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 with filter_fid upgrade"
2946 ff_convert=yes t32_test $tarball || rc=$?
2950 run_test 32d "convert filter_fid upgrade test"
2953 [[ "$MDS1_VERSION" -ge $(version_code 2.10.56) ]] ||
2954 skip "Need MDS version at least 2.10.56"
2961 for tarball in $tarballs; do
2962 [[ "$tarball" =~ "2_9" ]] ||
2963 { echo "skip $(basename $tarball)"; continue; }
2965 banner "testing $tarball upgrade with DoM"
2966 dom_upgrade=yes t32_test $tarball writeconf || let "rc += $?"
2970 run_test 32e "dom upgrade test"
2973 [[ $MDS1_VERSION -ge $(version_code 2.10.56) ]] ||
2974 skip "Need MDS version at least 2.10.56"
2981 for tarball in $tarballs; do
2982 [[ "$tarball" =~ "2_10" ]] ||
2983 { echo "skip $(basename $tarball)"; continue; }
2984 pfl_upgrade=yes project_quota_upgrade=yes \
2985 t32_test $tarball writeconf || let "rc += $?"
2989 run_test 32f "pfl upgrade test"
2992 [[ $MDS1_VERSION -ge $(version_code 2.10.56) ]] ||
2993 skip "Need MDS version at least 2.10.56"
3000 for tarball in $tarballs; do
3001 [[ $tarball =~ "2_12" ]] ||
3002 { echo "skip $(basename $tarball)"; continue; }
3003 flr_upgrade=yes dom_new_upgrade=yes \
3004 t32_test $tarball writeconf || let "rc += $?"
3008 run_test 32g "flr/dom upgrade test"
3010 test_33a() { # bug 12333, was test_33
3011 local FSNAME2=test-$testnum
3013 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3016 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
3018 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
3019 local dev=${SINGLEMDS}_dev
3020 local MDSDEV=${!dev}
3022 is_blkdev $SINGLEMDS $MDSDEV &&
3023 skip_env "mixed loopback and real device not working"
3026 local fs2mdsdev=$(mdsdevname 1_2)
3027 local fs2ostdev=$(ostdevname 1_2)
3028 local fs2mdsvdev=$(mdsvdevname 1_2)
3029 local fs2ostvdev=$(ostvdevname 1_2)
3031 if [ "$mds1_FSTYPE" == ldiskfs ]; then
3032 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
3035 if combined_mgs_mds; then
3036 local mgs_flag="--mgs"
3039 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --fsname=${FSNAME2} \
3040 --reformat $mgs_flag $mkfsoptions $fs2mdsdev $fs2mdsvdev ||
3042 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
3043 --fsname=${FSNAME2} --index=0x1fff --reformat $fs2ostdev \
3044 $fs2ostvdev || exit 10
3046 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
3047 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
3049 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3050 do_facet mgs "$PERM_CMD timeout=200" ||
3051 error "$PERM_CMD timeout=200 failed"
3053 do_facet mgs "$PERM_CMD $FSNAME2.sys.timeout=200" ||
3054 error "$PERM_CMD $FSNAME2.sys.timeout=200 failed"
3056 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
3057 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
3060 cp /etc/hosts $MOUNT2/ || error "copy /etc/hosts $MOUNT2/ failed"
3061 $LFS getstripe $MOUNT2/hosts ||
3062 error "$LFS getstripe $MOUNT2/hosts failed"
3066 if (( "$MGS_VERSION" >= $(version_code 2.15.51) &&
3067 "$MDS1_VERSION" >= $(version_code 2.15.51) )); then
3068 # test lctl del_ost on large index
3069 do_facet mgs "$LCTL del_ost -t ${FSNAME2}-OST1fff" ||
3070 error "del_ost failed with $?"
3071 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 ||
3072 error "$MOUNT_CMD failed"
3074 $LFS df | grep -q OST1fff &&
3075 error "del_ost did not remove OST1fff!"
3081 cleanup_nocli || error "cleanup_nocli failed with $?"
3083 run_test 33a "Mount ost with a large index number"
3085 test_33b() { # was test_34
3088 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
3089 # Drop lock cancelation reply during umount
3090 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
3091 do_facet client $LCTL set_param fail_loc=0x80000304
3092 #lctl set_param debug=-1
3093 umount_client $MOUNT
3094 cleanup || error "cleanup failed with $?"
3096 run_test 33b "Drop cancel during umount"
3099 (( MDS1_VERSION >= $(version_code 2.15.57) )) ||
3100 skip "Need MDS version at least 2.15.57"
3101 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3102 local tstid=${TSTID:-"$(id -u $TSTUSR)"}
3104 local qpool="qpool1"
3106 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
3108 local fs2mdsdev=$(mdsdevname 1_2)
3109 local fs2ostdev=$(ostdevname 1_2)
3110 local fs2mdsvdev=$(mdsvdevname 1_2)
3111 local fs2ostvdev=$(ostvdevname 1_2)
3113 if [ "$mds1_FSTYPE" == ldiskfs ]; then
3114 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
3117 if combined_mgs_mds; then
3118 local mgs_flag="--mgs"
3122 stack_trap unload_modules_conf
3124 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --fsname=${FSNAME} \
3125 --reformat $mgs_flag $mkfsoptions $fs2mdsdev $fs2mdsvdev ||
3127 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
3128 --fsname=${FSNAME} --index=0x7c6 --reformat $fs2ostdev \
3129 $fs2ostvdev || exit 10
3132 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
3133 stack_trap "stop fs2mds -f"
3134 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
3135 stack_trap "stop fs2ost -f"
3137 mount_client $MOUNT || error "client start failed"
3138 stack_trap "umount_client $MOUNT"
3139 mkdir_on_mdt0 $DIR/$tdir || error "cannot create $DIR/$tdir"
3140 chmod 0777 $DIR/$tdir || error "chown failed"
3141 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3142 do_facet mgs $PERM_CMD \
3143 osd-*.$FSNAME-OST*.quota_slave.enable=$QUOTA_TYPE
3145 do_facet mgs $PERM_CMD $FSNAME.quota.ost=$QUOTA_TYPE ||
3146 error "set ost quota type failed"
3149 local old_MDSCOUNT=$MDSCOUNT
3151 stack_trap "MDSCOUNT=$old_MDSCOUNT"
3153 pool_add $qpool || error "pool_add failed"
3154 pool_add_targets $qpool 0x7c6
3156 $LFS setquota -u $tstid -B20M -b 0 $MOUNT
3157 $LFS setquota -g $tstid -B20M -b 0 $MOUNT
3158 $LFS setquota -u $tstid -B20M -b 0 --pool $qpool $MOUNT
3159 $LFS setquota -g $tstid -B20M -b 0 --pool $qpool $MOUNT
3161 for i in {1..10}; do
3162 runas -u $tstid -g $tstid dd if=/dev/zero of=$DIR/$tdir/f1 \
3163 bs=1M count=30 oflag=direct
3169 #umount_client $MOUNT || error "client start failed"
3172 #cleanup_nocli || error "cleanup_nocli failed with $?"
3174 run_test 33c "Mount ost with a large index number"
3178 do_facet client "bash runmultiop_bg_pause $DIR/file O_c"
3179 manual_umount_client
3181 do_facet client killall -USR1 multiop
3182 if [ $rc -eq 0 ]; then
3183 error "umount not fail!"
3186 cleanup || error "cleanup failed with rc $?"
3188 run_test 34a "umount with opened file should be fail"
3192 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3193 stop_mds || error "Unable to stop MDS"
3195 manual_umount_client --force || error "mtab after failed umount with $?"
3197 cleanup || error "cleanup failed with $?"
3199 run_test 34b "force umount with failed mds should be normal"
3203 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3204 stop_ost || error "Unable to stop OST1"
3206 manual_umount_client --force || error "mtab after failed umount with $?"
3208 cleanup || error "cleanup failed with $?"
3210 run_test 34c "force umount with failed ost should be normal"
3212 test_35a() { # bug 12459
3215 DBG_SAVE="`$LCTL get_param -n debug`"
3216 $LCTL set_param debug="ha"
3218 log "Set up a fake failnode for the MDS"
3220 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
3221 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
3223 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3224 do_facet mgs "$PERM_CMD \
3225 mdc.*${device}*.import=connection=$(h2nettype $FAKENID)" ||
3226 error "Setting mdc.*${device}*.import=connection=\
3227 $(h2nettype $FAKENID) failed."
3229 do_facet mgs "$PERM_CMD \
3230 ${device}.failover.node=$(h2nettype $FAKENID)" ||
3231 error "Setting ${device}.failover.node=\
3232 $(h2nettype $FAKENID) failed."
3234 log "Wait for RECONNECT_INTERVAL seconds (10s)"
3237 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
3240 log "Stopping the MDT: $device"
3241 stop_mdt 1 || error "MDT0 stop fail"
3243 df $MOUNT > /dev/null 2>&1 &
3245 log "Restarting the MDT: $device"
3246 start_mdt 1 || error "MDT0 start fail"
3247 log "Wait for df ($DFPID) ... "
3250 $LCTL set_param debug="$DBG_SAVE"
3252 # retrieve from the log the first server that the client tried to
3253 # contact after the connection loss
3254 $LCTL dk $TMP/lustre-log-$TESTNAME.log
3255 NEXTCONN=`awk "/${MSG}/ {start = 1;}
3256 /import_select_connection.*$device-mdc.* using connection/ {
3258 if (\\\$NF ~ /$FAKENID/)
3264 }" $TMP/lustre-log-$TESTNAME.log`
3265 [ "$NEXTCONN" != "0" ] &&
3266 error "Tried to connect to ${NEXTCONN} not last active server"
3267 cleanup || error "cleanup failed with $?"
3268 # remove nid settings
3269 writeconf_or_reformat
3271 run_test 35a "Reconnect to the last active server first"
3273 test_35b() { # bug 18674
3274 remote_mds || skip "local MDS"
3278 $LCTL set_param debug="ha"
3280 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
3283 log "Set up a fake failnode for the MDS"
3285 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
3286 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
3288 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3289 do_facet mgs "$PERM_CMD \
3290 mdc.*${device}*.import=connection=$(h2nettype $FAKENID)" ||
3291 error "Set mdc.*${device}*.import=connection=\
3292 $(h2nettype $FAKENID) failed"
3294 do_facet mgs "$PERM_CMD \
3295 ${device}.failover.node=$(h2nettype $FAKENID)" ||
3296 error "Set ${device}.failover.node=\
3297 $(h2nettype $FAKENID) failed"
3300 local at_max_saved=0
3301 # adaptive timeouts may prevent seeing the issue
3302 if at_is_enabled; then
3303 at_max_saved=$(at_max_get mds)
3304 at_max_set 0 mds client
3307 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
3309 log "Injecting EBUSY on MDS"
3310 # Setting OBD_FAIL_MDS_RESEND=0x136
3311 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" ||
3312 error "unable to set param fail_loc=0x80000136"
3314 $LCTL set_param mdc.${FSNAME}*.stats=clear
3316 log "Creating a test file and stat it"
3317 touch $MOUNT/$tdir/$tfile || error "touch $MOUNT/$tdir/$tfile failed"
3318 stat $MOUNT/$tdir/$tfile
3320 log "Stop injecting EBUSY on MDS"
3321 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" ||
3322 error "unable to set param fail_loc=0"
3323 rm -f $MOUNT/$tdir/$tfile || error "remove $MOUNT/$tdir/$tfile failed"
3326 # restore adaptive timeout
3327 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
3329 $LCTL dk $TMP/lustre-log-$TESTNAME.log
3331 CONNCNT=$($LCTL get_param mdc.${FSNAME}*.stats |
3332 awk '/mds_connect/{print $2}')
3334 # retrieve from the log if the client has ever tried to
3335 # contact the fake server after the loss of connection
3336 FAILCONN=`awk "BEGIN {ret = 0;}
3337 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
3339 if (\\\$NF ~ /$FAKENID/) {
3344 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
3346 [ "$FAILCONN" == "0" ] &&
3347 error "The client reconnection has not been triggered"
3348 [ "$FAILCONN" == "2" ] &&
3349 error "Primary server busy, client reconnect to failover failed"
3352 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
3353 # Reconnects are supposed to be rate limited to one every 5s
3354 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] &&
3355 error "Too many reconnects $CONNCNT"
3357 cleanup || error "cleanup failed with $?"
3358 # remove nid settings
3359 writeconf_or_reformat
3361 run_test 35b "Continue reconnection retries, if the active server is busy"
3364 [ $OSTCOUNT -lt 2 ] && skip_env "needs >= 2 OSTs"
3366 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] ||
3370 local FSNAME2=test1234
3371 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3373 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
3375 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
3376 is_blkdev $SINGLEMDS $MDSDEV &&
3377 skip_env "mixed loopback and real device not working"
3380 local fs2mdsdev=$(mdsdevname 1_2)
3381 local fs2ostdev=$(ostdevname 1_2)
3382 local fs3ostdev=$(ostdevname 2_2)
3383 local fs2mdsvdev=$(mdsvdevname 1_2)
3384 local fs2ostvdev=$(ostvdevname 1_2)
3385 local fs3ostvdev=$(ostvdevname 2_2)
3388 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
3389 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
3390 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
3391 # different one than the default value here.
3392 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
3393 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
3394 add fs3ost $(mkfs_opts ost2 ${fs3ostdev}) --mgsnode=$MGSNID \
3395 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
3397 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
3398 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
3399 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
3400 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
3401 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
3403 sleep 5 # until 11778 fixed
3405 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || error "dd failed"
3407 BKTOTAL=$($LCTL get_param -n obdfilter.*.kbytestotal |
3408 awk 'BEGIN{total=0}; {total+=$1}; END{print total}')
3409 BKFREE=$($LCTL get_param -n obdfilter.*.kbytesfree |
3410 awk 'BEGIN{free=0}; {free+=$1}; END{print free}')
3411 BKAVAIL=$($LCTL get_param -n obdfilter.*.kbytesavail |
3412 awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}')
3413 STRING=$(df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}')
3414 DFTOTAL=$(echo $STRING | cut -d, -f1)
3415 DFUSED=$(echo $STRING | cut -d, -f2)
3416 DFAVAIL=$(echo $STRING | cut -d, -f3)
3417 DFFREE=$(($DFTOTAL - $DFUSED))
3419 ALLOWANCE=$((64 * $OSTCOUNT))
3421 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
3422 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
3423 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
3426 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
3427 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
3428 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
3431 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
3432 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
3433 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
3438 stop fs3ost -f || error "unable to stop OST3"
3439 stop fs2ost -f || error "unable to stop OST2"
3440 stop fs2mds -f || error "unable to stop second MDS"
3441 unload_modules_conf || error "unable unload modules"
3444 run_test 36 "df report consistency on OSTs with different block size"
3447 local mntpt=$(facet_mntpt $SINGLEMDS)
3448 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
3449 local mdsdev_sym="$TMP/sym_mdt.img"
3450 local opts=$MDS_MOUNT_OPTS
3453 if [ "$mds1_FSTYPE" != ldiskfs ]; then
3454 skip "ldiskfs only test"
3457 echo "MDS : $mdsdev"
3458 echo "SYMLINK : $mdsdev_sym"
3459 do_facet $SINGLEMDS rm -f $mdsdev_sym
3461 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
3463 echo "mount symlink device - $mdsdev_sym"
3465 if ! do_facet $SINGLEMDS test -b $mdsdev; then
3466 opts=$(csa_add "$opts" -o loop)
3470 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
3471 $mdsdev_sym $mntpt 2>&1)
3474 echo mount_op=$mount_op
3476 do_facet $SINGLEMDS "$UMOUNT $mntpt && rm -f $mdsdev_sym"
3478 if $(echo $mount_op | grep -q "unable to set tunable"); then
3479 error "set tunables failed for symlink device"
3482 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
3484 run_test 37 "verify set tunables works for symlink device"
3486 test_38() { # bug 14222
3487 local mntpt=$(facet_mntpt $SINGLEMDS)
3492 local SRC="/etc /bin"
3493 local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT)
3495 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
3496 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3497 tar cf - $FILES | tar xf - -C $DIR/$tdir ||
3498 error "copying $SRC to $DIR/$tdir"
3500 umount_client $MOUNT || error "umount_client $MOUNT failed"
3501 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3502 stop_mds || error "Unable to stop MDS"
3503 log "delete lov_objid file on MDS"
3505 mount_fstype $SINGLEMDS || error "mount MDS failed (1)"
3507 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid; rm $mntpt/lov_objid"
3509 unmount_fstype $SINGLEMDS || error "umount failed (1)"
3511 # check create in mds_lov_connect
3512 start_mds || error "unable to start MDS"
3513 mount_client $MOUNT || error "mount_client $MOUNT failed"
3515 [ $V ] && log "verifying $DIR/$tdir/$f"
3516 diff -q $f $DIR/$tdir/$f || ERROR=y
3518 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3519 if [ "$ERROR" = "y" ]; then
3520 # check it's updates in sync
3521 umount_client $MOUNT
3523 mount_fstype $SIGNLEMDS
3524 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3525 unmount_fstype $SINGLEMDS
3526 error "old and new files are different after connect" || true
3528 touch $DIR/$tdir/f2 || error "f2 file create failed"
3530 # check it's updates in sync
3531 umount_client $MOUNT || error "second umount_client $MOUNT failed"
3534 mount_fstype $SINGLEMDS || error "mount MDS failed (3)"
3536 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3537 do_facet $SINGLEMDS dd if=/dev/zero of=$mntpt/lov_objid.clear count=8
3539 unmount_fstype $SINGLEMDS || error "umount failed (3)"
3541 start_mds || error "unable to start MDS"
3542 mount_client $MOUNT || error "mount_client $MOUNT failed"
3544 [ $V ] && log "verifying $DIR/$tdir/$f"
3545 diff -q $f $DIR/$tdir/$f || ERROR=y
3547 touch $DIR/$tdir/f3 || error "f3 file create failed"
3548 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
3549 umount_client $MOUNT || error "third umount_client $MOUNT failed"
3551 mount_fstype $SINGLEMDS || error "mount MDS failed (4)"
3552 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
3553 unmount_fstype $SINGLEMDS || error "umount failed (4)"
3555 [[ "$ERROR" != "y" ]] ||
3556 error "old and new files are different after sync"
3558 log "files compared the same"
3559 cleanup || error "cleanup failed with $?"
3561 run_test 38 "MDS recreates missing lov_objid file from OST data"
3564 [[ -n "$(type -p perl)" ]] || skip_env "need perl for leak_finder.pl"
3568 cleanup || error "cleanup failed with $?"
3569 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
3570 error "memory leak detected" || true
3572 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
3574 test_40() { # bug 15759
3575 start_ost || error "Unable to start OST1"
3576 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
3577 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
3579 cleanup || error "cleanup failed with rc $?"
3581 run_test 40 "race during service thread startup"
3583 test_41a() { #bug 14134
3584 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3585 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
3586 skip "Loop devices does not work with nosvc option"
3589 combined_mgs_mds || skip "needs combined MGT and MDT device"
3591 start_mdt 1 -o nosvc -n
3592 if [ $MDSCOUNT -ge 2 ]; then
3593 for num in $(seq 2 $MDSCOUNT); do
3594 start_mdt $num || return
3597 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
3598 start_mdt 1 -o nomgs,force
3599 mount_client $MOUNT || error "mount_client $MOUNT failed"
3602 echo "blah blah" > $MOUNT/$tfile
3605 umount_client $MOUNT || error "umount_client $MOUNT failed"
3606 stop ost1 -f || error "unable to stop OST1"
3607 stop_mds || error "Unable to stop MDS"
3608 stop_mds || error "Unable to stop MDS on second try"
3610 run_test 41a "mount mds with --nosvc and --nomgs"
3613 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3614 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
3615 skip "Loop devices does not work with nosvc option"
3618 ! combined_mgs_mds && skip "needs combined mgs device"
3622 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
3624 start_mdt 1 -o nosvc -n
3625 if [ $MDSCOUNT -ge 2 ]; then
3626 for num in $(seq 2 $MDSCOUNT); do
3627 start_mdt $num || return
3630 start_ost || error "Unable to start OST1"
3631 start_mdt 1 -o nomgs,force
3632 mount_client $MOUNT || error "mount_client $MOUNT failed"
3635 echo "blah blah" > $MOUNT/$tfile
3636 cat $MOUNT/$tfile || error "cat $MOUNT/$tfile failed"
3638 umount_client $MOUNT -f || error "umount_client $MOUNT failed"
3639 stop_ost || error "Unable to stop OST1"
3640 stop_mds || error "Unable to stop MDS"
3641 stop_mds || error "Unable to stop MDS on second try"
3643 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
3646 local oss_list=$(comma_list $(osts_nodes))
3648 [[ "$MDS1_VERSION" -ge $(version_code 2.6.52) ]] ||
3649 [[ "$MDS1_VERSION" -ge $(version_code 2.5.26) &&
3650 "$MDS1_VERSION" -lt $(version_code 2.5.50) ]] ||
3651 [[ "$MDS1_VERSION" -ge $(version_code 2.5.4) &&
3652 "$MDS1_VERSION" -lt $(version_code 2.5.11) ]] ||
3653 skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"
3655 # ensure mds1 ost1 have been created even if running sub-test standalone
3658 cleanup || error "cleanup failed"
3660 # using directly mount command instead of start() function to avoid
3661 # any side effect of // with others/externals tools/features
3662 # ("zpool import", ...)
3664 # MDT concurrent start
3666 LOAD_MODULES_REMOTE=true load_modules
3667 do_facet $SINGLEMDS "lsmod | grep -q libcfs" ||
3668 error "MDT concurrent start: libcfs module not loaded"
3670 local mds1dev=$(mdsdevname 1)
3671 local mds1mnt=$(facet_mntpt mds1)
3672 local mds1opts=$MDS_MOUNT_OPTS
3674 if [ "$mds1_FSTYPE" == ldiskfs ] &&
3675 ! do_facet mds1 test -b $mds1dev; then
3676 mds1opts=$(csa_add "$mds1opts" -o loop)
3678 if [[ "$mds1_FSTYPE" == zfs ]]; then
3679 import_zpool mds1 || return ${PIPESTATUS[0]}
3682 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3683 do_facet mds1 "$LCTL set_param fail_loc=0x80000716"
3685 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts &
3688 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts
3693 do_facet mds1 "$LCTL set_param fail_loc=0x0"
3694 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3695 echo "1st MDT start succeed"
3696 echo "2nd MDT start failed with $rc2"
3697 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3698 echo "1st MDT start failed with $rc"
3699 echo "2nd MDT start succeed"
3702 error "unexpected concurrent MDT mounts result, rc=$rc rc2=$rc2"
3705 if [ $MDSCOUNT -ge 2 ]; then
3706 for num in $(seq 2 $MDSCOUNT); do
3707 start_mdt $num || return
3711 # OST concurrent start
3713 do_rpc_nodes $oss_list "lsmod | grep -q libcfs" ||
3714 error "OST concurrent start: libcfs module not loaded"
3716 local ost1dev=$(ostdevname 1)
3717 local ost1mnt=$(facet_mntpt ost1)
3718 local ost1opts=$OST_MOUNT_OPTS
3720 if [ "$ost1_FSTYPE" == ldiskfs ] &&
3721 ! do_facet ost1 test -b $ost1dev; then
3722 ost1opts=$(csa_add "$ost1opts" -o loop)
3724 if [[ "$ost1_FSTYPE" == zfs ]]; then
3725 import_zpool ost1 || return ${PIPESTATUS[0]}
3728 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3729 do_facet ost1 "$LCTL set_param fail_loc=0x80000716"
3731 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts &
3734 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts
3738 do_facet ost1 "$LCTL set_param fail_loc=0x0"
3739 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3740 echo "1st OST start succeed"
3741 echo "2nd OST start failed with $rc2"
3742 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3743 echo "1st OST start failed with $rc"
3744 echo "2nd OST start succeed"
3748 error "unexpected concurrent OST mounts result, rc=$rc rc2=$rc2"
3754 # verify everything ok
3759 error "MDT(s) start failed"
3767 error "OST(s) start failed"
3775 error "client start failed"
3782 error "client mount failed"
3786 run_test 41c "concurrent mounts of MDT/OST should all fail but one"
3788 test_42() { #bug 14693
3792 check_mount || error "client was not mounted"
3794 if [[ $PERM_CMD == *"set_param -P"* ]]; then
3795 PARAM="llite.$FSNAME-*.some_wrong_param"
3797 PARAM="$FSNAME.llite.some_wrong_param"
3800 do_facet mgs $PERM_CMD $PARAM=10
3801 umount_client $MOUNT ||
3802 error "unmounting client failed with invalid llite param"
3803 mount_client $MOUNT ||
3804 error "mounting client failed with invalid llite param"
3806 do_facet mgs $PERM_CMD $PARAM=20
3807 cleanup || error "stopping $FSNAME failed with invalid sys param"
3809 check_mount || error "client was not mounted with invalid sys param"
3810 cleanup || error "stopping $FSNAME failed with invalid sys param"
3812 run_test 42 "allow client/server mount/unmount with invalid config param"
3815 [[ "$MGS_VERSION" -ge $(version_code 2.5.58) ]] ||
3816 skip "Need MDS version at least 2.5.58"
3817 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
3820 USER1=$(getent passwd | grep :$ID1:$ID1: | cut -d: -f1)
3821 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1"
3824 chmod ugo+x $DIR || error "chmod 0 failed"
3825 set_persistent_param_and_check mds1 \
3826 "mdt.$FSNAME-MDT0000.root_squash" \
3827 "$FSNAME.mdt.root_squash" \
3829 wait_update $HOSTNAME \
3830 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3832 error "check llite root_squash failed!"
3833 set_persistent_param_and_check mds1 \
3834 "mdt.$FSNAME-MDT0000.nosquash_nids" \
3835 "$FSNAME.mdt.nosquash_nids" \
3837 wait_update $HOSTNAME \
3838 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3840 error "check llite nosquash_nids failed!"
3843 # create set of test files
3845 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
3846 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
3847 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
3849 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
3850 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
3852 mkdir_on_mdt0 $DIR/$tdir-rootdir || error "mkdir failed"
3853 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
3854 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
3856 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
3857 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
3858 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
3861 # check root_squash:
3862 # set root squash UID:GID to RUNAS_ID
3863 # root should be able to access only files owned by RUNAS_ID
3865 set_persistent_param_and_check mds1 \
3866 "mdt.$FSNAME-MDT0000.root_squash" \
3867 "$FSNAME.mdt.root_squash" \
3868 "$RUNAS_ID:$RUNAS_ID"
3869 wait_update $HOSTNAME \
3870 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3871 "$RUNAS_ID:$RUNAS_ID" ||
3872 error "check llite root_squash failed!"
3874 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
3875 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3876 error "$ST: root read permission is denied"
3877 echo "$ST: root read permission is granted - ok"
3880 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3881 error "$ST: root write permission is denied"
3882 echo "$ST: root write permission is granted - ok"
3884 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3885 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3886 error "$ST: root read permission is granted"
3887 echo "$ST: root read permission is denied - ok"
3890 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3891 error "$ST: root write permission is granted"
3892 echo "$ST: root write permission is denied - ok"
3894 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3895 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null &&
3896 error "$ST: root unlink permission is granted"
3897 echo "$ST: root unlink permission is denied - ok"
3899 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null &&
3900 error "$ST: root create permission is granted"
3901 echo "$ST: root create permission is denied - ok"
3904 # check root_squash is enforced independently
3905 # of client cache content
3907 # access file by USER1, keep access open
3908 # root should be denied access to user file
3910 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
3914 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
3915 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3916 { kill $pid; error "$ST: root read permission is granted"; }
3917 echo "$ST: root read permission is denied - ok"
3920 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3921 { kill $pid; error "$ST: root write permission is granted"; }
3922 echo "$ST: root write permission is denied - ok"
3928 # check nosquash_nids:
3929 # put client's NID into nosquash_nids list,
3930 # root should be able to access root file after that
3932 local NIDLIST=$($LCTL list_nids all | tr '\n' ' ')
3933 NIDLIST="2@gni $NIDLIST 192.168.0.[2,10]@tcp"
3934 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
3935 set_persistent_param_and_check mds1 \
3936 "mdt.$FSNAME-MDT0000.nosquash_nids" \
3937 "$FSNAME-MDTall.mdt.nosquash_nids" \
3939 wait_update $HOSTNAME \
3940 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3942 error "check llite nosquash_nids failed!"
3944 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3945 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3946 error "$ST: root read permission is denied"
3947 echo "$ST: root read permission is granted - ok"
3950 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3951 error "$ST: root write permission is denied"
3952 echo "$ST: root write permission is granted - ok"
3954 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3955 rm $DIR/$tdir-rootdir/tfile-1 ||
3956 error "$ST: root unlink permission is denied"
3957 echo "$ST: root unlink permission is granted - ok"
3958 touch $DIR/$tdir-rootdir/tfile-2 ||
3959 error "$ST: root create permission is denied"
3960 echo "$ST: root create permission is granted - ok"
3961 cleanup || error "cleanup failed with $?"
3963 run_test 43a "check root_squash and nosquash_nids"
3965 test_43b() { # LU-5690
3966 [[ "$MGS_VERSION" -ge $(version_code 2.7.62) ]] ||
3967 skip "Need MGS version 2.7.62+"
3969 if [[ -z "$fs2mds_DEV" ]]; then
3970 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
3971 skip_env "mixed loopback and real device not working"
3974 local fs2mdsdev=$(mdsdevname 1_2)
3975 local fs2mdsvdev=$(mdsvdevname 1_2)
3977 # temporarily use fs2mds as fs2mgs
3979 local fs2mgsdev=$fs2mdsdev
3980 local fs2mgsvdev=$fs2mdsvdev
3982 local fsname=test1234
3984 load_module llite/lustre
3985 local client_ip=$(host_nids_address $HOSTNAME $NETTYPE)
3986 local host=${client_ip//*./}
3987 local net=${client_ip/%$host/}
3988 local nosquash_nids=$(h2nettype $net[$host,$host,$host])
3990 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
3991 --param mdt.root_squash=$RUNAS_ID:$RUNAS_ID \
3992 --param mdt.nosquash_nids=$nosquash_nids \
3993 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
3994 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS || error "start fs2mgs failed"
3995 stop $fs2mgs -f || error "stop fs2mgs failed"
3996 cleanup || error "cleanup failed with $?"
3998 run_test 43b "parse nosquash_nids with commas in expr_list"
4002 check_mount || error "check_mount"
4003 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
4005 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
4006 for VAL in $UUIDS; do
4007 NID=$(echo $VAL | cut -d= -f1)
4008 CLUUID=$(echo $VAL | cut -d= -f2)
4009 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
4011 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
4012 cleanup || error "cleanup failed with $?"
4014 run_test 44 "mounted client proc entry exists"
4018 check_mount || error "check_mount"
4019 stop_mds || error "Unable to stop MDS"
4023 #define OBD_FAIL_PTLRPC_LONG_REPL_UNLINK 0x50f
4024 do_facet client "$LCTL set_param fail_loc=0x8000050f"
4027 manual_umount_client --force || error "manual_umount_client failed"
4028 do_facet client "$LCTL set_param fail_loc=0x0"
4029 start_mds || error "unable to start MDS"
4030 mount_client $MOUNT || error "mount_client $MOUNT failed"
4031 cleanup || error "cleanup failed with $?"
4033 run_test 45 "long unlink handling in ptlrpcd"
4040 umount_client $MOUNT2 || rc=$?
4041 umount_client $MOUNT || rc=$?
4042 while [ $count -gt 0 ]; do
4043 stop ost${count} -f || rc=$?
4047 cleanup_nocli || rc=$?
4048 #writeconf to remove all ost2 traces for subsequent tests
4049 writeconf_or_reformat
4054 echo "Testing with $OSTCOUNT OSTs"
4056 start_mds || error "unable to start MDS"
4057 #first client should see only one ost
4058 start_ost || error "Unable to start OST1"
4059 wait_osc_import_state mds ost FULL
4061 mount_client $MOUNT || error "mount_client $MOUNT failed"
4062 trap "cleanup_46a $OSTCOUNT" EXIT ERR
4065 for (( i=2; i<=$OSTCOUNT; i++ )); do
4066 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
4067 error "start_ost$i $(ostdevname $i) failed"
4070 # wait until osts in sync
4071 for (( i=2; i<=$OSTCOUNT; i++ )); do
4072 wait_osc_import_state mds ost$i FULL
4073 wait_osc_import_ready client ost$i
4076 #second client see all ost's
4078 mount_client $MOUNT2 || error "mount_client failed"
4079 $LFS setstripe -c -1 $MOUNT2 ||
4080 error "$LFS setstripe -c -1 $MOUNT2 failed"
4081 $LFS getstripe $MOUNT2 || error "$LFS getstripe $MOUNT2 failed"
4083 echo "ok" > $MOUNT2/widestripe
4084 $LFS getstripe $MOUNT2/widestripe ||
4085 error "$LFS getstripe $MOUNT2/widestripe failed"
4086 # fill acl buffer for avoid expand lsm to them
4087 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd |
4089 setfacl -m $acl $MOUNT2/widestripe
4093 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
4095 cleanup_46a $OSTCOUNT || error "cleanup_46a failed"
4097 run_test 46a "handle ost additional - wide striped file"
4102 check_mount || error "check_mount failed"
4103 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
4107 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
4108 if echo $ns | grep "MDT[[:digit:]]*"; then
4111 lrs=$(echo $ns | sed 's/.*lru_size=//')
4112 lru_size[count]=$lrs
4117 facet_failover $SINGLEMDS
4118 client_up || error "client_up failed"
4121 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
4122 if echo $ns | grep "MDT[[:digit:]]*"; then
4125 lrs=$(echo $ns | sed 's/.*lru_size=//')
4126 if ! test "$lrs" -eq "${lru_size[count]}"; then
4127 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
4128 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
4133 cleanup || error "cleanup failed with $?"
