8 PTLDEBUG=${PTLDEBUG:--1}
9 LUSTRE=${LUSTRE:-$(dirname $0)/..}
10 . $LUSTRE/tests/test-framework.sh
14 # tool to create lustre filesystem images
15 ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT 32newtarball"
17 # bug number for skipped test: LU-11915
18 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 110"
19 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
22 # bug number for skipped tests: LU-9795 (all below)
23 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 84 86 103"
26 if ! combined_mgs_mds; then
27 # bug number for skipped test: LU-11991 LU-11990
28 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 32a 32b 32c 32d 32e 66"
29 # bug number for skipped test: LU-9897 LU-12032
30 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 84 123F"
34 [ "$SLOW" = "no" ] && EXCEPT_SLOW="45 69 106 111 114"
38 # use small MDS + OST size to speed formatting time
39 # do not use too small MDSSIZE/OSTSIZE, which affect the default journal size
40 # STORED_MDSSIZE is used in test_18
41 STORED_MDSSIZE=$MDSSIZE
42 STORED_OSTSIZE=$OSTSIZE
44 [ "$mds1_FSTYPE" = zfs ] && MDSSIZE=400000
46 [ "$ost1_FSTYPE" = zfs ] && OSTSIZE=400000
56 # pass "-E lazy_itable_init" to mke2fs to speed up the formatting time
57 if [[ "$LDISKFS_MKFS_OPTS" != *lazy_itable_init* ]]; then
58 LDISKFS_MKFS_OPTS=$(csa_add "$LDISKFS_MKFS_OPTS" -E lazy_itable_init)
62 require_dsh_mds || exit 0
63 require_dsh_ost || exit 0
68 # The MGS must be started before the OSTs for a new fs, so start
69 # and stop to generate the startup logs.
72 wait_osc_import_state mds ost FULL
77 reformat_and_config() {
79 if ! combined_mgs_mds ; then
85 writeconf_or_reformat() {
86 # There are at most 2 OSTs for write_conf test
87 # who knows if/where $TUNEFS is installed?
88 # Better reformat if it fails...
89 writeconf_all $MDSCOUNT 2 ||
90 { echo "tunefs failed, reformatting instead" &&
91 reformat_and_config && return 0; }
100 echo "start mgs service on $(facet_active_host mgs)"
101 start mgs $(mgsdevname) $MGS_MOUNT_OPTS "$@"
107 local dev=$(mdsdevname $num)
110 echo "start mds service on `facet_active_host $facet`"
111 start $facet ${dev} $MDS_MOUNT_OPTS "$@" || return 94
114 stop_mdt_no_force() {
117 local dev=$(mdsdevname $num)
120 echo "stop mds service on `facet_active_host $facet`"
121 stop $facet || return 97
127 local dev=$(mdsdevname $num)
130 echo "stop mds service on `facet_active_host $facet`"
131 # These tests all use non-failover stop
132 stop $facet -f || return 97
138 for num in $(seq $MDSCOUNT); do
139 start_mdt $num "$@" || return 94
141 for num in $(seq $MDSCOUNT); do
142 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds${num} FULL
147 if ! combined_mgs_mds ; then
155 for num in $(seq $MDSCOUNT); do
156 stop_mdt $num || return 97
161 echo "stop mgs service on `facet_active_host mgs`"
162 # These tests all use non-failover stop
163 stop mgs -f || return 97
167 echo "start ost1 service on `facet_active_host ost1`"
168 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS "$@" || return 95
169 wait_clients_import_ready ${CLIENTS:-$HOSTNAME} ost1
173 echo "stop ost1 service on `facet_active_host ost1`"
174 # These tests all use non-failover stop
175 stop ost1 -f || return 98
179 echo "start ost2 service on `facet_active_host ost2`"
180 start ost2 $(ostdevname 2) $OST_MOUNT_OPTS "$@" || return 92
181 wait_clients_import_ready ${CLIENTS:-$HOSTNAME} ost2
185 echo "stop ost2 service on `facet_active_host ost2`"
186 # These tests all use non-failover stop
187 stop ost2 -f || return 93
194 echo "mount $FSNAME ${mountopt:+with opts $mountopt} on $mountpath....."
195 zconf_mount $HOSTNAME $mountpath $mountopt || return 96
201 echo "umount lustre on $mountpath....."
202 zconf_umount $HOSTNAME $mountpath "$@" || return 97
205 manual_umount_client(){
208 echo "manual umount lustre on ${MOUNT}...."
209 do_facet client "umount ${FORCE} $MOUNT"
215 start_mds || error "MDT start failed"
216 start_ost || error "Unable to start OST1"
217 mount_client $MOUNT || error "client start failed"
218 client_up || error "client_up failed"
227 unload_modules_conf () {
228 if combined_mgs_mds || ! local_mode; then
229 unload_modules || return 1
234 stop_ost || return 202
235 stop_mds || return 201
236 unload_modules_conf || return 203
241 [ "x$1" != "x" ] && force='-f'
242 umount_client $MOUNT $force|| return 200
243 cleanup_nocli || return $?
248 echo "umount $MOUNT2 ..."
249 umount $MOUNT2 || true
250 echo "stopping fs2mds ..."
251 stop fs2mds -f || true
252 echo "stopping fs2ost ..."
253 stop fs2ost -f || true
257 do_facet client "cp /etc/passwd $DIR/a" || return 71
258 do_facet client "rm $DIR/a" || return 72
259 # make sure lustre is actually mounted (touch will block,
260 # but grep won't, so do it after)
261 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
262 echo "setup single mount lustre success"
266 do_facet client "touch $DIR/a" || return 71
267 do_facet client "rm $DIR/a" || return 72
268 do_facet client "touch $DIR2/a" || return 73
269 do_facet client "rm $DIR2/a" || return 74
270 echo "setup double mount lustre success"
274 cat /dev/urandom | tr -dc 'a-zA-Z0-9' | fold -w $1 | head -n 1
277 if [ "$ONLY" == "setup" ]; then
282 if [ "$ONLY" == "cleanup" ]; then
289 #create single point mountpoint
295 check_mount || error "check_mount failed"
296 cleanup || error "cleanup failed with $?"
298 run_test 0 "single mount setup"
301 start_mds || error "MDS start failed"
302 start_ost || error "unable to start OST"
303 echo "start ost second time..."
304 start_ost && error "2nd OST start should fail"
305 mount_client $MOUNT || error "client start failed"
306 check_mount || error "check_mount failed"
307 cleanup || error "cleanup failed with $?"
309 run_test 1 "start up ost twice (should return errors)"
312 start_mds || error "MDT start failed"
313 echo "start mds second time.."
314 start_mds && error "2nd MDT start should fail"
315 start_ost || error "OST start failed"
316 mount_client $MOUNT || error "mount_client failed to start client"
317 check_mount || error "check_mount failed"
318 cleanup || error "cleanup failed with $?"
320 run_test 2 "start up mds twice (should return err)"
324 #mount.lustre returns an error if already in mtab
325 mount_client $MOUNT && error "2nd client mount should fail"
326 check_mount || error "check_mount failed"
327 cleanup || error "cleanup failed with $?"
329 run_test 3 "mount client twice (should return err)"
333 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
334 stop_ost || error "Unable to stop OST1"
335 umount_client $MOUNT -f || error "unmount $MOUNT failed"
338 # ok for ost to fail shutdown
339 if [ 202 -ne $eno ] && [ 0 -ne $eno ]; then
340 error "cleanup failed with $?"
343 run_test 4 "force cleanup ost, then cleanup"
345 test_5a() { # was test_5
347 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
348 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
350 stop_mds || error "Unable to stop MDS"
352 # cleanup may return an error from the failed
353 # disconnects; for now I'll consider this successful
354 # if all the modules have unloaded.
358 echo "killing umount"
359 kill -TERM $UMOUNT_PID
360 echo "waiting for umount to finish"
362 if grep " $MOUNT " /proc/mounts; then
363 echo "test 5: /proc/mounts after failed umount"
367 echo "killing umount"
368 kill -TERM $UMOUNT_PID
369 echo "waiting for umount to finish"
371 grep " $MOUNT " /proc/mounts &&
372 error "/proc/mounts after second umount"
375 # manual_mount_client may fail due to umount succeeding above
377 # stop_mds is a no-op here, and should not fail
378 cleanup_nocli || error "cleanup_nocli failed with $?"
379 # df may have lingering entry
381 # mtab may have lingering entry
385 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
387 grep -q $MOUNT" " /etc/mtab || break
388 echo "Waiting /etc/mtab updated ... "
389 WAIT=$(( WAIT + sleep))
391 [ "$WAIT" -eq "$MAX_WAIT" ] &&
392 error "/etc/mtab is not updated in $WAIT secs"
393 echo "/etc/mtab updated in $WAIT secs"
395 run_test 5a "force cleanup mds, then cleanup"
403 grep " $MOUNT " /etc/mtab &&
404 error false "unexpected entry in mtab before mount" && return 10
406 start_ost || error "OST start failed"
407 if ! combined_mgs_mds ; then
408 trap cleanup_5b EXIT ERR
409 start_mds || error "MDS start failed"
413 mount_client $MOUNT && error "mount_client $MOUNT should fail"
414 grep " $MOUNT " /etc/mtab &&
415 error "$MOUNT entry in mtab after failed mount"
417 # stop_mds is a no-op here, and should not fail
418 cleanup_nocli || error "cleanup_nocli failed with $?"
419 if ! combined_mgs_mds ; then
423 run_test 5b "Try to start a client with no MGS (should return errs)"
426 grep " $MOUNT " /etc/mtab &&
427 error false "unexpected entry in mtab before mount" && return 10
429 start_mds || error "MDS start failed"
430 start_ost || error "OST start failed"
431 local oldfs="${FSNAME}"
432 FSNAME="wrong.${FSNAME}"
433 mount_client $MOUNT || :
435 grep " $MOUNT " /etc/mtab &&
436 error "$MOUNT entry in mtab after failed mount"
438 cleanup_nocli || error "cleanup_nocli failed with $?"
440 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
443 grep " $MOUNT " /etc/mtab &&
444 error "unexpected entry in mtab before mount"
446 start_ost || error "OST start failed"
447 start_mds || error "MDS start failed"
448 stop_ost -f || error "Unable to stop OST1"
449 mount_client $MOUNT || error "mount_client $MOUNT failed"
450 umount_client $MOUNT -f || error "umount_client $MOUNT failed"
451 cleanup_nocli || error "cleanup_nocli failed with $?"
452 ! grep " $MOUNT " /etc/mtab ||
453 error "$MOUNT entry in mtab after unmount"
455 run_test 5d "mount with ost down"
458 grep " $MOUNT " /etc/mtab &&
459 error false "unexpected entry in mtab before mount" && return 10
461 start_mds || error "MDS start failed"
462 start_ost || error "OST start failed"
464 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
465 do_facet client "$LCTL set_param fail_loc=0x80000506"
466 mount_client $MOUNT || echo "mount failed (not fatal)"
467 cleanup || error "cleanup failed with $?"
468 grep " $MOUNT " /etc/mtab &&
469 error "$MOUNT entry in mtab after unmount"
472 run_test 5e "delayed connect, don't crash (bug 10268)"
475 combined_mgs_mds && skip "needs separate mgs and mds"
477 grep " $MOUNT " /etc/mtab &&
478 error false "unexpected entry in mtab before mount" && return 10
481 start_ost || error "OST start failed"
482 mount_client $MOUNT &
484 echo client_mount pid is $pid
488 if ! ps -f -p $pid >/dev/null; then
491 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
492 error "mount returns $rc, expected to hang"
494 cleanup || error "cleanup failed with $?"
499 start_mds || error "start MDS failed"
501 # mount should succeed after start mds
503 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
504 cleanup || error "final call to cleanup failed with rc $?"
506 run_test 5f "mds down, cleanup after failed mount (bug 2712)"
510 [ "$CLIENT_VERSION" -lt $(version_code 2.9.53) ] &&
511 skip "automount of debugfs missing before 2.9.53"
512 umount /sys/kernel/debug
513 $LCTL get_param -n devices | egrep -v "error" && \
514 error "lctl can't access debugfs data"
515 grep " debugfs " /etc/mtab || error "debugfs failed to remount"
517 run_test 5g "handle missing debugfs"
523 #define OBD_FAIL_MDS_FS_SETUP 0x135
524 do_facet mds1 "$LCTL set_param fail_loc=0x80000135"
525 start_mdt 1 && error "start mdt should fail"
526 start_mdt 1 || error "start mdt failed"
527 client_up || error "client_up failed"
530 run_test 5h "start mdt failure at mdt_fs_setup()"
533 (( $MDS1_VERSION >= $(version_code 2.12.54) )) ||
534 skip "Need MDS version at least 2.12.54"
539 #define OBD_FAIL_QUOTA_INIT 0xA05
540 do_facet mds1 "$LCTL set_param fail_loc=0x80000A05"
541 start_mdt 1 && error "start mdt should fail"
542 start_mdt 1 || error "start mdt failed"
543 client_up || error "client_up failed"
546 run_test 5i "start mdt failure at mdt_quota_init()"
551 mount_client $MOUNT || error "mount_client $MOUNT failed"
552 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
553 cleanup || error "cleanup failed with rc $?"
555 run_test 6 "manual umount, then mount again"
560 cleanup_nocli || error "cleanup_nocli failed with $?"
562 run_test 7 "manual umount, then cleanup"
566 mount_client $MOUNT2 || error "mount_client $MOUNT2 failed"
567 check_mount2 || error "check_mount2 failed"
568 umount_client $MOUNT2 || error "umount_client $MOUNT2 failed"
569 cleanup || error "cleanup failed with rc $?"
571 run_test 8 "double mount setup"
574 start_ost || error "OST start failed"
576 do_facet ost1 $LCTL set_param debug=\'inode trace\' ||
577 error "do_facet ost1 set_param inode trace failed."
578 do_facet ost1 $LCTL set_param subsystem_debug=\'mds ost\' ||
579 error "do_facet ost1 set_param debug mds ost failed."
581 CHECK_PTLDEBUG="`do_facet ost1 $LCTL get_param -n debug`"
582 if [ "$CHECK_PTLDEBUG" ] && { \
583 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
584 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
585 echo "lnet.debug success"
587 error "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
589 CHECK_SUBSYS="`do_facet ost1 $LCTL get_param -n subsystem_debug`"
590 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
591 echo "lnet.subsystem_debug success"
593 error "lnet.subsystem_debug: want 'mds ost' got '$CHECK_SUBSYS'"
595 stop_ost || error "Unable to stop OST1"
597 run_test 9 "test ptldebug and subsystem for mkfs"
602 if ! combined_mgs_mds; then
603 files=$(do_facet mgs "find /{proc/fs,sys/fs,sys/kernel/debug}/lustre -type l -exec test ! -e {} \; -print")
604 [ -z $files ] || echo "MGS $files is a broken symlink"
607 files=$(do_facet mds1 "find /{proc/fs,sys/fs,sys/kernel/debug}/lustre -type l -exec test ! -e {} \; -print")
608 [ -z $files ] || echo "MDS $files is a broken symlink"
610 files=$(do_facet ost1 "find /{proc/fs,sys/fs,sys/kernel/debug}/lustre -type l -exec test ! -e {} \; -print")
611 [ -z $files ] || echo "OSS $files is a broken symlink"
613 files=$(do_facet client "find /{proc/fs,sys/fs,sys/kernel/debug}/lustre -type l -exec test ! -e {} \; -print")
614 [ -z $files ] || echo "clients $files is a broken symlink"
616 cleanup || error "cleanup failed with rc $?"
618 run_test 10a "find lctl param broken symlinks"
621 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
622 # But with new MDS stack we don't care about the mode of local objects
623 # anymore, so this test is removed. See bug 22944 for more details.
627 if [ "$mds1_FSTYPE" != ldiskfs ]; then
628 skip "ldiskfs only test"
632 check_mount || error "check_mount failed"
633 cleanup || error "cleanup failed with rc $?"
635 echo "Remove mds config log"
636 if ! combined_mgs_mds ; then
640 do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' \
641 $(mgsdevname) || return \$?" ||
642 error "do_facet mgs failed with $?"
644 if ! combined_mgs_mds ; then
648 start_ost || error "OST start failed"
649 start_mds && error "MDS start succeeded, but should fail"
652 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
655 if [ "$mds1_FSTYPE" != ldiskfs ]; then
656 skip "ldiskfs only test"
659 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
664 # check if current MDSSIZE is large enough
665 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE &&
666 log "use MDSSIZE=$MDSSIZE"
668 # check if the global config has a large enough MDSSIZE
669 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] &&
670 [ $STORED_MDSSIZE -ge $MIN ] &&
671 OK=1 && myMDSSIZE=$STORED_MDSSIZE &&
672 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
674 # check if the block device is large enough
675 is_blkdev $SINGLEMDS $MDSDEV $MIN
676 local large_enough=$?
677 if [ -n "$OK" ]; then
678 [ $large_enough -ne 0 ] && OK=""
680 [ $large_enough -eq 0 ] && OK=1 && myMDSSIZE=$MIN &&
681 log "use device $MDSDEV with MIN=$MIN"
684 # check if a loopback device has enough space for fs metadata (5%)
686 if [ -z "$OK" ]; then
687 local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! \
688 -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
689 awk '($1 != "Filesystem") { print $4 }')
690 ! [ -z "$SPACE" ] && [ $SPACE -gt $((MIN / 20)) ] &&
691 OK=1 && myMDSSIZE=$MIN &&
692 log "use file $MDSDEV with MIN=$MIN"
695 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS"
697 echo "mount mds with large journal..."
699 local OLD_MDSSIZE=$MDSSIZE
703 echo "mount lustre system..."
705 check_mount || error "check_mount failed"
707 echo "check journal size..."
708 local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
709 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
710 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
712 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
715 cleanup || error "cleanup failed with rc $?"
720 run_test 18 "check mkfs creates large journals"
723 start_mds || error "MDS start failed"
724 stop_mds || error "Unable to stop MDS"
726 run_test 19a "start/stop MDS without OSTs"
729 start_ost || error "Unable to start OST1"
730 stop_ost -f || error "Unable to stop OST1"
732 run_test 19b "start/stop OSTs without MDS"
735 # first format the ost/mdt
736 start_mds || error "MDS start failed"
737 start_ost || error "Unable to start OST1"
738 mount_client $MOUNT || error "mount_client $MOUNT failed"
739 check_mount || error "check_mount failed"
740 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed."
741 mount_client $MOUNT remount,ro || error "remount client with ro failed"
742 touch $DIR/$tfile && error "$DIR/$tfile created incorrectly"
743 [ -e $DIR/$tfile ] && error "$DIR/$tfile exists incorrectly"
744 mount_client $MOUNT remount,rw || error "remount client with rw failed"
745 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
746 MCNT=$(grep -c $MOUNT' ' /etc/mtab)
747 [ "$MCNT" -ne 1 ] && error "$MOUNT in /etc/mtab $MCNT times"
749 stop_mds || error "Unable to stop MDS"
750 stop_ost || error "Unable to stop OST1"
752 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
755 start_mds || error "MDS start failed"
756 start_ost || error "unable to start OST1"
757 wait_osc_import_state mds ost FULL
758 stop_ost || error "unable to stop OST1"
759 stop_mds || error "unable to stop MDS"
761 run_test 21a "start mds before ost, stop ost first"
764 start_ost || error "unable to start OST1"
765 start_mds || error "MDS start failed"
766 wait_osc_import_state mds ost FULL
767 stop_mds || error "unable to stop MDS"
768 stop_ost || error "unable to stop OST1"
770 run_test 21b "start ost before mds, stop mds first"
773 start_ost || error "Unable to start OST1"
774 start_mds || error "MDS start failed"
775 start_ost2 || error "Unable to start OST2"
776 wait_osc_import_state mds ost2 FULL
777 stop_ost || error "Unable to stop OST1"
778 stop_ost2 || error "Unable to stop OST2"
779 stop_mds || error "Unable to stop MDS"
780 #writeconf to remove all ost2 traces for subsequent tests
781 writeconf_or_reformat
783 run_test 21c "start mds between two osts, stop mds last"
786 combined_mgs_mds && skip "need separate mgs device"
791 start_mgs || error "unable to start MGS"
792 start_ost || error "unable to start OST1"
793 start_ost2 || error "unable to start OST2"
794 start_mds || error "MDS start failed"
795 wait_osc_import_state mds ost2 FULL
797 local zkeeper=${KEEP_ZPOOL}
798 stack_trap "KEEP_ZPOOL=$zkeeper" EXIT
801 stop_ost || error "Unable to stop OST1"
802 stop_ost2 || error "Unable to stop OST2"
803 stop_mds || error "Unable to stop MDS"
805 #writeconf to remove all ost2 traces for subsequent tests
806 writeconf_or_reformat
807 KEEP_ZPOOL="${zkeeper}"
809 start_mgs || error "unable to start MGS"
811 run_test 21d "start mgs then ost and then mds"
814 MGSNID="$saved_mgsnid"
816 echo "stopping fs2mgs ..."
817 stop $fs2mgs -f || true
820 test_21e() { # LU-5863
821 if [[ -z "$fs3ost_DEV" || -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
822 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
823 skip_env "mixed loopback and real device not working"
826 local fs2mdsdev=$(mdsdevname 1_2)
827 local fs2ostdev=$(ostdevname 1_2)
828 local fs3ostdev=$(ostdevname 2_2)
830 local fs2mdsvdev=$(mdsvdevname 1_2)
831 local fs2ostvdev=$(ostvdevname 1_2)
832 local fs3ostvdev=$(ostvdevname 2_2)
834 # temporarily use fs3ost as fs2mgs
836 local fs2mgsdev=$fs3ostdev
837 local fs2mgsvdev=$fs3ostvdev
839 local fsname=test1234
841 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
842 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
843 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS && trap cleanup_21e EXIT INT ||
844 error "start fs2mgs failed"
846 local saved_mgsnid="$MGSNID"
847 MGSNID=$(do_facet $fs2mgs $LCTL list_nids | xargs | tr ' ' ,)
849 add fs2mds $(mkfs_opts mds1 $fs2mdsdev $fsname) \
850 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
851 add fs2ost $(mkfs_opts ost1 $fs2ostdev $fsname) \
852 --reformat $fs2ostdev $fs2ostvdev || error "add fs2ost failed"
854 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
855 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS || error "start fs2mds failed"
857 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
858 $MOUNT_CMD $MGSNID:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
859 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
863 run_test 21e "separate MGS and MDS"
866 start_mds || error "MDS start failed"
868 echo "Client mount with ost in logs, but none running"
869 start_ost || error "unable to start OST1"
870 # wait until mds connected to ost and open client connection
871 wait_osc_import_state mds ost FULL
872 stop_ost || error "unable to stop OST1"
873 mount_client $MOUNT || error "mount_client $MOUNT failed"
874 # check_mount will block trying to contact ost
875 mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed"
876 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed"
877 umount_client $MOUNT -f
880 echo "Client mount with a running ost"
881 start_ost || error "unable to start OST1"
883 # if gss enabled, wait full time to let connection from
884 # mds to ost be established, due to the mismatch between
885 # initial connect timeout and gss context negotiation timeout.
886 # This perhaps could be remove after AT landed.
887 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
888 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
890 mount_client $MOUNT || error "mount_client $MOUNT failed"
891 wait_osc_import_state mds ost FULL
892 wait_osc_import_ready client ost
893 check_mount || error "check_mount failed"
896 cleanup || error "cleanup failed with rc $?"
898 run_test 22 "start a client before osts (should return errs)"
900 test_23a() { # was test_23
903 stop $SINGLEMDS || error "failed to stop $SINGLEMDS"
904 # force down client so that recovering mds waits for reconnect
905 local running=$(grep -c $MOUNT /proc/mounts) || true
906 if [ $running -ne 0 ]; then
907 echo "Stopping client $MOUNT (opts: -f)"
911 # enter recovery on failed mds
912 local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/})
913 start $SINGLEMDS $MDT_DEV $MDS_MOUNT_OPTS || error "MDS start failed"
914 # try to start a new client
915 mount_client $MOUNT &
917 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
918 MOUNT_LUSTRE_PID=$(ps -ef | grep mount.lustre |
919 grep -v grep | awk '{print $2}')
920 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
922 ps --ppid $MOUNT_LUSTRE_PID
923 echo "waiting for mount to finish"
925 # "ctrl-c" sends SIGINT but it usually (in script) does not work on child process
926 # SIGTERM works but it does not spread to offspring processses
927 kill -s TERM $MOUNT_PID
928 kill -s TERM $MOUNT_LUSTRE_PID
929 # we can not wait $MOUNT_PID because it is not a child of this shell
935 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
937 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
938 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
941 [ -z "$PID1" -a -z "$PID2" ] && break
942 echo "waiting for mount to finish ... "
943 WAIT=$(( WAIT + sleep))
945 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
946 error "MOUNT_PID $MOUNT_PID and "\
947 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
950 cleanup || error "cleanup failed with rc $?"
952 run_test 23a "interrupt client during recovery mount delay"
954 test_23b() { # was test_23
955 start_mds || error "MDS start failed"
956 start_ost || error "Unable to start OST1"
957 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
958 $LCTL set_param fail_loc=0x80000313
960 cleanup || error "cleanup failed with rc $?"
962 run_test 23b "Simulate -EINTR during mount"
965 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
967 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
968 is_blkdev $SINGLEMDS $MDSDEV &&
969 skip_env "mixed loopback and real device not working"
972 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
974 local fs2mdsdev=$(mdsdevname 1_2)
975 local fs2ostdev=$(ostdevname 1_2)
976 local fs2mdsvdev=$(mdsvdevname 1_2)
977 local fs2ostvdev=$(ostvdevname 1_2)
980 # LU-9733 test fsname started with numbers as well
981 local FSNAME2=969362ae
983 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --nomgs --mgsnode=$MGSNID \
984 --fsname=${FSNAME2} --reformat $fs2mdsdev $fs2mdsvdev || exit 10
986 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --fsname=${FSNAME2} \
987 --reformat $fs2ostdev $fs2ostvdev || exit 10
990 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
991 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
992 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
993 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
995 # LU-9733 test fsname started with numbers
996 cl_user=$(do_facet $SINGLEMDS lctl --device $FSNAME2-MDT0000 \
997 changelog_register -n) ||
998 error "register changelog failed"
1000 do_facet $SINGLEMDS lctl --device $FSNAME2-MDT0000 \
1001 changelog_deregister $cl_user ||
1002 error "deregister changelog failed"
1004 check_mount || error "check_mount failed"
1005 # files written on 1 should not show up on 2
1006 cp /etc/passwd $DIR/$tfile
1008 [ -e $MOUNT2/$tfile ] && error "File bleed"
1011 cp /etc/passwd $MOUNT2/$tfile ||
1012 error "cp /etc/passwd $MOUNT2/$tfile failed"
1013 rm $MOUNT2/$tfile || error "remove $MOUNT2/$tfile failed"
1014 # 2 is actually mounted
1015 grep $MOUNT2' ' /proc/mounts > /dev/null || error "$MOUNT2 not mounted"
1017 facet_failover fs2mds
1018 facet_failover fs2ost
1020 umount_client $MOUNT
1021 # the MDS must remain up until last MDT
1023 MDS=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
1024 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
1025 [ -z "$MDS" ] && error "No MDT"
1027 cleanup_nocli || error "cleanup_nocli failed with rc $?"
1029 run_test 24a "Multiple MDTs on a single node"
1032 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1034 skip "needs combined MGT and MDT device"
1036 if [ -z "$fs2mds_DEV" ]; then
1037 local dev=${SINGLEMDS}_dev
1038 local MDSDEV=${!dev}
1039 is_blkdev $SINGLEMDS $MDSDEV &&
1040 skip_env "mixed loopback and real device not working"
1043 local fs2mdsdev=$(mdsdevname 1_2)
1044 local fs2mdsvdev=$(mdsvdevname 1_2)
1046 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --mgs --fsname=${FSNAME}2 \
1047 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
1049 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && {
1051 error "start MDS should fail"
1054 cleanup || error "cleanup failed with rc $?"
1056 run_test 24b "Multiple MGSs on a single node (should return err)"
1060 check_mount || error "check_mount failed"
1061 local MODULES=$($LCTL modules | awk '{ print $2 }')
1062 rmmod $MODULES 2>/dev/null || true
1063 cleanup || error "cleanup failed with $?"
1065 run_test 25 "Verify modules are referenced"
1069 # we need modules before mount for sysctl, so make sure...
1070 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
1071 #define OBD_FAIL_MDS_FS_SETUP 0x135
1072 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000135"
1073 start_mds && error "MDS started but should not have started"
1074 $LCTL get_param -n devices
1075 DEVS=$($LCTL get_param -n devices | egrep -v MG | wc -l)
1076 [ $DEVS -gt 0 ] && error "number of devices is $DEVS, should be zero"
1077 # start mds to drop writeconf setting
1078 start_mds || error "Unable to start MDS"
1079 stop_mds || error "Unable to stop MDS"
1080 unload_modules_conf || error "unload_modules_conf failed with $?"
1082 run_test 26 "MDT startup failure cleans LOV (should return errs)"
1086 start_ost || error "Unable to start OST1"
1087 start_mds || error "Unable to start MDS"
1088 echo "Requeue thread should have started: "
1089 ps -e | grep ll_cfg_requeue
1090 set_persistent_param_and_check ost1 \
1091 "obdfilter.$FSNAME-OST0000.client_cache_seconds" \
1092 "$FSNAME-OST0000.ost.client_cache_seconds"
1093 cleanup_nocli || error "cleanup_nocli failed with rc $?"
1095 run_test 27a "Reacquire MGS lock if OST started first"
1100 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
1101 awk '($3 ~ "mdt" && $4 ~ "MDT0000") { print $4 }')
1103 facet_failover $SINGLEMDS
1104 set_persistent_param_and_check $SINGLEMDS \
1105 "mdt.$device.identity_acquire_expire" \
1106 "$device.mdt.identity_acquire_expire"
1107 set_persistent_param_and_check client \
1108 "mdc.$device-mdc-*.max_rpcs_in_flight" \
1109 "$device.mdc.max_rpcs_in_flight"
1111 cleanup || error "cleanup failed with $?"
1113 run_test 27b "Reacquire MGS lock after failover"
1115 test_28A() { # was test_28
1118 local TEST="llite.$FSNAME-*.max_read_ahead_whole_mb"
1119 local PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
1120 local orig=$($LCTL get_param -n $TEST)
1121 local max=$($LCTL get_param -n \
1122 llite.$FSNAME-*.max_read_ahead_per_file_mb)
1124 orig=${orig%%.[0-9]*}
1126 echo "ORIG:$orig MAX:$max"
1127 [[ $max -le $orig ]] && orig=$((max - 3))
1128 echo "ORIG:$orig MAX:$max"
1130 local final=$((orig + 1))
1132 set_persistent_param_and_check client "$TEST" "$PARAM" $final
1133 final=$((final + 1))
1134 set_persistent_param_and_check client "$TEST" "$PARAM" $final
1135 umount_client $MOUNT || error "umount_client $MOUNT failed"
1136 mount_client $MOUNT || error "mount_client $MOUNT failed"
1138 local result=$($LCTL get_param -n $TEST)
1140 if [ $result -ne $final ]; then
1141 error "New config not seen: wanted $final got $result"
1143 echo "New config success: got $result"
1145 set_persistent_param_and_check client "$TEST" "$PARAM" $orig
1146 cleanup || error "cleanup failed with rc $?"
1148 run_test 28A "permanent parameter setting"
1150 test_28a() { # LU-4221
1151 [[ "$OST1_VERSION" -ge $(version_code 2.5.52) ]] ||
1152 skip "Need OST version at least 2.5.52"
1153 [ "$ost1_FSTYPE" = zfs ] &&
1154 skip "LU-4221: no such proc params for ZFS OSTs"
1161 local device="$FSNAME-OST0000"
1165 # In this test we will set three kinds of proc parameters with
1166 # lctl set_param -P or lctl conf_param:
1167 # 1. non-symlink ones in the OFD
1168 # 2. non-symlink ones in the OSD
1171 # prepare a non-symlink parameter in the OFD
1172 name="client_cache_seconds"
1173 param="$device.ost.$name"
1174 cmd="obdfilter.$device.$name"
1176 # permanently setting the parameter in the OFD
1177 old=$(do_facet ost1 $LCTL get_param -n $cmd)
1179 set_persistent_param_and_check ost1 "$cmd" "$param" $new
1180 set_persistent_param_and_check ost1 "$cmd" "$param" $old
1183 # prepare a non-symlink parameter in the OSD
1185 param="$device.osd.$name"
1186 cmd="osd-*.$device.$name"
1188 # conf_param the parameter in the OSD
1189 old=$(do_facet ost1 $LCTL get_param -n $cmd)
1190 new=$(((old + 1) % 2))
1191 set_persistent_param_and_check ost1 "$cmd" "$param" $new
1192 set_persistent_param_and_check ost1 "$cmd" "$param" $old
1194 cleanup || error "cleanup failed with $?"
1196 run_test 28a "set symlink parameters permanently with lctl"
1199 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >= 2 OSTs"
1200 setup_noconfig > /dev/null 2>&1
1201 start_ost2 || error "Unable to start OST2"
1204 local PARAM="$FSNAME-OST0001.osc.active"
1205 # With lctl set_param -P the value $PROC_ACT will be sent to
1206 # all nodes. The [!M] filter out the ability to set active
1207 # on the MDS servers which is tested with wait_osp_* below.
1208 # For ost_server_uuid that only exist on client so filtering
1210 local PROC_ACT="os[cp].$FSNAME-OST0001-osc-[^M]*.active"
1211 local PROC_UUID="os[cp].$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
1213 ACTV=$($LCTL get_param -n $PROC_ACT)
1215 set_persistent_param_and_check client $PROC_ACT $PARAM $DEAC
1216 # also check ost_server_uuid status
1217 RESULT=$($LCTL get_param -n $PROC_UUID | grep DEACTIV)
1218 if [ -z "$RESULT" ]; then
1219 error "Client not deactivated: $($LCTL get_param \
1222 echo "Live client success: got $RESULT"
1226 wait_osp_active ost ${FSNAME}-OST0001 1 0
1228 # test new client starts deactivated
1229 umount_client $MOUNT || error "umount_client $MOUNT failed"
1230 mount_client $MOUNT || error "mount_client $MOUNT failed"
1232 # the 2nd and 3rd field of ost_server_uuid do not update at the same
1233 # time when using lctl set_param -P
1234 wait_update_facet client \
1235 "$LCTL get_param -n $PROC_UUID | awk '{print \\\$3 }'" \
1237 error "New client start active: $($LCTL get_param -n $PROC_UUID)"
1239 echo "New client success: got '$($LCTL get_param -n $PROC_UUID)'"
1241 # make sure it reactivates
1242 set_persistent_param_and_check client $PROC_ACT $PARAM $ACTV
1244 umount_client $MOUNT
1245 stop_ost2 || error "Unable to stop OST2"
1246 cleanup_nocli || error "cleanup_nocli failed with $?"
1247 #writeconf to remove all ost2 traces for subsequent tests
1248 writeconf_or_reformat
1250 run_test 29 "permanently remove an OST"
1255 echo Big config llog
1256 local path="llite.$FSNAME-*"
1257 local cpath="$FSNAME.llite"
1258 local param="max_read_ahead_whole_mb"
1260 local test="${path}.$param"
1261 local conf="${cpath}.$param"
1263 local orig=$($LCTL get_param -n $test)
1264 local list=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1265 for i in ${list[@]}; do
1266 set_persistent_param_and_check client $test $conf $i
1268 # make sure client restart still works
1269 umount_client $MOUNT
1270 mount_client $MOUNT || error "mount_client $MOUNT failed"
1271 [ "$($LCTL get_param -n $test)" -ne "$i" ] &&
1272 error "Param didn't stick across restart $($test) != $i"
1275 echo Erase parameter setting
1276 if [[ $PERM_CMD == *"set_param -P"* ]]; then
1277 do_facet mgs "$PERM_CMD -d $test" ||
1278 error "Erase param $test failed"
1280 do_facet mgs "$PERM_CMD -d $conf" ||
1281 error "Erase param $conf failed"
1283 umount_client $MOUNT
1284 mount_client $MOUNT || error "mount_client $MOUNT failed"
1285 local final=$($LCTL get_param -n $test)
1286 echo "deleted (default) value=$final, orig=$orig"
1287 orig=${orig%%.[0-9]*}
1288 final=${final%%.[0-9]*}
1289 # assumes this parameter started at the default value
1290 [ "$final" -eq "$orig" ] ||
1291 error "Deleted value=$final -ne orig=$orig"
1293 cleanup || error "cleanup failed with rc $?"
1295 run_test 30a "Big config llog and permanent parameter deletion"
1300 local orignids=$($LCTL get_param -n \
1301 osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1303 local orignidcount=$(echo "$orignids" | wc -w)
1305 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1306 # numerical part of it. Hopefully that's not already a failover address
1308 local OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | \
1310 local ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1311 local NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1312 local NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1313 echo "Using fake nid $NEW"
1315 local TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1316 grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1317 if [[ $PERM_CMD == *"set_param -P"* ]]; then
1318 PARAM="osc.$FSNAME-OST0000-osc-[^M]*.import"
1319 echo "Setting $PARAM from $TEST to $NEW"
1320 do_facet mgs "$PERM_CMD $PARAM='connection=$NEW'" ||
1321 error "$PERM_CMD $PARAM failed"
1323 PARAM="$FSNAME-OST0000.failover.node"
1324 echo "Setting $PARAM from $TEST to $NEW"
1325 do_facet mgs "$PERM_CMD $PARAM='$NEW'" ||
1326 error "$PARAM $PARAM failed"
1328 wait_update_facet client "$TEST" "$NEW" ||
1329 error "check $PARAM failed!"
1331 local NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1333 local NIDCOUNT=$(echo "$NIDS" | wc -w)
1334 echo "should have $((orignidcount + 1)) entries \
1335 in failover nids string, have $NIDCOUNT"
1336 [ $NIDCOUNT -eq $((orignidcount + 1)) ] ||
1337 error "Failover nid not added"
1339 if [[ $PERM_CMD == *"set_param -P"* ]]; then
1340 do_facet mgs "$PERM_CMD -d osc.$FSNAME-OST0000-osc-*.import"
1342 do_facet mgs "$PERM_CMD -d $FSNAME-OST0000.failover.node" ||
1343 error "$PERM_CMD delete failed"
1345 umount_client $MOUNT
1346 mount_client $MOUNT || error "mount_client $MOUNT failed"
1348 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1350 NIDCOUNT=$(echo "$NIDS" | wc -w)
1351 echo "only $orignidcount final entries should remain \
1352 in failover nids string, have $NIDCOUNT"
1353 [ $NIDCOUNT -eq $orignidcount ] || error "Failover nids not removed"
1355 cleanup || error "cleanup failed with rc $?"
1357 run_test 30b "Remove failover nids"
1359 test_31() { # bug 10734
1360 # ipaddr must not exist
1361 $MOUNT_CMD 4.3.2.1@tcp:/lustre $MOUNT || true
1362 cleanup || error "cleanup failed with rc $?"
1364 run_test 31 "Connect to non-existent node (shouldn't crash)"
1368 T32_BLIMIT=40960 # Kbytes
1371 T32_PROLIMIT=$((T32_BLIMIT/10))
1373 # This is not really a test but a tool to create new disk
1374 # image tarballs for the upgrade tests.
1376 # Disk image tarballs should be created on single-node
1377 # clusters by running this test with default configurations
1378 # plus a few mandatory environment settings that are verified
1379 # at the beginning of the test.
