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 always_except LU-11915 110
18 always_except LU-17269 41c
21 always_except LU-9795 84 86 103
24 if ! combined_mgs_mds; then
25 always_except LU-11991 32a 32b 32c 32d 32e
26 always_except LU-9897 84
27 always_except LU-12032 123F
31 [ "$SLOW" = "no" ] && EXCEPT_SLOW="45 69 106 111 114"
35 # use small MDS + OST size to speed formatting time
36 # do not use too small MDSSIZE/OSTSIZE, which affect the default journal size
37 # STORED_MDSSIZE is used in test_18
38 STORED_MDSSIZE=$MDSSIZE
39 STORED_OSTSIZE=$OSTSIZE
41 [ "$mds1_FSTYPE" = zfs ] && MDSSIZE=400000
43 [ "$ost1_FSTYPE" = zfs ] && OSTSIZE=400000
53 # pass "-E lazy_itable_init" to mke2fs to speed up the formatting time
54 if [[ "$LDISKFS_MKFS_OPTS" != *lazy_itable_init* ]]; then
55 LDISKFS_MKFS_OPTS=$(csa_add "$LDISKFS_MKFS_OPTS" -E lazy_itable_init)
59 require_dsh_mds || exit 0
60 require_dsh_ost || exit 0
65 # The MGS must be started before the OSTs for a new fs, so start
66 # and stop to generate the startup logs.
69 wait_osc_import_state mds ost FULL
74 reformat_and_config() {
76 if ! combined_mgs_mds ; then
82 writeconf_or_reformat() {
83 # There are at most 2 OSTs for write_conf test
84 # who knows if/where $TUNEFS is installed?
85 # Better reformat if it fails...
86 writeconf_all $MDSCOUNT 2 ||
87 { echo "tunefs failed, reformatting instead" &&
88 reformat_and_config && return 0; }
97 echo "start mgs service on $(facet_active_host mgs)"
98 start mgs $(mgsdevname) $MGS_MOUNT_OPTS "$@"
104 local dev=$(mdsdevname $num)
107 echo "start mds service on `facet_active_host $facet`"
108 start $facet ${dev} $MDS_MOUNT_OPTS "$@" || return 94
111 stop_mdt_no_force() {
114 local dev=$(mdsdevname $num)
117 echo "stop mds service on `facet_active_host $facet`"
118 stop $facet || return 97
124 local dev=$(mdsdevname $num)
127 echo "stop mds service on `facet_active_host $facet`"
128 # These tests all use non-failover stop
129 stop $facet -f || return 97
133 local mdscount=$MDSCOUNT
136 [[ "$1" == "--mdscount" ]] && mdscount=$2 && shift 2
138 for ((num=1; num <= $mdscount; num++ )); do
139 start_mdt $num "$@" || return 94
141 for ((num=1; num <= $mdscount; num++ )); 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()"
550 load_module ../libcfs/libcfs/libcfs ||
551 error "Failed to load modules libcfs rc = $?"
552 $LCTL set_param debug=all
555 run_test 5j "verify libcfs doesn't crash when setting debug flags"
560 mount_client $MOUNT || error "mount_client $MOUNT failed"
561 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
562 cleanup || error "cleanup failed with rc $?"
564 run_test 6 "manual umount, then mount again"
569 cleanup_nocli || error "cleanup_nocli failed with $?"
571 run_test 7 "manual umount, then cleanup"
575 mount_client $MOUNT2 || error "mount_client $MOUNT2 failed"
576 check_mount2 || error "check_mount2 failed"
577 umount_client $MOUNT2 || error "umount_client $MOUNT2 failed"
578 cleanup || error "cleanup failed with rc $?"
580 run_test 8 "double mount setup"
583 start_ost || error "OST start failed"
585 do_facet ost1 $LCTL set_param debug=\'inode trace\' ||
586 error "do_facet ost1 set_param inode trace failed."
587 do_facet ost1 $LCTL set_param subsystem_debug=\'mds ost\' ||
588 error "do_facet ost1 set_param debug mds ost failed."
590 CHECK_PTLDEBUG="`do_facet ost1 $LCTL get_param -n debug`"
591 if [ "$CHECK_PTLDEBUG" ] && { \
592 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
593 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
594 echo "lnet.debug success"
596 error "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
598 CHECK_SUBSYS="`do_facet ost1 $LCTL get_param -n subsystem_debug`"
599 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
600 echo "lnet.subsystem_debug success"
602 error "lnet.subsystem_debug: want 'mds ost' got '$CHECK_SUBSYS'"
604 stop_ost || error "Unable to stop OST1"
606 run_test 9 "test ptldebug and subsystem for mkfs"
611 if ! combined_mgs_mds; then
612 files=$(do_facet mgs "find /{proc/fs,sys/fs,sys/kernel/debug}/lustre -type l -exec test ! -e {} \; -print")
613 [ -z $files ] || echo "MGS $files is a broken symlink"
616 files=$(do_facet mds1 "find /{proc/fs,sys/fs,sys/kernel/debug}/lustre -type l -exec test ! -e {} \; -print")
617 [ -z $files ] || echo "MDS $files is a broken symlink"
619 files=$(do_facet ost1 "find /{proc/fs,sys/fs,sys/kernel/debug}/lustre -type l -exec test ! -e {} \; -print")
620 [ -z $files ] || echo "OSS $files is a broken symlink"
622 files=$(do_facet client "find /{proc/fs,sys/fs,sys/kernel/debug}/lustre -type l -exec test ! -e {} \; -print")
623 [ -z $files ] || echo "clients $files is a broken symlink"
625 cleanup || error "cleanup failed with rc $?"
627 run_test 10a "find lctl param broken symlinks"
630 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
631 # But with new MDS stack we don't care about the mode of local objects
632 # anymore, so this test is removed. See bug 22944 for more details.
636 if [ "$mds1_FSTYPE" != ldiskfs ]; then
637 skip "ldiskfs only test"
641 check_mount || error "check_mount failed"
642 cleanup || error "cleanup failed with rc $?"
644 echo "Remove mds config log"
645 if ! combined_mgs_mds ; then
649 do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' \
650 $(mgsdevname) || return \$?" ||
651 error "do_facet mgs failed with $?"
653 if ! combined_mgs_mds ; then
657 start_ost || error "OST start failed"
658 start_mds && error "MDS start succeeded, but should fail"
661 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
664 if [ "$mds1_FSTYPE" != ldiskfs ]; then
665 skip "ldiskfs only test"
668 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
673 # check if current MDSSIZE is large enough
674 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE &&
675 log "use MDSSIZE=$MDSSIZE"
677 # check if the global config has a large enough MDSSIZE
678 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] &&
679 [ $STORED_MDSSIZE -ge $MIN ] &&
680 OK=1 && myMDSSIZE=$STORED_MDSSIZE &&
681 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
683 # check if the block device is large enough
684 is_blkdev $SINGLEMDS $MDSDEV $MIN
685 local large_enough=$?
686 if [ -n "$OK" ]; then
687 [ $large_enough -ne 0 ] && OK=""
689 [ $large_enough -eq 0 ] && OK=1 && myMDSSIZE=$MIN &&
690 log "use device $MDSDEV with MIN=$MIN"
693 # check if a loopback device has enough space for fs metadata (5%)
695 if [ -z "$OK" ]; then
696 local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! \
697 -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
698 awk '($1 != "Filesystem") { print $4 }')
699 ! [ -z "$SPACE" ] && [ $SPACE -gt $((MIN / 20)) ] &&
700 OK=1 && myMDSSIZE=$MIN &&
701 log "use file $MDSDEV with MIN=$MIN"
704 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS"
706 echo "mount mds with large journal..."
708 local OLD_MDSSIZE=$MDSSIZE
712 echo "mount lustre system..."
714 check_mount || error "check_mount failed"
716 echo "check journal size..."
717 local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
718 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
719 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
721 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
724 cleanup || error "cleanup failed with rc $?"
729 run_test 18 "check mkfs creates large journals"
732 start_mds || error "MDS start failed"
733 stop_mds || error "Unable to stop MDS"
735 run_test 19a "start/stop MDS without OSTs"
738 start_ost || error "Unable to start OST1"
739 stop_ost -f || error "Unable to stop OST1"
741 run_test 19b "start/stop OSTs without MDS"
744 # first format the ost/mdt
745 start_mds || error "MDS start failed"
746 start_ost || error "Unable to start OST1"
747 mount_client $MOUNT || error "mount_client $MOUNT failed"
748 check_mount || error "check_mount failed"
749 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed."
750 mount_client $MOUNT remount,ro || error "remount client with ro failed"
751 touch $DIR/$tfile && error "$DIR/$tfile created incorrectly"
752 [ -e $DIR/$tfile ] && error "$DIR/$tfile exists incorrectly"
753 mount_client $MOUNT remount,rw || error "remount client with rw failed"
754 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
755 MCNT=$(grep -c $MOUNT' ' /etc/mtab)
756 [ "$MCNT" -ne 1 ] && error "$MOUNT in /etc/mtab $MCNT times"
758 stop_mds || error "Unable to stop MDS"
759 stop_ost || error "Unable to stop OST1"
761 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
764 start_mds || error "MDS start failed"
765 start_ost || error "unable to start OST1"
766 wait_osc_import_state mds ost FULL
767 stop_ost || error "unable to stop OST1"
768 stop_mds || error "unable to stop MDS"
770 run_test 21a "start mds before ost, stop ost first"
773 start_ost || error "unable to start OST1"
774 start_mds || error "MDS start failed"
775 wait_osc_import_state mds ost FULL
776 stop_mds || error "unable to stop MDS"
777 stop_ost || error "unable to stop OST1"
779 run_test 21b "start ost before mds, stop mds first"
782 start_ost || error "Unable to start OST1"
783 start_mds || error "MDS start failed"
784 start_ost2 || error "Unable to start OST2"
785 wait_osc_import_state mds ost2 FULL
786 stop_ost || error "Unable to stop OST1"
787 stop_ost2 || error "Unable to stop OST2"
788 stop_mds || error "Unable to stop MDS"
789 #writeconf to remove all ost2 traces for subsequent tests
790 writeconf_or_reformat
792 run_test 21c "start mds between two osts, stop mds last"
795 combined_mgs_mds && skip "need separate mgs device"
800 start_mgs || error "unable to start MGS"
801 start_ost || error "unable to start OST1"
802 start_ost2 || error "unable to start OST2"
803 start_mds || error "MDS start failed"
804 wait_osc_import_state mds ost2 FULL
806 local zkeeper=${KEEP_ZPOOL}
807 stack_trap "KEEP_ZPOOL=$zkeeper" EXIT
810 stop_ost || error "Unable to stop OST1"
811 stop_ost2 || error "Unable to stop OST2"
812 stop_mds || error "Unable to stop MDS"
814 #writeconf to remove all ost2 traces for subsequent tests
815 writeconf_or_reformat
816 KEEP_ZPOOL="${zkeeper}"
818 start_mgs || error "unable to start MGS"
820 run_test 21d "start mgs then ost and then mds"
823 MGSNID="$saved_mgsnid"
825 echo "stopping fs2mgs ..."
826 stop $fs2mgs -f || true
829 test_21e() { # LU-5863
830 if [[ -z "$fs3ost_DEV" || -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
831 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
832 skip_env "mixed loopback and real device not working"
835 local fs2mdsdev=$(mdsdevname 1_2)
836 local fs2ostdev=$(ostdevname 1_2)
837 local fs3ostdev=$(ostdevname 2_2)
839 local fs2mdsvdev=$(mdsvdevname 1_2)
840 local fs2ostvdev=$(ostvdevname 1_2)
841 local fs3ostvdev=$(ostvdevname 2_2)
843 # temporarily use fs3ost as fs2mgs
845 local fs2mgsdev=$fs3ostdev
846 local fs2mgsvdev=$fs3ostvdev
848 local fsname=test1234
850 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
851 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
852 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS && trap cleanup_21e EXIT INT ||
853 error "start fs2mgs failed"
855 local saved_mgsnid="$MGSNID"
856 MGSNID=$(do_facet $fs2mgs $LCTL list_nids | xargs | tr ' ' ,)
858 add fs2mds $(mkfs_opts mds1 $fs2mdsdev $fsname) \
859 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
860 add fs2ost $(mkfs_opts ost1 $fs2ostdev $fsname) \
861 --reformat $fs2ostdev $fs2ostvdev || error "add fs2ost failed"
863 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
864 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS || error "start fs2mds failed"
866 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
867 $MOUNT_CMD $MGSNID:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
868 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
872 run_test 21e "separate MGS and MDS"
875 start_mds || error "MDS start failed"
877 echo "Client mount with ost in logs, but none running"
878 start_ost || error "unable to start OST1"
879 # wait until mds connected to ost and open client connection
880 wait_osc_import_state mds ost FULL
881 stop_ost || error "unable to stop OST1"
882 mount_client $MOUNT || error "mount_client $MOUNT failed"
883 # check_mount will block trying to contact ost
884 mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed"
885 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed"
886 umount_client $MOUNT -f
889 echo "Client mount with a running ost"
890 start_ost || error "unable to start OST1"
892 # if gss enabled, wait full time to let connection from
893 # mds to ost be established, due to the mismatch between
894 # initial connect timeout and gss context negotiation timeout.
895 # This perhaps could be remove after AT landed.
896 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
897 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
899 mount_client $MOUNT || error "mount_client $MOUNT failed"
900 wait_osc_import_state mds ost FULL
901 wait_osc_import_ready client ost
902 check_mount || error "check_mount failed"
905 cleanup || error "cleanup failed with rc $?"
907 run_test 22 "start a client before osts (should return errs)"
909 test_23a() { # was test_23
912 stop $SINGLEMDS || error "failed to stop $SINGLEMDS"
913 # force down client so that recovering mds waits for reconnect
914 local running=$(grep -c $MOUNT /proc/mounts) || true
915 if [ $running -ne 0 ]; then
916 echo "Stopping client $MOUNT (opts: -f)"
920 # enter recovery on failed mds
921 local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/})
922 start $SINGLEMDS $MDT_DEV $MDS_MOUNT_OPTS || error "MDS start failed"
923 # try to start a new client
924 mount_client $MOUNT &
926 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
927 MOUNT_LUSTRE_PID=$(ps -ef | grep mount.lustre |
928 grep -v grep | awk '{print $2}')
929 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
931 ps --ppid $MOUNT_LUSTRE_PID
932 echo "waiting for mount to finish"
934 # "ctrl-c" sends SIGINT but it usually (in script) does not work on child process
935 # SIGTERM works but it does not spread to offspring processses
936 kill -s TERM $MOUNT_PID
937 kill -s TERM $MOUNT_LUSTRE_PID
938 # we can not wait $MOUNT_PID because it is not a child of this shell
944 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
946 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
947 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
950 [ -z "$PID1" -a -z "$PID2" ] && break
951 echo "waiting for mount to finish ... "
952 WAIT=$(( WAIT + sleep))
954 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
955 error "MOUNT_PID $MOUNT_PID and "\
956 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
959 cleanup || error "cleanup failed with rc $?"
961 run_test 23a "interrupt client during recovery mount delay"
963 test_23b() { # was test_23
964 start_mds || error "MDS start failed"
965 start_ost || error "Unable to start OST1"
966 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
967 $LCTL set_param fail_loc=0x80000313
969 cleanup || error "cleanup failed with rc $?"
971 run_test 23b "Simulate -EINTR during mount"
974 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
976 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
977 is_blkdev $SINGLEMDS $MDSDEV &&
978 skip_env "mixed loopback and real device not working"
981 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
983 local fs2mdsdev=$(mdsdevname 1_2)
984 local fs2ostdev=$(ostdevname 1_2)
985 local fs2mdsvdev=$(mdsvdevname 1_2)
986 local fs2ostvdev=$(ostvdevname 1_2)
989 # LU-9733 test fsname started with numbers as well
990 local FSNAME2=969362ae
992 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --nomgs --mgsnode=$MGSNID \
993 --fsname=${FSNAME2} --reformat $fs2mdsdev $fs2mdsvdev || exit 10
995 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --fsname=${FSNAME2} \
996 --reformat $fs2ostdev $fs2ostvdev || exit 10
999 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
1000 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1001 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
1002 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
1004 # LU-9733 test fsname started with numbers
1005 cl_user=$(do_facet $SINGLEMDS lctl --device $FSNAME2-MDT0000 \
1006 changelog_register -n) ||
1007 error "register changelog failed"
1009 do_facet $SINGLEMDS lctl --device $FSNAME2-MDT0000 \
1010 changelog_deregister $cl_user ||
1011 error "deregister changelog failed"
1013 check_mount || error "check_mount failed"
1014 # files written on 1 should not show up on 2
1015 cp /etc/passwd $DIR/$tfile
1017 [ -e $MOUNT2/$tfile ] && error "File bleed"
1020 cp /etc/passwd $MOUNT2/$tfile ||
1021 error "cp /etc/passwd $MOUNT2/$tfile failed"
1022 rm $MOUNT2/$tfile || error "remove $MOUNT2/$tfile failed"
1023 # 2 is actually mounted
1024 grep $MOUNT2' ' /proc/mounts > /dev/null || error "$MOUNT2 not mounted"
1026 facet_failover fs2mds
1027 facet_failover fs2ost
1029 umount_client $MOUNT
1030 # the MDS must remain up until last MDT
1032 MDS=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
1033 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
1034 [ -z "$MDS" ] && error "No MDT"
1036 cleanup_nocli || error "cleanup_nocli failed with rc $?"
1038 run_test 24a "Multiple MDTs on a single node"
1041 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1043 skip "needs combined MGT and MDT device"
1045 if [ -z "$fs2mds_DEV" ]; then
1046 local dev=${SINGLEMDS}_dev
1047 local MDSDEV=${!dev}
1048 is_blkdev $SINGLEMDS $MDSDEV &&
1049 skip_env "mixed loopback and real device not working"
1052 local fs2mdsdev=$(mdsdevname 1_2)
1053 local fs2mdsvdev=$(mdsvdevname 1_2)
1055 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --mgs --fsname=${FSNAME}2 \
1056 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
1058 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && {
1060 error "start MDS should fail"
1063 cleanup || error "cleanup failed with rc $?"
1065 run_test 24b "Multiple MGSs on a single node (should return err)"
1069 check_mount || error "check_mount failed"
1070 local MODULES=$($LCTL modules | awk '{ print $2 }')
1071 rmmod $MODULES 2>/dev/null || true
1072 cleanup || error "cleanup failed with $?"
1074 run_test 25 "Verify modules are referenced"
1078 # we need modules before mount for sysctl, so make sure...
1079 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
1080 #define OBD_FAIL_MDS_FS_SETUP 0x135
1081 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000135"
1082 start_mds && error "MDS started but should not have started"
1083 $LCTL get_param -n devices
1084 DEVS=$($LCTL get_param -n devices | egrep -v MG | wc -l)
1085 [ $DEVS -gt 0 ] && error "number of devices is $DEVS, should be zero"
1086 # start mds to drop writeconf setting
1087 start_mds || error "Unable to start MDS"
1088 stop_mds || error "Unable to stop MDS"
1089 unload_modules_conf || error "unload_modules_conf failed with $?"
1091 run_test 26 "MDT startup failure cleans LOV (should return errs)"
1095 start_ost || error "Unable to start OST1"
1096 start_mds || error "Unable to start MDS"
1097 echo "Requeue thread should have started: "
1098 ps -e | grep ll_cfg_requeue
1099 set_persistent_param_and_check ost1 \
1100 "obdfilter.$FSNAME-OST0000.client_cache_seconds" \
1101 "$FSNAME-OST0000.ost.client_cache_seconds"
1102 cleanup_nocli || error "cleanup_nocli failed with rc $?"
1104 run_test 27a "Reacquire MGS lock if OST started first"
1109 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
1110 awk '($3 ~ "mdt" && $4 ~ "MDT0000") { print $4 }')
1112 facet_failover $SINGLEMDS
1113 set_persistent_param_and_check $SINGLEMDS \
1114 "mdt.$device.identity_acquire_expire" \
1115 "$device.mdt.identity_acquire_expire"
1116 set_persistent_param_and_check client \
1117 "mdc.$device-mdc-*.max_rpcs_in_flight" \
1118 "$device.mdc.max_rpcs_in_flight"
1120 cleanup || error "cleanup failed with $?"
1122 run_test 27b "Reacquire MGS lock after failover"
1124 test_28A() { # was test_28
1127 local TEST="llite.$FSNAME-*.max_read_ahead_whole_mb"
1128 local PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
1129 local orig=$($LCTL get_param -n $TEST)
1130 local max=$($LCTL get_param -n \
1131 llite.$FSNAME-*.max_read_ahead_per_file_mb)
1133 orig=${orig%%.[0-9]*}
1135 echo "ORIG:$orig MAX:$max"
1136 [[ $max -le $orig ]] && orig=$((max - 3))
1137 echo "ORIG:$orig MAX:$max"
1139 local final=$((orig + 1))
1141 set_persistent_param_and_check client "$TEST" "$PARAM" $final
1142 final=$((final + 1))
1143 set_persistent_param_and_check client "$TEST" "$PARAM" $final
1144 umount_client $MOUNT || error "umount_client $MOUNT failed"
1145 mount_client $MOUNT || error "mount_client $MOUNT failed"
1147 local result=$($LCTL get_param -n $TEST)
1149 if [ $result -ne $final ]; then
1150 error "New config not seen: wanted $final got $result"
1152 echo "New config success: got $result"
1154 set_persistent_param_and_check client "$TEST" "$PARAM" $orig
1155 cleanup || error "cleanup failed with rc $?"
1157 run_test 28A "permanent parameter setting"
1159 test_28a() { # LU-4221
1160 [[ "$OST1_VERSION" -ge $(version_code 2.5.52) ]] ||
1161 skip "Need OST version at least 2.5.52"
1162 [ "$ost1_FSTYPE" = zfs ] &&
1163 skip "LU-4221: no such proc params for ZFS OSTs"
1170 local device="$FSNAME-OST0000"
1174 # In this test we will set three kinds of proc parameters with
1175 # lctl set_param -P or lctl conf_param:
1176 # 1. non-symlink ones in the OFD
1177 # 2. non-symlink ones in the OSD
1180 # prepare a non-symlink parameter in the OFD
1181 name="client_cache_seconds"
1182 param="$device.ost.$name"
1183 cmd="obdfilter.$device.$name"
1185 # permanently setting the parameter in the OFD
1186 old=$(do_facet ost1 $LCTL get_param -n $cmd)
1188 set_persistent_param_and_check ost1 "$cmd" "$param" $new
1189 set_persistent_param_and_check ost1 "$cmd" "$param" $old
1192 # prepare a non-symlink parameter in the OSD
1194 param="$device.osd.$name"
1195 cmd="osd-*.$device.$name"
1197 # conf_param the parameter in the OSD
1198 old=$(do_facet ost1 $LCTL get_param -n $cmd)
1199 new=$(((old + 1) % 2))
1200 set_persistent_param_and_check ost1 "$cmd" "$param" $new
1201 set_persistent_param_and_check ost1 "$cmd" "$param" $old
1203 cleanup || error "cleanup failed with $?"
1205 run_test 28a "set symlink parameters permanently with lctl"
1208 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >= 2 OSTs"
1209 setup_noconfig > /dev/null 2>&1
1210 start_ost2 || error "Unable to start OST2"
1213 local PARAM="$FSNAME-OST0001.osc.active"
1214 # With lctl set_param -P the value $PROC_ACT will be sent to
1215 # all nodes. The [!M] filter out the ability to set active
1216 # on the MDS servers which is tested with wait_osp_* below.
1217 # For ost_server_uuid that only exist on client so filtering
1219 local PROC_ACT="os[cp].$FSNAME-OST0001-osc-[^M]*.active"
1220 local PROC_UUID="os[cp].$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
1222 ACTV=$($LCTL get_param -n $PROC_ACT)
1224 set_persistent_param_and_check client $PROC_ACT $PARAM $DEAC
1225 # also check ost_server_uuid status
1226 RESULT=$($LCTL get_param -n $PROC_UUID | grep DEACTIV)
1227 if [ -z "$RESULT" ]; then
1228 error "Client not deactivated: $($LCTL get_param \
1231 echo "Live client success: got $RESULT"
1235 wait_osp_active ost ${FSNAME}-OST0001 1 0
1237 # test new client starts deactivated
1238 umount_client $MOUNT || error "umount_client $MOUNT failed"
1239 mount_client $MOUNT || error "mount_client $MOUNT failed"
1241 # the 2nd and 3rd field of ost_server_uuid do not update at the same
1242 # time when using lctl set_param -P
1243 wait_update_facet client \
1244 "$LCTL get_param -n $PROC_UUID | awk '{print \\\$3 }'" \
1246 error "New client start active: $($LCTL get_param -n $PROC_UUID)"
1248 echo "New client success: got '$($LCTL get_param -n $PROC_UUID)'"
1250 # make sure it reactivates
1251 set_persistent_param_and_check client $PROC_ACT $PARAM $ACTV
1253 umount_client $MOUNT
1254 stop_ost2 || error "Unable to stop OST2"
1255 cleanup_nocli || error "cleanup_nocli failed with $?"
1256 #writeconf to remove all ost2 traces for subsequent tests
1257 writeconf_or_reformat
1259 run_test 29 "permanently remove an OST"
1264 echo Big config llog
1265 local path="llite.$FSNAME-*"
1266 local cpath="$FSNAME.llite"
1267 local param="max_read_ahead_whole_mb"
1269 local test="${path}.$param"
1270 local conf="${cpath}.$param"
1272 local orig=$($LCTL get_param -n $test)
1273 local list=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1274 for i in ${list[@]}; do
1275 set_persistent_param_and_check client $test $conf $i
1277 # make sure client restart still works
1278 umount_client $MOUNT
1279 mount_client $MOUNT || error "mount_client $MOUNT failed"
1280 [ "$($LCTL get_param -n $test)" -ne "$i" ] &&
1281 error "Param didn't stick across restart $($test) != $i"
1284 echo Erase parameter setting
1285 if [[ $PERM_CMD == *"set_param -P"* ]]; then
1286 do_facet mgs "$PERM_CMD -d $test" ||
1287 error "Erase param $test failed"
1289 do_facet mgs "$PERM_CMD -d $conf" ||
1290 error "Erase param $conf failed"
1292 umount_client $MOUNT
1293 mount_client $MOUNT || error "mount_client $MOUNT failed"
1294 local final=$($LCTL get_param -n $test)
1295 echo "deleted (default) value=$final, orig=$orig"
1296 orig=${orig%%.[0-9]*}
1297 final=${final%%.[0-9]*}
1298 # assumes this parameter started at the default value
1299 [ "$final" -eq "$orig" ] ||
1300 error "Deleted value=$final -ne orig=$orig"
1302 cleanup || error "cleanup failed with rc $?"
1304 run_test 30a "Big config llog and permanent parameter deletion"
1309 local orignids=$($LCTL get_param -n \
1310 osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1312 local orignidcount=$(echo "$orignids" | wc -w)
1314 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1315 # numerical part of it. Hopefully that's not already a failover address
1317 local OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | \
1319 local ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1320 local NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1321 local NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1322 echo "Using fake nid $NEW"
1324 local TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1325 grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1326 if [[ $PERM_CMD == *"set_param -P"* ]]; then
1327 PARAM="osc.$FSNAME-OST0000-osc-[^M]*.import"
1328 echo "Setting $PARAM from $TEST to $NEW"
1329 do_facet mgs "$PERM_CMD $PARAM='connection=$NEW'" ||
1330 error "$PERM_CMD $PARAM failed"
1332 PARAM="$FSNAME-OST0000.failover.node"
1333 echo "Setting $PARAM from $TEST to $NEW"
1334 do_facet mgs "$PERM_CMD $PARAM='$NEW'" ||
1335 error "$PARAM $PARAM failed"
1337 wait_update_facet client "$TEST" "$NEW" ||
1338 error "check $PARAM failed!"
1340 local NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1342 local NIDCOUNT=$(echo "$NIDS" | wc -w)
1343 echo "should have $((orignidcount + 1)) entries \
1344 in failover nids string, have $NIDCOUNT"
1345 [ $NIDCOUNT -eq $((orignidcount + 1)) ] ||
1346 error "Failover nid not added"
1348 if [[ $PERM_CMD == *"set_param -P"* ]]; then
1349 do_facet mgs "$PERM_CMD -d osc.$FSNAME-OST0000-osc-*.import"
1351 do_facet mgs "$PERM_CMD -d $FSNAME-OST0000.failover.node" ||
1352 error "$PERM_CMD delete failed"
1354 umount_client $MOUNT
1355 mount_client $MOUNT || error "mount_client $MOUNT failed"
1357 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1359 NIDCOUNT=$(echo "$NIDS" | wc -w)
1360 echo "only $orignidcount final entries should remain \
1361 in failover nids string, have $NIDCOUNT"
1362 [ $NIDCOUNT -eq $orignidcount ] || error "Failover nids not removed"
1364 cleanup || error "cleanup failed with rc $?"
1366 run_test 30b "Remove failover nids"
1368 test_31() { # bug 10734
1369 # ipaddr must not exist
1370 $MOUNT_CMD 4.3.2.1@tcp:/lustre $MOUNT || true
1371 cleanup || error "cleanup failed with rc $?"
1373 run_test 31 "Connect to non-existent node (shouldn't crash)"
1377 T32_BLIMIT=40960 # Kbytes
1380 T32_PROLIMIT=$((T32_BLIMIT/10))
1382 # This is not really a test but a tool to create new disk
1383 # image tarballs for the upgrade tests.
1385 # Disk image tarballs should be created on single-node
1386 # clusters by running this test with default configurations
1387 # plus a few mandatory environment settings that are verified
1388 # at the beginning of the test.
1390 test_32newtarball() {
1394 local tmp=$TMP/t32_image_create
1405 local pj_quota_file_old
1408 if [ $FSNAME != t32fs -o \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o \
1409 $OSTCOUNT -ne 2 -o -z "$OSTDEV1" ]; then
1410 error "Needs FSNAME=t32fs MDSCOUNT=2 " \
1411 "MDSDEV1=<nonexistent_file> " \
1412 "MDSDEV2=<nonexistent_file> " \
1413 "(or MDSDEV, in the case of b1_8) " \
1414 "OSTCOUNT=2 OSTDEV1=<nonexistent_file> " \
1415 "OSTDEV2=<nonexistent_file>"
1419 echo "Found stale $tmp"
1423 mkdir $tmp/src || return 1
1424 tar cf - -C $src . | tar xf - -C $tmp/src
1426 # format ost with comma-separated NIDs to verify LU-4460
1427 local failnid="$(h2nettype 1.2.3.4),$(h2nettype 4.3.2.1)"
1428 MGSNID="$MGSNID,$MGSNID" OSTOPT="--failnode=$failnid" formatall
1432 [[ "$MDS1_VERSION" -ge $(version_code 2.3.50) ]] ||
1433 $LFS quotacheck -ug /mnt/$FSNAME
1434 $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
1437 if [[ $MDSCOUNT -ge 2 ]]; then
1438 remote_dir=/mnt/$FSNAME/remote_dir
1439 $LFS mkdir -i 1 $remote_dir
1440 tar cf - -C $tmp/src . | tar xf - -C $remote_dir
1442 target_dir=$remote_dir
1443 if [[ $MDS1_VERSION -ge $(version_code 2.7.0) ]]; then
1444 striped_dir=/mnt/$FSNAME/striped_dir_old
1445 $LFS mkdir -i 1 -c 2 $striped_dir
1446 tar cf - -C $tmp/src . | tar xf - -C $striped_dir
1449 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1450 target_dir=/mnt/$FSNAME
1453 # add project quota #
1454 [[ $mds1_FSTYPE == "ldiskfs" &&
1455 $MDS1_VERSION -gt $(version_code 2.9.55) ]] ||
1456 [[ $mds1_FSTYPE == "zfs" &&
1457 $MDS1_VERSION -gt $(version_code 2.10.53) ]] && {
1458 pj_quota_dir=$target_dir/project_quota_dir
1459 pj_quota_file_old=$pj_quota_dir/pj_quota_file_old
1461 enable_project_quota
1462 set_mdt_qtype ugp ||
1463 error "enable mdt quota failed"
1465 set_ost_qtype ugp ||
1466 error "enable ost quota failed"
1468 mkdir -p $pj_quota_dir
1469 $LFS setquota -p $T32_PRJID -b 0 -B $T32_PROLIMIT -i 0 \
1470 -I $T32_ILIMIT $pj_quota_dir ||
1471 error "setquota -p $T32_PRJID failed"
1473 $LFS setstripe $pj_quota_file_old -c 1 ||
1474 error "setstripe $pj_quota_file_old failed"
1476 chown $T32_QID:$T32_QID $pj_quota_file_old ||
1477 error "chown $pj_quota_file_old failed"
1479 change_project -p $T32_PRJID $pj_quota_file_old
1481 mkdir -p $tmp/src/project_quota_dir
1482 cp $pj_quota_file_old $tmp/src/project_quota_dir/
1485 #####################
1486 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1488 #if [[ $MDSCOUNT -ge 2 ]]; then
1489 # remote_dir=/mnt/$FSNAME/remote_dir
1490 # $LFS mkdir -i 1 $remote_dir
1491 # tar cf - -C $tmp/src . | tar xf - -C $remote_dir
1493 # if [[ "$MDS1_VERSION" -ge $(version_code 2.7.0) ]]; then
1494 # striped_dir=/mnt/$FSNAME/striped_dir_old
1495 # $LFS mkdir -i 1 -c 2 $striped_dir
1496 # tar cf - -C $tmp/src . | tar xf - -C $striped_dir
1501 if [[ $MDS1_VERSION -ge $(version_code 2.9.51) ]]; then
1502 pfl_dir=$target_dir/pfl_dir
1503 pfl_file=$pfl_dir/pfl_file
1505 $LFS setstripe -E 2M -c 1 -o 0 -E -1 -S 2M -c 1 -o 1 \
1507 error "Create PFL file failed"
1509 dd if=/dev/urandom of=$pfl_file bs=1k count=3k
1510 mkdir -p $tmp/src/pfl_dir
1511 cp $pfl_file $tmp/src/pfl_dir/
1516 if [[ $MDS1_VERSION -ge $(version_code 2.10.56) ]]; then
1517 dom_dir=$target_dir/dom_dir
1518 dom_file=$dom_dir/dom_file
1519 flr_dir=$target_dir/flr_dir
1520 flr_file=$flr_dir/flr_file
1523 $LFS setstripe -E 1M -L mdt -E -1 -S 4M $dom_file
1524 dd if=/dev/urandom of=$dom_file bs=1k count=2k
1525 mkdir -p $tmp/src/dom_dir
1526 cp $dom_file $tmp/src/dom_dir
1529 $LFS mirror create -N2 $flr_file
1530 dd if=/dev/urandom of=$flr_file bs=1k count=1
1531 mkdir -p $tmp/src/flr_dir
1532 cp $flr_file $tmp/src/flr_dir
1536 local large_xattr_dir=/mnt/$FSNAME/large_xattr_test_dir
1537 local xattr_file=$large_xattr_dir/large_xattr_file
1539 mkdir $large_xattr_dir
1541 setfattr -n user.fooattr -v $(printf "%c" {1..4096} ) $xattr_file ||
1546 mkdir $tmp/img || return 1
1550 pushd_dir=/mnt/$FSNAME
1551 if [[ $MDSCOUNT -ge 2 ]]; then
1552 pushd_dir=$remote_dir
1553 if [[ "$MDS1_VERSION" -ge $(version_code 2.7.0) ]]; then
1555 BLOCKSIZE=1024 ls -Rni --time-style=+%s >$tmp/img/list2
1561 BLOCKSIZE=1024 ls -Rni --time-style=+%s >$tmp/img/list
1562 find ! -name .lustre -type f -exec sha1sum {} \; |
1563 sort -k 2 >$tmp/img/sha1sums
1565 $LCTL get_param -n version | head -n 1 |
1566 sed -e 's/^lustre: *//' >$tmp/img/commit
1568 [[ "$MDS1_VERSION" -ge $(version_code 2.3.50) ]] ||
1569 $LFS quotaon -ug /mnt/$FSNAME
1570 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1571 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1572 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1573 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1574 | tr -d "*" > $tmp/img/bspace
1575 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1576 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1577 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1578 | tr -d "*" > $tmp/img/ispace
1579 echo $T32_BLIMIT > $tmp/img/blimit
1580 echo $T32_ILIMIT > $tmp/img/ilimit
1582 $MULTIOP /mnt/$FSNAME/orph_file Ouw_c&
1586 debugfs -R "ls /PENDING" ${MDSDEV1:-$MDSDEV}
1587 cp ${MDSDEV1:-$MDSDEV} $tmp/img
1596 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1599 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1600 echo "Data verification failed"
1603 uname -r >$tmp/img/kernel
1604 uname -m >$tmp/img/arch
1606 for ((num=1; num <= $MDSCOUNT; num++)); do
1607 local devname=$(mdsdevname $num)
1611 [[ num -eq 1 ]] && image_name=mdt || image_name=mdt$num
1612 [[ $(facet_fstype $facet) != zfs ]] ||
1613 devname=$(mdsvdevname $num)
1614 dd conv=sparse bs=4k if=$devname of=$tmp/img/$image_name
1617 for ((num=1; num <= $OSTCOUNT; num++)); do
1618 local devname=$(ostdevname $num)
1622 [[ num -eq 1 ]] && image_name=ost || image_name=ost$num
1623 [[ $(facet_fstype $facet) != zfs ]] ||
1624 devname=$(ostdevname $num)
1625 dd conv=sparse bs=4k if=$devname of=$tmp/img/$image_name
1628 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1629 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1632 tar cjvf $dst/disk$version-"$mds1_FSTYPE".tar.bz2 -S *
1637 # run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1640 # The list of applicable tarballs is returned via the caller's
1641 # variable "tarballs".
1644 [ "$CLIENTONLY" ] && skip "Client-only testing"
1646 check_versions || skip "do not upgrade with mismatched client/server"
1648 local node=$(facet_active_host $SINGLEMDS)
1649 local r="do_node $node"
1651 ! $r which "$TUNEFS" && skip_env "tunefs.lustre required on $node"
1653 local IMGTYPE="$mds1_FSTYPE"
1655 [[ -n "$T32IMAGE" ]] && tarballs=$T32IMAGE ||
1656 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 \
1657 -name \'disk*-$IMGTYPE.tar.bz2\')
1659 [ -z "$tarballs" ] && skip "No applicable tarballs found"
1662 t32_test_cleanup() {
1664 local facet=$SINGLEMDS
1667 echo "== cleanup with rc=$rc =="
1668 if $shall_cleanup_lustre; then
1669 umount $tmp/mnt/lustre || rc=$?
1671 if $shall_cleanup_mdt; then
1672 $r $UMOUNT $tmp/mnt/mdt || rc=$?
1674 if $shall_cleanup_mdt1; then
1675 $r $UMOUNT $tmp/mnt/mdt1 || rc=$?
1677 if $shall_cleanup_ost; then
1678 $r $UMOUNT $tmp/mnt/ost || rc=$?
1680 if $shall_cleanup_ost1; then
1681 $r $UMOUNT $tmp/mnt/ost1 || rc=$?
1686 if [[ "$mds1_FSTYPE" == zfs ]]; then
1688 local poolname_list="t32fs-mdt1 t32fs-ost1"
1690 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1692 for poolname in $poolname_list; do
1693 destroy_zpool $facet $poolname
1696 combined_mgs_mds || start_mgs || rc=$?
1700 t32_bits_per_long() {
1702 # Yes, this is not meant to be perfect.
1712 t32_reload_modules() {
1714 local all_removed=false
1717 [ "$mds1_FSTYPE" == zfs ] && do_rpc_nodes $node "service zed stop"
1719 while ((i < 20)); do
1720 echo "Unloading modules on $node: Attempt $i"
1721 do_rpc_nodes $node $LUSTRE_RMMOD "$mds1_FSTYPE" &&
1723 do_rpc_nodes $node check_mem_leak || return 1
1724 if $all_removed; then
1725 do_rpc_nodes $node load_modules
1728 if [ "$mds1_FSTYPE" == zfs ]; then
1729 do_rpc_nodes $node "$ZPOOL status -v"
1734 echo "Unloading modules on $node: Given up"
1738 t32_wait_til_devices_gone() {
1744 echo wait for devices to go
1745 while ((i < 20)); do
1746 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1747 loops=$(do_rpc_nodes $node losetup -a | grep -c t32)
1748 ((devices == 0 && loops == 0)) && return 0
1752 echo "waiting for dev on $node: dev $devices loop $loops given up"
1753 do_rpc_nodes $node "losetup -a"
1754 do_rpc_nodes $node "$LCTL device_list"
1758 t32_verify_quota() {
1765 # LU-2435: if the underlying zfs doesn't support userobj_accounting,
1766 # lustre will estimate the object count usage. This fails quota
1767 # verification in 32b. The object quota usage should be accurate after
1768 # zfs-0.7.0 is released.