4135 run_test 47 "server restart does not make client loss lru_resize settings"
4140 # reformat after this test is needed - if the test fails,
4141 # we will have unkillable file at FS
4145 test_48() { # bz-17636 LU-7473
4149 check_mount || error "check_mount failed"
4151 $LFS setstripe -c -1 $MOUNT ||
4152 error "$LFS setstripe -c -1 $MOUNT failed"
4153 $LFS getstripe $MOUNT || error "$LFS getstripe $MOUNT failed"
4155 echo "ok" > $MOUNT/widestripe
4156 $LFS getstripe $MOUNT/widestripe ||
4157 error "$LFS getstripe $MOUNT/widestripe failed"
4159 # In the future, we may introduce more EAs, such as selinux, enlarged
4160 # LOV EA, and so on. These EA will use some EA space that is shared by
4161 # ACL entries. So here we only check some reasonable ACL entries count,
4162 # instead of the max number that is calculated from the max_ea_size.
4163 if [ "$MDS1_VERSION" -lt $(version_code 2.8.57) ]; then
4164 count=28 # hard coded of RPC protocol
4165 elif large_xattr_enabled; then
4166 count=4500 # max_num 8187 max_ea_size = 65452
4167 # not create too many (4500) to save test time
4169 count=450 # max_num 497 max_ea_size = 4012
4172 echo "It is expected to hold at least $count ACL entries"
4173 trap cleanup_48 EXIT ERR
4174 for ((i = 0; i < $count; i++)) do
4175 setfacl -m u:$((i + 100)):rw $MOUNT/widestripe ||
4176 error "Fail to setfacl for $MOUNT/widestripe at $i"
4179 cancel_lru_locks mdc
4180 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
4181 local r_count=$(getfacl $MOUNT/widestripe | grep "user:" | wc -l)
4183 count=$((count + 1)) # for the entry "user::rw-"
4185 [ $count -eq $r_count ] ||
4186 error "Expected ACL entries $count, but got $r_count"
4190 run_test 48 "too many acls on file"
4192 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
4193 test_49a() { # bug 17710
4194 local timeout_orig=$TIMEOUT
4195 local ldlm_timeout_orig=$LDLM_TIMEOUT
4196 local LOCAL_TIMEOUT=20
4198 LDLM_TIMEOUT=$LOCAL_TIMEOUT
4199 TIMEOUT=$LOCAL_TIMEOUT
4203 check_mount || error "client mount failed"
4205 echo "check ldlm_timout..."
4206 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
4207 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
4208 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
4210 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
4211 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
4214 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
4215 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
4218 umount_client $MOUNT || error "umount_client $MOUNT failed"
4219 stop_ost || error "problem stopping OSS"
4220 stop_mds || error "problem stopping MDS"
4222 LDLM_TIMEOUT=$ldlm_timeout_orig
4223 TIMEOUT=$timeout_orig
4225 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
4227 test_49b() { # bug 17710
4228 local timeout_orig=$TIMEOUT
4229 local ldlm_timeout_orig=$LDLM_TIMEOUT
4230 local LOCAL_TIMEOUT=20
4232 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
4233 TIMEOUT=$LOCAL_TIMEOUT
4237 check_mount || error "client mount failed"
4239 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
4240 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
4241 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
4243 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
4244 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
4247 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
4248 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
4251 cleanup || error "cleanup failed"
4253 LDLM_TIMEOUT=$ldlm_timeout_orig
4254 TIMEOUT=$timeout_orig
4256 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
4259 # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1
4261 # Test both statfs and lfs df and fail if either one fails
4262 multiop_bg_pause $1 f_
4265 killall -USR1 multiop
4266 [ $RC -ne 0 ] && log "lazystatfs multiop failed"
4267 wait $PID || { RC=$?; log "multiop return error "; }
4269 # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1
4274 if kill -s 0 $PID; then
4277 log "lazystatfs lfs df failed to complete in 5s"
4285 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4286 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4288 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
4290 cleanup || error "cleanup failed with rc $?"
4292 run_test 50a "lazystatfs all servers available"
4296 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4297 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4299 # Wait for client to detect down OST
4300 stop_ost || error "Unable to stop OST1"
4301 wait_osc_import_state client ost DISCONN
4303 log "OSCs should all be DISCONN"
4305 lazystatfs $MOUNT || error "lazystatfs should not return EIO"
4307 umount_client $MOUNT || error "Unable to unmount client"
4308 stop_mds || error "Unable to stop MDS"
4310 run_test 50b "lazystatfs all servers down"
4313 start_mds || error "Unable to start MDS"
4314 start_ost || error "Unable to start OST1"
4315 start_ost2 || error "Unable to start OST2"
4316 mount_client $MOUNT || error "Unable to mount client"
4317 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4318 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4320 # Wait for client to detect down OST
4321 stop_ost || error "Unable to stop OST1"
4322 wait_osc_import_state mds ost DISCONN
4323 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
4325 umount_client $MOUNT || error "Unable to unmount client"
4326 stop_ost2 || error "Unable to stop OST2"
4327 stop_mds || error "Unable to stop MDS"
4328 #writeconf to remove all ost2 traces for subsequent tests
4329 writeconf_or_reformat
4331 run_test 50c "lazystatfs one server down"
4334 start_mds || error "Unable to start MDS"
4335 start_ost || error "Unable to start OST1"
4336 start_ost2 || error "Unable to start OST2"
4337 mount_client $MOUNT || error "Unable to mount client"
4338 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
4339 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4341 # Issue the statfs during the window where the client still
4342 # belives the OST to be available but it is in fact down.
4343 # No failure just a statfs which hangs for a timeout interval.
4344 stop_ost || error "Unable to stop OST1"
4345 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
4347 umount_client $MOUNT || error "Unable to unmount client"
4348 stop_ost2 || error "Unable to stop OST2"
4349 stop_mds || error "Unable to stop MDS"
4350 #writeconf to remove all ost2 traces for subsequent tests
4351 writeconf_or_reformat
4353 run_test 50d "lazystatfs client/server conn race"
4360 start_mds || error "Unable to start MDS"
4361 #first client should see only one ost
4362 start_ost || error "Unable to start OST1"
4363 wait_osc_import_state mds ost FULL
4365 # Wait for client to detect down OST
4366 stop_ost || error "Unable to stop OST1"
4367 wait_osc_import_state mds ost DISCONN
4369 mount_client $MOUNT || error "Unable to mount client"
4370 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
4372 multiop_bg_pause $MOUNT _f
4376 if [ $RC1 -ne 0 ]; then
4377 log "multiop failed $RC1"
4380 sleep $(( $TIMEOUT+1 ))
4382 [ $? -ne 0 ] && error "process isn't sleep"
4383 start_ost || error "Unable to start OST1"
4384 wait $pid || error "statfs failed"
4387 umount_client $MOUNT || error "Unable to unmount client"
4388 stop_ost || error "Unable to stop OST1"
4389 stop_mds || error "Unable to stop MDS"
4391 run_test 50e "normal statfs all servers down"
4396 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
4398 start_mds || error "Unable to start mds"
4399 #first client should see only one ost
4400 start_ost || error "Unable to start OST1"
4401 wait_osc_import_state mds ost FULL
4403 start_ost2 || error "Unable to start OST2"
4404 wait_osc_import_state mds ost2 FULL
4406 # Wait for client to detect down OST
4407 stop_ost2 || error "Unable to stop OST2"
4409 wait_osc_import_state mds ost2 DISCONN
4410 mount_client $MOUNT || error "Unable to mount client"
4411 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
4413 multiop_bg_pause $MOUNT _f
4417 if [ $RC1 -ne 0 ]; then
4418 log "lazystatfs multiop failed $RC1"
4421 sleep $(( $TIMEOUT+1 ))
4423 [ $? -ne 0 ] && error "process isn't sleep"
4424 start_ost2 || error "Unable to start OST2"
4425 wait $pid || error "statfs failed"
4426 stop_ost2 || error "Unable to stop OST2"
4429 umount_client $MOUNT -f || error "Unable to unmount client"
4430 stop_ost || error "Unable to stop OST1"
4431 stop_mds || error "Unable to stop MDS"
4432 #writeconf to remove all ost2 traces for subsequent tests
4433 writeconf_or_reformat
4435 run_test 50f "normal statfs one server in down"
4438 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs"
4440 start_ost2 || error "Unable to start OST2"
4441 wait_osc_import_state mds ost2 FULL
4442 wait_osc_import_ready client ost2
4444 if [[ $PERM_CMD == *"set_param -P"* ]]; then
4445 local PARAM="osc.${FSNAME}-OST0001*.active"
4447 local PARAM="${FSNAME}-OST0001.osc.active"
4450 $LFS setstripe -c -1 $DIR/$tfile || error "$LFS setstripe failed"
4451 do_facet mgs $PERM_CMD $PARAM=0 || error "Unable to deactivate OST"
4453 umount_client $MOUNT || error "Unable to unmount client"
4454 mount_client $MOUNT || error "Unable to mount client"
4455 # This df should not cause a panic
4458 do_facet mgs $PERM_CMD $PARAM=1 || error "Unable to activate OST"
4459 rm -f $DIR/$tfile || error "unable to remove file $DIR/$tfile"
4460 umount_client $MOUNT || error "Unable to unmount client"
4461 stop_ost2 || error "Unable to stop OST2"
4462 stop_ost || error "Unable to stop OST1"
4463 stop_mds || error "Unable to stop MDS"
4464 #writeconf to remove all ost2 traces for subsequent tests
4465 writeconf_or_reformat
4467 run_test 50g "deactivated OST should not cause panic"
4471 # prepare MDT/OST, make OSC inactive for OST1
4472 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs"
4474 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
4475 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
4476 error "tunefs OST1 failed"
4477 start_mds || error "Unable to start MDT"
4478 start_ost || error "Unable to start OST1"
4479 start_ost2 || error "Unable to start OST2"
4480 mount_client $MOUNT || error "client start failed"
4482 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4484 # activatate OSC for OST1
4485 set_persistent_param_and_check client \
4486 "osc.${FSNAME}-OST0000-osc-[!M]*.active" \
4487 "${FSNAME}-OST0000.osc.active" 1
4489 mkdir $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
4490 $LFS setstripe -c -1 -i 0 $DIR/$tdir/2 ||
4491 error "$LFS setstripe $DIR/$tdir/2 failed"
4492 sleep 1 && echo "create a file after OST1 is activated"
4493 # doing some io, shouldn't crash
4494 dd if=/dev/zero of=$DIR/$tdir/2/$tfile-io bs=1M count=10
4496 # check OSC import is working
4497 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
4498 error "some OSC imports are still not connected"
4502 umount_client $MOUNT || error "Unable to umount client"
4503 stop_ost2 || error "Unable to stop OST2"
4504 cleanup_nocli || error "cleanup_nocli failed with $?"
4506 run_test 50h "LU-642: activate deactivated OST"
4509 # prepare MDT/OST, make OSC inactive for OST1
4510 [ "$MDSCOUNT" -lt "2" ] && skip "needs >= 2 MDTs"
4513 [ $(facet_fstype mds2) == zfs ] && import_zpool mds2
4514 do_facet mds2 "$TUNEFS --param mdc.active=0 $(mdsdevname 2)" ||
4515 error "tunefs MDT2 failed"
4516 start_mds || error "Unable to start MDT"
4517 start_ost || error "Unable to start OST1"
4518 start_ost2 || error "Unable to start OST2"
4519 mount_client $MOUNT || error "client start failed"
4521 mkdir_on_mdt0 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4523 if [[ $PERM_CMD == *"set_param -P"* ]]; then
4524 $PERM_CMD mdc.${FSNAME}-MDT0001-mdc-*.active=0 &&
4525 error "deactive MDC0 succeeds"
4527 $PERM_CMD ${FSNAME}-MDT0000.mdc.active=0 &&
4528 error "deactive MDC0 succeeds"
4531 # activate MDC for MDT2
4532 set_persistent_param_and_check client \
4533 "mdc.${FSNAME}-MDT0001-mdc-*.active" \
4534 "${FSNAME}-MDT0001.mdc.active" 1
4536 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
4537 if [ "$MDS1_VERSION" -ge $(version_code 2.7.60) ]
4539 wait_dne_interconnect
4541 $LFS mkdir -i1 $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
4543 createmany -o $DIR/$tdir/2/$tfile-%d 1 || error "create files failed"
4545 rm -rf $DIR/$tdir/2 || error "unlink dir failed"
4547 # deactivate MDC for MDT2
4548 set_persistent_param_and_check client \
4549 "mdc.${FSNAME}-MDT0001-mdc-*.active" \
4550 "${FSNAME}-MDT0001.mdc.active" 0
4552 wait_osp_active mds ${FSNAME}-MDT0001 1 0
4554 $LFS mkdir -i1 $DIR/$tdir/2 &&
4555 error "mkdir $DIR/$tdir/2 succeeds after deactive MDT"
4557 $LFS mkdir -i0 -c$MDSCOUNT $DIR/$tdir/striped_dir ||
4558 error "mkdir $DIR/$tdir/striped_dir fails after deactive MDT2"
4560 local stripe_count=$($LFS getdirstripe -c $DIR/$tdir/striped_dir)
4561 [ $stripe_count -eq $((MDSCOUNT - 1)) ] ||
4562 error "wrong $stripe_count != $((MDSCOUNT -1)) for striped_dir"
4565 umount_client $MOUNT || error "Unable to umount client"
4570 run_test 50i "activate deactivated MDT"
4573 local LOCAL_TIMEOUT=20
4577 check_mount || error "check_mount failed"
4579 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
4580 $LFS setstripe -c -1 $MOUNT/$tdir ||
4581 error "$LFS setstripe -c -1 $MOUNT/$tdir failed"
4582 #define OBD_FAIL_MDS_REINT_DELAY 0x142
4583 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x142"
4584 touch $MOUNT/$tdir/$tfile &
4587 start_ost2 || error "Unable to start OST1"
4589 stop_ost2 || error "Unable to stop OST1"
4590 umount_client $MOUNT -f || error "unmount $MOUNT failed"
4591 cleanup_nocli || error "stop server failed"
4592 #writeconf to remove all ost2 traces for subsequent tests
4593 writeconf_or_reformat
4595 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
4604 do_node $node mkdir -p $dest
4605 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
4607 do_node $node 'tar cf - '$*' | tar xf - -C '$dest';
4608 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
4609 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
4611 do_node $node 'getfattr -d -m "[a-z]*\\." '$*' > '$xattrs
4612 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
4622 local backup2=${TMP}/backup2
4624 do_node $node mkdir -p $backup2
4625 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
4627 do_node $node 'tar cf - '$*' | tar xf - -C '$backup2';
4628 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
4629 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
4631 do_node $node "diff -rq $backup $backup2"
4632 [ $? -eq 0 ] || { error "contents differ"; return 3; }
4634 local xattrs2=${TMP}/xattrs2
4635 do_node $node 'getfattr -d -m "[a-z]*\\." '$*' > '$xattrs2
4636 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
4638 do_node $node "diff $xattrs $xattrs2"
4639 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
4641 do_node $node "rm -rf $backup2 $xattrs2"
4642 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
4646 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4647 skip "ldiskfs only test"
4650 start_mds || error "Unable to start MDS"
4651 start_ost || error "Unable to start OST1"
4652 mount_client $MOUNT || error "Unable to mount client"
4655 local ost1mnt=$(facet_mntpt ost1)
4656 local ost1node=$(facet_active_host ost1)
4657 local ost1tmp=$TMP/conf52
4660 mkdir $DIR/$tdir || error "Unable to create $DIR/$tdir"
4661 touch $TMP/modified_first || error "Unable to create temporary file"
4662 local mtime=$(stat -c %Y $TMP/modified_first)
4663 do_node $ost1node "mkdir -p $ost1tmp &&
4664 touch -m -d @$mtime $ost1tmp/modified_first" ||
4665 error "Unable to create temporary file"
4668 $LFS setstripe -c -1 -S 1M $DIR/$tdir || error "$LFS setstripe failed"
4670 for (( i=0; i < nrfiles; i++ )); do
4671 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c ||
4672 error "multiop failed"
4677 # sync all the data and make sure no pending data on the client,
4678 # thus the SOM xattr would not be changed any more.
4679 cancel_lru_locks osc
4682 echo backup files to $TMP/$tdir
4683 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
4684 copy_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
4685 error "Unable to copy files"
4687 umount_client $MOUNT || error "Unable to umount client"
4688 stop_ost || error "Unable to stop ost1"
4690 echo mount ost1 as ldiskfs
4691 do_node $ost1node mkdir -p $ost1mnt || error "Unable to create $ost1mnt"
4692 if ! do_node $ost1node test -b $ost1_dev; then
4695 do_node $ost1node mount -t "$ost1_FSTYPE" $loop $ost1_dev \
4697 error "Unable to mount ost1 as ldiskfs"
4700 echo backup objects to $ost1tmp/objects
4701 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
4702 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
4703 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
4705 error "Unable to copy objects"
4707 # move objects to lost+found
4708 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
4709 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
4711 do_node $ost1node "umount $ost1mnt" ||
4712 error "Unable to umount ost1 as ldiskfs"
4714 start_ost || error "Unable to start OST1"
4715 mount_client $MOUNT || error "Unable to mount client"
4717 local REPAIRED=$(do_node $ost1node "$LCTL get_param \
4718 -n osd-ldiskfs.$FSNAME-OST0000.oi_scrub" |
4719 awk '/^lf_repa[ri]*ed/ { print $2 }')
4720 [ $REPAIRED -gt 0 ] ||
4721 error "Some entry under /lost+found should be repaired"
4724 diff_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
4725 error "Unable to diff files"
4727 rm -rf $TMP/$tdir $TMP/file_xattrs ||
4728 error "Unable to delete temporary files"
4729 do_node $ost1node "rm -rf $ost1tmp" ||
4730 error "Unable to delete temporary files"
4731 cleanup || error "cleanup failed with $?"
4733 run_test 52 "check recovering objects from lost+found"
4735 # Checks threads_min/max/started for some service
4737 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
4738 # parameter pattern prefix like 'ost.*.ost'.
4751 local msg="Insane $modname thread counts"
4752 local ncpts=$(check_cpt_number $facet)
4756 check_mount || return 41
4758 # We need to expand $parampat, but it may match multiple parameters, so
4759 # we'll pick the first one
4760 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
4761 error "Couldn't expand ${parampat}.threads_min parameter name"
4765 # Remove the .threads_min part
4766 paramp=${paramp%.threads_min}
4768 # Check for sanity in defaults
4769 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4771 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4773 tstarted=$(do_facet $facet "$LCTL get_param \
4774 -n ${paramp}.threads_started" || echo 0)
4775 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' ||
4777 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4779 nthrs=$(expr $tmax - $tmin)
4780 if [ $nthrs -lt $ncpts ]; then
4786 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
4787 skip_env "module parameter forced $facet thread count"
4789 # Check that we can change min/max
4790 do_facet $facet "$LCTL set_param \
4791 ${paramp}.threads_min=$((tmin + nthrs))"
4792 do_facet $facet "$LCTL set_param \
4793 ${paramp}.threads_max=$((tmax - nthrs))"
4794 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4796 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4798 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) &&
4799 $tmax2 == ($tmax - $nthrs)))' || return $?
4801 # Check that we can set min/max to the same value
4802 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4804 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$tmin"
4805 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4807 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4809 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
4811 # Check that we can't set max < min
4812 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$((tmin - 1))"
4813 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4815 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4817 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
4819 # We need to ensure that we get the module options desired; to do this
4820 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
4821 LOAD_MODULES_REMOTE=true
4824 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
4826 setmodopts -a $modname "$newvalue" oldvalue
4829 check_mount || return 41
4831 # Restore previous setting of MODOPTS_*
4832 setmodopts $modname "$oldvalue"
4834 (( $MDS1_VERSION > $(version_code 2.12.52.91) )) || {
4835 echo "skip interop for MDS < v2_12_52-91-g183cb1e3cdd2"
4839 # Check that $opts took
4840 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4842 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4844 tstarted=$(do_facet $facet \
4845 "$LCTL get_param -n ${paramp}.threads_started" || echo 0)
4846 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4852 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
4853 cleanup || error "cleanup failed with rc $?"
4855 run_test 53a "check OSS thread count params"
4859 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
4860 cleanup || error "cleanup failed with $?"
4862 run_test 53b "check MDS thread count params"
4865 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4866 skip "ldiskfs only test"
4869 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p ||
4870 error "llverdev failed with rc=$?"
4873 run_test 54a "test llverdev and partial verify of device"
4876 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4877 skip "ldiskfs only test"
4881 run_llverfs $MOUNT -p || error "llverfs failed with rc=$?"
4882 cleanup || error "cleanup failed with rc=$?"
4884 run_test 54b "test llverfs and partial verify of filesystem"
4888 local max_ost_index=$1
4890 echo -n $(((max_ost_index + 1) * 8))
4894 if [ "$mds1_FSTYPE" != ldiskfs ]; then
4895 skip "ldiskfs only test"
4898 local mdsdev=$(mdsdevname 1)
4899 local mdsvdev=$(mdsvdevname 1)
4903 if ! combined_mgs_mds; then
4904 stop_mgs || error "stopping MGS service failed"
4905 format_mgs || error "formatting MGT failed"
4907 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
4909 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
4910 --reformat $(ostdevname 1) $(ostvdevname 1)
4916 echo checking size of lov_objid for ost index $i
4917 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" |
4918 grep ^User | awk -F 'Size: ' '{print $2}')
4919 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
4920 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
4922 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
4929 run_test 55 "check lov_objid size"
4932 local mds_journal_size_orig=$MDSJOURNALSIZE
4938 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
4939 $(ostdevname 1) $(ostvdevname 1)
4940 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
4941 $(ostdevname 2) $(ostvdevname 2)
4944 start_ost || error "Unable to start first ost (idx 10000)"
4945 start_ost2 || error "Unable to start second ost (idx 1000)"
4946 mount_client $MOUNT || error "Unable to mount client"
4950 # test instantiating PFL components with sparse index LU-15513
4951 mkdir -p $MOUNT/$tdir
4952 $LFS setstripe -E 4M -c 1 -E 1G -c 4 -S4M -E eof -c -1 $MOUNT/$tdir
4953 dd if=/dev/zero of=$MOUNT/$tdir/$tfile bs=4K count=1 seek=10k ||
4954 error "dd to second component failed"
4956 if [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
4957 [[ "$MDS1_VERSION" -ge $(version_code 2.5.4) &&
4958 "$MDS1_VERSION" -lt $(version_code 2.5.11) ]]; then
4959 wait_osc_import_state mds ost1 FULL
4960 wait_osc_import_state mds ost2 FULL
4961 $LFS setstripe --stripe-count=-1 $DIR/$tfile ||
4962 error "Unable to setstripe $DIR/$tfile"
4963 n=$($LFS getstripe --stripe-count $DIR/$tfile)
4964 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
4969 MDSJOURNALSIZE=$mds_journal_size_orig
4972 run_test 56a "check big OST indexes and out-of-index-order start"
4977 umount_client $MOUNT -f || error "unmount client failed"
4986 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs"
4988 trap cleanup_56b EXIT RETURN ERR
4991 if ! combined_mgs_mds ; then
4996 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) --index=0 --reformat \
4997 $(mdsdevname 1) $(mdsvdevname 1)
4998 add mds2 $(mkfs_opts mds2 $(mdsdevname 2)) --index=1 --reformat \
4999 $(mdsdevname 2) $(mdsvdevname 2)
5000 add mds3 $(mkfs_opts mds3 $(mdsdevname 3)) --index=1000 --reformat \
5001 $(mdsdevname 3) $(mdsvdevname 3)
5005 start_mdt 1 || error "MDT 1 (idx 0) start failed"
5006 start_mdt 2 || error "MDT 2 (idx 1) start failed"
5007 start_mdt 3 || error "MDT 3 (idx 1000) start failed"
5008 start_ost || error "Unable to start first ost"
5009 start_ost2 || error "Unable to start second ost"
5011 do_nodes $(comma_list $(mdts_nodes)) \
5012 "$LCTL set_param mdt.*.enable_remote_dir=1 \
5013 mdt.*.enable_remote_dir_gid=-1"
5015 mount_client $MOUNT || error "Unable to mount client"
5017 $LFS mkdir -c3 $MOUNT/$tdir || error "failed to make testdir"
5019 echo "This is test file 1!" > $MOUNT/$tdir/$tfile.1 ||
5020 error "failed to make test file 1"
5021 echo "This is test file 2!" > $MOUNT/$tdir/$tfile.2 ||
5022 error "failed to make test file 2"
5023 echo "This is test file 1000!" > $MOUNT/$tdir/$tfile.1000 ||
5024 error "failed to make test file 1000"
5026 rm -rf $MOUNT/$tdir || error "failed to remove testdir"
5028 $LFS mkdir -i1000 $MOUNT/$tdir.1000 ||
5029 error "create remote dir at idx 1000 failed"
5032 echo "=== START lfs df OUTPUT ==="
5034 echo "==== END lfs df OUTPUT ===="
5036 mdtcnt=$(echo -e "$output" | grep $FSNAME-MDT | wc -l)
5037 ostcnt=$(echo -e "$output" | grep $FSNAME-OST | wc -l)
5039 echo "lfs df returned mdt count $mdtcnt and ost count $ostcnt"
5040 [ $mdtcnt -eq 3 ] || error "lfs df returned wrong mdt count"
5041 [ $ostcnt -eq 2 ] || error "lfs df returned wrong ost count"
5043 echo "This is test file 1!" > $MOUNT/$tdir.1000/$tfile.1 ||
5044 error "failed to make test file 1"
5045 echo "This is test file 2!" > $MOUNT/$tdir.1000/$tfile.2 ||
5046 error "failed to make test file 2"
5047 echo "This is test file 1000!" > $MOUNT/$tdir.1000/$tfile.1000 ||
5048 error "failed to make test file 1000"
5049 rm -rf $MOUNT/$tdir.1000 || error "failed to remove remote_dir"
5052 echo "=== START lfs mdts OUTPUT ==="
5054 echo "==== END lfs mdts OUTPUT ===="
5056 echo -e "$output" | grep -v "MDTS:" | awk '{print $1}' |
5057 sed 's/://g' > $TMP/mdts-actual.txt
5058 sort $TMP/mdts-actual.txt -o $TMP/mdts-actual.txt
5060 echo -e "0\n1\n1000" > $TMP/mdts-expected.txt
5062 diff $TMP/mdts-expected.txt $TMP/mdts-actual.txt
5065 rm $TMP/mdts-expected.txt $TMP/mdts-actual.txt
5067 [ $result -eq 0 ] || error "target_obd proc file is incorrect!"
5069 run_test 56b "test target_obd correctness with nonconsecutive MDTs"
5071 test_57a() { # bug 22656
5072 do_rpc_nodes $(facet_active_host ost1) load_modules_local
5073 local NID=$(do_facet ost1 "$LCTL get_param nis" |
5074 tail -1 | awk '{print $1}')
5075 writeconf_or_reformat
5076 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
5077 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" ||
5078 error "tunefs failed"
5080 start_ost && error "OST registration from failnode should fail"
5083 run_test 57a "initial registration from failnode should fail (should return errs)"
5086 do_rpc_nodes $(facet_active_host ost1) load_modules_local
5087 local NID=$(do_facet ost1 "$LCTL get_param nis" |
5088 tail -1 | awk '{print $1}')
5089 writeconf_or_reformat
5090 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
5091 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" ||
5092 error "tunefs failed"
5094 start_ost || error "OST registration from servicenode should not fail"
5097 run_test 57b "initial registration from servicenode should not fail"
5100 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
5103 test_58() { # bug 22658
5104 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
5106 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5107 createmany -o $DIR/$tdir/$tfile-%d 100
5108 unlinkmany $DIR/$tdir/$tfile-%d 100
5109 stop_mds || error "Unable to stop MDS"
5111 local MNTDIR=$(facet_mntpt $SINGLEMDS)
5112 local devname=$(mdsdevname ${SINGLEMDS//mds/})
5114 # remove all files from the OBJECTS dir
5115 mount_fstype $SINGLEMDS
5117 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
5119 unmount_fstype $SINGLEMDS
5120 # restart MDS with missing llog files
5121 start_mds || error "unable to start MDS"
5122 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0"
5125 run_test 58 "missing llog files must not prevent MDT from mounting"
5128 start_mgsmds >> /dev/null
5129 local C1=$(count_osts)
5130 if [ $C1 -eq 0 ]; then
5131 start_ost >> /dev/null
5135 echo "original ost count: $C1 (expect > 0)"
5136 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
5137 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
5138 local C2=$(count_osts)
5139 echo "after mdt writeconf count: $C2 (expect 0)"
5140 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
5141 echo "OST start without writeconf should fail:"
5142 start_ost >> /dev/null &&
5143 error "OST start without writeconf didn't fail"
5144 echo "OST start with writeconf should succeed:"
5145 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
5146 local C3=$(count_osts)
5147 echo "after ost writeconf count: $C3 (expect 1)"
5148 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
5149 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
5150 local C4=$(count_osts)
5151 echo "after ost2 writeconf count: $C4 (expect 2)"
5152 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
5153 stop_ost2 >> /dev/null
5154 cleanup_nocli >> /dev/null
5155 #writeconf to remove all ost2 traces for subsequent tests
5156 writeconf_or_reformat
5158 run_test 59 "writeconf mount option"
5160 test_60a() { # LU-471
5161 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5162 skip "ldiskfs only test"
5167 for num in $(seq $MDSCOUNT); do
5168 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
5169 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
5170 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
5174 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
5175 [ ${PIPESTATUS[0]} -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
5177 # MDT default has dirdata feature
5178 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
5179 # we disable uninit_bg feature
5180 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
5181 # we set stride extended options
5182 echo $dump | grep stride > /dev/null || error "stride is not set"
5186 run_test 60a "check mkfs.lustre --mkfsoptions -E -O options setting"
5189 [[ "$mds1_FSTYPE" == ldiskfs ]] || skip "ldiskfs only test"
5191 local features=$(do_facet $SINGLEMDS $DUMPE2FS $(mdsdevname 1) |
5193 [ ${PIPESTATUS[0]} -eq 0 ] || error "$DUMPE2FS $(mdsdevname 1) failed"
5196 # ea_inode feature should be enabled by default for MDTs
5197 [[ "$features" =~ "ea_inode" ]] || error "ea_inode is not set"
5198 # large_dir feature should be enabled by default for MDTs
5199 [[ "$features" =~ "large_dir" ]] || error "large_dir is not set"
5201 run_test 60b "check mkfs.lustre MDT default features"
5203 test_61a() { # LU-80
5204 local lxattr=$(large_xattr_enabled)
5206 (( "$MDS1_VERSION" >= $(version_code 2.1.53) )) ||
5207 skip "Need MDS version at least 2.1.53 for large_xattr"
5209 if [[ "$mds1_FSTYPE" == ldiskfs ]] && ! large_xattr_enabled; then
5212 for ((num=1; num <= $MDSCOUNT; num++)); do
5213 do_facet mds${num} $TUNE2FS -O ea_inode \
5214 $(mdsdevname $num) ||
5215 error "tune2fs on mds $num failed"
5219 setup || error "setting up the filesystem failed"
5220 client_up || error "starting client failed"
5222 local file=$DIR/$tfile
5223 touch $file || error "touch $file failed"
5225 local large_value="$(generate_string $(max_xattr_size))"
5226 local small_value="bar"
5228 local name="trusted.big"
5229 log "save large xattr of $(max_xattr_size) bytes on $name on $file"
5230 setfattr -n $name -v $large_value $file ||
5231 error "saving $name on $file failed"
5233 local new_value=$(get_xattr_value $name $file)
5234 [[ "$new_value" != "$large_value" ]] &&
5235 error "$name different after saving"
5237 log "shrink value of $name on $file"
5238 setfattr -n $name -v $small_value $file ||
5239 error "shrinking value of $name on $file failed"
5241 new_value=$(get_xattr_value $name $file)
5242 [[ "$new_value" != "$small_value" ]] &&
5243 error "$name different after shrinking"
5245 log "grow value of $name on $file"
5246 setfattr -n $name -v $large_value $file ||
5247 error "growing value of $name on $file failed"
5249 new_value=$(get_xattr_value $name $file)
5250 [[ "$new_value" != "$large_value" ]] &&
5251 error "$name different after growing"
5253 log "check value of $name on $file after remounting MDS"
5255 new_value=$(get_xattr_value $name $file)
5256 [[ "$new_value" != "$large_value" ]] &&
5257 error "$name different after remounting MDS"
5259 log "remove large xattr $name from $file"
5260 setfattr -x $name $file || error "removing $name from $file failed"
5262 if $lxattr && [ "$mds1_FSTYPE" == ldiskfs ]; then
5263 stopall || error "stopping for e2fsck run"
5264 for num in $(seq $MDSCOUNT); do
5265 run_e2fsck $(facet_active_host mds$num) \
5266 $(mdsdevname $num) "-y" ||
5267 error "e2fsck MDT$num failed"
5269 setup_noconfig || error "remounting the filesystem failed"
5272 # need to delete this file to avoid problems in other tests
5274 cleanup || error "stopping systems failed"
5276 run_test 61a "large xattr"
5278 test_61b() { # LU-80
5279 local lxattr=$(large_xattr_enabled)
5281 (( $MDS1_VERSION >= $(version_code 2.15.51) )) ||
5282 skip "Need MDS version at least 2.15.51 for large_xattr fix"
5284 [[ "$mds1_FSTYPE" == "ldiskfs" ]] || skip "ldiskfs specific bug"
5286 if ! large_xattr_enabled; then
5289 for (( num=1; num <= $MDSCOUNT; num++ )); do
5290 do_facet mds${num} $TUNE2FS -O ea_inode \
5291 $(mdsdevname $num) ||
5292 error "tune2fs on mds $num failed"
5296 setup || error "setting up the filesystem failed"
5297 client_up || error "starting client failed"
5299 local _file=$MOUNT/panda
5300 local large_value="$(generate_string $(max_xattr_size))"
5301 local name="trusted.big"
5303 touch ${_file} || error "touch ${_file} failed"
5304 setfattr -n $name -v $large_value ${_file} ||
5305 error "saving $name on $file failed"
5307 MDT_DEV="${FSNAME}-MDT0000"
5308 MDT_DEVNAME=$(mdsdevname ${SINGLEMDS//mds/})
5310 stopall || error "stopping for e2fsck run"
5313 ino=$(do_facet $SINGLEMDS "$DEBUGFS -R 'stat /ROOT/panda' \
5314 ${MDT_DEVNAME} | grep trusted.big")
5315 ino=$(echo "${ino}" | awk '{print $2;}')
5316 echo "large ea "${ino}
5318 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"ln $ino /lost+found\\\" \
5321 setup_noconfig || error "remounting the filesystem failed"
5323 do_facet $SINGLEMDS $LCTL lfsck_start -M ${MDT_DEV} -t namespace || {
5324 error "can't start lfsck namespace"
5328 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
5329 mdd.${MDT_DEV}.lfsck_namespace |
5330 awk '/^status/ { print \\\$2 }'" "completed" 32 || {
5331 error "(2) unexpected status"
5334 stopall || error "stopping for e2fsck run"
5335 for num in $(seq $MDSCOUNT); do
5336 run_e2fsck $(facet_active_host mds$num) \
5337 $(mdsdevname $num) "-y" ||
5338 error "e2fsck MDT$num failed"
5340 setup_noconfig || error "remounting the filesystem failed"
5342 # need to delete this file to avoid problems in other tests
5344 cleanup || error "stopping systems failed"
5346 run_test 61b "large xattr"
5349 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5350 skip "ldiskfs only test"
5352 [[ "$MDS1_VERSION" -ge $(version_code 2.2.51) ]] ||
5353 skip "Need MDS version at least 2.2.51"
5356 local mdsdev=$(mdsdevname 1)
5357 local ostdev=$(ostdevname 1)
5359 echo "disable journal for mds"
5360 do_facet mds1 $TUNE2FS -O ^has_journal $mdsdev || error "tune2fs failed"
5361 start_mds && error "MDT start should fail"
5362 echo "disable journal for ost"
5363 do_facet ost1 $TUNE2FS -O ^has_journal $ostdev || error "tune2fs failed"
5364 start_ost && error "OST start should fail"
5365 cleanup || error "cleanup failed with rc $?"
