost1mnt=$(facet_mntpt ost1)
3564 local ost1opts=$OST_MOUNT_OPTS
3566 if [ "$ost1_FSTYPE" == ldiskfs ] &&
3567 ! do_facet ost1 test -b $ost1dev; then
3568 ost1opts=$(csa_add "$ost1opts" -o loop)
3570 if [[ "$ost1_FSTYPE" == zfs ]]; then
3571 import_zpool ost1 || return ${PIPESTATUS[0]}
3574 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3575 do_facet ost1 "$LCTL set_param fail_loc=0x80000716"
3577 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts &
3580 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts
3584 do_facet ost1 "$LCTL set_param fail_loc=0x0"
3585 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3586 echo "1st OST start succeed"
3587 echo "2nd OST start failed with $rc2"
3588 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3589 echo "1st OST start failed with $rc"
3590 echo "2nd OST start succeed"
3594 error "unexpected concurrent OST mounts result, rc=$rc rc2=$rc2"
3600 # verify everything ok
3605 error "MDT(s) start failed"
3613 error "OST(s) start failed"
3621 error "client start failed"
3628 error "client mount failed"
3632 run_test 41c "concurrent mounts of MDT/OST should all fail but one"
3634 test_42() { #bug 14693
3638 check_mount || error "client was not mounted"
3640 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3641 PARAM="llite.$FSNAME-*.some_wrong_param"
3643 PARAM="$FSNAME.llite.some_wrong_param"
3646 do_facet mgs $PERM_CMD $PARAM=10
3647 umount_client $MOUNT ||
3648 error "unmounting client failed with invalid llite param"
3649 mount_client $MOUNT ||
3650 error "mounting client failed with invalid llite param"
3652 do_facet mgs $PERM_CMD $PARAM=20
3653 cleanup || error "stopping $FSNAME failed with invalid sys param"
3655 check_mount || error "client was not mounted with invalid sys param"
3656 cleanup || error "stopping $FSNAME failed with invalid sys param"
3658 run_test 42 "allow client/server mount/unmount with invalid config param"
3661 [[ "$MGS_VERSION" -ge $(version_code 2.5.58) ]] ||
3662 skip "Need MDS version at least 2.5.58"
3663 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
3666 USER1=$(getent passwd | grep :$ID1:$ID1: | cut -d: -f1)
3667 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1"
3670 chmod ugo+x $DIR || error "chmod 0 failed"
3671 set_persistent_param_and_check mds1 \
3672 "mdt.$FSNAME-MDT0000.root_squash" \
3673 "$FSNAME.mdt.root_squash" \
3675 wait_update $HOSTNAME \
3676 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3678 error "check llite root_squash failed!"
3679 set_persistent_param_and_check mds1 \
3680 "mdt.$FSNAME-MDT0000.nosquash_nids" \
3681 "$FSNAME.mdt.nosquash_nids" \
3683 wait_update $HOSTNAME \
3684 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3686 error "check llite nosquash_nids failed!"
3689 # create set of test files
3691 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
3692 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
3693 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
3695 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
3696 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
3698 mkdir_on_mdt0 $DIR/$tdir-rootdir || error "mkdir failed"
3699 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
3700 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
3702 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
3703 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
3704 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
3707 # check root_squash:
3708 # set root squash UID:GID to RUNAS_ID
3709 # root should be able to access only files owned by RUNAS_ID
3711 set_persistent_param_and_check mds1 \
3712 "mdt.$FSNAME-MDT0000.root_squash" \
3713 "$FSNAME.mdt.root_squash" \
3714 "$RUNAS_ID:$RUNAS_ID"
3715 wait_update $HOSTNAME \
3716 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3717 "$RUNAS_ID:$RUNAS_ID" ||
3718 error "check llite root_squash failed!"
3720 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
3721 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3722 error "$ST: root read permission is denied"
3723 echo "$ST: root read permission is granted - ok"
3726 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3727 error "$ST: root write permission is denied"
3728 echo "$ST: root write permission is granted - ok"
3730 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3731 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3732 error "$ST: root read permission is granted"
3733 echo "$ST: root read permission is denied - ok"
3736 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3737 error "$ST: root write permission is granted"
3738 echo "$ST: root write permission is denied - ok"
3740 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3741 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null &&
3742 error "$ST: root unlink permission is granted"
3743 echo "$ST: root unlink permission is denied - ok"
3745 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null &&
3746 error "$ST: root create permission is granted"
3747 echo "$ST: root create permission is denied - ok"
3750 # check root_squash is enforced independently
3751 # of client cache content
3753 # access file by USER1, keep access open
3754 # root should be denied access to user file
3756 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
3760 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
3761 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3762 { kill $pid; error "$ST: root read permission is granted"; }
3763 echo "$ST: root read permission is denied - ok"
3766 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3767 { kill $pid; error "$ST: root write permission is granted"; }
3768 echo "$ST: root write permission is denied - ok"
3774 # check nosquash_nids:
3775 # put client's NID into nosquash_nids list,
3776 # root should be able to access root file after that
3778 local NIDLIST=$($LCTL list_nids all | tr '\n' ' ')
3779 NIDLIST="2@gni $NIDLIST 192.168.0.[2,10]@tcp"
3780 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
3781 set_persistent_param_and_check mds1 \
3782 "mdt.$FSNAME-MDT0000.nosquash_nids" \
3783 "$FSNAME-MDTall.mdt.nosquash_nids" \
3785 wait_update $HOSTNAME \
3786 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3788 error "check llite nosquash_nids failed!"
3790 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3791 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3792 error "$ST: root read permission is denied"
3793 echo "$ST: root read permission is granted - ok"
3796 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3797 error "$ST: root write permission is denied"
3798 echo "$ST: root write permission is granted - ok"
3800 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3801 rm $DIR/$tdir-rootdir/tfile-1 ||
3802 error "$ST: root unlink permission is denied"
3803 echo "$ST: root unlink permission is granted - ok"
3804 touch $DIR/$tdir-rootdir/tfile-2 ||
3805 error "$ST: root create permission is denied"
3806 echo "$ST: root create permission is granted - ok"
3807 cleanup || error "cleanup failed with $?"
3809 run_test 43a "check root_squash and nosquash_nids"
3811 test_43b() { # LU-5690
3812 [[ "$MGS_VERSION" -ge $(version_code 2.7.62) ]] ||
3813 skip "Need MGS version 2.7.62+"
3815 if [[ -z "$fs2mds_DEV" ]]; then
3816 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
3817 skip_env "mixed loopback and real device not working"
3820 local fs2mdsdev=$(mdsdevname 1_2)
3821 local fs2mdsvdev=$(mdsvdevname 1_2)
3823 # temporarily use fs2mds as fs2mgs
3825 local fs2mgsdev=$fs2mdsdev
3826 local fs2mgsvdev=$fs2mdsvdev
3828 local fsname=test1234
3830 load_module llite/lustre
3831 local client_ip=$(host_nids_address $HOSTNAME $NETTYPE)
3832 local host=${client_ip//*./}
3833 local net=${client_ip/%$host/}
3834 local nosquash_nids=$(h2nettype $net[$host,$host,$host])
3836 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
3837 --param mdt.root_squash=$RUNAS_ID:$RUNAS_ID \
3838 --param mdt.nosquash_nids=$nosquash_nids \
3839 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
3840 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS || error "start fs2mgs failed"
3841 stop $fs2mgs -f || error "stop fs2mgs failed"
3842 cleanup || error "cleanup failed with $?"
3844 run_test 43b "parse nosquash_nids with commas in expr_list"
3848 check_mount || error "check_mount"
3849 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
3851 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
3852 for VAL in $UUIDS; do
3853 NID=$(echo $VAL | cut -d= -f1)
3854 CLUUID=$(echo $VAL | cut -d= -f2)
3855 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
3857 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
3858 cleanup || error "cleanup failed with $?"
3860 run_test 44 "mounted client proc entry exists"
3864 check_mount || error "check_mount"
3865 stop_mds || error "Unable to stop MDS"
3869 #define OBD_FAIL_PTLRPC_LONG_REPL_UNLINK 0x50f
3870 do_facet client "$LCTL set_param fail_loc=0x8000050f"
3873 manual_umount_client --force || error "manual_umount_client failed"
3874 do_facet client "$LCTL set_param fail_loc=0x0"
3875 start_mds || error "unable to start MDS"
3876 mount_client $MOUNT || error "mount_client $MOUNT failed"
3877 cleanup || error "cleanup failed with $?"
3879 run_test 45 "long unlink handling in ptlrpcd"
3886 umount_client $MOUNT2 || rc=$?
3887 umount_client $MOUNT || rc=$?
3888 while [ $count -gt 0 ]; do
3889 stop ost${count} -f || rc=$?
3893 cleanup_nocli || rc=$?
3894 #writeconf to remove all ost2 traces for subsequent tests
3895 writeconf_or_reformat
3900 echo "Testing with $OSTCOUNT OSTs"
3902 start_mds || error "unable to start MDS"
3903 #first client should see only one ost
3904 start_ost || error "Unable to start OST1"
3905 wait_osc_import_state mds ost FULL
3907 mount_client $MOUNT || error "mount_client $MOUNT failed"
3908 trap "cleanup_46a $OSTCOUNT" EXIT ERR
3911 for (( i=2; i<=$OSTCOUNT; i++ )); do
3912 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
3913 error "start_ost$i $(ostdevname $i) failed"
3916 # wait until osts in sync
3917 for (( i=2; i<=$OSTCOUNT; i++ )); do
3918 wait_osc_import_state mds ost$i FULL
3919 wait_osc_import_ready client ost$i
3922 #second client see all ost's
3924 mount_client $MOUNT2 || error "mount_client failed"
3925 $LFS setstripe -c -1 $MOUNT2 ||
3926 error "$LFS setstripe -c -1 $MOUNT2 failed"
3927 $LFS getstripe $MOUNT2 || error "$LFS getstripe $MOUNT2 failed"
3929 echo "ok" > $MOUNT2/widestripe
3930 $LFS getstripe $MOUNT2/widestripe ||
3931 error "$LFS getstripe $MOUNT2/widestripe failed"
3932 # fill acl buffer for avoid expand lsm to them
3933 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd |
3935 setfacl -m $acl $MOUNT2/widestripe
3939 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3941 cleanup_46a $OSTCOUNT || error "cleanup_46a failed"
3943 run_test 46a "handle ost additional - wide striped file"
3948 check_mount || error "check_mount failed"
3949 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
3953 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3954 if echo $ns | grep "MDT[[:digit:]]*"; then
3957 lrs=$(echo $ns | sed 's/.*lru_size=//')
3958 lru_size[count]=$lrs
3963 facet_failover $SINGLEMDS
3964 client_up || error "client_up failed"
3967 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3968 if echo $ns | grep "MDT[[:digit:]]*"; then
3971 lrs=$(echo $ns | sed 's/.*lru_size=//')
3972 if ! test "$lrs" -eq "${lru_size[count]}"; then
3973 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
3974 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
3979 cleanup || error "cleanup failed with $?"
3981 run_test 47 "server restart does not make client loss lru_resize settings"
3986 # reformat after this test is needed - if the test fails,
3987 # we will have unkillable file at FS
3991 test_48() { # bz-17636 LU-7473
3995 check_mount || error "check_mount failed"
3997 $LFS setstripe -c -1 $MOUNT ||
3998 error "$LFS setstripe -c -1 $MOUNT failed"
3999 $LFS getstripe $MOUNT || error "$LFS getstripe $MOUNT failed"
4001 echo "ok" > $MOUNT/widestripe
4002 $LFS getstripe $MOUNT/widestripe ||
4003 error "$LFS getstripe $MOUNT/widestripe failed"
4005 # In the future, we may introduce more EAs, such as selinux, enlarged
4006 # LOV EA, and so on. These EA will use some EA space that is shared by
4007 # ACL entries. So here we only check some reasonable ACL entries count,
4008 # instead of the max number that is calculated from the max_ea_size.
4009 if [ "$MDS1_VERSION" -lt $(version_code 2.8.57) ]; then
4010 count=28 # hard coded of RPC protocol
4011 elif large_xattr_enabled; then
4012 count=4500 # max_num 8187 max_ea_size = 65452
4013 # not create too many (4500) to save test time
4015 count=450 # max_num 497 max_ea_size = 4012
4018 echo "It is expected to hold at least $count ACL entries"
4019 trap cleanup_48 EXIT ERR
4020 for ((i = 0; i < $count; i++)) do
4021 setfacl -m u:$((i + 100)):rw $MOUNT/widestripe ||
4022 error "Fail to setfacl for $MOUNT/widestripe at $i"
4025 cancel_lru_locks mdc
4026 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
4027 local r_count=$(getfacl $MOUNT/widestripe | grep "user:" | wc -l)
4028 count=$((count + 1)) # for the entry "user::rw-"
4030 [ $count -eq $r_count ] ||
4031 error "Expected ACL entries $count, but got $r_count"
4035 run_test 48 "too many acls on file"
4037 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
4038 test_49a() { # bug 17710
4039 local timeout_orig=$TIMEOUT
4040 local ldlm_timeout_orig=$LDLM_TIMEOUT
4041 local LOCAL_TIMEOUT=20
4043 LDLM_TIMEOUT=$LOCAL_TIMEOUT
4044 TIMEOUT=$LOCAL_TIMEOUT
4048 check_mount || error "client mount failed"
4050 echo "check ldlm_timout..."
4051 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
4052 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
4053 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
4055 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
4056 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
4059 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
4060 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
4063 umount_client $MOUNT || error "umount_client $MOUNT failed"
4064 stop_ost || error "problem stopping OSS"
4065 stop_mds || error "problem stopping MDS"
4067 LDLM_TIMEOUT=$ldlm_timeout_orig
4068 TIMEOUT=$timeout_orig
4070 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
4072 test_49b() { # bug 17710
4073 local timeout_orig=$TIMEOUT
4074 local ldlm_timeout_orig=$LDLM_TIMEOUT
4075 local LOCAL_TIMEOUT=20
4077 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
4078 TIMEOUT=$LOCAL_TIMEOUT
4082 check_mount || error "client mount failed"
4084 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
4085 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
4086 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
4088 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
4089 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
4092 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
4093 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
4096 cleanup || error "cleanup failed"
4098 LDLM_TIMEOUT=$ldlm_timeout_orig
4099 TIMEOUT=$timeout_orig
4101 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
4104 # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1
4106 # Test both statfs and lfs df and fail if either one fails
4107 multiop_bg_pause $1 f_
4110 killall -USR1 multiop
4111 [ $RC -ne 0 ] && log "lazystatfs multiop failed"
4112 wait $PID || { RC=$?; log "multiop return error "; }
4114 # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1
4119 if kill -s 0 $PID; then
4122 log "lazystatfs lfs df failed to complete in 5s"
4130 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4131 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4133 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
4135 cleanup || error "cleanup failed with rc $?"
4137 run_test 50a "lazystatfs all servers available"
4141 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4142 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4144 # Wait for client to detect down OST
4145 stop_ost || error "Unable to stop OST1"
4146 wait_osc_import_state client ost DISCONN
4148 log "OSCs should all be DISCONN"
4150 lazystatfs $MOUNT || error "lazystatfs should not return EIO"
4152 umount_client $MOUNT || error "Unable to unmount client"
4153 stop_mds || error "Unable to stop MDS"
4155 run_test 50b "lazystatfs all servers down"
4158 start_mds || error "Unable to start MDS"
4159 start_ost || error "Unable to start OST1"
4160 start_ost2 || error "Unable to start OST2"
4161 mount_client $MOUNT || error "Unable to mount client"
4162 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4163 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4165 # Wait for client to detect down OST
4166 stop_ost || error "Unable to stop OST1"
4167 wait_osc_import_state mds ost DISCONN
4168 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
4170 umount_client $MOUNT || error "Unable to unmount client"
4171 stop_ost2 || error "Unable to stop OST2"
4172 stop_mds || error "Unable to stop MDS"
4173 #writeconf to remove all ost2 traces for subsequent tests
4174 writeconf_or_reformat
4176 run_test 50c "lazystatfs one server down"
4179 start_mds || error "Unable to start MDS"
4180 start_ost || error "Unable to start OST1"
4181 start_ost2 || error "Unable to start OST2"
4182 mount_client $MOUNT || error "Unable to mount client"
4183 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4184 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4186 # Issue the statfs during the window where the client still
4187 # belives the OST to be available but it is in fact down.
4188 # No failure just a statfs which hangs for a timeout interval.
4189 stop_ost || error "Unable to stop OST1"
4190 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
4192 umount_client $MOUNT || error "Unable to unmount client"
4193 stop_ost2 || error "Unable to stop OST2"
4194 stop_mds || error "Unable to stop MDS"
4195 #writeconf to remove all ost2 traces for subsequent tests
4196 writeconf_or_reformat
4198 run_test 50d "lazystatfs client/server conn race"
4205 start_mds || error "Unable to start MDS"
4206 #first client should see only one ost
4207 start_ost || error "Unable to start OST1"
4208 wait_osc_import_state mds ost FULL
4210 # Wait for client to detect down OST
4211 stop_ost || error "Unable to stop OST1"
4212 wait_osc_import_state mds ost DISCONN
4214 mount_client $MOUNT || error "Unable to mount client"
4215 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
4217 multiop_bg_pause $MOUNT _f
4221 if [ $RC1 -ne 0 ]; then
4222 log "multiop failed $RC1"
4225 sleep $(( $TIMEOUT+1 ))
4227 [ $? -ne 0 ] && error "process isn't sleep"
4228 start_ost || error "Unable to start OST1"
4229 wait $pid || error "statfs failed"
4232 umount_client $MOUNT || error "Unable to unmount client"
4233 stop_ost || error "Unable to stop OST1"
4234 stop_mds || error "Unable to stop MDS"
4236 run_test 50e "normal statfs all servers down"
4241 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
4243 start_mds || error "Unable to start mds"
4244 #first client should see only one ost
4245 start_ost || error "Unable to start OST1"
4246 wait_osc_import_state mds ost FULL
4248 start_ost2 || error "Unable to start OST2"
4249 wait_osc_import_state mds ost2 FULL
4251 # Wait for client to detect down OST
4252 stop_ost2 || error "Unable to stop OST2"
4254 wait_osc_import_state mds ost2 DISCONN
4255 mount_client $MOUNT || error "Unable to mount client"
4256 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
4258 multiop_bg_pause $MOUNT _f
4262 if [ $RC1 -ne 0 ]; then
4263 log "lazystatfs multiop failed $RC1"
4266 sleep $(( $TIMEOUT+1 ))
4268 [ $? -ne 0 ] && error "process isn't sleep"
4269 start_ost2 || error "Unable to start OST2"
4270 wait $pid || error "statfs failed"
4271 stop_ost2 || error "Unable to stop OST2"
4274 umount_client $MOUNT -f || error "Unable to unmount client"
4275 stop_ost || error "Unable to stop OST1"
4276 stop_mds || error "Unable to stop MDS"
4277 #writeconf to remove all ost2 traces for subsequent tests
4278 writeconf_or_reformat
4280 run_test 50f "normal statfs one server in down"
4283 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs"
4285 start_ost2 || error "Unable to start OST2"
4286 wait_osc_import_state mds ost2 FULL
4287 wait_osc_import_ready client ost2
4289 if [[ $PERM_CMD == *"set_param -P"* ]]; then
4290 local PARAM="osc.${FSNAME}-OST0001*.active"
4292 local PARAM="${FSNAME}-OST0001.osc.active"
4295 $LFS setstripe -c -1 $DIR/$tfile || error "$LFS setstripe failed"
4296 do_facet mgs $PERM_CMD $PARAM=0 || error "Unable to deactivate OST"
4298 umount_client $MOUNT || error "Unable to unmount client"
4299 mount_client $MOUNT || error "Unable to mount client"
4300 # This df should not cause a panic
4303 do_facet mgs $PERM_CMD $PARAM=1 || error "Unable to activate OST"
4304 rm -f $DIR/$tfile || error "unable to remove file $DIR/$tfile"
4305 umount_client $MOUNT || error "Unable to unmount client"
4306 stop_ost2 || error "Unable to stop OST2"
4307 stop_ost || error "Unable to stop OST1"
4308 stop_mds || error "Unable to stop MDS"
4309 #writeconf to remove all ost2 traces for subsequent tests
4310 writeconf_or_reformat
4312 run_test 50g "deactivated OST should not cause panic"
4316 # prepare MDT/OST, make OSC inactive for OST1
4317 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs"
4319 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
4320 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
4321 error "tunefs OST1 failed"
4322 start_mds || error "Unable to start MDT"
4323 start_ost || error "Unable to start OST1"
4324 start_ost2 || error "Unable to start OST2"
4325 mount_client $MOUNT || error "client start failed"
4327 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4329 # activatate OSC for OST1
4330 set_persistent_param_and_check client \
4331 "osc.${FSNAME}-OST0000-osc-[!M]*.active" \
4332 "${FSNAME}-OST0000.osc.active" 1
4334 mkdir $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
4335 $LFS setstripe -c -1 -i 0 $DIR/$tdir/2 ||
4336 error "$LFS setstripe $DIR/$tdir/2 failed"
4337 sleep 1 && echo "create a file after OST1 is activated"
4338 # doing some io, shouldn't crash
4339 dd if=/dev/zero of=$DIR/$tdir/2/$tfile-io bs=1M count=10
4341 # check OSC import is working
4342 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
4343 error "some OSC imports are still not connected"
4347 umount_client $MOUNT || error "Unable to umount client"
4348 stop_ost2 || error "Unable to stop OST2"
4349 cleanup_nocli || error "cleanup_nocli failed with $?"
4351 run_test 50h "LU-642: activate deactivated OST"
4354 # prepare MDT/OST, make OSC inactive for OST1
4355 [ "$MDSCOUNT" -lt "2" ] && skip "needs >= 2 MDTs"
4358 [ $(facet_fstype mds2) == zfs ] && import_zpool mds2
4359 do_facet mds2 "$TUNEFS --param mdc.active=0 $(mdsdevname 2)" ||
4360 error "tunefs MDT2 failed"
4361 start_mds || error "Unable to start MDT"
4362 start_ost || error "Unable to start OST1"
4363 start_ost2 || error "Unable to start OST2"
4364 mount_client $MOUNT || error "client start failed"
4366 mkdir_on_mdt0 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4368 if [[ $PERM_CMD == *"set_param -P"* ]]; then
4369 $PERM_CMD mdc.${FSNAME}-MDT0001-mdc-*.active=0 &&
4370 error "deactive MDC0 succeeds"
4372 $PERM_CMD ${FSNAME}-MDT0000.mdc.active=0 &&
4373 error "deactive MDC0 succeeds"
4376 # activate MDC for MDT2
4377 set_persistent_param_and_check client \
4378 "mdc.${FSNAME}-MDT0001-mdc-*.active" \
4379 "${FSNAME}-MDT0001.mdc.active" 1
4381 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
4382 if [ "$MDS1_VERSION" -ge $(version_code 2.7.60) ]
4384 wait_dne_interconnect
4386 $LFS mkdir -i1 $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
4388 createmany -o $DIR/$tdir/2/$tfile-%d 1 || error "create files failed"
4390 rm -rf $DIR/$tdir/2 || error "unlink dir failed"
4392 # deactivate MDC for MDT2
4393 set_persistent_param_and_check client \
4394 "mdc.${FSNAME}-MDT0001-mdc-*.active" \
4395 "${FSNAME}-MDT0001.mdc.active" 0
4397 wait_osp_active mds ${FSNAME}-MDT0001 1 0
4399 $LFS mkdir -i1 $DIR/$tdir/2 &&
4400 error "mkdir $DIR/$tdir/2 succeeds after deactive MDT"
4402 $LFS mkdir -i0 -c$MDSCOUNT $DIR/$tdir/striped_dir ||
4403 error "mkdir $DIR/$tdir/striped_dir fails after deactive MDT2"
4405 local stripe_count=$($LFS getdirstripe -c $DIR/$tdir/striped_dir)
4406 [ $stripe_count -eq $((MDSCOUNT - 1)) ] ||
4407 error "wrong $stripe_count != $((MDSCOUNT -1)) for striped_dir"
4410 umount_client $MOUNT || error "Unable to umount client"
4415 run_test 50i "activate deactivated MDT"
4418 local LOCAL_TIMEOUT=20
4422 check_mount || error "check_mount failed"
4424 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
4425 $LFS setstripe -c -1 $MOUNT/$tdir ||
4426 error "$LFS setstripe -c -1 $MOUNT/$tdir failed"
4427 #define OBD_FAIL_MDS_REINT_DELAY 0x142
4428 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x142"
4429 touch $MOUNT/$tdir/$tfile &
4432 start_ost2 || error "Unable to start OST1"
4434 stop_ost2 || error "Unable to stop OST1"
4435 umount_client $MOUNT -f || error “unmount $MOUNT failed”
4436 cleanup_nocli || error “stop server failed”
4437 #writeconf to remove all ost2 traces for subsequent tests
4438 writeconf_or_reformat
4440 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
4449 do_node $node mkdir -p $dest
4450 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
4452 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
4453 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
4454 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
4456 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
4457 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
4467 local backup2=${TMP}/backup2
4469 do_node $node mkdir -p $backup2
4470 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
4472 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
4473 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
4474 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
4476 do_node $node "diff -rq $backup $backup2"
4477 [ $? -eq 0 ] || { error "contents differ"; return 3; }
4479 local xattrs2=${TMP}/xattrs2
4480 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
4481 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
4483 do_node $node "diff $xattrs $xattrs2"
4484 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
4486 do_node $node "rm -rf $backup2 $xattrs2"
4487 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
4491 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4492 skip "ldiskfs only test"
4495 start_mds || error "Unable to start MDS"
4496 start_ost || error "Unable to start OST1"
4497 mount_client $MOUNT || error "Unable to mount client"
4500 local ost1mnt=$(facet_mntpt ost1)
4501 local ost1node=$(facet_active_host ost1)
4502 local ost1tmp=$TMP/conf52
4505 mkdir $DIR/$tdir || error "Unable to create $DIR/$tdir"
4506 touch $TMP/modified_first || error "Unable to create temporary file"
4507 local mtime=$(stat -c %Y $TMP/modified_first)
4508 do_node $ost1node "mkdir -p $ost1tmp &&
4509 touch -m -d @$mtime $ost1tmp/modified_first" ||
4510 error "Unable to create temporary file"
4513 $LFS setstripe -c -1 -S 1M $DIR/$tdir || error "$LFS setstripe failed"
4515 for (( i=0; i < nrfiles; i++ )); do
4516 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c ||
4517 error "multiop failed"
4522 # sync all the data and make sure no pending data on the client,
4523 # thus the SOM xattr would not be changed any more.