1769 [ "$mds1_FSTYPE" == zfs ] && {
1770 local zfs_version=$(do_facet $facet cat /sys/module/zfs/version)
1772 [ $(version_code $zfs_version) -lt $(version_code 0.7.0) ] && {
1773 echo "Skip quota verify for zfs: $zfs_version"
1778 $LFS quota -u $T32_QID -v $mnt
1780 qval=$($LFS quota -v -u $T32_QID $mnt |
1781 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1782 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1784 [ $qval -eq $img_bspace ] || {
1785 echo "bspace, act:$qval, exp:$img_bspace"
1789 qval=$($LFS quota -v -u $T32_QID $mnt |
1790 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1791 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1793 [ $qval -eq $img_ispace ] || {
1794 echo "ispace, act:$qval, exp:$img_ispace"
1798 qval=$($LFS quota -v -u $T32_QID $mnt |
1799 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1800 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1802 [ $qval -eq $img_blimit ] || {
1803 echo "blimit, act:$qval, exp:$img_blimit"
1807 qval=$($LFS quota -v -u $T32_QID $mnt |
1808 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1809 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1811 [ $qval -eq $img_ilimit ] || {
1812 echo "ilimit, act:$qval, exp:$img_ilimit"
1816 set_persistent_param_and_check $facet \
1817 "osd-$mds1_FSTYPE.$fsname-MDT0000.quota_slave.enabled" \
1818 "$fsname.quota.mdt" ug
1820 set_persistent_param_and_check $facet \
1821 "osd-$mds1_FSTYPE.$fsname-OST0000.quota_slave.enabled" \
1822 "$fsname.quota.ost" ug
1825 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1826 bs=1M count=$((img_blimit / 1024)) oflag=sync && {
1827 echo "Write succeed, but expect -EDQUOT"
1830 rm -f $mnt/t32_qf_new
1832 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1834 echo "Create succeed, but expect -EDQUOT"
1837 unlinkmany $mnt/t32_qf_ $img_ilimit
1847 sync_all_data > /dev/null 2>&1 || true
1849 [ "$uuid" != "global" ] || uuid=$mnt
1851 $LFS quota -v "$1" "$2" $mnt |
1852 awk 'BEGIN { num='$spec' } { if ($1 == "'$uuid'") \
1853 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1860 local dne_upgrade=${dne_upgrade:-"no"}
1861 local dom_upgrade=${dom_upgrade:-"no"}
1862 local project_upgrade=${project_upgrade:-"no"}
1863 local ff_convert=${ff_convert:-"no"}
1864 local pfl_upgrade=${pfl_upgrade:-"no"}
1865 local project_quota_upgrade=${project_quota_upgrade:-"no"}
1866 local dom_new_upgrade=${dom_new_upgrade:-"no"}
1867 local flr_upgrade=${flr_upgrade:-"no"}
1868 local shall_cleanup_mdt=false
1869 local shall_cleanup_mdt1=false
1870 local shall_cleanup_ost=false
1871 local shall_cleanup_ost1=false
1872 local shall_cleanup_lustre=false
1873 local mdt2_is_available=false
1874 local ost2_is_available=false
1875 local node=$(facet_active_host $SINGLEMDS)
1876 local r="do_node $node"
1892 local mdt_dev=$tmp/mdt
1893 local mdt2_dev=$tmp/mdt2
1894 local ost_dev=$tmp/ost
1895 local ost2_dev=$tmp/ost2
1899 local pfl_file=$tmp/mnt/lustre/remote_dir/pfl_dir/pfl_file
1900 local flr_file=$tmp/mnt/lustre/remote_dir/flr_dir/flr_file
1901 local dom_file=$tmp/mnt/lustre/remote_dir/dom_dir/dom_file
1902 local quota_dir=$tmp/mnt/lustre/remote_dir/project_quota_dir
1904 combined_mgs_mds || stop_mgs || error "Unable to stop MGS"
1905 trap 'trap - RETURN; t32_test_cleanup' RETURN
1908 nid=$($r $LCTL list_nids | head -1)
1910 mkdir -p $tmp/mnt/lustre || error "mkdir $tmp/mnt/lustre failed"
1911 $r mkdir -p $tmp/mnt/{mdt,mdt1,ost,ost1}
1912 $r tar xjvf $tarball -S -C $tmp || {
1913 error_noexit "Unpacking the disk image tarball"
1916 img_commit=$($r cat $tmp/commit)
1917 img_kernel=$($r cat $tmp/kernel)
1918 img_arch=$($r cat $tmp/arch)
1919 img_bspace=$($r cat $tmp/bspace)
1920 img_ispace=$($r cat $tmp/ispace)
1922 # older images did not have "blimit" and "ilimit" files
1923 # use old values for T32_BLIMIT and T32_ILIMIT
1924 $r test -f $tmp/blimit && img_blimit=$($r cat $tmp/blimit) ||
1926 $r test -f $tmp/ilimit && img_ilimit=$($r cat $tmp/ilimit) ||
1929 echo "Upgrading from $(basename $tarball), created with:"
1930 echo " Commit: $img_commit"
1931 echo " Kernel: $img_kernel"
1932 echo " Arch: $img_arch"
1933 echo "OST version: $(lustre_build_version ost1)"
1935 # The conversion can be made only when both of the following
1936 # conditions are satisfied:
1937 # - ost device img version < 2.3.64
1938 # - ost server version >= 2.5
1939 [ $(version_code $img_commit) -ge $(version_code 2.3.64) -o \
1940 "$OST1_VERSION" -lt $(version_code 2.5.0) ] &&
1943 ! $r test -f $mdt2_dev || mdt2_is_available=true
1944 ! $r test -f $ost2_dev || ost2_is_available=true
1946 if [[ "$mds1_FSTYPE" == zfs ]]; then
1947 echo "== fstype is zfs =="
1950 local poolname_list="t32fs-mdt1 t32fs-ost1"
1952 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1954 for poolname in $poolname_list; do
1956 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
1957 $ZPOOL import -f -d $tmp $poolname"
1960 # upgrade zpool to latest supported features, including
1961 # dnode quota accounting in 0.7.0
1962 $r "$ZPOOL upgrade -a"
1964 mdt_dev=t32fs-mdt1/mdt1
1965 ost_dev=t32fs-ost1/ost1
1966 ! $mdt2_is_available || mdt2_dev=t32fs-mdt2/mdt2
1967 ! $ost2_is_available || ost2_dev=t32fs-ost2/ost2
1968 wait_update_facet $SINGLEMDS "$ZPOOL list |
1969 awk '/^t32fs-mdt1/ { print \\\$1 }'" "t32fs-mdt1" || {
1970 error_noexit "import zfs pool failed"
1973 elif [ "$project_upgrade" != "no" ]; then
1974 echo "== project upgrade =="
1975 ! $r tune2fs -O project $mdt_dev &&
1976 error_noexit "enable project on mdt0 failed" &&
1978 $mdt2_is_available && ! $r tune2fs "-O project" $mdt2_dev &&
1979 error_noexit "enable project on mdt failed" &&
1981 ! $r tune2fs -O project $ost_dev &&
1982 error_noexit "enable project on ost failed" &&
1984 $ost2_is_available && ! $r tune2fs -O project $ost2_dev &&
1985 error_noexit "enable project on ost2 failed" &&
1989 $r $LCTL set_param debug="$PTLDEBUG"
1991 $r $TUNEFS --dryrun $mdt_dev || {
1993 error_noexit "tunefs.lustre before mounting the MDT"
1997 if $mdt2_is_available; then
1998 $r $TUNEFS --dryrun $mdt2_dev || {
2000 error_noexit "tunefs.lustre before mounting the MDT"
2005 if [ "$writeconf" ]; then
2006 echo "== writeconf mdt tunefs and quota =="
2008 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2010 $r $TUNEFS --quota $mdt_dev || {
2012 error_noexit "Enable mdt quota feature"
2016 if [ "$project_quota_upgrade" != "no" ]; then
2017 echo "== enable mdt2 project quota =="
2018 $r $TUNE2FS -O project $mdt_dev || {
2020 error_noexit "tune2fs $mdt_dev failed"
2025 if $mdt2_is_available; then
2026 $r $TUNEFS --quota $mdt2_dev || {
2028 error_noexit "Enable mdt2 quota feature"
2031 if [ "$project_quota_upgrade" != "no" ]; then
2032 echo "== enable mdt2 project quota =="
2033 $r $TUNE2FS -O project $mdt2_dev || {
2036 "tune2fs $mdt2_dev failed"
2043 echo "== replace nids =="
2044 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
2045 [[ "$MGS_VERSION" -ge $(version_code 2.3.59) ]] ||
2046 skip "LU-2200: Cannot run over IB w/o lctl replace_nids "
2047 "(Need MGS version at least 2.3.59)"
2050 local osthost=$(facet_active_host ost1)
2051 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
2054 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2057 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt
2058 $r $LCTL replace_nids $fsname-OST0000 $ostnid || {
2060 error_noexit "replace_nids $fsname-OST0000 $ostnid failed"
2063 if $ost2_is_available; then
2064 $r $LCTL replace_nids $fsname-OST0001 $ostnid || {
2066 error_noexit "replace_nids $fsname-OST0001 $ostnid failed"
2070 $r $LCTL replace_nids $fsname-MDT0000 $nid || {
2072 error_noexit "replace_nids $fsname-MDT0000 $nid failed"
2075 if $mdt2_is_available; then
2076 $r $LCTL replace_nids $fsname-MDT0001 $nid || {
2078 error_noexit "replace_nids $fsname-MDT0001 $nid failed"
2082 $r $UMOUNT $tmp/mnt/mdt
2084 mopts=exclude=$fsname-OST0000
2085 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2090 t32_wait_til_devices_gone $node
2092 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2094 error_noexit "Mounting the MDT"
2097 shall_cleanup_mdt=true
2099 if $mdt2_is_available; then
2100 echo "== mdt2 available =="
2101 mopts=mgsnode=$nid,$mopts
2102 $r $MOUNT_CMD -o $mopts $mdt2_dev $tmp/mnt/mdt1 || {
2104 error_noexit "Mounting the MDT"
2107 shall_cleanup_mdt1=true
2108 echo "== mount new MDT $mdt2_dev =="
2110 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 || {
2111 error_noexit "enable remote dir create failed"
2115 elif [ "$dne_upgrade" != "no" ]; then
2116 local fs2mdsdev=$(mdsdevname 1_2)
2117 local fs2mdsvdev=$(mdsvdevname 1_2)
2119 echo "== mkfs new MDT on ${fs2mdsdev} =="
2120 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2121 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
2124 add $SINGLEMDS $(mds2failover_HOST="" \
2125 mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
2126 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
2127 error_noexit "Mkfs new MDT failed"
2131 [[ "$mds1_FSTYPE" != zfs ]] || import_zpool mds1
2133 $r $TUNEFS --dryrun $fs2mdsdev || {
2134 error_noexit "tunefs.lustre before mounting the MDT"
2138 echo "== mount new MDT....$fs2mdsdev =="
2139 $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
2140 error_noexit "mount mdt1 failed"
2144 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
2145 error_noexit "enable remote dir create failed"
2147 shall_cleanup_mdt1=true
2150 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
2151 error_noexit "Getting MDT UUID"
2154 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
2155 error_noexit "Unexpected MDT UUID: \"$uuid\""
2159 $r $TUNEFS --dryrun $ost_dev || {
2160 error_noexit "tunefs.lustre before mounting the OST"
2164 if $ost2_is_available; then
2165 $r $TUNEFS --dryrun $ost2_dev || {
2166 error_noexit "tunefs.lustre before mounting the OST"
2171 if [ "$writeconf" ]; then
2172 echo "== writeconf and ost quota tunefs =="
2173 mopts=mgsnode=$nid,$writeconf
2174 if [ "$ost1_FSTYPE" == ldiskfs ]; then
2176 $r $TUNEFS --quota $ost_dev || {
2178 error_noexit "Enable ost quota feature"
2182 if [ "$project_quota_upgrade" != "no" ]; then
2183 echo "== enable ost project quota =="
2184 $r $TUNE2FS -O project $ost_dev || {
2186 error_noexit "tune2fs $ost_dev failed"
2191 if $ost2_is_available; then
2192 $r $TUNEFS --quota $ost2_dev || {
2194 error_noexit "Enable ost2 quota feature"
2197 if [ "$project_quota_upgrade" != "no" ]; then
2198 echo "== enable ost2 project quota =="
2199 $r $TUNE2FS -O project $ost2_dev || {
2202 "tune2fs $ost2_dev failed"
2210 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2215 $r $MOUNT_CMD -onomgs -o$mopts $ost_dev $tmp/mnt/ost || {
2216 error_noexit "Mounting the OST"
2220 if $ost2_is_available; then
2221 $r $MOUNT_CMD -onomgs -o$mopts $ost2_dev $tmp/mnt/ost1 || {
2222 error_noexit "Mounting the OST2"
2225 shall_cleanup_ost1=true
2228 shall_cleanup_ost=true
2230 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
2231 error_noexit "Getting OST UUID"
2234 if [ "$uuid" != $fsname-OST0000_UUID ]; then
2235 error_noexit "Unexpected OST UUID: \"$uuid\""
2239 if $ost2_is_available; then
2240 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0001.uuid) ||{
2241 error_noexit "Getting OST1 UUID"
2244 if [ "$uuid" != $fsname-OST0001_UUID ]; then
2245 error_noexit "Unexpected OST1 UUID: \"$uuid\""
2250 if [[ $PERM_CMD =~ "set_param" ]]; then
2251 echo "== perm cmd = '$PERM_CMD' =="
2252 $r $PERM_CMD osc.$fsname-OST0000*.import=connection=$nid || {
2253 error_noexit "Setting OST \"failover.node\""
2256 $r $PERM_CMD mdc.$fsname-MDT0000*.import=connection=$nid || {
2257 error_noexit "Setting MDT \"failover.node\""
2260 $r $PERM_CMD osc.$fsname-OST0000-*.max_dirty_mb=15 || {
2261 error_noexit "Setting \"max_dirty_mb\""
2264 $r $PERM_CMD mdc.$fsname-MDT0000-*.max_rpcs_in_flight=9 || {
2265 error_noexit "Setting \"max_rpcs_in_flight\""
2268 $r $PERM_CMD lov.$fsname-MDT0000-*.stripesize=4M || {
2269 error_noexit "Setting \"lov.stripesize\""
2272 $r $PERM_CMD mdd.$fsname-MDT0000-*.atime_diff=70 || {
2273 error_noexit "Setting \"mdd.atime_diff\""
2277 echo "== perm cmd = '$PERM_CMD' =="
2278 $r $PERM_CMD $fsname-OST0000.failover.node=$nid || {
2279 error_noexit "Setting OST \"failover.node\""
2283 $r $PERM_CMD $fsname-MDT0000.failover.node=$nid || {
2284 error_noexit "Setting MDT \"failover.node\""
2288 $r $PERM_CMD $fsname-OST0000.osc.max_dirty_mb=15 || {
2289 error_noexit "Setting \"max_dirty_mb\""
2292 $r $PERM_CMD $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
2293 error_noexit "Setting \"max_rpcs_in_flight\""
2296 $r $PERM_CMD $fsname-MDT0000.lov.stripesize=4M || {
2297 error_noexit "Setting \"lov.stripesize\""
2300 $r $PERM_CMD $fsname-MDT0000.mdd.atime_diff=70 || {
2301 error_noexit "Setting \"mdd.atime_diff\""
2306 $r $LCTL pool_new $fsname.interop || {
2307 error_noexit "Setting \"interop\""
2311 if [ "$ff_convert" != "no" -a "$ost1_FSTYPE" == ldiskfs ]; then
2312 echo "== ff convert ($ff_convert) =="
2313 $r $LCTL lfsck_start -M $fsname-OST0000 || {
2314 error_noexit "Start OI scrub on OST0"
2318 # The oi_scrub should be on ost1, but for test_32(),
2319 # all on the SINGLEMDS.
2320 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
2321 osd-ldiskfs.$fsname-OST0000.oi_scrub |
2322 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
2323 error_noexit "Failed to get the expected 'completed'"
2327 local UPDATED=$($r $LCTL get_param -n \
2328 osd-ldiskfs.$fsname-OST0000.oi_scrub |
2329 awk '/^updated/ { print $2 }')
2330 [ $UPDATED -ge 1 ] || {
2331 error_noexit "Only $UPDATED objects have been converted"
2336 if [[ "$dne_upgrade" != "no" ]] || $mdt2_is_available; then
2337 echo "== dne upgrate ($dne_upgrade) or mdt2 available =="
2338 if [[ $PERM_CMD =~ "set_param" ]]; then
2339 $r $PERM_CMD mdc.$fsname-MDT0001*.import=connection=$nid || {
2340 error_noexit "Setting MDT1 \"failover.node\""
2344 $r $PERM_CMD mdc.$fsname-MDT0001-*.max_rpcs_in_flight=9 || {
2345 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
2348 $r $PERM_CMD lov.$fsname-MDT0001-*.stripesize=4M || {
2349 error_noexit "Setting MDT1 \"lov.stripesize\""
2353 $r $PERM_CMD $fsname-MDT0001.failover.node=$nid || {
2354 error_noexit "Setting MDT1 \"failover.node\""
2357 $r $PERM_CMD $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
2358 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
2361 $r $PERM_CMD $fsname-MDT0001.lov.stripesize=4M || {
2362 error_noexit "Setting MDT1 \"lov.stripesize\""
2368 if [ "$writeconf" ]; then
2369 echo "== writeconf and client mount =="
2370 $MOUNT_CMD $nid:/$fsname -o user_xattr $tmp/mnt/lustre || {
2371 error_noexit "Mounting the client"
2375 shall_cleanup_lustre=true
2376 $r $LCTL set_param debug="$PTLDEBUG"
2378 t32_verify_quota $SINGLEMDS $fsname $tmp/mnt/lustre || {
2379 error_noexit "verify quota failed"
2383 if $r test -f $tmp/list; then
2384 echo "== list verification =="
2386 # There is not a Test Framework API to copy files to or
2387 # from a remote node.
2389 # LU-2393 - do both sorts on same node to ensure locale
2391 local list_file=$tmp/list
2393 if $mdt2_is_available; then
2394 if [[ -d $tmp/mnt/lustre/striped_dir_old ]] &&
2395 $r test -f $tmp/list2; then
2396 list_file=$tmp/list2
2397 pushd $tmp/mnt/lustre/striped_dir_old
2399 pushd $tmp/mnt/lustre/remote_dir
2402 pushd $tmp/mnt/lustre
2404 $r cat $list_file | sort -k 6 >$tmp/list.orig
2405 BLOCKSIZE=1024 ls -Rni --time-style=+%s | sort -k 6 |
2406 sed 's/\. / /' >$tmp/list || {
2412 # 32-bit and 64-bit clients use different algorithms to
2413 # convert FIDs into inode numbers. Hence, remove the
2414 # inode numbers from the lists, if the original list was
2415 # created on an architecture with different number of
2418 if [ $(t32_bits_per_long $(uname -m)) != \
2419 $(t32_bits_per_long $img_arch) ]; then
2420 echo "Different number of bits per \"long\"" \
2421 "from the disk image"
2422 for list in list.orig list; do
2423 sed -i -e 's/^[0-9]\+[ \t]\+//' \
2427 if ! diff -ub $tmp/list.orig $tmp/list; then
2428 error_noexit "list verification failed"
2432 echo "== list verification skipped =="
2435 #non-dom upgrade to dom
2436 if [ "$dom_upgrade" != "no" ]; then
2437 echo "== dom upgrade =="
2438 $LFS setstripe -E 1M -L mdt -E EOF $tmp/mnt/lustre/dom || {
2439 error_noexit "Verify DoM creation"
2442 [ $($LFS getstripe -L $tmp/mnt/lustre/dom) == "mdt" ] || {
2443 error_noexit "Verify a DoM file"
2446 dd if=/dev/urandom of=$tmp/mnt/lustre/dom bs=4096 \
2447 count=1 conv=fsync || {
2448 error_noexit "Cannot write to DoM file"
2451 [ $(stat -c%s $tmp/mnt/lustre/dom) == 4096 ] || {
2452 error_noexit "DoM: bad size after write"
2455 rm $tmp/mnt/lustre/dom
2457 set_persistent_param_and_check mds \
2458 "lod.*MDT0000*.dom_stripesize" \
2459 "$fsname-MDT0000.lod.dom_stripesize" 0 || {
2460 error_noexit "Changing \"dom_stripesize\""
2466 #$LFS setstripe -E 1M -L mdt -E -1 -S 4M $dom_file
2467 if [ "$dom_new_upgrade" != "no" ]; then
2468 if ! $mdt2_is_available; then
2469 dom_file=$tmp/mnt/lustre/dom_dir/dom_file
2471 echo "== check DoM file can be accessed =="
2472 [ $($LFS getstripe -I1 -L $dom_file) == "mdt" ] || {
2473 error_noexit "Verify a DoM file"
2476 [ $(stat -c%s $dom_file) == $((2 * 1024 * 1024)) ] || {
2477 error_noexit "DoM: bad size after write"
2482 if [ "$flr_upgrade" != "no" ]; then
2483 if ! $mdt2_is_available; then
2484 flr_file=$tmp/mnt/lustre/flr_dir/flr_file
2486 local mirror_count=$($LFS getstripe -N $flr_file)
2487 echo "== check FLR file =="
2488 [ $mirror_count == 2 ] || {
2489 error_noexit "FLR mirror count wrong"
2494 if [ "$dne_upgrade" != "no" ]; then
2495 echo "== dne upgrade striped_dir =="
2496 $LFS mkdir -i 1 -c2 $tmp/mnt/lustre/striped_dir || {
2497 error_noexit "set striped dir failed"
2500 $LFS setdirstripe -D -c2 $tmp/mnt/lustre/striped_dir
2502 pushd $tmp/mnt/lustre
2503 tar -c --exclude=./striped_dir \
2504 --exclude=./striped_dir_old \
2505 --exclude=./remote_dir -f - .|
2506 tar -xvf - -C striped_dir 1>/dev/null || {
2507 error_noexit "cp to striped dir failed"
2513 if [ "$pfl_upgrade" != "no" ]; then
2520 echo "== check PFL file =="
2521 if ! $mdt2_is_available; then
2522 pfl_file=$tmp/mnt/lustre/pfl_dir/pfl_file
2524 comp_cnt=$($LFS getstripe --component-count $pfl_file)
2525 [ $comp_cnt == 2 ] || {
2526 error_noexit "wrong comp_cnt $comp_cnt"
2530 comp_size=$($LFS getstripe -I1 -E $pfl_file)
2531 [ $comp_size == 2097152 ] || {
2532 error_noexit "wrong component size $comp_size"
2536 comp_id=$($LFS getstripe -I1 -i $pfl_file)
2537 [ $comp_id == 0 ] || {
2538 error_noexit "wrong comp id $comp_id"
2542 comp_id=$($LFS getstripe -I2 -i $pfl_file)
2543 [ $comp_id -eq 1 ] || {
2544 error_noexit "wrong comp id $comp_id"
2548 stripe_size=$($LFS getstripe -I1 -S $pfl_file)
2549 [ $stripe_size -eq 1048576 ] || {
2550 error_noexit "wrong stripe size $stripe_size"
2554 stripe_size=$($LFS getstripe -I2 -S $pfl_file)
2555 [ $comp_size -eq 2097152 ] || {
2556 error_noexit "wrong component size $comp_size"
2561 local large_xattr_dir=$tmp/mnt/lustre/large_xattr_test_dir
2563 if [[ -d $large_xattr_dir ]]; then
2564 echo "== check Large EA =="
2565 local xattr_file=$large_xattr_dir/large_xattr_file
2567 xattr_val=$(getfattr --only-values\
2568 -n user.fooattr $xattr_file) || {
2569 error_noexit "Large EA cannot be read"
2573 [[ $xattr_val == $(printf "%c" {1..4096}) ]] || {
2574 error_noexit "Wrong large EA value"
2579 if [ "$project_quota_upgrade" != "no" ]; then
2580 if ! $mdt2_is_available; then
2581 quota_dir=$tmp/mnt/lustre/project_quota_dir
2585 echo "== check Project Quota =="
2586 hardlimit=$(getquota -p $T32_PRJID global 3 \
2588 [ $hardlimit == $T32_PROLIMIT ] || {
2589 error_noexit "wrong hardlimit $hardlimit"
2594 # If it is upgrade from DNE (2.5), then rename the remote dir,
2595 # which is created in 2.5 to striped dir.
2596 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]]; then
2597 echo "== rename remote_dir =="
2598 stripe_index=$($LFS getdirstripe -i \
2599 $tmp/mnt/lustre/remote_dir)
2601 [[ $stripe_index -eq 1 ]] || {
2602 error_noexit "get index \"$stripe_index\"" \
2603 "from remote dir failed"
2606 mv $tmp/mnt/lustre/remote_dir \
2607 $tmp/mnt/lustre/striped_dir/ || {
2608 error_noexit "mv remote dir failed"
2613 # If it is upgraded from DNE (2.7), then move the striped dir
2614 # which was created in 2.7 to the new striped dir.
2615 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]] &&
2616 echo "== rename striped dir =="
2617 [[ -d $tmp/mnt/lustre/striped_dir_old ]]; then
2618 stripe_count=$($LFS getdirstripe -c \
2619 $tmp/mnt/lustre/striped_dir_old)
2620 [[ $stripe_count -eq 2 ]] || {
2621 error_noexit "get count $stripe_count" \
2622 "from striped dir failed"
2625 mv $tmp/mnt/lustre/striped_dir_old \
2626 $tmp/mnt/lustre/striped_dir/ || {
2627 error_noexit "mv striped dir failed"
2633 $r $LCTL set_param -n osd*.*.force_sync=1
2634 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
2635 error_noexit "dd failed"
2638 rm -rf $tmp/mnt/lustre/tmp_file || {
2639 error_noexit "rm failed"
2643 if $r test -f $tmp/sha1sums; then
2644 echo "== checking sha1sums =="
2645 # LU-2393 - do both sorts on same node to ensure locale
2647 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
2648 if [[ "$dne_upgrade" != "no" ]]; then
2649 if [[ -d $tmp/mnt/lustre/striped_dir/remote_dir ]]; then
2650 pushd $tmp/mnt/lustre/striped_dir/remote_dir
2652 pushd $tmp/mnt/lustre/striped_dir
2654 elif [[ "$pfl_upgrade" != "no" ]] ||
2655 [[ "$flr_upgrade" != "no" ]] ||
2656 [[ "$dom_new_upgrade" != "no" ]] ||
2657 [[ "$project_quota_upgrade" != "no" ]] ||
2658 [[ -d $tmp/mnt/lustre/remote_dir ]]; then
2659 pushd $tmp/mnt/lustre/remote_dir
2661 pushd $tmp/mnt/lustre
2664 find ! -path "*remote_dir*" ! -path "*striped_dir*" \
2665 ! -name .lustre -type f -exec sha1sum {} \; |
2666 sort -k 2 >$tmp/sha1sums || {
2668 error_noexit "sha1sum"
2674 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
2675 echo "==** find $save files start **=="
2676 find $save -type f -ls
2677 echo "==** find $save files finish **=="
2678 error_noexit "sha1sum verification failed"
2682 # if upgrade from DNE(2.5), then check remote directory
2683 # if upgrade from DNE(2.7), then check striped directory
2684 if $mdt2_is_available &&
2685 [[ "$dne_upgrade" != "no" ]]; then
2686 echo "== mdt2 upgrade sha1sum check =="
2687 local new_dir="$tmp/mnt/lustre/striped_dir"
2688 local striped_dir_old="$new_dir/striped_dir_old"
2690 local dir_list="$new_dir/remote_dir"
2691 [[ ! -d $triped_dir_old ]] ||
2692 dir_list+=" $striped_dir_old"
2694 for dir in $dir_list; do
2695 echo "== dir = $dir =="
2697 find ! -name .lustre -type f \
2698 -exec sha1sum {} \; |
2699 sort -k 2 >$tmp/sha1sums || {
2702 error_noexit "sha1sum"
2706 if ! diff -ub $tmp/sha1sums.orig \
2709 error_noexit "sha1sum $dir" \
2716 echo "== sha1sum verification skipped =="
2719 # PFL write test after sha1sum check
2720 if [ "$pfl_upgrade" != "no" ]; then
2721 local rw_len=$((3 * 1034 * 1024))
2723 if ! $mdt2_is_available; then
2724 pfl_file=$tmp/mnt/lustre/pfl_dir/pfl_file
2726 small_write $pfl_file $rw_len || {
2727 error_noexit "PFL RW Failed"
2732 if [ "$dne_upgrade" != "no" ]; then
2733 rm -rf $tmp/mnt/lustre/striped_dir || {
2734 error_noexit "remove remote dir failed"
2739 # migrate files/dirs to remote MDT, then move them back
2740 if [ "$MDS1_VERSION" -ge $(version_code 2.7.50) -a \
2741 $dne_upgrade != "no" ]; then
2742 $r $LCTL set_param -n \
2743 mdt.${fsname}*.enable_remote_dir=1 2>/dev/null
2745 echo "== test migration =="
2746 pushd $tmp/mnt/lustre
2747 for dir in $(find ! -name .lustre ! -name . -type d); do
2748 mdt_index=$($LFS getdirstripe -i $dir)
2749 stripe_cnt=$($LFS getdirstripe -c $dir)
2750 if [ $mdt_index = 0 -a $stripe_cnt -le 1 ]; then
2751 $LFS migrate -m 1 $dir || {
2753 error_noexit "migrate MDT1 failed"
2759 for dir in $(find ! -name . ! -name .lustre -type d); do
2760 mdt_index=$($LFS getdirstripe -i $dir)
2761 stripe_cnt=$($LFS getdirstripe -c $dir)
2762 if [ $mdt_index = 1 -a $stripe_cnt -le 1 ]; then
2763 $LFS migrate -m 0 $dir || {
2765 error_noexit "migrate MDT0 failed"
2774 # When adding new data verification tests, please check for
2775 # the presence of the required reference files first, like
2776 # the "sha1sums" and "list" tests above, to avoid the need to
2777 # regenerate every image for each test addition.
2780 nrpcs_orig=$($LCTL get_param \
2781 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
2782 error_noexit "Getting \"max_rpcs_in_flight\""
2785 nrpcs=$((nrpcs_orig + 5))
2787 set_persistent_param_and_check client \
2788 "mdc.$fsname-MDT0000*.max_rpcs_in_flight" \
2789 "$fsname-MDT0000.mdc.max_rpcs_in_flight" $nrpcs || {
2790 error_noexit "Changing \"max_rpcs_in_flight\""
2794 umount $tmp/mnt/lustre || {
2795 error_noexit "Unmounting the client"
2798 shall_cleanup_lustre=false
2800 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
2801 error_noexit "Mounting the client"
2805 if [ "$mds1_FSTYPE" == ldiskfs -a \
2806 "$project_upgrade" != "no" ]; then
2807 ! $LFS project -d -p 1 $tmp/mnt/lustre/* &&
2808 error_noexit "set project failed" &&
2812 [[ $(do_facet mds1 pgrep orph_.*-MDD | wc -l) == 0 ]] ||
2813 error "MDD orphan cleanup thread not quit"
2815 umount $tmp/mnt/lustre || {
2816 error_noexit "Unmounting the client"
2820 if [[ "$dne_upgrade" != "no" ]] || $mdt2_is_available; then
2821 $r $UMOUNT $tmp/mnt/mdt1 || {
2822 error_noexit "Unmounting the MDT2"
2825 if [[ "$mds1_FSTYPE" == zfs ]]; then
2826 $r "$ZPOOL export t32fs-mdt2"
2828 shall_cleanup_mdt1=false
2831 $r $UMOUNT $tmp/mnt/mdt || {
2832 error_noexit "Unmounting the MDT"
2835 if [[ "$mds1_FSTYPE" == zfs ]]; then
2836 $r "$ZPOOL export t32fs-mdt1"
2838 shall_cleanup_mdt=false
2840 $r $UMOUNT $tmp/mnt/ost || {
2841 error_noexit "Unmounting the OST"
2844 if $ost2_is_available; then
2845 $r $UMOUNT $tmp/mnt/ost1 || {
2846 error_noexit "Unmounting the OST1"
2849 shall_cleanup_ost1=false
2852 if [[ $ost1_FSTYPE == zfs ]]; then
2853 $r "$ZPOOL export t32fs-ost1"
2854 if $ost2_is_available; then
2855 $r "$ZPOOL export t32fs-ost2"
2858 shall_cleanup_ost=false
2860 t32_reload_modules $node || {
2861 error_noexit "Reloading modules"
2865 if [[ "$mds1_FSTYPE" == zfs ]]; then
2866 local poolname=t32fs-mdt1
2868 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
2869 $ZPOOL import -f -d $tmp $poolname"
2871 # upgrade zpool to latest supported features,
2872 # including dnode quota accounting in 0.7.0
2873 $r "$ZPOOL upgrade $poolname"
2876 # mount a second time to make sure we didnt leave upgrade flag on
2877 $r $TUNEFS --dryrun $mdt_dev || {
2879 error_noexit "tunefs.lustre before remounting the MDT"
2883 mopts=exclude=$fsname-OST0000
2884 if [ "$mds1_FSTYPE" == ldiskfs ]; then
2887 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2888 error_noexit "Remounting the MDT"
2891 shall_cleanup_mdt=true
2901 for tarball in $tarballs; do
2902 banner "testing $tarball upgrade"
2903 t32_test $tarball || let "rc += $?"
2907 run_test 32a "Upgrade (not live)"
2915 for tarball in $tarballs; do
2916 banner "testing $tarball upgrade with writeconf"
2917 t32_test $tarball writeconf || let "rc += $?"
2921 run_test 32b "Upgrade with writeconf"
2929 for tarball in $tarballs; do
2930 banner "testing $tarball upgrade with DNE"
2932 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
2936 run_test 32c "dne upgrade test"
2944 for tarball in $tarballs; do
2945 banner "testing $tarball upgrade with ff and project upgrade"
2946 project_upgrade="no"
2947 [[ "$MDS1_VERSION" -ge $(version_code 2.13.54) ]] &&
2948 [[ "$tarball" =~ "disk2_4-ldiskfs" ]] &&
2949 project_upgrade="yes"
2950 project_upgrade=$project_upgrade ff_convert=yes t32_test \
2955 run_test 32d "convert ff and project quota upgrade test"
2958 [[ "$MDS1_VERSION" -ge $(version_code 2.10.56) ]] ||
2959 skip "Need MDS version at least 2.10.56"
2966 for tarball in $tarballs; do
2967 [[ "$tarball" =~ "2_9" ]] ||
2968 { echo "skip $(basename $tarball)"; continue; }
2970 banner "testing $tarball upgrade with DoM"
2971 dom_upgrade=yes t32_test $tarball writeconf || let "rc += $?"
2975 run_test 32e "dom upgrade test"
2978 [[ $MDS1_VERSION -ge $(version_code 2.10.56) ]] ||
2979 skip "Need MDS version at least 2.10.56"
2986 for tarball in $tarballs; do
2987 [[ "$tarball" =~ "2_10" ]] ||
2988 { echo "skip $(basename $tarball)"; continue; }
2989 pfl_upgrade=yes project_quota_upgrade=yes \
2990 t32_test $tarball writeconf || let "rc += $?"
2994 run_test 32f "pfl upgrade test"
2997 [[ $MDS1_VERSION -ge $(version_code 2.10.56) ]] ||
2998 skip "Need MDS version at least 2.10.56"
3005 for tarball in $tarballs; do
3006 [[ $tarball =~ "2_12" ]] ||
3007 { echo "skip $(basename $tarball)"; continue; }
3008 flr_upgrade=yes dom_new_upgrade=yes \
3009 t32_test $tarball writeconf || let "rc += $?"
3013 run_test 32g "flr/dom upgrade test"
3015 test_33a() { # bug 12333, was test_33
3016 local FSNAME2=test-$testnum
3018 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3021 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
3023 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
3024 local dev=${SINGLEMDS}_dev
3025 local MDSDEV=${!dev}
3027 is_blkdev $SINGLEMDS $MDSDEV &&
3028 skip_env "mixed loopback and real device not working"
3031 local fs2mdsdev=$(mdsdevname 1_2)
3032 local fs2ostdev=$(ostdevname 1_2)
3033 local fs2mdsvdev=$(mdsvdevname 1_2)
3034 local fs2ostvdev=$(ostvdevname 1_2)
3036 if [ "$mds1_FSTYPE" == ldiskfs ]; then
3037 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
3040 if combined_mgs_mds; then
3041 local mgs_flag="--mgs"
3044 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --fsname=${FSNAME2} \
3045 --reformat $mgs_flag $mkfsoptions $fs2mdsdev $fs2mdsvdev ||
3047 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
3048 --fsname=${FSNAME2} --index=0x1fff --reformat $fs2ostdev \
3049 $fs2ostvdev || exit 10
3051 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
3052 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
3054 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3055 do_facet mgs "$PERM_CMD timeout=200" ||
3056 error "$PERM_CMD timeout=200 failed"
3058 do_facet mgs "$PERM_CMD $FSNAME2.sys.timeout=200" ||
3059 error "$PERM_CMD $FSNAME2.sys.timeout=200 failed"
3061 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
3062 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
3065 cp /etc/hosts $MOUNT2/ || error "copy /etc/hosts $MOUNT2/ failed"
3066 $LFS getstripe $MOUNT2/hosts ||
3067 error "$LFS getstripe $MOUNT2/hosts failed"
3071 if (( "$MGS_VERSION" >= $(version_code 2.15.51) &&
3072 "$MDS1_VERSION" >= $(version_code 2.15.51) )); then
3073 # test lctl del_ost on large index
3074 do_facet mgs "$LCTL del_ost -t ${FSNAME2}-OST1fff" ||
3075 error "del_ost failed with $?"
3076 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 ||
3077 error "$MOUNT_CMD failed"
3079 $LFS df | grep -q OST1fff &&
3080 error "del_ost did not remove OST1fff!"
3086 cleanup_nocli || error "cleanup_nocli failed with $?"
3088 run_test 33a "Mount ost with a large index number"
3090 test_33b() { # was test_34
3093 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
3094 # Drop lock cancelation reply during umount
3095 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
3096 do_facet client $LCTL set_param fail_loc=0x80000304
3097 #lctl set_param debug=-1
3098 umount_client $MOUNT
3099 cleanup || error "cleanup failed with $?"
3101 run_test 33b "Drop cancel during umount"
3104 (( MDS1_VERSION >= $(version_code 2.15.57) )) ||
3105 skip "Need MDS version at least 2.15.57"
3106 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3107 local tstid=${TSTID:-"$(id -u $TSTUSR)"}
3109 local qpool="qpool1"
3111 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
3113 local fs2mdsdev=$(mdsdevname 1_2)
3114 local fs2ostdev=$(ostdevname 1_2)
3115 local fs2mdsvdev=$(mdsvdevname 1_2)
3116 local fs2ostvdev=$(ostvdevname 1_2)
3118 if [ "$mds1_FSTYPE" == ldiskfs ]; then
3119 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
3122 if combined_mgs_mds; then
3123 local mgs_flag="--mgs"
3127 stack_trap unload_modules_conf
3129 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --fsname=${FSNAME} \
3130 --reformat $mgs_flag $mkfsoptions $fs2mdsdev $fs2mdsvdev ||
3132 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
3133 --fsname=${FSNAME} --index=0x7c6 --reformat $fs2ostdev \
3134 $fs2ostvdev || exit 10
3137 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
3138 stack_trap "stop fs2mds -f"
3139 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
3140 stack_trap "stop fs2ost -f"
3142 mount_client $MOUNT || error "client start failed"
3143 stack_trap "umount_client $MOUNT"
3144 mkdir_on_mdt0 $DIR/$tdir || error "cannot create $DIR/$tdir"
3145 chmod 0777 $DIR/$tdir || error "chown failed"
3146 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3147 do_facet mgs $PERM_CMD \
3148 osd-*.$FSNAME-OST*.quota_slave.enable=$QUOTA_TYPE
3150 do_facet mgs $PERM_CMD $FSNAME.quota.ost=$QUOTA_TYPE ||
3151 error "set ost quota type failed"
3154 local old_MDSCOUNT=$MDSCOUNT
3156 stack_trap "MDSCOUNT=$old_MDSCOUNT"
3158 pool_add $qpool || error "pool_add failed"
3159 pool_add_targets $qpool 0x7c6
3161 $LFS setquota -u $tstid -B20M -b 0 $MOUNT
3162 $LFS setquota -g $tstid -B20M -b 0 $MOUNT
3163 $LFS setquota -u $tstid -B20M -b 0 --pool $qpool $MOUNT
3164 $LFS setquota -g $tstid -B20M -b 0 --pool $qpool $MOUNT
3166 for i in {1..10}; do
3167 runas -u $tstid -g $tstid dd if=/dev/zero of=$DIR/$tdir/f1 \
3168 bs=1M count=30 oflag=direct
3174 #umount_client $MOUNT || error "client start failed"
3177 #cleanup_nocli || error "cleanup_nocli failed with $?"
3179 run_test 33c "Mount ost with a large index number"
3183 do_facet client "bash runmultiop_bg_pause $DIR/file O_c"
3184 manual_umount_client
3186 do_facet client killall -USR1 multiop
3187 if [ $rc -eq 0 ]; then
3188 error "umount not fail!"
3191 cleanup || error "cleanup failed with rc $?"
3193 run_test 34a "umount with opened file should be fail"
3197 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3198 stop_mds || error "Unable to stop MDS"
3200 manual_umount_client --force || error "mtab after failed umount with $?"
3202 cleanup || error "cleanup failed with $?"
3204 run_test 34b "force umount with failed mds should be normal"
3208 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3209 stop_ost || error "Unable to stop OST1"
3211 manual_umount_client --force || error "mtab after failed umount with $?"
3213 cleanup || error "cleanup failed with $?"
3215 run_test 34c "force umount with failed ost should be normal"
3217 test_35a() { # bug 12459
3220 DBG_SAVE="`$LCTL get_param -n debug`"
3221 $LCTL set_param debug="ha"
3223 log "Set up a fake failnode for the MDS"
3225 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
3226 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
3228 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3229 do_facet mgs "$PERM_CMD \
3230 mdc.*${device}*.import=connection=$(h2nettype $FAKENID)" ||
3231 error "Setting mdc.*${device}*.import=connection=\
3232 $(h2nettype $FAKENID) failed."
3234 do_facet mgs "$PERM_CMD \
3235 ${device}.failover.node=$(h2nettype $FAKENID)" ||
3236 error "Setting ${device}.failover.node=\
3237 $(h2nettype $FAKENID) failed."
3239 log "Wait for RECONNECT_INTERVAL seconds (10s)"
3242 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
3245 log "Stopping the MDT: $device"
3246 stop_mdt 1 || error "MDT0 stop fail"
3248 df $MOUNT > /dev/null 2>&1 &
3250 log "Restarting the MDT: $device"
3251 start_mdt 1 || error "MDT0 start fail"
3252 log "Wait for df ($DFPID) ... "
3255 $LCTL set_param debug="$DBG_SAVE"
3257 # retrieve from the log the first server that the client tried to
3258 # contact after the connection loss
3259 $LCTL dk $TMP/lustre-log-$TESTNAME.log
3260 NEXTCONN=`awk "/${MSG}/ {start = 1;}
3261 /import_select_connection.*$device-mdc.* using connection/ {
3263 if (\\\$NF ~ /$FAKENID/)
3269 }" $TMP/lustre-log-$TESTNAME.log`
3270 [ "$NEXTCONN" != "0" ] &&
3271 error "Tried to connect to ${NEXTCONN} not last active server"
3272 cleanup || error "cleanup failed with $?"
3273 # remove nid settings
3274 writeconf_or_reformat
3276 run_test 35a "Reconnect to the last active server first"
3278 test_35b() { # bug 18674
3279 remote_mds || skip "local MDS"
3283 $LCTL set_param debug="ha"
3285 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
3288 log "Set up a fake failnode for the MDS"
3290 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
3291 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
3293 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3294 do_facet mgs "$PERM_CMD \
3295 mdc.*${device}*.import=connection=$(h2nettype $FAKENID)" ||
3296 error "Set mdc.*${device}*.import=connection=\
3297 $(h2nettype $FAKENID) failed"
3299 do_facet mgs "$PERM_CMD \
3300 ${device}.failover.node=$(h2nettype $FAKENID)" ||
3301 error "Set ${device}.failover.node=\
3302 $(h2nettype $FAKENID) failed"
3305 local at_max_saved=0
3306 # adaptive timeouts may prevent seeing the issue
3307 if at_is_enabled; then
3308 at_max_saved=$(at_max_get mds)
3309 at_max_set 0 mds client
3312 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
3314 log "Injecting EBUSY on MDS"
3315 # Setting OBD_FAIL_MDS_RESEND=0x136
3316 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" ||
3317 error "unable to set param fail_loc=0x80000136"
3319 $LCTL set_param mdc.${FSNAME}*.stats=clear
3321 log "Creating a test file and stat it"
3322 touch $MOUNT/$tdir/$tfile || error "touch $MOUNT/$tdir/$tfile failed"
3323 stat $MOUNT/$tdir/$tfile
3325 log "Stop injecting EBUSY on MDS"
3326 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" ||
3327 error "unable to set param fail_loc=0"
3328 rm -f $MOUNT/$tdir/$tfile || error "remove $MOUNT/$tdir/$tfile failed"
3331 # restore adaptive timeout
3332 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
3334 $LCTL dk $TMP/lustre-log-$TESTNAME.log
3336 CONNCNT=$($LCTL get_param mdc.${FSNAME}*.stats |
3337 awk '/mds_connect/{print $2}')
3339 # retrieve from the log if the client has ever tried to
3340 # contact the fake server after the loss of connection
3341 FAILCONN=`awk "BEGIN {ret = 0;}
3342 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
3344 if (\\\$NF ~ /$FAKENID/) {
3349 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
3351 [ "$FAILCONN" == "0" ] &&
3352 error "The client reconnection has not been triggered"
3353 [ "$FAILCONN" == "2" ] &&
3354 error "Primary server busy, client reconnect to failover failed"
3357 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
3358 # Reconnects are supposed to be rate limited to one every 5s
3359 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] &&
3360 error "Too many reconnects $CONNCNT"
3362 cleanup || error "cleanup failed with $?"