1381 test_32newtarball() {
1385 local tmp=$TMP/t32_image_create
1396 local pj_quota_file_old
1399 if [ $FSNAME != t32fs -o \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o \
1400 $OSTCOUNT -ne 2 -o -z "$OSTDEV1" ]; then
1401 error "Needs FSNAME=t32fs MDSCOUNT=2 " \
1402 "MDSDEV1=<nonexistent_file> " \
1403 "MDSDEV2=<nonexistent_file> " \
1404 "(or MDSDEV, in the case of b1_8) " \
1405 "OSTCOUNT=2 OSTDEV1=<nonexistent_file> " \
1406 "OSTDEV2=<nonexistent_file>"
1410 echo "Found stale $tmp"
1414 mkdir $tmp/src || return 1
1415 tar cf - -C $src . | tar xf - -C $tmp/src
1417 # format ost with comma-separated NIDs to verify LU-4460
1418 local failnid="$(h2nettype 1.2.3.4),$(h2nettype 4.3.2.1)"
1419 MGSNID="$MGSNID,$MGSNID" OSTOPT="--failnode=$failnid" formatall
1423 [[ "$MDS1_VERSION" -ge $(version_code 2.3.50) ]] ||
1424 $LFS quotacheck -ug /mnt/$FSNAME
1425 $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
1428 if [[ $MDSCOUNT -ge 2 ]]; then
1429 remote_dir=/mnt/$FSNAME/remote_dir
1430 $LFS mkdir -i 1 $remote_dir
1431 tar cf - -C $tmp/src . | tar xf - -C $remote_dir
1433 target_dir=$remote_dir
1434 if [[ $MDS1_VERSION -ge $(version_code 2.7.0) ]]; then
1435 striped_dir=/mnt/$FSNAME/striped_dir_old
1436 $LFS mkdir -i 1 -c 2 $striped_dir
1437 tar cf - -C $tmp/src . | tar xf - -C $striped_dir
1440 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1441 target_dir=/mnt/$FSNAME
1444 # add project quota #
1445 [[ $mds1_FSTYPE == "ldiskfs" &&
1446 $MDS1_VERSION -gt $(version_code 2.9.55) ]] ||
1447 [[ $mds1_FSTYPE == "zfs" &&
1448 $MDS1_VERSION -gt $(version_code 2.10.53) ]] && {
1449 pj_quota_dir=$target_dir/project_quota_dir
1450 pj_quota_file_old=$pj_quota_dir/pj_quota_file_old
1452 enable_project_quota
1453 set_mdt_qtype ugp ||
1454 error "enable mdt quota failed"
1456 set_ost_qtype ugp ||
1457 error "enable ost quota failed"
1459 mkdir -p $pj_quota_dir
1460 $LFS setquota -p $T32_PRJID -b 0 -B $T32_PROLIMIT -i 0 \
1461 -I $T32_ILIMIT $pj_quota_dir ||
1462 error "setquota -p $T32_PRJID failed"
1464 $LFS setstripe $pj_quota_file_old -c 1 ||
1465 error "setstripe $pj_quota_file_old failed"
1467 chown $T32_QID:$T32_QID $pj_quota_file_old ||
1468 error "chown $pj_quota_file_old failed"
1470 change_project -p $T32_PRJID $pj_quota_file_old
1472 mkdir -p $tmp/src/project_quota_dir
1473 cp $pj_quota_file_old $tmp/src/project_quota_dir/
1476 #####################
1477 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1479 #if [[ $MDSCOUNT -ge 2 ]]; then
1480 # remote_dir=/mnt/$FSNAME/remote_dir
1481 # $LFS mkdir -i 1 $remote_dir
1482 # tar cf - -C $tmp/src . | tar xf - -C $remote_dir
1484 # if [[ "$MDS1_VERSION" -ge $(version_code 2.7.0) ]]; then
1485 # striped_dir=/mnt/$FSNAME/striped_dir_old
1486 # $LFS mkdir -i 1 -c 2 $striped_dir
1487 # tar cf - -C $tmp/src . | tar xf - -C $striped_dir
1492 if [[ $MDS1_VERSION -ge $(version_code 2.9.51) ]]; then
1493 pfl_dir=$target_dir/pfl_dir
1494 pfl_file=$pfl_dir/pfl_file
1496 $LFS setstripe -E 2M -c 1 -o 0 -E -1 -S 2M -c 1 -o 1 \
1498 error "Create PFL file failed"
1500 dd if=/dev/urandom of=$pfl_file bs=1k count=3k
1501 mkdir -p $tmp/src/pfl_dir
1502 cp $pfl_file $tmp/src/pfl_dir/
1507 if [[ $MDS1_VERSION -ge $(version_code 2.10.56) ]]; then
1508 dom_dir=$target_dir/dom_dir
1509 dom_file=$dom_dir/dom_file
1510 flr_dir=$target_dir/flr_dir
1511 flr_file=$flr_dir/flr_file
1514 $LFS setstripe -E 1M -L mdt -E -1 -S 4M $dom_file
1515 dd if=/dev/urandom of=$dom_file bs=1k count=2k
1516 mkdir -p $tmp/src/dom_dir
1517 cp $dom_file $tmp/src/dom_dir
1520 $LFS mirror create -N2 $flr_file
1521 dd if=/dev/urandom of=$flr_file bs=1k count=1
1522 mkdir -p $tmp/src/flr_dir
1523 cp $flr_file $tmp/src/flr_dir
1527 local large_xattr_dir=/mnt/$FSNAME/large_xattr_test_dir
1528 local xattr_file=$large_xattr_dir/large_xattr_file
1530 mkdir $large_xattr_dir
1532 setfattr -n user.fooattr -v $(printf "%c" {1..4096} ) $xattr_file ||
1537 mkdir $tmp/img || return 1
1541 pushd_dir=/mnt/$FSNAME
1542 if [[ $MDSCOUNT -ge 2 ]]; then
1543 pushd_dir=$remote_dir
1544 if [[ "$MDS1_VERSION" -ge $(version_code 2.7.0) ]]; then
1546 BLOCKSIZE=1024 ls -Rni --time-style=+%s >$tmp/img/list2
1552 BLOCKSIZE=1024 ls -Rni --time-style=+%s >$tmp/img/list
1553 find ! -name .lustre -type f -exec sha1sum {} \; |
1554 sort -k 2 >$tmp/img/sha1sums
1556 $LCTL get_param -n version | head -n 1 |
1557 sed -e 's/^lustre: *//' >$tmp/img/commit
1559 [[ "$MDS1_VERSION" -ge $(version_code 2.3.50) ]] ||
1560 $LFS quotaon -ug /mnt/$FSNAME
1561 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1562 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1563 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1564 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1565 | tr -d "*" > $tmp/img/bspace
1566 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1567 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1568 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1569 | tr -d "*" > $tmp/img/ispace
1570 echo $T32_BLIMIT > $tmp/img/blimit
1571 echo $T32_ILIMIT > $tmp/img/ilimit
1573 $MULTIOP /mnt/$FSNAME/orph_file Ouw_c&
1577 debugfs -R "ls /PENDING" ${MDSDEV1:-$MDSDEV}
1578 cp ${MDSDEV1:-$MDSDEV} $tmp/img
1587 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1590 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1591 echo "Data verification failed"
1594 uname -r >$tmp/img/kernel
1595 uname -m >$tmp/img/arch
1597 for ((num=1; num <= $MDSCOUNT; num++)); do
1598 local devname=$(mdsdevname $num)
1602 [[ num -eq 1 ]] && image_name=mdt || image_name=mdt$num
1603 [[ $(facet_fstype $facet) != zfs ]] ||
1604 devname=$(mdsvdevname $num)
1605 dd conv=sparse bs=4k if=$devname of=$tmp/img/$image_name
1608 for ((num=1; num <= $OSTCOUNT; num++)); do
1609 local devname=$(ostdevname $num)
1613 [[ num -eq 1 ]] && image_name=ost || image_name=ost$num
1614 [[ $(facet_fstype $facet) != zfs ]] ||
1615 devname=$(ostdevname $num)
1616 dd conv=sparse bs=4k if=$devname of=$tmp/img/$image_name
1619 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1620 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1623 tar cjvf $dst/disk$version-"$mds1_FSTYPE".tar.bz2 -S *
1628 # run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1631 # The list of applicable tarballs is returned via the caller's
1632 # variable "tarballs".
1635 [ "$CLIENTONLY" ] && skip "Client-only testing"
1637 local node=$(facet_active_host $SINGLEMDS)
1638 local r="do_node $node"
1640 ! $r which "$TUNEFS" && skip_env "tunefs.lustre required on $node"
1642 local IMGTYPE="$mds1_FSTYPE"
1644 [[ -n "$T32IMAGE" ]] && tarballs=$T32IMAGE ||
1645 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 \
1646 -name \'disk*-$IMGTYPE.tar.bz2\')
1648 [ -z "$tarballs" ] && skip "No applicable tarballs found"
1651 t32_test_cleanup() {
1653 local facet=$SINGLEMDS
1656 echo "== cleanup with rc=$rc =="
1657 if $shall_cleanup_lustre; then
1658 umount $tmp/mnt/lustre || rc=$?
1660 if $shall_cleanup_mdt; then
1661 $r $UMOUNT $tmp/mnt/mdt || rc=$?
1663 if $shall_cleanup_mdt1; then
1664 $r $UMOUNT $tmp/mnt/mdt1 || rc=$?
1666 if $shall_cleanup_ost; then
1667 $r $UMOUNT $tmp/mnt/ost || rc=$?
1669 if $shall_cleanup_ost1; then
1670 $r $UMOUNT $tmp/mnt/ost1 || rc=$?
1675 if [[ "$mds1_FSTYPE" == zfs ]]; then
1677 local poolname_list="t32fs-mdt1 t32fs-ost1"
1679 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1681 for poolname in $poolname_list; do
1682 destroy_zpool $facet $poolname
1685 combined_mgs_mds || start_mgs || rc=$?
1689 t32_bits_per_long() {
1691 # Yes, this is not meant to be perfect.
1701 t32_reload_modules() {
1703 local all_removed=false
1706 [ "$mds1_FSTYPE" == zfs ] && do_rpc_nodes $node "service zed stop"
1708 while ((i < 20)); do
1709 echo "Unloading modules on $node: Attempt $i"
1710 do_rpc_nodes $node $LUSTRE_RMMOD "$mds1_FSTYPE" &&
1712 do_rpc_nodes $node check_mem_leak || return 1
1713 if $all_removed; then
1714 do_rpc_nodes $node load_modules
1717 if [ "$mds1_FSTYPE" == zfs ]; then
1718 do_rpc_nodes $node "$ZPOOL status -v"
1723 echo "Unloading modules on $node: Given up"
1727 t32_wait_til_devices_gone() {
1733 echo wait for devices to go
1734 while ((i < 20)); do
1735 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1736 loops=$(do_rpc_nodes $node losetup -a | grep -c t32)
1737 ((devices == 0 && loops == 0)) && return 0
1741 echo "waiting for dev on $node: dev $devices loop $loops given up"
1742 do_rpc_nodes $node "losetup -a"
1743 do_rpc_nodes $node "$LCTL device_list"
1747 t32_verify_quota() {
1754 # LU-2435: if the underlying zfs doesn't support userobj_accounting,
1755 # lustre will estimate the object count usage. This fails quota
1756 # verification in 32b. The object quota usage should be accurate after
1757 # zfs-0.7.0 is released.
1758 [ "$mds1_FSTYPE" == zfs ] && {
1759 local zfs_version=$(do_facet $facet cat /sys/module/zfs/version)
1761 [ $(version_code $zfs_version) -lt $(version_code 0.7.0) ] && {
1762 echo "Skip quota verify for zfs: $zfs_version"
1767 $LFS quota -u $T32_QID -v $mnt
1769 qval=$($LFS quota -v -u $T32_QID $mnt |
1770 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1771 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1773 [ $qval -eq $img_bspace ] || {
1774 echo "bspace, act:$qval, exp:$img_bspace"
1778 qval=$($LFS quota -v -u $T32_QID $mnt |
1779 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1780 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1782 [ $qval -eq $img_ispace ] || {
1783 echo "ispace, act:$qval, exp:$img_ispace"
1787 qval=$($LFS quota -v -u $T32_QID $mnt |
1788 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1789 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1791 [ $qval -eq $img_blimit ] || {
1792 echo "blimit, act:$qval, exp:$img_blimit"
1796 qval=$($LFS quota -v -u $T32_QID $mnt |
1797 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1798 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1800 [ $qval -eq $img_ilimit ] || {
1801 echo "ilimit, act:$qval, exp:$img_ilimit"
1805 set_persistent_param_and_check $facet \
1806 "osd-$mds1_FSTYPE.$fsname-MDT0000.quota_slave.enabled" \
1807 "$fsname.quota.mdt" ug
1809 set_persistent_param_and_check $facet \
1810 "osd-$mds1_FSTYPE.$fsname-OST0000.quota_slave.enabled" \
1811 "$fsname.quota.ost" ug
1814 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1815 bs=1M count=$((img_blimit / 1024)) oflag=sync && {
1816 echo "Write succeed, but expect -EDQUOT"
1819 rm -f $mnt/t32_qf_new
1821 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1823 echo "Create succeed, but expect -EDQUOT"
1826 unlinkmany $mnt/t32_qf_ $img_ilimit
1836 sync_all_data > /dev/null 2>&1 || true
1838 [ "$uuid" != "global" ] || uuid=$mnt
1840 $LFS quota -v "$1" "$2" $mnt |
1841 awk 'BEGIN { num='$spec' } { if ($1 == "'$uuid'") \
1842 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1849 local dne_upgrade=${dne_upgrade:-"no"}
1850 local dom_upgrade=${dom_upgrade:-"no"}
1851 local project_upgrade=${project_upgrade:-"no"}
1852 local ff_convert=${ff_convert:-"no"}
1853 local pfl_upgrade=${pfl_upgrade:-"no"}
1854 local project_quota_upgrade=${project_quota_upgrade:-"no"}
1855 local dom_new_upgrade=${dom_new_upgrade:-"no"}
1856 local flr_upgrade=${flr_upgrade:-"no"}
1857 local shall_cleanup_mdt=false
1858 local shall_cleanup_mdt1=false
1859 local shall_cleanup_ost=false
1860 local shall_cleanup_ost1=false
1861 local shall_cleanup_lustre=false
1862 local mdt2_is_available=false
1863 local ost2_is_available=false
1864 local node=$(facet_active_host $SINGLEMDS)
1865 local r="do_node $node"
1881 local mdt_dev=$tmp/mdt
1882 local mdt2_dev=$tmp/mdt2
1883 local ost_dev=$tmp/ost
1884 local ost2_dev=$tmp/ost2
1888 local pfl_file=$tmp/mnt/lustre/remote_dir/pfl_dir/pfl_file
1889 local flr_file=$tmp/mnt/lustre/remote_dir/flr_dir/flr_file
1890 local dom_file=$tmp/mnt/lustre/remote_dir/dom_dir/dom_file
1891 local quota_dir=$tmp/mnt/lustre/remote_dir/project_quota_dir
1893 combined_mgs_mds || stop_mgs || error "Unable to stop MGS"
1894 trap 'trap - RETURN; t32_test_cleanup' RETURN
1897 nid=$($r $LCTL list_nids | head -1)
1899 mkdir -p $tmp/mnt/lustre || error "mkdir $tmp/mnt/lustre failed"
1900 $r mkdir -p $tmp/mnt/{mdt,mdt1,ost,ost1}
1901 $r tar xjvf $tarball -S -C $tmp || {
1902 error_noexit "Unpacking the disk image tarball"
1905 img_commit=$($r cat $tmp/commit)
1906 img_kernel=$($r cat $tmp/kernel)
1907 img_arch=$($r cat $tmp/arch)
1908 img_bspace=$($r cat $tmp/bspace)
1909 img_ispace=$($r cat $tmp/ispace)
1911 # older images did not have "blimit" and "ilimit" files
1912 # use old values for T32_BLIMIT and T32_ILIMIT
1913 $r test -f $tmp/blimit && img_blimit=$($r cat $tmp/blimit) ||
1915 $r test -f $tmp/ilimit && img_ilimit=$($r cat $tmp/ilimit) ||
1918 echo "Upgrading from $(basename $tarball), created with:"
1919 echo " Commit: $img_commit"
1920 echo " Kernel: $img_kernel"
1921 echo " Arch: $img_arch"
1922 echo "OST version: $(lustre_build_version ost1)"
1924 # The conversion can be made only when both of the following
1925 # conditions are satisfied:
1926 # - ost device img version < 2.3.64
1927 # - ost server version >= 2.5
1928 [ $(version_code $img_commit) -ge $(version_code 2.3.64) -o \
1929 "$OST1_VERSION" -lt $(version_code 2.5.0) ] &&
1932 ! $r test -f $mdt2_dev || mdt2_is_available=true
1933 ! $r test -f $ost2_dev || ost2_is_available=true
1935 if [[ "$mds1_FSTYPE" == zfs ]]; then
1936 echo "== fstype is zfs =="
1939 local poolname_list="t32fs-mdt1 t32fs-ost1"
1941 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1943 for poolname in $poolname_list; do
1945 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
1946 $ZPOOL import -f -d $tmp $poolname"
1949 # upgrade zpool to latest supported features, including
1950 # dnode quota accounting in 0.7.0
1951 $r "$ZPOOL upgrade -a"
1953 mdt_dev=t32fs-mdt1/mdt1
1954 ost_dev=t32fs-ost1/ost1
1955 ! $mdt2_is_available || mdt2_dev=t32fs-mdt2/mdt2
1956 ! $ost2_is_available || ost2_dev=t32fs-ost2/ost2
1957 wait_update_facet $SINGLEMDS "$ZPOOL list |
1958 awk '/^t32fs-mdt1/ { print \\\$1 }'" "t32fs-mdt1" || {
1959 error_noexit "import zfs pool failed"
1962 elif [ "$project_upgrade" != "no" ]; then
1963 echo "== project upgrade =="
1964 ! $r tune2fs -O project $mdt_dev &&
1965 error_noexit "enable project on mdt0 failed" &&
1967 $mdt2_is_available && ! $r tune2fs "-O project" $mdt2_dev &&
1968 error_noexit "enable project on mdt failed" &&
1970 ! $r tune2fs -O project $ost_dev &&
1971 error_noexit "enable project on ost failed" &&
1973 $ost2_is_available && ! $r tune2fs -O project $ost2_dev &&
1974 error_noexit "enable project on ost2 failed" &&
1978 $r $LCTL set_param debug="$PTLDEBUG"
1980 $r $TUNEFS --dryrun $mdt_dev || {
1982 error_noexit "tunefs.lustre before mounting the MDT"
1986 if $mdt2_is_available; then
1987 $r $TUNEFS --dryrun $mdt2_dev || {
1989 error_noexit "tunefs.lustre before mounting the MDT"
1994 if [ "$writeconf" ]; then
1995 echo "== writeconf mdt tunefs and quota =="
1997 if [ "$mds1_FSTYPE" == ldiskfs ]; then
1999 $r $TUNEFS --quota $mdt_dev || {
2001 error_noexit "Enable mdt quota feature"
2005 if [ "$project_quota_upgrade" != "no" ]; then
2006 echo "== enable mdt2 project quota =="
2007 $r $TUNE2FS -O project $mdt_dev || {
2009 error_noexit "tune2fs $mdt_dev failed"
2014 if $mdt2_is_available; then
2015 $r $TUNEFS --quota $mdt2_dev || {
2017 error_noexit "Enable mdt2 quota feature"
2020 if [ "$project_quota_upgrade" != "no" ]; then
2021 echo "== enable mdt2 project quota =="
2022 $r $TUNE2FS -O project $mdt2_dev || {
2025 "tune2fs $mdt2_dev failed"
2032 echo "== replace nids =="
2033 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
2034 [[ "$MGS_VERSION" -ge $(version_code 2.3.59) ]] ||
2035 skip "LU-2200: Cannot run over IB w/o lctl replace_nids "
2036 "(Need MGS version at least 2.3.59)"
2039 local osthost=$(facet_active_host ost1)
2040 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
2043 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2046 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt
2047 $r $LCTL replace_nids $fsname-OST0000 $ostnid || {
2048 error_noexit "replace_nids $fsname-OST0000 $ostnid failed"
2051 if $ost2_is_available; then
2052 $r $LCTL replace_nids $fsname-OST0001 $ostnid || {
2053 error_noexit "replace_nids $fsname-OST0001 $ostnid failed"
2057 $r $LCTL replace_nids $fsname-MDT0000 $nid || {
2058 error_noexit "replace_nids $fsname-MDT0000 $nid failed"
2061 if $mdt2_is_available; then
2062 $r $LCTL replace_nids $fsname-MDT0001 $nid || {
2063 error_noexit "replace_nids $fsname-MDT0001 $nid failed"
2067 $r $UMOUNT $tmp/mnt/mdt
2069 mopts=exclude=$fsname-OST0000
2070 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2075 t32_wait_til_devices_gone $node
2077 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2079 error_noexit "Mounting the MDT"
2082 shall_cleanup_mdt=true
2084 if $mdt2_is_available; then
2085 echo "== mdt2 available =="
2086 mopts=mgsnode=$nid,$mopts
2087 $r $MOUNT_CMD -o $mopts $mdt2_dev $tmp/mnt/mdt1 || {
2089 error_noexit "Mounting the MDT"
2092 shall_cleanup_mdt1=true
2093 echo "== mount new MDT $mdt2_dev =="
2095 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 || {
2096 error_noexit "enable remote dir create failed"
2100 elif [ "$dne_upgrade" != "no" ]; then
2101 local fs2mdsdev=$(mdsdevname 1_2)
2102 local fs2mdsvdev=$(mdsvdevname 1_2)
2104 echo "== mkfs new MDT on ${fs2mdsdev} =="
2105 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2106 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
2109 add $SINGLEMDS $(mds2failover_HOST="" \
2110 mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
2111 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
2112 error_noexit "Mkfs new MDT failed"
2116 [[ "$mds1_FSTYPE" != zfs ]] || import_zpool mds1
2118 $r $TUNEFS --dryrun $fs2mdsdev || {
2119 error_noexit "tunefs.lustre before mounting the MDT"
2123 echo "== mount new MDT....$fs2mdsdev =="
2124 $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
2125 error_noexit "mount mdt1 failed"
2129 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
2130 error_noexit "enable remote dir create failed"
2132 shall_cleanup_mdt1=true
2135 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
2136 error_noexit "Getting MDT UUID"
2139 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
2140 error_noexit "Unexpected MDT UUID: \"$uuid\""
2144 $r $TUNEFS --dryrun $ost_dev || {
2145 error_noexit "tunefs.lustre before mounting the OST"
2149 if $ost2_is_available; then
2150 $r $TUNEFS --dryrun $ost2_dev || {
2151 error_noexit "tunefs.lustre before mounting the OST"
2156 if [ "$writeconf" ]; then
2157 echo "== writeconf and ost quota tunefs =="
2158 mopts=mgsnode=$nid,$writeconf
2159 if [ "$ost1_FSTYPE" == ldiskfs ]; then
2161 $r $TUNEFS --quota $ost_dev || {
2163 error_noexit "Enable ost quota feature"
2167 if [ "$project_quota_upgrade" != "no" ]; then
2168 echo "== enable ost project quota =="
2169 $r $TUNE2FS -O project $ost_dev || {
2171 error_noexit "tune2fs $ost_dev failed"
2176 if $ost2_is_available; then
2177 $r $TUNEFS --quota $ost2_dev || {
2179 error_noexit "Enable ost2 quota feature"
2182 if [ "$project_quota_upgrade" != "no" ]; then
2183 echo "== enable ost2 project quota =="
2184 $r $TUNE2FS -O project $ost2_dev || {
2187 "tune2fs $ost2_dev failed"
2195 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2200 $r $MOUNT_CMD -onomgs -o$mopts $ost_dev $tmp/mnt/ost || {
2201 error_noexit "Mounting the OST"
2205 if $ost2_is_available; then
2206 $r $MOUNT_CMD -onomgs -o$mopts $ost2_dev $tmp/mnt/ost1 || {
2207 error_noexit "Mounting the OST2"
2210 shall_cleanup_ost1=true
2213 shall_cleanup_ost=true
2215 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
2216 error_noexit "Getting OST UUID"
2219 if [ "$uuid" != $fsname-OST0000_UUID ]; then
2220 error_noexit "Unexpected OST UUID: \"$uuid\""
2224 if $ost2_is_available; then
2225 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0001.uuid) ||{
2226 error_noexit "Getting OST1 UUID"
2229 if [ "$uuid" != $fsname-OST0001_UUID ]; then
2230 error_noexit "Unexpected OST1 UUID: \"$uuid\""
2235 if [[ $PERM_CMD =~ "set_param" ]]; then
2236 echo "== perm cmd = '$PERM_CMD' =="
2237 $r $PERM_CMD osc.$fsname-OST0000*.import=connection=$nid || {
2238 error_noexit "Setting OST \"failover.node\""
2241 $r $PERM_CMD mdc.$fsname-MDT0000*.import=connection=$nid || {
2242 error_noexit "Setting MDT \"failover.node\""
2245 $r $PERM_CMD osc.$fsname-OST0000-*.max_dirty_mb=15 || {
2246 error_noexit "Setting \"max_dirty_mb\""
2249 $r $PERM_CMD mdc.$fsname-MDT0000-*.max_rpcs_in_flight=9 || {
2250 error_noexit "Setting \"max_rpcs_in_flight\""
2253 $r $PERM_CMD lov.$fsname-MDT0000-*.stripesize=4M || {
2254 error_noexit "Setting \"lov.stripesize\""
2257 $r $PERM_CMD mdd.$fsname-MDT0000-*.atime_diff=70 || {
2258 error_noexit "Setting \"mdd.atime_diff\""
2262 echo "== perm cmd = '$PERM_CMD' =="
2263 $r $PERM_CMD $fsname-OST0000.failover.node=$nid || {
2264 error_noexit "Setting OST \"failover.node\""
2268 $r $PERM_CMD $fsname-MDT0000.failover.node=$nid || {
2269 error_noexit "Setting MDT \"failover.node\""
2273 $r $PERM_CMD $fsname-OST0000.osc.max_dirty_mb=15 || {
2274 error_noexit "Setting \"max_dirty_mb\""
2277 $r $PERM_CMD $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
2278 error_noexit "Setting \"max_rpcs_in_flight\""
2281 $r $PERM_CMD $fsname-MDT0000.lov.stripesize=4M || {
2282 error_noexit "Setting \"lov.stripesize\""
2285 $r $PERM_CMD $fsname-MDT0000.mdd.atime_diff=70 || {
2286 error_noexit "Setting \"mdd.atime_diff\""
2291 $r $LCTL pool_new $fsname.interop || {
2292 error_noexit "Setting \"interop\""
2296 if [ "$ff_convert" != "no" -a "$ost1_FSTYPE" == ldiskfs ]; then
2297 echo "== ff convert ($ff_convert) =="
2298 $r $LCTL lfsck_start -M $fsname-OST0000 || {
2299 error_noexit "Start OI scrub on OST0"
2303 # The oi_scrub should be on ost1, but for test_32(),
2304 # all on the SINGLEMDS.
2305 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
2306 osd-ldiskfs.$fsname-OST0000.oi_scrub |
2307 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
2308 error_noexit "Failed to get the expected 'completed'"
2312 local UPDATED=$($r $LCTL get_param -n \
2313 osd-ldiskfs.$fsname-OST0000.oi_scrub |
2314 awk '/^updated/ { print $2 }')
2315 [ $UPDATED -ge 1 ] || {
2316 error_noexit "Only $UPDATED objects have been converted"
2321 if [[ "$dne_upgrade" != "no" ]] || $mdt2_is_available; then
2322 echo "== dne upgrate ($dne_upgrade) or mdt2 available =="
2323 if [[ $PERM_CMD =~ "set_param" ]]; then
2324 $r $PERM_CMD mdc.$fsname-MDT0001*.import=connection=$nid || {
2325 error_noexit "Setting MDT1 \"failover.node\""
2329 $r $PERM_CMD mdc.$fsname-MDT0001-*.max_rpcs_in_flight=9 || {
2330 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
2333 $r $PERM_CMD lov.$fsname-MDT0001-*.stripesize=4M || {
2334 error_noexit "Setting MDT1 \"lov.stripesize\""
2338 $r $PERM_CMD $fsname-MDT0001.failover.node=$nid || {
2339 error_noexit "Setting MDT1 \"failover.node\""
2342 $r $PERM_CMD $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
2343 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
2346 $r $PERM_CMD $fsname-MDT0001.lov.stripesize=4M || {
2347 error_noexit "Setting MDT1 \"lov.stripesize\""
2353 if [ "$writeconf" ]; then
2354 echo "== writeconf and client mount =="
2355 $MOUNT_CMD $nid:/$fsname -o user_xattr $tmp/mnt/lustre || {
2356 error_noexit "Mounting the client"
2360 shall_cleanup_lustre=true
2361 $r $LCTL set_param debug="$PTLDEBUG"
2363 t32_verify_quota $SINGLEMDS $fsname $tmp/mnt/lustre || {
2364 error_noexit "verify quota failed"
2368 if $r test -f $tmp/list; then
2369 echo "== list verification =="
2371 # There is not a Test Framework API to copy files to or
2372 # from a remote node.
2374 # LU-2393 - do both sorts on same node to ensure locale
2376 local list_file=$tmp/list
2378 if $mdt2_is_available; then
2379 if [[ -d $tmp/mnt/lustre/striped_dir_old ]] &&
2380 $r test -f $tmp/list2; then
2381 list_file=$tmp/list2
2382 pushd $tmp/mnt/lustre/striped_dir_old
2384 pushd $tmp/mnt/lustre/remote_dir
2387 pushd $tmp/mnt/lustre
2389 $r cat $list_file | sort -k 6 >$tmp/list.orig
2390 BLOCKSIZE=1024 ls -Rni --time-style=+%s | sort -k 6 |
2391 sed 's/\. / /' >$tmp/list || {
2397 # 32-bit and 64-bit clients use different algorithms to
2398 # convert FIDs into inode numbers. Hence, remove the
2399 # inode numbers from the lists, if the original list was
2400 # created on an architecture with different number of
2403 if [ $(t32_bits_per_long $(uname -m)) != \
2404 $(t32_bits_per_long $img_arch) ]; then
2405 echo "Different number of bits per \"long\"" \
2406 "from the disk image"
2407 for list in list.orig list; do
2408 sed -i -e 's/^[0-9]\+[ \t]\+//' \
2412 if ! diff -ub $tmp/list.orig $tmp/list; then
2413 error_noexit "list verification failed"
2417 echo "== list verification skipped =="
2420 #non-dom upgrade to dom
2421 if [ "$dom_upgrade" != "no" ]; then
2422 echo "== dom upgrade =="
2423 $LFS setstripe -E 1M -L mdt -E EOF $tmp/mnt/lustre/dom || {
2424 error_noexit "Verify DoM creation"
2427 [ $($LFS getstripe -L $tmp/mnt/lustre/dom) == "mdt" ] || {
2428 error_noexit "Verify a DoM file"
2431 dd if=/dev/urandom of=$tmp/mnt/lustre/dom bs=4096 \
2432 count=1 conv=fsync || {
2433 error_noexit "Cannot write to DoM file"
2436 [ $(stat -c%s $tmp/mnt/lustre/dom) == 4096 ] || {
2437 error_noexit "DoM: bad size after write"
2440 rm $tmp/mnt/lustre/dom
2442 set_persistent_param_and_check mds \
2443 "lod.*MDT0000*.dom_stripesize" \
2444 "$fsname-MDT0000.lod.dom_stripesize" 0 || {
2445 error_noexit "Changing \"dom_stripesize\""
2451 #$LFS setstripe -E 1M -L mdt -E -1 -S 4M $dom_file
2452 if [ "$dom_new_upgrade" != "no" ]; then
2453 if ! $mdt2_is_available; then
2454 dom_file=$tmp/mnt/lustre/dom_dir/dom_file
2456 echo "== check DoM file can be accessed =="
2457 [ $($LFS getstripe -I1 -L $dom_file) == "mdt" ] || {
2458 error_noexit "Verify a DoM file"
2461 [ $(stat -c%s $dom_file) == $((2 * 1024 * 1024)) ] || {
2462 error_noexit "DoM: bad size after write"
2467 if [ "$flr_upgrade" != "no" ]; then
2468 if ! $mdt2_is_available; then
2469 flr_file=$tmp/mnt/lustre/flr_dir/flr_file
2471 local mirror_count=$($LFS getstripe -N $flr_file)
2472 echo "== check FLR file =="
2473 [ $mirror_count == 2 ] || {
2474 error_noexit "FLR mirror count wrong"
2479 if [ "$dne_upgrade" != "no" ]; then
2480 echo "== dne upgrade striped_dir =="
2481 $LFS mkdir -i 1 -c2 $tmp/mnt/lustre/striped_dir || {
2482 error_noexit "set striped dir failed"
2485 $LFS setdirstripe -D -c2 $tmp/mnt/lustre/striped_dir
2487 pushd $tmp/mnt/lustre
2488 tar -c --exclude=./striped_dir \
2489 --exclude=./striped_dir_old \
2490 --exclude=./remote_dir -f - .|
2491 tar -xvf - -C striped_dir 1>/dev/null || {
2492 error_noexit "cp to striped dir failed"
2498 if [ "$pfl_upgrade" != "no" ]; then
2505 echo "== check PFL file =="
2506 if ! $mdt2_is_available; then
2507 pfl_file=$tmp/mnt/lustre/pfl_dir/pfl_file
2509 comp_cnt=$($LFS getstripe --component-count $pfl_file)
2510 [ $comp_cnt == 2 ] || {
2511 error_noexit "wrong comp_cnt $comp_cnt"
2515 comp_size=$($LFS getstripe -I1 -E $pfl_file)
2516 [ $comp_size == 2097152 ] || {
2517 error_noexit "wrong component size $comp_size"
2521 comp_id=$($LFS getstripe -I1 -i $pfl_file)
2522 [ $comp_id == 0 ] || {
2523 error_noexit "wrong comp id $comp_id"
2527 comp_id=$($LFS getstripe -I2 -i $pfl_file)
2528 [ $comp_id -eq 1 ] || {
2529 error_noexit "wrong comp id $comp_id"
2533 stripe_size=$($LFS getstripe -I1 -S $pfl_file)
2534 [ $stripe_size -eq 1048576 ] || {
2535 error_noexit "wrong stripe size $stripe_size"
2539 stripe_size=$($LFS getstripe -I2 -S $pfl_file)
2540 [ $comp_size -eq 2097152 ] || {
2541 error_noexit "wrong component size $comp_size"
2546 local large_xattr_dir=$tmp/mnt/lustre/large_xattr_test_dir
2548 if [[ -d $large_xattr_dir ]]; then
2549 echo "== check Large EA =="
2550 local xattr_file=$large_xattr_dir/large_xattr_file
2552 xattr_val=$(getfattr --only-values\
2553 -n user.fooattr $xattr_file) || {
2554 error_noexit "Large EA cannot be read"
2558 [[ $xattr_val == $(printf "%c" {1..4096}) ]] || {
2559 error_noexit "Wrong large EA value"
2564 if [ "$project_quota_upgrade" != "no" ]; then
2565 if ! $mdt2_is_available; then
2566 quota_dir=$tmp/mnt/lustre/project_quota_dir
2570 echo "== check Project Quota =="
2571 hardlimit=$(getquota -p $T32_PRJID global 3 \
2573 [ $hardlimit == $T32_PROLIMIT ] || {
2574 error_noexit "wrong hardlimit $hardlimit"
2579 # If it is upgrade from DNE (2.5), then rename the remote dir,
2580 # which is created in 2.5 to striped dir.
2581 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]]; then
2582 echo "== rename remote_dir =="
2583 stripe_index=$($LFS getdirstripe -i \
2584 $tmp/mnt/lustre/remote_dir)
2586 [[ $stripe_index -eq 1 ]] || {
2587 error_noexit "get index \"$stripe_index\"" \
2588 "from remote dir failed"
2591 mv $tmp/mnt/lustre/remote_dir \
2592 $tmp/mnt/lustre/striped_dir/ || {
2593 error_noexit "mv remote dir failed"
2598 # If it is upgraded from DNE (2.7), then move the striped dir
2599 # which was created in 2.7 to the new striped dir.
2600 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]] &&
2601 echo "== rename striped dir =="
2602 [[ -d $tmp/mnt/lustre/striped_dir_old ]]; then
2603 stripe_count=$($LFS getdirstripe -c \
2604 $tmp/mnt/lustre/striped_dir_old)
2605 [[ $stripe_count -eq 2 ]] || {
2606 error_noexit "get count $stripe_count" \
2607 "from striped dir failed"
2610 mv $tmp/mnt/lustre/striped_dir_old \
2611 $tmp/mnt/lustre/striped_dir/ || {
2612 error_noexit "mv striped dir failed"
2618 $r $LCTL set_param -n osd*.*.force_sync=1
2619 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
2620 error_noexit "dd failed"
2623 rm -rf $tmp/mnt/lustre/tmp_file || {
2624 error_noexit "rm failed"
2628 if $r test -f $tmp/sha1sums; then
2629 echo "== checking sha1sums =="
2630 # LU-2393 - do both sorts on same node to ensure locale
2632 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
2633 if [[ "$dne_upgrade" != "no" ]]; then
2634 if [[ -d $tmp/mnt/lustre/striped_dir/remote_dir ]]; then
2635 pushd $tmp/mnt/lustre/striped_dir/remote_dir
2637 pushd $tmp/mnt/lustre/striped_dir
2639 elif [[ "$pfl_upgrade" != "no" ]] ||
2640 [[ "$flr_upgrade" != "no" ]] ||
2641 [[ "$dom_new_upgrade" != "no" ]] ||
2642 [[ "$project_quota_upgrade" != "no" ]] ||
2643 [[ -d $tmp/mnt/lustre/remote_dir ]]; then
2644 pushd $tmp/mnt/lustre/remote_dir
2646 pushd $tmp/mnt/lustre
2649 find ! -path "*remote_dir*" ! -path "*striped_dir*" \
2650 ! -name .lustre -type f -exec sha1sum {} \; |
2651 sort -k 2 >$tmp/sha1sums || {
2653 error_noexit "sha1sum"
2659 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
2660 echo "==** find $save files start **=="
2661 find $save -type f -ls
2662 echo "==** find $save files finish **=="
2663 error_noexit "sha1sum verification failed"
2667 # if upgrade from DNE(2.5), then check remote directory
2668 # if upgrade from DNE(2.7), then check striped directory
2669 if $mdt2_is_available &&
2670 [[ "$dne_upgrade" != "no" ]]; then
2671 echo "== mdt2 upgrade sha1sum check =="
2672 local new_dir="$tmp/mnt/lustre/striped_dir"
2673 local striped_dir_old="$new_dir/striped_dir_old"
2675 local dir_list="$new_dir/remote_dir"
2676 [[ ! -d $triped_dir_old ]] ||
2677 dir_list+=" $striped_dir_old"
2679 for dir in $dir_list; do
2680 echo "== dir = $dir =="
2682 find ! -name .lustre -type f \
2683 -exec sha1sum {} \; |
2684 sort -k 2 >$tmp/sha1sums || {
2687 error_noexit "sha1sum"
2691 if ! diff -ub $tmp/sha1sums.orig \
2694 error_noexit "sha1sum $dir" \
2701 echo "== sha1sum verification skipped =="
2704 # PFL write test after sha1sum check
2705 if [ "$pfl_upgrade" != "no" ]; then
2706 local rw_len=$((3 * 1034 * 1024))
2708 if ! $mdt2_is_available; then
2709 pfl_file=$tmp/mnt/lustre/pfl_dir/pfl_file
2711 small_write $pfl_file $rw_len || {
2712 error_noexit "PFL RW Failed"
2717 if [ "$dne_upgrade" != "no" ]; then
2718 rm -rf $tmp/mnt/lustre/striped_dir || {
2719 error_noexit "remove remote dir failed"
2724 # migrate files/dirs to remote MDT, then move them back
2725 if [ "$MDS1_VERSION" -ge $(version_code 2.7.50) -a \
2726 $dne_upgrade != "no" ]; then
2727 $r $LCTL set_param -n \
2728 mdt.${fsname}*.enable_remote_dir=1 2>/dev/null
2730 echo "== test migration =="
2731 pushd $tmp/mnt/lustre
2732 for dir in $(find ! -name .lustre ! -name . -type d); do
2733 mdt_index=$($LFS getdirstripe -i $dir)
2734 stripe_cnt=$($LFS getdirstripe -c $dir)
2735 if [ $mdt_index = 0 -a $stripe_cnt -le 1 ]; then
2736 $LFS migrate -m 1 $dir || {
2738 error_noexit "migrate MDT1 failed"
2744 for dir in $(find ! -name . ! -name .lustre -type d); do
2745 mdt_index=$($LFS getdirstripe -i $dir)
2746 stripe_cnt=$($LFS getdirstripe -c $dir)
2747 if [ $mdt_index = 1 -a $stripe_cnt -le 1 ]; then
2748 $LFS migrate -m 0 $dir || {
2750 error_noexit "migrate MDT0 failed"
2759 # When adding new data verification tests, please check for
2760 # the presence of the required reference files first, like
2761 # the "sha1sums" and "list" tests above, to avoid the need to
2762 # regenerate every image for each test addition.
2765 nrpcs_orig=$($LCTL get_param \
2766 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
2767 error_noexit "Getting \"max_rpcs_in_flight\""
2770 nrpcs=$((nrpcs_orig + 5))
2772 set_persistent_param_and_check client \
2773 "mdc.$fsname-MDT0000*.max_rpcs_in_flight" \
2774 "$fsname-MDT0000.mdc.max_rpcs_in_flight" $nrpcs || {
2775 error_noexit "Changing \"max_rpcs_in_flight\""
2779 umount $tmp/mnt/lustre || {
2780 error_noexit "Unmounting the client"
2783 shall_cleanup_lustre=false
2785 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
2786 error_noexit "Mounting the client"
2790 if [ "$mds1_FSTYPE" == ldiskfs -a \
2791 "$project_upgrade" != "no" ]; then
2792 ! $LFS project -d -p 1 $tmp/mnt/lustre/* &&
2793 error_noexit "set project failed" &&
2797 [[ $(do_facet mds1 pgrep orph_.*-MDD | wc -l) == 0 ]] ||
2798 error "MDD orphan cleanup thread not quit"
2800 umount $tmp/mnt/lustre || {
2801 error_noexit "Unmounting the client"
2805 if [[ "$dne_upgrade" != "no" ]] || $mdt2_is_available; then
2806 $r $UMOUNT $tmp/mnt/mdt1 || {
2807 error_noexit "Unmounting the MDT2"
2810 if [[ "$mds1_FSTYPE" == zfs ]]; then
2811 $r "$ZPOOL export t32fs-mdt2"
2813 shall_cleanup_mdt1=false
2816 $r $UMOUNT $tmp/mnt/mdt || {
2817 error_noexit "Unmounting the MDT"
2820 if [[ "$mds1_FSTYPE" == zfs ]]; then
2821 $r "$ZPOOL export t32fs-mdt1"
2823 shall_cleanup_mdt=false
2825 $r $UMOUNT $tmp/mnt/ost || {
2826 error_noexit "Unmounting the OST"
2829 if $ost2_is_available; then
2830 $r $UMOUNT $tmp/mnt/ost1 || {
2831 error_noexit "Unmounting the OST1"
2834 shall_cleanup_ost1=false
2837 if [[ $ost1_FSTYPE == zfs ]]; then
2838 $r "$ZPOOL export t32fs-ost1"
2839 if $ost2_is_available; then
2840 $r "$ZPOOL export t32fs-ost2"
2843 shall_cleanup_ost=false
2845 t32_reload_modules $node || {
2846 error_noexit "Reloading modules"
2850 if [[ "$mds1_FSTYPE" == zfs ]]; then
2851 local poolname=t32fs-mdt1
2853 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
2854 $ZPOOL import -f -d $tmp $poolname"
2856 # upgrade zpool to latest supported features,
2857 # including dnode quota accounting in 0.7.0
2858 $r "$ZPOOL upgrade $poolname"
2861 # mount a second time to make sure we didnt leave upgrade flag on
2862 $r $TUNEFS --dryrun $mdt_dev || {
2864 error_noexit "tunefs.lustre before remounting the MDT"
2868 mopts=exclude=$fsname-OST0000
2869 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2872 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2873 error_noexit "Remounting the MDT"
2876 shall_cleanup_mdt=true
2886 for tarball in $tarballs; do
2887 banner "testing $tarball upgrade"
2888 t32_test $tarball || let "rc += $?"
2892 run_test 32a "Upgrade (not live)"
2900 for tarball in $tarballs; do
2901 banner "testing $tarball upgrade with writeconf"
2902 t32_test $tarball writeconf || let "rc += $?"
2906 run_test 32b "Upgrade with writeconf"
2914 for tarball in $tarballs; do
2915 banner "testing $tarball upgrade with DNE"
2917 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
2921 run_test 32c "dne upgrade test"
2929 for tarball in $tarballs; do
2930 banner "testing $tarball upgrade with ff and project upgrade"
2931 project_upgrade="no"
2932 [[ "$MDS1_VERSION" -ge $(version_code 2.13.54) ]] &&
2933 [[ "$tarball" =~ "disk2_4-ldiskfs" ]] &&
2934 project_upgrade="yes"
2935 project_upgrade=$project_upgrade ff_convert=yes t32_test \
2940 run_test 32d "convert ff and project quota upgrade test"
2943 [[ "$MDS1_VERSION" -ge $(version_code 2.10.56) ]] ||
2944 skip "Need MDS version at least 2.10.56"
2951 for tarball in $tarballs; do
2952 [[ "$tarball" =~ "2_9" ]] ||
2953 { echo "skip $(basename $tarball)"; continue; }
2955 banner "testing $tarball upgrade with DoM"
2956 dom_upgrade=yes t32_test $tarball writeconf || let "rc += $?"
2960 run_test 32e "dom upgrade test"
2963 [[ $MDS1_VERSION -ge $(version_code 2.10.56) ]] ||
2964 skip "Need MDS version at least 2.10.56"
2971 for tarball in $tarballs; do
2972 [[ "$tarball" =~ "2_10" ]] ||
2973 { echo "skip $(basename $tarball)"; continue; }
2974 pfl_upgrade=yes project_quota_upgrade=yes \
2975 t32_test $tarball writeconf || let "rc += $?"