5368 run_test 62 "start with disabled journal"
5371 if [ "$mds1_FSTYPE" != ldiskfs ]; then
5372 skip "ldiskfs only test"
5375 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ldiskfs
5376 local inode_slab=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
5377 awk '/ldiskfs_inode_cache/ { print $5 / $6 }')
5378 if [ -z "$inode_slab" ]; then
5379 skip "ldiskfs module has not been loaded"
5382 if grep -q "CONFIG_DEBUG_LOCK_ALLOC=y" /boot/config-$(uname -r); then
5383 skip "test is not compatible with CONFIG_DEBUG_LOCK_ALLOC=y"
5386 echo "$inode_slab ldiskfs inodes per page"
5387 [ "${inode_slab%.*}" -ge "3" ] && return 0
5389 # If kmalloc-128 is also 1 per page - this is a debug kernel
5390 # and so this is not an error.
5391 local kmalloc128=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
5392 awk '/^(dma-kmalloc|size)-128 / { print $5 / $6 }')
5393 # 32 128-byte chunks in 4k
5394 [ "${kmalloc128%.*}" -lt "32" ] ||
5395 error "ldiskfs inode too big, only $inode_slab objs/page, " \
5396 "kmalloc128 = $kmalloc128 objs/page"
5398 run_test 63 "Verify each page can at least hold 3 ldiskfs inodes"
5401 start_mds || error "unable to start MDS"
5402 start_ost || error "Unable to start OST1"
5403 start_ost2 || error "Unable to start second ost"
5404 mount_client $MOUNT || error "Unable to mount client"
5405 stop_ost2 || error "Unable to stop second ost"
5408 umount_client $MOUNT -f || error "unmount $MOUNT failed"
5409 cleanup_nocli || error "cleanup_nocli failed with $?"
5410 #writeconf to remove all ost2 traces for subsequent tests
5411 writeconf_or_reformat
5413 run_test 64 "check lfs df --lazy "
5415 test_65() { # LU-2237
5416 # Currently, the test is only valid for ldiskfs backend
5417 [ "$mds1_FSTYPE" != ldiskfs ] &&
5418 skip "ldiskfs only test"
5420 local devname=$(mdsdevname ${SINGLEMDS//mds/})
5421 local brpt=$(facet_mntpt brpt)
5424 if ! do_facet $SINGLEMDS "test -b $devname"; then
5428 stop_mds || error "Unable to stop MDS"
5429 local obj=$(do_facet $SINGLEMDS \
5430 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
5432 if [ -z "$obj" ]; then
5433 # The MDT may be just re-formatted, mount the MDT for the
5434 # first time to guarantee the "last_rcvd" file is there.
5435 start_mds || error "fail to mount the MDS for the first time"
5436 stop_mds || error "Unable to stop MDS"
5439 # remove the "last_rcvd" file
5440 do_facet $SINGLEMDS "mkdir -p $brpt"
5441 do_facet $SINGLEMDS \
5442 "mount -t $mds1_FSTYPE $opts $devname $brpt"
5443 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
5444 do_facet $SINGLEMDS "$UMOUNT $brpt"
5446 # restart MDS, the "last_rcvd" file should be recreated.
5447 start_mds || error "fail to restart the MDS"
5448 stop_mds || error "Unable to stop MDS"
5449 obj=$(do_facet $SINGLEMDS \
5450 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
5451 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
5453 run_test 65 "re-create the lost last_rcvd file when server mount"
5456 [[ "$MGS_VERSION" -ge $(version_code 2.3.59) ]] ||
5457 skip "Need MGS version at least 2.3.59"
5459 check_versions || skip "do not replace_nids with mismatched versions"
5462 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
5463 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
5465 # add EXCLUDE records to config log, they are not to be
5466 # removed by lctl replace_nids
5467 set_conf_param_and_check mds \
5468 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
5469 "$FSNAME-OST0000.osc.active" \
5472 echo "replace_nids should fail if MDS, OSTs and clients are UP"
5473 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5474 error "replace_nids fail"
5476 umount_client $MOUNT || error "unmounting client failed"
5477 echo "replace_nids should fail if MDS and OSTs are UP"
5478 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5479 error "replace_nids fail"
5481 stop_ost || error "Unable to stop OST1"
5482 echo "replace_nids should fail if MDS is UP"
5483 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
5484 error "replace_nids fail"
5486 stop_mds || error "stopping mds failed"
5488 if combined_mgs_mds; then
5489 start_mdt 1 "-o nosvc" ||
5490 error "starting mds with nosvc option failed"
5493 echo "command should accept two parameters"
5494 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
5495 error "command should accept two params"
5497 echo "correct device name should be passed"
5498 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
5499 error "wrong devname"
5501 echo "wrong nids list should not destroy the system"
5502 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
5504 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "asdfasdf, asdfadf" &&
5507 echo "replace OST nid"
5508 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
5509 error "replace nids failed"
5511 echo "command should accept two parameters"
5512 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
5513 error "command should accept two params"
5515 echo "wrong nids list should not destroy the system"
5516 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
5519 local FAKE_NIDS="192.168.0.112@tcp1,192.168.0.112@tcp2"
5520 local FAKE_FAILOVER="192.168.0.113@tcp1,192.168.0.113@tcp2"
5521 local NIDS_AND_FAILOVER="$MDS_NID,$FAKE_NIDS:$FAKE_FAILOVER"
5522 echo "set NIDs with failover"
5523 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $NIDS_AND_FAILOVER ||
5524 error "replace nids failed"
5527 echo "replace MDS nid"
5528 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
5529 error "replace nids failed"
5531 if ! combined_mgs_mds ; then
5534 stop_mds || error "Unable to stop MDS"
5537 start_mgsmds || error "start mgsmds failed"
5538 set_conf_param_and_check mds \
5539 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
5540 "$FSNAME-OST0000.osc.active" \
5542 start_ost || error "unable to start OST"
5543 mount_client $MOUNT || error "mount client failed"
5545 check_mount || error "error after nid replace"
5546 cleanup || error "cleanup failed"
5549 run_test 66 "replace nids"
5551 test_67() { #LU-2950
5552 local legacy="$TMP/legacy_lnet_config"
5553 local new="$TMP/new_routes_test"
5554 local out="$TMP/config_out_file"
5555 local verify="$TMP/conv_verify"
5556 local verify_conf="$TMP/conf_verify"
5558 # Create the legacy file that will be run through the
5559 # lustre_routes_conversion script
5560 cat <<- LEGACY_LNET_CONFIG > $legacy
5561 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
5562 tcp2 54 10.1.3.2@tcp;
5563 tcp3 10.3.4.3@tcp:3;
5567 # Create the verification file to verify the output of
5568 # lustre_routes_conversion script against.
5569 cat <<- VERIFY_LNET_CONFIG > $verify
5570 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
5571 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
5572 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
5573 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
5574 tcp4: { gateway: 10.3.3.4@tcp }
5577 # Create the verification file to verify the output of
5578 # lustre_routes_config script against
5579 cat <<- VERIFY_LNET_CONFIG > $verify_conf
5580 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
5581 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
5582 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
5583 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
5584 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
5587 $LUSTRE_ROUTES_CONVERSION $legacy $new > /dev/null
5588 if [ -f $new ]; then
5589 # verify the conversion output
5590 cmp -s $new $verify > /dev/null
5591 if [ $? -eq 1 ]; then
5592 error "routes conversion failed"
5595 lustre_routes_config --dry-run --verbose $new > $out
5596 # check that the script succeeded
5597 cmp -s $out $verify_conf > /dev/null
5598 if [ $? -eq 1 ]; then
5599 error "routes config failed"
5602 error "routes conversion test failed"
5604 # remove generated files
5605 rm -f $new $legacy $verify $verify_conf $out
5607 run_test 67 "test routes conversion and configuration"
5615 [ "$MDS1_VERSION" -ge $(version_code 2.4.53) ] ||
5616 skip "Need MDS version at least 2.4.53"
5618 umount_client $MOUNT || error "umount client failed"
5623 # START-END - the sequences we'll be reserving
5624 START=$(do_facet $SINGLEMDS \
5625 $LCTL get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
5626 END=$((START + (1 << 30)))
5627 do_facet $SINGLEMDS \
5628 $LCTL set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
5630 # reset the sequences MDT0000 has already assigned
5631 do_facet $SINGLEMDS \
5632 $LCTL set_param seq.srv*MDT0000.space=clear
5634 # remount to let the client allocate new sequence
5635 mount_client $MOUNT || error "mount client failed"
5637 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
5638 do_facet $SINGLEMDS \
5639 $LCTL get_param seq.srv*MDT0000.space
5640 $LFS path2fid $DIR/$tfile
5642 local old_ifs="$IFS"
5644 fid=($($LFS path2fid $DIR/$tfile))
5648 if [[ $seq < $END ]]; then
5649 error "used reserved sequence $seq?"
5651 cleanup || error "cleanup failed with $?"
5653 run_test 68 "be able to reserve specific sequences in FLDB"
5655 # Test 69: is about the total number of objects ever created on an OST.
5656 # so that when it is reformatted the normal MDS->OST orphan recovery won't
5657 # just "precreate" the missing objects. In the past it might try to recreate
5658 # millions of objects after an OST was reformatted
5660 [[ "$MDS1_VERSION" -lt $(version_code 2.4.2) ]] &&
5661 skip "Need MDS version at least 2.4.2"
5663 [[ "$MDS1_VERSION" -ge $(version_code 2.4.50) ]] &&
5664 [[ "$MDS1_VERSION" -lt $(version_code 2.5.0) ]] &&
5665 skip "Need MDS version at least 2.5.0"
5668 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5669 do_nodes $(comma_list $(osts_nodes)) $LCTL set_param \
5670 seq.*OST*-super.width=$DATA_SEQ_MAX_WIDTH
5672 # use OST0000 since it probably has the most creations
5673 local OSTNAME=$(ostname_from_index 0)
5674 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
5675 local last_id=$(do_facet mds1 $LCTL get_param -n \
5676 osp.$mdtosc_proc1.prealloc_last_id)
5678 # Want to have OST LAST_ID over 5 * OST_MAX_PRECREATE to
5679 # verify that the LAST_ID recovery is working properly. If
5680 # not, then the OST will refuse to allow the MDS connect
5681 # because the LAST_ID value is too different from the MDS
5682 #define OST_MAX_PRECREATE=20000
5683 local ost_max_pre=20000
5684 local num_create=$(( ost_max_pre * 5 + 1 - last_id))
5686 # If the LAST_ID is already over 5 * OST_MAX_PRECREATE, we don't
5687 # need to create any files. So, skip this section.
5688 if [ $num_create -gt 0 ]; then
5689 # Check the number of inodes available on OST0
5691 local ifree=$($LFS df -i $MOUNT |
5692 awk '/OST0000/ { print $4 }'; exit ${PIPESTATUS[0]})
5693 log "On OST0, $ifree inodes available. Want $num_create. rc=$?"
5695 $LFS setstripe -i 0 $DIR/$tdir ||
5696 error "$LFS setstripe -i 0 $DIR/$tdir failed"
5697 if [ $ifree -lt 10000 ]; then
5698 files=$(( ifree - 50 ))
5703 local j=$((num_create / files + 1))
5704 for i in $(seq 1 $j); do
5705 createmany -o $DIR/$tdir/$tfile-$i- $files ||
5706 error "createmany fail create $files files: $?"
5707 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
5708 error "unlinkmany failed unlink $files files"
5712 # delete all of the files with objects on OST0 so the
5713 # filesystem is not inconsistent later on
5714 $LFS find $MOUNT --ost 0 -print0 | xargs -0 rm
5716 umount_client $MOUNT || error "umount client failed"
5717 stop_ost || error "OST0 stop failure"
5718 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
5719 $(ostdevname 1) $(ostvdevname 1) ||
5720 error "reformat and replace $ostdev failed"
5721 start_ost || error "OST0 restart failure"
5722 wait_osc_import_state mds ost FULL
5724 mount_client $MOUNT || error "mount client failed"
5725 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
5726 local idx=$($LFS getstripe -i $DIR/$tdir/$tfile-last)
5727 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
5729 local iused=$($LFS df -i $MOUNT |
5730 awk '/OST0000/ { print $3 }'; exit ${PIPESTATUS[0]})
5731 log "On OST0, $iused used inodes rc=$?"
5732 [ $iused -ge $((ost_max_pre + 1000)) ] &&
5733 error "OST replacement created too many inodes; $iused"
5734 cleanup || error "cleanup failed with $?"
5736 run_test 69 "replace an OST with the same index"
5739 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5742 cleanup || error "cleanup failed with $?"
5744 start_mdt 1 || error "MDT0 start fail"
5746 start_ost || error "OST0 start fail"
5747 for num in $(seq 2 $MDSCOUNT); do
5748 start_mdt $num || return
5751 mount_client $MOUNT || error "mount client fails"
5753 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
5755 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5756 error "create remote dir fail"
5758 rm -rf $DIR/$tdir || error "delete dir fail"
5759 cleanup || error "cleanup failed with $?"
5761 run_test 70a "start MDT0, then OST, then MDT1"
5764 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5767 start_ost || error "OST0 start fail"
5769 start_mds || error "MDS start fail"
5771 mount_client $MOUNT || error "mount client fails"
5773 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
5775 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5776 error "create remote dir fail"
5778 rm -rf $DIR/$tdir || error "delete dir fail"
5780 cleanup || error "cleanup failed with $?"
5782 run_test 70b "start OST, MDT1, MDT0"
5785 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5788 start_mds || error "MDS start fail"
5789 start_ost || error "OST0 start fail"
5791 mount_client $MOUNT || error "mount client fails"
5792 stop_mdt 1 || error "MDT1 start fail"
5794 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
5795 echo "deactivate $mdc_for_mdt1"
5796 $LCTL --device $mdc_for_mdt1 deactivate ||
5797 error "set $mdc_for_mdt1 deactivate failed"
5799 mkdir $DIR/$tdir && error "mkdir succeed"
5801 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
5802 error "create remote dir succeed"
5804 cleanup || error "cleanup failed with $?"
5806 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
5809 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5812 start_mds || error "MDS start fail"
5813 start_ost || error "OST0 start fail"
5815 mount_client $MOUNT || error "mount client fails"
5817 stop_mdt 2 || error "MDT1 start fail"
5819 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
5821 echo "deactivate $mdc_for_mdt2"
5822 $LCTL --device $mdc_for_mdt2 deactivate ||
5823 error "set $mdc_for_mdt2 deactivate failed"
5825 mkdir $DIR/$tdir || error "mkdir fail"
5826 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
5827 error "create remote dir succeed"
5829 rm -rf $DIR/$tdir || error "delete dir fail"
5831 cleanup || error "cleanup failed with $?"
5833 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
5836 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5838 [ "$MDS1_VERSION" -ge $(version_code 2.7.62) ] ||
5839 skip "Need MDS version at least 2.7.62"
5841 reformat || error "reformat failed with $?"
5845 local mdsdev=$(mdsdevname 1)
5846 local ostdev=$(ostdevname 1)
5847 local mdsvdev=$(mdsvdevname 1)
5848 local ostvdev=$(ostvdevname 1)
5849 local opts_mds="$(mkfs_opts mds1 $mdsdev) --reformat $mdsdev $mdsvdev"
5850 local opts_ost="$(mkfs_opts ost1 $ostdev) --reformat $ostdev $ostvdev"
5852 if ! combined_mgs_mds ; then
5856 add mds1 $opts_mds || error "add mds1 failed"
5857 start_mdt 1 || error "start mdt1 failed"
5858 add ost1 $opts_ost || error "add ost1 failed"
5859 start_ost || error "start ost failed"
5860 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
5862 local soc=$(do_facet mds1 "$LCTL get_param -n \
5863 mdt.*MDT0000.sync_lock_cancel")
5864 [ $soc == "never" ] || error "SoC enabled on single MDS"
5866 for i in $(seq 2 $MDSCOUNT); do
5867 mdsdev=$(mdsdevname $i)
5868 mdsvdev=$(mdsvdevname $i)
5869 opts_mds="$(mkfs_opts mds$i $mdsdev) --reformat $mdsdev \
5871 add mds$i $opts_mds || error "add mds$i failed"
5872 start_mdt $i || error "start mdt$i fail"
5875 wait_dne_interconnect
5877 for i in $(seq $MDSCOUNT); do
5878 soc=$(do_facet mds$i "$LCTL get_param -n \
5879 mdt.*MDT000$((i - 1)).sync_lock_cancel")
5880 [ $soc == "blocking" ] || error "SoC not enabled on DNE"
5883 for i in $(seq 2 $MDSCOUNT); do
5884 stop_mdt $i || error "stop mdt$i fail"
5886 soc=$(do_facet mds1 "$LCTL get_param -n \
5887 mdt.*MDT0000.sync_lock_cancel")
5888 [ $soc == "never" ] || error "SoC enabled on single MDS"
5889 umount_client $MOUNT -f > /dev/null
5891 cleanup || error "cleanup failed with $?"
5893 run_test 70e "Sync-on-Cancel will be enabled by default on DNE"
5896 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5897 if combined_mgs_mds; then
5898 skip "needs separate MGS/MDT"
5902 start_mdt 1 || error "MDT0 start fail"
5903 start_ost || error "OST0 start fail"
5904 for num in $(seq 2 $MDSCOUNT); do
5905 start_mdt $num || return
5908 start_ost2 || error "OST1 start fail"
5910 mount_client $MOUNT || error "mount client fails"
5912 mkdir $DIR/$tdir || error "mkdir fail"
5913 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5914 error "create remote dir succeed"
5916 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5917 rm -rf $DIR/$tdir || error "delete dir fail"
5919 umount_client $MOUNT || error "umount_client failed"
5920 stop_mds || error "MDS stop fail"
5921 stop_ost || error "OST0 stop fail"
5922 stop_ost2 || error "OST1 stop fail"
5924 run_test 71a "start MDT0 OST0, MDT1, OST1"
5927 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5928 if combined_mgs_mds; then
5929 skip "needs separate MGS/MDT"
5933 for num in $(seq 2 $MDSCOUNT); do
5934 start_mdt $num || return
5936 start_ost || error "OST0 start fail"
5937 start_mdt 1 || error "MDT0 start fail"
5938 start_ost2 || error "OST1 start fail"
5940 mount_client $MOUNT || error "mount client fails"
5942 mkdir $DIR/$tdir || error "mkdir fail"
5943 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5944 error "create remote dir succeed"
5946 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5947 rm -rf $DIR/$tdir || error "delete dir fail"
5949 umount_client $MOUNT || error "umount_client failed"
5950 stop_mds || error "MDT0 stop fail"
5951 stop_ost || error "OST0 stop fail"
5952 stop_ost2 || error "OST1 stop fail"
5954 run_test 71b "start MDT1, OST0, MDT0, OST1"
5957 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5958 combined_mgs_mds && skip "needs separate MGS/MDT"
5962 start_ost || error "OST0 start fail"
5963 start_ost2 || error "OST1 start fail"
5964 for num in $(seq 2 $MDSCOUNT); do
5965 start_mdt $num || return
5967 start_mdt 1 || error "MDT0 start fail"
5969 mount_client $MOUNT || error "mount client fails"
5971 mkdir $DIR/$tdir || error "mkdir fail"
5972 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5973 error "create remote dir succeed"
5975 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5976 rm -rf $DIR/$tdir || error "delete dir fail"
5978 umount_client $MOUNT || error "umount_client failed"
5979 stop_mds || error "MDS stop fail"
5980 stop_ost || error "OST0 stop fail"
5981 stop_ost2 || error "OST1 stop fail"
5984 run_test 71c "start OST0, OST1, MDT1, MDT0"
5987 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
5988 combined_mgs_mds && skip "needs separate MGS/MDT"
5992 start_ost || error "OST0 start fail"
5993 for num in $(seq 2 $MDSCOUNT); do
5994 start_mdt $num || return
5996 start_mdt 1 || error "MDT0 start fail"
5997 start_ost2 || error "OST1 start fail"
5999 mount_client $MOUNT || error "mount client fails"
6001 mkdir $DIR/$tdir || error "mkdir fail"
6002 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
6003 error "create remote dir succeed"
6005 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
6006 rm -rf $DIR/$tdir || error "delete dir fail"
6008 umount_client $MOUNT || error "umount_client failed"
6009 stop_mds || error "MDS stop fail"
6010 stop_ost || error "OST0 stop fail"
6011 stop_ost2 || error "OST1 stop fail"
6014 run_test 71d "start OST0, MDT1, MDT0, OST1"
6017 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
6018 combined_mgs_mds && skip "needs separate MGS/MDT"
6022 start_ost || error "OST0 start fail"
6023 for num in $(seq 2 $MDSCOUNT); do
6024 start_mdt $num || return
6026 start_ost2 || error "OST1 start fail"
6027 start_mdt 1 || error "MDT0 start fail"
6029 mount_client $MOUNT || error "mount client fails"
6031 mkdir $DIR/$tdir || error "mkdir fail"
6032 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
6033 error "create remote dir succeed"
6035 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
6036 rm -rf $DIR/$tdir || error "delete dir fail"
6038 umount_client $MOUNT || error "umount_client failed"
6039 stop_mds || error "MDS stop fail"
6040 stop_ost || error "OST0 stop fail"
6041 stop_ost2 || error "OST1 stop fail"
6044 run_test 71e "start OST0, MDT1, OST1, MDT0"
6046 test_72() { #LU-2634
6047 [ "$mds1_FSTYPE" != ldiskfs ] &&
6048 skip "ldiskfs only test"
6050 local mdsdev=$(mdsdevname 1)
6051 local ostdev=$(ostdevname 1)
6052 local cmd="$E2FSCK -fnvd $mdsdev"
6059 if combined_mgs_mds; then
6060 add_options='--reformat'
6062 add_options='--reformat --replace'
6065 #tune MDT with "-O extents"
6067 for num in $(seq $MDSCOUNT); do
6068 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
6069 $add_options $(mdsdevname $num) $(mdsvdevname $num) ||
6070 error "add mds $num failed"
6071 do_facet mds${num} "$TUNE2FS -O extents $(mdsdevname $num)" ||
6072 error "$TUNE2FS failed on mds${num}"
6075 add ost1 $(mkfs_opts ost1 $ostdev) $add_options $ostdev ||
6076 error "add $ostdev failed"
6077 start_mds || error "start mds failed"
6078 start_ost || error "start ost failed"
6079 mount_client $MOUNT || error "mount client failed"
6081 #create some short symlinks
6082 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6083 createmany -o $DIR/$tdir/$tfile-%d $fn
6084 echo "create $fn short symlinks"
6085 for i in $(seq -w 1 $fn); do
6086 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
6091 umount_client $MOUNT || error "umount client failed"
6092 stop_mds || error "stop mds failed"
6093 stop_ost || error "stop ost failed"
6096 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
6098 run_test 72 "test fast symlink with extents flag enabled"
6100 test_73() { #LU-3006
6101 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
6102 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
6103 error "1st tunefs failed"
6104 start_mgsmds || error "start mds failed"
6105 start_ost || error "start ost failed"
6106 mount_client $MOUNT || error "mount client failed"
6107 $LCTL get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
6108 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
6109 umount_client $MOUNT || error "umount client failed"
6113 run_test 73 "failnode to update from mountdata properly"
6117 (( $MDS1_VERSION >= $(version_code 2.15.57.16) )) ||
6118 skip "need MDS version >= 2.15.57.16 for per-device timeouts"
6121 stack_trap "cleanup"
6123 # Prepare fs2, share the mgs of fs
6124 local FSNAME2=fs15246
6125 local fs2mdsdev=$(mdsdevname 1_2)
6126 local fs2ostdev=$(ostdevname 1_2)
6127 local fs2mdsvdev=$(mdsvdevname 1_2)
6128 local fs2ostvdev=$(ostvdevname 1_2)
6130 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --fsname=$FSNAME2 \
6131 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
6132 add fs2ost $(mkfs_opts ost1 $fs2ostdev) --fsname=$FSNAME2 \
6133 --reformat $fs2ostdev $fs2ostvdev || error "add fs2ost failed"
6135 stack_trap "cleanup_fs2"
6137 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
6138 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS || error "start fs2mds failed"
6140 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
6141 $MOUNT_CMD $MGSNID:/$FSNAME2 $MOUNT2 || error "mount $MOUNT2 failed"
6143 echo "========== All mounted lustre fs ===================="
6144 mount | grep 'type lustre'
6145 echo "====================================================="
6147 # Set and check osc/ldlm_enqueue_min
6148 $LCTL set_param osc.${FSNAME}-*.ldlm_enqueue_min=99 ||
6149 error "failed to set per-device adaptive parameters on client"
6150 stack_trap "$LCTL set_param osc.${FSNAME}-*.ldlm_enqueue_min=0"
6152 local ldlm_enqueue_min
6153 ldlm_enqueue_min=$($LCTL get_param -n osc.${FSNAME}-*.ldlm_enqueue_min \
6155 (( $ldlm_enqueue_min == 99 )) ||
6156 error "wrong ldlm_enqueue_min value for osc.${FSNAME}-*"
6158 # Check fs2 as reference
6159 ldlm_enqueue_min=$($LCTL get_param -n osc.${FSNAME2}-*.ldlm_enqueue_min\
6161 (( $ldlm_enqueue_min == 0 )) ||
6162 error "wrong ldlm_enqueue_min value for osc.${FSNAME2}-*"
6164 # Set and check obdfilter/at_min
6165 do_facet ost1 $LCTL set_param obdfilter.${FSNAME}-*.at_min=1 ||
6166 error "failed to set per-device adaptive parameters on ost"
6167 stack_trap "do_facet ost1 $LCTL set_param obdfilter.${FSNAME}-*.at_min=0"
6170 at_min=$(do_facet ost1 $LCTL get_param -n obdfilter.${FSNAME}-*.at_min \
6172 (( $at_min == 1 )) ||
6173 error "wrong at_min value for obdfilter.${FSNAME}-*"
6175 # set and check mdc/at_max
6176 $LCTL set_param mdc.${FSNAME}-*.at_max=599 ||
6177 error "failed to set per-device adaptive parameters on client"
6178 stack_trap "$LCTL set_param mdc.${FSNAME}-*.at_max=0"
6181 at_max=$($LCTL get_param -n mdc.${FSNAME}-*.at_max | uniq)
6182 (( $at_max == 599 )) ||
6183 error "wrong at_max value for osc.${FSNAME}-*"
6185 # Check fs2 as reference
6186 at_max=$($LCTL get_param -n mdc.${FSNAME2}-*.at_max | uniq)
6187 (( $at_max == 0 )) ||
6188 error "wrong at_max value for osc.${FSNAME2}-*"
6190 # Set and check mds/at_max
6191 do_facet mds1 $LCTL set_param *.${FSNAME}-*.at_max=599 ||
6192 error "failed to set per-device adaptive parameters on mds"
6193 stack_trap "do_facet mds1 $LCTL set_param *.${FSNAME}-*.at_max=0"
6196 at_max=$(do_facet mds1 $LCTL get_param -n mdt.${FSNAME}-*.at_max | uniq)
6197 (( at_max == 599 )) ||
6198 error "wrong at_max value for mdt.${FSNAME}-*"
6200 # Set and check mgs&mgc/at_history
6201 local mgs_nid=$(do_facet $SINGLEMDS $LCTL list_nids | tail -1)
6202 $LCTL set_param mgc.MGC$mgs_nid.at_history=588
6203 stack_trap "$LCTL set_param mgc.MGC$mgs_nid.at_history=0"
6206 at_history=$($LCTL get_param -n mgc.MGC$mgs_nid.at_history)
6207 (( $at_history == 588 )) ||
6208 error "wrong at_history value for mgc.MGC$mgs_nid"
6210 do_facet mgs $LCTL set_param mgs.MGS.at_history=588
6211 stack_trap "do_facet mgs $LCTL set_param mgs.MGS.at_history=0"
6212 at_history=$(do_facet mgs $LCTL get_param -n mgs.MGS.at_history)
6213 (( $at_history == 588 )) ||
6214 error "wrong at_history value for mgs.MGS"
6216 run_test 74 "Test per-device adaptive timeout parameters"
6218 test_75() { # LU-2374
6219 [[ "$MDS1_VERSION" -lt $(version_code 2.4.1) ]] &&
6220 skip "Need MDS version at least 2.4.1"
6223 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
6224 --replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
6225 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
6226 --replace --reformat $(ostdevname 1) $(ostvdevname 1)"
6229 #check with default parameters
6230 add mds1 $opts_mds || error "add mds1 failed for default params"
6231 add ost1 $opts_ost || error "add ost1 failed for default params"
6233 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
6234 opts_mds=$(echo $opts_mds |
6235 sed -e "s/--index=$index/--index=$index --mdt/")
6236 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
6237 opts_ost=$(echo $opts_ost |
6238 sed -e "s/--index=$index/--index=$index --ost/")
6240 add mds1 $opts_mds || error "add mds1 failed for new params"
6241 add ost1 $opts_ost || error "add ost1 failed for new params"
6246 run_test 75 "The order of --index should be irrelevant"
6249 [[ "$MGS_VERSION" -ge $(version_code 2.4.52) ]] ||
6250 skip "Need MDS version at least 2.4.52"
6253 local MDMB_PARAM="osc.*.max_dirty_mb"
6254 echo "Change MGS params"
6255 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
6257 echo "max_dirty_mb: $MAX_DIRTY_MB"
6258 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB - 10))
6259 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
6260 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
6261 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
6262 head -1" $NEW_MAX_DIRTY_MB
6263 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
6264 echo "$MAX_DIRTY_MB"
6265 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
6266 error "error while apply max_dirty_mb"
6268 echo "Check the value is stored after remount"
6271 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
6272 head -1" $NEW_MAX_DIRTY_MB
6273 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
6274 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
6275 error "max_dirty_mb is not saved after remount"
6277 echo "Change OST params"
6278 CLIENT_PARAM="obdfilter.*.client_cache_count"
6279 local CLIENT_CACHE_COUNT
6280 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
6282 echo "client_cache_count: $CLIENT_CACHE_COUNT"
6283 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
6284 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
6285 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
6286 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
6287 head -1" $NEW_CLIENT_CACHE_COUNT
6288 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
6290 echo "$CLIENT_CACHE_COUNT"
6291 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
6292 error "error while apply client_cache_count"
6294 echo "Check the value is stored after remount"
6297 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
6298 head -1" $NEW_CLIENT_CACHE_COUNT
6299 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
6301 echo "$CLIENT_CACHE_COUNT"
6302 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
6303 error "client_cache_count is not saved after remount"
6306 run_test 76a "set permanent params with lctl across mounts"
6308 test_76b() { # LU-4783
6309 [[ "$MGS_VERSION" -ge $(version_code 2.5.57) ]] ||
6310 skip "Need MGS version at least 2.5.57"
6313 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
6314 error "start params log failed"
6317 run_test 76b "verify params log setup correctly"
6320 [[ "$MGS_VERSION" -ge $(version_code 2.8.54) ]] ||
6321 skip "Need MDS version at least 2.4.52"
6323 local MASK_PARAM="mdd.*.changelog_mask"
6324 echo "Change changelog_mask"
6325 do_facet mgs $LCTL set_param -P $MASK_PARAM=-CLOSE ||
6326 error "Can't change changlog_mask"
6327 wait_update $(facet_host mds) "$LCTL get_param -n $MASK_PARAM |
6330 echo "Check the value is stored after mds remount"
6331 stop_mds || error "Failed to stop MDS"
6332 start_mds || error "Failed to start MDS"
6333 local CHANGELOG_MASK=$(do_facet mgs $LCTL get_param -n $MASK_PARAM)
6334 echo $CHANGELOG_MASK | grep CLOSE > /dev/null &&
6335 error "changelog_mask is not changed"
6339 run_test 76c "verify changelog_mask is applied with lctl set_param -P"
6341 test_76d() { #LU-9399
6344 local xattr_cache="llite.*.xattr_cache"
6345 local cmd="$LCTL get_param -n $xattr_cache | head -1"
6346 local new=$((($(eval $cmd) + 1) % 2))
6348 echo "lctl set_param -P llite.*.xattr_cache=$new"
6349 do_facet mgs $LCTL set_param -P $xattr_cache=$new ||
6350 error "Can't change xattr_cache"
6351 wait_update $HOSTNAME "$cmd" "$new"
6353 echo "Check $xattr_cache on client $MOUNT"
6354 umount_client $MOUNT || error "umount $MOUNT failed"
6355 mount_client $MOUNT || error "mount $MOUNT failed"
6356 [ $(eval $cmd) -eq $new ] ||
6357 error "$xattr_cache != $new on client $MOUNT"
6359 echo "Check $xattr_cache on the new client $MOUNT2"
6360 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
6361 [ $(eval $cmd) -eq $new ] ||