4524 cancel_lru_locks osc
4527 echo backup files to $TMP/$tdir
4528 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
4529 copy_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
4530 error "Unable to copy files"
4532 umount_client $MOUNT || error "Unable to umount client"
4533 stop_ost || error "Unable to stop ost1"
4535 echo mount ost1 as ldiskfs
4536 do_node $ost1node mkdir -p $ost1mnt || error "Unable to create $ost1mnt"
4537 if ! do_node $ost1node test -b $ost1_dev; then
4540 do_node $ost1node mount -t "$ost1_FSTYPE" $loop $ost1_dev \
4542 error "Unable to mount ost1 as ldiskfs"
4545 echo backup objects to $ost1tmp/objects
4546 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
4547 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
4548 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
4550 error "Unable to copy objects"
4552 # move objects to lost+found
4553 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
4554 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
4556 do_node $ost1node "umount $ost1mnt" ||
4557 error "Unable to umount ost1 as ldiskfs"
4559 start_ost || error "Unable to start OST1"
4560 mount_client $MOUNT || error "Unable to mount client"
4562 local REPAIRED=$(do_node $ost1node "$LCTL get_param \
4563 -n osd-ldiskfs.$FSNAME-OST0000.oi_scrub" |
4564 awk '/^lf_repa[ri]*ed/ { print $2 }')
4565 [ $REPAIRED -gt 0 ] ||
4566 error "Some entry under /lost+found should be repaired"
4569 diff_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
4570 error "Unable to diff files"
4572 rm -rf $TMP/$tdir $TMP/file_xattrs ||
4573 error "Unable to delete temporary files"
4574 do_node $ost1node "rm -rf $ost1tmp" ||
4575 error "Unable to delete temporary files"
4576 cleanup || error "cleanup failed with $?"
4578 run_test 52 "check recovering objects from lost+found"
4580 # Checks threads_min/max/started for some service
4582 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
4583 # parameter pattern prefix like 'ost.*.ost'.
4596 local msg="Insane $modname thread counts"
4597 local ncpts=$(check_cpt_number $facet)
4601 check_mount || return 41
4603 # We need to expand $parampat, but it may match multiple parameters, so
4604 # we'll pick the first one
4605 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
4606 error "Couldn't expand ${parampat}.threads_min parameter name"
4610 # Remove the .threads_min part
4611 paramp=${paramp%.threads_min}
4613 # Check for sanity in defaults
4614 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4616 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4618 tstarted=$(do_facet $facet "$LCTL get_param \
4619 -n ${paramp}.threads_started" || echo 0)
4620 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' ||
4622 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4624 nthrs=$(expr $tmax - $tmin)
4625 if [ $nthrs -lt $ncpts ]; then
4631 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
4632 skip_env "module parameter forced $facet thread count"
4634 # Check that we can change min/max
4635 do_facet $facet "$LCTL set_param \
4636 ${paramp}.threads_min=$((tmin + nthrs))"
4637 do_facet $facet "$LCTL set_param \
4638 ${paramp}.threads_max=$((tmax - nthrs))"
4639 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4641 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4643 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) &&
4644 $tmax2 == ($tmax - $nthrs)))' || return $?
4646 # Check that we can set min/max to the same value
4647 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4649 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$tmin"
4650 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4652 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4654 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
4656 # Check that we can't set max < min
4657 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$((tmin - 1))"
4658 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4660 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4662 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
4664 # We need to ensure that we get the module options desired; to do this
4665 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
4666 LOAD_MODULES_REMOTE=true
4669 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
4670 setmodopts -a $modname "$newvalue" oldvalue
4673 check_mount || return 41
4675 # Restore previous setting of MODOPTS_*
4676 setmodopts $modname "$oldvalue"
4678 # Check that $opts took
4679 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4681 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4683 tstarted=$(do_facet $facet \
4684 "$LCTL get_param -n ${paramp}.threads_started" || echo 0)
4685 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4694 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
4695 cleanup || error "cleanup failed with rc $?"
4697 run_test 53a "check OSS thread count params"
4701 local mds=$(do_facet $SINGLEMDS "$LCTL get_param \
4702 -N mds.*.*.threads_max 2>/dev/null")
4703 if [ -z "$mds" ]; then
4704 #running this on an old MDT
4705 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
4707 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
4709 cleanup || error "cleanup failed with $?"
4711 run_test 53b "check MDS thread count params"
4714 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4715 skip "ldiskfs only test"
4718 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p ||
4719 error "llverdev failed with rc=$?"
4722 run_test 54a "test llverdev and partial verify of device"
4725 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4726 skip "ldiskfs only test"
4730 run_llverfs $MOUNT -p || error "llverfs failed with rc=$?"
4731 cleanup || error "cleanup failed with rc=$?"
4733 run_test 54b "test llverfs and partial verify of filesystem"
4737 local max_ost_index=$1
4738 echo -n $(((max_ost_index + 1) * 8))
4742 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4743 skip "ldiskfs only test"
4746 local mdsdev=$(mdsdevname 1)
4747 local mdsvdev=$(mdsvdevname 1)
4751 if ! combined_mgs_mds; then
4752 stop_mgs || error "stopping MGS service failed"
4753 format_mgs || error "formatting MGT failed"
4755 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
4757 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
4758 --reformat $(ostdevname 1) $(ostvdevname 1)
4764 echo checking size of lov_objid for ost index $i
4765 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" |
4766 grep ^User | awk -F 'Size: ' '{print $2}')
4767 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
4768 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
4770 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
4777 run_test 55 "check lov_objid size"
4780 local mds_journal_size_orig=$MDSJOURNALSIZE
4786 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
4787 $(ostdevname 1) $(ostvdevname 1)
4788 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
4789 $(ostdevname 2) $(ostvdevname 2)
4792 start_ost || error "Unable to start first ost (idx 10000)"
4793 start_ost2 || error "Unable to start second ost (idx 1000)"
4794 mount_client $MOUNT || error "Unable to mount client"
4798 # test instantiating PFL components with sparse index LU-15513
4799 mkdir -p $MOUNT/$tdir
4800 $LFS setstripe -E 4M -c 1 -E 1G -c 4 -S4M -E eof -c -1 $MOUNT/$tdir
4801 dd if=/dev/zero of=$MOUNT/$tdir/$tfile bs=4K count=1 seek=10k ||
4802 error "dd to second component failed"
4804 if [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
4805 [[ "$MDS1_VERSION" -ge $(version_code 2.5.4) &&
4806 "$MDS1_VERSION" -lt $(version_code 2.5.11) ]]; then
4807 wait_osc_import_state mds ost1 FULL
4808 wait_osc_import_state mds ost2 FULL
4809 $LFS setstripe --stripe-count=-1 $DIR/$tfile ||
4810 error "Unable to setstripe $DIR/$tfile"
4811 n=$($LFS getstripe --stripe-count $DIR/$tfile)
4812 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
4817 MDSJOURNALSIZE=$mds_journal_size_orig
4820 run_test 56a "check big OST indexes and out-of-index-order start"
4825 umount_client $MOUNT -f || error "unmount client failed"
4834 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs"
4836 trap cleanup_56b EXIT RETURN ERR
4839 if ! combined_mgs_mds ; then
4844 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) --index=0 --reformat \
4845 $(mdsdevname 1) $(mdsvdevname 1)
4846 add mds2 $(mkfs_opts mds2 $(mdsdevname 2)) --index=1 --reformat \
4847 $(mdsdevname 2) $(mdsvdevname 2)
4848 add mds3 $(mkfs_opts mds3 $(mdsdevname 3)) --index=1000 --reformat \
4849 $(mdsdevname 3) $(mdsvdevname 3)
4853 start_mdt 1 || error "MDT 1 (idx 0) start failed"
4854 start_mdt 2 || error "MDT 2 (idx 1) start failed"
4855 start_mdt 3 || error "MDT 3 (idx 1000) start failed"
4856 start_ost || error "Unable to start first ost"
4857 start_ost2 || error "Unable to start second ost"
4859 do_nodes $(comma_list $(mdts_nodes)) \
4860 "$LCTL set_param mdt.*.enable_remote_dir=1 \
4861 mdt.*.enable_remote_dir_gid=-1"
4863 mount_client $MOUNT || error "Unable to mount client"
4865 $LFS mkdir -c3 $MOUNT/$tdir || error "failed to make testdir"
4867 echo "This is test file 1!" > $MOUNT/$tdir/$tfile.1 ||
4868 error "failed to make test file 1"
4869 echo "This is test file 2!" > $MOUNT/$tdir/$tfile.2 ||
4870 error "failed to make test file 2"
4871 echo "This is test file 1000!" > $MOUNT/$tdir/$tfile.1000 ||
4872 error "failed to make test file 1000"
4874 rm -rf $MOUNT/$tdir || error "failed to remove testdir"
4876 $LFS mkdir -i1000 $MOUNT/$tdir.1000 ||
4877 error "create remote dir at idx 1000 failed"
4880 echo "=== START lfs df OUTPUT ==="
4882 echo "==== END lfs df OUTPUT ===="
4884 mdtcnt=$(echo -e "$output" | grep $FSNAME-MDT | wc -l)
4885 ostcnt=$(echo -e "$output" | grep $FSNAME-OST | wc -l)
4887 echo "lfs df returned mdt count $mdtcnt and ost count $ostcnt"
4888 [ $mdtcnt -eq 3 ] || error "lfs df returned wrong mdt count"
4889 [ $ostcnt -eq 2 ] || error "lfs df returned wrong ost count"
4891 echo "This is test file 1!" > $MOUNT/$tdir.1000/$tfile.1 ||
4892 error "failed to make test file 1"
4893 echo "This is test file 2!" > $MOUNT/$tdir.1000/$tfile.2 ||
4894 error "failed to make test file 2"
4895 echo "This is test file 1000!" > $MOUNT/$tdir.1000/$tfile.1000 ||
4896 error "failed to make test file 1000"
4897 rm -rf $MOUNT/$tdir.1000 || error "failed to remove remote_dir"
4900 echo "=== START lfs mdts OUTPUT ==="
4902 echo "==== END lfs mdts OUTPUT ===="
4904 echo -e "$output" | grep -v "MDTS:" | awk '{print $1}' |
4905 sed 's/://g' > $TMP/mdts-actual.txt
4906 sort $TMP/mdts-actual.txt -o $TMP/mdts-actual.txt
4908 echo -e "0\n1\n1000" > $TMP/mdts-expected.txt
4910 diff $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4913 rm $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4915 [ $result -eq 0 ] || error "target_obd proc file is incorrect!"
4917 run_test 56b "test target_obd correctness with nonconsecutive MDTs"
4919 test_57a() { # bug 22656
4920 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4921 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4922 tail -1 | awk '{print $1}')
4923 writeconf_or_reformat
4924 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
4925 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" ||
4926 error "tunefs failed"
4928 start_ost && error "OST registration from failnode should fail"
4931 run_test 57a "initial registration from failnode should fail (should return errs)"
4934 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4935 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4936 tail -1 | awk '{print $1}')
4937 writeconf_or_reformat
4938 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
4939 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" ||
4940 error "tunefs failed"
4942 start_ost || error "OST registration from servicenode should not fail"
4945 run_test 57b "initial registration from servicenode should not fail"
4948 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
4951 test_58() { # bug 22658
4952 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
4954 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4955 createmany -o $DIR/$tdir/$tfile-%d 100
4956 unlinkmany $DIR/$tdir/$tfile-%d 100
4957 stop_mds || error "Unable to stop MDS"
4959 local MNTDIR=$(facet_mntpt $SINGLEMDS)
4960 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4962 # remove all files from the OBJECTS dir
4963 mount_fstype $SINGLEMDS
4965 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
4967 unmount_fstype $SINGLEMDS
4968 # restart MDS with missing llog files
4969 start_mds || error "unable to start MDS"
4970 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0"
4973 run_test 58 "missing llog files must not prevent MDT from mounting"
4976 start_mgsmds >> /dev/null
4977 local C1=$(count_osts)
4978 if [ $C1 -eq 0 ]; then
4979 start_ost >> /dev/null
4983 echo "original ost count: $C1 (expect > 0)"
4984 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
4985 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
4986 local C2=$(count_osts)
4987 echo "after mdt writeconf count: $C2 (expect 0)"
4988 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
4989 echo "OST start without writeconf should fail:"
4990 start_ost >> /dev/null &&
4991 error "OST start without writeconf didn't fail"
4992 echo "OST start with writeconf should succeed:"
4993 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
4994 local C3=$(count_osts)
4995 echo "after ost writeconf count: $C3 (expect 1)"
4996 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
4997 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
4998 local C4=$(count_osts)
4999 echo "after ost2 writeconf count: $C4 (expect 2)"
5000 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
5001 stop_ost2 >> /dev/null
5002 cleanup_nocli >> /dev/null
5003 #writeconf to remove all ost2 traces for subsequent tests
5004 writeconf_or_reformat
5006 run_test 59 "writeconf mount option"
5008 test_60() { # LU-471
5009 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5010 skip "ldiskfs only test"
5015 for num in $(seq $MDSCOUNT); do
5016 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
5017 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
5018 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
5022 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
5023 [ ${PIPESTATUS[0]} -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
5025 # MDT default has dirdata feature
5026 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
5027 # we disable uninit_bg feature
5028 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
5029 # we set stride extended options
5030 echo $dump | grep stride > /dev/null || error "stride is not set"
5034 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
5037 local lxattr=$(large_xattr_enabled)
5039 [ "$MDS1_VERSION" -ge $(version_code 2.1.53) ] ||
5040 skip "Need MDS version at least 2.1.53"
5042 if [ "$mds1_FSTYPE" == ldiskfs ] &&
5043 ! large_xattr_enabled; then
5046 for num in $(seq $MDSCOUNT); do
5047 do_facet mds${num} $TUNE2FS -O ea_inode \
5048 $(mdsdevname $num) ||
5049 error "tune2fs on mds $num failed"
5053 setup || error "setting up the filesystem failed"
5054 client_up || error "starting client failed"
5056 local file=$DIR/$tfile
5057 touch $file || error "touch $file failed"
5059 local large_value="$(generate_string $(max_xattr_size))"
5060 local small_value="bar"
5062 local name="trusted.big"
5063 log "save large xattr of $(max_xattr_size) bytes on $name on $file"
5064 setfattr -n $name -v $large_value $file ||
5065 error "saving $name on $file failed"
5067 local new_value=$(get_xattr_value $name $file)
5068 [[ "$new_value" != "$large_value" ]] &&
5069 error "$name different after saving"
5071 log "shrink value of $name on $file"
5072 setfattr -n $name -v $small_value $file ||
5073 error "shrinking value of $name on $file failed"
5075 new_value=$(get_xattr_value $name $file)
5076 [[ "$new_value" != "$small_value" ]] &&
5077 error "$name different after shrinking"
5079 log "grow value of $name on $file"
5080 setfattr -n $name -v $large_value $file ||
5081 error "growing value of $name on $file failed"
5083 new_value=$(get_xattr_value $name $file)
5084 [[ "$new_value" != "$large_value" ]] &&
5085 error "$name different after growing"
5087 log "check value of $name on $file after remounting MDS"
5089 new_value=$(get_xattr_value $name $file)
5090 [[ "$new_value" != "$large_value" ]] &&
5091 error "$name different after remounting MDS"
5093 log "remove large xattr $name from $file"
5094 setfattr -x $name $file || error "removing $name from $file failed"
5096 if $lxattr && [ "$mds1_FSTYPE" == ldiskfs ]; then
5097 stopall || error "stopping for e2fsck run"
5098 for num in $(seq $MDSCOUNT); do
5099 run_e2fsck $(facet_active_host mds$num) \
5100 $(mdsdevname $num) "-y" ||
5101 error "e2fsck MDT$num failed"
5103 setup_noconfig || error "remounting the filesystem failed"
5106 # need to delete this file to avoid problems in other tests
5108 cleanup || error "stopping systems failed"
5110 run_test 61 "large xattr"
5113 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5114 skip "ldiskfs only test"
5116 [[ "$MDS1_VERSION" -ge $(version_code 2.2.51) ]] ||
5117 skip "Need MDS version at least 2.2.51"
5120 local mdsdev=$(mdsdevname 1)
5121 local ostdev=$(ostdevname 1)
5123 echo "disable journal for mds"
5124 do_facet mds1 $TUNE2FS -O ^has_journal $mdsdev || error "tune2fs failed"
5125 start_mds && error "MDT start should fail"
5126 echo "disable journal for ost"
5127 do_facet ost1 $TUNE2FS -O ^has_journal $ostdev || error "tune2fs failed"
5128 start_ost && error "OST start should fail"
5129 cleanup || error "cleanup failed with rc $?"
5132 run_test 62 "start with disabled journal"
5135 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5136 skip "ldiskfs only test"
5139 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ldiskfs
5140 local inode_slab=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
5141 awk '/ldiskfs_inode_cache/ { print $5 / $6 }')
5142 if [ -z "$inode_slab" ]; then
5143 skip "ldiskfs module has not been loaded"
5146 if grep -q "CONFIG_DEBUG_LOCK_ALLOC=y" /boot/config-$(uname -r); then
5147 skip "test is not compatible with CONFIG_DEBUG_LOCK_ALLOC=y"
5150 echo "$inode_slab ldiskfs inodes per page"
5151 [ "${inode_slab%.*}" -ge "3" ] && return 0
5153 # If kmalloc-128 is also 1 per page - this is a debug kernel
5154 # and so this is not an error.
5155 local kmalloc128=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
5156 awk '/^(dma-kmalloc|size)-128 / { print $5 / $6 }')
5157 # 32 128-byte chunks in 4k
5158 [ "${kmalloc128%.*}" -lt "32" ] ||
5159 error "ldiskfs inode too big, only $inode_slab objs/page, " \
5160 "kmalloc128 = $kmalloc128 objs/page"
5162 run_test 63 "Verify each page can at least hold 3 ldiskfs inodes"
5165 start_mds || error "unable to start MDS"
5166 start_ost || error "Unable to start OST1"
5167 start_ost2 || error "Unable to start second ost"
5168 mount_client $MOUNT || error "Unable to mount client"
5169 stop_ost2 || error "Unable to stop second ost"
5172 umount_client $MOUNT -f || error “unmount $MOUNT failed”
5173 cleanup_nocli || error "cleanup_nocli failed with $?"
5174 #writeconf to remove all ost2 traces for subsequent tests
5175 writeconf_or_reformat
5177 run_test 64 "check lfs df --lazy "
5179 test_65() { # LU-2237
5180 # Currently, the test is only valid for ldiskfs backend
5181 [ "$mds1_FSTYPE" != ldiskfs ] &&
5182 skip "ldiskfs only test"
5184 local devname=$(mdsdevname ${SINGLEMDS//mds/})
5185 local brpt=$(facet_mntpt brpt)
5188 if ! do_facet $SINGLEMDS "test -b $devname"; then
5192 stop_mds || error "Unable to stop MDS"
5193 local obj=$(do_facet $SINGLEMDS \
5194 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
5196 if [ -z "$obj" ]; then
5197 # The MDT may be just re-formatted, mount the MDT for the
5198 # first time to guarantee the "last_rcvd" file is there.
5199 start_mds || error "fail to mount the MDS for the first time"
5200 stop_mds || error "Unable to stop MDS"
5203 # remove the "last_rcvd" file
5204 do_facet $SINGLEMDS "mkdir -p $brpt"
5205 do_facet $SINGLEMDS \
5206 "mount -t $mds1_FSTYPE $opts $devname $brpt"
5207 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
5208 do_facet $SINGLEMDS "$UMOUNT $brpt"
5210 # restart MDS, the "last_rcvd" file should be recreated.
5211 start_mds || error "fail to restart the MDS"
5212 stop_mds || error "Unable to stop MDS"
5213 obj=$(do_facet $SINGLEMDS \
5214 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
5215 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
5217 run_test 65 "re-create the lost last_rcvd file when server mount"
5220 [[ "$MGS_VERSION" -ge $(version_code 2.3.59) ]] ||
5221 skip "Need MGS version at least 2.3.59"
5224 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
5225 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
5227 # add EXCLUDE records to config log, they are not to be
5228 # removed by lctl replace_nids
5229 set_conf_param_and_check mds \
5230 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
5231 "$FSNAME-OST0000.osc.active" \
5234 echo "replace_nids should fail if MDS, OSTs and clients are UP"
5235 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5236 error "replace_nids fail"
5238 umount_client $MOUNT || error "unmounting client failed"
5239 echo "replace_nids should fail if MDS and OSTs are UP"
5240 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5241 error "replace_nids fail"
5243 stop_ost || error "Unable to stop OST1"
5244 echo "replace_nids should fail if MDS is UP"
5245 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5246 error "replace_nids fail"
5248 stop_mds || error "stopping mds failed"
5250 if combined_mgs_mds; then
5251 start_mdt 1 "-o nosvc" ||
5252 error "starting mds with nosvc option failed"
5255 echo "command should accept two parameters"
5256 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
5257 error "command should accept two params"
5259 echo "correct device name should be passed"
5260 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
5261 error "wrong devname"
5263 echo "wrong nids list should not destroy the system"
5264 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
5266 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "asdfasdf, asdfadf" &&
5269 echo "replace OST nid"
5270 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
5271 error "replace nids failed"
5273 echo "command should accept two parameters"
5274 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
5275 error "command should accept two params"
5277 echo "wrong nids list should not destroy the system"
5278 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
5281 local FAKE_NIDS="192.168.0.112@tcp1,192.168.0.112@tcp2"
5282 local FAKE_FAILOVER="192.168.0.113@tcp1,192.168.0.113@tcp2"
5283 local NIDS_AND_FAILOVER="$MDS_NID,$FAKE_NIDS:$FAKE_FAILOVER"
5284 echo "set NIDs with failover"
5285 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $NIDS_AND_FAILOVER ||
5286 error "replace nids failed"
5289 echo "replace MDS nid"
5290 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
5291 error "replace nids failed"
5293 if ! combined_mgs_mds ; then
5296 stop_mds || error "Unable to stop MDS"
5299 start_mgsmds || error "start mgsmds failed"
5300 set_conf_param_and_check mds \
5301 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
5302 "$FSNAME-OST0000.osc.active" \
5304 start_ost || error "unable to start OST"
5305 mount_client $MOUNT || error "mount client failed"
5307 check_mount || error "error after nid replace"
5308 cleanup || error "cleanup failed"
5311 run_test 66 "replace nids"
5313 test_67() { #LU-2950
5314 local legacy="$TMP/legacy_lnet_config"
5315 local new="$TMP/new_routes_test"
5316 local out="$TMP/config_out_file"
5317 local verify="$TMP/conv_verify"
5318 local verify_conf="$TMP/conf_verify"
5320 # Create the legacy file that will be run through the
5321 # lustre_routes_conversion script
5322 cat <<- LEGACY_LNET_CONFIG > $legacy
5323 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
5324 tcp2 54 10.1.3.2@tcp;
5325 tcp3 10.3.4.3@tcp:3;
5329 # Create the verification file to verify the output of
5330 # lustre_routes_conversion script against.