3363 # remove nid settings
3364 writeconf_or_reformat
3366 run_test 35b "Continue reconnection retries, if the active server is busy"
3369 [ $OSTCOUNT -lt 2 ] && skip_env "needs >= 2 OSTs"
3371 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] ||
3375 local FSNAME2=test1234
3376 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3378 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
3380 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
3381 is_blkdev $SINGLEMDS $MDSDEV &&
3382 skip_env "mixed loopback and real device not working"
3385 local fs2mdsdev=$(mdsdevname 1_2)
3386 local fs2ostdev=$(ostdevname 1_2)
3387 local fs3ostdev=$(ostdevname 2_2)
3388 local fs2mdsvdev=$(mdsvdevname 1_2)
3389 local fs2ostvdev=$(ostvdevname 1_2)
3390 local fs3ostvdev=$(ostvdevname 2_2)
3393 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
3394 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
3395 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
3396 # different one than the default value here.
3397 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
3398 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
3399 add fs3ost $(mkfs_opts ost2 ${fs3ostdev}) --mgsnode=$MGSNID \
3400 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
3402 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
3403 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
3404 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
3405 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
3406 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
3408 sleep 5 # until 11778 fixed
3410 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || error "dd failed"
3412 BKTOTAL=$($LCTL get_param -n obdfilter.*.kbytestotal |
3413 awk 'BEGIN{total=0}; {total+=$1}; END{print total}')
3414 BKFREE=$($LCTL get_param -n obdfilter.*.kbytesfree |
3415 awk 'BEGIN{free=0}; {free+=$1}; END{print free}')
3416 BKAVAIL=$($LCTL get_param -n obdfilter.*.kbytesavail |
3417 awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}')
3418 STRING=$(df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}')
3419 DFTOTAL=$(echo $STRING | cut -d, -f1)
3420 DFUSED=$(echo $STRING | cut -d, -f2)
3421 DFAVAIL=$(echo $STRING | cut -d, -f3)
3422 DFFREE=$(($DFTOTAL - $DFUSED))
3424 ALLOWANCE=$((64 * $OSTCOUNT))
3426 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
3427 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
3428 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
3431 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
3432 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
3433 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
3436 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
3437 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
3438 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
3443 stop fs3ost -f || error "unable to stop OST3"
3444 stop fs2ost -f || error "unable to stop OST2"
3445 stop fs2mds -f || error "unable to stop second MDS"
3446 unload_modules_conf || error "unable unload modules"
3449 run_test 36 "df report consistency on OSTs with different block size"
3452 local mntpt=$(facet_mntpt $SINGLEMDS)
3453 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
3454 local mdsdev_sym="$TMP/sym_mdt.img"
3455 local opts=$MDS_MOUNT_OPTS
3458 if [ "$mds1_FSTYPE" != ldiskfs ]; then
3459 skip "ldiskfs only test"
3462 echo "MDS : $mdsdev"
3463 echo "SYMLINK : $mdsdev_sym"
3464 do_facet $SINGLEMDS rm -f $mdsdev_sym
3466 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
3468 echo "mount symlink device - $mdsdev_sym"
3470 if ! do_facet $SINGLEMDS test -b $mdsdev; then
3471 opts=$(csa_add "$opts" -o loop)
3475 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
3476 $mdsdev_sym $mntpt 2>&1)
3479 echo mount_op=$mount_op
3481 do_facet $SINGLEMDS "$UMOUNT $mntpt && rm -f $mdsdev_sym"
3483 if $(echo $mount_op | grep -q "unable to set tunable"); then
3484 error "set tunables failed for symlink device"
3487 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
3489 run_test 37 "verify set tunables works for symlink device"
3491 test_38() { # bug 14222
3492 local mntpt=$(facet_mntpt $SINGLEMDS)
3497 local SRC="/etc /bin"
3498 local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT)
3500 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
3501 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3502 tar cf - $FILES | tar xf - -C $DIR/$tdir ||
3503 error "copying $SRC to $DIR/$tdir"
3505 umount_client $MOUNT || error "umount_client $MOUNT failed"
3506 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3507 stop_mds || error "Unable to stop MDS"
3508 log "delete lov_objid file on MDS"
3510 mount_fstype $SINGLEMDS || error "mount MDS failed (1)"
3512 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid; rm $mntpt/lov_objid"
3514 unmount_fstype $SINGLEMDS || error "umount failed (1)"
3516 # check create in mds_lov_connect
3517 start_mds || error "unable to start MDS"
3518 mount_client $MOUNT || error "mount_client $MOUNT failed"
3520 [ $V ] && log "verifying $DIR/$tdir/$f"
3521 diff -q $f $DIR/$tdir/$f || ERROR=y
3523 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3524 if [ "$ERROR" = "y" ]; then
3525 # check it's updates in sync
3526 umount_client $MOUNT
3528 mount_fstype $SIGNLEMDS
3529 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3530 unmount_fstype $SINGLEMDS
3531 error "old and new files are different after connect" || true
3533 touch $DIR/$tdir/f2 || error "f2 file create failed"
3535 # check it's updates in sync
3536 umount_client $MOUNT || error "second umount_client $MOUNT failed"
3539 mount_fstype $SINGLEMDS || error "mount MDS failed (3)"
3541 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3542 do_facet $SINGLEMDS dd if=/dev/zero of=$mntpt/lov_objid.clear count=8
3544 unmount_fstype $SINGLEMDS || error "umount failed (3)"
3546 start_mds || error "unable to start MDS"
3547 mount_client $MOUNT || error "mount_client $MOUNT failed"
3549 [ $V ] && log "verifying $DIR/$tdir/$f"
3550 diff -q $f $DIR/$tdir/$f || ERROR=y
3552 touch $DIR/$tdir/f3 || error "f3 file create failed"
3553 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3554 umount_client $MOUNT || error "third umount_client $MOUNT failed"
3556 mount_fstype $SINGLEMDS || error "mount MDS failed (4)"
3557 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3558 unmount_fstype $SINGLEMDS || error "umount failed (4)"
3560 [[ "$ERROR" != "y" ]] ||
3561 error "old and new files are different after sync"
3563 log "files compared the same"
3564 cleanup || error "cleanup failed with $?"
3566 run_test 38 "MDS recreates missing lov_objid file from OST data"
3569 [[ -n "$(type -p perl)" ]] || skip_env "need perl for leak_finder.pl"
3573 cleanup || error "cleanup failed with $?"
3574 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
3575 error "memory leak detected" || true
3577 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
3579 test_40() { # bug 15759
3580 start_ost || error "Unable to start OST1"
3581 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
3582 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
3584 cleanup || error "cleanup failed with rc $?"
3586 run_test 40 "race during service thread startup"
3588 test_41a() { #bug 14134
3589 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3590 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
3591 skip "Loop devices does not work with nosvc option"
3594 combined_mgs_mds || skip "needs combined MGT and MDT device"
3596 start_mdt 1 -o nosvc -n
3597 if [ $MDSCOUNT -ge 2 ]; then
3598 for num in $(seq 2 $MDSCOUNT); do
3599 start_mdt $num || return
3602 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
3603 start_mdt 1 -o nomgs,force
3604 mount_client $MOUNT || error "mount_client $MOUNT failed"
3607 echo "blah blah" > $MOUNT/$tfile
3610 umount_client $MOUNT || error "umount_client $MOUNT failed"
3611 stop ost1 -f || error "unable to stop OST1"
3612 stop_mds || error "Unable to stop MDS"
3613 stop_mds || error "Unable to stop MDS on second try"
3615 run_test 41a "mount mds with --nosvc and --nomgs"
3618 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3619 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
3620 skip "Loop devices does not work with nosvc option"
3623 ! combined_mgs_mds && skip "needs combined mgs device"
3627 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3629 start_mdt 1 -o nosvc -n
3630 if [ $MDSCOUNT -ge 2 ]; then
3631 for num in $(seq 2 $MDSCOUNT); do
3632 start_mdt $num || return
3635 start_ost || error "Unable to start OST1"
3636 start_mdt 1 -o nomgs,force
3637 mount_client $MOUNT || error "mount_client $MOUNT failed"
3640 echo "blah blah" > $MOUNT/$tfile
3641 cat $MOUNT/$tfile || error "cat $MOUNT/$tfile failed"
3643 umount_client $MOUNT -f || error "umount_client $MOUNT failed"
3644 stop_ost || error "Unable to stop OST1"
3645 stop_mds || error "Unable to stop MDS"
3646 stop_mds || error "Unable to stop MDS on second try"
3648 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
3651 local oss_list=$(comma_list $(osts_nodes))
3653 [[ "$MDS1_VERSION" -ge $(version_code 2.6.52) ]] ||
3654 [[ "$MDS1_VERSION" -ge $(version_code 2.5.26) &&
3655 "$MDS1_VERSION" -lt $(version_code 2.5.50) ]] ||
3656 [[ "$MDS1_VERSION" -ge $(version_code 2.5.4) &&
3657 "$MDS1_VERSION" -lt $(version_code 2.5.11) ]] ||
3658 skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"
3660 # ensure mds1 ost1 have been created even if running sub-test standalone
3663 cleanup || error "cleanup failed"
3665 # using directly mount command instead of start() function to avoid
3666 # any side effect of // with others/externals tools/features
3667 # ("zpool import", ...)
3669 # MDT concurrent start
3671 LOAD_MODULES_REMOTE=true load_modules
3672 do_facet $SINGLEMDS "lsmod | grep -q libcfs" ||
3673 error "MDT concurrent start: libcfs module not loaded"
3675 local mds1dev=$(mdsdevname 1)
3676 local mds1mnt=$(facet_mntpt mds1)
3677 local mds1opts=$MDS_MOUNT_OPTS
3679 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3680 ! do_facet mds1 test -b $mds1dev; then
3681 mds1opts=$(csa_add "$mds1opts" -o loop)
3683 if [[ "$mds1_FSTYPE" == zfs ]]; then
3684 import_zpool mds1 || return ${PIPESTATUS[0]}
3687 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3688 do_facet mds1 "$LCTL set_param fail_loc=0x80000716"
3690 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts &
3693 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts
3698 do_facet mds1 "$LCTL set_param fail_loc=0x0"
3699 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3700 echo "1st MDT start succeed"
3701 echo "2nd MDT start failed with $rc2"
3702 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3703 echo "1st MDT start failed with $rc"
3704 echo "2nd MDT start succeed"
3707 error "unexpected concurrent MDT mounts result, rc=$rc rc2=$rc2"
3710 if [ $MDSCOUNT -ge 2 ]; then
3711 for num in $(seq 2 $MDSCOUNT); do
3712 start_mdt $num || return
3716 # OST concurrent start
3718 do_rpc_nodes $oss_list "lsmod | grep -q libcfs" ||
3719 error "OST concurrent start: libcfs module not loaded"
3721 local ost1dev=$(ostdevname 1)
3722 local ost1mnt=$(facet_mntpt ost1)
3723 local ost1opts=$OST_MOUNT_OPTS
3725 if [ "$ost1_FSTYPE" == ldiskfs ] &&
3726 ! do_facet ost1 test -b $ost1dev; then
3727 ost1opts=$(csa_add "$ost1opts" -o loop)
3729 if [[ "$ost1_FSTYPE" == zfs ]]; then
3730 import_zpool ost1 || return ${PIPESTATUS[0]}
3733 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3734 do_facet ost1 "$LCTL set_param fail_loc=0x80000716"
3736 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts &
3739 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts
3743 do_facet ost1 "$LCTL set_param fail_loc=0x0"
3744 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3745 echo "1st OST start succeed"
3746 echo "2nd OST start failed with $rc2"
3747 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3748 echo "1st OST start failed with $rc"
3749 echo "2nd OST start succeed"
3753 error "unexpected concurrent OST mounts result, rc=$rc rc2=$rc2"
3759 # verify everything ok
3764 error "MDT(s) start failed"
3772 error "OST(s) start failed"
3780 error "client start failed"
3787 error "client mount failed"
3791 run_test 41c "concurrent mounts of MDT/OST should all fail but one"
3793 test_42() { #bug 14693
3797 check_mount || error "client was not mounted"
3799 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3800 PARAM="llite.$FSNAME-*.some_wrong_param"
3802 PARAM="$FSNAME.llite.some_wrong_param"
3805 do_facet mgs $PERM_CMD $PARAM=10
3806 umount_client $MOUNT ||
3807 error "unmounting client failed with invalid llite param"
3808 mount_client $MOUNT ||
3809 error "mounting client failed with invalid llite param"
3811 do_facet mgs $PERM_CMD $PARAM=20
3812 cleanup || error "stopping $FSNAME failed with invalid sys param"
3814 check_mount || error "client was not mounted with invalid sys param"
3815 cleanup || error "stopping $FSNAME failed with invalid sys param"
3817 run_test 42 "allow client/server mount/unmount with invalid config param"
3820 [[ "$MGS_VERSION" -ge $(version_code 2.5.58) ]] ||
3821 skip "Need MDS version at least 2.5.58"
3822 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
3825 USER1=$(getent passwd | grep :$ID1:$ID1: | cut -d: -f1)
3826 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1"
3829 chmod ugo+x $DIR || error "chmod 0 failed"
3830 set_persistent_param_and_check mds1 \
3831 "mdt.$FSNAME-MDT0000.root_squash" \
3832 "$FSNAME.mdt.root_squash" \
3834 wait_update $HOSTNAME \
3835 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3837 error "check llite root_squash failed!"
3838 set_persistent_param_and_check mds1 \
3839 "mdt.$FSNAME-MDT0000.nosquash_nids" \
3840 "$FSNAME.mdt.nosquash_nids" \
3842 wait_update $HOSTNAME \
3843 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3845 error "check llite nosquash_nids failed!"
3848 # create set of test files
3850 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
3851 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
3852 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
3854 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
3855 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
3857 mkdir_on_mdt0 $DIR/$tdir-rootdir || error "mkdir failed"
3858 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
3859 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
3861 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
3862 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
3863 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
3866 # check root_squash:
3867 # set root squash UID:GID to RUNAS_ID
3868 # root should be able to access only files owned by RUNAS_ID
3870 set_persistent_param_and_check mds1 \
3871 "mdt.$FSNAME-MDT0000.root_squash" \
3872 "$FSNAME.mdt.root_squash" \
3873 "$RUNAS_ID:$RUNAS_ID"
3874 wait_update $HOSTNAME \
3875 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3876 "$RUNAS_ID:$RUNAS_ID" ||
3877 error "check llite root_squash failed!"
3879 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
3880 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3881 error "$ST: root read permission is denied"
3882 echo "$ST: root read permission is granted - ok"
3885 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3886 error "$ST: root write permission is denied"
3887 echo "$ST: root write permission is granted - ok"
3889 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3890 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3891 error "$ST: root read permission is granted"
3892 echo "$ST: root read permission is denied - ok"
3895 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3896 error "$ST: root write permission is granted"
3897 echo "$ST: root write permission is denied - ok"
3899 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3900 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null &&
3901 error "$ST: root unlink permission is granted"
3902 echo "$ST: root unlink permission is denied - ok"
3904 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null &&
3905 error "$ST: root create permission is granted"
3906 echo "$ST: root create permission is denied - ok"
3909 # check root_squash is enforced independently
3910 # of client cache content
3912 # access file by USER1, keep access open
3913 # root should be denied access to user file
3915 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
3919 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
3920 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3921 { kill $pid; error "$ST: root read permission is granted"; }
3922 echo "$ST: root read permission is denied - ok"
3925 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3926 { kill $pid; error "$ST: root write permission is granted"; }
3927 echo "$ST: root write permission is denied - ok"
3933 # check nosquash_nids:
3934 # put client's NID into nosquash_nids list,
3935 # root should be able to access root file after that
3937 local NIDLIST=$($LCTL list_nids all | tr '\n' ' ')
3938 NIDLIST="2@gni $NIDLIST 192.168.0.[2,10]@tcp"
3939 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
3940 set_persistent_param_and_check mds1 \
3941 "mdt.$FSNAME-MDT0000.nosquash_nids" \
3942 "$FSNAME-MDTall.mdt.nosquash_nids" \
3944 wait_update $HOSTNAME \
3945 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3947 error "check llite nosquash_nids failed!"
3949 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3950 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3951 error "$ST: root read permission is denied"
3952 echo "$ST: root read permission is granted - ok"
3955 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3956 error "$ST: root write permission is denied"
3957 echo "$ST: root write permission is granted - ok"
3959 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3960 rm $DIR/$tdir-rootdir/tfile-1 ||
3961 error "$ST: root unlink permission is denied"
3962 echo "$ST: root unlink permission is granted - ok"
3963 touch $DIR/$tdir-rootdir/tfile-2 ||
3964 error "$ST: root create permission is denied"
3965 echo "$ST: root create permission is granted - ok"
3966 cleanup || error "cleanup failed with $?"
3968 run_test 43a "check root_squash and nosquash_nids"
3970 test_43b() { # LU-5690
3971 [[ "$MGS_VERSION" -ge $(version_code 2.7.62) ]] ||
3972 skip "Need MGS version 2.7.62+"
3974 if [[ -z "$fs2mds_DEV" ]]; then
3975 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
3976 skip_env "mixed loopback and real device not working"
3979 local fs2mdsdev=$(mdsdevname 1_2)
3980 local fs2mdsvdev=$(mdsvdevname 1_2)
3982 # temporarily use fs2mds as fs2mgs
3984 local fs2mgsdev=$fs2mdsdev
3985 local fs2mgsvdev=$fs2mdsvdev
3987 local fsname=test1234
3989 load_module llite/lustre
3990 local client_ip=$(host_nids_address $HOSTNAME $NETTYPE)
3991 local host=${client_ip//*./}
3992 local net=${client_ip/%$host/}
3993 local nosquash_nids=$(h2nettype $net[$host,$host,$host])
3995 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
3996 --param mdt.root_squash=$RUNAS_ID:$RUNAS_ID \
3997 --param mdt.nosquash_nids=$nosquash_nids \
3998 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
3999 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS || error "start fs2mgs failed"
4000 stop $fs2mgs -f || error "stop fs2mgs failed"
4001 cleanup || error "cleanup failed with $?"
4003 run_test 43b "parse nosquash_nids with commas in expr_list"
4007 check_mount || error "check_mount"
4008 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
4010 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
4011 for VAL in $UUIDS; do
4012 NID=$(echo $VAL | cut -d= -f1)
4013 CLUUID=$(echo $VAL | cut -d= -f2)
4014 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
4016 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
4017 cleanup || error "cleanup failed with $?"
4019 run_test 44 "mounted client proc entry exists"
4023 check_mount || error "check_mount"
4024 stop_mds || error "Unable to stop MDS"
4028 #define OBD_FAIL_PTLRPC_LONG_REPL_UNLINK 0x50f
4029 do_facet client "$LCTL set_param fail_loc=0x8000050f"
4032 manual_umount_client --force || error "manual_umount_client failed"
4033 do_facet client "$LCTL set_param fail_loc=0x0"
4034 start_mds || error "unable to start MDS"
4035 mount_client $MOUNT || error "mount_client $MOUNT failed"
4036 cleanup || error "cleanup failed with $?"
4038 run_test 45 "long unlink handling in ptlrpcd"
4045 umount_client $MOUNT2 || rc=$?
4046 umount_client $MOUNT || rc=$?
4047 while [ $count -gt 0 ]; do
4048 stop ost${count} -f || rc=$?
4052 cleanup_nocli || rc=$?
4053 #writeconf to remove all ost2 traces for subsequent tests
4054 writeconf_or_reformat
4059 echo "Testing with $OSTCOUNT OSTs"
4061 start_mds || error "unable to start MDS"
4062 #first client should see only one ost
4063 start_ost || error "Unable to start OST1"
4064 wait_osc_import_state mds ost FULL
4066 mount_client $MOUNT || error "mount_client $MOUNT failed"
4067 trap "cleanup_46a $OSTCOUNT" EXIT ERR
4070 for (( i=2; i<=$OSTCOUNT; i++ )); do
4071 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
4072 error "start_ost$i $(ostdevname $i) failed"
4075 # wait until osts in sync
4076 for (( i=2; i<=$OSTCOUNT; i++ )); do
4077 wait_osc_import_state mds ost$i FULL
4078 wait_osc_import_ready client ost$i
4081 #second client see all ost's
4083 mount_client $MOUNT2 || error "mount_client failed"
4084 $LFS setstripe -c -1 $MOUNT2 ||
4085 error "$LFS setstripe -c -1 $MOUNT2 failed"
4086 $LFS getstripe $MOUNT2 || error "$LFS getstripe $MOUNT2 failed"
4088 echo "ok" > $MOUNT2/widestripe
4089 $LFS getstripe $MOUNT2/widestripe ||
4090 error "$LFS getstripe $MOUNT2/widestripe failed"
4091 # fill acl buffer for avoid expand lsm to them
4092 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd |
4094 setfacl -m $acl $MOUNT2/widestripe
4098 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
4100 cleanup_46a $OSTCOUNT || error "cleanup_46a failed"
4102 run_test 46a "handle ost additional - wide striped file"
4107 check_mount || error "check_mount failed"
4108 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
4112 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
4113 if echo $ns | grep "MDT[[:digit:]]*"; then
4116 lrs=$(echo $ns | sed 's/.*lru_size=//')
4117 lru_size[count]=$lrs
4122 facet_failover $SINGLEMDS
4123 client_up || error "client_up failed"
4126 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
4127 if echo $ns | grep "MDT[[:digit:]]*"; then
4130 lrs=$(echo $ns | sed 's/.*lru_size=//')
4131 if ! test "$lrs" -eq "${lru_size[count]}"; then
4132 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
4133 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
4138 cleanup || error "cleanup failed with $?"
4140 run_test 47 "server restart does not make client loss lru_resize settings"
4145 # reformat after this test is needed - if the test fails,
4146 # we will have unkillable file at FS
4150 test_48() { # bz-17636 LU-7473
4154 check_mount || error "check_mount failed"
4156 $LFS setstripe -c -1 $MOUNT ||
4157 error "$LFS setstripe -c -1 $MOUNT failed"
4158 $LFS getstripe $MOUNT || error "$LFS getstripe $MOUNT failed"
4160 echo "ok" > $MOUNT/widestripe
4161 $LFS getstripe $MOUNT/widestripe ||
4162 error "$LFS getstripe $MOUNT/widestripe failed"
4164 # In the future, we may introduce more EAs, such as selinux, enlarged
4165 # LOV EA, and so on. These EA will use some EA space that is shared by
4166 # ACL entries. So here we only check some reasonable ACL entries count,
4167 # instead of the max number that is calculated from the max_ea_size.
4168 if [ "$MDS1_VERSION" -lt $(version_code 2.8.57) ]; then
4169 count=28 # hard coded of RPC protocol
4170 elif large_xattr_enabled; then
4171 count=4500 # max_num 8187 max_ea_size = 65452
4172 # not create too many (4500) to save test time
4174 count=450 # max_num 497 max_ea_size = 4012
4177 echo "It is expected to hold at least $count ACL entries"
4178 trap cleanup_48 EXIT ERR
4179 for ((i = 0; i < $count; i++)) do
4180 setfacl -m u:$((i + 100)):rw $MOUNT/widestripe ||
4181 error "Fail to setfacl for $MOUNT/widestripe at $i"
4184 cancel_lru_locks mdc
4185 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
4186 local r_count=$(getfacl $MOUNT/widestripe | grep "user:" | wc -l)
4188 count=$((count + 1)) # for the entry "user::rw-"
4190 [ $count -eq $r_count ] ||
4191 error "Expected ACL entries $count, but got $r_count"
4195 run_test 48 "too many acls on file"
4197 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
4198 test_49a() { # bug 17710
4199 local timeout_orig=$TIMEOUT
4200 local ldlm_timeout_orig=$LDLM_TIMEOUT
4201 local LOCAL_TIMEOUT=20
4203 LDLM_TIMEOUT=$LOCAL_TIMEOUT
4204 TIMEOUT=$LOCAL_TIMEOUT
4208 check_mount || error "client mount failed"
4210 echo "check ldlm_timout..."
4211 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
4212 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
4213 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
4215 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
4216 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
4219 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
4220 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
4223 umount_client $MOUNT || error "umount_client $MOUNT failed"
4224 stop_ost || error "problem stopping OSS"
4225 stop_mds || error "problem stopping MDS"
4227 LDLM_TIMEOUT=$ldlm_timeout_orig
4228 TIMEOUT=$timeout_orig
4230 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
4232 test_49b() { # bug 17710
4233 local timeout_orig=$TIMEOUT
4234 local ldlm_timeout_orig=$LDLM_TIMEOUT
4235 local LOCAL_TIMEOUT=20
4237 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
4238 TIMEOUT=$LOCAL_TIMEOUT
4242 check_mount || error "client mount failed"
4244 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
4245 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
4246 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
4248 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
4249 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
4252 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
4253 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
4256 cleanup || error "cleanup failed"
4258 LDLM_TIMEOUT=$ldlm_timeout_orig
4259 TIMEOUT=$timeout_orig
4261 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
4264 # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1
4266 # Test both statfs and lfs df and fail if either one fails
4267 multiop_bg_pause $1 f_
4270 killall -USR1 multiop
4271 [ $RC -ne 0 ] && log "lazystatfs multiop failed"
4272 wait $PID || { RC=$?; log "multiop return error "; }
4274 # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1
4279 if kill -s 0 $PID; then
4282 log "lazystatfs lfs df failed to complete in 5s"
4290 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4291 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4293 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
4295 cleanup || error "cleanup failed with rc $?"
4297 run_test 50a "lazystatfs all servers available"
4301 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4302 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4304 # Wait for client to detect down OST
4305 stop_ost || error "Unable to stop OST1"
4306 wait_osc_import_state client ost DISCONN
4308 log "OSCs should all be DISCONN"
4310 lazystatfs $MOUNT || error "lazystatfs should not return EIO"
4312 umount_client $MOUNT || error "Unable to unmount client"
4313 stop_mds || error "Unable to stop MDS"
4315 run_test 50b "lazystatfs all servers down"
4318 start_mds || error "Unable to start MDS"
4319 start_ost || error "Unable to start OST1"
4320 start_ost2 || error "Unable to start OST2"
4321 mount_client $MOUNT || error "Unable to mount client"
4322 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4323 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4325 # Wait for client to detect down OST
4326 stop_ost || error "Unable to stop OST1"
4327 wait_osc_import_state mds ost DISCONN
4328 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
4330 umount_client $MOUNT || error "Unable to unmount client"
4331 stop_ost2 || error "Unable to stop OST2"
4332 stop_mds || error "Unable to stop MDS"
4333 #writeconf to remove all ost2 traces for subsequent tests
4334 writeconf_or_reformat
4336 run_test 50c "lazystatfs one server down"
4339 start_mds || error "Unable to start MDS"
4340 start_ost || error "Unable to start OST1"
4341 start_ost2 || error "Unable to start OST2"
4342 mount_client $MOUNT || error "Unable to mount client"
4343 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4344 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4346 # Issue the statfs during the window where the client still
4347 # belives the OST to be available but it is in fact down.
4348 # No failure just a statfs which hangs for a timeout interval.
4349 stop_ost || error "Unable to stop OST1"
4350 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
4352 umount_client $MOUNT || error "Unable to unmount client"
4353 stop_ost2 || error "Unable to stop OST2"
4354 stop_mds || error "Unable to stop MDS"
4355 #writeconf to remove all ost2 traces for subsequent tests
4356 writeconf_or_reformat
4358 run_test 50d "lazystatfs client/server conn race"
4365 start_mds || error "Unable to start MDS"
4366 #first client should see only one ost
4367 start_ost || error "Unable to start OST1"
4368 wait_osc_import_state mds ost FULL
4370 # Wait for client to detect down OST
4371 stop_ost || error "Unable to stop OST1"
4372 wait_osc_import_state mds ost DISCONN
4374 mount_client $MOUNT || error "Unable to mount client"
4375 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
4377 multiop_bg_pause $MOUNT _f
4381 if [ $RC1 -ne 0 ]; then
4382 log "multiop failed $RC1"
4385 sleep $(( $TIMEOUT+1 ))
4387 [ $? -ne 0 ] && error "process isn't sleep"
4388 start_ost || error "Unable to start OST1"
4389 wait $pid || error "statfs failed"
4392 umount_client $MOUNT || error "Unable to unmount client"
4393 stop_ost || error "Unable to stop OST1"
4394 stop_mds || error "Unable to stop MDS"
4396 run_test 50e "normal statfs all servers down"
4401 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
4403 start_mds || error "Unable to start mds"
4404 #first client should see only one ost
4405 start_ost || error "Unable to start OST1"
4406 wait_osc_import_state mds ost FULL
4408 start_ost2 || error "Unable to start OST2"
4409 wait_osc_import_state mds ost2 FULL
4411 # Wait for client to detect down OST
4412 stop_ost2 || error "Unable to stop OST2"
4414 wait_osc_import_state mds ost2 DISCONN
4415 mount_client $MOUNT || error "Unable to mount client"
4416 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
4418 multiop_bg_pause $MOUNT _f
4422 if [ $RC1 -ne 0 ]; then
4423 log "lazystatfs multiop failed $RC1"
4426 sleep $(( $TIMEOUT+1 ))
4428 [ $? -ne 0 ] && error "process isn't sleep"
4429 start_ost2 || error "Unable to start OST2"
4430 wait $pid || error "statfs failed"
4431 stop_ost2 || error "Unable to stop OST2"
4434 umount_client $MOUNT -f || error "Unable to unmount client"
4435 stop_ost || error "Unable to stop OST1"
4436 stop_mds || error "Unable to stop MDS"
4437 #writeconf to remove all ost2 traces for subsequent tests
4438 writeconf_or_reformat
4440 run_test 50f "normal statfs one server in down"
4443 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs"
4445 start_ost2 || error "Unable to start OST2"
4446 wait_osc_import_state mds ost2 FULL
4447 wait_osc_import_ready client ost2
4449 if [[ $PERM_CMD == *"set_param -P"* ]]; then
4450 local PARAM="osc.${FSNAME}-OST0001*.active"
4452 local PARAM="${FSNAME}-OST0001.osc.active"
4455 $LFS setstripe -c -1 $DIR/$tfile || error "$LFS setstripe failed"
4456 do_facet mgs $PERM_CMD $PARAM=0 || error "Unable to deactivate OST"
4458 umount_client $MOUNT || error "Unable to unmount client"
4459 mount_client $MOUNT || error "Unable to mount client"
4460 # This df should not cause a panic
4463 do_facet mgs $PERM_CMD $PARAM=1 || error "Unable to activate OST"
4464 rm -f $DIR/$tfile || error "unable to remove file $DIR/$tfile"
4465 umount_client $MOUNT || error "Unable to unmount client"
4466 stop_ost2 || error "Unable to stop OST2"
4467 stop_ost || error "Unable to stop OST1"
4468 stop_mds || error "Unable to stop MDS"
4469 #writeconf to remove all ost2 traces for subsequent tests
4470 writeconf_or_reformat
4472 run_test 50g "deactivated OST should not cause panic"
4476 # prepare MDT/OST, make OSC inactive for OST1
4477 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs"
4479 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
4480 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
4481 error "tunefs OST1 failed"
4482 start_mds || error "Unable to start MDT"
4483 start_ost || error "Unable to start OST1"
4484 start_ost2 || error "Unable to start OST2"
4485 mount_client $MOUNT || error "client start failed"
4487 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4489 # activatate OSC for OST1
4490 set_persistent_param_and_check client \
4491 "osc.${FSNAME}-OST0000-osc-[!M]*.active" \
4492 "${FSNAME}-OST0000.osc.active" 1
4494 mkdir $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
4495 $LFS setstripe -c -1 -i 0 $DIR/$tdir/2 ||
4496 error "$LFS setstripe $DIR/$tdir/2 failed"
4497 sleep 1 && echo "create a file after OST1 is activated"
4498 # doing some io, shouldn't crash
4499 dd if=/dev/zero of=$DIR/$tdir/2/$tfile-io bs=1M count=10
4501 # check OSC import is working
4502 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
4503 error "some OSC imports are still not connected"
4507 umount_client $MOUNT || error "Unable to umount client"
4508 stop_ost2 || error "Unable to stop OST2"
4509 cleanup_nocli || error "cleanup_nocli failed with $?"
4511 run_test 50h "LU-642: activate deactivated OST"
4514 # prepare MDT/OST, make OSC inactive for OST1
4515 [ "$MDSCOUNT" -lt "2" ] && skip "needs >= 2 MDTs"
4518 [ $(facet_fstype mds2) == zfs ] && import_zpool mds2
4519 do_facet mds2 "$TUNEFS --param mdc.active=0 $(mdsdevname 2)" ||
4520 error "tunefs MDT2 failed"
4521 start_mds || error "Unable to start MDT"
4522 start_ost || error "Unable to start OST1"
4523 start_ost2 || error "Unable to start OST2"
4524 mount_client $MOUNT || error "client start failed"
4526 mkdir_on_mdt0 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4528 if [[ $PERM_CMD == *"set_param -P"* ]]; then
4529 $PERM_CMD mdc.${FSNAME}-MDT0001-mdc-*.active=0 &&
4530 error "deactive MDC0 succeeds"
4532 $PERM_CMD ${FSNAME}-MDT0000.mdc.active=0 &&
4533 error "deactive MDC0 succeeds"
4536 # activate MDC for MDT2
4537 set_persistent_param_and_check client \
4538 "mdc.${FSNAME}-MDT0001-mdc-*.active" \
4539 "${FSNAME}-MDT0001.mdc.active" 1
4541 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
4542 if [ "$MDS1_VERSION" -ge $(version_code 2.7.60) ]
4544 wait_dne_interconnect
4546 $LFS mkdir -i1 $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
4548 createmany -o $DIR/$tdir/2/$tfile-%d 1 || error "create files failed"
4550 rm -rf $DIR/$tdir/2 || error "unlink dir failed"
4552 # deactivate MDC for MDT2
4553 set_persistent_param_and_check client \
4554 "mdc.${FSNAME}-MDT0001-mdc-*.active" \
4555 "${FSNAME}-MDT0001.mdc.active" 0
4557 wait_osp_active mds ${FSNAME}-MDT0001 1 0
4559 $LFS mkdir -i1 $DIR/$tdir/2 &&
4560 error "mkdir $DIR/$tdir/2 succeeds after deactive MDT"
4562 $LFS mkdir -i0 -c$MDSCOUNT $DIR/$tdir/striped_dir ||
4563 error "mkdir $DIR/$tdir/striped_dir fails after deactive MDT2"
4565 local stripe_count=$($LFS getdirstripe -c $DIR/$tdir/striped_dir)
4566 [ $stripe_count -eq $((MDSCOUNT - 1)) ] ||
4567 error "wrong $stripe_count != $((MDSCOUNT -1)) for striped_dir"
4570 umount_client $MOUNT || error "Unable to umount client"
4575 run_test 50i "activate deactivated MDT"
4578 local LOCAL_TIMEOUT=20
4582 check_mount || error "check_mount failed"
4584 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
4585 $LFS setstripe -c -1 $MOUNT/$tdir ||
4586 error "$LFS setstripe -c -1 $MOUNT/$tdir failed"
4587 #define OBD_FAIL_MDS_REINT_DELAY 0x142
4588 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x142"
4589 touch $MOUNT/$tdir/$tfile &
4592 start_ost2 || error "Unable to start OST1"
4594 stop_ost2 || error "Unable to stop OST1"
4595 umount_client $MOUNT -f || error "unmount $MOUNT failed"
4596 cleanup_nocli || error "stop server failed"
4597 #writeconf to remove all ost2 traces for subsequent tests
4598 writeconf_or_reformat
4600 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
4609 do_node $node mkdir -p $dest
4610 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
4612 do_node $node 'tar cf - '$*' | tar xf - -C '$dest';
4613 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
4614 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
4616 do_node $node 'getfattr -d -m "[a-z]*\\." '$*' > '$xattrs
4617 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
4627 local backup2=${TMP}/backup2
4629 do_node $node mkdir -p $backup2
4630 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
4632 do_node $node 'tar cf - '$*' | tar xf - -C '$backup2';
4633 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
4634 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
4636 do_node $node "diff -rq $backup $backup2"
4637 [ $? -eq 0 ] || { error "contents differ"; return 3; }
4639 local xattrs2=${TMP}/xattrs2
4640 do_node $node 'getfattr -d -m "[a-z]*\\." '$*' > '$xattrs2
4641 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
4643 do_node $node "diff $xattrs $xattrs2"
4644 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
4646 do_node $node "rm -rf $backup2 $xattrs2"
4647 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
4651 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4652 skip "ldiskfs only test"
4655 start_mds || error "Unable to start MDS"
4656 start_ost || error "Unable to start OST1"
4657 mount_client $MOUNT || error "Unable to mount client"
4660 local ost1mnt=$(facet_mntpt ost1)
4661 local ost1node=$(facet_active_host ost1)
4662 local ost1tmp=$TMP/conf52
4665 mkdir $DIR/$tdir || error "Unable to create $DIR/$tdir"
4666 touch $TMP/modified_first || error "Unable to create temporary file"
4667 local mtime=$(stat -c %Y $TMP/modified_first)
4668 do_node $ost1node "mkdir -p $ost1tmp &&
4669 touch -m -d @$mtime $ost1tmp/modified_first" ||
4670 error "Unable to create temporary file"
4673 $LFS setstripe -c -1 -S 1M $DIR/$tdir || error "$LFS setstripe failed"
4675 for (( i=0; i < nrfiles; i++ )); do
4676 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c ||
4677 error "multiop failed"
4682 # sync all the data and make sure no pending data on the client,
4683 # thus the SOM xattr would not be changed any more.
4684 cancel_lru_locks osc
4687 echo backup files to $TMP/$tdir
4688 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
4689 copy_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
4690 error "Unable to copy files"
4692 umount_client $MOUNT || error "Unable to umount client"
4693 stop_ost || error "Unable to stop ost1"
4695 echo mount ost1 as ldiskfs
4696 do_node $ost1node mkdir -p $ost1mnt || error "Unable to create $ost1mnt"
4697 if ! do_node $ost1node test -b $ost1_dev; then
4700 do_node $ost1node mount -t "$ost1_FSTYPE" $loop $ost1_dev \
4702 error "Unable to mount ost1 as ldiskfs"
4705 echo backup objects to $ost1tmp/objects
4706 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
4707 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
4708 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
4710 error "Unable to copy objects"
4712 # move objects to lost+found
4713 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
4714 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
4716 do_node $ost1node "umount $ost1mnt" ||
4717 error "Unable to umount ost1 as ldiskfs"
4719 start_ost || error "Unable to start OST1"
4720 mount_client $MOUNT || error "Unable to mount client"
4722 local REPAIRED=$(do_node $ost1node "$LCTL get_param \
4723 -n osd-ldiskfs.$FSNAME-OST0000.oi_scrub" |
4724 awk '/^lf_repa[ri]*ed/ { print $2 }')
4725 [ $REPAIRED -gt 0 ] ||
4726 error "Some entry under /lost+found should be repaired"
4729 diff_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
4730 error "Unable to diff files"
4732 rm -rf $TMP/$tdir $TMP/file_xattrs ||
4733 error "Unable to delete temporary files"
4734 do_node $ost1node "rm -rf $ost1tmp" ||
4735 error "Unable to delete temporary files"
4736 cleanup || error "cleanup failed with $?"
4738 run_test 52 "check recovering objects from lost+found"
4740 # Checks threads_min/max/started for some service
4742 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
4743 # parameter pattern prefix like 'ost.*.ost'.
4756 local msg="Insane $modname thread counts"
4757 local ncpts=$(check_cpt_number $facet)
4761 check_mount || return 41
4763 # We need to expand $parampat, but it may match multiple parameters, so
4764 # we'll pick the first one
4765 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
4766 error "Couldn't expand ${parampat}.threads_min parameter name"
4770 # Remove the .threads_min part
4771 paramp=${paramp%.threads_min}
4773 # Check for sanity in defaults
4774 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4776 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4778 tstarted=$(do_facet $facet "$LCTL get_param \
4779 -n ${paramp}.threads_started" || echo 0)
4780 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' ||
4782 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4784 nthrs=$(expr $tmax - $tmin)
4785 if [ $nthrs -lt $ncpts ]; then
4791 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
4792 skip_env "module parameter forced $facet thread count"
4794 # Check that we can change min/max
4795 do_facet $facet "$LCTL set_param \
4796 ${paramp}.threads_min=$((tmin + nthrs))"
4797 do_facet $facet "$LCTL set_param \
4798 ${paramp}.threads_max=$((tmax - nthrs))"
4799 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4801 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4803 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) &&
4804 $tmax2 == ($tmax - $nthrs)))' || return $?
4806 # Check that we can set min/max to the same value
4807 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4809 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$tmin"
4810 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4812 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4814 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
4816 # Check that we can't set max < min
4817 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$((tmin - 1))"
4818 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4820 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4822 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
4824 # We need to ensure that we get the module options desired; to do this
4825 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
4826 LOAD_MODULES_REMOTE=true
4829 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
4831 setmodopts -a $modname "$newvalue" oldvalue
4834 check_mount || return 41
4836 # Restore previous setting of MODOPTS_*
4837 setmodopts $modname "$oldvalue"
4839 (( $MDS1_VERSION > $(version_code 2.12.52.91) )) || {
4840 echo "skip interop for MDS < v2_12_52-91-g183cb1e3cdd2"
4844 # Check that $opts took
4845 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4847 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4849 tstarted=$(do_facet $facet \
4850 "$LCTL get_param -n ${paramp}.threads_started" || echo 0)
4851 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4857 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
4858 cleanup || error "cleanup failed with rc $?"
4860 run_test 53a "check OSS thread count params"
4864 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
4865 cleanup || error "cleanup failed with $?"
4867 run_test 53b "check MDS thread count params"
4870 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4871 skip "ldiskfs only test"
4874 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p ||
4875 error "llverdev failed with rc=$?"
4878 run_test 54a "test llverdev and partial verify of device"
4881 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4882 skip "ldiskfs only test"
4886 run_llverfs $MOUNT -p || error "llverfs failed with rc=$?"