6362 error "$xattr_cache != $new on client $MOUNT2"
6363 umount_client $MOUNT2 || error "umount $MOUNT2 failed"
6367 run_test 76d "verify llite.*.xattr_cache can be set by 'lctl set_param -P' correctly"
6369 test_77() { # LU-3445
6370 [[ "$MDS1_VERSION" -ge $(version_code 2.8.55) ]] ||
6371 skip "Need MDS version 2.8.55+ "
6373 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
6374 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
6375 skip_env "mixed loopback and real device not working"
6378 local fs2mdsdev=$(mdsdevname 1_2)
6379 local fs2ostdev=$(ostdevname 1_2)
6380 local fs2mdsvdev=$(mdsvdevname 1_2)
6381 local fs2ostvdev=$(ostvdevname 1_2)
6382 local fsname=test1234
6384 local failnid="$(h2nettype 1.2.3.4),$(h2nettype 4.3.2.1)"
6386 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
6388 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
6389 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
6390 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
6391 error "start fs2mds failed"
6393 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
6394 mgsnid="0.0.0.0@tcp,$mgsnid,$mgsnid:$mgsnid"
6396 add fs2ost --mgsnode=$mgsnid $(mkfs_opts ost1 $fs2ostdev) \
6397 --failnode=$failnid --fsname=$fsname \
6398 --reformat $fs2ostdev $fs2ostvdev ||
6399 error "add fs2ost failed"
6400 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
6402 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
6403 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
6404 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
6407 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
6410 [[ "$mds1_FSTYPE" != ldiskfs ||
6411 "$ost1_FSTYPE" != ldiskfs ]] &&
6412 skip "ldiskfs only test"
6414 # reformat the Lustre filesystem with a smaller size
6415 local saved_MDSCOUNT=$MDSCOUNT
6416 local saved_MDSSIZE=$MDSSIZE
6417 local saved_OSTCOUNT=$OSTCOUNT
6418 local saved_OSTSIZE=$OSTSIZE
6421 MDSSIZE=$((MDSSIZE - 20000))
6422 OSTSIZE=$((OSTSIZE - 20000))
6423 reformat || error "(1) reformat Lustre filesystem failed"
6424 MDSSIZE=$saved_MDSSIZE
6425 OSTSIZE=$saved_OSTSIZE
6427 # mount the Lustre filesystem
6428 setup_noconfig || error "(2) setup Lustre filesystem failed"
6431 log "create test files"
6436 mkdir $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
6438 for i in $(seq $num_files); do
6439 file=$MOUNT/$tdir/$tfile-$i
6440 dd if=/dev/urandom of=$file count=1 bs=1M || {
6441 $LCTL get_param osc.*.cur*grant*
6442 $LFS df; $LFS df -i;
6443 # stop creating files if there is no more space
6444 if [ ! -e $file ]; then
6445 num_files=$((i - 1))
6449 $LFS getstripe -v $file
6450 local ost_idx=$(LFS getstripe -i $file)
6451 do_facet ost$((ost_idx + 1)) \
6452 $LCTL get_param obdfilter.*.*grant*
6453 error "(4) create $file failed"
6457 # unmount the Lustre filesystem
6458 cleanup || error "(5) cleanup Lustre filesystem failed"
6460 # run e2fsck on the MDT and OST devices
6461 local mds_host=$(facet_active_host $SINGLEMDS)
6462 local ost_host=$(facet_active_host ost1)
6463 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
6464 local ost_dev=$(ostdevname 1)
6466 run_e2fsck $mds_host $mds_dev "-y"
6467 run_e2fsck $ost_host $ost_dev "-y"
6469 # get the original block count of the MDT and OST filesystems
6470 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
6471 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
6473 # expand the MDT and OST filesystems to the device size
6474 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
6475 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
6477 # run e2fsck on the MDT and OST devices again
6478 run_e2fsck $mds_host $mds_dev "-y"
6479 run_e2fsck $ost_host $ost_dev "-y"
6481 # mount the Lustre filesystem
6485 log "check files after expanding the MDT and OST filesystems"
6486 for i in $(seq $num_files); do
6487 file=$MOUNT/$tdir/$tfile-$i
6488 $CHECKSTAT -t file -s 1048576 $file ||
6489 error "(6) checkstat $file failed"
6493 log "create more files after expanding the MDT and OST filesystems"
6494 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
6495 file=$MOUNT/$tdir/$tfile-$i
6496 dd if=/dev/urandom of=$file count=1 bs=1M ||
6497 error "(7) create $file failed"
6500 # unmount the Lustre filesystem
6501 cleanup || error "(8) cleanup Lustre filesystem failed"
6503 # run e2fsck on the MDT and OST devices
6504 run_e2fsck $mds_host $mds_dev "-y"
6505 run_e2fsck $ost_host $ost_dev "-y"
6507 # get the maximum block count of the MDT and OST filesystems
6508 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
6509 local ost_max_blks=$(get_block_count ost1 $ost_dev)
6511 # get the minimum block count of the MDT and OST filesystems
6512 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
6513 grep minimum | sed -e 's/^.*filesystem: //g')
6514 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
6515 grep minimum | sed -e 's/^.*filesystem: //g')
6517 # shrink the MDT and OST filesystems to a smaller size
6521 if [[ $mds_max_blks -gt $mds_min_blks &&
6522 $mds_max_blks -gt $mds_orig_blks ]]; then
6523 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
6524 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
6525 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
6526 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
6527 error "shrink $SINGLEMDS to $new_blks failed"
6531 if [[ $ost_max_blks -gt $ost_min_blks &&
6532 $ost_max_blks -gt $ost_orig_blks ]]; then
6533 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
6534 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
6535 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
6536 run_resize2fs ost1 $ost_dev $new_blks ||
6537 error "shrink ost1 to $new_blks failed"
6541 # check whether the MDT or OST filesystem was shrunk or not
6543 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
6544 reformat_and_config ||
6545 error "(10) reformat Lustre filesystem failed"
6549 # run e2fsck on the MDT and OST devices again
6550 run_e2fsck $mds_host $mds_dev "-y"
6551 run_e2fsck $ost_host $ost_dev "-y"
6553 # mount the Lustre filesystem again
6557 log "check files after shrinking the MDT and OST filesystems"
6558 for i in $(seq $((num_files + 10))); do
6559 file=$MOUNT/$tdir/$tfile-$i
6560 $CHECKSTAT -t file -s 1048576 $file ||
6561 error "(11) checkstat $file failed"
6564 # unmount and reformat the Lustre filesystem
6565 cleanup || error "(12) cleanup Lustre filesystem failed"
6566 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
6568 MDSCOUNT=$saved_MDSCOUNT
6569 OSTCOUNT=$saved_OSTCOUNT
6570 reformat_and_config || error "(14) reformat Lustre filesystem failed"
6572 run_test 78 "run resize2fs on MDT and OST filesystems"
6574 test_79() { # LU-4227
6575 [[ "$MDS1_VERSION" -ge $(version_code 2.5.59) ]] ||
6576 skip "Need MDS version at least 2.5.59"
6578 local mdsdev1=$(mdsdevname 1)
6579 local mdsvdev1=$(mdsvdevname 1)
6580 local mdsdev2=$(mdsdevname 2)
6581 local mdsvdev2=$(mdsvdevname 2)
6582 local ostdev1=$(ostdevname 1)
6583 local ostvdev1=$(ostvdevname 1)
6584 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
6585 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
6586 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
6589 # remove --mgs/--mgsnode from mkfs.lustre options
6590 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
6592 mgsnode_opt=$(echo $opts_mds2 |
6593 awk '{ for ( i = 1; i < NF; i++ )
6594 if ( $i ~ "--mgsnode" ) { print $i; break } }')
6595 [ -n "$mgsnode_opt" ] &&
6596 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
6598 mgsnode_opt=$(echo $opts_ost1 |
6599 awk '{ for ( i = 1; i < NF; i++ )
6600 if ( $i ~ "--mgsnode" ) { print $i; break } }')
6601 [ -n "$mgsnode_opt" ] &&
6602 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
6604 # -MGS, format a mdt without --mgs option
6605 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
6606 error "Must specify --mgs when formatting mdt combined with mgs"
6608 # +MGS, format a mdt/ost without --mgsnode option
6609 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
6610 > /dev/null || error "start mds1 failed"
6611 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
6612 error "Must specify --mgsnode when formatting a mdt"
6613 add ost1 $opts_ost1 $ostdev1 $ostvdev1 &&
6614 error "Must specify --mgsnode when formatting an ost"
6618 run_test 79 "format MDT/OST without mgs option (should return errors)"
6621 start_mds || error "Failed to start MDT"
6622 start_ost || error "Failed to start OST1"
6623 uuid=$(do_facet ost1 $LCTL get_param -n mgc.*.uuid)
6624 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
6625 do_facet ost1 "$LCTL set_param fail_val=10 fail_loc=0x906"
6626 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x906"
6627 do_facet mgs "$LCTL set_param -n mgs/MGS/evict_client $uuid"
6629 start_ost2 || error "Failed to start OST2"
6631 do_facet ost1 "$LCTL set_param fail_loc=0"
6636 run_test 80 "mgc import reconnect race"
6638 #Save the original values of $OSTCOUNT and $OSTINDEX$i.
6640 local new_ostcount=$1
6641 saved_ostcount=$OSTCOUNT
6642 OSTCOUNT=$new_ostcount
6646 for ((i = 1; i <= $OSTCOUNT; i++ )); do
6648 eval saved_ostindex$i=${!index}
6653 # Restore the original values of $OSTCOUNT and $OSTINDEX$i.
6654 restore_ostindex() {
6658 echo "restoring OSTCOUNT=$saved_ostcount and OSTINDEXn"
6660 for ((i = 1; i <= $OSTCOUNT; i++ )); do
6661 index=saved_ostindex$i
6662 eval OSTINDEX$i=${!index}
6664 OSTCOUNT=$saved_ostcount
6667 if ! combined_mgs_mds ; then
6672 # The main purpose of this test is to ensure the OST_INDEX_LIST functions as
6673 # expected. This test uses OST_INDEX_LIST to format OSTs with a randomly
6674 # assigned index and ensures we can mount such a formatted file system
6675 test_81() { # LU-4665
6676 (( MDS1_VERSION >= $(version_code 2.6.54) )) ||
6677 skip "Need MDS version at least 2.6.54"
6678 (( OSTCOUNT >= 3 )) || skip_env "needs >= 3 OSTs"
6682 # Each time RANDOM is referenced, a random integer between 0 and 32767
6685 local saved_ostindex1=$OSTINDEX1
6686 for i in 65535 $((RANDOM + 65536)); do
6687 echo -e "\nFormat ost1 with --index=$i, should fail"
6689 if add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat \
6690 $(ostdevname 1) $(ostvdevname 1); then
6691 OSTINDEX1=$saved_ostindex1
6692 error "format ost1 with --index=$i should fail"
6695 OSTINDEX1=$saved_ostindex1
6698 stack_trap restore_ostindex
6700 # Format OSTs with random sparse indices.
6701 local rand_ost=$(((RANDOM * 2 % 65533) + 1))
6702 echo "Format $OSTCOUNT OSTs with OST_INDEX_LIST=[0,$rand_ost,65534]"
6703 OST_INDEX_LIST=[0,$rand_ost,65534] formatall ||
6704 error "formatall failed with $?"
6706 # Setup and check Lustre filesystem.
6707 start_mgsmds || error "start_mgsmds failed"
6708 for ((i = 1; i <= $OSTCOUNT; i++ )); do
6709 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6710 error "start ost$i failed"
6713 mount_client $MOUNT || error "mount client $MOUNT failed"
6714 check_mount || error "check client $MOUNT failed"
6718 local max_easize=$($LCTL get_param -n llite.*.max_easize)
6719 local xattr_size_max=$((4096 - 84)) # 4096 less ldiskfs ea overhead
6721 # XATTR_SIZE_MAX less ldiskfs ea overhead
6722 large_xattr_enabled && xattr_size_max=$((65536 - 84))
6723 (( max_easize >= xattr_size_max )) ||
6724 error "max_easize $max_easize < $xattr_size_max bytes"
6726 test_mkdir $DIR/$tdir
6727 $LFS setstripe -i $rand_ost $DIR/$tdir/$tfile ||
6728 error "error creating $tfile on ost$rand_ost"
6729 $LFS getstripe $DIR/$tdir/$tfile
6736 cmd="getstripe -i $DIR/$tdir/$tfile"
6739 (( $found == $rand_ost )) || error "index $found is not $rand_ost"
6741 cmd="find $DIR/$tdir -i $rand_ost"
6745 [[ "$found" == "$DIR/$tdir/$tfile" ]] ||
6746 error "'lfs find' returned '$found', not '$tfile' by index"
6749 uuid=$(ostuuid_from_index $rand_ost)
6751 cmd="find $DIR/$tdir -O $uuid"
6756 [[ "$found" == "$DIR/$tdir/$tfile" ]] ||
6757 error "'lfs find' returned '$found', not '$tfile' by UUID"
6759 run_test 81 "sparse OST indexing"
6761 # Here we exercise the stripe placement functionality on a file system that
6762 # has formatted the OST with a random index. With the file system the following
6763 # functionality is tested:
6765 # 1. Creating a new file with a specific stripe layout.
6767 # 2. Modifiy a existing empty file with a specific stripe layout.
6769 # 3. Ensure we fail to set the stripe layout of a file that already has one.
6771 # 4. If ost-index is defined we need to ensure it is the first entry in the
6772 # ost index list returned by lfs getstripe.
6774 # 5. Lastly ensure this functionality fails with directories.
6775 test_82a() { # LU-4665
6776 [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
6777 skip "Need MDS version at least 2.6.54"
6778 [[ $OSTCOUNT -ge 3 ]] || skip_env "needs >= 3 OSTs"
6784 # Format OSTs with random sparse indices.
6788 local LOV_V1_INSANE_STRIPE_COUNT=65532
6789 for i in $(seq $OSTCOUNT); do
6790 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
6791 ost_indices+=" $index"
6793 ost_indices=$(comma_list $ost_indices)
6795 stack_trap "restore_ostindex" EXIT
6796 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
6797 OST_INDEX_LIST=[$ost_indices] formatall
6799 # Setup Lustre filesystem.
6800 start_mgsmds || error "start_mgsmds failed"
6801 for i in $(seq $OSTCOUNT); do
6802 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6803 error "start ost$i failed"
6806 # Collect debug information - start of test
6807 do_nodes $(comma_list $(mdts_nodes)) \
6808 $LCTL get_param osp.*.prealloc_*_id
6810 mount_client $MOUNT || error "mount client $MOUNT failed"
6814 check_lfs_df_ret_val $? || error "$LFS df $MOUNT failed"
6815 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6817 stack_trap "do_nodes $(comma_list $(mdts_nodes)) \
6818 $LCTL get_param osp.*.prealloc_*_id || true" EXIT
6820 # 1. If the file does not exist, new file will be created
6821 # with specified OSTs.
6822 local file=$DIR/$tdir/$tfile-1
6823 local cmd="$LFS setstripe -o $ost_indices $file"
6825 eval $cmd || error "$cmd failed"
6826 check_stripe_count $file $OSTCOUNT
6827 check_obdidx $file $ost_indices
6828 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6829 error "write $file failed"
6831 # 2. If the file already exists and is an empty file, the file
6832 # will be attached with specified layout.
6833 file=$DIR/$tdir/$tfile-2
6834 mcreate $file || error "mcreate $file failed"
6835 cmd="$LFS setstripe -o $ost_indices $file"
6837 eval $cmd || error "$cmd failed"
6838 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6839 error "write $file failed"
6840 check_stripe_count $file $OSTCOUNT
6841 check_obdidx $file $ost_indices
6843 # 3. If the file already has a valid layout attached, the command
6844 # should fail with EBUSY.
6846 eval $cmd && error "stripe is already set on $file, $cmd should fail"
6848 # 4. If [--stripe-index|-i <start_ost_idx>] is used, the index must
6849 # be in the OST indices list.
6850 local start_ost_idx=${ost_indices##*,}
6851 file=$DIR/$tdir/$tfile-3
6852 cmd="$LFS setstripe -o $ost_indices -i $start_ost_idx $file"
6854 eval $cmd || error "$cmd failed"
6855 check_stripe_count $file $OSTCOUNT
6856 check_obdidx $file $ost_indices
6857 check_start_ost_idx $file $start_ost_idx
6859 file=$DIR/$tdir/$tfile-4
6860 cmd="$LFS setstripe"
6861 cmd+=" -o $(exclude_items_from_list $ost_indices $start_ost_idx)"
6862 cmd+=" -i $start_ost_idx $file"
6864 eval $cmd && error "index $start_ost_idx should be in $ost_indices"
6866 # 5. Specifying OST indices for directory should succeed.
6867 local dir=$DIR/$tdir/$tdir
6868 mkdir $dir || error "mkdir $dir failed"
6869 cmd="$LFS setstripe -o $ost_indices $dir"
6870 if [ "$MDS1_VERSION" -gt $(version_code 2.11.53) ] &&
6871 [ "$CLIENT_VERSION" -gt $(version_code 2.11.53) ]; then
6873 eval $cmd || error "unable to specify OST indices on directory"
6875 echo "need MDS+client version at least 2.11.53"
6878 run_test 82a "specify OSTs for file (succeed) or directory (succeed)"
6880 # Test 82b is run to ensure that if the user supplies a pool with a specific
6881 # stripe layout that it behaves proprerly. It should fail in the case that
6882 # the supplied OST index list points to OSTs not contained in the user
6884 test_82b() { # LU-4665
6885 [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] ||
6886 skip "Need MDS version at least 2.6.54"
6887 [[ $OSTCOUNT -ge 4 ]] || skip_env "needs >= 4 OSTs"
6893 # Format OSTs with random sparse indices.
6897 local LOV_V1_INSANE_STRIPE_COUNT=65532
6898 for i in $(seq $OSTCOUNT); do
6899 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
6900 ost_indices+=" $index"
6902 ost_indices=$(comma_list $ost_indices)
6904 stack_trap "restore_ostindex" EXIT
6905 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
6906 OST_INDEX_LIST=[$ost_indices] formatall
6908 # Setup Lustre filesystem.
6909 start_mgsmds || error "start_mgsmds failed"
6910 for i in $(seq $OSTCOUNT); do
6911 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
6912 error "start ost$i failed"
6915 mount_client $MOUNT || error "mount client $MOUNT failed"
6919 check_lfs_df_ret_val $? || error "$LFS df $MOUNT failed"
6920 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6922 # Create a new pool and add OSTs into it.
6923 local ost_pool=$FSNAME.$TESTNAME
6924 create_pool $ost_pool || error "create OST pool $ost_pool failed"
6926 local ost_idx_in_list=${ost_indices##*,}
6927 local ost_idx_in_pool=$(exclude_items_from_list $ost_indices \
6930 local ost_targets="$FSNAME-OST["
6931 for i in ${ost_idx_in_pool//,/ }; do
6932 ost_targets=$ost_targets$(printf "%04x," $i)
6934 ost_targets="${ost_targets%,}]"
6936 local ost_targets_uuid=$(for i in ${ost_idx_in_pool//,/ }; \
6937 do printf "$FSNAME-OST%04x_UUID\n" $i; done |
6938 sort -u | tr '\n' ' ')
6940 local cmd="$LCTL pool_add $ost_pool $ost_targets"
6941 do_facet mgs $cmd || error "$cmd failed"
6942 wait_update $HOSTNAME "$LCTL get_param -n lov.$FSNAME-*.pools.$TESTNAME|
6943 sort -u | tr '\n' ' ' " "$ost_targets_uuid" ||
6944 error "wait_update $ost_pool failed"
6945 wait_update_facet $SINGLEMDS "$LCTL pool_list $ost_pool | wc -l" 4 ||
6946 error "wait_update pool_list $ost_pool failed"
6948 # If [--pool|-p <pool_name>] is set with [--ost-list|-o <ost_indices>],
6949 # then the OSTs must be the members of the pool.
6950 local file=$DIR/$tdir/$tfile
6951 cmd="$LFS setstripe -p $ost_pool -o $ost_idx_in_list $file"
6953 eval $cmd && error "OST with index $ost_idx_in_list should be" \
6954 "in OST pool $ost_pool"
6956 # Only select OST $ost_idx_in_list from $ost_pool for file.
6957 ost_idx_in_list=${ost_idx_in_pool#*,}
6958 cmd="$LFS setstripe -p $ost_pool -o $ost_idx_in_list $file"
6960 eval $cmd || error "$cmd failed"
6961 cmd="$LFS getstripe $file"
6963 eval $cmd || error "$cmd failed"
6964 check_stripe_count $file 2
6965 check_obdidx $file $ost_idx_in_list
6966 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6967 error "write $file failed"
6969 run_test 82b "specify OSTs for file with --pool and --ost-list options"
6972 [[ "$OST1_VERSION" -ge $(version_code 2.6.91) ]] ||
6973 skip "Need OST version at least 2.6.91"
6974 if [ "$ost1_FSTYPE" != ldiskfs ]; then
6975 skip "ldiskfs only test"
6984 ostmnt=$(facet_mntpt ost1)
6986 # Mount the OST as an ldiskfs filesystem.
6987 log "mount the OST $dev as a $ost1_FSTYPE filesystem"
6988 add ost1 $(mkfs_opts ost1 $dev) $FSTYPE_OPT \
6989 --reformat $dev > /dev/null ||
6990 error "format ost1 error"
6992 if ! test -b $dev; then
6993 mnt_opts=$(csa_add "$OST_MOUNT_FS_OPTS" -o loop)
6995 echo "mnt_opts $mnt_opts"
6996 do_facet ost1 mount -t "$ost1_FSTYPE" $dev \
6998 # Run llverfs on the mounted ldiskfs filesystem.
6999 # It is needed to get ENOSPACE.
7000 log "run llverfs in partial mode on the OST $ost1_FSTYPE $ostmnt"
7001 do_rpc_nodes $(facet_host ost1) run_llverfs $ostmnt -vpl \
7002 "no" || error "run_llverfs error on $ost1_FSTYPE"
7005 log "unmount the OST $dev"
7008 # Delete file IO_scrub. Later osd_scrub_setup will try to
7009 # create "IO_scrub" but will get ENOSPACE.
7011 echo "start ost1 service on `facet_active_host ost1`"
7012 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
7015 err=$(do_facet ost1 dmesg | grep "VFS: Busy inodes after unmount of")
7016 echo "string err $err"
7017 [ -z "$err" ] || error $err
7020 run_test 83 "ENOSPACE on OST doesn't cause message VFS: \
7021 Busy inodes after unmount ..."
7024 local facet=$SINGLEMDS
7025 local num=$(echo $facet | tr -d "mds")
7026 local dev=$(mdsdevname $num)
7027 local time_min=$(recovery_time_min)
7028 local recovery_duration
7029 local completed_clients
7030 local correct_clients
7033 echo "start mds service on $(facet_active_host $facet)"
7035 "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
7036 error "start MDS failed"
7038 start_ost || error "start OST0000 failed"
7039 wait_osc_import_state mds ost1 FULL
7040 start_ost2 || error "start OST0001 failed"
7041 wait_osc_import_state mds ost2 FULL
7043 echo "recovery_time=$time_min, timeout=$TIMEOUT, wrap_up=$wrap_up"
7045 mount_client $MOUNT1 || error "mount $MOUNT1 failed"
7046 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
7047 # make sure new superblock labels are sync'd before disabling writes
7051 replay_barrier $SINGLEMDS
7052 createmany -o $DIR1/$tfile-%d 1000
7054 # We need to catch the end of recovery window to extend it.
7055 # Skip 5 requests and add delay to request handling.
7056 #define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP
7057 do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5"
7059 facet_failover --fsck $SINGLEMDS || error "failover: $?"
7062 echo "recovery status"
7063 do_facet $SINGLEMDS \
7064 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status"
7066 recovery_duration=$(do_facet $SINGLEMDS \
7067 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
7068 awk '/recovery_duration/ { print $2 }')
7069 (( $recovery_duration > $time_min + $wrap_up )) &&
7070 error "recovery_duration > recovery_time_hard + wrap up"
7071 completed_clients=$(do_facet $SINGLEMDS \
7072 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
7073 awk '/completed_clients/ { print $2 }')
7075 correct_clients="$MDSCOUNT/$((MDSCOUNT+1))"
7076 [ "$completed_clients" = "${correct_clients}" ] ||
7077 error "$completed_clients != $correct_clients"
7079 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
7080 umount_client $MOUNT1
7081 umount_client $MOUNT2
7087 run_test 84 "check recovery_hard_time"
7090 [[ "$OST1_VERSION" -ge $(version_code 2.7.55) ]] ||
7091 skip "Need OST version at least 2.7.55"
7092 ##define OBD_FAIL_OSD_OST_EA_FID_SET 0x197
7093 do_facet ost1 "lctl set_param fail_loc=0x197"
7097 run_test 85 "osd_ost init: fail ea_fid_set"
7102 # ost1 has already registered to the MGS before the reformat.
7103 # So after reformatting it with option "-G", it could not be
7104 # mounted to the MGS. Cleanup the system for subsequent tests.
7109 [ "$ost1_FSTYPE" = zfs ] &&
7110 skip "LU-6442: no such mkfs params for ZFS OSTs"
7111 [[ "$MDS1_VERSION" -ge $(version_code 2.7.56) ]] ||
7112 skip "Need server version newer than 2.7.55"
7115 local OLDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
7116 awk '/Flex block group size: / { print $NF; exit; }')
7118 [ "$OLDSIZE" == "$NEWSIZE" ] && skip "$NEWSIZE groups already"
7120 local opts=" -O flex_bg -G $NEWSIZE"
7121 opts=$(OST_FS_MKFS_OPTS+="$opts" mkfs_opts ost1 $(ostdevname 1))
7122 opts+=" --reformat $(ostdevname 1) $(ostvdevname 1)"
7123 echo "params: $opts"
7125 trap cleanup_86 EXIT ERR
7128 add ost1 $opts || error "add ost1 failed with new params"
7130 local FOUNDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
7131 awk '/Flex block group size: / { print $NF; exit; }')
7133 [[ $FOUNDSIZE == $NEWSIZE ]] ||
7134 error "Flex block group size: $FOUNDSIZE, expected: $NEWSIZE"
7138 run_test 86 "Replacing mkfs.lustre -G option"
7140 test_87() { #LU-6544
7141 [[ "$MDS1_VERSION" -ge $(version_code 2.9.51) ]] ||
7142 skip "Need MDS version at least 2.9.51"
7143 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
7144 skip "ldiskfs only test"
7145 [[ $OSTCOUNT -gt 59 ]] &&
7146 skip "Ignore wide striping situation"
7147 [ -n "$FILESET" ] && skip "Not functional for FILESET set"
7149 local mdsdev=$(mdsdevname 1)
7150 local mdsvdev=$(mdsvdevname 1)
7151 local file=$DIR/$tfile
7152 local mntpt=$(facet_mntpt $SINGLEMDS)
7153 local used_xattr_blk=0
7154 local inode_size=${1:-1024}
7156 local xtest="trusted.test"
7160 local stripe_cnt=$(($OSTCOUNT + 2))
7162 #Please see ldiskfs_make_lustre() for MDT inode size calculation
7163 if [ $stripe_cnt -gt 16 ]; then
7166 left_size=$(expr $inode_size - \
7169 32 - 40 \* 3 - 32 \* 3 - $stripe_cnt \* 24 - 16 - 3 - \
7171 24 - 18 - $(expr length $tfile) - 16 - 4)
7172 if [ $left_size -le 0 ]; then
7173 echo "No space($left_size) is expected in inode."
7174 echo "Try 1-byte xattr instead to verify this."
7177 echo "Estimate: at most $left_size-byte space left in inode."
7183 if ! combined_mgs_mds ; then
7187 add mds1 $(mkfs_opts mds1 ${mdsdev}) --stripe-count-hint=$stripe_cnt \
7188 --reformat $mdsdev $mdsvdev || error "add mds1 failed"
7189 start_mdt 1 > /dev/null || error "start mdt1 failed"
7190 for i in $(seq $OSTCOUNT); do
7191 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS > /dev/null ||
7192 error "start ost$i failed"
7194 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
7195 check_mount || error "check client $MOUNT failed"
7198 $LFS setstripe -E 1M -S 1M -c 1 -E 64M -c 1 -E -1 -c -1 $file ||
7199 error "Create file with 3 components failed"
7200 $TRUNCATE $file $((1024*1024*64+1)) || error "truncate file failed"
7201 i=$($LFS getstripe -I3 -c $file) || error "get 3rd stripe count failed"
7202 if [ $i -ne $OSTCOUNT ]; then
7203 left_size=$(expr $left_size + $(expr $OSTCOUNT - $i) \* 24)
7204 echo -n "Since only $i out $OSTCOUNT OSTs are used, "
7205 echo -n "the expected left space is changed to "
7206 echo "$left_size bytes at most."
7208 value=$(generate_string $left_size)
7209 setfattr -n $xtest -v $value $file
7210 orig=$(get_xattr_value $xtest $file)
7211 [[ "$orig" != "$value" ]] && error "$xtest changed"
7213 #Verify if inode has some expected space left
7214 umount $MOUNT > /dev/null || error "umount $MOUNT failed"
7215 stop_mdt 1 > /dev/null || error "stop mdt1 failed"
7216 mount_ldiskfs $SINGLEMDS || error "mount -t ldiskfs $SINGLEMDS failed"
7218 do_facet $SINGLEMDS ls -sal $mntpt/ROOT/$tfile
7219 used_xattr_blk=$(do_facet $SINGLEMDS ls -s $mntpt/ROOT/$tfile |
7221 [[ $used_xattr_blk -eq 0 ]] &&
7222 error "Please check MDS inode size calculation: \
7223 more than $left_size-byte space left in inode."
7224 echo "Verified: at most $left_size-byte space left in inode."
7226 unmount_ldiskfs $SINGLEMDS
7228 for i in $(seq $OSTCOUNT); do
7229 stop ost$i -f || error "stop ost$i failed"
7232 run_test 87 "check if MDT inode can hold EAs with N stripes properly"
7235 [ "$mds1_FSTYPE" == zfs ] &&
7236 skip "LU-6662: no implementation for ZFS"
7240 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
7241 --reformat $(mdsdevname 1) || error "add mds1 failed"
7243 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
7244 grep -e \".*opts:.*errors=remount-ro.*\"" ||
7245 error "default mount options is missing"
7247 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
7248 --mountfsoptions="user_xattr,errors=panic" \
7249 --reformat $(mdsdevname 1) || error "add mds1 failed"
7251 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
7252 grep -e \".*opts:.*errors=panic.*\"" ||
7253 error "user can't override default mount options"
7255 run_test 88 "check the default mount options can be overridden"
7257 test_89() { # LU-7131
7258 [[ "$MDS1_VERSION" -ge $(version_code 2.9.54) ]] ||
7259 skip "Need MDT version at least 2.9.54"
7261 local key=failover.node
7262 local val1=192.0.2.254@tcp0 # Reserved IPs, see RFC 5735
7263 local val2=192.0.2.255@tcp0
7264 local mdsdev=$(mdsdevname 1)
7269 if [[ "$mds1_FSTYPE" == zfs ]]; then
7270 import_zpool mds1 || return ${PIPESTATUS[0]}
7273 # Check that parameters are added correctly
7274 echo "tunefs --param $key=$val1"
7275 do_facet mds1 "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
7276 error "tunefs --param $key=$val1 failed"
7277 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7278 error "tunefs --dryrun failed"
7279 params=${params##*Parameters:}
7280 params=${params%%exiting*}
7281 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val1) = "1" ] ||
7282 error "on-disk parameter not added correctly via tunefs"
7284 # Check that parameters replace existing instances when added
7285 echo "tunefs --param $key=$val2"
7286 do_facet mds1 "$TUNEFS --param $key=$val2 $mdsdev >/dev/null" ||
7287 error "tunefs --param $key=$val2 failed"
7288 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7289 error "tunefs --dryrun failed"
7290 params=${params##*Parameters:}
7291 params=${params%%exiting*}
7292 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "1" ] ||
7293 error "on-disk parameter not replaced via tunefs"
7294 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val2) = "1" ] ||
7295 error "on-disk parameter not replaced correctly via tunefs"
7297 # Check that a parameter is erased properly
7298 echo "tunefs --erase-param $key"
7299 do_facet mds1 "$TUNEFS --erase-param $key $mdsdev >/dev/null" ||
7300 error "tunefs --erase-param $key failed"
7301 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7302 error "tunefs --dryrun failed"
7303 params=${params##*Parameters:}
7304 params=${params%%exiting*}
7305 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "0" ] ||
7306 error "on-disk parameter not erased correctly via tunefs"
7308 # Check that all the parameters are erased
7309 do_facet mds1 "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
7310 error "tunefs --param $key=$val1 failed"
7311 echo "tunefs --erase-params"
7312 do_facet mds1 "$TUNEFS --erase-params $mdsdev >/dev/null" ||
7313 error "tunefs --erase-params failed"
7314 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7315 error "tunefs --dryrun failed"
7316 params=${params##*Parameters:}
7317 params=${params%%exiting*}
7318 params=$(echo $params | tr ' ' '\n')
7320 error "all on-disk parameters not erased correctly via tunefs $params"
7322 # Check the order of options --erase-params and --param
7323 echo "tunefs --param $key=$val1 --erase-params"
7325 "$TUNEFS --param $key=$val1 --erase-params $mdsdev >/dev/null"||
7326 error "tunefs --param $key=$val1 --erase-params failed"
7327 params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) ||
7328 error "tunefs --dryrun failed"
7329 params=${params##*Parameters:}
7330 params=${params%%exiting*}
7331 [ $(echo $params | tr ' ' '\n') == "$key=$val1" ] ||
7332 error "on-disk param not added correctly with --erase-params"
7336 run_test 89 "check tunefs --param and --erase-param{s} options"
7339 # $2 (optional) value of max_mod_rpcs_in_flight to set
7340 check_max_mod_rpcs_in_flight() {
7348 idx=$(printf "%04x" $($LFS getdirstripe -i $dir))
7349 facet="mds$((0x$idx + 1))"
7351 if [ -z "$mmr" ]; then
7352 # get value of max_mod_rcps_in_flight
7353 mmr=$($LCTL get_param -n \
7354 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) ||
7355 error "Unable to get max_mod_rpcs_in_flight"
7356 echo "max_mod_rcps_in_flight is $mmr"
7358 # set value of max_mod_rpcs_in_flight
7360 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight=$mmr ||
7361 error "Unable to set max_mod_rpcs_in_flight to $mmr"
7362 echo "max_mod_rpcs_in_flight set to $mmr"
7365 # create mmr+1 files
7366 echo "creating $((mmr + 1)) files ..."