2979 run_test 32f "pfl upgrade test"
2982 [[ $MDS1_VERSION -ge $(version_code 2.10.56) ]] ||
2983 skip "Need MDS version at least 2.10.56"
2990 for tarball in $tarballs; do
2991 [[ $tarball =~ "2_12" ]] ||
2992 { echo "skip $(basename $tarball)"; continue; }
2993 flr_upgrade=yes dom_new_upgrade=yes \
2994 t32_test $tarball writeconf || let "rc += $?"
2998 run_test 32g "flr/dom upgrade test"
3000 test_33a() { # bug 12333, was test_33
3001 local FSNAME2=test-$testnum
3003 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3006 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
3008 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
3009 local dev=${SINGLEMDS}_dev
3010 local MDSDEV=${!dev}
3012 is_blkdev $SINGLEMDS $MDSDEV &&
3013 skip_env "mixed loopback and real device not working"
3016 local fs2mdsdev=$(mdsdevname 1_2)
3017 local fs2ostdev=$(ostdevname 1_2)
3018 local fs2mdsvdev=$(mdsvdevname 1_2)
3019 local fs2ostvdev=$(ostvdevname 1_2)
3021 if [ "$mds1_FSTYPE" == ldiskfs ]; then
3022 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
3025 if combined_mgs_mds; then
3026 local mgs_flag="--mgs"
3029 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --fsname=${FSNAME2} \
3030 --reformat $mgs_flag $mkfsoptions $fs2mdsdev $fs2mdsvdev ||
3032 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
3033 --fsname=${FSNAME2} --index=0x1fff --reformat $fs2ostdev \
3034 $fs2ostvdev || exit 10
3036 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
3037 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
3039 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3040 do_facet mgs "$PERM_CMD timeout=200" ||
3041 error "$PERM_CMD timeout=200 failed"
3043 do_facet mgs "$PERM_CMD $FSNAME2.sys.timeout=200" ||
3044 error "$PERM_CMD $FSNAME2.sys.timeout=200 failed"
3046 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
3047 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
3050 cp /etc/hosts $MOUNT2/ || error "copy /etc/hosts $MOUNT2/ failed"
3051 $LFS getstripe $MOUNT2/hosts ||
3052 error "$LFS getstripe $MOUNT2/hosts failed"
3056 if (( "$MGS_VERSION" >= $(version_code 2.15.51) &&
3057 "$MDS1_VERSION" >= $(version_code 2.15.51) )); then
3058 # test lctl del_ost on large index
3059 do_facet mgs "$LCTL del_ost -t ${FSNAME2}-OST1fff" ||
3060 error "del_ost failed with $?"
3061 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 ||
3062 error "$MOUNT_CMD failed"
3064 $LFS df | grep -q OST1fff &&
3065 error "del_ost did not remove OST1fff!"
3071 cleanup_nocli || error "cleanup_nocli failed with $?"
3073 run_test 33a "Mount ost with a large index number"
3075 test_33b() { # was test_34
3078 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
3079 # Drop lock cancelation reply during umount
3080 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
3081 do_facet client $LCTL set_param fail_loc=0x80000304
3082 #lctl set_param debug=-1
3083 umount_client $MOUNT
3084 cleanup || error "cleanup failed with $?"
3086 run_test 33b "Drop cancel during umount"
3090 do_facet client "bash runmultiop_bg_pause $DIR/file O_c"
3091 manual_umount_client
3093 do_facet client killall -USR1 multiop
3094 if [ $rc -eq 0 ]; then
3095 error "umount not fail!"
3098 cleanup || error "cleanup failed with rc $?"
3100 run_test 34a "umount with opened file should be fail"
3104 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3105 stop_mds || error "Unable to stop MDS"
3107 manual_umount_client --force || error "mtab after failed umount with $?"
3109 cleanup || error "cleanup failed with $?"
3111 run_test 34b "force umount with failed mds should be normal"
3115 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3116 stop_ost || error "Unable to stop OST1"
3118 manual_umount_client --force || error "mtab after failed umount with $?"
3120 cleanup || error "cleanup failed with $?"
3122 run_test 34c "force umount with failed ost should be normal"
3124 test_35a() { # bug 12459
3127 DBG_SAVE="`$LCTL get_param -n debug`"
3128 $LCTL set_param debug="ha"
3130 log "Set up a fake failnode for the MDS"
3132 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
3133 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
3135 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3136 do_facet mgs "$PERM_CMD \
3137 mdc.*${device}*.import=connection=$(h2nettype $FAKENID)" ||
3138 error "Setting mdc.*${device}*.import=connection=\
3139 $(h2nettype $FAKENID) failed."
3141 do_facet mgs "$PERM_CMD \
3142 ${device}.failover.node=$(h2nettype $FAKENID)" ||
3143 error "Setting ${device}.failover.node=\
3144 $(h2nettype $FAKENID) failed."
3146 log "Wait for RECONNECT_INTERVAL seconds (10s)"
3149 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
3152 log "Stopping the MDT: $device"
3153 stop_mdt 1 || error "MDT0 stop fail"
3155 df $MOUNT > /dev/null 2>&1 &
3157 log "Restarting the MDT: $device"
3158 start_mdt 1 || error "MDT0 start fail"
3159 log "Wait for df ($DFPID) ... "
3162 $LCTL set_param debug="$DBG_SAVE"
3164 # retrieve from the log the first server that the client tried to
3165 # contact after the connection loss
3166 $LCTL dk $TMP/lustre-log-$TESTNAME.log
3167 NEXTCONN=`awk "/${MSG}/ {start = 1;}
3168 /import_select_connection.*$device-mdc.* using connection/ {
3170 if (\\\$NF ~ /$FAKENID/)
3176 }" $TMP/lustre-log-$TESTNAME.log`
3177 [ "$NEXTCONN" != "0" ] &&
3178 error "Tried to connect to ${NEXTCONN} not last active server"
3179 cleanup || error "cleanup failed with $?"
3180 # remove nid settings
3181 writeconf_or_reformat
3183 run_test 35a "Reconnect to the last active server first"
3185 test_35b() { # bug 18674
3186 remote_mds || skip "local MDS"
3190 $LCTL set_param debug="ha"
3192 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
3195 log "Set up a fake failnode for the MDS"
3197 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
3198 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
3200 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3201 do_facet mgs "$PERM_CMD \
3202 mdc.*${device}*.import=connection=$(h2nettype $FAKENID)" ||
3203 error "Set mdc.*${device}*.import=connection=\
3204 $(h2nettype $FAKENID) failed"
3206 do_facet mgs "$PERM_CMD \
3207 ${device}.failover.node=$(h2nettype $FAKENID)" ||
3208 error "Set ${device}.failover.node=\
3209 $(h2nettype $FAKENID) failed"
3212 local at_max_saved=0
3213 # adaptive timeouts may prevent seeing the issue
3214 if at_is_enabled; then
3215 at_max_saved=$(at_max_get mds)
3216 at_max_set 0 mds client
3219 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
3221 log "Injecting EBUSY on MDS"
3222 # Setting OBD_FAIL_MDS_RESEND=0x136
3223 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" ||
3224 error "unable to set param fail_loc=0x80000136"
3226 $LCTL set_param mdc.${FSNAME}*.stats=clear
3228 log "Creating a test file and stat it"
3229 touch $MOUNT/$tdir/$tfile || error "touch $MOUNT/$tdir/$tfile failed"
3230 stat $MOUNT/$tdir/$tfile
3232 log "Stop injecting EBUSY on MDS"
3233 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" ||
3234 error "unable to set param fail_loc=0"
3235 rm -f $MOUNT/$tdir/$tfile || error "remove $MOUNT/$tdir/$tfile failed"
3238 # restore adaptive timeout
3239 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
3241 $LCTL dk $TMP/lustre-log-$TESTNAME.log
3243 CONNCNT=$($LCTL get_param mdc.${FSNAME}*.stats |
3244 awk '/mds_connect/{print $2}')
3246 # retrieve from the log if the client has ever tried to
3247 # contact the fake server after the loss of connection
3248 FAILCONN=`awk "BEGIN {ret = 0;}
3249 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
3251 if (\\\$NF ~ /$FAKENID/) {
3256 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
3258 [ "$FAILCONN" == "0" ] &&
3259 error "The client reconnection has not been triggered"
3260 [ "$FAILCONN" == "2" ] &&
3261 error "Primary server busy, client reconnect to failover failed"
3264 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
3265 # Reconnects are supposed to be rate limited to one every 5s
3266 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] &&
3267 error "Too many reconnects $CONNCNT"
3269 cleanup || error "cleanup failed with $?"
3270 # remove nid settings
3271 writeconf_or_reformat
3273 run_test 35b "Continue reconnection retries, if the active server is busy"
3276 [ $OSTCOUNT -lt 2 ] && skip_env "needs >= 2 OSTs"
3278 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] ||
3282 local FSNAME2=test1234
3283 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3285 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
3287 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
3288 is_blkdev $SINGLEMDS $MDSDEV &&
3289 skip_env "mixed loopback and real device not working"
3292 local fs2mdsdev=$(mdsdevname 1_2)
3293 local fs2ostdev=$(ostdevname 1_2)
3294 local fs3ostdev=$(ostdevname 2_2)
3295 local fs2mdsvdev=$(mdsvdevname 1_2)
3296 local fs2ostvdev=$(ostvdevname 1_2)
3297 local fs3ostvdev=$(ostvdevname 2_2)
3300 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
3301 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
3302 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
3303 # different one than the default value here.
3304 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
3305 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
3306 add fs3ost $(mkfs_opts ost2 ${fs3ostdev}) --mgsnode=$MGSNID \
3307 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
3309 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
3310 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
3311 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
3312 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
3313 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
3315 sleep 5 # until 11778 fixed
3317 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || error "dd failed"
3319 BKTOTAL=$($LCTL get_param -n obdfilter.*.kbytestotal |
3320 awk 'BEGIN{total=0}; {total+=$1}; END{print total}')
3321 BKFREE=$($LCTL get_param -n obdfilter.*.kbytesfree |
3322 awk 'BEGIN{free=0}; {free+=$1}; END{print free}')
3323 BKAVAIL=$($LCTL get_param -n obdfilter.*.kbytesavail |
3324 awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}')
3325 STRING=$(df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}')
3326 DFTOTAL=$(echo $STRING | cut -d, -f1)
3327 DFUSED=$(echo $STRING | cut -d, -f2)
3328 DFAVAIL=$(echo $STRING | cut -d, -f3)
3329 DFFREE=$(($DFTOTAL - $DFUSED))
3331 ALLOWANCE=$((64 * $OSTCOUNT))
3333 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
3334 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
3335 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
3338 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
3339 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
3340 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
3343 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
3344 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
3345 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
3350 stop fs3ost -f || error "unable to stop OST3"
3351 stop fs2ost -f || error "unable to stop OST2"
3352 stop fs2mds -f || error "unable to stop second MDS"
3353 unload_modules_conf || error "unable unload modules"
3356 run_test 36 "df report consistency on OSTs with different block size"
3359 local mntpt=$(facet_mntpt $SINGLEMDS)
3360 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
3361 local mdsdev_sym="$TMP/sym_mdt.img"
3362 local opts=$MDS_MOUNT_OPTS
3365 if [ "$mds1_FSTYPE" != ldiskfs ]; then
3366 skip "ldiskfs only test"
3369 echo "MDS : $mdsdev"
3370 echo "SYMLINK : $mdsdev_sym"
3371 do_facet $SINGLEMDS rm -f $mdsdev_sym
3373 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
3375 echo "mount symlink device - $mdsdev_sym"
3377 if ! do_facet $SINGLEMDS test -b $mdsdev; then
3378 opts=$(csa_add "$opts" -o loop)
3382 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
3383 $mdsdev_sym $mntpt 2>&1)
3386 echo mount_op=$mount_op
3388 do_facet $SINGLEMDS "$UMOUNT $mntpt && rm -f $mdsdev_sym"
3390 if $(echo $mount_op | grep -q "unable to set tunable"); then
3391 error "set tunables failed for symlink device"
3394 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
3396 run_test 37 "verify set tunables works for symlink device"
3398 test_38() { # bug 14222
3399 local mntpt=$(facet_mntpt $SINGLEMDS)
3404 local SRC="/etc /bin"
3405 local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT)
3407 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
3408 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3409 tar cf - $FILES | tar xf - -C $DIR/$tdir ||
3410 error "copying $SRC to $DIR/$tdir"
3412 umount_client $MOUNT || error "umount_client $MOUNT failed"
3413 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3414 stop_mds || error "Unable to stop MDS"
3415 log "delete lov_objid file on MDS"
3417 mount_fstype $SINGLEMDS || error "mount MDS failed (1)"
3419 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid; rm $mntpt/lov_objid"
3421 unmount_fstype $SINGLEMDS || error "umount failed (1)"
3423 # check create in mds_lov_connect
3424 start_mds || error "unable to start MDS"
3425 mount_client $MOUNT || error "mount_client $MOUNT failed"
3427 [ $V ] && log "verifying $DIR/$tdir/$f"
3428 diff -q $f $DIR/$tdir/$f || ERROR=y
3430 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3431 if [ "$ERROR" = "y" ]; then
3432 # check it's updates in sync
3433 umount_client $MOUNT
3435 mount_fstype $SIGNLEMDS
3436 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3437 unmount_fstype $SINGLEMDS
3438 error "old and new files are different after connect" || true
3440 touch $DIR/$tdir/f2 || error "f2 file create failed"
3442 # check it's updates in sync
3443 umount_client $MOUNT || error "second umount_client $MOUNT failed"
3446 mount_fstype $SINGLEMDS || error "mount MDS failed (3)"
3448 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3449 do_facet $SINGLEMDS dd if=/dev/zero of=$mntpt/lov_objid.clear count=8
3451 unmount_fstype $SINGLEMDS || error "umount failed (3)"
3453 start_mds || error "unable to start MDS"
3454 mount_client $MOUNT || error "mount_client $MOUNT failed"
3456 [ $V ] && log "verifying $DIR/$tdir/$f"
3457 diff -q $f $DIR/$tdir/$f || ERROR=y
3459 touch $DIR/$tdir/f3 || error "f3 file create failed"
3460 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3461 umount_client $MOUNT || error "third umount_client $MOUNT failed"
3463 mount_fstype $SINGLEMDS || error "mount MDS failed (4)"
3464 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3465 unmount_fstype $SINGLEMDS || error "umount failed (4)"
3467 [[ "$ERROR" != "y" ]] ||
3468 error "old and new files are different after sync"
3470 log "files compared the same"
3471 cleanup || error "cleanup failed with $?"
3473 run_test 38 "MDS recreates missing lov_objid file from OST data"
3476 [[ -n "$(type -p perl)" ]] || skip_env "need perl for leak_finder.pl"
3480 cleanup || error "cleanup failed with $?"
3481 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
3482 error "memory leak detected" || true
3484 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
3486 test_40() { # bug 15759
3487 start_ost || error "Unable to start OST1"
3488 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
3489 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
3491 cleanup || error "cleanup failed with rc $?"
3493 run_test 40 "race during service thread startup"
3495 test_41a() { #bug 14134
3496 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3497 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
3498 skip "Loop devices does not work with nosvc option"
3501 combined_mgs_mds || skip "needs combined MGT and MDT device"
3503 start_mdt 1 -o nosvc -n
3504 if [ $MDSCOUNT -ge 2 ]; then
3505 for num in $(seq 2 $MDSCOUNT); do
3506 start_mdt $num || return
3509 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
3510 start_mdt 1 -o nomgs,force
3511 mount_client $MOUNT || error "mount_client $MOUNT failed"
3514 echo "blah blah" > $MOUNT/$tfile
3517 umount_client $MOUNT || error "umount_client $MOUNT failed"
3518 stop ost1 -f || error "unable to stop OST1"
3519 stop_mds || error "Unable to stop MDS"
3520 stop_mds || error "Unable to stop MDS on second try"
3522 run_test 41a "mount mds with --nosvc and --nomgs"
3525 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3526 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
3527 skip "Loop devices does not work with nosvc option"
3530 ! combined_mgs_mds && skip "needs combined mgs device"
3534 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3536 start_mdt 1 -o nosvc -n
3537 if [ $MDSCOUNT -ge 2 ]; then
3538 for num in $(seq 2 $MDSCOUNT); do
3539 start_mdt $num || return
3542 start_ost || error "Unable to start OST1"
3543 start_mdt 1 -o nomgs,force
3544 mount_client $MOUNT || error "mount_client $MOUNT failed"
3547 echo "blah blah" > $MOUNT/$tfile
3548 cat $MOUNT/$tfile || error "cat $MOUNT/$tfile failed"
3550 umount_client $MOUNT -f || error "umount_client $MOUNT failed"
3551 stop_ost || error "Unable to stop OST1"
3552 stop_mds || error "Unable to stop MDS"
3553 stop_mds || error "Unable to stop MDS on second try"
3555 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
3558 local oss_list=$(comma_list $(osts_nodes))
3560 [[ "$MDS1_VERSION" -ge $(version_code 2.6.52) ]] ||
3561 [[ "$MDS1_VERSION" -ge $(version_code 2.5.26) &&
3562 "$MDS1_VERSION" -lt $(version_code 2.5.50) ]] ||
3563 [[ "$MDS1_VERSION" -ge $(version_code 2.5.4) &&
3564 "$MDS1_VERSION" -lt $(version_code 2.5.11) ]] ||
3565 skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"
3567 # ensure mds1 ost1 have been created even if running sub-test standalone
3570 cleanup || error "cleanup failed"
3572 # using directly mount command instead of start() function to avoid
3573 # any side effect of // with others/externals tools/features
3574 # ("zpool import", ...)
3576 # MDT concurrent start
3578 LOAD_MODULES_REMOTE=true load_modules
3579 do_facet $SINGLEMDS "lsmod | grep -q libcfs" ||
3580 error "MDT concurrent start: libcfs module not loaded"
3582 local mds1dev=$(mdsdevname 1)
3583 local mds1mnt=$(facet_mntpt mds1)
3584 local mds1opts=$MDS_MOUNT_OPTS
3586 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3587 ! do_facet mds1 test -b $mds1dev; then
3588 mds1opts=$(csa_add "$mds1opts" -o loop)
3590 if [[ "$mds1_FSTYPE" == zfs ]]; then
3591 import_zpool mds1 || return ${PIPESTATUS[0]}
3594 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3595 do_facet mds1 "$LCTL set_param fail_loc=0x80000716"
3597 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts &
3600 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts
3605 do_facet mds1 "$LCTL set_param fail_loc=0x0"
3606 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3607 echo "1st MDT start succeed"
3608 echo "2nd MDT start failed with $rc2"
3609 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3610 echo "1st MDT start failed with $rc"
3611 echo "2nd MDT start succeed"
3614 error "unexpected concurrent MDT mounts result, rc=$rc rc2=$rc2"
3617 if [ $MDSCOUNT -ge 2 ]; then
3618 for num in $(seq 2 $MDSCOUNT); do
3619 start_mdt $num || return
3623 # OST concurrent start
3625 do_rpc_nodes $oss_list "lsmod | grep -q libcfs" ||
3626 error "OST concurrent start: libcfs module not loaded"
3628 local ost1dev=$(ostdevname 1)
3629 local ost1mnt=$(facet_mntpt ost1)
3630 local ost1opts=$OST_MOUNT_OPTS
3632 if [ "$ost1_FSTYPE" == ldiskfs ] &&
3633 ! do_facet ost1 test -b $ost1dev; then
3634 ost1opts=$(csa_add "$ost1opts" -o loop)
3636 if [[ "$ost1_FSTYPE" == zfs ]]; then
3637 import_zpool ost1 || return ${PIPESTATUS[0]}
3640 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3641 do_facet ost1 "$LCTL set_param fail_loc=0x80000716"
3643 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts &
3646 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts
3650 do_facet ost1 "$LCTL set_param fail_loc=0x0"
3651 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3652 echo "1st OST start succeed"
3653 echo "2nd OST start failed with $rc2"
3654 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3655 echo "1st OST start failed with $rc"
3656 echo "2nd OST start succeed"
3660 error "unexpected concurrent OST mounts result, rc=$rc rc2=$rc2"
3666 # verify everything ok
3671 error "MDT(s) start failed"
3679 error "OST(s) start failed"
3687 error "client start failed"
3694 error "client mount failed"
3698 run_test 41c "concurrent mounts of MDT/OST should all fail but one"
3700 test_42() { #bug 14693
3704 check_mount || error "client was not mounted"
3706 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3707 PARAM="llite.$FSNAME-*.some_wrong_param"
3709 PARAM="$FSNAME.llite.some_wrong_param"
3712 do_facet mgs $PERM_CMD $PARAM=10
3713 umount_client $MOUNT ||
3714 error "unmounting client failed with invalid llite param"
3715 mount_client $MOUNT ||
3716 error "mounting client failed with invalid llite param"
3718 do_facet mgs $PERM_CMD $PARAM=20
3719 cleanup || error "stopping $FSNAME failed with invalid sys param"
3721 check_mount || error "client was not mounted with invalid sys param"
3722 cleanup || error "stopping $FSNAME failed with invalid sys param"
3724 run_test 42 "allow client/server mount/unmount with invalid config param"
3727 [[ "$MGS_VERSION" -ge $(version_code 2.5.58) ]] ||
3728 skip "Need MDS version at least 2.5.58"
3729 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
3732 USER1=$(getent passwd | grep :$ID1:$ID1: | cut -d: -f1)
3733 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1"
3736 chmod ugo+x $DIR || error "chmod 0 failed"
3737 set_persistent_param_and_check mds1 \
3738 "mdt.$FSNAME-MDT0000.root_squash" \
3739 "$FSNAME.mdt.root_squash" \
3741 wait_update $HOSTNAME \
3742 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3744 error "check llite root_squash failed!"
3745 set_persistent_param_and_check mds1 \
3746 "mdt.$FSNAME-MDT0000.nosquash_nids" \
3747 "$FSNAME.mdt.nosquash_nids" \
3749 wait_update $HOSTNAME \
3750 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3752 error "check llite nosquash_nids failed!"
3755 # create set of test files
3757 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
3758 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
3759 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
3761 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
3762 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
3764 mkdir_on_mdt0 $DIR/$tdir-rootdir || error "mkdir failed"
3765 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
3766 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
3768 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
3769 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
3770 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
3773 # check root_squash:
3774 # set root squash UID:GID to RUNAS_ID
3775 # root should be able to access only files owned by RUNAS_ID
3777 set_persistent_param_and_check mds1 \
3778 "mdt.$FSNAME-MDT0000.root_squash" \
3779 "$FSNAME.mdt.root_squash" \
3780 "$RUNAS_ID:$RUNAS_ID"
3781 wait_update $HOSTNAME \
3782 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3783 "$RUNAS_ID:$RUNAS_ID" ||
3784 error "check llite root_squash failed!"
3786 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
3787 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3788 error "$ST: root read permission is denied"
3789 echo "$ST: root read permission is granted - ok"
3792 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3793 error "$ST: root write permission is denied"
3794 echo "$ST: root write permission is granted - ok"
3796 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3797 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3798 error "$ST: root read permission is granted"
3799 echo "$ST: root read permission is denied - ok"
3802 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3803 error "$ST: root write permission is granted"
3804 echo "$ST: root write permission is denied - ok"
3806 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3807 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null &&
3808 error "$ST: root unlink permission is granted"
3809 echo "$ST: root unlink permission is denied - ok"
3811 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null &&
3812 error "$ST: root create permission is granted"
3813 echo "$ST: root create permission is denied - ok"
3816 # check root_squash is enforced independently
3817 # of client cache content
3819 # access file by USER1, keep access open
3820 # root should be denied access to user file
3822 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
3826 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
3827 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3828 { kill $pid; error "$ST: root read permission is granted"; }
3829 echo "$ST: root read permission is denied - ok"
3832 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3833 { kill $pid; error "$ST: root write permission is granted"; }
3834 echo "$ST: root write permission is denied - ok"
3840 # check nosquash_nids:
3841 # put client's NID into nosquash_nids list,
3842 # root should be able to access root file after that
3844 local NIDLIST=$($LCTL list_nids all | tr '\n' ' ')
3845 NIDLIST="2@gni $NIDLIST 192.168.0.[2,10]@tcp"
3846 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
3847 set_persistent_param_and_check mds1 \
3848 "mdt.$FSNAME-MDT0000.nosquash_nids" \
3849 "$FSNAME-MDTall.mdt.nosquash_nids" \
3851 wait_update $HOSTNAME \
3852 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3854 error "check llite nosquash_nids failed!"
3856 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3857 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3858 error "$ST: root read permission is denied"
3859 echo "$ST: root read permission is granted - ok"
3862 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3863 error "$ST: root write permission is denied"
3864 echo "$ST: root write permission is granted - ok"
3866 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3867 rm $DIR/$tdir-rootdir/tfile-1 ||
3868 error "$ST: root unlink permission is denied"
3869 echo "$ST: root unlink permission is granted - ok"
3870 touch $DIR/$tdir-rootdir/tfile-2 ||
3871 error "$ST: root create permission is denied"
3872 echo "$ST: root create permission is granted - ok"
3873 cleanup || error "cleanup failed with $?"
3875 run_test 43a "check root_squash and nosquash_nids"
3877 test_43b() { # LU-5690
3878 [[ "$MGS_VERSION" -ge $(version_code 2.7.62) ]] ||
3879 skip "Need MGS version 2.7.62+"
3881 if [[ -z "$fs2mds_DEV" ]]; then
3882 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
3883 skip_env "mixed loopback and real device not working"
3886 local fs2mdsdev=$(mdsdevname 1_2)
3887 local fs2mdsvdev=$(mdsvdevname 1_2)
3889 # temporarily use fs2mds as fs2mgs
3891 local fs2mgsdev=$fs2mdsdev
3892 local fs2mgsvdev=$fs2mdsvdev
3894 local fsname=test1234
3896 load_module llite/lustre
3897 local client_ip=$(host_nids_address $HOSTNAME $NETTYPE)
3898 local host=${client_ip//*./}
3899 local net=${client_ip/%$host/}
3900 local nosquash_nids=$(h2nettype $net[$host,$host,$host])
3902 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
3903 --param mdt.root_squash=$RUNAS_ID:$RUNAS_ID \
3904 --param mdt.nosquash_nids=$nosquash_nids \
3905 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
3906 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS || error "start fs2mgs failed"
3907 stop $fs2mgs -f || error "stop fs2mgs failed"
3908 cleanup || error "cleanup failed with $?"
3910 run_test 43b "parse nosquash_nids with commas in expr_list"
3914 check_mount || error "check_mount"
3915 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
3917 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
3918 for VAL in $UUIDS; do
3919 NID=$(echo $VAL | cut -d= -f1)
3920 CLUUID=$(echo $VAL | cut -d= -f2)
3921 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
3923 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
3924 cleanup || error "cleanup failed with $?"
3926 run_test 44 "mounted client proc entry exists"
3930 check_mount || error "check_mount"
3931 stop_mds || error "Unable to stop MDS"
3935 #define OBD_FAIL_PTLRPC_LONG_REPL_UNLINK 0x50f
3936 do_facet client "$LCTL set_param fail_loc=0x8000050f"
3939 manual_umount_client --force || error "manual_umount_client failed"
3940 do_facet client "$LCTL set_param fail_loc=0x0"
3941 start_mds || error "unable to start MDS"
3942 mount_client $MOUNT || error "mount_client $MOUNT failed"
3943 cleanup || error "cleanup failed with $?"
3945 run_test 45 "long unlink handling in ptlrpcd"
3952 umount_client $MOUNT2 || rc=$?
3953 umount_client $MOUNT || rc=$?
3954 while [ $count -gt 0 ]; do
3955 stop ost${count} -f || rc=$?
3959 cleanup_nocli || rc=$?
3960 #writeconf to remove all ost2 traces for subsequent tests
3961 writeconf_or_reformat
3966 echo "Testing with $OSTCOUNT OSTs"
3968 start_mds || error "unable to start MDS"
3969 #first client should see only one ost
3970 start_ost || error "Unable to start OST1"
3971 wait_osc_import_state mds ost FULL
3973 mount_client $MOUNT || error "mount_client $MOUNT failed"
3974 trap "cleanup_46a $OSTCOUNT" EXIT ERR
3977 for (( i=2; i<=$OSTCOUNT; i++ )); do
3978 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
3979 error "start_ost$i $(ostdevname $i) failed"
3982 # wait until osts in sync
3983 for (( i=2; i<=$OSTCOUNT; i++ )); do
3984 wait_osc_import_state mds ost$i FULL
3985 wait_osc_import_ready client ost$i
3988 #second client see all ost's
3990 mount_client $MOUNT2 || error "mount_client failed"
3991 $LFS setstripe -c -1 $MOUNT2 ||
3992 error "$LFS setstripe -c -1 $MOUNT2 failed"
3993 $LFS getstripe $MOUNT2 || error "$LFS getstripe $MOUNT2 failed"
3995 echo "ok" > $MOUNT2/widestripe
3996 $LFS getstripe $MOUNT2/widestripe ||
3997 error "$LFS getstripe $MOUNT2/widestripe failed"
3998 # fill acl buffer for avoid expand lsm to them
3999 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd |
4001 setfacl -m $acl $MOUNT2/widestripe
4005 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
4007 cleanup_46a $OSTCOUNT || error "cleanup_46a failed"
4009 run_test 46a "handle ost additional - wide striped file"
4014 check_mount || error "check_mount failed"
4015 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
4019 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
4020 if echo $ns | grep "MDT[[:digit:]]*"; then
4023 lrs=$(echo $ns | sed 's/.*lru_size=//')
4024 lru_size[count]=$lrs
4029 facet_failover $SINGLEMDS
4030 client_up || error "client_up failed"
4033 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
4034 if echo $ns | grep "MDT[[:digit:]]*"; then
4037 lrs=$(echo $ns | sed 's/.*lru_size=//')
4038 if ! test "$lrs" -eq "${lru_size[count]}"; then
4039 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
4040 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
4045 cleanup || error "cleanup failed with $?"
4047 run_test 47 "server restart does not make client loss lru_resize settings"
4052 # reformat after this test is needed - if the test fails,
4053 # we will have unkillable file at FS
4057 test_48() { # bz-17636 LU-7473
4061 check_mount || error "check_mount failed"
4063 $LFS setstripe -c -1 $MOUNT ||
4064 error "$LFS setstripe -c -1 $MOUNT failed"
4065 $LFS getstripe $MOUNT || error "$LFS getstripe $MOUNT failed"
4067 echo "ok" > $MOUNT/widestripe
4068 $LFS getstripe $MOUNT/widestripe ||
4069 error "$LFS getstripe $MOUNT/widestripe failed"
4071 # In the future, we may introduce more EAs, such as selinux, enlarged
4072 # LOV EA, and so on. These EA will use some EA space that is shared by
4073 # ACL entries. So here we only check some reasonable ACL entries count,
4074 # instead of the max number that is calculated from the max_ea_size.
4075 if [ "$MDS1_VERSION" -lt $(version_code 2.8.57) ]; then
4076 count=28 # hard coded of RPC protocol
4077 elif large_xattr_enabled; then
4078 count=4500 # max_num 8187 max_ea_size = 65452
4079 # not create too many (4500) to save test time
4081 count=450 # max_num 497 max_ea_size = 4012
4084 echo "It is expected to hold at least $count ACL entries"
4085 trap cleanup_48 EXIT ERR
4086 for ((i = 0; i < $count; i++)) do
4087 setfacl -m u:$((i + 100)):rw $MOUNT/widestripe ||
4088 error "Fail to setfacl for $MOUNT/widestripe at $i"
4091 cancel_lru_locks mdc
4092 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
4093 local r_count=$(getfacl $MOUNT/widestripe | grep "user:" | wc -l)
4095 count=$((count + 1)) # for the entry "user::rw-"
4097 [ $count -eq $r_count ] ||
4098 error "Expected ACL entries $count, but got $r_count"
4102 run_test 48 "too many acls on file"
4104 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
4105 test_49a() { # bug 17710
4106 local timeout_orig=$TIMEOUT
4107 local ldlm_timeout_orig=$LDLM_TIMEOUT
4108 local LOCAL_TIMEOUT=20
4110 LDLM_TIMEOUT=$LOCAL_TIMEOUT
4111 TIMEOUT=$LOCAL_TIMEOUT
4115 check_mount || error "client mount failed"
4117 echo "check ldlm_timout..."
4118 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
4119 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
4120 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
4122 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
4123 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
4126 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
4127 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
4130 umount_client $MOUNT || error "umount_client $MOUNT failed"
4131 stop_ost || error "problem stopping OSS"
4132 stop_mds || error "problem stopping MDS"
4134 LDLM_TIMEOUT=$ldlm_timeout_orig
4135 TIMEOUT=$timeout_orig
4137 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
4139 test_49b() { # bug 17710
4140 local timeout_orig=$TIMEOUT
4141 local ldlm_timeout_orig=$LDLM_TIMEOUT
4142 local LOCAL_TIMEOUT=20
4144 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
4145 TIMEOUT=$LOCAL_TIMEOUT
4149 check_mount || error "client mount failed"
4151 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
4152 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
4153 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
4155 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
4156 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
4159 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
4160 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
4163 cleanup || error "cleanup failed"
4165 LDLM_TIMEOUT=$ldlm_timeout_orig
4166 TIMEOUT=$timeout_orig
4168 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
4171 # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1
4173 # Test both statfs and lfs df and fail if either one fails
4174 multiop_bg_pause $1 f_
4177 killall -USR1 multiop
4178 [ $RC -ne 0 ] && log "lazystatfs multiop failed"
4179 wait $PID || { RC=$?; log "multiop return error "; }
4181 # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1
4186 if kill -s 0 $PID; then
4189 log "lazystatfs lfs df failed to complete in 5s"
4197 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4198 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4200 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
4202 cleanup || error "cleanup failed with rc $?"
4204 run_test 50a "lazystatfs all servers available"
4208 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4209 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4211 # Wait for client to detect down OST
4212 stop_ost || error "Unable to stop OST1"
4213 wait_osc_import_state client ost DISCONN
4215 log "OSCs should all be DISCONN"
4217 lazystatfs $MOUNT || error "lazystatfs should not return EIO"
4219 umount_client $MOUNT || error "Unable to unmount client"
4220 stop_mds || error "Unable to stop MDS"
4222 run_test 50b "lazystatfs all servers down"
4225 start_mds || error "Unable to start MDS"
4226 start_ost || error "Unable to start OST1"
4227 start_ost2 || error "Unable to start OST2"
4228 mount_client $MOUNT || error "Unable to mount client"
4229 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4230 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4232 # Wait for client to detect down OST
4233 stop_ost || error "Unable to stop OST1"
4234 wait_osc_import_state mds ost DISCONN
4235 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
4237 umount_client $MOUNT || error "Unable to unmount client"
4238 stop_ost2 || error "Unable to stop OST2"
4239 stop_mds || error "Unable to stop MDS"
4240 #writeconf to remove all ost2 traces for subsequent tests
4241 writeconf_or_reformat
4243 run_test 50c "lazystatfs one server down"
4246 start_mds || error "Unable to start MDS"
4247 start_ost || error "Unable to start OST1"
4248 start_ost2 || error "Unable to start OST2"
4249 mount_client $MOUNT || error "Unable to mount client"
4250 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4251 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4253 # Issue the statfs during the window where the client still
4254 # belives the OST to be available but it is in fact down.
4255 # No failure just a statfs which hangs for a timeout interval.
4256 stop_ost || error "Unable to stop OST1"
4257 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
4259 umount_client $MOUNT || error "Unable to unmount client"
4260 stop_ost2 || error "Unable to stop OST2"
4261 stop_mds || error "Unable to stop MDS"
4262 #writeconf to remove all ost2 traces for subsequent tests
4263 writeconf_or_reformat
4265 run_test 50d "lazystatfs client/server conn race"
4272 start_mds || error "Unable to start MDS"
4273 #first client should see only one ost
4274 start_ost || error "Unable to start OST1"
4275 wait_osc_import_state mds ost FULL
4277 # Wait for client to detect down OST
4278 stop_ost || error "Unable to stop OST1"
4279 wait_osc_import_state mds ost DISCONN
4281 mount_client $MOUNT || error "Unable to mount client"
4282 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
4284 multiop_bg_pause $MOUNT _f
4288 if [ $RC1 -ne 0 ]; then
4289 log "multiop failed $RC1"
4292 sleep $(( $TIMEOUT+1 ))
4294 [ $? -ne 0 ] && error "process isn't sleep"
4295 start_ost || error "Unable to start OST1"
4296 wait $pid || error "statfs failed"
4299 umount_client $MOUNT || error "Unable to unmount client"
4300 stop_ost || error "Unable to stop OST1"
4301 stop_mds || error "Unable to stop MDS"
4303 run_test 50e "normal statfs all servers down"
4308 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
4310 start_mds || error "Unable to start mds"
4311 #first client should see only one ost
4312 start_ost || error "Unable to start OST1"
4313 wait_osc_import_state mds ost FULL
4315 start_ost2 || error "Unable to start OST2"
4316 wait_osc_import_state mds ost2 FULL
4318 # Wait for client to detect down OST
4319 stop_ost2 || error "Unable to stop OST2"
4321 wait_osc_import_state mds ost2 DISCONN
4322 mount_client $MOUNT || error "Unable to mount client"
4323 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
4325 multiop_bg_pause $MOUNT _f
4329 if [ $RC1 -ne 0 ]; then
4330 log "lazystatfs multiop failed $RC1"
4333 sleep $(( $TIMEOUT+1 ))
4335 [ $? -ne 0 ] && error "process isn't sleep"
4336 start_ost2 || error "Unable to start OST2"
4337 wait $pid || error "statfs failed"
4338 stop_ost2 || error "Unable to stop OST2"
4341 umount_client $MOUNT -f || error "Unable to unmount client"
4342 stop_ost || error "Unable to stop OST1"
4343 stop_mds || error "Unable to stop MDS"
4344 #writeconf to remove all ost2 traces for subsequent tests
4345 writeconf_or_reformat
4347 run_test 50f "normal statfs one server in down"
4350 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs"
4352 start_ost2 || error "Unable to start OST2"
4353 wait_osc_import_state mds ost2 FULL
4354 wait_osc_import_ready client ost2
4356 if [[ $PERM_CMD == *"set_param -P"* ]]; then
4357 local PARAM="osc.${FSNAME}-OST0001*.active"
4359 local PARAM="${FSNAME}-OST0001.osc.active"
4362 $LFS setstripe -c -1 $DIR/$tfile || error "$LFS setstripe failed"
4363 do_facet mgs $PERM_CMD $PARAM=0 || error "Unable to deactivate OST"
4365 umount_client $MOUNT || error "Unable to unmount client"
4366 mount_client $MOUNT || error "Unable to mount client"
4367 # This df should not cause a panic
4370 do_facet mgs $PERM_CMD $PARAM=1 || error "Unable to activate OST"
4371 rm -f $DIR/$tfile || error "unable to remove file $DIR/$tfile"
4372 umount_client $MOUNT || error "Unable to unmount client"
4373 stop_ost2 || error "Unable to stop OST2"
4374 stop_ost || error "Unable to stop OST1"
4375 stop_mds || error "Unable to stop MDS"
4376 #writeconf to remove all ost2 traces for subsequent tests
4377 writeconf_or_reformat
4379 run_test 50g "deactivated OST should not cause panic"
4383 # prepare MDT/OST, make OSC inactive for OST1
4384 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs"
4386 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
4387 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
4388 error "tunefs OST1 failed"
4389 start_mds || error "Unable to start MDT"
4390 start_ost || error "Unable to start OST1"
4391 start_ost2 || error "Unable to start OST2"
4392 mount_client $MOUNT || error "client start failed"
4394 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4396 # activatate OSC for OST1
4397 set_persistent_param_and_check client \
4398 "osc.${FSNAME}-OST0000-osc-[!M]*.active" \
4399 "${FSNAME}-OST0000.osc.active" 1
4401 mkdir $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
4402 $LFS setstripe -c -1 -i 0 $DIR/$tdir/2 ||
4403 error "$LFS setstripe $DIR/$tdir/2 failed"
4404 sleep 1 && echo "create a file after OST1 is activated"
4405 # doing some io, shouldn't crash
4406 dd if=/dev/zero of=$DIR/$tdir/2/$tfile-io bs=1M count=10
4408 # check OSC import is working
4409 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
4410 error "some OSC imports are still not connected"
4414 umount_client $MOUNT || error "Unable to umount client"
4415 stop_ost2 || error "Unable to stop OST2"
4416 cleanup_nocli || error "cleanup_nocli failed with $?"