4887 cleanup || error "cleanup failed with rc=$?"
4889 run_test 54b "test llverfs and partial verify of filesystem"
4893 local max_ost_index=$1
4895 echo -n $(((max_ost_index + 1) * 8))
4899 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4900 skip "ldiskfs only test"
4903 local mdsdev=$(mdsdevname 1)
4904 local mdsvdev=$(mdsvdevname 1)
4908 if ! combined_mgs_mds; then
4909 stop_mgs || error "stopping MGS service failed"
4910 format_mgs || error "formatting MGT failed"
4912 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
4914 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
4915 --reformat $(ostdevname 1) $(ostvdevname 1)
4921 echo checking size of lov_objid for ost index $i
4922 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" |
4923 grep ^User | awk -F 'Size: ' '{print $2}')
4924 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
4925 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
4927 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
4934 run_test 55 "check lov_objid size"
4937 local mds_journal_size_orig=$MDSJOURNALSIZE
4943 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
4944 $(ostdevname 1) $(ostvdevname 1)
4945 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
4946 $(ostdevname 2) $(ostvdevname 2)
4949 start_ost || error "Unable to start first ost (idx 10000)"
4950 start_ost2 || error "Unable to start second ost (idx 1000)"
4951 mount_client $MOUNT || error "Unable to mount client"
4955 # test instantiating PFL components with sparse index LU-15513
4956 mkdir -p $MOUNT/$tdir
4957 $LFS setstripe -E 4M -c 1 -E 1G -c 4 -S4M -E eof -c -1 $MOUNT/$tdir
4958 dd if=/dev/zero of=$MOUNT/$tdir/$tfile bs=4K count=1 seek=10k ||
4959 error "dd to second component failed"
4961 if [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
4962 [[ "$MDS1_VERSION" -ge $(version_code 2.5.4) &&
4963 "$MDS1_VERSION" -lt $(version_code 2.5.11) ]]; then
4964 wait_osc_import_state mds ost1 FULL
4965 wait_osc_import_state mds ost2 FULL
4966 $LFS setstripe --stripe-count=-1 $DIR/$tfile ||
4967 error "Unable to setstripe $DIR/$tfile"
4968 n=$($LFS getstripe --stripe-count $DIR/$tfile)
4969 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
4974 MDSJOURNALSIZE=$mds_journal_size_orig
4977 run_test 56a "check big OST indexes and out-of-index-order start"
4982 umount_client $MOUNT -f || error "unmount client failed"
4991 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs"
4993 trap cleanup_56b EXIT RETURN ERR
4996 if ! combined_mgs_mds ; then
5001 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) --index=0 --reformat \
5002 $(mdsdevname 1) $(mdsvdevname 1)
5003 add mds2 $(mkfs_opts mds2 $(mdsdevname 2)) --index=1 --reformat \
5004 $(mdsdevname 2) $(mdsvdevname 2)
5005 add mds3 $(mkfs_opts mds3 $(mdsdevname 3)) --index=1000 --reformat \
5006 $(mdsdevname 3) $(mdsvdevname 3)
5010 start_mdt 1 || error "MDT 1 (idx 0) start failed"
5011 start_mdt 2 || error "MDT 2 (idx 1) start failed"
5012 start_mdt 3 || error "MDT 3 (idx 1000) start failed"
5013 start_ost || error "Unable to start first ost"
5014 start_ost2 || error "Unable to start second ost"
5016 do_nodes $(comma_list $(mdts_nodes)) \
5017 "$LCTL set_param mdt.*.enable_remote_dir=1 \
5018 mdt.*.enable_remote_dir_gid=-1"
5020 mount_client $MOUNT || error "Unable to mount client"
5022 $LFS mkdir -c3 $MOUNT/$tdir || error "failed to make testdir"
5024 echo "This is test file 1!" > $MOUNT/$tdir/$tfile.1 ||
5025 error "failed to make test file 1"
5026 echo "This is test file 2!" > $MOUNT/$tdir/$tfile.2 ||
5027 error "failed to make test file 2"
5028 echo "This is test file 1000!" > $MOUNT/$tdir/$tfile.1000 ||
5029 error "failed to make test file 1000"
5031 rm -rf $MOUNT/$tdir || error "failed to remove testdir"
5033 $LFS mkdir -i1000 $MOUNT/$tdir.1000 ||
5034 error "create remote dir at idx 1000 failed"
5037 echo "=== START lfs df OUTPUT ==="
5039 echo "==== END lfs df OUTPUT ===="
5041 mdtcnt=$(echo -e "$output" | grep $FSNAME-MDT | wc -l)
5042 ostcnt=$(echo -e "$output" | grep $FSNAME-OST | wc -l)
5044 echo "lfs df returned mdt count $mdtcnt and ost count $ostcnt"
5045 [ $mdtcnt -eq 3 ] || error "lfs df returned wrong mdt count"
5046 [ $ostcnt -eq 2 ] || error "lfs df returned wrong ost count"
5048 echo "This is test file 1!" > $MOUNT/$tdir.1000/$tfile.1 ||
5049 error "failed to make test file 1"
5050 echo "This is test file 2!" > $MOUNT/$tdir.1000/$tfile.2 ||
5051 error "failed to make test file 2"
5052 echo "This is test file 1000!" > $MOUNT/$tdir.1000/$tfile.1000 ||
5053 error "failed to make test file 1000"
5054 rm -rf $MOUNT/$tdir.1000 || error "failed to remove remote_dir"
5057 echo "=== START lfs mdts OUTPUT ==="
5059 echo "==== END lfs mdts OUTPUT ===="
5061 echo -e "$output" | grep -v "MDTS:" | awk '{print $1}' |
5062 sed 's/://g' > $TMP/mdts-actual.txt
5063 sort $TMP/mdts-actual.txt -o $TMP/mdts-actual.txt
5065 echo -e "0\n1\n1000" > $TMP/mdts-expected.txt
5067 diff $TMP/mdts-expected.txt $TMP/mdts-actual.txt
5070 rm $TMP/mdts-expected.txt $TMP/mdts-actual.txt
5072 [ $result -eq 0 ] || error "target_obd proc file is incorrect!"
5074 run_test 56b "test target_obd correctness with nonconsecutive MDTs"
5076 test_57a() { # bug 22656
5077 do_rpc_nodes $(facet_active_host ost1) load_modules_local
5078 local NID=$(do_facet ost1 "$LCTL get_param nis" |
5079 tail -1 | awk '{print $1}')
5080 writeconf_or_reformat
5081 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
5082 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" ||
5083 error "tunefs failed"
5085 start_ost && error "OST registration from failnode should fail"
5088 run_test 57a "initial registration from failnode should fail (should return errs)"
5091 do_rpc_nodes $(facet_active_host ost1) load_modules_local
5092 local NID=$(do_facet ost1 "$LCTL get_param nis" |
5093 tail -1 | awk '{print $1}')
5094 writeconf_or_reformat
5095 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
5096 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" ||
5097 error "tunefs failed"
5099 start_ost || error "OST registration from servicenode should not fail"
5102 run_test 57b "initial registration from servicenode should not fail"
5105 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
5108 test_58() { # bug 22658
5109 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
5111 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5112 createmany -o $DIR/$tdir/$tfile-%d 100
5113 unlinkmany $DIR/$tdir/$tfile-%d 100
5114 stop_mds || error "Unable to stop MDS"
5116 local MNTDIR=$(facet_mntpt $SINGLEMDS)
5117 local devname=$(mdsdevname ${SINGLEMDS//mds/})
5119 # remove all files from the OBJECTS dir
5120 mount_fstype $SINGLEMDS
5122 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
5124 unmount_fstype $SINGLEMDS
5125 # restart MDS with missing llog files
5126 start_mds || error "unable to start MDS"
5127 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0"
5130 run_test 58 "missing llog files must not prevent MDT from mounting"
5133 start_mgsmds >> /dev/null
5134 local C1=$(count_osts)
5135 if [ $C1 -eq 0 ]; then
5136 start_ost >> /dev/null
5140 echo "original ost count: $C1 (expect > 0)"
5141 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
5142 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
5143 local C2=$(count_osts)
5144 echo "after mdt writeconf count: $C2 (expect 0)"
5145 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
5146 echo "OST start without writeconf should fail:"
5147 start_ost >> /dev/null &&
5148 error "OST start without writeconf didn't fail"
5149 echo "OST start with writeconf should succeed:"
5150 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
5151 local C3=$(count_osts)
5152 echo "after ost writeconf count: $C3 (expect 1)"
5153 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
5154 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
5155 local C4=$(count_osts)
5156 echo "after ost2 writeconf count: $C4 (expect 2)"
5157 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
5158 stop_ost2 >> /dev/null
5159 cleanup_nocli >> /dev/null
5160 #writeconf to remove all ost2 traces for subsequent tests
5161 writeconf_or_reformat
5163 run_test 59 "writeconf mount option"
5165 test_60a() { # LU-471
5166 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5167 skip "ldiskfs only test"
5172 for num in $(seq $MDSCOUNT); do
5173 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
5174 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
5175 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
5179 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
5180 [ ${PIPESTATUS[0]} -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
5182 # MDT default has dirdata feature
5183 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
5184 # we disable uninit_bg feature
5185 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
5186 # we set stride extended options
5187 echo $dump | grep stride > /dev/null || error "stride is not set"
5191 run_test 60a "check mkfs.lustre --mkfsoptions -E -O options setting"
5194 [[ "$mds1_FSTYPE" == ldiskfs ]] || skip "ldiskfs only test"
5196 local features=$(do_facet $SINGLEMDS $DUMPE2FS $(mdsdevname 1) |
5198 [ ${PIPESTATUS[0]} -eq 0 ] || error "$DUMPE2FS $(mdsdevname 1) failed"
5201 # ea_inode feature should be enabled by default for MDTs
5202 [[ "$features" =~ "ea_inode" ]] || error "ea_inode is not set"
5203 # large_dir feature should be enabled by default for MDTs
5204 [[ "$features" =~ "large_dir" ]] || error "large_dir is not set"
5206 run_test 60b "check mkfs.lustre MDT default features"
5208 test_61a() { # LU-80
5209 local lxattr=$(large_xattr_enabled)
5211 (( "$MDS1_VERSION" >= $(version_code 2.1.53) )) ||
5212 skip "Need MDS version at least 2.1.53 for large_xattr"
5214 if [[ "$mds1_FSTYPE" == ldiskfs ]] && ! large_xattr_enabled; then
5217 for ((num=1; num <= $MDSCOUNT; num++)); do
5218 do_facet mds${num} $TUNE2FS -O ea_inode \
5219 $(mdsdevname $num) ||
5220 error "tune2fs on mds $num failed"
5224 setup || error "setting up the filesystem failed"
5225 client_up || error "starting client failed"
5227 local file=$DIR/$tfile
5228 touch $file || error "touch $file failed"
5230 local large_value="$(generate_string $(max_xattr_size))"
5231 local small_value="bar"
5233 local name="trusted.big"
5234 log "save large xattr of $(max_xattr_size) bytes on $name on $file"
5235 setfattr -n $name -v $large_value $file ||
5236 error "saving $name on $file failed"
5238 local new_value=$(get_xattr_value $name $file)
5239 [[ "$new_value" != "$large_value" ]] &&
5240 error "$name different after saving"
5242 log "shrink value of $name on $file"
5243 setfattr -n $name -v $small_value $file ||
5244 error "shrinking value of $name on $file failed"
5246 new_value=$(get_xattr_value $name $file)
5247 [[ "$new_value" != "$small_value" ]] &&
5248 error "$name different after shrinking"
5250 log "grow value of $name on $file"
5251 setfattr -n $name -v $large_value $file ||
5252 error "growing value of $name on $file failed"
5254 new_value=$(get_xattr_value $name $file)
5255 [[ "$new_value" != "$large_value" ]] &&
5256 error "$name different after growing"
5258 log "check value of $name on $file after remounting MDS"
5260 new_value=$(get_xattr_value $name $file)
5261 [[ "$new_value" != "$large_value" ]] &&
5262 error "$name different after remounting MDS"
5264 log "remove large xattr $name from $file"
5265 setfattr -x $name $file || error "removing $name from $file failed"
5267 if $lxattr && [ "$mds1_FSTYPE" == ldiskfs ]; then
5268 stopall || error "stopping for e2fsck run"
5269 for num in $(seq $MDSCOUNT); do
5270 run_e2fsck $(facet_active_host mds$num) \
5271 $(mdsdevname $num) "-y" ||
5272 error "e2fsck MDT$num failed"
5274 setup_noconfig || error "remounting the filesystem failed"
5277 # need to delete this file to avoid problems in other tests
5279 cleanup || error "stopping systems failed"
5281 run_test 61a "large xattr"
5283 test_61b() { # LU-80
5284 local lxattr=$(large_xattr_enabled)
5286 (( $MDS1_VERSION >= $(version_code 2.15.51) )) ||
5287 skip "Need MDS version at least 2.15.51 for large_xattr fix"
5289 [[ "$mds1_FSTYPE" == "ldiskfs" ]] || skip "ldiskfs specific bug"
5291 if ! large_xattr_enabled; then
5294 for (( num=1; num <= $MDSCOUNT; num++ )); do
5295 do_facet mds${num} $TUNE2FS -O ea_inode \
5296 $(mdsdevname $num) ||
5297 error "tune2fs on mds $num failed"
5301 setup || error "setting up the filesystem failed"
5302 client_up || error "starting client failed"
5304 local _file=$MOUNT/panda
5305 local large_value="$(generate_string $(max_xattr_size))"
5306 local name="trusted.big"
5308 touch ${_file} || error "touch ${_file} failed"
5309 setfattr -n $name -v $large_value ${_file} ||
5310 error "saving $name on $file failed"
5312 MDT_DEV="${FSNAME}-MDT0000"
5313 MDT_DEVNAME=$(mdsdevname ${SINGLEMDS//mds/})
5315 stopall || error "stopping for e2fsck run"
5318 ino=$(do_facet $SINGLEMDS "$DEBUGFS -R 'stat /ROOT/panda' \
5319 ${MDT_DEVNAME} | grep trusted.big")
5320 ino=$(echo "${ino}" | awk '{print $2;}')
5321 echo "large ea "${ino}
5323 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"ln $ino /lost+found\\\" \
5326 setup_noconfig || error "remounting the filesystem failed"
5328 do_facet $SINGLEMDS $LCTL lfsck_start -M ${MDT_DEV} -t namespace || {
5329 error "can't start lfsck namespace"
5333 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
5334 mdd.${MDT_DEV}.lfsck_namespace |
5335 awk '/^status/ { print \\\$2 }'" "completed" 32 || {
5336 error "(2) unexpected status"
5339 stopall || error "stopping for e2fsck run"
5340 for num in $(seq $MDSCOUNT); do
5341 run_e2fsck $(facet_active_host mds$num) \
5342 $(mdsdevname $num) "-y" ||
5343 error "e2fsck MDT$num failed"
5345 setup_noconfig || error "remounting the filesystem failed"
5347 # need to delete this file to avoid problems in other tests
5349 cleanup || error "stopping systems failed"
5351 run_test 61b "large xattr"
5354 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5355 skip "ldiskfs only test"
5357 [[ "$MDS1_VERSION" -ge $(version_code 2.2.51) ]] ||
5358 skip "Need MDS version at least 2.2.51"
5361 local mdsdev=$(mdsdevname 1)
5362 local ostdev=$(ostdevname 1)
5364 echo "disable journal for mds"
5365 do_facet mds1 $TUNE2FS -O ^has_journal $mdsdev || error "tune2fs failed"
5366 start_mds && error "MDT start should fail"
5367 echo "disable journal for ost"
5368 do_facet ost1 $TUNE2FS -O ^has_journal $ostdev || error "tune2fs failed"
5369 start_ost && error "OST start should fail"
5370 cleanup || error "cleanup failed with rc $?"
5373 run_test 62 "start with disabled journal"
5376 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5377 skip "ldiskfs only test"
5380 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ldiskfs
5381 local inode_slab=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
5382 awk '/ldiskfs_inode_cache/ { print $5 / $6 }')
5383 if [ -z "$inode_slab" ]; then
5384 skip "ldiskfs module has not been loaded"
5387 if grep -q "CONFIG_DEBUG_LOCK_ALLOC=y" /boot/config-$(uname -r); then
5388 skip "test is not compatible with CONFIG_DEBUG_LOCK_ALLOC=y"
5391 echo "$inode_slab ldiskfs inodes per page"
5392 [ "${inode_slab%.*}" -ge "3" ] && return 0
5394 # If kmalloc-128 is also 1 per page - this is a debug kernel
5395 # and so this is not an error.
5396 local kmalloc128=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
5397 awk '/^(dma-kmalloc|size)-128 / { print $5 / $6 }')
5398 # 32 128-byte chunks in 4k
5399 [ "${kmalloc128%.*}" -lt "32" ] ||
5400 error "ldiskfs inode too big, only $inode_slab objs/page, " \
5401 "kmalloc128 = $kmalloc128 objs/page"
5403 run_test 63 "Verify each page can at least hold 3 ldiskfs inodes"
5406 start_mds || error "unable to start MDS"
5407 start_ost || error "Unable to start OST1"
5408 start_ost2 || error "Unable to start second ost"
5409 mount_client $MOUNT || error "Unable to mount client"
5410 stop_ost2 || error "Unable to stop second ost"
5413 umount_client $MOUNT -f || error "unmount $MOUNT failed"
5414 cleanup_nocli || error "cleanup_nocli failed with $?"
5415 #writeconf to remove all ost2 traces for subsequent tests
5416 writeconf_or_reformat
5418 run_test 64 "check lfs df --lazy "
5420 test_65() { # LU-2237
5421 # Currently, the test is only valid for ldiskfs backend
5422 [ "$mds1_FSTYPE" != ldiskfs ] &&
5423 skip "ldiskfs only test"
5425 local devname=$(mdsdevname ${SINGLEMDS//mds/})
5426 local brpt=$(facet_mntpt brpt)
5429 if ! do_facet $SINGLEMDS "test -b $devname"; then
5433 stop_mds || error "Unable to stop MDS"
5434 local obj=$(do_facet $SINGLEMDS \
5435 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
5437 if [ -z "$obj" ]; then
5438 # The MDT may be just re-formatted, mount the MDT for the
5439 # first time to guarantee the "last_rcvd" file is there.
5440 start_mds || error "fail to mount the MDS for the first time"
5441 stop_mds || error "Unable to stop MDS"
5444 # remove the "last_rcvd" file
5445 do_facet $SINGLEMDS "mkdir -p $brpt"
5446 do_facet $SINGLEMDS \
5447 "mount -t $mds1_FSTYPE $opts $devname $brpt"
5448 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
5449 do_facet $SINGLEMDS "$UMOUNT $brpt"
5451 # restart MDS, the "last_rcvd" file should be recreated.
5452 start_mds || error "fail to restart the MDS"
5453 stop_mds || error "Unable to stop MDS"
5454 obj=$(do_facet $SINGLEMDS \
5455 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
5456 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
5458 run_test 65 "re-create the lost last_rcvd file when server mount"
5461 [[ "$MGS_VERSION" -ge $(version_code 2.3.59) ]] ||
5462 skip "Need MGS version at least 2.3.59"
5464 check_versions || skip "do not replace_nids with mismatched versions"
5467 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
5468 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
5470 # add EXCLUDE records to config log, they are not to be
5471 # removed by lctl replace_nids
5472 set_conf_param_and_check mds \
5473 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
5474 "$FSNAME-OST0000.osc.active" \
5477 echo "replace_nids should fail if MDS, OSTs and clients are UP"
5478 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5479 error "replace_nids fail"
5481 umount_client $MOUNT || error "unmounting client failed"
5482 echo "replace_nids should fail if MDS and OSTs are UP"
5483 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5484 error "replace_nids fail"
5486 stop_ost || error "Unable to stop OST1"
5487 echo "replace_nids should fail if MDS is UP"
5488 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5489 error "replace_nids fail"
5491 stop_mds || error "stopping mds failed"
5493 if combined_mgs_mds; then
5494 start_mdt 1 "-o nosvc" ||
5495 error "starting mds with nosvc option failed"
5498 echo "command should accept two parameters"
5499 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
5500 error "command should accept two params"
5502 echo "correct device name should be passed"
5503 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
5504 error "wrong devname"
5506 echo "wrong nids list should not destroy the system"
5507 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
5509 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "asdfasdf, asdfadf" &&
5512 echo "replace OST nid"
5513 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
5514 error "replace nids failed"
5516 echo "command should accept two parameters"
5517 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
5518 error "command should accept two params"
5520 echo "wrong nids list should not destroy the system"
5521 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
5524 local FAKE_NIDS="192.168.0.112@tcp1,192.168.0.112@tcp2"
5525 local FAKE_FAILOVER="192.168.0.113@tcp1,192.168.0.113@tcp2"
5526 local NIDS_AND_FAILOVER="$MDS_NID,$FAKE_NIDS:$FAKE_FAILOVER"
5527 echo "set NIDs with failover"
5528 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $NIDS_AND_FAILOVER ||
5529 error "replace nids failed"
5532 echo "replace MDS nid"
5533 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
5534 error "replace nids failed"
5536 if ! combined_mgs_mds ; then
5539 stop_mds || error "Unable to stop MDS"
5542 start_mgsmds || error "start mgsmds failed"
5543 set_conf_param_and_check mds \
5544 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
5545 "$FSNAME-OST0000.osc.active" \
5547 start_ost || error "unable to start OST"
5548 mount_client $MOUNT || error "mount client failed"
5550 check_mount || error "error after nid replace"
5551 cleanup || error "cleanup failed"
5554 run_test 66 "replace nids"
5556 test_67() { #LU-2950
5557 local legacy="$TMP/legacy_lnet_config"
5558 local new="$TMP/new_routes_test"
5559 local out="$TMP/config_out_file"
5560 local verify="$TMP/conv_verify"
5561 local verify_conf="$TMP/conf_verify"
5563 # Create the legacy file that will be run through the
5564 # lustre_routes_conversion script
5565 cat <<- LEGACY_LNET_CONFIG > $legacy
5566 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
5567 tcp2 54 10.1.3.2@tcp;
5568 tcp3 10.3.4.3@tcp:3;
5572 # Create the verification file to verify the output of
5573 # lustre_routes_conversion script against.
5574 cat <<- VERIFY_LNET_CONFIG > $verify
5575 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
5576 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
5577 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
5578 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
5579 tcp4: { gateway: 10.3.3.4@tcp }
5582 # Create the verification file to verify the output of
5583 # lustre_routes_config script against
5584 cat <<- VERIFY_LNET_CONFIG > $verify_conf
5585 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
5586 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
5587 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
5588 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
5589 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
5592 $LUSTRE_ROUTES_CONVERSION $legacy $new > /dev/null
5593 if [ -f $new ]; then
5594 # verify the conversion output
5595 cmp -s $new $verify > /dev/null
5596 if [ $? -eq 1 ]; then
5597 error "routes conversion failed"
5600 lustre_routes_config --dry-run --verbose $new > $out
5601 # check that the script succeeded
5602 cmp -s $out $verify_conf > /dev/null
5603 if [ $? -eq 1 ]; then
5604 error "routes config failed"
5607 error "routes conversion test failed"
5609 # remove generated files
5610 rm -f $new $legacy $verify $verify_conf $out
5612 run_test 67 "test routes conversion and configuration"
5620 [ "$MDS1_VERSION" -ge $(version_code 2.4.53) ] ||
5621 skip "Need MDS version at least 2.4.53"
5623 umount_client $MOUNT || error "umount client failed"
5628 # START-END - the sequences we'll be reserving
5629 START=$(do_facet $SINGLEMDS \
5630 $LCTL get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
5631 END=$((START + (1 << 30)))
5632 do_facet $SINGLEMDS \
5633 $LCTL set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
5635 # reset the sequences MDT0000 has already assigned
5636 do_facet $SINGLEMDS \
5637 $LCTL set_param seq.srv*MDT0000.space=clear
5639 # remount to let the client allocate new sequence
5640 mount_client $MOUNT || error "mount client failed"
5642 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
5643 do_facet $SINGLEMDS \
5644 $LCTL get_param seq.srv*MDT0000.space
5645 $LFS path2fid $DIR/$tfile
5647 local old_ifs="$IFS"
5649 fid=($($LFS path2fid $DIR/$tfile))
5653 if [[ $seq < $END ]]; then
5654 error "used reserved sequence $seq?"
5656 cleanup || error "cleanup failed with $?"
5658 run_test 68 "be able to reserve specific sequences in FLDB"
5660 # Test 69: is about the total number of objects ever created on an OST.
5661 # so that when it is reformatted the normal MDS->OST orphan recovery won't
5662 # just "precreate" the missing objects. In the past it might try to recreate
5663 # millions of objects after an OST was reformatted
5665 [[ "$MDS1_VERSION" -lt $(version_code 2.4.2) ]] &&
5666 skip "Need MDS version at least 2.4.2"
5668 [[ "$MDS1_VERSION" -ge $(version_code 2.4.50) ]] &&
5669 [[ "$MDS1_VERSION" -lt $(version_code 2.5.0) ]] &&
5670 skip "Need MDS version at least 2.5.0"
5673 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5674 do_nodes $(comma_list $(osts_nodes)) $LCTL set_param \
5675 seq.*OST*-super.width=$DATA_SEQ_MAX_WIDTH
5677 # use OST0000 since it probably has the most creations
5678 local OSTNAME=$(ostname_from_index 0)
5679 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
5680 local last_id=$(do_facet mds1 $LCTL get_param -n \
5681 osp.$mdtosc_proc1.prealloc_last_id)
5683 # Want to have OST LAST_ID over 5 * OST_MAX_PRECREATE to
5684 # verify that the LAST_ID recovery is working properly. If
5685 # not, then the OST will refuse to allow the MDS connect
5686 # because the LAST_ID value is too different from the MDS
5687 #define OST_MAX_PRECREATE=20000
5688 local ost_max_pre=20000
5689 local num_create=$(( ost_max_pre * 5 + 1 - last_id))
5691 # If the LAST_ID is already over 5 * OST_MAX_PRECREATE, we don't
5692 # need to create any files. So, skip this section.
5693 if [ $num_create -gt 0 ]; then
5694 # Check the number of inodes available on OST0
5696 local ifree=$($LFS df -i $MOUNT |
5697 awk '/OST0000/ { print $4 }'; exit ${PIPESTATUS[0]})
5698 log "On OST0, $ifree inodes available. Want $num_create. rc=$?"
5700 $LFS setstripe -i 0 $DIR/$tdir ||
5701 error "$LFS setstripe -i 0 $DIR/$tdir failed"
5702 if [ $ifree -lt 10000 ]; then
5703 files=$(( ifree - 50 ))
5708 local j=$((num_create / files + 1))
5709 for i in $(seq 1 $j); do
5710 createmany -o $DIR/$tdir/$tfile-$i- $files ||
5711 error "createmany fail create $files files: $?"
5712 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
5713 error "unlinkmany failed unlink $files files"
5717 # delete all of the files with objects on OST0 so the
5718 # filesystem is not inconsistent later on
5719 $LFS find $MOUNT --ost 0 -print0 | xargs -0 rm
5721 umount_client $MOUNT || error "umount client failed"
5722 stop_ost || error "OST0 stop failure"
5723 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
5724 $(ostdevname 1) $(ostvdevname 1) ||
5725 error "reformat and replace $ostdev failed"
5726 start_ost || error "OST0 restart failure"
5727 wait_osc_import_state mds ost FULL
5729 mount_client $MOUNT || error "mount client failed"
5730 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
5731 local idx=$($LFS getstripe -i $DIR/$tdir/$tfile-last)
5732 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
5734 local iused=$($LFS df -i $MOUNT |
5735 awk '/OST0000/ { print $3 }'; exit ${PIPESTATUS[0]})
5736 log "On OST0, $iused used inodes rc=$?"
5737 [ $iused -ge $((ost_max_pre + 1000)) ] &&
5738 error "OST replacement created too many inodes; $iused"
5739 cleanup || error "cleanup failed with $?"
5741 run_test 69 "replace an OST with the same index"
5744 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5747 cleanup || error "cleanup failed with $?"
5749 start_mdt 1 || error "MDT0 start fail"
5751 start_ost || error "OST0 start fail"
5752 for num in $(seq 2 $MDSCOUNT); do
5753 start_mdt $num || return
5756 mount_client $MOUNT || error "mount client fails"
5758 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
5760 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5761 error "create remote dir fail"
5763 rm -rf $DIR/$tdir || error "delete dir fail"
5764 cleanup || error "cleanup failed with $?"
5766 run_test 70a "start MDT0, then OST, then MDT1"
5769 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5772 start_ost || error "OST0 start fail"
5774 start_mds || error "MDS start fail"
5776 mount_client $MOUNT || error "mount client fails"
5778 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
5780 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5781 error "create remote dir fail"
5783 rm -rf $DIR/$tdir || error "delete dir fail"
5785 cleanup || error "cleanup failed with $?"
5787 run_test 70b "start OST, MDT1, MDT0"
5790 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5793 start_mds || error "MDS start fail"
5794 start_ost || error "OST0 start fail"
5796 mount_client $MOUNT || error "mount client fails"
5797 stop_mdt 1 || error "MDT1 start fail"
5799 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
5800 echo "deactivate $mdc_for_mdt1"
5801 $LCTL --device $mdc_for_mdt1 deactivate ||
5802 error "set $mdc_for_mdt1 deactivate failed"
5804 mkdir $DIR/$tdir && error "mkdir succeed"
5806 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
5807 error "create remote dir succeed"
5809 cleanup || error "cleanup failed with $?"
5811 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
5814 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5817 start_mds || error "MDS start fail"
5818 start_ost || error "OST0 start fail"
5820 mount_client $MOUNT || error "mount client fails"
5822 stop_mdt 2 || error "MDT1 start fail"
5824 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
5826 echo "deactivate $mdc_for_mdt2"
5827 $LCTL --device $mdc_for_mdt2 deactivate ||
5828 error "set $mdc_for_mdt2 deactivate failed"
5830 mkdir $DIR/$tdir || error "mkdir fail"
5831 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
5832 error "create remote dir succeed"
5834 rm -rf $DIR/$tdir || error "delete dir fail"
5836 cleanup || error "cleanup failed with $?"
5838 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
5841 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5843 [ "$MDS1_VERSION" -ge $(version_code 2.7.62) ] ||
5844 skip "Need MDS version at least 2.7.62"
5846 reformat || error "reformat failed with $?"
5850 local mdsdev=$(mdsdevname 1)
5851 local ostdev=$(ostdevname 1)
5852 local mdsvdev=$(mdsvdevname 1)
5853 local ostvdev=$(ostvdevname 1)
5854 local opts_mds="$(mkfs_opts mds1 $mdsdev) --reformat $mdsdev $mdsvdev"
5855 local opts_ost="$(mkfs_opts ost1 $ostdev) --reformat $ostdev $ostvdev"
5857 if ! combined_mgs_mds ; then
5861 add mds1 $opts_mds || error "add mds1 failed"
5862 start_mdt 1 || error "start mdt1 failed"
5863 add ost1 $opts_ost || error "add ost1 failed"
5864 start_ost || error "start ost failed"
5865 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
5867 local soc=$(do_facet mds1 "$LCTL get_param -n \
5868 mdt.*MDT0000.sync_lock_cancel")
5869 [ $soc == "never" ] || error "SoC enabled on single MDS"
5871 for i in $(seq 2 $MDSCOUNT); do
5872 mdsdev=$(mdsdevname $i)
5873 mdsvdev=$(mdsvdevname $i)
5874 opts_mds="$(mkfs_opts mds$i $mdsdev) --reformat $mdsdev \
5876 add mds$i $opts_mds || error "add mds$i failed"
5877 start_mdt $i || error "start mdt$i fail"
5880 wait_dne_interconnect
5882 for i in $(seq $MDSCOUNT); do
5883 soc=$(do_facet mds$i "$LCTL get_param -n \
5884 mdt.*MDT000$((i - 1)).sync_lock_cancel")
5885 [ $soc == "blocking" ] || error "SoC not enabled on DNE"
5888 for i in $(seq 2 $MDSCOUNT); do
5889 stop_mdt $i || error "stop mdt$i fail"
5891 soc=$(do_facet mds1 "$LCTL get_param -n \
5892 mdt.*MDT0000.sync_lock_cancel")
5893 [ $soc == "never" ] || error "SoC enabled on single MDS"
5894 umount_client $MOUNT -f > /dev/null
5896 cleanup || error "cleanup failed with $?"
