2 # -*- mode: Bash; tab-width: 4; indent-tabs-mode: t; -*-
3 # vim:shiftwidth=4:softtabstop=4:tabstop=4:
7 # bug number: LU-2012 10124
8 ALWAYS_EXCEPT="14b 15c $REPLAY_DUAL_EXCEPT"
11 PTLDEBUG=${PTLDEBUG:--1}
12 LUSTRE=${LUSTRE:-$(cd $(dirname $0)/..; echo $PWD)}
14 CLEANUP=${CLEANUP:-""}
15 MOUNT_2=${MOUNT_2:-"yes"}
16 export MULTIOP=${MULTIOP:-multiop}
17 . $LUSTRE/tests/test-framework.sh
20 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
23 remote_mds_nodsh && skip "remote MDS with nodsh" && exit 0
25 [ "$SLOW" = "no" ] && EXCEPT_SLOW="21b"
27 [[ $(facet_fstype $SINGLEMDS) == zfs ]] &&
28 # bug number for skipped test: LU-2230
29 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 21b"
33 check_and_setup_lustre
34 MOUNTED=$(mounted_lustre_filesystems)
35 if ! $(echo $MOUNTED' ' | grep -w -q $MOUNT2' '); then
36 zconf_mount $HOSTNAME $MOUNT2
41 rm -rf $DIR/[df][0-9]*
43 [ "$DAEMONFILE" ] && $LCTL debug_daemon start $DAEMONFILE $DAEMONSIZE
45 # if there is no CLIENT1 defined, some tests can be ran on localhost
46 CLIENT1=${CLIENT1:-$HOSTNAME}
47 # if CLIENT2 doesn't exist then use CLIENT1 instead
48 # All tests should use CLIENT2 with MOUNT2 only therefore it will work if
50 # Exception is the test which need two separate nodes
51 CLIENT2=${CLIENT2:-$CLIENT1}
53 # LU-482 Avert LVM and VM inability to flush caches in pre .33 kernels
54 if [ $LINUX_VERSION_CODE -lt $(version_code 2.6.33) ]; then
56 do_facet $SINGLEMDS "sync; sleep 10; sync; sleep 10; sync"
59 LU482_FAILED=$(mktemp -u $TMP/$TESTSUITE.lu482.XXXXXX)
61 echo "Check file is LU482_FAILED=$LU482_FAILED"
62 touch $MOUNT2/$tfile-A # force sync FLD/SEQ update before barrier
63 replay_barrier $SINGLEMDS
64 #define OBD_FAIL_PTLRPC_FINISH_REPLAY | OBD_FAIL_ONCE
66 createmany -o $MOUNT1/$tfile- 50
67 $LCTL set_param fail_loc=0x80000514
68 facet_failover $SINGLEMDS
69 [ -f "$LU482_FAILED" ] && skip "LU-482 failure" && return 0
73 zconf_mount `hostname` $MOUNT2 || error "mount2 fais"
74 unlinkmany $MOUNT1/$tfile- 50 || return 2
75 rm $MOUNT2/$tfile || return 3
76 rm $MOUNT2/$tfile-A || return 4
78 run_test 0a "expired recovery with lost client"
80 if [ -f "$LU482_FAILED" ]; then
81 log "Found check file $LU482_FAILED, aborting test script"
82 rm -vf "$LU482_FAILED"
84 do_nodes $CLIENTS umount -f $MOUNT2 || true
85 do_nodes $CLIENTS umount -f $MOUNT || true
86 # copied from stopall, but avoid the MDS recovery
87 for num in `seq $OSTCOUNT`; do
89 rm -f $TMP/ost${num}active
91 if ! combined_mgs_mds ; then
99 replay_barrier $SINGLEMDS
101 touch $MOUNT1/$tfile-2
103 facet_failover $SINGLEMDS
105 zconf_mount `hostname` $MOUNT1 || error "mount1 fais"
106 zconf_mount `hostname` $MOUNT2 || error "mount2 fais"
107 # it is uncertain if file-2 exists or not, remove it if it does
108 checkstat $MOUNT1/$tfile-2 && rm $MOUNT1/$tfile-2
109 checkstat $MOUNT2/$tfile && return 2
112 run_test 0b "lost client during waiting for next transno"
116 replay_barrier $SINGLEMDS
120 checkstat $MOUNT2/a || return 1
121 checkstat $MOUNT1/b || return 2
122 rm $MOUNT2/a $MOUNT1/b
123 checkstat $MOUNT1/a && return 3
124 checkstat $MOUNT2/b && return 4
128 run_test 1 "|X| simple create"
132 replay_barrier $SINGLEMDS
136 checkstat $MOUNT2/adir || return 1
138 checkstat $MOUNT2/adir && return 2
141 run_test 2 "|X| mkdir adir"
144 replay_barrier $SINGLEMDS