4418 run_test 50h "LU-642: activate deactivated OST"
4421 # prepare MDT/OST, make OSC inactive for OST1
4422 [ "$MDSCOUNT" -lt "2" ] && skip "needs >= 2 MDTs"
4425 [ $(facet_fstype mds2) == zfs ] && import_zpool mds2
4426 do_facet mds2 "$TUNEFS --param mdc.active=0 $(mdsdevname 2)" ||
4427 error "tunefs MDT2 failed"
4428 start_mds || error "Unable to start MDT"
4429 start_ost || error "Unable to start OST1"
4430 start_ost2 || error "Unable to start OST2"
4431 mount_client $MOUNT || error "client start failed"
4433 mkdir_on_mdt0 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4435 if [[ $PERM_CMD == *"set_param -P"* ]]; then
4436 $PERM_CMD mdc.${FSNAME}-MDT0001-mdc-*.active=0 &&
4437 error "deactive MDC0 succeeds"
4439 $PERM_CMD ${FSNAME}-MDT0000.mdc.active=0 &&
4440 error "deactive MDC0 succeeds"
4443 # activate MDC for MDT2
4444 set_persistent_param_and_check client \
4445 "mdc.${FSNAME}-MDT0001-mdc-*.active" \
4446 "${FSNAME}-MDT0001.mdc.active" 1
4448 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
4449 if [ "$MDS1_VERSION" -ge $(version_code 2.7.60) ]
4451 wait_dne_interconnect
4453 $LFS mkdir -i1 $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
4455 createmany -o $DIR/$tdir/2/$tfile-%d 1 || error "create files failed"
4457 rm -rf $DIR/$tdir/2 || error "unlink dir failed"
4459 # deactivate MDC for MDT2
4460 set_persistent_param_and_check client \
4461 "mdc.${FSNAME}-MDT0001-mdc-*.active" \
4462 "${FSNAME}-MDT0001.mdc.active" 0
4464 wait_osp_active mds ${FSNAME}-MDT0001 1 0
4466 $LFS mkdir -i1 $DIR/$tdir/2 &&
4467 error "mkdir $DIR/$tdir/2 succeeds after deactive MDT"
4469 $LFS mkdir -i0 -c$MDSCOUNT $DIR/$tdir/striped_dir ||
4470 error "mkdir $DIR/$tdir/striped_dir fails after deactive MDT2"
4472 local stripe_count=$($LFS getdirstripe -c $DIR/$tdir/striped_dir)
4473 [ $stripe_count -eq $((MDSCOUNT - 1)) ] ||
4474 error "wrong $stripe_count != $((MDSCOUNT -1)) for striped_dir"
4477 umount_client $MOUNT || error "Unable to umount client"
4482 run_test 50i "activate deactivated MDT"
4485 local LOCAL_TIMEOUT=20
4489 check_mount || error "check_mount failed"
4491 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
4492 $LFS setstripe -c -1 $MOUNT/$tdir ||
4493 error "$LFS setstripe -c -1 $MOUNT/$tdir failed"
4494 #define OBD_FAIL_MDS_REINT_DELAY 0x142
4495 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x142"
4496 touch $MOUNT/$tdir/$tfile &
4499 start_ost2 || error "Unable to start OST1"
4501 stop_ost2 || error "Unable to stop OST1"
4502 umount_client $MOUNT -f || error "unmount $MOUNT failed"
4503 cleanup_nocli || error "stop server failed"
4504 #writeconf to remove all ost2 traces for subsequent tests
4505 writeconf_or_reformat
4507 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
4516 do_node $node mkdir -p $dest
4517 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
4519 do_node $node 'tar cf - '$*' | tar xf - -C '$dest';
4520 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
4521 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
4523 do_node $node 'getfattr -d -m "[a-z]*\\." '$*' > '$xattrs
4524 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
4534 local backup2=${TMP}/backup2
4536 do_node $node mkdir -p $backup2
4537 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
4539 do_node $node 'tar cf - '$*' | tar xf - -C '$backup2';
4540 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
4541 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
4543 do_node $node "diff -rq $backup $backup2"
4544 [ $? -eq 0 ] || { error "contents differ"; return 3; }
4546 local xattrs2=${TMP}/xattrs2
4547 do_node $node 'getfattr -d -m "[a-z]*\\." '$*' > '$xattrs2
4548 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
4550 do_node $node "diff $xattrs $xattrs2"
4551 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
4553 do_node $node "rm -rf $backup2 $xattrs2"
4554 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
4558 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4559 skip "ldiskfs only test"
4562 start_mds || error "Unable to start MDS"
4563 start_ost || error "Unable to start OST1"
4564 mount_client $MOUNT || error "Unable to mount client"
4567 local ost1mnt=$(facet_mntpt ost1)
4568 local ost1node=$(facet_active_host ost1)
4569 local ost1tmp=$TMP/conf52
4572 mkdir $DIR/$tdir || error "Unable to create $DIR/$tdir"
4573 touch $TMP/modified_first || error "Unable to create temporary file"
4574 local mtime=$(stat -c %Y $TMP/modified_first)
4575 do_node $ost1node "mkdir -p $ost1tmp &&
4576 touch -m -d @$mtime $ost1tmp/modified_first" ||
4577 error "Unable to create temporary file"
4580 $LFS setstripe -c -1 -S 1M $DIR/$tdir || error "$LFS setstripe failed"
4582 for (( i=0; i < nrfiles; i++ )); do
4583 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c ||
4584 error "multiop failed"
4589 # sync all the data and make sure no pending data on the client,
4590 # thus the SOM xattr would not be changed any more.
4591 cancel_lru_locks osc
4594 echo backup files to $TMP/$tdir
4595 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
4596 copy_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
4597 error "Unable to copy files"
4599 umount_client $MOUNT || error "Unable to umount client"
4600 stop_ost || error "Unable to stop ost1"
4602 echo mount ost1 as ldiskfs
4603 do_node $ost1node mkdir -p $ost1mnt || error "Unable to create $ost1mnt"
4604 if ! do_node $ost1node test -b $ost1_dev; then
4607 do_node $ost1node mount -t "$ost1_FSTYPE" $loop $ost1_dev \
4609 error "Unable to mount ost1 as ldiskfs"
4612 echo backup objects to $ost1tmp/objects
4613 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
4614 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
4615 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
4617 error "Unable to copy objects"
4619 # move objects to lost+found
4620 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
4621 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
4623 do_node $ost1node "umount $ost1mnt" ||
4624 error "Unable to umount ost1 as ldiskfs"
4626 start_ost || error "Unable to start OST1"
4627 mount_client $MOUNT || error "Unable to mount client"
4629 local REPAIRED=$(do_node $ost1node "$LCTL get_param \
4630 -n osd-ldiskfs.$FSNAME-OST0000.oi_scrub" |
4631 awk '/^lf_repa[ri]*ed/ { print $2 }')
4632 [ $REPAIRED -gt 0 ] ||
4633 error "Some entry under /lost+found should be repaired"
4636 diff_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
4637 error "Unable to diff files"
4639 rm -rf $TMP/$tdir $TMP/file_xattrs ||
4640 error "Unable to delete temporary files"
4641 do_node $ost1node "rm -rf $ost1tmp" ||
4642 error "Unable to delete temporary files"
4643 cleanup || error "cleanup failed with $?"
4645 run_test 52 "check recovering objects from lost+found"
4647 # Checks threads_min/max/started for some service
4649 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
4650 # parameter pattern prefix like 'ost.*.ost'.
4663 local msg="Insane $modname thread counts"
4664 local ncpts=$(check_cpt_number $facet)
4668 check_mount || return 41
4670 # We need to expand $parampat, but it may match multiple parameters, so
4671 # we'll pick the first one
4672 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
4673 error "Couldn't expand ${parampat}.threads_min parameter name"
4677 # Remove the .threads_min part
4678 paramp=${paramp%.threads_min}
4680 # Check for sanity in defaults
4681 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4683 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4685 tstarted=$(do_facet $facet "$LCTL get_param \
4686 -n ${paramp}.threads_started" || echo 0)
4687 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' ||
4689 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4691 nthrs=$(expr $tmax - $tmin)
4692 if [ $nthrs -lt $ncpts ]; then
4698 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
4699 skip_env "module parameter forced $facet thread count"
4701 # Check that we can change min/max
4702 do_facet $facet "$LCTL set_param \
4703 ${paramp}.threads_min=$((tmin + nthrs))"
4704 do_facet $facet "$LCTL set_param \
4705 ${paramp}.threads_max=$((tmax - nthrs))"
4706 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4708 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4710 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) &&
4711 $tmax2 == ($tmax - $nthrs)))' || return $?
4713 # Check that we can set min/max to the same value
4714 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4716 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$tmin"
4717 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4719 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4721 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
4723 # Check that we can't set max < min
4724 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$((tmin - 1))"
4725 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4727 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4729 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
4731 # We need to ensure that we get the module options desired; to do this
4732 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
4733 LOAD_MODULES_REMOTE=true
4736 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
4738 setmodopts -a $modname "$newvalue" oldvalue
4741 check_mount || return 41
4743 # Restore previous setting of MODOPTS_*
4744 setmodopts $modname "$oldvalue"
4746 # Check that $opts took
4747 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4749 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4751 tstarted=$(do_facet $facet \
4752 "$LCTL get_param -n ${paramp}.threads_started" || echo 0)
4753 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4762 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
4763 cleanup || error "cleanup failed with rc $?"
4765 run_test 53a "check OSS thread count params"
4769 local mds=$(do_facet $SINGLEMDS "$LCTL get_param \
4770 -N mds.*.*.threads_max 2>/dev/null")
4771 if [ -z "$mds" ]; then
4772 #running this on an old MDT
4773 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
4775 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
4777 cleanup || error "cleanup failed with $?"
4779 run_test 53b "check MDS thread count params"
4782 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4783 skip "ldiskfs only test"
4786 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p ||
4787 error "llverdev failed with rc=$?"
4790 run_test 54a "test llverdev and partial verify of device"
4793 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4794 skip "ldiskfs only test"
4798 run_llverfs $MOUNT -p || error "llverfs failed with rc=$?"
4799 cleanup || error "cleanup failed with rc=$?"
4801 run_test 54b "test llverfs and partial verify of filesystem"
4805 local max_ost_index=$1
4807 echo -n $(((max_ost_index + 1) * 8))
4811 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4812 skip "ldiskfs only test"
4815 local mdsdev=$(mdsdevname 1)
4816 local mdsvdev=$(mdsvdevname 1)
4820 if ! combined_mgs_mds; then
4821 stop_mgs || error "stopping MGS service failed"
4822 format_mgs || error "formatting MGT failed"
4824 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
4826 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
4827 --reformat $(ostdevname 1) $(ostvdevname 1)
4833 echo checking size of lov_objid for ost index $i
4834 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" |
4835 grep ^User | awk -F 'Size: ' '{print $2}')
4836 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
4837 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
4839 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
4846 run_test 55 "check lov_objid size"
4849 local mds_journal_size_orig=$MDSJOURNALSIZE
4855 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
4856 $(ostdevname 1) $(ostvdevname 1)
4857 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
4858 $(ostdevname 2) $(ostvdevname 2)
4861 start_ost || error "Unable to start first ost (idx 10000)"
4862 start_ost2 || error "Unable to start second ost (idx 1000)"
4863 mount_client $MOUNT || error "Unable to mount client"
4867 # test instantiating PFL components with sparse index LU-15513
4868 mkdir -p $MOUNT/$tdir
4869 $LFS setstripe -E 4M -c 1 -E 1G -c 4 -S4M -E eof -c -1 $MOUNT/$tdir
4870 dd if=/dev/zero of=$MOUNT/$tdir/$tfile bs=4K count=1 seek=10k ||
4871 error "dd to second component failed"
4873 if [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
4874 [[ "$MDS1_VERSION" -ge $(version_code 2.5.4) &&
4875 "$MDS1_VERSION" -lt $(version_code 2.5.11) ]]; then
4876 wait_osc_import_state mds ost1 FULL
4877 wait_osc_import_state mds ost2 FULL
4878 $LFS setstripe --stripe-count=-1 $DIR/$tfile ||
4879 error "Unable to setstripe $DIR/$tfile"
4880 n=$($LFS getstripe --stripe-count $DIR/$tfile)
4881 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
4886 MDSJOURNALSIZE=$mds_journal_size_orig
4889 run_test 56a "check big OST indexes and out-of-index-order start"
4894 umount_client $MOUNT -f || error "unmount client failed"
4903 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs"
4905 trap cleanup_56b EXIT RETURN ERR
4908 if ! combined_mgs_mds ; then
4913 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) --index=0 --reformat \
4914 $(mdsdevname 1) $(mdsvdevname 1)
4915 add mds2 $(mkfs_opts mds2 $(mdsdevname 2)) --index=1 --reformat \
4916 $(mdsdevname 2) $(mdsvdevname 2)
4917 add mds3 $(mkfs_opts mds3 $(mdsdevname 3)) --index=1000 --reformat \
4918 $(mdsdevname 3) $(mdsvdevname 3)
4922 start_mdt 1 || error "MDT 1 (idx 0) start failed"
4923 start_mdt 2 || error "MDT 2 (idx 1) start failed"
4924 start_mdt 3 || error "MDT 3 (idx 1000) start failed"
4925 start_ost || error "Unable to start first ost"
4926 start_ost2 || error "Unable to start second ost"
4928 do_nodes $(comma_list $(mdts_nodes)) \
4929 "$LCTL set_param mdt.*.enable_remote_dir=1 \
4930 mdt.*.enable_remote_dir_gid=-1"
4932 mount_client $MOUNT || error "Unable to mount client"
4934 $LFS mkdir -c3 $MOUNT/$tdir || error "failed to make testdir"
4936 echo "This is test file 1!" > $MOUNT/$tdir/$tfile.1 ||
4937 error "failed to make test file 1"
4938 echo "This is test file 2!" > $MOUNT/$tdir/$tfile.2 ||
4939 error "failed to make test file 2"
4940 echo "This is test file 1000!" > $MOUNT/$tdir/$tfile.1000 ||
4941 error "failed to make test file 1000"
4943 rm -rf $MOUNT/$tdir || error "failed to remove testdir"
4945 $LFS mkdir -i1000 $MOUNT/$tdir.1000 ||
4946 error "create remote dir at idx 1000 failed"
4949 echo "=== START lfs df OUTPUT ==="
4951 echo "==== END lfs df OUTPUT ===="
4953 mdtcnt=$(echo -e "$output" | grep $FSNAME-MDT | wc -l)
4954 ostcnt=$(echo -e "$output" | grep $FSNAME-OST | wc -l)
4956 echo "lfs df returned mdt count $mdtcnt and ost count $ostcnt"
4957 [ $mdtcnt -eq 3 ] || error "lfs df returned wrong mdt count"
4958 [ $ostcnt -eq 2 ] || error "lfs df returned wrong ost count"
4960 echo "This is test file 1!" > $MOUNT/$tdir.1000/$tfile.1 ||
4961 error "failed to make test file 1"
4962 echo "This is test file 2!" > $MOUNT/$tdir.1000/$tfile.2 ||
4963 error "failed to make test file 2"
4964 echo "This is test file 1000!" > $MOUNT/$tdir.1000/$tfile.1000 ||
4965 error "failed to make test file 1000"
4966 rm -rf $MOUNT/$tdir.1000 || error "failed to remove remote_dir"
4969 echo "=== START lfs mdts OUTPUT ==="
4971 echo "==== END lfs mdts OUTPUT ===="
4973 echo -e "$output" | grep -v "MDTS:" | awk '{print $1}' |
4974 sed 's/://g' > $TMP/mdts-actual.txt
4975 sort $TMP/mdts-actual.txt -o $TMP/mdts-actual.txt
4977 echo -e "0\n1\n1000" > $TMP/mdts-expected.txt
4979 diff $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4982 rm $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4984 [ $result -eq 0 ] || error "target_obd proc file is incorrect!"
4986 run_test 56b "test target_obd correctness with nonconsecutive MDTs"
4988 test_57a() { # bug 22656
4989 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4990 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4991 tail -1 | awk '{print $1}')
4992 writeconf_or_reformat
4993 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
4994 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" ||
4995 error "tunefs failed"
4997 start_ost && error "OST registration from failnode should fail"
5000 run_test 57a "initial registration from failnode should fail (should return errs)"
5003 do_rpc_nodes $(facet_active_host ost1) load_modules_local
5004 local NID=$(do_facet ost1 "$LCTL get_param nis" |
5005 tail -1 | awk '{print $1}')
5006 writeconf_or_reformat
5007 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
5008 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" ||
5009 error "tunefs failed"
5011 start_ost || error "OST registration from servicenode should not fail"
5014 run_test 57b "initial registration from servicenode should not fail"
5017 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
5020 test_58() { # bug 22658
5021 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
5023 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5024 createmany -o $DIR/$tdir/$tfile-%d 100
5025 unlinkmany $DIR/$tdir/$tfile-%d 100
5026 stop_mds || error "Unable to stop MDS"
5028 local MNTDIR=$(facet_mntpt $SINGLEMDS)
5029 local devname=$(mdsdevname ${SINGLEMDS//mds/})
5031 # remove all files from the OBJECTS dir
5032 mount_fstype $SINGLEMDS
5034 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
5036 unmount_fstype $SINGLEMDS
5037 # restart MDS with missing llog files
5038 start_mds || error "unable to start MDS"
5039 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0"
5042 run_test 58 "missing llog files must not prevent MDT from mounting"
5045 start_mgsmds >> /dev/null
5046 local C1=$(count_osts)
5047 if [ $C1 -eq 0 ]; then
5048 start_ost >> /dev/null
5052 echo "original ost count: $C1 (expect > 0)"
5053 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
5054 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
5055 local C2=$(count_osts)
5056 echo "after mdt writeconf count: $C2 (expect 0)"
5057 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
5058 echo "OST start without writeconf should fail:"
5059 start_ost >> /dev/null &&
5060 error "OST start without writeconf didn't fail"
5061 echo "OST start with writeconf should succeed:"
5062 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
5063 local C3=$(count_osts)
5064 echo "after ost writeconf count: $C3 (expect 1)"
5065 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
5066 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
5067 local C4=$(count_osts)
5068 echo "after ost2 writeconf count: $C4 (expect 2)"
5069 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
5070 stop_ost2 >> /dev/null
5071 cleanup_nocli >> /dev/null
5072 #writeconf to remove all ost2 traces for subsequent tests
5073 writeconf_or_reformat
5075 run_test 59 "writeconf mount option"
5077 test_60a() { # LU-471
5078 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5079 skip "ldiskfs only test"
5084 for num in $(seq $MDSCOUNT); do
5085 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
5086 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
5087 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
5091 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
5092 [ ${PIPESTATUS[0]} -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
5094 # MDT default has dirdata feature
5095 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
5096 # we disable uninit_bg feature
5097 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
5098 # we set stride extended options
5099 echo $dump | grep stride > /dev/null || error "stride is not set"
5103 run_test 60a "check mkfs.lustre --mkfsoptions -E -O options setting"
5106 [[ "$mds1_FSTYPE" == ldiskfs ]] || skip "ldiskfs only test"
5108 local features=$(do_facet $SINGLEMDS $DUMPE2FS $(mdsdevname 1) |
5110 [ ${PIPESTATUS[0]} -eq 0 ] || error "$DUMPE2FS $(mdsdevname 1) failed"
5113 # ea_inode feature should be enabled by default for MDTs
5114 [[ "$features" =~ "ea_inode" ]] || error "ea_inode is not set"
5115 # large_dir feature should be enabled by default for MDTs
5116 [[ "$features" =~ "large_dir" ]] || error "large_dir is not set"
5118 run_test 60b "check mkfs.lustre MDT default features"
5120 test_61a() { # LU-80
5121 local lxattr=$(large_xattr_enabled)
5123 (( "$MDS1_VERSION" >= $(version_code 2.1.53) )) ||
5124 skip "Need MDS version at least 2.1.53 for large_xattr"
5126 if [[ "$mds1_FSTYPE" == ldiskfs ]] && ! large_xattr_enabled; then
5129 for ((num=1; num <= $MDSCOUNT; num++)); do
5130 do_facet mds${num} $TUNE2FS -O ea_inode \
5131 $(mdsdevname $num) ||
5132 error "tune2fs on mds $num failed"
5136 setup || error "setting up the filesystem failed"
5137 client_up || error "starting client failed"
5139 local file=$DIR/$tfile
5140 touch $file || error "touch $file failed"
5142 local large_value="$(generate_string $(max_xattr_size))"
5143 local small_value="bar"
5145 local name="trusted.big"
5146 log "save large xattr of $(max_xattr_size) bytes on $name on $file"
5147 setfattr -n $name -v $large_value $file ||
5148 error "saving $name on $file failed"
5150 local new_value=$(get_xattr_value $name $file)
5151 [[ "$new_value" != "$large_value" ]] &&
5152 error "$name different after saving"
5154 log "shrink value of $name on $file"
5155 setfattr -n $name -v $small_value $file ||
5156 error "shrinking value of $name on $file failed"
5158 new_value=$(get_xattr_value $name $file)
5159 [[ "$new_value" != "$small_value" ]] &&
5160 error "$name different after shrinking"
5162 log "grow value of $name on $file"
5163 setfattr -n $name -v $large_value $file ||
5164 error "growing value of $name on $file failed"
5166 new_value=$(get_xattr_value $name $file)
5167 [[ "$new_value" != "$large_value" ]] &&
5168 error "$name different after growing"
5170 log "check value of $name on $file after remounting MDS"
5172 new_value=$(get_xattr_value $name $file)
5173 [[ "$new_value" != "$large_value" ]] &&
5174 error "$name different after remounting MDS"
5176 log "remove large xattr $name from $file"
5177 setfattr -x $name $file || error "removing $name from $file failed"
5179 if $lxattr && [ "$mds1_FSTYPE" == ldiskfs ]; then
5180 stopall || error "stopping for e2fsck run"
5181 for num in $(seq $MDSCOUNT); do
5182 run_e2fsck $(facet_active_host mds$num) \
5183 $(mdsdevname $num) "-y" ||
5184 error "e2fsck MDT$num failed"
5186 setup_noconfig || error "remounting the filesystem failed"
5189 # need to delete this file to avoid problems in other tests
5191 cleanup || error "stopping systems failed"
5193 run_test 61a "large xattr"
5195 test_61b() { # LU-80
5196 local lxattr=$(large_xattr_enabled)
5198 (( $MDS1_VERSION >= $(version_code 2.15.51) )) ||
5199 skip "Need MDS version at least 2.15.51 for large_xattr fix"
5201 [[ "$mds1_FSTYPE" == "ldiskfs" ]] || skip "ldiskfs specific bug"
5203 if ! large_xattr_enabled; then
5206 for (( num=1; num <= $MDSCOUNT; num++ )); do
5207 do_facet mds${num} $TUNE2FS -O ea_inode \
5208 $(mdsdevname $num) ||
5209 error "tune2fs on mds $num failed"
5213 setup || error "setting up the filesystem failed"
5214 client_up || error "starting client failed"
5216 local _file=$MOUNT/panda
5217 local large_value="$(generate_string $(max_xattr_size))"
5218 local name="trusted.big"
5220 touch ${_file} || error "touch ${_file} failed"
5221 setfattr -n $name -v $large_value ${_file} ||
5222 error "saving $name on $file failed"
5224 MDT_DEV="${FSNAME}-MDT0000"
5225 MDT_DEVNAME=$(mdsdevname ${SINGLEMDS//mds/})
5227 stopall || error "stopping for e2fsck run"
5230 ino=$(do_facet $SINGLEMDS "$DEBUGFS -R 'stat /ROOT/panda' \
5231 ${MDT_DEVNAME} | grep trusted.big")
5232 ino=$(echo "${ino}" | awk '{print $2;}')
5233 echo "large ea "${ino}
5235 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"ln $ino /lost+found\\\" \
5238 setup_noconfig || error "remounting the filesystem failed"
5240 do_facet $SINGLEMDS $LCTL lfsck_start -M ${MDT_DEV} -t namespace || {
5241 error "can't start lfsck namespace"
5245 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
5246 mdd.${MDT_DEV}.lfsck_namespace |
5247 awk '/^status/ { print \\\$2 }'" "completed" 32 || {
5248 error "(2) unexpected status"
5251 stopall || error "stopping for e2fsck run"
5252 for num in $(seq $MDSCOUNT); do
5253 run_e2fsck $(facet_active_host mds$num) \
5254 $(mdsdevname $num) "-y" ||
5255 error "e2fsck MDT$num failed"
5257 setup_noconfig || error "remounting the filesystem failed"
5259 # need to delete this file to avoid problems in other tests
5261 cleanup || error "stopping systems failed"
5263 run_test 61b "large xattr"
5266 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5267 skip "ldiskfs only test"
5269 [[ "$MDS1_VERSION" -ge $(version_code 2.2.51) ]] ||
5270 skip "Need MDS version at least 2.2.51"
5273 local mdsdev=$(mdsdevname 1)
5274 local ostdev=$(ostdevname 1)
5276 echo "disable journal for mds"
5277 do_facet mds1 $TUNE2FS -O ^has_journal $mdsdev || error "tune2fs failed"
5278 start_mds && error "MDT start should fail"
5279 echo "disable journal for ost"
5280 do_facet ost1 $TUNE2FS -O ^has_journal $ostdev || error "tune2fs failed"
5281 start_ost && error "OST start should fail"
5282 cleanup || error "cleanup failed with rc $?"
5285 run_test 62 "start with disabled journal"
5288 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5289 skip "ldiskfs only test"
5292 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ldiskfs
5293 local inode_slab=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
5294 awk '/ldiskfs_inode_cache/ { print $5 / $6 }')
5295 if [ -z "$inode_slab" ]; then
5296 skip "ldiskfs module has not been loaded"
5299 if grep -q "CONFIG_DEBUG_LOCK_ALLOC=y" /boot/config-$(uname -r); then
5300 skip "test is not compatible with CONFIG_DEBUG_LOCK_ALLOC=y"
5303 echo "$inode_slab ldiskfs inodes per page"
5304 [ "${inode_slab%.*}" -ge "3" ] && return 0
5306 # If kmalloc-128 is also 1 per page - this is a debug kernel
5307 # and so this is not an error.
5308 local kmalloc128=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
5309 awk '/^(dma-kmalloc|size)-128 / { print $5 / $6 }')
5310 # 32 128-byte chunks in 4k
5311 [ "${kmalloc128%.*}" -lt "32" ] ||
5312 error "ldiskfs inode too big, only $inode_slab objs/page, " \
5313 "kmalloc128 = $kmalloc128 objs/page"
5315 run_test 63 "Verify each page can at least hold 3 ldiskfs inodes"
5318 start_mds || error "unable to start MDS"
5319 start_ost || error "Unable to start OST1"
5320 start_ost2 || error "Unable to start second ost"
5321 mount_client $MOUNT || error "Unable to mount client"
5322 stop_ost2 || error "Unable to stop second ost"
5325 umount_client $MOUNT -f || error "unmount $MOUNT failed"
5326 cleanup_nocli || error "cleanup_nocli failed with $?"
5327 #writeconf to remove all ost2 traces for subsequent tests
5328 writeconf_or_reformat
5330 run_test 64 "check lfs df --lazy "
5332 test_65() { # LU-2237
5333 # Currently, the test is only valid for ldiskfs backend
5334 [ "$mds1_FSTYPE" != ldiskfs ] &&
5335 skip "ldiskfs only test"
5337 local devname=$(mdsdevname ${SINGLEMDS//mds/})
5338 local brpt=$(facet_mntpt brpt)
5341 if ! do_facet $SINGLEMDS "test -b $devname"; then
5345 stop_mds || error "Unable to stop MDS"
5346 local obj=$(do_facet $SINGLEMDS \
5347 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
5349 if [ -z "$obj" ]; then
5350 # The MDT may be just re-formatted, mount the MDT for the
5351 # first time to guarantee the "last_rcvd" file is there.
5352 start_mds || error "fail to mount the MDS for the first time"
5353 stop_mds || error "Unable to stop MDS"
5356 # remove the "last_rcvd" file
5357 do_facet $SINGLEMDS "mkdir -p $brpt"
5358 do_facet $SINGLEMDS \
5359 "mount -t $mds1_FSTYPE $opts $devname $brpt"
5360 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
5361 do_facet $SINGLEMDS "$UMOUNT $brpt"
5363 # restart MDS, the "last_rcvd" file should be recreated.
5364 start_mds || error "fail to restart the MDS"
5365 stop_mds || error "Unable to stop MDS"
5366 obj=$(do_facet $SINGLEMDS \
5367 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
5368 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
5370 run_test 65 "re-create the lost last_rcvd file when server mount"
5373 [[ "$MGS_VERSION" -ge $(version_code 2.3.59) ]] ||
5374 skip "Need MGS version at least 2.3.59"
5377 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
5378 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
5380 # add EXCLUDE records to config log, they are not to be
5381 # removed by lctl replace_nids
5382 set_conf_param_and_check mds \
5383 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
5384 "$FSNAME-OST0000.osc.active" \
5387 echo "replace_nids should fail if MDS, OSTs and clients are UP"
5388 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5389 error "replace_nids fail"
5391 umount_client $MOUNT || error "unmounting client failed"
5392 echo "replace_nids should fail if MDS and OSTs are UP"
5393 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5394 error "replace_nids fail"
5396 stop_ost || error "Unable to stop OST1"
5397 echo "replace_nids should fail if MDS is UP"
5398 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5399 error "replace_nids fail"
5401 stop_mds || error "stopping mds failed"
5403 if combined_mgs_mds; then
5404 start_mdt 1 "-o nosvc" ||
5405 error "starting mds with nosvc option failed"
5408 echo "command should accept two parameters"
5409 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
5410 error "command should accept two params"
5412 echo "correct device name should be passed"
5413 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
5414 error "wrong devname"
5416 echo "wrong nids list should not destroy the system"
5417 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
5419 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "asdfasdf, asdfadf" &&
5422 echo "replace OST nid"
5423 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
5424 error "replace nids failed"
5426 echo "command should accept two parameters"
5427 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
5428 error "command should accept two params"
5430 echo "wrong nids list should not destroy the system"
5431 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
5434 local FAKE_NIDS="192.168.0.112@tcp1,192.168.0.112@tcp2"
5435 local FAKE_FAILOVER="192.168.0.113@tcp1,192.168.0.113@tcp2"
5436 local NIDS_AND_FAILOVER="$MDS_NID,$FAKE_NIDS:$FAKE_FAILOVER"
5437 echo "set NIDs with failover"
5438 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $NIDS_AND_FAILOVER ||
5439 error "replace nids failed"
5442 echo "replace MDS nid"
5443 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
5444 error "replace nids failed"
5446 if ! combined_mgs_mds ; then
5449 stop_mds || error "Unable to stop MDS"
5452 start_mgsmds || error "start mgsmds failed"
5453 set_conf_param_and_check mds \
5454 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
5455 "$FSNAME-OST0000.osc.active" \
5457 start_ost || error "unable to start OST"
5458 mount_client $MOUNT || error "mount client failed"
5460 check_mount || error "error after nid replace"
5461 cleanup || error "cleanup failed"
5464 run_test 66 "replace nids"
5466 test_67() { #LU-2950
5467 local legacy="$TMP/legacy_lnet_config"
5468 local new="$TMP/new_routes_test"
5469 local out="$TMP/config_out_file"
5470 local verify="$TMP/conv_verify"
5471 local verify_conf="$TMP/conf_verify"
5473 # Create the legacy file that will be run through the
5474 # lustre_routes_conversion script
5475 cat <<- LEGACY_LNET_CONFIG > $legacy
5476 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
5477 tcp2 54 10.1.3.2@tcp;
5478 tcp3 10.3.4.3@tcp:3;
5482 # Create the verification file to verify the output of
5483 # lustre_routes_conversion script against.
5484 cat <<- VERIFY_LNET_CONFIG > $verify
5485 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
5486 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
5487 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
5488 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
5489 tcp4: { gateway: 10.3.3.4@tcp }
5492 # Create the verification file to verify the output of
5493 # lustre_routes_config script against
5494 cat <<- VERIFY_LNET_CONFIG > $verify_conf
5495 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
5496 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
5497 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
5498 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
5499 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
5502 $LUSTRE_ROUTES_CONVERSION $legacy $new > /dev/null
5503 if [ -f $new ]; then
5504 # verify the conversion output
5505 cmp -s $new $verify > /dev/null
5506 if [ $? -eq 1 ]; then
5507 error "routes conversion failed"
5510 lustre_routes_config --dry-run --verbose $new > $out
5511 # check that the script succeeded
5512 cmp -s $out $verify_conf > /dev/null
5513 if [ $? -eq 1 ]; then
5514 error "routes config failed"
5517 error "routes conversion test failed"
5519 # remove generated files
5520 rm -f $new $legacy $verify $verify_conf $out
5522 run_test 67 "test routes conversion and configuration"
5530 [ "$MDS1_VERSION" -ge $(version_code 2.4.53) ] ||
5531 skip "Need MDS version at least 2.4.53"
5533 umount_client $MOUNT || error "umount client failed"
5538 # START-END - the sequences we'll be reserving
5539 START=$(do_facet $SINGLEMDS \
5540 $LCTL get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
5541 END=$((START + (1 << 30)))
5542 do_facet $SINGLEMDS \
5543 $LCTL set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
5545 # reset the sequences MDT0000 has already assigned
5546 do_facet $SINGLEMDS \
5547 $LCTL set_param seq.srv*MDT0000.space=clear
5549 # remount to let the client allocate new sequence
5550 mount_client $MOUNT || error "mount client failed"
5552 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
5553 do_facet $SINGLEMDS \
5554 $LCTL get_param seq.srv*MDT0000.space
5555 $LFS path2fid $DIR/$tfile
5557 local old_ifs="$IFS"
5559 fid=($($LFS path2fid $DIR/$tfile))
5563 if [[ $seq < $END ]]; then
5564 error "used reserved sequence $seq?"
5566 cleanup || error "cleanup failed with $?"
5568 run_test 68 "be able to reserve specific sequences in FLDB"
5570 # Test 69: is about the total number of objects ever created on an OST.
5571 # so that when it is reformatted the normal MDS->OST orphan recovery won't
5572 # just "precreate" the missing objects. In the past it might try to recreate
5573 # millions of objects after an OST was reformatted
5575 [[ "$MDS1_VERSION" -lt $(version_code 2.4.2) ]] &&
5576 skip "Need MDS version at least 2.4.2"
5578 [[ "$MDS1_VERSION" -ge $(version_code 2.4.50) ]] &&
5579 [[ "$MDS1_VERSION" -lt $(version_code 2.5.0) ]] &&
5580 skip "Need MDS version at least 2.5.0"
5583 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5584 do_nodes $(comma_list $(osts_nodes)) $LCTL set_param \
5585 seq.*OST*-super.width=$DATA_SEQ_MAX_WIDTH
5587 # use OST0000 since it probably has the most creations
5588 local OSTNAME=$(ostname_from_index 0)
5589 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
5590 local last_id=$(do_facet mds1 $LCTL get_param -n \
5591 osp.$mdtosc_proc1.prealloc_last_id)
5593 # Want to have OST LAST_ID over 5 * OST_MAX_PRECREATE to
5594 # verify that the LAST_ID recovery is working properly. If
5595 # not, then the OST will refuse to allow the MDS connect
5596 # because the LAST_ID value is too different from the MDS
5597 #define OST_MAX_PRECREATE=20000
5598 local ost_max_pre=20000
5599 local num_create=$(( ost_max_pre * 5 + 1 - last_id))
5601 # If the LAST_ID is already over 5 * OST_MAX_PRECREATE, we don't
5602 # need to create any files. So, skip this section.
5603 if [ $num_create -gt 0 ]; then
5604 # Check the number of inodes available on OST0
5606 local ifree=$($LFS df -i $MOUNT |
5607 awk '/OST0000/ { print $4 }'; exit ${PIPESTATUS[0]})
5608 log "On OST0, $ifree inodes available. Want $num_create. rc=$?"
5610 $LFS setstripe -i 0 $DIR/$tdir ||
5611 error "$LFS setstripe -i 0 $DIR/$tdir failed"
5612 if [ $ifree -lt 10000 ]; then
5613 files=$(( ifree - 50 ))
5618 local j=$((num_create / files + 1))
5619 for i in $(seq 1 $j); do
5620 createmany -o $DIR/$tdir/$tfile-$i- $files ||
5621 error "createmany fail create $files files: $?"
5622 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
5623 error "unlinkmany failed unlink $files files"
5627 # delete all of the files with objects on OST0 so the
5628 # filesystem is not inconsistent later on
5629 $LFS find $MOUNT --ost 0 -print0 | xargs -0 rm
5631 umount_client $MOUNT || error "umount client failed"
5632 stop_ost || error "OST0 stop failure"
5633 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
5634 $(ostdevname 1) $(ostvdevname 1) ||
5635 error "reformat and replace $ostdev failed"
5636 start_ost || error "OST0 restart failure"
5637 wait_osc_import_state mds ost FULL
5639 mount_client $MOUNT || error "mount client failed"
5640 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
5641 local idx=$($LFS getstripe -i $DIR/$tdir/$tfile-last)
5642 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
5644 local iused=$($LFS df -i $MOUNT |
5645 awk '/OST0000/ { print $3 }'; exit ${PIPESTATUS[0]})
5646 log "On OST0, $iused used inodes rc=$?"
5647 [ $iused -ge $((ost_max_pre + 1000)) ] &&
5648 error "OST replacement created too many inodes; $iused"
5649 cleanup || error "cleanup failed with $?"
5651 run_test 69 "replace an OST with the same index"
5654 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5657 cleanup || error "cleanup failed with $?"
5659 start_mdt 1 || error "MDT0 start fail"
5661 start_ost || error "OST0 start fail"
5662 for num in $(seq 2 $MDSCOUNT); do
5663 start_mdt $num || return
5666 mount_client $MOUNT || error "mount client fails"
5668 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
5670 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5671 error "create remote dir fail"
5673 rm -rf $DIR/$tdir || error "delete dir fail"
5674 cleanup || error "cleanup failed with $?"
5676 run_test 70a "start MDT0, then OST, then MDT1"
5679 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5682 start_ost || error "OST0 start fail"
5684 start_mds || error "MDS start fail"
5686 mount_client $MOUNT || error "mount client fails"
5688 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
5690 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5691 error "create remote dir fail"
5693 rm -rf $DIR/$tdir || error "delete dir fail"
5695 cleanup || error "cleanup failed with $?"
5697 run_test 70b "start OST, MDT1, MDT0"
5700 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5703 start_mds || error "MDS start fail"
5704 start_ost || error "OST0 start fail"
5706 mount_client $MOUNT || error "mount client fails"
5707 stop_mdt 1 || error "MDT1 start fail"
5709 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
5710 echo "deactivate $mdc_for_mdt1"
5711 $LCTL --device $mdc_for_mdt1 deactivate ||
5712 error "set $mdc_for_mdt1 deactivate failed"
5714 mkdir $DIR/$tdir && error "mkdir succeed"
5716 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
5717 error "create remote dir succeed"
5719 cleanup || error "cleanup failed with $?"
5721 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
5724 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5727 start_mds || error "MDS start fail"
5728 start_ost || error "OST0 start fail"
5730 mount_client $MOUNT || error "mount client fails"
5732 stop_mdt 2 || error "MDT1 start fail"
5734 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
5736 echo "deactivate $mdc_for_mdt2"
5737 $LCTL --device $mdc_for_mdt2 deactivate ||
5738 error "set $mdc_for_mdt2 deactivate failed"
5740 mkdir $DIR/$tdir || error "mkdir fail"
5741 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
5742 error "create remote dir succeed"
5744 rm -rf $DIR/$tdir || error "delete dir fail"
5746 cleanup || error "cleanup failed with $?"
5748 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
5751 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5753 [ "$MDS1_VERSION" -ge $(version_code 2.7.62) ] ||
5754 skip "Need MDS version at least 2.7.62"
5756 reformat || error "reformat failed with $?"
5760 local mdsdev=$(mdsdevname 1)
5761 local ostdev=$(ostdevname 1)
5762 local mdsvdev=$(mdsvdevname 1)
5763 local ostvdev=$(ostvdevname 1)
5764 local opts_mds="$(mkfs_opts mds1 $mdsdev) --reformat $mdsdev $mdsvdev"
5765 local opts_ost="$(mkfs_opts ost1 $ostdev) --reformat $ostdev $ostvdev"
5767 if ! combined_mgs_mds ; then
5771 add mds1 $opts_mds || error "add mds1 failed"
5772 start_mdt 1 || error "start mdt1 failed"
5773 add ost1 $opts_ost || error "add ost1 failed"
5774 start_ost || error "start ost failed"
5775 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
5777 local soc=$(do_facet mds1 "$LCTL get_param -n \
5778 mdt.*MDT0000.sync_lock_cancel")
5779 [ $soc == "never" ] || error "SoC enabled on single MDS"
5781 for i in $(seq 2 $MDSCOUNT); do
5782 mdsdev=$(mdsdevname $i)
5783 mdsvdev=$(mdsvdevname $i)
5784 opts_mds="$(mkfs_opts mds$i $mdsdev) --reformat $mdsdev \
5786 add mds$i $opts_mds || error "add mds$i failed"
5787 start_mdt $i || error "start mdt$i fail"
5790 wait_dne_interconnect
5792 for i in $(seq $MDSCOUNT); do
5793 soc=$(do_facet mds$i "$LCTL get_param -n \
5794 mdt.*MDT000$((i - 1)).sync_lock_cancel")
5795 [ $soc == "blocking" ] || error "SoC not enabled on DNE"
5798 for i in $(seq 2 $MDSCOUNT); do
5799 stop_mdt $i || error "stop mdt$i fail"
5801 soc=$(do_facet mds1 "$LCTL get_param -n \
5802 mdt.*MDT0000.sync_lock_cancel")
5803 [ $soc == "never" ] || error "SoC enabled on single MDS"
5804 umount_client $MOUNT -f > /dev/null
5806 cleanup || error "cleanup failed with $?"