5898 run_test 70e "Sync-on-Cancel will be enabled by default on DNE"
5901 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5902 if combined_mgs_mds; then
5903 skip "needs separate MGS/MDT"
5907 start_mdt 1 || error "MDT0 start fail"
5908 start_ost || error "OST0 start fail"
5909 for num in $(seq 2 $MDSCOUNT); do
5910 start_mdt $num || return
5913 start_ost2 || error "OST1 start fail"
5915 mount_client $MOUNT || error "mount client fails"
5917 mkdir $DIR/$tdir || error "mkdir fail"
5918 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5919 error "create remote dir succeed"
5921 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5922 rm -rf $DIR/$tdir || error "delete dir fail"
5924 umount_client $MOUNT || error "umount_client failed"
5925 stop_mds || error "MDS stop fail"
5926 stop_ost || error "OST0 stop fail"
5927 stop_ost2 || error "OST1 stop fail"
5929 run_test 71a "start MDT0 OST0, MDT1, OST1"
5932 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5933 if combined_mgs_mds; then
5934 skip "needs separate MGS/MDT"
5938 for num in $(seq 2 $MDSCOUNT); do
5939 start_mdt $num || return
5941 start_ost || error "OST0 start fail"
5942 start_mdt 1 || error "MDT0 start fail"
5943 start_ost2 || error "OST1 start fail"
5945 mount_client $MOUNT || error "mount client fails"
5947 mkdir $DIR/$tdir || error "mkdir fail"
5948 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5949 error "create remote dir succeed"
5951 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5952 rm -rf $DIR/$tdir || error "delete dir fail"
5954 umount_client $MOUNT || error "umount_client failed"
5955 stop_mds || error "MDT0 stop fail"
5956 stop_ost || error "OST0 stop fail"
5957 stop_ost2 || error "OST1 stop fail"
5959 run_test 71b "start MDT1, OST0, MDT0, OST1"
5962 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5963 combined_mgs_mds && skip "needs separate MGS/MDT"
5967 start_ost || error "OST0 start fail"
5968 start_ost2 || error "OST1 start fail"
5969 for num in $(seq 2 $MDSCOUNT); do
5970 start_mdt $num || return
5972 start_mdt 1 || error "MDT0 start fail"
5974 mount_client $MOUNT || error "mount client fails"
5976 mkdir $DIR/$tdir || error "mkdir fail"
5977 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5978 error "create remote dir succeed"
5980 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5981 rm -rf $DIR/$tdir || error "delete dir fail"
5983 umount_client $MOUNT || error "umount_client failed"
5984 stop_mds || error "MDS stop fail"
5985 stop_ost || error "OST0 stop fail"
5986 stop_ost2 || error "OST1 stop fail"
5989 run_test 71c "start OST0, OST1, MDT1, MDT0"
5992 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5993 combined_mgs_mds && skip "needs separate MGS/MDT"
5997 start_ost || error "OST0 start fail"
5998 for num in $(seq 2 $MDSCOUNT); do
5999 start_mdt $num || return
6001 start_mdt 1 || error "MDT0 start fail"
6002 start_ost2 || error "OST1 start fail"
6004 mount_client $MOUNT || error "mount client fails"
6006 mkdir $DIR/$tdir || error "mkdir fail"
6007 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
6008 error "create remote dir succeed"
6010 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
6011 rm -rf $DIR/$tdir || error "delete dir fail"
6013 umount_client $MOUNT || error "umount_client failed"
6014 stop_mds || error "MDS stop fail"
6015 stop_ost || error "OST0 stop fail"
6016 stop_ost2 || error "OST1 stop fail"
6019 run_test 71d "start OST0, MDT1, MDT0, OST1"
6022 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
6023 combined_mgs_mds && skip "needs separate MGS/MDT"
6027 start_ost || error "OST0 start fail"
6028 for num in $(seq 2 $MDSCOUNT); do
6029 start_mdt $num || return
6031 start_ost2 || error "OST1 start fail"
6032 start_mdt 1 || error "MDT0 start fail"
6034 mount_client $MOUNT || error "mount client fails"
6036 mkdir $DIR/$tdir || error "mkdir fail"
6037 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
6038 error "create remote dir succeed"
6040 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
6041 rm -rf $DIR/$tdir || error "delete dir fail"
6043 umount_client $MOUNT || error "umount_client failed"
6044 stop_mds || error "MDS stop fail"
6045 stop_ost || error "OST0 stop fail"
6046 stop_ost2 || error "OST1 stop fail"
6049 run_test 71e "start OST0, MDT1, OST1, MDT0"
6051 test_72() { #LU-2634
6052 [ "$mds1_FSTYPE" != ldiskfs ] &&
6053 skip "ldiskfs only test"
6055 local mdsdev=$(mdsdevname 1)
6056 local ostdev=$(ostdevname 1)
6057 local cmd="$E2FSCK -fnvd $mdsdev"
6064 if combined_mgs_mds; then
6065 add_options='--reformat'
6067 add_options='--reformat --replace'
6070 #tune MDT with "-O extents"
6072 for num in $(seq $MDSCOUNT); do
6073 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
6074 $add_options $(mdsdevname $num) $(mdsvdevname $num) ||
6075 error "add mds $num failed"
6076 do_facet mds${num} "$TUNE2FS -O extents $(mdsdevname $num)" ||
6077 error "$TUNE2FS failed on mds${num}"
6080 add ost1 $(mkfs_opts ost1 $ostdev) $add_options $ostdev ||
6081 error "add $ostdev failed"
6082 start_mds || error "start mds failed"
6083 start_ost || error "start ost failed"
6084 mount_client $MOUNT || error "mount client failed"
6086 #create some short symlinks
6087 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6088 createmany -o $DIR/$tdir/$tfile-%d $fn
6089 echo "create $fn short symlinks"
6090 for i in $(seq -w 1 $fn); do
6091 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
6096 umount_client $MOUNT || error "umount client failed"
6097 stop_mds || error "stop mds failed"
6098 stop_ost || error "stop ost failed"
6101 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
6103 run_test 72 "test fast symlink with extents flag enabled"
6105 test_73() { #LU-3006
6106 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
6107 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
6108 error "1st tunefs failed"
6109 start_mgsmds || error "start mds failed"
6110 start_ost || error "start ost failed"
6111 mount_client $MOUNT || error "mount client failed"
6112 $LCTL get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
6113 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
6114 umount_client $MOUNT || error "umount client failed"
6118 run_test 73 "failnode to update from mountdata properly"
6122 (( $MDS1_VERSION >= $(version_code 2.15.57.16) )) ||
6123 skip "need MDS version >= 2.15.57.16 for per-device timeouts"
6126 stack_trap "cleanup"
6128 # Prepare fs2, share the mgs of fs
6129 local FSNAME2=fs15246
6130 local fs2mdsdev=$(mdsdevname 1_2)
6131 local fs2ostdev=$(ostdevname 1_2)
6132 local fs2mdsvdev=$(mdsvdevname 1_2)
6133 local fs2ostvdev=$(ostvdevname 1_2)
6135 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --fsname=$FSNAME2 \
6136 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
6137 add fs2ost $(mkfs_opts ost1 $fs2ostdev) --fsname=$FSNAME2 \
6138 --reformat $fs2ostdev $fs2ostvdev || error "add fs2ost failed"
6140 stack_trap "cleanup_fs2"
6142 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
6143 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS || error "start fs2mds failed"
6145 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
6146 $MOUNT_CMD $MGSNID:/$FSNAME2 $MOUNT2 || error "mount $MOUNT2 failed"
6148 echo "========== All mounted lustre fs ===================="
6149 mount | grep 'type lustre'
6150 echo "====================================================="
6152 # Set and check osc/ldlm_enqueue_min
6153 $LCTL set_param osc.${FSNAME}-*.ldlm_enqueue_min=99 ||
6154 error "failed to set per-device adaptive parameters on client"
6155 stack_trap "$LCTL set_param osc.${FSNAME}-*.ldlm_enqueue_min=0"
6157 local ldlm_enqueue_min
6158 ldlm_enqueue_min=$($LCTL get_param -n osc.${FSNAME}-*.ldlm_enqueue_min \
6160 (( $ldlm_enqueue_min == 99 )) ||
6161 error "wrong ldlm_enqueue_min value for osc.${FSNAME}-*"
6163 # Check fs2 as reference
6164 ldlm_enqueue_min=$($LCTL get_param -n osc.${FSNAME2}-*.ldlm_enqueue_min\
6166 (( $ldlm_enqueue_min == 0 )) ||
6167 error "wrong ldlm_enqueue_min value for osc.${FSNAME2}-*"
6169 # Set and check obdfilter/at_min
6170 do_facet ost1 $LCTL set_param obdfilter.${FSNAME}-*.at_min=1 ||
6171 error "failed to set per-device adaptive parameters on ost"
6172 stack_trap "do_facet ost1 $LCTL set_param obdfilter.${FSNAME}-*.at_min=0"
6175 at_min=$(do_facet ost1 $LCTL get_param -n obdfilter.${FSNAME}-*.at_min \
6177 (( $at_min == 1 )) ||
6178 error "wrong at_min value for obdfilter.${FSNAME}-*"
6180 # set and check mdc/at_max
6181 $LCTL set_param mdc.${FSNAME}-*.at_max=599 ||
6182 error "failed to set per-device adaptive parameters on client"
6183 stack_trap "$LCTL set_param mdc.${FSNAME}-*.at_max=0"
6186 at_max=$($LCTL get_param -n mdc.${FSNAME}-*.at_max | uniq)
6187 (( $at_max == 599 )) ||
6188 error "wrong at_max value for osc.${FSNAME}-*"
6190 # Check fs2 as reference
6191 at_max=$($LCTL get_param -n mdc.${FSNAME2}-*.at_max | uniq)
6192 (( $at_max == 0 )) ||
6193 error "wrong at_max value for osc.${FSNAME2}-*"
6195 # Set and check mds/at_max
6196 do_facet mds1 $LCTL set_param *.${FSNAME}-*.at_max=599 ||
6197 error "failed to set per-device adaptive parameters on mds"
6198 stack_trap "do_facet mds1 $LCTL set_param *.${FSNAME}-*.at_max=0"
6201 at_max=$(do_facet mds1 $LCTL get_param -n mdt.${FSNAME}-*.at_max | uniq)
6202 (( at_max == 599 )) ||
6203 error "wrong at_max value for mdt.${FSNAME}-*"
6205 # Set and check mgs&mgc/at_history
6206 local mgs_nid=$(do_facet $SINGLEMDS $LCTL list_nids | tail -1)
6207 $LCTL set_param mgc.MGC$mgs_nid.at_history=588
6208 stack_trap "$LCTL set_param mgc.MGC$mgs_nid.at_history=0"
6211 at_history=$($LCTL get_param -n mgc.MGC$mgs_nid.at_history)
6212 (( $at_history == 588 )) ||
6213 error "wrong at_history value for mgc.MGC$mgs_nid"
6215 do_facet mgs $LCTL set_param mgs.MGS.at_history=588
6216 stack_trap "do_facet mgs $LCTL set_param mgs.MGS.at_history=0"
6217 at_history=$(do_facet mgs $LCTL get_param -n mgs.MGS.at_history)
6218 (( $at_history == 588 )) ||
6219 error "wrong at_history value for mgs.MGS"
6221 run_test 74 "Test per-device adaptive timeout parameters"
6223 test_75() { # LU-2374
6224 [[ "$MDS1_VERSION" -lt $(version_code 2.4.1) ]] &&
6225 skip "Need MDS version at least 2.4.1"
6228 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
6229 --replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
6230 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
6231 --replace --reformat $(ostdevname 1) $(ostvdevname 1)"
6234 #check with default parameters
6235 add mds1 $opts_mds || error "add mds1 failed for default params"
6236 add ost1 $opts_ost || error "add ost1 failed for default params"
6238 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
6239 opts_mds=$(echo $opts_mds |
6240 sed -e "s/--index=$index/--index=$index --mdt/")
6241 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
6242 opts_ost=$(echo $opts_ost |
6243 sed -e "s/--index=$index/--index=$index --ost/")
6245 add mds1 $opts_mds || error "add mds1 failed for new params"
6246 add ost1 $opts_ost || error "add ost1 failed for new params"
6251 run_test 75 "The order of --index should be irrelevant"
6254 [[ "$MGS_VERSION" -ge $(version_code 2.4.52) ]] ||
6255 skip "Need MDS version at least 2.4.52"
6258 local MDMB_PARAM="osc.*.max_dirty_mb"
6259 echo "Change MGS params"
6260 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
6262 echo "max_dirty_mb: $MAX_DIRTY_MB"
6263 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB - 10))
6264 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
6265 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
6266 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
6267 head -1" $NEW_MAX_DIRTY_MB
6268 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
6269 echo "$MAX_DIRTY_MB"
6270 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
6271 error "error while apply max_dirty_mb"
6273 echo "Check the value is stored after remount"
6276 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
6277 head -1" $NEW_MAX_DIRTY_MB
6278 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
6279 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
6280 error "max_dirty_mb is not saved after remount"
6282 echo "Change OST params"
6283 CLIENT_PARAM="obdfilter.*.client_cache_count"
6284 local CLIENT_CACHE_COUNT
6285 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
6287 echo "client_cache_count: $CLIENT_CACHE_COUNT"
6288 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
6289 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
6290 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
6291 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
6292 head -1" $NEW_CLIENT_CACHE_COUNT
6293 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
6295 echo "$CLIENT_CACHE_COUNT"
6296 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
6297 error "error while apply client_cache_count"
6299 echo "Check the value is stored after remount"
6302 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
6303 head -1" $NEW_CLIENT_CACHE_COUNT
6304 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
6306 echo "$CLIENT_CACHE_COUNT"
6307 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
6308 error "client_cache_count is not saved after remount"
6311 run_test 76a "set permanent params with lctl across mounts"
6313 test_76b() { # LU-4783
6314 [[ "$MGS_VERSION" -ge $(version_code 2.5.57) ]] ||
6315 skip "Need MGS version at least 2.5.57"
6318 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
6319 error "start params log failed"
6322 run_test 76b "verify params log setup correctly"
6325 [[ "$MGS_VERSION" -ge $(version_code 2.8.54) ]] ||
6326 skip "Need MDS version at least 2.4.52"
6328 local MASK_PARAM="mdd.*.changelog_mask"
6329 echo "Change changelog_mask"
6330 do_facet mgs $LCTL set_param -P $MASK_PARAM=-CLOSE ||
6331 error "Can't change changlog_mask"
6332 wait_update $(facet_host mds) "$LCTL get_param -n $MASK_PARAM |
6335 echo "Check the value is stored after mds remount"
6336 stop_mds || error "Failed to stop MDS"
6337 start_mds || error "Failed to start MDS"
6338 local CHANGELOG_MASK=$(do_facet mgs $LCTL get_param -n $MASK_PARAM)
6339 echo $CHANGELOG_MASK | grep CLOSE > /dev/null &&
6340 error "changelog_mask is not changed"
6344 run_test 76c "verify changelog_mask is applied with lctl set_param -P"
6346 test_76d() { #LU-9399
6349 local xattr_cache="llite.*.xattr_cache"
6350 local cmd="$LCTL get_param -n $xattr_cache | head -1"
6351 local new=$((($(eval $cmd) + 1) % 2))
6353 echo "lctl set_param -P llite.*.xattr_cache=$new"
6354 do_facet mgs $LCTL set_param -P $xattr_cache=$new ||
6355 error "Can't change xattr_cache"
6356 wait_update $HOSTNAME "$cmd" "$new"
6358 echo "Check $xattr_cache on client $MOUNT"
6359 umount_client $MOUNT || error "umount $MOUNT failed"
6360 mount_client $MOUNT || error "mount $MOUNT failed"
6361 [ $(eval $cmd) -eq $new ] ||
6362 error "$xattr_cache != $new on client $MOUNT"
6364 echo "Check $xattr_cache on the new client $MOUNT2"
6365 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
6366 [ $(eval $cmd) -eq $new ] ||
6367 error "$xattr_cache != $new on client $MOUNT2"
6368 umount_client $MOUNT2 || error "umount $MOUNT2 failed"
6372 run_test 76d "verify llite.*.xattr_cache can be set by 'lctl set_param -P' correctly"
6374 test_77() { # LU-3445
6375 [[ "$MDS1_VERSION" -ge $(version_code 2.8.55) ]] ||
6376 skip "Need MDS version 2.8.55+ "
6378 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
6379 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
6380 skip_env "mixed loopback and real device not working"
6383 local fs2mdsdev=$(mdsdevname 1_2)
6384 local fs2ostdev=$(ostdevname 1_2)
6385 local fs2mdsvdev=$(mdsvdevname 1_2)
6386 local fs2ostvdev=$(ostvdevname 1_2)
6387 local fsname=test1234
6389 local failnid="$(h2nettype 1.2.3.4),$(h2nettype 4.3.2.1)"
6391 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
6393 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
6394 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
6395 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
6396 error "start fs2mds failed"
6398 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
6399 mgsnid="0.0.0.0@tcp,$mgsnid,$mgsnid:$mgsnid"
6401 add fs2ost --mgsnode=$mgsnid $(mkfs_opts ost1 $fs2ostdev) \
6402 --failnode=$failnid --fsname=$fsname \
6403 --reformat $fs2ostdev $fs2ostvdev ||
6404 error "add fs2ost failed"
6405 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
6407 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
6408 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
6409 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
6412 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
6415 [[ "$mds1_FSTYPE" != ldiskfs ||
6416 "$ost1_FSTYPE" != ldiskfs ]] &&
6417 skip "ldiskfs only test"
6419 # reformat the Lustre filesystem with a smaller size
6420 local saved_MDSCOUNT=$MDSCOUNT
6421 local saved_MDSSIZE=$MDSSIZE
6422 local saved_OSTCOUNT=$OSTCOUNT
6423 local saved_OSTSIZE=$OSTSIZE
6426 MDSSIZE=$((MDSSIZE - 20000))
6427 OSTSIZE=$((OSTSIZE - 20000))
6428 reformat || error "(1) reformat Lustre filesystem failed"
6429 MDSSIZE=$saved_MDSSIZE
6430 OSTSIZE=$saved_OSTSIZE
6432 # mount the Lustre filesystem
6433 setup_noconfig || error "(2) setup Lustre filesystem failed"
6436 log "create test files"
6441 mkdir $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
6443 for i in $(seq $num_files); do
6444 file=$MOUNT/$tdir/$tfile-$i
6445 dd if=/dev/urandom of=$file count=1 bs=1M || {
6446 $LCTL get_param osc.*.cur*grant*
6447 $LFS df; $LFS df -i;
6448 # stop creating files if there is no more space
6449 if [ ! -e $file ]; then
6450 num_files=$((i - 1))
6454 $LFS getstripe -v $file
6455 local ost_idx=$(LFS getstripe -i $file)
6456 do_facet ost$((ost_idx + 1)) \
6457 $LCTL get_param obdfilter.*.*grant*
6458 error "(4) create $file failed"
6462 # unmount the Lustre filesystem
6463 cleanup || error "(5) cleanup Lustre filesystem failed"
6465 # run e2fsck on the MDT and OST devices
6466 local mds_host=$(facet_active_host $SINGLEMDS)
6467 local ost_host=$(facet_active_host ost1)
6468 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
6469 local ost_dev=$(ostdevname 1)
6471 run_e2fsck $mds_host $mds_dev "-y"
6472 run_e2fsck $ost_host $ost_dev "-y"
6474 # get the original block count of the MDT and OST filesystems
6475 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
6476 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
6478 # expand the MDT and OST filesystems to the device size
6479 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
6480 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
6482 # run e2fsck on the MDT and OST devices again
6483 run_e2fsck $mds_host $mds_dev "-y"
6484 run_e2fsck $ost_host $ost_dev "-y"
6486 # mount the Lustre filesystem
6490 log "check files after expanding the MDT and OST filesystems"
6491 for i in $(seq $num_files); do
6492 file=$MOUNT/$tdir/$tfile-$i
6493 $CHECKSTAT -t file -s 1048576 $file ||
6494 error "(6) checkstat $file failed"
6498 log "create more files after expanding the MDT and OST filesystems"
6499 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
6500 file=$MOUNT/$tdir/$tfile-$i
6501 dd if=/dev/urandom of=$file count=1 bs=1M ||
6502 error "(7) create $file failed"
6505 # unmount the Lustre filesystem
6506 cleanup || error "(8) cleanup Lustre filesystem failed"
6508 # run e2fsck on the MDT and OST devices
6509 run_e2fsck $mds_host $mds_dev "-y"
6510 run_e2fsck $ost_host $ost_dev "-y"
6512 # get the maximum block count of the MDT and OST filesystems
6513 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
6514 local ost_max_blks=$(get_block_count ost1 $ost_dev)
6516 # get the minimum block count of the MDT and OST filesystems
6517 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
6518 grep minimum | sed -e 's/^.*filesystem: //g')
6519 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
6520 grep minimum | sed -e 's/^.*filesystem: //g')
6522 # shrink the MDT and OST filesystems to a smaller size
6526 if [[ $mds_max_blks -gt $mds_min_blks &&
6527 $mds_max_blks -gt $mds_orig_blks ]]; then
6528 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
6529 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
6530 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
6531 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
6532 error "shrink $SINGLEMDS to $new_blks failed"
6536 if [[ $ost_max_blks -gt $ost_min_blks &&
6537 $ost_max_blks -gt $ost_orig_blks ]]; then
6538 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
6539 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
6540 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
6541 run_resize2fs ost1 $ost_dev $new_blks ||
6542 error "shrink ost1 to $new_blks failed"
6546 # check whether the MDT or OST filesystem was shrunk or not
6548 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
6549 reformat_and_config ||
6550 error "(10) reformat Lustre filesystem failed"
6554 # run e2fsck on the MDT and OST devices again
6555 run_e2fsck $mds_host $mds_dev "-y"
6556 run_e2fsck $ost_host $ost_dev "-y"
6558 # mount the Lustre filesystem again
6562 log "check files after shrinking the MDT and OST filesystems"
6563 for i in $(seq $((num_files + 10))); do
6564 file=$MOUNT/$tdir/$tfile-$i
6565 $CHECKSTAT -t file -s 1048576 $file ||
6566 error "(11) checkstat $file failed"
6569 # unmount and reformat the Lustre filesystem
6570 cleanup || error "(12) cleanup Lustre filesystem failed"
6571 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
6573 MDSCOUNT=$saved_MDSCOUNT
6574 OSTCOUNT=$saved_OSTCOUNT
6575 reformat_and_config || error "(14) reformat Lustre filesystem failed"
6577 run_test 78 "run resize2fs on MDT and OST filesystems"
6579 test_79() { # LU-4227
6580 [[ "$MDS1_VERSION" -ge $(version_code 2.5.59) ]] ||
6581 skip "Need MDS version at least 2.5.59"
6583 local mdsdev1=$(mdsdevname 1)
6584 local mdsvdev1=$(mdsvdevname 1)
6585 local mdsdev2=$(mdsdevname 2)
6586 local mdsvdev2=$(mdsvdevname 2)
6587 local ostdev1=$(ostdevname 1)
6588 local ostvdev1=$(ostvdevname 1)
6589 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
6590 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
6591 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
6594 # remove --mgs/--mgsnode from mkfs.lustre options
6595 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
6597 mgsnode_opt=$(echo $opts_mds2 |
6598 awk '{ for ( i = 1; i < NF; i++ )
6599 if ( $i ~ "--mgsnode" ) { print $i; break } }')
6600 [ -n "$mgsnode_opt" ] &&
6601 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
6603 mgsnode_opt=$(echo $opts_ost1 |
6604 awk '{ for ( i = 1; i < NF; i++ )
6605 if ( $i ~ "--mgsnode" ) { print $i; break } }')
6606 [ -n "$mgsnode_opt" ] &&
6607 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
6609 # -MGS, format a mdt without --mgs option
6610 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
6611 error "Must specify --mgs when formatting mdt combined with mgs"
6613 # +MGS, format a mdt/ost without --mgsnode option
6614 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
6615 > /dev/null || error "start mds1 failed"
6616 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
6617 error "Must specify --mgsnode when formatting a mdt"
6618 add ost1 $opts_ost1 $ostdev1 $ostvdev1 &&
6619 error "Must specify --mgsnode when formatting an ost"
6623 run_test 79 "format MDT/OST without mgs option (should return errors)"
6626 start_mds || error "Failed to start MDT"
6627 start_ost || error "Failed to start OST1"
6628 uuid=$(do_facet ost1 $LCTL get_param -n mgc.*.uuid)
6629 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
6630 do_facet ost1 "$LCTL set_param fail_val=10 fail_loc=0x906"
6631 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x906"
6632 do_facet mgs "$LCTL set_param -n mgs/MGS/evict_client $uuid"
6634 start_ost2 || error "Failed to start OST2"
6636 do_facet ost1 "$LCTL set_param fail_loc=0"
6641 run_test 80 "mgc import reconnect race"
6643 #Save the original values of $OSTCOUNT and $OSTINDEX$i.
6645 local new_ostcount=$1
6646 saved_ostcount=$OSTCOUNT
6647 OSTCOUNT=$new_ostcount
6651 for ((i = 1; i <= $OSTCOUNT; i++ )); do
6653 eval saved_ostindex$i=${!index}
6658 # Restore the original values of $OSTCOUNT and $OSTINDEX$i.
6659 restore_ostindex() {
6663 echo "restoring OSTCOUNT=$saved_ostcount and OSTINDEXn"
6665 for ((i = 1; i <= $OSTCOUNT; i++ )); do
6666 index=saved_ostindex$i
6667 eval OSTINDEX$i=${!index}
6669 OSTCOUNT=$saved_ostcount
6672 if ! combined_mgs_mds ; then
6677 # The main purpose of this test is to ensure the OST_INDEX_LIST functions as
6678 # expected. This test uses OST_INDEX_LIST to format OSTs with a randomly
6679 # assigned index and ensures we can mount such a formatted file system
6680 test_81() { # LU-4665
6681 (( MDS1_VERSION >= $(version_code 2.6.54) )) ||
6682 skip "Need MDS version at least 2.6.54"
6683 (( OSTCOUNT >= 3 )) || skip_env "needs >= 3 OSTs"
6687 # Each time RANDOM is referenced, a random integer between 0 and 32767
6690 local saved_ostindex1=$OSTINDEX1
6691 for i in 65535 $((RANDOM + 65536)); do
6692 echo -e "\nFormat ost1 with --index=$i, should fail"
6694 if add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat \
6695 $(ostdevname 1) $(ostvdevname 1); then
6696 OSTINDEX1=$saved_ostindex1
6697 error "format ost1 with --index=$i should fail"
6700 OSTINDEX1=$saved_ostindex1
6703 stack_trap restore_ostindex
6705 # Format OSTs with random sparse indices.
6706 local rand_ost=$(((RANDOM * 2 % 65533) + 1))
6707 echo "Format $OSTCOUNT OSTs with OST_INDEX_LIST=[0,$rand_ost,65534]"
6708 OST_INDEX_LIST=[0,$rand_ost,65534] formatall ||
6709 error "formatall failed with $?"
6711 # Setup and check Lustre filesystem.
6712 start_mgsmds || error "start_mgsmds failed"
6713 for ((i = 1; i <= $OSTCOUNT; i++ )); do
6714 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6715 error "start ost$i failed"
6718 mount_client $MOUNT || error "mount client $MOUNT failed"
6719 check_mount || error "check client $MOUNT failed"
6723 local max_easize=$($LCTL get_param -n llite.*.max_easize)
6724 local xattr_size_max=$((4096 - 84)) # 4096 less ldiskfs ea overhead
6726 # XATTR_SIZE_MAX less ldiskfs ea overhead
6727 large_xattr_enabled && xattr_size_max=$((65536 - 84))
6728 (( max_easize >= xattr_size_max )) ||
6729 error "max_easize $max_easize < $xattr_size_max bytes"
6731 test_mkdir $DIR/$tdir
6732 $LFS setstripe -i $rand_ost $DIR/$tdir/$tfile ||
6733 error "error creating $tfile on ost$rand_ost"
6734 $LFS getstripe $DIR/$tdir/$tfile
6741 cmd="getstripe -i $DIR/$tdir/$tfile"
6744 (( $found == $rand_ost )) || error "index $found is not $rand_ost"
6746 cmd="find $DIR/$tdir -i $rand_ost"
6750 [[ "$found" == "$DIR/$tdir/$tfile" ]] ||
6751 error "'lfs find' returned '$found', not '$tfile' by index"
6754 uuid=$(ostuuid_from_index $rand_ost)
6756 cmd="find $DIR/$tdir -O $uuid"
6761 [[ "$found" == "$DIR/$tdir/$tfile" ]] ||
6762 error "'lfs find' returned '$found', not '$tfile' by UUID"
6764 run_test 81 "sparse OST indexing"
6766 # Here we exercise the stripe placement functionality on a file system that
6767 # has formatted the OST with a random index. With the file system the following
6768 # functionality is tested:
6770 # 1. Creating a new file with a specific stripe layout.
6772 # 2. Modifiy a existing empty file with a specific stripe layout.
6774 # 3. Ensure we fail to set the stripe layout of a file that already has one.
6776 # 4. If ost-index is defined we need to ensure it is the first entry in the
6777 # ost index list returned by lfs getstripe.
6779 # 5. Lastly ensure this functionality fails with directories.
6780 test_82a() { # LU-4665
6781 [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
6782 skip "Need MDS version at least 2.6.54"
6783 [[ $OSTCOUNT -ge 3 ]] || skip_env "needs >= 3 OSTs"
6789 # Format OSTs with random sparse indices.
6793 local LOV_V1_INSANE_STRIPE_COUNT=65532
6794 for i in $(seq $OSTCOUNT); do
6795 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
6796 ost_indices+=" $index"
6798 ost_indices=$(comma_list $ost_indices)
6800 stack_trap "restore_ostindex" EXIT
6801 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
6802 OST_INDEX_LIST=[$ost_indices] formatall
6804 # Setup Lustre filesystem.
6805 start_mgsmds || error "start_mgsmds failed"
6806 for i in $(seq $OSTCOUNT); do
6807 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6808 error "start ost$i failed"
6811 # Collect debug information - start of test
6812 do_nodes $(comma_list $(mdts_nodes)) \
6813 $LCTL get_param osp.*.prealloc_*_id
6815 mount_client $MOUNT || error "mount client $MOUNT failed"
6819 check_lfs_df_ret_val $? || error "$LFS df $MOUNT failed"
6820 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6822 stack_trap "do_nodes $(comma_list $(mdts_nodes)) \
6823 $LCTL get_param osp.*.prealloc_*_id || true" EXIT
6825 # 1. If the file does not exist, new file will be created
6826 # with specified OSTs.
6827 local file=$DIR/$tdir/$tfile-1
6828 local cmd="$LFS setstripe -o $ost_indices $file"
6830 eval $cmd || error "$cmd failed"
6831 check_stripe_count $file $OSTCOUNT
6832 check_obdidx $file $ost_indices
6833 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6834 error "write $file failed"
6836 # 2. If the file already exists and is an empty file, the file
6837 # will be attached with specified layout.
6838 file=$DIR/$tdir/$tfile-2
6839 mcreate $file || error "mcreate $file failed"
6840 cmd="$LFS setstripe -o $ost_indices $file"
6842 eval $cmd || error "$cmd failed"
6843 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6844 error "write $file failed"
6845 check_stripe_count $file $OSTCOUNT
6846 check_obdidx $file $ost_indices
6848 # 3. If the file already has a valid layout attached, the command
6849 # should fail with EBUSY.
6851 eval $cmd && error "stripe is already set on $file, $cmd should fail"
6853 # 4. If [--stripe-index|-i <start_ost_idx>] is used, the index must
6854 # be in the OST indices list.
6855 local start_ost_idx=${ost_indices##*,}
6856 file=$DIR/$tdir/$tfile-3
6857 cmd="$LFS setstripe -o $ost_indices -i $start_ost_idx $file"
6859 eval $cmd || error "$cmd failed"
6860 check_stripe_count $file $OSTCOUNT
6861 check_obdidx $file $ost_indices
6862 check_start_ost_idx $file $start_ost_idx
6864 file=$DIR/$tdir/$tfile-4
6865 cmd="$LFS setstripe"
6866 cmd+=" -o $(exclude_items_from_list $ost_indices $start_ost_idx)"
6867 cmd+=" -i $start_ost_idx $file"
6869 eval $cmd && error "index $start_ost_idx should be in $ost_indices"
6871 # 5. Specifying OST indices for directory should succeed.
6872 local dir=$DIR/$tdir/$tdir
6873 mkdir $dir || error "mkdir $dir failed"
6874 cmd="$LFS setstripe -o $ost_indices $dir"
6875 if [ "$MDS1_VERSION" -gt $(version_code 2.11.53) ] &&
6876 [ "$CLIENT_VERSION" -gt $(version_code 2.11.53) ]; then
6878 eval $cmd || error "unable to specify OST indices on directory"
6880 echo "need MDS+client version at least 2.11.53"
6883 run_test 82a "specify OSTs for file (succeed) or directory (succeed)"
6885 # Test 82b is run to ensure that if the user supplies a pool with a specific
6886 # stripe layout that it behaves proprerly. It should fail in the case that
6887 # the supplied OST index list points to OSTs not contained in the user
6889 test_82b() { # LU-4665
6890 [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
6891 skip "Need MDS version at least 2.6.54"
6892 [[ $OSTCOUNT -ge 4 ]] || skip_env "needs >= 4 OSTs"
6898 # Format OSTs with random sparse indices.
6902 local LOV_V1_INSANE_STRIPE_COUNT=65532
6903 for i in $(seq $OSTCOUNT); do
6904 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
6905 ost_indices+=" $index"
6907 ost_indices=$(comma_list $ost_indices)
6909 stack_trap "restore_ostindex" EXIT
6910 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
6911 OST_INDEX_LIST=[$ost_indices] formatall
6913 # Setup Lustre filesystem.
6914 start_mgsmds || error "start_mgsmds failed"
6915 for i in $(seq $OSTCOUNT); do
6916 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6917 error "start ost$i failed"
6920 mount_client $MOUNT || error "mount client $MOUNT failed"
6924 check_lfs_df_ret_val $? || error "$LFS df $MOUNT failed"
6925 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6927 # Create a new pool and add OSTs into it.
6928 local ost_pool=$FSNAME.$TESTNAME
6929 create_pool $ost_pool || error "create OST pool $ost_pool failed"
6931 local ost_idx_in_list=${ost_indices##*,}
6932 local ost_idx_in_pool=$(exclude_items_from_list $ost_indices \
6935 local ost_targets="$FSNAME-OST["
6936 for i in ${ost_idx_in_pool//,/ }; do
6937 ost_targets=$ost_targets$(printf "%04x," $i)
6939 ost_targets="${ost_targets%,}]"
6941 local ost_targets_uuid=$(for i in ${ost_idx_in_pool//,/ }; \
6942 do printf "$FSNAME-OST%04x_UUID\n" $i; done |
6943 sort -u | tr '\n' ' ')
6945 local cmd="$LCTL pool_add $ost_pool $ost_targets"
6946 do_facet mgs $cmd || error "$cmd failed"
6947 wait_update $HOSTNAME "$LCTL get_param -n lov.$FSNAME-*.pools.$TESTNAME|
6948 sort -u | tr '\n' ' ' " "$ost_targets_uuid" ||
6949 error "wait_update $ost_pool failed"
6950 wait_update_facet $SINGLEMDS "$LCTL pool_list $ost_pool | wc -l" 4 ||
6951 error "wait_update pool_list $ost_pool failed"
6953 # If [--pool|-p <pool_name>] is set with [--ost-list|-o <ost_indices>],
6954 # then the OSTs must be the members of the pool.
6955 local file=$DIR/$tdir/$tfile
6956 cmd="$LFS setstripe -p $ost_pool -o $ost_idx_in_list $file"
6958 eval $cmd && error "OST with index $ost_idx_in_list should be" \
6959 "in OST pool $ost_pool"
6961 # Only select OST $ost_idx_in_list from $ost_pool for file.
6962 ost_idx_in_list=${ost_idx_in_pool#*,}
6963 cmd="$LFS setstripe -p $ost_pool -o $ost_idx_in_list $file"
6965 eval $cmd || error "$cmd failed"
6966 cmd="$LFS getstripe $file"
6968 eval $cmd || error "$cmd failed"
6969 check_stripe_count $file 2
6970 check_obdidx $file $ost_idx_in_list
6971 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6972 error "write $file failed"
6974 run_test 82b "specify OSTs for file with --pool and --ost-list options"
6977 [[ "$OST1_VERSION" -ge $(version_code 2.6.91) ]] ||
6978 skip "Need OST version at least 2.6.91"
6979 if [ "$ost1_FSTYPE" != ldiskfs ]; then
6980 skip "ldiskfs only test"
6989 ostmnt=$(facet_mntpt ost1)
6991 # Mount the OST as an ldiskfs filesystem.
6992 log "mount the OST $dev as a $ost1_FSTYPE filesystem"
6993 add ost1 $(mkfs_opts ost1 $dev) $FSTYPE_OPT \
6994 --reformat $dev > /dev/null ||
6995 error "format ost1 error"
6997 if ! test -b $dev; then
6998 mnt_opts=$(csa_add "$OST_MOUNT_FS_OPTS" -o loop)
7000 echo "mnt_opts $mnt_opts"
7001 do_facet ost1 mount -t "$ost1_FSTYPE" $dev \
7003 # Run llverfs on the mounted ldiskfs filesystem.
7004 # It is needed to get ENOSPACE.
7005 log "run llverfs in partial mode on the OST $ost1_FSTYPE $ostmnt"
7006 do_rpc_nodes $(facet_host ost1) run_llverfs $ostmnt -vpl \
7007 "no" || error "run_llverfs error on $ost1_FSTYPE"
7010 log "unmount the OST $dev"
7013 # Delete file IO_scrub. Later osd_scrub_setup will try to
7014 # create "IO_scrub" but will get ENOSPACE.
7016 echo "start ost1 service on `facet_active_host ost1`"
7017 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
7020 err=$(do_facet ost1 dmesg | grep "VFS: Busy inodes after unmount of")
7021 echo "string err $err"
7022 [ -z "$err" ] || error $err
7025 run_test 83 "ENOSPACE on OST doesn't cause message VFS: \
7026 Busy inodes after unmount ..."
7029 local facet=$SINGLEMDS
7030 local num=$(echo $facet | tr -d "mds")
7031 local dev=$(mdsdevname $num)
7032 local time_min=$(recovery_time_min)
7033 local recovery_duration
7034 local completed_clients
7035 local correct_clients
7038 echo "start mds service on $(facet_active_host $facet)"
7040 "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
7041 error "start MDS failed"
7043 start_ost || error "start OST0000 failed"
7044 wait_osc_import_state mds ost1 FULL
7045 start_ost2 || error "start OST0001 failed"
7046 wait_osc_import_state mds ost2 FULL
7048 echo "recovery_time=$time_min, timeout=$TIMEOUT, wrap_up=$wrap_up"
7050 mount_client $MOUNT1 || error "mount $MOUNT1 failed"
7051 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
7052 # make sure new superblock labels are sync'd before disabling writes
7056 replay_barrier $SINGLEMDS
7057 createmany -o $DIR1/$tfile-%d 1000
7059 # We need to catch the end of recovery window to extend it.
7060 # Skip 5 requests and add delay to request handling.
7061 #define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP
7062 do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5"
7064 facet_failover --fsck $SINGLEMDS || error "failover: $?"
7067 echo "recovery status"
7068 do_facet $SINGLEMDS \
7069 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status"
7071 recovery_duration=$(do_facet $SINGLEMDS \
7072 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
7073 awk '/recovery_duration/ { print $2 }')
7074 (( $recovery_duration > $time_min + $wrap_up )) &&
7075 error "recovery_duration > recovery_time_hard + wrap up"
7076 completed_clients=$(do_facet $SINGLEMDS \
7077 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
7078 awk '/completed_clients/ { print $2 }')
7080 correct_clients="$MDSCOUNT/$((MDSCOUNT+1))"
7081 [ "$completed_clients" = "${correct_clients}" ] ||
7082 error "$completed_clients != $correct_clients"
7084 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
7085 umount_client $MOUNT1
7086 umount_client $MOUNT2
7092 run_test 84 "check recovery_hard_time"
7095 [[ "$OST1_VERSION" -ge $(version_code 2.7.55) ]] ||
7096 skip "Need OST version at least 2.7.55"
7097 ##define OBD_FAIL_OSD_OST_EA_FID_SET 0x197
7098 do_facet ost1 "lctl set_param fail_loc=0x197"
7102 run_test 85 "osd_ost init: fail ea_fid_set"
7107 # ost1 has already registered to the MGS before the reformat.
7108 # So after reformatting it with option "-G", it could not be
7109 # mounted to the MGS. Cleanup the system for subsequent tests.
7114 [ "$ost1_FSTYPE" = zfs ] &&
7115 skip "LU-6442: no such mkfs params for ZFS OSTs"
7116 [[ "$MDS1_VERSION" -ge $(version_code 2.7.56) ]] ||
7117 skip "Need server version newer than 2.7.55"
7120 local OLDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
7121 awk '/Flex block group size: / { print $NF; exit; }')
7123 [ "$OLDSIZE" == "$NEWSIZE" ] && skip "$NEWSIZE groups already"
7125 local opts=" -O flex_bg -G $NEWSIZE"
7126 opts=$(OST_FS_MKFS_OPTS+="$opts" mkfs_opts ost1 $(ostdevname 1))
7127 opts+=" --reformat $(ostdevname 1) $(ostvdevname 1)"
7128 echo "params: $opts"
7130 trap cleanup_86 EXIT ERR
7133 add ost1 $opts || error "add ost1 failed with new params"
7135 local FOUNDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
7136 awk '/Flex block group size: / { print $NF; exit; }')
7138 [[ $FOUNDSIZE == $NEWSIZE ]] ||
7139 error "Flex block group size: $FOUNDSIZE, expected: $NEWSIZE"
7143 run_test 86 "Replacing mkfs.lustre -G option"
7145 test_87() { #LU-6544
7146 [[ "$MDS1_VERSION" -ge $(version_code 2.9.51) ]] ||
7147 skip "Need MDS version at least 2.9.51"
7148 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
7149 skip "ldiskfs only test"
7150 [[ $OSTCOUNT -gt 59 ]] &&
7151 skip "Ignore wide striping situation"
7152 [ -n "$FILESET" ] && skip "Not functional for FILESET set"
7154 local mdsdev=$(mdsdevname 1)
7155 local mdsvdev=$(mdsvdevname 1)
7156 local file=$DIR/$tfile
7157 local mntpt=$(facet_mntpt $SINGLEMDS)
7158 local used_xattr_blk=0
7159 local inode_size=${1:-1024}
7161 local xtest="trusted.test"
7165 local stripe_cnt=$(($OSTCOUNT + 2))
7167 #Please see ldiskfs_make_lustre() for MDT inode size calculation
7168 if [ $stripe_cnt -gt 16 ]; then
7171 left_size=$(expr $inode_size - \
7174 32 - 40 \* 3 - 32 \* 3 - $stripe_cnt \* 24 - 16 - 3 - \
7176 24 - 18 - $(expr length $tfile) - 16 - 4)
7177 if [ $left_size -le 0 ]; then
7178 echo "No space($left_size) is expected in inode."
7179 echo "Try 1-byte xattr instead to verify this."
7182 echo "Estimate: at most $left_size-byte space left in inode."
7188 if ! combined_mgs_mds ; then
7192 add mds1 $(mkfs_opts mds1 ${mdsdev}) --stripe-count-hint=$stripe_cnt \
7193 --reformat $mdsdev $mdsvdev || error "add mds1 failed"
7194 start_mdt 1 > /dev/null || error "start mdt1 failed"
7195 for i in $(seq $OSTCOUNT); do
7196 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS > /dev/null ||
7197 error "start ost$i failed"
7199 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
7200 check_mount || error "check client $MOUNT failed"
7203 $LFS setstripe -E 1M -S 1M -c 1 -E 64M -c 1 -E -1 -c -1 $file ||
7204 error "Create file with 3 components failed"
7205 $TRUNCATE $file $((1024*1024*64+1)) || error "truncate file failed"
7206 i=$($LFS getstripe -I3 -c $file) || error "get 3rd stripe count failed"
7207 if [ $i -ne $OSTCOUNT ]; then
7208 left_size=$(expr $left_size + $(expr $OSTCOUNT - $i) \* 24)
7209 echo -n "Since only $i out $OSTCOUNT OSTs are used, "
7210 echo -n "the expected left space is changed to "
7211 echo "$left_size bytes at most."
7213 value=$(generate_string $left_size)
7214 setfattr -n $xtest -v $value $file
7215 orig=$(get_xattr_value $xtest $file)
7216 [[ "$orig" != "$value" ]] && error "$xtest changed"
7218 #Verify if inode has some expected space left
7219 umount $MOUNT > /dev/null || error "umount $MOUNT failed"
7220 stop_mdt 1 > /dev/null || error "stop mdt1 failed"
7221 mount_ldiskfs $SINGLEMDS || error "mount -t ldiskfs $SINGLEMDS failed"
7223 do_facet $SINGLEMDS ls -sal $mntpt/ROOT/$tfile
7224 used_xattr_blk=$(do_facet $SINGLEMDS ls -s $mntpt/ROOT/$tfile |
7226 [[ $used_xattr_blk -eq 0 ]] &&
7227 error "Please check MDS inode size calculation: \
7228 more than $left_size-byte space left in inode."
7229 echo "Verified: at most $left_size-byte space left in inode."
7231 unmount_ldiskfs $SINGLEMDS
7233 for i in $(seq $OSTCOUNT); do
7234 stop ost$i -f || error "stop ost$i failed"
7237 run_test 87 "check if MDT inode can hold EAs with N stripes properly"
7240 [ "$mds1_FSTYPE" == zfs ] &&
7241 skip "LU-6662: no implementation for ZFS"
7245 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
7246 --reformat $(mdsdevname 1) || error "add mds1 failed"
7248 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
7249 grep -e \".*opts:.*errors=remount-ro.*\"" ||
7250 error "default mount options is missing"
7252 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
7253 --mountfsoptions="user_xattr,errors=panic" \
7254 --reformat $(mdsdevname 1) || error "add mds1 failed"
7256 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
7257 grep -e \".*opts:.*errors=panic.*\"" ||
7258 error "user can't override default mount options"
7260 run_test 88 "check the default mount options can be overridden"
7262 test_89() { # LU-7131
7263 [[ "$MDS1_VERSION" -ge $(version_code 2.9.54) ]] ||
7264 skip "Need MDT version at least 2.9.54"
7266 local key=failover.node
7267 local val1=192.0.2.254@tcp0 # Reserved IPs, see RFC 5735
7268 local val2=192.0.2.255@tcp0
7269 local mdsdev=$(mdsdevname 1)
7274 if [[ "$mds1_FSTYPE" == zfs ]]; then
7275 import_zpool mds1 || return ${PIPESTATUS[0]}
7278 # Check that parameters are added correctly
7279 echo "tunefs --param $key=$val1"
7280 do_facet mds1 "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
7281 error "tunefs --param $key=$val1 failed"
7282 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7283 error "tunefs --dryrun failed"
7284 params=${params##*Parameters:}
7285 params=${params%%exiting*}
7286 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val1) = "1" ] ||
7287 error "on-disk parameter not added correctly via tunefs"
7289 # Check that parameters replace existing instances when added
7290 echo "tunefs --param $key=$val2"
7291 do_facet mds1 "$TUNEFS --param $key=$val2 $mdsdev >/dev/null" ||
7292 error "tunefs --param $key=$val2 failed"
7293 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7294 error "tunefs --dryrun failed"
7295 params=${params##*Parameters:}
7296 params=${params%%exiting*}
7297 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "1" ] ||
7298 error "on-disk parameter not replaced via tunefs"
7299 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val2) = "1" ] ||
7300 error "on-disk parameter not replaced correctly via tunefs"
7302 # Check that a parameter is erased properly
7303 echo "tunefs --erase-param $key"
7304 do_facet mds1 "$TUNEFS --erase-param $key $mdsdev >/dev/null" ||
7305 error "tunefs --erase-param $key failed"
7306 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7307 error "tunefs --dryrun failed"
7308 params=${params##*Parameters:}
7309 params=${params%%exiting*}
7310 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "0" ] ||
7311 error "on-disk parameter not erased correctly via tunefs"
7313 # Check that all the parameters are erased
7314 do_facet mds1 "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
7315 error "tunefs --param $key=$val1 failed"
7316 echo "tunefs --erase-params"
7317 do_facet mds1 "$TUNEFS --erase-params $mdsdev >/dev/null" ||
7318 error "tunefs --erase-params failed"
7319 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7320 error "tunefs --dryrun failed"
7321 params=${params##*Parameters:}
7322 params=${params%%exiting*}
7323 params=$(echo $params | tr ' ' '\n')
7325 error "all on-disk parameters not erased correctly via tunefs $params"
7327 # Check the order of options --erase-params and --param
7328 echo "tunefs --param $key=$val1 --erase-params"
7330 "$TUNEFS --param $key=$val1 --erase-params $mdsdev >/dev/null"||
7331 error "tunefs --param $key=$val1 --erase-params failed"
7332 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7333 error "tunefs --dryrun failed"
7334 params=${params##*Parameters:}
7335 params=${params%%exiting*}
7336 [ $(echo $params | tr ' ' '\n') == "$key=$val1" ] ||
7337 error "on-disk param not added correctly with --erase-params"
7341 run_test 89 "check tunefs --param and --erase-param{s} options"
7344 # $2 (optional) value of max_mod_rpcs_in_flight to set
7345 check_max_mod_rpcs_in_flight() {
7353 idx=$(printf "%04x" $($LFS getdirstripe -i $dir))
7354 facet="mds$((0x$idx + 1))"
7356 if [ -z "$mmr" ]; then
7357 # get value of max_mod_rcps_in_flight
7358 mmr=$($LCTL get_param -n \
7359 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) ||
7360 error "Unable to get max_mod_rpcs_in_flight"
7361 echo "max_mod_rcps_in_flight is $mmr"
7363 # set value of max_mod_rpcs_in_flight
7365 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight=$mmr ||
7366 error "Unable to set max_mod_rpcs_in_flight to $mmr"
7367 echo "max_mod_rpcs_in_flight set to $mmr"
7370 # create mmr+1 files
7371 echo "creating $((mmr + 1)) files ..."
7373 for i in $(seq $((mmr + 1))); do
7379 # consumes mmr-1 modify RPC slots
7380 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7381 # drop requests on MDT so that RPC slots are consumed
7382 # during all the request resend interval
7383 do_facet $facet "$LCTL set_param fail_loc=0x159"
7384 echo "launch $((mmr - 1)) chmod in parallel ..."
7385 for i in $(seq $((mmr - 1))); do
7386 chmod 0600 $dir/file-$i &
7390 # send one additional modify RPC
7391 do_facet $facet "$LCTL set_param fail_loc=0"
7392 echo "launch 1 additional chmod in parallel ..."
7393 chmod 0600 $dir/file-$mmr &
7396 # check this additional modify RPC get a modify RPC slot
7397 # and succeed its operation
7398 checkstat -vp 0600 $dir/file-$mmr ||
7399 error "Unable to send $mmr modify RPCs in parallel"
7404 # consumes mmr modify RPC slots
7405 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7406 # drop requests on MDT so that RPC slots are consumed
7407 # during all the request resend interval
7408 do_facet $facet "$LCTL set_param fail_loc=0x159"
7409 echo "launch $mmr chmod in parallel ..."
7410 for i in $(seq $mmr); do
7411 chmod 0666 $dir/file-$i &
7415 # send one additional modify RPC
7416 do_facet $facet "$LCTL set_param fail_loc=0"
7417 echo "launch 1 additional chmod in parallel ..."
7418 chmod 0666 $dir/file-$((mmr + 1)) &
7421 # check this additional modify RPC blocked getting a modify RPC slot
7422 checkstat -vp 0644 $dir/file-$((mmr + 1)) ||
7423 error "Unexpectedly send $(($mmr + 1)) modify RPCs in parallel"
7427 get_mdt_max_mod_rpcs_in_flight_val() {
7431 # It will be enough to get one value from one MDT
7432 max_mrif=$(do_facet $facet \
7433 "$LCTL get_param -n \
7434 mdt.$FSNAME-MDT*.max_mod_rpcs_in_flight | head -n 1")
7435 if [[ $max_mrif ]]; then
7438 max_mrif=$(do_facet $facet \
7439 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
7444 set_mdt_max_mod_rpcs_in_flight() {
7449 lctl_op=$($LCTL get_param \
7450 mdt.*.max_mod_rpcs_in_flight)
7451 if [[ $lctl_op ]]; then
7454 mdt.$FSNAME-MDT*.max_mod_rpcs_in_flight=$max_mrif"
7458 /sys/module/mdt/parameters/max_mod_rpcs_per_client"
7459 echo "the deprecated max_mod_rpcs_per_client \
7460 parameter was involved"
7467 [[ $($LCTL get_param mdc.*.import |
7468 grep "connect_flags:.*multi_mod_rpc") ]] ||
7469 skip "Need MDC with 'multi_mod_rpcs' feature"
7471 # check default value
7472 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
7473 check_max_mod_rpcs_in_flight $DIR/$tdir
7477 run_test 90a "check max_mod_rpcs_in_flight is enforced"
7487 [[ $($LCTL get_param mdc.*.import |
7488 grep "connect_flags:.*multi_mod_rpc") ]] ||
7489 skip "Need MDC with 'multi_mod_rpcs' feature"
7492 # update max_mod_rpcs_in_flight
7493 $LFS mkdir -c1 $DIR/${tdir}1 || error "mkdir $DIR/${tdir}1 failed"
7494 check_max_mod_rpcs_in_flight $DIR/${tdir}1 1
7497 # check client is able to send multiple modify RPCs in paralell
7498 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
7499 grep -c "multi_mod_rpcs")
7500 if [ "$tmp" -ne $MDSCOUNT ]; then
7502 skip "Client not able to send multiple modify RPCs in parallel"
7505 # update max_mod_rpcs_in_flight
7506 $LFS mkdir -c1 $DIR/${tdir}2 || error "mkdir $DIR/${tdir}2 failed"
7507 check_max_mod_rpcs_in_flight $DIR/${tdir}2 5
7510 $LFS mkdir -c1 $DIR/${tdir}3 || error "mkdir $DIR/${tdir}3 failed"
7511 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/${tdir}3))
7512 facet="mds$((0x$idx + 1))"
7514 mmrpc=$(get_mdt_max_mod_rpcs_in_flight_val $facet)
7515 echo "mdt_max_mod_rpcs_in_flight is $mmrpc"
7516 # update max_mod_rpcs_in_flight
7517 umount_client $MOUNT
7518 set_mdt_max_mod_rpcs_in_flight 16 $facet
7520 $LCTL set_param mdc.$FSNAME-MDT$idx-mdc-*.max_rpcs_in_flight=17
7521 check_max_mod_rpcs_in_flight $DIR/${tdir}3 16
7523 # restore MDT max_mod_rpcs_in_flight initial value
7524 set_mdt_max_mod_rpcs_in_flight $mmrpc $facet
7526 rm -rf $DIR/${tdir}?