7368 for i in $(seq $((mmr + 1))); do
7374 # consumes mmr-1 modify RPC slots
7375 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7376 # drop requests on MDT so that RPC slots are consumed
7377 # during all the request resend interval
7378 do_facet $facet "$LCTL set_param fail_loc=0x159"
7379 echo "launch $((mmr - 1)) chmod in parallel ..."
7380 for i in $(seq $((mmr - 1))); do
7381 chmod 0600 $dir/file-$i &
7385 # send one additional modify RPC
7386 do_facet $facet "$LCTL set_param fail_loc=0"
7387 echo "launch 1 additional chmod in parallel ..."
7388 chmod 0600 $dir/file-$mmr &
7391 # check this additional modify RPC get a modify RPC slot
7392 # and succeed its operation
7393 checkstat -vp 0600 $dir/file-$mmr ||
7394 error "Unable to send $mmr modify RPCs in parallel"
7399 # consumes mmr modify RPC slots
7400 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7401 # drop requests on MDT so that RPC slots are consumed
7402 # during all the request resend interval
7403 do_facet $facet "$LCTL set_param fail_loc=0x159"
7404 echo "launch $mmr chmod in parallel ..."
7405 for i in $(seq $mmr); do
7406 chmod 0666 $dir/file-$i &
7410 # send one additional modify RPC
7411 do_facet $facet "$LCTL set_param fail_loc=0"
7412 echo "launch 1 additional chmod in parallel ..."
7413 chmod 0666 $dir/file-$((mmr + 1)) &
7416 # check this additional modify RPC blocked getting a modify RPC slot
7417 checkstat -vp 0644 $dir/file-$((mmr + 1)) ||
7418 error "Unexpectedly send $(($mmr + 1)) modify RPCs in parallel"
7422 get_mdt_max_mod_rpcs_in_flight_val() {
7426 # It will be enough to get one value from one MDT
7427 max_mrif=$(do_facet $facet \
7428 "$LCTL get_param -n \
7429 mdt.$FSNAME-MDT*.max_mod_rpcs_in_flight | head -n 1")
7430 if [[ $max_mrif ]]; then
7433 max_mrif=$(do_facet $facet \
7434 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
7439 set_mdt_max_mod_rpcs_in_flight() {
7444 lctl_op=$($LCTL get_param \
7445 mdt.*.max_mod_rpcs_in_flight)
7446 if [[ $lctl_op ]]; then
7449 mdt.$FSNAME-MDT*.max_mod_rpcs_in_flight=$max_mrif"
7453 /sys/module/mdt/parameters/max_mod_rpcs_per_client"
7454 echo "the deprecated max_mod_rpcs_per_client \
7455 parameter was involved"
7462 [[ $($LCTL get_param mdc.*.import |
7463 grep "connect_flags:.*multi_mod_rpc") ]] ||
7464 skip "Need MDC with 'multi_mod_rpcs' feature"
7466 # check default value
7467 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
7468 check_max_mod_rpcs_in_flight $DIR/$tdir
7472 run_test 90a "check max_mod_rpcs_in_flight is enforced"
7482 [[ $($LCTL get_param mdc.*.import |
7483 grep "connect_flags:.*multi_mod_rpc") ]] ||
7484 skip "Need MDC with 'multi_mod_rpcs' feature"
7487 # update max_mod_rpcs_in_flight
7488 $LFS mkdir -c1 $DIR/${tdir}1 || error "mkdir $DIR/${tdir}1 failed"
7489 check_max_mod_rpcs_in_flight $DIR/${tdir}1 1
7492 # check client is able to send multiple modify RPCs in paralell
7493 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
7494 grep -c "multi_mod_rpcs")
7495 if [ "$tmp" -ne $MDSCOUNT ]; then
7497 skip "Client not able to send multiple modify RPCs in parallel"
7500 # update max_mod_rpcs_in_flight
7501 $LFS mkdir -c1 $DIR/${tdir}2 || error "mkdir $DIR/${tdir}2 failed"
7502 check_max_mod_rpcs_in_flight $DIR/${tdir}2 5
7505 $LFS mkdir -c1 $DIR/${tdir}3 || error "mkdir $DIR/${tdir}3 failed"
7506 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/${tdir}3))
7507 facet="mds$((0x$idx + 1))"
7509 mmrpc=$(get_mdt_max_mod_rpcs_in_flight_val $facet)
7510 echo "mdt_max_mod_rpcs_in_flight is $mmrpc"
7511 # update max_mod_rpcs_in_flight
7512 umount_client $MOUNT
7513 set_mdt_max_mod_rpcs_in_flight 16 $facet
7515 $LCTL set_param mdc.$FSNAME-MDT$idx-mdc-*.max_rpcs_in_flight=17
7516 check_max_mod_rpcs_in_flight $DIR/${tdir}3 16
7518 # restore MDT max_mod_rpcs_in_flight initial value
7519 set_mdt_max_mod_rpcs_in_flight $mmrpc $facet
7521 rm -rf $DIR/${tdir}?
7524 run_test 90b "check max_mod_rpcs_in_flight is enforced after update"
7527 # get max_rpcs_in_flight value
7528 mrif_90c=$($LCTL get_param -n \
7529 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
7530 echo "max_rpcs_in_flight is $mrif_90c"
7532 # get MDC max_mod_rpcs_in_flight value
7533 mmrif_90c=$($LCTL get_param -n \
7534 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight)
7535 echo "MDC max_mod_rpcs_in_flight is $mmrif_90c"
7537 # get MDT max_mod_rpcs_in_flight value
7538 mmrpc_90c=$(get_mdt_max_mod_rpcs_in_flight_val "mds1")
7539 echo "mdt_max_mod_rpcs_in_flight is $mmrpc_90c"
7542 restore_params_90c() {
7545 # restore max_rpcs_in_flight value
7546 do_facet mgs $LCTL set_param -P \
7547 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$mrif_90c
7549 # restore max_mod_rpcs_in_flight value
7550 do_facet mgs $LCTL set_param -P \
7551 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mmrif_90c
7553 # restore MDT max_mod_rpcs_in_flight value
7554 set_mdt_max_mod_rpcs_in_flight $mmrpc_90c "mds1"
7562 [[ $($LCTL get_param mdc.*.import |
7563 grep "connect_flags:.*multi_mod_rpc") ]] ||
7564 skip "Need MDC with 'multi_mod_rpcs' feature"
7566 # check client is able to send multiple modify RPCs in paralell
7567 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
7568 grep -c "multi_mod_rpcs")
7569 if [ "$tmp" -ne $MDSCOUNT ]; then
7571 skip "Client not able to send multiple modify RPCs in parallel"
7575 stack_trap restore_params_90c
7578 # attempt to set max_mod_rpcs_in_flight to max_rpcs_in_flight value
7579 # prerequisite: set MDT max_mod_rpcs_in_flight to
7580 # max_rpcs_in_flight value
7581 set_mdt_max_mod_rpcs_in_flight $mrif_90c "mds1"
7583 # if max_mod_rpcs_in_flight is set to be equal to or larger than
7584 # max_rpcs_in_flight, then max_rpcs_in_flight will be increased
7585 if [[ "$CLIENT_VERSION" -ge $(version_code 2.13.53) ]]; then
7587 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif_90c ||
7588 error "set max_mod_rpcs_in_flight to $mrif_90c failed"
7590 local new_mrif=$($LCTL get_param -n \
7591 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
7592 ((new_mrif == mrif_90c + 1)) ||
7593 error "max_rpcs_in_flight was not increased"
7596 umount_client $MOUNT
7597 set_mdt_max_mod_rpcs_in_flight $mmrpc_90c "mds1"
7601 # attempt to set max_mod_rpcs_in_flight to MDT max_mod_rpcs_in_flight+1
7602 # prerequisite: set max_rpcs_in_flight to MDT max_mod_rpcs_in_flight+2
7604 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$((mmrpc_90c + 2))
7607 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$((mmrpc_90c + 1)) &&
7608 error "set max_mod_rpcs_in_flight to $((mmrpc_90c + 1)) should fail"
7611 # attempt to set max_mod_rpcs_in_flight permanently
7612 do_facet mgs $LCTL set_param -P \
7613 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$mrif_90c
7615 do_facet mgs $LCTL set_param -P \
7616 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif_90c
7618 remount_client $MOUNT
7620 wait_update_facet --verbose client "$LCTL get_param -n \
7621 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight" \
7622 "$((mrif_90c + 1))" ||
7623 error "expected '$((mrif_90c + 1))' for max_rpcs_in_flight"
7625 wait_update_facet --verbose client "$LCTL get_param -n \
7626 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight" \
7628 error "expected '$mrif_90c' for max_mod_rpcs_in_flight"
7633 run_test 90c "check max_mod_rpcs_in_flight update limits"
7644 [[ $($LCTL get_param mdc.*.import |
7645 grep "connect_flags:.*multi_mod_rpc") ]] ||
7646 skip "Need MDC with 'multi_mod_rpcs' feature"
7648 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
7649 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
7650 facet="mds$((0x$idx + 1))"
7652 # check client version supports multislots
7653 tmp=$($LCTL get_param -N \
7654 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
7655 if [ -z "$tmp" ]; then
7657 skip "Client does not support multiple modify RPCs in flight"
7660 # get current value of max_mod_rcps_in_flight
7661 mmr=$($LCTL get_param -n \
7662 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
7663 echo "max_mod_rcps_in_flight is $mmr"
7666 echo "creating $mmr files ..."
7668 for i in $(seq $mmr); do
7669 touch $DIR/$tdir/file-$i
7672 # prepare for close RPC
7673 multiop_bg_pause $DIR/$tdir/file-close O_c
7676 # consumes mmr modify RPC slots
7677 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
7678 # drop requests on MDT so that RPC slots are consumed
7679 # during all the request resend interval
7680 do_facet $facet "$LCTL set_param fail_loc=0x159"
7681 echo "launch $mmr chmod in parallel ..."
7682 for i in $(seq $mmr); do
7683 chmod 0600 $DIR/$tdir/file-$i &
7686 # send one additional close RPC
7687 do_facet $facet "$LCTL set_param fail_loc=0"
7688 echo "launch 1 additional close in parallel ..."
7690 cancel_lru_locks mdc
7693 # check this additional close RPC get a modify RPC slot
7694 # and multiop process completed
7695 [ -d /proc/$pid ] &&
7696 error "Unable to send the additional close RPC in parallel"
7701 run_test 90d "check one close RPC is allowed above max_mod_rpcs_in_flight"
7703 check_uuid_on_ost() {
7705 do_facet ost1 "$LCTL get_param obdfilter.${FSNAME}*.exports.'$nid'.uuid"
7708 check_uuid_on_mdt() {
7710 do_facet $SINGLEMDS "$LCTL get_param mdt.${mds1_svc}*.exports.'$nid'.uuid"
7718 [[ "$OST1_VERSION" -ge $(version_code 2.7.63) ]] ||
7719 skip "Need OST version at least 2.7.63"
7720 [[ "$MDS1_VERSION" -ge $(version_code 2.7.63) ]] ||
7721 skip "Need MDT version at least 2.7.63"
7723 start_mds || error "MDS start failed"
7724 start_ost || error "unable to start OST"
7725 mount_client $MOUNT || error "client start failed"
7726 check_mount || error "check_mount failed"
7729 nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g")
7733 uuid=$(get_client_uuid $MOUNT)
7735 echo "list nids on mdt:"
7736 do_facet $SINGLEMDS "$LCTL list_param mdt.${FSNAME}*.exports.*"
7737 echo "uuid from $nid:"
7738 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
7740 found=$(check_uuid_on_mdt $nid | grep $uuid)
7741 [ -z "$found" ] && error "can't find $uuid $nid on MDT"
7742 found=$(check_uuid_on_ost $nid | grep $uuid)
7743 [ -z "$found" ] && error "can't find $uuid $nid on OST"
7745 # umount the client so it won't reconnect
7746 manual_umount_client --force || error "failed to umount $?"
7747 # shouldn't disappear on MDS after forced umount
7748 found=$(check_uuid_on_mdt $nid | grep $uuid)
7749 [ -z "$found" ] && error "can't find $uuid $nid"
7752 do_facet $SINGLEMDS \
7753 "$LCTL set_param -n mdt.${mds1_svc}.evict_client nid:$nid"
7754 sleep 1 # eviction above is async, give it some time to proceed
7756 found=$(check_uuid_on_mdt $nid | grep $uuid)
7757 [ -n "$found" ] && error "found $uuid $nid on MDT"
7758 found=$(check_uuid_on_ost $nid | grep $uuid)
7759 [ -n "$found" ] && error "found $uuid $nid on OST"
7761 # check it didn't reconnect (being umounted)
7762 sleep $((TIMEOUT+1))
7763 found=$(check_uuid_on_mdt $nid | grep $uuid)
7764 [ -n "$found" ] && error "found $uuid $nid on MDT"
7765 found=$(check_uuid_on_ost $nid | grep $uuid)
7766 [ -n "$found" ] && error "found $uuid $nid on OST"
7770 run_test 91 "evict-by-nid support"
7772 generate_ldev_conf() {
7773 # generate an ldev.conf file
7774 local ldevconfpath=$1
7776 local fsldevformat=""
7779 local facets="mgs,$(get_facets OST),$(get_facets MDS)"
7780 for facet in ${facets//,/ }; do
7782 fstype=$(facet_fstype $facet)
7783 [ "$fstype" = zfs ] && fsldevformat="$fstype:"
7784 local host=$(facet_host $facet)
7786 local varfo=${facet}failover_HOST
7787 if [ $facet == mgs ] && combined_mgs_mds; then
7788 varfo=mds1failover_HOST
7790 [ -n "${!varfo}" ] && fo=${!varfo}
7791 local type=$(echo $facet | tr -d "[:digit:]" | \
7792 tr "[:lower:]" "[:upper:]" | sed s/MDS/MDT/ )
7794 [ ${facet} == mgs ] ||
7795 num=$(facet_number $facet)
7796 printf "%s\t%s\t%s-%s%04x\t%s%s\n" \
7803 $(facet_device $facet) >> $ldevconfpath
7806 echo "----- $ldevconfpath -----"
7808 echo "--- END $ldevconfpath ---"
7813 # generate a nids file (mapping between hostname to nid)
7814 # looks like we only have the MGS nid available to us
7815 # so just echo that to a file
7817 echo -e "${mgs_HOST}\t${MGSNID}" > $nidspath
7819 echo "----- $nidspath -----"
7821 echo "--- END $nidspath ---"
7824 compare_ldev_output() {
7825 local ldev_output=$1
7826 local expected_output=$2
7828 sort $expected_output -o $expected_output
7829 sort $ldev_output -o $ldev_output
7831 echo "-- START OF LDEV OUTPUT --"
7833 echo "--- END OF LDEV OUTPUT ---"
7835 echo "-- START OF EXPECTED OUTPUT --"
7836 cat $expected_output
7837 echo "--- END OF EXPECTED OUTPUT ---"
7839 diff $expected_output $ldev_output
7844 if [ -z "$LDEV" ]; then
7845 error "ldev is missing!"
7848 local LDEVCONFPATH=$TMP/ldev.conf
7849 local NIDSPATH=$TMP/nids
7851 echo "Host is $(hostname)"
7853 generate_ldev_conf $LDEVCONFPATH
7854 generate_nids $NIDSPATH
7856 # echo the mgs nid and compare it to environment variable MGSNID
7857 local facets="$(get_facets OST),$(get_facets MDS),mgs"
7858 for facet in ${facets//,/ }; do
7859 local host=$(facet_host $facet)
7860 local output=$($LDEV -c $LDEVCONFPATH -H $host -n $NIDSPATH echo %m)
7862 echo "-- START OF LDEV OUTPUT --"
7864 echo "--- END OF LDEV OUTPUT ---"
7867 error "ldev failed to execute!"
7869 # need to process multiple lines because of
7870 # several targets on host
7871 echo -e $output | awk '{ print $2 }' | while read -r line ; do
7872 [ "$line" = "$MGSNID" ] ||
7873 error "ldev failed mgs nid '$line', \
7877 rm -f $LDEVCONFPATH $NIDSPATH
7879 run_test 92 "ldev returns MGS NID correctly in command substitution"
7882 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs"
7885 #start mgs or mgs/mdt0
7886 if ! combined_mgs_mds ; then
7893 start_ost || error "OST0 start fail"
7895 #define OBD_FAIL_MGS_WRITE_TARGET_DELAY 0x90e
7896 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x8000090e"
7897 for num in $(seq 2 $MDSCOUNT); do
7901 mount_client $MOUNT || error "mount client fails"
7902 wait_osc_import_state mds ost FULL
7903 wait_osc_import_ready client ost
7904 check_mount || error "check_mount failed"
7906 cleanup || error "cleanup failed with $?"
7908 run_test 93 "register mulitple MDT at the same time"
7911 if [ -z "$LDEV" ]; then
7912 error "ldev is missing!"
7915 local LDEVCONFPATH=$TMP/ldev.conf
7916 local NIDSPATH=$TMP/nids
7918 generate_ldev_conf $LDEVCONFPATH
7919 generate_nids $NIDSPATH
7921 local LDEV_OUTPUT=$TMP/ldev-output.txt
7922 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME > $LDEV_OUTPUT
7924 # ldev failed, error
7925 if [ $? -ne 0 ]; then
7926 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7927 error "ldev failed to execute!"
7931 local EXPECTED_OUTPUT=$TMP/ldev-expected.txt
7933 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
7935 for num in $(seq $MDSCOUNT); do
7936 printf "%s-MDT%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7939 for num in $(seq $OSTCOUNT); do
7940 printf "%s-OST%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
7943 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7945 if [ $? -ne 0 ]; then
7946 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7947 error "ldev failed to produce the correct hostlist!"
7950 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7952 run_test 94 "ldev outputs correct labels for file system name query"
7955 if [ -z "$LDEV" ]; then
7956 error "ldev is missing!"
7959 local LDEVCONFPATH=$TMP/ldev.conf
7960 local NIDSPATH=$TMP/nids
7962 generate_ldev_conf $LDEVCONFPATH
7963 generate_nids $NIDSPATH
7967 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME &>/dev/null
7968 if [ $? -ne 0 ]; then
7969 rm $LDEVCONFPATH $NIDSPATH
7970 error "ldev label filtering w/ -F failed!"
7974 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -l &>/dev/null
7975 if [ $? -ne 0 ]; then
7976 rm $LDEVCONFPATH $NIDSPATH
7977 error "ldev label filtering w/ -l failed!"
7981 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f &>/dev/null
7982 if [ $? -ne 0 ]; then
7983 rm $LDEVCONFPATH $NIDSPATH
7984 error "ldev label filtering w/ -f failed!"
7988 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a &>/dev/null
7989 if [ $? -ne 0 ]; then
7990 rm $LDEVCONFPATH $NIDSPATH
7991 error "ldev label filtering w/ -a failed!"
7996 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -F $FSNAME &>/dev/null
7997 if [ $? -eq 0 ]; then
7998 rm $LDEVCONFPATH $NIDSPATH
7999 error "ldev label filtering w/ -a and -F incorrectly succeeded"
8003 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -f &>/dev/null
8004 if [ $? -eq 0 ]; then
8005 rm $LDEVCONFPATH $NIDSPATH
8006 error "ldev label filtering w/ -a and -f incorrectly succeeded"
8010 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -l &>/dev/null
8011 if [ $? -eq 0 ]; then
8012 rm $LDEVCONFPATH $NIDSPATH
8013 error "ldev label filtering w/ -a and -l incorrectly succeeded"
8017 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f -l &>/dev/null
8018 if [ $? -eq 0 ]; then
8019 rm $LDEVCONFPATH $NIDSPATH
8020 error "ldev label filtering w/ -f and -l incorrectly succeeded"
8024 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -l &>/dev/null
8025 if [ $? -eq 0 ]; then
8026 rm $LDEVCONFPATH $NIDSPATH
8027 error "ldev label filtering w/ -F and -l incorrectly succeeded"
8030 # file sys & foreign
8031 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -f &>/dev/null
8032 if [ $? -eq 0 ]; then
8033 rm $LDEVCONFPATH $NIDSPATH
8034 error "ldev label filtering w/ -F and -f incorrectly succeeded"
8037 rm $LDEVCONFPATH $NIDSPATH
8039 run_test 95 "ldev should only allow one label filter"
8042 if [ -z "$LDEV" ]; then
8043 error "ldev is missing!"
8046 local LDEVCONFPATH=$TMP/ldev.conf
8047 local NIDSPATH=$TMP/nids
8049 generate_ldev_conf $LDEVCONFPATH
8050 generate_nids $NIDSPATH
8052 local LDEV_OUTPUT=$TMP/ldev-output.txt
8053 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -H $mgs_HOST \
8055 awk '{print $2}' > $LDEV_OUTPUT
8057 # ldev failed, error
8058 if [ $? -ne 0 ]; then
8059 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
8060 error "ldev failed to execute!"
8064 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
8066 echo "$mgs_HOST-$(facet_fstype mgs)" > $EXPECTED_OUTPUT
8068 local facets="$(get_facets OST),$(get_facets MDS)"
8069 for facet in ${facets//,/ }; do
8070 local host=$(facet_host $facet)
8071 [ "$mgs_HOST" == "$host" ] &&
8072 echo "$host-$(facet_fstype $facet)" \
8076 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
8078 if [ $? -ne 0 ]; then
8079 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8080 error "ldev failed to produce the correct output!"
8083 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8085 run_test 96 "ldev returns hostname and backend fs correctly in command sub"
8088 if [ -z "$LDEV" ]; then
8089 error "ldev is missing!"
8092 local LDEVCONFPATH=$TMP/ldev.conf
8093 local NIDSPATH=$TMP/nids
8095 generate_ldev_conf $LDEVCONFPATH
8096 generate_nids $NIDSPATH
8098 local LDEV_OUTPUT=$TMP/ldev-output.txt
8099 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
8101 echo -e "\nMDT role"
8102 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mdt > $LDEV_OUTPUT
8104 if [ $? -ne 0 ]; then
8105 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
8106 error "ldev failed to execute for mdt role!"
8109 for num in $(seq $MDSCOUNT); do
8110 printf "%s-MDT%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
8113 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
8115 if [ $? -ne 0 ]; then
8116 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8117 error "ldev failed to produce the correct output for mdt role!"
8120 echo -e "\nOST role"
8121 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R ost > $LDEV_OUTPUT
8123 if [ $? -ne 0 ]; then
8124 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
8125 error "ldev failed to execute for ost role!"
8129 for num in $(seq $OSTCOUNT); do
8130 printf "%s-OST%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT
8133 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
8135 if [ $? -ne 0 ]; then
8136 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8137 error "ldev failed to produce the correct output for ost role!"
8140 echo -e "\nMGS role"
8141 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mgs > $LDEV_OUTPUT
8143 if [ $? -ne 0 ]; then
8144 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
8145 error "ldev failed to execute for mgs role!"
8148 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
8150 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
8152 if [ $? -ne 0 ]; then
8153 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8154 error "ldev failed to produce the correct output for mgs role!"
8157 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
8159 run_test 97 "ldev returns correct ouput when querying based on role"
8166 check_mount || error "mount failed"
8167 mountopt="user_xattr"
8168 for ((x = 1; x <= PAGE_SIZE/11; x++)); do
8169 mountopt="$mountopt,user_xattr"
8171 mount_client $MOUNT remount,$mountopt 2>&1 | grep "too long" ||
8172 error "Buffer overflow check failed"
8173 cleanup || error "cleanup failed"
8175 run_test 98 "Buffer-overflow check while parsing mount_opts"
8179 [[ "$ost1_FSTYPE" != ldiskfs ]] &&
8180 skip "ldiskfs only test"
8181 [[ "$OST1_VERSION" -ge $(version_code 2.8.57) ]] ||
8182 skip "Need OST version at least 2.8.57"
8184 local ost_opts="$(mkfs_opts ost1 $(ostdevname 1)) \
8185 --reformat $(ostdevname 1) $(ostvdevname 1)"
8186 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" &&
8187 skip "meta_bg already set"
8190 if [[ ${!opts} != *mkfsoptions* ]]; then
8191 eval opts=\"${!opts} \
8192 --mkfsoptions='\\\"-O ^resize_inode,meta_bg\\\"'\"
8194 local val=${!opts//--mkfsoptions=\\\"/ \
8195 --mkfsoptions=\\\"-O ^resize_inode,meta_bg }
8199 echo "params: $opts"
8202 add ost1 $opts || error "add ost1 failed with new params"
8204 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" ||
8205 error "meta_bg is not set"
8207 run_test 99 "Adding meta_bg option"
8212 start_mgsmds || error "MDS start failed"
8213 start_ost || error "unable to start OST"
8214 mount_client $MOUNT || error "client start failed"
8215 check_mount || error "check_mount failed"
8224 do_facet mgs 'lshowmount -v' | awk 'BEGIN {NR == 0; rc=1} /MGS:/ {rc=0}
8225 END {exit rc}' || error "lshowmount have no output MGS"
8227 do_facet mds1 'lshowmount -v' | awk 'BEGIN {NR == 2; rc=1} /-MDT0000:/
8228 {rc=0} END {exit rc}' || error "lshowmount have no output MDT0"
8230 do_facet ost1 'lshowmount -v' | awk 'BEGIN {NR == 4; rc=1} /-OST0000:/
8231 {rc=0} END {exit rc}' || error "lshowmount have no output OST0"
8233 cleanup || error "cleanup failed with $?"
8235 run_test 100 "check lshowmount lists MGS, MDT, OST and 0@lo"
8238 local createmany_pid
8239 local dev=$FSNAME-OST0000-osc-MDT0000
8243 do_nodes $(comma_list $(osts_nodes)) $LCTL set_param \
8244 seq.*OST*-super.width=$DATA_SEQ_MAX_WIDTH
8245 createmany -o $DIR1/$tdir/$tfile-%d 50000 &
8247 # MDT->OST reconnection causes MDT<->OST last_id synchornisation
8248 # via osp_precreate_cleanup_orphans.
8249 for ((i = 0; i < 100; i++)); do
8250 for ((k = 0; k < 10; k++)); do
8251 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;" \
8252 "$LCTL --device $dev activate"
8255 ls -asl $MOUNT | grep '???' &&
8256 { kill -9 $createmany_pid &>/dev/null;
8257 error "File has no object on OST"; }
8259 kill -s 0 $createmany_pid || break
8261 wait $createmany_pid
8263 unlinkmany $DIR1/$tdir/$tfile-%d 50000
8266 run_test 101a "Race MDT->OST reconnection with create"
8269 local dev=$FSNAME-OST0000-osc-MDT0000
8270 local dir=$DIR1/$tdir
8274 $LFS setstripe -c 1 -i 0 $dir
8275 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;"
8276 #define OBD_FAIL_OSP_CON_EVENT_DELAY 0x2107
8277 do_facet mds1 "$LCTL set_param fail_loc=0x80002107 fail_val=20"
8278 do_facet $SINGLEMDS "$LCTL --device $dev activate;"
8283 wait_osc_import_ready client ost1
8284 touch $dir/$tfile || error "Can't create file"
8288 run_test 101b "Race events DISCONNECT and ACTIVE in osp"
8291 [[ "$MDS1_VERSION" -gt $(version_code 2.9.53) ]] ||
8292 skip "Need server version greater than 2.9.53"
8293 [[ “$(mdsdevname 1)” != “$(mgsdevname)” ]] &&
8294 [[ “$(facet_host mds1)” = “$(facet_host mgs)” ]] &&
8295 skip "MGS must be on different node or combined"
8297 cleanup || error "cleanup failed with $?"
8299 local mds1dev=$(mdsdevname 1)
8300 local mds1mnt=$(facet_mntpt mds1)
8301 local mds1opts=$MDS_MOUNT_OPTS
8303 if [ "$mds1_FSTYPE" == ldiskfs ] &&
8304 ! do_facet mds1 test -b $mds1dev; then
8305 mds1opts=$(csa_add "$mds1opts" -o loop)
8307 if [[ "$mds1_FSTYPE" == zfs ]]; then
8308 import_zpool mds1 || return ${PIPESTATUS[0]}
8311 # unload all and only load libcfs to allow fail_loc setting
8312 do_facet mds1 $LUSTRE_RMMOD || error "unable to unload modules"
8313 do_facet mds1 modprobe libcfs || error "libcfs not loaded"
8314 do_facet mds1 lsmod \| grep libcfs || error "libcfs not loaded"
8316 #define OBD_FAIL_OBDCLASS_MODULE_LOAD 0x60a
8317 do_facet mds1 "$LCTL set_param fail_loc=0x8000060a"
8319 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts &&
8320 error "mdt start must fail"
8321 do_facet mds1 lsmod \| grep obdclass && error "obdclass must not load"
8323 do_facet mds1 "$LCTL set_param fail_loc=0x0"
8325 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts ||
8326 error "mdt start must not fail"
8328 cleanup || error "cleanup failed with $?"
8330 run_test 102 "obdclass module cleanup upon error"
8335 echo "rename $FSNAME to $newname"
8337 if ! combined_mgs_mds ; then
8338 local dev=$(mgsdevname)
8341 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
8342 error "(7) Fail to rename MGS"
8343 if [ "$(facet_fstype mgs)" = "zfs" ]; then
8344 reimport_zpool mgs $newname-mgs
8348 for num in $(seq $MDSCOUNT); do
8349 local dev=$(mdsdevname $num)
8351 do_facet mds${num} \
8352 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
8353 error "(8) Fail to rename MDT $num"
8354 if [ "$(facet_fstype mds${num})" = "zfs" ]; then
8355 reimport_zpool mds${num} $newname-mdt${num}
8359 for num in $(seq $OSTCOUNT); do
8360 local dev=$(ostdevname $num)
8362 do_facet ost${num} \
8363 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $dev" ||
8364 error "(9) Fail to rename OST $num"
8365 if [ "$(facet_fstype ost${num})" = "zfs" ]; then
8366 reimport_zpool ost${num} $newname-ost${num}
8371 test_103_set_pool() {
8375 do_facet mgs $LCTL pool_add $FSNAME.$pname ${FSNAME}-$ost_x ||
8376 error "Fail to add $ost_x to $FSNAME.$pname"
8377 wait_update $HOSTNAME \
8378 "lctl get_param -n lov.$FSNAME-clilov-*.pools.$pname |
8379 grep $ost_x" "$FSNAME-${ost_x}_UUID" ||
8380 error "$ost_x is NOT in pool $FSNAME.$pname"
8383 test_103_check_pool() {
8384 local save_fsname=$1
8387 stat $DIR/$tdir/test-framework.sh ||
8388 error "($errno) Fail to stat"
8389 do_facet mgs $LCTL pool_list $FSNAME.pool1 ||
8390 error "($errno) Fail to list $FSNAME.pool1"
8391 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname ||
8392 error "($errno) Fail to list $FSNAME.$save_fsname"
8393 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname |
8394 grep ${FSNAME}-OST0000 ||
8395 error "($errno) List $FSNAME.$save_fsname is invalid"
8397 local pname=$($LFS getstripe --pool $DIR/$tdir/d0)
8398 [ "$pname" = "$save_fsname" ] ||
8399 error "($errno) Unexpected pool name $pname"
8403 check_mount_and_prep
8405 mkdir $DIR/$tdir || error "(1) Fail to mkdir $DIR/$tdir"
8406 cp $LUSTRE/tests/test-framework.sh $DIR/$tdir ||
8407 error "(2) Fail to copy test-framework.sh"
8409 do_facet mgs $LCTL pool_new $FSNAME.pool1 ||
8410 error "(3) Fail to create $FSNAME.pool1"
8411 # name the pool name as the fsname
8412 do_facet mgs $LCTL pool_new $FSNAME.$FSNAME ||
8413 error "(4) Fail to create $FSNAME.$FSNAME"
8415 test_103_set_pool $FSNAME OST0000
8417 $LFS setstripe -p $FSNAME $DIR/$tdir/d0 ||
8418 error "(6) Fail to setstripe on $DIR/$tdir/d0"
8423 test_renamefs mylustre
8425 local save_fsname=$FSNAME
8429 test_103_check_pool $save_fsname 7
8431 if [ $OSTCOUNT -ge 2 ]; then
8432 test_103_set_pool $save_fsname OST0001
8435 $LFS setstripe -p $save_fsname $DIR/$tdir/f0 ||
8436 error "(16) Fail to setstripe on $DIR/$tdir/f0"
8445 test_103_check_pool $save_fsname 17
8449 test_renamefs $save_fsname
8455 run_test 103 "rename filesystem name"
8457 test_104a() { # LU-6952
8458 local mds_mountopts=$MDS_MOUNT_OPTS
8459 local ost_mountopts=$OST_MOUNT_OPTS
8460 local mds_mountfsopts=$MDS_MOUNT_FS_OPTS
8461 local lctl_ver=$(do_facet $SINGLEMDS $LCTL --version |
8464 [[ $(version_code $lctl_ver) -lt $(version_code 2.9.55) ]] &&
8465 skip "this test needs utils above 2.9.55"
8467 # specify "acl" in mount options used by mkfs.lustre
8468 if [ -z "$MDS_MOUNT_FS_OPTS" ]; then
8469 MDS_MOUNT_FS_OPTS="acl,user_xattr"
8472 MDS_MOUNT_FS_OPTS="${MDS_MOUNT_FS_OPTS},acl,user_xattr"
8475 echo "mountfsopt: $MDS_MOUNT_FS_OPTS"
8477 #reformat/remount the MDT to apply the MDT_MOUNT_FS_OPT options
8479 if ! combined_mgs_mds ; then
8483 if [ -z "$MDS_MOUNT_OPTS" ]; then
8484 MDS_MOUNT_OPTS="-o noacl"
8486 MDS_MOUNT_OPTS="${MDS_MOUNT_OPTS},noacl"
8489 for num in $(seq $MDSCOUNT); do
8490 start mds$num $(mdsdevname $num) $MDS_MOUNT_OPTS ||
8491 error "Failed to start MDS"
8494 for num in $(seq $OSTCOUNT); do
8495 start ost$num $(ostdevname $num) $OST_MOUNT_OPTS ||
8496 error "Failed to start OST"
8500 setfacl -m "d:$RUNAS_ID:rwx" $MOUNT &&
8501 error "ACL is applied when FS is mounted with noacl."