5808 run_test 70e "Sync-on-Cancel will be enabled by default on DNE"
5811 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5812 if combined_mgs_mds; then
5813 skip "needs separate MGS/MDT"
5817 start_mdt 1 || error "MDT0 start fail"
5818 start_ost || error "OST0 start fail"
5819 for num in $(seq 2 $MDSCOUNT); do
5820 start_mdt $num || return
5823 start_ost2 || error "OST1 start fail"
5825 mount_client $MOUNT || error "mount client fails"
5827 mkdir $DIR/$tdir || error "mkdir fail"
5828 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5829 error "create remote dir succeed"
5831 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5832 rm -rf $DIR/$tdir || error "delete dir fail"
5834 umount_client $MOUNT || error "umount_client failed"
5835 stop_mds || error "MDS stop fail"
5836 stop_ost || error "OST0 stop fail"
5837 stop_ost2 || error "OST1 stop fail"
5839 run_test 71a "start MDT0 OST0, MDT1, OST1"
5842 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5843 if combined_mgs_mds; then
5844 skip "needs separate MGS/MDT"
5848 for num in $(seq 2 $MDSCOUNT); do
5849 start_mdt $num || return
5851 start_ost || error "OST0 start fail"
5852 start_mdt 1 || error "MDT0 start fail"
5853 start_ost2 || error "OST1 start fail"
5855 mount_client $MOUNT || error "mount client fails"
5857 mkdir $DIR/$tdir || error "mkdir fail"
5858 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5859 error "create remote dir succeed"
5861 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5862 rm -rf $DIR/$tdir || error "delete dir fail"
5864 umount_client $MOUNT || error "umount_client failed"
5865 stop_mds || error "MDT0 stop fail"
5866 stop_ost || error "OST0 stop fail"
5867 stop_ost2 || error "OST1 stop fail"
5869 run_test 71b "start MDT1, OST0, MDT0, OST1"
5872 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5873 combined_mgs_mds && skip "needs separate MGS/MDT"
5877 start_ost || error "OST0 start fail"
5878 start_ost2 || error "OST1 start fail"
5879 for num in $(seq 2 $MDSCOUNT); do
5880 start_mdt $num || return
5882 start_mdt 1 || error "MDT0 start fail"
5884 mount_client $MOUNT || error "mount client fails"
5886 mkdir $DIR/$tdir || error "mkdir fail"
5887 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5888 error "create remote dir succeed"
5890 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5891 rm -rf $DIR/$tdir || error "delete dir fail"
5893 umount_client $MOUNT || error "umount_client failed"
5894 stop_mds || error "MDS stop fail"
5895 stop_ost || error "OST0 stop fail"
5896 stop_ost2 || error "OST1 stop fail"
5899 run_test 71c "start OST0, OST1, MDT1, MDT0"
5902 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5903 combined_mgs_mds && skip "needs separate MGS/MDT"
5907 start_ost || error "OST0 start fail"
5908 for num in $(seq 2 $MDSCOUNT); do
5909 start_mdt $num || return
5911 start_mdt 1 || error "MDT0 start fail"
5912 start_ost2 || error "OST1 start fail"
5914 mount_client $MOUNT || error "mount client fails"
5916 mkdir $DIR/$tdir || error "mkdir fail"
5917 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5918 error "create remote dir succeed"
5920 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5921 rm -rf $DIR/$tdir || error "delete dir fail"
5923 umount_client $MOUNT || error "umount_client failed"
5924 stop_mds || error "MDS stop fail"
5925 stop_ost || error "OST0 stop fail"
5926 stop_ost2 || error "OST1 stop fail"
5929 run_test 71d "start OST0, MDT1, MDT0, OST1"
5932 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5933 combined_mgs_mds && skip "needs separate MGS/MDT"
5937 start_ost || error "OST0 start fail"
5938 for num in $(seq 2 $MDSCOUNT); do
5939 start_mdt $num || return
5941 start_ost2 || error "OST1 start fail"
5942 start_mdt 1 || error "MDT0 start fail"
5944 mount_client $MOUNT || error "mount client fails"
5946 mkdir $DIR/$tdir || error "mkdir fail"
5947 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5948 error "create remote dir succeed"
5950 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5951 rm -rf $DIR/$tdir || error "delete dir fail"
5953 umount_client $MOUNT || error "umount_client failed"
5954 stop_mds || error "MDS stop fail"
5955 stop_ost || error "OST0 stop fail"
5956 stop_ost2 || error "OST1 stop fail"
5959 run_test 71e "start OST0, MDT1, OST1, MDT0"
5961 test_72() { #LU-2634
5962 [ "$mds1_FSTYPE" != ldiskfs ] &&
5963 skip "ldiskfs only test"
5965 local mdsdev=$(mdsdevname 1)
5966 local ostdev=$(ostdevname 1)
5967 local cmd="$E2FSCK -fnvd $mdsdev"
5974 if combined_mgs_mds; then
5975 add_options='--reformat'
5977 add_options='--reformat --replace'
5980 #tune MDT with "-O extents"
5982 for num in $(seq $MDSCOUNT); do
5983 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
5984 $add_options $(mdsdevname $num) $(mdsvdevname $num) ||
5985 error "add mds $num failed"
5986 do_facet mds${num} "$TUNE2FS -O extents $(mdsdevname $num)" ||
5987 error "$TUNE2FS failed on mds${num}"
5990 add ost1 $(mkfs_opts ost1 $ostdev) $add_options $ostdev ||
5991 error "add $ostdev failed"
5992 start_mds || error "start mds failed"
5993 start_ost || error "start ost failed"
5994 mount_client $MOUNT || error "mount client failed"
5996 #create some short symlinks
5997 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5998 createmany -o $DIR/$tdir/$tfile-%d $fn
5999 echo "create $fn short symlinks"
6000 for i in $(seq -w 1 $fn); do
6001 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
6006 umount_client $MOUNT || error "umount client failed"
6007 stop_mds || error "stop mds failed"
6008 stop_ost || error "stop ost failed"
6011 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
6013 run_test 72 "test fast symlink with extents flag enabled"
6015 test_73() { #LU-3006
6016 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
6017 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
6018 error "1st tunefs failed"
6019 start_mgsmds || error "start mds failed"
6020 start_ost || error "start ost failed"
6021 mount_client $MOUNT || error "mount client failed"
6022 $LCTL get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
6023 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
6024 umount_client $MOUNT || error "umount client failed"
6028 run_test 73 "failnode to update from mountdata properly"
6032 (( $MDS1_VERSION >= $(version_code 2.15.57.16) )) ||
6033 skip "need MDS version >= 2.15.57.16 for per-device timeouts"
6036 stack_trap "cleanup"
6038 # Prepare fs2, share the mgs of fs
6039 local FSNAME2=fs15246
6040 local fs2mdsdev=$(mdsdevname 1_2)
6041 local fs2ostdev=$(ostdevname 1_2)
6042 local fs2mdsvdev=$(mdsvdevname 1_2)
6043 local fs2ostvdev=$(ostvdevname 1_2)
6045 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --fsname=$FSNAME2 \
6046 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
6047 add fs2ost $(mkfs_opts ost1 $fs2ostdev) --fsname=$FSNAME2 \
6048 --reformat $fs2ostdev $fs2ostvdev || error "add fs2ost failed"
6050 stack_trap "cleanup_fs2"
6052 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
6053 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS || error "start fs2mds failed"
6055 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
6056 $MOUNT_CMD $MGSNID:/$FSNAME2 $MOUNT2 || error "mount $MOUNT2 failed"
6058 echo "========== All mounted lustre fs ===================="
6059 mount | grep 'type lustre'
6060 echo "====================================================="
6062 # Set and check osc/ldlm_enqueue_min
6063 $LCTL set_param osc.${FSNAME}-*.ldlm_enqueue_min=99 ||
6064 error "failed to set per-device adaptive parameters on client"
6065 stack_trap "$LCTL set_param osc.${FSNAME}-*.ldlm_enqueue_min=0"
6067 local ldlm_enqueue_min
6068 ldlm_enqueue_min=$($LCTL get_param -n osc.${FSNAME}-*.ldlm_enqueue_min \
6070 (( $ldlm_enqueue_min == 99 )) ||
6071 error "wrong ldlm_enqueue_min value for osc.${FSNAME}-*"
6073 # Check fs2 as reference
6074 ldlm_enqueue_min=$($LCTL get_param -n osc.${FSNAME2}-*.ldlm_enqueue_min\
6076 (( $ldlm_enqueue_min == 0 )) ||
6077 error "wrong ldlm_enqueue_min value for osc.${FSNAME2}-*"
6079 # Set and check obdfilter/at_min
6080 do_facet ost1 $LCTL set_param obdfilter.${FSNAME}-*.at_min=1 ||
6081 error "failed to set per-device adaptive parameters on ost"
6082 stack_trap "do_facet ost1 $LCTL set_param obdfilter.${FSNAME}-*.at_min=0"
6085 at_min=$(do_facet ost1 $LCTL get_param -n obdfilter.${FSNAME}-*.at_min \
6087 (( $at_min == 1 )) ||
6088 error "wrong at_min value for obdfilter.${FSNAME}-*"
6090 # set and check mdc/at_max
6091 $LCTL set_param mdc.${FSNAME}-*.at_max=599 ||
6092 error "failed to set per-device adaptive parameters on client"
6093 stack_trap "$LCTL set_param mdc.${FSNAME}-*.at_max=0"
6096 at_max=$($LCTL get_param -n mdc.${FSNAME}-*.at_max | uniq)
6097 (( $at_max == 599 )) ||
6098 error "wrong at_max value for osc.${FSNAME}-*"
6100 # Check fs2 as reference
6101 at_max=$($LCTL get_param -n mdc.${FSNAME2}-*.at_max | uniq)
6102 (( $at_max == 0 )) ||
6103 error "wrong at_max value for osc.${FSNAME2}-*"
6105 # Set and check mds/at_max
6106 do_facet mds1 $LCTL set_param *.${FSNAME}-*.at_max=599 ||
6107 error "failed to set per-device adaptive parameters on mds"
6108 stack_trap "do_facet mds1 $LCTL set_param *.${FSNAME}-*.at_max=0"
6111 at_max=$(do_facet mds1 $LCTL get_param -n mdt.${FSNAME}-*.at_max | uniq)
6112 (( at_max == 599 )) ||
6113 error "wrong at_max value for mdt.${FSNAME}-*"
6115 # Set and check mgs&mgc/at_history
6116 local mgs_nid=$(do_facet $SINGLEMDS $LCTL list_nids | tail -1)
6117 $LCTL set_param mgc.MGC$mgs_nid.at_history=588
6118 stack_trap "$LCTL set_param mgc.MGC$mgs_nid.at_history=0"
6121 at_history=$($LCTL get_param -n mgc.MGC$mgs_nid.at_history)
6122 (( $at_history == 588 )) ||
6123 error "wrong at_history value for mgc.MGC$mgs_nid"
6125 do_facet mgs $LCTL set_param mgs.MGS.at_history=588
6126 stack_trap "do_facet mgs $LCTL set_param mgs.MGS.at_history=0"
6127 at_history=$(do_facet mgs $LCTL get_param -n mgs.MGS.at_history)
6128 (( $at_history == 588 )) ||
6129 error "wrong at_history value for mgs.MGS"
6131 run_test 74 "Test per-device adaptive timeout parameters"
6133 test_75() { # LU-2374
6134 [[ "$MDS1_VERSION" -lt $(version_code 2.4.1) ]] &&
6135 skip "Need MDS version at least 2.4.1"
6138 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
6139 --replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
6140 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
6141 --replace --reformat $(ostdevname 1) $(ostvdevname 1)"
6144 #check with default parameters
6145 add mds1 $opts_mds || error "add mds1 failed for default params"
6146 add ost1 $opts_ost || error "add ost1 failed for default params"
6148 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
6149 opts_mds=$(echo $opts_mds |
6150 sed -e "s/--index=$index/--index=$index --mdt/")
6151 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
6152 opts_ost=$(echo $opts_ost |
6153 sed -e "s/--index=$index/--index=$index --ost/")
6155 add mds1 $opts_mds || error "add mds1 failed for new params"
6156 add ost1 $opts_ost || error "add ost1 failed for new params"
6161 run_test 75 "The order of --index should be irrelevant"
6164 [[ "$MGS_VERSION" -ge $(version_code 2.4.52) ]] ||
6165 skip "Need MDS version at least 2.4.52"
6168 local MDMB_PARAM="osc.*.max_dirty_mb"
6169 echo "Change MGS params"
6170 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
6172 echo "max_dirty_mb: $MAX_DIRTY_MB"
6173 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB - 10))
6174 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
6175 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
6176 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
6177 head -1" $NEW_MAX_DIRTY_MB
6178 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
6179 echo "$MAX_DIRTY_MB"
6180 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
6181 error "error while apply max_dirty_mb"
6183 echo "Check the value is stored after remount"
6186 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
6187 head -1" $NEW_MAX_DIRTY_MB
6188 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
6189 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
6190 error "max_dirty_mb is not saved after remount"
6192 echo "Change OST params"
6193 CLIENT_PARAM="obdfilter.*.client_cache_count"
6194 local CLIENT_CACHE_COUNT
6195 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
6197 echo "client_cache_count: $CLIENT_CACHE_COUNT"
6198 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
6199 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
6200 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
6201 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
6202 head -1" $NEW_CLIENT_CACHE_COUNT
6203 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
6205 echo "$CLIENT_CACHE_COUNT"
6206 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
6207 error "error while apply client_cache_count"
6209 echo "Check the value is stored after remount"
6212 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
6213 head -1" $NEW_CLIENT_CACHE_COUNT
6214 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
6216 echo "$CLIENT_CACHE_COUNT"
6217 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
6218 error "client_cache_count is not saved after remount"
6221 run_test 76a "set permanent params with lctl across mounts"
6223 test_76b() { # LU-4783
6224 [[ "$MGS_VERSION" -ge $(version_code 2.5.57) ]] ||
6225 skip "Need MGS version at least 2.5.57"
6228 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
6229 error "start params log failed"
6232 run_test 76b "verify params log setup correctly"
6235 [[ "$MGS_VERSION" -ge $(version_code 2.8.54) ]] ||
6236 skip "Need MDS version at least 2.4.52"
6238 local MASK_PARAM="mdd.*.changelog_mask"
6239 echo "Change changelog_mask"
6240 do_facet mgs $LCTL set_param -P $MASK_PARAM=-CLOSE ||
6241 error "Can't change changlog_mask"
6242 wait_update $(facet_host mds) "$LCTL get_param -n $MASK_PARAM |
6245 echo "Check the value is stored after mds remount"
6246 stop_mds || error "Failed to stop MDS"
6247 start_mds || error "Failed to start MDS"
6248 local CHANGELOG_MASK=$(do_facet mgs $LCTL get_param -n $MASK_PARAM)
6249 echo $CHANGELOG_MASK | grep CLOSE > /dev/null &&
6250 error "changelog_mask is not changed"
6254 run_test 76c "verify changelog_mask is applied with lctl set_param -P"
6256 test_76d() { #LU-9399
6259 local xattr_cache="llite.*.xattr_cache"
6260 local cmd="$LCTL get_param -n $xattr_cache | head -1"
6261 local new=$((($(eval $cmd) + 1) % 2))
6263 echo "lctl set_param -P llite.*.xattr_cache=$new"
6264 do_facet mgs $LCTL set_param -P $xattr_cache=$new ||
6265 error "Can't change xattr_cache"
6266 wait_update $HOSTNAME "$cmd" "$new"
6268 echo "Check $xattr_cache on client $MOUNT"
6269 umount_client $MOUNT || error "umount $MOUNT failed"
6270 mount_client $MOUNT || error "mount $MOUNT failed"
6271 [ $(eval $cmd) -eq $new ] ||
6272 error "$xattr_cache != $new on client $MOUNT"
6274 echo "Check $xattr_cache on the new client $MOUNT2"
6275 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
6276 [ $(eval $cmd) -eq $new ] ||
6277 error "$xattr_cache != $new on client $MOUNT2"
6278 umount_client $MOUNT2 || error "umount $MOUNT2 failed"
6282 run_test 76d "verify llite.*.xattr_cache can be set by 'lctl set_param -P' correctly"
6284 test_77() { # LU-3445
6285 [[ "$MDS1_VERSION" -ge $(version_code 2.8.55) ]] ||
6286 skip "Need MDS version 2.8.55+ "
6288 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
6289 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
6290 skip_env "mixed loopback and real device not working"
6293 local fs2mdsdev=$(mdsdevname 1_2)
6294 local fs2ostdev=$(ostdevname 1_2)
6295 local fs2mdsvdev=$(mdsvdevname 1_2)
6296 local fs2ostvdev=$(ostvdevname 1_2)
6297 local fsname=test1234
6299 local failnid="$(h2nettype 1.2.3.4),$(h2nettype 4.3.2.1)"
6301 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
6303 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
6304 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
6305 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
6306 error "start fs2mds failed"
6308 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
6309 mgsnid="0.0.0.0@tcp,$mgsnid,$mgsnid:$mgsnid"
6311 add fs2ost --mgsnode=$mgsnid $(mkfs_opts ost1 $fs2ostdev) \
6312 --failnode=$failnid --fsname=$fsname \
6313 --reformat $fs2ostdev $fs2ostvdev ||
6314 error "add fs2ost failed"
6315 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
6317 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
6318 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
6319 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
6322 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
6325 [[ "$mds1_FSTYPE" != ldiskfs ||
6326 "$ost1_FSTYPE" != ldiskfs ]] &&
6327 skip "ldiskfs only test"
6329 # reformat the Lustre filesystem with a smaller size
6330 local saved_MDSCOUNT=$MDSCOUNT
6331 local saved_MDSSIZE=$MDSSIZE
6332 local saved_OSTCOUNT=$OSTCOUNT
6333 local saved_OSTSIZE=$OSTSIZE
6336 MDSSIZE=$((MDSSIZE - 20000))
6337 OSTSIZE=$((OSTSIZE - 20000))
6338 reformat || error "(1) reformat Lustre filesystem failed"
6339 MDSSIZE=$saved_MDSSIZE
6340 OSTSIZE=$saved_OSTSIZE
6342 # mount the Lustre filesystem
6343 setup_noconfig || error "(2) setup Lustre filesystem failed"
6346 log "create test files"
6351 mkdir $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
6353 for i in $(seq $num_files); do
6354 file=$MOUNT/$tdir/$tfile-$i
6355 dd if=/dev/urandom of=$file count=1 bs=1M || {
6356 $LCTL get_param osc.*.cur*grant*
6357 $LFS df; $LFS df -i;
6358 # stop creating files if there is no more space
6359 if [ ! -e $file ]; then
6360 num_files=$((i - 1))
6364 $LFS getstripe -v $file
6365 local ost_idx=$(LFS getstripe -i $file)
6366 do_facet ost$((ost_idx + 1)) \
6367 $LCTL get_param obdfilter.*.*grant*
6368 error "(4) create $file failed"
6372 # unmount the Lustre filesystem
6373 cleanup || error "(5) cleanup Lustre filesystem failed"
6375 # run e2fsck on the MDT and OST devices
6376 local mds_host=$(facet_active_host $SINGLEMDS)
6377 local ost_host=$(facet_active_host ost1)
6378 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
6379 local ost_dev=$(ostdevname 1)
6381 run_e2fsck $mds_host $mds_dev "-y"
6382 run_e2fsck $ost_host $ost_dev "-y"
6384 # get the original block count of the MDT and OST filesystems
6385 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
6386 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
6388 # expand the MDT and OST filesystems to the device size
6389 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
6390 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
6392 # run e2fsck on the MDT and OST devices again
6393 run_e2fsck $mds_host $mds_dev "-y"
6394 run_e2fsck $ost_host $ost_dev "-y"
6396 # mount the Lustre filesystem
6400 log "check files after expanding the MDT and OST filesystems"
6401 for i in $(seq $num_files); do
6402 file=$MOUNT/$tdir/$tfile-$i
6403 $CHECKSTAT -t file -s 1048576 $file ||
6404 error "(6) checkstat $file failed"
6408 log "create more files after expanding the MDT and OST filesystems"
6409 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
6410 file=$MOUNT/$tdir/$tfile-$i
6411 dd if=/dev/urandom of=$file count=1 bs=1M ||
6412 error "(7) create $file failed"
6415 # unmount the Lustre filesystem
6416 cleanup || error "(8) cleanup Lustre filesystem failed"
6418 # run e2fsck on the MDT and OST devices
6419 run_e2fsck $mds_host $mds_dev "-y"
6420 run_e2fsck $ost_host $ost_dev "-y"
6422 # get the maximum block count of the MDT and OST filesystems
6423 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
6424 local ost_max_blks=$(get_block_count ost1 $ost_dev)
6426 # get the minimum block count of the MDT and OST filesystems
6427 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
6428 grep minimum | sed -e 's/^.*filesystem: //g')
6429 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
6430 grep minimum | sed -e 's/^.*filesystem: //g')
6432 # shrink the MDT and OST filesystems to a smaller size
6436 if [[ $mds_max_blks -gt $mds_min_blks &&
6437 $mds_max_blks -gt $mds_orig_blks ]]; then
6438 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
6439 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
6440 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
6441 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
6442 error "shrink $SINGLEMDS to $new_blks failed"
6446 if [[ $ost_max_blks -gt $ost_min_blks &&
6447 $ost_max_blks -gt $ost_orig_blks ]]; then
6448 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
6449 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
6450 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
6451 run_resize2fs ost1 $ost_dev $new_blks ||
6452 error "shrink ost1 to $new_blks failed"
6456 # check whether the MDT or OST filesystem was shrunk or not
6458 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
6459 reformat_and_config ||
6460 error "(10) reformat Lustre filesystem failed"
6464 # run e2fsck on the MDT and OST devices again
6465 run_e2fsck $mds_host $mds_dev "-y"
6466 run_e2fsck $ost_host $ost_dev "-y"
6468 # mount the Lustre filesystem again
6472 log "check files after shrinking the MDT and OST filesystems"
6473 for i in $(seq $((num_files + 10))); do
6474 file=$MOUNT/$tdir/$tfile-$i
6475 $CHECKSTAT -t file -s 1048576 $file ||
6476 error "(11) checkstat $file failed"
6479 # unmount and reformat the Lustre filesystem
6480 cleanup || error "(12) cleanup Lustre filesystem failed"
6481 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
6483 MDSCOUNT=$saved_MDSCOUNT
6484 OSTCOUNT=$saved_OSTCOUNT
6485 reformat_and_config || error "(14) reformat Lustre filesystem failed"
6487 run_test 78 "run resize2fs on MDT and OST filesystems"
6489 test_79() { # LU-4227
6490 [[ "$MDS1_VERSION" -ge $(version_code 2.5.59) ]] ||
6491 skip "Need MDS version at least 2.5.59"
6493 local mdsdev1=$(mdsdevname 1)
6494 local mdsvdev1=$(mdsvdevname 1)
6495 local mdsdev2=$(mdsdevname 2)
6496 local mdsvdev2=$(mdsvdevname 2)
6497 local ostdev1=$(ostdevname 1)
6498 local ostvdev1=$(ostvdevname 1)
6499 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
6500 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
6501 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
6504 # remove --mgs/--mgsnode from mkfs.lustre options
6505 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
6507 mgsnode_opt=$(echo $opts_mds2 |
6508 awk '{ for ( i = 1; i < NF; i++ )
6509 if ( $i ~ "--mgsnode" ) { print $i; break } }')
6510 [ -n "$mgsnode_opt" ] &&
6511 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
6513 mgsnode_opt=$(echo $opts_ost1 |
6514 awk '{ for ( i = 1; i < NF; i++ )
6515 if ( $i ~ "--mgsnode" ) { print $i; break } }')
6516 [ -n "$mgsnode_opt" ] &&
6517 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
6519 # -MGS, format a mdt without --mgs option
6520 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
6521 error "Must specify --mgs when formatting mdt combined with mgs"
6523 # +MGS, format a mdt/ost without --mgsnode option
6524 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
6525 > /dev/null || error "start mds1 failed"
6526 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
6527 error "Must specify --mgsnode when formatting a mdt"
6528 add ost1 $opts_ost1 $ostdev1 $ostvdev1 &&
6529 error "Must specify --mgsnode when formatting an ost"
6533 run_test 79 "format MDT/OST without mgs option (should return errors)"
6536 start_mds || error "Failed to start MDT"
6537 start_ost || error "Failed to start OST1"
6538 uuid=$(do_facet ost1 $LCTL get_param -n mgc.*.uuid)
6539 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
6540 do_facet ost1 "$LCTL set_param fail_val=10 fail_loc=0x906"
6541 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x906"
6542 do_facet mgs "$LCTL set_param -n mgs/MGS/evict_client $uuid"
6544 start_ost2 || error "Failed to start OST2"
6546 do_facet ost1 "$LCTL set_param fail_loc=0"
6551 run_test 80 "mgc import reconnect race"
6553 #Save the original values of $OSTCOUNT and $OSTINDEX$i.
6555 local new_ostcount=$1
6556 saved_ostcount=$OSTCOUNT
6557 OSTCOUNT=$new_ostcount
6561 for ((i = 1; i <= $OSTCOUNT; i++ )); do
6563 eval saved_ostindex$i=${!index}
6568 # Restore the original values of $OSTCOUNT and $OSTINDEX$i.
6569 restore_ostindex() {
6573 echo "restoring OSTCOUNT=$saved_ostcount and OSTINDEXn"
6575 for ((i = 1; i <= $OSTCOUNT; i++ )); do
6576 index=saved_ostindex$i
6577 eval OSTINDEX$i=${!index}
6579 OSTCOUNT=$saved_ostcount
6582 if ! combined_mgs_mds ; then
6587 # The main purpose of this test is to ensure the OST_INDEX_LIST functions as
6588 # expected. This test uses OST_INDEX_LIST to format OSTs with a randomly
6589 # assigned index and ensures we can mount such a formatted file system
6590 test_81() { # LU-4665
6591 (( MDS1_VERSION >= $(version_code 2.6.54) )) ||
6592 skip "Need MDS version at least 2.6.54"
6593 (( OSTCOUNT >= 3 )) || skip_env "needs >= 3 OSTs"
6597 # Each time RANDOM is referenced, a random integer between 0 and 32767
6600 local saved_ostindex1=$OSTINDEX1
6601 for i in 65535 $((RANDOM + 65536)); do
6602 echo -e "\nFormat ost1 with --index=$i, should fail"
6604 if add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat \
6605 $(ostdevname 1) $(ostvdevname 1); then
6606 OSTINDEX1=$saved_ostindex1
6607 error "format ost1 with --index=$i should fail"
6610 OSTINDEX1=$saved_ostindex1
6613 stack_trap restore_ostindex
6615 # Format OSTs with random sparse indices.
6616 local rand_ost=$(((RANDOM * 2 % 65533) + 1))
6617 echo "Format $OSTCOUNT OSTs with OST_INDEX_LIST=[0,$rand_ost,65534]"
6618 OST_INDEX_LIST=[0,$rand_ost,65534] formatall ||
6619 error "formatall failed with $?"
6621 # Setup and check Lustre filesystem.
6622 start_mgsmds || error "start_mgsmds failed"
6623 for ((i = 1; i <= $OSTCOUNT; i++ )); do
6624 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6625 error "start ost$i failed"
6628 mount_client $MOUNT || error "mount client $MOUNT failed"
6629 check_mount || error "check client $MOUNT failed"
6633 local max_easize=$($LCTL get_param -n llite.*.max_easize)
6634 local xattr_size_max=$((4096 - 84)) # 4096 less ldiskfs ea overhead
6636 # XATTR_SIZE_MAX less ldiskfs ea overhead
6637 large_xattr_enabled && xattr_size_max=$((65536 - 84))
6638 (( max_easize >= xattr_size_max )) ||
6639 error "max_easize $max_easize < $xattr_size_max bytes"
6641 test_mkdir $DIR/$tdir
6642 $LFS setstripe -i $rand_ost $DIR/$tdir/$tfile ||
6643 error "error creating $tfile on ost$rand_ost"
6644 $LFS getstripe $DIR/$tdir/$tfile
6651 cmd="getstripe -i $DIR/$tdir/$tfile"
6654 (( $found == $rand_ost )) || error "index $found is not $rand_ost"
6656 cmd="find $DIR/$tdir -i $rand_ost"
6660 [[ "$found" == "$DIR/$tdir/$tfile" ]] ||
6661 error "'lfs find' returned '$found', not '$tfile' by index"
6664 uuid=$(ostuuid_from_index $rand_ost)
6666 cmd="find $DIR/$tdir -O $uuid"
6671 [[ "$found" == "$DIR/$tdir/$tfile" ]] ||
6672 error "'lfs find' returned '$found', not '$tfile' by UUID"
6674 run_test 81 "sparse OST indexing"
6676 # Here we exercise the stripe placement functionality on a file system that
6677 # has formatted the OST with a random index. With the file system the following
6678 # functionality is tested:
6680 # 1. Creating a new file with a specific stripe layout.
6682 # 2. Modifiy a existing empty file with a specific stripe layout.
6684 # 3. Ensure we fail to set the stripe layout of a file that already has one.
6686 # 4. If ost-index is defined we need to ensure it is the first entry in the
6687 # ost index list returned by lfs getstripe.
6689 # 5. Lastly ensure this functionality fails with directories.
6690 test_82a() { # LU-4665
6691 [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
6692 skip "Need MDS version at least 2.6.54"
6693 [[ $OSTCOUNT -ge 3 ]] || skip_env "needs >= 3 OSTs"
6699 # Format OSTs with random sparse indices.
6703 local LOV_V1_INSANE_STRIPE_COUNT=65532
6704 for i in $(seq $OSTCOUNT); do
6705 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
6706 ost_indices+=" $index"
6708 ost_indices=$(comma_list $ost_indices)
6710 stack_trap "restore_ostindex" EXIT
6711 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
6712 OST_INDEX_LIST=[$ost_indices] formatall
6714 # Setup Lustre filesystem.
6715 start_mgsmds || error "start_mgsmds failed"
6716 for i in $(seq $OSTCOUNT); do
6717 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6718 error "start ost$i failed"
6721 # Collect debug information - start of test
6722 do_nodes $(comma_list $(mdts_nodes)) \
6723 $LCTL get_param osp.*.prealloc_*_id
6725 mount_client $MOUNT || error "mount client $MOUNT failed"
6729 check_lfs_df_ret_val $? || error "$LFS df $MOUNT failed"
6730 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6732 stack_trap "do_nodes $(comma_list $(mdts_nodes)) \
6733 $LCTL get_param osp.*.prealloc_*_id || true" EXIT
6735 # 1. If the file does not exist, new file will be created
6736 # with specified OSTs.
6737 local file=$DIR/$tdir/$tfile-1
6738 local cmd="$LFS setstripe -o $ost_indices $file"
6740 eval $cmd || error "$cmd failed"
6741 check_stripe_count $file $OSTCOUNT
6742 check_obdidx $file $ost_indices
6743 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6744 error "write $file failed"
6746 # 2. If the file already exists and is an empty file, the file
6747 # will be attached with specified layout.
6748 file=$DIR/$tdir/$tfile-2
6749 mcreate $file || error "mcreate $file failed"
6750 cmd="$LFS setstripe -o $ost_indices $file"
6752 eval $cmd || error "$cmd failed"
6753 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6754 error "write $file failed"
6755 check_stripe_count $file $OSTCOUNT
6756 check_obdidx $file $ost_indices
6758 # 3. If the file already has a valid layout attached, the command
6759 # should fail with EBUSY.
6761 eval $cmd && error "stripe is already set on $file, $cmd should fail"
6763 # 4. If [--stripe-index|-i <start_ost_idx>] is used, the index must
6764 # be in the OST indices list.
6765 local start_ost_idx=${ost_indices##*,}
6766 file=$DIR/$tdir/$tfile-3
6767 cmd="$LFS setstripe -o $ost_indices -i $start_ost_idx $file"
6769 eval $cmd || error "$cmd failed"
6770 check_stripe_count $file $OSTCOUNT
6771 check_obdidx $file $ost_indices
6772 check_start_ost_idx $file $start_ost_idx
6774 file=$DIR/$tdir/$tfile-4
6775 cmd="$LFS setstripe"
6776 cmd+=" -o $(exclude_items_from_list $ost_indices $start_ost_idx)"
6777 cmd+=" -i $start_ost_idx $file"
6779 eval $cmd && error "index $start_ost_idx should be in $ost_indices"
6781 # 5. Specifying OST indices for directory should succeed.
6782 local dir=$DIR/$tdir/$tdir
6783 mkdir $dir || error "mkdir $dir failed"
6784 cmd="$LFS setstripe -o $ost_indices $dir"
6785 if [ "$MDS1_VERSION" -gt $(version_code 2.11.53) ] &&
6786 [ "$CLIENT_VERSION" -gt $(version_code 2.11.53) ]; then
6788 eval $cmd || error "unable to specify OST indices on directory"
6790 echo "need MDS+client version at least 2.11.53"
6793 run_test 82a "specify OSTs for file (succeed) or directory (succeed)"
6795 # Test 82b is run to ensure that if the user supplies a pool with a specific
6796 # stripe layout that it behaves proprerly. It should fail in the case that
6797 # the supplied OST index list points to OSTs not contained in the user
6799 test_82b() { # LU-4665
6800 [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
6801 skip "Need MDS version at least 2.6.54"
6802 [[ $OSTCOUNT -ge 4 ]] || skip_env "needs >= 4 OSTs"
6808 # Format OSTs with random sparse indices.
6812 local LOV_V1_INSANE_STRIPE_COUNT=65532
6813 for i in $(seq $OSTCOUNT); do
6814 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
6815 ost_indices+=" $index"
6817 ost_indices=$(comma_list $ost_indices)
6819 stack_trap "restore_ostindex" EXIT
6820 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
6821 OST_INDEX_LIST=[$ost_indices] formatall
6823 # Setup Lustre filesystem.
6824 start_mgsmds || error "start_mgsmds failed"
6825 for i in $(seq $OSTCOUNT); do
6826 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6827 error "start ost$i failed"
6830 mount_client $MOUNT || error "mount client $MOUNT failed"
6834 check_lfs_df_ret_val $? || error "$LFS df $MOUNT failed"
6835 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6837 # Create a new pool and add OSTs into it.
6838 local ost_pool=$FSNAME.$TESTNAME
6839 create_pool $ost_pool || error "create OST pool $ost_pool failed"
6841 local ost_idx_in_list=${ost_indices##*,}
6842 local ost_idx_in_pool=$(exclude_items_from_list $ost_indices \
6845 local ost_targets="$FSNAME-OST["
6846 for i in ${ost_idx_in_pool//,/ }; do
6847 ost_targets=$ost_targets$(printf "%04x," $i)
6849 ost_targets="${ost_targets%,}]"
6851 local ost_targets_uuid=$(for i in ${ost_idx_in_pool//,/ }; \
6852 do printf "$FSNAME-OST%04x_UUID\n" $i; done |
6853 sort -u | tr '\n' ' ')
6855 local cmd="$LCTL pool_add $ost_pool $ost_targets"
6856 do_facet mgs $cmd || error "$cmd failed"
6857 wait_update $HOSTNAME "$LCTL get_param -n lov.$FSNAME-*.pools.$TESTNAME|
6858 sort -u | tr '\n' ' ' " "$ost_targets_uuid" ||
6859 error "wait_update $ost_pool failed"
6860 wait_update_facet $SINGLEMDS "$LCTL pool_list $ost_pool | wc -l" 4 ||
6861 error "wait_update pool_list $ost_pool failed"
6863 # If [--pool|-p <pool_name>] is set with [--ost-list|-o <ost_indices>],
6864 # then the OSTs must be the members of the pool.
6865 local file=$DIR/$tdir/$tfile
6866 cmd="$LFS setstripe -p $ost_pool -o $ost_idx_in_list $file"
6868 eval $cmd && error "OST with index $ost_idx_in_list should be" \
6869 "in OST pool $ost_pool"
6871 # Only select OST $ost_idx_in_list from $ost_pool for file.
6872 ost_idx_in_list=${ost_idx_in_pool#*,}
6873 cmd="$LFS setstripe -p $ost_pool -o $ost_idx_in_list $file"
6875 eval $cmd || error "$cmd failed"
6876 cmd="$LFS getstripe $file"
6878 eval $cmd || error "$cmd failed"
6879 check_stripe_count $file 2
6880 check_obdidx $file $ost_idx_in_list
6881 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6882 error "write $file failed"
6884 run_test 82b "specify OSTs for file with --pool and --ost-list options"
6887 [[ "$OST1_VERSION" -ge $(version_code 2.6.91) ]] ||
6888 skip "Need OST version at least 2.6.91"
6889 if [ "$ost1_FSTYPE" != ldiskfs ]; then
6890 skip "ldiskfs only test"
6899 ostmnt=$(facet_mntpt ost1)
6901 # Mount the OST as an ldiskfs filesystem.
6902 log "mount the OST $dev as a $ost1_FSTYPE filesystem"
6903 add ost1 $(mkfs_opts ost1 $dev) $FSTYPE_OPT \
6904 --reformat $dev > /dev/null ||
6905 error "format ost1 error"
6907 if ! test -b $dev; then
6908 mnt_opts=$(csa_add "$OST_MOUNT_FS_OPTS" -o loop)
6910 echo "mnt_opts $mnt_opts"
6911 do_facet ost1 mount -t "$ost1_FSTYPE" $dev \
6913 # Run llverfs on the mounted ldiskfs filesystem.
6914 # It is needed to get ENOSPACE.
6915 log "run llverfs in partial mode on the OST $ost1_FSTYPE $ostmnt"
6916 do_rpc_nodes $(facet_host ost1) run_llverfs $ostmnt -vpl \
6917 "no" || error "run_llverfs error on $ost1_FSTYPE"
6920 log "unmount the OST $dev"
6923 # Delete file IO_scrub. Later osd_scrub_setup will try to
6924 # create "IO_scrub" but will get ENOSPACE.
6926 echo "start ost1 service on `facet_active_host ost1`"
6927 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
6930 err=$(do_facet ost1 dmesg | grep "VFS: Busy inodes after unmount of")
6931 echo "string err $err"
6932 [ -z "$err" ] || error $err
6935 run_test 83 "ENOSPACE on OST doesn't cause message VFS: \
6936 Busy inodes after unmount ..."
6939 local facet=$SINGLEMDS
6940 local num=$(echo $facet | tr -d "mds")
6941 local dev=$(mdsdevname $num)
6942 local time_min=$(recovery_time_min)
6943 local recovery_duration
6944 local completed_clients
6945 local correct_clients
6948 echo "start mds service on $(facet_active_host $facet)"
6950 "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
6951 error "start MDS failed"
6953 start_ost || error "start OST0000 failed"
6954 wait_osc_import_state mds ost1 FULL
6955 start_ost2 || error "start OST0001 failed"
6956 wait_osc_import_state mds ost2 FULL
6958 echo "recovery_time=$time_min, timeout=$TIMEOUT, wrap_up=$wrap_up"
6960 mount_client $MOUNT1 || error "mount $MOUNT1 failed"
6961 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
6962 # make sure new superblock labels are sync'd before disabling writes
6966 replay_barrier $SINGLEMDS
6967 createmany -o $DIR1/$tfile-%d 1000
6969 # We need to catch the end of recovery window to extend it.
6970 # Skip 5 requests and add delay to request handling.
6971 #define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP
6972 do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5"
6974 facet_failover --fsck $SINGLEMDS || error "failover: $?"
6977 echo "recovery status"
6978 do_facet $SINGLEMDS \
6979 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status"
6981 recovery_duration=$(do_facet $SINGLEMDS \
6982 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
6983 awk '/recovery_duration/ { print $2 }')
6984 (( $recovery_duration > $time_min + $wrap_up )) &&
6985 error "recovery_duration > recovery_time_hard + wrap up"
6986 completed_clients=$(do_facet $SINGLEMDS \
6987 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
6988 awk '/completed_clients/ { print $2 }')
6990 correct_clients="$MDSCOUNT/$((MDSCOUNT+1))"
6991 [ "$completed_clients" = "${correct_clients}" ] ||
6992 error "$completed_clients != $correct_clients"
6994 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
6995 umount_client $MOUNT1
6996 umount_client $MOUNT2
7002 run_test 84 "check recovery_hard_time"
7005 [[ "$OST1_VERSION" -ge $(version_code 2.7.55) ]] ||
7006 skip "Need OST version at least 2.7.55"
7007 ##define OBD_FAIL_OSD_OST_EA_FID_SET 0x197
7008 do_facet ost1 "lctl set_param fail_loc=0x197"
7012 run_test 85 "osd_ost init: fail ea_fid_set"
7017 # ost1 has already registered to the MGS before the reformat.
7018 # So after reformatting it with option "-G", it could not be
7019 # mounted to the MGS. Cleanup the system for subsequent tests.
7024 [ "$ost1_FSTYPE" = zfs ] &&
7025 skip "LU-6442: no such mkfs params for ZFS OSTs"
7026 [[ "$MDS1_VERSION" -ge $(version_code 2.7.56) ]] ||
7027 skip "Need server version newer than 2.7.55"
7030 local OLDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
7031 awk '/Flex block group size: / { print $NF; exit; }')
7033 [ "$OLDSIZE" == "$NEWSIZE" ] && skip "$NEWSIZE groups already"
7035 local opts=" -O flex_bg -G $NEWSIZE"
7036 opts=$(OST_FS_MKFS_OPTS+="$opts" mkfs_opts ost1 $(ostdevname 1))
7037 opts+=" --reformat $(ostdevname 1) $(ostvdevname 1)"
7038 echo "params: $opts"
7040 trap cleanup_86 EXIT ERR
7043 add ost1 $opts || error "add ost1 failed with new params"
7045 local FOUNDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
7046 awk '/Flex block group size: / { print $NF; exit; }')
7048 [[ $FOUNDSIZE == $NEWSIZE ]] ||
7049 error "Flex block group size: $FOUNDSIZE, expected: $NEWSIZE"
7053 run_test 86 "Replacing mkfs.lustre -G option"
7055 test_87() { #LU-6544
7056 [[ "$MDS1_VERSION" -ge $(version_code 2.9.51) ]] ||
7057 skip "Need MDS version at least 2.9.51"
7058 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
7059 skip "ldiskfs only test"
7060 [[ $OSTCOUNT -gt 59 ]] &&
7061 skip "Ignore wide striping situation"
7062 [ -n "$FILESET" ] && skip "Not functional for FILESET set"
7064 local mdsdev=$(mdsdevname 1)
7065 local mdsvdev=$(mdsvdevname 1)
7066 local file=$DIR/$tfile
7067 local mntpt=$(facet_mntpt $SINGLEMDS)
7068 local used_xattr_blk=0
7069 local inode_size=${1:-1024}
7071 local xtest="trusted.test"
7075 local stripe_cnt=$(($OSTCOUNT + 2))
7077 #Please see ldiskfs_make_lustre() for MDT inode size calculation
7078 if [ $stripe_cnt -gt 16 ]; then
7081 left_size=$(expr $inode_size - \
7084 32 - 40 \* 3 - 32 \* 3 - $stripe_cnt \* 24 - 16 - 3 - \
7086 24 - 18 - $(expr length $tfile) - 16 - 4)
7087 if [ $left_size -le 0 ]; then
7088 echo "No space($left_size) is expected in inode."
7089 echo "Try 1-byte xattr instead to verify this."
7092 echo "Estimate: at most $left_size-byte space left in inode."
7098 if ! combined_mgs_mds ; then
7102 add mds1 $(mkfs_opts mds1 ${mdsdev}) --stripe-count-hint=$stripe_cnt \
7103 --reformat $mdsdev $mdsvdev || error "add mds1 failed"
7104 start_mdt 1 > /dev/null || error "start mdt1 failed"
7105 for i in $(seq $OSTCOUNT); do
7106 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS > /dev/null ||
7107 error "start ost$i failed"
7109 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
7110 check_mount || error "check client $MOUNT failed"
7113 $LFS setstripe -E 1M -S 1M -c 1 -E 64M -c 1 -E -1 -c -1 $file ||
7114 error "Create file with 3 components failed"
7115 $TRUNCATE $file $((1024*1024*64+1)) || error "truncate file failed"
7116 i=$($LFS getstripe -I3 -c $file) || error "get 3rd stripe count failed"
7117 if [ $i -ne $OSTCOUNT ]; then
7118 left_size=$(expr $left_size + $(expr $OSTCOUNT - $i) \* 24)
7119 echo -n "Since only $i out $OSTCOUNT OSTs are used, "
7120 echo -n "the expected left space is changed to "
7121 echo "$left_size bytes at most."
7123 value=$(generate_string $left_size)
7124 setfattr -n $xtest -v $value $file
7125 orig=$(get_xattr_value $xtest $file)
7126 [[ "$orig" != "$value" ]] && error "$xtest changed"
7128 #Verify if inode has some expected space left
7129 umount $MOUNT > /dev/null || error "umount $MOUNT failed"
7130 stop_mdt 1 > /dev/null || error "stop mdt1 failed"
7131 mount_ldiskfs $SINGLEMDS || error "mount -t ldiskfs $SINGLEMDS failed"
7133 do_facet $SINGLEMDS ls -sal $mntpt/ROOT/$tfile
7134 used_xattr_blk=$(do_facet $SINGLEMDS ls -s $mntpt/ROOT/$tfile |
7136 [[ $used_xattr_blk -eq 0 ]] &&
7137 error "Please check MDS inode size calculation: \
7138 more than $left_size-byte space left in inode."
7139 echo "Verified: at most $left_size-byte space left in inode."