7529 run_test 90b "check max_mod_rpcs_in_flight is enforced after update"
7532 # get max_rpcs_in_flight value
7533 mrif_90c=$($LCTL get_param -n \
7534 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
7535 echo "max_rpcs_in_flight is $mrif_90c"
7537 # get MDC max_mod_rpcs_in_flight value
7538 mmrif_90c=$($LCTL get_param -n \
7539 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight)
7540 echo "MDC max_mod_rpcs_in_flight is $mmrif_90c"
7542 # get MDT max_mod_rpcs_in_flight value
7543 mmrpc_90c=$(get_mdt_max_mod_rpcs_in_flight_val "mds1")
7544 echo "mdt_max_mod_rpcs_in_flight is $mmrpc_90c"
7547 restore_params_90c() {
7550 # restore max_rpcs_in_flight value
7551 do_facet mgs $LCTL set_param -P \
7552 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$mrif_90c
7554 # restore max_mod_rpcs_in_flight value
7555 do_facet mgs $LCTL set_param -P \
7556 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mmrif_90c
7558 # restore MDT max_mod_rpcs_in_flight value
7559 set_mdt_max_mod_rpcs_in_flight $mmrpc_90c "mds1"
7567 [[ $($LCTL get_param mdc.*.import |
7568 grep "connect_flags:.*multi_mod_rpc") ]] ||
7569 skip "Need MDC with 'multi_mod_rpcs' feature"
7571 # check client is able to send multiple modify RPCs in paralell
7572 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
7573 grep -c "multi_mod_rpcs")
7574 if [ "$tmp" -ne $MDSCOUNT ]; then
7576 skip "Client not able to send multiple modify RPCs in parallel"
7580 stack_trap restore_params_90c
7583 # attempt to set max_mod_rpcs_in_flight to max_rpcs_in_flight value
7584 # prerequisite: set MDT max_mod_rpcs_in_flight to
7585 # max_rpcs_in_flight value
7586 set_mdt_max_mod_rpcs_in_flight $mrif_90c "mds1"
7588 # if max_mod_rpcs_in_flight is set to be equal to or larger than
7589 # max_rpcs_in_flight, then max_rpcs_in_flight will be increased
7590 if [[ "$CLIENT_VERSION" -ge $(version_code 2.13.53) ]]; then
7592 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif_90c ||
7593 error "set max_mod_rpcs_in_flight to $mrif_90c failed"
7595 local new_mrif=$($LCTL get_param -n \
7596 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
7597 ((new_mrif == mrif_90c + 1)) ||
7598 error "max_rpcs_in_flight was not increased"
7601 umount_client $MOUNT
7602 set_mdt_max_mod_rpcs_in_flight $mmrpc_90c "mds1"
7606 # attempt to set max_mod_rpcs_in_flight to MDT max_mod_rpcs_in_flight+1
7607 # prerequisite: set max_rpcs_in_flight to MDT max_mod_rpcs_in_flight+2
7609 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$((mmrpc_90c + 2))
7612 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$((mmrpc_90c + 1)) &&
7613 error "set max_mod_rpcs_in_flight to $((mmrpc_90c + 1)) should fail"
7616 # attempt to set max_mod_rpcs_in_flight permanently
7617 do_facet mgs $LCTL set_param -P \
7618 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$mrif_90c
7620 do_facet mgs $LCTL set_param -P \
7621 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif_90c
7623 remount_client $MOUNT
7625 wait_update_facet --verbose client "$LCTL get_param -n \
7626 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight" \
7627 "$((mrif_90c + 1))" ||
7628 error "expected '$((mrif_90c + 1))' for max_rpcs_in_flight"
7630 wait_update_facet --verbose client "$LCTL get_param -n \
7631 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight" \
7633 error "expected '$mrif_90c' for max_mod_rpcs_in_flight"
7638 run_test 90c "check max_mod_rpcs_in_flight update limits"
7649 [[ $($LCTL get_param mdc.*.import |
7650 grep "connect_flags:.*multi_mod_rpc") ]] ||
7651 skip "Need MDC with 'multi_mod_rpcs' feature"
7653 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
7654 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
7655 facet="mds$((0x$idx + 1))"
7657 # check client version supports multislots
7658 tmp=$($LCTL get_param -N \
7659 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
7660 if [ -z "$tmp" ]; then
7662 skip "Client does not support multiple modify RPCs in flight"
7665 # get current value of max_mod_rcps_in_flight
7666 mmr=$($LCTL get_param -n \
7667 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
7668 echo "max_mod_rcps_in_flight is $mmr"
7671 echo "creating $mmr files ..."
7673 for i in $(seq $mmr); do
7674 touch $DIR/$tdir/file-$i
7677 # prepare for close RPC
7678 multiop_bg_pause $DIR/$tdir/file-close O_c
7681 # consumes mmr modify RPC slots
7682 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7683 # drop requests on MDT so that RPC slots are consumed
7684 # during all the request resend interval
7685 do_facet $facet "$LCTL set_param fail_loc=0x159"
7686 echo "launch $mmr chmod in parallel ..."
7687 for i in $(seq $mmr); do
7688 chmod 0600 $DIR/$tdir/file-$i &
7691 # send one additional close RPC
7692 do_facet $facet "$LCTL set_param fail_loc=0"
7693 echo "launch 1 additional close in parallel ..."
7695 cancel_lru_locks mdc
7698 # check this additional close RPC get a modify RPC slot
7699 # and multiop process completed
7700 [ -d /proc/$pid ] &&
7701 error "Unable to send the additional close RPC in parallel"
7706 run_test 90d "check one close RPC is allowed above max_mod_rpcs_in_flight"
7708 check_uuid_on_ost() {
7710 do_facet ost1 "$LCTL get_param obdfilter.${FSNAME}*.exports.'$nid'.uuid"
7713 check_uuid_on_mdt() {
7715 do_facet $SINGLEMDS "$LCTL get_param mdt.${mds1_svc}*.exports.'$nid'.uuid"
7723 [[ "$OST1_VERSION" -ge $(version_code 2.7.63) ]] ||
7724 skip "Need OST version at least 2.7.63"
7725 [[ "$MDS1_VERSION" -ge $(version_code 2.7.63) ]] ||
7726 skip "Need MDT version at least 2.7.63"
7728 start_mds || error "MDS start failed"
7729 start_ost || error "unable to start OST"
7730 mount_client $MOUNT || error "client start failed"
7731 check_mount || error "check_mount failed"
7734 nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g")
7738 uuid=$(get_client_uuid $MOUNT)
7740 echo "list nids on mdt:"
7741 do_facet $SINGLEMDS "$LCTL list_param mdt.${FSNAME}*.exports.*"
7742 echo "uuid from $nid:"
7743 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
7745 found=$(check_uuid_on_mdt $nid | grep $uuid)
7746 [ -z "$found" ] && error "can't find $uuid $nid on MDT"
7747 found=$(check_uuid_on_ost $nid | grep $uuid)
7748 [ -z "$found" ] && error "can't find $uuid $nid on OST"
7750 # umount the client so it won't reconnect
7751 manual_umount_client --force || error "failed to umount $?"
7752 # shouldn't disappear on MDS after forced umount
7753 found=$(check_uuid_on_mdt $nid | grep $uuid)
7754 [ -z "$found" ] && error "can't find $uuid $nid"
7757 do_facet $SINGLEMDS \
7758 "$LCTL set_param -n mdt.${mds1_svc}.evict_client nid:$nid"
7759 sleep 1 # eviction above is async, give it some time to proceed
7761 found=$(check_uuid_on_mdt $nid | grep $uuid)
7762 [ -n "$found" ] && error "found $uuid $nid on MDT"
7763 found=$(check_uuid_on_ost $nid | grep $uuid)
7764 [ -n "$found" ] && error "found $uuid $nid on OST"
7766 # check it didn't reconnect (being umounted)
7767 sleep $((TIMEOUT+1))
7768 found=$(check_uuid_on_mdt $nid | grep $uuid)
7769 [ -n "$found" ] && error "found $uuid $nid on MDT"
7770 found=$(check_uuid_on_ost $nid | grep $uuid)
7771 [ -n "$found" ] && error "found $uuid $nid on OST"
7775 run_test 91 "evict-by-nid support"
7777 generate_ldev_conf() {
7778 # generate an ldev.conf file
7779 local ldevconfpath=$1
7781 local fsldevformat=""
7784 local facets="mgs,$(get_facets OST),$(get_facets MDS)"
7785 for facet in ${facets//,/ }; do
7787 fstype=$(facet_fstype $facet)
7788 [ "$fstype" = zfs ] && fsldevformat="$fstype:"
7789 local host=$(facet_host $facet)
7791 local varfo=${facet}failover_HOST
7792 if [ $facet == mgs ] && combined_mgs_mds; then
7793 varfo=mds1failover_HOST
7795 [ -n "${!varfo}" ] && fo=${!varfo}
7796 local type=$(echo $facet | tr -d "[:digit:]" | \
7797 tr "[:lower:]" "[:upper:]" | sed s/MDS/MDT/ )
7799 [ ${facet} == mgs ] ||
7800 num=$(facet_number $facet)
7801 printf "%s\t%s\t%s-%s%04x\t%s%s\n" \
7808 $(facet_device $facet) >> $ldevconfpath
7811 echo "----- $ldevconfpath -----"
7813 echo "--- END $ldevconfpath ---"
7818 # generate a nids file (mapping between hostname to nid)
7819 # looks like we only have the MGS nid available to us
7820 # so just echo that to a file
7822 echo -e "${mgs_HOST}\t${MGSNID}" > $nidspath
7824 echo "----- $nidspath -----"
7826 echo "--- END $nidspath ---"
7829 compare_ldev_output() {
7830 local ldev_output=$1
7831 local expected_output=$2
7833 sort $expected_output -o $expected_output
7834 sort $ldev_output -o $ldev_output
7836 echo "-- START OF LDEV OUTPUT --"
7838 echo "--- END OF LDEV OUTPUT ---"
7840 echo "-- START OF EXPECTED OUTPUT --"
7841 cat $expected_output
7842 echo "--- END OF EXPECTED OUTPUT ---"
7844 diff $expected_output $ldev_output
7849 if [ -z "$LDEV" ]; then
7850 error "ldev is missing!"
7853 local LDEVCONFPATH=$TMP/ldev.conf
7854 local NIDSPATH=$TMP/nids
7856 echo "Host is $(hostname)"
7858 generate_ldev_conf $LDEVCONFPATH
7859 generate_nids $NIDSPATH
7861 # echo the mgs nid and compare it to environment variable MGSNID
7862 local facets="$(get_facets OST),$(get_facets MDS),mgs"
7863 for facet in ${facets//,/ }; do
7864 local host=$(facet_host $facet)
7865 local output=$($LDEV -c $LDEVCONFPATH -H $host -n $NIDSPATH echo %m)
7867 echo "-- START OF LDEV OUTPUT --"
7869 echo "--- END OF LDEV OUTPUT ---"
7872 error "ldev failed to execute!"
7874 # need to process multiple lines because of
7875 # several targets on host
7876 echo -e $output | awk '{ print $2 }' | while read -r line ; do
7877 [ "$line" = "$MGSNID" ] ||
7878 error "ldev failed mgs nid '$line', \
7882 rm -f $LDEVCONFPATH $NIDSPATH
7884 run_test 92 "ldev returns MGS NID correctly in command substitution"
7887 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs"
7890 #start mgs or mgs/mdt0
7891 if ! combined_mgs_mds ; then
7898 start_ost || error "OST0 start fail"
7900 #define OBD_FAIL_MGS_WRITE_TARGET_DELAY 0x90e
7901 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x8000090e"
7902 for num in $(seq 2 $MDSCOUNT); do
7906 mount_client $MOUNT || error "mount client fails"
7907 wait_osc_import_state mds ost FULL
7908 wait_osc_import_ready client ost
7909 check_mount || error "check_mount failed"
7911 cleanup || error "cleanup failed with $?"
7913 run_test 93 "register mulitple MDT at the same time"
7916 if [ -z "$LDEV" ]; then
7917 error "ldev is missing!"
7920 local LDEVCONFPATH=$TMP/ldev.conf
7921 local NIDSPATH=$TMP/nids
7923 generate_ldev_conf $LDEVCONFPATH
7924 generate_nids $NIDSPATH
7926 local LDEV_OUTPUT=$TMP/ldev-output.txt
7927 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME > $LDEV_OUTPUT
7929 # ldev failed, error
7930 if [ $? -ne 0 ]; then
7931 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7932 error "ldev failed to execute!"
7936 local EXPECTED_OUTPUT=$TMP/ldev-expected.txt
7938 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
7940 for num in $(seq $MDSCOUNT); do
7941 printf "%s-MDT%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7944 for num in $(seq $OSTCOUNT); do
7945 printf "%s-OST%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7948 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7950 if [ $? -ne 0 ]; then
7951 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7952 error "ldev failed to produce the correct hostlist!"
7955 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7957 run_test 94 "ldev outputs correct labels for file system name query"
7960 if [ -z "$LDEV" ]; then
7961 error "ldev is missing!"
7964 local LDEVCONFPATH=$TMP/ldev.conf
7965 local NIDSPATH=$TMP/nids
7967 generate_ldev_conf $LDEVCONFPATH
7968 generate_nids $NIDSPATH
7972 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME &>/dev/null
7973 if [ $? -ne 0 ]; then
7974 rm $LDEVCONFPATH $NIDSPATH
7975 error "ldev label filtering w/ -F failed!"
7979 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -l &>/dev/null
7980 if [ $? -ne 0 ]; then
7981 rm $LDEVCONFPATH $NIDSPATH
7982 error "ldev label filtering w/ -l failed!"
7986 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f &>/dev/null
7987 if [ $? -ne 0 ]; then
7988 rm $LDEVCONFPATH $NIDSPATH
7989 error "ldev label filtering w/ -f failed!"
7993 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a &>/dev/null
7994 if [ $? -ne 0 ]; then
7995 rm $LDEVCONFPATH $NIDSPATH
7996 error "ldev label filtering w/ -a failed!"
8001 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -F $FSNAME &>/dev/null
8002 if [ $? -eq 0 ]; then
8003 rm $LDEVCONFPATH $NIDSPATH
8004 error "ldev label filtering w/ -a and -F incorrectly succeeded"
8008 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -f &>/dev/null
8009 if [ $? -eq 0 ]; then
8010 rm $LDEVCONFPATH $NIDSPATH
8011 error "ldev label filtering w/ -a and -f incorrectly succeeded"
8015 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -l &>/dev/null
8016 if [ $? -eq 0 ]; then
8017 rm $LDEVCONFPATH $NIDSPATH
8018 error "ldev label filtering w/ -a and -l incorrectly succeeded"
8022 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f -l &>/dev/null
8023 if [ $? -eq 0 ]; then
8024 rm $LDEVCONFPATH $NIDSPATH
8025 error "ldev label filtering w/ -f and -l incorrectly succeeded"
8029 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -l &>/dev/null
8030 if [ $? -eq 0 ]; then
8031 rm $LDEVCONFPATH $NIDSPATH
8032 error "ldev label filtering w/ -F and -l incorrectly succeeded"
8035 # file sys & foreign
8036 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -f &>/dev/null
8037 if [ $? -eq 0 ]; then
8038 rm $LDEVCONFPATH $NIDSPATH
8039 error "ldev label filtering w/ -F and -f incorrectly succeeded"
8042 rm $LDEVCONFPATH $NIDSPATH
8044 run_test 95 "ldev should only allow one label filter"
8047 if [ -z "$LDEV" ]; then
8048 error "ldev is missing!"
8051 local LDEVCONFPATH=$TMP/ldev.conf
8052 local NIDSPATH=$TMP/nids
8054 generate_ldev_conf $LDEVCONFPATH
8055 generate_nids $NIDSPATH
8057 local LDEV_OUTPUT=$TMP/ldev-output.txt
8058 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -H $mgs_HOST \
8060 awk '{print $2}' > $LDEV_OUTPUT
8062 # ldev failed, error
8063 if [ $? -ne 0 ]; then
8064 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
8065 error "ldev failed to execute!"
8069 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
8071 echo "$mgs_HOST-$(facet_fstype mgs)" > $EXPECTED_OUTPUT
8073 local facets="$(get_facets OST),$(get_facets MDS)"
8074 for facet in ${facets//,/ }; do
8075 local host=$(facet_host $facet)
8076 [ "$mgs_HOST" == "$host" ] &&
8077 echo "$host-$(facet_fstype $facet)" \
8081 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
8083 if [ $? -ne 0 ]; then
8084 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8085 error "ldev failed to produce the correct output!"
8088 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8090 run_test 96 "ldev returns hostname and backend fs correctly in command sub"
8093 if [ -z "$LDEV" ]; then
8094 error "ldev is missing!"
8097 local LDEVCONFPATH=$TMP/ldev.conf
8098 local NIDSPATH=$TMP/nids
8100 generate_ldev_conf $LDEVCONFPATH
8101 generate_nids $NIDSPATH
8103 local LDEV_OUTPUT=$TMP/ldev-output.txt
8104 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
8106 echo -e "\nMDT role"
8107 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mdt > $LDEV_OUTPUT
8109 if [ $? -ne 0 ]; then
8110 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
8111 error "ldev failed to execute for mdt role!"
8114 for num in $(seq $MDSCOUNT); do
8115 printf "%s-MDT%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
8118 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
8120 if [ $? -ne 0 ]; then
8121 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8122 error "ldev failed to produce the correct output for mdt role!"
8125 echo -e "\nOST role"
8126 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R ost > $LDEV_OUTPUT
8128 if [ $? -ne 0 ]; then
8129 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
8130 error "ldev failed to execute for ost role!"
8134 for num in $(seq $OSTCOUNT); do
8135 printf "%s-OST%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
8138 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
8140 if [ $? -ne 0 ]; then
8141 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8142 error "ldev failed to produce the correct output for ost role!"
8145 echo -e "\nMGS role"
8146 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mgs > $LDEV_OUTPUT
8148 if [ $? -ne 0 ]; then
8149 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
8150 error "ldev failed to execute for mgs role!"
8153 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
8155 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
8157 if [ $? -ne 0 ]; then
8158 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8159 error "ldev failed to produce the correct output for mgs role!"
8162 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8164 run_test 97 "ldev returns correct ouput when querying based on role"
8171 check_mount || error "mount failed"
8172 mountopt="user_xattr"
8173 for ((x = 1; x <= PAGE_SIZE/11; x++)); do
8174 mountopt="$mountopt,user_xattr"
8176 mount_client $MOUNT remount,$mountopt 2>&1 | grep "too long" ||
8177 error "Buffer overflow check failed"
8178 cleanup || error "cleanup failed"
8180 run_test 98 "Buffer-overflow check while parsing mount_opts"
8184 [[ "$ost1_FSTYPE" != ldiskfs ]] &&
8185 skip "ldiskfs only test"
8186 [[ "$OST1_VERSION" -ge $(version_code 2.8.57) ]] ||
8187 skip "Need OST version at least 2.8.57"
8189 local ost_opts="$(mkfs_opts ost1 $(ostdevname 1)) \
8190 --reformat $(ostdevname 1) $(ostvdevname 1)"
8191 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" &&
8192 skip "meta_bg already set"
8195 if [[ ${!opts} != *mkfsoptions* ]]; then
8196 eval opts=\"${!opts} \
8197 --mkfsoptions='\\\"-O ^resize_inode,meta_bg\\\"'\"
8199 local val=${!opts//--mkfsoptions=\\\"/ \
8200 --mkfsoptions=\\\"-O ^resize_inode,meta_bg }
8204 echo "params: $opts"
8207 add ost1 $opts || error "add ost1 failed with new params"
8209 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" ||
8210 error "meta_bg is not set"
8212 run_test 99 "Adding meta_bg option"
8217 start_mgsmds || error "MDS start failed"
8218 start_ost || error "unable to start OST"
8219 mount_client $MOUNT || error "client start failed"
8220 check_mount || error "check_mount failed"
8229 do_facet mgs 'lshowmount -v' | awk 'BEGIN {NR == 0; rc=1} /MGS:/ {rc=0}
8230 END {exit rc}' || error "lshowmount have no output MGS"
8232 do_facet mds1 'lshowmount -v' | awk 'BEGIN {NR == 2; rc=1} /-MDT0000:/
8233 {rc=0} END {exit rc}' || error "lshowmount have no output MDT0"
8235 do_facet ost1 'lshowmount -v' | awk 'BEGIN {NR == 4; rc=1} /-OST0000:/
8236 {rc=0} END {exit rc}' || error "lshowmount have no output OST0"
8238 cleanup || error "cleanup failed with $?"
8240 run_test 100 "check lshowmount lists MGS, MDT, OST and 0@lo"
8243 local createmany_pid
8244 local dev=$FSNAME-OST0000-osc-MDT0000
8248 do_nodes $(comma_list $(osts_nodes)) $LCTL set_param \
8249 seq.*OST*-super.width=$DATA_SEQ_MAX_WIDTH
8250 createmany -o $DIR1/$tdir/$tfile-%d 50000 &
8252 # MDT->OST reconnection causes MDT<->OST last_id synchornisation
8253 # via osp_precreate_cleanup_orphans.
8254 for ((i = 0; i < 100; i++)); do
8255 for ((k = 0; k < 10; k++)); do
8256 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;" \
8257 "$LCTL --device $dev activate"
8260 ls -asl $MOUNT | grep '???' &&
8261 { kill -9 $createmany_pid &>/dev/null;
8262 error "File has no object on OST"; }
8264 kill -s 0 $createmany_pid || break
8266 wait $createmany_pid
8268 unlinkmany $DIR1/$tdir/$tfile-%d 50000
8271 run_test 101a "Race MDT->OST reconnection with create"
8274 local dev=$FSNAME-OST0000-osc-MDT0000
8275 local dir=$DIR1/$tdir
8279 $LFS setstripe -c 1 -i 0 $dir
8280 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;"
8281 #define OBD_FAIL_OSP_CON_EVENT_DELAY 0x2107
8282 do_facet mds1 "$LCTL set_param fail_loc=0x80002107 fail_val=20"
8283 do_facet $SINGLEMDS "$LCTL --device $dev activate;"
8288 wait_osc_import_ready client ost1
8289 touch $dir/$tfile || error "Can't create file"
8293 run_test 101b "Race events DISCONNECT and ACTIVE in osp"
8296 [[ "$MDS1_VERSION" -gt $(version_code 2.9.53) ]] ||
8297 skip "Need server version greater than 2.9.53"
8298 [[ “$(mdsdevname 1)” != “$(mgsdevname)” ]] &&
8299 [[ “$(facet_host mds1)” = “$(facet_host mgs)” ]] &&
8300 skip "MGS must be on different node or combined"
8302 cleanup || error "cleanup failed with $?"
8304 local mds1dev=$(mdsdevname 1)
8305 local mds1mnt=$(facet_mntpt mds1)
8306 local mds1opts=$MDS_MOUNT_OPTS
8308 if [ "$mds1_FSTYPE" == ldiskfs ] &&
8309 ! do_facet mds1 test -b $mds1dev; then
8310 mds1opts=$(csa_add "$mds1opts" -o loop)
8312 if [[ "$mds1_FSTYPE" == zfs ]]; then
8313 import_zpool mds1 || return ${PIPESTATUS[0]}
8316 # unload all and only load libcfs to allow fail_loc setting
8317 do_facet mds1 $LUSTRE_RMMOD || error "unable to unload modules"
8318 do_facet mds1 modprobe libcfs || error "libcfs not loaded"
8319 do_facet mds1 lsmod \| grep libcfs || error "libcfs not loaded"
8321 #define OBD_FAIL_OBDCLASS_MODULE_LOAD 0x60a
8322 do_facet mds1 "$LCTL set_param fail_loc=0x8000060a"
8324 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts &&
8325 error "mdt start must fail"
8326 do_facet mds1 lsmod \| grep obdclass && error "obdclass must not load"
8328 do_facet mds1 "$LCTL set_param fail_loc=0x0"
8330 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts ||
8331 error "mdt start must not fail"
8333 cleanup || error "cleanup failed with $?"
8335 run_test 102 "obdclass module cleanup upon error"
8340 echo "rename $FSNAME to $newname"
8342 if ! combined_mgs_mds ; then
8343 local dev=$(mgsdevname)
8346 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
8347 error "(7) Fail to rename MGS"
8348 if [ "$(facet_fstype mgs)" = "zfs" ]; then
8349 reimport_zpool mgs $newname-mgs
8353 for num in $(seq $MDSCOUNT); do
8354 local dev=$(mdsdevname $num)
8356 do_facet mds${num} \
8357 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
8358 error "(8) Fail to rename MDT $num"
8359 if [ "$(facet_fstype mds${num})" = "zfs" ]; then
8360 reimport_zpool mds${num} $newname-mdt${num}
8364 for num in $(seq $OSTCOUNT); do
8365 local dev=$(ostdevname $num)
8367 do_facet ost${num} \
8368 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
8369 error "(9) Fail to rename OST $num"
8370 if [ "$(facet_fstype ost${num})" = "zfs" ]; then
8371 reimport_zpool ost${num} $newname-ost${num}
8376 test_103_set_pool() {
8380 do_facet mgs $LCTL pool_add $FSNAME.$pname ${FSNAME}-$ost_x ||
8381 error "Fail to add $ost_x to $FSNAME.$pname"
8382 wait_update $HOSTNAME \
8383 "lctl get_param -n lov.$FSNAME-clilov-*.pools.$pname |
8384 grep $ost_x" "$FSNAME-${ost_x}_UUID" ||
8385 error "$ost_x is NOT in pool $FSNAME.$pname"
8388 test_103_check_pool() {
8389 local save_fsname=$1
8392 stat $DIR/$tdir/test-framework.sh ||
8393 error "($errno) Fail to stat"
8394 do_facet mgs $LCTL pool_list $FSNAME.pool1 ||
8395 error "($errno) Fail to list $FSNAME.pool1"
8396 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname ||
8397 error "($errno) Fail to list $FSNAME.$save_fsname"
8398 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname |
8399 grep ${FSNAME}-OST0000 ||
8400 error "($errno) List $FSNAME.$save_fsname is invalid"
8402 local pname=$($LFS getstripe --pool $DIR/$tdir/d0)
8403 [ "$pname" = "$save_fsname" ] ||
8404 error "($errno) Unexpected pool name $pname"
8408 check_mount_and_prep
8410 mkdir $DIR/$tdir || error "(1) Fail to mkdir $DIR/$tdir"
8411 cp $LUSTRE/tests/test-framework.sh $DIR/$tdir ||
8412 error "(2) Fail to copy test-framework.sh"
8414 do_facet mgs $LCTL pool_new $FSNAME.pool1 ||
8415 error "(3) Fail to create $FSNAME.pool1"
8416 # name the pool name as the fsname
8417 do_facet mgs $LCTL pool_new $FSNAME.$FSNAME ||
8418 error "(4) Fail to create $FSNAME.$FSNAME"
8420 test_103_set_pool $FSNAME OST0000
8422 $LFS setstripe -p $FSNAME $DIR/$tdir/d0 ||
8423 error "(6) Fail to setstripe on $DIR/$tdir/d0"
8428 test_renamefs mylustre
8430 local save_fsname=$FSNAME
8434 test_103_check_pool $save_fsname 7
8436 if [ $OSTCOUNT -ge 2 ]; then
8437 test_103_set_pool $save_fsname OST0001
8440 $LFS setstripe -p $save_fsname $DIR/$tdir/f0 ||
8441 error "(16) Fail to setstripe on $DIR/$tdir/f0"
8450 test_103_check_pool $save_fsname 17
8454 test_renamefs $save_fsname
8460 run_test 103 "rename filesystem name"
8462 test_104a() { # LU-6952
8463 local mds_mountopts=$MDS_MOUNT_OPTS
8464 local ost_mountopts=$OST_MOUNT_OPTS
8465 local mds_mountfsopts=$MDS_MOUNT_FS_OPTS
8466 local lctl_ver=$(do_facet $SINGLEMDS $LCTL --version |
8469 [[ $(version_code $lctl_ver) -lt $(version_code 2.9.55) ]] &&
8470 skip "this test needs utils above 2.9.55"
8472 # specify "acl" in mount options used by mkfs.lustre
8473 if [ -z "$MDS_MOUNT_FS_OPTS" ]; then
8474 MDS_MOUNT_FS_OPTS="acl,user_xattr"
8477 MDS_MOUNT_FS_OPTS="${MDS_MOUNT_FS_OPTS},acl,user_xattr"
8480 echo "mountfsopt: $MDS_MOUNT_FS_OPTS"
8482 #reformat/remount the MDT to apply the MDT_MOUNT_FS_OPT options
8484 if ! combined_mgs_mds ; then
8488 if [ -z "$MDS_MOUNT_OPTS" ]; then
8489 MDS_MOUNT_OPTS="-o noacl"
8491 MDS_MOUNT_OPTS="${MDS_MOUNT_OPTS},noacl"
8494 for num in $(seq $MDSCOUNT); do
8495 start mds$num $(mdsdevname $num) $MDS_MOUNT_OPTS ||
8496 error "Failed to start MDS"
8499 for num in $(seq $OSTCOUNT); do
8500 start ost$num $(ostdevname $num) $OST_MOUNT_OPTS ||
8501 error "Failed to start OST"
8505 setfacl -m "d:$RUNAS_ID:rwx" $MOUNT &&
8506 error "ACL is applied when FS is mounted with noacl."
8508 MDS_MOUNT_OPTS=$mds_mountopts
8509 OST_MOUNT_OPTS=$ost_mountopts
8510 MDS_MOUNT_FS_OPTS=$mds_mountfsopts
8512 run_test 104a "Make sure user defined options are reflected in mount"
8514 test_104b() { # LU-12859
8515 mount_client $MOUNT3 flock,localflock
8516 stack_trap "umount_client $MOUNT3" EXIT
8517 mount | grep "$MOUNT3 .*,flock" && error "flock is still set"
8518 mount | grep "$MOUNT3 .*,localflock" || error "localflock is not set"
8519 umount_client $MOUNT3
8520 mount_client $MOUNT3 localflock,flock
8521 mount | grep "$MOUNT3 .*,localflock" && error "localflock is still set"
8522 mount | grep "$MOUNT3 .*,flock" || error "flock is not set"
8523 umount_client $MOUNT3
8524 mount_client $MOUNT3 localflock,flock,noflock
8525 flock_is_enabled $MOUNT3 && error "some flock is still enabled" || true
8527 run_test 104b "Mount uses last flock argument"
8529 error_and_umount() {
8540 mount --bind $DIR $TMP/$tdir || error "mount bind mnt pt failed"
8541 rm -f $TMP/$tdir/$tfile
8542 rm -f $TMP/$tdir/${tfile}1
8544 # Files should not be created in ro bind mount point
8545 # remounting from rw to ro
8546 mount -o remount,ro $TMP/$tdir ||
8547 error_and_umount "readonly remount of bind mnt pt failed"
8548 touch $TMP/$tdir/$tfile &&
8549 error_and_umount "touch succeeds on ro bind mnt pt"
8550 [ -e $TMP/$tdir/$tfile ] &&
8551 error_and_umount "file created on ro bind mnt pt"
8553 # Files should be created in rw bind mount point
8554 # remounting from ro to rw
8555 mount -o remount,rw $TMP/$tdir ||
8556 error_and_umount "read-write remount of bind mnt pt failed"
8557 touch $TMP/$tdir/${tfile}1 ||
8558 error_and_umount "touch fails on rw bind mnt pt"
8559 [ -e $TMP/$tdir/${tfile}1 ] ||
8560 error_and_umount "file not created on rw bind mnt pt"
8561 umount $TMP/$tdir || error "umount of bind mnt pt failed"
8563 cleanup || error "cleanup failed with $?"
8565 run_test 105 "check file creation for ro and rw bind mnt pt"
8569 local creates=64768 # one full plain llog
8571 # ensure there are enough inodes in the filesystem
8572 (( OSTSIZE < (creates + 1024) * 8)) && OSTSIZE=$(((creates + 1024) * 8))
8576 mkdir -p $DIR/$tdir || error "create $tdir failed"
8577 do_nodes $(comma_list $(osts_nodes)) $LCTL set_param \
8578 seq.*OST*-super.width=$DATA_SEQ_MAX_WIDTH
8579 lfs setstripe -c 1 -i 0 $DIR/$tdir
8580 #define OBD_FAIL_CAT_RECORDS 0x1312
8581 do_facet mds1 $LCTL set_param fail_loc=0x1312 fail_val=$repeat
8583 for ((i = 1; i <= $repeat; i++)); do
8584 createmany -o $DIR/$tdir/f- $creates || lfs df -i
8585 createmany -u $DIR/$tdir/f- $creates
8586 wait_delete_completed $((TIMEOUT * 7))
8588 #ASSERTION osp_sync_thread() ( thread->t_flags != SVC_RUNNING ) failed
8589 #shows that osp code is buggy
8590 do_facet mds1 $LCTL set_param fail_loc=0 fail_val=0
8594 run_test 106 "check osp llog processing when catalog is wrapped"
8597 [[ "$MDS1_VERSION" -ge $(version_code 2.10.50) ]] ||
8598 skip "Need MDS version > 2.10.50"
8601 start_mgsmds || error "start_mgsmds failed"
8602 start_ost || error "unable to start OST"
8604 # add unknown configuration parameter.
8605 if [[ $PERM_CMD == *"set_param -P"* ]]; then
8606 cmd="$PERM_CMD ost.$FSNAME-OST0000*.unknown_param"
8608 cmd="$PERM_CMD $FSNAME-OST0000*.ost.unknown_param"
8610 do_facet mgs "$cmd=50"
8611 cleanup_nocli || error "cleanup_nocli failed with $?"
8614 # unknown param should be ignored while mounting.
8615 start_ost || error "unable to start OST after unknown param set"
8617 cleanup || error "cleanup failed with $?"
8619 run_test 107 "Unknown config param should not fail target mounting"
8624 $rcmd rm -rf $tmp > /dev/null 2>&1
8625 $rcmd mkdir -p $tmp/{mnt,images} || error "failed to mkdir remotely"
8627 for facet in $facets; do
8628 [ "$mds1_FSTYPE" = zfs ] &&
8629 $rcmd $ZPOOL -f export lustre-$facet > /dev/null 2>&1
8630 $rcmd mkdir $tmp/mnt/$facet ||
8631 error "failed to mkdir $tmp/mnt/$facet"
8632 $rcmd dd if=/dev/zero of=$tmp/images/$facet \
8633 seek=199 bs=1M count=1 ||
8634 error "failed to create $tmp/images/$facet"
8643 local facet=${role}$((idx + 1))
8645 [ $# -eq 5 ] && pool=$5
8647 do_facet $SINGLEMDS $MKFS --fsname=lustre --$mgs \
8648 --$role --index=$idx --replace --backfstype=$bkfs \
8649 --device-size=200000 --reformat $pool $tmp/images/$facet ||
8650 error "failed to mkfs for $facet"
8654 echo "mounting client..."
8655 mount -t lustre ${nid}:/lustre $MOUNT ||
8656 error "failed to mount lustre"
8659 ls -l $MOUNT/local_dir || error "failed to list"
8661 echo "check truncate && write"
8662 echo "dummmmmmmmmmmmm" > $MOUNT/remote_dir/fsx.c ||
8663 error "failed to tuncate & write"
8666 touch $MOUNT/foooo ||
8667 error "failed to create"
8669 echo "check read && write && append"
8670 sha1sum $MOUNT/conf-sanity.sh |
8671 awk '{ print $1 }' > $MOUNT/checksum.new ||
8672 error "failed to read(1)"
8673 sha1sum $MOUNT/remote_dir/unlinkmany.c |
8674 awk '{ print $1 }' >> $MOUNT/checksum.new ||
8675 error "failed to read(2)"
8676 sha1sum $MOUNT/striped_dir/lockahead_test.o |
8677 awk '{ print $1 }' >> $MOUNT/checksum.new ||
8678 error "failed to read(3)"
8681 diff $MOUNT/checksum.new $MOUNT/checksum.src ||
8682 error "failed to verify data"
8692 umount -f $MOUNT || error "failed to umount client"
8693 for facet in $facets; do
8694 $rcmd umount -f $tmp/mnt/$facet ||
8695 error "failed to umount $facet"
8696 if [ "$mds1_FSTYPE" = zfs ]; then
8697 $rcmd $ZPOOL export -f lustre-$facet ||
8698 error "failed to export lustre-$facet"
8702 $rcmd rm -rf $tmp || error "failed to rm the dir $tmp"
8706 [ "$CLIENTONLY" ] && skip "Client-only testing"
8707 [ "$mds1_FSTYPE" != zfs ] && skip "zfs only test"
8708 [ "$MDS1_VERSION" -lt $(version_code 2.10.58) ] &&
8709 skip "Need server version at least 2.10.58"
8714 local tmp=$TMP/$tdir
8715 local rcmd="do_facet $SINGLEMDS"
8716 local facets="mdt1 mdt2 ost1 ost2"
8717 local nid=$($rcmd $LCTL list_nids | head -1)
8720 trap t_108_cleanup EXIT ERR
8723 t_108_mkfs mdt 0 zfs mgs lustre-mdt1/mdt1
8724 t_108_mkfs mdt 1 zfs mgsnode=$nid lustre-mdt2/mdt2
8725 t_108_mkfs ost 0 zfs mgsnode=$nid lustre-ost1/ost1
8726 t_108_mkfs ost 1 zfs mgsnode=$nid lustre-ost2/ost2
8728 for facet in $facets; do
8729 $rcmd zfs set mountpoint=$tmp/mnt/$facet canmount=on \
8730 lustre-$facet/$facet ||
8731 error "failed to zfs set for $facet (1)"
8732 $rcmd zfs mount lustre-$facet/$facet ||
8733 error "failed to local mount $facet"
8734 $rcmd tar jxf $LUSTRE/tests/ldiskfs_${facet}_2_11.tar.bz2 \
8735 --xattrs --xattrs-include="trusted.*" \
8736 -C $tmp/mnt/$facet/ > /dev/null 2>&1 ||
8737 error "failed to untar image for $facet"
8738 $rcmd "cd $tmp/mnt/$facet && rm -rf oi.* OI_* lfsck_* LFSCK" ||
8739 error "failed to cleanup for $facet"
8740 $rcmd zfs umount lustre-$facet/$facet ||
8741 error "failed to local umount $facet"
8742 $rcmd zfs set canmount=off lustre-$facet/$facet ||
8743 error "failed to zfs set $facet (2)"
8746 echo "changing server nid..."
8747 $rcmd mount -t lustre -o nosvc lustre-mdt1/mdt1 $tmp/mnt/mdt1
8748 $rcmd lctl replace_nids $FSNAME-MDT0000 $nid
8749 $rcmd lctl replace_nids $FSNAME-MDT0001 $nid
8750 $rcmd lctl replace_nids $FSNAME-OST0000 $nid
8751 $rcmd lctl replace_nids $FSNAME-OST0001 $nid
8752 $rcmd umount $tmp/mnt/mdt1
8754 for facet in $facets; do
8755 echo "mounting $facet from backup..."
8756 $rcmd mount -t lustre -o abort_recov lustre-$facet/$facet \
8757 $tmp/mnt/$facet || error "failed to mount $facet"
8760 # ZFS backend can detect migration and trigger OI scrub automatically
8761 # sleep 3 seconds for scrub done
8767 run_test 108a "migrate from ldiskfs to ZFS"
8770 [ "$CLIENTONLY" ] && skip "Client-only testing"
8771 [ "$mds1_FSTYPE" != ldiskfs ] && skip "ldiskfs only test"
8772 [ "$MDS1_VERSION" -lt $(version_code 2.10.58) ] &&
8773 skip "Need server version at least 2.10.58"
8778 local tmp=$TMP/$tdir
8779 local rcmd="do_facet $SINGLEMDS"
8780 local facets="mdt1 mdt2 ost1 ost2"
8781 local scrub_list="MDT0000 MDT0001 OST0000 OST0001"
8782 local nid=$($rcmd $LCTL list_nids | head -1)
8785 trap t_108_cleanup EXIT ERR
8788 t_108_mkfs mdt 0 ldiskfs mgs
8789 t_108_mkfs mdt 1 ldiskfs mgsnode=$nid
8790 t_108_mkfs ost 0 ldiskfs mgsnode=$nid
8791 t_108_mkfs ost 1 ldiskfs mgsnode=$nid
8793 for facet in $facets; do
8794 $rcmd mount -t ldiskfs -o loop $tmp/images/$facet \
8796 error "failed to local mount $facet"
8798 $rcmd tar jxf $LUSTRE/tests/zfs_${facet}_2_11.tar.bz2 \
8799 --xattrs --xattrs-include="*.*" \
8800 -C $tmp/mnt/$facet/ > /dev/null 2>&1 ||
8801 error "failed to untar image for $facet"
8802 $rcmd "cd $tmp/mnt/$facet && rm -rf oi.* OI_* lfsck_* LFSCK" ||
8803 error "failed to cleanup for $facet"
8804 $rcmd umount $tmp/mnt/$facet ||
8805 error "failed to local umount $facet"
8808 echo "changing server nid..."
8809 $rcmd mount -t lustre -o nosvc,loop $tmp/images/mdt1 $tmp/mnt/mdt1
8810 $rcmd lctl replace_nids $FSNAME-MDT0000 $nid
8811 $rcmd lctl replace_nids $FSNAME-MDT0001 $nid
8812 $rcmd lctl replace_nids $FSNAME-OST0000 $nid
8813 $rcmd lctl replace_nids $FSNAME-OST0001 $nid
8814 $rcmd umount $tmp/mnt/mdt1
8816 for facet in $facets; do
8817 echo "mounting $facet from backup..."