5331 cat <<- VERIFY_LNET_CONFIG > $verify
5332 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
5333 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
5334 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
5335 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
5336 tcp4: { gateway: 10.3.3.4@tcp }
5339 # Create the verification file to verify the output of
5340 # lustre_routes_config script against
5341 cat <<- VERIFY_LNET_CONFIG > $verify_conf
5342 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
5343 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
5344 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
5345 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
5346 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
5349 $LUSTRE_ROUTES_CONVERSION $legacy $new > /dev/null
5350 if [ -f $new ]; then
5351 # verify the conversion output
5352 cmp -s $new $verify > /dev/null
5353 if [ $? -eq 1 ]; then
5354 error "routes conversion failed"
5357 lustre_routes_config --dry-run --verbose $new > $out
5358 # check that the script succeeded
5359 cmp -s $out $verify_conf > /dev/null
5360 if [ $? -eq 1 ]; then
5361 error "routes config failed"
5364 error "routes conversion test failed"
5366 # remove generated files
5367 rm -f $new $legacy $verify $verify_conf $out
5369 run_test 67 "test routes conversion and configuration"
5377 [ "$MDS1_VERSION" -ge $(version_code 2.4.53) ] ||
5378 skip "Need MDS version at least 2.4.53"
5380 umount_client $MOUNT || error "umount client failed"
5385 # START-END - the sequences we'll be reserving
5386 START=$(do_facet $SINGLEMDS \
5387 $LCTL get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
5388 END=$((START + (1 << 30)))
5389 do_facet $SINGLEMDS \
5390 $LCTL set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
5392 # reset the sequences MDT0000 has already assigned
5393 do_facet $SINGLEMDS \
5394 $LCTL set_param seq.srv*MDT0000.space=clear
5396 # remount to let the client allocate new sequence
5397 mount_client $MOUNT || error "mount client failed"
5399 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
5400 do_facet $SINGLEMDS \
5401 $LCTL get_param seq.srv*MDT0000.space
5402 $LFS path2fid $DIR/$tfile
5404 local old_ifs="$IFS"
5406 fid=($($LFS path2fid $DIR/$tfile))
5410 if [[ $seq < $END ]]; then
5411 error "used reserved sequence $seq?"
5413 cleanup || error "cleanup failed with $?"
5415 run_test 68 "be able to reserve specific sequences in FLDB"
5417 # Test 69: is about the total number of objects ever created on an OST.
5418 # so that when it is reformatted the normal MDS->OST orphan recovery won't
5419 # just "precreate" the missing objects. In the past it might try to recreate
5420 # millions of objects after an OST was reformatted
5422 [[ "$MDS1_VERSION" -lt $(version_code 2.4.2) ]] &&
5423 skip "Need MDS version at least 2.4.2"
5425 [[ "$MDS1_VERSION" -ge $(version_code 2.4.50) ]] &&
5426 [[ "$MDS1_VERSION" -lt $(version_code 2.5.0) ]] &&
5427 skip "Need MDS version at least 2.5.0"
5430 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5432 # use OST0000 since it probably has the most creations
5433 local OSTNAME=$(ostname_from_index 0)
5434 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
5435 local last_id=$(do_facet mds1 $LCTL get_param -n \
5436 osp.$mdtosc_proc1.prealloc_last_id)
5438 # Want to have OST LAST_ID over 5 * OST_MAX_PRECREATE to
5439 # verify that the LAST_ID recovery is working properly. If
5440 # not, then the OST will refuse to allow the MDS connect
5441 # because the LAST_ID value is too different from the MDS
5442 #define OST_MAX_PRECREATE=20000
5443 local ost_max_pre=20000
5444 local num_create=$(( ost_max_pre * 5 + 1 - last_id))
5446 # If the LAST_ID is already over 5 * OST_MAX_PRECREATE, we don't
5447 # need to create any files. So, skip this section.
5448 if [ $num_create -gt 0 ]; then
5449 # Check the number of inodes available on OST0
5451 local ifree=$($LFS df -i $MOUNT |
5452 awk '/OST0000/ { print $4 }'; exit ${PIPESTATUS[0]})
5453 log "On OST0, $ifree inodes available. Want $num_create. rc=$?"
5455 $LFS setstripe -i 0 $DIR/$tdir ||
5456 error "$LFS setstripe -i 0 $DIR/$tdir failed"
5457 if [ $ifree -lt 10000 ]; then
5458 files=$(( ifree - 50 ))
5463 local j=$((num_create / files + 1))
5464 for i in $(seq 1 $j); do
5465 createmany -o $DIR/$tdir/$tfile-$i- $files ||
5466 error "createmany fail create $files files: $?"
5467 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
5468 error "unlinkmany failed unlink $files files"
5472 # delete all of the files with objects on OST0 so the
5473 # filesystem is not inconsistent later on
5474 $LFS find $MOUNT --ost 0 -print0 | xargs -0 rm
5476 umount_client $MOUNT || error "umount client failed"
5477 stop_ost || error "OST0 stop failure"
5478 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
5479 $(ostdevname 1) $(ostvdevname 1) ||
5480 error "reformat and replace $ostdev failed"
5481 start_ost || error "OST0 restart failure"
5482 wait_osc_import_state mds ost FULL
5484 mount_client $MOUNT || error "mount client failed"
5485 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
5486 local idx=$($LFS getstripe -i $DIR/$tdir/$tfile-last)
5487 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
5489 local iused=$($LFS df -i $MOUNT |
5490 awk '/OST0000/ { print $3 }'; exit ${PIPESTATUS[0]})
5491 log "On OST0, $iused used inodes rc=$?"
5492 [ $iused -ge $((ost_max_pre/2 + 1000)) ] &&
5493 error "OST replacement created too many inodes; $iused"
5494 cleanup || error "cleanup failed with $?"
5496 run_test 69 "replace an OST with the same index"
5499 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5502 cleanup || error "cleanup failed with $?"
5504 start_mdt 1 || error "MDT0 start fail"
5506 start_ost || error "OST0 start fail"
5507 for num in $(seq 2 $MDSCOUNT); do
5508 start_mdt $num || return
5511 mount_client $MOUNT || error "mount client fails"
5513 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
5515 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5516 error "create remote dir fail"
5518 rm -rf $DIR/$tdir || error "delete dir fail"
5519 cleanup || error "cleanup failed with $?"
5521 run_test 70a "start MDT0, then OST, then MDT1"
5524 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5527 start_ost || error "OST0 start fail"
5529 start_mds || error "MDS start fail"
5531 mount_client $MOUNT || error "mount client fails"
5533 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
5535 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5536 error "create remote dir fail"
5538 rm -rf $DIR/$tdir || error "delete dir fail"
5540 cleanup || error "cleanup failed with $?"
5542 run_test 70b "start OST, MDT1, MDT0"
5545 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5548 start_mds || error "MDS start fail"
5549 start_ost || error "OST0 start fail"
5551 mount_client $MOUNT || error "mount client fails"
5552 stop_mdt 1 || error "MDT1 start fail"
5554 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
5555 echo "deactivate $mdc_for_mdt1"
5556 $LCTL --device $mdc_for_mdt1 deactivate ||
5557 error "set $mdc_for_mdt1 deactivate failed"
5559 mkdir $DIR/$tdir && error "mkdir succeed"
5561 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
5562 error "create remote dir succeed"
5564 cleanup || error "cleanup failed with $?"
5566 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
5569 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5572 start_mds || error "MDS start fail"
5573 start_ost || error "OST0 start fail"
5575 mount_client $MOUNT || error "mount client fails"
5577 stop_mdt 2 || error "MDT1 start fail"
5579 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
5581 echo "deactivate $mdc_for_mdt2"
5582 $LCTL --device $mdc_for_mdt2 deactivate ||
5583 error "set $mdc_for_mdt2 deactivate failed"
5585 mkdir $DIR/$tdir || error "mkdir fail"
5586 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
5587 error "create remote dir succeed"
5589 rm -rf $DIR/$tdir || error "delete dir fail"
5591 cleanup || error "cleanup failed with $?"
5593 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
5596 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5598 [ "$MDS1_VERSION" -ge $(version_code 2.7.62) ] ||
5599 skip "Need MDS version at least 2.7.62"
5601 reformat || error "reformat failed with $?"
5605 local mdsdev=$(mdsdevname 1)
5606 local ostdev=$(ostdevname 1)
5607 local mdsvdev=$(mdsvdevname 1)
5608 local ostvdev=$(ostvdevname 1)
5609 local opts_mds="$(mkfs_opts mds1 $mdsdev) --reformat $mdsdev $mdsvdev"
5610 local opts_ost="$(mkfs_opts ost1 $ostdev) --reformat $ostdev $ostvdev"
5612 if ! combined_mgs_mds ; then
5616 add mds1 $opts_mds || error "add mds1 failed"
5617 start_mdt 1 || error "start mdt1 failed"
5618 add ost1 $opts_ost || error "add ost1 failed"
5619 start_ost || error "start ost failed"
5620 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
5622 local soc=$(do_facet mds1 "$LCTL get_param -n \
5623 mdt.*MDT0000.sync_lock_cancel")
5624 [ $soc == "never" ] || error "SoC enabled on single MDS"
5626 for i in $(seq 2 $MDSCOUNT); do
5627 mdsdev=$(mdsdevname $i)
5628 mdsvdev=$(mdsvdevname $i)
5629 opts_mds="$(mkfs_opts mds$i $mdsdev) --reformat $mdsdev \
5631 add mds$i $opts_mds || error "add mds$i failed"
5632 start_mdt $i || error "start mdt$i fail"
5635 wait_dne_interconnect
5637 for i in $(seq $MDSCOUNT); do
5638 soc=$(do_facet mds$i "$LCTL get_param -n \
5639 mdt.*MDT000$((i - 1)).sync_lock_cancel")
5640 [ $soc == "blocking" ] || error "SoC not enabled on DNE"
5643 for i in $(seq 2 $MDSCOUNT); do
5644 stop_mdt $i || error "stop mdt$i fail"
5646 soc=$(do_facet mds1 "$LCTL get_param -n \
5647 mdt.*MDT0000.sync_lock_cancel")
5648 [ $soc == "never" ] || error "SoC enabled on single MDS"
5649 umount_client $MOUNT -f > /dev/null
5651 cleanup || error "cleanup failed with $?"
5653 run_test 70e "Sync-on-Cancel will be enabled by default on DNE"
5656 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5657 if combined_mgs_mds; then
5658 skip "needs separate MGS/MDT"
5662 start_mdt 1 || error "MDT0 start fail"
5663 start_ost || error "OST0 start fail"
5664 for num in $(seq 2 $MDSCOUNT); do
5665 start_mdt $num || return
5668 start_ost2 || error "OST1 start fail"
5670 mount_client $MOUNT || error "mount client fails"
5672 mkdir $DIR/$tdir || error "mkdir fail"
5673 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5674 error "create remote dir succeed"
5676 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5677 rm -rf $DIR/$tdir || error "delete dir fail"
5679 umount_client $MOUNT || error "umount_client failed"
5680 stop_mds || error "MDS stop fail"
5681 stop_ost || error "OST0 stop fail"
5682 stop_ost2 || error "OST1 stop fail"
5684 run_test 71a "start MDT0 OST0, MDT1, OST1"
5687 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5688 if combined_mgs_mds; then
5689 skip "needs separate MGS/MDT"
5693 for num in $(seq 2 $MDSCOUNT); do
5694 start_mdt $num || return
5696 start_ost || error "OST0 start fail"
5697 start_mdt 1 || error "MDT0 start fail"
5698 start_ost2 || error "OST1 start fail"
5700 mount_client $MOUNT || error "mount client fails"
5702 mkdir $DIR/$tdir || error "mkdir fail"
5703 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5704 error "create remote dir succeed"
5706 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5707 rm -rf $DIR/$tdir || error "delete dir fail"
5709 umount_client $MOUNT || error "umount_client failed"
5710 stop_mds || error "MDT0 stop fail"
5711 stop_ost || error "OST0 stop fail"
5712 stop_ost2 || error "OST1 stop fail"
5714 run_test 71b "start MDT1, OST0, MDT0, OST1"
5717 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5718 combined_mgs_mds && skip "needs separate MGS/MDT"
5722 start_ost || error "OST0 start fail"
5723 start_ost2 || error "OST1 start fail"
5724 for num in $(seq 2 $MDSCOUNT); do
5725 start_mdt $num || return
5727 start_mdt 1 || error "MDT0 start fail"
5729 mount_client $MOUNT || error "mount client fails"
5731 mkdir $DIR/$tdir || error "mkdir fail"
5732 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5733 error "create remote dir succeed"
5735 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5736 rm -rf $DIR/$tdir || error "delete dir fail"
5738 umount_client $MOUNT || error "umount_client failed"
5739 stop_mds || error "MDS stop fail"
5740 stop_ost || error "OST0 stop fail"
5741 stop_ost2 || error "OST1 stop fail"
5744 run_test 71c "start OST0, OST1, MDT1, MDT0"
5747 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5748 combined_mgs_mds && skip "needs separate MGS/MDT"
5752 start_ost || error "OST0 start fail"
5753 for num in $(seq 2 $MDSCOUNT); do
5754 start_mdt $num || return
5756 start_mdt 1 || error "MDT0 start fail"
5757 start_ost2 || error "OST1 start fail"
5759 mount_client $MOUNT || error "mount client fails"
5761 mkdir $DIR/$tdir || error "mkdir fail"
5762 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5763 error "create remote dir succeed"
5765 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5766 rm -rf $DIR/$tdir || error "delete dir fail"
5768 umount_client $MOUNT || error "umount_client failed"
5769 stop_mds || error "MDS stop fail"
5770 stop_ost || error "OST0 stop fail"
5771 stop_ost2 || error "OST1 stop fail"
5774 run_test 71d "start OST0, MDT1, MDT0, OST1"
5777 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5778 combined_mgs_mds && skip "needs separate MGS/MDT"
5782 start_ost || error "OST0 start fail"
5783 for num in $(seq 2 $MDSCOUNT); do
5784 start_mdt $num || return
5786 start_ost2 || error "OST1 start fail"
5787 start_mdt 1 || error "MDT0 start fail"
5789 mount_client $MOUNT || error "mount client fails"
5791 mkdir $DIR/$tdir || error "mkdir fail"
5792 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5793 error "create remote dir succeed"
5795 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5796 rm -rf $DIR/$tdir || error "delete dir fail"
5798 umount_client $MOUNT || error "umount_client failed"
5799 stop_mds || error "MDS stop fail"
5800 stop_ost || error "OST0 stop fail"
5801 stop_ost2 || error "OST1 stop fail"
5804 run_test 71e "start OST0, MDT1, OST1, MDT0"
5806 test_72() { #LU-2634
5807 [ "$mds1_FSTYPE" != ldiskfs ] &&
5808 skip "ldiskfs only test"
5810 local mdsdev=$(mdsdevname 1)
5811 local ostdev=$(ostdevname 1)
5812 local cmd="$E2FSCK -fnvd $mdsdev"
5819 if combined_mgs_mds; then
5820 add_options='--reformat'
5822 add_options='--reformat --replace'
5825 #tune MDT with "-O extents"
5827 for num in $(seq $MDSCOUNT); do
5828 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
5829 $add_options $(mdsdevname $num) $(mdsvdevname $num) ||
5830 error "add mds $num failed"
5831 do_facet mds${num} "$TUNE2FS -O extents $(mdsdevname $num)" ||
5832 error "$TUNE2FS failed on mds${num}"
5835 add ost1 $(mkfs_opts ost1 $ostdev) $add_options $ostdev ||
5836 error "add $ostdev failed"
5837 start_mds || error "start mds failed"
5838 start_ost || error "start ost failed"
5839 mount_client $MOUNT || error "mount client failed"
5841 #create some short symlinks
5842 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5843 createmany -o $DIR/$tdir/$tfile-%d $fn
5844 echo "create $fn short symlinks"
5845 for i in $(seq -w 1 $fn); do
5846 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
5851 umount_client $MOUNT || error "umount client failed"
5852 stop_mds || error "stop mds failed"
5853 stop_ost || error "stop ost failed"
5856 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
5858 run_test 72 "test fast symlink with extents flag enabled"
5860 test_73() { #LU-3006
5861 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
5862 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
5863 error "1st tunefs failed"
5864 start_mgsmds || error "start mds failed"
5865 start_ost || error "start ost failed"
5866 mount_client $MOUNT || error "mount client failed"
5867 $LCTL get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
5868 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
5869 umount_client $MOUNT || error "umount client failed"
5873 run_test 73 "failnode to update from mountdata properly"
5875 test_75() { # LU-2374
5876 [[ "$MDS1_VERSION" -lt $(version_code 2.4.1) ]] &&
5877 skip "Need MDS version at least 2.4.1"
5880 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
5881 --replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
5882 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
5883 --replace --reformat $(ostdevname 1) $(ostvdevname 1)"
5885 #check with default parameters
5886 add mds1 $opts_mds || error "add mds1 failed for default params"
5887 add ost1 $opts_ost || error "add ost1 failed for default params"
5889 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
5890 opts_mds=$(echo $opts_mds |
5891 sed -e "s/--index=$index/--index=$index --mdt/")
5892 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
5893 opts_ost=$(echo $opts_ost |
5894 sed -e "s/--index=$index/--index=$index --ost/")
5896 add mds1 $opts_mds || error "add mds1 failed for new params"
5897 add ost1 $opts_ost || error "add ost1 failed for new params"
5902 run_test 75 "The order of --index should be irrelevant"
5905 [[ "$MGS_VERSION" -ge $(version_code 2.4.52) ]] ||
5906 skip "Need MDS version at least 2.4.52"
5909 local MDMB_PARAM="osc.*.max_dirty_mb"
5910 echo "Change MGS params"
5911 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
5913 echo "max_dirty_mb: $MAX_DIRTY_MB"
5914 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB - 10))
5915 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
5916 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
5917 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5918 head -1" $NEW_MAX_DIRTY_MB
5919 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5920 echo "$MAX_DIRTY_MB"
5921 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5922 error "error while apply max_dirty_mb"
5924 echo "Check the value is stored after remount"
5927 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5928 head -1" $NEW_MAX_DIRTY_MB
5929 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5930 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5931 error "max_dirty_mb is not saved after remount"
5933 echo "Change OST params"
5934 CLIENT_PARAM="obdfilter.*.client_cache_count"
5935 local CLIENT_CACHE_COUNT
5936 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5938 echo "client_cache_count: $CLIENT_CACHE_COUNT"
5939 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
5940 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
5941 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
5942 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5943 head -1" $NEW_CLIENT_CACHE_COUNT
5944 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5946 echo "$CLIENT_CACHE_COUNT"
5947 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5948 error "error while apply client_cache_count"
5950 echo "Check the value is stored after remount"
5953 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5954 head -1" $NEW_CLIENT_CACHE_COUNT
5955 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5957 echo "$CLIENT_CACHE_COUNT"
5958 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5959 error "client_cache_count is not saved after remount"
5962 run_test 76a "set permanent params with lctl across mounts"
5964 test_76b() { # LU-4783
5965 [[ "$MGS_VERSION" -ge $(version_code 2.5.57) ]] ||
5966 skip "Need MGS version at least 2.5.57"
5969 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
5970 error "start params log failed"
5973 run_test 76b "verify params log setup correctly"
5976 [[ "$MGS_VERSION" -ge $(version_code 2.8.54) ]] ||
5977 skip "Need MDS version at least 2.4.52"
5979 local MASK_PARAM="mdd.*.changelog_mask"
5980 echo "Change changelog_mask"
5981 do_facet mgs $LCTL set_param -P $MASK_PARAM=-CLOSE ||
5982 error "Can't change changlog_mask"
5983 wait_update $(facet_host mds) "$LCTL get_param -n $MASK_PARAM |
5986 echo "Check the value is stored after mds remount"
5987 stop_mds || error "Failed to stop MDS"
5988 start_mds || error "Failed to start MDS"
5989 local CHANGELOG_MASK=$(do_facet mgs $LCTL get_param -n $MASK_PARAM)
5990 echo $CHANGELOG_MASK | grep CLOSE > /dev/null &&
5991 error "changelog_mask is not changed"
5995 run_test 76c "verify changelog_mask is applied with lctl set_param -P"
5997 test_76d() { #LU-9399
6000 local xattr_cache="llite.*.xattr_cache"
6001 local cmd="$LCTL get_param -n $xattr_cache | head -1"
6002 local new=$((($(eval $cmd) + 1) % 2))
6004 echo "lctl set_param -P llite.*.xattr_cache=$new"
6005 do_facet mgs $LCTL set_param -P $xattr_cache=$new ||
6006 error "Can't change xattr_cache"
6007 wait_update $HOSTNAME "$cmd" "$new"
6009 echo "Check $xattr_cache on client $MOUNT"
6010 umount_client $MOUNT || error "umount $MOUNT failed"
6011 mount_client $MOUNT || error "mount $MOUNT failed"
6012 [ $(eval $cmd) -eq $new ] ||
6013 error "$xattr_cache != $new on client $MOUNT"
6015 echo "Check $xattr_cache on the new client $MOUNT2"
6016 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
6017 [ $(eval $cmd) -eq $new ] ||
6018 error "$xattr_cache != $new on client $MOUNT2"
6019 umount_client $MOUNT2 || error "umount $MOUNT2 failed"
6023 run_test 76d "verify llite.*.xattr_cache can be set by 'lctl set_param -P' correctly"
6025 test_77() { # LU-3445
6026 [[ "$MDS1_VERSION" -ge $(version_code 2.8.55) ]] ||
6027 skip "Need MDS version 2.8.55+ "
6029 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
6030 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
6031 skip_env "mixed loopback and real device not working"
6034 local fs2mdsdev=$(mdsdevname 1_2)
6035 local fs2ostdev=$(ostdevname 1_2)
6036 local fs2mdsvdev=$(mdsvdevname 1_2)
6037 local fs2ostvdev=$(ostvdevname 1_2)
6038 local fsname=test1234
6040 local failnid="$(h2nettype 1.2.3.4),$(h2nettype 4.3.2.1)"
6042 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
6044 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
6045 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
6046 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
6047 error "start fs2mds failed"
6049 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
6050 mgsnid="0.0.0.0@tcp,$mgsnid,$mgsnid:$mgsnid"
6052 add fs2ost --mgsnode=$mgsnid $(mkfs_opts ost1 $fs2ostdev) \
6053 --failnode=$failnid --fsname=$fsname \
6054 --reformat $fs2ostdev $fs2ostvdev ||
6055 error "add fs2ost failed"
6056 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
6058 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
6059 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
6060 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
6063 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
6066 [[ "$mds1_FSTYPE" != ldiskfs ||
6067 "$ost1_FSTYPE" != ldiskfs ]] &&
6068 skip "ldiskfs only test"
6070 # reformat the Lustre filesystem with a smaller size
6071 local saved_MDSCOUNT=$MDSCOUNT
6072 local saved_MDSSIZE=$MDSSIZE
6073 local saved_OSTCOUNT=$OSTCOUNT
6074 local saved_OSTSIZE=$OSTSIZE
6077 MDSSIZE=$((MDSSIZE - 20000))
6078 OSTSIZE=$((OSTSIZE - 20000))
6079 reformat || error "(1) reformat Lustre filesystem failed"
6080 MDSSIZE=$saved_MDSSIZE
6081 OSTSIZE=$saved_OSTSIZE
6083 # mount the Lustre filesystem
6084 setup_noconfig || error "(2) setup Lustre filesystem failed"
6087 log "create test files"
6092 mkdir $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
6094 for i in $(seq $num_files); do
6095 file=$MOUNT/$tdir/$tfile-$i
6096 dd if=/dev/urandom of=$file count=1 bs=1M || {
6097 $LCTL get_param osc.*.cur*grant*
6098 $LFS df; $LFS df -i;
6099 # stop creating files if there is no more space
6100 if [ ! -e $file ]; then
6101 num_files=$((i - 1))
6105 $LFS getstripe -v $file
6106 local ost_idx=$(LFS getstripe -i $file)
6107 do_facet ost$((ost_idx + 1)) \
6108 $LCTL get_param obdfilter.*.*grant*
6109 error "(4) create $file failed"
6113 # unmount the Lustre filesystem
6114 cleanup || error "(5) cleanup Lustre filesystem failed"
6116 # run e2fsck on the MDT and OST devices
6117 local mds_host=$(facet_active_host $SINGLEMDS)
6118 local ost_host=$(facet_active_host ost1)
6119 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
6120 local ost_dev=$(ostdevname 1)
6122 run_e2fsck $mds_host $mds_dev "-y"
6123 run_e2fsck $ost_host $ost_dev "-y"
6125 # get the original block count of the MDT and OST filesystems
6126 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
6127 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
6129 # expand the MDT and OST filesystems to the device size
6130 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
6131 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
6133 # run e2fsck on the MDT and OST devices again
6134 run_e2fsck $mds_host $mds_dev "-y"
6135 run_e2fsck $ost_host $ost_dev "-y"
6137 # mount the Lustre filesystem
6141 log "check files after expanding the MDT and OST filesystems"
6142 for i in $(seq $num_files); do
6143 file=$MOUNT/$tdir/$tfile-$i
6144 $CHECKSTAT -t file -s 1048576 $file ||
6145 error "(6) checkstat $file failed"
6149 log "create more files after expanding the MDT and OST filesystems"
6150 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
6151 file=$MOUNT/$tdir/$tfile-$i
6152 dd if=/dev/urandom of=$file count=1 bs=1M ||
6153 error "(7) create $file failed"
6156 # unmount the Lustre filesystem
6157 cleanup || error "(8) cleanup Lustre filesystem failed"
6159 # run e2fsck on the MDT and OST devices
6160 run_e2fsck $mds_host $mds_dev "-y"
6161 run_e2fsck $ost_host $ost_dev "-y"
6163 # get the maximum block count of the MDT and OST filesystems
6164 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
6165 local ost_max_blks=$(get_block_count ost1 $ost_dev)
6167 # get the minimum block count of the MDT and OST filesystems
6168 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
6169 grep minimum | sed -e 's/^.*filesystem: //g')
6170 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
6171 grep minimum | sed -e 's/^.*filesystem: //g')
6173 # shrink the MDT and OST filesystems to a smaller size
6177 if [[ $mds_max_blks -gt $mds_min_blks &&
6178 $mds_max_blks -gt $mds_orig_blks ]]; then
6179 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
6180 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
6181 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
6182 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
6183 error "shrink $SINGLEMDS to $new_blks failed"
6187 if [[ $ost_max_blks -gt $ost_min_blks &&
6188 $ost_max_blks -gt $ost_orig_blks ]]; then
6189 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
6190 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
6191 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
6192 run_resize2fs ost1 $ost_dev $new_blks ||
6193 error "shrink ost1 to $new_blks failed"
6197 # check whether the MDT or OST filesystem was shrunk or not
6199 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
6200 reformat_and_config ||
6201 error "(10) reformat Lustre filesystem failed"
6205 # run e2fsck on the MDT and OST devices again
6206 run_e2fsck $mds_host $mds_dev "-y"
6207 run_e2fsck $ost_host $ost_dev "-y"
6209 # mount the Lustre filesystem again
6213 log "check files after shrinking the MDT and OST filesystems"
6214 for i in $(seq $((num_files + 10))); do
6215 file=$MOUNT/$tdir/$tfile-$i
6216 $CHECKSTAT -t file -s 1048576 $file ||
6217 error "(11) checkstat $file failed"
6220 # unmount and reformat the Lustre filesystem
6221 cleanup || error "(12) cleanup Lustre filesystem failed"
6222 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
6224 MDSCOUNT=$saved_MDSCOUNT
6225 OSTCOUNT=$saved_OSTCOUNT
6226 reformat_and_config || error "(14) reformat Lustre filesystem failed"
6228 run_test 78 "run resize2fs on MDT and OST filesystems"
6230 test_79() { # LU-4227
6231 [[ "$MDS1_VERSION" -ge $(version_code 2.5.59) ]] ||
6232 skip "Need MDS version at least 2.5.59"
6234 local mdsdev1=$(mdsdevname 1)
6235 local mdsvdev1=$(mdsvdevname 1)
6236 local mdsdev2=$(mdsdevname 2)
6237 local mdsvdev2=$(mdsvdevname 2)
6238 local ostdev1=$(ostdevname 1)
6239 local ostvdev1=$(ostvdevname 1)
6240 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
6241 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
6242 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
6245 # remove --mgs/--mgsnode from mkfs.lustre options
6246 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
6248 mgsnode_opt=$(echo $opts_mds2 |
6249 awk '{ for ( i = 1; i < NF; i++ )
6250 if ( $i ~ "--mgsnode" ) { print $i; break } }')
6251 [ -n "$mgsnode_opt" ] &&
6252 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
6254 mgsnode_opt=$(echo $opts_ost1 |
6255 awk '{ for ( i = 1; i < NF; i++ )
6256 if ( $i ~ "--mgsnode" ) { print $i; break } }')
6257 [ -n "$mgsnode_opt" ] &&
6258 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
6260 # -MGS, format a mdt without --mgs option
6261 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
6262 error "Must specify --mgs when formatting mdt combined with mgs"
6264 # +MGS, format a mdt/ost without --mgsnode option
6265 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
6266 > /dev/null || error "start mds1 failed"
6267 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
6268 error "Must specify --mgsnode when formatting a mdt"
6269 add ost1 $opts_ost1 $ostdev1 $ostvdev2 &&
6270 error "Must specify --mgsnode when formatting an ost"
6274 run_test 79 "format MDT/OST without mgs option (should return errors)"
6277 start_mds || error "Failed to start MDT"
6278 start_ost || error "Failed to start OST1"
6279 uuid=$(do_facet ost1 $LCTL get_param -n mgc.*.uuid)
6280 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
6281 do_facet ost1 "$LCTL set_param fail_val=10 fail_loc=0x906"
6282 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x906"
6283 do_facet mgs "$LCTL set_param -n mgs/MGS/evict_client $uuid"
6285 start_ost2 || error "Failed to start OST2"
6287 do_facet ost1 "$LCTL set_param fail_loc=0"
6292 run_test 80 "mgc import reconnect race"
6294 #Save the original values of $OSTCOUNT and $OSTINDEX$i.
