7 # bug number for skipped test: LU-7428
8 ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT 84"
9 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
13 if [ -r /etc/SuSE-release ]
15 local vers=$(grep VERSION /etc/SuSE-release | awk '{print $3}')
16 local patchlev=$(grep PATCHLEVEL /etc/SuSE-release |
18 if [ $vers -eq 11 ] && [ $patchlev -eq 2 ]
26 if is_sles11; then # LU-2181
27 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 23a 34b"
30 if [ "$FAILURE_MODE" = "HARD" ]; then
31 CONFIG_EXCEPTIONS="24a " &&
32 echo "Except the tests: $CONFIG_EXCEPTIONS for " \
33 "FAILURE_MODE=$FAILURE_MODE, b=23573" &&
34 ALWAYS_EXCEPT="$ALWAYS_EXCEPT $CONFIG_EXCEPTIONS"
37 # bug number for skipped test:
38 # a tool to create lustre filesystem images
39 ALWAYS_EXCEPT="32newtarball $ALWAYS_EXCEPT"
42 PATH=$PWD/$SRCDIR:$SRCDIR:$SRCDIR/../utils:$PATH
44 PTLDEBUG=${PTLDEBUG:--1}
46 LUSTRE=${LUSTRE:-$(dirname $0)/..}
47 RLUSTRE=${RLUSTRE:-$LUSTRE}
48 export MULTIOP=${MULTIOP:-multiop}
50 . $LUSTRE/tests/test-framework.sh
52 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
54 # use small MDS + OST size to speed formatting time
55 # do not use too small MDSSIZE/OSTSIZE, which affect the default journal size
56 # STORED_MDSSIZE is used in test_18
57 STORED_MDSSIZE=$MDSSIZE
58 STORED_OSTSIZE=$OSTSIZE
70 if ! combined_mgs_mds; then
71 # bug number for skipped test: 23954
72 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 24b"
75 # pass "-E lazy_itable_init" to mke2fs to speed up the formatting time
76 if [[ "$LDISKFS_MKFS_OPTS" != *lazy_itable_init* ]]; then
77 LDISKFS_MKFS_OPTS=$(csa_add "$LDISKFS_MKFS_OPTS" -E lazy_itable_init)
80 [ $(facet_fstype $SINGLEMDS) = "zfs" ] &&
81 # bug number for skipped test: LU-4444
82 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 69"
87 require_dsh_mds || exit 0
88 require_dsh_ost || exit 0
90 [ "$SLOW" = "no" ] && EXCEPT_SLOW="30a 31 45 69"
95 # The MGS must be started before the OSTs for a new fs, so start
96 # and stop to generate the startup logs.
99 wait_osc_import_state mds ost FULL
104 reformat_and_config() {
106 if ! combined_mgs_mds ; then
112 writeconf_or_reformat() {
113 # There are at most 2 OSTs for write_conf test
114 # who knows if/where $TUNEFS is installed?
115 # Better reformat if it fails...
116 writeconf_all $MDSCOUNT 2 ||
117 { echo "tunefs failed, reformatting instead" &&
118 reformat_and_config && return 0; }
128 start mgs $(mgsdevname) $MGS_MOUNT_OPTS
134 local dev=$(mdsdevname $num)
137 echo "start mds service on `facet_active_host $facet`"
138 start $facet ${dev} $MDS_MOUNT_OPTS $@ || return 94
144 local dev=$(mdsdevname $num)
147 echo "stop mds service on `facet_active_host $facet`"
148 # These tests all use non-failover stop
149 stop $facet -f || return 97
155 for num in $(seq $MDSCOUNT); do
156 start_mdt $num $@ || return 94
161 if ! combined_mgs_mds ; then
169 for num in $(seq $MDSCOUNT); do
170 stop_mdt $num || return 97
175 echo "stop mgs service on `facet_active_host mgs`"
176 # These tests all use non-failover stop
177 stop mgs -f || return 97
181 echo "start ost1 service on `facet_active_host ost1`"
182 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS $@ || return 95
186 echo "stop ost1 service on `facet_active_host ost1`"
187 # These tests all use non-failover stop
188 stop ost1 -f || return 98
192 echo "start ost2 service on `facet_active_host ost2`"
193 start ost2 $(ostdevname 2) $OST_MOUNT_OPTS $@ || return 92
197 echo "stop ost2 service on `facet_active_host ost2`"
198 # These tests all use non-failover stop
199 stop ost2 -f || return 93
204 echo "mount $FSNAME on ${MOUNTPATH}....."
205 zconf_mount $(hostname) $MOUNTPATH || return 96
209 local mountopt="remount,$1"
211 echo "remount '$1' lustre on ${MOUNTPATH}....."
212 zconf_mount $(hostname) $MOUNTPATH "$mountopt" || return 96
218 echo "umount lustre on $mountpath....."
219 zconf_umount $HOSTNAME $mountpath $@ || return 97
222 manual_umount_client(){
225 echo "manual umount lustre on ${MOUNT}...."
226 do_facet client "umount ${FORCE} $MOUNT"
232 start_mds || error "MDT start failed"
233 start_ost || error "Unable to start OST1"
234 mount_client $MOUNT || error "client start failed"
235 client_up || error "client_up failed"
244 unload_modules_conf () {
245 if combined_mgs_mds || ! local_mode; then
246 unload_modules || return 1
251 stop_ost || return 202
252 stop_mds || return 201
253 unload_modules_conf || return 203
257 umount_client $MOUNT || return 200
258 cleanup_nocli || return $?
263 echo "umount $MOUNT2 ..."
264 umount $MOUNT2 || true
265 echo "stopping fs2mds ..."
266 stop fs2mds -f || true
267 echo "stopping fs2ost ..."
268 stop fs2ost -f || true
272 do_facet client "cp /etc/passwd $DIR/a" || return 71
273 do_facet client "rm $DIR/a" || return 72
274 # make sure lustre is actually mounted (touch will block,
275 # but grep won't, so do it after)
276 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
277 echo "setup single mount lustre success"
281 do_facet client "touch $DIR/a" || return 71
282 do_facet client "rm $DIR/a" || return 72
283 do_facet client "touch $DIR2/a" || return 73
284 do_facet client "rm $DIR2/a" || return 74
285 echo "setup double mount lustre success"
290 if [ "$ONLY" == "setup" ]; then
295 if [ "$ONLY" == "cleanup" ]; then
302 #create single point mountpoint
308 check_mount || error "check_mount failed"
309 cleanup || error "cleanup failed with $?"
311 run_test 0 "single mount setup"
314 start_mds || error "MDS start failed"
315 start_ost || error "unable to start OST"
316 echo "start ost second time..."
317 start_ost && error "2nd OST start should fail"
318 mount_client $MOUNT || error "client start failed"
319 check_mount || error "check_mount failed"
320 cleanup || error "cleanup failed with $?"
322 run_test 1 "start up ost twice (should return errors)"
325 start_mds || error "MDT start failed"
326 echo "start mds second time.."
327 start_mds && error "2nd MDT start should fail"
328 start_ost || error "OST start failed"
329 mount_client $MOUNT || error "mount_client failed to start client"
330 check_mount || error "check_mount failed"
331 cleanup || error "cleanup failed with $?"
333 run_test 2 "start up mds twice (should return err)"
337 #mount.lustre returns an error if already in mtab
338 mount_client $MOUNT && error "2nd client mount should fail"
339 check_mount || error "check_mount failed"
340 cleanup || error "cleanup failed with $?"
342 run_test 3 "mount client twice (should return err)"
346 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
347 stop_ost || error "Unable to stop OST1"
348 umount_client $MOUNT -f || error “unmount $MOUNT failed”
351 # ok for ost to fail shutdown
352 if [ 202 -ne $eno ] && [ 0 -ne $eno ]; then
353 error "cleanup failed with $?"
356 run_test 4 "force cleanup ost, then cleanup"
358 test_5a() { # was test_5
360 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
361 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
363 stop_mds || error "Unable to stop MDS"
365 # cleanup may return an error from the failed
366 # disconnects; for now I'll consider this successful
367 # if all the modules have unloaded.
371 echo "killing umount"
372 kill -TERM $UMOUNT_PID
373 echo "waiting for umount to finish"
375 if grep " $MOUNT " /proc/mounts; then
376 echo "test 5: /proc/mounts after failed umount"
380 echo "killing umount"
381 kill -TERM $UMOUNT_PID
382 echo "waiting for umount to finish"
384 grep " $MOUNT " /proc/mounts &&
385 error "/proc/mounts after second umount"
388 # manual_mount_client may fail due to umount succeeding above
390 # stop_mds is a no-op here, and should not fail
391 cleanup_nocli || error "cleanup_nocli failed with $?"
392 # df may have lingering entry
394 # mtab may have lingering entry
398 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
400 grep -q $MOUNT" " /etc/mtab || break
401 echo "Waiting /etc/mtab updated ... "
402 WAIT=$(( WAIT + sleep))
404 [ "$WAIT" -eq "$MAX_WAIT" ] &&
405 error "/etc/mtab is not updated in $WAIT secs"
406 echo "/etc/mtab updated in $WAIT secs"
408 run_test 5a "force cleanup mds, then cleanup"
416 grep " $MOUNT " /etc/mtab &&
417 error false "unexpected entry in mtab before mount" && return 10
419 start_ost || error "OST start failed"
420 if ! combined_mgs_mds ; then
421 trap cleanup_5b EXIT ERR
422 start_mds || error "MDS start failed"
426 mount_client $MOUNT && error "mount_client $MOUNT should fail"
427 grep " $MOUNT " /etc/mtab &&
428 error "$MOUNT entry in mtab after failed mount"
430 # stop_mds is a no-op here, and should not fail
431 cleanup_nocli || error "cleanup_nocli failed with $?"
432 if ! combined_mgs_mds ; then
436 run_test 5b "Try to start a client with no MGS (should return errs)"
439 grep " $MOUNT " /etc/mtab &&
440 error false "unexpected entry in mtab before mount" && return 10
442 start_mds || error "MDS start failed"
443 start_ost || error "OST start failed"
444 local oldfs="${FSNAME}"
445 FSNAME="wrong.${FSNAME}"
446 mount_client $MOUNT || :
448 grep " $MOUNT " /etc/mtab &&
449 error "$MOUNT entry in mtab after failed mount"
451 cleanup_nocli || error "cleanup_nocli failed with $?"
453 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
456 grep " $MOUNT " /etc/mtab &&
457 error false "unexpected entry in mtab before mount" && return 10
459 start_ost || error "OST start failed"
460 start_mds || error "MDS start failed"
461 stop_ost || error "Unable to stop OST1"
462 mount_client $MOUNT || error "mount_client $MOUNT failed"
463 umount_client $MOUNT -f || error "umount_client $MOUNT failed"
464 cleanup_nocli || error "cleanup_nocli failed with $?"
465 grep " $MOUNT " /etc/mtab &&
466 error "$MOUNT entry in mtab after unmount"
469 run_test 5d "mount with ost down"
472 grep " $MOUNT " /etc/mtab &&
473 error false "unexpected entry in mtab before mount" && return 10
475 start_mds || error "MDS start failed"
476 start_ost || error "OST start failed"
478 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
479 do_facet client "$LCTL set_param fail_loc=0x80000506"
480 mount_client $MOUNT || echo "mount failed (not fatal)"
481 cleanup || error "cleanup failed with $?"
482 grep " $MOUNT " /etc/mtab &&
483 error "$MOUNT entry in mtab after unmount"
486 run_test 5e "delayed connect, don't crash (bug 10268)"
489 if combined_mgs_mds ; then
490 skip "combined mgs and mds"
494 grep " $MOUNT " /etc/mtab &&
495 error false "unexpected entry in mtab before mount" && return 10
498 start_ost || error "OST start failed"
499 mount_client $MOUNT &
501 echo client_mount pid is $pid
505 if ! ps -f -p $pid >/dev/null; then
508 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
509 error "mount returns $rc, expected to hang"
511 cleanup || error "cleanup failed with $?"
516 start_mds || error "start MDS failed"
518 # mount should succeed after start mds
520 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
521 cleanup || error "final call to cleanup failed with rc $?"
523 run_test 5f "mds down, cleanup after failed mount (bug 2712)"
528 mount_client $MOUNT || error "mount_client $MOUNT failed"
529 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
530 cleanup || error "cleanup failed with rc $?"
532 run_test 6 "manual umount, then mount again"
537 cleanup_nocli || error "cleanup_nocli failed with $?"
539 run_test 7 "manual umount, then cleanup"
543 mount_client $MOUNT2 || error "mount_client $MOUNT2 failed"
544 check_mount2 || error "check_mount2 failed"
545 umount_client $MOUNT2 || error "umount_client $MOUNT2 failed"
546 cleanup || error "cleanup failed with rc $?"
548 run_test 8 "double mount setup"
551 start_ost || error "OST start failed"
553 do_facet ost1 $LCTL set_param debug=\'inode trace\' ||
554 error "do_facet ost1 set_param inode trace failed."
555 do_facet ost1 $LCTL set_param subsystem_debug=\'mds ost\' ||
556 error "do_facet ost1 set_param debug mds ost failed."
558 CHECK_PTLDEBUG="`do_facet ost1 $LCTL get_param -n debug`"
559 if [ "$CHECK_PTLDEBUG" ] && { \
560 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
561 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
562 echo "lnet.debug success"
564 error "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
566 CHECK_SUBSYS="`do_facet ost1 $LCTL get_param -n subsystem_debug`"
567 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
568 echo "lnet.subsystem_debug success"
570 error "lnet.subsystem_debug: want 'mds ost' got '$CHECK_SUBSYS'"
572 stop_ost || error "Unable to stop OST1"
574 run_test 9 "test ptldebug and subsystem for mkfs"
582 do_facet $facet "test -b $dev" || rc=1
583 if [[ "$size" ]]; then
584 local in=$(do_facet $facet "dd if=$dev of=/dev/null bs=1k \
585 count=1 skip=$size 2>&1" |
586 awk '($3 == "in") { print $1 }')
587 [[ $in = "1+0" ]] || rc=1
593 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
594 # But with new MDS stack we don't care about the mode of local objects
595 # anymore, so this test is removed. See bug 22944 for more details.
599 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
600 skip "Only applicable to ldiskfs-based MDTs"
605 check_mount || error "check_mount failed"
606 cleanup || error "cleanup failed with rc $?"
608 echo "Remove mds config log"
609 if ! combined_mgs_mds ; then
613 do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' \
614 $(mgsdevname) || return \$?" ||
615 error "do_facet mgs failed with $?"
617 if ! combined_mgs_mds ; then
621 start_ost || error "OST start failed"
622 start_mds && error "MDS start succeeded, but should fail"
625 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
628 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
629 skip "Only applicable to ldiskfs-based MDTs"
633 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
638 # check if current MDSSIZE is large enough
639 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE &&
640 log "use MDSSIZE=$MDSSIZE"
642 # check if the global config has a large enough MDSSIZE
643 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] &&
644 [ $STORED_MDSSIZE -ge $MIN ] &&
645 OK=1 && myMDSSIZE=$STORED_MDSSIZE &&
646 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
648 # check if the block device is large enough
649 is_blkdev $SINGLEMDS $MDSDEV $MIN
650 local large_enough=$?
651 if [ -n "$OK" ]; then
652 [ $large_enough -ne 0 ] && OK=""
654 [ $large_enough -eq 0 ] && OK=1 && myMDSSIZE=$MIN &&
655 log "use device $MDSDEV with MIN=$MIN"
658 # check if a loopback device has enough space for fs metadata (5%)
660 if [ -z "$OK" ]; then
661 local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! \
662 -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
663 awk '($1 != "Filesystem") { print $4 }')
664 ! [ -z "$SPACE" ] && [ $SPACE -gt $((MIN / 20)) ] &&
665 OK=1 && myMDSSIZE=$MIN &&
666 log "use file $MDSDEV with MIN=$MIN"
669 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS" && return
671 echo "mount mds with large journal..."
673 local OLD_MDSSIZE=$MDSSIZE
677 echo "mount lustre system..."
679 check_mount || error "check_mount failed"
681 echo "check journal size..."
682 local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
683 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
684 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
686 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
689 cleanup || error "cleanup failed with rc $?"
694 run_test 18 "check mkfs creates large journals"
697 start_mds || error "MDS start failed"
698 stop_mds || error "Unable to stop MDS"
700 run_test 19a "start/stop MDS without OSTs"
703 start_ost || error "Unable to start OST1"
704 stop_ost || error "Unable to stop OST1"
706 run_test 19b "start/stop OSTs without MDS"
709 # first format the ost/mdt
710 start_mds || error "MDS start failed"
711 start_ost || error "Unable to start OST1"
712 mount_client $MOUNT || error "mount_client $MOUNT failed"
713 check_mount || error "check_mount failed"
714 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed."
715 remount_client ro $MOUNT || error "remount_client with ro failed"
716 touch $DIR/$tfile && error "$DIR/$tfile created incorrectly"
717 [ -e $DIR/$tfile ] && error "$DIR/$tfile exists incorrectly"
718 remount_client rw $MOUNT || error "remount_client with rw failed"
719 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
720 MCNT=$(grep -c $MOUNT /etc/mtab)
721 [ "$MCNT" -ne 1 ] && error "$MOUNT in /etc/mtab $MCNT times"
723 stop_mds || error "Unable to stop MDS"
724 stop_ost || error "Unable to stop OST1"
726 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
729 start_mds || error "MDS start failed"
730 start_ost || error "unable to start OST1"
731 wait_osc_import_state mds ost FULL
732 stop_ost || error "unable to stop OST1"
733 stop_mds || error "unable to stop MDS"
735 run_test 21a "start mds before ost, stop ost first"
738 start_ost || error "unable to start OST1"
739 start_mds || error "MDS start failed"
740 wait_osc_import_state mds ost FULL
741 stop_mds || error "unable to stop MDS"
742 stop_ost || error "unable to stop OST1"
744 run_test 21b "start ost before mds, stop mds first"
747 start_ost || error "Unable to start OST1"
748 start_mds || error "MDS start failed"
749 start_ost2 || error "Unable to start OST2"
750 wait_osc_import_state mds ost2 FULL
751 stop_ost || error "Unable to stop OST1"
752 stop_ost2 || error "Unable to stop OST2"
753 stop_mds || error "Unable to stop MDS"
754 #writeconf to remove all ost2 traces for subsequent tests
755 writeconf_or_reformat
757 run_test 21c "start mds between two osts, stop mds last"
760 if combined_mgs_mds ; then
761 skip "need separate mgs device" && return 0
767 start_mgs || error "unable to start MGS"
768 start_ost || error "unable to start OST1"
769 start_ost2 || error "unable to start OST2"
770 start_mds || error "MDS start failed"
771 wait_osc_import_state mds ost2 FULL
773 stop_ost || error "Unable to stop OST1"
774 stop_ost2 || error "Unable to stop OST2"
775 stop_mds || error "Unable to stop MDS"
777 #writeconf to remove all ost2 traces for subsequent tests
778 writeconf_or_reformat
779 start_mgs || error "unable to start MGS"
781 run_test 21d "start mgs then ost and then mds"
784 MGSNID="$saved_mgsnid"
786 echo "stopping fs2mgs ..."
787 stop $fs2mgs -f || true
790 test_21e() { # LU-5863
791 if [[ -z "$fs3ost_DEV" || -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
792 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
793 skip_env "mixed loopback and real device not working" && return
796 local fs2mdsdev=$(mdsdevname 1_2)
797 local fs2ostdev=$(ostdevname 1_2)
798 local fs3ostdev=$(ostdevname 2_2)
800 local fs2mdsvdev=$(mdsvdevname 1_2)
801 local fs2ostvdev=$(ostvdevname 1_2)
802 local fs3ostvdev=$(ostvdevname 2_2)
804 # temporarily use fs3ost as fs2mgs
806 local fs2mgsdev=$fs3ostdev
807 local fs2mgsvdev=$fs3ostvdev
809 local fsname=test1234
811 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
812 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
813 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS && trap cleanup_21e EXIT INT ||
814 error "start fs2mgs failed"
816 local saved_mgsnid="$MGSNID"
817 MGSNID=$(do_facet $fs2mgs $LCTL list_nids | xargs | tr ' ' ,)
819 add fs2mds $(mkfs_opts mds1 $fs2mdsdev $fsname) \
820 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
821 add fs2ost $(mkfs_opts ost1 $fs2ostdev $fsname) \
822 --reformat $fs2ostdev $fs2ostvdev || error "add fs2ost failed"
824 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
825 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS || error "start fs2mds failed"
827 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
828 $MOUNT_CMD $MGSNID:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
829 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
833 run_test 21e "separate MGS and MDS"
836 start_mds || error "MDS start failed"
838 echo "Client mount with ost in logs, but none running"
839 start_ost || error "unable to start OST1"
840 # wait until mds connected to ost and open client connection
841 wait_osc_import_state mds ost FULL
842 stop_ost || error "unable to stop OST1"
843 mount_client $MOUNT || error "mount_client $MOUNT failed"
844 # check_mount will block trying to contact ost
845 mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed"
846 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed"
847 umount_client $MOUNT -f
850 echo "Client mount with a running ost"
851 start_ost || error "unable to start OST1"
853 # if gss enabled, wait full time to let connection from
854 # mds to ost be established, due to the mismatch between
855 # initial connect timeout and gss context negotiation timeout.
856 # This perhaps could be remove after AT landed.
857 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
858 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
860 mount_client $MOUNT || error "mount_client $MOUNT failed"
861 wait_osc_import_state mds ost FULL
862 wait_osc_import_state client ost FULL
863 check_mount || error "check_mount failed"
866 cleanup || error "cleanup failed with rc $?"