8818 $rcmd mount -t lustre -o loop,abort_recov $tmp/images/$facet \
8819 $tmp/mnt/$facet || error "failed to mount $facet"
8822 for facet in $scrub_list; do
8823 $rcmd $LCTL lfsck_start -M $FSNAME-$facet -t scrub ||
8824 error "failed to start OI scrub on $facet"
8827 # sleep 3 seconds for scrub done
8833 run_test 108b "migrate from ZFS to ldiskfs"
8836 # set number of permanent parameters
8838 test_109_set_params() {
8841 set_persistent_param_and_check mds \
8842 "mdd.$fsname-MDT0000.atime_diff" \
8843 "$fsname-MDT0000.mdd.atime_diff" \
8845 set_persistent_param_and_check mds \
8846 "mdd.$fsname-MDT0000.atime_diff" \
8847 "$fsname-MDT0000.mdd.atime_diff" \
8849 set_persistent_param_and_check client \
8850 "llite.$fsname*.max_read_ahead_mb" \
8851 "$fsname.llite.max_read_ahead_mb" \
8853 set_persistent_param_and_check client \
8854 "llite.$fsname*.max_read_ahead_mb" \
8855 "$fsname.llite.max_read_ahead_mb" \
8857 create_pool $fsname.pool1 || error "create pool failed"
8858 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
8859 error "pool_add failed"
8860 do_facet mgs $LCTL pool_remove $fsname.pool1 OST0000 ||
8861 error "pool_remove failed"
8862 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
8863 error "pool_add failed"
8867 # check permanent parameters
8869 test_109_test_params() {
8872 local atime_diff=$(do_facet mds $LCTL \
8873 get_param -n mdd.$fsname-MDT0000.atime_diff)
8874 [ $atime_diff == 63 ] || error "wrong mdd parameter after clear_conf"
8875 local max_read_ahead_mb=$(do_facet client $LCTL \
8876 get_param -n llite.$fsname*.max_read_ahead_mb)
8877 [ $max_read_ahead_mb == 64 ] ||
8878 error "wrong llite parameter after clear_conf"
8879 local ost_in_pool=$(do_facet mds $LCTL pool_list $fsname.pool1 |
8880 grep -v "^Pool:" | sed 's/_UUID//')
8881 [ $ost_in_pool = "$fsname-OST0000" ] ||
8882 error "wrong pool after clear_conf"
8886 # run lctl clear_conf, store CONFIGS before and after that
8888 test_109_clear_conf()
8890 local clear_conf_arg=$1
8893 if ! combined_mgs_mds ; then
8894 mgsdev=$(mgsdevname)
8895 stop_mgs || error "stop_mgs failed"
8896 start_mgs "-o nosvc" || error "start_mgs nosvc failed"
8898 mgsdev=$(mdsdevname 1)
8899 start_mdt 1 "-o nosvc" || error "start_mdt 1 nosvc failed"
8902 do_facet mgs "rm -rf $TMP/${tdir}/conf1; mkdir -p $TMP/${tdir}/conf1;" \
8903 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf1\\\" \
8907 # the command being tested
8909 do_facet mgs $LCTL clear_conf $clear_conf_arg ||
8910 error "clear_conf failed"
8911 if ! combined_mgs_mds ; then
8912 stop_mgs || error "stop_mgs failed"
8914 stop_mdt 1 || error "stop_mdt 1 failed"
8917 do_facet mgs "rm -rf $TMP/${tdir}/conf2; mkdir -p $TMP/${tdir}/conf2;" \
8918 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf2\\\" \
8922 test_109_file_shortened() {
8924 local sizes=($(do_facet mgs "stat -c %s " \
8925 "$TMP/${tdir}/conf1/CONFIGS/$file" \
8926 "$TMP/${tdir}/conf2/CONFIGS/$file"))
8927 [ ${sizes[1]} -lt ${sizes[0]} ] && return 0
8933 [ $MDS1_VERSION -lt $(version_code 2.10.59) ] &&
8934 skip "Needs MDS version 2.10.59 or later."
8935 [ "$(facet_fstype mgs)" == zfs ] &&
8936 skip "LU-8727: no implementation for ZFS"
8941 client_up || error "client_up failed"
8944 # set number of permanent parameters
8946 test_109_set_params $FSNAME
8948 umount_client $MOUNT || error "umount_client failed"
8949 stop_ost || error "stop_ost failed"
8950 stop_mds || error "stop_mds failed"
8952 test_109_clear_conf $FSNAME
8954 # make sure that all configs are cleared
8956 test_109_file_shortened $FSNAME-MDT0000 ||
8957 error "failed to clear MDT0000 config"
8958 test_109_file_shortened $FSNAME-client ||
8959 error "failed to clear client config"
8964 # check that configurations are intact
8966 test_109_test_params $FSNAME
8971 destroy_test_pools || error "destroy test pools failed"
8975 run_test 109a "test lctl clear_conf fsname"
8979 [ $MDS1_VERSION -lt $(version_code 2.10.59) ] &&
8980 skip "Needs MDS version 2.10.59 or later."
8981 [ "$(facet_fstype mgs)" == zfs ] &&
8982 skip "LU-8727: no implementation for ZFS"
8987 client_up || error "client_up failed"
8990 # set number of permanent parameters
8992 test_109_set_params $FSNAME
8994 umount_client $MOUNT || error "umount_client failed"
8995 stop_ost || error "stop_ost failed"
8996 stop_mds || error "stop_mds failed"
8998 test_109_clear_conf $FSNAME-MDT0000
9000 # make sure that only one config is cleared
9002 test_109_file_shortened $FSNAME-MDT0000 ||
9003 error "failed to clear MDT0000 config"
9004 test_109_file_shortened $FSNAME-client &&
9005 error "failed to clear client config"
9009 # check that configurations are intact
9011 test_109_test_params $FSNAME
9016 destroy_test_pools || error "destroy test pools failed"
9020 run_test 109b "test lctl clear_conf one config"
9024 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
9025 skip "Only applicable to ldiskfs-based MDTs"
9027 do_facet $SINGLEMDS $DEBUGFS -w -R supported_features |grep large_dir ||
9028 skip "large_dir option is not supported on MDS"
9029 do_facet ost1 $DEBUGFS -w -R supported_features | grep large_dir ||
9030 skip "large_dir option is not supported on OSS"
9032 stopall # stop all targets before modifying the target counts
9033 local old_mdscount=$MDSCOUNT
9034 local old_ostcount=$OSTCOUNT
9036 stack_trap "MDSCOUNT=$old_mdscount OSTCOUNT=$old_ostcount; reformat"
9040 # ext4_dir_entry_2 struct size:264
9041 # dx_root struct size:8
9042 # dx_node struct size:8
9043 # dx_entry struct size:8
9044 # For 1024 bytes block size.
9045 # First level directory entries: 126
9046 # Second level directory entries: 127
9047 # Entries in leaf: 3
9048 # For 2 levels limit: 48006
9049 # For 3 levels limit : 6096762
9050 # Create 80000 files to safely exceed 2-level htree limit.
9051 CONF_SANITY_110_LINKS=${CONF_SANITY_110_LINKS:-80000}
9053 # can fit at most 3 filenames per 1KB leaf block, but each
9054 # leaf/index block will only be 3/4 full before split at each level
9055 (( MDSSIZE < CONF_SANITY_110_LINKS / 3 * 4/3 * 4/3 )) &&
9056 CONF_SANITY_110_LINKS=$((MDSSIZE * 3 * 3/4 * 3/4))
9058 combined_mgs_mds || replace=" --replace "
9059 local opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
9060 $replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
9061 if [[ $opts != *large_dir* ]]; then
9062 if [[ $opts != *mkfsoptions* ]]; then
9063 opts+=" --mkfsoptions=\\\"-O large_dir -b 1024 -i 65536\\\""
9065 opts="${opts//--mkfsoptions=\\\"/ \
9066 --mkfsoptions=\\\"-O large_dir -b 1024 -i 65536 }"
9069 echo "MDT params: $opts"
9071 combined_mgs_mds || start_mgs
9072 add mds1 $opts || error "add mds1 failed with new params"
9073 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
9075 opts="$(mkfs_opts ost1 $(ostdevname 1)) \
9076 $replace --reformat $(ostdevname 1) $(ostvdevname 1)"
9078 if [[ $opts != *large_dir* ]]; then
9079 if [[ $opts != *mkfsoptions* ]]; then
9080 opts+=" --mkfsoptions=\\\"-O large_dir\\\" "
9082 opts="${opts//--mkfsoptions=\\\"/ \
9083 --mkfsoptions=\\\"-O large_dir }"
9086 echo "OST params: $opts"
9087 add ost1 $opts || error "add ost1 failed with new params"
9088 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
9090 MOUNT_2=yes mountcli || error "mount clients failed"
9092 mkdir -v $DIR/$tdir || error "cannot create $DIR/$tdir"
9093 local pids count=0 group=0
9095 echo "creating $CONF_SANITY_110_LINKS in total"
9096 while (( count < CONF_SANITY_110_LINKS )); do
9097 local len=$((253 - $(wc -c <<<"$tfile-$group-40000-")))
9098 local dir=DIR$((group % 2 + 1))
9099 local target=${!dir}/$tdir/$tfile-$group
9100 local long=$target-$(generate_name $len)-
9101 local create=$((CONF_SANITY_110_LINKS - count))
9103 (( create > 40000 )) && create=40000
9104 touch $target || error "creating $target failed"
9105 echo "creating $create hard links to $target"
9106 createmany -l $target $long $create &
9109 count=$((count + create))
9110 group=$((group + 1))
9112 echo "waiting for PIDs$pids to complete"
9113 wait $pids || error "createmany failed after $group groups"
9115 umount_client $MOUNT2 -f
9118 run_e2fsck $(facet_active_host mds1) $(mdsdevname 1) -n
9119 MDSCOUNT=$old_mdscount
9120 OSTCOUNT=$old_ostcount
9122 run_test 110 "Adding large_dir with 3-level htree"
9125 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
9126 skip "Only applicable to ldiskfs-based MDTs"
9128 is_dm_flakey_dev $SINGLEMDS $(mdsdevname 1) &&
9129 skip "This test can not be executed on flakey dev"
9131 do_facet $SINGLEMDS $DEBUGFS -w -R supported_features |grep large_dir ||
9132 skip "large_dir option is not supported on MDS"
9134 do_facet ost1 $DEBUGFS -w -R supported_features | grep large_dir ||
9135 skip "large_dir option is not supported on OSS"
9137 # cleanup before changing target counts
9139 local old_mdscount=$MDSCOUNT
9140 local old_ostcount=$OSTCOUNT
9141 local old_mdssize=$MDSSIZE
9143 stack_trap "MDSSIZE=$old_mdssize MDSCOUNT=$old_mdscount \
9144 OSTCOUNT=$old_ostcount; reformat"
9147 (( MDSSIZE < 2400000 )) && MDSSIZE=2400000 # need at least 2.4GB
9149 local mdsdev=$(mdsdevname 1)
9150 combined_mgs_mds || replace=" --replace "
9151 local opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
9152 $replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
9153 if [[ $opts != *large_dir* ]]; then
9154 if [[ $opts != *mkfsoptions* ]]; then
9155 opts+=" --mkfsoptions=\\\"-O large_dir -i 1048576 \\\" "
9157 opts="${opts//--mkfsoptions=\\\"/ \
9158 --mkfsoptions=\\\"-O large_dir -i 1048576 }"
9161 echo "MDT params: $opts"
9163 combined_mgs_mds || start_mgs
9164 __touch_device mds 1
9165 add mds1 $opts || error "add mds1 failed with new params"
9166 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
9168 opts="$(mkfs_opts ost1 $(ostdevname 1)) \
9169 $replace --reformat $(ostdevname 1) $(ostvdevname 1)"
9170 if [[ $opts != *large_dir* ]]; then
9171 if [[ $opts != *mkfsoptions* ]]; then
9172 opts+=" --mkfsoptions=\\\"-O large_dir \\\""
9174 opts="${opts//--mkfsoptions=\\\"/ --mkfsoptions=\\\"-O large_dir }"
9177 echo "OST params: $opts"
9178 __touch_device ost 1
9179 add ost1 $opts || error "add ost1 failed with new params"
9180 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
9182 MOUNT_2=yes mountcli
9183 mkdir $DIR/$tdir || error "cannot create $DIR/$tdir"
9185 lfs df -i $DIR/$tdir
9189 local start=$SECONDS
9191 local dirmax=$((2 << 30))
9197 while (( !needskip & dirsize < dirmax )); do
9200 for cli in ${CLIENTS//,/ }; do
9201 local len=$((253 - $(wc -c <<<"$cli-$group-60000-")))
9202 local target=$cli-$group
9203 local long=$DIR/$tdir/$target-$(generate_name $len)-
9205 RPWD=$DIR/$tdir do_node $cli touch $target ||
9206 error "creating $target failed"
9207 echo "creating 60000 hardlinks to $target"
9208 RPWD=$DIR/$tdir do_node $cli createmany -l $target $long 60000 &
9211 group=$((group + 1))
9213 long=$DIR2/$tdir/$target-$(generate_name $len)-
9215 RPWD=$DIR2/$tdir do_node $cli touch $target ||
9216 error "creating $target failed"
9217 echo "creating 60000 hardlinks to $target"
9218 RPWD=$DIR2/$tdir do_node $cli createmany -l $target $long 60000 &
9221 group=$((group + 1))
9223 echo "waiting for PIDs$pids to complete"
9224 wait $pids || error "createmany failed after $group groups"
9225 dirsize=$(stat -c %s $DIR/$tdir)
9226 taken=$((SECONDS - start))
9227 rate=$((dirsize / taken))
9228 left=$(((dirmax - dirsize) / rate))
9229 num=$((group * 60000))
9230 echo "estimate ${left}s left after $num files / ${taken}s"
9231 # if the estimated time remaining is too large (it may change
9232 # over time as the create rate is not constant) then exit
9233 # without declaring a failure.
9234 (( left > 1200 )) && needskip=1
9237 umount_client $MOUNT2 -f
9240 ! (( $needskip )) ||
9241 echo "ETA ${left}s after $num files / ${taken}s is too long"
9243 run_e2fsck $(facet_active_host mds1) $(mdsdevname 1) -n
9244 MDSCOUNT=$old_mdscount
9245 OSTCOUNT=$old_ostcount
9246 MDSSIZE=$old_mdssize
9248 run_test 111 "Adding large_dir with over 2GB directory"
9251 local param="no_create"
9253 (( $OST1_VERSION > $(version_code 2.14.0) )) ||
9254 skip "need OSS at least 2.14.0"
9255 (( $OST1_VERSION >= $(version_code 2.15.56.125) )) ||
9256 param="no_precreate"
9258 start_mds || error "MDS start failed"
9259 start_ost || error "OSS start failed"
9260 echo "start ost2 service on $(facet_active_host ost2)"
9261 start ost2 $(ostdevname 2) $(csa_add "$OST_MOUNT_OPTS" -o $param) ||
9262 error "start ost2 facet failed"
9263 local val=$(do_facet ost2 \
9264 "$LCTL get_param -n obdfilter.$FSNAME-OST0001*.$param")
9265 (( $val == 1 )) || error "obdfilter.$FSNAME-OST0001*.$param=$val"
9267 mount_client $MOUNT || error "mount client failed"
9268 wait_osc_import_state mds1 ost1 FULL
9269 wait_osc_import_ready client ost1
9270 wait_osc_import_ready client ost2
9272 $LFS setstripe -i 0 $DIR/$tfile.0 ||
9273 error "problem creating $tfile.0 on OST0000"
9274 $LFS setstripe -i 1 $DIR/$tfile.1 && $LFS getstripe $DIR/$tfile.1 &&
9275 (( $($LFS getstripe -i $DIR/$tfile.1) == 1 )) &&
9276 error "allowed to create $tfile.1 on OST0001"
9278 $LFS df -v $MOUNT | grep -q "OST:1.*N" ||
9279 error "NOCREATE not in 'lfs df'"
9281 do_facet ost2 $LCTL set_param obdfilter.$FSNAME-OST0001*.$param=0
9283 $LFS setstripe -i 1 $DIR/$tfile.2 ||
9284 error "failed to create $tfile.2 on ost1 facet"
9285 # files not cleaned with ONLY_REPEAT because of client unmount below
9286 rm -f $DIR/$tfile.[012]
9287 stop_ost2 || error "stop ost2 facet failed"
9290 run_test 112a "mount OST with no_create option"
9293 (( MDSCOUNT >= 2 )) || skip "need at least 2 MDTs"
9294 (( $MDS1_VERSION >= $(version_code 2.15.56.125) )) ||
9295 skip "need MDS >= 2.15.56.125"
9296 local mdsnum=$MDSCOUNT
9297 local facet=mds$mdsnum
9298 local mdtidx=$((mdsnum - 1))
9299 local mdtname=$FSNAME-MDT$(printf %04x $mdtidx)
9301 start_mds --mdscount $((mdsnum - 1)) || error "MDS start failed"
9302 start_mdt $mdsnum -o no_create || error "start $facet failed"
9303 local val=$(do_facet $facet \
9304 "$LCTL get_param -n mdt.$mdtname*.no_create")
9305 (( $val == 1 )) || error "mdt.$mdtname*.no_create=$val"
9306 start_ost || error "ost1 start failed"
9307 start_ost2 || error "ost1 start failed"
9309 mount_client $MOUNT || error "mount client failed"
9310 wait_osc_import_ready $facet ost2
9313 $LFS df -v $MOUNT | grep -q "MDT:$mdtidx.*N" ||
9314 error "NOCREATE not in 'lfs df'"
9316 $LFS mkdir -i $mdtidx $DIR/$tdir ||
9317 $LFS setdirstripe -D -c 1 -i -1 --max-inherit-rr 2 $DIR/$tdir ||
9318 error "error creating $tdir on $mdtname"
9319 stack_trap "rm -rf $DIR/$tdir"
9321 mkdir $DIR/$tdir/d1.{1..100} || error "mkdir $tdir/d1.{1..100} failed"
9322 $LFS getdirstripe -i $DIR/$tdir/d1.* | sort | uniq -c
9323 do_facet $facet $LCTL set_param mdt.$mdtname*.no_create=0
9324 # allow one initial create for delayed statfs on client
9325 (( $($LFS getdirstripe -i $DIR/$tdir/d1.* | grep -c $mdtidx) < 2 )) ||
9326 error "allowed create on $mdtname"
9329 mkdir $DIR/$tdir/d2.{1..100} || error "mkdir $tdir/d2.{1..100} failed"
9330 $LFS getdirstripe -i $DIR/$tdir/d2.{1..100} | sort | uniq -c
9331 (( $($LFS getdirstripe -i $DIR/$tdir/d2.* | grep -c $mdtidx) > 10 )) ||
9332 error "no create on $mdtname"
9333 # files not cleaned with ONLY_REPEAT because of client unmount below
9335 stop_ost2 || error "ost1 start failed"
9338 run_test 112b "mount MDT with no_create option"
9341 SAVE_MGS_MOUNT_OPTS=$MGS_MOUNT_OPTS
9342 SAVE_MDS_MOUNT_OPTS=$MDS_MOUNT_OPTS
9343 SAVE_OST_MOUNT_OPTS=$OST_MOUNT_OPTS
9349 MGS_MOUNT_OPTS=$SAVE_MGS_MOUNT_OPTS
9350 MDS_MOUNT_OPTS=$SAVE_MDS_MOUNT_OPTS
9351 OST_MOUNT_OPTS=$SAVE_OST_MOUNT_OPTS
9352 # Revert old mount options back
9354 # Subsequent following test requires
9355 # conf-sanity to be in stopall state.
9356 # Force 'stopall' so others following
9361 # Error out with mount info
9363 local server_nodes=$(comma_list $(mdts_nodes) $(osts_nodes))
9366 echo "--Client Mount Info--"
9367 mount | grep -i lustre
9368 echo "--Server Mount Info--"
9369 do_nodes $server_nodes mount | grep -i lustre
9375 local ost_version="2.15.51" # Minimum version required
9377 (( OST1_VERSION >= $(version_code $ost_version) )) ||
9378 skip "Need server version at least $ost_version"
9380 stack_trap cleanup_113 EXIT
9382 # Reset before starting
9386 # Verify MDS's should start with "rw"
9387 do_facet $SINGLEMDS mount | grep "lustre.*rw,.*MDT" ||
9388 error_113 "$SINGLEMDS should be read-write"
9390 # Verify OST's should start with "rw"
9391 for (( i=1; i <= OSTCOUNT; i++ )); do
9392 do_facet ost$i mount | grep "lustre.*rw,.*OST" ||
9393 error_113 "ost$i should be read-write"
9396 # rdonly_dev does not currently work for ldiskfs
9397 # We skip the rdonly_dev check until then.
9398 if [[ $ost1_FSTYPE == ldiskfs ]]; then
9399 echo "Shadow Mountpoint correctly reports rw for ldiskfs"
9404 # Only ZFS specific tests below.
9407 # Must stop all (server+client) and restart to verify new
9411 # add rdonly_dev to mount option
9412 MGS_MOUNT_OPTS=$(csa_add "$MGS_MOUNT_OPTS" -o rdonly_dev)
9413 MDS_MOUNT_OPTS=$(csa_add "$MDS_MOUNT_OPTS" -o rdonly_dev)
9414 OST_MOUNT_OPTS=$(csa_add "$OST_MOUNT_OPTS" -o rdonly_dev)
9416 # Only restart server(mds/ost). Sufficient for test
9417 setupall server_only || error "Fail to start servers"
9419 # Verify MDS's should be "ro"
9420 do_facet $SINGLEMDS mount | grep "lustre.*ro,.*MDT.*rdonly_dev" ||
9421 error_113 "$SINGLEMDS should be read-only"
9423 # Verify OST's should be "ro"
9424 for (( i=1; i <= OSTCOUNT; i++ )); do
9425 do_facet ost$i mount | grep "lustre.*ro,.*OST.*rdonly_dev" ||
9426 error_113 "ost$i should be read-only"
9429 run_test 113 "Shadow mountpoint correctly report ro/rw for mounts"
9432 # Purpose: To verify dynamic thread (OSS) creation.
9433 # (This was sanity/115)
9436 [ $PARALLEL == "yes" ] && skip "skip parallel run"
9437 remote_ost_nodsh && skip "remote OST with nodsh"
9439 # Lustre does not stop service threads once they are started.
9440 # Reset number of running threads to default.
9445 local save_params="$TMP/sanity-$TESTNAME.parameters"
9447 # Get ll_ost_io count before I/O
9448 OSTIO_pre=$(do_facet ost1 \
9449 "$LCTL get_param ost.OSS.ost_io.threads_started | cut -d= -f2")
9450 # Exit if lustre is not running (ll_ost_io not running).
9451 [ -z "$OSTIO_pre" ] && error "no OSS threads"
9453 echo "Starting with $OSTIO_pre threads"
9454 local thread_max=$((OSTIO_pre * 2))
9455 local rpc_in_flight=$((thread_max * 2))
9457 # this is limited to OSC_MAX_RIF_MAX (256)
9458 [ $rpc_in_flight -gt 256 ] && rpc_in_flight=256
9459 thread_max=$((rpc_in_flight / 2))
9460 [ $thread_max -le $OSTIO_pre ] && skip "Too many ost_io threads" &&
9463 # Number of I/O Process proposed to be started.
9465 local facets=$(get_facets OST)
9467 save_lustre_params client "osc.*OST*.max_rpcs_in_flight" > $save_params
9468 save_lustre_params $facets "ost.OSS.ost_io.threads_max" >> $save_params
9470 # Set in_flight to $rpc_in_flight
9471 $LCTL set_param osc.*OST*.max_rpcs_in_flight=$rpc_in_flight ||
9472 error "Failed to set max_rpcs_in_flight to $rpc_in_flight"
9473 nfiles=${rpc_in_flight}
9474 # Set ost thread_max to $thread_max
9475 do_facet ost1 "$LCTL set_param ost.OSS.ost_io.threads_max=$thread_max"
9477 # 5 Minutes should be sufficient for max number of OSS
9478 # threads(thread_max) to be created.
9482 local wtl=${WTL:-"$LUSTRE/tests/write_time_limit"}
9484 test_mkdir $DIR/$tdir
9485 for ((i = 1; i <= nfiles; i++)); do
9486 local file=$DIR/$tdir/${tfile}-$i
9488 $LFS setstripe -c -1 -i 0 $file
9489 ($wtl $file $timeout)&
9492 # I/O Started - Wait for thread_started to reach thread_max or report
9493 # error if thread_started is more than thread_max.
9494 echo "Waiting for thread_started to reach thread_max"
9495 local thread_started=0
9496 local end_time=$((SECONDS + timeout))
9498 while [ $SECONDS -le $end_time ] ; do
9500 # Get ost i/o thread_started count.
9501 thread_started=$(do_facet ost1 \
9503 ost.OSS.ost_io.threads_started | cut -d= -f2")
9504 # Break out if thread_started is equal/greater than thread_max
9505 if (( $thread_started >= $thread_max )); then
9506 echo ll_ost_io thread_started $thread_started, \
9507 equal/greater than thread_max $thread_max
9513 # Cleanup - We have the numbers, Kill i/o jobs if running.
9514 jobcount=($(jobs -p))
9516 for ((i=0; i < ${#jobcount[*]}; i++)); do
9517 kill -9 ${jobcount[$i]}
9518 if [ $? -ne 0 ] ; then
9519 echo "warning: cannot kill WTL pid ${jobcount[$i]}"
9523 # Cleanup files left by WTL binary.
9524 for ((i = 1; i <= nfiles; i++)); do
9525 local file=$DIR/$tdir/${tfile}-$i
9528 if [ $? -ne 0 ] ; then
9529 echo "Warning: Failed to delete file $file"
9533 restore_lustre_params <$save_params
9534 rm -f $save_params || echo "Warning: delete file '$save_params' failed"
9536 # Error out if no new thread has started or Thread started is greater
9538 if (( $thread_started <= $OSTIO_pre ||
9539 $thread_started > $thread_max )); then
9540 error "ll_ost_io: thread_started $thread_started" \
9541 "OSTIO_pre $OSTIO_pre, thread_max $thread_max." \
9542 "No new thread started or thread started greater " \
9546 run_test 114 "verify dynamic thread creation===================="
9552 do_facet mds1 rm -f $1
9556 if [ "$mds1_FSTYPE" != ldiskfs ]; then
9557 skip "Only applicable to ldiskfs-based MDTs"
9559 [ -n "$FILESET" ] && skip "Not functional for FILESET set"
9561 local dbfs_ver=$(do_facet mds1 $DEBUGFS -V 2>&1)
9563 echo "debugfs version: $dbfs_ver"
9564 echo "$dbfs_ver" | egrep -w "1.44.3.wc1|1.44.5.wc1|1.45.2.wc1" &&
9565 skip_env "This version of debugfs doesn't show inode number"
9567 local IMAGESIZE=$((3072 << 30)) # 3072 GiB
9577 do_facet mds1 "hostname; $LCTL dl; mount"
9579 do_facet ost1 "hostname; $LCTL dl; mount"
9580 # We need MDT size 3072GB, because it is smallest
9581 # partition that can store 2B inodes
9582 do_facet mds1 "mkdir -p $TMP/$tdir"
9583 local mdsimgname=$TMP/$tdir/lustre-mdt
9585 do_facet mds1 "rm -f $mdsimgname"
9586 do_facet mds1 "touch $mdsimgname"
9587 stack_trap "cleanup_115 $mdsimgname" EXIT
9588 do_facet mds1 "$TRUNCATE $mdsimgname $IMAGESIZE" ||
9589 skip "Backend FS doesn't support sparse files"
9590 local mdsdev=$(do_facet mds1 "losetup -f")
9592 do_facet mds1 "losetup $mdsdev $mdsimgname"
9594 local mds_opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
9595 --mkfsoptions='-O ea_inode,^resize_inode,meta_bg \
9596 -N 2247484000 -E lazy_itable_init' --device-size=$IMAGESIZE"
9597 add mds1 $mds_opts --mgs --reformat $mdsdev || {
9598 do_facet $SINGLEMDS \
9599 "losetup -d $mdsdev && rm -f $mdsimgname"
9600 skip_env "format large MDT failed"
9603 local ostdev=$(ostdevname 1)
9605 local opts="$(mkfs_opts ost1 $ostdev) \
9606 --reformat $ostdev $ostdev"
9607 add ost1 $opts || error "add ost1 failed with new params"
9608 start mds1 $mdsdev $MDS_MOUNT_OPTS || error "start MDS failed"
9609 start_ost || error "start OSS failed"
9610 mount_client $MOUNT || error "mount client failed"
9612 mkdir -p $DIR/$tdir || error "mkdir $DIR/$tdir fail"
9613 local goal="/sys/fs/ldiskfs/$(basename $mdsdev)/inode_goal"
9615 # 2147483648 is 0x80000000
9616 do_facet mds1 "echo 2147483648 >> $goal; grep . $goal"
9617 touch $DIR/$tdir/$tfile
9619 # attrs from 1 to 16 go to block, 17th - to inode
9622 for i in {1..17}; do
9623 local nm="trusted.ea$i"
9624 setfattr -n $nm -v $(printf "xattr%0250d" $i) $DIR/$tdir/$tfile
9627 do_facet mds1 "$DEBUGFS -c -R 'stat ROOT/$tdir/$tfile' $mdsdev"
9629 # inode <2147483649> trusted.ea16 (255)
9630 local inode_num=$(do_facet mds1 \
9631 "$DEBUGFS -c -R 'stat ROOT/$tdir/$tfile' $mdsdev" |
9632 awk '/ea17/ { print $2 }' |
9633 sed -e 's/>//' -e 's/<//' -e 's/\"//')
9634 echo "inode num: $inode_num"
9635 [ $inode_num -ge 2147483648 ] || error "inode $inode_num too small"
9636 do_facet mds1 "losetup -d $mdsdev"
9638 run_test 115 "Access large xattr with inodes number over 2TB"
9641 [ "$mds1_FSTYPE" != ldiskfs ] && skip "ldiskfs only test"
9642 [ "$MDS1_VERSION" -lt $(version_code 2.10.59) ] &&
9643 skip "Need server version at least 2.10.59"
9644 do_facet $SINGLEMDS which mkfs.xfs ||
9645 skip_env "No mkfs.xfs installed"
9650 local tmpmnt=$TMP/$tdir
9651 local mdtimg=$tfile-mdt0
9653 do_facet $SINGLEMDS mkdir -p $tmpmnt
9654 stack_trap "do_facet $SINGLEMDS rmdir $tmpmnt" EXIT
9656 do_facet $SINGLEMDS touch $TMP/$mdtimg
9657 stack_trap "do_facet $SINGLEMDS rm -f $TMP/$mdtimg" EXIT
9658 do_facet $SINGLEMDS mkfs -t xfs -d file,size=1t,name=$TMP/$mdtimg ||
9659 error "mkfs temporary xfs image"
9661 do_facet $SINGLEMDS mount $TMP/$mdtimg $tmpmnt ||
9662 error "mount temporary xfs image"
9663 stack_trap "do_facet $SINGLEMDS umount $tmpmnt" EXIT
9664 local old_mdssize=$MDSSIZE
9665 local old_mdsisize=$MDSISIZE
9667 MDSSIZE=$((17 * 1024 * 1024 * 1024)) # 17T MDT
9668 MDSISIZE=$((16 << 20))
9669 local opts17t="$(mkfs_opts $SINGLEMDS)"
9671 MDSSIZE=$old_mdssize
9672 MDSISIZE=$old_mdsisize
9674 do_facet $SINGLEMDS $MKFS \
9675 ${opts17t/-E /-E lazy_itable_init,lazy_journal_init,} \
9677 error "failed to mkfs for $tmpmnt/$mdtimg"
9679 do_facet $SINGLEMDS $TUNE2FS -l $tmpmnt/$mdtimg |
9680 grep -qw 'features.*extent' || error "extent should be enabled"
9683 run_test 116 "big size MDT support"
9686 # Call setup only if LustreFS is not mounted
9687 check_mount || setup
9689 do_facet ost1 "$LCTL set_param ost.OSS.ost_io.nrs_policies=fifo"
9690 do_facet ost1 "$LCTL get_param -n ost.OSS.ost_io.nrs_tbf_rule" &&
9691 error "get_param should fail"
9692 cleanup || error "cleanup failed with rc $?"
9694 run_test 117 "lctl get_param return errors properly"
9700 [[ "$MDSCOUNT" -ge 2 ]] || skip "Need more at least 2 MDTs"
9702 had_config=$(do_facet mds1 "$LCTL get_param debug | grep config")
9703 do_facet mds1 "$LCTL set_param debug=+config"
9704 do_facet mds1 "$LCTL clear"
9707 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)" &>/dev/null
9708 # mount after writeconf will make "add osp" added to mdt0 config:
9709 # 53 (224)marker 60 (flags=0x01, v2.5.1.0) lustre-MDT0001 'add osp'
9710 # 54 (080)add_uuid nid=... 0: 1:...
9711 # 55 (144)attach 0:lustre-MDT0001-osp-MDT0000 1:osp 2:...
9712 # 56 (144)setup 0:lustre-MDT0001-osp-MDT0000 1:... 2:...
9713 # 57 (136)modify_mdc_tgts add 0:lustre-MDT0000-mdtlov 1:... 2:1 3:1
9714 # duplicate modify_mdc_tgts caused crashes
9717 # using larger debug_mb size to avoid lctl dk log truncation
9718 size_mb=$((DEBUG_SIZE_SAVED * 4))
9721 # though config processing stops after failed attach and setup
9722 # it will proceed after the failed command after each writeconf
9723 # this is the original scenario of the issue
9724 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)" &>/dev/null
9725 do_facet mds1 "$LCTL set_param debug_mb=$size_mb"
9728 wait_update_facet_cond mds1 \
9729 "$LCTL dk | grep -c Processed.log.$FSNAME-MDT0000" \
9734 [[ -z "$had_config" ]] && do_facet mds1 lctl set_param debug=-config
9738 run_test 119 "writeconf on slave mdt shouldn't duplicate mdc/osp and crash"
9740 test_120() { # LU-11130
9741 [ "$MDSCOUNT" -lt 2 ] && skip "mdt count < 2"
9742 [ "$mds1_FSTYPE" != ldiskfs ] &&
9743 skip "ldiskfs only test"
9744 [ "$MDS1_VERSION" -lt $(version_code 2.11.56) ] &&
9745 skip "Need DNE2 capable MD target with LU-11130 fix"
9749 local mds1host=$(facet_active_host mds1)
9750 local mds1dev=$(mdsdevname 1)
9752 $LFS mkdir -i 1 $DIR/$tdir
9753 $LFS mkdir -i 0 $DIR/$tdir/mds1dir
9755 ln -s foo $DIR/$tdir/bar
9756 mv $DIR/$tdir/bar $DIR/$tdir/mds1dir/bar2 ||
9757 error "cross-target rename failed"
9761 run_e2fsck $mds1host $mds1dev "-n"
9763 run_test 120 "cross-target rename should not create bad symlinks"
9767 start_mgsmds || error "MGS MDS Start failed"
9769 stop_mds || error "Stopping MDSes failed"
9773 stop_mds || error "Stopping MDSes failed"
9775 run_test 121 "failover MGS"
9778 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
9779 [[ "$OST1_VERSION" -ge $(version_code 2.11.53) ]] ||
9780 skip "Need OST version at least 2.11.53"
9783 LOAD_MODULES_REMOTE=true load_modules
9784 #define OBD_FAIL_OFD_SET_OID 0x1e0
9785 do_facet ost1 $LCTL set_param fail_loc=0x00001e0
9788 $LFS mkdir -i1 -c1 $DIR/$tdir
9789 $LFS setstripe -i0 -c1 $DIR/$tdir
9790 do_facet ost1 $LCTL set_param fail_loc=0
9791 createmany -o $DIR/$tdir/file_ 1000 ||
9792 error "Fail to create a new sequence"
9796 run_test 122a "Check OST sequence update"
9799 remote_mgs_nodsh && skip "remote MGS with nodsh"
9800 [ -d $MOUNT/.lustre ] || setup
9802 # test old logid format until removal from llog_ioctl.c::str2logid()
9803 if (( $MGS_VERSION < $(version_code 2.99.53) )); then
9804 do_facet mgs $LCTL dl | grep MGS
9805 do_facet mgs "$LCTL --device %MGS llog_print \
9806 \\\\\\\$$FSNAME-client 1 10" ||
9807 error "old llog_print failed"
9810 # test new logid format
9811 if [ $MGS_VERSION -ge $(version_code 2.9.53) ]; then
9812 do_facet mgs "$LCTL --device MGS llog_print $FSNAME-client" ||
9813 error "new llog_print failed"
9816 run_test 123aa "llog_print works with FIDs and simple names"
9819 remote_mgs_nodsh && skip "remote MGS with nodsh"
9820 [[ $MGS_VERSION -gt $(version_code 2.11.51) ]] ||
9821 skip "Need server with working llog_print support"
9823 [ -d $MOUNT/.lustre ] || setup
9829 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9830 mgs_arg="--device MGS"
9832 orig_val=$(do_facet mgs $LCTL get_param jobid_name)
9833 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME"
9835 yaml=$(do_facet mgs $LCTL $mgs_arg llog_print params |
9836 grep jobid_name | tail -n 1)
9838 local param=$(awk '{ print $10 }' <<< "$yaml")
9839 local val=$(awk '{ print $12 }' <<< "$yaml")
9840 #return to the default
9841 do_facet mgs $LCTL set_param -P jobid_name=$orig_val
9842 [ $val = "TESTNAME" ] || error "bad value: $val"
9843 [ $param = "jobid_name," ] || error "Bad param: $param"
9845 run_test 123ab "llog_print params output values from set_param -P"
9847 test_123ac() { # LU-11566
9848 remote_mgs_nodsh && skip "remote MGS with nodsh"
9849 do_facet mgs "$LCTL help llog_print" 2>&1 | grep -q -- --start ||
9850 skip "Need 'lctl llog_print --start' on MGS"
9856 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9857 mgs_arg="--device MGS"
9859 [ -d $MOUNT/.lustre ] || setup
9861 # - { index: 10, event: add_uuid, nid: 192.168.20.1@tcp(0x20000c0a81401,
9862 # node: 192.168.20.1@tcp }
9863 do_facet mgs $LCTL $mgs_arg \
9864 llog_print --start $start --end $end $FSNAME-client | tr -d , |
9865 while read DASH BRACE INDEX idx EVENT BLAH BLAH BLAH; do
9866 (( idx >= start )) || error "llog_print index $idx < $start"
9867 (( idx <= end )) || error "llog_print index $idx > $end"
9870 run_test 123ac "llog_print with --start and --end"
9872 test_123ad() { # LU-11566
9873 remote_mgs_nodsh && skip "remote MGS with nodsh"
9874 # older versions of lctl may not print all records properly
9875 do_facet mgs "$LCTL help llog_print" 2>&1 | grep -q -- --start ||
9876 skip "Need 'lctl llog_print --start' on MGS"
9878 [ -d $MOUNT/.lustre ] || setup
9880 # append a new record, to avoid issues if last record was cancelled
9881 local old=$($LCTL get_param -n osc.*-OST0000-*.max_dirty_mb | head -1)
9882 do_facet mgs $LCTL conf_param $FSNAME-OST0000.osc.max_dirty_mb=$old
9884 # logid: [0x3:0xa:0x0]:0
9888 local num=$(do_facet mgs $LCTL --device MGS llog_info $FSNAME-client |
9889 awk '/last_index:/ { print $2 - 1 }')
9891 # - { index: 71, event: set_timeout, num: 0x14, param: sys.timeout=20 }
9892 local last=$(do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
9893 tail -1 | awk '{ print $4 }' | tr -d , )
9894 (( last == num )) || error "llog_print only showed $last/$num records"
9896 run_test 123ad "llog_print shows all records"
9898 test_123ae() { # LU-11566
9906 remote_mgs_nodsh && skip "remote MGS with nodsh"
9907 [ -d $MOUNT/.lustre ] || setupall
9909 max=$($LCTL get_param -n osc.*-OST0000-*.max_dirty_mb | head -1)
9910 pgs=$($LCTL get_param -n osc.*-OST0000-*.max_pages_per_rpc | head -1)
9911 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9912 mgs_arg="--device MGS"
9914 if do_facet mgs "$LCTL help llog_cancel" 2>&1| grep -q -- --log_id; then
9915 # save one set_param -P record in case none exist
9917 do_facet mgs $LCTL set_param -P osc.*.max_pages_per_rpc=$pgs
9918 stack_trap "do_facet mgs $LCTL set_param -P -d \
9919 osc.*.max_pages_per_rpc"
9922 orig=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9923 tail -1 | awk '{ print $4 }' | tr -d , )
9924 do_facet mgs $LCTL set_param -P osc.*.max_dirty_mb=$max
9925 do_facet mgs $LCTL $mgs_arg llog_print $log | tail -1 |
9926 grep "parameter: osc.*.max_dirty_mb" ||
9927 error "new set_param -P wasn't stored in params log"
9929 # - { index: 71, event: set_param, device: general,
9930 # param: osc.*.max_dirty_mb, value: 256 }
9931 id=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9932 tail -1 | awk '{ print $4 }' | tr -d , )
9933 do_facet mgs $LCTL $mgs_arg llog_cancel $log --log_idx=$id
9934 local new=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9935 tail -1 | awk '{ print $4 }' | tr -d , )
9936 (( new == orig )) ||
9937 error "new llog_cancel now $new, not at $orig records"
9940 # test old positional parameters for a while still
9941 if [ "$MGS_VERSION" -le $(version_code 3.1.53) ]; then
9944 do_facet mgs $LCTL conf_param \
9945 $FSNAME-OST0000.osc.max_pages_per_rpc=$pgs
9946 stack_trap "do_facet mgs $LCTL conf_param -d \
9947 $FSNAME-OST0000.osc.max_pages_per_rpc"
9949 orig=$(do_facet mgs $LCTL --device MGS llog_print $log |
9950 tail -1 | awk '{ print $4 }' | tr -d , )
9951 do_facet mgs $LCTL conf_param $FSNAME-OST0000.osc.max_dirty_mb=$max
9952 do_facet mgs $LCTL --device MGS llog_print $log |
9953 tail -1 | grep "parameter: osc.max_dirty_mb" ||
9954 error "old conf_param wasn't stored in params log"
9955 do_facet mgs $LCTL --device MGS llog_print $log
9956 # - { index: 71, event: conf_param, device: testfs-OST0000-osc,
9957 # param: osc.max_dirty_mb=256 }
9958 id=$(do_facet mgs $LCTL --device MGS llog_print $log |
9959 tail -1 | awk '{ print $4 }' | tr -d , )
9960 do_facet mgs $LCTL --device MGS llog_cancel $log $id
9961 do_facet mgs $LCTL --device MGS llog_print $log
9962 new=$(do_facet mgs $LCTL --device MGS llog_print $log |
9963 tail -1 | awk '{ print $4 }' | tr -d , )
9964 (( new == orig )) ||
9965 error "old llog_cancel now $new, not at $orig records"
9968 run_test 123ae "llog_cancel can cancel requested record"
9970 test_123af() { #LU-13609
9971 [ "$MGS_VERSION" -ge $(version_code 2.13.54) -a \
9972 "$MDS1_VERSION" -ge $(version_code 2.13.54) ] ||
9973 skip "Need both MGS and MDS version at least 2.13.54"
9975 [ -d $MOUNT/.lustre ] || setupall
9976 stack_trap "do_facet mds1 $LCTL set_param fail_loc=0" EXIT
9986 for device in "MGS" "$FSNAME-MDT0000"; do
9987 cmd="--device $device llog_catlist"
9988 echo "lctl $cmd ..."