6296 local new_ostcount=$1
6297 saved_ostcount=$OSTCOUNT
6298 OSTCOUNT=$new_ostcount
6302 for i in $(seq $OSTCOUNT); do
6304 eval saved_ostindex$i=${!index}
6309 # Restore the original values of $OSTCOUNT and $OSTINDEX$i.
6310 restore_ostindex() {
6315 for i in $(seq $OSTCOUNT); do
6316 index=saved_ostindex$i
6317 eval OSTINDEX$i=${!index}
6319 OSTCOUNT=$saved_ostcount
6322 if ! combined_mgs_mds ; then
6327 # The main purpose of this test is to ensure the OST_INDEX_LIST functions as
6328 # expected. This test uses OST_INDEX_LIST to format OSTs with a randomly
6329 # assigned index and ensures we can mount such a formatted file system
6330 test_81() { # LU-4665
6331 [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
6332 skip "Need MDS version at least 2.6.54"
6333 [[ $OSTCOUNT -ge 3 ]] || skip_env "needs >= 3 OSTs"
6337 # Each time RANDOM is referenced, a random integer between 0 and 32767
6340 local saved_ostindex1=$OSTINDEX1
6341 for i in 65535 $((RANDOM + 65536)); do
6342 echo -e "\nFormat ost1 with --index=$i, should fail"
6344 if add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat \
6345 $(ostdevname 1) $(ostvdevname 1); then
6346 OSTINDEX1=$saved_ostindex1
6347 error "format ost1 with --index=$i should fail"
6350 OSTINDEX1=$saved_ostindex1
6354 # Format OSTs with random sparse indices.
6355 trap "restore_ostindex" EXIT
6356 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices"
6357 OST_INDEX_LIST=[0,$((RANDOM * 2 % 65533 + 1)),65534] formatall
6359 # Setup and check Lustre filesystem.
6360 start_mgsmds || error "start_mgsmds failed"
6361 for i in $(seq $OSTCOUNT); do
6362 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6363 error "start ost$i failed"
6366 mount_client $MOUNT || error "mount client $MOUNT failed"
6367 check_mount || error "check client $MOUNT failed"
6370 local max_easize=$($LCTL get_param -n llite.*.max_easize)
6371 # 65452 is XATTR_SIZE_MAX less ldiskfs ea overhead
6372 if large_xattr_enabled; then
6373 [[ $max_easize -ge 65452 ]] ||
6374 error "max_easize is $max_easize, should be at least 65452 bytes"
6377 # 4012 is 4096 less ldiskfs ea overhead
6378 [[ $max_easize -ge 4012 ]] ||
6379 error "max_easize is $max_easize, should be at least 4012 bytes"
6384 run_test 81 "sparse OST indexing"
6386 # Here we exercise the stripe placement functionality on a file system that
6387 # has formatted the OST with a random index. With the file system the following
6388 # functionality is tested:
6390 # 1. Creating a new file with a specific stripe layout.
6392 # 2. Modifiy a existing empty file with a specific stripe layout.
6394 # 3. Ensure we fail to set the stripe layout of a file that already has one.
6396 # 4. If ost-index is defined we need to ensure it is the first entry in the
6397 # ost index list returned by lfs getstripe.
6399 # 5. Lastly ensure this functionality fails with directories.
6400 test_82a() { # LU-4665
6401 [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
6402 skip "Need MDS version at least 2.6.54"
6403 [[ $OSTCOUNT -ge 3 ]] || skip_env "needs >= 3 OSTs"
6409 # Format OSTs with random sparse indices.
6413 local LOV_V1_INSANE_STRIPE_COUNT=65532
6414 for i in $(seq $OSTCOUNT); do
6415 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
6416 ost_indices+=" $index"
6418 ost_indices=$(comma_list $ost_indices)
6420 stack_trap "restore_ostindex" EXIT
6421 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
6422 OST_INDEX_LIST=[$ost_indices] formatall
6424 # Setup Lustre filesystem.
6425 start_mgsmds || error "start_mgsmds failed"
6426 for i in $(seq $OSTCOUNT); do
6427 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6428 error "start ost$i failed"
6431 # Collect debug information - start of test
6432 do_nodes $(comma_list $(mdts_nodes)) \
6433 $LCTL get_param osp.*.prealloc_*_id
6435 mount_client $MOUNT || error "mount client $MOUNT failed"
6439 check_lfs_df_ret_val $? || error "$LFS df $MOUNT failed"
6440 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6442 stack_trap "do_nodes $(comma_list $(mdts_nodes)) \
6443 $LCTL get_param osp.*.prealloc_*_id" EXIT
6445 # 1. If the file does not exist, new file will be created
6446 # with specified OSTs.
6447 local file=$DIR/$tdir/$tfile-1
6448 local cmd="$LFS setstripe -o $ost_indices $file"
6450 eval $cmd || error "$cmd failed"
6451 check_stripe_count $file $OSTCOUNT
6452 check_obdidx $file $ost_indices
6453 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6454 error "write $file failed"
6456 # 2. If the file already exists and is an empty file, the file
6457 # will be attached with specified layout.
6458 file=$DIR/$tdir/$tfile-2
6459 mcreate $file || error "mcreate $file failed"
6460 cmd="$LFS setstripe -o $ost_indices $file"
6462 eval $cmd || error "$cmd failed"
6463 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6464 error "write $file failed"
6465 check_stripe_count $file $OSTCOUNT
6466 check_obdidx $file $ost_indices
6468 # 3. If the file already has a valid layout attached, the command
6469 # should fail with EBUSY.
6471 eval $cmd && error "stripe is already set on $file, $cmd should fail"
6473 # 4. If [--stripe-index|-i <start_ost_idx>] is used, the index must
6474 # be in the OST indices list.
6475 local start_ost_idx=${ost_indices##*,}
6476 file=$DIR/$tdir/$tfile-3
6477 cmd="$LFS setstripe -o $ost_indices -i $start_ost_idx $file"
6479 eval $cmd || error "$cmd failed"
6480 check_stripe_count $file $OSTCOUNT
6481 check_obdidx $file $ost_indices
6482 check_start_ost_idx $file $start_ost_idx
6484 file=$DIR/$tdir/$tfile-4
6485 cmd="$LFS setstripe"
6486 cmd+=" -o $(exclude_items_from_list $ost_indices $start_ost_idx)"
6487 cmd+=" -i $start_ost_idx $file"
6489 eval $cmd && error "index $start_ost_idx should be in $ost_indices"
6491 # 5. Specifying OST indices for directory should succeed.
6492 local dir=$DIR/$tdir/$tdir
6493 mkdir $dir || error "mkdir $dir failed"
6494 cmd="$LFS setstripe -o $ost_indices $dir"
6495 if [ "$MDS1_VERSION" -gt $(version_code 2.11.53) ] &&
6496 [ "$CLIENT_VERSION" -gt $(version_code 2.11.53) ]; then
6498 eval $cmd || error "unable to specify OST indices on directory"
6500 echo "need MDS+client version at least 2.11.53"
6505 run_test 82a "specify OSTs for file (succeed) or directory (succeed)"
6510 # Remove OSTs from a pool and destroy the pool.
6511 destroy_pool $ost_pool || true
6516 # Test 82b is run to ensure that if the user supplies a pool with a specific
6517 # stripe layout that it behaves proprerly. It should fail in the case that
6518 # the supplied OST index list points to OSTs not contained in the user
6520 test_82b() { # LU-4665
6521 [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
6522 skip "Need MDS version at least 2.6.54"
6523 [[ $OSTCOUNT -ge 4 ]] || skip_env "needs >= 4 OSTs"
6529 # Format OSTs with random sparse indices.
6533 local LOV_V1_INSANE_STRIPE_COUNT=65532
6534 for i in $(seq $OSTCOUNT); do
6535 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
6536 ost_indices+=" $index"
6538 ost_indices=$(comma_list $ost_indices)
6540 trap "restore_ostindex" EXIT
6541 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
6542 OST_INDEX_LIST=[$ost_indices] formatall
6544 # Setup Lustre filesystem.
6545 start_mgsmds || error "start_mgsmds failed"
6546 for i in $(seq $OSTCOUNT); do
6547 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6548 error "start ost$i failed"
6551 mount_client $MOUNT || error "mount client $MOUNT failed"
6555 check_lfs_df_ret_val $? || error "$LFS df $MOUNT failed"
6556 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6558 # Create a new pool and add OSTs into it.
6559 local ost_pool=$FSNAME.$TESTNAME
6560 create_pool $ost_pool || error "create OST pool $ost_pool failed"
6563 trap "cleanup_82b" EXIT
6565 local ost_idx_in_list=${ost_indices##*,}
6566 local ost_idx_in_pool=$(exclude_items_from_list $ost_indices \
6569 local ost_targets="$FSNAME-OST["
6570 for i in ${ost_idx_in_pool//,/ }; do
6571 ost_targets=$ost_targets$(printf "%04x," $i)
6573 ost_targets="${ost_targets%,}]"
6575 local ost_targets_uuid=$(for i in ${ost_idx_in_pool//,/ }; \
6576 do printf "$FSNAME-OST%04x_UUID\n" $i; done |
6577 sort -u | tr '\n' ' ')
6579 local cmd="$LCTL pool_add $ost_pool $ost_targets"
6580 do_facet mgs $cmd || error "$cmd failed"
6581 wait_update $HOSTNAME "$LCTL get_param -n lov.$FSNAME-*.pools.$TESTNAME|
6582 sort -u | tr '\n' ' ' " "$ost_targets_uuid" ||
6583 error "wait_update $ost_pool failed"
6584 wait_update_facet $SINGLEMDS "$LCTL pool_list $ost_pool | wc -l" 4 ||
6585 error "wait_update pool_list $ost_pool failed"
6587 # If [--pool|-p <pool_name>] is set with [--ost-list|-o <ost_indices>],
6588 # then the OSTs must be the members of the pool.
6589 local file=$DIR/$tdir/$tfile
6590 cmd="$LFS setstripe -p $ost_pool -o $ost_idx_in_list $file"
6592 eval $cmd && error "OST with index $ost_idx_in_list should be" \
6593 "in OST pool $ost_pool"
6595 # Only select OST $ost_idx_in_list from $ost_pool for file.
6596 ost_idx_in_list=${ost_idx_in_pool#*,}
6597 cmd="$LFS setstripe -p $ost_pool -o $ost_idx_in_list $file"
6599 eval $cmd || error "$cmd failed"
6600 cmd="$LFS getstripe $file"
6602 eval $cmd || error "$cmd failed"
6603 check_stripe_count $file 2
6604 check_obdidx $file $ost_idx_in_list
6605 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6606 error "write $file failed"
6610 run_test 82b "specify OSTs for file with --pool and --ost-list options"
6613 [[ "$OST1_VERSION" -ge $(version_code 2.6.91) ]] ||
6614 skip "Need OST version at least 2.6.91"
6615 if [ "$ost1_FSTYPE" != ldiskfs ]; then
6616 skip "ldiskfs only test"
6625 ostmnt=$(facet_mntpt ost1)
6627 # Mount the OST as an ldiskfs filesystem.
6628 log "mount the OST $dev as a $ost1_FSTYPE filesystem"
6629 add ost1 $(mkfs_opts ost1 $dev) $FSTYPE_OPT \
6630 --reformat $dev > /dev/null ||
6631 error "format ost1 error"
6633 if ! test -b $dev; then
6634 mnt_opts=$(csa_add "$OST_MOUNT_FS_OPTS" -o loop)
6636 echo "mnt_opts $mnt_opts"
6637 do_facet ost1 mount -t "$ost1_FSTYPE" $dev \
6639 # Run llverfs on the mounted ldiskfs filesystem.
6640 # It is needed to get ENOSPACE.
6641 log "run llverfs in partial mode on the OST $ost1_FSTYPE $ostmnt"
6642 do_rpc_nodes $(facet_host ost1) run_llverfs $ostmnt -vpl \
6643 "no" || error "run_llverfs error on $ost1_FSTYPE"
6646 log "unmount the OST $dev"
6649 # Delete file IO_scrub. Later osd_scrub_setup will try to
6650 # create "IO_scrub" but will get ENOSPACE.
6652 echo "start ost1 service on `facet_active_host ost1`"
6653 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
6656 err=$(do_facet ost1 dmesg | grep "VFS: Busy inodes after unmount of")
6657 echo "string err $err"
6658 [ -z "$err" ] || error $err
6661 run_test 83 "ENOSPACE on OST doesn't cause message VFS: \
6662 Busy inodes after unmount ..."
6665 local facet=$SINGLEMDS
6666 local num=$(echo $facet | tr -d "mds")
6667 local dev=$(mdsdevname $num)
6668 local time_min=$(recovery_time_min)
6669 local recovery_duration
6670 local completed_clients
6671 local correct_clients
6674 echo "start mds service on $(facet_active_host $facet)"
6676 "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
6677 error "start MDS failed"
6679 start_ost || error "start OST0000 failed"
6680 start_ost2 || error "start OST0001 failed"
6682 echo "recovery_time=$time_min, timeout=$TIMEOUT, wrap_up=$wrap_up"
6684 mount_client $MOUNT1 || error "mount $MOUNT1 failed"
6685 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
6686 # make sure new superblock labels are sync'd before disabling writes
6690 replay_barrier $SINGLEMDS
6691 createmany -o $DIR1/$tfile-%d 1000
6693 # We need to catch the end of recovery window to extend it.
6694 # Skip 5 requests and add delay to request handling.
6695 #define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP
6696 do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5"
6698 facet_failover --fsck $SINGLEMDS || error "failover: $?"
6701 echo "recovery status"
6702 do_facet $SINGLEMDS \
6703 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status"
6705 recovery_duration=$(do_facet $SINGLEMDS \
6706 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
6707 awk '/recovery_duration/ { print $2 }')
6708 (( $recovery_duration > $time_min + $wrap_up )) &&
6709 error "recovery_duration > recovery_time_hard + wrap up"
6710 completed_clients=$(do_facet $SINGLEMDS \
6711 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
6712 awk '/completed_clients/ { print $2 }')
6714 correct_clients="$MDSCOUNT/$((MDSCOUNT+1))"
6715 [ "$completed_clients" = "${correct_clients}" ] ||
6716 error "$completed_clients != $correct_clients"
6718 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
6719 umount_client $MOUNT1
6720 umount_client $MOUNT2
6726 run_test 84 "check recovery_hard_time"
6729 [[ "$OST1_VERSION" -ge $(version_code 2.7.55) ]] ||
6730 skip "Need OST version at least 2.7.55"
6731 ##define OBD_FAIL_OSD_OST_EA_FID_SET 0x197
6732 do_facet ost1 "lctl set_param fail_loc=0x197"
6736 run_test 85 "osd_ost init: fail ea_fid_set"
6741 # ost1 has already registered to the MGS before the reformat.
6742 # So after reformatting it with option "-G", it could not be
6743 # mounted to the MGS. Cleanup the system for subsequent tests.
6748 [ "$ost1_FSTYPE" = zfs ] &&
6749 skip "LU-6442: no such mkfs params for ZFS OSTs"
6750 [[ "$MDS1_VERSION" -ge $(version_code 2.7.56) ]] ||
6751 skip "Need server version newer than 2.7.55"
6754 local OLDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6755 awk '/Flex block group size: / { print $NF; exit; }')
6757 [ "$OLDSIZE" == "$NEWSIZE" ] && skip "$NEWSIZE groups already"
6759 local opts=" -O flex_bg -G $NEWSIZE"
6760 opts=$(OST_FS_MKFS_OPTS+="$opts" mkfs_opts ost1 $(ostdevname 1))
6761 opts+=" --reformat $(ostdevname 1) $(ostvdevname 1)"
6762 echo "params: $opts"
6764 trap cleanup_86 EXIT ERR
6767 add ost1 $opts || error "add ost1 failed with new params"
6769 local FOUNDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6770 awk '/Flex block group size: / { print $NF; exit; }')
6772 [[ $FOUNDSIZE == $NEWSIZE ]] ||
6773 error "Flex block group size: $FOUNDSIZE, expected: $NEWSIZE"
6777 run_test 86 "Replacing mkfs.lustre -G option"
6779 test_87() { #LU-6544
6780 [[ "$MDS1_VERSION" -ge $(version_code 2.9.51) ]] ||
6781 skip "Need MDS version at least 2.9.51"
6782 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
6783 skip "ldiskfs only test"
6784 [[ $OSTCOUNT -gt 59 ]] &&
6785 skip "Ignore wide striping situation"
6786 [ -n "$FILESET" ] && skip "Not functional for FILESET set"
6788 local mdsdev=$(mdsdevname 1)
6789 local mdsvdev=$(mdsvdevname 1)
6790 local file=$DIR/$tfile
6791 local mntpt=$(facet_mntpt $SINGLEMDS)
6792 local used_xattr_blk=0
6793 local inode_size=${1:-1024}
6795 local xtest="trusted.test"
6799 local stripe_cnt=$(($OSTCOUNT + 2))
6801 #Please see ldiskfs_make_lustre() for MDT inode size calculation
6802 if [ $stripe_cnt -gt 16 ]; then
6805 left_size=$(expr $inode_size - \
6808 32 - 40 \* 3 - 32 \* 3 - $stripe_cnt \* 24 - 16 - 3 - \
6810 24 - 18 - $(expr length $tfile) - 16 - 4)
6811 if [ $left_size -le 0 ]; then
6812 echo "No space($left_size) is expected in inode."
6813 echo "Try 1-byte xattr instead to verify this."
6816 echo "Estimate: at most $left_size-byte space left in inode."
6822 if ! combined_mgs_mds ; then
6826 add mds1 $(mkfs_opts mds1 ${mdsdev}) --stripe-count-hint=$stripe_cnt \
6827 --reformat $mdsdev $mdsvdev || error "add mds1 failed"
6828 start_mdt 1 > /dev/null || error "start mdt1 failed"
6829 for i in $(seq $OSTCOUNT); do
6830 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS > /dev/null ||
6831 error "start ost$i failed"
6833 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
6834 check_mount || error "check client $MOUNT failed"
6837 $LFS setstripe -E 1M -S 1M -c 1 -E 64M -c 1 -E -1 -c -1 $file ||
6838 error "Create file with 3 components failed"
6839 $TRUNCATE $file $((1024*1024*64+1)) || error "truncate file failed"
6840 i=$($LFS getstripe -I3 -c $file) || error "get 3rd stripe count failed"
6841 if [ $i -ne $OSTCOUNT ]; then
6842 left_size=$(expr $left_size + $(expr $OSTCOUNT - $i) \* 24)
6843 echo -n "Since only $i out $OSTCOUNT OSTs are used, "
6844 echo -n "the expected left space is changed to "
6845 echo "$left_size bytes at most."
6847 value=$(generate_string $left_size)
6848 setfattr -n $xtest -v $value $file
6849 orig=$(get_xattr_value $xtest $file)
6850 [[ "$orig" != "$value" ]] && error "$xtest changed"
6852 #Verify if inode has some expected space left
6853 umount $MOUNT > /dev/null || error "umount $MOUNT failed"
6854 stop_mdt 1 > /dev/null || error "stop mdt1 failed"
6855 mount_ldiskfs $SINGLEMDS || error "mount -t ldiskfs $SINGLEMDS failed"
6857 do_facet $SINGLEMDS ls -sal $mntpt/ROOT/$tfile
6858 used_xattr_blk=$(do_facet $SINGLEMDS ls -s $mntpt/ROOT/$tfile |
6860 [[ $used_xattr_blk -eq 0 ]] &&
6861 error "Please check MDS inode size calculation: \
6862 more than $left_size-byte space left in inode."
6863 echo "Verified: at most $left_size-byte space left in inode."