146 mkdir $MOUNT2/adir/bdir
149 checkstat $MOUNT2/adir || return 1
150 checkstat $MOUNT1/adir/bdir || return 2
151 rmdir $MOUNT2/adir/bdir $MOUNT1/adir
152 checkstat $MOUNT1/adir && return 3
153 checkstat $MOUNT2/adir/bdir && return 4
156 run_test 3 "|X| mkdir adir, mkdir adir/bdir "
160 replay_barrier $SINGLEMDS
161 mkdir $MOUNT1/adir && return 1
162 mkdir $MOUNT2/adir/bdir
165 checkstat $MOUNT2/adir || return 2
166 checkstat $MOUNT1/adir/bdir || return 3
168 rmdir $MOUNT2/adir/bdir $MOUNT1/adir
169 checkstat $MOUNT1/adir && return 4
170 checkstat $MOUNT2/adir/bdir && return 5
173 run_test 4 "|X| mkdir adir (-EEXIST), mkdir adir/bdir "
177 # multiclient version of replay_single.sh/test_8
179 multiop_bg_pause $MOUNT2/a o_tSc || return 1
182 replay_barrier $SINGLEMDS
184 wait $pid || return 1
187 [ -e $MOUNT2/a ] && return 2
190 run_test 5 "open, unlink |X| close"
195 multiop_bg_pause $MOUNT2/a o_c || return 1
197 multiop_bg_pause $MOUNT1/a o_c || return 1
200 replay_barrier $SINGLEMDS
202 wait $pid1 || return 1
206 wait $pid2 || return 1
207 [ -e $MOUNT2/a ] && return 2
210 run_test 6 "open1, open2, unlink |X| close1 [fail $SINGLEMDS] close2"
213 replay_barrier $SINGLEMDS
214 drop_reint_reply "mcreate $MOUNT1/$tfile" || return 1
216 checkstat $MOUNT2/$tfile || return 2
217 rm $MOUNT1/$tfile || return 3
221 run_test 8 "replay of resent request"
224 replay_barrier $SINGLEMDS
225 mcreate $MOUNT1/$tfile-1
226 mcreate $MOUNT2/$tfile-2
227 # drop first reint reply
228 do_facet $SINGLEMDS lctl set_param fail_loc=0x80000119
230 do_facet $SINGLEMDS lctl set_param fail_loc=0
232 rm $MOUNT1/$tfile-[1,2] || return 1
236 run_test 9 "resending a replayed create"
239 mcreate $MOUNT1/$tfile-1
240 replay_barrier $SINGLEMDS
241 munlink $MOUNT1/$tfile-1
242 mcreate $MOUNT2/$tfile-2
243 # drop first reint reply
244 do_facet $SINGLEMDS lctl set_param fail_loc=0x80000119
246 do_facet $SINGLEMDS lctl set_param fail_loc=0
248 checkstat $MOUNT1/$tfile-1 && return 1
249 checkstat $MOUNT1/$tfile-2 || return 2
254 run_test 10 "resending a replayed unlink"
257 replay_barrier $SINGLEMDS
258 mcreate $MOUNT1/$tfile-1
259 mcreate $MOUNT2/$tfile-2
260 mcreate $MOUNT1/$tfile-3
261 mcreate $MOUNT2/$tfile-4
262 mcreate $MOUNT1/$tfile-5
263 # drop all reint replies for a while
264 do_facet $SINGLEMDS lctl set_param fail_loc=0x0119
265 # note that with this fail_loc set, facet_failover df will fail
266 facet_failover $SINGLEMDS
267 #sleep for while, let both clients reconnect and timeout
268 sleep $((TIMEOUT * 2))
269 do_facet $SINGLEMDS lctl set_param fail_loc=0
271 rm $MOUNT1/$tfile-[1-5] || return 1
275 run_test 11 "both clients timeout during replay"
278 replay_barrier $SINGLEMDS
280 multiop_bg_pause $DIR/$tfile mo_c || return 1
283 #define OBD_FAIL_LDLM_ENQUEUE_NET 0x302
284 do_facet $SINGLEMDS lctl set_param fail_loc=0x80000302
285 facet_failover $SINGLEMDS
286 do_facet $SINGLEMDS lctl set_param fail_loc=0
287 clients_up || return 1
290 kill -USR1 $MULTIPID || return 3
291 wait $MULTIPID || return 4
292 $CHECKSTAT -t file $DIR/$tfile || return 2
297 run_test 12 "open resend timeout"
300 multiop_bg_pause $DIR/$tfile mo_c || return 1
303 replay_barrier $SINGLEMDS
305 kill -USR1 $MULTIPID || return 3
306 wait $MULTIPID || return 4
309 do_facet $SINGLEMDS lctl set_param fail_loc=0x80000115