9989 if [ "$device" = "MGS" ]; then
9994 orig_clist=($(do_facet $facet $LCTL $cmd | awk '{ print $2 }'))
9995 orig_count=${#orig_clist[@]}
9996 echo "orig_clist: ${orig_clist[*]}"
9998 #define OBD_FAIL_CATLIST 0x131b
9999 #fetch to llog records from the second one
10000 do_facet $facet $LCTL set_param fail_loc=0x131b fail_val=2
10002 new_clist=($(do_facet $facet $LCTL $cmd | awk '{ print $2 }'))
10003 new_count=${#new_clist[@]}
10004 echo "new_clist: ${new_clist[*]}"
10006 [ $new_count -eq $((orig_count - 1)) ] ||
10007 error "$new_count != $orig_count - 1"
10008 for i in $(seq 0 $new_count); do
10010 [ "${orig_clist[$j]}" = "${new_clist[$i]}" ] ||
10011 error "${orig_clist[$j]} != ${new_clist[$i]}"
10013 do_facet mds1 $LCTL set_param fail_loc=0
10017 run_test 123af "llog_catlist can show all config files correctly"
10019 test_123ag() { # LU-15142
10023 remote_mgs_nodsh && skip "remote MGS with nodsh"
10024 (( $MGS_VERSION >= $(version_code 2.14.55) )) ||
10025 skip "Need server version least 2.14.55"
10027 [ -d $MOUNT/.lustre ] || setup
10029 orig_val=$(do_facet mgs $LCTL get_param jobid_name)
10030 stack_trap "do_facet mgs $LCTL set_param -P jobid_name=$orig_val"
10032 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME1"
10033 do_facet mgs $LCTL set_param -P -d jobid_name
10034 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
10035 grep -c jobid_name)
10036 (( rec == 0 )) || error "parameter was not deleted, check #1"
10037 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME1"
10038 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
10039 grep -c jobid_name)
10040 (( rec == 1)) || error "parameter is not set"
10041 # usage with ordinary set_param format works too
10042 do_facet mgs $LCTL set_param -P -d jobid_name="ANY"
10043 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
10044 grep -c jobid_name)
10045 (( rec == 0 )) || error "parameter was not deleted, check #2"
10047 run_test 123ag "llog_print skips values deleted by set_param -P -d"
10049 test_123ah() { #LU-7668 del_ost
10050 (( "$MGS_VERSION" >= $(version_code 2.15.51) &&
10051 "$MDS1_VERSION" >= $(version_code 2.15.51) )) ||
10052 skip "Need MGS/MDS version >= 2.15.51 for del_ost"
10054 [ -d $MOUNT/.lustre ] || setupall
10055 stack_trap "do_facet mds1 $LCTL set_param fail_loc=0" EXIT
10057 local cmd="--device MGS llog_print"
10059 cli_llogcnt_orig=$(do_facet mgs $LCTL $cmd $FSNAME-client |
10060 grep -c $FSNAME-OST0000)
10061 mds1_llogcnt_orig=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 |
10062 grep -c $FSNAME-OST0000)
10064 [ $cli_llogcnt_orig -gt 0 ] ||
10065 error "$FSNAME-OST0000 not found (client)"
10066 [ $mds1_llogcnt_orig -gt 0 ] || error "$FSNAME-OST0000 not found (MDT)"
10068 # -n/--dryrun should NOT modify catalog
10069 do_facet mgs "$LCTL del_ost -n -t $FSNAME-OST0000" ||
10070 error "del_ost --dryrun failed with $?"
10072 local cli_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-client |
10073 grep -c $FSNAME-OST0000)
10074 local mds1_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 |
10075 grep -c $FSNAME-OST0000)
10077 [ $cli_llogcnt -eq $cli_llogcnt_orig ] ||
10078 error "--dryrun error: $cli_llogcnt != $cli_llogcnt_orig"
10079 [ $mds1_llogcnt -eq $mds1_llogcnt_orig ] ||
10080 error "--dryrun error: $mds1_llogcnt != $mds1_llogcnt_orig"
10083 do_facet mgs "$LCTL del_ost --target $FSNAME-OST0000" ||
10084 error "del_ost failed with $?"
10086 local cli_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-client |
10087 grep -c $FSNAME-OST0000)
10088 local mds1_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 |
10089 grep -c $FSNAME-OST0000)
10091 # every catalog entry for OST0000 should have been cancelled
10092 [ $cli_llogcnt -eq 0 ] || error "$FSNAME-OST0000 not cancelled (cli)"
10093 [ $mds1_llogcnt -eq 0 ] || error "$FSNAME-OST0000 not cancelled (MDT)"
10095 umount_client $MOUNT
10096 mount_client $MOUNT
10098 $LFS df | grep -q OST0000 && error "del_ost did not remove OST0000!"
10100 reformat_and_config
10102 run_test 123ah "del_ost cancels config log entries correctly"
10107 echo "cleanup test 123ai"
10109 # cancel last timeout record
10110 do_facet mgs "$LCTL set_param -P -d timeout"
10112 # restore timeout value
10113 do_nodes $(comma_list $(all_nodes)) "$LCTL set_param timeout=$timeout"
10116 test_123ai() { #LU-16167
10122 remote_mgs_nodsh && skip "remote MGS with nodsh"
10123 (( MDS1_VERSION >= $(version_code 2.15.51) )) ||
10124 skip "Need MDS version at least 2.15.51"
10126 [ -d $MOUNT/.lustre ] || setup
10128 old_count=$(do_facet mgs "$LCTL --device MGS llog_print -r params" |
10129 grep -c "parameter: timeout")
10131 old_timeout=$($LCTL get_param -n timeout)
10132 stack_trap "cleanup_123ai $old_timeout" EXIT
10134 # add and cancel params (2 * MAX_IOC_BUFLEN / 128 + 1 = 129)
10135 for i in {1..129}; do
10136 do_facet mgs "$LCTL set_param -P timeout=$i" ||
10137 error "fail to set timeout=$i on MGS"
10140 count=$(do_facet mgs "$LCTL --device MGS llog_print -r params" |
10141 grep -c "parameter: timeout")
10143 ((count - old_count == 129)) ||
10144 error "number timeout record missmatch ($count - $old_count != 129)"
10146 do_facet mgs "$LCTL --device MGS llog_print params" |
10147 tail -1 | grep "timeout, value: 129" ||
10148 error "llog_print could not display the last record (timeout=129)"
10151 run_test 123ai "llog_print display all non skipped records"
10154 remote_mgs_nodsh && skip "remote MGS with nodsh"
10156 [ -d $MOUNT/.lustre ] || setup
10157 yaml_file="$TMP/$tfile.yaml"
10159 do_facet mgs rm "$yaml_file"
10160 cfgfiles=$(do_facet mgs "lctl --device MGS llog_catlist" |
10161 sed 's/config_log://')
10163 # set jobid_var to a different value for test
10164 orig_val=$(do_facet mgs $LCTL get_param jobid_var)
10166 do_facet mgs $LCTL set_param -P jobid_var="TESTNAME"
10168 for i in $cfgfiles params; do
10169 do_facet mgs "lctl --device MGS llog_print ${i} >> $yaml_file"
10172 echo "Unmounting FS"
10180 test_123_restore() {
10181 set_val=$(do_facet mgs $LCTL get_param jobid_var)
10183 do_facet mgs $LCTL set_param -P $orig_val
10185 [ $set_val == "jobid_var=TESTNAME" ] ||
10186 error "$set_val is not TESTNAME"
10188 do_facet mgs rm "$yaml_file"
10200 # Reapply the config from before
10201 echo "Setting configuration parameters"
10202 do_facet mgs "lctl set_param -F $yaml_file"
10206 run_test 123F "clear and reset all parameters using set_param -F"
10216 # Reapply the config from before
10217 echo "Setting configuration parameters"
10218 do_facet mgs "lctl apply_yaml $yaml_file"
10222 run_test 123G "clear and reset all parameters using apply_yaml"
10227 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
10228 [ -z $mds2failover_HOST ] && skip "needs MDT failover setup"
10231 do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
10232 grep 1.2.3.4@tcp && error "Should not be fake nid"
10233 do_facet mgs $LCTL conf_param $FSNAME-MDT0001.failover.node=1.2.3.4@tcp\
10234 || error "Set params error"
10235 do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
10236 grep 1.2.3.4@tcp || error "Fake nid should be added"
10240 if combined_mgs_mds; then
10241 start_mdt 1 "-o nosvc" ||
10242 error "starting mds with nosvc option failed"
10244 local nid=$(do_facet mds2 $LCTL list_nids | head -1)
10245 local failover_nid=$(do_node $mds2failover_HOST $LCTL list_nids | head -1)
10247 do_facet mgs $LCTL replace_nids $FSNAME-MDT0001 $nid:$failover_nid ||
10248 error "replace_nids execution error"
10250 if combined_mgs_mds; then
10256 echo "lfs setdirstripe"
10257 $LFS setdirstripe -i 1 $MOUNT/$tdir || error "setdirstirpe error"
10260 run_test 124 "check failover after replace_nids"
10262 get_max_sectors_kb() {
10265 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
10266 local max_sectors_path="/sys/block/${dev_base}/queue/max_sectors_kb"
10268 do_facet ${facet} "[[ -e ${max_sectors_path} ]] &&
10269 cat ${max_sectors_path}"
10272 get_max_hw_sectors_kb() {
10275 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
10276 local max_hw_path="/sys/block/${dev_base}/queue/max_hw_sectors_kb"
10278 do_facet ${facet} "[[ -e ${max_hw_path} ]] && cat ${max_hw_path}"
10281 set_max_sectors_kb() {
10285 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
10286 local max_sectors_path="/sys/block/${dev_base}/queue/max_sectors_kb"
10288 do_facet ${facet} "[[ -e ${max_sectors_path} ]] &&
10289 echo ${value} > ${max_sectors_path}"
10292 [[ $rc -ne 0 ]] && echo "Failed to set ${max_sectors_path} to ${value}"
10297 # Return 0 if all slave devices have max_sectors_kb == max_hw_sectors_kb
10298 # Otherwise return > 0
10299 check_slaves_max_sectors_kb()
10303 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
10304 local slaves_dir=/sys/block/${dev_base}/slaves
10305 local slave_devices=$(do_facet ${facet} "ls ${slaves_dir} 2>/dev/null")
10306 [[ -z ${slave_devices} ]] && return 0
10308 local slave max_sectors new_max_sectors max_hw_sectors path
10311 for slave in ${slave_devices}; do
10312 path="/dev/${slave}"
10313 ! is_blkdev ${facet} ${path} && continue
10314 max_sectors=$(get_max_sectors_kb ${facet} ${path})
10315 max_hw_sectors=$(get_max_hw_sectors_kb ${facet} ${path})
10316 new_max_sectors=${max_hw_sectors}
10317 [[ ${new_max_sectors} -gt ${RQ_SIZE_LIMIT} ]] &&
10318 new_max_sectors=${RQ_SIZE_LIMIT}
10320 if [[ ${max_sectors} -ne ${new_max_sectors} ]]; then
10321 echo "${path} ${max_sectors} ${new_max_sectors}"
10324 check_slaves_max_sectors_kb ${facet} ${path}
10333 local facet_list="mgs mds1 ost1"
10334 combined_mgs_mds && facet_list="mgs ost1"
10337 for facet in ${facet_list}; do
10338 [[ $(facet_fstype ${facet}) != ldiskfs ]] &&
10339 skip "ldiskfs only test" &&
10341 ! is_blkdev ${facet} $(facet_device ${facet}) &&
10342 skip "requires all real devices" &&
10347 # We don't increase IO request size limit past 16MB. See comments in
10348 # lustre/utils/libmount_utils_ldiskfs.c:tune_max_sectors_kb()
10349 RQ_SIZE_LIMIT=$((16 * 1024))
10350 local device old_max_sectors new_max_sectors max_hw_sectors
10351 for facet in ${facet_list}; do
10352 device=$(facet_device ${facet})
10353 old_max_sectors=$(get_max_sectors_kb ${facet} ${device})
10354 max_hw_sectors=$(get_max_hw_sectors_kb ${facet} ${device})
10356 # The expected value after l_tunedisk is executed
10357 new_max_sectors=$old_max_sectors
10358 [[ ${new_max_sectors_kb} -gt ${RQ_SIZE_LIMIT} ]] &&
10359 new_max_sectors_kb=${RQ_SIZE_LIMIT}
10361 # Ensure the current value of max_sectors_kb does not equal
10362 # max_hw_sectors_kb, so we can tell whether l_tunedisk did
10364 set_max_sectors_kb ${facet} ${device} $((new_max_sectors - 1))
10366 # Value before l_tunedisk
10367 local pre_max_sectors=$(get_max_sectors_kb ${facet} ${device})
10368 if [[ ${pre_max_sectors} -ne $((new_max_sectors - 1)) ]]; then
10369 echo "unable to satsify test pre-condition:"
10370 echo "${pre_max_sectors} != $((new_max_sectors - 1))"
10375 echo "Before: ${facet} ${device} ${pre_max_sectors} ${max_hw_sectors}"
10377 do_facet ${facet} "l_tunedisk ${device}"
10379 # Value after l_tunedisk
10380 local post_max_sectors=$(get_max_sectors_kb ${facet} ${device})
10382 echo "After: ${facet} ${device} ${post_max_sectors} ${max_hw_sectors}"
10384 if [[ ${facet} != ost1 ]]; then
10385 if [[ ${post_max_sectors} -ne ${pre_max_sectors} ]]; then
10386 echo "l_tunedisk modified max_sectors_kb of ${facet}"
10390 set_max_sectors_kb ${facet} ${device} ${old_max_sectors}
10392 if [[ ${post_max_sectors} -eq ${pre_max_sectors} ]]; then
10393 echo "l_tunedisk failed to modify max_sectors_kb of ${facet}"
10397 check_slaves_max_sectors_kb ${facet} ${device} ||
10404 run_test 125 "check l_tunedisk only tunes OSTs and their slave devices"
10407 [[ "$MDS1_VERSION" -ge $(version_code 2.13.52) ]] ||
10408 skip "Need MDS version at least 2.13.52"
10411 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ../libcfs/libcfs/libcfs
10412 #define OBD_FAIL_OBD_SETUP 0x60d
10413 do_facet mds1 $LCTL set_param fail_loc=0x60d
10414 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_modules &
10415 for i in {1..40}; do
10416 do_facet mds1 lsmod | grep -q osd_$mds1_FSTYPE && break
10419 clear_failloc $SINGLEMDS 20 &
10420 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
10422 run_test 126 "mount in parallel shouldn't cause a crash"
10425 [[ "$ost1_FSTYPE" == ldiskfs ]] || skip "ldiskfs only test"
10429 zconf_umount_clients $RCLIENTS $MOUNT
10431 wait_osp_active ost ${FSNAME}-OST0000 0 1
10432 local osc_tgt="$FSNAME-OST0000-osc-$($LFS getname -i $DIR)"
10433 local avail1=($($LCTL get_param -n osc.${osc_tgt}.kbytesavail))
10435 $LFS setstripe -i 0 $DIR/$tfile || error "failed creating $DIR/$tfile"
10436 dd if=/dev/zero of=$DIR/$tfile bs=1M oflag=direct || true
10438 local avail2=($($LCTL get_param -n osc.${osc_tgt}.kbytesavail))
10440 if ((avail2 * 100 / avail1 > 1)); then
10443 error "more than 1% space left: before=$avail1 after=$avail2"
10446 local mbs=$(($(stat -c %s $DIR/$tfile) / (1024 * 1024)))
10448 dd if=/dev/zero of=$DIR/$tfile bs=1M count=$mbs conv=notrunc \
10449 oflag=direct || error "overwrite failed"
10451 run_test 127 "direct io overwrite on full ost"
10455 combined_mgs_mds && skip "need separate mgs device"
10456 [ "$ost2_FSTYPE" == zfs ] && import_zpool ost2
10459 # Try to apply nolocallogs to the virgin OST. Should fail.
10460 do_facet ost2 "$TUNEFS --nolocallogs $(ostdevname 2)" &&
10461 error "nolocallogs should not be allowed on the virgin target"
10466 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
10467 # Start OST without MGS (local configs)
10468 do_facet ost1 "$TUNEFS --dryrun $(ostdevname 1)"
10469 start_ost || error "unable to start OST1"
10470 stop_ost || error "Unable to stop OST1"
10472 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
10473 # Do not allow reading local configs, should fail
10474 do_facet ost1 "$TUNEFS --nolocallogs $(ostdevname 1)" ||
10475 error "Can not set nolocallogs"
10476 start_ost && error "OST1 started, but should fail"
10478 # Connect to MGS successfully, reset nolocallogs flag
10479 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
10480 start_mgs || error "unable to start MGS"
10481 start_ost || error "unable to start OST1"
10483 do_facet ost1 "$TUNEFS --dryrun $(ostdevname 1)" | grep "nolocallogs" &&
10484 error "nolocallogs expected to be reset"
10486 stop_ost || error "Unable to stop OST1"
10488 run_test 128 "Force using remote logs with --nolocallogs"
10492 (( MDS1_VERSION >= $(version_code 2.14.57) )) ||
10493 skip "Need MDS version at least 2.14.57"
10495 start_mgsmds || error "MGS/MDS start failed"
10497 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &&
10498 error "start ost1 should fail" || true
10499 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &&
10500 error "second start ost1 should fail" || true
10501 do_facet ost1 "$TUNEFS --writeconf $(ostdevname 1)"
10502 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS ||
10503 error "start ost1 failed"
10507 run_test 129 "attempt to connect an OST with the same index should fail"
10511 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
10513 stop_mdt 2 || error "mdt2 stop failed"
10514 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)"
10515 start_mdt 2 || error "mdt2 start failed"
10516 do_facet mds2 "$LCTL dl" | grep MDT0001-osp-MDT0001 &&
10517 error "Illegal OSP device created" || true
10519 run_test 130 "re-register an MDT after writeconf"
10522 [ "$mds1_FSTYPE" == "ldiskfs" ] || skip "ldiskfs only test"
10523 (( $MDS1_VERSION >= $(version_code 2.14.56.35) )) ||
10524 skip "Need MDS version at least 2.14.56.35"
10525 do_facet mds1 $DEBUGFS -R features $(mdsdevname 1) |
10526 grep -q project || skip "skip project quota not supported"
10531 test_mkdir -c $MDSCOUNT -p $DIR/$tdir
10532 $LFS project -p 1000 $DIR/$tdir || error "set dir project id failed"
10533 createmany -o $DIR/$tdir/f 512
10534 for ((i = 0; i < 512; ++i)); do
10535 $LFS project -p $i $DIR/$tdir/f${i} ||
10536 error "set f${i} project id failed"
10539 test_mkdir -c $MDSCOUNT -p $DIR/$tdir.inherit
10540 $LFS project -p 1001 -s $DIR/$tdir.inherit
10541 createmany -o $DIR/$tdir.inherit/f 128
10542 (( $($LFS project $DIR/$tdir.inherit/f* |
10543 awk '$1 == 1001 { print }' | wc -l) == 128 )) ||
10544 error "files did not inherit projid 1001"
10548 for i in $(seq $MDSCOUNT); do
10549 mds_backup_restore mds$i ||
10550 error "Backup/restore on mds$i failed"
10555 projid=($($LFS project -d $DIR/$tdir))
10556 [ ${projid[0]} == "1000" ] ||
10557 error "projid expected 1000 not ${projid[0]}"
10558 for ((i = 0; i < 512; ++i)); do
10559 projid=($($LFS project $DIR/$tdir/f${i}))
10560 [ ${projid[0]} == "$i" ] ||
10561 error "projid expected $i not ${projid[0]}"
10564 (( $($LFS project $DIR/$tdir.inherit/f* |
10565 awk '$1 == 1001 { print }' | wc -l) == 128 )) ||
10566 error "restore did not copy projid 1001"
10568 run_test 131 "MDT backup restore with project ID"
10574 (( MDS1_VERSION >= $(version_code 2.14.57) )) ||
10575 skip "Need MDS version at least 2.14.57"
10577 combined_mgs_mds || start_mgs || error "unable to start MGS"
10578 start_mdt 1 || error "unable to start mdt1"
10580 err_cnt=$(do_facet mds1 dmesg | grep -c "cannot take the layout locks")
10581 stop_mdt 1 || error "stop mdt1 failed"
10583 [ "$mds1_FSTYPE" == zfs ] && import_zpool mds1
10584 do_facet mds1 $TUNEFS --param mdt.hsm_control=enabled $(mdsdevname 1) ||
10585 error "tunefs failed"
10586 start_mdt 1 || error "cannot start mdt1"
10588 err_cnt2=$(do_facet mds1 dmesg | grep -c "cannot take the layout locks")
10589 [ $err_cnt -eq $err_cnt2 ] || error "Can not take the layout lock"
10590 stop_mdt 1 || error "stop mdt1 failed"
10592 run_test 132 "hsm_actions processed after failover"
10594 # This test verifies we do RR allocation within a pool even if there is a
10595 # significant imbalance vs an OST outside the pool
10597 [[ $OSTCOUNT -lt 4 ]] && skip_env "needs >= 4 OSTs"
10598 [[ "$OST1_VERSION" -ge $(version_code 2.15.51) ]] ||
10599 skip "Need OST version at least 2.15.51"
10600 # This is the easiest way to ensure OSTs start out balanced
10601 reformat_and_config
10604 check_set_fallocate_or_skip
10606 local testfile=$DIR/$tdir/$tfile
10607 local pool="testpool"
10608 local ostrange=$((OSTCOUNT - 1))
10609 # Select all but the last OST to add to the pool
10610 local poolostrange=$((OSTCOUNT - 2))
10618 [ $MINV -eq 0 ] && error "no free space in OST$MINI"
10619 [ $MAXV -gt $((2 * $MINV)) ] &&
10620 error "OSTs badly unbalanced after reformat"
10622 create_pool $FSNAME.$pool || error "failed to create a pool"
10623 do_facet mgs $LCTL pool_add $FSNAME.$pool OST[0-$poolostrange] ||
10624 error "failed to add OST[0-$poolostrange] to the pool"
10626 test_mkdir -p $DIR/$tdir || error "failed to mkdir $DIR/$tdir"
10627 # Consume space on the OSTs in the pool so they are unbalanced with the
10628 # OST outside of the pool
10629 # fill each OST 90% with fallocate so they are widely
10631 local size=$(((MINV * 9 / 10) * 1024))
10632 for ((i = 0; i <= poolostrange; i++)); do
10633 $LFS setstripe -c 1 -i $i $testfile$i ||
10634 error "failed to setstripe $testfile$i"
10635 fallocate -l $size $testfile$i || error "fallocate failed"
10641 # Create files in the pool now that there is an imbalance
10642 filecount=$(((OSTCOUNT - 1) * filenum))
10643 for ((i = 0; i < filecount; i++)); do
10644 $LFS setstripe -p $pool $testfile-$i ||
10645 error "failed to setstripe -p $pool $testfile-$i"
10647 $LFS getstripe -i $testfile-* > /tmp/$tfile.log
10648 # Count the number of files with a stripe on each OST to verify the
10649 # pool allocated with round-robin
10650 for ((i = 0; i <= poolostrange; i++)); do
10651 stripecount=$(grep -c $i /tmp/$tfile.log)
10652 # Allow a little leeway
10653 if (( stripecount < filenum - 1 ||
10654 stripecount > filenum + 1 )); then
10655 cat /tmp/$tfile.log
10656 error "$stripecount != $filenum files on OST$i"
10660 # Create files across the system now that there is an imbalance
10661 filecount=$((OSTCOUNT * filenum))
10662 for ((i = 1; i < filecount; i++)); do
10663 $LFS setstripe $testfile-$i.2 ||
10664 error "failed to setstripe $testilfe-$i.2"
10666 $LFS getstripe -i $testfile-*.2 > /tmp/$tfile.log
10668 # Count the number of files with a stripe on each OST to verify the
10669 # files are *NOT* allocated with round-robin
10670 for ((i = 0; i <= ostrange; i++)); do
10671 stripecount=$(grep -c $i /tmp/$tfile.log)
10672 if [[ $stripecount -ne $filenum ]]; then
10674 echo "QOS: $stripecount != $filenum files on OST$i"
10677 if [ -z "$qos_used" ]; then
10678 error "QOS not used on imbalanced OSTs!"
10681 rm -rf /tmp/$tfile.log $DIR/$tdir
10682 do_facet mgs $LCTL pool_remove $FSNAME.$pool OST[0-$poolostrange] ||
10683 "failed to remove OST[0-$poolostrange] from the pool"
10684 do_facet mgs $LCTL pool_destroy $FSNAME.$pool ||
10685 error "failed to destroy pool"
10687 run_test 133 "stripe QOS: free space balance in a pool"
10690 [ "$mds1_FSTYPE" == "ldiskfs" ] || skip "ldiskfs only test"
10693 local mdt_dev=$(facet_device mds1)
10694 local tmp_dir=$TMP/$tdir
10695 local out=$tmp_dir/check_iam.txt
10696 local CHECK_IAM=${CHECK_IAM:-$(do_facet mds1 "which check_iam 2> /dev/null || true")}
10698 [[ -n "$CHECK_IAM" ]] || skip "check_iam not found"
10701 for ((i=0; i<64; i++)); do
10703 #cmd introduce a random corruption to IAM file
10704 local cmd="dd if=/dev/urandom of=$tmp_dir/$f bs=2 conv=notrunc count=1 seek=$((RANDOM % 36))"
10705 do_facet mds1 "mkdir -p $tmp_dir; \
10706 $DEBUGFS -c -R 'dump $f $tmp_dir/$f' $mdt_dev 2>&1; \
10707 $CHECK_IAM -v $tmp_dir/$f 2>&1; \
10708 echo $cmd; eval $cmd 2>/dev/null;
10709 $CHECK_IAM -v $tmp_dir/$f 2>&1; echo \\\$?" >> $out 2>&1
10714 stack_trap "rm -rf $tmp_dir && do_facet mds1 rm -rf $tmp_dir" EXIT
10716 rc=$(grep -c "fault\|except" $out)
10717 (( rc == 0 )) || { cat $out &&
10718 error "check_iam failed with fault or exception $rc"; }
10720 rc=$(grep -c "^255" $out)
10721 errors=$(grep -c "FINISHED WITH ERRORS" $out)
10723 (( rc == errors )) || { cat $out &&
10724 error "check_iam errcode does not fit with errors $rc $errors"; }
10726 run_test 134 "check_iam works without faults"
10728 cleanup_test_135(){
10731 printf "\nCleanup test_135\n" >&2
10732 do_facet mds1 "$LCTL set_param -n $oldgc"
10733 rm -rf $DIR/$tdir &> /dev/null
10737 __test_135_file_thread() {
10739 local init_time=$(awk '{print $1}' /proc/uptime)
10740 local awkcmd="/crosses index zero/ {if (\$1 > $init_time) exit(1);}"
10742 #Generate a full plain llogs
10743 while dmesg | sed -r 's/(\[|\])//g' | awk "$awkcmd" ; do
10744 createmany -o $DIR/$tdir/f 4500 >&2
10745 createmany -u $DIR/$tdir/f 4500 >&2
10749 __test_135_reader() {
10752 local firstidx=$(changelog_user_rec mds1 $cl_user)
10753 local oldidx=$firstidx
10758 while read -t10 -u$fd newidx other; do
10759 (( (newidx - oldidx) == 1 )) ||
10760 error "changelog jump detected (last: $oldidx, current: $newidx)"
10762 if (( (newidx - firstidx + 1) % 13000 == 0 )); then
10763 [[ $pid -eq 0 ]] ||
10764 wait $pid || error "changelog_clear failed"
10765 changelog_clear $((newidx - 1)) mds1 >&2 & pid=$!
10770 [[ $pid -eq 0 ]] ||
10771 wait $pid || error "changelog_clear failed"
10777 (( MDS1_VERSION >= $(version_code 2.15.52) )) ||
10778 skip "need MDS version at least 2.15.52"
10780 local service=$(facet_svc mds1)
10782 local lastread lastidx
10783 local files_pid reader_pid
10788 # Need to reformat because we are changing llog catalog sizes to 5.
10789 # Otherwise, processing could fail with existing catalogs (last_idx>5).
10793 # Disable changelog garbage colector
10794 local oldgc=$(do_facet mds1 "$LCTL get_param mdd.${service}.changelog_gc")
10795 do_facet mds1 "$LCTL set_param -n mdd.${service}.changelog_gc=0"
10796 stack_trap "cleanup_test_135 $oldgc" EXIT INT
10798 # change the changelog_catalog size to 5 entries for everybody
10799 #define OBD_FAIL_CAT_RECORDS 0x1312
10800 do_node $(comma_list $(all_nodes)) $LCTL set_param fail_loc=0x1312 fail_val=5
10802 # disable console ratelimit
10803 local rl=$(cat /sys/module/libcfs/parameters/libcfs_console_ratelimit)
10804 echo 0 > /sys/module/libcfs/parameters/libcfs_console_ratelimit
10805 stack_trap "echo $rl > /sys/module/libcfs/parameters/libcfs_console_ratelimit" EXIT
10807 test_mkdir -c 1 -i 0 $DIR/$tdir || error "Failed to create directory"
10808 changelog_chmask "ALL" || error "changelog_chmask failed"
10809 changelog_register || error "changelog_register failed"
10811 cl_user="${CL_USERS[mds1]%% *}"
10812 changelog_users mds1 | grep -q $cl_user ||
10813 error "User $cl_user not found in changelog_users"
10815 # Start reader thread
10816 coproc $LFS changelog --follow $service
10819 stack_trap "echo kill changelog reader; kill $reader_pid" EXIT
10821 echo -e "\nWrap arround changelog catalog"
10823 # Start file writer thread
10824 __test_135_file_thread "$service" & files_pid=$!
10825 stack_trap "(pkill -P$files_pid; kill $files_pid) &> /dev/null || true" EXIT
10827 # Check changelog entries
10828 lastread=$(__test_135_reader $fd $cl_user) || exit $?
10829 ! kill -0 $files_pid 2>/dev/null ||
10830 error "creation thread is running. Is changelog reader stuck?"
10832 lastidx=$(changelog_users mds1 | awk '/current_index/ {print $NF}' )
10833 [[ "$lastread" -eq "$lastidx" ]] ||
10834 error "invalid changelog lastidx (read: $lastread, mds: $lastidx)"
10836 run_test 135 "check the behavior when changelog is wrapped around"
10839 do_facet mds2 "$LCTL --device ec cleanup" || true
10840 do_facet mds2 "$LCTL --device ec detach" || true
10843 reformat_and_config
10847 (( MDSCOUNT >= 2 )) || skip "needs >= 2 MDTs"
10852 do_facet mds2 "$LCTL attach echo_client ec ec_uuid" ||
10853 error "echo attach fail"
10855 stack_trap cleanup_136 EXIT
10857 do_facet mds2 "$LCTL --device ec setup lustre-MDT0001 mdt" &&
10858 error "attach to MDT should fail!"
10860 do_facet mds2 "$LCTL --device ec setup lustre-MDT0001 mdd" ||
10861 error "attach to MDD should OK"
10862 do_facet mds2 "$LCTL --device ec test_mkdir /tt" &&
10863 error "mkdir test should fail with remote object"
10867 run_test 136 "don't panic with bad obdecho setup"
10870 (( MDS1_VERSION >= $(version_code 2.15.55) )) ||
10871 skip "need MDS version at least 2.15.55"
10872 (( MDSCOUNT >= 2 )) || skip "needs >= 2 MDTs"
10874 RM_UPDATELOG=$(do_facet mds2 "which remove_updatelog 2> /dev/null")
10875 RM_UPDATELOG=${RM_UPDATELOG:-"$LUSTRE/scripts/remove_updatelog"}
10877 [ -f "$RM_UPDATELOG" ] ||
10878 skip_env "remove_updatelog is not found on mds2"
10880 local mntpt=$(facet_mntpt mds2)
10883 test_mkdir -c $MDSCOUNT -p $DIR/$tdir ||
10884 error "mkdir $DIR/$tdir failed"
10885 mkdir $DIR/$tdir/d{1..256}
10890 mount_fstype mds2 || error "mount as fstype mds2 failed"
10891 do_facet mds2 $RM_UPDATELOG -n $mntpt
10892 MDTS=($(do_facet mds2 $RM_UPDATELOG -n $mntpt |
10893 grep -o "Processing MDT[0-9]*" | awk -F'MDT' '{print $2}'))
10894 (( ${#MDTS[@]} == MDSCOUNT )) ||
10895 error "Processed ${#MDTS[@]} from $MDSCOUNT"
10897 do_facet mds2 $RM_UPDATELOG -n -m 1,0 $mntpt
10898 MDTS=($(do_facet mds2 $RM_UPDATELOG -n -m 1,0 $mntpt |
10899 grep -o "Processing MDT[0-9]*" | awk -F'MDT' '{print $2}'))
10900 (( ${#MDTS[@]} == 2 )) ||
10901 error "Processed ${#MDTS[@]} instead of 2"
10902 (( ${MDTS[0]} == 1 && ${MDTS[1]} == 0 )) ||
10903 error "Processed: ${MDTS[*]}, expected: 1 0"
10905 do_facet mds2 $RM_UPDATELOG -m 0 $mntpt
10906 unmount_fstype mds2
10907 start_mdt 2 || error "mds2 start fail"
10908 start_mdt 1 || error "mds1 start fail"
10909 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds1 FULL
10910 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
10912 rm -rf $DIR/$tdir || error "Can't remove $tdir"
10914 reformat_and_config
10916 run_test 140 "remove_updatelog script actions"
10921 local max_cached_mb=$($LCTL get_param llite.*.max_cached_mb |
10922 awk '/^max_cached_mb/ { print $2 }')
10923 stack_trap "$LCTL set_param -n llite.*.max_cached_mb=$max_cached_mb"
10925 $LCTL set_param llite.*.max_cached_mb='100%'
10927 local new_max_cached_mb=$($LCTL get_param llite.*.max_cached_mb |
10928 awk '/^max_cached_mb/ { print $2 }')
10929 local total_ram_mb=$(free -m | grep 'Mem:' | awk '{print $2}')
10931 $LCTL get_param llite.*.max_cached_mb
10932 echo "total ram mb: $total_ram_mb"
10933 (( new_max_cached_mb == total_ram_mb )) ||
10934 error "setting cache to 100% not equal to total RAM"
10936 $LCTL set_param llite.*.max_cached_mb='50%'
10937 new_max_cached_mb=$($LCTL get_param llite.*.max_cached_mb |
10938 awk '/^max_cached_mb/ { print $2 }')
10940 $LCTL get_param llite.*.max_cached_mb
10941 (( new_max_cached_mb == $((total_ram_mb / 2)) )) ||
10942 error "setting cache to 50% not equal to 50% of RAM"
10944 $LCTL set_param llite.*.max_cached_mb='105%' &&
10945 error "should not be able to set insane value"
10947 $LCTL set_param llite.*.max_cached_mb='0%'
10948 new_max_cached_mb=$($LCTL get_param llite.*.max_cached_mb |
10949 awk '/^max_cached_mb/ { print $2 }')
10950 # Minimum cache size is 64 MiB
10951 $LCTL get_param llite.*.max_cached_mb
10952 (( new_max_cached_mb == 64 )) ||
10953 error "setting cache to 0% != minimum cache size"
10955 run_test 150 "test setting max_cached_mb to a %"
10958 (( MDS1_VERSION >= $(version_code 2.15.58) )) ||
10959 skip "need MDS version at least 2.15.58"
10960 [[ "$ost1_FSTYPE" == ldiskfs ]] || skip "ldiskfs only test"
10963 if ! combined_mgs_mds ; then
10967 echo "Damage ost1 local config log"
10968 do_facet ost1 "$DEBUGFS -w -R 'punch CONFIGS/$FSNAME-OST0000 0 1' \
10969 $(ostdevname 1) || return \$?" ||
10970 error "do_facet ost1 failed with $?"
10972 # expect OST to fail mount with no MGS and bad local config
10973 start_ost && error "OST start should fail"
10975 if ! combined_mgs_mds ; then
10978 start_mds || error "MDS start failed"
10979 # now it should start with MGS config
10980 start_ost || error "OST start failed"
10981 reformat_and_config
10983 run_test 151 "damaged local config doesn't prevent mounting"
10986 (( MDS1_VERSION >= $(version_code 2.15.59.53) )) ||
10987 skip "need MDS >= 2.15.59.53 for sequence allocation retry"
10988 (( MDSCOUNT >= 2 )) || skip "needs >= 2 MDTs"
10989 local tf=$DIR/$tdir/$tfile
10990 local nost=$((OSTCOUNT+1))
10991 local nostdevname=$(ostdevname $nost)
10994 test_mkdir -i 1 -c1 $DIR/$tdir || error "can't mkdir"
10997 add ost$nost $(mkfs_opts ost1 $nostdevname) --index=$nost \
10998 --reformat $nostdevname $(ostvdevname $nost)
10999 [[ -d "$nostdevname" ]] || stack_trap "do_facet mds1 rm -f $nostdevname"
11001 #define OBD_FAIL_OPS_FAIL_SEQ_ALLOC 0x2109
11002 do_facet mds1 $LCTL set_param fail_loc=0x80002109 fail_val=2
11003 echo "START OST$nost"
11004 stack_trap "stop ost$nost"
11005 start ost$nost $nostdevname $OST_MOUNT_OPTS &
11009 $LFS setstripe -c -1 $tf &
11013 log "STOP OST$nost"
11014 # probably mount hasn't completed yet, so stop races with it
11017 jobs -pr | grep -E "^$PID\$" && sleep 0.5 && continue
11022 do_facet mds1 $LCTL set_param fail_loc=0
11023 log "START OST$nost again"
11024 start ost$nost $nostdevname $OST_MOUNT_OPTS ||
11025 error "can't start ost$nost"
11027 $LFS setstripe -c -1 $tf-2 || error "can't touch $tf-2"
11028 $LFS getstripe -v $tf-2
11029 local stripes=$($LFS getstripe -c $tf-2)
11030 (( stripes == $nost )) || error "$tf-2 $stripes != $nost"
11032 run_test 152 "seq allocation error in OSP"
11035 reformat_and_config
11037 start_mds || error "MDS start failed"
11038 start_ost || error "OST start failed"
11040 local nid=$($LCTL list_nids | grep ${NETTYPE} | head -n1)
11041 local net=${nid#*@}
11042 local MGS_NID=$(do_facet mgs $LCTL list_nids | head -1)
11043 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
11044 local FAKE_PNID="192.168.252.112@${net}"
11045 local FAKE_NIDS="${FAKE_PNID},${FAKE_PNID}2"
11046 local FAKE_FAILOVER="10.252.252.113@${net},10.252.252.113@${net}2"
11047 local NIDS_AND_FAILOVER="$FAKE_NIDS:$FAKE_FAILOVER:$OST1_NID:$MGS_NID"
11052 mount -t lustre $NIDS_AND_FAILOVER:/lustre $MOUNT &
11054 while (( period < 30 )); do
11055 [[ -n "$(ps -p $pid -o pid=)" ]] || break
11056 echo "waiting for mount ..."
11058 period=$((period + 5))
11060 $LCTL get_param mgc.MGC${FAKE_PNID}.import | grep "uptodate:"
11061 check_mount || error "check_mount failed"
11063 cleanup || error "cleanup failed with rc $?"
11065 run_test 153a "bypass invalid NIDs quickly"
11068 # (This was sanity/802a)
11070 saved_MGS_MOUNT_OPTS=$MGS_MOUNT_OPTS
11071 saved_MDS_MOUNT_OPTS=$MDS_MOUNT_OPTS
11072 saved_OST_MOUNT_OPTS=$OST_MOUNT_OPTS
11073 saved_MOUNT_OPTS=$MOUNT_OPTS
11077 MGS_MOUNT_OPTS=$saved_MGS_MOUNT_OPTS
11078 MDS_MOUNT_OPTS=$saved_MDS_MOUNT_OPTS
11079 OST_MOUNT_OPTS=$saved_OST_MOUNT_OPTS
11080 MOUNT_OPTS=$saved_MOUNT_OPTS
11085 [[ $mds1_FSTYPE = zfs ]] || skip "ZFS specific test"
11086 [[ $MDS1_VERSION -lt $(version_code 2.9.55) ]] ||
11087 [[ $OST1_VERSION -lt $(version_code 2.9.55) ]] &&
11088 skip "Need server version at least 2.9.55"
11090 [[ $ENABLE_QUOTA ]] && skip "Quota enabled for read-only test"
11092 # Reset before starting
11096 mkdir $DIR/$tdir || error "(1) fail to mkdir"
11098 cp $LUSTRE/tests/test-framework.sh $DIR/$tdir/ ||
11099 error "(2) Fail to copy"
11101 stack_trap cleanup_802a EXIT
11103 # sync by force before remount as readonly
11104 sync; sync_all_data; sleep 3; sync_all_data
11108 MGS_MOUNT_OPTS=$(csa_add "$MGS_MOUNT_OPTS" -o rdonly_dev)
11109 MDS_MOUNT_OPTS=$(csa_add "$MDS_MOUNT_OPTS" -o rdonly_dev)
11110 OST_MOUNT_OPTS=$(csa_add "$OST_MOUNT_OPTS" -o rdonly_dev)
11112 echo "Mount the server as read only"
11113 setupall server_only || error "(3) Fail to start servers"
11115 echo "Mount client without ro should fail"
11116 mount_client $MOUNT &&
11117 error "(4) Mount client without 'ro' should fail"
11119 echo "Mount client with ro should succeed"
11120 MOUNT_OPTS=$(csa_add "$MOUNT_OPTS" -o ro)
11121 mount_client $MOUNT ||
11122 error "(5) Mount client with 'ro' should succeed"
11124 echo "Modify should be refused"
11125 touch $DIR/$tdir/guard && error "(6) Touch should fail under ro mode"
11127 echo "Read should be allowed"
11128 diff $LUSTRE/tests/test-framework.sh $DIR/$tdir/test-framework.sh ||
11129 error "(7) Read should succeed under ro mode"
11131 run_test 802a "simulate readonly device"
11133 if ! combined_mgs_mds ; then
11139 # restore the values of MDSSIZE and OSTSIZE
11140 MDSSIZE=$STORED_MDSSIZE
11141 OSTSIZE=$STORED_OSTSIZE
11144 complete_test $SECONDS
11145 check_and_cleanup_lustre