868 run_test 22 "start a client before osts (should return errs)"
870 test_23a() { # was test_23
873 stop $SINGLEMDS || error "failed to stop $SINGLEMDS"
874 # force down client so that recovering mds waits for reconnect
875 local running=$(grep -c $MOUNT /proc/mounts) || true
876 if [ $running -ne 0 ]; then
877 echo "Stopping client $MOUNT (opts: -f)"
881 # enter recovery on failed mds
882 local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/})
883 start $SINGLEMDS $MDT_DEV $MDS_MOUNT_OPTS || error "MDS start failed"
884 # try to start a new client
885 mount_client $MOUNT &
887 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
888 MOUNT_LUSTRE_PID=$(ps -ef | grep mount.lustre |
889 grep -v grep | awk '{print $2}')
890 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
892 ps --ppid $MOUNT_LUSTRE_PID
893 echo "waiting for mount to finish"
895 # "ctrl-c" sends SIGINT but it usually (in script) does not work on child process
896 # SIGTERM works but it does not spread to offspring processses
897 kill -s TERM $MOUNT_PID
898 kill -s TERM $MOUNT_LUSTRE_PID
899 # we can not wait $MOUNT_PID because it is not a child of this shell
905 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
907 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
908 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
911 [ -z "$PID1" -a -z "$PID2" ] && break
912 echo "waiting for mount to finish ... "
913 WAIT=$(( WAIT + sleep))
915 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
916 error "MOUNT_PID $MOUNT_PID and "\
917 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
920 stop_mds || error "stopping MDSes failed"
921 stop_ost || error "stopping OSSes failed"
923 run_test 23a "interrupt client during recovery mount delay"
928 test_23b() { # was test_23
929 start_mds || error "MDS start failed"
930 start_ost || error "Unable to start OST1"
931 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
932 $LCTL set_param fail_loc=0x80000313
934 cleanup || error "cleanup failed with rc $?"
936 run_test 23b "Simulate -EINTR during mount"
939 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
941 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
942 is_blkdev $SINGLEMDS $MDSDEV &&
943 skip_env "mixed loopback and real device not working" && return
946 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
948 local fs2mdsdev=$(mdsdevname 1_2)
949 local fs2ostdev=$(ostdevname 1_2)
950 local fs2mdsvdev=$(mdsvdevname 1_2)
951 local fs2ostvdev=$(ostvdevname 1_2)
953 # test 8-char fsname as well
954 local FSNAME2=test1234
956 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --nomgs --mgsnode=$MGSNID \
957 --fsname=${FSNAME2} --reformat $fs2mdsdev $fs2mdsvdev || exit 10
959 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --fsname=${FSNAME2} \
960 --reformat $fs2ostdev $fs2ostvdev || exit 10
963 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
964 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
965 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
966 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
968 check_mount || error "check_mount failed"
969 # files written on 1 should not show up on 2
970 cp /etc/passwd $DIR/$tfile
972 [ -e $MOUNT2/$tfile ] && error "File bleed"
975 cp /etc/passwd $MOUNT2/$tfile ||
976 error "cp /etc/passwd $MOUNT2/$tfile failed"
977 rm $MOUNT2/$tfile || error "remove $MOUNT2/$tfile failed"
978 # 2 is actually mounted
979 grep $MOUNT2' ' /proc/mounts > /dev/null || error "$MOUNT2 not mounted"
981 facet_failover fs2mds
982 facet_failover fs2ost
985 # the MDS must remain up until last MDT
987 MDS=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
988 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
989 [ -z "$MDS" ] && error "No MDT"
991 cleanup_nocli || error "cleanup_nocli failed with rc $?"
993 run_test 24a "Multiple MDTs on a single node"
996 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
998 if [ -z "$fs2mds_DEV" ]; then
999 local dev=${SINGLEMDS}_dev
1000 local MDSDEV=${!dev}
1001 is_blkdev $SINGLEMDS $MDSDEV &&
1002 skip_env "mixed loopback and real device not working" && return
1005 local fs2mdsdev=$(mdsdevname 1_2)
1006 local fs2mdsvdev=$(mdsvdevname 1_2)
1008 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --mgs --fsname=${FSNAME}2 \
1009 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
1011 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS &&
1012 error "start MDS should fail"
1014 cleanup || error "cleanup failed with rc $?"
1016 run_test 24b "Multiple MGSs on a single node (should return err)"
1020 check_mount || error "check_mount failed"
1021 local MODULES=$($LCTL modules | awk '{ print $2 }')
1022 rmmod $MODULES 2>/dev/null || true
1023 cleanup || error "cleanup failed with $?"
1025 run_test 25 "Verify modules are referenced"
1029 # we need modules before mount for sysctl, so make sure...
1030 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
1031 #define OBD_FAIL_MDS_FS_SETUP 0x135
1032 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000135"
1033 start_mds && error "MDS started but should not have started"
1034 $LCTL get_param -n devices
1035 DEVS=$($LCTL get_param -n devices | egrep -v MG | wc -l)
1036 [ $DEVS -gt 0 ] && error "number of devices is $DEVS, should be zero"
1037 # start mds to drop writeconf setting
1038 start_mds || error "Unable to start MDS"
1039 stop_mds || error "Unable to stop MDS"
1040 unload_modules_conf || error "unload_modules_conf failed with $?"
1042 run_test 26 "MDT startup failure cleans LOV (should return errs)"
1045 start_ost || error "Unable to start OST1"
1046 start_mds || error "Unable to start MDS"
1047 echo "Requeue thread should have started: "
1048 ps -e | grep ll_cfg_requeue
1049 set_conf_param_and_check ost1 \
1050 "$LCTL get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" \
1051 "$FSNAME-OST0000.ost.client_cache_seconds" ||
1052 error "set_conf_param_and_check ost1 failed"
1053 cleanup_nocli || error "cleanup_nocli failed with rc $?"
1055 run_test 27a "Reacquire MGS lock if OST started first"
1060 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
1061 awk '($3 ~ "mdt" && $4 ~ "MDT0000") { print $4 }')
1063 facet_failover $SINGLEMDS
1064 set_conf_param_and_check $SINGLEMDS \
1065 "$LCTL get_param -n mdt.$device.identity_acquire_expire" \
1066 "$device.mdt.identity_acquire_expire" ||
1067 error "set_conf_param_and_check $SINGLEMDS failed"
1068 set_conf_param_and_check client \
1069 "$LCTL get_param -n mdc.$device-mdc-*.max_rpcs_in_flight"\
1070 "$device.mdc.max_rpcs_in_flight" ||
1071 error "set_conf_param_and_check client failed"
1073 cleanup || error "cleanup failed with $?"
1075 run_test 27b "Reacquire MGS lock after failover"
1079 TEST="$LCTL get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1080 PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
1082 FINAL=$(($ORIG + 1))
1083 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL ||
1084 error "first set_conf_param_and_check client failed"
1085 FINAL=$(($FINAL + 1))
1086 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL ||
1087 error "second set_conf_param_and_check client failed"
1088 umount_client $MOUNT || error "umount_client $MOUNT failed"
1089 mount_client $MOUNT || error "mount_client $MOUNT failed"
1091 if [ $RESULT -ne $FINAL ]; then
1092 error "New config not seen: wanted $FINAL got $RESULT"
1094 echo "New config success: got $RESULT"
1096 set_conf_param_and_check client "$TEST" "$PARAM" $ORIG ||
1097 error "third set_conf_param_and_check client failed"
1098 cleanup || error "cleanup failed with rc $?"
1100 run_test 28 "permanent parameter setting"
1102 test_28a() { # LU-4221
1103 [[ $(lustre_version_code ost1) -ge $(version_code 2.5.52) ]] ||
1104 { skip "Need OST version at least 2.5.52" && return 0; }
1105 [ "$(facet_fstype ost1)" = "zfs" ] &&
1106 skip "LU-4221: no such proc params for ZFS OSTs" && return
1113 local device="$FSNAME-OST0000"
1117 # In this test we will set three kinds of proc parameters with
1119 # 1. the ones moved from the OFD to the OSD, and only their
1120 # symlinks kept in obdfilter
1121 # 2. non-symlink ones in the OFD
1122 # 3. non-symlink ones in the OSD
1125 # prepare a symlink parameter in the OFD
1126 name="writethrough_cache_enable"
1127 param="$device.ost.$name"
1128 cmd="$LCTL get_param -n obdfilter.$device.$name"
1130 # conf_param the symlink parameter in the OFD
1131 old=$(do_facet ost1 $cmd)
1132 new=$(((old + 1) % 2))
1133 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1134 error "lctl conf_param $device.ost.$param=$new failed"
1136 # conf_param the target parameter in the OSD
1137 param="$device.osd.$name"
1138 cmd="$LCTL get_param -n osd-*.$device.$name"
1139 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1140 error "lctl conf_param $device.osd.$param=$old failed"
1143 # prepare a non-symlink parameter in the OFD
1144 name="client_cache_seconds"
1145 param="$device.ost.$name"
1146 cmd="$LCTL get_param -n obdfilter.$device.$name"
1148 # conf_param the parameter in the OFD
1149 old=$(do_facet ost1 $cmd)
1151 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1152 error "lctl conf_param $device.ost.$param=$new failed"
1153 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1154 error "lctl conf_param $device.ost.$param=$old failed"
1157 # prepare a non-symlink parameter in the OSD
1159 param="$device.osd.$name"
1160 cmd="$LCTL get_param -n osd-*.$device.$name"
1162 # conf_param the parameter in the OSD
1163 old=$(do_facet ost1 $cmd)
1164 new=$(((old + 1) % 2))
1165 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1166 error "lctl conf_param $device.osd.$param=$new failed"
1167 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1168 error "lctl conf_param $device.osd.$param=$old failed"
1170 cleanup || error "cleanup failed with $?"
1172 run_test 28a "set symlink parameters permanently with conf_param"
1175 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
1176 setup > /dev/null 2>&1
1177 start_ost2 || error "Unable to start OST2"
1180 local PARAM="$FSNAME-OST0001.osc.active"
1181 local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
1182 local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
1184 ACTV=$($LCTL get_param -n $PROC_ACT)
1186 set_conf_param_and_check client \
1187 "$LCTL get_param -n $PROC_ACT" "$PARAM" $DEAC ||
1188 error "set_conf_param_and_check client failed"
1189 # also check ost_server_uuid status
1190 RESULT=$($LCTL get_param -n $PROC_UUID | grep DEACTIV)
1191 if [ -z "$RESULT" ]; then
1192 error "Client not deactivated: $($LCTL get_param \
1195 echo "Live client success: got $RESULT"
1199 for num in $(seq $MDSCOUNT); do
1200 local mdtosc=$(get_mdtosc_proc_path mds${num} $FSNAME-OST0001)
1201 local MPROC="osc.$mdtosc.active"
1206 RESULT=$(do_facet mds${num} "$LCTL get_param -n $MPROC")
1207 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
1208 if [ $RESULT -eq $DEAC ]; then
1209 echo -n "MDT deactivated also after"
1210 echo "$WAIT sec (got $RESULT)"
1214 if [ $WAIT -eq $MAX ]; then
1215 error "MDT active: wanted $DEAC got $RESULT"
1217 echo "Waiting $(($MAX - $WAIT))secs for MDT deactivated"
1220 # test new client starts deactivated
1221 umount_client $MOUNT || error "umount_client $MOUNT failed"
1222 mount_client $MOUNT || error "mount_client $MOUNT failed"
1223 RESULT=$($LCTL get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
1224 if [ -z "$RESULT" ]; then
1225 error "New client start active: $(lctl get_param -n $PROC_UUID)"
1227 echo "New client success: got $RESULT"
1230 # make sure it reactivates
1231 set_conf_param_and_check client \
1232 "$LCTL get_param -n $PROC_ACT" "$PARAM" $ACTV ||
1233 error "lctl get_param $PROC_ACT $PARAM $ACTV failed"
1235 umount_client $MOUNT
1236 stop_ost2 || error "Unable to stop OST2"
1237 cleanup_nocli || error "cleanup_nocli failed with $?"
1238 #writeconf to remove all ost2 traces for subsequent tests
1239 writeconf_or_reformat
1241 run_test 29 "permanently remove an OST"
1246 echo Big config llog
1247 TEST="$LCTL get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1249 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1250 for i in ${LIST[@]}; do
1251 set_conf_param_and_check client "$TEST" \
1252 "$FSNAME.llite.max_read_ahead_whole_mb" $i ||
1253 error "Set $FSNAME.llite.max_read_ahead_whole_mb failed"
1255 # make sure client restart still works
1256 umount_client $MOUNT
1257 mount_client $MOUNT || error "mount_client $MOUNT failed"
1258 [ "$($TEST)" -ne "$i" ] &&
1259 error "Param didn't stick across restart $($TEST) != $i"
1262 echo Erase parameter setting
1263 do_facet mgs "$LCTL conf_param \
1264 -d $FSNAME.llite.max_read_ahead_whole_mb" ||
1265 error "Erase param $FSNAME.llite.max_read_ahead_whole_mb failed"
1266 umount_client $MOUNT
1267 mount_client $MOUNT || error "mount_client $MOUNT failed"
1269 echo "deleted (default) value=$FINAL, orig=$ORIG"
1270 # assumes this parameter started at the default value
1271 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
1273 cleanup || error "cleanup failed with rc $?"
1275 run_test 30a "Big config llog and conf_param deletion"
1280 local orignids=$($LCTL get_param -n \
1281 osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1283 local orignidcount=$(echo "$orignids" | wc -w)
1285 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1286 # numerical part of it. Hopefully that's not already a failover address
1288 local OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | \
1290 local ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1291 local NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1292 local NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1293 echo "Using fake nid $NEW"
1295 local TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1296 grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1297 set_conf_param_and_check client "$TEST" \
1298 "$FSNAME-OST0000.failover.node" $NEW ||
1299 error "didn't add failover nid $NEW"
1300 local NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1303 local NIDCOUNT=$(echo "$NIDS" | wc -w)
1304 echo "should have $((orignidcount + 1)) entries \
1305 in failover nids string, have $NIDCOUNT"
1306 [ $NIDCOUNT -eq $((orignidcount + 1)) ] ||
1307 error "Failover nid not added"
1309 do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" ||
1310 error "conf_param delete failed"
1311 umount_client $MOUNT
1312 mount_client $MOUNT || error "mount_client $MOUNT failed"
1314 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1317 NIDCOUNT=$(echo "$NIDS" | wc -w)
1318 echo "only $orignidcount final entries should remain \
1319 in failover nids string, have $NIDCOUNT"
1320 [ $NIDCOUNT -eq $orignidcount ] || error "Failover nids not removed"
1322 cleanup || error "cleanup failed with rc $?"
1324 run_test 30b "Remove failover nids"
1326 test_31() { # bug 10734
1327 # ipaddr must not exist
1328 $MOUNT_CMD 4.3.2.1@tcp:/lustre $MOUNT || true
1329 cleanup || error "cleanup failed with rc $?"
1331 run_test 31 "Connect to non-existent node (shouldn't crash)"
1335 T32_BLIMIT=40960 # Kbytes
1339 # This is not really a test but a tool to create new disk
1340 # image tarballs for the upgrade tests.
1342 # Disk image tarballs should be created on single-node
1343 # clusters by running this test with default configurations
1344 # plus a few mandatory environment settings that are verified
1345 # at the beginning of the test.
1347 test_32newtarball() {
1351 local tmp=$TMP/t32_image_create
1352 local server_version=$(lustre_version_code $SINGLEMDS)
1357 if [ $FSNAME != t32fs -o \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o \
1358 $OSTCOUNT -ne 1 -o -z "$OSTDEV1" ]; then
1359 error "Needs FSNAME=t32fs MDSCOUNT=2 " \
1360 "MDSDEV1=<nonexistent_file>" \
1361 "MDSDEV2=<nonexistent_file>" \
1362 "(or MDSDEV, in the case of b1_8)" \
1363 "OSTCOUNT=1 OSTDEV1=<nonexistent_file>"
1367 echo "Found stale $tmp"
1371 mkdir $tmp/src || return 1
1372 tar cf - -C $src . | tar xf - -C $tmp/src
1373 dd if=/dev/zero of=$tmp/src/t32_qf_old bs=1M \
1374 count=$(($T32_BLIMIT / 1024 / 4))
1375 chown $T32_QID.$T32_QID $tmp/src/t32_qf_old
1377 # format ost with comma-separated NIDs to verify LU-4460
1378 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
1379 MGSNID="$MGSNID,$MGSNID" OSTOPT="--failnode=$failnid" formatall
1383 [[ $server_version -ge $(version_code 2.3.50) ]] ||
1384 $LFS quotacheck -ug /mnt/$FSNAME
1385 $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
1388 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1390 if [[ $MDSCOUNT -ge 2 ]]; then
1391 remote_dir=/mnt/$FSNAME/remote_dir
1392 $LFS mkdir -i 1 $remote_dir
1393 tar cf - -C $tmp/src . | tar xf - -C $remote_dir
1395 if [[ $server_version -ge $(version_code 2.7.0) ]]; then
1396 striped_dir=/mnt/$FSNAME/striped_dir_old
1397 $LFS mkdir -i 1 -c 2 $striped_dir
1398 tar cf - -C $tmp/src . | tar xf - -C $striped_dir
1404 mkdir $tmp/img || return 1
1408 pushd_dir=/mnt/$FSNAME
1409 if [[ $MDSCOUNT -ge 2 ]]; then
1410 pushd_dir=$remote_dir
1411 if [[ $server_version -ge $(version_code 2.7.0) ]]; then
1413 ls -Rni --time-style=+%s >$tmp/img/list2
1419 ls -Rni --time-style=+%s >$tmp/img/list
1420 find ! -name .lustre -type f -exec sha1sum {} \; |
1421 sort -k 2 >$tmp/img/sha1sums
1423 $LCTL get_param -n version | head -n 1 |
1424 sed -e 's/^lustre: *//' >$tmp/img/commit
1426 [[ $server_version -ge $(version_code 2.3.50) ]] ||
1427 $LFS quotaon -ug /mnt/$FSNAME
1428 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1429 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1430 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1431 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1432 | tr -d "*" > $tmp/img/bspace
1433 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1434 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1435 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1436 | tr -d "*" > $tmp/img/ispace
1437 echo $T32_BLIMIT > $tmp/img/blimit
1438 echo $T32_ILIMIT > $tmp/img/ilimit
1443 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1446 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1447 echo "Data verification failed"
1450 uname -r >$tmp/img/kernel
1451 uname -m >$tmp/img/arch
1453 mv ${MDSDEV1:-$MDSDEV} $tmp/img
1454 for num in $(seq 2 $MDSCOUNT); do
1455 local devname=$(mdsdevname $num)
1457 [[ $(facet_fstype $facet) != zfs ]] ||
1458 devname=$(mdsvdevname $num)
1459 mv $devname $tmp/img
1461 mv $OSTDEV1 $tmp/img
1463 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1464 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1467 tar cjvf $dst/disk$version-$(facet_fstype $SINGLEMDS).tar.bz2 -S *
1472 #run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1475 # The list of applicable tarballs is returned via the caller's
1476 # variable "tarballs".
1479 local node=$(facet_active_host $SINGLEMDS)
1480 local r="do_node $node"
1482 if [ "$CLIENTONLY" ]; then
1483 skip "Client-only testing"
1487 if ! $r which $TUNEFS; then
1488 skip_env "tunefs.lustre required on $node"
1492 local IMGTYPE=$(facet_fstype $SINGLEMDS)
1494 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 -name \'disk*-$IMGTYPE.tar.bz2\')
1496 if [ -z "$tarballs" ]; then
1497 skip "No applicable tarballs found"
1502 t32_test_cleanup() {
1504 local facet=$SINGLEMDS
1505 local fstype=$(facet_fstype $facet)
1508 if $shall_cleanup_lustre; then
1509 umount $tmp/mnt/lustre || rc=$?
1511 if $shall_cleanup_mdt; then
1512 $r $UMOUNT $tmp/mnt/mdt || rc=$?
1514 if $shall_cleanup_mdt1; then
1515 $r $UMOUNT $tmp/mnt/mdt1 || rc=$?
1517 if $shall_cleanup_ost; then
1518 $r $UMOUNT $tmp/mnt/ost || rc=$?
1523 if [[ $fstype == zfs ]]; then
1525 local poolname_list="t32fs-mdt1 t32fs-ost1"
1527 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1529 for poolname in $poolname_list; do
1530 destroy_zpool $facet $poolname
1536 t32_bits_per_long() {
1538 # Yes, this is not meant to be perfect.