310 facet_failover $SINGLEMDS
311 do_facet $SINGLEMDS lctl set_param fail_loc=0
312 clients_up || return 1
315 $CHECKSTAT -t file $DIR/$tfile || return 2
320 run_test 13 "close resend timeout"
322 # test 14a removed after 18143 because it shouldn't fail anymore and do the same
327 wait_delete_completed
329 local BEFOREUSED=$(df -P $DIR | tail -1 | awk '{ print $3 }')
331 mkdir -p $MOUNT1/$tdir
332 $SETSTRIPE -i 0 $MOUNT1/$tdir
333 replay_barrier $SINGLEMDS
334 createmany -o $MOUNT1/$tdir/$tfile- 5
336 $SETSTRIPE -i 0 $MOUNT2/$tfile-2
337 dd if=/dev/zero of=$MOUNT2/$tfile-2 bs=1M count=5
338 createmany -o $MOUNT1/$tdir/$tfile-3- 5
342 wait_recovery_complete $SINGLEMDS || error "MDS recovery not done"
344 # first set of files should have been replayed
345 unlinkmany $MOUNT1/$tdir/$tfile- 5 || error "first unlinks failed"
346 unlinkmany $MOUNT1/$tdir/$tfile-3- 5 || error "second unlinks failed"
348 zconf_mount $HOSTNAME $MOUNT2 || error "mount $MOUNT2 failed"
349 [ -f $MOUNT2/$tfile-2 ] && error "$MOUNT2/$tfile-2 exists!"
351 wait_mds_ost_sync || error "wait_mds_ost_sync failed"
352 wait_delete_completed || error "wait_delete_complete failed"
354 local AFTERUSED=$(df -P $DIR | tail -1 | awk '{ print $3 }')
355 log "before $BEFOREUSED, after $AFTERUSED"
356 # leave some margin for some files/dirs to be modified (OI, llog, etc)
357 [ $AFTERUSED -gt $((BEFOREUSED + 128)) ] &&
358 error "after $AFTERUSED > before $BEFOREUSED" || true
360 run_test 14b "delete ost orphans if gap occured in objids due to VBR"
362 test_15a() { # was test_15
363 replay_barrier $SINGLEMDS
364 createmany -o $MOUNT1/$tfile- 25
365 createmany -o $MOUNT2/$tfile-2- 1
370 unlinkmany $MOUNT1/$tfile- 25 || return 2
371 [ -e $MOUNT1/$tfile-2-0 ] && error "$tfile-2-0 exists"
373 zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
376 run_test 15a "timeout waiting for lost client during replay, 1 client completes"
379 replay_barrier $SINGLEMDS
380 for ((i = 0; i < 2000; i++)); do
381 echo "data" > "$MOUNT2/${tfile}-$i" || error "create ${tfile}-$i failed"
387 zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
390 run_test 15c "remove multiple OST orphans"
393 replay_barrier $SINGLEMDS
394 createmany -o $MOUNT1/$tfile- 25
395 createmany -o $MOUNT2/$tfile-2- 1
398 facet_failover $SINGLEMDS
402 unlinkmany $MOUNT1/$tfile- 25 || return 2
404 zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
408 run_test 16 "fail MDS during recovery (3571)"
411 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
413 createmany -o $MOUNT1/$tfile- 25
414 createmany -o $MOUNT2/$tfile-2- 1
416 # Make sure the disconnect is lost
424 unlinkmany $MOUNT1/$tfile- 25 || return 2
426 zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
430 run_test 17 "fail OST during recovery (3571)"
432 # cleanup with blocked enqueue fails until timer elapses (MDS busy), wait for it
435 test_18() { # bug 3822 - evicting client with enqueued lock
437 local DLMTRACE=$(do_facet $SINGLEMDS lctl get_param debug)
438 do_facet $SINGLEMDS lctl set_param debug=+dlmtrace
439 mkdir -p $MOUNT1/$tdir || error "mkdir $MOUNT1/$tdir failed"
440 touch $MOUNT1/$tdir/$tfile
441 #define OBD_FAIL_LDLM_ENQUEUE_BLOCKED 0x30b
442 statmany -s $MOUNT1/$tdir/f 1 500 &
445 do_facet $SINGLEMDS lctl set_param fail_loc=0x8000030b # hold enqueue
447 #define OBD_FAIL_LDLM_BL_CALLBACK_NET 0x305