6865 unmount_ldiskfs $SINGLEMDS
6867 for i in $(seq $OSTCOUNT); do
6868 stop ost$i -f || error "stop ost$i failed"
6871 run_test 87 "check if MDT inode can hold EAs with N stripes properly"
6874 [ "$mds1_FSTYPE" == zfs ] &&
6875 skip "LU-6662: no implementation for ZFS"
6879 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
6880 --reformat $(mdsdevname 1) || error "add mds1 failed"
6882 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
6883 grep -e \".*opts:.*errors=remount-ro.*\"" ||
6884 error "default mount options is missing"
6886 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
6887 --mountfsoptions="user_xattr,errors=panic" \
6888 --reformat $(mdsdevname 1) || error "add mds1 failed"
6890 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
6891 grep -e \".*opts:.*errors=panic.*\"" ||
6892 error "user can't override default mount options"
6894 run_test 88 "check the default mount options can be overridden"
6896 test_89() { # LU-7131
6897 [[ "$MDS1_VERSION" -ge $(version_code 2.9.54) ]] ||
6898 skip "Need MDT version at least 2.9.54"
6900 local key=failover.node
6901 local val1=192.0.2.254@tcp0 # Reserved IPs, see RFC 5735
6902 local val2=192.0.2.255@tcp0
6903 local mdsdev=$(mdsdevname 1)
6908 if [[ "$mds1_FSTYPE" == zfs ]]; then
6909 import_zpool mds1 || return ${PIPESTATUS[0]}
6912 # Check that parameters are added correctly
6913 echo "tunefs --param $key=$val1"
6914 do_facet mds1 "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
6915 error "tunefs --param $key=$val1 failed"
6916 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
6917 error "tunefs --dryrun failed"
6918 params=${params##*Parameters:}
6919 params=${params%%exiting*}
6920 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val1) = "1" ] ||
6921 error "on-disk parameter not added correctly via tunefs"
6923 # Check that parameters replace existing instances when added
6924 echo "tunefs --param $key=$val2"
6925 do_facet mds1 "$TUNEFS --param $key=$val2 $mdsdev >/dev/null" ||
6926 error "tunefs --param $key=$val2 failed"
6927 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
6928 error "tunefs --dryrun failed"
6929 params=${params##*Parameters:}
6930 params=${params%%exiting*}
6931 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "1" ] ||
6932 error "on-disk parameter not replaced via tunefs"
6933 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val2) = "1" ] ||
6934 error "on-disk parameter not replaced correctly via tunefs"
6936 # Check that a parameter is erased properly
6937 echo "tunefs --erase-param $key"
6938 do_facet mds1 "$TUNEFS --erase-param $key $mdsdev >/dev/null" ||
6939 error "tunefs --erase-param $key failed"
6940 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
6941 error "tunefs --dryrun failed"
6942 params=${params##*Parameters:}
6943 params=${params%%exiting*}
6944 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "0" ] ||
6945 error "on-disk parameter not erased correctly via tunefs"
6947 # Check that all the parameters are erased
6948 do_facet mds1 "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
6949 error "tunefs --param $key=$val1 failed"
6950 echo "tunefs --erase-params"
6951 do_facet mds1 "$TUNEFS --erase-params $mdsdev >/dev/null" ||
6952 error "tunefs --erase-params failed"
6953 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
6954 error "tunefs --dryrun failed"
6955 params=${params##*Parameters:}
6956 params=${params%%exiting*}
6957 params=$(echo $params | tr ' ' '\n')
6959 error "all on-disk parameters not erased correctly via tunefs $params"
6961 # Check the order of options --erase-params and --param
6962 echo "tunefs --param $key=$val1 --erase-params"
6964 "$TUNEFS --param $key=$val1 --erase-params $mdsdev >/dev/null"||
6965 error "tunefs --param $key=$val1 --erase-params failed"
6966 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
6967 error "tunefs --dryrun failed"
6968 params=${params##*Parameters:}
6969 params=${params%%exiting*}
6970 [ $(echo $params | tr ' ' '\n') == "$key=$val1" ] ||
6971 error "on-disk param not added correctly with --erase-params"
6975 run_test 89 "check tunefs --param and --erase-param{s} options"
6978 # $2 (optional) value of max_mod_rpcs_in_flight to set
6979 check_max_mod_rpcs_in_flight() {
6987 idx=$(printf "%04x" $($LFS getdirstripe -i $dir))
6988 facet="mds$((0x$idx + 1))"
6990 if [ -z "$mmr" ]; then
6991 # get value of max_mod_rcps_in_flight
6992 mmr=$($LCTL get_param -n \
6993 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) ||
6994 error "Unable to get max_mod_rpcs_in_flight"
6995 echo "max_mod_rcps_in_flight is $mmr"
6997 # set value of max_mod_rpcs_in_flight
6999 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight=$mmr ||
7000 error "Unable to set max_mod_rpcs_in_flight to $mmr"
7001 echo "max_mod_rpcs_in_flight set to $mmr"
7004 # create mmr+1 files
7005 echo "creating $((mmr + 1)) files ..."
7007 for i in $(seq $((mmr + 1))); do
7013 # consumes mmr-1 modify RPC slots
7014 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7015 # drop requests on MDT so that RPC slots are consumed
7016 # during all the request resend interval
7017 do_facet $facet "$LCTL set_param fail_loc=0x159"
7018 echo "launch $((mmr - 1)) chmod in parallel ..."
7019 for i in $(seq $((mmr - 1))); do
7020 chmod 0600 $dir/file-$i &
7024 # send one additional modify RPC
7025 do_facet $facet "$LCTL set_param fail_loc=0"
7026 echo "launch 1 additional chmod in parallel ..."
7027 chmod 0600 $dir/file-$mmr &
7030 # check this additional modify RPC get a modify RPC slot
7031 # and succeed its operation
7032 checkstat -vp 0600 $dir/file-$mmr ||
7033 error "Unable to send $mmr modify RPCs in parallel"
7038 # consumes mmr modify RPC slots
7039 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7040 # drop requests on MDT so that RPC slots are consumed
7041 # during all the request resend interval
7042 do_facet $facet "$LCTL set_param fail_loc=0x159"
7043 echo "launch $mmr chmod in parallel ..."
7044 for i in $(seq $mmr); do
7045 chmod 0666 $dir/file-$i &
7049 # send one additional modify RPC
7050 do_facet $facet "$LCTL set_param fail_loc=0"
7051 echo "launch 1 additional chmod in parallel ..."
7052 chmod 0666 $dir/file-$((mmr + 1)) &
7055 # check this additional modify RPC blocked getting a modify RPC slot
7056 checkstat -vp 0644 $dir/file-$((mmr + 1)) ||
7057 error "Unexpectedly send $(($mmr + 1)) modify RPCs in parallel"
7064 [[ $($LCTL get_param mdc.*.import |
7065 grep "connect_flags:.*multi_mod_rpc") ]] ||
7066 skip "Need MDC with 'multi_mod_rpcs' feature"
7068 # check default value
7069 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
7070 check_max_mod_rpcs_in_flight $DIR/$tdir
7074 run_test 90a "check max_mod_rpcs_in_flight is enforced"
7084 [[ $($LCTL get_param mdc.*.import |
7085 grep "connect_flags:.*multi_mod_rpc") ]] ||
7086 skip "Need MDC with 'multi_mod_rpcs' feature"
7089 # update max_mod_rpcs_in_flight
7090 $LFS mkdir -c1 $DIR/${tdir}1 || error "mkdir $DIR/${tdir}1 failed"
7091 check_max_mod_rpcs_in_flight $DIR/${tdir}1 1
7094 # check client is able to send multiple modify RPCs in paralell
7095 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
7096 grep -c "multi_mod_rpcs")
7097 if [ "$tmp" -ne $MDSCOUNT ]; then
7099 skip "Client not able to send multiple modify RPCs in parallel"
7102 # update max_mod_rpcs_in_flight
7103 $LFS mkdir -c1 $DIR/${tdir}2 || error "mkdir $DIR/${tdir}2 failed"
7104 check_max_mod_rpcs_in_flight $DIR/${tdir}2 5
7107 $LFS mkdir -c1 $DIR/${tdir}3 || error "mkdir $DIR/${tdir}3 failed"
7108 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/${tdir}3))
7109 facet="mds$((0x$idx + 1))"
7111 # save MDT max_mod_rpcs_per_client
7112 mmrpc=$(do_facet $facet \
7113 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
7115 # update max_mod_rpcs_in_flight
7116 umount_client $MOUNT
7118 "echo 16 > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
7120 $LCTL set_param mdc.$FSNAME-MDT$idx-mdc-*.max_rpcs_in_flight=17
7121 check_max_mod_rpcs_in_flight $DIR/${tdir}3 16
7123 # restore MDT max_mod_rpcs_per_client initial value
7125 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
7127 rm -rf $DIR/${tdir}?
7130 run_test 90b "check max_mod_rpcs_in_flight is enforced after update"
7133 # get max_rpcs_in_flight value
7134 mrif_90c=$($LCTL get_param -n \
7135 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
7136 echo "max_rpcs_in_flight is $mrif_90c"
7138 # get max_mod_rpcs_in_flight value
7139 mmrif_90c=$($LCTL get_param -n \
7140 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight)
7141 echo "max_mod_rpcs_in_flight is $mmrif_90c"
7143 # get MDT max_mod_rpcs_per_client value
7144 mmrpc_90c=$(do_facet mds1 \
7145 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
7146 echo "max_mod_rpcs_per_client is $mmrpc_90c"
7149 restore_params_90c() {
7152 # restore max_rpcs_in_flight value
7153 do_facet mgs $LCTL set_param -P \
7154 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$mrif_90c
7156 # restore max_mod_rpcs_in_flight value
7157 do_facet mgs $LCTL set_param -P \
7158 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mmrif_90c
7160 # restore MDT max_mod_rpcs_per_client value
7161 do_facet mds1 "echo $mmrpc_90c > \
7162 /sys/module/mdt/parameters/max_mod_rpcs_per_client"
7170 [[ $($LCTL get_param mdc.*.import |
7171 grep "connect_flags:.*multi_mod_rpc") ]] ||
7172 skip "Need MDC with 'multi_mod_rpcs' feature"
7174 # check client is able to send multiple modify RPCs in paralell
7175 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
7176 grep -c "multi_mod_rpcs")
7177 if [ "$tmp" -ne $MDSCOUNT ]; then
7179 skip "Client not able to send multiple modify RPCs in parallel"
7183 stack_trap restore_params_90c
7186 # attempt to set max_mod_rpcs_in_flight to max_rpcs_in_flight value
7187 # prerequisite: set max_mod_rpcs_per_client to max_rpcs_in_flight value
7188 do_facet mds1 "echo $mrif_90c > \
7189 /sys/module/mdt/parameters/max_mod_rpcs_per_client"
7191 # if max_mod_rpcs_in_flight is set to be equal to or larger than
7192 # max_rpcs_in_flight, then max_rpcs_in_flight will be increased
7193 if [[ "$CLIENT_VERSION" -ge $(version_code 2.13.53) ]]; then
7195 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif_90c ||
7196 error "set max_mod_rpcs_in_flight to $mrif_90c failed"
7198 local new_mrif=$($LCTL get_param -n \
7199 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
7200 ((new_mrif == mrif_90c + 1)) ||
7201 error "max_rpcs_in_flight was not increased"
7204 umount_client $MOUNT
7205 do_facet mds1 "echo $mmrpc_90c > \
7206 /sys/module/mdt/parameters/max_mod_rpcs_per_client"
7210 # attempt to set max_mod_rpcs_in_flight to max_mod_rpcs_per_client+1
7211 # prerequisite: set max_rpcs_in_flight to max_mod_rpcs_per_client+2
7213 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$((mmrpc_90c + 2))
7216 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$((mmrpc_90c + 1)) &&
7217 error "set max_mod_rpcs_in_flight to $((mmrpc_90c + 1)) should fail"
7220 # attempt to set max_mod_rpcs_in_flight permanently
7221 do_facet mgs $LCTL set_param -P \
7222 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$mrif_90c
7224 do_facet mgs $LCTL set_param -P \
7225 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif_90c
7227 remount_client $MOUNT
7229 wait_update_facet --verbose client "$LCTL get_param -n \
7230 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight" \
7231 "$((mrif_90c + 1))" ||
7232 error "expected '$((mrif_90c + 1))' for max_rpcs_in_flight"
7234 wait_update_facet --verbose client "$LCTL get_param -n \
7235 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight" \
7237 error "expected '$mrif_90c' for max_mod_rpcs_in_flight"
7242 run_test 90c "check max_mod_rpcs_in_flight update limits"
7253 [[ $($LCTL get_param mdc.*.import |
7254 grep "connect_flags:.*multi_mod_rpc") ]] ||
7255 skip "Need MDC with 'multi_mod_rpcs' feature"
7257 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
7258 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
7259 facet="mds$((0x$idx + 1))"
7261 # check client version supports multislots
7262 tmp=$($LCTL get_param -N \
7263 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
7264 if [ -z "$tmp" ]; then
7266 skip "Client does not support multiple modify RPCs in flight"
7269 # get current value of max_mod_rcps_in_flight
7270 mmr=$($LCTL get_param -n \
7271 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
7272 echo "max_mod_rcps_in_flight is $mmr"
7275 echo "creating $mmr files ..."
7277 for i in $(seq $mmr); do
7278 touch $DIR/$tdir/file-$i
7281 # prepare for close RPC
7282 multiop_bg_pause $DIR/$tdir/file-close O_c
7285 # consumes mmr modify RPC slots
7286 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7287 # drop requests on MDT so that RPC slots are consumed
7288 # during all the request resend interval
7289 do_facet $facet "$LCTL set_param fail_loc=0x159"
7290 echo "launch $mmr chmod in parallel ..."
7291 for i in $(seq $mmr); do
7292 chmod 0600 $DIR/$tdir/file-$i &
7295 # send one additional close RPC
7296 do_facet $facet "$LCTL set_param fail_loc=0"
7297 echo "launch 1 additional close in parallel ..."
7299 cancel_lru_locks mdc
7302 # check this additional close RPC get a modify RPC slot
7303 # and multiop process completed
7304 [ -d /proc/$pid ] &&
7305 error "Unable to send the additional close RPC in parallel"
7310 run_test 90d "check one close RPC is allowed above max_mod_rpcs_in_flight"
7312 check_uuid_on_ost() {
7314 do_facet ost1 "$LCTL get_param obdfilter.${FSNAME}*.exports.'$nid'.uuid"
7317 check_uuid_on_mdt() {
7319 do_facet $SINGLEMDS "$LCTL get_param mdt.${mds1_svc}*.exports.'$nid'.uuid"
7327 [[ "$OST1_VERSION" -ge $(version_code 2.7.63) ]] ||
7328 skip "Need OST version at least 2.7.63"
7329 [[ "$MDS1_VERSION" -ge $(version_code 2.7.63) ]] ||
7330 skip "Need MDT version at least 2.7.63"
7332 start_mds || error "MDS start failed"
7333 start_ost || error "unable to start OST"
7334 mount_client $MOUNT || error "client start failed"
7335 check_mount || error "check_mount failed"
7338 nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g")
7342 uuid=$(get_client_uuid $MOUNT)
7344 echo "list nids on mdt:"
7345 do_facet $SINGLEMDS "$LCTL list_param mdt.${FSNAME}*.exports.*"
7346 echo "uuid from $nid:"
7347 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
7349 found=$(check_uuid_on_mdt $nid | grep $uuid)
7350 [ -z "$found" ] && error "can't find $uuid $nid on MDT"
7351 found=$(check_uuid_on_ost $nid | grep $uuid)
7352 [ -z "$found" ] && error "can't find $uuid $nid on OST"
7354 # umount the client so it won't reconnect
7355 manual_umount_client --force || error "failed to umount $?"
7356 # shouldn't disappear on MDS after forced umount
7357 found=$(check_uuid_on_mdt $nid | grep $uuid)
7358 [ -z "$found" ] && error "can't find $uuid $nid"
7361 do_facet $SINGLEMDS \
7362 "$LCTL set_param -n mdt.${mds1_svc}.evict_client nid:$nid"
7363 sleep 1 # eviction above is async, give it some time to proceed
7365 found=$(check_uuid_on_mdt $nid | grep $uuid)
7366 [ -n "$found" ] && error "found $uuid $nid on MDT"
7367 found=$(check_uuid_on_ost $nid | grep $uuid)
7368 [ -n "$found" ] && error "found $uuid $nid on OST"
7370 # check it didn't reconnect (being umounted)
7371 sleep $((TIMEOUT+1))
7372 found=$(check_uuid_on_mdt $nid | grep $uuid)
7373 [ -n "$found" ] && error "found $uuid $nid on MDT"
7374 found=$(check_uuid_on_ost $nid | grep $uuid)
7375 [ -n "$found" ] && error "found $uuid $nid on OST"
7379 run_test 91 "evict-by-nid support"
7381 generate_ldev_conf() {
7382 # generate an ldev.conf file
7383 local ldevconfpath=$1
7385 local fsldevformat=""
7388 local facets="mgs,$(get_facets OST),$(get_facets MDS)"
7389 for facet in ${facets//,/ }; do
7391 fstype=$(facet_fstype $facet)
7392 [ "$fstype" = zfs ] && fsldevformat="$fstype:"
7393 local host=$(facet_host $facet)
7395 local varfo=${facet}failover_HOST
7396 if [ $facet == mgs ] && combined_mgs_mds; then
7397 varfo=mds1failover_HOST
7399 [ -n "${!varfo}" ] && fo=${!varfo}
7400 local type=$(echo $facet | tr -d "[:digit:]" | \
7401 tr "[:lower:]" "[:upper:]" | sed s/MDS/MDT/ )
7403 [ ${facet} == mgs ] ||
7404 num=$(facet_number $facet)
7405 printf "%s\t%s\t%s-%s%04x\t%s%s\n" \
7412 $(facet_device $facet) >> $ldevconfpath
7415 echo "----- $ldevconfpath -----"
7417 echo "--- END $ldevconfpath ---"
7422 # generate a nids file (mapping between hostname to nid)
7423 # looks like we only have the MGS nid available to us
7424 # so just echo that to a file
7426 echo -e "${mgs_HOST}\t${MGSNID}" > $nidspath
7428 echo "----- $nidspath -----"
7430 echo "--- END $nidspath ---"
7433 compare_ldev_output() {
7434 local ldev_output=$1
7435 local expected_output=$2
7437 sort $expected_output -o $expected_output
7438 sort $ldev_output -o $ldev_output
7440 echo "-- START OF LDEV OUTPUT --"
7442 echo "--- END OF LDEV OUTPUT ---"
7444 echo "-- START OF EXPECTED OUTPUT --"
7445 cat $expected_output
7446 echo "--- END OF EXPECTED OUTPUT ---"
7448 diff $expected_output $ldev_output
7453 if [ -z "$LDEV" ]; then
7454 error "ldev is missing!"
7457 local LDEVCONFPATH=$TMP/ldev.conf
7458 local NIDSPATH=$TMP/nids
7460 echo "Host is $(hostname)"
7462 generate_ldev_conf $LDEVCONFPATH
7463 generate_nids $NIDSPATH
7465 # echo the mgs nid and compare it to environment variable MGSNID
7466 local facets="$(get_facets OST),$(get_facets MDS),mgs"
7467 for facet in ${facets//,/ }; do
7468 local host=$(facet_host $facet)
7469 local output=$($LDEV -c $LDEVCONFPATH -H $host -n $NIDSPATH echo %m)
7471 echo "-- START OF LDEV OUTPUT --"
7473 echo "--- END OF LDEV OUTPUT ---"
7476 error "ldev failed to execute!"
7478 # need to process multiple lines because of
7479 # several targets on host
7480 echo -e $output | awk '{ print $2 }' | while read -r line ; do
7481 [ "$line" = "$MGSNID" ] ||
7482 error "ldev failed mgs nid '$line', \
7486 rm -f $LDEVCONFPATH $NIDSPATH
7488 run_test 92 "ldev returns MGS NID correctly in command substitution"
7491 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs"
7494 #start mgs or mgs/mdt0
7495 if ! combined_mgs_mds ; then
7502 start_ost || error "OST0 start fail"
7504 #define OBD_FAIL_MGS_WRITE_TARGET_DELAY 0x90e
7505 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x8000090e"
7506 for num in $(seq 2 $MDSCOUNT); do
7510 mount_client $MOUNT || error "mount client fails"
7511 wait_osc_import_state mds ost FULL
7512 wait_osc_import_ready client ost
7513 check_mount || error "check_mount failed"
7515 cleanup || error "cleanup failed with $?"
7517 run_test 93 "register mulitple MDT at the same time"
7520 if [ -z "$LDEV" ]; then
7521 error "ldev is missing!"
7524 local LDEVCONFPATH=$TMP/ldev.conf
7525 local NIDSPATH=$TMP/nids
7527 generate_ldev_conf $LDEVCONFPATH
7528 generate_nids $NIDSPATH
7530 local LDEV_OUTPUT=$TMP/ldev-output.txt
7531 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME > $LDEV_OUTPUT
7533 # ldev failed, error
7534 if [ $? -ne 0 ]; then
7535 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7536 error "ldev failed to execute!"
7540 local EXPECTED_OUTPUT=$TMP/ldev-expected.txt
7542 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
7544 for num in $(seq $MDSCOUNT); do
7545 printf "%s-MDT%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7548 for num in $(seq $OSTCOUNT); do
7549 printf "%s-OST%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7552 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7554 if [ $? -ne 0 ]; then
7555 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7556 error "ldev failed to produce the correct hostlist!"
7559 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7561 run_test 94 "ldev outputs correct labels for file system name query"
7564 if [ -z "$LDEV" ]; then
7565 error "ldev is missing!"
7568 local LDEVCONFPATH=$TMP/ldev.conf
7569 local NIDSPATH=$TMP/nids
7571 generate_ldev_conf $LDEVCONFPATH
7572 generate_nids $NIDSPATH
7576 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME &>/dev/null
7577 if [ $? -ne 0 ]; then
7578 rm $LDEVCONFPATH $NIDSPATH
7579 error "ldev label filtering w/ -F failed!"
7583 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -l &>/dev/null
7584 if [ $? -ne 0 ]; then
7585 rm $LDEVCONFPATH $NIDSPATH
7586 error "ldev label filtering w/ -l failed!"
7590 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f &>/dev/null
7591 if [ $? -ne 0 ]; then
7592 rm $LDEVCONFPATH $NIDSPATH
7593 error "ldev label filtering w/ -f failed!"
7597 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a &>/dev/null
7598 if [ $? -ne 0 ]; then
7599 rm $LDEVCONFPATH $NIDSPATH
7600 error "ldev label filtering w/ -a failed!"
7605 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -F $FSNAME &>/dev/null
7606 if [ $? -eq 0 ]; then
7607 rm $LDEVCONFPATH $NIDSPATH
7608 error "ldev label filtering w/ -a and -F incorrectly succeeded"
7612 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -f &>/dev/null
7613 if [ $? -eq 0 ]; then
7614 rm $LDEVCONFPATH $NIDSPATH
7615 error "ldev label filtering w/ -a and -f incorrectly succeeded"
7619 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -l &>/dev/null
7620 if [ $? -eq 0 ]; then
7621 rm $LDEVCONFPATH $NIDSPATH
7622 error "ldev label filtering w/ -a and -l incorrectly succeeded"
7626 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f -l &>/dev/null
7627 if [ $? -eq 0 ]; then
7628 rm $LDEVCONFPATH $NIDSPATH
7629 error "ldev label filtering w/ -f and -l incorrectly succeeded"
7633 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -l &>/dev/null
7634 if [ $? -eq 0 ]; then
7635 rm $LDEVCONFPATH $NIDSPATH
7636 error "ldev label filtering w/ -F and -l incorrectly succeeded"
7639 # file sys & foreign
7640 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -f &>/dev/null
7641 if [ $? -eq 0 ]; then
7642 rm $LDEVCONFPATH $NIDSPATH
7643 error "ldev label filtering w/ -F and -f incorrectly succeeded"
7646 rm $LDEVCONFPATH $NIDSPATH
7648 run_test 95 "ldev should only allow one label filter"
7651 if [ -z "$LDEV" ]; then
7652 error "ldev is missing!"
7655 local LDEVCONFPATH=$TMP/ldev.conf
7656 local NIDSPATH=$TMP/nids
7658 generate_ldev_conf $LDEVCONFPATH
7659 generate_nids $NIDSPATH
7661 local LDEV_OUTPUT=$TMP/ldev-output.txt
7662 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -H $mgs_HOST \
7664 awk '{print $2}' > $LDEV_OUTPUT
7666 # ldev failed, error
7667 if [ $? -ne 0 ]; then
7668 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7669 error "ldev failed to execute!"
7673 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
7675 echo "$mgs_HOST-$(facet_fstype mgs)" > $EXPECTED_OUTPUT
7677 local facets="$(get_facets OST),$(get_facets MDS)"
7678 for facet in ${facets//,/ }; do
7679 local host=$(facet_host $facet)
7680 [ "$mgs_HOST" == "$host" ] &&
7681 echo "$host-$(facet_fstype $facet)" \
7685 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7687 if [ $? -ne 0 ]; then
7688 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7689 error "ldev failed to produce the correct output!"
7692 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7694 run_test 96 "ldev returns hostname and backend fs correctly in command sub"
7697 if [ -z "$LDEV" ]; then
7698 error "ldev is missing!"
7701 local LDEVCONFPATH=$TMP/ldev.conf
7702 local NIDSPATH=$TMP/nids
7704 generate_ldev_conf $LDEVCONFPATH
7705 generate_nids $NIDSPATH
7707 local LDEV_OUTPUT=$TMP/ldev-output.txt
7708 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
7710 echo -e "\nMDT role"
7711 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mdt > $LDEV_OUTPUT
7713 if [ $? -ne 0 ]; then
7714 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7715 error "ldev failed to execute for mdt role!"
7718 for num in $(seq $MDSCOUNT); do
7719 printf "%s-MDT%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7722 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7724 if [ $? -ne 0 ]; then
7725 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7726 error "ldev failed to produce the correct output for mdt role!"
7729 echo -e "\nOST role"
7730 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R ost > $LDEV_OUTPUT
7732 if [ $? -ne 0 ]; then
7733 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
7734 error "ldev failed to execute for ost role!"
7738 for num in $(seq $OSTCOUNT); do
7739 printf "%s-OST%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7742 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7744 if [ $? -ne 0 ]; then
7745 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7746 error "ldev failed to produce the correct output for ost role!"
7749 echo -e "\nMGS role"
7750 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mgs > $LDEV_OUTPUT
7752 if [ $? -ne 0 ]; then
7753 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
7754 error "ldev failed to execute for mgs role!"