7141 unmount_ldiskfs $SINGLEMDS
7143 for i in $(seq $OSTCOUNT); do
7144 stop ost$i -f || error "stop ost$i failed"
7147 run_test 87 "check if MDT inode can hold EAs with N stripes properly"
7150 [ "$mds1_FSTYPE" == zfs ] &&
7151 skip "LU-6662: no implementation for ZFS"
7155 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
7156 --reformat $(mdsdevname 1) || error "add mds1 failed"
7158 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
7159 grep -e \".*opts:.*errors=remount-ro.*\"" ||
7160 error "default mount options is missing"
7162 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
7163 --mountfsoptions="user_xattr,errors=panic" \
7164 --reformat $(mdsdevname 1) || error "add mds1 failed"
7166 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
7167 grep -e \".*opts:.*errors=panic.*\"" ||
7168 error "user can't override default mount options"
7170 run_test 88 "check the default mount options can be overridden"
7172 test_89() { # LU-7131
7173 [[ "$MDS1_VERSION" -ge $(version_code 2.9.54) ]] ||
7174 skip "Need MDT version at least 2.9.54"
7176 local key=failover.node
7177 local val1=192.0.2.254@tcp0 # Reserved IPs, see RFC 5735
7178 local val2=192.0.2.255@tcp0
7179 local mdsdev=$(mdsdevname 1)
7184 if [[ "$mds1_FSTYPE" == zfs ]]; then
7185 import_zpool mds1 || return ${PIPESTATUS[0]}
7188 # Check that parameters are added correctly
7189 echo "tunefs --param $key=$val1"
7190 do_facet mds1 "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
7191 error "tunefs --param $key=$val1 failed"
7192 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7193 error "tunefs --dryrun failed"
7194 params=${params##*Parameters:}
7195 params=${params%%exiting*}
7196 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val1) = "1" ] ||
7197 error "on-disk parameter not added correctly via tunefs"
7199 # Check that parameters replace existing instances when added
7200 echo "tunefs --param $key=$val2"
7201 do_facet mds1 "$TUNEFS --param $key=$val2 $mdsdev >/dev/null" ||
7202 error "tunefs --param $key=$val2 failed"
7203 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7204 error "tunefs --dryrun failed"
7205 params=${params##*Parameters:}
7206 params=${params%%exiting*}
7207 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "1" ] ||
7208 error "on-disk parameter not replaced via tunefs"
7209 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val2) = "1" ] ||
7210 error "on-disk parameter not replaced correctly via tunefs"
7212 # Check that a parameter is erased properly
7213 echo "tunefs --erase-param $key"
7214 do_facet mds1 "$TUNEFS --erase-param $key $mdsdev >/dev/null" ||
7215 error "tunefs --erase-param $key failed"
7216 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7217 error "tunefs --dryrun failed"
7218 params=${params##*Parameters:}
7219 params=${params%%exiting*}
7220 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "0" ] ||
7221 error "on-disk parameter not erased correctly via tunefs"
7223 # Check that all the parameters are erased
7224 do_facet mds1 "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
7225 error "tunefs --param $key=$val1 failed"
7226 echo "tunefs --erase-params"
7227 do_facet mds1 "$TUNEFS --erase-params $mdsdev >/dev/null" ||
7228 error "tunefs --erase-params failed"
7229 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7230 error "tunefs --dryrun failed"
7231 params=${params##*Parameters:}
7232 params=${params%%exiting*}
7233 params=$(echo $params | tr ' ' '\n')
7235 error "all on-disk parameters not erased correctly via tunefs $params"
7237 # Check the order of options --erase-params and --param
7238 echo "tunefs --param $key=$val1 --erase-params"
7240 "$TUNEFS --param $key=$val1 --erase-params $mdsdev >/dev/null"||
7241 error "tunefs --param $key=$val1 --erase-params failed"
7242 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7243 error "tunefs --dryrun failed"
7244 params=${params##*Parameters:}
7245 params=${params%%exiting*}
7246 [ $(echo $params | tr ' ' '\n') == "$key=$val1" ] ||
7247 error "on-disk param not added correctly with --erase-params"
7251 run_test 89 "check tunefs --param and --erase-param{s} options"
7254 # $2 (optional) value of max_mod_rpcs_in_flight to set
7255 check_max_mod_rpcs_in_flight() {
7263 idx=$(printf "%04x" $($LFS getdirstripe -i $dir))
7264 facet="mds$((0x$idx + 1))"
7266 if [ -z "$mmr" ]; then
7267 # get value of max_mod_rcps_in_flight
7268 mmr=$($LCTL get_param -n \
7269 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) ||
7270 error "Unable to get max_mod_rpcs_in_flight"
7271 echo "max_mod_rcps_in_flight is $mmr"
7273 # set value of max_mod_rpcs_in_flight
7275 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight=$mmr ||
7276 error "Unable to set max_mod_rpcs_in_flight to $mmr"
7277 echo "max_mod_rpcs_in_flight set to $mmr"
7280 # create mmr+1 files
7281 echo "creating $((mmr + 1)) files ..."
7283 for i in $(seq $((mmr + 1))); do
7289 # consumes mmr-1 modify RPC slots
7290 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7291 # drop requests on MDT so that RPC slots are consumed
7292 # during all the request resend interval
7293 do_facet $facet "$LCTL set_param fail_loc=0x159"
7294 echo "launch $((mmr - 1)) chmod in parallel ..."
7295 for i in $(seq $((mmr - 1))); do
7296 chmod 0600 $dir/file-$i &
7300 # send one additional modify RPC
7301 do_facet $facet "$LCTL set_param fail_loc=0"
7302 echo "launch 1 additional chmod in parallel ..."
7303 chmod 0600 $dir/file-$mmr &
7306 # check this additional modify RPC get a modify RPC slot
7307 # and succeed its operation
7308 checkstat -vp 0600 $dir/file-$mmr ||
7309 error "Unable to send $mmr modify RPCs in parallel"
7314 # consumes mmr modify RPC slots
7315 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7316 # drop requests on MDT so that RPC slots are consumed
7317 # during all the request resend interval
7318 do_facet $facet "$LCTL set_param fail_loc=0x159"
7319 echo "launch $mmr chmod in parallel ..."
7320 for i in $(seq $mmr); do
7321 chmod 0666 $dir/file-$i &
7325 # send one additional modify RPC
7326 do_facet $facet "$LCTL set_param fail_loc=0"
7327 echo "launch 1 additional chmod in parallel ..."
7328 chmod 0666 $dir/file-$((mmr + 1)) &
7331 # check this additional modify RPC blocked getting a modify RPC slot
7332 checkstat -vp 0644 $dir/file-$((mmr + 1)) ||
7333 error "Unexpectedly send $(($mmr + 1)) modify RPCs in parallel"
7337 get_mdt_max_mod_rpcs_in_flight_val() {
7341 # It will be enough to get one value from one MDT
7342 max_mrif=$(do_facet $facet \
7343 "$LCTL get_param -n \
7344 mdt.$FSNAME-MDT*.max_mod_rpcs_in_flight | head -n 1")
7345 if [[ $max_mrif ]]; then
7348 max_mrif=$(do_facet $facet \
7349 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
7354 set_mdt_max_mod_rpcs_in_flight() {
7359 lctl_op=$($LCTL get_param \
7360 mdt.*.max_mod_rpcs_in_flight)
7361 if [[ $lctl_op ]]; then
7364 mdt.$FSNAME-MDT*.max_mod_rpcs_in_flight=$max_mrif"
7368 /sys/module/mdt/parameters/max_mod_rpcs_per_client"
7369 echo "the deprecated max_mod_rpcs_per_client \
7370 parameter was involved"
7377 [[ $($LCTL get_param mdc.*.import |
7378 grep "connect_flags:.*multi_mod_rpc") ]] ||
7379 skip "Need MDC with 'multi_mod_rpcs' feature"
7381 # check default value
7382 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
7383 check_max_mod_rpcs_in_flight $DIR/$tdir
7387 run_test 90a "check max_mod_rpcs_in_flight is enforced"
7397 [[ $($LCTL get_param mdc.*.import |
7398 grep "connect_flags:.*multi_mod_rpc") ]] ||
7399 skip "Need MDC with 'multi_mod_rpcs' feature"
7402 # update max_mod_rpcs_in_flight
7403 $LFS mkdir -c1 $DIR/${tdir}1 || error "mkdir $DIR/${tdir}1 failed"
7404 check_max_mod_rpcs_in_flight $DIR/${tdir}1 1
7407 # check client is able to send multiple modify RPCs in paralell
7408 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
7409 grep -c "multi_mod_rpcs")
7410 if [ "$tmp" -ne $MDSCOUNT ]; then
7412 skip "Client not able to send multiple modify RPCs in parallel"
7415 # update max_mod_rpcs_in_flight
7416 $LFS mkdir -c1 $DIR/${tdir}2 || error "mkdir $DIR/${tdir}2 failed"
7417 check_max_mod_rpcs_in_flight $DIR/${tdir}2 5
7420 $LFS mkdir -c1 $DIR/${tdir}3 || error "mkdir $DIR/${tdir}3 failed"
7421 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/${tdir}3))
7422 facet="mds$((0x$idx + 1))"
7424 mmrpc=$(get_mdt_max_mod_rpcs_in_flight_val $facet)
7425 echo "mdt_max_mod_rpcs_in_flight is $mmrpc"
7426 # update max_mod_rpcs_in_flight
7427 umount_client $MOUNT
7428 set_mdt_max_mod_rpcs_in_flight 16 $facet
7430 $LCTL set_param mdc.$FSNAME-MDT$idx-mdc-*.max_rpcs_in_flight=17
7431 check_max_mod_rpcs_in_flight $DIR/${tdir}3 16
7433 # restore MDT max_mod_rpcs_in_flight initial value
7434 set_mdt_max_mod_rpcs_in_flight $mmrpc $facet
7436 rm -rf $DIR/${tdir}?
7439 run_test 90b "check max_mod_rpcs_in_flight is enforced after update"
7442 # get max_rpcs_in_flight value
7443 mrif_90c=$($LCTL get_param -n \
7444 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
7445 echo "max_rpcs_in_flight is $mrif_90c"
7447 # get MDC max_mod_rpcs_in_flight value
7448 mmrif_90c=$($LCTL get_param -n \
7449 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight)
7450 echo "MDC max_mod_rpcs_in_flight is $mmrif_90c"
7452 # get MDT max_mod_rpcs_in_flight value
7453 mmrpc_90c=$(get_mdt_max_mod_rpcs_in_flight_val "mds1")
7454 echo "mdt_max_mod_rpcs_in_flight is $mmrpc_90c"
7457 restore_params_90c() {
7460 # restore max_rpcs_in_flight value
7461 do_facet mgs $LCTL set_param -P \
7462 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$mrif_90c
7464 # restore max_mod_rpcs_in_flight value
7465 do_facet mgs $LCTL set_param -P \
7466 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mmrif_90c
7468 # restore MDT max_mod_rpcs_in_flight value
7469 set_mdt_max_mod_rpcs_in_flight $mmrpc_90c "mds1"
7477 [[ $($LCTL get_param mdc.*.import |
7478 grep "connect_flags:.*multi_mod_rpc") ]] ||
7479 skip "Need MDC with 'multi_mod_rpcs' feature"
7481 # check client is able to send multiple modify RPCs in paralell
7482 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
7483 grep -c "multi_mod_rpcs")
7484 if [ "$tmp" -ne $MDSCOUNT ]; then
7486 skip "Client not able to send multiple modify RPCs in parallel"
7490 stack_trap restore_params_90c
7493 # attempt to set max_mod_rpcs_in_flight to max_rpcs_in_flight value
7494 # prerequisite: set MDT max_mod_rpcs_in_flight to
7495 # max_rpcs_in_flight value
7496 set_mdt_max_mod_rpcs_in_flight $mrif_90c "mds1"
7498 # if max_mod_rpcs_in_flight is set to be equal to or larger than
7499 # max_rpcs_in_flight, then max_rpcs_in_flight will be increased
7500 if [[ "$CLIENT_VERSION" -ge $(version_code 2.13.53) ]]; then
7502 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif_90c ||
7503 error "set max_mod_rpcs_in_flight to $mrif_90c failed"
7505 local new_mrif=$($LCTL get_param -n \
7506 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
7507 ((new_mrif == mrif_90c + 1)) ||
7508 error "max_rpcs_in_flight was not increased"
7511 umount_client $MOUNT
7512 set_mdt_max_mod_rpcs_in_flight $mmrpc_90c "mds1"
7516 # attempt to set max_mod_rpcs_in_flight to MDT max_mod_rpcs_in_flight+1
7517 # prerequisite: set max_rpcs_in_flight to MDT max_mod_rpcs_in_flight+2
7519 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$((mmrpc_90c + 2))
7522 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$((mmrpc_90c + 1)) &&
7523 error "set max_mod_rpcs_in_flight to $((mmrpc_90c + 1)) should fail"
7526 # attempt to set max_mod_rpcs_in_flight permanently
7527 do_facet mgs $LCTL set_param -P \
7528 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$mrif_90c
7530 do_facet mgs $LCTL set_param -P \
7531 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif_90c
7533 remount_client $MOUNT
7535 wait_update_facet --verbose client "$LCTL get_param -n \
7536 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight" \
7537 "$((mrif_90c + 1))" ||
7538 error "expected '$((mrif_90c + 1))' for max_rpcs_in_flight"
7540 wait_update_facet --verbose client "$LCTL get_param -n \
7541 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight" \
7543 error "expected '$mrif_90c' for max_mod_rpcs_in_flight"
7548 run_test 90c "check max_mod_rpcs_in_flight update limits"
7559 [[ $($LCTL get_param mdc.*.import |
7560 grep "connect_flags:.*multi_mod_rpc") ]] ||
7561 skip "Need MDC with 'multi_mod_rpcs' feature"
7563 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
7564 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
7565 facet="mds$((0x$idx + 1))"
7567 # check client version supports multislots
7568 tmp=$($LCTL get_param -N \
7569 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
7570 if [ -z "$tmp" ]; then
7572 skip "Client does not support multiple modify RPCs in flight"
7575 # get current value of max_mod_rcps_in_flight
7576 mmr=$($LCTL get_param -n \
7577 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
7578 echo "max_mod_rcps_in_flight is $mmr"
7581 echo "creating $mmr files ..."
7583 for i in $(seq $mmr); do
7584 touch $DIR/$tdir/file-$i
7587 # prepare for close RPC
7588 multiop_bg_pause $DIR/$tdir/file-close O_c
7591 # consumes mmr modify RPC slots
7592 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7593 # drop requests on MDT so that RPC slots are consumed
7594 # during all the request resend interval
7595 do_facet $facet "$LCTL set_param fail_loc=0x159"
7596 echo "launch $mmr chmod in parallel ..."
7597 for i in $(seq $mmr); do
7598 chmod 0600 $DIR/$tdir/file-$i &
7601 # send one additional close RPC
7602 do_facet $facet "$LCTL set_param fail_loc=0"
7603 echo "launch 1 additional close in parallel ..."
7605 cancel_lru_locks mdc
7608 # check this additional close RPC get a modify RPC slot
7609 # and multiop process completed
7610 [ -d /proc/$pid ] &&
7611 error "Unable to send the additional close RPC in parallel"
7616 run_test 90d "check one close RPC is allowed above max_mod_rpcs_in_flight"
7618 check_uuid_on_ost() {
7620 do_facet ost1 "$LCTL get_param obdfilter.${FSNAME}*.exports.'$nid'.uuid"
7623 check_uuid_on_mdt() {
7625 do_facet $SINGLEMDS "$LCTL get_param mdt.${mds1_svc}*.exports.'$nid'.uuid"
7633 [[ "$OST1_VERSION" -ge $(version_code 2.7.63) ]] ||
7634 skip "Need OST version at least 2.7.63"
7635 [[ "$MDS1_VERSION" -ge $(version_code 2.7.63) ]] ||
7636 skip "Need MDT version at least 2.7.63"
7638 start_mds || error "MDS start failed"
7639 start_ost || error "unable to start OST"
7640 mount_client $MOUNT || error "client start failed"
7641 check_mount || error "check_mount failed"
7644 nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g")
7648 uuid=$(get_client_uuid $MOUNT)
7650 echo "list nids on mdt:"
7651 do_facet $SINGLEMDS "$LCTL list_param mdt.${FSNAME}*.exports.*"
7652 echo "uuid from $nid:"
7653 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
7655 found=$(check_uuid_on_mdt $nid | grep $uuid)
7656 [ -z "$found" ] && error "can't find $uuid $nid on MDT"
7657 found=$(check_uuid_on_ost $nid | grep $uuid)
7658 [ -z "$found" ] && error "can't find $uuid $nid on OST"
7660 # umount the client so it won't reconnect
7661 manual_umount_client --force || error "failed to umount $?"
7662 # shouldn't disappear on MDS after forced umount
7663 found=$(check_uuid_on_mdt $nid | grep $uuid)
7664 [ -z "$found" ] && error "can't find $uuid $nid"
7667 do_facet $SINGLEMDS \
7668 "$LCTL set_param -n mdt.${mds1_svc}.evict_client nid:$nid"
7669 sleep 1 # eviction above is async, give it some time to proceed
7671 found=$(check_uuid_on_mdt $nid | grep $uuid)
7672 [ -n "$found" ] && error "found $uuid $nid on MDT"
7673 found=$(check_uuid_on_ost $nid | grep $uuid)
7674 [ -n "$found" ] && error "found $uuid $nid on OST"
7676 # check it didn't reconnect (being umounted)
7677 sleep $((TIMEOUT+1))
7678 found=$(check_uuid_on_mdt $nid | grep $uuid)
7679 [ -n "$found" ] && error "found $uuid $nid on MDT"
7680 found=$(check_uuid_on_ost $nid | grep $uuid)
7681 [ -n "$found" ] && error "found $uuid $nid on OST"
7685 run_test 91 "evict-by-nid support"
7687 generate_ldev_conf() {
7688 # generate an ldev.conf file
7689 local ldevconfpath=$1
7691 local fsldevformat=""
7694 local facets="mgs,$(get_facets OST),$(get_facets MDS)"
7695 for facet in ${facets//,/ }; do
7697 fstype=$(facet_fstype $facet)
7698 [ "$fstype" = zfs ] && fsldevformat="$fstype:"
7699 local host=$(facet_host $facet)
7701 local varfo=${facet}failover_HOST
7702 if [ $facet == mgs ] && combined_mgs_mds; then
7703 varfo=mds1failover_HOST
7705 [ -n "${!varfo}" ] && fo=${!varfo}
7706 local type=$(echo $facet | tr -d "[:digit:]" | \
7707 tr "[:lower:]" "[:upper:]" | sed s/MDS/MDT/ )
7709 [ ${facet} == mgs ] ||
7710 num=$(facet_number $facet)
7711 printf "%s\t%s\t%s-%s%04x\t%s%s\n" \
7718 $(facet_device $facet) >> $ldevconfpath
7721 echo "----- $ldevconfpath -----"
7723 echo "--- END $ldevconfpath ---"
7728 # generate a nids file (mapping between hostname to nid)
7729 # looks like we only have the MGS nid available to us
7730 # so just echo that to a file
7732 echo -e "${mgs_HOST}\t${MGSNID}" > $nidspath
7734 echo "----- $nidspath -----"
7736 echo "--- END $nidspath ---"
7739 compare_ldev_output() {
7740 local ldev_output=$1
7741 local expected_output=$2
7743 sort $expected_output -o $expected_output
7744 sort $ldev_output -o $ldev_output
7746 echo "-- START OF LDEV OUTPUT --"
7748 echo "--- END OF LDEV OUTPUT ---"
7750 echo "-- START OF EXPECTED OUTPUT --"
7751 cat $expected_output
7752 echo "--- END OF EXPECTED OUTPUT ---"
7754 diff $expected_output $ldev_output
7759 if [ -z "$LDEV" ]; then
7760 error "ldev is missing!"
7763 local LDEVCONFPATH=$TMP/ldev.conf
7764 local NIDSPATH=$TMP/nids
7766 echo "Host is $(hostname)"
7768 generate_ldev_conf $LDEVCONFPATH
7769 generate_nids $NIDSPATH
7771 # echo the mgs nid and compare it to environment variable MGSNID
7772 local facets="$(get_facets OST),$(get_facets MDS),mgs"
7773 for facet in ${facets//,/ }; do
7774 local host=$(facet_host $facet)
7775 local output=$($LDEV -c $LDEVCONFPATH -H $host -n $NIDSPATH echo %m)
7777 echo "-- START OF LDEV OUTPUT --"
7779 echo "--- END OF LDEV OUTPUT ---"
7782 error "ldev failed to execute!"
7784 # need to process multiple lines because of
7785 # several targets on host
7786 echo -e $output | awk '{ print $2 }' | while read -r line ; do
7787 [ "$line" = "$MGSNID" ] ||
7788 error "ldev failed mgs nid '$line', \
7792 rm -f $LDEVCONFPATH $NIDSPATH
7794 run_test 92 "ldev returns MGS NID correctly in command substitution"
7797 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs"
7800 #start mgs or mgs/mdt0
7801 if ! combined_mgs_mds ; then
7808 start_ost || error "OST0 start fail"
7810 #define OBD_FAIL_MGS_WRITE_TARGET_DELAY 0x90e
7811 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x8000090e"
7812 for num in $(seq 2 $MDSCOUNT); do
7816 mount_client $MOUNT || error "mount client fails"
7817 wait_osc_import_state mds ost FULL
7818 wait_osc_import_ready client ost
7819 check_mount || error "check_mount failed"
7821 cleanup || error "cleanup failed with $?"
7823 run_test 93 "register mulitple MDT at the same time"
7826 if [ -z "$LDEV" ]; then
7827 error "ldev is missing!"
7830 local LDEVCONFPATH=$TMP/ldev.conf
7831 local NIDSPATH=$TMP/nids
7833 generate_ldev_conf $LDEVCONFPATH
7834 generate_nids $NIDSPATH
7836 local LDEV_OUTPUT=$TMP/ldev-output.txt
7837 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME > $LDEV_OUTPUT
7839 # ldev failed, error
7840 if [ $? -ne 0 ]; then
7841 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7842 error "ldev failed to execute!"
7846 local EXPECTED_OUTPUT=$TMP/ldev-expected.txt
7848 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
7850 for num in $(seq $MDSCOUNT); do
7851 printf "%s-MDT%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7854 for num in $(seq $OSTCOUNT); do
7855 printf "%s-OST%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7858 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7860 if [ $? -ne 0 ]; then
7861 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7862 error "ldev failed to produce the correct hostlist!"
7865 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7867 run_test 94 "ldev outputs correct labels for file system name query"
7870 if [ -z "$LDEV" ]; then
7871 error "ldev is missing!"
7874 local LDEVCONFPATH=$TMP/ldev.conf
7875 local NIDSPATH=$TMP/nids
7877 generate_ldev_conf $LDEVCONFPATH
7878 generate_nids $NIDSPATH
7882 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME &>/dev/null
7883 if [ $? -ne 0 ]; then
7884 rm $LDEVCONFPATH $NIDSPATH
7885 error "ldev label filtering w/ -F failed!"
7889 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -l &>/dev/null
7890 if [ $? -ne 0 ]; then
7891 rm $LDEVCONFPATH $NIDSPATH
7892 error "ldev label filtering w/ -l failed!"
7896 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f &>/dev/null
7897 if [ $? -ne 0 ]; then
7898 rm $LDEVCONFPATH $NIDSPATH
7899 error "ldev label filtering w/ -f failed!"
7903 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a &>/dev/null
7904 if [ $? -ne 0 ]; then
7905 rm $LDEVCONFPATH $NIDSPATH
7906 error "ldev label filtering w/ -a failed!"
7911 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -F $FSNAME &>/dev/null
7912 if [ $? -eq 0 ]; then
7913 rm $LDEVCONFPATH $NIDSPATH
7914 error "ldev label filtering w/ -a and -F incorrectly succeeded"
7918 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -f &>/dev/null
7919 if [ $? -eq 0 ]; then
7920 rm $LDEVCONFPATH $NIDSPATH
7921 error "ldev label filtering w/ -a and -f incorrectly succeeded"
7925 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -l &>/dev/null
7926 if [ $? -eq 0 ]; then
7927 rm $LDEVCONFPATH $NIDSPATH
7928 error "ldev label filtering w/ -a and -l incorrectly succeeded"
7932 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f -l &>/dev/null
7933 if [ $? -eq 0 ]; then
7934 rm $LDEVCONFPATH $NIDSPATH
7935 error "ldev label filtering w/ -f and -l incorrectly succeeded"
7939 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -l &>/dev/null
7940 if [ $? -eq 0 ]; then
7941 rm $LDEVCONFPATH $NIDSPATH
7942 error "ldev label filtering w/ -F and -l incorrectly succeeded"
7945 # file sys & foreign
7946 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -f &>/dev/null
7947 if [ $? -eq 0 ]; then
7948 rm $LDEVCONFPATH $NIDSPATH
7949 error "ldev label filtering w/ -F and -f incorrectly succeeded"
7952 rm $LDEVCONFPATH $NIDSPATH
7954 run_test 95 "ldev should only allow one label filter"
7957 if [ -z "$LDEV" ]; then
7958 error "ldev is missing!"
7961 local LDEVCONFPATH=$TMP/ldev.conf
7962 local NIDSPATH=$TMP/nids
7964 generate_ldev_conf $LDEVCONFPATH
7965 generate_nids $NIDSPATH
7967 local LDEV_OUTPUT=$TMP/ldev-output.txt
7968 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -H $mgs_HOST \
7970 awk '{print $2}' > $LDEV_OUTPUT
7972 # ldev failed, error
7973 if [ $? -ne 0 ]; then
7974 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7975 error "ldev failed to execute!"
7979 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
7981 echo "$mgs_HOST-$(facet_fstype mgs)" > $EXPECTED_OUTPUT
7983 local facets="$(get_facets OST),$(get_facets MDS)"
7984 for facet in ${facets//,/ }; do
7985 local host=$(facet_host $facet)
7986 [ "$mgs_HOST" == "$host" ] &&
7987 echo "$host-$(facet_fstype $facet)" \
7991 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7993 if [ $? -ne 0 ]; then
7994 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7995 error "ldev failed to produce the correct output!"
7998 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8000 run_test 96 "ldev returns hostname and backend fs correctly in command sub"
8003 if [ -z "$LDEV" ]; then
8004 error "ldev is missing!"
8007 local LDEVCONFPATH=$TMP/ldev.conf
8008 local NIDSPATH=$TMP/nids
8010 generate_ldev_conf $LDEVCONFPATH
8011 generate_nids $NIDSPATH
8013 local LDEV_OUTPUT=$TMP/ldev-output.txt
8014 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
8016 echo -e "\nMDT role"
8017 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mdt > $LDEV_OUTPUT
8019 if [ $? -ne 0 ]; then
8020 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
8021 error "ldev failed to execute for mdt role!"
8024 for num in $(seq $MDSCOUNT); do
8025 printf "%s-MDT%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
8028 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
8030 if [ $? -ne 0 ]; then
8031 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8032 error "ldev failed to produce the correct output for mdt role!"
8035 echo -e "\nOST role"
8036 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R ost > $LDEV_OUTPUT
8038 if [ $? -ne 0 ]; then
8039 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
8040 error "ldev failed to execute for ost role!"
8044 for num in $(seq $OSTCOUNT); do
8045 printf "%s-OST%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
8048 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
8050 if [ $? -ne 0 ]; then
8051 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8052 error "ldev failed to produce the correct output for ost role!"
8055 echo -e "\nMGS role"
8056 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mgs > $LDEV_OUTPUT
8058 if [ $? -ne 0 ]; then
8059 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
8060 error "ldev failed to execute for mgs role!"
8063 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
8065 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
8067 if [ $? -ne 0 ]; then
8068 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8069 error "ldev failed to produce the correct output for mgs role!"
8072 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8074 run_test 97 "ldev returns correct ouput when querying based on role"
8081 check_mount || error "mount failed"
8082 mountopt="user_xattr"
8083 for ((x = 1; x <= PAGE_SIZE/11; x++)); do
8084 mountopt="$mountopt,user_xattr"
8086 mount_client $MOUNT remount,$mountopt 2>&1 | grep "too long" ||
8087 error "Buffer overflow check failed"
8088 cleanup || error "cleanup failed"
8090 run_test 98 "Buffer-overflow check while parsing mount_opts"
8094 [[ "$ost1_FSTYPE" != ldiskfs ]] &&
8095 skip "ldiskfs only test"
8096 [[ "$OST1_VERSION" -ge $(version_code 2.8.57) ]] ||
8097 skip "Need OST version at least 2.8.57"
8099 local ost_opts="$(mkfs_opts ost1 $(ostdevname 1)) \
8100 --reformat $(ostdevname 1) $(ostvdevname 1)"
8101 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" &&
8102 skip "meta_bg already set"
8105 if [[ ${!opts} != *mkfsoptions* ]]; then
8106 eval opts=\"${!opts} \
8107 --mkfsoptions='\\\"-O ^resize_inode,meta_bg\\\"'\"
8109 local val=${!opts//--mkfsoptions=\\\"/ \
8110 --mkfsoptions=\\\"-O ^resize_inode,meta_bg }
8114 echo "params: $opts"
8117 add ost1 $opts || error "add ost1 failed with new params"
8119 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" ||
8120 error "meta_bg is not set"
8122 run_test 99 "Adding meta_bg option"
8127 start_mgsmds || error "MDS start failed"
8128 start_ost || error "unable to start OST"
8129 mount_client $MOUNT || error "client start failed"
8130 check_mount || error "check_mount failed"
8139 do_facet mgs 'lshowmount -v' | awk 'BEGIN {NR == 0; rc=1} /MGS:/ {rc=0}
8140 END {exit rc}' || error "lshowmount have no output MGS"
8142 do_facet mds1 'lshowmount -v' | awk 'BEGIN {NR == 2; rc=1} /-MDT0000:/
8143 {rc=0} END {exit rc}' || error "lshowmount have no output MDT0"
8145 do_facet ost1 'lshowmount -v' | awk 'BEGIN {NR == 4; rc=1} /-OST0000:/
8146 {rc=0} END {exit rc}' || error "lshowmount have no output OST0"
8148 cleanup || error "cleanup failed with $?"
8150 run_test 100 "check lshowmount lists MGS, MDT, OST and 0@lo"
8153 local createmany_pid
8154 local dev=$FSNAME-OST0000-osc-MDT0000
8158 do_nodes $(comma_list $(osts_nodes)) $LCTL set_param \
8159 seq.*OST*-super.width=$DATA_SEQ_MAX_WIDTH
8160 createmany -o $DIR1/$tdir/$tfile-%d 50000 &
8162 # MDT->OST reconnection causes MDT<->OST last_id synchornisation
8163 # via osp_precreate_cleanup_orphans.
8164 for ((i = 0; i < 100; i++)); do
8165 for ((k = 0; k < 10; k++)); do
8166 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;" \
8167 "$LCTL --device $dev activate"
8170 ls -asl $MOUNT | grep '???' &&
8171 { kill -9 $createmany_pid &>/dev/null;
8172 error "File has no object on OST"; }
8174 kill -s 0 $createmany_pid || break
8176 wait $createmany_pid
8178 unlinkmany $DIR1/$tdir/$tfile-%d 50000
8181 run_test 101a "Race MDT->OST reconnection with create"
8184 local dev=$FSNAME-OST0000-osc-MDT0000
8185 local dir=$DIR1/$tdir
8189 $LFS setstripe -c 1 -i 0 $dir
8190 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;"
8191 #define OBD_FAIL_OSP_CON_EVENT_DELAY 0x2107
8192 do_facet mds1 "$LCTL set_param fail_loc=0x80002107 fail_val=20"
8193 do_facet $SINGLEMDS "$LCTL --device $dev activate;"
8198 wait_osc_import_ready client ost1
8199 touch $dir/$tfile || error "Can't create file"
8203 run_test 101b "Race events DISCONNECT and ACTIVE in osp"
8206 [[ "$MDS1_VERSION" -gt $(version_code 2.9.53) ]] ||
8207 skip "Need server version greater than 2.9.53"
8208 [[ “$(mdsdevname 1)” != “$(mgsdevname)” ]] &&
8209 [[ “$(facet_host mds1)” = “$(facet_host mgs)” ]] &&
8210 skip "MGS must be on different node or combined"
8212 cleanup || error "cleanup failed with $?"
8214 local mds1dev=$(mdsdevname 1)
8215 local mds1mnt=$(facet_mntpt mds1)
8216 local mds1opts=$MDS_MOUNT_OPTS
8218 if [ "$mds1_FSTYPE" == ldiskfs ] &&
8219 ! do_facet mds1 test -b $mds1dev; then
8220 mds1opts=$(csa_add "$mds1opts" -o loop)
8222 if [[ "$mds1_FSTYPE" == zfs ]]; then
8223 import_zpool mds1 || return ${PIPESTATUS[0]}
8226 # unload all and only load libcfs to allow fail_loc setting
8227 do_facet mds1 $LUSTRE_RMMOD || error "unable to unload modules"
8228 do_facet mds1 modprobe libcfs || error "libcfs not loaded"
8229 do_facet mds1 lsmod \| grep libcfs || error "libcfs not loaded"
8231 #define OBD_FAIL_OBDCLASS_MODULE_LOAD 0x60a
8232 do_facet mds1 "$LCTL set_param fail_loc=0x8000060a"
8234 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts &&
8235 error "mdt start must fail"
8236 do_facet mds1 lsmod \| grep obdclass && error "obdclass must not load"
8238 do_facet mds1 "$LCTL set_param fail_loc=0x0"
8240 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts ||
8241 error "mdt start must not fail"
8243 cleanup || error "cleanup failed with $?"
8245 run_test 102 "obdclass module cleanup upon error"
8250 echo "rename $FSNAME to $newname"
8252 if ! combined_mgs_mds ; then
8253 local dev=$(mgsdevname)
8256 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
8257 error "(7) Fail to rename MGS"
8258 if [ "$(facet_fstype mgs)" = "zfs" ]; then
8259 reimport_zpool mgs $newname-mgs
8263 for num in $(seq $MDSCOUNT); do
8264 local dev=$(mdsdevname $num)
8266 do_facet mds${num} \
8267 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
8268 error "(8) Fail to rename MDT $num"
8269 if [ "$(facet_fstype mds${num})" = "zfs" ]; then
8270 reimport_zpool mds${num} $newname-mdt${num}
8274 for num in $(seq $OSTCOUNT); do
8275 local dev=$(ostdevname $num)
8277 do_facet ost${num} \
8278 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
8279 error "(9) Fail to rename OST $num"
8280 if [ "$(facet_fstype ost${num})" = "zfs" ]; then
8281 reimport_zpool ost${num} $newname-ost${num}
8286 test_103_set_pool() {
8290 do_facet mgs $LCTL pool_add $FSNAME.$pname ${FSNAME}-$ost_x ||
8291 error "Fail to add $ost_x to $FSNAME.$pname"
8292 wait_update $HOSTNAME \
8293 "lctl get_param -n lov.$FSNAME-clilov-*.pools.$pname |
8294 grep $ost_x" "$FSNAME-${ost_x}_UUID" ||
8295 error "$ost_x is NOT in pool $FSNAME.$pname"
8298 test_103_check_pool() {
8299 local save_fsname=$1
8302 stat $DIR/$tdir/test-framework.sh ||
8303 error "($errno) Fail to stat"
8304 do_facet mgs $LCTL pool_list $FSNAME.pool1 ||
8305 error "($errno) Fail to list $FSNAME.pool1"
8306 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname ||
8307 error "($errno) Fail to list $FSNAME.$save_fsname"
8308 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname |
8309 grep ${FSNAME}-OST0000 ||
8310 error "($errno) List $FSNAME.$save_fsname is invalid"
8312 local pname=$($LFS getstripe --pool $DIR/$tdir/d0)
8313 [ "$pname" = "$save_fsname" ] ||
8314 error "($errno) Unexpected pool name $pname"
8318 check_mount_and_prep
8320 mkdir $DIR/$tdir || error "(1) Fail to mkdir $DIR/$tdir"
8321 cp $LUSTRE/tests/test-framework.sh $DIR/$tdir ||
8322 error "(2) Fail to copy test-framework.sh"
8324 do_facet mgs $LCTL pool_new $FSNAME.pool1 ||
8325 error "(3) Fail to create $FSNAME.pool1"
8326 # name the pool name as the fsname
8327 do_facet mgs $LCTL pool_new $FSNAME.$FSNAME ||
8328 error "(4) Fail to create $FSNAME.$FSNAME"
8330 test_103_set_pool $FSNAME OST0000
8332 $LFS setstripe -p $FSNAME $DIR/$tdir/d0 ||
8333 error "(6) Fail to setstripe on $DIR/$tdir/d0"
8338 test_renamefs mylustre
8340 local save_fsname=$FSNAME
8344 test_103_check_pool $save_fsname 7
8346 if [ $OSTCOUNT -ge 2 ]; then
8347 test_103_set_pool $save_fsname OST0001
8350 $LFS setstripe -p $save_fsname $DIR/$tdir/f0 ||
8351 error "(16) Fail to setstripe on $DIR/$tdir/f0"
8360 test_103_check_pool $save_fsname 17
8364 test_renamefs $save_fsname
8370 run_test 103 "rename filesystem name"
8372 test_104a() { # LU-6952
8373 local mds_mountopts=$MDS_MOUNT_OPTS
8374 local ost_mountopts=$OST_MOUNT_OPTS
8375 local mds_mountfsopts=$MDS_MOUNT_FS_OPTS
8376 local lctl_ver=$(do_facet $SINGLEMDS $LCTL --version |
8379 [[ $(version_code $lctl_ver) -lt $(version_code 2.9.55) ]] &&
8380 skip "this test needs utils above 2.9.55"
8382 # specify "acl" in mount options used by mkfs.lustre
8383 if [ -z "$MDS_MOUNT_FS_OPTS" ]; then
8384 MDS_MOUNT_FS_OPTS="acl,user_xattr"
8387 MDS_MOUNT_FS_OPTS="${MDS_MOUNT_FS_OPTS},acl,user_xattr"
8390 echo "mountfsopt: $MDS_MOUNT_FS_OPTS"
8392 #reformat/remount the MDT to apply the MDT_MOUNT_FS_OPT options
8394 if ! combined_mgs_mds ; then
8398 if [ -z "$MDS_MOUNT_OPTS" ]; then
8399 MDS_MOUNT_OPTS="-o noacl"
8401 MDS_MOUNT_OPTS="${MDS_MOUNT_OPTS},noacl"
8404 for num in $(seq $MDSCOUNT); do
8405 start mds$num $(mdsdevname $num) $MDS_MOUNT_OPTS ||
8406 error "Failed to start MDS"
8409 for num in $(seq $OSTCOUNT); do
8410 start ost$num $(ostdevname $num) $OST_MOUNT_OPTS ||
8411 error "Failed to start OST"
8415 setfacl -m "d:$RUNAS_ID:rwx" $MOUNT &&
8416 error "ACL is applied when FS is mounted with noacl."
8418 MDS_MOUNT_OPTS=$mds_mountopts
8419 OST_MOUNT_OPTS=$ost_mountopts
8420 MDS_MOUNT_FS_OPTS=$mds_mountfsopts
8422 run_test 104a "Make sure user defined options are reflected in mount"
8424 test_104b() { # LU-12859
8425 mount_client $MOUNT3 flock,localflock
8426 stack_trap "umount_client $MOUNT3" EXIT
8427 mount | grep "$MOUNT3 .*,flock" && error "flock is still set"
8428 mount | grep "$MOUNT3 .*,localflock" || error "localflock is not set"
8429 umount_client $MOUNT3
8430 mount_client $MOUNT3 localflock,flock
8431 mount | grep "$MOUNT3 .*,localflock" && error "localflock is still set"
8432 mount | grep "$MOUNT3 .*,flock" || error "flock is not set"
8433 umount_client $MOUNT3
8434 mount_client $MOUNT3 localflock,flock,noflock
8435 flock_is_enabled $MOUNT3 && error "some flock is still enabled" || true
8437 run_test 104b "Mount uses last flock argument"
8439 error_and_umount() {
8450 mount --bind $DIR $TMP/$tdir || error "mount bind mnt pt failed"
8451 rm -f $TMP/$tdir/$tfile
8452 rm -f $TMP/$tdir/${tfile}1
8454 # Files should not be created in ro bind mount point
8455 # remounting from rw to ro
8456 mount -o remount,ro $TMP/$tdir ||
8457 error_and_umount "readonly remount of bind mnt pt failed"
8458 touch $TMP/$tdir/$tfile &&
8459 error_and_umount "touch succeeds on ro bind mnt pt"
8460 [ -e $TMP/$tdir/$tfile ] &&
8461 error_and_umount "file created on ro bind mnt pt"
8463 # Files should be created in rw bind mount point
8464 # remounting from ro to rw
8465 mount -o remount,rw $TMP/$tdir ||
8466 error_and_umount "read-write remount of bind mnt pt failed"
8467 touch $TMP/$tdir/${tfile}1 ||
8468 error_and_umount "touch fails on rw bind mnt pt"
8469 [ -e $TMP/$tdir/${tfile}1 ] ||
8470 error_and_umount "file not created on rw bind mnt pt"
8471 umount $TMP/$tdir || error "umount of bind mnt pt failed"
8473 cleanup || error "cleanup failed with $?"
8475 run_test 105 "check file creation for ro and rw bind mnt pt"
8479 local creates=64768 # one full plain llog
8481 # ensure there are enough inodes in the filesystem
8482 (( OSTSIZE < (creates + 1024) * 8)) && OSTSIZE=$(((creates + 1024) * 8))
8486 mkdir -p $DIR/$tdir || error "create $tdir failed"
8487 do_nodes $(comma_list $(osts_nodes)) $LCTL set_param \
8488 seq.*OST*-super.width=$DATA_SEQ_MAX_WIDTH
8489 lfs setstripe -c 1 -i 0 $DIR/$tdir
8490 #define OBD_FAIL_CAT_RECORDS 0x1312
8491 do_facet mds1 $LCTL set_param fail_loc=0x1312 fail_val=$repeat
8493 for ((i = 1; i <= $repeat; i++)); do
8494 createmany -o $DIR/$tdir/f- $creates || lfs df -i
8495 createmany -u $DIR/$tdir/f- $creates
8496 wait_delete_completed $((TIMEOUT * 7))
8498 #ASSERTION osp_sync_thread() ( thread->t_flags != SVC_RUNNING ) failed
8499 #shows that osp code is buggy
8500 do_facet mds1 $LCTL set_param fail_loc=0 fail_val=0
8504 run_test 106 "check osp llog processing when catalog is wrapped"
8507 [[ "$MDS1_VERSION" -ge $(version_code 2.10.50) ]] ||
8508 skip "Need MDS version > 2.10.50"
8511 start_mgsmds || error "start_mgsmds failed"
8512 start_ost || error "unable to start OST"
8514 # add unknown configuration parameter.
8515 if [[ $PERM_CMD == *"set_param -P"* ]]; then
8516 cmd="$PERM_CMD ost.$FSNAME-OST0000*.unknown_param"
8518 cmd="$PERM_CMD $FSNAME-OST0000*.ost.unknown_param"
8520 do_facet mgs "$cmd=50"
8521 cleanup_nocli || error "cleanup_nocli failed with $?"
8524 # unknown param should be ignored while mounting.
8525 start_ost || error "unable to start OST after unknown param set"
8527 cleanup || error "cleanup failed with $?"