448 do_facet client lctl set_param fail_loc=0x80000305 # drop cb, evict
450 usleep 500 # wait to ensure first client is one that will be evicted
451 openfile -f O_RDONLY $MOUNT2/$tdir/$tfile
453 do_facet $SINGLEMDS lctl debug_kernel |
454 grep "not entering recovery" && error "client not evicted"
455 do_facet client "lctl set_param fail_loc=0"
456 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
458 run_test 18 "ldlm_handle_enqueue succeeds on evicted export (3822)"
460 test_19() { # Bug 10991 - resend of open request does not fail assertion.
461 replay_barrier $SINGLEMDS
462 drop_ldlm_reply "createmany -o $DIR/$tfile 1" || return 1
464 checkstat $DIR2/${tfile}0 || return 2
465 rm $DIR/${tfile}0 || return 3
469 run_test 19 "resend of open request"
473 replay_barrier $SINGLEMDS
479 zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
480 TIER1=$((`date +%s` - BEFORE))
482 replay_barrier $SINGLEMDS
488 zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
489 TIER2=$((`date +%s` - BEFORE))
490 [ $TIER2 -ge $((TIER1 * 2)) ] && \
491 error "recovery time is growing $TIER2 > $TIER1"
494 run_test 20 "recovery time is not increasing"
496 # commit on sharing tests
498 local param_file=$TMP/$tfile-params
500 save_lustre_params $SINGLEMDS "mdt.*.commit_on_sharing" > $param_file
501 do_facet $SINGLEMDS lctl set_param mdt.*.commit_on_sharing=1
502 touch $MOUNT1/$tfile-1
503 mv $MOUNT2/$tfile-1 $MOUNT2/$tfile-2
504 mv $MOUNT1/$tfile-2 $MOUNT1/$tfile-3
505 replay_barrier_nosync $SINGLEMDS
508 facet_failover $SINGLEMDS
510 # all renames are replayed
511 unlink $MOUNT1/$tfile-3 || return 2
513 zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
515 do_facet $SINGLEMDS lctl set_param mdt.*.commit_on_sharing=0
516 rm -rf $MOUNT1/$tfile-*
517 restore_lustre_params < $param_file
521 run_test 21a "commit on sharing"
525 do_node $CLIENT1 rm -f $MOUNT1/$tfile-*
528 do_node $CLIENT1 touch $MOUNT1/$tfile-1
529 do_node $CLIENT2 mv $MOUNT1/$tfile-1 $MOUNT1/$tfile-2
530 do_node $CLIENT1 mv $MOUNT1/$tfile-2 $MOUNT1/$tfile-3
532 replay_barrier_nosync $mds
533 shutdown_client $CLIENT2 $MOUNT1
537 # were renames replayed?
539 echo UNLINK $MOUNT1/$tfile-3
540 do_node $CLIENT1 unlink $MOUNT1/$tfile-3 || \
541 { echo "unlink $tfile-3 fail!" && rc=1; }
544 zconf_mount_clients $CLIENT2 $MOUNT1 || error "mount $CLIENT2 $MOUNT1 fail"
550 [ -z "$CLIENTS" ] && skip "Need two or more clients." && return
551 [ $CLIENTCOUNT -lt 2 ] && \
552 { skip "Need two or more clients, have $CLIENTCOUNT" && return; }
554 if [ "$FAILURE_MODE" = "HARD" ] && mixed_mdt_devs; then
555 skip "Several mdt services on one mds node are used with FAILURE_MODE=$FAILURE_MODE. "
560 zconf_umount_clients $CLIENTS $MOUNT2
561 zconf_mount_clients $CLIENTS $MOUNT1
563 local param_file=$TMP/$tfile-params
565 local num=$(get_mds_dir $MOUNT1)
567 save_lustre_params mds$num "mdt.*.commit_on_sharing" > $param_file
571 do_facet mds$num lctl set_param mdt.*.commit_on_sharing=$COS
573 test_21b_sub mds$num || error "Not all renames are replayed. COS=$COS"
575 # COS disabled (should fail)
577 do_facet mds$num lctl set_param mdt.*.commit_on_sharing=$COS
579 # there is still a window when transactions may be written to disk before
580 # the mds device is set R/O. To avoid such a rare test failure, the check
581 # is repeated several times.