8503 MDS_MOUNT_OPTS=$mds_mountopts
8504 OST_MOUNT_OPTS=$ost_mountopts
8505 MDS_MOUNT_FS_OPTS=$mds_mountfsopts
8507 run_test 104a "Make sure user defined options are reflected in mount"
8509 test_104b() { # LU-12859
8510 mount_client $MOUNT3 flock,localflock
8511 stack_trap "umount_client $MOUNT3" EXIT
8512 mount | grep "$MOUNT3 .*,flock" && error "flock is still set"
8513 mount | grep "$MOUNT3 .*,localflock" || error "localflock is not set"
8514 umount_client $MOUNT3
8515 mount_client $MOUNT3 localflock,flock
8516 mount | grep "$MOUNT3 .*,localflock" && error "localflock is still set"
8517 mount | grep "$MOUNT3 .*,flock" || error "flock is not set"
8518 umount_client $MOUNT3
8519 mount_client $MOUNT3 localflock,flock,noflock
8520 flock_is_enabled $MOUNT3 && error "some flock is still enabled" || true
8522 run_test 104b "Mount uses last flock argument"
8524 error_and_umount() {
8535 mount --bind $DIR $TMP/$tdir || error "mount bind mnt pt failed"
8536 rm -f $TMP/$tdir/$tfile
8537 rm -f $TMP/$tdir/${tfile}1
8539 # Files should not be created in ro bind mount point
8540 # remounting from rw to ro
8541 mount -o remount,ro $TMP/$tdir ||
8542 error_and_umount "readonly remount of bind mnt pt failed"
8543 touch $TMP/$tdir/$tfile &&
8544 error_and_umount "touch succeeds on ro bind mnt pt"
8545 [ -e $TMP/$tdir/$tfile ] &&
8546 error_and_umount "file created on ro bind mnt pt"
8548 # Files should be created in rw bind mount point
8549 # remounting from ro to rw
8550 mount -o remount,rw $TMP/$tdir ||
8551 error_and_umount "read-write remount of bind mnt pt failed"
8552 touch $TMP/$tdir/${tfile}1 ||
8553 error_and_umount "touch fails on rw bind mnt pt"
8554 [ -e $TMP/$tdir/${tfile}1 ] ||
8555 error_and_umount "file not created on rw bind mnt pt"
8556 umount $TMP/$tdir || error "umount of bind mnt pt failed"
8558 cleanup || error "cleanup failed with $?"
8560 run_test 105 "check file creation for ro and rw bind mnt pt"
8564 local creates=64768 # one full plain llog
8566 # ensure there are enough inodes in the filesystem
8567 (( OSTSIZE < (creates + 1024) * 8)) && OSTSIZE=$(((creates + 1024) * 8))
8571 mkdir -p $DIR/$tdir || error "create $tdir failed"
8572 do_nodes $(comma_list $(osts_nodes)) $LCTL set_param \
8573 seq.*OST*-super.width=$DATA_SEQ_MAX_WIDTH
8574 lfs setstripe -c 1 -i 0 $DIR/$tdir
8575 #define OBD_FAIL_CAT_RECORDS 0x1312
8576 do_facet mds1 $LCTL set_param fail_loc=0x1312 fail_val=$repeat
8578 for ((i = 1; i <= $repeat; i++)); do
8579 createmany -o $DIR/$tdir/f- $creates || lfs df -i
8580 createmany -u $DIR/$tdir/f- $creates
8581 wait_delete_completed $((TIMEOUT * 7))
8583 #ASSERTION osp_sync_thread() ( thread->t_flags != SVC_RUNNING ) failed
8584 #shows that osp code is buggy
8585 do_facet mds1 $LCTL set_param fail_loc=0 fail_val=0
8589 run_test 106 "check osp llog processing when catalog is wrapped"
8592 [[ "$MDS1_VERSION" -ge $(version_code 2.10.50) ]] ||
8593 skip "Need MDS version > 2.10.50"
8596 start_mgsmds || error "start_mgsmds failed"
8597 start_ost || error "unable to start OST"
8599 # add unknown configuration parameter.
8600 if [[ $PERM_CMD == *"set_param -P"* ]]; then
8601 cmd="$PERM_CMD ost.$FSNAME-OST0000*.unknown_param"
8603 cmd="$PERM_CMD $FSNAME-OST0000*.ost.unknown_param"
8605 do_facet mgs "$cmd=50"
8606 cleanup_nocli || error "cleanup_nocli failed with $?"
8609 # unknown param should be ignored while mounting.
8610 start_ost || error "unable to start OST after unknown param set"
8612 cleanup || error "cleanup failed with $?"
8614 run_test 107 "Unknown config param should not fail target mounting"
8619 $rcmd rm -rf $tmp > /dev/null 2>&1
8620 $rcmd mkdir -p $tmp/{mnt,images} || error "failed to mkdir remotely"
8622 for facet in $facets; do
8623 [ "$mds1_FSTYPE" = zfs ] &&
8624 $rcmd $ZPOOL -f export lustre-$facet > /dev/null 2>&1
8625 $rcmd mkdir $tmp/mnt/$facet ||
8626 error "failed to mkdir $tmp/mnt/$facet"
8627 $rcmd dd if=/dev/zero of=$tmp/images/$facet \
8628 seek=199 bs=1M count=1 ||
8629 error "failed to create $tmp/images/$facet"
8638 local facet=${role}$((idx + 1))
8640 [ $# -eq 5 ] && pool=$5
8642 do_facet $SINGLEMDS $MKFS --fsname=lustre --$mgs \
8643 --$role --index=$idx --replace --backfstype=$bkfs \
8644 --device-size=200000 --reformat $pool $tmp/images/$facet ||
8645 error "failed to mkfs for $facet"
8649 echo "mounting client..."
8650 mount -t lustre ${nid}:/lustre $MOUNT ||
8651 error "failed to mount lustre"
8654 ls -l $MOUNT/local_dir || error "failed to list"
8656 echo "check truncate && write"
8657 echo "dummmmmmmmmmmmm" > $MOUNT/remote_dir/fsx.c ||
8658 error "failed to tuncate & write"
8661 touch $MOUNT/foooo ||
8662 error "failed to create"
8664 echo "check read && write && append"
8665 sha1sum $MOUNT/conf-sanity.sh |
8666 awk '{ print $1 }' > $MOUNT/checksum.new ||
8667 error "failed to read(1)"
8668 sha1sum $MOUNT/remote_dir/unlinkmany.c |
8669 awk '{ print $1 }' >> $MOUNT/checksum.new ||
8670 error "failed to read(2)"
8671 sha1sum $MOUNT/striped_dir/lockahead_test.o |
8672 awk '{ print $1 }' >> $MOUNT/checksum.new ||
8673 error "failed to read(3)"
8676 diff $MOUNT/checksum.new $MOUNT/checksum.src ||
8677 error "failed to verify data"
8687 umount -f $MOUNT || error "failed to umount client"
8688 for facet in $facets; do
8689 $rcmd umount -f $tmp/mnt/$facet ||
8690 error "failed to umount $facet"
8691 if [ "$mds1_FSTYPE" = zfs ]; then
8692 $rcmd $ZPOOL export -f lustre-$facet ||
8693 error "failed to export lustre-$facet"
8697 $rcmd rm -rf $tmp || error "failed to rm the dir $tmp"
8701 [ "$CLIENTONLY" ] && skip "Client-only testing"
8702 [ "$mds1_FSTYPE" != zfs ] && skip "zfs only test"
8703 [ "$MDS1_VERSION" -lt $(version_code 2.10.58) ] &&
8704 skip "Need server version at least 2.10.58"
8709 local tmp=$TMP/$tdir
8710 local rcmd="do_facet $SINGLEMDS"
8711 local facets="mdt1 mdt2 ost1 ost2"
8712 local nid=$($rcmd $LCTL list_nids | head -1)
8715 trap t_108_cleanup EXIT ERR
8718 t_108_mkfs mdt 0 zfs mgs lustre-mdt1/mdt1
8719 t_108_mkfs mdt 1 zfs mgsnode=$nid lustre-mdt2/mdt2
8720 t_108_mkfs ost 0 zfs mgsnode=$nid lustre-ost1/ost1
8721 t_108_mkfs ost 1 zfs mgsnode=$nid lustre-ost2/ost2
8723 for facet in $facets; do
8724 $rcmd zfs set mountpoint=$tmp/mnt/$facet canmount=on \
8725 lustre-$facet/$facet ||
8726 error "failed to zfs set for $facet (1)"
8727 $rcmd zfs mount lustre-$facet/$facet ||
8728 error "failed to local mount $facet"
8729 $rcmd tar jxf $LUSTRE/tests/ldiskfs_${facet}_2_11.tar.bz2 \
8730 --xattrs --xattrs-include="trusted.*" \
8731 -C $tmp/mnt/$facet/ > /dev/null 2>&1 ||
8732 error "failed to untar image for $facet"
8733 $rcmd "cd $tmp/mnt/$facet && rm -rf oi.* OI_* lfsck_* LFSCK" ||
8734 error "failed to cleanup for $facet"
8735 $rcmd zfs umount lustre-$facet/$facet ||
8736 error "failed to local umount $facet"
8737 $rcmd zfs set canmount=off lustre-$facet/$facet ||
8738 error "failed to zfs set $facet (2)"
8741 echo "changing server nid..."
8742 $rcmd mount -t lustre -o nosvc lustre-mdt1/mdt1 $tmp/mnt/mdt1
8743 $rcmd lctl replace_nids $FSNAME-MDT0000 $nid
8744 $rcmd lctl replace_nids $FSNAME-MDT0001 $nid
8745 $rcmd lctl replace_nids $FSNAME-OST0000 $nid
8746 $rcmd lctl replace_nids $FSNAME-OST0001 $nid
8747 $rcmd umount $tmp/mnt/mdt1
8749 for facet in $facets; do
8750 echo "mounting $facet from backup..."
8751 $rcmd mount -t lustre -o abort_recov lustre-$facet/$facet \
8752 $tmp/mnt/$facet || error "failed to mount $facet"
8755 # ZFS backend can detect migration and trigger OI scrub automatically
8756 # sleep 3 seconds for scrub done
8762 run_test 108a "migrate from ldiskfs to ZFS"
8765 [ "$CLIENTONLY" ] && skip "Client-only testing"
8766 [ "$mds1_FSTYPE" != ldiskfs ] && skip "ldiskfs only test"
8767 [ "$MDS1_VERSION" -lt $(version_code 2.10.58) ] &&
8768 skip "Need server version at least 2.10.58"
8773 local tmp=$TMP/$tdir
8774 local rcmd="do_facet $SINGLEMDS"
8775 local facets="mdt1 mdt2 ost1 ost2"
8776 local scrub_list="MDT0000 MDT0001 OST0000 OST0001"
8777 local nid=$($rcmd $LCTL list_nids | head -1)
8780 trap t_108_cleanup EXIT ERR
8783 t_108_mkfs mdt 0 ldiskfs mgs
8784 t_108_mkfs mdt 1 ldiskfs mgsnode=$nid
8785 t_108_mkfs ost 0 ldiskfs mgsnode=$nid
8786 t_108_mkfs ost 1 ldiskfs mgsnode=$nid
8788 for facet in $facets; do
8789 $rcmd mount -t ldiskfs -o loop $tmp/images/$facet \
8791 error "failed to local mount $facet"
8793 $rcmd tar jxf $LUSTRE/tests/zfs_${facet}_2_11.tar.bz2 \
8794 --xattrs --xattrs-include="*.*" \
8795 -C $tmp/mnt/$facet/ > /dev/null 2>&1 ||
8796 error "failed to untar image for $facet"
8797 $rcmd "cd $tmp/mnt/$facet && rm -rf oi.* OI_* lfsck_* LFSCK" ||
8798 error "failed to cleanup for $facet"
8799 $rcmd umount $tmp/mnt/$facet ||
8800 error "failed to local umount $facet"
8803 echo "changing server nid..."
8804 $rcmd mount -t lustre -o nosvc,loop $tmp/images/mdt1 $tmp/mnt/mdt1
8805 $rcmd lctl replace_nids $FSNAME-MDT0000 $nid
8806 $rcmd lctl replace_nids $FSNAME-MDT0001 $nid
8807 $rcmd lctl replace_nids $FSNAME-OST0000 $nid
8808 $rcmd lctl replace_nids $FSNAME-OST0001 $nid
8809 $rcmd umount $tmp/mnt/mdt1
8811 for facet in $facets; do
8812 echo "mounting $facet from backup..."
8813 $rcmd mount -t lustre -o loop,abort_recov $tmp/images/$facet \
8814 $tmp/mnt/$facet || error "failed to mount $facet"
8817 for facet in $scrub_list; do
8818 $rcmd $LCTL lfsck_start -M $FSNAME-$facet -t scrub ||
8819 error "failed to start OI scrub on $facet"
8822 # sleep 3 seconds for scrub done
8828 run_test 108b "migrate from ZFS to ldiskfs"
8831 # set number of permanent parameters
8833 test_109_set_params() {
8836 set_persistent_param_and_check mds \
8837 "mdd.$fsname-MDT0000.atime_diff" \
8838 "$fsname-MDT0000.mdd.atime_diff" \
8840 set_persistent_param_and_check mds \
8841 "mdd.$fsname-MDT0000.atime_diff" \
8842 "$fsname-MDT0000.mdd.atime_diff" \
8844 set_persistent_param_and_check client \
8845 "llite.$fsname*.max_read_ahead_mb" \
8846 "$fsname.llite.max_read_ahead_mb" \
8848 set_persistent_param_and_check client \
8849 "llite.$fsname*.max_read_ahead_mb" \
8850 "$fsname.llite.max_read_ahead_mb" \
8852 create_pool $fsname.pool1 || error "create pool failed"
8853 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
8854 error "pool_add failed"
8855 do_facet mgs $LCTL pool_remove $fsname.pool1 OST0000 ||
8856 error "pool_remove failed"
8857 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
8858 error "pool_add failed"
8862 # check permanent parameters
8864 test_109_test_params() {
8867 local atime_diff=$(do_facet mds $LCTL \
8868 get_param -n mdd.$fsname-MDT0000.atime_diff)
8869 [ $atime_diff == 63 ] || error "wrong mdd parameter after clear_conf"
8870 local max_read_ahead_mb=$(do_facet client $LCTL \
8871 get_param -n llite.$fsname*.max_read_ahead_mb)
8872 [ $max_read_ahead_mb == 64 ] ||
8873 error "wrong llite parameter after clear_conf"
8874 local ost_in_pool=$(do_facet mds $LCTL pool_list $fsname.pool1 |
8875 grep -v "^Pool:" | sed 's/_UUID//')
8876 [ $ost_in_pool = "$fsname-OST0000" ] ||
8877 error "wrong pool after clear_conf"
8881 # run lctl clear_conf, store CONFIGS before and after that
8883 test_109_clear_conf()
8885 local clear_conf_arg=$1
8888 if ! combined_mgs_mds ; then
8889 mgsdev=$(mgsdevname)
8890 stop_mgs || error "stop_mgs failed"
8891 start_mgs "-o nosvc" || error "start_mgs nosvc failed"
8893 mgsdev=$(mdsdevname 1)
8894 start_mdt 1 "-o nosvc" || error "start_mdt 1 nosvc failed"
8897 do_facet mgs "rm -rf $TMP/${tdir}/conf1; mkdir -p $TMP/${tdir}/conf1;" \
8898 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf1\\\" \
8902 # the command being tested
8904 do_facet mgs $LCTL clear_conf $clear_conf_arg ||
8905 error "clear_conf failed"
8906 if ! combined_mgs_mds ; then
8907 stop_mgs || error "stop_mgs failed"
8909 stop_mdt 1 || error "stop_mdt 1 failed"
8912 do_facet mgs "rm -rf $TMP/${tdir}/conf2; mkdir -p $TMP/${tdir}/conf2;" \
8913 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf2\\\" \
8917 test_109_file_shortened() {
8919 local sizes=($(do_facet mgs "stat -c %s " \
8920 "$TMP/${tdir}/conf1/CONFIGS/$file" \
8921 "$TMP/${tdir}/conf2/CONFIGS/$file"))
8922 [ ${sizes[1]} -lt ${sizes[0]} ] && return 0
8928 [ $MDS1_VERSION -lt $(version_code 2.10.59) ] &&
8929 skip "Needs MDS version 2.10.59 or later."
8930 [ "$(facet_fstype mgs)" == zfs ] &&
8931 skip "LU-8727: no implementation for ZFS"
8936 client_up || error "client_up failed"
8939 # set number of permanent parameters
8941 test_109_set_params $FSNAME
8943 umount_client $MOUNT || error "umount_client failed"
8944 stop_ost || error "stop_ost failed"
8945 stop_mds || error "stop_mds failed"
8947 test_109_clear_conf $FSNAME
8949 # make sure that all configs are cleared
8951 test_109_file_shortened $FSNAME-MDT0000 ||
8952 error "failed to clear MDT0000 config"
8953 test_109_file_shortened $FSNAME-client ||
8954 error "failed to clear client config"
8959 # check that configurations are intact
8961 test_109_test_params $FSNAME
8966 destroy_test_pools || error "destroy test pools failed"
8970 run_test 109a "test lctl clear_conf fsname"
8974 [ $MDS1_VERSION -lt $(version_code 2.10.59) ] &&
8975 skip "Needs MDS version 2.10.59 or later."
8976 [ "$(facet_fstype mgs)" == zfs ] &&
8977 skip "LU-8727: no implementation for ZFS"
8982 client_up || error "client_up failed"
8985 # set number of permanent parameters
8987 test_109_set_params $FSNAME
8989 umount_client $MOUNT || error "umount_client failed"
8990 stop_ost || error "stop_ost failed"
8991 stop_mds || error "stop_mds failed"
8993 test_109_clear_conf $FSNAME-MDT0000
8995 # make sure that only one config is cleared
8997 test_109_file_shortened $FSNAME-MDT0000 ||
8998 error "failed to clear MDT0000 config"
8999 test_109_file_shortened $FSNAME-client &&
9000 error "failed to clear client config"
9004 # check that configurations are intact
9006 test_109_test_params $FSNAME
9011 destroy_test_pools || error "destroy test pools failed"
9015 run_test 109b "test lctl clear_conf one config"
9019 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
9020 skip "Only applicable to ldiskfs-based MDTs"
9022 do_facet $SINGLEMDS $DEBUGFS -w -R supported_features |grep large_dir ||
9023 skip "large_dir option is not supported on MDS"
9024 do_facet ost1 $DEBUGFS -w -R supported_features | grep large_dir ||
9025 skip "large_dir option is not supported on OSS"
9027 stopall # stop all targets before modifying the target counts
9028 local old_mdscount=$MDSCOUNT
9029 local old_ostcount=$OSTCOUNT
9031 stack_trap "MDSCOUNT=$old_mdscount OSTCOUNT=$old_ostcount; reformat"
9035 # ext4_dir_entry_2 struct size:264
9036 # dx_root struct size:8
9037 # dx_node struct size:8
9038 # dx_entry struct size:8
9039 # For 1024 bytes block size.
9040 # First level directory entries: 126
9041 # Second level directory entries: 127
9042 # Entries in leaf: 3
9043 # For 2 levels limit: 48006
9044 # For 3 levels limit : 6096762
9045 # Create 80000 files to safely exceed 2-level htree limit.
9046 CONF_SANITY_110_LINKS=${CONF_SANITY_110_LINKS:-80000}
9048 # can fit at most 3 filenames per 1KB leaf block, but each
9049 # leaf/index block will only be 3/4 full before split at each level
9050 (( MDSSIZE < CONF_SANITY_110_LINKS / 3 * 4/3 * 4/3 )) &&
9051 CONF_SANITY_110_LINKS=$((MDSSIZE * 3 * 3/4 * 3/4))
9053 combined_mgs_mds || replace=" --replace "
9054 local opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
9055 $replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
9056 if [[ $opts != *large_dir* ]]; then
9057 if [[ $opts != *mkfsoptions* ]]; then
9058 opts+=" --mkfsoptions=\\\"-O large_dir -b 1024 -i 65536\\\""
9060 opts="${opts//--mkfsoptions=\\\"/ \
9061 --mkfsoptions=\\\"-O large_dir -b 1024 -i 65536 }"
9064 echo "MDT params: $opts"
9066 combined_mgs_mds || start_mgs
9067 add mds1 $opts || error "add mds1 failed with new params"
9068 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
9070 opts="$(mkfs_opts ost1 $(ostdevname 1)) \
9071 $replace --reformat $(ostdevname 1) $(ostvdevname 1)"
9073 if [[ $opts != *large_dir* ]]; then
9074 if [[ $opts != *mkfsoptions* ]]; then
9075 opts+=" --mkfsoptions=\\\"-O large_dir\\\" "
9077 opts="${opts//--mkfsoptions=\\\"/ \
9078 --mkfsoptions=\\\"-O large_dir }"
9081 echo "OST params: $opts"
9082 add ost1 $opts || error "add ost1 failed with new params"
9083 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
9085 MOUNT_2=yes mountcli || error "mount clients failed"
9087 mkdir -v $DIR/$tdir || error "cannot create $DIR/$tdir"
9088 local pids count=0 group=0
9090 echo "creating $CONF_SANITY_110_LINKS in total"
9091 while (( count < CONF_SANITY_110_LINKS )); do
9092 local len=$((253 - $(wc -c <<<"$tfile-$group-40000-")))
9093 local dir=DIR$((group % 2 + 1))
9094 local target=${!dir}/$tdir/$tfile-$group
9095 local long=$target-$(generate_name $len)-
9096 local create=$((CONF_SANITY_110_LINKS - count))
9098 (( create > 40000 )) && create=40000
9099 touch $target || error "creating $target failed"
9100 echo "creating $create hard links to $target"
9101 createmany -l $target $long $create &
9104 count=$((count + create))
9105 group=$((group + 1))
9107 echo "waiting for PIDs$pids to complete"
9108 wait $pids || error "createmany failed after $group groups"
9110 umount_client $MOUNT2 -f
9113 run_e2fsck $(facet_active_host mds1) $(mdsdevname 1) -n
9114 MDSCOUNT=$old_mdscount
9115 OSTCOUNT=$old_ostcount
9117 run_test 110 "Adding large_dir with 3-level htree"
9120 [[ "$mds1_FSTYPE" != ldiskfs ]] &&
9121 skip "Only applicable to ldiskfs-based MDTs"
9123 is_dm_flakey_dev $SINGLEMDS $(mdsdevname 1) &&
9124 skip "This test can not be executed on flakey dev"
9126 do_facet $SINGLEMDS $DEBUGFS -w -R supported_features |grep large_dir ||
9127 skip "large_dir option is not supported on MDS"
9129 do_facet ost1 $DEBUGFS -w -R supported_features | grep large_dir ||
9130 skip "large_dir option is not supported on OSS"
9132 # cleanup before changing target counts
9134 local old_mdscount=$MDSCOUNT
9135 local old_ostcount=$OSTCOUNT
9136 local old_mdssize=$MDSSIZE
9138 stack_trap "MDSSIZE=$old_mdssize MDSCOUNT=$old_mdscount \
9139 OSTCOUNT=$old_ostcount; reformat"
9142 (( MDSSIZE < 2400000 )) && MDSSIZE=2400000 # need at least 2.4GB
9144 local mdsdev=$(mdsdevname 1)
9145 combined_mgs_mds || replace=" --replace "
9146 local opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
9147 $replace --reformat $(mdsdevname 1) $(mdsvdevname 1)"
9148 if [[ $opts != *large_dir* ]]; then
9149 if [[ $opts != *mkfsoptions* ]]; then
9150 opts+=" --mkfsoptions=\\\"-O large_dir -i 1048576 \\\" "
9152 opts="${opts//--mkfsoptions=\\\"/ \
9153 --mkfsoptions=\\\"-O large_dir -i 1048576 }"
9156 echo "MDT params: $opts"
9158 combined_mgs_mds || start_mgs
9159 __touch_device mds 1
9160 add mds1 $opts || error "add mds1 failed with new params"
9161 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
9163 opts="$(mkfs_opts ost1 $(ostdevname 1)) \
9164 $replace --reformat $(ostdevname 1) $(ostvdevname 1)"
9165 if [[ $opts != *large_dir* ]]; then
9166 if [[ $opts != *mkfsoptions* ]]; then
9167 opts+=" --mkfsoptions=\\\"-O large_dir \\\""
9169 opts="${opts//--mkfsoptions=\\\"/ --mkfsoptions=\\\"-O large_dir }"
9172 echo "OST params: $opts"
9173 __touch_device ost 1
9174 add ost1 $opts || error "add ost1 failed with new params"
9175 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
9177 MOUNT_2=yes mountcli
9178 mkdir $DIR/$tdir || error "cannot create $DIR/$tdir"
9180 lfs df -i $DIR/$tdir
9184 local start=$SECONDS
9186 local dirmax=$((2 << 30))
9192 while (( !needskip & dirsize < dirmax )); do
9195 for cli in ${CLIENTS//,/ }; do
9196 local len=$((253 - $(wc -c <<<"$cli-$group-60000-")))
9197 local target=$cli-$group
9198 local long=$DIR/$tdir/$target-$(generate_name $len)-
9200 RPWD=$DIR/$tdir do_node $cli touch $target ||
9201 error "creating $target failed"
9202 echo "creating 60000 hardlinks to $target"
9203 RPWD=$DIR/$tdir do_node $cli createmany -l $target $long 60000 &
9206 group=$((group + 1))
9208 long=$DIR2/$tdir/$target-$(generate_name $len)-
9210 RPWD=$DIR2/$tdir do_node $cli touch $target ||
9211 error "creating $target failed"
9212 echo "creating 60000 hardlinks to $target"
9213 RPWD=$DIR2/$tdir do_node $cli createmany -l $target $long 60000 &
9216 group=$((group + 1))
9218 echo "waiting for PIDs$pids to complete"
9219 wait $pids || error "createmany failed after $group groups"
9220 dirsize=$(stat -c %s $DIR/$tdir)
9221 taken=$((SECONDS - start))
9222 rate=$((dirsize / taken))
9223 left=$(((dirmax - dirsize) / rate))
9224 num=$((group * 60000))
9225 echo "estimate ${left}s left after $num files / ${taken}s"
9226 # if the estimated time remaining is too large (it may change
9227 # over time as the create rate is not constant) then exit
9228 # without declaring a failure.
9229 (( left > 1200 )) && needskip=1
9232 umount_client $MOUNT2 -f
9235 ! (( $needskip )) ||
9236 echo "ETA ${left}s after $num files / ${taken}s is too long"
9238 run_e2fsck $(facet_active_host mds1) $(mdsdevname 1) -n
9239 MDSCOUNT=$old_mdscount
9240 OSTCOUNT=$old_ostcount
9241 MDSSIZE=$old_mdssize
9243 run_test 111 "Adding large_dir with over 2GB directory"
9246 local param="no_create"
9248 (( $OST1_VERSION > $(version_code 2.14.0) )) ||
9249 skip "need OSS at least 2.14.0"
9250 (( $OST1_VERSION >= $(version_code 2.15.56.125) )) ||
9251 param="no_precreate"
9253 start_mds || error "MDS start failed"
9254 start_ost || error "OSS start failed"
9255 echo "start ost2 service on $(facet_active_host ost2)"
9256 start ost2 $(ostdevname 2) $(csa_add "$OST_MOUNT_OPTS" -o $param) ||
9257 error "start ost2 facet failed"
9258 local val=$(do_facet ost2 \
9259 "$LCTL get_param -n obdfilter.$FSNAME-OST0001*.$param")
9260 (( $val == 1 )) || error "obdfilter.$FSNAME-OST0001*.$param=$val"
9262 mount_client $MOUNT || error "mount client failed"
9263 wait_osc_import_state mds1 ost1 FULL
9264 wait_osc_import_ready client ost1
9265 wait_osc_import_ready client ost2
9267 $LFS setstripe -i 0 $DIR/$tfile.0 ||
9268 error "problem creating $tfile.0 on OST0000"
9269 $LFS setstripe -i 1 $DIR/$tfile.1 && $LFS getstripe $DIR/$tfile.1 &&
9270 (( $($LFS getstripe -i $DIR/$tfile.1) == 1 )) &&
9271 error "allowed to create $tfile.1 on OST0001"
9273 $LFS df -v $MOUNT | grep -q "OST:1.*N" ||
9274 error "NOCREATE not in 'lfs df'"
9276 do_facet ost2 $LCTL set_param obdfilter.$FSNAME-OST0001*.$param=0
9278 $LFS setstripe -i 1 $DIR/$tfile.2 ||
9279 error "failed to create $tfile.2 on ost1 facet"
9280 # files not cleaned with ONLY_REPEAT because of client unmount below
9281 rm -f $DIR/$tfile.[012]
9282 stop_ost2 || error "stop ost2 facet failed"
9285 run_test 112a "mount OST with no_create option"
9288 (( MDSCOUNT >= 2 )) || skip "need at least 2 MDTs"
9289 (( $MDS1_VERSION >= $(version_code 2.15.56.125) )) ||
9290 skip "need MDS >= 2.15.56.125"
9291 local mdsnum=$MDSCOUNT
9292 local facet=mds$mdsnum
9293 local mdtidx=$((mdsnum - 1))
9294 local mdtname=$FSNAME-MDT$(printf %04x $mdtidx)
9296 start_mds --mdscount $((mdsnum - 1)) || error "MDS start failed"
9297 start_mdt $mdsnum -o no_create || error "start $facet failed"
9298 local val=$(do_facet $facet \
9299 "$LCTL get_param -n mdt.$mdtname*.no_create")
9300 (( $val == 1 )) || error "mdt.$mdtname*.no_create=$val"
9301 start_ost || error "ost1 start failed"
9302 start_ost2 || error "ost1 start failed"
9304 mount_client $MOUNT || error "mount client failed"
9305 wait_osc_import_ready $facet ost2
9308 $LFS df -v $MOUNT | grep -q "MDT:$mdtidx.*N" ||
9309 error "NOCREATE not in 'lfs df'"
9311 $LFS mkdir -i $mdtidx $DIR/$tdir ||
9312 $LFS setdirstripe -D -c 1 -i -1 --max-inherit-rr 2 $DIR/$tdir ||
9313 error "error creating $tdir on $mdtname"
9314 stack_trap "rm -rf $DIR/$tdir"
9316 mkdir $DIR/$tdir/d1.{1..100} || error "mkdir $tdir/d1.{1..100} failed"
9317 $LFS getdirstripe -i $DIR/$tdir/d1.* | sort | uniq -c
9318 do_facet $facet $LCTL set_param mdt.$mdtname*.no_create=0
9319 # allow one initial create for delayed statfs on client
9320 (( $($LFS getdirstripe -i $DIR/$tdir/d1.* | grep -c $mdtidx) < 2 )) ||
9321 error "allowed create on $mdtname"
9324 mkdir $DIR/$tdir/d2.{1..100} || error "mkdir $tdir/d2.{1..100} failed"
9325 $LFS getdirstripe -i $DIR/$tdir/d2.{1..100} | sort | uniq -c
9326 (( $($LFS getdirstripe -i $DIR/$tdir/d2.* | grep -c $mdtidx) > 10 )) ||
9327 error "no create on $mdtname"
9328 # files not cleaned with ONLY_REPEAT because of client unmount below
9330 stop_ost2 || error "ost1 start failed"
9333 run_test 112b "mount MDT with no_create option"
9336 SAVE_MGS_MOUNT_OPTS=$MGS_MOUNT_OPTS
9337 SAVE_MDS_MOUNT_OPTS=$MDS_MOUNT_OPTS
9338 SAVE_OST_MOUNT_OPTS=$OST_MOUNT_OPTS
9344 MGS_MOUNT_OPTS=$SAVE_MGS_MOUNT_OPTS
9345 MDS_MOUNT_OPTS=$SAVE_MDS_MOUNT_OPTS
9346 OST_MOUNT_OPTS=$SAVE_OST_MOUNT_OPTS
9347 # Revert old mount options back
9349 # Subsequent following test requires
9350 # conf-sanity to be in stopall state.