8529 run_test 107 "Unknown config param should not fail target mounting"
8534 $rcmd rm -rf $tmp > /dev/null 2>&1
8535 $rcmd mkdir -p $tmp/{mnt,images} || error "failed to mkdir remotely"
8537 for facet in $facets; do
8538 [ "$mds1_FSTYPE" = zfs ] &&
8539 $rcmd $ZPOOL -f export lustre-$facet > /dev/null 2>&1
8540 $rcmd mkdir $tmp/mnt/$facet ||
8541 error "failed to mkdir $tmp/mnt/$facet"
8542 $rcmd dd if=/dev/zero of=$tmp/images/$facet \
8543 seek=199 bs=1M count=1 ||
8544 error "failed to create $tmp/images/$facet"
8553 local facet=${role}$((idx + 1))
8555 [ $# -eq 5 ] && pool=$5
8557 do_facet $SINGLEMDS $MKFS --fsname=lustre --$mgs \
8558 --$role --index=$idx --replace --backfstype=$bkfs \
8559 --device-size=200000 --reformat $pool $tmp/images/$facet ||
8560 error "failed to mkfs for $facet"
8564 echo "mounting client..."
8565 mount -t lustre ${nid}:/lustre $MOUNT ||
8566 error "failed to mount lustre"
8569 ls -l $MOUNT/local_dir || error "failed to list"
8571 echo "check truncate && write"
8572 echo "dummmmmmmmmmmmm" > $MOUNT/remote_dir/fsx.c ||
8573 error "failed to tuncate & write"
8576 touch $MOUNT/foooo ||
8577 error "failed to create"
8579 echo "check read && write && append"
8580 sha1sum $MOUNT/conf-sanity.sh |
8581 awk '{ print $1 }' > $MOUNT/checksum.new ||
8582 error "failed to read(1)"
8583 sha1sum $MOUNT/remote_dir/unlinkmany.c |
8584 awk '{ print $1 }' >> $MOUNT/checksum.new ||
8585 error "failed to read(2)"
8586 sha1sum $MOUNT/striped_dir/lockahead_test.o |
8587 awk '{ print $1 }' >> $MOUNT/checksum.new ||
8588 error "failed to read(3)"
8591 diff $MOUNT/checksum.new $MOUNT/checksum.src ||
8592 error "failed to verify data"
8602 umount -f $MOUNT || error "failed to umount client"
8603 for facet in $facets; do
8604 $rcmd umount -f $tmp/mnt/$facet ||
8605 error "failed to umount $facet"
8606 if [ "$mds1_FSTYPE" = zfs ]; then
8607 $rcmd $ZPOOL export -f lustre-$facet ||
8608 error "failed to export lustre-$facet"
8612 $rcmd rm -rf $tmp || error "failed to rm the dir $tmp"
8616 [ "$CLIENTONLY" ] && skip "Client-only testing"
8617 [ "$mds1_FSTYPE" != zfs ] && skip "zfs only test"
8618 [ "$MDS1_VERSION" -lt $(version_code 2.10.58) ] &&
8619 skip "Need server version at least 2.10.58"
8624 local tmp=$TMP/$tdir
8625 local rcmd="do_facet $SINGLEMDS"
8626 local facets="mdt1 mdt2 ost1 ost2"
8627 local nid=$($rcmd $LCTL list_nids | head -1)
8630 trap t_108_cleanup EXIT ERR
8633 t_108_mkfs mdt 0 zfs mgs lustre-mdt1/mdt1
8634 t_108_mkfs mdt 1 zfs mgsnode=$nid lustre-mdt2/mdt2
8635 t_108_mkfs ost 0 zfs mgsnode=$nid lustre-ost1/ost1
8636 t_108_mkfs ost 1 zfs mgsnode=$nid lustre-ost2/ost2
8638 for facet in $facets; do
8639 $rcmd zfs set mountpoint=$tmp/mnt/$facet canmount=on \
8640 lustre-$facet/$facet ||
8641 error "failed to zfs set for $facet (1)"
8642 $rcmd zfs mount lustre-$facet/$facet ||
8643 error "failed to local mount $facet"
8644 $rcmd tar jxf $LUSTRE/tests/ldiskfs_${facet}_2_11.tar.bz2 \
8645 --xattrs --xattrs-include="trusted.*" \
8646 -C $tmp/mnt/$facet/ > /dev/null 2>&1 ||
8647 error "failed to untar image for $facet"
8648 $rcmd "cd $tmp/mnt/$facet && rm -rf oi.* OI_* lfsck_* LFSCK" ||
8649 error "failed to cleanup for $facet"
8650 $rcmd zfs umount lustre-$facet/$facet ||
8651 error "failed to local umount $facet"
8652 $rcmd zfs set canmount=off lustre-$facet/$facet ||
8653 error "failed to zfs set $facet (2)"
8656 echo "changing server nid..."
8657 $rcmd mount -t lustre -o nosvc lustre-mdt1/mdt1 $tmp/mnt/mdt1
8658 $rcmd lctl replace_nids $FSNAME-MDT0000 $nid
8659 $rcmd lctl replace_nids $FSNAME-MDT0001 $nid
8660 $rcmd lctl replace_nids $FSNAME-OST0000 $nid
8661 $rcmd lctl replace_nids $FSNAME-OST0001 $nid
8662 $rcmd umount $tmp/mnt/mdt1
8664 for facet in $facets; do
8665 echo "mounting $facet from backup..."
8666 $rcmd mount -t lustre -o abort_recov lustre-$facet/$facet \
8667 $tmp/mnt/$facet || error "failed to mount $facet"
8670 # ZFS backend can detect migration and trigger OI scrub automatically
8671 # sleep 3 seconds for scrub done
8677 run_test 108a "migrate from ldiskfs to ZFS"
8680 [ "$CLIENTONLY" ] && skip "Client-only testing"
8681 [ "$mds1_FSTYPE" != ldiskfs ] && skip "ldiskfs only test"
8682 [ "$MDS1_VERSION" -lt $(version_code 2.10.58) ] &&
8683 skip "Need server version at least 2.10.58"
8688 local tmp=$TMP/$tdir
8689 local rcmd="do_facet $SINGLEMDS"
8690 local facets="mdt1 mdt2 ost1 ost2"
8691 local scrub_list="MDT0000 MDT0001 OST0000 OST0001"
8692 local nid=$($rcmd $LCTL list_nids | head -1)
8695 trap t_108_cleanup EXIT ERR
8698 t_108_mkfs mdt 0 ldiskfs mgs
8699 t_108_mkfs mdt 1 ldiskfs mgsnode=$nid
8700 t_108_mkfs ost 0 ldiskfs mgsnode=$nid
8701 t_108_mkfs ost 1 ldiskfs mgsnode=$nid
8703 for facet in $facets; do
8704 $rcmd mount -t ldiskfs -o loop $tmp/images/$facet \
8706 error "failed to local mount $facet"
8708 $rcmd tar jxf $LUSTRE/tests/zfs_${facet}_2_11.tar.bz2 \
8709 --xattrs --xattrs-include="*.*" \
8710 -C $tmp/mnt/$facet/ > /dev/null 2>&1 ||
8711 error "failed to untar image for $facet"
8712 $rcmd "cd $tmp/mnt/$facet && rm -rf oi.* OI_* lfsck_* LFSCK" ||
8713 error "failed to cleanup for $facet"
8714 $rcmd umount $tmp/mnt/$facet ||
8715 error "failed to local umount $facet"
8718 echo "changing server nid..."
8719 $rcmd mount -t lustre -o nosvc,loop $tmp/images/mdt1 $tmp/mnt/mdt1
8720 $rcmd lctl replace_nids $FSNAME-MDT0000 $nid
8721 $rcmd lctl replace_nids $FSNAME-MDT0001 $nid
8722 $rcmd lctl replace_nids $FSNAME-OST0000 $nid
8723 $rcmd lctl replace_nids $FSNAME-OST0001 $nid
8724 $rcmd umount $tmp/mnt/mdt1
8726 for facet in $facets; do
8727 echo "mounting $facet from backup..."
8728 $rcmd mount -t lustre -o loop,abort_recov $tmp/images/$facet \
8729 $tmp/mnt/$facet || error "failed to mount $facet"
8732 for facet in $scrub_list; do
8733 $rcmd $LCTL lfsck_start -M $FSNAME-$facet -t scrub ||
8734 error "failed to start OI scrub on $facet"
8737 # sleep 3 seconds for scrub done
8743 run_test 108b "migrate from ZFS to ldiskfs"
8746 # set number of permanent parameters
8748 test_109_set_params() {
8751 set_persistent_param_and_check mds \
8752 "mdd.$fsname-MDT0000.atime_diff" \
8753 "$fsname-MDT0000.mdd.atime_diff" \
8755 set_persistent_param_and_check mds \
8756 "mdd.$fsname-MDT0000.atime_diff" \
8757 "$fsname-MDT0000.mdd.atime_diff" \
8759 set_persistent_param_and_check client \
8760 "llite.$fsname*.max_read_ahead_mb" \
8761 "$fsname.llite.max_read_ahead_mb" \
8763 set_persistent_param_and_check client \
8764 "llite.$fsname*.max_read_ahead_mb" \
8765 "$fsname.llite.max_read_ahead_mb" \
8767 create_pool $fsname.pool1 || error "create pool failed"
8768 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
8769 error "pool_add failed"
8770 do_facet mgs $LCTL pool_remove $fsname.pool1 OST0000 ||
8771 error "pool_remove failed"
8772 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
8773 error "pool_add failed"
8777 # check permanent parameters
8779 test_109_test_params() {
8782 local atime_diff=$(do_facet mds $LCTL \
8783 get_param -n mdd.$fsname-MDT0000.atime_diff)
8784 [ $atime_diff == 63 ] || error "wrong mdd parameter after clear_conf"
8785 local max_read_ahead_mb=$(do_facet client $LCTL \
8786 get_param -n llite.$fsname*.max_read_ahead_mb)
8787 [ $max_read_ahead_mb == 64 ] ||
8788 error "wrong llite parameter after clear_conf"
8789 local ost_in_pool=$(do_facet mds $LCTL pool_list $fsname.pool1 |
8790 grep -v "^Pool:" | sed 's/_UUID//')
8791 [ $ost_in_pool = "$fsname-OST0000" ] ||
8792 error "wrong pool after clear_conf"
8796 # run lctl clear_conf, store CONFIGS before and after that
8798 test_109_clear_conf()
8800 local clear_conf_arg=$1
8803 if ! combined_mgs_mds ; then
8804 mgsdev=$(mgsdevname)
8805 stop_mgs || error "stop_mgs failed"
8806 start_mgs "-o nosvc" || error "start_mgs nosvc failed"
8808 mgsdev=$(mdsdevname 1)
8809 start_mdt 1 "-o nosvc" || error "start_mdt 1 nosvc failed"
8812 do_facet mgs "rm -rf $TMP/${tdir}/conf1; mkdir -p $TMP/${tdir}/conf1;" \
8813 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf1\\\" \
8817 # the command being tested
8819 do_facet mgs $LCTL clear_conf $clear_conf_arg ||
8820 error "clear_conf failed"
8821 if ! combined_mgs_mds ; then
8822 stop_mgs || error "stop_mgs failed"
8824 stop_mdt 1 || error "stop_mdt 1 failed"
8827 do_facet mgs "rm -rf $TMP/${tdir}/conf2; mkdir -p $TMP/${tdir}/conf2;" \
8828 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf2\\\" \
8832 test_109_file_shortened() {
8834 local sizes=($(do_facet mgs "stat -c %s " \
8835 "$TMP/${tdir}/conf1/CONFIGS/$file" \
8836 "$TMP/${tdir}/conf2/CONFIGS/$file"))
8837 [ ${sizes[1]} -lt ${sizes[0]} ] && return 0
8843 [ $MDS1_VERSION -lt $(version_code 2.10.59) ] &&
8844 skip "Needs MDS version 2.10.59 or later."
8845 [ "$(facet_fstype mgs)" == zfs ] &&
8846 skip "LU-8727: no implementation for ZFS"
8851 client_up || error "client_up failed"
8854 # set number of permanent parameters
8856 test_109_set_params $FSNAME
8858 umount_client $MOUNT || error "umount_client failed"
8859 stop_ost || error "stop_ost failed"
8860 stop_mds || error "stop_mds failed"
8862 test_109_clear_conf $FSNAME
8864 # make sure that all configs are cleared
8866 test_109_file_shortened $FSNAME-MDT0000 ||
8867 error "failed to clear MDT0000 config"
8868 test_109_file_shortened $FSNAME-client ||
8869 error "failed to clear client config"
8874 # check that configurations are intact
8876 test_109_test_params $FSNAME
8881 destroy_test_pools || error "destroy test pools failed"
8885 run_test 109a "test lctl clear_conf fsname"
8889 [ $MDS1_VERSION -lt $(version_code 2.10.59) ] &&
8890 skip "Needs MDS version 2.10.59 or later."
8891 [ "$(facet_fstype mgs)" == zfs ] &&
8892 skip "LU-8727: no implementation for ZFS"
8897 client_up || error "client_up failed"
8900 # set number of permanent parameters
8902 test_109_set_params $FSNAME
8904 umount_client $MOUNT || error "umount_client failed"
8905 stop_ost || error "stop_ost failed"
8906 stop_mds || error "stop_mds failed"
8908 test_109_clear_conf $FSNAME-MDT0000
8910 # make sure that only one config is cleared
8912 test_109_file_shortened $FSNAME-MDT0000 ||
8913 error "failed to clear MDT0000 config"
8914 test_109_file_shortened $FSNAME-client &&
8915 error "failed to clear client config"
8919 # check that configurations are intact
8921 test_109_test_params $FSNAME
8926 destroy_test_pools || error "destroy test pools failed"
8930 run_test 109b "test lctl clear_conf one config"
8934 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
8935 skip "Only applicable to ldiskfs-based MDTs"
8937 do_facet $SINGLEMDS $DEBUGFS -w -R supported_features |grep large_dir ||
8938 skip "large_dir option is not supported on MDS"
8939 do_facet ost1 $DEBUGFS -w -R supported_features | grep large_dir ||
8940 skip "large_dir option is not supported on OSS"
8942 stopall # stop all targets before modifying the target counts
8943 local old_mdscount=$MDSCOUNT
8944 local old_ostcount=$OSTCOUNT
8946 stack_trap "MDSCOUNT=$old_mdscount OSTCOUNT=$old_ostcount; reformat"
8950 # ext4_dir_entry_2 struct size:264
8951 # dx_root struct size:8
8952 # dx_node struct size:8
8953 # dx_entry struct size:8
8954 # For 1024 bytes block size.
8955 # First level directory entries: 126
8956 # Second level directory entries: 127
8957 # Entries in leaf: 3
8958 # For 2 levels limit: 48006
8959 # For 3 levels limit : 6096762
8960 # Create 80000 files to safely exceed 2-level htree limit.
8961 CONF_SANITY_110_LINKS=${CONF_SANITY_110_LINKS:-80000}
8963 # can fit at most 3 filenames per 1KB leaf block, but each
8964 # leaf/index block will only be 3/4 full before split at each level
8965 (( MDSSIZE < CONF_SANITY_110_LINKS / 3 * 4/3 * 4/3 )) &&
8966 CONF_SANITY_110_LINKS=$((MDSSIZE * 3 * 3/4 * 3/4))
8968 combined_mgs_mds || replace=" --replace "
8969 local opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
8970 $replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
8971 if [[ $opts != *large_dir* ]]; then
8972 if [[ $opts != *mkfsoptions* ]]; then
8973 opts+=" --mkfsoptions=\\\"-O large_dir -b 1024 -i 65536\\\""
8975 opts="${opts//--mkfsoptions=\\\"/ \
8976 --mkfsoptions=\\\"-O large_dir -b 1024 -i 65536 }"
8979 echo "MDT params: $opts"
8981 combined_mgs_mds || start_mgs
8982 add mds1 $opts || error "add mds1 failed with new params"
8983 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
8985 opts="$(mkfs_opts ost1 $(ostdevname 1)) \
8986 $replace --reformat $(ostdevname 1) $(ostvdevname 1)"
8988 if [[ $opts != *large_dir* ]]; then
8989 if [[ $opts != *mkfsoptions* ]]; then
8990 opts+=" --mkfsoptions=\\\"-O large_dir\\\" "
8992 opts="${opts//--mkfsoptions=\\\"/ \
8993 --mkfsoptions=\\\"-O large_dir }"
8996 echo "OST params: $opts"
8997 add ost1 $opts || error "add ost1 failed with new params"
8998 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
9000 MOUNT_2=yes mountcli || error "mount clients failed"
9002 mkdir -v $DIR/$tdir || error "cannot create $DIR/$tdir"
9003 local pids count=0 group=0
9005 echo "creating $CONF_SANITY_110_LINKS in total"
9006 while (( count < CONF_SANITY_110_LINKS )); do
9007 local len=$((253 - $(wc -c <<<"$tfile-$group-40000-")))
9008 local dir=DIR$((group % 2 + 1))
9009 local target=${!dir}/$tdir/$tfile-$group
9010 local long=$target-$(generate_name $len)-
9011 local create=$((CONF_SANITY_110_LINKS - count))
9013 (( create > 40000 )) && create=40000
9014 touch $target || error "creating $target failed"
9015 echo "creating $create hard links to $target"
9016 createmany -l $target $long $create &
9019 count=$((count + create))
9020 group=$((group + 1))
9022 echo "waiting for PIDs$pids to complete"
9023 wait $pids || error "createmany failed after $group groups"
9025 umount_client $MOUNT2 -f
9028 run_e2fsck $(facet_active_host mds1) $(mdsdevname 1) -n
9029 MDSCOUNT=$old_mdscount
9030 OSTCOUNT=$old_ostcount
9032 run_test 110 "Adding large_dir with 3-level htree"
9035 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
9036 skip "Only applicable to ldiskfs-based MDTs"
9038 is_dm_flakey_dev $SINGLEMDS $(mdsdevname 1) &&
9039 skip "This test can not be executed on flakey dev"
9041 do_facet $SINGLEMDS $DEBUGFS -w -R supported_features |grep large_dir ||
9042 skip "large_dir option is not supported on MDS"
9044 do_facet ost1 $DEBUGFS -w -R supported_features | grep large_dir ||
9045 skip "large_dir option is not supported on OSS"
9047 # cleanup before changing target counts
9049 local old_mdscount=$MDSCOUNT
9050 local old_ostcount=$OSTCOUNT
9051 local old_mdssize=$MDSSIZE
9053 stack_trap "MDSSIZE=$old_mdssize MDSCOUNT=$old_mdscount \
9054 OSTCOUNT=$old_ostcount; reformat"
9057 (( MDSSIZE < 2400000 )) && MDSSIZE=2400000 # need at least 2.4GB
9059 local mdsdev=$(mdsdevname 1)
9060 combined_mgs_mds || replace=" --replace "
9061 local opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
9062 $replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
9063 if [[ $opts != *large_dir* ]]; then
9064 if [[ $opts != *mkfsoptions* ]]; then
9065 opts+=" --mkfsoptions=\\\"-O large_dir -i 1048576 \\\" "
9067 opts="${opts//--mkfsoptions=\\\"/ \
9068 --mkfsoptions=\\\"-O large_dir -i 1048576 }"
9071 echo "MDT params: $opts"
9073 combined_mgs_mds || start_mgs
9074 __touch_device mds 1
9075 add mds1 $opts || error "add mds1 failed with new params"
9076 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
9078 opts="$(mkfs_opts ost1 $(ostdevname 1)) \
9079 $replace --reformat $(ostdevname 1) $(ostvdevname 1)"
9080 if [[ $opts != *large_dir* ]]; then
9081 if [[ $opts != *mkfsoptions* ]]; then
9082 opts+=" --mkfsoptions=\\\"-O large_dir \\\""
9084 opts="${opts//--mkfsoptions=\\\"/ --mkfsoptions=\\\"-O large_dir }"
9087 echo "OST params: $opts"
9088 __touch_device ost 1
9089 add ost1 $opts || error "add ost1 failed with new params"
9090 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
9092 MOUNT_2=yes mountcli
9093 mkdir $DIR/$tdir || error "cannot create $DIR/$tdir"
9095 lfs df -i $DIR/$tdir
9099 local start=$SECONDS
9101 local dirmax=$((2 << 30))
9107 while (( !needskip & dirsize < dirmax )); do
9110 for cli in ${CLIENTS//,/ }; do
9111 local len=$((253 - $(wc -c <<<"$cli-$group-60000-")))
9112 local target=$cli-$group
9113 local long=$DIR/$tdir/$target-$(generate_name $len)-
9115 RPWD=$DIR/$tdir do_node $cli touch $target ||
9116 error "creating $target failed"
9117 echo "creating 60000 hardlinks to $target"
9118 RPWD=$DIR/$tdir do_node $cli createmany -l $target $long 60000 &
9121 group=$((group + 1))
9123 long=$DIR2/$tdir/$target-$(generate_name $len)-
9125 RPWD=$DIR2/$tdir do_node $cli touch $target ||
9126 error "creating $target failed"
9127 echo "creating 60000 hardlinks to $target"
9128 RPWD=$DIR2/$tdir do_node $cli createmany -l $target $long 60000 &
9131 group=$((group + 1))
9133 echo "waiting for PIDs$pids to complete"
9134 wait $pids || error "createmany failed after $group groups"
9135 dirsize=$(stat -c %s $DIR/$tdir)
9136 taken=$((SECONDS - start))
9137 rate=$((dirsize / taken))
9138 left=$(((dirmax - dirsize) / rate))
9139 num=$((group * 60000))
9140 echo "estimate ${left}s left after $num files / ${taken}s"
9141 # if the estimated time remaining is too large (it may change
9142 # over time as the create rate is not constant) then exit
9143 # without declaring a failure.
9144 (( left > 1200 )) && needskip=1
9147 umount_client $MOUNT2 -f
9150 ! (( $needskip )) ||
9151 echo "ETA ${left}s after $num files / ${taken}s is too long"
9153 run_e2fsck $(facet_active_host mds1) $(mdsdevname 1) -n
9154 MDSCOUNT=$old_mdscount
9155 OSTCOUNT=$old_ostcount
9156 MDSSIZE=$old_mdssize
9158 run_test 111 "Adding large_dir with over 2GB directory"
9161 start_mds || error "MDS start failed"
9162 start_ost || error "OSS start failed"
9163 echo "start ost2 service on $(facet_active_host ost2)"
9164 start ost2 $(ostdevname 2) $(csa_add "$OST_MOUNT_OPTS" -o no_precreate) ||
9165 error "start ost2 facet failed"
9166 local val=$(do_facet ost2 \
9167 "$LCTL get_param -n obdfilter.$FSNAME-OST0001*.no_precreate")
9168 (( $val == 1 )) || error "obdfilter.$FSNAME-OST0001*.no_precreate=$val"
9170 mount_client $MOUNT || error "mount client failed"
9171 wait_osc_import_state mds1 ost1 FULL
9172 wait_osc_import_ready client ost1
9173 wait_osc_import_ready client ost2
9175 $LFS setstripe -i 0 $DIR/$tfile.0 ||
9176 error "problem creating $tfile.0 on OST0000"
9177 $LFS setstripe -i 1 $DIR/$tfile.1 && $LFS getstripe $DIR/$tfile.1 &&
9178 (( $($LFS getstripe -i $DIR/$tfile.1) == 1 )) &&
9179 error "allowed to create $tfile.1 on OST0001"
9180 do_facet ost2 $LCTL set_param obdfilter.*.no_precreate=0
9182 $LFS setstripe -i 1 $DIR/$tfile.2 ||
9183 error "failed to create $tfile.2 on ost1 facet"
9184 # files not cleaned with ONLY_REPEAT because of client unmount below
9185 rm -f $DIR/$tfile.[012]
9186 stop_ost2 || error "stop ost2 facet failed"
9189 run_test 112 "mount OST with nocreate option"
9192 SAVE_MGS_MOUNT_OPTS=$MGS_MOUNT_OPTS
9193 SAVE_MDS_MOUNT_OPTS=$MDS_MOUNT_OPTS
9194 SAVE_OST_MOUNT_OPTS=$OST_MOUNT_OPTS
9200 MGS_MOUNT_OPTS=$SAVE_MGS_MOUNT_OPTS
9201 MDS_MOUNT_OPTS=$SAVE_MDS_MOUNT_OPTS
9202 OST_MOUNT_OPTS=$SAVE_OST_MOUNT_OPTS
9203 # Revert old mount options back
9205 # Subsequent following test requires
9206 # conf-sanity to be in stopall state.
9207 # Force 'stopall' so others following
9212 # Error out with mount info
9214 local server_nodes=$(comma_list $(mdts_nodes) $(osts_nodes))
9217 echo "--Client Mount Info--"
9218 mount | grep -i lustre
9219 echo "--Server Mount Info--"
9220 do_nodes $server_nodes mount | grep -i lustre
9226 local ost_version="2.15.51" # Minimum version required
9228 (( OST1_VERSION >= $(version_code $ost_version) )) ||
9229 skip "Need server version at least $ost_version"
9231 stack_trap cleanup_113 EXIT
9233 # Reset before starting
9237 # Verify MDS's should start with "rw"
9238 do_facet $SINGLEMDS mount | grep "lustre.*rw,.*MDT" ||
9239 error_113 "$SINGLEMDS should be read-write"
9241 # Verify OST's should start with "rw"
9242 for (( i=1; i <= OSTCOUNT; i++ )); do
9243 do_facet ost$i mount | grep "lustre.*rw,.*OST" ||
9244 error_113 "ost$i should be read-write"
9247 # rdonly_dev does not currently work for ldiskfs
9248 # We skip the rdonly_dev check until then.
9249 if [[ $ost1_FSTYPE == ldiskfs ]]; then
9250 echo "Shadow Mountpoint correctly reports rw for ldiskfs"
9255 # Only ZFS specific tests below.
9258 # Must stop all (server+client) and restart to verify new
9262 # add rdonly_dev to mount option
9263 MGS_MOUNT_OPTS=$(csa_add "$MGS_MOUNT_OPTS" -o rdonly_dev)
9264 MDS_MOUNT_OPTS=$(csa_add "$MDS_MOUNT_OPTS" -o rdonly_dev)
9265 OST_MOUNT_OPTS=$(csa_add "$OST_MOUNT_OPTS" -o rdonly_dev)
9267 # Only restart server(mds/ost). Sufficient for test
9268 setupall server_only || error "Fail to start servers"
9270 # Verify MDS's should be "ro"
9271 do_facet $SINGLEMDS mount | grep "lustre.*ro,.*MDT.*rdonly_dev" ||
9272 error_113 "$SINGLEMDS should be read-only"
9274 # Verify OST's should be "ro"
9275 for (( i=1; i <= OSTCOUNT; i++ )); do
9276 do_facet ost$i mount | grep "lustre.*ro,.*OST.*rdonly_dev" ||
9277 error_113 "ost$i should be read-only"
9280 run_test 113 "Shadow mountpoint correctly report ro/rw for mounts"
9283 # Purpose: To verify dynamic thread (OSS) creation.
9284 # (This was sanity/115)
9287 [ $PARALLEL == "yes" ] && skip "skip parallel run"
9288 remote_ost_nodsh && skip "remote OST with nodsh"
9290 # Lustre does not stop service threads once they are started.
9291 # Reset number of running threads to default.
9296 local save_params="$TMP/sanity-$TESTNAME.parameters"
9298 # Get ll_ost_io count before I/O
9299 OSTIO_pre=$(do_facet ost1 \
9300 "$LCTL get_param ost.OSS.ost_io.threads_started | cut -d= -f2")
9301 # Exit if lustre is not running (ll_ost_io not running).
9302 [ -z "$OSTIO_pre" ] && error "no OSS threads"
9304 echo "Starting with $OSTIO_pre threads"
9305 local thread_max=$((OSTIO_pre * 2))
9306 local rpc_in_flight=$((thread_max * 2))
9308 # this is limited to OSC_MAX_RIF_MAX (256)
9309 [ $rpc_in_flight -gt 256 ] && rpc_in_flight=256
9310 thread_max=$((rpc_in_flight / 2))
9311 [ $thread_max -le $OSTIO_pre ] && skip "Too many ost_io threads" &&
9314 # Number of I/O Process proposed to be started.
9316 local facets=$(get_facets OST)
9318 save_lustre_params client "osc.*OST*.max_rpcs_in_flight" > $save_params
9319 save_lustre_params $facets "ost.OSS.ost_io.threads_max" >> $save_params
9321 # Set in_flight to $rpc_in_flight
9322 $LCTL set_param osc.*OST*.max_rpcs_in_flight=$rpc_in_flight ||
9323 error "Failed to set max_rpcs_in_flight to $rpc_in_flight"
9324 nfiles=${rpc_in_flight}
9325 # Set ost thread_max to $thread_max
9326 do_facet ost1 "$LCTL set_param ost.OSS.ost_io.threads_max=$thread_max"
9328 # 5 Minutes should be sufficient for max number of OSS
9329 # threads(thread_max) to be created.
9333 local wtl=${WTL:-"$LUSTRE/tests/write_time_limit"}
9335 test_mkdir $DIR/$tdir
9336 for ((i = 1; i <= nfiles; i++)); do
9337 local file=$DIR/$tdir/${tfile}-$i
9339 $LFS setstripe -c -1 -i 0 $file
9340 ($wtl $file $timeout)&
9343 # I/O Started - Wait for thread_started to reach thread_max or report
9344 # error if thread_started is more than thread_max.
9345 echo "Waiting for thread_started to reach thread_max"
9346 local thread_started=0
9347 local end_time=$((SECONDS + timeout))
9349 while [ $SECONDS -le $end_time ] ; do
9351 # Get ost i/o thread_started count.
9352 thread_started=$(do_facet ost1 \
9354 ost.OSS.ost_io.threads_started | cut -d= -f2")
9355 # Break out if thread_started is equal/greater than thread_max
9356 if (( $thread_started >= $thread_max )); then
9357 echo ll_ost_io thread_started $thread_started, \
9358 equal/greater than thread_max $thread_max
9364 # Cleanup - We have the numbers, Kill i/o jobs if running.
9365 jobcount=($(jobs -p))
9367 for ((i=0; i < ${#jobcount[*]}; i++)); do
9368 kill -9 ${jobcount[$i]}
9369 if [ $? -ne 0 ] ; then
9370 echo "warning: cannot kill WTL pid ${jobcount[$i]}"
9374 # Cleanup files left by WTL binary.
9375 for ((i = 1; i <= nfiles; i++)); do
9376 local file=$DIR/$tdir/${tfile}-$i
9379 if [ $? -ne 0 ] ; then
9380 echo "Warning: Failed to delete file $file"
9384 restore_lustre_params <$save_params
9385 rm -f $save_params || echo "Warning: delete file '$save_params' failed"
9387 # Error out if no new thread has started or Thread started is greater
9389 if (( $thread_started <= $OSTIO_pre ||
9390 $thread_started > $thread_max )); then
9391 error "ll_ost_io: thread_started $thread_started" \
9392 "OSTIO_pre $OSTIO_pre, thread_max $thread_max." \
9393 "No new thread started or thread started greater " \
9397 run_test 114 "verify dynamic thread creation===================="
9403 do_facet mds1 rm -f $1
9407 if [ "$mds1_FSTYPE" != ldiskfs ]; then
9408 skip "Only applicable to ldiskfs-based MDTs"
9410 [ -n "$FILESET" ] && skip "Not functional for FILESET set"
9412 local dbfs_ver=$(do_facet mds1 $DEBUGFS -V 2>&1)
9414 echo "debugfs version: $dbfs_ver"
9415 echo "$dbfs_ver" | egrep -w "1.44.3.wc1|1.44.5.wc1|1.45.2.wc1" &&
9416 skip_env "This version of debugfs doesn't show inode number"
9418 local IMAGESIZE=$((3072 << 30)) # 3072 GiB
9428 do_facet mds1 "hostname; $LCTL dl; mount"
9430 do_facet ost1 "hostname; $LCTL dl; mount"
9431 # We need MDT size 3072GB, because it is smallest
9432 # partition that can store 2B inodes
9433 do_facet mds1 "mkdir -p $TMP/$tdir"
9434 local mdsimgname=$TMP/$tdir/lustre-mdt
9436 do_facet mds1 "rm -f $mdsimgname"
9437 do_facet mds1 "touch $mdsimgname"
9438 stack_trap "cleanup_115 $mdsimgname" EXIT
9439 do_facet mds1 "$TRUNCATE $mdsimgname $IMAGESIZE" ||
9440 skip "Backend FS doesn't support sparse files"
9441 local mdsdev=$(do_facet mds1 "losetup -f")
9443 do_facet mds1 "losetup $mdsdev $mdsimgname"
9445 local mds_opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
9446 --mkfsoptions='-O ea_inode,^resize_inode,meta_bg \
9447 -N 2247484000 -E lazy_itable_init' --device-size=$IMAGESIZE"
9448 add mds1 $mds_opts --mgs --reformat $mdsdev || {
9449 do_facet $SINGLEMDS \
9450 "losetup -d $mdsdev && rm -f $mdsimgname"
9451 skip_env "format large MDT failed"
9454 local ostdev=$(ostdevname 1)
9456 local opts="$(mkfs_opts ost1 $ostdev) \
9457 --reformat $ostdev $ostdev"
9458 add ost1 $opts || error "add ost1 failed with new params"
9459 start mds1 $mdsdev $MDS_MOUNT_OPTS || error "start MDS failed"
9460 start_ost || error "start OSS failed"
9461 mount_client $MOUNT || error "mount client failed"
9463 mkdir -p $DIR/$tdir || error "mkdir $DIR/$tdir fail"
9464 local goal="/sys/fs/ldiskfs/$(basename $mdsdev)/inode_goal"
9466 # 2147483648 is 0x80000000
9467 do_facet mds1 "echo 2147483648 >> $goal; grep . $goal"
9468 touch $DIR/$tdir/$tfile
9470 # attrs from 1 to 16 go to block, 17th - to inode
9473 for i in {1..17}; do
9474 local nm="trusted.ea$i"
9475 setfattr -n $nm -v $(printf "xattr%0250d" $i) $DIR/$tdir/$tfile
9478 do_facet mds1 "$DEBUGFS -c -R 'stat ROOT/$tdir/$tfile' $mdsdev"
9480 # inode <2147483649> trusted.ea16 (255)
9481 local inode_num=$(do_facet mds1 \
9482 "$DEBUGFS -c -R 'stat ROOT/$tdir/$tfile' $mdsdev" |
9483 awk '/ea17/ { print $2 }' |
9484 sed -e 's/>//' -e 's/<//' -e 's/\"//')
9485 echo "inode num: $inode_num"
9486 [ $inode_num -ge 2147483648 ] || error "inode $inode_num too small"
9487 do_facet mds1 "losetup -d $mdsdev"
9489 run_test 115 "Access large xattr with inodes number over 2TB"
9492 [ "$mds1_FSTYPE" != ldiskfs ] && skip "ldiskfs only test"
9493 [ "$MDS1_VERSION" -lt $(version_code 2.10.59) ] &&
9494 skip "Need server version at least 2.10.59"
9495 do_facet $SINGLEMDS which mkfs.xfs ||
9496 skip_env "No mkfs.xfs installed"
9501 local tmpmnt=$TMP/$tdir
9502 local mdtimg=$tfile-mdt0
9504 do_facet $SINGLEMDS mkdir -p $tmpmnt
9505 stack_trap "do_facet $SINGLEMDS rmdir $tmpmnt" EXIT
9507 do_facet $SINGLEMDS touch $TMP/$mdtimg
9508 stack_trap "do_facet $SINGLEMDS rm -f $TMP/$mdtimg" EXIT
9509 do_facet $SINGLEMDS mkfs -t xfs -d file,size=1t,name=$TMP/$mdtimg ||
9510 error "mkfs temporary xfs image"
9512 do_facet $SINGLEMDS mount $TMP/$mdtimg $tmpmnt ||
9513 error "mount temporary xfs image"
9514 stack_trap "do_facet $SINGLEMDS umount $tmpmnt" EXIT
9515 local old_mdssize=$MDSSIZE
9516 local old_mdsisize=$MDSISIZE
9518 MDSSIZE=$((17 * 1024 * 1024 * 1024)) # 17T MDT
9519 MDSISIZE=$((16 << 20))
9520 local opts17t="$(mkfs_opts $SINGLEMDS)"
9522 MDSSIZE=$old_mdssize
9523 MDSISIZE=$old_mdsisize
9525 do_facet $SINGLEMDS $MKFS \
9526 ${opts17t/-E /-E lazy_itable_init,lazy_journal_init,} \
9528 error "failed to mkfs for $tmpmnt/$mdtimg"
9530 do_facet $SINGLEMDS $TUNE2FS -l $tmpmnt/$mdtimg |
9531 grep -qw 'features.*extent' || error "extent should be enabled"
9534 run_test 116 "big size MDT support"
9537 # Call setup only if LustreFS is not mounted
9538 check_mount || setup
9540 do_facet ost1 "$LCTL set_param ost.OSS.ost_io.nrs_policies=fifo"
9541 do_facet ost1 "$LCTL get_param -n ost.OSS.ost_io.nrs_tbf_rule" &&
9542 error "get_param should fail"
9543 cleanup || error "cleanup failed with rc $?"
9545 run_test 117 "lctl get_param return errors properly"
9551 [[ "$MDSCOUNT" -ge 2 ]] || skip "Need more at least 2 MDTs"
9553 had_config=$(do_facet mds1 "$LCTL get_param debug | grep config")
9554 do_facet mds1 "$LCTL set_param debug=+config"
9555 do_facet mds1 "$LCTL clear"
9558 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)" &>/dev/null
9559 # mount after writeconf will make "add osp" added to mdt0 config:
9560 # 53 (224)marker 60 (flags=0x01, v2.5.1.0) lustre-MDT0001 'add osp'
9561 # 54 (080)add_uuid nid=... 0: 1:...
9562 # 55 (144)attach 0:lustre-MDT0001-osp-MDT0000 1:osp 2:...
9563 # 56 (144)setup 0:lustre-MDT0001-osp-MDT0000 1:... 2:...