1548 t32_reload_modules() {
1550 local all_removed=false
1553 while ((i < 20)); do
1554 echo "Unloading modules on $node: Attempt $i"
1555 do_rpc_nodes $node $LUSTRE_RMMOD $(facet_fstype $SINGLEMDS) &&
1557 do_rpc_nodes $node check_mem_leak || return 1
1558 if $all_removed; then
1559 do_rpc_nodes $node load_modules
1565 echo "Unloading modules on $node: Given up"
1569 t32_wait_til_devices_gone() {
1575 echo wait for devices to go
1576 while ((i < 20)); do
1577 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1578 loops=$(do_rpc_nodes $node losetup -a | grep -c t32)
1579 ((devices == 0 && loops == 0)) && return 0
1583 echo "waiting for dev on $node: dev $devices loop $loops given up"
1584 do_rpc_nodes $node "losetup -a"
1585 do_rpc_nodes $node "$LCTL device_list"
1589 t32_verify_quota() {
1593 local fstype=$(facet_fstype $SINGLEMDS)
1597 $LFS quota -u $T32_QID -v $mnt
1599 qval=$($LFS quota -v -u $T32_QID $mnt |
1600 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1601 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1603 [ $qval -eq $img_bspace ] || {
1604 echo "bspace, act:$qval, exp:$img_bspace"
1608 qval=$($LFS quota -v -u $T32_QID $mnt |
1609 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1610 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1612 [ $qval -eq $img_ispace ] || {
1613 echo "ispace, act:$qval, exp:$img_ispace"
1617 qval=$($LFS quota -v -u $T32_QID $mnt |
1618 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1619 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1621 [ $qval -eq $img_blimit ] || {
1622 echo "blimit, act:$qval, exp:$img_blimit"
1626 qval=$($LFS quota -v -u $T32_QID $mnt |
1627 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1628 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1630 [ $qval -eq $img_ilimit ] || {
1631 echo "ilimit, act:$qval, exp:$img_ilimit"
1635 do_node $node $LCTL conf_param $fsname.quota.mdt=ug
1636 cmd="$LCTL get_param -n osd-$fstype.$fsname-MDT0000"
1637 cmd=$cmd.quota_slave.enabled
1638 wait_update $node "$cmd" "ug" || {
1639 echo "Enable mdt quota failed"
1643 do_node $node $LCTL conf_param $fsname.quota.ost=ug
1644 cmd="$LCTL get_param -n osd-$fstype.$fsname-OST0000"
1645 cmd=$cmd.quota_slave.enabled
1646 wait_update $node "$cmd" "ug" || {
1647 echo "Enable ost quota failed"
1652 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1653 bs=1M count=$((img_blimit / 1024)) oflag=sync && {
1654 echo "Write succeed, but expect -EDQUOT"
1657 rm -f $mnt/t32_qf_new
1659 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1661 echo "Create succeed, but expect -EDQUOT"
1664 unlinkmany $mnt/t32_qf_ $img_ilimit
1672 local dne_upgrade=${dne_upgrade:-"no"}
1673 local ff_convert=${ff_convert:-"no"}
1674 local shall_cleanup_mdt=false
1675 local shall_cleanup_mdt1=false
1676 local shall_cleanup_ost=false
1677 local shall_cleanup_lustre=false
1678 local mdt2_is_available=false
1679 local node=$(facet_active_host $SINGLEMDS)
1680 local r="do_node $node"
1681 local node2=$(facet_active_host mds2)
1691 local nid=$($r $LCTL list_nids | head -1)
1697 local fstype=$(facet_fstype $SINGLEMDS)
1698 local mdt_dev=$tmp/mdt
1699 local mdt2_dev=$tmp/mdt2
1700 local ost_dev=$tmp/ost
1705 trap 'trap - RETURN; t32_test_cleanup' RETURN
1708 mkdir -p $tmp/mnt/lustre || error "mkdir $tmp/mnt/lustre failed"
1709 $r mkdir -p $tmp/mnt/{mdt,mdt1,ost}
1710 $r tar xjvf $tarball -S -C $tmp || {
1711 error_noexit "Unpacking the disk image tarball"
1714 img_commit=$($r cat $tmp/commit)
1715 img_kernel=$($r cat $tmp/kernel)
1716 img_arch=$($r cat $tmp/arch)
1717 img_bspace=$($r cat $tmp/bspace)
1718 img_ispace=$($r cat $tmp/ispace)
1720 # older images did not have "blimit" and "ilimit" files
1721 # use old values for T32_BLIMIT and T32_ILIMIT
1722 $r test -f $tmp/blimit && img_blimit=$($r cat $tmp/blimit) ||
1724 $r test -f $tmp/ilimit && img_ilimit=$($r cat $tmp/ilimit) ||
1727 echo "Upgrading from $(basename $tarball), created with:"
1728 echo " Commit: $img_commit"
1729 echo " Kernel: $img_kernel"
1730 echo " Arch: $img_arch"
1731 echo "OST version: $(lustre_build_version ost1)"
1733 # The conversion can be made only when both of the following
1734 # conditions are satisfied:
1735 # - ost device img version < 2.3.64
1736 # - ost server version >= 2.5
1737 [ $(version_code $img_commit) -ge $(version_code 2.3.64) -o \
1738 $(lustre_version_code ost1) -lt $(version_code 2.5.0) ] &&
1741 ! $r test -f $mdt2_dev || mdt2_is_available=true
1743 if [[ $fstype == zfs ]]; then
1746 local poolname_list="t32fs-mdt1 t32fs-ost1"
1748 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1750 for poolname in $poolname_list; do
1751 $r "$ZPOOL list -H $poolname >/dev/null 2>&1 ||
1752 $ZPOOL import -f -d $tmp $poolname"
1755 mdt_dev=t32fs-mdt1/mdt1
1756 ost_dev=t32fs-ost1/ost1
1757 ! $mdt2_is_available || mdt2_dev=t32fs-mdt2/mdt2
1758 wait_update_facet $SINGLEMDS "$ZPOOL list |
1759 awk '/^t32fs-mdt1/ { print \\\$1 }'" "t32fs-mdt1" || {
1760 error_noexit "import zfs pool failed"
1765 $r $LCTL set_param debug="$PTLDEBUG"
1767 $r $TUNEFS --dryrun $mdt_dev || {
1769 error_noexit "tunefs.lustre before mounting the MDT"
1773 if $mdt2_is_available; then
1774 $r $TUNEFS --dryrun $mdt2_dev || {
1776 error_noexit "tunefs.lustre before mounting the MDT"
1781 if [ "$writeconf" ]; then
1783 if [ $fstype == "ldiskfs" ]; then
1785 $r $TUNEFS --quota $mdt_dev || {
1787 error_noexit "Enable mdt quota feature"
1790 if $mdt2_is_available; then
1791 $r $TUNEFS --quota $mdt2_dev || {
1793 error_noexit "Enable mdt quota feature"
1799 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
1800 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
1801 { skip "LU-2200: Cannot run over Inifiniband w/o lctl replace_nids "
1802 "(Need MGS version at least 2.3.59)"; return 0; }
1804 local osthost=$(facet_active_host ost1)
1805 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
1808 if [ $fstype == "ldiskfs" ]; then
1811 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt
1812 $r $LCTL replace_nids $fsname-OST0000 $ostnid
1813 $r $LCTL replace_nids $fsname-MDT0000 $nid
1814 $r $UMOUNT $tmp/mnt/mdt
1817 mopts=exclude=$fsname-OST0000
1818 if [ $fstype == "ldiskfs" ]; then
1823 t32_wait_til_devices_gone $node
1825 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
1827 error_noexit "Mounting the MDT"
1830 shall_cleanup_mdt=true
1832 if $mdt2_is_available; then
1833 mopts=mgsnode=$nid,$mopts
1834 $r $MOUNT_CMD -o $mopts $mdt2_dev $tmp/mnt/mdt1 || {
1836 error_noexit "Mounting the MDT"
1840 echo "mount new MDT....$mdt2_dev"
1841 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1842 error_noexit "enable remote dir create failed"
1844 shall_cleanup_mdt1=true
1845 elif [ "$dne_upgrade" != "no" ]; then
1846 local fs2mdsdev=$(mdsdevname 1_2)
1847 local fs2mdsvdev=$(mdsvdevname 1_2)
1849 echo "mkfs new MDT on ${fs2mdsdev}...."
1850 if [ $(facet_fstype mds1) == ldiskfs ]; then
1851 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
1854 add fs2mds $(mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
1855 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
1856 error_noexit "Mkfs new MDT failed"
1860 [[ $(facet_fstype mds1) != zfs ]] || import_zpool fs2mds
1862 $r $TUNEFS --dryrun $fs2mdsdev || {
1863 error_noexit "tunefs.lustre before mounting the MDT"
1867 echo "mount new MDT....$fs2mdsdev"
1868 $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
1869 error_noexit "mount mdt1 failed"
1873 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1874 error_noexit "enable remote dir create failed"
1876 shall_cleanup_mdt1=true
1879 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
1880 error_noexit "Getting MDT UUID"
1883 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
1884 error_noexit "Unexpected MDT UUID: \"$uuid\""
1888 $r $TUNEFS --dryrun $ost_dev || {
1889 error_noexit "tunefs.lustre before mounting the OST"
1892 if [ "$writeconf" ]; then
1893 mopts=mgsnode=$nid,$writeconf
1894 if [ $fstype == "ldiskfs" ]; then
1896 $r $TUNEFS --quota $ost_dev || {
1898 error_noexit "Enable ost quota feature"
1904 if [ $fstype == "ldiskfs" ]; then
1908 $r $MOUNT_CMD -o $mopts $ost_dev $tmp/mnt/ost || {
1909 error_noexit "Mounting the OST"
1912 shall_cleanup_ost=true
1914 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
1915 error_noexit "Getting OST UUID"
1918 if [ "$uuid" != $fsname-OST0000_UUID ]; then
1919 error_noexit "Unexpected OST UUID: \"$uuid\""
1923 $r $LCTL conf_param $fsname-OST0000.osc.max_dirty_mb=15 || {
1924 error_noexit "Setting \"max_dirty_mb\""
1927 $r $LCTL conf_param $fsname-OST0000.failover.node=$nid || {
1928 error_noexit "Setting OST \"failover.node\""
1931 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
1932 error_noexit "Setting \"max_rpcs_in_flight\""
1935 $r $LCTL conf_param $fsname-MDT0000.failover.node=$nid || {
1936 error_noexit "Setting MDT \"failover.node\""
1939 $r $LCTL pool_new $fsname.interop || {
1940 error_noexit "Setting \"interop\""
1943 $r $LCTL conf_param $fsname-MDT0000.lov.stripesize=4M || {
1944 error_noexit "Setting \"lov.stripesize\""
1947 $r $LCTL conf_param $fsname-MDT0000.mdd.atime_diff=70 || {
1948 error_noexit "Setting \"mdd.atime_diff\""
1952 if [ "$ff_convert" != "no" -a $(facet_fstype ost1) == "ldiskfs" ]; then
1953 $r $LCTL lfsck_start -M $fsname-OST0000 || {
1954 error_noexit "Start OI scrub on OST0"
1958 # The oi_scrub should be on ost1, but for test_32(),
1959 # all on the SINGLEMDS.
1960 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
1961 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1962 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
1963 error_noexit "Failed to get the expected 'completed'"
1967 local UPDATED=$($r $LCTL get_param -n \
1968 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1969 awk '/^updated/ { print $2 }')
1970 [ $UPDATED -ge 1 ] || {
1971 error_noexit "Only $UPDATED objects have been converted"
1976 if [ "$dne_upgrade" != "no" ]; then
1977 $r $LCTL conf_param \
1978 $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
1979 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
1982 $r $LCTL conf_param $fsname-MDT0001.failover.node=$nid || {
1983 error_noexit "Setting MDT1 \"failover.node\""
1986 $r $LCTL conf_param $fsname-MDT0001.lov.stripesize=4M || {
1987 error_noexit "Setting MDT1 \"lov.stripesize\""
1992 if [ "$writeconf" ]; then
1993 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
1994 error_noexit "Mounting the client"
1998 shall_cleanup_lustre=true
1999 $r $LCTL set_param debug="$PTLDEBUG"
2001 t32_verify_quota $node $fsname $tmp/mnt/lustre || {
2002 error_noexit "verify quota failed"
2006 if $r test -f $tmp/list; then
2008 # There is not a Test Framework API to copy files to or
2009 # from a remote node.
2011 # LU-2393 - do both sorts on same node to ensure locale
2013 local list_file=$tmp/list
2015 if $mdt2_is_available; then
2016 if [[ -d $tmp/mnt/lustre/striped_dir_old ]] &&
2017 $r test -f $tmp/list2; then
2018 list_file=$tmp/list2
2019 pushd $tmp/mnt/lustre/striped_dir_old
2021 pushd $tmp/mnt/lustre/remote_dir
2024 pushd $tmp/mnt/lustre
2026 $r cat $list_file | sort -k 6 >$tmp/list.orig
2027 ls -Rni --time-style=+%s | sort -k 6 |
2028 sed 's/\. / /' >$tmp/list || {
2034 # 32-bit and 64-bit clients use different algorithms to
2035 # convert FIDs into inode numbers. Hence, remove the
2036 # inode numbers from the lists, if the original list was
2037 # created on an architecture with different number of
2040 if [ $(t32_bits_per_long $(uname -m)) != \
2041 $(t32_bits_per_long $img_arch) ]; then
2042 echo "Different number of bits per \"long\"" \
2043 "from the disk image"
2044 for list in list.orig list; do
2045 sed -i -e 's/^[0-9]\+[ \t]\+//' \
2049 if ! diff -ub $tmp/list.orig $tmp/list; then
2050 error_noexit "list verification failed"
2054 echo "list verification skipped"
2057 if [ "$dne_upgrade" != "no" ]; then
2058 $LFS mkdir -i 1 -c2 $tmp/mnt/lustre/striped_dir || {
2059 error_noexit "set striped dir failed"
2063 $LFS setdirstripe -D -c2 $tmp/mnt/lustre/striped_dir
2065 pushd $tmp/mnt/lustre
2066 tar -cf - . --exclude=./striped_dir \
2067 --exclude=./striped_dir_old \
2068 --exclude=./remote_dir |
2069 tar -xvf - -C striped_dir 1>/dev/null || {
2070 error_noexit "cp to striped dir failed"
2076 # If it is upgrade from DNE (2.5), then rename the remote dir,
2077 # which is created in 2.5 to striped dir.
2078 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]]; then
2079 stripe_index=$($LFS getdirstripe -i \
2080 $tmp/mnt/lustre/remote_dir)
2082 [[ $stripe_index -eq 1 ]] || {
2083 error_noexit "get index \"$stripe_index\"" \
2084 "from remote dir failed"
2087 mv $tmp/mnt/lustre/remote_dir \
2088 $tmp/mnt/lustre/striped_dir/ || {
2089 error_noexit "mv remote dir failed"
2094 # If it is upgraded from DNE (2.7), then move the striped dir
2095 # which was created in 2.7 to the new striped dir.
2096 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]] &&
2097 [[ -d $tmp/mnt/lustre/striped_dir_old ]]; then
2098 stripe_count=$($LFS getdirstripe -c \
2099 $tmp/mnt/lustre/striped_dir_old)
2100 [[ $stripe_count -eq 2 ]] || {
2101 error_noexit "get count $stripe_count" \
2102 "from striped dir failed"
2105 mv $tmp/mnt/lustre/striped_dir_old \
2106 $tmp/mnt/lustre/striped_dir/ || {
2107 error_noexit "mv striped dir failed"
2113 $r $LCTL set_param -n osd*.*.force_sync=1
2114 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
2115 error_noexit "dd failed"
2118 rm -rf $tmp/mnt/lustre/tmp_file || {
2119 error_noexit "rm failed"
2123 if $r test -f $tmp/sha1sums; then
2124 # LU-2393 - do both sorts on same node to ensure locale
2126 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
2127 if [ "$dne_upgrade" != "no" ]; then
2128 pushd $tmp/mnt/lustre/striped_dir
2130 pushd $tmp/mnt/lustre
2133 find ! -path "*remote_dir*" ! -path "*striped_dir*" \
2134 ! -name .lustre -type f -exec sha1sum {} \; |
2135 sort -k 2 >$tmp/sha1sums || {
2137 error_noexit "sha1sum"
2141 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
2142 error_noexit "sha1sum verification failed"
2146 # if upgrade from DNE(2.5), then check remote directory
2147 # if upgrade from DNE(2.7), then check striped directory
2148 if $mdt2_is_available &&
2149 [[ "$dne_upgrade" != "no" ]]; then
2150 local new_dir="$tmp/mnt/lustre/striped_dir"
2151 local striped_dir_old="$new_dir/striped_dir_old"
2153 local dir_list="$new_dir/remote_dir"
2154 [[ ! -d $triped_dir_old ]] ||
2155 dir_list+=" $striped_dir_old"
2157 for dir in $dir_list; do
2159 find ! -name .lustre -type f \
2160 -exec sha1sum {} \; |
2161 sort -k 2 >$tmp/sha1sums || {
2163 error_noexit "sha1sum"
2167 if ! diff -ub $tmp/sha1sums.orig \
2169 error_noexit "sha1sum $dir" \
2176 echo "sha1sum verification skipped"
2179 if [ "$dne_upgrade" != "no" ]; then
2180 rm -rf $tmp/mnt/lustre/striped_dir || {
2181 error_noexit "remove remote dir failed"
2186 # migrate files/dirs to remote MDT, then move them back
2187 if [ $(lustre_version_code mds1) -ge $(version_code 2.7.50) -a \
2188 $dne_upgrade != "no" ]; then
2189 $r $LCTL set_param -n \
2190 mdt.${fsname}*.enable_remote_dir=1 2>/dev/null
2192 echo "test migration"
2193 pushd $tmp/mnt/lustre
2194 for dir in $(find ! -name .lustre ! -name . -type d); do
2195 mdt_index=$($LFS getdirstripe -i $dir)
2196 stripe_cnt=$($LFS getdirstripe -c $dir)
2197 if [ $mdt_index = 0 -a $stripe_cnt -le 1 ]; then
2198 $LFS mv -M 1 $dir || {
2200 error_noexit "migrate MDT1 failed"
2206 for dir in $(find ! -name . ! -name .lustre -type d); do
2207 mdt_index=$($LFS getdirstripe -i $dir)
2208 stripe_cnt=$($LFS getdirstripe -c $dir)
2209 if [ $mdt_index = 1 -a $stripe_cnt -le 1 ]; then
2210 $LFS mv -M 0 $dir || {
2212 error_noexit "migrate MDT0 failed"
2221 # When adding new data verification tests, please check for
2222 # the presence of the required reference files first, like
2223 # the "sha1sums" and "list" tests above, to avoid the need to
2224 # regenerate every image for each test addition.
2227 nrpcs_orig=$($LCTL get_param \
2228 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
2229 error_noexit "Getting \"max_rpcs_in_flight\""
2232 nrpcs=$((nrpcs_orig + 5))
2233 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=$nrpcs || {
2234 error_noexit "Changing \"max_rpcs_in_flight\""
2237 wait_update $HOSTNAME "$LCTL get_param \
2238 -n mdc.*MDT0000*.max_rpcs_in_flight" $nrpcs || {
2239 error_noexit "Verifying \"max_rpcs_in_flight\""
2243 umount $tmp/mnt/lustre || {
2244 error_noexit "Unmounting the client"
2247 shall_cleanup_lustre=false
2249 if [[ "$dne_upgrade" != "no" ]] || $mdt2_is_available; then
2250 $r $UMOUNT $tmp/mnt/mdt1 || {
2251 error_noexit "Unmounting the MDT2"
2254 shall_cleanup_mdt1=false
2257 $r $UMOUNT $tmp/mnt/mdt || {
2258 error_noexit "Unmounting the MDT"
2261 shall_cleanup_mdt=false
2263 $r $UMOUNT $tmp/mnt/ost || {
2264 error_noexit "Unmounting the OST"
2267 shall_cleanup_ost=false
2269 t32_reload_modules $node || {
2270 error_noexit "Reloading modules"
2274 if [[ $fstype == zfs ]]; then
2275 local poolname=t32fs-mdt1
2276 $r "$ZPOOL list -H $poolname >/dev/null 2>&1 ||
2277 $ZPOOL import -f -d $tmp $poolname"
2280 # mount a second time to make sure we didnt leave upgrade flag on
2281 $r $TUNEFS --dryrun $mdt_dev || {
2283 error_noexit "tunefs.lustre before remounting the MDT"
2287 mopts=exclude=$fsname-OST0000
2288 if [ $fstype == "ldiskfs" ]; then
2291 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2292 error_noexit "Remounting the MDT"
2295 shall_cleanup_mdt=true
2305 for tarball in $tarballs; do
2306 t32_test $tarball || let "rc += $?"
2310 run_test 32a "Upgrade (not live)"
2318 for tarball in $tarballs; do
2319 t32_test $tarball writeconf || let "rc += $?"
2323 run_test 32b "Upgrade with writeconf"
2330 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
2332 for tarball in $tarballs; do
2333 # Do not support 1_8 and 2_1 direct upgrade to DNE2 anymore */
2334 echo $tarball | grep "1_8" && continue
2335 echo $tarball | grep "2_1" && continue
2337 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
2341 run_test 32c "dne upgrade test"
2349 for tarball in $tarballs; do
2350 ff_convert=yes t32_test $tarball || rc=$?
2354 run_test 32d "convert ff test"
2356 test_33a() { # bug 12333, was test_33
2357 local FSNAME2=test-123
2358 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2361 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
2363 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
2364 local dev=${SINGLEMDS}_dev
2365 local MDSDEV=${!dev}
2366 is_blkdev $SINGLEMDS $MDSDEV &&
2367 skip_env "mixed loopback and real device not working" &&
2371 local fs2mdsdev=$(mdsdevname 1_2)
2372 local fs2ostdev=$(ostdevname 1_2)
2373 local fs2mdsvdev=$(mdsvdevname 1_2)
2374 local fs2ostvdev=$(ostvdevname 1_2)
2376 if [ $(facet_fstype mds1) == ldiskfs ]; then
2377 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
2380 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2381 --reformat $mkfsoptions $fs2mdsdev $fs2mdsvdev || exit 10
2382 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2383 --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \
2384 $fs2ostvdev || exit 10
2386 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
2387 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2388 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" ||
2389 error "$LCTL conf_param $FSNAME2.sys.timeout=200 failed"
2390 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2391 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2394 cp /etc/hosts $MOUNT2/ || error "copy /etc/hosts $MOUNT2/ failed"
2395 $GETSTRIPE $MOUNT2/hosts || error "$GETSTRIPE $MOUNT2/hosts failed"
2400 cleanup_nocli || error "cleanup_nocli failed with $?"
2402 run_test 33a "Mount ost with a large index number"
2404 test_33b() { # was test_34
2407 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
2408 # Drop lock cancelation reply during umount
2409 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
2410 do_facet client $LCTL set_param fail_loc=0x80000304
2411 #lctl set_param debug=-1
2412 umount_client $MOUNT
2413 cleanup || error "cleanup failed with $?"
2415 run_test 33b "Drop cancel during umount"
2419 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
2420 manual_umount_client
2422 do_facet client killall -USR1 multiop
2423 if [ $rc -eq 0 ]; then
2424 error "umount not fail!"
2427 cleanup || error "cleanup failed with rc $?"
2429 run_test 34a "umount with opened file should be fail"
2433 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2434 stop_mds || error "Unable to stop MDS"
2436 manual_umount_client --force || error "mtab after failed umount with $?"
2438 cleanup || error "cleanup failed with $?"
2440 run_test 34b "force umount with failed mds should be normal"
2444 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2445 stop_ost || error "Unable to stop OST1"
2447 manual_umount_client --force || error "mtab after failed umount with $?"
2449 cleanup || error "cleanup failed with $?"
2451 run_test 34c "force umount with failed ost should be normal"
2453 test_35a() { # bug 12459
2456 DBG_SAVE="`$LCTL get_param -n debug`"
2457 $LCTL set_param debug="ha"
2459 log "Set up a fake failnode for the MDS"
2461 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2462 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2463 do_facet mgs "$LCTL conf_param \
2464 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" ||
2465 error "Setting ${device}.failover.node=\
2466 $(h2$NETTYPE $FAKENID) failed."
2468 log "Wait for RECONNECT_INTERVAL seconds (10s)"
2471 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
2474 log "Stopping the MDT: $device"
2475 stop_mdt 1 || error "MDT0 stop fail"
2477 df $MOUNT > /dev/null 2>&1 &
2479 log "Restarting the MDT: $device"
2480 start_mdt 1 || error "MDT0 start fail"
2481 log "Wait for df ($DFPID) ... "
2484 $LCTL set_param debug="$DBG_SAVE"
2486 # retrieve from the log the first server that the client tried to
2487 # contact after the connection loss
2488 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2489 NEXTCONN=`awk "/${MSG}/ {start = 1;}
2490 /import_select_connection.*$device-mdc.* using connection/ {
2492 if (\\\$NF ~ /$FAKENID/)
2498 }" $TMP/lustre-log-$TESTNAME.log`
2499 [ "$NEXTCONN" != "0" ] &&
2500 error "Tried to connect to ${NEXTCONN} not last active server"
2501 cleanup || error "cleanup failed with $?"