9351 # Force 'stopall' so others following
9356 # Error out with mount info
9358 local server_nodes=$(comma_list $(mdts_nodes) $(osts_nodes))
9361 echo "--Client Mount Info--"
9362 mount | grep -i lustre
9363 echo "--Server Mount Info--"
9364 do_nodes $server_nodes mount | grep -i lustre
9370 local ost_version="2.15.51" # Minimum version required
9372 (( OST1_VERSION >= $(version_code $ost_version) )) ||
9373 skip "Need server version at least $ost_version"
9375 stack_trap cleanup_113 EXIT
9377 # Reset before starting
9381 # Verify MDS's should start with "rw"
9382 do_facet $SINGLEMDS mount | grep "lustre.*rw,.*MDT" ||
9383 error_113 "$SINGLEMDS should be read-write"
9385 # Verify OST's should start with "rw"
9386 for (( i=1; i <= OSTCOUNT; i++ )); do
9387 do_facet ost$i mount | grep "lustre.*rw,.*OST" ||
9388 error_113 "ost$i should be read-write"
9391 # rdonly_dev does not currently work for ldiskfs
9392 # We skip the rdonly_dev check until then.
9393 if [[ $ost1_FSTYPE == ldiskfs ]]; then
9394 echo "Shadow Mountpoint correctly reports rw for ldiskfs"
9399 # Only ZFS specific tests below.
9402 # Must stop all (server+client) and restart to verify new
9406 # add rdonly_dev to mount option
9407 MGS_MOUNT_OPTS=$(csa_add "$MGS_MOUNT_OPTS" -o rdonly_dev)
9408 MDS_MOUNT_OPTS=$(csa_add "$MDS_MOUNT_OPTS" -o rdonly_dev)
9409 OST_MOUNT_OPTS=$(csa_add "$OST_MOUNT_OPTS" -o rdonly_dev)
9411 # Only restart server(mds/ost). Sufficient for test
9412 setupall server_only || error "Fail to start servers"
9414 # Verify MDS's should be "ro"
9415 do_facet $SINGLEMDS mount | grep "lustre.*ro,.*MDT.*rdonly_dev" ||
9416 error_113 "$SINGLEMDS should be read-only"
9418 # Verify OST's should be "ro"
9419 for (( i=1; i <= OSTCOUNT; i++ )); do
9420 do_facet ost$i mount | grep "lustre.*ro,.*OST.*rdonly_dev" ||
9421 error_113 "ost$i should be read-only"
9424 run_test 113 "Shadow mountpoint correctly report ro/rw for mounts"
9427 # Purpose: To verify dynamic thread (OSS) creation.
9428 # (This was sanity/115)
9431 [ $PARALLEL == "yes" ] && skip "skip parallel run"
9432 remote_ost_nodsh && skip "remote OST with nodsh"
9434 # Lustre does not stop service threads once they are started.
9435 # Reset number of running threads to default.
9440 local save_params="$TMP/sanity-$TESTNAME.parameters"
9442 # Get ll_ost_io count before I/O
9443 OSTIO_pre=$(do_facet ost1 \
9444 "$LCTL get_param ost.OSS.ost_io.threads_started | cut -d= -f2")
9445 # Exit if lustre is not running (ll_ost_io not running).
9446 [ -z "$OSTIO_pre" ] && error "no OSS threads"
9448 echo "Starting with $OSTIO_pre threads"
9449 local thread_max=$((OSTIO_pre * 2))
9450 local rpc_in_flight=$((thread_max * 2))
9452 # this is limited to OSC_MAX_RIF_MAX (256)
9453 [ $rpc_in_flight -gt 256 ] && rpc_in_flight=256
9454 thread_max=$((rpc_in_flight / 2))
9455 [ $thread_max -le $OSTIO_pre ] && skip "Too many ost_io threads" &&
9458 # Number of I/O Process proposed to be started.
9460 local facets=$(get_facets OST)
9462 save_lustre_params client "osc.*OST*.max_rpcs_in_flight" > $save_params
9463 save_lustre_params $facets "ost.OSS.ost_io.threads_max" >> $save_params
9465 # Set in_flight to $rpc_in_flight
9466 $LCTL set_param osc.*OST*.max_rpcs_in_flight=$rpc_in_flight ||
9467 error "Failed to set max_rpcs_in_flight to $rpc_in_flight"
9468 nfiles=${rpc_in_flight}
9469 # Set ost thread_max to $thread_max
9470 do_facet ost1 "$LCTL set_param ost.OSS.ost_io.threads_max=$thread_max"
9472 # 5 Minutes should be sufficient for max number of OSS
9473 # threads(thread_max) to be created.
9477 local wtl=${WTL:-"$LUSTRE/tests/write_time_limit"}
9479 test_mkdir $DIR/$tdir
9480 for ((i = 1; i <= nfiles; i++)); do
9481 local file=$DIR/$tdir/${tfile}-$i
9483 $LFS setstripe -c -1 -i 0 $file
9484 ($wtl $file $timeout)&
9487 # I/O Started - Wait for thread_started to reach thread_max or report
9488 # error if thread_started is more than thread_max.
9489 echo "Waiting for thread_started to reach thread_max"
9490 local thread_started=0
9491 local end_time=$((SECONDS + timeout))
9493 while [ $SECONDS -le $end_time ] ; do
9495 # Get ost i/o thread_started count.
9496 thread_started=$(do_facet ost1 \
9498 ost.OSS.ost_io.threads_started | cut -d= -f2")
9499 # Break out if thread_started is equal/greater than thread_max
9500 if (( $thread_started >= $thread_max )); then
9501 echo ll_ost_io thread_started $thread_started, \
9502 equal/greater than thread_max $thread_max
9508 # Cleanup - We have the numbers, Kill i/o jobs if running.
9509 jobcount=($(jobs -p))
9511 for ((i=0; i < ${#jobcount[*]}; i++)); do
9512 kill -9 ${jobcount[$i]}
9513 if [ $? -ne 0 ] ; then
9514 echo "warning: cannot kill WTL pid ${jobcount[$i]}"
9518 # Cleanup files left by WTL binary.
9519 for ((i = 1; i <= nfiles; i++)); do
9520 local file=$DIR/$tdir/${tfile}-$i
9523 if [ $? -ne 0 ] ; then
9524 echo "Warning: Failed to delete file $file"
9528 restore_lustre_params <$save_params
9529 rm -f $save_params || echo "Warning: delete file '$save_params' failed"
9531 # Error out if no new thread has started or Thread started is greater
9533 if (( $thread_started <= $OSTIO_pre ||
9534 $thread_started > $thread_max )); then
9535 error "ll_ost_io: thread_started $thread_started" \
9536 "OSTIO_pre $OSTIO_pre, thread_max $thread_max." \
9537 "No new thread started or thread started greater " \
9541 run_test 114 "verify dynamic thread creation===================="
9547 do_facet mds1 rm -f $1
9551 if [ "$mds1_FSTYPE" != ldiskfs ]; then
9552 skip "Only applicable to ldiskfs-based MDTs"
9554 [ -n "$FILESET" ] && skip "Not functional for FILESET set"
9556 local dbfs_ver=$(do_facet mds1 $DEBUGFS -V 2>&1)
9558 echo "debugfs version: $dbfs_ver"
9559 echo "$dbfs_ver" | egrep -w "1.44.3.wc1|1.44.5.wc1|1.45.2.wc1" &&
9560 skip_env "This version of debugfs doesn't show inode number"
9562 local IMAGESIZE=$((3072 << 30)) # 3072 GiB
9572 do_facet mds1 "hostname; $LCTL dl; mount"
9574 do_facet ost1 "hostname; $LCTL dl; mount"
9575 # We need MDT size 3072GB, because it is smallest
9576 # partition that can store 2B inodes
9577 do_facet mds1 "mkdir -p $TMP/$tdir"
9578 local mdsimgname=$TMP/$tdir/lustre-mdt
9580 do_facet mds1 "rm -f $mdsimgname"
9581 do_facet mds1 "touch $mdsimgname"
9582 stack_trap "cleanup_115 $mdsimgname" EXIT
9583 do_facet mds1 "$TRUNCATE $mdsimgname $IMAGESIZE" ||
9584 skip "Backend FS doesn't support sparse files"
9585 local mdsdev=$(do_facet mds1 "losetup -f")
9587 do_facet mds1 "losetup $mdsdev $mdsimgname"
9589 local mds_opts="$(mkfs_opts mds1 $(mdsdevname 1)) \
9590 --mkfsoptions='-O ea_inode,^resize_inode,meta_bg \
9591 -N 2247484000 -E lazy_itable_init' --device-size=$IMAGESIZE"
9592 add mds1 $mds_opts --mgs --reformat $mdsdev || {
9593 do_facet $SINGLEMDS \
9594 "losetup -d $mdsdev && rm -f $mdsimgname"
9595 skip_env "format large MDT failed"
9598 local ostdev=$(ostdevname 1)
9600 local opts="$(mkfs_opts ost1 $ostdev) \
9601 --reformat $ostdev $ostdev"
9602 add ost1 $opts || error "add ost1 failed with new params"
9603 start mds1 $mdsdev $MDS_MOUNT_OPTS || error "start MDS failed"
9604 start_ost || error "start OSS failed"
9605 mount_client $MOUNT || error "mount client failed"
9607 mkdir -p $DIR/$tdir || error "mkdir $DIR/$tdir fail"
9608 local goal="/sys/fs/ldiskfs/$(basename $mdsdev)/inode_goal"
9610 # 2147483648 is 0x80000000
9611 do_facet mds1 "echo 2147483648 >> $goal; grep . $goal"
9612 touch $DIR/$tdir/$tfile
9614 # attrs from 1 to 16 go to block, 17th - to inode
9617 for i in {1..17}; do
9618 local nm="trusted.ea$i"
9619 setfattr -n $nm -v $(printf "xattr%0250d" $i) $DIR/$tdir/$tfile
9622 do_facet mds1 "$DEBUGFS -c -R 'stat ROOT/$tdir/$tfile' $mdsdev"
9624 # inode <2147483649> trusted.ea16 (255)
9625 local inode_num=$(do_facet mds1 \
9626 "$DEBUGFS -c -R 'stat ROOT/$tdir/$tfile' $mdsdev" |
9627 awk '/ea17/ { print $2 }' |
9628 sed -e 's/>//' -e 's/<//' -e 's/\"//')
9629 echo "inode num: $inode_num"
9630 [ $inode_num -ge 2147483648 ] || error "inode $inode_num too small"
9631 do_facet mds1 "losetup -d $mdsdev"
9633 run_test 115 "Access large xattr with inodes number over 2TB"
9636 [ "$mds1_FSTYPE" != ldiskfs ] && skip "ldiskfs only test"
9637 [ "$MDS1_VERSION" -lt $(version_code 2.10.59) ] &&
9638 skip "Need server version at least 2.10.59"
9639 do_facet $SINGLEMDS which mkfs.xfs ||
9640 skip_env "No mkfs.xfs installed"
9645 local tmpmnt=$TMP/$tdir
9646 local mdtimg=$tfile-mdt0
9648 do_facet $SINGLEMDS mkdir -p $tmpmnt
9649 stack_trap "do_facet $SINGLEMDS rmdir $tmpmnt" EXIT
9651 do_facet $SINGLEMDS touch $TMP/$mdtimg
9652 stack_trap "do_facet $SINGLEMDS rm -f $TMP/$mdtimg" EXIT
9653 do_facet $SINGLEMDS mkfs -t xfs -d file,size=1t,name=$TMP/$mdtimg ||
9654 error "mkfs temporary xfs image"
9656 do_facet $SINGLEMDS mount $TMP/$mdtimg $tmpmnt ||
9657 error "mount temporary xfs image"
9658 stack_trap "do_facet $SINGLEMDS umount $tmpmnt" EXIT
9659 local old_mdssize=$MDSSIZE
9660 local old_mdsisize=$MDSISIZE
9662 MDSSIZE=$((17 * 1024 * 1024 * 1024)) # 17T MDT
9663 MDSISIZE=$((16 << 20))
9664 local opts17t="$(mkfs_opts $SINGLEMDS)"
9666 MDSSIZE=$old_mdssize
9667 MDSISIZE=$old_mdsisize
9669 do_facet $SINGLEMDS $MKFS \
9670 ${opts17t/-E /-E lazy_itable_init,lazy_journal_init,} \
9672 error "failed to mkfs for $tmpmnt/$mdtimg"
9674 do_facet $SINGLEMDS $TUNE2FS -l $tmpmnt/$mdtimg |
9675 grep -qw 'features.*extent' || error "extent should be enabled"
9678 run_test 116 "big size MDT support"
9681 # Call setup only if LustreFS is not mounted
9682 check_mount || setup
9684 do_facet ost1 "$LCTL set_param ost.OSS.ost_io.nrs_policies=fifo"
9685 do_facet ost1 "$LCTL get_param -n ost.OSS.ost_io.nrs_tbf_rule" &&
9686 error "get_param should fail"
9687 cleanup || error "cleanup failed with rc $?"
9689 run_test 117 "lctl get_param return errors properly"
9695 [[ "$MDSCOUNT" -ge 2 ]] || skip "Need more at least 2 MDTs"
9697 had_config=$(do_facet mds1 "$LCTL get_param debug | grep config")
9698 do_facet mds1 "$LCTL set_param debug=+config"
9699 do_facet mds1 "$LCTL clear"
9702 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)" &>/dev/null
9703 # mount after writeconf will make "add osp" added to mdt0 config:
9704 # 53 (224)marker 60 (flags=0x01, v2.5.1.0) lustre-MDT0001 'add osp'
9705 # 54 (080)add_uuid nid=... 0: 1:...
9706 # 55 (144)attach 0:lustre-MDT0001-osp-MDT0000 1:osp 2:...
9707 # 56 (144)setup 0:lustre-MDT0001-osp-MDT0000 1:... 2:...
9708 # 57 (136)modify_mdc_tgts add 0:lustre-MDT0000-mdtlov 1:... 2:1 3:1
9709 # duplicate modify_mdc_tgts caused crashes
9712 # using larger debug_mb size to avoid lctl dk log truncation
9713 size_mb=$((DEBUG_SIZE_SAVED * 4))
9716 # though config processing stops after failed attach and setup
9717 # it will proceed after the failed command after each writeconf
9718 # this is the original scenario of the issue
9719 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)" &>/dev/null
9720 do_facet mds1 "$LCTL set_param debug_mb=$size_mb"
9723 wait_update_facet_cond mds1 \
9724 "$LCTL dk | grep -c Processed.log.$FSNAME-MDT0000" \
9729 [[ -z "$had_config" ]] && do_facet mds1 lctl set_param debug=-config
9733 run_test 119 "writeconf on slave mdt shouldn't duplicate mdc/osp and crash"
9735 test_120() { # LU-11130
9736 [ "$MDSCOUNT" -lt 2 ] && skip "mdt count < 2"
9737 [ "$mds1_FSTYPE" != ldiskfs ] &&
9738 skip "ldiskfs only test"
9739 [ "$MDS1_VERSION" -lt $(version_code 2.11.56) ] &&
9740 skip "Need DNE2 capable MD target with LU-11130 fix"
9744 local mds1host=$(facet_active_host mds1)
9745 local mds1dev=$(mdsdevname 1)
9747 $LFS mkdir -i 1 $DIR/$tdir
9748 $LFS mkdir -i 0 $DIR/$tdir/mds1dir
9750 ln -s foo $DIR/$tdir/bar
9751 mv $DIR/$tdir/bar $DIR/$tdir/mds1dir/bar2 ||
9752 error "cross-target rename failed"
9756 run_e2fsck $mds1host $mds1dev "-n"
9758 run_test 120 "cross-target rename should not create bad symlinks"
9762 start_mgsmds || error "MGS MDS Start failed"
9764 stop_mds || error "Stopping MDSes failed"
9768 stop_mds || error "Stopping MDSes failed"
9770 run_test 121 "failover MGS"
9773 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
9774 [[ "$OST1_VERSION" -ge $(version_code 2.11.53) ]] ||
9775 skip "Need OST version at least 2.11.53"
9778 LOAD_MODULES_REMOTE=true load_modules
9779 #define OBD_FAIL_OFD_SET_OID 0x1e0
9780 do_facet ost1 $LCTL set_param fail_loc=0x00001e0
9783 $LFS mkdir -i1 -c1 $DIR/$tdir
9784 $LFS setstripe -i0 -c1 $DIR/$tdir
9785 do_facet ost1 $LCTL set_param fail_loc=0
9786 createmany -o $DIR/$tdir/file_ 1000 ||
9787 error "Fail to create a new sequence"
9791 run_test 122a "Check OST sequence update"
9794 remote_mgs_nodsh && skip "remote MGS with nodsh"
9795 [ -d $MOUNT/.lustre ] || setup
9797 # test old logid format until removal from llog_ioctl.c::str2logid()
9798 if (( $MGS_VERSION < $(version_code 2.99.53) )); then
9799 do_facet mgs $LCTL dl | grep MGS
9800 do_facet mgs "$LCTL --device %MGS llog_print \
9801 \\\\\\\$$FSNAME-client 1 10" ||
9802 error "old llog_print failed"
9805 # test new logid format
9806 if [ $MGS_VERSION -ge $(version_code 2.9.53) ]; then
9807 do_facet mgs "$LCTL --device MGS llog_print $FSNAME-client" ||
9808 error "new llog_print failed"
9811 run_test 123aa "llog_print works with FIDs and simple names"
9814 remote_mgs_nodsh && skip "remote MGS with nodsh"
9815 [[ $MGS_VERSION -gt $(version_code 2.11.51) ]] ||
9816 skip "Need server with working llog_print support"
9818 [ -d $MOUNT/.lustre ] || setup
9824 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9825 mgs_arg="--device MGS"
9827 orig_val=$(do_facet mgs $LCTL get_param jobid_name)
9828 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME"
9830 yaml=$(do_facet mgs $LCTL $mgs_arg llog_print params |
9831 grep jobid_name | tail -n 1)
9833 local param=$(awk '{ print $10 }' <<< "$yaml")
9834 local val=$(awk '{ print $12 }' <<< "$yaml")
9835 #return to the default
9836 do_facet mgs $LCTL set_param -P jobid_name=$orig_val
9837 [ $val = "TESTNAME" ] || error "bad value: $val"
9838 [ $param = "jobid_name," ] || error "Bad param: $param"
9840 run_test 123ab "llog_print params output values from set_param -P"
9842 test_123ac() { # LU-11566
9843 remote_mgs_nodsh && skip "remote MGS with nodsh"
9844 do_facet mgs "$LCTL help llog_print" 2>&1 | grep -q -- --start ||
9845 skip "Need 'lctl llog_print --start' on MGS"
9851 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9852 mgs_arg="--device MGS"
9854 [ -d $MOUNT/.lustre ] || setup
9856 # - { index: 10, event: add_uuid, nid: 192.168.20.1@tcp(0x20000c0a81401,
9857 # node: 192.168.20.1@tcp }
9858 do_facet mgs $LCTL $mgs_arg \
9859 llog_print --start $start --end $end $FSNAME-client | tr -d , |
9860 while read DASH BRACE INDEX idx EVENT BLAH BLAH BLAH; do
9861 (( idx >= start )) || error "llog_print index $idx < $start"
9862 (( idx <= end )) || error "llog_print index $idx > $end"
9865 run_test 123ac "llog_print with --start and --end"
9867 test_123ad() { # LU-11566
9868 remote_mgs_nodsh && skip "remote MGS with nodsh"
9869 # older versions of lctl may not print all records properly
9870 do_facet mgs "$LCTL help llog_print" 2>&1 | grep -q -- --start ||
9871 skip "Need 'lctl llog_print --start' on MGS"
9873 [ -d $MOUNT/.lustre ] || setup
9875 # append a new record, to avoid issues if last record was cancelled
9876 local old=$($LCTL get_param -n osc.*-OST0000-*.max_dirty_mb | head -1)
9877 do_facet mgs $LCTL conf_param $FSNAME-OST0000.osc.max_dirty_mb=$old
9879 # logid: [0x3:0xa:0x0]:0
9883 local num=$(do_facet mgs $LCTL --device MGS llog_info $FSNAME-client |
9884 awk '/last_index:/ { print $2 - 1 }')
9886 # - { index: 71, event: set_timeout, num: 0x14, param: sys.timeout=20 }
9887 local last=$(do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
9888 tail -1 | awk '{ print $4 }' | tr -d , )
9889 (( last == num )) || error "llog_print only showed $last/$num records"
9891 run_test 123ad "llog_print shows all records"
9893 test_123ae() { # LU-11566
9901 remote_mgs_nodsh && skip "remote MGS with nodsh"
9902 [ -d $MOUNT/.lustre ] || setupall
9904 max=$($LCTL get_param -n osc.*-OST0000-*.max_dirty_mb | head -1)
9905 pgs=$($LCTL get_param -n osc.*-OST0000-*.max_pages_per_rpc | head -1)
9906 [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] ||
9907 mgs_arg="--device MGS"
9909 if do_facet mgs "$LCTL help llog_cancel" 2>&1| grep -q -- --log_id; then
9910 # save one set_param -P record in case none exist
9912 do_facet mgs $LCTL set_param -P osc.*.max_pages_per_rpc=$pgs
9913 stack_trap "do_facet mgs $LCTL set_param -P -d \
9914 osc.*.max_pages_per_rpc"
9917 orig=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9918 tail -1 | awk '{ print $4 }' | tr -d , )
9919 do_facet mgs $LCTL set_param -P osc.*.max_dirty_mb=$max
9920 do_facet mgs $LCTL $mgs_arg llog_print $log | tail -1 |
9921 grep "parameter: osc.*.max_dirty_mb" ||
9922 error "new set_param -P wasn't stored in params log"
9924 # - { index: 71, event: set_param, device: general,
9925 # param: osc.*.max_dirty_mb, value: 256 }
9926 id=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9927 tail -1 | awk '{ print $4 }' | tr -d , )
9928 do_facet mgs $LCTL $mgs_arg llog_cancel $log --log_idx=$id
9929 local new=$(do_facet mgs $LCTL $mgs_arg llog_print $log |
9930 tail -1 | awk '{ print $4 }' | tr -d , )
9931 (( new == orig )) ||
9932 error "new llog_cancel now $new, not at $orig records"
9935 # test old positional parameters for a while still
9936 if [ "$MGS_VERSION" -le $(version_code 3.1.53) ]; then
9939 do_facet mgs $LCTL conf_param \
9940 $FSNAME-OST0000.osc.max_pages_per_rpc=$pgs
9941 stack_trap "do_facet mgs $LCTL conf_param -d \
9942 $FSNAME-OST0000.osc.max_pages_per_rpc"
9944 orig=$(do_facet mgs $LCTL --device MGS llog_print $log |
9945 tail -1 | awk '{ print $4 }' | tr -d , )
9946 do_facet mgs $LCTL conf_param $FSNAME-OST0000.osc.max_dirty_mb=$max
9947 do_facet mgs $LCTL --device MGS llog_print $log |
9948 tail -1 | grep "parameter: osc.max_dirty_mb" ||
9949 error "old conf_param wasn't stored in params log"
9950 do_facet mgs $LCTL --device MGS llog_print $log
9951 # - { index: 71, event: conf_param, device: testfs-OST0000-osc,
9952 # param: osc.max_dirty_mb=256 }
9953 id=$(do_facet mgs $LCTL --device MGS llog_print $log |
9954 tail -1 | awk '{ print $4 }' | tr -d , )
9955 do_facet mgs $LCTL --device MGS llog_cancel $log $id
9956 do_facet mgs $LCTL --device MGS llog_print $log
9957 new=$(do_facet mgs $LCTL --device MGS llog_print $log |
9958 tail -1 | awk '{ print $4 }' | tr -d , )
9959 (( new == orig )) ||
9960 error "old llog_cancel now $new, not at $orig records"
9963 run_test 123ae "llog_cancel can cancel requested record"
9965 test_123af() { #LU-13609
9966 [ "$MGS_VERSION" -ge $(version_code 2.13.54) -a \
9967 "$MDS1_VERSION" -ge $(version_code 2.13.54) ] ||
9968 skip "Need both MGS and MDS version at least 2.13.54"
9970 [ -d $MOUNT/.lustre ] || setupall
9971 stack_trap "do_facet mds1 $LCTL set_param fail_loc=0" EXIT
9981 for device in "MGS" "$FSNAME-MDT0000"; do
9982 cmd="--device $device llog_catlist"
9983 echo "lctl $cmd ..."
9984 if [ "$device" = "MGS" ]; then
9989 orig_clist=($(do_facet $facet $LCTL $cmd | awk '{ print $2 }'))
9990 orig_count=${#orig_clist[@]}
9991 echo "orig_clist: ${orig_clist[*]}"
9993 #define OBD_FAIL_CATLIST 0x131b
9994 #fetch to llog records from the second one
9995 do_facet $facet $LCTL set_param fail_loc=0x131b fail_val=2
9997 new_clist=($(do_facet $facet $LCTL $cmd | awk '{ print $2 }'))
9998 new_count=${#new_clist[@]}
9999 echo "new_clist: ${new_clist[*]}"
10001 [ $new_count -eq $((orig_count - 1)) ] ||
10002 error "$new_count != $orig_count - 1"
10003 for i in $(seq 0 $new_count); do
10005 [ "${orig_clist[$j]}" = "${new_clist[$i]}" ] ||
10006 error "${orig_clist[$j]} != ${new_clist[$i]}"
10008 do_facet mds1 $LCTL set_param fail_loc=0
10012 run_test 123af "llog_catlist can show all config files correctly"
10014 test_123ag() { # LU-15142
10018 remote_mgs_nodsh && skip "remote MGS with nodsh"
10019 (( $MGS_VERSION >= $(version_code 2.14.55) )) ||
10020 skip "Need server version least 2.14.55"
10022 [ -d $MOUNT/.lustre ] || setup
10024 orig_val=$(do_facet mgs $LCTL get_param jobid_name)
10025 stack_trap "do_facet mgs $LCTL set_param -P jobid_name=$orig_val"
10027 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME1"
10028 do_facet mgs $LCTL set_param -P -d jobid_name
10029 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
10030 grep -c jobid_name)
10031 (( rec == 0 )) || error "parameter was not deleted, check #1"
10032 do_facet mgs $LCTL set_param -P jobid_name="TESTNAME1"
10033 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
10034 grep -c jobid_name)
10035 (( rec == 1)) || error "parameter is not set"
10036 # usage with ordinary set_param format works too
10037 do_facet mgs $LCTL set_param -P -d jobid_name="ANY"
10038 rec=$(do_facet mgs $LCTL --device MGS llog_print params |
10039 grep -c jobid_name)
10040 (( rec == 0 )) || error "parameter was not deleted, check #2"
10042 run_test 123ag "llog_print skips values deleted by set_param -P -d"
10044 test_123ah() { #LU-7668 del_ost
10045 (( "$MGS_VERSION" >= $(version_code 2.15.51) &&
10046 "$MDS1_VERSION" >= $(version_code 2.15.51) )) ||
10047 skip "Need MGS/MDS version >= 2.15.51 for del_ost"
10049 [ -d $MOUNT/.lustre ] || setupall
10050 stack_trap "do_facet mds1 $LCTL set_param fail_loc=0" EXIT
10052 local cmd="--device MGS llog_print"
10054 cli_llogcnt_orig=$(do_facet mgs $LCTL $cmd $FSNAME-client |
10055 grep -c $FSNAME-OST0000)
10056 mds1_llogcnt_orig=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 |
10057 grep -c $FSNAME-OST0000)
10059 [ $cli_llogcnt_orig -gt 0 ] ||
10060 error "$FSNAME-OST0000 not found (client)"
10061 [ $mds1_llogcnt_orig -gt 0 ] || error "$FSNAME-OST0000 not found (MDT)"
10063 # -n/--dryrun should NOT modify catalog
10064 do_facet mgs "$LCTL del_ost -n -t $FSNAME-OST0000" ||
10065 error "del_ost --dryrun failed with $?"
10067 local cli_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-client |
10068 grep -c $FSNAME-OST0000)
10069 local mds1_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 |
10070 grep -c $FSNAME-OST0000)
10072 [ $cli_llogcnt -eq $cli_llogcnt_orig ] ||
10073 error "--dryrun error: $cli_llogcnt != $cli_llogcnt_orig"
10074 [ $mds1_llogcnt -eq $mds1_llogcnt_orig ] ||
10075 error "--dryrun error: $mds1_llogcnt != $mds1_llogcnt_orig"
10078 do_facet mgs "$LCTL del_ost --target $FSNAME-OST0000" ||
10079 error "del_ost failed with $?"
10081 local cli_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-client |
10082 grep -c $FSNAME-OST0000)
10083 local mds1_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 |
10084 grep -c $FSNAME-OST0000)
10086 # every catalog entry for OST0000 should have been cancelled
10087 [ $cli_llogcnt -eq 0 ] || error "$FSNAME-OST0000 not cancelled (cli)"
10088 [ $mds1_llogcnt -eq 0 ] || error "$FSNAME-OST0000 not cancelled (MDT)"
10090 umount_client $MOUNT
10091 mount_client $MOUNT
10093 $LFS df | grep -q OST0000 && error "del_ost did not remove OST0000!"