9564 # 57 (136)modify_mdc_tgts add 0:lustre-MDT0000-mdtlov 1:... 2:1 3:1
9565 # duplicate modify_mdc_tgts caused crashes
9568 # using larger debug_mb size to avoid lctl dk log truncation
9569 size_mb=$((DEBUG_SIZE_SAVED * 4))
9572 # though config processing stops after failed attach and setup
9573 # it will proceed after the failed command after each writeconf
9574 # this is the original scenario of the issue
9575 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)" &>/dev/null
9576 do_facet mds1 "$LCTL set_param debug_mb=$size_mb"
9579 wait_update_facet_cond mds1 \
9580 "$LCTL dk | grep -c Processed.log.$FSNAME-MDT0000" \
9585 [[ -z "$had_config" ]] && do_facet mds1 lctl set_param debug=-config
9589 run_test 119 "writeconf on slave mdt shouldn't duplicate mdc/osp and crash"
9591 test_120() { # LU-11130
9592 [ "$MDSCOUNT" -lt 2 ] && skip "mdt count < 2"
9593 [ "$mds1_FSTYPE" != ldiskfs ] &&
9594 skip "ldiskfs only test"
9595 [ "$MDS1_VERSION" -lt $(version_code 2.11.56) ] &&
9596 skip "Need DNE2 capable MD target with LU-11130 fix"
9600 local mds1host=$(facet_active_host mds1)
9601 local mds1dev=$(mdsdevname 1)
9603 $LFS mkdir -i 1 $DIR/$tdir
9604 $LFS mkdir -i 0 $DIR/$tdir/mds1dir
9606 ln -s foo $DIR/$tdir/bar
9607 mv $DIR/$tdir/bar $DIR/$tdir/mds1dir/bar2 ||
9608 error "cross-target rename failed"
9612 run_e2fsck $mds1host $mds1dev "-n"
9614 run_test 120 "cross-target rename should not create bad symlinks"
9618 start_mgsmds || error "MGS MDS Start failed"
9620 stop_mds || error "Stopping MDSes failed"
9624 stop_mds || error "Stopping MDSes failed"
9626 run_test 121 "failover MGS"
9629 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
9630 [[ "$OST1_VERSION" -ge $(version_code 2.11.53) ]] ||
9631 skip "Need OST version at least 2.11.53"
9634 LOAD_MODULES_REMOTE=true load_modules
9635 #define OBD_FAIL_OFD_SET_OID 0x1e0
9636 do_facet ost1 $LCTL set_param fail_loc=0x00001e0
9639 $LFS mkdir -i1 -c1 $DIR/$tdir
9640 $LFS setstripe -i0 -c1 $DIR/$tdir
9641 do_facet ost1 $LCTL set_param fail_loc=0
9642 createmany -o $DIR/$tdir/file_ 1000 ||
9643 error "Fail to create a new sequence"
9647 run_test 122a "Check OST sequence update"
9650 remote_mgs_nodsh && skip "remote MGS with nodsh"
9651 [ -d $MOUNT/.lustre ] || setup
9653 # test old logid format until removal from llog_ioctl.c::str2logid()
9654 if (( $MGS_VERSION < $(version_code 2.99.53) )); then
9655 do_facet mgs $LCTL dl | grep MGS
9656 do_facet mgs "$LCTL --device %MGS llog_print \
9657 \\\\\\\$$FSNAME-client 1 10" ||
9658 error "old llog_print failed"
9661 # test new logid format
9662 if [ $MGS_VERSION -ge $(version_code 2.9.53) ]; then
9663 do_facet mgs "$LCTL --device MGS llog_print $FSNAME-client" ||
9664 error "new llog_print failed"
9667 run_test 123aa "llog_print works with FIDs and simple names"
9670 remote_mgs_nodsh && skip "remote MGS with nodsh"
9671 [[ $MGS_VERSION -gt $(version_code 2.11.51) ]] ||
9672 skip "Need server with working llog_print support"
9674 [ -d $MOUNT/.lustre ] || setup
9680 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9681 mgs_arg="--device MGS"
9683 orig_val=$(do_facet mgs $LCTL get_param jobid_name)
9684 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME"
9686 yaml=$(do_facet mgs $LCTL $mgs_arg llog_print params |
9687 grep jobid_name | tail -n 1)
9689 local param=$(awk '{ print $10 }' <<< "$yaml")
9690 local val=$(awk '{ print $12 }' <<< "$yaml")
9691 #return to the default
9692 do_facet mgs $LCTL set_param -P jobid_name=$orig_val
9693 [ $val = "TESTNAME" ] || error "bad value: $val"
9694 [ $param = "jobid_name," ] || error "Bad param: $param"
9696 run_test 123ab "llog_print params output values from set_param -P"
9698 test_123ac() { # LU-11566
9699 remote_mgs_nodsh && skip "remote MGS with nodsh"
9700 do_facet mgs "$LCTL help llog_print" 2>&1 | grep -q -- --start ||
9701 skip "Need 'lctl llog_print --start' on MGS"
9707 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9708 mgs_arg="--device MGS"
9710 [ -d $MOUNT/.lustre ] || setup
9712 # - { index: 10, event: add_uuid, nid: 192.168.20.1@tcp(0x20000c0a81401,
9713 # node: 192.168.20.1@tcp }
9714 do_facet mgs $LCTL $mgs_arg \
9715 llog_print --start $start --end $end $FSNAME-client | tr -d , |
9716 while read DASH BRACE INDEX idx EVENT BLAH BLAH BLAH; do
9717 (( idx >= start )) || error "llog_print index $idx < $start"
9718 (( idx <= end )) || error "llog_print index $idx > $end"
9721 run_test 123ac "llog_print with --start and --end"
9723 test_123ad() { # LU-11566
9724 remote_mgs_nodsh && skip "remote MGS with nodsh"
9725 # older versions of lctl may not print all records properly
9726 do_facet mgs "$LCTL help llog_print" 2>&1 | grep -q -- --start ||
9727 skip "Need 'lctl llog_print --start' on MGS"
9729 [ -d $MOUNT/.lustre ] || setup
9731 # append a new record, to avoid issues if last record was cancelled
9732 local old=$($LCTL get_param -n osc.*-OST0000-*.max_dirty_mb | head -1)
9733 do_facet mgs $LCTL conf_param $FSNAME-OST0000.osc.max_dirty_mb=$old
9735 # logid: [0x3:0xa:0x0]:0
9739 local num=$(do_facet mgs $LCTL --device MGS llog_info $FSNAME-client |
9740 awk '/last_index:/ { print $2 - 1 }')
9742 # - { index: 71, event: set_timeout, num: 0x14, param: sys.timeout=20 }
9743 local last=$(do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
9744 tail -1 | awk '{ print $4 }' | tr -d , )
9745 (( last == num )) || error "llog_print only showed $last/$num records"
9747 run_test 123ad "llog_print shows all records"
9749 test_123ae() { # LU-11566
9757 remote_mgs_nodsh && skip "remote MGS with nodsh"
9758 [ -d $MOUNT/.lustre ] || setupall
9760 max=$($LCTL get_param -n osc.*-OST0000-*.max_dirty_mb | head -1)
9761 pgs=$($LCTL get_param -n osc.*-OST0000-*.max_pages_per_rpc | head -1)
9762 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9763 mgs_arg="--device MGS"
9765 if do_facet mgs "$LCTL help llog_cancel" 2>&1| grep -q -- --log_id; then
9766 # save one set_param -P record in case none exist
9768 do_facet mgs $LCTL set_param -P osc.*.max_pages_per_rpc=$pgs
9769 stack_trap "do_facet mgs $LCTL set_param -P -d \
9770 osc.*.max_pages_per_rpc"
9773 orig=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9774 tail -1 | awk '{ print $4 }' | tr -d , )
9775 do_facet mgs $LCTL set_param -P osc.*.max_dirty_mb=$max
9776 do_facet mgs $LCTL $mgs_arg llog_print $log | tail -1 |
9777 grep "parameter: osc.*.max_dirty_mb" ||
9778 error "new set_param -P wasn't stored in params log"
9780 # - { index: 71, event: set_param, device: general,
9781 # param: osc.*.max_dirty_mb, value: 256 }
9782 id=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9783 tail -1 | awk '{ print $4 }' | tr -d , )
9784 do_facet mgs $LCTL $mgs_arg llog_cancel $log --log_idx=$id
9785 local new=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9786 tail -1 | awk '{ print $4 }' | tr -d , )
9787 (( new == orig )) ||
9788 error "new llog_cancel now $new, not at $orig records"
9791 # test old positional parameters for a while still
9792 if [ "$MGS_VERSION" -le $(version_code 3.1.53) ]; then
9795 do_facet mgs $LCTL conf_param \
9796 $FSNAME-OST0000.osc.max_pages_per_rpc=$pgs
9797 stack_trap "do_facet mgs $LCTL conf_param -d \
9798 $FSNAME-OST0000.osc.max_pages_per_rpc"
9800 orig=$(do_facet mgs $LCTL --device MGS llog_print $log |
9801 tail -1 | awk '{ print $4 }' | tr -d , )
9802 do_facet mgs $LCTL conf_param $FSNAME-OST0000.osc.max_dirty_mb=$max
9803 do_facet mgs $LCTL --device MGS llog_print $log |
9804 tail -1 | grep "parameter: osc.max_dirty_mb" ||
9805 error "old conf_param wasn't stored in params log"
9806 do_facet mgs $LCTL --device MGS llog_print $log
9807 # - { index: 71, event: conf_param, device: testfs-OST0000-osc,
9808 # param: osc.max_dirty_mb=256 }
9809 id=$(do_facet mgs $LCTL --device MGS llog_print $log |
9810 tail -1 | awk '{ print $4 }' | tr -d , )
9811 do_facet mgs $LCTL --device MGS llog_cancel $log $id
9812 do_facet mgs $LCTL --device MGS llog_print $log
9813 new=$(do_facet mgs $LCTL --device MGS llog_print $log |
9814 tail -1 | awk '{ print $4 }' | tr -d , )
9815 (( new == orig )) ||
9816 error "old llog_cancel now $new, not at $orig records"
9819 run_test 123ae "llog_cancel can cancel requested record"
9821 test_123af() { #LU-13609
9822 [ "$MGS_VERSION" -ge $(version_code 2.13.54) -a \
9823 "$MDS1_VERSION" -ge $(version_code 2.13.54) ] ||
9824 skip "Need both MGS and MDS version at least 2.13.54"
9826 [ -d $MOUNT/.lustre ] || setupall
9827 stack_trap "do_facet mds1 $LCTL set_param fail_loc=0" EXIT
9837 for device in "MGS" "$FSNAME-MDT0000"; do
9838 cmd="--device $device llog_catlist"
9839 echo "lctl $cmd ..."
9840 if [ "$device" = "MGS" ]; then
9845 orig_clist=($(do_facet $facet $LCTL $cmd | awk '{ print $2 }'))
9846 orig_count=${#orig_clist[@]}
9847 echo "orig_clist: ${orig_clist[*]}"
9849 #define OBD_FAIL_CATLIST 0x131b
9850 #fetch to llog records from the second one
9851 do_facet $facet $LCTL set_param fail_loc=0x131b fail_val=2
9853 new_clist=($(do_facet $facet $LCTL $cmd | awk '{ print $2 }'))
9854 new_count=${#new_clist[@]}
9855 echo "new_clist: ${new_clist[*]}"
9857 [ $new_count -eq $((orig_count - 1)) ] ||
9858 error "$new_count != $orig_count - 1"
9859 for i in $(seq 0 $new_count); do
9861 [ "${orig_clist[$j]}" = "${new_clist[$i]}" ] ||
9862 error "${orig_clist[$j]} != ${new_clist[$i]}"
9864 do_facet mds1 $LCTL set_param fail_loc=0
9868 run_test 123af "llog_catlist can show all config files correctly"
9870 test_123ag() { # LU-15142
9874 remote_mgs_nodsh && skip "remote MGS with nodsh"
9875 (( $MGS_VERSION >= $(version_code 2.14.55) )) ||
9876 skip "Need server version least 2.14.55"
9878 [ -d $MOUNT/.lustre ] || setup
9880 orig_val=$(do_facet mgs $LCTL get_param jobid_name)
9881 stack_trap "do_facet mgs $LCTL set_param -P jobid_name=$orig_val"
9883 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME1"
9884 do_facet mgs $LCTL set_param -P -d jobid_name
9885 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
9887 (( rec == 0 )) || error "parameter was not deleted, check #1"
9888 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME1"
9889 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
9891 (( rec == 1)) || error "parameter is not set"
9892 # usage with ordinary set_param format works too
9893 do_facet mgs $LCTL set_param -P -d jobid_name="ANY"
9894 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
9896 (( rec == 0 )) || error "parameter was not deleted, check #2"
9898 run_test 123ag "llog_print skips values deleted by set_param -P -d"
9900 test_123ah() { #LU-7668 del_ost
9901 (( "$MGS_VERSION" >= $(version_code 2.15.51) &&
9902 "$MDS1_VERSION" >= $(version_code 2.15.51) )) ||
9903 skip "Need MGS/MDS version >= 2.15.51 for del_ost"
9905 [ -d $MOUNT/.lustre ] || setupall
9906 stack_trap "do_facet mds1 $LCTL set_param fail_loc=0" EXIT
9908 local cmd="--device MGS llog_print"
9910 cli_llogcnt_orig=$(do_facet mgs $LCTL $cmd $FSNAME-client |
9911 grep -c $FSNAME-OST0000)
9912 mds1_llogcnt_orig=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 |
9913 grep -c $FSNAME-OST0000)
9915 [ $cli_llogcnt_orig -gt 0 ] ||
9916 error "$FSNAME-OST0000 not found (client)"
9917 [ $mds1_llogcnt_orig -gt 0 ] || error "$FSNAME-OST0000 not found (MDT)"
9919 # -n/--dryrun should NOT modify catalog
9920 do_facet mgs "$LCTL del_ost -n -t $FSNAME-OST0000" ||
9921 error "del_ost --dryrun failed with $?"
9923 local cli_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-client |
9924 grep -c $FSNAME-OST0000)
9925 local mds1_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 |
9926 grep -c $FSNAME-OST0000)
9928 [ $cli_llogcnt -eq $cli_llogcnt_orig ] ||
9929 error "--dryrun error: $cli_llogcnt != $cli_llogcnt_orig"
9930 [ $mds1_llogcnt -eq $mds1_llogcnt_orig ] ||
9931 error "--dryrun error: $mds1_llogcnt != $mds1_llogcnt_orig"
9934 do_facet mgs "$LCTL del_ost --target $FSNAME-OST0000" ||
9935 error "del_ost failed with $?"
9937 local cli_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-client |
9938 grep -c $FSNAME-OST0000)
9939 local mds1_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 |
9940 grep -c $FSNAME-OST0000)
9942 # every catalog entry for OST0000 should have been cancelled
9943 [ $cli_llogcnt -eq 0 ] || error "$FSNAME-OST0000 not cancelled (cli)"
9944 [ $mds1_llogcnt -eq 0 ] || error "$FSNAME-OST0000 not cancelled (MDT)"
9946 umount_client $MOUNT
9949 $LFS df | grep -q OST0000 && error "del_ost did not remove OST0000!"
9953 run_test 123ah "del_ost cancels config log entries correctly"
9958 echo "cleanup test 123ai"
9960 # cancel last timeout record
9961 do_facet mgs "$LCTL set_param -P -d timeout"
9963 # restore timeout value
9964 do_nodes $(comma_list $(all_nodes)) "$LCTL set_param timeout=$timeout"
9967 test_123ai() { #LU-16167
9973 remote_mgs_nodsh && skip "remote MGS with nodsh"
9974 (( MDS1_VERSION >= $(version_code 2.15.51) )) ||
9975 skip "Need MDS version at least 2.15.51"
9977 [ -d $MOUNT/.lustre ] || setup
9979 old_count=$(do_facet mgs "$LCTL --device MGS llog_print -r params" |
9980 grep -c "parameter: timeout")
9982 old_timeout=$($LCTL get_param -n timeout)
9983 stack_trap "cleanup_123ai $old_timeout" EXIT
9985 # add and cancel params (2 * MAX_IOC_BUFLEN / 128 + 1 = 129)
9986 for i in {1..129}; do
9987 do_facet mgs "$LCTL set_param -P timeout=$i" ||
9988 error "fail to set timeout=$i on MGS"
9991 count=$(do_facet mgs "$LCTL --device MGS llog_print -r params" |
9992 grep -c "parameter: timeout")
9994 ((count - old_count == 129)) ||
9995 error "number timeout record missmatch ($count - $old_count != 129)"
9997 do_facet mgs "$LCTL --device MGS llog_print params" |
9998 tail -1 | grep "timeout, value: 129" ||
9999 error "llog_print could not display the last record (timeout=129)"
10002 run_test 123ai "llog_print display all non skipped records"
10005 remote_mgs_nodsh && skip "remote MGS with nodsh"
10007 [ -d $MOUNT/.lustre ] || setup
10008 yaml_file="$TMP/$tfile.yaml"
10010 do_facet mgs rm "$yaml_file"
10011 cfgfiles=$(do_facet mgs "lctl --device MGS llog_catlist" |
10012 sed 's/config_log://')
10014 # set jobid_var to a different value for test
10015 orig_val=$(do_facet mgs $LCTL get_param jobid_var)
10017 do_facet mgs $LCTL set_param -P jobid_var="TESTNAME"
10019 for i in $cfgfiles params; do
10020 do_facet mgs "lctl --device MGS llog_print ${i} >> $yaml_file"
10023 echo "Unmounting FS"
10031 test_123_restore() {
10032 set_val=$(do_facet mgs $LCTL get_param jobid_var)
10034 do_facet mgs $LCTL set_param -P $orig_val
10036 [ $set_val == "jobid_var=TESTNAME" ] ||
10037 error "$set_val is not TESTNAME"
10039 do_facet mgs rm "$yaml_file"
10051 # Reapply the config from before
10052 echo "Setting configuration parameters"
10053 do_facet mgs "lctl set_param -F $yaml_file"
10057 run_test 123F "clear and reset all parameters using set_param -F"
10067 # Reapply the config from before
10068 echo "Setting configuration parameters"
10069 do_facet mgs "lctl apply_yaml $yaml_file"
10073 run_test 123G "clear and reset all parameters using apply_yaml"
10078 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
10079 [ -z $mds2failover_HOST ] && skip "needs MDT failover setup"
10082 do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
10083 grep 1.2.3.4@tcp && error "Should not be fake nid"
10084 do_facet mgs $LCTL conf_param $FSNAME-MDT0001.failover.node=1.2.3.4@tcp\
10085 || error "Set params error"
10086 do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
10087 grep 1.2.3.4@tcp || error "Fake nid should be added"
10091 if combined_mgs_mds; then
10092 start_mdt 1 "-o nosvc" ||
10093 error "starting mds with nosvc option failed"
10095 local nid=$(do_facet mds2 $LCTL list_nids | head -1)
10096 local failover_nid=$(do_node $mds2failover_HOST $LCTL list_nids | head -1)
10098 do_facet mgs $LCTL replace_nids $FSNAME-MDT0001 $nid:$failover_nid ||
10099 error "replace_nids execution error"
10101 if combined_mgs_mds; then
10107 echo "lfs setdirstripe"
10108 $LFS setdirstripe -i 1 $MOUNT/$tdir || error "setdirstirpe error"
10111 run_test 124 "check failover after replace_nids"
10113 get_max_sectors_kb() {
10116 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
10117 local max_sectors_path="/sys/block/${dev_base}/queue/max_sectors_kb"
10119 do_facet ${facet} "[[ -e ${max_sectors_path} ]] &&
10120 cat ${max_sectors_path}"
10123 get_max_hw_sectors_kb() {
10126 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
10127 local max_hw_path="/sys/block/${dev_base}/queue/max_hw_sectors_kb"
10129 do_facet ${facet} "[[ -e ${max_hw_path} ]] && cat ${max_hw_path}"
10132 set_max_sectors_kb() {
10136 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
10137 local max_sectors_path="/sys/block/${dev_base}/queue/max_sectors_kb"
10139 do_facet ${facet} "[[ -e ${max_sectors_path} ]] &&
10140 echo ${value} > ${max_sectors_path}"
10143 [[ $rc -ne 0 ]] && echo "Failed to set ${max_sectors_path} to ${value}"
10148 # Return 0 if all slave devices have max_sectors_kb == max_hw_sectors_kb
10149 # Otherwise return > 0
10150 check_slaves_max_sectors_kb()
10154 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
10155 local slaves_dir=/sys/block/${dev_base}/slaves
10156 local slave_devices=$(do_facet ${facet} "ls ${slaves_dir} 2>/dev/null")
10157 [[ -z ${slave_devices} ]] && return 0
10159 local slave max_sectors new_max_sectors max_hw_sectors path
10162 for slave in ${slave_devices}; do
10163 path="/dev/${slave}"
10164 ! is_blkdev ${facet} ${path} && continue
10165 max_sectors=$(get_max_sectors_kb ${facet} ${path})
10166 max_hw_sectors=$(get_max_hw_sectors_kb ${facet} ${path})
10167 new_max_sectors=${max_hw_sectors}
10168 [[ ${new_max_sectors} -gt ${RQ_SIZE_LIMIT} ]] &&
10169 new_max_sectors=${RQ_SIZE_LIMIT}
10171 if [[ ${max_sectors} -ne ${new_max_sectors} ]]; then
10172 echo "${path} ${max_sectors} ${new_max_sectors}"
10175 check_slaves_max_sectors_kb ${facet} ${path}
10184 local facet_list="mgs mds1 ost1"
10185 combined_mgs_mds && facet_list="mgs ost1"
10188 for facet in ${facet_list}; do
10189 [[ $(facet_fstype ${facet}) != ldiskfs ]] &&
10190 skip "ldiskfs only test" &&
10192 ! is_blkdev ${facet} $(facet_device ${facet}) &&
10193 skip "requires all real devices" &&
10198 # We don't increase IO request size limit past 16MB. See comments in
10199 # lustre/utils/libmount_utils_ldiskfs.c:tune_max_sectors_kb()
10200 RQ_SIZE_LIMIT=$((16 * 1024))
10201 local device old_max_sectors new_max_sectors max_hw_sectors
10202 for facet in ${facet_list}; do
10203 device=$(facet_device ${facet})
10204 old_max_sectors=$(get_max_sectors_kb ${facet} ${device})
10205 max_hw_sectors=$(get_max_hw_sectors_kb ${facet} ${device})
10207 # The expected value after l_tunedisk is executed
10208 new_max_sectors=$old_max_sectors
10209 [[ ${new_max_sectors_kb} -gt ${RQ_SIZE_LIMIT} ]] &&
10210 new_max_sectors_kb=${RQ_SIZE_LIMIT}
10212 # Ensure the current value of max_sectors_kb does not equal
10213 # max_hw_sectors_kb, so we can tell whether l_tunedisk did
10215 set_max_sectors_kb ${facet} ${device} $((new_max_sectors - 1))
10217 # Value before l_tunedisk
10218 local pre_max_sectors=$(get_max_sectors_kb ${facet} ${device})
10219 if [[ ${pre_max_sectors} -ne $((new_max_sectors - 1)) ]]; then
10220 echo "unable to satsify test pre-condition:"
10221 echo "${pre_max_sectors} != $((new_max_sectors - 1))"
10226 echo "Before: ${facet} ${device} ${pre_max_sectors} ${max_hw_sectors}"
10228 do_facet ${facet} "l_tunedisk ${device}"
10230 # Value after l_tunedisk
10231 local post_max_sectors=$(get_max_sectors_kb ${facet} ${device})
10233 echo "After: ${facet} ${device} ${post_max_sectors} ${max_hw_sectors}"
10235 if [[ ${facet} != ost1 ]]; then
10236 if [[ ${post_max_sectors} -ne ${pre_max_sectors} ]]; then
10237 echo "l_tunedisk modified max_sectors_kb of ${facet}"
10241 set_max_sectors_kb ${facet} ${device} ${old_max_sectors}
10243 if [[ ${post_max_sectors} -eq ${pre_max_sectors} ]]; then
10244 echo "l_tunedisk failed to modify max_sectors_kb of ${facet}"
10248 check_slaves_max_sectors_kb ${facet} ${device} ||
10255 run_test 125 "check l_tunedisk only tunes OSTs and their slave devices"
10258 [[ "$MDS1_VERSION" -ge $(version_code 2.13.52) ]] ||
10259 skip "Need MDS version at least 2.13.52"
10262 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ../libcfs/libcfs/libcfs
10263 #define OBD_FAIL_OBD_SETUP 0x60d
10264 do_facet mds1 $LCTL set_param fail_loc=0x60d
10265 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_modules &
10266 for i in {1..40}; do
10267 do_facet mds1 lsmod | grep -q osd_$mds1_FSTYPE && break
10270 clear_failloc $SINGLEMDS 20 &
10271 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
10273 run_test 126 "mount in parallel shouldn't cause a crash"
10276 [[ "$ost1_FSTYPE" == ldiskfs ]] || skip "ldiskfs only test"
10280 zconf_umount_clients $RCLIENTS $MOUNT
10282 wait_osp_active ost ${FSNAME}-OST0000 0 1
10283 local osc_tgt="$FSNAME-OST0000-osc-$($LFS getname -i $DIR)"
10284 local avail1=($($LCTL get_param -n osc.${osc_tgt}.kbytesavail))
10286 $LFS setstripe -i 0 $DIR/$tfile || error "failed creating $DIR/$tfile"
10287 dd if=/dev/zero of=$DIR/$tfile bs=1M oflag=direct || true
10289 local avail2=($($LCTL get_param -n osc.${osc_tgt}.kbytesavail))
10291 if ((avail2 * 100 / avail1 > 1)); then
10294 error "more than 1% space left: before=$avail1 after=$avail2"
10297 local mbs=$(($(stat -c %s $DIR/$tfile) / (1024 * 1024)))
10299 dd if=/dev/zero of=$DIR/$tfile bs=1M count=$mbs conv=notrunc \
10300 oflag=direct || error "overwrite failed"
10302 run_test 127 "direct io overwrite on full ost"
10306 combined_mgs_mds && skip "need separate mgs device"
10307 [ "$ost2_FSTYPE" == zfs ] && import_zpool ost2
10310 # Try to apply nolocallogs to the virgin OST. Should fail.
10311 do_facet ost2 "$TUNEFS --nolocallogs $(ostdevname 2)" &&
10312 error "nolocallogs should not be allowed on the virgin target"
10317 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
10318 # Start OST without MGS (local configs)
10319 do_facet ost1 "$TUNEFS --dryrun $(ostdevname 1)"
10320 start_ost || error "unable to start OST1"
10321 stop_ost || error "Unable to stop OST1"
10323 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
10324 # Do not allow reading local configs, should fail
10325 do_facet ost1 "$TUNEFS --nolocallogs $(ostdevname 1)" ||
10326 error "Can not set nolocallogs"
10327 start_ost && error "OST1 started, but should fail"
10329 # Connect to MGS successfully, reset nolocallogs flag
10330 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
10331 start_mgs || error "unable to start MGS"
10332 start_ost || error "unable to start OST1"
10334 do_facet ost1 "$TUNEFS --dryrun $(ostdevname 1)" | grep "nolocallogs" &&
10335 error "nolocallogs expected to be reset"
10337 stop_ost || error "Unable to stop OST1"
10339 run_test 128 "Force using remote logs with --nolocallogs"
10343 (( MDS1_VERSION >= $(version_code 2.14.57) )) ||
10344 skip "Need MDS version at least 2.14.57"
10346 start_mgsmds || error "MGS/MDS start failed"
10348 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &&
10349 error "start ost1 should fail" || true
10350 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &&
10351 error "second start ost1 should fail" || true
10352 do_facet ost1 "$TUNEFS --writeconf $(ostdevname 1)"
10353 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS ||
10354 error "start ost1 failed"
10358 run_test 129 "attempt to connect an OST with the same index should fail"
10362 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
10364 stop_mdt 2 || error "mdt2 stop failed"
10365 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)"
10366 start_mdt 2 || error "mdt2 start failed"
10367 do_facet mds2 "$LCTL dl" | grep MDT0001-osp-MDT0001 &&
10368 error "Illegal OSP device created" || true
10370 run_test 130 "re-register an MDT after writeconf"
10373 [ "$mds1_FSTYPE" == "ldiskfs" ] || skip "ldiskfs only test"
10374 (( $MDS1_VERSION >= $(version_code 2.14.56.35) )) ||
10375 skip "Need MDS version at least 2.14.56.35"
10376 do_facet mds1 $DEBUGFS -R features $(mdsdevname 1) |
10377 grep -q project || skip "skip project quota not supported"
10382 test_mkdir -c $MDSCOUNT -p $DIR/$tdir
10383 $LFS project -p 1000 $DIR/$tdir || error "set dir project id failed"
10384 createmany -o $DIR/$tdir/f 512
10385 for ((i = 0; i < 512; ++i)); do
10386 $LFS project -p $i $DIR/$tdir/f${i} ||
10387 error "set f${i} project id failed"
10390 test_mkdir -c $MDSCOUNT -p $DIR/$tdir.inherit
10391 $LFS project -p 1001 -s $DIR/$tdir.inherit
10392 createmany -o $DIR/$tdir.inherit/f 128
10393 (( $($LFS project $DIR/$tdir.inherit/f* |
10394 awk '$1 == 1001 { print }' | wc -l) == 128 )) ||
10395 error "files did not inherit projid 1001"
10399 for i in $(seq $MDSCOUNT); do
10400 mds_backup_restore mds$i ||
10401 error "Backup/restore on mds$i failed"
10406 projid=($($LFS project -d $DIR/$tdir))
10407 [ ${projid[0]} == "1000" ] ||
10408 error "projid expected 1000 not ${projid[0]}"
10409 for ((i = 0; i < 512; ++i)); do
10410 projid=($($LFS project $DIR/$tdir/f${i}))
10411 [ ${projid[0]} == "$i" ] ||
10412 error "projid expected $i not ${projid[0]}"
10415 (( $($LFS project $DIR/$tdir.inherit/f* |
10416 awk '$1 == 1001 { print }' | wc -l) == 128 )) ||
10417 error "restore did not copy projid 1001"
10419 run_test 131 "MDT backup restore with project ID"
10425 (( MDS1_VERSION >= $(version_code 2.14.57) )) ||
10426 skip "Need MDS version at least 2.14.57"
10428 combined_mgs_mds || start_mgs || error "unable to start MGS"
10429 start_mdt 1 || error "unable to start mdt1"
10431 err_cnt=$(do_facet mds1 dmesg | grep -c "cannot take the layout locks")
10432 stop_mdt 1 || error "stop mdt1 failed"
10434 [ "$mds1_FSTYPE" == zfs ] && import_zpool mds1
10435 do_facet mds1 $TUNEFS --param mdt.hsm_control=enabled $(mdsdevname 1) ||
10436 error "tunefs failed"
10437 start_mdt 1 || error "cannot start mdt1"
10439 err_cnt2=$(do_facet mds1 dmesg | grep -c "cannot take the layout locks")
10440 [ $err_cnt -eq $err_cnt2 ] || error "Can not take the layout lock"
10441 stop_mdt 1 || error "stop mdt1 failed"
10443 run_test 132 "hsm_actions processed after failover"
10445 # This test verifies we do RR allocation within a pool even if there is a
10446 # significant imbalance vs an OST outside the pool
10448 [[ $OSTCOUNT -lt 4 ]] && skip_env "needs >= 4 OSTs"
10449 [[ "$OST1_VERSION" -ge $(version_code 2.15.51) ]] ||
10450 skip "Need OST version at least 2.15.51"
10451 # This is the easiest way to ensure OSTs start out balanced
10452 reformat_and_config
10455 check_set_fallocate_or_skip
10457 local testfile=$DIR/$tdir/$tfile
10458 local pool="testpool"
10459 local ostrange=$((OSTCOUNT - 1))
10460 # Select all but the last OST to add to the pool
10461 local poolostrange=$((OSTCOUNT - 2))
10469 [ $MINV -eq 0 ] && error "no free space in OST$MINI"
10470 [ $MAXV -gt $((2 * $MINV)) ] &&
10471 error "OSTs badly unbalanced after reformat"
10473 create_pool $FSNAME.$pool || error "failed to create a pool"
10474 do_facet mgs $LCTL pool_add $FSNAME.$pool OST[0-$poolostrange] ||
10475 error "failed to add OST[0-$poolostrange] to the pool"
10477 test_mkdir -p $DIR/$tdir || error "failed to mkdir $DIR/$tdir"
10478 # Consume space on the OSTs in the pool so they are unbalanced with the
10479 # OST outside of the pool
10480 # fill each OST 90% with fallocate so they are widely
10482 local size=$(((MINV * 9 / 10) * 1024))
10483 for ((i = 0; i <= poolostrange; i++)); do
10484 $LFS setstripe -c 1 -i $i $testfile$i ||
10485 error "failed to setstripe $testfile$i"
10486 fallocate -l $size $testfile$i || error "fallocate failed"
10492 # Create files in the pool now that there is an imbalance
10493 filecount=$(((OSTCOUNT - 1) * filenum))
10494 for ((i = 0; i < filecount; i++)); do
10495 $LFS setstripe -p $pool $testfile-$i ||
10496 error "failed to setstripe -p $pool $testfile-$i"
10498 $LFS getstripe -i $testfile-* > /tmp/$tfile.log
10499 # Count the number of files with a stripe on each OST to verify the
10500 # pool allocated with round-robin
10501 for ((i = 0; i <= poolostrange; i++)); do
10502 stripecount=$(grep -c $i /tmp/$tfile.log)
10503 # Allow a little leeway
10504 if (( stripecount < filenum - 1 ||
10505 stripecount > filenum + 1 )); then
10506 cat /tmp/$tfile.log
10507 error "$stripecount != $filenum files on OST$i"
10511 # Create files across the system now that there is an imbalance
10512 filecount=$((OSTCOUNT * filenum))
10513 for ((i = 1; i < filecount; i++)); do
10514 $LFS setstripe $testfile-$i.2 ||
10515 error "failed to setstripe $testilfe-$i.2"
10517 $LFS getstripe -i $testfile-*.2 > /tmp/$tfile.log
10519 # Count the number of files with a stripe on each OST to verify the
10520 # files are *NOT* allocated with round-robin
10521 for ((i = 0; i <= ostrange; i++)); do
10522 stripecount=$(grep -c $i /tmp/$tfile.log)
10523 if [[ $stripecount -ne $filenum ]]; then
10525 echo "QOS: $stripecount != $filenum files on OST$i"
10528 if [ -z "$qos_used" ]; then
10529 error "QOS not used on imbalanced OSTs!"
10532 rm -rf /tmp/$tfile.log $DIR/$tdir
10533 do_facet mgs $LCTL pool_remove $FSNAME.$pool OST[0-$poolostrange] ||
10534 "failed to remove OST[0-$poolostrange] from the pool"
10535 do_facet mgs $LCTL pool_destroy $FSNAME.$pool ||
10536 error "failed to destroy pool"
10538 run_test 133 "stripe QOS: free space balance in a pool"
10541 [ "$mds1_FSTYPE" == "ldiskfs" ] || skip "ldiskfs only test"
10544 local mdt_dev=$(facet_device mds1)
10545 local tmp_dir=$TMP/$tdir
10546 local out=$tmp_dir/check_iam.txt
10547 local CHECK_IAM=${CHECK_IAM:-$(do_facet mds1 "which check_iam 2> /dev/null || true")}
10549 [[ -n "$CHECK_IAM" ]] || skip "check_iam not found"
10552 for ((i=0; i<64; i++)); do
10554 #cmd introduce a random corruption to IAM file
10555 local cmd="dd if=/dev/urandom of=$tmp_dir/$f bs=2 conv=notrunc count=1 seek=$((RANDOM % 36))"
10556 do_facet mds1 "mkdir -p $tmp_dir; \
10557 $DEBUGFS -c -R 'dump $f $tmp_dir/$f' $mdt_dev 2>&1; \
10558 $CHECK_IAM -v $tmp_dir/$f 2>&1; \
10559 echo $cmd; eval $cmd 2>/dev/null;
10560 $CHECK_IAM -v $tmp_dir/$f 2>&1; echo \\\$?" >> $out 2>&1
10565 stack_trap "rm -rf $tmp_dir && do_facet mds1 rm -rf $tmp_dir" EXIT
10567 rc=$(grep -c "fault\|except" $out)
10568 (( rc == 0 )) || { cat $out &&
10569 error "check_iam failed with fault or exception $rc"; }
10571 rc=$(grep -c "^255" $out)
10572 errors=$(grep -c "FINISHED WITH ERRORS" $out)
10574 (( rc == errors )) || { cat $out &&
10575 error "check_iam errcode does not fit with errors $rc $errors"; }
10577 run_test 134 "check_iam works without faults"
10579 cleanup_test_135(){
10582 printf "\nCleanup test_135\n" >&2
10583 do_facet mds1 "$LCTL set_param -n $oldgc"
10584 rm -rf $DIR/$tdir &> /dev/null
10588 __test_135_file_thread() {
10590 local init_time=$(awk '{print $1}' /proc/uptime)
10591 local awkcmd="/crosses index zero/ {if (\$1 > $init_time) exit(1);}"
10593 #Generate a full plain llogs
10594 while dmesg | sed -r 's/(\[|\])//g' | awk "$awkcmd" ; do
10595 createmany -o $DIR/$tdir/f 4500 >&2
10596 createmany -u $DIR/$tdir/f 4500 >&2
10600 __test_135_reader() {
10603 local firstidx=$(changelog_user_rec mds1 $cl_user)
10604 local oldidx=$firstidx
10609 while read -t10 -u$fd newidx other; do
10610 (( (newidx - oldidx) == 1 )) ||
10611 error "changelog jump detected (last: $oldidx, current: $newidx)"
10613 if (( (newidx - firstidx + 1) % 13000 == 0 )); then
10614 [[ $pid -eq 0 ]] ||
10615 wait $pid || error "changelog_clear failed"
10616 changelog_clear $((newidx - 1)) mds1 >&2 & pid=$!
10621 [[ $pid -eq 0 ]] ||
10622 wait $pid || error "changelog_clear failed"
10628 (( MDS1_VERSION >= $(version_code 2.15.52) )) ||
10629 skip "need MDS version at least 2.15.52"
10631 local service=$(facet_svc mds1)
10633 local lastread lastidx
10634 local files_pid reader_pid
10639 # Need to reformat because we are changing llog catalog sizes to 5.
10640 # Otherwise, processing could fail with existing catalogs (last_idx>5).
10644 # Disable changelog garbage colector
10645 local oldgc=$(do_facet mds1 "$LCTL get_param mdd.${service}.changelog_gc")
10646 do_facet mds1 "$LCTL set_param -n mdd.${service}.changelog_gc=0"
10647 stack_trap "cleanup_test_135 $oldgc" EXIT INT
10649 # change the changelog_catalog size to 5 entries for everybody
10650 #define OBD_FAIL_CAT_RECORDS 0x1312
10651 do_node $(comma_list $(all_nodes)) $LCTL set_param fail_loc=0x1312 fail_val=5
10653 # disable console ratelimit
10654 local rl=$(cat /sys/module/libcfs/parameters/libcfs_console_ratelimit)
10655 echo 0 > /sys/module/libcfs/parameters/libcfs_console_ratelimit
10656 stack_trap "echo $rl > /sys/module/libcfs/parameters/libcfs_console_ratelimit" EXIT
10658 test_mkdir -c 1 -i 0 $DIR/$tdir || error "Failed to create directory"
10659 changelog_chmask "ALL" || error "changelog_chmask failed"
10660 changelog_register || error "changelog_register failed"
10662 cl_user="${CL_USERS[mds1]%% *}"
10663 changelog_users mds1 | grep -q $cl_user ||
10664 error "User $cl_user not found in changelog_users"
10666 # Start reader thread
10667 coproc $LFS changelog --follow $service
10670 stack_trap "echo kill changelog reader; kill $reader_pid" EXIT
10672 echo -e "\nWrap arround changelog catalog"
10674 # Start file writer thread
10675 __test_135_file_thread "$service" & files_pid=$!
10676 stack_trap "(pkill -P$files_pid; kill $files_pid) &> /dev/null || true" EXIT
10678 # Check changelog entries
10679 lastread=$(__test_135_reader $fd $cl_user) || exit $?
10680 ! kill -0 $files_pid 2>/dev/null ||
10681 error "creation thread is running. Is changelog reader stuck?"
10683 lastidx=$(changelog_users mds1 | awk '/current_index/ {print $NF}' )
10684 [[ "$lastread" -eq "$lastidx" ]] ||
10685 error "invalid changelog lastidx (read: $lastread, mds: $lastidx)"
10687 run_test 135 "check the behavior when changelog is wrapped around"
10690 (( MDS1_VERSION >= $(version_code 2.15.55) )) ||
10691 skip "need MDS version at least 2.15.55"
10692 (( MDSCOUNT >= 2 )) || skip "needs >= 2 MDTs"
10694 RM_UPDATELOG=$(do_facet mds2 "which remove_updatelog 2> /dev/null")
10695 RM_UPDATELOG=${RM_UPDATELOG:-"$LUSTRE/scripts/remove_updatelog"}
10697 [ -f "$RM_UPDATELOG" ] ||
10698 skip_env "remove_updatelog is not found on mds2"
10700 local mntpt=$(facet_mntpt mds2)
10703 test_mkdir -c $MDSCOUNT -p $DIR/$tdir ||
10704 error "mkdir $DIR/$tdir failed"
10705 mkdir $DIR/$tdir/d{1..256}
10710 mount_fstype mds2 || error "mount as fstype mds2 failed"
10711 do_facet mds2 $RM_UPDATELOG -n $mntpt
10712 MDTS=($(do_facet mds2 $RM_UPDATELOG -n $mntpt |
10713 grep -o "Processing MDT[0-9]*" | awk -F'MDT' '{print $2}'))
10714 (( ${#MDTS[@]} == MDSCOUNT )) ||
10715 error "Processed ${#MDTS[@]} from $MDSCOUNT"
10717 do_facet mds2 $RM_UPDATELOG -n -m 1,0 $mntpt
10718 MDTS=($(do_facet mds2 $RM_UPDATELOG -n -m 1,0 $mntpt |
10719 grep -o "Processing MDT[0-9]*" | awk -F'MDT' '{print $2}'))
10720 (( ${#MDTS[@]} == 2 )) ||
10721 error "Processed ${#MDTS[@]} instead of 2"
10722 (( ${MDTS[0]} == 1 && ${MDTS[1]} == 0 )) ||
10723 error "Processed: ${MDTS[*]}, expected: 1 0"
10725 do_facet mds2 $RM_UPDATELOG -m 0 $mntpt
10726 unmount_fstype mds2
10727 start_mdt 2 || error "mds2 start fail"
10728 start_mdt 1 || error "mds1 start fail"
10729 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds1 FULL
10730 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
10732 rm -rf $DIR/$tdir || error "Can't remove $tdir"
10734 reformat_and_config
10736 run_test 140 "remove_updatelog script actions"
10741 local max_cached_mb=$($LCTL get_param llite.*.max_cached_mb |
10742 awk '/^max_cached_mb/ { print $2 }')
10743 stack_trap "$LCTL set_param -n llite.*.max_cached_mb=$max_cached_mb"
10745 $LCTL set_param llite.*.max_cached_mb='100%'
10747 local new_max_cached_mb=$($LCTL get_param llite.*.max_cached_mb |
10748 awk '/^max_cached_mb/ { print $2 }')
10749 local total_ram_mb=$(free -m | grep 'Mem:' | awk '{print $2}')
10751 $LCTL get_param llite.*.max_cached_mb
10752 echo "total ram mb: $total_ram_mb"
10753 (( new_max_cached_mb == total_ram_mb )) ||
10754 error "setting cache to 100% not equal to total RAM"
10756 $LCTL set_param llite.*.max_cached_mb='50%'
10757 new_max_cached_mb=$($LCTL get_param llite.*.max_cached_mb |
10758 awk '/^max_cached_mb/ { print $2 }')
10760 $LCTL get_param llite.*.max_cached_mb
10761 (( new_max_cached_mb == $((total_ram_mb / 2)) )) ||
10762 error "setting cache to 50% not equal to 50% of RAM"
10764 $LCTL set_param llite.*.max_cached_mb='105%' &&
10765 error "should not be able to set insane value"
10767 $LCTL set_param llite.*.max_cached_mb='0%'
10768 new_max_cached_mb=$($LCTL get_param llite.*.max_cached_mb |
10769 awk '/^max_cached_mb/ { print $2 }')
10770 # Minimum cache size is 64 MiB
10771 $LCTL get_param llite.*.max_cached_mb
10772 (( new_max_cached_mb == 64 )) ||
10773 error "setting cache to 0% != minimum cache size"
10775 run_test 150 "test setting max_cached_mb to a %"
10778 # (This was sanity/802a)
10780 saved_MGS_MOUNT_OPTS=$MGS_MOUNT_OPTS
10781 saved_MDS_MOUNT_OPTS=$MDS_MOUNT_OPTS
10782 saved_OST_MOUNT_OPTS=$OST_MOUNT_OPTS
10783 saved_MOUNT_OPTS=$MOUNT_OPTS
10787 MGS_MOUNT_OPTS=$saved_MGS_MOUNT_OPTS
10788 MDS_MOUNT_OPTS=$saved_MDS_MOUNT_OPTS
10789 OST_MOUNT_OPTS=$saved_OST_MOUNT_OPTS
10790 MOUNT_OPTS=$saved_MOUNT_OPTS
10795 [[ $mds1_FSTYPE = zfs ]] || skip "ZFS specific test"
10796 [[ $MDS1_VERSION -lt $(version_code 2.9.55) ]] ||
10797 [[ $OST1_VERSION -lt $(version_code 2.9.55) ]] &&
10798 skip "Need server version at least 2.9.55"
10800 [[ $ENABLE_QUOTA ]] && skip "Quota enabled for read-only test"
10802 # Reset before starting
10806 mkdir $DIR/$tdir || error "(1) fail to mkdir"
10808 cp $LUSTRE/tests/test-framework.sh $DIR/$tdir/ ||
10809 error "(2) Fail to copy"
10811 stack_trap cleanup_802a EXIT
10813 # sync by force before remount as readonly
10814 sync; sync_all_data; sleep 3; sync_all_data
10818 MGS_MOUNT_OPTS=$(csa_add "$MGS_MOUNT_OPTS" -o rdonly_dev)
10819 MDS_MOUNT_OPTS=$(csa_add "$MDS_MOUNT_OPTS" -o rdonly_dev)
10820 OST_MOUNT_OPTS=$(csa_add "$OST_MOUNT_OPTS" -o rdonly_dev)
10822 echo "Mount the server as read only"
10823 setupall server_only || error "(3) Fail to start servers"
10825 echo "Mount client without ro should fail"
10826 mount_client $MOUNT &&
10827 error "(4) Mount client without 'ro' should fail"
10829 echo "Mount client with ro should succeed"
10830 MOUNT_OPTS=$(csa_add "$MOUNT_OPTS" -o ro)
10831 mount_client $MOUNT ||
10832 error "(5) Mount client with 'ro' should succeed"
10834 echo "Modify should be refused"
10835 touch $DIR/$tdir/guard && error "(6) Touch should fail under ro mode"
10837 echo "Read should be allowed"
10838 diff $LUSTRE/tests/test-framework.sh $DIR/$tdir/test-framework.sh ||
10839 error "(7) Read should succeed under ro mode"
10841 run_test 802a "simulate readonly device"
10843 if ! combined_mgs_mds ; then
10849 # restore the values of MDSSIZE and OSTSIZE
10850 MDSSIZE=$STORED_MDSSIZE
10851 OSTSIZE=$STORED_OSTSIZE
10854 complete_test $SECONDS
10855 check_and_cleanup_lustre