2502 # remove nid settings
2503 writeconf_or_reformat
2505 run_test 35a "Reconnect to the last active server first"
2507 test_35b() { # bug 18674
2508 remote_mds || { skip "local MDS" && return 0; }
2512 $LCTL set_param debug="ha"
2514 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
2517 log "Set up a fake failnode for the MDS"
2519 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2520 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2521 do_facet mgs "$LCTL conf_param \
2522 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" ||
2523 error "Set ${device}.failover.node=\
2524 $(h2$NETTYPE $FAKENID) failed"
2526 local at_max_saved=0
2527 # adaptive timeouts may prevent seeing the issue
2528 if at_is_enabled; then
2529 at_max_saved=$(at_max_get mds)
2530 at_max_set 0 mds client
2533 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
2535 log "Injecting EBUSY on MDS"
2536 # Setting OBD_FAIL_MDS_RESEND=0x136
2537 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" ||
2538 error "unable to set param fail_loc=0x80000136"
2540 $LCTL set_param mdc.${FSNAME}*.stats=clear
2542 log "Creating a test file and stat it"
2543 touch $MOUNT/$tdir/$tfile || error "touch $MOUNT/$tdir/$tfile failed"
2544 stat $MOUNT/$tdir/$tfile
2546 log "Stop injecting EBUSY on MDS"
2547 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" ||
2548 error "unable to set param fail_loc=0"
2549 rm -f $MOUNT/$tdir/$tfile || error "remove $MOUNT/$tdir/$tfile failed"
2552 # restore adaptive timeout
2553 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
2555 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2557 CONNCNT=$($LCTL get_param mdc.${FSNAME}*.stats |
2558 awk '/mds_connect/{print $2}')
2560 # retrieve from the log if the client has ever tried to
2561 # contact the fake server after the loss of connection
2562 FAILCONN=`awk "BEGIN {ret = 0;}
2563 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
2565 if (\\\$NF ~ /$FAKENID/) {
2570 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
2572 [ "$FAILCONN" == "0" ] &&
2573 error "The client reconnection has not been triggered"
2574 [ "$FAILCONN" == "2" ] &&
2575 error "Primary server busy, client reconnect to failover failed"
2578 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
2579 # Reconnects are supposed to be rate limited to one every 5s
2580 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] &&
2581 error "Too many reconnects $CONNCNT"
2583 cleanup || error "cleanup failed with $?"
2584 # remove nid settings
2585 writeconf_or_reformat
2587 run_test 35b "Continue reconnection retries, if the active server is busy"
2590 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
2592 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] ||
2593 { skip "remote OST" && return 0; }
2596 local FSNAME2=test1234
2597 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2599 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
2601 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
2602 is_blkdev $SINGLEMDS $MDSDEV &&
2603 skip_env "mixed loopback and real device not working" && return
2606 local fs2mdsdev=$(mdsdevname 1_2)
2607 local fs2ostdev=$(ostdevname 1_2)
2608 local fs3ostdev=$(ostdevname 2_2)
2609 local fs2mdsvdev=$(mdsvdevname 1_2)
2610 local fs2ostvdev=$(ostvdevname 1_2)
2611 local fs3ostvdev=$(ostvdevname 2_2)
2613 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2614 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
2615 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
2616 # different one than the default value here.
2617 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2618 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
2619 add fs3ost $(mkfs_opts ost2 ${fs3ostdev}) --mgsnode=$MGSNID \
2620 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
2622 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
2623 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2624 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
2625 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2626 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2628 sleep 5 # until 11778 fixed
2630 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || error "dd failed"
2632 BKTOTAL=$($LCTL get_param -n obdfilter.*.kbytestotal |
2633 awk 'BEGIN{total=0}; {total+=$1}; END{print total}')
2634 BKFREE=$($LCTL get_param -n obdfilter.*.kbytesfree |
2635 awk 'BEGIN{free=0}; {free+=$1}; END{print free}')
2636 BKAVAIL=$($LCTL get_param -n obdfilter.*.kbytesavail |
2637 awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}')
2638 STRING=$(df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}')
2639 DFTOTAL=$(echo $STRING | cut -d, -f1)
2640 DFUSED=$(echo $STRING | cut -d, -f2)
2641 DFAVAIL=$(echo $STRING | cut -d, -f3)
2642 DFFREE=$(($DFTOTAL - $DFUSED))
2644 ALLOWANCE=$((64 * $OSTCOUNT))
2646 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
2647 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
2648 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
2651 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
2652 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
2653 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
2656 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
2657 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
2658 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
2663 stop fs3ost -f || error "unable to stop OST3"
2664 stop fs2ost -f || error "unable to stop OST2"
2665 stop fs2mds -f || error "unable to stop second MDS"
2666 unload_modules_conf || error "unable unload modules"
2669 run_test 36 "df report consistency on OSTs with different block size"
2672 local mntpt=$(facet_mntpt $SINGLEMDS)
2673 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
2674 local mdsdev_sym="$TMP/sym_mdt.img"
2675 local opts=$MDS_MOUNT_OPTS
2678 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2679 skip "Currently only applicable to ldiskfs-based MDTs"
2683 echo "MDS : $mdsdev"
2684 echo "SYMLINK : $mdsdev_sym"
2685 do_facet $SINGLEMDS rm -f $mdsdev_sym
2687 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
2689 echo "mount symlink device - $mdsdev_sym"
2691 if ! do_facet $SINGLEMDS test -b $mdsdev; then
2692 opts=$(csa_add "$opts" -o loop)
2694 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
2695 $mdsdev_sym $mntpt 2>&1)
2698 echo mount_op=$mount_op
2700 do_facet $SINGLEMDS "$UMOUNT $mntpt && rm -f $mdsdev_sym"
2702 if $(echo $mount_op | grep -q "unable to set tunable"); then
2703 error "set tunables failed for symlink device"
2706 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
2708 run_test 37 "verify set tunables works for symlink device"
2710 test_38() { # bug 14222
2711 local fstype=$(facet_fstype $SINGLEMDS)
2712 local mntpt=$(facet_mntpt $SINGLEMDS)
2717 local SRC="/etc /bin"
2718 local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT)
2719 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
2720 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2721 tar cf - $FILES | tar xf - -C $DIR/$tdir ||
2722 error "copying $SRC to $DIR/$tdir"
2724 umount_client $MOUNT || error "umount_client $MOUNT failed"
2725 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2726 stop_mds || error "Unable to stop MDS"
2727 log "delete lov_objid file on MDS"
2729 mount_fstype $SINGLEMDS || error "mount MDS failed (1)"
2731 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid; rm $mntpt/lov_objid"
2733 unmount_fstype $SINGLEMDS || error "umount failed (1)"
2735 # check create in mds_lov_connect
2736 start_mds || error "unable to start MDS"
2737 mount_client $MOUNT || error "mount_client $MOUNT failed"
2739 [ $V ] && log "verifying $DIR/$tdir/$f"
2740 diff -q $f $DIR/$tdir/$f || ERROR=y
2742 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2743 if [ "$ERROR" = "y" ]; then
2744 # check it's updates in sync
2745 umount_client $MOUNT
2747 mount_fstype $SIGNLEMDS
2748 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2749 unmount_fstype $SINGLEMDS
2750 error "old and new files are different after connect" || true
2752 touch $DIR/$tdir/f2 || error "f2 file create failed"
2754 # check it's updates in sync
2755 umount_client $MOUNT || error "second umount_client $MOUNT failed"
2758 mount_fstype $SINGLEMDS || error "mount MDS failed (3)"
2760 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2761 do_facet $SINGLEMDS dd if=/dev/zero of=$mntpt/lov_objid.clear count=8
2763 unmount_fstype $SINGLEMDS || error "umount failed (3)"
2765 start_mds || error "unable to start MDS"
2766 mount_client $MOUNT || error "mount_client $MOUNT failed"
2768 [ $V ] && log "verifying $DIR/$tdir/$f"
2769 diff -q $f $DIR/$tdir/$f || ERROR=y
2771 touch $DIR/$tdir/f3 || error "f3 file create failed"
2772 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2773 umount_client $MOUNT || error "third umount_client $MOUNT failed"
2775 mount_fstype $SINGLEMDS || error "mount MDS failed (4)"
2776 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2777 unmount_fstype $SINGLEMDS || error "umount failed (4)"
2779 [ "$ERROR" = "y" ] &&
2780 error "old and new files are different after sync" || true
2782 log "files compared the same"
2783 cleanup || error "cleanup failed with $?"
2785 run_test 38 "MDS recreates missing lov_objid file from OST data"
2790 cleanup || error "cleanup failed with $?"
2791 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
2792 error "memory leak detected" || true
2794 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
2796 test_40() { # bug 15759
2797 start_ost || error "Unable to start OST1"
2798 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
2799 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
2801 cleanup || error "cleanup failed with rc $?"
2803 run_test 40 "race during service thread startup"
2805 test_41a() { #bug 14134
2806 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2807 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2808 skip "Loop devices does not work with nosvc option"
2812 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2814 start_mdt 1 -o nosvc -n
2815 if [ $MDSCOUNT -ge 2 ]; then
2816 for num in $(seq 2 $MDSCOUNT); do
2817 start_mdt $num || return
2820 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
2821 start_mdt 1 -o nomgs,force
2822 mount_client $MOUNT || error "mount_client $MOUNT failed"
2825 echo "blah blah" > $MOUNT/$tfile
2828 umount_client $MOUNT || error "umount_client $MOUNT failed"
2829 stop ost1 -f || error "unable to stop OST1"
2830 stop_mds || error "Unable to stop MDS"
2831 stop_mds || error "Unable to stop MDS on second try"
2833 run_test 41a "mount mds with --nosvc and --nomgs"
2836 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2837 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2838 skip "Loop devices does not work with nosvc option"
2842 ! combined_mgs_mds && skip "needs combined mgs device" && return 0
2846 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2848 start_mdt 1 -o nosvc -n
2849 if [ $MDSCOUNT -ge 2 ]; then
2850 for num in $(seq 2 $MDSCOUNT); do
2851 start_mdt $num || return
2854 start_ost || error "Unable to start OST1"
2855 start_mdt 1 -o nomgs,force
2856 mount_client $MOUNT || error "mount_client $MOUNT failed"
2859 echo "blah blah" > $MOUNT/$tfile
2860 cat $MOUNT/$tfile || error "cat $MOUNT/$tfile failed"
2862 umount_client $MOUNT || error "umount_client $MOUNT failed"
2863 stop_ost || error "Unable to stop OST1"
2864 stop_mds || error "Unable to stop MDS"
2865 stop_mds || error "Unable to stop MDS on second try"
2867 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
2870 local server_version=$(lustre_version_code $SINGLEMDS)
2872 [[ $server_version -ge $(version_code 2.6.52) ]] ||
2873 [[ $server_version -ge $(version_code 2.5.26) &&
2874 $server_version -lt $(version_code 2.5.50) ]] ||
2875 [[ $server_version -ge $(version_code 2.5.4) &&
2876 $server_version -lt $(version_code 2.5.11) ]] ||
2877 { skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"; return; }
2880 # MDT concurrent start
2881 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
2882 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x716"
2883 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS &
2885 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS &
2886 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x0"
2892 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
2893 echo "1st MDT start succeed"
2894 echo "2nd MDT start failed with $rc2"
2895 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
2896 echo "1st MDT start failed with $rc"
2897 echo "2nd MDT start succeed"
2900 error "unexpected concurrent MDT mounts result, rc=$rc rc2=$rc2"
2903 if [ $MDSCOUNT -ge 2 ]; then
2904 for num in $(seq 2 $MDSCOUNT); do
2905 start_mdt $num || return
2909 # OST concurrent start
2911 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
2912 do_facet ost1 "$LCTL set_param fail_loc=0x716"
2913 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &
2915 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &
2916 do_facet ost1 "$LCTL set_param fail_loc=0x0"
2922 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
2923 echo "1st OST start succeed"
2924 echo "2nd OST start failed with $rc2"
2925 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
2926 echo "1st OST start failed with $rc"
2927 echo "2nd OST start succeed"
2931 error "unexpected concurrent OST mounts result, rc=$rc rc2=$rc2"
2937 # verify everything ok
2942 error "MDT(s) start failed"
2950 error "OST(s) start failed"
2958 error "client start failed"
2965 error "client mount failed"
2969 run_test 41c "concurrent mounts of MDT/OST should all fail but one"
2971 test_42() { #bug 14693
2973 check_mount || error "client was not mounted"
2975 do_facet mgs $LCTL conf_param $FSNAME.llite.some_wrong_param=10
2976 umount_client $MOUNT ||
2977 error "unmounting client failed with invalid llite param"
2978 mount_client $MOUNT ||
2979 error "mounting client failed with invalid llite param"
2981 do_facet mgs $LCTL conf_param $FSNAME.sys.some_wrong_param=20
2982 cleanup || error "stopping $FSNAME failed with invalid sys param"
2984 check_mount || error "client was not mounted with invalid sys param"
2985 cleanup || error "stopping $FSNAME failed with invalid sys param"
2987 run_test 42 "allow client/server mount/unmount with invalid config param"
2990 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.58) ]] ||
2991 { skip "Need MDS version at least 2.5.58" && return 0; }
2992 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
2995 USER1=$(getent passwd | grep :$ID1:$ID1: | cut -d: -f1)
2996 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1" &&
3000 chmod ugo+x $DIR || error "chmod 0 failed"
3001 set_conf_param_and_check mds \
3002 "$LCTL get_param -n mdt.$FSNAME-MDT0000.root_squash" \
3003 "$FSNAME.mdt.root_squash" \
3005 wait_update $HOSTNAME \
3006 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3008 error "check llite root_squash failed!"
3009 set_conf_param_and_check mds \
3010 "$LCTL get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
3011 "$FSNAME.mdt.nosquash_nids" \
3013 wait_update $HOSTNAME \
3014 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3016 error "check llite nosquash_nids failed!"
3019 # create set of test files
3021 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
3022 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
3023 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
3025 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
3026 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
3028 mkdir $DIR/$tdir-rootdir || error "mkdir failed"
3029 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
3030 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
3032 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
3033 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
3034 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
3037 # check root_squash:
3038 # set root squash UID:GID to RUNAS_ID
3039 # root should be able to access only files owned by RUNAS_ID
3041 set_conf_param_and_check mds \
3042 "$LCTL get_param -n mdt.$FSNAME-MDT0000.root_squash" \
3043 "$FSNAME.mdt.root_squash" \
3044 "$RUNAS_ID:$RUNAS_ID"
3045 wait_update $HOSTNAME \
3046 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3047 "$RUNAS_ID:$RUNAS_ID" ||
3048 error "check llite root_squash failed!"
3050 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
3051 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3052 error "$ST: root read permission is denied"
3053 echo "$ST: root read permission is granted - ok"
3056 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3057 error "$ST: root write permission is denied"
3058 echo "$ST: root write permission is granted - ok"
3060 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3061 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3062 error "$ST: root read permission is granted"
3063 echo "$ST: root read permission is denied - ok"
3066 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3067 error "$ST: root write permission is granted"
3068 echo "$ST: root write permission is denied - ok"
3070 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3071 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null &&
3072 error "$ST: root unlink permission is granted"
3073 echo "$ST: root unlink permission is denied - ok"
3075 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null &&
3076 error "$ST: root create permission is granted"
3077 echo "$ST: root create permission is denied - ok"
3080 # check root_squash is enforced independently
3081 # of client cache content
3083 # access file by USER1, keep access open
3084 # root should be denied access to user file
3086 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
3090 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
3091 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3092 { kill $pid; error "$ST: root read permission is granted"; }
3093 echo "$ST: root read permission is denied - ok"
3096 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3097 { kill $pid; error "$ST: root write permission is granted"; }
3098 echo "$ST: root write permission is denied - ok"
3104 # check nosquash_nids:
3105 # put client's NID into nosquash_nids list,
3106 # root should be able to access root file after that
3108 local NIDLIST=$($LCTL list_nids all | tr '\n' ' ')
3109 NIDLIST="2@gni $NIDLIST 192.168.0.[2,10]@tcp"
3110 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
3111 set_conf_param_and_check mds \
3112 "$LCTL get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
3113 "$FSNAME-MDTall.mdt.nosquash_nids" \
3115 wait_update $HOSTNAME \
3116 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3118 error "check llite nosquash_nids failed!"
3120 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3121 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3122 error "$ST: root read permission is denied"
3123 echo "$ST: root read permission is granted - ok"
3126 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3127 error "$ST: root write permission is denied"
3128 echo "$ST: root write permission is granted - ok"
3130 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3131 rm $DIR/$tdir-rootdir/tfile-1 ||
3132 error "$ST: root unlink permission is denied"
3133 echo "$ST: root unlink permission is granted - ok"
3134 touch $DIR/$tdir-rootdir/tfile-2 ||
3135 error "$ST: root create permission is denied"
3136 echo "$ST: root create permission is granted - ok"
3137 cleanup || error "cleanup failed with $?"
3139 run_test 43a "check root_squash and nosquash_nids"
3141 test_43b() { # LU-5690
3142 [[ $(lustre_version_code mgs) -ge $(version_code 2.7.62) ]] ||
3143 { skip "Need MGS version 2.7.62+"; return; }
3145 if [[ -z "$fs2mds_DEV" ]]; then
3146 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
3147 skip_env "mixed loopback and real device not working" && return
3150 local fs2mdsdev=$(mdsdevname 1_2)
3151 local fs2mdsvdev=$(mdsvdevname 1_2)
3153 # temporarily use fs2mds as fs2mgs
3155 local fs2mgsdev=$fs2mdsdev
3156 local fs2mgsvdev=$fs2mdsvdev
3158 local fsname=test1234
3160 load_module llite/lustre
3161 local client_ip=$(host_nids_address $HOSTNAME $NETTYPE)
3162 local host=${client_ip//*./}
3163 local net=${client_ip/%$host/}
3164 local nosquash_nids=$(h2$NETTYPE $net[$host,$host,$host])
3166 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
3167 --param mdt.root_squash=$RUNAS_ID:$RUNAS_ID \
3168 --param mdt.nosquash_nids=$nosquash_nids \
3169 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
3170 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS || error "start fs2mgs failed"
3171 stop $fs2mgs -f || error "stop fs2mgs failed"
3173 run_test 43b "parse nosquash_nids with commas in expr_list"
3175 umount_client $MOUNT
3180 check_mount || error "check_mount"
3181 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
3183 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
3184 for VAL in $UUIDS; do
3185 NID=$(echo $VAL | cut -d= -f1)
3186 CLUUID=$(echo $VAL | cut -d= -f2)
3187 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
3189 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
3190 cleanup || error "cleanup failed with $?"
3192 run_test 44 "mounted client proc entry exists"
3196 check_mount || error "check_mount"
3197 stop_mds || error "Unable to stop MDS"
3201 #define OBD_FAIL_PTLRPC_LONG_UNLINK 0x50f
3202 do_facet client "$LCTL set_param fail_loc=0x50f"
3205 manual_umount_client --force || error "manual_umount_client failed"
3206 do_facet client "$LCTL set_param fail_loc=0x0"
3207 start_mds || error "unable to start MDS"
3208 mount_client $MOUNT || error "mount_client $MOUNT failed"
3209 cleanup || error "cleanup failed with $?"
3211 run_test 45 "long unlink handling in ptlrpcd"
3218 umount_client $MOUNT2 || rc=$?
3219 umount_client $MOUNT || rc=$?
3220 while [ $count -gt 0 ]; do
3221 stop ost${count} -f || rc=$?
3225 cleanup_nocli || rc=$?
3226 #writeconf to remove all ost2 traces for subsequent tests
3227 writeconf_or_reformat
3232 echo "Testing with $OSTCOUNT OSTs"
3234 start_mds || error "unable to start MDS"
3235 #first client should see only one ost
3236 start_ost || error "Unable to start OST1"
3237 wait_osc_import_state mds ost FULL
3239 mount_client $MOUNT || error "mount_client $MOUNT failed"
3240 trap "cleanup_46a $OSTCOUNT" EXIT ERR
3243 for (( i=2; i<=$OSTCOUNT; i++ )); do
3244 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
3245 error "start_ost$i $(ostdevname $i) failed"
3248 # wait until osts in sync
3249 for (( i=2; i<=$OSTCOUNT; i++ )); do
3250 wait_osc_import_state mds ost$i FULL
3251 wait_osc_import_state client ost$i FULL
3254 #second client see all ost's
3256 mount_client $MOUNT2 || error "mount_client failed"
3257 $SETSTRIPE -c -1 $MOUNT2 || error "$SETSTRIPE -c -1 $MOUNT2 failed"
3258 $GETSTRIPE $MOUNT2 || error "$GETSTRIPE $MOUNT2 failed"
3260 echo "ok" > $MOUNT2/widestripe
3261 $GETSTRIPE $MOUNT2/widestripe ||
3262 error "$GETSTRIPE $MOUNT2/widestripe failed"
3263 # fill acl buffer for avoid expand lsm to them
3264 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd |
3266 setfacl -m $acl $MOUNT2/widestripe
3270 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3272 cleanup_46a $OSTCOUNT || error "cleanup_46a failed"
3274 run_test 46a "handle ost additional - wide striped file"
3279 check_mount || error "check_mount failed"
3280 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
3284 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3285 if echo $ns | grep "MDT[[:digit:]]*"; then
3288 lrs=$(echo $ns | sed 's/.*lru_size=//')
3289 lru_size[count]=$lrs
3294 facet_failover $SINGLEMDS
3295 client_up || error "client_up failed"
3298 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3299 if echo $ns | grep "MDT[[:digit:]]*"; then
3302 lrs=$(echo $ns | sed 's/.*lru_size=//')
3303 if ! test "$lrs" -eq "${lru_size[count]}"; then
3304 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
3305 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
3310 cleanup || error "cleanup failed with $?"
3312 run_test 47 "server restart does not make client loss lru_resize settings"
3317 # reformat after this test is needed - if the test fails,
3318 # we will have unkillable file at FS
3322 test_48() { # bug 17636
3325 check_mount || error "check_mount failed"
3327 $SETSTRIPE -c -1 $MOUNT || error "$SETSTRIPE -c -1 $MOUNT failed"
3328 $GETSTRIPE $MOUNT || error "$GETSTRIPE $MOUNT failed"
3330 echo "ok" > $MOUNT/widestripe
3331 $GETSTRIPE $MOUNT/widestripe ||
3332 error "$GETSTRIPE $MOUNT/widestripe failed"
3334 trap cleanup_48 EXIT ERR
3336 # fill acl buffer for avoid expand lsm to them
3337 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
3338 setfacl -m $acl $MOUNT/widestripe
3341 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3345 run_test 48 "too many acls on file"
3347 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
3348 test_49a() { # bug 17710
3349 local timeout_orig=$TIMEOUT
3350 local ldlm_timeout_orig=$LDLM_TIMEOUT
3351 local LOCAL_TIMEOUT=20
3353 LDLM_TIMEOUT=$LOCAL_TIMEOUT
3354 TIMEOUT=$LOCAL_TIMEOUT
3358 check_mount || error "client mount failed"
3360 echo "check ldlm_timout..."