10095 reformat_and_config
10097 run_test 123ah "del_ost cancels config log entries correctly"
10102 echo "cleanup test 123ai"
10104 # cancel last timeout record
10105 do_facet mgs "$LCTL set_param -P -d timeout"
10107 # restore timeout value
10108 do_nodes $(comma_list $(all_nodes)) "$LCTL set_param timeout=$timeout"
10111 test_123ai() { #LU-16167
10117 remote_mgs_nodsh && skip "remote MGS with nodsh"
10118 (( MDS1_VERSION >= $(version_code 2.15.51) )) ||
10119 skip "Need MDS version at least 2.15.51"
10121 [ -d $MOUNT/.lustre ] || setup
10123 old_count=$(do_facet mgs "$LCTL --device MGS llog_print -r params" |
10124 grep -c "parameter: timeout")
10126 old_timeout=$($LCTL get_param -n timeout)
10127 stack_trap "cleanup_123ai $old_timeout" EXIT
10129 # add and cancel params (2 * MAX_IOC_BUFLEN / 128 + 1 = 129)
10130 for i in {1..129}; do
10131 do_facet mgs "$LCTL set_param -P timeout=$i" ||
10132 error "fail to set timeout=$i on MGS"
10135 count=$(do_facet mgs "$LCTL --device MGS llog_print -r params" |
10136 grep -c "parameter: timeout")
10138 ((count - old_count == 129)) ||
10139 error "number timeout record missmatch ($count - $old_count != 129)"
10141 do_facet mgs "$LCTL --device MGS llog_print params" |
10142 tail -1 | grep "timeout, value: 129" ||
10143 error "llog_print could not display the last record (timeout=129)"
10146 run_test 123ai "llog_print display all non skipped records"
10149 remote_mgs_nodsh && skip "remote MGS with nodsh"
10151 [ -d $MOUNT/.lustre ] || setup
10152 yaml_file="$TMP/$tfile.yaml"
10154 do_facet mgs rm "$yaml_file"
10155 cfgfiles=$(do_facet mgs "lctl --device MGS llog_catlist" |
10156 sed 's/config_log://')
10158 # set jobid_var to a different value for test
10159 orig_val=$(do_facet mgs $LCTL get_param jobid_var)
10161 do_facet mgs $LCTL set_param -P jobid_var="TESTNAME"
10163 for i in $cfgfiles params; do
10164 do_facet mgs "lctl --device MGS llog_print ${i} >> $yaml_file"
10167 echo "Unmounting FS"
10175 test_123_restore() {
10176 set_val=$(do_facet mgs $LCTL get_param jobid_var)
10178 do_facet mgs $LCTL set_param -P $orig_val
10180 [ $set_val == "jobid_var=TESTNAME" ] ||
10181 error "$set_val is not TESTNAME"
10183 do_facet mgs rm "$yaml_file"
10195 # Reapply the config from before
10196 echo "Setting configuration parameters"
10197 do_facet mgs "lctl set_param -F $yaml_file"
10201 run_test 123F "clear and reset all parameters using set_param -F"
10211 # Reapply the config from before
10212 echo "Setting configuration parameters"
10213 do_facet mgs "lctl apply_yaml $yaml_file"
10217 run_test 123G "clear and reset all parameters using apply_yaml"
10222 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
10223 [ -z $mds2failover_HOST ] && skip "needs MDT failover setup"
10226 do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
10227 grep 1.2.3.4@tcp && error "Should not be fake nid"
10228 do_facet mgs $LCTL conf_param $FSNAME-MDT0001.failover.node=1.2.3.4@tcp\
10229 || error "Set params error"
10230 do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
10231 grep 1.2.3.4@tcp || error "Fake nid should be added"
10235 if combined_mgs_mds; then
10236 start_mdt 1 "-o nosvc" ||
10237 error "starting mds with nosvc option failed"
10239 local nid=$(do_facet mds2 $LCTL list_nids | head -1)
10240 local failover_nid=$(do_node $mds2failover_HOST $LCTL list_nids | head -1)
10242 do_facet mgs $LCTL replace_nids $FSNAME-MDT0001 $nid:$failover_nid ||
10243 error "replace_nids execution error"
10245 if combined_mgs_mds; then
10251 echo "lfs setdirstripe"
10252 $LFS setdirstripe -i 1 $MOUNT/$tdir || error "setdirstirpe error"
10255 run_test 124 "check failover after replace_nids"
10257 get_max_sectors_kb() {
10260 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
10261 local max_sectors_path="/sys/block/${dev_base}/queue/max_sectors_kb"
10263 do_facet ${facet} "[[ -e ${max_sectors_path} ]] &&
10264 cat ${max_sectors_path}"
10267 get_max_hw_sectors_kb() {
10270 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
10271 local max_hw_path="/sys/block/${dev_base}/queue/max_hw_sectors_kb"
10273 do_facet ${facet} "[[ -e ${max_hw_path} ]] && cat ${max_hw_path}"
10276 set_max_sectors_kb() {
10280 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
10281 local max_sectors_path="/sys/block/${dev_base}/queue/max_sectors_kb"
10283 do_facet ${facet} "[[ -e ${max_sectors_path} ]] &&
10284 echo ${value} > ${max_sectors_path}"
10287 [[ $rc -ne 0 ]] && echo "Failed to set ${max_sectors_path} to ${value}"
10292 # Return 0 if all slave devices have max_sectors_kb == max_hw_sectors_kb
10293 # Otherwise return > 0
10294 check_slaves_max_sectors_kb()
10298 local dev_base=$(basename $(do_facet ${facet} "readlink -f ${device}"))
10299 local slaves_dir=/sys/block/${dev_base}/slaves
10300 local slave_devices=$(do_facet ${facet} "ls ${slaves_dir} 2>/dev/null")
10301 [[ -z ${slave_devices} ]] && return 0
10303 local slave max_sectors new_max_sectors max_hw_sectors path
10306 for slave in ${slave_devices}; do
10307 path="/dev/${slave}"
10308 ! is_blkdev ${facet} ${path} && continue
10309 max_sectors=$(get_max_sectors_kb ${facet} ${path})
10310 max_hw_sectors=$(get_max_hw_sectors_kb ${facet} ${path})
10311 new_max_sectors=${max_hw_sectors}
10312 [[ ${new_max_sectors} -gt ${RQ_SIZE_LIMIT} ]] &&
10313 new_max_sectors=${RQ_SIZE_LIMIT}
10315 if [[ ${max_sectors} -ne ${new_max_sectors} ]]; then
10316 echo "${path} ${max_sectors} ${new_max_sectors}"
10319 check_slaves_max_sectors_kb ${facet} ${path}
10328 local facet_list="mgs mds1 ost1"
10329 combined_mgs_mds && facet_list="mgs ost1"
10332 for facet in ${facet_list}; do
10333 [[ $(facet_fstype ${facet}) != ldiskfs ]] &&
10334 skip "ldiskfs only test" &&
10336 ! is_blkdev ${facet} $(facet_device ${facet}) &&
10337 skip "requires all real devices" &&
10342 # We don't increase IO request size limit past 16MB. See comments in
10343 # lustre/utils/libmount_utils_ldiskfs.c:tune_max_sectors_kb()
10344 RQ_SIZE_LIMIT=$((16 * 1024))
10345 local device old_max_sectors new_max_sectors max_hw_sectors
10346 for facet in ${facet_list}; do
10347 device=$(facet_device ${facet})
10348 old_max_sectors=$(get_max_sectors_kb ${facet} ${device})
10349 max_hw_sectors=$(get_max_hw_sectors_kb ${facet} ${device})
10351 # The expected value after l_tunedisk is executed
10352 new_max_sectors=$old_max_sectors
10353 [[ ${new_max_sectors_kb} -gt ${RQ_SIZE_LIMIT} ]] &&
10354 new_max_sectors_kb=${RQ_SIZE_LIMIT}
10356 # Ensure the current value of max_sectors_kb does not equal
10357 # max_hw_sectors_kb, so we can tell whether l_tunedisk did
10359 set_max_sectors_kb ${facet} ${device} $((new_max_sectors - 1))
10361 # Value before l_tunedisk
10362 local pre_max_sectors=$(get_max_sectors_kb ${facet} ${device})
10363 if [[ ${pre_max_sectors} -ne $((new_max_sectors - 1)) ]]; then
10364 echo "unable to satsify test pre-condition:"
10365 echo "${pre_max_sectors} != $((new_max_sectors - 1))"
10370 echo "Before: ${facet} ${device} ${pre_max_sectors} ${max_hw_sectors}"
10372 do_facet ${facet} "l_tunedisk ${device}"
10374 # Value after l_tunedisk
10375 local post_max_sectors=$(get_max_sectors_kb ${facet} ${device})
10377 echo "After: ${facet} ${device} ${post_max_sectors} ${max_hw_sectors}"
10379 if [[ ${facet} != ost1 ]]; then
10380 if [[ ${post_max_sectors} -ne ${pre_max_sectors} ]]; then
10381 echo "l_tunedisk modified max_sectors_kb of ${facet}"
10385 set_max_sectors_kb ${facet} ${device} ${old_max_sectors}
10387 if [[ ${post_max_sectors} -eq ${pre_max_sectors} ]]; then
10388 echo "l_tunedisk failed to modify max_sectors_kb of ${facet}"
10392 check_slaves_max_sectors_kb ${facet} ${device} ||
10399 run_test 125 "check l_tunedisk only tunes OSTs and their slave devices"
10402 [[ "$MDS1_VERSION" -ge $(version_code 2.13.52) ]] ||
10403 skip "Need MDS version at least 2.13.52"
10406 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ../libcfs/libcfs/libcfs
10407 #define OBD_FAIL_OBD_SETUP 0x60d
10408 do_facet mds1 $LCTL set_param fail_loc=0x60d
10409 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_modules &
10410 for i in {1..40}; do
10411 do_facet mds1 lsmod | grep -q osd_$mds1_FSTYPE && break
10414 clear_failloc $SINGLEMDS 20 &
10415 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
10417 run_test 126 "mount in parallel shouldn't cause a crash"
10420 [[ "$ost1_FSTYPE" == ldiskfs ]] || skip "ldiskfs only test"
10424 zconf_umount_clients $RCLIENTS $MOUNT
10426 wait_osp_active ost ${FSNAME}-OST0000 0 1
10427 local osc_tgt="$FSNAME-OST0000-osc-$($LFS getname -i $DIR)"
10428 local avail1=($($LCTL get_param -n osc.${osc_tgt}.kbytesavail))
10430 $LFS setstripe -i 0 $DIR/$tfile || error "failed creating $DIR/$tfile"
10431 dd if=/dev/zero of=$DIR/$tfile bs=1M oflag=direct || true
10433 local avail2=($($LCTL get_param -n osc.${osc_tgt}.kbytesavail))
10435 if ((avail2 * 100 / avail1 > 1)); then
10438 error "more than 1% space left: before=$avail1 after=$avail2"
10441 local mbs=$(($(stat -c %s $DIR/$tfile) / (1024 * 1024)))
10443 dd if=/dev/zero of=$DIR/$tfile bs=1M count=$mbs conv=notrunc \
10444 oflag=direct || error "overwrite failed"
10446 run_test 127 "direct io overwrite on full ost"
10450 combined_mgs_mds && skip "need separate mgs device"
10451 [ "$ost2_FSTYPE" == zfs ] && import_zpool ost2
10454 # Try to apply nolocallogs to the virgin OST. Should fail.
10455 do_facet ost2 "$TUNEFS --nolocallogs $(ostdevname 2)" &&
10456 error "nolocallogs should not be allowed on the virgin target"
10461 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
10462 # Start OST without MGS (local configs)
10463 do_facet ost1 "$TUNEFS --dryrun $(ostdevname 1)"
10464 start_ost || error "unable to start OST1"
10465 stop_ost || error "Unable to stop OST1"
10467 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
10468 # Do not allow reading local configs, should fail
10469 do_facet ost1 "$TUNEFS --nolocallogs $(ostdevname 1)" ||
10470 error "Can not set nolocallogs"
10471 start_ost && error "OST1 started, but should fail"
10473 # Connect to MGS successfully, reset nolocallogs flag
10474 [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1
10475 start_mgs || error "unable to start MGS"
10476 start_ost || error "unable to start OST1"
10478 do_facet ost1 "$TUNEFS --dryrun $(ostdevname 1)" | grep "nolocallogs" &&
10479 error "nolocallogs expected to be reset"
10481 stop_ost || error "Unable to stop OST1"
10483 run_test 128 "Force using remote logs with --nolocallogs"
10487 (( MDS1_VERSION >= $(version_code 2.14.57) )) ||
10488 skip "Need MDS version at least 2.14.57"
10490 start_mgsmds || error "MGS/MDS start failed"
10492 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &&
10493 error "start ost1 should fail" || true
10494 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &&
10495 error "second start ost1 should fail" || true
10496 do_facet ost1 "$TUNEFS --writeconf $(ostdevname 1)"
10497 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS ||
10498 error "start ost1 failed"
10502 run_test 129 "attempt to connect an OST with the same index should fail"
10506 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
10508 stop_mdt 2 || error "mdt2 stop failed"
10509 do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)"
10510 start_mdt 2 || error "mdt2 start failed"
10511 do_facet mds2 "$LCTL dl" | grep MDT0001-osp-MDT0001 &&
10512 error "Illegal OSP device created" || true
10514 run_test 130 "re-register an MDT after writeconf"
10517 [ "$mds1_FSTYPE" == "ldiskfs" ] || skip "ldiskfs only test"
10518 (( $MDS1_VERSION >= $(version_code 2.14.56.35) )) ||
10519 skip "Need MDS version at least 2.14.56.35"
10520 do_facet mds1 $DEBUGFS -R features $(mdsdevname 1) |
10521 grep -q project || skip "skip project quota not supported"
10526 test_mkdir -c $MDSCOUNT -p $DIR/$tdir
10527 $LFS project -p 1000 $DIR/$tdir || error "set dir project id failed"
10528 createmany -o $DIR/$tdir/f 512
10529 for ((i = 0; i < 512; ++i)); do
10530 $LFS project -p $i $DIR/$tdir/f${i} ||
10531 error "set f${i} project id failed"
10534 test_mkdir -c $MDSCOUNT -p $DIR/$tdir.inherit
10535 $LFS project -p 1001 -s $DIR/$tdir.inherit
10536 createmany -o $DIR/$tdir.inherit/f 128
10537 (( $($LFS project $DIR/$tdir.inherit/f* |
10538 awk '$1 == 1001 { print }' | wc -l) == 128 )) ||
10539 error "files did not inherit projid 1001"
10543 for i in $(seq $MDSCOUNT); do
10544 mds_backup_restore mds$i ||
10545 error "Backup/restore on mds$i failed"
10550 projid=($($LFS project -d $DIR/$tdir))
10551 [ ${projid[0]} == "1000" ] ||
10552 error "projid expected 1000 not ${projid[0]}"
10553 for ((i = 0; i < 512; ++i)); do
10554 projid=($($LFS project $DIR/$tdir/f${i}))
10555 [ ${projid[0]} == "$i" ] ||
10556 error "projid expected $i not ${projid[0]}"
10559 (( $($LFS project $DIR/$tdir.inherit/f* |
10560 awk '$1 == 1001 { print }' | wc -l) == 128 )) ||
10561 error "restore did not copy projid 1001"
10563 run_test 131 "MDT backup restore with project ID"
10569 (( MDS1_VERSION >= $(version_code 2.14.57) )) ||
10570 skip "Need MDS version at least 2.14.57"
10572 combined_mgs_mds || start_mgs || error "unable to start MGS"
10573 start_mdt 1 || error "unable to start mdt1"
10575 err_cnt=$(do_facet mds1 dmesg | grep -c "cannot take the layout locks")
10576 stop_mdt 1 || error "stop mdt1 failed"
10578 [ "$mds1_FSTYPE" == zfs ] && import_zpool mds1
10579 do_facet mds1 $TUNEFS --param mdt.hsm_control=enabled $(mdsdevname 1) ||
10580 error "tunefs failed"
10581 start_mdt 1 || error "cannot start mdt1"
10583 err_cnt2=$(do_facet mds1 dmesg | grep -c "cannot take the layout locks")
10584 [ $err_cnt -eq $err_cnt2 ] || error "Can not take the layout lock"
10585 stop_mdt 1 || error "stop mdt1 failed"
10587 run_test 132 "hsm_actions processed after failover"
10589 # This test verifies we do RR allocation within a pool even if there is a
10590 # significant imbalance vs an OST outside the pool
10592 [[ $OSTCOUNT -lt 4 ]] && skip_env "needs >= 4 OSTs"
10593 [[ "$OST1_VERSION" -ge $(version_code 2.15.51) ]] ||
10594 skip "Need OST version at least 2.15.51"
10595 # This is the easiest way to ensure OSTs start out balanced
10596 reformat_and_config
10599 check_set_fallocate_or_skip
10601 local testfile=$DIR/$tdir/$tfile
10602 local pool="testpool"
10603 local ostrange=$((OSTCOUNT - 1))
10604 # Select all but the last OST to add to the pool
10605 local poolostrange=$((OSTCOUNT - 2))
10613 [ $MINV -eq 0 ] && error "no free space in OST$MINI"
10614 [ $MAXV -gt $((2 * $MINV)) ] &&
10615 error "OSTs badly unbalanced after reformat"
10617 create_pool $FSNAME.$pool || error "failed to create a pool"
10618 do_facet mgs $LCTL pool_add $FSNAME.$pool OST[0-$poolostrange] ||
10619 error "failed to add OST[0-$poolostrange] to the pool"
10621 test_mkdir -p $DIR/$tdir || error "failed to mkdir $DIR/$tdir"
10622 # Consume space on the OSTs in the pool so they are unbalanced with the
10623 # OST outside of the pool
10624 # fill each OST 90% with fallocate so they are widely
10626 local size=$(((MINV * 9 / 10) * 1024))
10627 for ((i = 0; i <= poolostrange; i++)); do
10628 $LFS setstripe -c 1 -i $i $testfile$i ||
10629 error "failed to setstripe $testfile$i"
10630 fallocate -l $size $testfile$i || error "fallocate failed"
10636 # Create files in the pool now that there is an imbalance
10637 filecount=$(((OSTCOUNT - 1) * filenum))
10638 for ((i = 0; i < filecount; i++)); do
10639 $LFS setstripe -p $pool $testfile-$i ||
10640 error "failed to setstripe -p $pool $testfile-$i"
10642 $LFS getstripe -i $testfile-* > /tmp/$tfile.log
10643 # Count the number of files with a stripe on each OST to verify the
10644 # pool allocated with round-robin
10645 for ((i = 0; i <= poolostrange; i++)); do
10646 stripecount=$(grep -c $i /tmp/$tfile.log)
10647 # Allow a little leeway
10648 if (( stripecount < filenum - 1 ||
10649 stripecount > filenum + 1 )); then
10650 cat /tmp/$tfile.log
10651 error "$stripecount != $filenum files on OST$i"
10655 # Create files across the system now that there is an imbalance
10656 filecount=$((OSTCOUNT * filenum))
10657 for ((i = 1; i < filecount; i++)); do
10658 $LFS setstripe $testfile-$i.2 ||
10659 error "failed to setstripe $testilfe-$i.2"
10661 $LFS getstripe -i $testfile-*.2 > /tmp/$tfile.log
10663 # Count the number of files with a stripe on each OST to verify the
10664 # files are *NOT* allocated with round-robin
10665 for ((i = 0; i <= ostrange; i++)); do
10666 stripecount=$(grep -c $i /tmp/$tfile.log)
10667 if [[ $stripecount -ne $filenum ]]; then
10669 echo "QOS: $stripecount != $filenum files on OST$i"
10672 if [ -z "$qos_used" ]; then
10673 error "QOS not used on imbalanced OSTs!"
10676 rm -rf /tmp/$tfile.log $DIR/$tdir
10677 do_facet mgs $LCTL pool_remove $FSNAME.$pool OST[0-$poolostrange] ||
10678 "failed to remove OST[0-$poolostrange] from the pool"
10679 do_facet mgs $LCTL pool_destroy $FSNAME.$pool ||
10680 error "failed to destroy pool"
10682 run_test 133 "stripe QOS: free space balance in a pool"
10685 [ "$mds1_FSTYPE" == "ldiskfs" ] || skip "ldiskfs only test"
10688 local mdt_dev=$(facet_device mds1)
10689 local tmp_dir=$TMP/$tdir
10690 local out=$tmp_dir/check_iam.txt
10691 local CHECK_IAM=${CHECK_IAM:-$(do_facet mds1 "which check_iam 2> /dev/null || true")}
10693 [[ -n "$CHECK_IAM" ]] || skip "check_iam not found"
10696 for ((i=0; i<64; i++)); do
10698 #cmd introduce a random corruption to IAM file
10699 local cmd="dd if=/dev/urandom of=$tmp_dir/$f bs=2 conv=notrunc count=1 seek=$((RANDOM % 36))"
10700 do_facet mds1 "mkdir -p $tmp_dir; \
10701 $DEBUGFS -c -R 'dump $f $tmp_dir/$f' $mdt_dev 2>&1; \
10702 $CHECK_IAM -v $tmp_dir/$f 2>&1; \
10703 echo $cmd; eval $cmd 2>/dev/null;
10704 $CHECK_IAM -v $tmp_dir/$f 2>&1; echo \\\$?" >> $out 2>&1
10709 stack_trap "rm -rf $tmp_dir && do_facet mds1 rm -rf $tmp_dir" EXIT
10711 rc=$(grep -c "fault\|except" $out)
10712 (( rc == 0 )) || { cat $out &&
10713 error "check_iam failed with fault or exception $rc"; }
10715 rc=$(grep -c "^255" $out)
10716 errors=$(grep -c "FINISHED WITH ERRORS" $out)
10718 (( rc == errors )) || { cat $out &&
10719 error "check_iam errcode does not fit with errors $rc $errors"; }
10721 run_test 134 "check_iam works without faults"
10723 cleanup_test_135(){
10726 printf "\nCleanup test_135\n" >&2
10727 do_facet mds1 "$LCTL set_param -n $oldgc"
10728 rm -rf $DIR/$tdir &> /dev/null
10732 __test_135_file_thread() {
10734 local init_time=$(awk '{print $1}' /proc/uptime)
10735 local awkcmd="/crosses index zero/ {if (\$1 > $init_time) exit(1);}"
10737 #Generate a full plain llogs
10738 while dmesg | sed -r 's/(\[|\])//g' | awk "$awkcmd" ; do
10739 createmany -o $DIR/$tdir/f 4500 >&2
10740 createmany -u $DIR/$tdir/f 4500 >&2
10744 __test_135_reader() {
10747 local firstidx=$(changelog_user_rec mds1 $cl_user)
10748 local oldidx=$firstidx
10753 while read -t10 -u$fd newidx other; do
10754 (( (newidx - oldidx) == 1 )) ||
10755 error "changelog jump detected (last: $oldidx, current: $newidx)"
10757 if (( (newidx - firstidx + 1) % 13000 == 0 )); then
10758 [[ $pid -eq 0 ]] ||
10759 wait $pid || error "changelog_clear failed"
10760 changelog_clear $((newidx - 1)) mds1 >&2 & pid=$!
10765 [[ $pid -eq 0 ]] ||
10766 wait $pid || error "changelog_clear failed"
10772 (( MDS1_VERSION >= $(version_code 2.15.52) )) ||
10773 skip "need MDS version at least 2.15.52"
10775 local service=$(facet_svc mds1)
10777 local lastread lastidx
10778 local files_pid reader_pid
10783 # Need to reformat because we are changing llog catalog sizes to 5.
10784 # Otherwise, processing could fail with existing catalogs (last_idx>5).
10788 # Disable changelog garbage colector
10789 local oldgc=$(do_facet mds1 "$LCTL get_param mdd.${service}.changelog_gc")
10790 do_facet mds1 "$LCTL set_param -n mdd.${service}.changelog_gc=0"
10791 stack_trap "cleanup_test_135 $oldgc" EXIT INT
10793 # change the changelog_catalog size to 5 entries for everybody
10794 #define OBD_FAIL_CAT_RECORDS 0x1312
10795 do_node $(comma_list $(all_nodes)) $LCTL set_param fail_loc=0x1312 fail_val=5
10797 # disable console ratelimit
10798 local rl=$(cat /sys/module/libcfs/parameters/libcfs_console_ratelimit)
10799 echo 0 > /sys/module/libcfs/parameters/libcfs_console_ratelimit
10800 stack_trap "echo $rl > /sys/module/libcfs/parameters/libcfs_console_ratelimit" EXIT
10802 test_mkdir -c 1 -i 0 $DIR/$tdir || error "Failed to create directory"
10803 changelog_chmask "ALL" || error "changelog_chmask failed"
10804 changelog_register || error "changelog_register failed"
10806 cl_user="${CL_USERS[mds1]%% *}"
10807 changelog_users mds1 | grep -q $cl_user ||
10808 error "User $cl_user not found in changelog_users"
10810 # Start reader thread
10811 coproc $LFS changelog --follow $service
10814 stack_trap "echo kill changelog reader; kill $reader_pid" EXIT
10816 echo -e "\nWrap arround changelog catalog"
10818 # Start file writer thread
10819 __test_135_file_thread "$service" & files_pid=$!
10820 stack_trap "(pkill -P$files_pid; kill $files_pid) &> /dev/null || true" EXIT
10822 # Check changelog entries
10823 lastread=$(__test_135_reader $fd $cl_user) || exit $?
10824 ! kill -0 $files_pid 2>/dev/null ||
10825 error "creation thread is running. Is changelog reader stuck?"
10827 lastidx=$(changelog_users mds1 | awk '/current_index/ {print $NF}' )
10828 [[ "$lastread" -eq "$lastidx" ]] ||
10829 error "invalid changelog lastidx (read: $lastread, mds: $lastidx)"
10831 run_test 135 "check the behavior when changelog is wrapped around"
10834 do_facet mds2 "$LCTL --device ec cleanup" || true
10835 do_facet mds2 "$LCTL --device ec detach" || true
10838 reformat_and_config
10842 (( MDSCOUNT >= 2 )) || skip "needs >= 2 MDTs"
10847 do_facet mds2 "$LCTL attach echo_client ec ec_uuid" ||
10848 error "echo attach fail"
10850 stack_trap cleanup_136 EXIT
10852 do_facet mds2 "$LCTL --device ec setup lustre-MDT0001 mdt" &&
10853 error "attach to MDT should fail!"
10855 do_facet mds2 "$LCTL --device ec setup lustre-MDT0001 mdd" ||
10856 error "attach to MDD should OK"
10857 do_facet mds2 "$LCTL --device ec test_mkdir /tt" &&
10858 error "mkdir test should fail with remote object"
10862 run_test 136 "don't panic with bad obdecho setup"
10865 (( MDS1_VERSION >= $(version_code 2.15.55) )) ||
10866 skip "need MDS version at least 2.15.55"
10867 (( MDSCOUNT >= 2 )) || skip "needs >= 2 MDTs"
10869 RM_UPDATELOG=$(do_facet mds2 "which remove_updatelog 2> /dev/null")
10870 RM_UPDATELOG=${RM_UPDATELOG:-"$LUSTRE/scripts/remove_updatelog"}
10872 [ -f "$RM_UPDATELOG" ] ||
10873 skip_env "remove_updatelog is not found on mds2"
10875 local mntpt=$(facet_mntpt mds2)
10878 test_mkdir -c $MDSCOUNT -p $DIR/$tdir ||
10879 error "mkdir $DIR/$tdir failed"
10880 mkdir $DIR/$tdir/d{1..256}
10885 mount_fstype mds2 || error "mount as fstype mds2 failed"
10886 do_facet mds2 $RM_UPDATELOG -n $mntpt
10887 MDTS=($(do_facet mds2 $RM_UPDATELOG -n $mntpt |
10888 grep -o "Processing MDT[0-9]*" | awk -F'MDT' '{print $2}'))
10889 (( ${#MDTS[@]} == MDSCOUNT )) ||
10890 error "Processed ${#MDTS[@]} from $MDSCOUNT"
10892 do_facet mds2 $RM_UPDATELOG -n -m 1,0 $mntpt
10893 MDTS=($(do_facet mds2 $RM_UPDATELOG -n -m 1,0 $mntpt |
10894 grep -o "Processing MDT[0-9]*" | awk -F'MDT' '{print $2}'))
10895 (( ${#MDTS[@]} == 2 )) ||
10896 error "Processed ${#MDTS[@]} instead of 2"
10897 (( ${MDTS[0]} == 1 && ${MDTS[1]} == 0 )) ||
10898 error "Processed: ${MDTS[*]}, expected: 1 0"
10900 do_facet mds2 $RM_UPDATELOG -m 0 $mntpt
10901 unmount_fstype mds2
10902 start_mdt 2 || error "mds2 start fail"
10903 start_mdt 1 || error "mds1 start fail"
10904 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds1 FULL
10905 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
10907 rm -rf $DIR/$tdir || error "Can't remove $tdir"
10909 reformat_and_config
10911 run_test 140 "remove_updatelog script actions"
10916 local max_cached_mb=$($LCTL get_param llite.*.max_cached_mb |
10917 awk '/^max_cached_mb/ { print $2 }')
10918 stack_trap "$LCTL set_param -n llite.*.max_cached_mb=$max_cached_mb"
10920 $LCTL set_param llite.*.max_cached_mb='100%'
10922 local new_max_cached_mb=$($LCTL get_param llite.*.max_cached_mb |
10923 awk '/^max_cached_mb/ { print $2 }')
10924 local total_ram_mb=$(free -m | grep 'Mem:' | awk '{print $2}')
10926 $LCTL get_param llite.*.max_cached_mb
10927 echo "total ram mb: $total_ram_mb"
10928 (( new_max_cached_mb == total_ram_mb )) ||
10929 error "setting cache to 100% not equal to total RAM"
10931 $LCTL set_param llite.*.max_cached_mb='50%'
10932 new_max_cached_mb=$($LCTL get_param llite.*.max_cached_mb |
10933 awk '/^max_cached_mb/ { print $2 }')
10935 $LCTL get_param llite.*.max_cached_mb
10936 (( new_max_cached_mb == $((total_ram_mb / 2)) )) ||
10937 error "setting cache to 50% not equal to 50% of RAM"
10939 $LCTL set_param llite.*.max_cached_mb='105%' &&
10940 error "should not be able to set insane value"
10942 $LCTL set_param llite.*.max_cached_mb='0%'
10943 new_max_cached_mb=$($LCTL get_param llite.*.max_cached_mb |
10944 awk '/^max_cached_mb/ { print $2 }')
10945 # Minimum cache size is 64 MiB
10946 $LCTL get_param llite.*.max_cached_mb
10947 (( new_max_cached_mb == 64 )) ||
10948 error "setting cache to 0% != minimum cache size"
10950 run_test 150 "test setting max_cached_mb to a %"
10953 (( MDS1_VERSION >= $(version_code 2.15.58) )) ||
10954 skip "need MDS version at least 2.15.58"
10955 [[ "$ost1_FSTYPE" == ldiskfs ]] || skip "ldiskfs only test"
10958 if ! combined_mgs_mds ; then
10962 echo "Damage ost1 local config log"
10963 do_facet ost1 "$DEBUGFS -w -R 'punch CONFIGS/$FSNAME-OST0000 0 1' \
10964 $(ostdevname 1) || return \$?" ||
10965 error "do_facet ost1 failed with $?"
10967 # expect OST to fail mount with no MGS and bad local config
10968 start_ost && error "OST start should fail"
10970 if ! combined_mgs_mds ; then
10973 start_mds || error "MDS start failed"
10974 # now it should start with MGS config
10975 start_ost || error "OST start failed"
10976 reformat_and_config
10978 run_test 151 "damaged local config doesn't prevent mounting"
10981 (( MDS1_VERSION >= $(version_code 2.15.59.53) )) ||
10982 skip "need MDS >= 2.15.59.53 for sequence allocation retry"
10983 (( MDSCOUNT >= 2 )) || skip "needs >= 2 MDTs"
10984 local tf=$DIR/$tdir/$tfile
10985 local nost=$((OSTCOUNT+1))
10986 local nostdevname=$(ostdevname $nost)
10989 test_mkdir -i 1 -c1 $DIR/$tdir || error "can't mkdir"
10992 add ost$nost $(mkfs_opts ost1 $nostdevname) --index=$nost \
10993 --reformat $nostdevname $(ostvdevname $nost)
10994 [[ -d "$nostdevname" ]] || stack_trap "do_facet mds1 rm -f $nostdevname"
10996 #define OBD_FAIL_OPS_FAIL_SEQ_ALLOC 0x2109
10997 do_facet mds1 $LCTL set_param fail_loc=0x80002109 fail_val=2
10998 echo "START OST$nost"
10999 stack_trap "stop ost$nost"
11000 start ost$nost $nostdevname $OST_MOUNT_OPTS &
11004 $LFS setstripe -c -1 $tf &
11008 log "STOP OST$nost"
11009 # probably mount hasn't completed yet, so stop races with it
11012 jobs -pr | grep -E "^$PID\$" && sleep 0.5 && continue
11017 do_facet mds1 $LCTL set_param fail_loc=0
11018 log "START OST$nost again"
11019 start ost$nost $nostdevname $OST_MOUNT_OPTS ||
11020 error "can't start ost$nost"
11022 $LFS setstripe -c -1 $tf-2 || error "can't touch $tf-2"
11023 $LFS getstripe -v $tf-2
11024 local stripes=$($LFS getstripe -c $tf-2)
11025 (( stripes == $nost )) || error "$tf-2 $stripes != $nost"
11027 run_test 152 "seq allocation error in OSP"
11030 reformat_and_config
11032 start_mds || error "MDS start failed"
11033 start_ost || error "OST start failed"
11035 local nid=$($LCTL list_nids | grep ${NETTYPE} | head -n1)
11036 local net=${nid#*@}
11037 local MGS_NID=$(do_facet mgs $LCTL list_nids | head -1)
11038 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
11039 local FAKE_PNID="192.168.252.112@${net}"
11040 local FAKE_NIDS="${FAKE_PNID},${FAKE_PNID}2"
11041 local FAKE_FAILOVER="10.252.252.113@${net},10.252.252.113@${net}2"
11042 local NIDS_AND_FAILOVER="$FAKE_NIDS:$FAKE_FAILOVER:$OST1_NID:$MGS_NID"
11047 mount -t lustre $NIDS_AND_FAILOVER:/lustre $MOUNT &
11049 while (( period < 30 )); do
11050 [[ -n "$(ps -p $pid -o pid=)" ]] || break
11051 echo "waiting for mount ..."
11053 period=$((period + 5))
11055 $LCTL get_param mgc.MGC${FAKE_PNID}.import | grep "uptodate:"
11056 check_mount || error "check_mount failed"
11058 cleanup || error "cleanup failed with rc $?"
11060 run_test 153a "bypass invalid NIDs quickly"
11063 # (This was sanity/802a)
11065 saved_MGS_MOUNT_OPTS=$MGS_MOUNT_OPTS
11066 saved_MDS_MOUNT_OPTS=$MDS_MOUNT_OPTS
11067 saved_OST_MOUNT_OPTS=$OST_MOUNT_OPTS
11068 saved_MOUNT_OPTS=$MOUNT_OPTS
11072 MGS_MOUNT_OPTS=$saved_MGS_MOUNT_OPTS
11073 MDS_MOUNT_OPTS=$saved_MDS_MOUNT_OPTS
11074 OST_MOUNT_OPTS=$saved_OST_MOUNT_OPTS
11075 MOUNT_OPTS=$saved_MOUNT_OPTS
11080 [[ $mds1_FSTYPE = zfs ]] || skip "ZFS specific test"
11081 [[ $MDS1_VERSION -lt $(version_code 2.9.55) ]] ||
11082 [[ $OST1_VERSION -lt $(version_code 2.9.55) ]] &&
11083 skip "Need server version at least 2.9.55"
11085 [[ $ENABLE_QUOTA ]] && skip "Quota enabled for read-only test"
11087 # Reset before starting
11091 mkdir $DIR/$tdir || error "(1) fail to mkdir"
11093 cp $LUSTRE/tests/test-framework.sh $DIR/$tdir/ ||
11094 error "(2) Fail to copy"
11096 stack_trap cleanup_802a EXIT
11098 # sync by force before remount as readonly
11099 sync; sync_all_data; sleep 3; sync_all_data
11103 MGS_MOUNT_OPTS=$(csa_add "$MGS_MOUNT_OPTS" -o rdonly_dev)
11104 MDS_MOUNT_OPTS=$(csa_add "$MDS_MOUNT_OPTS" -o rdonly_dev)
11105 OST_MOUNT_OPTS=$(csa_add "$OST_MOUNT_OPTS" -o rdonly_dev)
11107 echo "Mount the server as read only"
11108 setupall server_only || error "(3) Fail to start servers"
11110 echo "Mount client without ro should fail"
11111 mount_client $MOUNT &&
11112 error "(4) Mount client without 'ro' should fail"
11114 echo "Mount client with ro should succeed"
11115 MOUNT_OPTS=$(csa_add "$MOUNT_OPTS" -o ro)
11116 mount_client $MOUNT ||
11117 error "(5) Mount client with 'ro' should succeed"
11119 echo "Modify should be refused"
11120 touch $DIR/$tdir/guard && error "(6) Touch should fail under ro mode"
11122 echo "Read should be allowed"
11123 diff $LUSTRE/tests/test-framework.sh $DIR/$tdir/test-framework.sh ||
11124 error "(7) Read should succeed under ro mode"
11126 run_test 802a "simulate readonly device"
11128 if ! combined_mgs_mds ; then
11134 # restore the values of MDSSIZE and OSTSIZE
11135 MDSSIZE=$STORED_MDSSIZE
11136 OSTSIZE=$STORED_OSTSIZE
11139 complete_test $SECONDS
11140 check_and_cleanup_lustre