584 test_21b_sub mds$num || break;
585 let n_attempts=n_attempts+1
586 [ $n_attempts -gt 3 ] &&
587 error "The test cannot check whether COS works or not: all renames are replied w/o COS"
589 zconf_mount_clients $CLIENTS $MOUNT2
590 restore_lustre_params < $param_file
594 run_test 21b "commit on sharing, two clients"
597 checkstat $MOUNT1/$remote_dir || return 1
598 checkstat $MOUNT1/$remote_dir/dir || return 2
599 checkstat $MOUNT1/$remote_dir/$tfile-1 || return 3
600 checkstat $MOUNT1/$remote_dir/dir/$tfile-1 || return 4
604 create_remote_dir_files_22() {
605 do_node $CLIENT2 mkdir ${MOUNT2}/$remote_dir/dir || return 1
606 do_node $CLIENT1 createmany -o $MOUNT1/$remote_dir/dir/$tfile- 2 ||
608 do_node $CLIENT2 createmany -o $MOUNT2/$remote_dir/$tfile- 2 ||
614 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
615 ([ $FAILURE_MODE == "HARD" ] &&
616 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
617 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
621 local remote_dir=${tdir}/remote_dir
623 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
625 # OBD_FAIL_MDS_REINT_NET_REP 0x119
626 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
627 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir &
630 fail mds$((MDTIDX + 1))
631 wait $CLIENT_PID || error "lfs mkdir failed"
633 replay_barrier mds$MDTIDX
634 create_remote_dir_files_22 || error "Remote creation failed $?"
637 checkstat_22 || error "check stat failed $?"
639 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
642 run_test 22a "c1 lfs mkdir -i 1 dir1, M1 drop reply & fail, c2 mkdir dir1/dir"
645 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
647 local remote_dir=$tdir/remote_dir
649 # OBD_FAIL_MDS_REINT_NET_REP 0x119
650 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
652 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
653 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir &
656 fail mds${MDTIDX},mds$((MDTIDX + 1))
657 wait $CLIENT_PID || error "lfs mkdir failed"
659 replay_barrier mds$MDTIDX
660 create_remote_dir_files_22 || error "Remote creation failed $?"
663 checkstat_22 || error "check stat failed $?"
665 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
668 run_test 22b "c1 lfs mkdir -i 1 d1, M1 drop reply & fail M0/M1, c2 mkdir d1/dir"
671 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
672 ([ $FAILURE_MODE == "HARD" ] &&
673 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
674 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
677 local remote_dir=${tdir}/remote_dir
679 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
681 # OBD_FAIL_UPDATE_OBJ_NET_REP 0x1701
682 do_facet mds$MDTIDX lctl set_param fail_loc=0x1701
683 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir &
685 do_facet mds$MDTIDX lctl set_param fail_loc=0
688 wait $CLIENT_PID || error "lfs mkdir failed"
690 replay_barrier mds$MDTIDX
691 create_remote_dir_files_22 || error "Remote creation failed $?"
694 checkstat_22 || error "check stat failed $?"
696 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
699 run_test 22c "c1 lfs mkdir -i 1 d1, M1 drop update & fail M1, c2 mkdir d1/dir"
702 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
704 local remote_dir=${tdir}/remote_dir
706 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
708 # OBD_FAIL_UPDATE_OBJ_NET_REP 0x1701
709 do_facet mds$MDTIDX lctl set_param fail_loc=0x1701
710 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir &
712 do_facet mds$MDTIDX lctl set_param fail_loc=0
714 fail mds${MDTIDX},mds$((MDTIDX + 1))
715 wait $CLIENT_PID || error "lfs mkdir failed"
717 replay_barrier mds$MDTIDX
718 create_remote_dir_files_22 || error "Remote creation failed $?"