3361 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3362 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3363 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3365 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3366 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3369 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
3370 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
3373 umount_client $MOUNT || error "umount_client $MOUNT failed"
3374 stop_ost || error "problem stopping OSS"
3375 stop_mds || error "problem stopping MDS"
3377 LDLM_TIMEOUT=$ldlm_timeout_orig
3378 TIMEOUT=$timeout_orig
3380 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3382 test_49b() { # bug 17710
3383 local timeout_orig=$TIMEOUT
3384 local ldlm_timeout_orig=$LDLM_TIMEOUT
3385 local LOCAL_TIMEOUT=20
3387 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
3388 TIMEOUT=$LOCAL_TIMEOUT
3392 check_mount || error "client mount failed"
3394 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3395 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3396 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3398 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3399 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3402 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
3403 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
3406 cleanup || error "cleanup failed"
3408 LDLM_TIMEOUT=$ldlm_timeout_orig
3409 TIMEOUT=$timeout_orig
3411 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3414 # Test both statfs and lfs df and fail if either one fails
3415 multiop_bg_pause $1 f_
3418 killall -USR1 multiop
3419 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
3420 wait $PID || { RC1=$?; log "multiop return error "; }
3427 if [ $RC2 -eq 0 ]; then
3429 log "lazystatfs df failed"
3433 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
3439 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3440 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3442 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
3444 cleanup || error "cleanup failed with rc $?"
3446 run_test 50a "lazystatfs all servers available"
3450 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3451 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3453 # Wait for client to detect down OST
3454 stop_ost || error "Unable to stop OST1"
3455 wait_osc_import_state mds ost DISCONN
3457 lazystatfs $MOUNT || error "lazystatfs should not return EIO"
3459 umount_client $MOUNT || error "Unable to unmount client"
3460 stop_mds || error "Unable to stop MDS"
3462 run_test 50b "lazystatfs all servers down"
3465 start_mds || error "Unable to start MDS"
3466 start_ost || error "Unable to start OST1"
3467 start_ost2 || error "Unable to start OST2"
3468 mount_client $MOUNT || error "Unable to mount client"
3469 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3470 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3472 # Wait for client to detect down OST
3473 stop_ost || error "Unable to stop OST1"
3474 wait_osc_import_state mds ost DISCONN
3475 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3477 umount_client $MOUNT || error "Unable to unmount client"
3478 stop_ost2 || error "Unable to stop OST2"
3479 stop_mds || error "Unable to stop MDS"
3480 #writeconf to remove all ost2 traces for subsequent tests
3481 writeconf_or_reformat
3483 run_test 50c "lazystatfs one server down"
3486 start_mds || error "Unable to start MDS"
3487 start_ost || error "Unable to start OST1"
3488 start_ost2 || error "Unable to start OST2"
3489 mount_client $MOUNT || error "Unable to mount client"
3490 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3491 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3493 # Issue the statfs during the window where the client still
3494 # belives the OST to be available but it is in fact down.
3495 # No failure just a statfs which hangs for a timeout interval.
3496 stop_ost || error "Unable to stop OST1"
3497 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3499 umount_client $MOUNT || error "Unable to unmount client"
3500 stop_ost2 || error "Unable to stop OST2"
3501 stop_mds || error "Unable to stop MDS"
3502 #writeconf to remove all ost2 traces for subsequent tests
3503 writeconf_or_reformat
3505 run_test 50d "lazystatfs client/server conn race"
3512 start_mds || error "Unable to start MDS"
3513 #first client should see only one ost
3514 start_ost || error "Unable to start OST1"
3515 wait_osc_import_state mds ost FULL
3517 # Wait for client to detect down OST
3518 stop_ost || error "Unable to stop OST1"
3519 wait_osc_import_state mds ost DISCONN
3521 mount_client $MOUNT || error "Unable to mount client"
3522 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3524 multiop_bg_pause $MOUNT _f
3528 if [ $RC1 -ne 0 ]; then
3529 log "multiop failed $RC1"
3532 sleep $(( $TIMEOUT+1 ))
3534 [ $? -ne 0 ] && error "process isn't sleep"
3535 start_ost || error "Unable to start OST1"
3536 wait $pid || error "statfs failed"
3539 umount_client $MOUNT || error "Unable to unmount client"
3540 stop_ost || error "Unable to stop OST1"
3541 stop_mds || error "Unable to stop MDS"
3543 run_test 50e "normal statfs all servers down"
3548 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
3550 start_mds || error "Unable to start mds"
3551 #first client should see only one ost
3552 start_ost || error "Unable to start OST1"
3553 wait_osc_import_state mds ost FULL
3555 start_ost2 || error "Unable to start OST2"
3556 wait_osc_import_state mds ost2 FULL
3558 # Wait for client to detect down OST
3559 stop_ost2 || error "Unable to stop OST2"
3561 wait_osc_import_state mds ost2 DISCONN
3562 mount_client $MOUNT || error "Unable to mount client"
3563 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3565 multiop_bg_pause $MOUNT _f
3569 if [ $RC1 -ne 0 ]; then
3570 log "lazystatfs multiop failed $RC1"
3573 sleep $(( $TIMEOUT+1 ))
3575 [ $? -ne 0 ] && error "process isn't sleep"
3576 start_ost2 || error "Unable to start OST2"
3577 wait $pid || error "statfs failed"
3578 stop_ost2 || error "Unable to stop OST2"
3581 umount_client $MOUNT -f || error "Unable to unmount client"
3582 stop_ost || error "Unable to stop OST1"
3583 stop_mds || error "Unable to stop MDS"
3584 #writeconf to remove all ost2 traces for subsequent tests
3585 writeconf_or_reformat
3587 run_test 50f "normal statfs one server in down"
3590 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3592 start_ost2 || error "Unable to start OST2"
3593 wait_osc_import_state mds ost2 FULL
3594 wait_osc_import_state client ost2 FULL
3596 local PARAM="${FSNAME}-OST0001.osc.active"
3598 $SETSTRIPE -c -1 $DIR/$tfile || error "$SETSTRIPE failed"
3599 do_facet mgs $LCTL conf_param $PARAM=0 ||
3600 error "Unable to deactivate OST"
3602 umount_client $MOUNT || error "Unable to unmount client"
3603 mount_client $MOUNT || error "Unable to mount client"
3604 # This df should not cause a panic
3607 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
3608 rm -f $DIR/$tfile || error "unable to remove file $DIR/$tfile"
3609 umount_client $MOUNT || error "Unable to unmount client"
3610 stop_ost2 || error "Unable to stop OST2"
3611 stop_ost || error "Unable to stop OST1"
3612 stop_mds || error "Unable to stop MDS"
3613 #writeconf to remove all ost2 traces for subsequent tests
3614 writeconf_or_reformat
3616 run_test 50g "deactivated OST should not cause panic"
3620 # prepare MDT/OST, make OSC inactive for OST1
3621 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3623 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3624 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
3625 error "tunefs OST1 failed"
3626 start_mds || error "Unable to start MDT"
3627 start_ost || error "Unable to start OST1"
3628 start_ost2 || error "Unable to start OST2"
3629 mount_client $MOUNT || error "client start failed"
3631 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3633 # activatate OSC for OST1
3634 local TEST="$LCTL get_param -n osc.${FSNAME}-OST0000-osc-[!M]*.active"
3635 set_conf_param_and_check client \
3636 "$TEST" "${FSNAME}-OST0000.osc.active" 1 ||
3637 error "Unable to activate OST1"
3639 mkdir $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3640 $SETSTRIPE -c -1 -i 0 $DIR/$tdir/2 ||
3641 error "$SETSTRIPE $DIR/$tdir/2 failed"
3642 sleep 1 && echo "create a file after OST1 is activated"
3644 createmany -o $DIR/$tdir/2/$tfile-%d 1
3646 # check OSC import is working
3647 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
3648 error "some OSC imports are still not connected"
3651 umount_client $MOUNT || error "Unable to umount client"
3652 stop_ost2 || error "Unable to stop OST2"
3653 cleanup_nocli || error "cleanup_nocli failed with $?"
3655 run_test 50h "LU-642: activate deactivated OST"
3658 # prepare MDT/OST, make OSC inactive for OST1
3659 [ "$MDSCOUNT" -lt "2" ] && skip_env "$MDSCOUNT < 2, skipping" && return
3661 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3662 do_facet mds2 "$TUNEFS --param mdc.active=0 $(mdsdevname 2)" ||
3663 error "tunefs MDT2 failed"
3664 start_mds || error "Unable to start MDT"
3665 start_ost || error "Unable to start OST1"
3666 start_ost2 || error "Unable to start OST2"
3667 mount_client $MOUNT || error "client start failed"
3669 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3671 $LCTL conf_param ${FSNAME}-MDT0000.mdc.active=0 &&
3672 error "deactive MDC0 succeeds"
3673 # activate MDC for MDT2
3674 local TEST="$LCTL get_param -n mdc.${FSNAME}-MDT0001-mdc-[!M]*.active"
3675 set_conf_param_and_check client \
3676 "$TEST" "${FSNAME}-MDT0001.mdc.active" 1 ||
3677 error "Unable to activate MDT2"
3679 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
3680 if [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.60) ]
3682 wait_dne_interconnect
3684 $LFS mkdir -i1 $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3686 createmany -o $DIR/$tdir/2/$tfile-%d 1 || error "create files failed"
3688 rm -rf $DIR/$tdir/2 || error "unlink dir failed"
3690 # deactivate MDC for MDT2
3691 local TEST="$LCTL get_param -n mdc.${FSNAME}-MDT0001-mdc-[!M]*.active"
3692 set_conf_param_and_check client \
3693 "$TEST" "${FSNAME}-MDT0001.mdc.active" 0 ||
3694 error "Unable to deactivate MDT2"
3696 $LFS mkdir -i1 $DIR/$tdir/2 &&
3697 error "mkdir $DIR/$tdir/2 succeeds after deactive MDT"
3700 umount_client $MOUNT || error "Unable to umount client"
3705 run_test 50i "activate deactivated MDT"
3708 local LOCAL_TIMEOUT=20
3712 check_mount || error "check_mount failed"
3714 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
3715 $SETSTRIPE -c -1 $MOUNT/$tdir ||
3716 error "$SETSTRIPE -c -1 $MOUNT/$tdir failed"
3717 #define OBD_FAIL_MDS_REINT_DELAY 0x142
3718 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x142"
3719 touch $MOUNT/$tdir/$tfile &
3722 start_ost2 || error "Unable to start OST1"
3724 stop_ost2 || error "Unable to stop OST1"
3725 umount_client $MOUNT -f || error “unmount $MOUNT failed”
3726 cleanup_nocli || error “stop server failed”
3727 #writeconf to remove all ost2 traces for subsequent tests
3728 writeconf_or_reformat
3730 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
3739 do_node $node mkdir -p $dest
3740 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3742 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
3743 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3744 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
3746 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
3747 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
3757 local backup2=${TMP}/backup2
3759 do_node $node mkdir -p $backup2
3760 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3762 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
3763 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3764 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
3766 do_node $node "diff -rq $backup $backup2"
3767 [ $? -eq 0 ] || { error "contents differ"; return 3; }
3769 local xattrs2=${TMP}/xattrs2
3770 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
3771 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
3773 do_node $node "diff $xattrs $xattrs2"
3774 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
3776 do_node $node "rm -rf $backup2 $xattrs2"
3777 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
3781 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3782 skip "Only applicable to ldiskfs-based MDTs"
3786 start_mds || error "Unable to start MDS"
3787 start_ost || error "Unable to start OST1"
3788 mount_client $MOUNT || error "Unable to mount client"
3791 local ost1mnt=$(facet_mntpt ost1)
3792 local ost1node=$(facet_active_host ost1)
3793 local ost1tmp=$TMP/conf52
3796 mkdir $DIR/$tdir || error "Unable to create $DIR/$tdir"
3797 touch $TMP/modified_first || error "Unable to create temporary file"
3798 local mtime=$(stat -c %Y $TMP/modified_first)
3799 do_node $ost1node "mkdir -p $ost1tmp &&
3800 touch -m -d @$mtime $ost1tmp/modified_first" ||
3801 error "Unable to create temporary file"
3804 $SETSTRIPE -c -1 -S 1M $DIR/$tdir || error "$SETSTRIPE failed"
3806 for (( i=0; i < nrfiles; i++ )); do
3807 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c ||
3808 error "multiop failed"
3814 echo backup files to $TMP/files
3815 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
3816 copy_files_xattrs $(hostname) $TMP/files $TMP/file_xattrs $files ||
3817 error "Unable to copy files"
3819 umount_client $MOUNT || error "Unable to umount client"
3820 stop_ost || error "Unable to stop ost1"
3822 echo mount ost1 as ldiskfs
3823 do_node $ost1node mkdir -p $ost1mnt || error "Unable to create $ost1mnt"
3824 if ! do_node $ost1node test -b $ost1_dev; then
3827 do_node $ost1node mount -t $(facet_fstype ost1) $loop $ost1_dev \
3829 error "Unable to mount ost1 as ldiskfs"
3832 echo backup objects to $ost1tmp/objects
3833 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
3834 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
3835 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
3837 error "Unable to copy objects"
3839 # move objects to lost+found
3840 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
3841 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
3843 do_node $ost1node "umount $ost1mnt" ||
3844 error "Unable to umount ost1 as ldiskfs"
3846 start_ost || error "Unable to start OST1"
3847 mount_client $MOUNT || error "Unable to mount client"
3849 local REPAIRED=$(do_node $ost1node "$LCTL get_param \
3850 -n osd-ldiskfs.$FSNAME-OST0000.oi_scrub" |
3851 awk '/^lf_repa[ri]*ed/ { print $2 }')
3852 [ $REPAIRED -gt 0 ] ||
3853 error "Some entry under /lost+found should be repaired"
3856 diff_files_xattrs $(hostname) $TMP/files $TMP/file_xattrs $files ||
3857 error "Unable to diff files"
3859 rm -rf $TMP/files $TMP/file_xattrs ||
3860 error "Unable to delete temporary files"
3861 do_node $ost1node "rm -rf $ost1tmp" ||
3862 error "Unable to delete temporary files"
3863 cleanup || error "cleanup failed with $?"
3865 run_test 52 "check recovering objects from lost+found"
3867 # Checks threads_min/max/started for some service
3869 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
3870 # parameter pattern prefix like 'ost.*.ost'.
3883 local msg="Insane $modname thread counts"
3884 local ncpts=$(check_cpt_number $facet)
3888 check_mount || return 41
3890 # We need to expand $parampat, but it may match multiple parameters, so
3891 # we'll pick the first one
3892 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
3893 error "Couldn't expand ${parampat}.threads_min parameter name"
3897 # Remove the .threads_min part
3898 paramp=${paramp%.threads_min}
3900 # Check for sanity in defaults
3901 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3903 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3905 tstarted=$(do_facet $facet "$LCTL get_param \
3906 -n ${paramp}.threads_started" || echo 0)
3907 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' ||
3909 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
3911 nthrs=$(expr $tmax - $tmin)
3912 if [ $nthrs -lt $ncpts ]; then
3918 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
3919 skip_env "module parameter forced $facet thread count" &&
3920 tmin=3 && tmax=$((3 * tmax))
3922 # Check that we can change min/max
3923 do_facet $facet "$LCTL set_param \
3924 ${paramp}.threads_min=$((tmin + nthrs))"
3925 do_facet $facet "$LCTL set_param \
3926 ${paramp}.threads_max=$((tmax - nthrs))"
3927 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3929 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3931 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) &&
3932 $tmax2 == ($tmax - $nthrs)))' || return $?
3934 # Check that we can set min/max to the same value
3935 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3937 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$tmin"
3938 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3940 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3942 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
3944 # Check that we can't set max < min
3945 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$((tmin - 1))"
3946 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3948 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3950 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
3952 # We need to ensure that we get the module options desired; to do this
3953 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
3954 LOAD_MODULES_REMOTE=true
3957 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
3958 setmodopts -a $modname "$newvalue" oldvalue
3961 check_mount || return 41
3963 # Restore previous setting of MODOPTS_*
3964 setmodopts $modname "$oldvalue"
3966 # Check that $opts took
3967 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min")
3968 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max")
3969 tstarted=$(do_facet $facet \
3970 "$LCTL get_param -n ${paramp}.threads_started")
3971 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
3980 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
3981 cleanup || error "cleanup failed with rc $?"
3983 run_test 53a "check OSS thread count params"
3987 local mds=$(do_facet $SINGLEMDS "$LCTL get_param \
3988 -N mds.*.*.threads_max 2>/dev/null")
3989 if [ -z "$mds" ]; then
3990 #running this on an old MDT
3991 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
3993 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
3995 cleanup || error "cleanup failed with $?"
3997 run_test 53b "check MDS thread count params"
4000 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4001 skip "Only applicable to ldiskfs-based MDTs"
4005 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p ||
4006 error "llverdev failed with rc=$?"
4009 run_test 54a "test llverdev and partial verify of device"
4012 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4013 skip "Only applicable to ldiskfs-based MDTs"
4018 run_llverfs $MOUNT -p || error "llverfs failed with rc=$?"
4019 cleanup || error "cleanup failed with rc=$?"
4021 run_test 54b "test llverfs and partial verify of filesystem"
4025 local max_ost_index=$1
4026 echo -n $(((max_ost_index + 1) * 8))
4030 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4031 skip "Only applicable to ldiskfs-based MDTs"
4035 local mdsdev=$(mdsdevname 1)
4036 local mdsvdev=$(mdsvdevname 1)
4040 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
4042 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
4043 --reformat $(ostdevname 1) $(ostvdevname 1)
4049 echo checking size of lov_objid for ost index $i
4050 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
4051 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
4052 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
4054 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
4061 run_test 55 "check lov_objid size"
4064 local server_version=$(lustre_version_code $SINGLEMDS)
4065 local mds_journal_size_orig=$MDSJOURNALSIZE
4070 for num in $(seq 1 $MDSCOUNT); do
4073 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
4074 $(ostdevname 1) $(ostvdevname 1)
4075 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
4076 $(ostdevname 2) $(ostvdevname 2)
4079 start_ost || error "Unable to start first ost (idx 10000)"
4080 start_ost2 || error "Unable to start second ost (idx 1000)"
4081 mount_client $MOUNT || error "Unable to mount client"
4085 if [[ $server_version -ge $(version_code 2.6.54) ]] ||
4086 [[ $server_version -ge $(version_code 2.5.4) &&
4087 $server_version -lt $(version_code 2.5.11) ]]; then
4088 wait_osc_import_state mds ost1 FULL
4089 wait_osc_import_state mds ost2 FULL
4090 $SETSTRIPE --stripe-count=-1 $DIR/$tfile ||
4091 error "Unable to setstripe $DIR/$tfile"
4092 n=$($LFS getstripe --stripe-count $DIR/$tfile)
4093 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
4098 MDSJOURNALSIZE=$mds_journal_size_orig
4101 run_test 56 "check big OST indexes and out-of-index-order start"
4103 test_57a() { # bug 22656
4104 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4105 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4106 tail -1 | awk '{print $1}')
4107 writeconf_or_reformat
4108 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4109 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" ||
4110 error "tunefs failed"
4112 start_ost && error "OST registration from failnode should fail"
4115 run_test 57a "initial registration from failnode should fail (should return errs)"
4118 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4119 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4120 tail -1 | awk '{print $1}')
4121 writeconf_or_reformat
4122 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4123 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" ||
4124 error "tunefs failed"
4126 start_ost || error "OST registration from servicenode should not fail"
4129 run_test 57b "initial registration from servicenode should not fail"
4132 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
4135 test_58() { # bug 22658
4137 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4138 createmany -o $DIR/$tdir/$tfile-%d 100
4139 # make sure that OSTs do not cancel llog cookies before we unmount the MDS
4140 #define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601
4141 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x601"
4142 unlinkmany $DIR/$tdir/$tfile-%d 100
4143 stop_mds || error "Unable to stop MDS"
4145 local MNTDIR=$(facet_mntpt $SINGLEMDS)
4146 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4148 # remove all files from the OBJECTS dir
4149 mount_fstype $SINGLEMDS
4151 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
4153 unmount_fstype $SINGLEMDS
4154 # restart MDS with missing llog files
4155 start_mds || error "unable to start MDS"
4156 do_facet mds "$LCTL set_param fail_loc=0"
4159 run_test 58 "missing llog files must not prevent MDT from mounting"
4162 start_mgsmds >> /dev/null
4163 local C1=$(count_osts)
4164 if [ $C1 -eq 0 ]; then
4165 start_ost >> /dev/null
4169 echo "original ost count: $C1 (expect > 0)"
4170 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
4171 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
4172 local C2=$(count_osts)
4173 echo "after mdt writeconf count: $C2 (expect 0)"
4174 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
4175 echo "OST start without writeconf should fail:"
4176 start_ost >> /dev/null &&
4177 error "OST start without writeconf didn't fail"
4178 echo "OST start with writeconf should succeed:"
4179 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
4180 local C3=$(count_osts)
4181 echo "after ost writeconf count: $C3 (expect 1)"
4182 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
4183 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
4184 local C4=$(count_osts)
4185 echo "after ost2 writeconf count: $C4 (expect 2)"
4186 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
4187 stop_ost2 >> /dev/null
4188 cleanup_nocli >> /dev/null
4189 #writeconf to remove all ost2 traces for subsequent tests
4190 writeconf_or_reformat
4192 run_test 59 "writeconf mount option"
4194 test_60() { # LU-471
4197 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4198 skip "Only applicable to ldiskfs-based MDTs"
4202 for num in $(seq $MDSCOUNT); do
4203 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
4204 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
4205 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4209 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
4210 [ ${PIPESTATUS[0]} -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
4212 # MDT default has dirdata feature
4213 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
4214 # we disable uninit_bg feature
4215 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
4216 # we set stride extended options
4217 echo $dump | grep stride > /dev/null || error "stride is not set"
4220 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
4225 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] ||
4226 { skip "Need MDS version at least 2.1.53"; return 0; }
4228 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
4229 ! large_xattr_enabled; then
4232 for num in $(seq $MDSCOUNT); do
4233 do_facet mds${num} $TUNE2FS -O large_xattr \
4234 $(mdsdevname $num) ||
4235 error "tune2fs on mds $num failed"
4239 setup_noconfig || error "setting up the filesystem failed"
4240 client_up || error "starting client failed"
4242 local file=$DIR/$tfile
4243 touch $file || error "touch $file failed"
4245 local large_value="$(generate_string $(max_xattr_size))"
4246 local small_value="bar"
4248 local name="trusted.big"
4249 log "save large xattr $name on $file"
4250 setfattr -n $name -v $large_value $file ||
4251 error "saving $name on $file failed"
4253 local new_value=$(get_xattr_value $name $file)
4254 [[ "$new_value" != "$large_value" ]] &&
4255 error "$name different after saving"
4257 log "shrink value of $name on $file"
4258 setfattr -n $name -v $small_value $file ||
4259 error "shrinking value of $name on $file failed"
4261 new_value=$(get_xattr_value $name $file)
4262 [[ "$new_value" != "$small_value" ]] &&
4263 error "$name different after shrinking"
4265 log "grow value of $name on $file"
4266 setfattr -n $name -v $large_value $file ||
4267 error "growing value of $name on $file failed"
4269 new_value=$(get_xattr_value $name $file)
4270 [[ "$new_value" != "$large_value" ]] &&
4271 error "$name different after growing"
4273 log "check value of $name on $file after remounting MDS"
4275 new_value=$(get_xattr_value $name $file)
4276 [[ "$new_value" != "$large_value" ]] &&
4277 error "$name different after remounting MDS"
4279 log "remove large xattr $name from $file"
4280 setfattr -x $name $file || error "removing $name from $file failed"
4283 stopall || error "stopping for e2fsck run"
4284 for num in $(seq $MDSCOUNT); do
4285 run_e2fsck $(facet_active_host mds$num) \
4286 $(mdsdevname $num) "-y" ||
4287 error "e2fsck MDT$num failed"
4289 setup_noconfig || error "remounting the filesystem failed"
4292 # need to delete this file to avoid problems in other tests
4294 stopall || error "stopping systems to turn off large_xattr"
4296 for num in $(seq $MDSCOUNT); do
4297 do_facet mds${num} $TUNE2FS -O ^large_xattr \
4298 $(mdsdevname $num) ||
4299 error "tune2fs on mds $num failed"
4303 run_test 61 "large xattr"
4306 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4307 skip "Only applicable to ldiskfs-based MDTs"
4312 local mdsdev=$(mdsdevname 1)
4313 local ostdev=$(ostdevname 1)
4315 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] ||
4316 { skip "Need MDS version at least 2.2.51"; return 0; }
4318 echo "disable journal for mds"
4319 do_facet mds $TUNE2FS -O ^has_journal $mdsdev || error "tune2fs failed"
4320 start_mds && error "MDT start should fail"
4321 echo "disable journal for ost"
4322 do_facet ost1 $TUNE2FS -O ^has_journal $ostdev || error "tune2fs failed"
4323 start_ost && error "OST start should fail"
4324 cleanup || error "cleanup failed with rc $?"