7757 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
7759 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7761 if [ $? -ne 0 ]; then
7762 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7763 error "ldev failed to produce the correct output for mgs role!"
7766 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7768 run_test 97 "ldev returns correct ouput when querying based on role"
7773 local temp=$MDS_MOUNT_OPTS
7776 check_mount || error "mount failed"
7777 mountopt="user_xattr"
7778 for ((x = 1; x <= 400; x++)); do
7779 mountopt="$mountopt,user_xattr"
7781 mount_client $MOUNT remount,$mountopt 2>&1 | grep "too long" ||
7782 error "Buffer overflow check failed"
7783 cleanup || error "cleanup failed"
7785 run_test 98 "Buffer-overflow check while parsing mount_opts"
7789 [[ "$ost1_FSTYPE" != ldiskfs ]] &&
7790 skip "ldiskfs only test"
7791 [[ "$OST1_VERSION" -ge $(version_code 2.8.57) ]] ||
7792 skip "Need OST version at least 2.8.57"
7794 local ost_opts="$(mkfs_opts ost1 $(ostdevname 1)) \
7795 --reformat $(ostdevname 1) $(ostvdevname 1)"
7796 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" &&
7797 skip "meta_bg already set"
7800 if [[ ${!opts} != *mkfsoptions* ]]; then
7801 eval opts=\"${!opts} \
7802 --mkfsoptions='\\\"-O ^resize_inode,meta_bg\\\"'\"
7804 local val=${!opts//--mkfsoptions=\\\"/ \
7805 --mkfsoptions=\\\"-O ^resize_inode,meta_bg }
7809 echo "params: $opts"
7812 add ost1 $opts || error "add ost1 failed with new params"
7814 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" ||
7815 error "meta_bg is not set"
7817 run_test 99 "Adding meta_bg option"
7822 start_mgsmds || error "MDS start failed"
7823 start_ost || error "unable to start OST"
7824 mount_client $MOUNT || error "client start failed"
7825 check_mount || error "check_mount failed"
7834 do_facet mgs 'lshowmount -v' | awk 'BEGIN {NR == 0; rc=1} /MGS:/ {rc=0}
7835 END {exit rc}' || error "lshowmount have no output MGS"
7837 do_facet mds1 'lshowmount -v' | awk 'BEGIN {NR == 2; rc=1} /-MDT0000:/
7838 {rc=0} END {exit rc}' || error "lshowmount have no output MDT0"
7840 do_facet ost1 'lshowmount -v' | awk 'BEGIN {NR == 4; rc=1} /-OST0000:/
7841 {rc=0} END {exit rc}' || error "lshowmount have no output OST0"
7843 cleanup || error "cleanup failed with $?"
7845 run_test 100 "check lshowmount lists MGS, MDT, OST and 0@lo"
7848 local createmany_pid
7849 local dev=$FSNAME-OST0000-osc-MDT0000
7853 createmany -o $DIR1/$tdir/$tfile-%d 50000 &
7855 # MDT->OST reconnection causes MDT<->OST last_id synchornisation
7856 # via osp_precreate_cleanup_orphans.
7857 for ((i = 0; i < 100; i++)); do
7858 for ((k = 0; k < 10; k++)); do
7859 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;" \
7860 "$LCTL --device $dev activate"
7863 ls -asl $MOUNT | grep '???' &&
7864 { kill -9 $createmany_pid &>/dev/null;
7865 error "File has no object on OST"; }
7867 kill -s 0 $createmany_pid || break
7869 wait $createmany_pid
7871 unlinkmany $DIR1/$tdir/$tfile-%d 50000
7874 run_test 101a "Race MDT->OST reconnection with create"
7877 local dev=$FSNAME-OST0000-osc-MDT0000
7878 local dir=$DIR1/$tdir
7882 $LFS setstripe -c 1 -i 0 $dir
7883 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;"
7884 #define OBD_FAIL_OSP_CON_EVENT_DELAY 0x2107
7885 do_facet mds1 "$LCTL set_param fail_loc=0x80002107 fail_val=20"
7886 do_facet $SINGLEMDS "$LCTL --device $dev activate;"
7891 wait_osc_import_state client ost1 FULL
7892 touch $dir/$tfile || error "Can't create file"
7896 run_test 101b "Race events DISCONNECT and ACTIVE in osp"
7899 [[ "$MDS1_VERSION" -gt $(version_code 2.9.53) ]] ||
7900 skip "Need server version greater than 2.9.53"
7901 [[ “$(mdsdevname 1)” != “$(mgsdevname)” ]] &&
7902 [[ “$(facet_host mds1)” = “$(facet_host mgs)” ]] &&
7903 skip "MGS must be on different node or combined"
7905 cleanup || error "cleanup failed with $?"
7907 local mds1dev=$(mdsdevname 1)
7908 local mds1mnt=$(facet_mntpt mds1)
7909 local mds1opts=$MDS_MOUNT_OPTS
7911 if [ "$mds1_FSTYPE" == ldiskfs ] &&
7912 ! do_facet mds1 test -b $mds1dev; then
7913 mds1opts=$(csa_add "$mds1opts" -o loop)
7915 if [[ "$mds1_FSTYPE" == zfs ]]; then
7916 import_zpool mds1 || return ${PIPESTATUS[0]}
7919 # unload all and only load libcfs to allow fail_loc setting
7920 do_facet mds1 $LUSTRE_RMMOD || error "unable to unload modules"
7921 do_facet mds1 modprobe libcfs || error "libcfs not loaded"
7922 do_facet mds1 lsmod \| grep libcfs || error "libcfs not loaded"
7924 #define OBD_FAIL_OBDCLASS_MODULE_LOAD 0x60a
7925 do_facet mds1 "$LCTL set_param fail_loc=0x8000060a"
7927 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts &&
7928 error "mdt start must fail"
7929 do_facet mds1 lsmod \| grep obdclass && error "obdclass must not load"
7931 do_facet mds1 "$LCTL set_param fail_loc=0x0"
7933 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts ||
7934 error "mdt start must not fail"
7936 cleanup || error "cleanup failed with $?"
7938 run_test 102 "obdclass module cleanup upon error"
7943 echo "rename $FSNAME to $newname"
7945 if ! combined_mgs_mds ; then
7946 local dev=$(mgsdevname)
7949 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
7950 error "(7) Fail to rename MGS"
7951 if [ "$(facet_fstype mgs)" = "zfs" ]; then
7952 reimport_zpool mgs $newname-mgs
7956 for num in $(seq $MDSCOUNT); do
7957 local dev=$(mdsdevname $num)
7959 do_facet mds${num} \
7960 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
7961 error "(8) Fail to rename MDT $num"
7962 if [ "$(facet_fstype mds${num})" = "zfs" ]; then
7963 reimport_zpool mds${num} $newname-mdt${num}
7967 for num in $(seq $OSTCOUNT); do
7968 local dev=$(ostdevname $num)
7970 do_facet ost${num} \
7971 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
7972 error "(9) Fail to rename OST $num"
7973 if [ "$(facet_fstype ost${num})" = "zfs" ]; then
7974 reimport_zpool ost${num} $newname-ost${num}
7979 test_103_set_pool() {
7983 do_facet mgs $LCTL pool_add $FSNAME.$pname ${FSNAME}-$ost_x ||
7984 error "Fail to add $ost_x to $FSNAME.$pname"
7985 wait_update $HOSTNAME \
7986 "lctl get_param -n lov.$FSNAME-clilov-*.pools.$pname |
7987 grep $ost_x" "$FSNAME-${ost_x}_UUID" ||
7988 error "$ost_x is NOT in pool $FSNAME.$pname"
7991 test_103_check_pool() {
7992 local save_fsname=$1
7995 stat $DIR/$tdir/test-framework.sh ||
7996 error "($errno) Fail to stat"
7997 do_facet mgs $LCTL pool_list $FSNAME.pool1 ||
7998 error "($errno) Fail to list $FSNAME.pool1"
7999 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname ||
8000 error "($errno) Fail to list $FSNAME.$save_fsname"
8001 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname |
8002 grep ${FSNAME}-OST0000 ||
8003 error "($errno) List $FSNAME.$save_fsname is invalid"
8005 local pname=$($LFS getstripe --pool $DIR/$tdir/d0)
8006 [ "$pname" = "$save_fsname" ] ||
8007 error "($errno) Unexpected pool name $pname"
8011 check_mount_and_prep
8013 mkdir $DIR/$tdir || error "(1) Fail to mkdir $DIR/$tdir"
8014 cp $LUSTRE/tests/test-framework.sh $DIR/$tdir ||
8015 error "(2) Fail to copy test-framework.sh"
8017 do_facet mgs $LCTL pool_new $FSNAME.pool1 ||
8018 error "(3) Fail to create $FSNAME.pool1"
8019 # name the pool name as the fsname
8020 do_facet mgs $LCTL pool_new $FSNAME.$FSNAME ||
8021 error "(4) Fail to create $FSNAME.$FSNAME"
8023 test_103_set_pool $FSNAME OST0000
8025 $LFS setstripe -p $FSNAME $DIR/$tdir/d0 ||
8026 error "(6) Fail to setstripe on $DIR/$tdir/d0"
8031 test_renamefs mylustre
8033 local save_fsname=$FSNAME
8037 test_103_check_pool $save_fsname 7
8039 if [ $OSTCOUNT -ge 2 ]; then
8040 test_103_set_pool $save_fsname OST0001
8043 $LFS setstripe -p $save_fsname $DIR/$tdir/f0 ||
8044 error "(16) Fail to setstripe on $DIR/$tdir/f0"
8053 test_103_check_pool $save_fsname 17
8057 test_renamefs $save_fsname
8063 run_test 103 "rename filesystem name"
8065 test_104a() { # LU-6952
8066 local mds_mountopts=$MDS_MOUNT_OPTS
8067 local ost_mountopts=$OST_MOUNT_OPTS
8068 local mds_mountfsopts=$MDS_MOUNT_FS_OPTS
8069 local lctl_ver=$(do_facet $SINGLEMDS $LCTL --version |
8072 [[ $(version_code $lctl_ver) -lt $(version_code 2.9.55) ]] &&
8073 skip "this test needs utils above 2.9.55"
8075 # specify "acl" in mount options used by mkfs.lustre
8076 if [ -z "$MDS_MOUNT_FS_OPTS" ]; then
8077 MDS_MOUNT_FS_OPTS="acl,user_xattr"
8080 MDS_MOUNT_FS_OPTS="${MDS_MOUNT_FS_OPTS},acl,user_xattr"
8083 echo "mountfsopt: $MDS_MOUNT_FS_OPTS"
8085 #reformat/remount the MDT to apply the MDT_MOUNT_FS_OPT options
8087 if ! combined_mgs_mds ; then
8091 if [ -z "$MDS_MOUNT_OPTS" ]; then
8092 MDS_MOUNT_OPTS="-o noacl"
8094 MDS_MOUNT_OPTS="${MDS_MOUNT_OPTS},noacl"
8097 for num in $(seq $MDSCOUNT); do
8098 start mds$num $(mdsdevname $num) $MDS_MOUNT_OPTS ||
8099 error "Failed to start MDS"
8102 for num in $(seq $OSTCOUNT); do
8103 start ost$num $(ostdevname $num) $OST_MOUNT_OPTS ||
8104 error "Failed to start OST"
8108 setfacl -m "d:$RUNAS_ID:rwx" $MOUNT &&
8109 error "ACL is applied when FS is mounted with noacl."
8111 MDS_MOUNT_OPTS=$mds_mountopts
8112 OST_MOUNT_OPTS=$ost_mountopts
8113 MDS_MOUNT_FS_OPTS=$mds_mountfsopts
8115 run_test 104a "Make sure user defined options are reflected in mount"
8117 test_104b() { # LU-12859
8118 mount_client $MOUNT3 flock,localflock
8119 stack_trap "umount_client $MOUNT3" EXIT
8120 mount | grep "$MOUNT3 .*,flock" && error "flock is still set"
8121 mount | grep "$MOUNT3 .*,localflock" || error "localflock is not set"
8122 umount_client $MOUNT3
8123 mount_client $MOUNT3 localflock,flock
8124 mount | grep "$MOUNT3 .*,localflock" && error "localflock is still set"
8125 mount | grep "$MOUNT3 .*,flock" || error "flock is not set"
8126 umount_client $MOUNT3
8127 mount_client $MOUNT3 localflock,flock,noflock
8128 flock_is_enabled $MOUNT3 && error "some flock is still enabled" || true
8130 run_test 104b "Mount uses last flock argument"
8132 error_and_umount() {
8143 mount --bind $DIR $TMP/$tdir || error "mount bind mnt pt failed"
8144 rm -f $TMP/$tdir/$tfile
8145 rm -f $TMP/$tdir/${tfile}1
8147 # Files should not be created in ro bind mount point
8148 # remounting from rw to ro
8149 mount -o remount,ro $TMP/$tdir ||
8150 error_and_umount "readonly remount of bind mnt pt failed"
8151 touch $TMP/$tdir/$tfile &&
8152 error_and_umount "touch succeeds on ro bind mnt pt"
8153 [ -e $TMP/$tdir/$tfile ] &&
8154 error_and_umount "file created on ro bind mnt pt"
8156 # Files should be created in rw bind mount point
8157 # remounting from ro to rw
8158 mount -o remount,rw $TMP/$tdir ||
8159 error_and_umount "read-write remount of bind mnt pt failed"
8160 touch $TMP/$tdir/${tfile}1 ||
8161 error_and_umount "touch fails on rw bind mnt pt"
8162 [ -e $TMP/$tdir/${tfile}1 ] ||
8163 error_and_umount "file not created on rw bind mnt pt"
8164 umount $TMP/$tdir || error "umount of bind mnt pt failed"
8166 cleanup || error "cleanup failed with $?"
8168 run_test 105 "check file creation for ro and rw bind mnt pt"
8175 mkdir -p $DIR/$tdir || error "create $tdir failed"
8176 lfs setstripe -c 1 -i 0 $DIR/$tdir
8177 #define OBD_FAIL_CAT_RECORDS 0x1312
8178 do_facet mds1 $LCTL set_param fail_loc=0x1312 fail_val=$repeat
8180 for ((i = 1; i <= $repeat; i++)); do
8182 #one full plain llog
8183 createmany -o $DIR/$tdir/f- 64768
8185 createmany -u $DIR/$tdir/f- 64768
8187 wait_delete_completed $((TIMEOUT * 7))
8188 #ASSERTION osp_sync_thread() ( thread->t_flags != SVC_RUNNING ) failed
8189 #shows that osp code is buggy
8190 do_facet mds1 $LCTL set_param fail_loc=0 fail_val=0
8194 run_test 106 "check osp llog processing when catalog is wrapped"
8197 [[ "$MDS1_VERSION" -ge $(version_code 2.10.50) ]] ||
8198 skip "Need MDS version > 2.10.50"
8201 start_mgsmds || error "start_mgsmds failed"
8202 start_ost || error "unable to start OST"
8204 # add unknown configuration parameter.
8205 if [[ $PERM_CMD == *"set_param -P"* ]]; then
8206 cmd="$PERM_CMD ost.$FSNAME-OST0000*.unknown_param"
8208 cmd="$PERM_CMD $FSNAME-OST0000*.ost.unknown_param"
8210 do_facet mgs "$cmd=50"
8211 cleanup_nocli || error "cleanup_nocli failed with $?"
8214 # unknown param should be ignored while mounting.
8215 start_ost || error "unable to start OST after unknown param set"
8217 cleanup || error "cleanup failed with $?"
8219 run_test 107 "Unknown config param should not fail target mounting"
8224 $rcmd rm -rf $tmp > /dev/null 2>&1
8225 $rcmd mkdir -p $tmp/{mnt,images} || error "failed to mkdir remotely"
8227 for facet in $facets; do
8228 [ "$mds1_FSTYPE" = zfs ] &&
8229 $rcmd $ZPOOL -f export lustre-$facet > /dev/null 2>&1
8230 $rcmd mkdir $tmp/mnt/$facet ||
8231 error "failed to mkdir $tmp/mnt/$facet"
8232 $rcmd dd if=/dev/zero of=$tmp/images/$facet \
8233 seek=199 bs=1M count=1 ||
8234 error "failed to create $tmp/images/$facet"
8243 local facet=${role}$((idx + 1))
8245 [ $# -eq 5 ] && pool=$5
8247 do_facet $SINGLEMDS $MKFS --fsname=lustre --$mgs \
8248 --$role --index=$idx --replace --backfstype=$bkfs \
8249 --device-size=200000 --reformat $pool $tmp/images/$facet ||
8250 error "failed to mkfs for $facet"
8254 echo "mounting client..."
8255 mount -t lustre ${nid}:/lustre $MOUNT ||
8256 error "failed to mount lustre"
8259 ls -l $MOUNT/local_dir || error "failed to list"
8261 echo "check truncate && write"
8262 echo "dummmmmmmmmmmmm" > $MOUNT/remote_dir/fsx.c ||
8263 error "failed to tuncate & write"
8266 touch $MOUNT/foooo ||
8267 error "failed to create"
8269 echo "check read && write && append"
8270 sha1sum $MOUNT/conf-sanity.sh |
8271 awk '{ print $1 }' > $MOUNT/checksum.new ||
8272 error "failed to read(1)"
8273 sha1sum $MOUNT/remote_dir/unlinkmany.c |
8274 awk '{ print $1 }' >> $MOUNT/checksum.new ||
8275 error "failed to read(2)"
8276 sha1sum $MOUNT/striped_dir/lockahead_test.o |
8277 awk '{ print $1 }' >> $MOUNT/checksum.new ||
8278 error "failed to read(3)"
8281 diff $MOUNT/checksum.new $MOUNT/checksum.src ||
8282 error "failed to verify data"
8292 umount -f $MOUNT || error "failed to umount client"
8293 for facet in $facets; do
8294 $rcmd umount -f $tmp/mnt/$facet ||
8295 error "failed to umount $facet"
8296 if [ "$mds1_FSTYPE" = zfs ]; then
8297 $rcmd $ZPOOL export -f lustre-$facet ||
8298 error "failed to export lustre-$facet"
8302 $rcmd rm -rf $tmp || error "failed to rm the dir $tmp"
8306 [ "$CLIENTONLY" ] && skip "Client-only testing"
8307 [ "$mds1_FSTYPE" != zfs ] && skip "zfs only test"
8308 [ "$MDS1_VERSION" -lt $(version_code 2.10.58) ] &&
8309 skip "Need server version at least 2.10.58"
8314 local tmp=$TMP/$tdir
8315 local rcmd="do_facet $SINGLEMDS"
8316 local facets="mdt1 mdt2 ost1 ost2"
8317 local nid=$($rcmd $LCTL list_nids | head -1)
8320 trap t_108_cleanup EXIT ERR
8323 t_108_mkfs mdt 0 zfs mgs lustre-mdt1/mdt1
8324 t_108_mkfs mdt 1 zfs mgsnode=$nid lustre-mdt2/mdt2
8325 t_108_mkfs ost 0 zfs mgsnode=$nid lustre-ost1/ost1
8326 t_108_mkfs ost 1 zfs mgsnode=$nid lustre-ost2/ost2
8328 for facet in $facets; do
8329 $rcmd zfs set mountpoint=$tmp/mnt/$facet canmount=on \
8330 lustre-$facet/$facet ||
8331 error "failed to zfs set for $facet (1)"
8332 $rcmd zfs mount lustre-$facet/$facet ||
8333 error "failed to local mount $facet"
8334 $rcmd tar jxf $LUSTRE/tests/ldiskfs_${facet}_2_11.tar.bz2 \
8335 --xattrs --xattrs-include="trusted.*" \
8336 -C $tmp/mnt/$facet/ > /dev/null 2>&1 ||
8337 error "failed to untar image for $facet"
8338 $rcmd "cd $tmp/mnt/$facet && rm -rf oi.* OI_* lfsck_* LFSCK" ||
8339 error "failed to cleanup for $facet"
8340 $rcmd zfs umount lustre-$facet/$facet ||
8341 error "failed to local umount $facet"
8342 $rcmd zfs set canmount=off lustre-$facet/$facet ||
8343 error "failed to zfs set $facet (2)"
8346 echo "changing server nid..."
8347 $rcmd mount -t lustre -o nosvc lustre-mdt1/mdt1 $tmp/mnt/mdt1
8348 $rcmd lctl replace_nids $FSNAME-MDT0000 $nid
8349 $rcmd lctl replace_nids $FSNAME-MDT0001 $nid
8350 $rcmd lctl replace_nids $FSNAME-OST0000 $nid
8351 $rcmd lctl replace_nids $FSNAME-OST0001 $nid
8352 $rcmd umount $tmp/mnt/mdt1
8354 for facet in $facets; do
8355 echo "mounting $facet from backup..."
8356 $rcmd mount -t lustre -o abort_recov lustre-$facet/$facet \
8357 $tmp/mnt/$facet || error "failed to mount $facet"
8360 # ZFS backend can detect migration and trigger OI scrub automatically
8361 # sleep 3 seconds for scrub done
8367 run_test 108a "migrate from ldiskfs to ZFS"
8370 [ "$CLIENTONLY" ] && skip "Client-only testing"
8371 [ "$mds1_FSTYPE" != ldiskfs ] && skip "ldiskfs only test"
8372 [ "$MDS1_VERSION" -lt $(version_code 2.10.58) ] &&
8373 skip "Need server version at least 2.10.58"
8378 local tmp=$TMP/$tdir
8379 local rcmd="do_facet $SINGLEMDS"
8380 local facets="mdt1 mdt2 ost1 ost2"
8381 local scrub_list="MDT0000 MDT0001 OST0000 OST0001"
8382 local nid=$($rcmd $LCTL list_nids | head -1)
8385 trap t_108_cleanup EXIT ERR
8388 t_108_mkfs mdt 0 ldiskfs mgs
8389 t_108_mkfs mdt 1 ldiskfs mgsnode=$nid
8390 t_108_mkfs ost 0 ldiskfs mgsnode=$nid
8391 t_108_mkfs ost 1 ldiskfs mgsnode=$nid
8393 for facet in $facets; do
8394 $rcmd mount -t ldiskfs -o loop $tmp/images/$facet \
8396 error "failed to local mount $facet"
8398 $rcmd tar jxf $LUSTRE/tests/zfs_${facet}_2_11.tar.bz2 \
8399 --xattrs --xattrs-include="*.*" \
8400 -C $tmp/mnt/$facet/ > /dev/null 2>&1 ||
8401 error "failed to untar image for $facet"
8402 $rcmd "cd $tmp/mnt/$facet && rm -rf oi.* OI_* lfsck_* LFSCK" ||
8403 error "failed to cleanup for $facet"
8404 $rcmd umount $tmp/mnt/$facet ||
8405 error "failed to local umount $facet"
8408 echo "changing server nid..."
8409 $rcmd mount -t lustre -o nosvc,loop $tmp/images/mdt1 $tmp/mnt/mdt1
8410 $rcmd lctl replace_nids $FSNAME-MDT0000 $nid
8411 $rcmd lctl replace_nids $FSNAME-MDT0001 $nid
8412 $rcmd lctl replace_nids $FSNAME-OST0000 $nid
8413 $rcmd lctl replace_nids $FSNAME-OST0001 $nid
8414 $rcmd umount $tmp/mnt/mdt1
8416 for facet in $facets; do
8417 echo "mounting $facet from backup..."