721 checkstat_22 || error "check stat failed $?"
723 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
726 run_test 22d "c1 lfs mkdir -i 1 d1, M1 drop update & fail M0/M1,c2 mkdir d1/dir"
729 checkstat $MOUNT1/$remote_dir || return 1
730 checkstat $MOUNT1/$remote_dir/$tfile-1 || return 2
734 create_remote_dir_files_23() {
735 do_node $CLIENT2 mkdir ${MOUNT2}/$remote_dir || return 1
736 do_node $CLIENT2 createmany -o $MOUNT2/$remote_dir/$tfile- 2 || return 2
741 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
742 ([ $FAILURE_MODE == "HARD" ] &&
743 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
744 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
747 local remote_dir=$tdir/remote_dir
749 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
750 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir ||
751 error "lfs mkdir failed"
752 # OBD_FAIL_MDS_REINT_NET_REP 0x119
753 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
754 do_node $CLIENT1 rmdir $MOUNT1/$remote_dir &
756 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0
758 fail mds$((MDTIDX + 1))
759 wait $CLIENT_PID || error "rmdir remote dir failed"
761 replay_barrier mds${MDTIDX}
762 create_remote_dir_files_23 || error "Remote creation failed $?"
765 checkstat_23 || error "check stat failed $?"
767 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
770 run_test 23a "c1 rmdir d1, M1 drop reply and fail, client2 mkdir d1"
773 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
775 local remote_dir=$tdir/remote_dir
777 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
778 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir ||
779 error "lfs mkdir failed"
781 # OBD_FAIL_MDS_REINT_NET_REP 0x119
782 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
783 do_node $CLIENT1 rmdir $MOUNT1/$remote_dir &
785 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0
787 fail mds${MDTIDX},mds$((MDTIDX + 1))
788 wait $CLIENT_PID || error "rmdir remote dir failed"
790 replay_barrier mds${MDTIDX}
791 create_remote_dir_files_23 || error "Remote creation failed $?"
794 checkstat_23 || error "check stat failed $?"
796 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
799 run_test 23b "c1 rmdir d1, M1 drop reply and fail M0/M1, c2 mkdir d1"
802 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
804 ([ $FAILURE_MODE == "HARD" ] &&
805 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
806 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
809 local remote_dir=$tdir/remote_dir
811 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
812 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir ||
813 error "lfs mkdir failed"
815 # OBD_FAIL_UPDATE_OBJ_NET_REP 0x1701
816 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
817 do_node $CLIENT1 rmdir $MOUNT1/$remote_dir &
819 do_facet mds${MDTIDX} lctl set_param fail_loc=0
822 wait $CLIENT_PID || error "rmdir remote dir failed"
824 replay_barrier mds${MDTIDX}
825 create_remote_dir_files_23 || error "Remote creation failed $?"
828 checkstat_23 || error "check stat failed $?"
830 rm -rf $MOUNT1/$tdir || return 6
833 run_test 23c "c1 rmdir d1, M0 drop update reply and fail M0, c2 mkdir d1"
836 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
838 local remote_dir=$tdir/remote_dir
840 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
841 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir ||
842 error "lfs mkdir failed"
844 # OBD_FAIL_UPDATE_OBJ_NET 0x1701
845 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
846 do_node $CLIENT1 rmdir $MOUNT1/$remote_dir &
848 do_facet mds${MDTIDX} lctl set_param fail_loc=0
850 fail mds${MDTIDX},mds$((MDTIDX + 1))
851 wait $CLIENT_PID || error "rmdir remote dir failed"
853 replay_barrier mds${MDTIDX}
854 create_remote_dir_files_23 || error "Remote creation failed $?"
857 checkstat_23 || error "check stat failed $?"
859 rm -rf $MOUNT1/$tdir || return 6
862 run_test 23d "c1 rmdir d1, M0 drop update reply and fail M0/M1, c2 mkdir d1"
864 # end commit on sharing tests
867 SLEEP=$((`date +%s` - $NOW))
868 [ $SLEEP -lt $TIMEOUT ] && sleep $SLEEP
869 [ "$MOUNTED2" = yes ] && zconf_umount $HOSTNAME $MOUNT2 || true
870 check_and_cleanup_lustre