4327 run_test 62 "start with disabled journal"
4330 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4331 skip "Only applicable to ldiskfs-based MDTs"
4335 local inode_slab=$(do_facet $SINGLEMDS \
4336 "awk '/ldiskfs_inode_cache/ { print \\\$5 }' /proc/slabinfo")
4337 if [ -z "$inode_slab" ]; then
4338 skip "ldiskfs module has not been loaded"
4342 echo "$inode_slab ldisk inodes per page"
4343 [ "$inode_slab" -ge "3" ] ||
4344 error "ldisk inode size is too big, $inode_slab objs per page"
4347 run_test 63 "Verify each page can at least hold 3 ldisk inodes"
4350 start_mds || error "unable to start MDS"
4351 start_ost || error "Unable to start OST1"
4352 start_ost2 || error "Unable to start second ost"
4353 mount_client $MOUNT || error "Unable to mount client"
4354 stop_ost2 || error "Unable to stop second ost"
4356 $LFS df --lazy || error "lfs df failed"
4357 umount_client $MOUNT -f || error “unmount $MOUNT failed”
4358 cleanup_nocli || error "cleanup_nocli failed with $?"
4359 #writeconf to remove all ost2 traces for subsequent tests
4360 writeconf_or_reformat
4362 run_test 64 "check lfs df --lazy "
4364 test_65() { # LU-2237
4365 # Currently, the test is only valid for ldiskfs backend
4366 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4367 skip "non-ldiskfs backend" && return
4369 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4370 local brpt=$(facet_mntpt brpt)
4373 if ! do_facet $SINGLEMDS "test -b $devname"; then
4377 stop_mds || error "Unable to stop MDS"
4378 local obj=$(do_facet $SINGLEMDS \
4379 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
4381 if [ -z "$obj" ]; then
4382 # The MDT may be just re-formatted, mount the MDT for the
4383 # first time to guarantee the "last_rcvd" file is there.
4384 start_mds || error "fail to mount the MDS for the first time"
4385 stop_mds || error "Unable to stop MDS"
4388 # remove the "last_rcvd" file
4389 do_facet $SINGLEMDS "mkdir -p $brpt"
4390 do_facet $SINGLEMDS \
4391 "mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
4392 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
4393 do_facet $SINGLEMDS "$UMOUNT $brpt"
4395 # restart MDS, the "last_rcvd" file should be recreated.
4396 start_mds || error "fail to restart the MDS"
4397 stop_mds || error "Unable to stop MDS"
4398 obj=$(do_facet $SINGLEMDS \
4399 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
4400 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
4402 run_test 65 "re-create the lost last_rcvd file when server mount"
4405 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
4406 { skip "Need MGS version at least 2.3.59"; return 0; }
4409 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
4410 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
4412 echo "replace_nids should fail if MDS, OSTs and clients are UP"
4413 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4414 error "replace_nids fail"
4416 umount_client $MOUNT || error "unmounting client failed"
4417 echo "replace_nids should fail if MDS and OSTs are UP"
4418 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4419 error "replace_nids fail"
4421 stop_ost || error "Unable to stop OST1"
4422 echo "replace_nids should fail if MDS is UP"
4423 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4424 error "replace_nids fail"
4426 stop_mds || error "stopping mds failed"
4428 if combined_mgs_mds; then
4429 start_mdt 1 "-o nosvc" ||
4430 error "starting mds with nosvc option failed"
4433 echo "command should accept two parameters"
4434 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
4435 error "command should accept two params"
4437 echo "correct device name should be passed"
4438 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
4439 error "wrong devname"
4441 echo "wrong nids list should not destroy the system"
4442 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
4445 echo "replace OST nid"
4446 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
4447 error "replace nids failed"
4449 echo "command should accept two parameters"
4450 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
4451 error "command should accept two params"
4453 echo "wrong nids list should not destroy the system"
4454 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
4457 echo "replace MDS nid"
4458 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
4459 error "replace nids failed"
4461 if ! combined_mgs_mds ; then
4464 stop_mds || error "Unable to stop MDS"
4468 check_mount || error "error after nid replace"
4469 cleanup || error "cleanup failed"
4472 run_test 66 "replace nids"
4474 test_67() { #LU-2950
4475 local legacy="$TMP/legacy_lnet_config"
4476 local new="$TMP/new_routes_test"
4477 local out="$TMP/config_out_file"
4478 local verify="$TMP/conv_verify"
4479 local verify_conf="$TMP/conf_verify"
4481 # Create the legacy file that will be run through the
4482 # lustre_routes_conversion script
4483 cat <<- LEGACY_LNET_CONFIG > $legacy
4484 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
4485 tcp2 54 10.1.3.2@tcp;
4486 tcp3 10.3.4.3@tcp:3;
4490 # Create the verification file to verify the output of
4491 # lustre_routes_conversion script against.
4492 cat <<- VERIFY_LNET_CONFIG > $verify
4493 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
4494 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
4495 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
4496 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
4497 tcp4: { gateway: 10.3.3.4@tcp }
4500 # Create the verification file to verify the output of
4501 # lustre_routes_config script against
4502 cat <<- VERIFY_LNET_CONFIG > $verify_conf
4503 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
4504 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
4505 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
4506 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
4507 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
4510 lustre_routes_conversion $legacy $new > /dev/null
4511 if [ -f $new ]; then
4512 # verify the conversion output
4513 cmp -s $new $verify > /dev/null
4514 if [ $? -eq 1 ]; then
4515 error "routes conversion failed"
4518 lustre_routes_config --dry-run --verbose $new > $out
4519 # check that the script succeeded
4520 cmp -s $out $verify_conf > /dev/null
4521 if [ $? -eq 1 ]; then
4522 error "routes config failed"
4525 error "routes conversion test failed"
4527 # remove generated files
4528 rm -f $new $legacy $verify $verify_conf $out
4530 run_test 67 "test routes conversion and configuration"
4538 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.4.53) ] ||
4539 { skip "Need MDS version at least 2.4.53"; return 0; }
4541 umount_client $MOUNT || error "umount client failed"
4543 start_mdt 1 || error "MDT start failed"
4544 start_ost || error "Unable to start OST1"
4546 # START-END - the sequences we'll be reserving
4547 START=$(do_facet $SINGLEMDS \
4548 $LCTL get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
4549 END=$((START + (1 << 30)))
4550 do_facet $SINGLEMDS \
4551 $LCTL set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
4553 # reset the sequences MDT0000 has already assigned
4554 do_facet $SINGLEMDS \
4555 $LCTL set_param seq.srv*MDT0000.space=clear
4557 # remount to let the client allocate new sequence
4558 mount_client $MOUNT || error "mount client failed"
4560 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4561 do_facet $SINGLEMDS \
4562 $LCTL get_param seq.srv*MDT0000.space
4563 $LFS path2fid $DIR/$tfile
4565 local old_ifs="$IFS"
4567 fid=($($LFS path2fid $DIR/$tfile))
4571 if [[ $seq < $END ]]; then
4572 error "used reserved sequence $seq?"
4574 cleanup || error "cleanup failed with $?"
4576 run_test 68 "be able to reserve specific sequences in FLDB"
4578 # Test 69: is about the total number of objects ever created on an OST.
4579 # so that when it is reformatted the normal MDS->OST orphan recovery won't
4580 # just "precreate" the missing objects. In the past it might try to recreate
4581 # millions of objects after an OST was reformatted
4583 local server_version=$(lustre_version_code $SINGLEMDS)
4585 [[ $server_version -lt $(version_code 2.4.2) ]] &&
4586 skip "Need MDS version at least 2.4.2" && return
4588 [[ $server_version -ge $(version_code 2.4.50) ]] &&
4589 [[ $server_version -lt $(version_code 2.5.0) ]] &&
4590 skip "Need MDS version at least 2.5.0" && return
4593 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4595 # use OST0000 since it probably has the most creations
4596 local OSTNAME=$(ostname_from_index 0)
4597 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
4598 local last_id=$(do_facet mds1 $LCTL get_param -n \
4599 osc.$mdtosc_proc1.prealloc_last_id)
4601 # Want to have OST LAST_ID over 5 * OST_MAX_PRECREATE to
4602 # verify that the LAST_ID recovery is working properly. If
4603 # not, then the OST will refuse to allow the MDS connect
4604 # because the LAST_ID value is too different from the MDS
4605 #define OST_MAX_PRECREATE=20000
4606 local ost_max_pre=20000
4607 local num_create=$(( ost_max_pre * 5 + 1 - last_id))
4609 # If the LAST_ID is already over 5 * OST_MAX_PRECREATE, we don't
4610 # need to create any files. So, skip this section.
4611 if [ $num_create -gt 0 ]; then
4612 # Check the number of inodes available on OST0
4614 local ifree=$($LFS df -i $MOUNT | awk '/OST0000/ { print $4 }')
4615 log "On OST0, $ifree inodes available. Want $num_create."
4617 $SETSTRIPE -i 0 $DIR/$tdir ||
4618 error "$SETSTRIPE -i 0 $DIR/$tdir failed"
4619 if [ $ifree -lt 10000 ]; then
4620 files=$(( ifree - 50 ))
4625 local j=$((num_create / files + 1))
4626 for i in $(seq 1 $j); do
4627 createmany -o $DIR/$tdir/$tfile-$i- $files ||
4628 error "createmany fail create $files files: $?"
4629 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
4630 error "unlinkmany failed unlink $files files"
4634 # delete all of the files with objects on OST0 so the
4635 # filesystem is not inconsistent later on
4636 $LFS find $MOUNT --ost 0 -print0 | xargs -0 rm
4638 umount_client $MOUNT || error "umount client failed"
4639 stop_ost || error "OST0 stop failure"
4640 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
4641 $(ostdevname 1) $(ostvdevname 1) ||
4642 error "reformat and replace $ostdev failed"
4643 start_ost || error "OST0 restart failure"
4644 wait_osc_import_state mds ost FULL
4646 mount_client $MOUNT || error "mount client failed"
4647 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
4648 local idx=$($GETSTRIPE -i $DIR/$tdir/$tfile-last)
4649 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
4651 local iused=$($LFS df -i $MOUNT | awk '/OST0000/ { print $3 }')
4652 log "On OST0, $iused used inodes"
4653 [ $iused -ge $((ost_max_pre/2 + 1000)) ] &&
4654 error "OST replacement created too many inodes; $iused"
4655 cleanup || error "cleanup failed with $?"
4657 run_test 69 "replace an OST with the same index"
4660 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4663 cleanup || error "cleanup failed with $?"
4665 start_mdt 1 || error "MDT0 start fail"
4667 start_ost || error "OST0 start fail"
4668 for num in $(seq 2 $MDSCOUNT); do
4669 start_mdt $num || return
4672 mount_client $MOUNT || error "mount client fails"
4674 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
4676 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4677 error "create remote dir fail"
4679 rm -rf $DIR/$tdir || error "delete dir fail"
4680 cleanup || error "cleanup failed with $?"
4682 run_test 70a "start MDT0, then OST, then MDT1"
4685 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4688 start_ost || error "OST0 start fail"
4690 start_mds || error "MDS start fail"
4692 mount_client $MOUNT || error "mount client fails"
4694 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
4696 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4697 error "create remote dir fail"
4699 rm -rf $DIR/$tdir || error "delete dir fail"
4701 cleanup || error "cleanup failed with $?"
4703 run_test 70b "start OST, MDT1, MDT0"
4706 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4709 start_mds || error "MDS start fail"
4710 start_ost || error "OST0 start fail"
4712 mount_client $MOUNT || error "mount client fails"
4713 stop_mdt 1 || error "MDT1 start fail"
4715 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
4716 echo "deactivate $mdc_for_mdt1"
4717 $LCTL --device $mdc_for_mdt1 deactivate ||
4718 error "set $mdc_for_mdt1 deactivate failed"
4720 mkdir $DIR/$tdir && error "mkdir succeed"
4722 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4723 error "create remote dir succeed"
4725 cleanup || error "cleanup failed with $?"
4727 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
4730 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4733 start_mds || error "MDS start fail"
4734 start_ost || error "OST0 start fail"
4736 mount_client $MOUNT || error "mount client fails"
4738 stop_mdt 2 || error "MDT1 start fail"
4740 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
4742 echo "deactivate $mdc_for_mdt2"
4743 $LCTL --device $mdc_for_mdt2 deactivate ||
4744 error "set $mdc_for_mdt2 deactivate failed"
4746 mkdir $DIR/$tdir || error "mkdir fail"
4747 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4748 error "create remote dir succeed"
4750 rm -rf $DIR/$tdir || error "delete dir fail"
4752 cleanup || error "cleanup failed with $?"
4754 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
4757 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4759 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.62) ] ||
4760 { skip "Need MDS version at least 2.7.62"; return 0; }
4762 cleanup || error "cleanup failed with $?"
4764 local mdsdev=$(mdsdevname 1)
4765 local ostdev=$(ostdevname 1)
4766 local mdsvdev=$(mdsvdevname 1)
4767 local ostvdev=$(ostvdevname 1)
4768 local opts_mds="$(mkfs_opts mds1 $mdsdev) --reformat $mdsdev $mdsvdev"
4769 local opts_ost="$(mkfs_opts ost1 $ostdev) --reformat $ostdev $ostvdev"
4771 add mds1 $opts_mds || error "add mds1 failed"
4772 start_mdt 1 || error "start mdt1 failed"
4773 add ost1 $opts_ost || error "add ost1 failed"
4774 start_ost || error "start ost failed"
4775 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
4777 local soc=$(do_facet mds1 "$LCTL get_param -n \
4778 mdt.*MDT0000.sync_lock_cancel")
4779 [ $soc == "never" ] || error "SoC enabled on single MDS"
4781 for i in $(seq 2 $MDSCOUNT); do
4782 mdsdev=$(mdsdevname $i)
4783 mdsvdev=$(mdsvdevname $i)
4784 opts_mds="$(mkfs_opts mds$i $mdsdev) --reformat $mdsdev \
4786 add mds$i $opts_mds || error "add mds$i failed"
4787 start_mdt $i || error "start mdt$i fail"
4790 wait_dne_interconnect
4792 for i in $(seq $MDSCOUNT); do
4793 soc=$(do_facet mds$i "$LCTL get_param -n \
4794 mdt.*MDT000$((i - 1)).sync_lock_cancel")
4795 [ $soc == "blocking" ] || error "SoC not enabled on DNE"
4798 for i in $(seq 2 $MDSCOUNT); do
4799 stop_mdt $i || error "stop mdt$i fail"
4801 soc=$(do_facet mds1 "$LCTL get_param -n \
4802 mdt.*MDT0000.sync_lock_cancel")
4803 [ $soc == "never" ] || error "SoC enabled on single MDS"
4805 cleanup || error "cleanup failed with $?"