8418 $rcmd mount -t lustre -o loop,abort_recov $tmp/images/$facet \
8419 $tmp/mnt/$facet || error "failed to mount $facet"
8422 for facet in $scrub_list; do
8423 $rcmd $LCTL lfsck_start -M $FSNAME-$facet -t scrub ||
8424 error "failed to start OI scrub on $facet"
8427 # sleep 3 seconds for scrub done
8433 run_test 108b "migrate from ZFS to ldiskfs"
8436 # set number of permanent parameters
8438 test_109_set_params() {
8441 set_persistent_param_and_check mds \
8442 "mdd.$fsname-MDT0000.atime_diff" \
8443 "$fsname-MDT0000.mdd.atime_diff" \
8445 set_persistent_param_and_check mds \
8446 "mdd.$fsname-MDT0000.atime_diff" \
8447 "$fsname-MDT0000.mdd.atime_diff" \
8449 set_persistent_param_and_check client \
8450 "llite.$fsname*.max_read_ahead_mb" \
8451 "$fsname.llite.max_read_ahead_mb" \
8453 set_persistent_param_and_check client \
8454 "llite.$fsname*.max_read_ahead_mb" \
8455 "$fsname.llite.max_read_ahead_mb" \
8457 create_pool $fsname.pool1 || error "create pool failed"
8458 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
8459 error "pool_add failed"
8460 do_facet mgs $LCTL pool_remove $fsname.pool1 OST0000 ||
8461 error "pool_remove failed"
8462 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
8463 error "pool_add failed"
8467 # check permanent parameters
8469 test_109_test_params() {
8472 local atime_diff=$(do_facet mds $LCTL \
8473 get_param -n mdd.$fsname-MDT0000.atime_diff)
8474 [ $atime_diff == 63 ] || error "wrong mdd parameter after clear_conf"
8475 local max_read_ahead_mb=$(do_facet client $LCTL \
8476 get_param -n llite.$fsname*.max_read_ahead_mb)
8477 [ $max_read_ahead_mb == 64 ] ||
8478 error "wrong llite parameter after clear_conf"
8479 local ost_in_pool=$(do_facet mds $LCTL pool_list $fsname.pool1 |
8480 grep -v "^Pool:" | sed 's/_UUID//')
8481 [ $ost_in_pool = "$fsname-OST0000" ] ||
8482 error "wrong pool after clear_conf"
8486 # run lctl clear_conf, store CONFIGS before and after that
8488 test_109_clear_conf()
8490 local clear_conf_arg=$1
8493 if ! combined_mgs_mds ; then
8494 mgsdev=$(mgsdevname)
8495 stop_mgs || error "stop_mgs failed"
8496 start_mgs "-o nosvc" || error "start_mgs nosvc failed"
8498 mgsdev=$(mdsdevname 1)
8499 start_mdt 1 "-o nosvc" || error "start_mdt 1 nosvc failed"
8502 do_facet mgs "rm -rf $TMP/${tdir}/conf1; mkdir -p $TMP/${tdir}/conf1;" \
8503 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf1\\\" \
8507 # the command being tested
8509 do_facet mgs $LCTL clear_conf $clear_conf_arg ||
8510 error "clear_conf failed"
8511 if ! combined_mgs_mds ; then
8512 stop_mgs || error "stop_mgs failed"
8514 stop_mdt 1 || error "stop_mdt 1 failed"
8517 do_facet mgs "rm -rf $TMP/${tdir}/conf2; mkdir -p $TMP/${tdir}/conf2;" \
8518 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf2\\\" \
8522 test_109_file_shortened() {
8524 local sizes=($(do_facet mgs "stat -c %s " \
8525 "$TMP/${tdir}/conf1/CONFIGS/$file" \
8526 "$TMP/${tdir}/conf2/CONFIGS/$file"))
8527 [ ${sizes[1]} -lt ${sizes[0]} ] && return 0
8533 [ $MDS1_VERSION -lt $(version_code 2.10.59) ] &&
8534 skip "Needs MDS version 2.10.59 or later."
8535 [ "$(facet_fstype mgs)" == zfs ] &&
8536 skip "LU-8727: no implementation for ZFS"
8541 client_up || error "client_up failed"
8544 # set number of permanent parameters
8546 test_109_set_params $FSNAME
8548 umount_client $MOUNT || error "umount_client failed"
8549 stop_ost || error "stop_ost failed"
8550 stop_mds || error "stop_mds failed"
8552 test_109_clear_conf $FSNAME
8554 # make sure that all configs are cleared
8556 test_109_file_shortened $FSNAME-MDT0000 ||
8557 error "failed to clear MDT0000 config"
8558 test_109_file_shortened $FSNAME-client ||
8559 error "failed to clear client config"
8564 # check that configurations are intact
8566 test_109_test_params $FSNAME
8571 destroy_test_pools || error "destroy test pools failed"
8575 run_test 109a "test lctl clear_conf fsname"
8579 [ $MDS1_VERSION -lt $(version_code 2.10.59) ] &&
8580 skip "Needs MDS version 2.10.59 or later."
8581 [ "$(facet_fstype mgs)" == zfs ] &&
8582 skip "LU-8727: no implementation for ZFS"
8587 client_up || error "client_up failed"
8590 # set number of permanent parameters
8592 test_109_set_params $FSNAME
8594 umount_client $MOUNT || error "umount_client failed"
8595 stop_ost || error "stop_ost failed"
8596 stop_mds || error "stop_mds failed"
8598 test_109_clear_conf $FSNAME-MDT0000
8600 # make sure that only one config is cleared
8602 test_109_file_shortened $FSNAME-MDT0000 ||
8603 error "failed to clear MDT0000 config"
8604 test_109_file_shortened $FSNAME-client &&
8605 error "failed to clear client config"
8609 # check that configurations are intact
8611 test_109_test_params $FSNAME
8616 destroy_test_pools || error "destroy test pools failed"
8620 run_test 109b "test lctl clear_conf one config"
8624 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
8625 skip "Only applicable to ldiskfs-based MDTs"
8627 do_facet $SINGLEMDS $DEBUGFS -w -R supported_features |grep large_dir ||
8628 skip "large_dir option is not supported on MDS"
8629 do_facet ost1 $DEBUGFS -w -R supported_features | grep large_dir ||
8630 skip "large_dir option is not supported on OSS"
8632 stopall # stop all targets before modifying the target counts
8633 local old_mdscount=$MDSCOUNT
8634 local old_ostcount=$OSTCOUNT
8636 stack_trap "MDSCOUNT=$old_mdscount OSTCOUNT=$old_ostcount" EXIT
8640 # ext4_dir_entry_2 struct size:264
8641 # dx_root struct size:8
8642 # dx_node struct size:8
8643 # dx_entry struct size:8
8644 # For 1024 bytes block size.
8645 # First level directory entries: 126
8646 # Second level directory entries: 127
8647 # Entries in leaf: 3
8648 # For 2 levels limit: 48006
8649 # For 3 levels limit : 6096762
8650 # Create 80000 files to safely exceed 2-level htree limit.
8651 CONF_SANITY_110_LINKS=${CONF_SANITY_110_LINKS:-80000}
8653 # can fit at most 3 filenames per 1KB leaf block, but each
8654 # leaf/index block will only be 3/4 full before split at each level
8655 (( MDSSIZE < CONF_SANITY_110_LINKS / 3 * 4/3 * 4/3 )) &&
8656 CONF_SANITY_110_LINKS=$((MDSSIZE * 3 * 3/4 * 3/4))
8658 combined_mgs_mds || replace=" --replace "
8659 local opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
8660 $replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
8661 if [[ $opts != *large_dir* ]]; then
8662 if [[ $opts != *mkfsoptions* ]]; then
8663 opts+=" --mkfsoptions=\\\"-O large_dir -b 1024 -i 65536\\\""
8665 opts="${opts//--mkfsoptions=\\\"/ \
8666 --mkfsoptions=\\\"-O large_dir -b 1024 -i 65536 }"
8669 echo "MDT params: $opts"
8671 combined_mgs_mds || start_mgs
8672 add mds1 $opts || error "add mds1 failed with new params"
8673 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
8675 opts="$(mkfs_opts ost1 $(ostdevname 1)) \
8676 $replace --reformat $(ostdevname 1) $(ostvdevname 1)"
8678 if [[ $opts != *large_dir* ]]; then
8679 if [[ $opts != *mkfsoptions* ]]; then
8680 opts+=" --mkfsoptions=\\\"-O large_dir\\\" "
8682 opts="${opts//--mkfsoptions=\\\"/ \
8683 --mkfsoptions=\\\"-O large_dir }"
8686 echo "OST params: $opts"
8687 add ost1 $opts || error "add ost1 failed with new params"
8688 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
8690 MOUNT_2=yes mountcli || error "mount clients failed"
8692 mkdir -v $DIR/$tdir || error "cannot create $DIR/$tdir"
8693 local pids count=0 group=0
8695 echo "creating $CONF_SANITY_110_LINKS in total"
8696 while (( count < CONF_SANITY_110_LINKS )); do
8697 local len=$((253 - $(wc -c <<<"$tfile-$group-40000-")))
8698 local dir=DIR$((group % 2 + 1))
8699 local target=${!dir}/$tdir/$tfile-$group
8700 local long=$target-$(generate_name $len)-
8701 local create=$((CONF_SANITY_110_LINKS - count))
8703 (( create > 40000 )) && create=40000
8704 touch $target || error "creating $target failed"
8705 echo "creating $create hard links to $target"
8706 createmany -l $target $long $create &
8709 count=$((count + create))
8710 group=$((group + 1))
8712 echo "waiting for PIDs$pids to complete"
8713 wait $pids || error "createmany failed after $group groups"
8715 umount_client $MOUNT2 -f
8718 run_e2fsck $(facet_active_host mds1) $(mdsdevname 1) -n
8719 MDSCOUNT=$old_mdscount
8720 OSTCOUNT=$old_ostcount
8722 run_test 110 "Adding large_dir with 3-level htree"
8725 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
8726 skip "Only applicable to ldiskfs-based MDTs"
8728 is_dm_flakey_dev $SINGLEMDS $(mdsdevname 1) &&
8729 skip "This test can not be executed on flakey dev"
8731 do_facet $SINGLEMDS $DEBUGFS -w -R supported_features |grep large_dir ||
8732 skip "large_dir option is not supported on MDS"
8734 do_facet ost1 $DEBUGFS -w -R supported_features | grep large_dir ||
8735 skip "large_dir option is not supported on OSS"
8737 # cleanup before changing target counts
8739 local old_mdscount=$MDSCOUNT
8740 local old_ostcount=$OSTCOUNT
8741 local old_mdssize=$MDSSIZE
8743 stack_trap "MDSSIZE=$MDSSIZE MDSCOUNT=$MDSCOUNT OSTCOUNT=$OSTCOUNT" EXIT
8746 (( MDSSIZE < 2400000 )) && MDSSIZE=2400000 # need at least 2.4GB
8748 local mdsdev=$(mdsdevname 1)
8749 combined_mgs_mds || replace=" --replace "
8750 local opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
8751 $replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
8752 if [[ $opts != *large_dir* ]]; then
8753 if [[ $opts != *mkfsoptions* ]]; then
8754 opts+=" --mkfsoptions=\\\"-O large_dir -i 1048576 \\\" "
8756 opts="${opts//--mkfsoptions=\\\"/ \
8757 --mkfsoptions=\\\"-O large_dir -i 1048576 }"
8760 echo "MDT params: $opts"
8762 combined_mgs_mds || start_mgs
8763 __touch_device mds 1
8764 add mds1 $opts || error "add mds1 failed with new params"
8765 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
8767 opts="$(mkfs_opts ost1 $(ostdevname 1)) \
8768 $replace --reformat $(ostdevname 1) $(ostvdevname 1)"
8769 if [[ $opts != *large_dir* ]]; then
8770 if [[ $opts != *mkfsoptions* ]]; then
8771 opts+=" --mkfsoptions=\\\"-O large_dir \\\""
8773 opts="${opts//--mkfsoptions=\\\"/ --mkfsoptions=\\\"-O large_dir }"
8776 echo "OST params: $opts"
8777 __touch_device ost 1
8778 add ost1 $opts || error "add ost1 failed with new params"
8779 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
8781 MOUNT_2=yes mountcli
8782 mkdir $DIR/$tdir || error "cannot create $DIR/$tdir"
8784 lfs df -i $DIR/$tdir
8788 local start=$SECONDS
8790 local dirmax=$((2 << 30))
8796 while (( !needskip & dirsize < dirmax )); do
8799 for cli in ${CLIENTS//,/ }; do
8800 local len=$((253 - $(wc -c <<<"$cli-$group-60000-")))
8801 local target=$cli-$group
8802 local long=$DIR/$tdir/$target-$(generate_name $len)-
8804 RPWD=$DIR/$tdir do_node $cli touch $target ||
8805 error "creating $target failed"
8806 echo "creating 60000 hardlinks to $target"
8807 RPWD=$DIR/$tdir do_node $cli createmany -l $target $long 60000 &
8810 group=$((group + 1))
8812 long=$DIR2/$tdir/$target-$(generate_name $len)-
8814 RPWD=$DIR2/$tdir do_node $cli touch $target ||
8815 error "creating $target failed"
8816 echo "creating 60000 hardlinks to $target"
8817 RPWD=$DIR2/$tdir do_node $cli createmany -l $target $long 60000 &
8820 group=$((group + 1))
8822 echo "waiting for PIDs$pids to complete"
8823 wait $pids || error "createmany failed after $group groups"
8824 dirsize=$(stat -c %s $DIR/$tdir)
8825 taken=$((SECONDS - start))
8826 rate=$((dirsize / taken))
8827 left=$(((dirmax - dirsize) / rate))
8828 num=$((group * 60000))
8829 echo "estimate ${left}s left after $num files / ${taken}s"
8830 # if the estimated time remaining is too large (it may change
8831 # over time as the create rate is not constant) then exit
8832 # without declaring a failure.
8833 (( left > 1200 )) && needskip=1
8836 umount_client $MOUNT2 -f
8839 (( $needskip )) && skip "ETA ${left}s after $num files / ${taken}s is too long"
8841 run_e2fsck $(facet_active_host mds1) $(mdsdevname 1) -n
8842 MDSCOUNT=$old_mdscount
8843 OSTCOUNT=$old_ostcount
8844 MDSSIZE=$old_mdssize
8846 run_test 111 "Adding large_dir with over 2GB directory"
8849 start_mds || error "MDS start failed"
8850 start_ost || error "OSS start failed"
8851 echo "start ost2 service on $(facet_active_host ost2)"
8852 start ost2 $(ostdevname 2) $(csa_add "$OST_MOUNT_OPTS" -o no_precreate) ||
8853 error "start ost2 facet failed"
8854 local val=$(do_facet ost2 \
8855 "$LCTL get_param -n obdfilter.$FSNAME-OST0001*.no_precreate")
8856 (( $val == 1 )) || error "obdfilter.$FSNAME-OST0001*.no_precreate=$val"
8858 mount_client $MOUNT || error "mount client failed"
8859 wait_osc_import_state mds1 ost1 FULL
8860 wait_osc_import_state client ost1 FULL
8861 wait_osc_import_state client ost2 FULL
8863 $LFS setstripe -i 0 $DIR/$tfile.0 ||
8864 error "problem creating $tfile.0 on OST0000"
8865 $LFS setstripe -i 1 $DIR/$tfile.1 && $LFS getstripe $DIR/$tfile.1 &&
8866 (( $($LFS getstripe -i $DIR/$tfile.1) == 1 )) &&
8867 error "allowed to create $tfile.1 on OST0001"
8868 do_facet ost2 $LCTL set_param obdfilter.*.no_precreate=0
8870 $LFS setstripe -i 1 $DIR/$tfile.2 ||
8871 error "failed to create $tfile.2 on ost1 facet"
8872 # files not cleaned with ONLY_REPEAT because of client unmount below
8873 rm -f $DIR/$tfile.[012]
8874 stop_ost2 || error "stop ost2 facet failed"
8877 run_test 112 "mount OST with nocreate option"
8883 rm -f $TMP/$tdir/lustre-mdt
8887 if [ "$mds1_FSTYPE" != ldiskfs ]; then
8888 skip "Only applicable to ldiskfs-based MDTs"
8890 [ -n "$FILESET" ] && skip "Not functional for FILESET set"
8892 local dbfs_ver=$(do_facet $SINGLEMDS $DEBUGFS -V 2>&1)
8894 echo "debugfs version: $dbfs_ver"
8895 echo "$dbfs_ver" | egrep -w "1.44.3.wc1|1.44.5.wc1|1.45.2.wc1" &&
8896 skip_env "This version of debugfs doesn't show inode number"
8898 IMAGESIZE=$((3072 << 30)) # 3072 GiB
8902 local saved_flakey=${FLAKEY}
8903 stack_trap "FLAKEY=$saved_flakey" EXIT
8910 do_facet mds1 "hostname; lctl dl; mount"
8912 do_facet ost1 "hostname; lctl dl; mount"
8913 # We need MDT size 3072GB, because it is smallest
8914 # partition that can store 2B inodes
8915 do_facet $SINGLEMDS "mkdir -p $TMP/$tdir"
8916 local mdsimgname=$TMP/$tdir/lustre-mdt
8917 do_facet $SINGLEMDS "rm -f $mdsimgname"
8918 do_facet $SINGLEMDS "touch $mdsimgname"
8919 trap cleanup_115 RETURN EXIT
8920 do_facet $SINGLEMDS "$TRUNCATE $mdsimgname $IMAGESIZE" ||
8921 skip "Backend FS doesn't support sparse files"
8922 local mdsdev=$(do_facet $SINGLEMDS "losetup -f")
8923 do_facet $SINGLEMDS "losetup $mdsdev $mdsimgname"
8925 local mds_opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
8926 --mkfsoptions='-O ea_inode,^resize_inode,meta_bg \
8927 -N 2247484000 -E lazy_itable_init' --device-size=$IMAGESIZE"
8928 add mds1 $mds_opts --mgs --reformat $mdsdev ||
8929 skip_env "format large MDT failed"
8930 opts="$(mkfs_opts ost1 $(ostdevname 1)) \
8931 $replace --reformat $(ostdevname 1) $(ostvdevname 1)"
8932 add ost1 $opts || error "add ost1 failed with new params"
8933 start $SINGLEMDS $mdsdev $MDS_MOUNT_OPTS || error "start MDS failed"
8934 start_ost || error "start OSS failed"
8935 mount_client $MOUNT || error "mount client failed"
8937 mkdir -p $DIR/$tdir || error "mkdir $DIR/$tdir fail"
8938 goal="/sys/fs/ldiskfs/$(basename $mdsdev)/inode_goal"
8940 # 2147483648 is 0x80000000
8941 do_facet $SINGLEMDS "echo 2147483648 >> $goal; grep . $goal"
8942 touch $DIR/$tdir/$tfile
8944 # attrs from 1 to 16 go to block, 17th - to inode
8945 for i in {1..17}; do
8946 local nm="trusted.ea$i"
8947 setfattr -n $nm -v $(printf "xattr%0250d" $i) $DIR/$tdir/$tfile
8950 do_facet $SINGLEMDS $DEBUGFS -c -R "stat ROOT/$tdir/$tfile" $mdsdev
8952 # inode <2147483649> trusted.ea16 (255)
8953 local inode_num=$(do_facet $SINGLEMDS \
8954 "$DEBUGFS -c -R 'stat ROOT/$tdir/$tfile' $mdsdev" |
8955 awk '/ea17/ { print $2 }' |
8956 sed -e 's/>//' -e 's/<//' -e 's/\"//')
8957 echo "inode num: $inode_num"
8958 [ $inode_num -ge 2147483648 ] || error "inode $inode_num too small"
8959 do_facet $SINGLEMDS "losetup -d $mdsdev"
8962 run_test 115 "Access large xattr with inodes number over 2TB"
8965 [ "$mds1_FSTYPE" != ldiskfs ] && skip "ldiskfs only test"
8966 [ "$MDS1_VERSION" -lt $(version_code 2.10.59) ] &&
8967 skip "Need server version at least 2.10.59"
8968 do_facet $SINGLEMDS which mkfs.xfs ||
8969 skip_env "No mkfs.xfs installed"
8974 local tmpmnt=$TMP/$tdir
8975 local mdtimg=$tfile-mdt0
8977 do_facet $SINGLEMDS mkdir -p $tmpmnt
8978 stack_trap "do_facet $SINGLEMDS rmdir $tmpmnt" EXIT
8980 do_facet $SINGLEMDS touch $TMP/$mdtimg
8981 stack_trap "do_facet $SINGLEMDS rm -f $TMP/$mdtimg" EXIT
8982 do_facet $SINGLEMDS mkfs -t xfs -d file,size=1t,name=$TMP/$mdtimg ||
8983 error "mkfs temporary xfs image"
8985 do_facet $SINGLEMDS mount $TMP/$mdtimg $tmpmnt ||
8986 error "mount temporary xfs image"
8987 stack_trap "do_facet $SINGLEMDS umount $tmpmnt" EXIT
8988 local old_mdssize=$MDSSIZE
8989 local old_mdsisize=$MDSISIZE
8991 MDSSIZE=$((17 * 1024 * 1024 * 1024)) # 17T MDT
8992 MDSISIZE=$((16 << 20))
8993 local opts17t="$(mkfs_opts $SINGLEMDS)"
8995 MDSSIZE=$old_mdssize
8996 MDSISIZE=$old_mdsisize
8998 do_facet $SINGLEMDS $MKFS \
8999 ${opts17t/-E /-E lazy_itable_init,lazy_journal_init,} \
9001 error "failed to mkfs for $tmpmnt/$mdtimg"
9003 do_facet $SINGLEMDS $TUNE2FS -l $tmpmnt/$mdtimg |
9004 grep -qw 'features.*extent' || error "extent should be enabled"
9007 run_test 116 "big size MDT support"
9011 do_facet ost1 "$LCTL set_param ost.OSS.ost_io.nrs_policies=fifo"
9012 do_facet ost1 "$LCTL get_param -n ost.OSS.ost_io.nrs_tbf_rule" &&
9013 error "get_param should fail"
9014 cleanup || error "cleanup failed with rc $?"
9016 run_test 117 "lctl get_param return errors properly"
9022 [[ "$MDSCOUNT" -ge 2 ]] || skip "Need more at least 2 MDTs"
9024 had_config=$(do_facet mds1 "$LCTL get_param debug | grep config")
9025 do_facet mds1 "$LCTL set_param debug=+config"
9026 do_facet mds1 "$LCTL clear"
9029 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)" &>/dev/null
9030 # mount after writeconf will make "add osp" added to mdt0 config:
9031 # 53 (224)marker 60 (flags=0x01, v2.5.1.0) lustre-MDT0001 'add osp'
9032 # 54 (080)add_uuid nid=... 0: 1:...
9033 # 55 (144)attach 0:lustre-MDT0001-osp-MDT0000 1:osp 2:...
9034 # 56 (144)setup 0:lustre-MDT0001-osp-MDT0000 1:... 2:...
9035 # 57 (136)modify_mdc_tgts add 0:lustre-MDT0000-mdtlov 1:... 2:1 3:1
9036 # duplicate modify_mdc_tgts caused crashes
9039 # using larger debug_mb size to avoid lctl dk log truncation
9040 size_mb=$((DEBUG_SIZE_SAVED * 4))
9043 # though config processing stops after failed attach and setup
9044 # it will proceed after the failed command after each writeconf
9045 # this is the original scenario of the issue
9046 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)" &>/dev/null
9047 do_facet mds1 "$LCTL set_param debug_mb=$size_mb"
9050 wait_update_facet_cond mds1 \
9051 "$LCTL dk | grep -c Processed.log.$FSNAME-MDT0000" \
9056 [[ -z "$had_config" ]] && do_facet mds1 lctl set_param debug=-config
9060 run_test 119 "writeconf on slave mdt shouldn't duplicate mdc/osp and crash"
9062 test_120() { # LU-11130
9063 [ "$MDSCOUNT" -lt 2 ] && skip "mdt count < 2"
9064 [ "$mds1_FSTYPE" != ldiskfs ] &&
9065 skip "ldiskfs only test"
9066 [ "$MDS1_VERSION" -lt $(version_code 2.11.56) ] &&
9067 skip "Need DNE2 capable MD target with LU-11130 fix"
9071 local mds1host=$(facet_active_host mds1)
9072 local mds1dev=$(mdsdevname 1)
9074 $LFS mkdir -i 1 $DIR/$tdir
9075 $LFS mkdir -i 0 $DIR/$tdir/mds1dir
9077 ln -s foo $DIR/$tdir/bar
9078 mv $DIR/$tdir/bar $DIR/$tdir/mds1dir/bar2 ||
9079 error "cross-target rename failed"
9083 run_e2fsck $mds1host $mds1dev "-n"
9085 run_test 120 "cross-target rename should not create bad symlinks"
9089 start_mgsmds || error "MGS MDS Start failed"
9091 stop_mds || error "Stopping MDSes failed"
9095 stop_mds || error "Stopping MDSes failed"
9097 run_test 121 "failover MGS"
9100 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
9101 [[ "$OST1_VERSION" -ge $(version_code 2.11.53) ]] ||
9102 skip "Need OST version at least 2.11.53"
9105 LOAD_MODULES_REMOTE=true load_modules
9106 #define OBD_FAIL_OFD_SET_OID 0x1e0
9107 do_facet ost1 $LCTL set_param fail_loc=0x00001e0
9110 $LFS mkdir -i1 -c1 $DIR/$tdir
9111 $LFS setstripe -i0 -c1 $DIR/$tdir
9112 do_facet ost1 $LCTL set_param fail_loc=0
9113 createmany -o $DIR/$tdir/file_ 1000 ||
9114 error "Fail to create a new sequence"
9118 run_test 122a "Check OST sequence update"
9121 [[ "$OST1_VERSION" -ge $(version_code 2.11.53) ]] ||
9122 skip "Need OST version at least 2.11.53"
9126 LOAD_MODULES_REMOTE=true load_modules
9127 #define OBD_FAIL_OFD_SET_OID 0x1e0
9128 do_facet ost1 $LCTL set_param fail_loc=0x00001e0
9130 stack_trap cleanup EXIT
9132 do_facet ost1 $LCTL set_param obdfilter.*.precreate_batch=256
9133 $LFS mkdir -i0 -c1 $DIR/$tdir || error "failed to create directory"
9134 $LFS setstripe -i0 -c1 $DIR/$tdir || error "failed to setstripe"
9135 do_facet ost1 $LCTL set_param fail_loc=0
9136 # overflow IDIF 32bit and create > OST_MAX_PRECREATE*5
9137 # so a new wrong sequence would differ from an original with error
9138 #define OST_MAX_PRECREATE 20000
9139 local ost_max_precreate=20100
9140 local num_create=$(( ost_max_precreate * 5 ))
9142 # Check the number of inodes available on OST0
9144 local ifree=$($LFS df -i $MOUNT |
9145 awk '/OST0000/ { print $4 }'; exit ${PIPESTATUS[0]})
9147 log "On OST0, $ifree inodes available. Want $num_create. rc=$?"