4807 run_test 70e "Sync-on-Cancel will be enabled by default on DNE"
4810 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4811 if combined_mgs_mds; then
4812 skip "needs separate MGS/MDT" && return
4816 start_mdt 1 || error "MDT0 start fail"
4817 start_ost || error "OST0 start fail"
4818 for num in $(seq 2 $MDSCOUNT); do
4819 start_mdt $num || return
4822 start_ost2 || error "OST1 start fail"
4824 mount_client $MOUNT || error "mount client fails"
4826 mkdir $DIR/$tdir || error "mkdir fail"
4827 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4828 error "create remote dir succeed"
4830 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4831 rm -rf $DIR/$tdir || error "delete dir fail"
4833 umount_client $MOUNT || error "umount_client failed"
4834 stop_mds || error "MDS stop fail"
4835 stop_ost || error "OST0 stop fail"
4836 stop_ost2 || error "OST1 stop fail"
4838 run_test 71a "start MDT0 OST0, MDT1, OST1"
4841 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4842 if combined_mgs_mds; then
4843 skip "needs separate MGS/MDT" && return
4847 for num in $(seq 2 $MDSCOUNT); do
4848 start_mdt $num || return
4850 start_ost || error "OST0 start fail"
4851 start_mdt 1 || error "MDT0 start fail"
4852 start_ost2 || error "OST1 start fail"
4854 mount_client $MOUNT || error "mount client fails"
4856 mkdir $DIR/$tdir || error "mkdir fail"
4857 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4858 error "create remote dir succeed"
4860 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4861 rm -rf $DIR/$tdir || error "delete dir fail"
4863 umount_client $MOUNT || error "umount_client failed"
4864 stop_mds || error "MDT0 stop fail"
4865 stop_ost || error "OST0 stop fail"
4866 stop_ost2 || error "OST1 stop fail"
4868 run_test 71b "start MDT1, OST0, MDT0, OST1"
4871 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4872 if combined_mgs_mds; then
4873 skip "needs separate MGS/MDT" && return
4877 start_ost || error "OST0 start fail"
4878 start_ost2 || error "OST1 start fail"
4879 for num in $(seq 2 $MDSCOUNT); do
4880 start_mdt $num || return
4882 start_mdt 1 || error "MDT0 start fail"
4884 mount_client $MOUNT || error "mount client fails"
4886 mkdir $DIR/$tdir || error "mkdir fail"
4887 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4888 error "create remote dir succeed"
4890 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4891 rm -rf $DIR/$tdir || error "delete dir fail"
4893 umount_client $MOUNT || error "umount_client failed"
4894 stop_mds || error "MDS stop fail"
4895 stop_ost || error "OST0 stop fail"
4896 stop_ost2 || error "OST1 stop fail"
4899 run_test 71c "start OST0, OST1, MDT1, MDT0"
4902 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4903 if combined_mgs_mds; then
4904 skip "needs separate MGS/MDT" && return
4908 start_ost || error "OST0 start fail"
4909 for num in $(seq 2 $MDSCOUNT); do
4910 start_mdt $num || return
4912 start_mdt 1 || error "MDT0 start fail"
4913 start_ost2 || error "OST1 start fail"
4915 mount_client $MOUNT || error "mount client fails"
4917 mkdir $DIR/$tdir || error "mkdir fail"
4918 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4919 error "create remote dir succeed"
4921 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4922 rm -rf $DIR/$tdir || error "delete dir fail"
4924 umount_client $MOUNT || error "umount_client failed"
4925 stop_mds || error "MDS stop fail"
4926 stop_ost || error "OST0 stop fail"
4927 stop_ost2 || error "OST1 stop fail"
4930 run_test 71d "start OST0, MDT1, MDT0, OST1"
4933 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4934 if combined_mgs_mds; then
4935 skip "needs separate MGS/MDT" && return
4939 start_ost || error "OST0 start fail"
4940 for num in $(seq 2 $MDSCOUNT); do
4941 start_mdt $num || return
4943 start_ost2 || error "OST1 start fail"
4944 start_mdt 1 || error "MDT0 start fail"
4946 mount_client $MOUNT || error "mount client fails"
4948 mkdir $DIR/$tdir || error "mkdir fail"
4949 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4950 error "create remote dir succeed"
4952 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4953 rm -rf $DIR/$tdir || error "delete dir fail"
4955 umount_client $MOUNT || error "umount_client failed"
4956 stop_mds || error "MDS stop fail"
4957 stop_ost || error "OST0 stop fail"
4958 stop_ost2 || error "OST1 stop fail"
4961 run_test 71e "start OST0, MDT1, OST1, MDT0"
4963 test_72() { #LU-2634
4964 local mdsdev=$(mdsdevname 1)
4965 local ostdev=$(ostdevname 1)
4966 local cmd="$E2FSCK -fnvd $mdsdev"
4969 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4970 skip "ldiskfs only test" && return
4972 #tune MDT with "-O extents"
4974 for num in $(seq $MDSCOUNT); do
4975 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
4976 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4977 error "add mds $num failed"
4978 do_facet mds${num} "$TUNE2FS -O extents $(mdsdevname $num)" ||
4979 error "$TUNE2FS failed on mds${num}"
4982 add ost1 $(mkfs_opts ost1 $ostdev) --reformat $ostdev ||
4983 error "add $ostdev failed"
4984 start_mgsmds || error "start mds failed"
4985 start_ost || error "start ost failed"
4986 mount_client $MOUNT || error "mount client failed"
4988 #create some short symlinks
4989 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4990 createmany -o $DIR/$tdir/$tfile-%d $fn
4991 echo "create $fn short symlinks"
4992 for i in $(seq -w 1 $fn); do
4993 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
4998 umount_client $MOUNT || error "umount client failed"
4999 stop_mds || error "stop mds failed"
5000 stop_ost || error "stop ost failed"
5003 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
5005 run_test 72 "test fast symlink with extents flag enabled"
5007 test_73() { #LU-3006
5008 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
5009 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
5010 error "1st tunefs failed"
5011 start_mgsmds || error "start mds failed"
5012 start_ost || error "start ost failed"
5013 mount_client $MOUNT || error "mount client failed"
5014 $LCTL get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
5015 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
5016 umount_client $MOUNT || error "umount client failed"
5020 run_test 73 "failnode to update from mountdata properly"
5022 test_75() { # LU-2374
5023 [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] &&
5024 skip "Need MDS version at least 2.4.1" && return
5027 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
5028 --reformat $(mdsdevname 1) $(mdsvdevname 1)"
5029 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
5030 --reformat $(ostdevname 1) $(ostvdevname 1)"
5032 #check with default parameters
5033 add mds1 $opts_mds || error "add mds1 failed for default params"
5034 add ost1 $opts_ost || error "add ost1 failed for default params"
5036 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
5037 opts_mds=$(echo $opts_mds |
5038 sed -e "s/--index=$index/--index=$index --mdt/")
5039 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
5040 opts_ost=$(echo $opts_ost |
5041 sed -e "s/--index=$index/--index=$index --ost/")
5043 add mds1 $opts_mds || error "add mds1 failed for new params"
5044 add ost1 $opts_ost || error "add ost1 failed for new params"
5047 run_test 75 "The order of --index should be irrelevant"
5050 [[ $(lustre_version_code mgs) -ge $(version_code 2.4.52) ]] ||
5051 { skip "Need MDS version at least 2.4.52" && return 0; }
5053 local MDMB_PARAM="osc.*.max_dirty_mb"
5054 echo "Change MGS params"
5055 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
5057 echo "max_dirty_mb: $MAX_DIRTY_MB"
5058 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB + MAX_DIRTY_MB))
5059 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
5060 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
5061 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5062 head -1" $NEW_MAX_DIRTY_MB
5063 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5064 echo "$MAX_DIRTY_MB"
5065 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5066 error "error while apply max_dirty_mb"
5068 echo "Check the value is stored after remount"
5071 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5072 head -1" $NEW_MAX_DIRTY_MB
5073 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5074 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5075 error "max_dirty_mb is not saved after remount"
5077 echo "Change OST params"
5078 CLIENT_PARAM="obdfilter.*.client_cache_count"
5079 local CLIENT_CACHE_COUNT
5080 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5082 echo "client_cache_count: $CLIENT_CACHE_COUNT"
5083 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
5084 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
5085 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
5086 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5087 head -1" $NEW_CLIENT_CACHE_COUNT
5088 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5090 echo "$CLIENT_CACHE_COUNT"
5091 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5092 error "error while apply client_cache_count"
5094 echo "Check the value is stored after remount"
5097 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5098 head -1" $NEW_CLIENT_CACHE_COUNT
5099 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5101 echo "$CLIENT_CACHE_COUNT"
5102 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5103 error "client_cache_count is not saved after remount"
5106 run_test 76a "set permanent params set_param -P"
5108 test_76b() { # LU-4783
5109 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.57) ]] ||
5110 { skip "Need MGS version at least 2.5.57" && return 0; }
5113 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
5114 error "start params log failed"
5117 run_test 76b "verify params log setup correctly"
5119 test_77() { # LU-3445
5120 local server_version=$(lustre_version_code $SINGLEMDS)
5122 [[ $server_version -ge $(version_code 2.2.60) ]] &&
5123 [[ $server_version -le $(version_code 2.4.0) ]] &&
5124 skip "Need MDS version < 2.2.60 or > 2.4.0" && return
5126 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
5127 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
5128 skip_env "mixed loopback and real device not working" && return
5131 local fs2mdsdev=$(mdsdevname 1_2)
5132 local fs2ostdev=$(ostdevname 1_2)
5133 local fs2mdsvdev=$(mdsvdevname 1_2)
5134 local fs2ostvdev=$(ostvdevname 1_2)
5135 local fsname=test1234
5137 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
5139 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
5140 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
5141 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
5142 error "start fs2mds failed"
5144 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
5145 [[ $mgsnid = *,* ]] || mgsnid+=",$mgsnid"
5147 add fs2ost $(mkfs_opts ost1 $fs2ostdev) --mgsnode=$mgsnid \
5148 --failnode=$failnid --fsname=$fsname \
5149 --reformat $fs2ostdev $fs2ostvdev ||
5150 error "add fs2ost failed"
5151 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
5153 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
5154 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
5155 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
5158 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
5161 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ||
5162 $(facet_fstype ost1) != ldiskfs ]] &&
5163 skip "only applicable to ldiskfs-based MDTs and OSTs" && return
5165 # reformat the Lustre filesystem with a smaller size
5166 local saved_MDSSIZE=$MDSSIZE
5167 local saved_OSTSIZE=$OSTSIZE
5168 MDSSIZE=$((MDSSIZE - 20000))
5169 OSTSIZE=$((OSTSIZE - 20000))
5170 reformat || error "(1) reformat Lustre filesystem failed"
5171 MDSSIZE=$saved_MDSSIZE
5172 OSTSIZE=$saved_OSTSIZE
5174 # mount the Lustre filesystem
5175 setup_noconfig || error "(2) setup Lustre filesystem failed"
5178 log "create test files"
5182 mkdir $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
5183 for i in $(seq $num_files); do
5184 file=$MOUNT/$tdir/$tfile-$i
5185 dd if=/dev/urandom of=$file count=1 bs=1M ||
5186 error "(4) create $file failed"
5189 # unmount the Lustre filesystem
5190 cleanup || error "(5) cleanup Lustre filesystem failed"
5192 # run e2fsck on the MDT and OST devices
5193 local mds_host=$(facet_active_host $SINGLEMDS)
5194 local ost_host=$(facet_active_host ost1)
5195 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
5196 local ost_dev=$(ostdevname 1)
5198 run_e2fsck $mds_host $mds_dev "-y"
5199 run_e2fsck $ost_host $ost_dev "-y"
5201 # get the original block count of the MDT and OST filesystems
5202 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
5203 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
5205 # expand the MDT and OST filesystems to the device size
5206 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
5207 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
5209 # run e2fsck on the MDT and OST devices again
5210 run_e2fsck $mds_host $mds_dev "-y"
5211 run_e2fsck $ost_host $ost_dev "-y"
5213 # mount the Lustre filesystem
5217 log "check files after expanding the MDT and OST filesystems"
5218 for i in $(seq $num_files); do
5219 file=$MOUNT/$tdir/$tfile-$i
5220 $CHECKSTAT -t file -s 1048576 $file ||
5221 error "(6) checkstat $file failed"
5225 log "create more files after expanding the MDT and OST filesystems"
5226 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
5227 file=$MOUNT/$tdir/$tfile-$i
5228 dd if=/dev/urandom of=$file count=1 bs=1M ||
5229 error "(7) create $file failed"
5232 # unmount the Lustre filesystem
5233 cleanup || error "(8) cleanup Lustre filesystem failed"
5235 # run e2fsck on the MDT and OST devices
5236 run_e2fsck $mds_host $mds_dev "-y"
5237 run_e2fsck $ost_host $ost_dev "-y"
5239 # get the maximum block count of the MDT and OST filesystems
5240 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
5241 local ost_max_blks=$(get_block_count ost1 $ost_dev)
5243 # get the minimum block count of the MDT and OST filesystems
5244 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
5245 grep minimum | sed -e 's/^.*filesystem: //g')
5246 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
5247 grep minimum | sed -e 's/^.*filesystem: //g')
5249 # shrink the MDT and OST filesystems to a smaller size
5253 if [[ $mds_max_blks -gt $mds_min_blks &&
5254 $mds_max_blks -gt $mds_orig_blks ]]; then
5255 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
5256 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
5257 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
5258 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
5259 error "shrink $SINGLEMDS to $new_blks failed"
5263 if [[ $ost_max_blks -gt $ost_min_blks &&
5264 $ost_max_blks -gt $ost_orig_blks ]]; then
5265 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
5266 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
5267 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
5268 run_resize2fs ost1 $ost_dev $new_blks ||
5269 error "shrink ost1 to $new_blks failed"
5273 # check whether the MDT or OST filesystem was shrunk or not
5275 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
5276 reformat || error "(10) reformat Lustre filesystem failed"
5280 # run e2fsck on the MDT and OST devices again
5281 run_e2fsck $mds_host $mds_dev "-y"
5282 run_e2fsck $ost_host $ost_dev "-y"
5284 # mount the Lustre filesystem again
5288 log "check files after shrinking the MDT and OST filesystems"
5289 for i in $(seq $((num_files + 10))); do
5290 file=$MOUNT/$tdir/$tfile-$i
5291 $CHECKSTAT -t file -s 1048576 $file ||
5292 error "(11) checkstat $file failed"
5295 # unmount and reformat the Lustre filesystem
5296 cleanup || error "(12) cleanup Lustre filesystem failed"
5297 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
5298 reformat || error "(14) reformat Lustre filesystem failed"
5300 run_test 78 "run resize2fs on MDT and OST filesystems"
5302 test_79() { # LU-4227
5303 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.5.59) ]] ||
5304 { skip "Need MDS version at least 2.5.59"; return 0; }
5306 local mdsdev1=$(mdsdevname 1)
5307 local mdsvdev1=$(mdsvdevname 1)
5308 local mdsdev2=$(mdsdevname 2)
5309 local mdsvdev2=$(mdsvdevname 2)
5310 local ostdev1=$(ostdevname 1)
5311 local ostvdev1=$(ostvdevname 1)
5312 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
5313 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
5314 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
5317 # remove --mgs/--mgsnode from mkfs.lustre options
5318 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
5320 mgsnode_opt=$(echo $opts_mds2 |
5321 awk '{ for ( i = 1; i < NF; i++ )
5322 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5323 [ -n $mgsnode_opt ] &&
5324 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
5326 mgsnode_opt=$(echo $opts_ost1 |
5327 awk '{ for ( i = 1; i < NF; i++ )
5328 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5329 [ -n $mgsnode_opt ] &&
5330 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
5332 # -MGS, format a mdt without --mgs option
5333 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
5334 error "Must specify --mgs when formatting mdt combined with mgs"
5336 # +MGS, format a mdt/ost without --mgsnode option
5337 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
5338 > /dev/null || error "start mds1 failed"
5339 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
5340 error "Must specify --mgsnode when formatting a mdt"
5341 add ost1 $opts_ost1 $ostdev1 $ostvdev2 &&
5342 error "Must specify --mgsnode when formatting an ost"
5346 run_test 79 "format MDT/OST without mgs option (should return errors)"
5349 start_mds || error "Failed to start MDT"
5350 start_ost || error "Failed to start OST1"
5351 uuid=$(do_facet ost1 $LCTL get_param -n mgc.*.uuid)
5352 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
5353 do_facet ost1 "$LCTL set_param fail_val=10 fail_loc=0x906"
5354 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x906"
5355 do_facet mgs "$LCTL set_param -n mgs/MGS/evict_client $uuid"
5357 start_ost2 || error "Failed to start OST2"
5359 do_facet ost1 "$LCTL set_param fail_loc=0"
5362 run_test 80 "mgc import reconnect race"
5364 #Save the original values of $OSTCOUNT and $OSTINDEX$i.
5366 local new_ostcount=$1
5367 saved_ostcount=$OSTCOUNT
5368 OSTCOUNT=$new_ostcount
5372 for i in $(seq $OSTCOUNT); do
5374 eval saved_ostindex$i=${!index}
5379 # Restore the original values of $OSTCOUNT and $OSTINDEX$i.
5380 restore_ostindex() {
5385 for i in $(seq $OSTCOUNT); do
5386 index=saved_ostindex$i
5387 eval OSTINDEX$i=${!index}
5389 OSTCOUNT=$saved_ostcount
5394 # The main purpose of this test is to ensure the OST_INDEX_LIST functions as
5395 # expected. This test uses OST_INDEX_LIST to format OSTs with a randomly
5396 # assigned index and ensures we can mount such a formatted file system
5397 test_81() { # LU-4665
5398 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5399 { skip "Need MDS version at least 2.6.54" && return; }
5400 [[ $OSTCOUNT -ge 3 ]] || { skip_env "Need at least 3 OSTs" && return; }
5404 # Each time RANDOM is referenced, a random integer between 0 and 32767
5407 local saved_ostindex1=$OSTINDEX1
5408 for i in 65535 $((RANDOM + 65536)); do
5409 echo -e "\nFormat ost1 with --index=$i, should fail"
5411 if add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat \
5412 $(ostdevname 1) $(ostvdevname 1); then
5413 OSTINDEX1=$saved_ostindex1
5414 error "format ost1 with --index=$i should fail"
5417 OSTINDEX1=$saved_ostindex1
5421 # Format OSTs with random sparse indices.
5422 trap "restore_ostindex" EXIT
5423 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices"
5424 OST_INDEX_LIST=[0,$((RANDOM * 2 % 65533 + 1)),65534] formatall
5426 # Setup and check Lustre filesystem.
5427 start_mgsmds || error "start_mgsmds failed"
5428 for i in $(seq $OSTCOUNT); do
5429 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5430 error "start ost$i failed"
5433 mount_client $MOUNT || error "mount client $MOUNT failed"
5434 check_mount || error "check client $MOUNT failed"
5437 local max_easize=$($LCTL get_param -n llite.*.max_easize)
5438 [[ $max_easize -eq 128 ]] ||
5439 error "max_easize is $max_easize, should be 128 bytes"
5443 run_test 81 "sparse OST indexing"
5445 # Wait OSTs to be active on both client and MDT side.
5447 local cmd="$LCTL get_param -n lov.$FSNAME-clilov-*.target_obd |
5448 awk 'BEGIN {c = 0} /ACTIVE/{c += 1} END {printf \\\"%d\\\", c}'"
5449 wait_update $HOSTNAME "eval $cmd" $OSTCOUNT ||
5450 error "wait_update OSTs up on client failed"
5452 cmd="$LCTL get_param -n lod.$FSNAME-MDT*-*.target_obd | sort -u |
5453 awk 'BEGIN {c = 0} /ACTIVE/{c += 1} END {printf \\\"%d\\\", c}'"
5454 wait_update_facet $SINGLEMDS "eval $cmd" $OSTCOUNT ||
5455 error "wait_update OSTs up on MDT failed"
5458 # Here we exercise the stripe placement functionality on a file system that
5459 # has formatted the OST with a random index. With the file system the following
5460 # functionality is tested:
5462 # 1. Creating a new file with a specific stripe layout.
5464 # 2. Modifiy a existing empty file with a specific stripe layout.
5466 # 3. Ensure we fail to set the stripe layout of a file that already has one.
5468 # 4. If ost-index is defined we need to ensure it is the first entry in the
5469 # ost index list returned by lfs getstripe.
5471 # 5. Lastly ensure this functionality fails with directories.
5472 test_82a() { # LU-4665
5473 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5474 { skip "Need MDS version at least 2.6.54" && return; }
5475 [[ $OSTCOUNT -ge 3 ]] || { skip_env "Need at least 3 OSTs" && return; }
5481 # Format OSTs with random sparse indices.
5485 local LOV_V1_INSANE_STRIPE_COUNT=65532
5486 for i in $(seq $OSTCOUNT); do
5487 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
5488 ost_indices+=" $index"
5490 ost_indices=$(comma_list $ost_indices)
5492 trap "restore_ostindex" EXIT
5493 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5494 OST_INDEX_LIST=[$ost_indices] formatall
5496 # Setup Lustre filesystem.
5497 start_mgsmds || error "start_mgsmds failed"
5498 for i in $(seq $OSTCOUNT); do
5499 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5500 error "start ost$i failed"
5503 mount_client $MOUNT || error "mount client $MOUNT failed"
5506 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5507 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5509 # 1. If the file does not exist, new file will be created
5510 # with specified OSTs.
5511 local file=$DIR/$tdir/$tfile-1
5512 local cmd="$SETSTRIPE -o $ost_indices $file"
5514 eval $cmd || error "$cmd failed"
5515 check_stripe_count $file $OSTCOUNT
5516 check_obdidx $file $ost_indices
5517 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5518 error "write $file failed"
5520 # 2. If the file already exists and is an empty file, the file
5521 # will be attached with specified layout.
5522 file=$DIR/$tdir/$tfile-2
5523 mcreate $file || error "mcreate $file failed"
5524 cmd="$SETSTRIPE -o $ost_indices $file"
5526 eval $cmd || error "$cmd failed"
5527 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5528 error "write $file failed"
5529 check_stripe_count $file $OSTCOUNT
5530 check_obdidx $file $ost_indices
5532 # 3. If the file already has a valid layout attached, the command
5533 # should fail with EBUSY.
5535 eval $cmd && error "stripe is already set on $file, $cmd should fail"
5537 # 4. If [--stripe-index|-i <start_ost_idx>] is used, the index must
5538 # be in the OST indices list.
5539 local start_ost_idx=${ost_indices##*,}
5540 file=$DIR/$tdir/$tfile-3
5541 cmd="$SETSTRIPE -o $ost_indices -i $start_ost_idx $file"
5543 eval $cmd || error "$cmd failed"
5544 check_stripe_count $file $OSTCOUNT
5545 check_obdidx $file $ost_indices
5546 check_start_ost_idx $file $start_ost_idx
5548 file=$DIR/$tdir/$tfile-4
5550 cmd+=" -o $(exclude_items_from_list $ost_indices $start_ost_idx)"
5551 cmd+=" -i $start_ost_idx $file"
5553 eval $cmd && error "index $start_ost_idx should be in $ost_indices"
5555 # 5. Specifying OST indices for directory should fail with ENOSUPP.
5556 local dir=$DIR/$tdir/$tdir
5557 mkdir $dir || error "mkdir $dir failed"
5558 cmd="$SETSTRIPE -o $ost_indices $dir"
5560 eval $cmd && error "$cmd should fail, specifying OST indices" \
5561 "for directory is not supported"
5565 run_test 82a "specify OSTs for file (succeed) or directory (fail)"
5570 # Remove OSTs from a pool and destroy the pool.
5571 destroy_pool $ost_pool || true
5576 # Test 82b is run to ensure that if the user supplies a pool with a specific
5577 # stripe layout that it behaves proprerly. It should fail in the case that
5578 # the supplied OST index list points to OSTs not contained in the user
5580 test_82b() { # LU-4665
5581 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5582 { skip "Need MDS version at least 2.6.54" && return; }
5583 [[ $OSTCOUNT -ge 4 ]] || { skip_env "Need at least 4 OSTs" && return; }
5589 # Format OSTs with random sparse indices.
5593 local LOV_V1_INSANE_STRIPE_COUNT=65532
5594 for i in $(seq $OSTCOUNT); do
5595 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
5596 ost_indices+=" $index"
5598 ost_indices=$(comma_list $ost_indices)
5600 trap "restore_ostindex" EXIT
5601 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5602 OST_INDEX_LIST=[$ost_indices] formatall
5604 # Setup Lustre filesystem.
5605 start_mgsmds || error "start_mgsmds failed"
5606 for i in $(seq $OSTCOUNT); do
5607 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5608 error "start ost$i failed"
5611 mount_client $MOUNT || error "mount client $MOUNT failed"
5613 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5614 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5616 # Create a new pool and add OSTs into it.
5617 local ost_pool=$FSNAME.$TESTNAME
5618 create_pool $ost_pool || error "create OST pool $ost_pool failed"
5621 trap "cleanup_82b" EXIT
5623 local ost_idx_in_list=${ost_indices##*,}
5624 local ost_idx_in_pool=$(exclude_items_from_list $ost_indices \
5627 local ost_targets="$FSNAME-OST["
5628 for i in ${ost_idx_in_pool//,/ }; do
5629 ost_targets=$ost_targets$(printf "%04x," $i)
5631 ost_targets="${ost_targets%,}]"
5633 local ost_targets_uuid=$(for i in ${ost_idx_in_pool//,/ }; \
5634 do printf "$FSNAME-OST%04x_UUID\n" $i; done |
5635 sort -u | tr '\n' ' ')
5637 local cmd="$LCTL pool_add $ost_pool $ost_targets"
5638 do_facet mgs $cmd || error "$cmd failed"
5639 wait_update $HOSTNAME "$LCTL get_param -n lov.$FSNAME-*.pools.$TESTNAME|
5640 sort -u | tr '\n' ' ' " "$ost_targets_uuid" ||
5641 error "wait_update $ost_pool failed"
5642 pool_list $ost_pool || error "list OST pool $ost_pool failed"
5644 # If [--pool|-p <pool_name>] is set with [--ost-list|-o <ost_indices>],
5645 # then the OSTs must be the members of the pool.
5646 local file=$DIR/$tdir/$tfile
5647 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
5649 eval $cmd && error "OST with index $ost_idx_in_list should be" \
5650 "in OST pool $ost_pool"
5652 # Only select OST $ost_idx_in_list from $ost_pool for file.
5653 ost_idx_in_list=${ost_idx_in_pool#*,}
5654 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
5656 eval $cmd || error "$cmd failed"
5657 cmd="$GETSTRIPE $file"
5659 eval $cmd || error "$cmd failed"
5660 check_stripe_count $file 2
5661 check_obdidx $file $ost_idx_in_list
5662 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5663 error "write $file failed"
5667 run_test 82b "specify OSTs for file with --pool and --ost-list options"
5670 [[ $(lustre_version_code ost1) -ge $(version_code 2.6.91) ]] ||
5671 { skip "Need OST version at least 2.6.91" && return 0; }
5672 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
5673 skip "Only applicable to ldiskfs-based MDTs"
5683 ostmnt=$(facet_mntpt ost1)
5684 fstype=$(facet_fstype ost1)
5686 # Mount the OST as an ldiskfs filesystem.
5687 log "mount the OST $dev as a $fstype filesystem"
5688 add ost1 $(mkfs_opts ost1 $dev) $FSTYPE_OPT \
5689 --reformat $dev $dev > /dev/null ||
5690 error "format ost1 error"
5692 if ! test -b $dev; then
5693 mnt_opts=$(csa_add "$OST_MOUNT_OPTS" -o loop)
5695 echo "mnt_opts $mnt_opts"
5696 do_facet ost1 mount -t $fstype $dev \
5698 # Run llverfs on the mounted ldiskfs filesystem.
5699 # It is needed to get ENOSPACE.
5700 log "run llverfs in partial mode on the OST $fstype $ostmnt"
5701 do_rpc_nodes $(facet_host ost1) run_llverfs $ostmnt -vpl \
5702 "no" || error "run_llverfs error on $fstype"
5705 log "unmount the OST $dev"
5708 # Delete file IO_scrub. Later osd_scrub_setup will try to
5709 # create "IO_scrub" but will get ENOSPACE.
5711 echo "start ost1 service on `facet_active_host ost1`"
5712 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
5715 err=$(do_facet ost1 dmesg | grep "VFS: Busy inodes after unmount of")
5716 echo "string err $err"
5717 [ -z "$err" ] || error $err
5720 run_test 83 "ENOSPACE on OST doesn't cause message VFS: \
5721 Busy inodes after unmount ..."
5724 local facet=$SINGLEMDS
5725 local num=$(echo $facet | tr -d "mds")
5726 local dev=$(mdsdevname $num)
5727 local time_min=$(recovery_time_min)
5728 local recovery_duration
5729 local completed_clients
5730 local correct_clients
5733 echo "start mds service on $(facet_active_host $facet)"
5735 "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
5736 error "start MDS failed"
5738 start_ost || error "start OST0000 failed"
5739 start_ost2 || error "start OST0001 failed"
5741 echo "recovery_time=$time_min, timeout=$TIMEOUT, wrap_up=$wrap_up"
5743 mount_client $MOUNT1 || error "mount $MOUNT1 failed"
5744 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
5745 # make sure new superblock labels are sync'd before disabling writes
5749 replay_barrier $SINGLEMDS
5750 createmany -o $DIR1/$tfile-%d 1000
5752 # We need to catch the end of recovery window to extend it.
5753 # Skip 5 requests and add delay to request handling.
5754 #define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP
5755 do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5"
5757 facet_failover --fsck $SINGLEMDS || error "failover: $?"
5760 echo "recovery status"
5761 do_facet $SINGLEMDS \
5762 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status"
5764 recovery_duration=$(do_facet $SINGLEMDS \
5765 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
5766 awk '/recovery_duration/ { print $2 }')
5767 (( $recovery_duration > $time_min + $wrap_up )) &&
5768 error "recovery_duration > recovery_time_hard + wrap up"
5769 completed_clients=$(do_facet $SINGLEMDS \
5770 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
5771 awk '/completed_clients/ { print $2 }')
5773 correct_clients="$MDSCOUNT/$((MDSCOUNT+1))"
5774 [ "$completed_clients" = "${correct_clients}" ] ||
5775 error "$completed_clients != $correct_clients"
5777 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
5778 umount_client $MOUNT1
5779 umount_client $MOUNT2
5785 run_test 84 "check recovery_hard_time"
5788 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.55) ]] ||
5789 { skip "Need OST version at least 2.7.55" && return 0; }
5790 ##define OBD_FAIL_OSD_OST_EA_FID_SET 0x197
5791 do_facet ost1 "lctl set_param fail_loc=0x197"
5795 run_test 85 "osd_ost init: fail ea_fid_set"
5798 [ "$(facet_fstype ost1)" = "zfs" ] &&
5799 skip "LU-6442: no such mkfs params for ZFS OSTs" && return
5801 local OST_OPTS="$(mkfs_opts ost1 $(ostdevname 1)) \
5802 --reformat $(ostdevname 1) $(ostvdevname 1)"
5805 local OLDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
5806 awk '/Flex block group size: / { print $NF; exit; }')
5809 if [[ ${!opts} != *mkfsoptions* ]]; then
5810 eval opts=\"${!opts} \
5811 --mkfsoptions='\\\"-O flex_bg -G $NEWSIZE\\\"'\"
5813 val=${!opts//--mkfsoptions=\\\"/ \
5814 --mkfsoptions=\\\"-O flex_bg -G $NEWSIZE }
5818 echo "params: $opts"
5820 add ost1 $opts || error "add ost1 failed with new params"
5822 local FOUNDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
5823 awk '/Flex block group size: / { print $NF; exit; }')
5825 [[ $FOUNDSIZE == $NEWSIZE ]] ||
5826 error "Flex block group size: $FOUNDSIZE, expected: $NEWSIZE"
5829 run_test 86 "Replacing mkfs.lustre -G option"
5831 test_87() { #LU-6544
5832 [[ $(lustre_version_code $SINGLEMDS1) -ge $(version_code 2.7.56) ]] ||
5833 { skip "Need MDS version at least 2.7.56" && return; }
5834 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ]] &&
5835 { skip "Only applicable to ldiskfs-based MDTs" && return; }
5836 [[ $OSTCOUNT -gt 69 ]] &&
5837 { skip "Ignore wide striping situation" && return; }
5839 local mdsdev=$(mdsdevname 1)
5840 local mdsvdev=$(mdsvdevname 1)
5841 local file=$DIR/$tfile
5842 local mntpt=$(facet_mntpt $SINGLEMDS)
5843 local used_xattr_blk=0
5844 local inode_size=${1:-512}
5846 local xtest="trusted.test"
5851 #Please see LU-6544 for MDT inode size calculation
5852 if [ $OSTCOUNT -gt 26 ]; then
5854 elif [ $OSTCOUNT -gt 5 ]; then
5857 left_size=$(expr $inode_size - \
5860 32 - $OSTCOUNT \* 24 - 16 - 3 - \
5862 24 - 18 - $(expr length $tfile) - 16 - 4)
5863 if [ $left_size -le 0 ]; then
5864 echo "No space($left_size) is expected in inode."
5865 echo "Try 1-byte xattr instead to verify this."
5868 echo "Estimate: at most $left_size-byte space left in inode."
5874 add mds1 $(mkfs_opts mds1 ${mdsdev}) --stripe-count-hint=$OSTCOUNT \
5875 --reformat $mdsdev $mdsvdev || error "add mds1 failed"
5876 start_mdt 1 > /dev/null || error "start mdt1 failed"
5877 for i in $(seq $OSTCOUNT); do
5878 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS > /dev/null ||
5879 error "start ost$i failed"
5881 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
5882 check_mount || error "check client $MOUNT failed"
5885 $SETSTRIPE -c -1 $file || error "$SETSTRIPE -c -1 $file failed"
5886 $GETSTRIPE $file || error "$GETSTRIPE $file failed"
5887 i=$($GETSTRIPE -c $file)
5888 if [ $i -ne $OSTCOUNT ]; then
5889 left_size=$(expr $left_size + $(expr $OSTCOUNT - $i) \* 24)
5890 echo -n "Since only $i out $OSTCOUNT OSTs are used, "
5891 echo -n "the expected left space is changed to "
5892 echo "$left_size bytes at most."
5894 value=$(generate_string $left_size)
5895 setfattr -n $xtest -v $value $file
5896 orig=$(get_xattr_value $xtest $file)
5897 [[ "$orig" != "$value" ]] && error "$xtest changed"
5899 #Verify if inode has some expected space left
5900 umount $MOUNT > /dev/null || error "umount $MOUNT failed"
5901 stop_mdt 1 > /dev/null || error "stop mdt1 failed"
5902 mount_ldiskfs $SINGLEMDS || error "mount -t ldiskfs $SINGLEMDS failed"
5904 do_facet $SINGLEMDS ls -sal $mntpt/ROOT/$tfile
5905 used_xattr_blk=$(do_facet $SINGLEMDS ls -s $mntpt/ROOT/$tfile |
5907 [[ $used_xattr_blk -eq 0 ]] &&
5908 error "Please check MDS inode size calculation: \
5909 more than $left_size-byte space left in inode."
5910 echo "Verified: at most $left_size-byte space left in inode."
5914 run_test 87 "check if MDT inode can hold EAs with N stripes properly"
5917 [ "$(facet_fstype mds1)" == "zfs" ] &&
5918 skip "LU-6662: no implementation for ZFS" && return
5922 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
5923 --reformat $(mdsdevname 1) || error "add mds1 failed"
5925 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
5926 grep -e \".*opts:.*errors=remount-ro.*\"" ||
5927 error "default mount options is missing"
5929 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
5930 --mountfsoptions="user_xattr,errors=panic" \
5931 --reformat $(mdsdevname 1) || error "add mds1 failed"
5933 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
5934 grep -e \".*opts:.*errors=panic.*\"" ||
5935 error "user can't override default mount options"
5937 run_test 88 "check the default mount options can be overridden"
5940 # $2 (optional) value of max_mod_rpcs_in_flight to set
5941 check_max_mod_rpcs_in_flight() {
5949 idx=$(printf "%04x" $($LFS getdirstripe -i $dir))
5950 facet="mds$((0x$idx + 1))"
5952 if [ -z "$mmr" ]; then
5953 # get value of max_mod_rcps_in_flight
5954 mmr=$($LCTL get_param -n \
5955 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) ||
5956 error "Unable to get max_mod_rpcs_in_flight"
5957 echo "max_mod_rcps_in_flight is $mmr"
5959 # set value of max_mod_rpcs_in_flight
5961 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight=$mmr ||
5962 error "Unable to set max_mod_rpcs_in_flight to $mmr"
5963 echo "max_mod_rpcs_in_flight set to $mmr"
5966 # create mmr+1 files
5967 echo "creating $((mmr + 1)) files ..."
5969 for i in $(seq $((mmr + 1))); do
5975 # consumes mmr-1 modify RPC slots
5976 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
5977 # drop requests on MDT so that RPC slots are consumed
5978 # during all the request resend interval
5979 do_facet $facet "$LCTL set_param fail_loc=0x159"
5980 echo "launch $((mmr - 1)) chmod in parallel ..."
5981 for i in $(seq $((mmr - 1))); do
5982 chmod 0600 $dir/file-$i &
5986 # send one additional modify RPC
5987 do_facet $facet "$LCTL set_param fail_loc=0"
5988 echo "launch 1 additional chmod in parallel ..."
5989 chmod 0600 $dir/file-$mmr &
5992 # check this additional modify RPC get a modify RPC slot
5993 # and succeed its operation
5994 checkstat -vp 0600 $dir/file-$mmr ||
5995 error "Unable to send $mmr modify RPCs in parallel"
6000 # consumes mmr modify RPC slots
6001 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6002 # drop requests on MDT so that RPC slots are consumed
6003 # during all the request resend interval
6004 do_facet $facet "$LCTL set_param fail_loc=0x159"
6005 echo "launch $mmr chmod in parallel ..."
6006 for i in $(seq $mmr); do
6007 chmod 0666 $dir/file-$i &
6011 # send one additional modify RPC
6012 do_facet $facet "$LCTL set_param fail_loc=0"
6013 echo "launch 1 additional chmod in parallel ..."
6014 chmod 0666 $dir/file-$((mmr + 1)) &
6017 # check this additional modify RPC blocked getting a modify RPC slot
6018 checkstat -vp 0644 $dir/file-$((mmr + 1)) ||
6019 error "Unexpectedly send $(($mmr + 1)) modify RPCs in parallel"
6025 if ! combined_mgs_mds ; then
6030 [[ $($LCTL get_param mdc.*.import |
6031 grep "connect_flags:.*multi_mod_rpc") ]] ||
6032 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6034 # check default value
6035 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6036 check_max_mod_rpcs_in_flight $DIR/$tdir
6040 run_test 90a "check max_mod_rpcs_in_flight is enforced"
6050 [[ $($LCTL get_param mdc.*.import |
6051 grep "connect_flags:.*multi_mod_rpc") ]] ||
6052 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6055 # update max_mod_rpcs_in_flight
6056 $LFS mkdir -c1 $DIR/${tdir}1 || error "mkdir $DIR/${tdir}1 failed"
6057 check_max_mod_rpcs_in_flight $DIR/${tdir}1 1
6060 # check client is able to send multiple modify RPCs in paralell
6061 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
6062 grep -c "multi_mod_rpcs")
6063 if [ "$tmp" -ne $MDSCOUNT ]; then
6064 echo "Client not able to send multiple modify RPCs in parallel"
6069 # update max_mod_rpcs_in_flight
6070 $LFS mkdir -c1 $DIR/${tdir}2 || error "mkdir $DIR/${tdir}2 failed"
6071 check_max_mod_rpcs_in_flight $DIR/${tdir}2 5
6074 $LFS mkdir -c1 $DIR/${tdir}3 || error "mkdir $DIR/${tdir}3 failed"
6075 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/${tdir}3))
6076 facet="mds$((0x$idx + 1))"
6078 # save MDT max_mod_rpcs_per_client
6079 mmrpc=$(do_facet $facet \
6080 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
6082 # update max_mod_rpcs_in_flight
6083 umount_client $MOUNT
6085 "echo 16 > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6087 $LCTL set_param mdc.$FSNAME-MDT$idx-mdc-*.max_rpcs_in_flight=17
6088 check_max_mod_rpcs_in_flight $DIR/${tdir}3 16
6090 # restore MDT max_mod_rpcs_per_client initial value
6092 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6094 rm -rf $DIR/${tdir}?
6097 run_test 90b "check max_mod_rpcs_in_flight is enforced after update"
6106 [[ $($LCTL get_param mdc.*.import |
6107 grep "connect_flags:.*multi_mod_rpc") ]] ||
6108 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6110 # check client is able to send multiple modify RPCs in paralell
6111 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
6112 grep -c "multi_mod_rpcs")
6113 if [ "$tmp" -ne $MDSCOUNT ]; then
6114 skip "Client not able to send multiple modify RPCs in parallel"
6119 # get max_rpcs_in_flight value
6120 mrif=$($LCTL get_param -n mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
6121 echo "max_rpcs_in_flight is $mrif"
6123 # get MDT max_mod_rpcs_per_client
6124 mmrpc=$(do_facet mds1 \
6125 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
6126 echo "max_mod_rpcs_per_client is $mmrpc"
6129 # attempt to set max_mod_rpcs_in_flight to max_rpcs_in_flight value
6130 # prerequisite: set max_mod_rpcs_per_client to max_rpcs_in_flight value
6131 umount_client $MOUNT
6133 "echo $mrif > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6137 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif &&
6138 error "set max_mod_rpcs_in_flight to $mrif should fail"
6140 umount_client $MOUNT
6142 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6146 # attempt to set max_mod_rpcs_in_flight to max_mod_rpcs_per_client+1
6147 # prerequisite: set max_rpcs_in_flight to max_mod_rpcs_per_client+2
6149 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$((mmrpc + 2))
6152 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$((mmrpc + 1)) &&
6153 error "set max_mod_rpcs_in_flight to $((mmrpc + 1)) should fail"
6157 run_test 90c "check max_mod_rpcs_in_flight update limits"
6168 [[ $($LCTL get_param mdc.*.import |
6169 grep "connect_flags:.*multi_mod_rpc") ]] ||
6170 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6172 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6173 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
6174 facet="mds$((0x$idx + 1))"
6176 # check client version supports multislots
6177 tmp=$($LCTL get_param -N \
6178 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
6179 if [ -z "$tmp" ]; then
6180 skip "Client does not support multiple modify RPCs in flight"
6185 # get current value of max_mod_rcps_in_flight
6186 mmr=$($LCTL get_param -n \
6187 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
6188 echo "max_mod_rcps_in_flight is $mmr"
6191 echo "creating $mmr files ..."
6193 for i in $(seq $mmr); do
6194 touch $DIR/$tdir/file-$i
6197 # prepare for close RPC
6198 multiop_bg_pause $DIR/$tdir/file-close O_c
6201 # consumes mmr modify RPC slots
6202 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6203 # drop requests on MDT so that RPC slots are consumed
6204 # during all the request resend interval
6205 do_facet $facet "$LCTL set_param fail_loc=0x159"
6206 echo "launch $mmr chmod in parallel ..."
6207 for i in $(seq $mmr); do
6208 chmod 0600 $DIR/$tdir/file-$i &
6211 # send one additional close RPC
6212 do_facet $facet "$LCTL set_param fail_loc=0"
6213 echo "launch 1 additional close in parallel ..."
6215 cancel_lru_locks mdc
6218 # check this additional close RPC get a modify RPC slot
6219 # and multiop process completed
6220 [ -d /proc/$pid ] &&
6221 error "Unable to send the additional close RPC in parallel"
6226 run_test 90d "check one close RPC is allowed above max_mod_rpcs_in_flight"
6228 check_uuid_on_ost() {
6230 do_facet ost1 "$LCTL get_param obdfilter.${FSNAME}*.exports.'$nid'.uuid"
6233 check_uuid_on_mdt() {
6235 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
6243 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.63) ]] ||
6244 { skip "Need OST version at least 2.7.63" && return 0; }
6245 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.63) ]] ||
6246 { skip "Need MDT version at least 2.7.63" && return 0; }
6248 start_mds || error "MDS start failed"
6249 start_ost || error "unable to start OST"
6250 mount_client $MOUNT || error "client start failed"
6251 check_mount || error "check_mount failed"
6254 nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g")
6258 uuid=$(get_client_uuid $MOUNT)
6260 echo "list nids on mdt:"
6261 do_facet $SINGLEMDS "$LCTL list_param mdt.${FSNAME}*.exports.*"
6262 echo "uuid from $nid:"
6263 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
6265 found=$(check_uuid_on_mdt $nid | grep $uuid)
6266 [ -z "$found" ] && error "can't find $uuid $nid on MDT"
6267 found=$(check_uuid_on_ost $nid | grep $uuid)
6268 [ -z "$found" ] && error "can't find $uuid $nid on OST"
6270 # umount the client so it won't reconnect
6271 manual_umount_client --force || error "failed to umount $?"
6272 # shouldn't disappear on MDS after forced umount
6273 found=$(check_uuid_on_mdt $nid | grep $uuid)
6274 [ -z "$found" ] && error "can't find $uuid $nid"
6277 do_facet $SINGLEMDS \
6278 "$LCTL set_param -n mdt.${mds1_svc}.evict_client nid:$nid"
6280 found=$(check_uuid_on_mdt $nid | grep $uuid)
6281 [ -n "$found" ] && error "found $uuid $nid on MDT"
6282 found=$(check_uuid_on_ost $nid | grep $uuid)
6283 [ -n "$found" ] && error "found $uuid $nid on OST"
6285 # check it didn't reconnect (being umounted)
6286 sleep $((TIMEOUT+1))
6287 found=$(check_uuid_on_mdt $nid | grep $uuid)
6288 [ -n "$found" ] && error "found $uuid $nid on MDT"
6289 found=$(check_uuid_on_ost $nid | grep $uuid)
6290 [ -n "$found" ] && error "found $uuid $nid on OST"
6294 run_test 91 "evict-by-nid support"
6296 generate_ldev_conf() {
6297 # generate an ldev.conf file
6298 local ldevconfpath=$1
6300 printf "%s\t-\t%s-MGS0000\t%s\n" \
6303 $(mgsdevname) >> $ldevconfpath
6305 local mdsfo_host=$mdsfailover_HOST;
6306 if [ -z "$mdsfo_host" ]; then
6310 for num in $(seq $MDSCOUNT); do
6311 printf "%s\t%s\t%s-MDT%04d\t%s\n" \
6316 $(mdsdevname $num) >> $ldevconfpath
6319 local ostfo_host=$ostfailover_HOST;
6320 if [ -z "$ostfo_host" ]; then
6324 for num in $(seq $OSTCOUNT); do
6325 printf "%s\t%s\t%s-OST%04d\t%s\n" \
6330 $(ostdevname $num) >> $ldevconfpath
6335 # generate a nids file (mapping between hostname to nid)
6336 # looks like we only have the MGS nid available to us
6337 # so just echo that to a file
6340 echo -e "${mgs_HOST}\t${MGSNID}" >> $nidspath
6344 local LDEVCONFPATH=$TMP/ldev.conf
6345 local NIDSPATH=$TMP/nids
6347 echo "Host is $(hostname)"
6349 generate_ldev_conf $LDEVCONFPATH
6350 generate_nids $NIDSPATH
6352 echo "----- ldev.conf -----"
6354 echo "--- END ldev.conf ---"
6356 echo "----- /etc/nids -----"
6358 echo "--- END /etc/nids ---"
6360 # ldev can be in our build tree and if we aren't in a
6361 # build tree, use 'which' to try and find it
6362 local LDEV=$LUSTRE/scripts/ldev
6363 [ ! -f "$LDEV" ] && local LDEV=$(which ldev 2> /dev/null)
6365 echo "ldev path is $LDEV"
6367 if [ ! -f "$LDEV" ]; then
6368 rm $LDEVCONFPATH $NIDSPATH
6369 error "failed to find ldev!"
6372 # echo the mgs nid and compare it to environment variable MGSNID
6373 # also, ldev.conf and nids is a server side thing, use the OSS
6376 output=$(perl $LDEV -c $LDEVCONFPATH -H \
6377 $ost_HOST -n $NIDSPATH echo %m)
6379 echo "-- START OF LDEV OUTPUT --"
6381 echo "--- END OF LDEV OUTPUT ---"
6383 # ldev failed, error
6384 if [ $? -ne 0 ]; then
6385 rm $LDEVCONFPATH $NIDSPATH
6386 error "ldev failed to execute!"
6389 # need to process multiple lines because of combined MGS and MDS
6390 echo -e $output | awk '{ print $2 }' | while read -r line ; do
6391 if [ "$line" != "$MGSNID" ]; then
6392 rm $LDEVCONFPATH $NIDSPATH
6393 error "ldev failed mgs nid '$line', expected '$MGSNID'"
6397 rm $LDEVCONFPATH $NIDSPATH
6399 run_test 92 "ldev returns MGS NID correctly in command substitution"
6401 if ! combined_mgs_mds ; then
6407 # restore the values of MDSSIZE and OSTSIZE
6408 MDSSIZE=$STORED_MDSSIZE
6409 OSTSIZE=$STORED_OSTSIZE