9149 if [ $ifree -lt 10000 ]; then
9150 files=$(( ifree - 50 ))
9155 local j=$((num_create / files + 1))
9157 for i in $(seq 1 $j); do
9158 createmany -o $DIR/$tdir/$tfile-$i- $files ||
9159 error "createmany fail create $files files: $?"
9160 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
9161 error "unlinkmany failed unlink $files files"
9164 touch $DIR/$tdir/$tfile
9166 #we need a write req during recovery for ofd_seq_load
9168 dd if=/dev/urandom of=$DIR/$tdir/$tfile bs=1024k count=1 oflag=sync ||
9169 error "failed to write file"
9171 # OBD_FAIL_OST_CREATE_NET 0x204
9172 do_facet ost1 $LCTL set_param fail_loc=0x80000204
9174 createmany -o $DIR/$tdir/file_ 100
9177 err=$(do_facet ost1 dmesg | tac | sed "/Recovery over/,$ d" |
9178 grep "OST replaced or reformatted")
9179 [ -z "$err" ] || error $err
9181 run_test 122b "Check OST sequence wouldn't change when IDIF 32bit overflows"
9184 remote_mgs_nodsh && skip "remote MGS with nodsh"
9185 [ -d $MOUNT/.lustre ] || setup
9187 # test old logid format until removal from llog_ioctl.c::str2logid()
9188 if [ $MGS_VERSION -lt $(version_code 3.1.53) ]; then
9189 do_facet mgs $LCTL dl | grep MGS
9190 do_facet mgs "$LCTL --device %MGS llog_print \
9191 \\\\\\\$$FSNAME-client 1 10" ||
9192 error "old llog_print failed"
9195 # test new logid format
9196 if [ $MGS_VERSION -ge $(version_code 2.9.53) ]; then
9197 do_facet mgs "$LCTL --device MGS llog_print $FSNAME-client" ||
9198 error "new llog_print failed"
9201 run_test 123aa "llog_print works with FIDs and simple names"
9204 remote_mgs_nodsh && skip "remote MGS with nodsh"
9205 [[ $MGS_VERSION -gt $(version_code 2.11.51) ]] ||
9206 skip "Need server with working llog_print support"
9208 [ -d $MOUNT/.lustre ] || setup
9214 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9215 mgs_arg="--device MGS"
9217 orig_val=$(do_facet mgs $LCTL get_param jobid_name)
9218 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME"
9220 yaml=$(do_facet mgs $LCTL $mgs_arg llog_print params |
9221 grep jobid_name | tail -n 1)
9223 local param=$(awk '{ print $10 }' <<< "$yaml")
9224 local val=$(awk '{ print $12 }' <<< "$yaml")
9225 #return to the default
9226 do_facet mgs $LCTL set_param -P jobid_name=$orig_val
9227 [ $val = "TESTNAME" ] || error "bad value: $val"
9228 [ $param = "jobid_name," ] || error "Bad param: $param"
9230 run_test 123ab "llog_print params output values from set_param -P"
9232 test_123ac() { # LU-11566
9233 remote_mgs_nodsh && skip "remote MGS with nodsh"
9234 do_facet mgs "$LCTL help llog_print" 2>&1 | grep -q -- --start ||
9235 skip "Need 'lctl llog_print --start' on MGS"
9241 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9242 mgs_arg="--device MGS"
9244 [ -d $MOUNT/.lustre ] || setup
9246 # - { index: 10, event: add_uuid, nid: 192.168.20.1@tcp(0x20000c0a81401,
9247 # node: 192.168.20.1@tcp }
9248 do_facet mgs $LCTL $mgs_arg \
9249 llog_print --start $start --end $end $FSNAME-client | tr -d , |
9250 while read DASH BRACE INDEX idx EVENT BLAH BLAH BLAH; do
9251 (( idx >= start )) || error "llog_print index $idx < $start"
9252 (( idx <= end )) || error "llog_print index $idx > $end"
9255 run_test 123ac "llog_print with --start and --end"
9257 test_123ad() { # LU-11566
9258 remote_mgs_nodsh && skip "remote MGS with nodsh"
9259 # older versions of lctl may not print all records properly
9260 do_facet mgs "$LCTL help llog_print" 2>&1 | grep -q -- --start ||
9261 skip "Need 'lctl llog_print --start' on MGS"
9263 [ -d $MOUNT/.lustre ] || setup
9265 # append a new record, to avoid issues if last record was cancelled
9266 local old=$($LCTL get_param -n osc.*-OST0000-*.max_dirty_mb | head -1)
9267 do_facet mgs $LCTL conf_param $FSNAME-OST0000.osc.max_dirty_mb=$old
9269 # logid: [0x3:0xa:0x0]:0
9273 local num=$(do_facet mgs $LCTL --device MGS llog_info $FSNAME-client |
9274 awk '/last_index:/ { print $2 - 1 }')
9276 # - { index: 71, event: set_timeout, num: 0x14, param: sys.timeout=20 }
9277 local last=$(do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
9278 tail -1 | awk '{ print $4 }' | tr -d , )
9279 (( last == num )) || error "llog_print only showed $last/$num records"
9281 run_test 123ad "llog_print shows all records"
9283 test_123ae() { # LU-11566
9292 remote_mgs_nodsh && skip "remote MGS with nodsh"
9293 [ -d $MOUNT/.lustre ] || setupall
9295 max=$($LCTL get_param -n osc.*-OST0000-*.max_dirty_mb | head -1)
9296 pgs=$($LCTL get_param -n osc.*-OST0000-*.max_pages_per_rpc | head -1)
9297 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9298 mgs_arg="--device MGS"
9300 if do_facet mgs "$LCTL help llog_cancel" 2>&1| grep -q -- --log_id; then
9301 # save one set_param -P record in case none exist
9303 do_facet mgs $LCTL set_param -P osc.*.max_pages_per_rpc=$pgs
9304 stack_trap "do_facet mgs $LCTL set_param -P -d \
9305 osc.*.max_pages_per_rpc"
9308 orig=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9309 tail -1 | awk '{ print $4 }' | tr -d , )
9310 do_facet mgs $LCTL set_param -P osc.*.max_dirty_mb=$max
9311 do_facet mgs $LCTL $mgs_arg llog_print $log | tail -1 |
9312 grep "parameter: osc.*.max_dirty_mb" ||
9313 error "new set_param -P wasn't stored in params log"
9315 # - { index: 71, event: set_param, device: general,
9316 # param: osc.*.max_dirty_mb, value: 256 }
9317 id=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9318 tail -1 | awk '{ print $4 }' | tr -d , )
9319 do_facet mgs $LCTL $mgs_arg llog_cancel $log --log_idx=$id
9320 local new=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9321 tail -1 | awk '{ print $4 }' | tr -d , )
9322 (( new == orig )) ||
9323 error "new llog_cancel now $new, not at $orig records"
9326 # test old positional parameters for a while still
9327 if [ "$MGS_VERSION" -le $(version_code 3.1.53) ]; then
9330 do_facet mgs $LCTL conf_param \
9331 $FSNAME-OST0000.osc.max_pages_per_rpc=$pgs
9332 stack_trap "do_facet mgs $LCTL conf_param -d \
9333 $FSNAME-OST0000.osc.max_pages_per_rpc"
9335 orig=$(do_facet mgs $LCTL --device MGS llog_print $log |
9336 tail -1 | awk '{ print $4 }' | tr -d , )
9337 do_facet mgs $LCTL conf_param $FSNAME-OST0000.osc.max_dirty_mb=$max
9338 do_facet mgs $LCTL --device MGS llog_print $log |
9339 tail -1 | grep "parameter: osc.max_dirty_mb" ||
9340 error "old conf_param wasn't stored in params log"
9341 do_facet mgs $LCTL --device MGS llog_print $log
9342 # - { index: 71, event: conf_param, device: testfs-OST0000-osc,
9343 # param: osc.max_dirty_mb=256 }
9344 id=$(do_facet mgs $LCTL --device MGS llog_print $log |
9345 tail -1 | awk '{ print $4 }' | tr -d , )
9346 do_facet mgs $LCTL --device MGS llog_cancel $log $id
9347 do_facet mgs $LCTL --device MGS llog_print $log
9348 new=$(do_facet mgs $LCTL --device MGS llog_print $log |
9349 tail -1 | awk '{ print $4 }' | tr -d , )
9350 (( new == orig )) ||
9351 error "old llog_cancel now $new, not at $orig records"
9354 run_test 123ae "llog_cancel can cancel requested record"
9356 test_123af() { #LU-13609
9357 [ "$MGS_VERSION" -ge $(version_code 2.13.54) -a \
9358 "$MDS1_VERSION" -ge $(version_code 2.13.54) ] ||
9359 skip "Need both MGS and MDS version at least 2.13.54"
9361 [ -d $MOUNT/.lustre ] || setupall
9362 stack_trap "do_facet mds1 $LCTL set_param fail_loc=0" EXIT
9372 for device in "MGS" "$FSNAME-MDT0000"; do
9373 cmd="--device $device llog_catlist"
9374 echo "lctl $cmd ..."
9375 if [ "$device" = "MGS" ]; then
9380 orig_clist=($(do_facet $facet $LCTL $cmd | awk '{ print $2 }'))
9381 orig_count=${#orig_clist[@]}
9382 echo "orig_clist: ${orig_clist[@]}"
9384 #define OBD_FAIL_CATLIST 0x131b
9385 #fetch to llog records from the second one
9386 do_facet $facet $LCTL set_param fail_loc=0x131b fail_val=2
9388 new_clist=($(do_facet $facet $LCTL $cmd | awk '{ print $2 }'))
9389 new_count=${#new_clist[@]}
9390 echo "new_clist: ${new_clist[@]}"
9392 [ $new_count -eq $((orig_count - 1)) ] ||
9393 error "$new_count != $orig_count - 1"
9394 for i in $(seq 0 $new_count); do
9396 [ "${orig_clist[$j]}" = "${new_clist[$i]}" ] ||
9397 error "${orig_clist[$j]} != ${new_clist[$i]}"
9399 do_facet mds1 $LCTL set_param fail_loc=0
9403 run_test 123af "llog_catlist can show all config files correctly"
9405 test_123ag() { # LU-15142
9409 remote_mgs_nodsh && skip "remote MGS with nodsh"
9410 (( $MGS_VERSION >= $(version_code 2.14.55) )) ||
9411 skip "Need server version least 2.14.55"
9413 [ -d $MOUNT/.lustre ] || setup
9415 orig_val=$(do_facet mgs $LCTL get_param jobid_name)
9416 stack_trap "do_facet mgs $LCTL set_param -P jobid_name=$orig_val"
9418 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME1"
9419 do_facet mgs $LCTL set_param -P -d jobid_name
9420 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
9422 (( rec == 0 )) || error "parameter was not deleted, check #1"
9423 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME1"
9424 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
9426 (( rec == 1)) || error "parameter is not set"
9427 # usage with ordinary set_param format works too
9428 do_facet mgs $LCTL set_param -P -d jobid_name="ANY"
9429 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
9431 (( rec == 0 )) || error "parameter was not deleted, check #2"
9433 run_test 123ag "llog_print skips values deleted by set_param -P -d"
9436 remote_mgs_nodsh && skip "remote MGS with nodsh"
9438 [ -d $MOUNT/.lustre ] || setup
9439 local yaml_file="$TMP/$tfile.yaml"
9440 do_facet mgs rm "$yaml_file"
9441 local cfgfiles=$(do_facet mgs "lctl --device MGS llog_catlist" |
9442 sed 's/config_log://')
9444 # set jobid_var to a different value for test
9445 local orig_val=$(do_facet mgs $LCTL get_param jobid_var)
9446 do_facet mgs $LCTL set_param -P jobid_var="TESTNAME"
9448 for i in $cfgfiles params; do
9449 do_facet mgs "lctl --device MGS llog_print ${i} >> $yaml_file"
9452 echo "Unmounting FS"
9459 # Reapply the config from before
9460 echo "Setting configuration parameters"
9461 do_facet mgs "lctl set_param -F $yaml_file"
9463 local set_val=$(do_facet mgs $LCTL get_param jobid_var)
9464 do_facet mgs $LCTL set_param -P $orig_val
9466 [ $set_val == "jobid_var=TESTNAME" ] ||
9467 error "$set_val is not TESTNAME"
9469 do_facet mgs rm "$yaml_file"
9472 run_test 123F "clear and reset all parameters using set_param -F"
9476 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
9477 [ -z $mds2failover_HOST ] && skip "needs MDT failover setup"
9480 do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
9481 grep 1.2.3.4@tcp && error "Should not be fake nid"
9482 do_facet mgs $LCTL conf_param $FSNAME-MDT0001.failover.node=1.2.3.4@tcp\
9483 || error "Set params error"
9484 do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
9485 grep 1.2.3.4@tcp || error "Fake nid should be added"
9489 if combined_mgs_mds; then
9490 start_mdt 1 "-o nosvc" ||
9491 error "starting mds with nosvc option failed"
9493 local nid=$(do_facet mds2 $LCTL list_nids | head -1)
9494 local failover_nid=$(do_node $mds2failover_HOST $LCTL list_nids | head -1)
9495 do_facet mgs $LCTL replace_nids $FSNAME-MDT0001 $nid:$failover_nid ||
9496 error "replace_nids execution error"
9498 if combined_mgs_mds; then
9504 echo "lfs setdirstripe"
9505 $LFS setdirstripe -i 1 $MOUNT/$tdir || error "setdirstirpe error"
9508 run_test 124 "check failover after replace_nids"
9510 get_max_sectors_kb() {
9513 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
9514 local max_sectors_path="/sys/block/${dev_base}/queue/max_sectors_kb"
9516 do_facet ${facet} "[[ -e ${max_sectors_path} ]] &&
9517 cat ${max_sectors_path}"
9520 get_max_hw_sectors_kb() {
9523 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
9524 local max_hw_path="/sys/block/${dev_base}/queue/max_hw_sectors_kb"
9526 do_facet ${facet} "[[ -e ${max_hw_path} ]] && cat ${max_hw_path}"
9529 set_max_sectors_kb() {
9533 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
9534 local max_sectors_path="/sys/block/${dev_base}/queue/max_sectors_kb"
9536 do_facet ${facet} "[[ -e ${max_sectors_path} ]] &&
9537 echo ${value} > ${max_sectors_path}"
9540 [[ $rc -ne 0 ]] && echo "Failed to set ${max_sectors_path} to ${value}"
9545 # Return 0 if all slave devices have max_sectors_kb == max_hw_sectors_kb
9546 # Otherwise return > 0
9547 check_slaves_max_sectors_kb()
9551 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
9552 local slaves_dir=/sys/block/${dev_base}/slaves
9553 local slave_devices=$(do_facet ${facet} "ls ${slaves_dir} 2>/dev/null")
9554 [[ -z ${slave_devices} ]] && return 0
9556 local slave max_sectors new_max_sectors max_hw_sectors path
9558 for slave in ${slave_devices}; do
9559 path="/dev/${slave}"
9560 ! is_blkdev ${facet} ${path} && continue
9561 max_sectors=$(get_max_sectors_kb ${facet} ${path})
9562 max_hw_sectors=$(get_max_hw_sectors_kb ${facet} ${path})
9563 new_max_sectors=${max_hw_sectors}
9564 [[ ${new_max_sectors} -gt ${RQ_SIZE_LIMIT} ]] &&
9565 new_max_sectors=${RQ_SIZE_LIMIT}
9567 if [[ ${max_sectors} -ne ${new_max_sectors} ]]; then
9568 echo "${path} ${max_sectors} ${new_max_sectors}"
9571 check_slaves_max_sectors_kb ${facet} ${path}
9580 local facet_list="mgs mds1 ost1"
9581 combined_mgs_mds && facet_list="mgs ost1"
9584 for facet in ${facet_list}; do
9585 [[ $(facet_fstype ${facet}) != ldiskfs ]] &&
9586 skip "ldiskfs only test" &&
9588 ! is_blkdev ${facet} $(facet_device ${facet}) &&
9589 skip "requires all real devices" &&
9594 # We don't increase IO request size limit past 16MB. See comments in
9595 # lustre/utils/libmount_utils_ldiskfs.c:tune_max_sectors_kb()
9596 RQ_SIZE_LIMIT=$((16 * 1024))
9597 local device old_max_sectors new_max_sectors max_hw_sectors
9598 for facet in ${facet_list}; do
9599 device=$(facet_device ${facet})
9600 old_max_sectors=$(get_max_sectors_kb ${facet} ${device})
9601 max_hw_sectors=$(get_max_hw_sectors_kb ${facet} ${device})
9603 # The expected value after l_tunedisk is executed
9604 new_max_sectors=$old_max_sectors
9605 [[ ${new_max_sectors_kb} -gt ${RQ_SIZE_LIMIT} ]] &&
9606 new_max_sectors_kb=${RQ_SIZE_LIMIT}
9608 # Ensure the current value of max_sectors_kb does not equal
9609 # max_hw_sectors_kb, so we can tell whether l_tunedisk did
9611 set_max_sectors_kb ${facet} ${device} $((new_max_sectors - 1))
9613 # Value before l_tunedisk
9614 local pre_max_sectors=$(get_max_sectors_kb ${facet} ${device})
9615 if [[ ${pre_max_sectors} -ne $((new_max_sectors - 1)) ]]; then
9616 echo "unable to satsify test pre-condition:"
9617 echo "${pre_max_sectors} != $((new_max_sectors - 1))"
9622 echo "Before: ${facet} ${device} ${pre_max_sectors} ${max_hw_sectors}"
9624 do_facet ${facet} "l_tunedisk ${device}"
9626 # Value after l_tunedisk
9627 local post_max_sectors=$(get_max_sectors_kb ${facet} ${device})
9629 echo "After: ${facet} ${device} ${post_max_sectors} ${max_hw_sectors}"
9631 if [[ ${facet} != ost1 ]]; then
9632 if [[ ${post_max_sectors} -ne ${pre_max_sectors} ]]; then
9633 echo "l_tunedisk modified max_sectors_kb of ${facet}"
9637 set_max_sectors_kb ${facet} ${device} ${old_max_sectors}
9639 if [[ ${post_max_sectors} -eq ${pre_max_sectors} ]]; then
9640 echo "l_tunedisk failed to modify max_sectors_kb of ${facet}"
9644 check_slaves_max_sectors_kb ${facet} ${device} ||
9651 run_test 125 "check l_tunedisk only tunes OSTs and their slave devices"
9654 [[ "$MDS1_VERSION" -ge $(version_code 2.13.52) ]] ||
9655 skip "Need MDS version at least 2.13.52"
9658 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ../libcfs/libcfs/libcfs
9659 #define OBD_FAIL_OBD_SETUP 0x60d
9660 do_facet mds1 $LCTL set_param fail_loc=0x60d
9661 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_modules &
9662 for i in {1..40}; do
9663 do_facet mds1 lsmod | grep -q osd_$mds1_FSTYPE && break
9666 clear_failloc $SINGLEMDS 20 &
9667 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
9669 run_test 126 "mount in parallel shouldn't cause a crash"
9672 [[ "$ost1_FSTYPE" == ldiskfs ]] || skip "ldiskfs only test"
9676 zconf_umount_clients $RCLIENTS $MOUNT
9678 wait_osp_active ost ${FSNAME}-OST0000 0 1
9679 local osc_tgt="$FSNAME-OST0000-osc-$($LFS getname -i $DIR)"
9680 local avail1=($($LCTL get_param -n osc.${osc_tgt}.kbytesavail))
9682 $LFS setstripe -i 0 $DIR/$tfile || error "failed creating $DIR/$tfile"
9683 dd if=/dev/zero of=$DIR/$tfile bs=1M oflag=direct || true
9685 local avail2=($($LCTL get_param -n osc.${osc_tgt}.kbytesavail))
9687 if ((avail2 * 100 / avail1 > 1)); then
9690 error "more than 1% space left: before=$avail1 after=$avail2"
9693 local mbs=$(($(stat -c %s $DIR/$tfile) / (1024 * 1024)))
9695 dd if=/dev/zero of=$DIR/$tfile bs=1M count=$mbs conv=notrunc \
9696 oflag=direct || error "overwrite failed"
9698 run_test 127 "direct io overwrite on full ost"
9702 combined_mgs_mds && skip "need separate mgs device"
9703 [ "$ost2_FSTYPE" == zfs ] && import_zpool ost2
9706 # Try to apply nolocallogs to the virgin OST. Should fail.
9707 do_facet ost2 "$TUNEFS --nolocallogs $(ostdevname 2)" &&
9708 error "nolocallogs should not be allowed on the virgin target"
9713 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
9714 # Start OST without MGS (local configs)
9715 do_facet ost1 "$TUNEFS --dryrun $(ostdevname 1)"
9716 start_ost || error "unable to start OST1"
9717 stop_ost || error "Unable to stop OST1"
9719 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
9720 # Do not allow reading local configs, should fail
9721 do_facet ost1 "$TUNEFS --nolocallogs $(ostdevname 1)" ||
9722 error "Can not set nolocallogs"
9723 start_ost && error "OST1 started, but should fail"
9725 # Connect to MGS successfully, reset nolocallogs flag
9726 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
9727 start_mgs || error "unable to start MGS"
9728 start_ost || error "unable to start OST1"
9730 do_facet ost1 "$TUNEFS --dryrun $(ostdevname 1)" | grep "nolocallogs" &&
9731 error "nolocallogs expected to be reset"
9733 stop_ost || error "Unable to stop OST1"
9735 run_test 128 "Force using remote logs with --nolocallogs"
9740 start_mds || error "MDS start failed"
9742 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &&
9743 error "start ost1 should fail" || true
9744 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &&
9745 error "second start ost1 should fail" || true
9746 do_facet ost1 "$TUNEFS --writeconf $(ostdevname 1)"
9747 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS ||
9748 error "start ost1 failed"
9752 run_test 129 "attempt to connect an OST with the same index should fail"
9756 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
9758 stop_mdt 2 || error "mdt2 stop failed"
9759 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)"
9760 start_mdt 2 || error "mdt2 start failed"
9761 do_facet mds2 "$LCTL dl" | grep MDT0001-osp-MDT0001 &&
9762 error "Illegal OSP device created" || true
9764 run_test 130 "re-register an MDT after writeconf"
9767 [ "$mds1_FSTYPE" == "ldiskfs" ] || skip "ldiskfs only test"
9768 do_facet mds1 $DEBUGFS -R features $(mdsdevname 1) |
9769 grep -q project || skip "skip project quota not supported"
9774 test_mkdir -c $MDSCOUNT -p $DIR/$tdir
9775 $LFS project -p 1000 $DIR/$tdir || error "set dir project id failed"
9776 createmany -o $DIR/$tdir/f 512
9777 for ((i = 0; i < 512; ++i)); do
9778 $LFS project -p $i $DIR/$tdir/f${i} ||
9779 error "set f${i} project id failed"
9782 test_mkdir -c $MDSCOUNT -p $DIR/$tdir.inherit
9783 $LFS project -p 1001 -s $DIR/$tdir.inherit
9784 createmany -o $DIR/$tdir.inherit/f 128
9785 (( $($LFS project $DIR/$tdir.inherit/f* |
9786 awk '$1 == 1001 { print }' | wc -l) == 128 )) ||
9787 error "files did not inherit projid 1001"
9791 for i in $(seq $MDSCOUNT); do
9792 mds_backup_restore mds$i ||
9793 error "Backup/restore on mds$i failed"
9798 projid=($($LFS project -d $DIR/$tdir))
9799 [ ${projid[0]} == "1000" ] ||
9800 error "projid expected 1000 not ${projid[0]}"
9801 for ((i = 0; i < 512; ++i)); do
9802 projid=($($LFS project $DIR/$tdir/f${i}))
9803 [ ${projid[0]} == "$i" ] ||
9804 error "projid expected $i not ${projid[0]}"
9807 (( $($LFS project $DIR/$tdir.inherit/f* |
9808 awk '$1 == 1001 { print }' | wc -l) == 128 )) ||
9809 error "restore did not copy projid 1001"
9811 run_test 131 "MDT backup restore with project ID"
9818 combined_mgs_mds || start_mgs || error "unable to start MGS"
9819 start_mdt 1 || error "unable to start mdt1"
9821 err_cnt=$(do_facet mds1 dmesg | grep -c "cannot take the layout locks")
9822 stop_mdt 1 || error "stop mdt1 failed"
9824 [ "$mds1_FSTYPE" == zfs ] && import_zpool mds1
9825 do_facet mds1 $TUNEFS --param mdt.hsm_control=enabled $(mdsdevname 1) ||
9826 error "tunefs failed"
9827 start_mdt 1 || error "cannot start mdt1"
9829 err_cnt2=$(do_facet mds1 dmesg | grep -c "cannot take the layout locks")
9830 [ $err_cnt -eq $err_cnt2 ] || error "Can not take the layout lock"
9831 stop_mdt 1 || error "stop mdt1 failed"
9833 run_test 132 "hsm_actions processed after failover"
9835 if ! combined_mgs_mds ; then
9841 # restore the values of MDSSIZE and OSTSIZE
9842 MDSSIZE=$STORED_MDSSIZE
9843 OSTSIZE=$STORED_OSTSIZE
9847 check_and_cleanup_lustre