5 # bug number: LU-2012 10124
6 ALWAYS_EXCEPT="14b 15c $REPLAY_DUAL_EXCEPT"
9 PTLDEBUG=${PTLDEBUG:--1}
10 LUSTRE=${LUSTRE:-$(cd $(dirname $0)/..; echo $PWD)}
12 CLEANUP=${CLEANUP:-""}
13 MOUNT_2=${MOUNT_2:-"yes"}
14 export MULTIOP=${MULTIOP:-multiop}
15 . $LUSTRE/tests/test-framework.sh
18 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
21 remote_mds_nodsh && skip "remote MDS with nodsh" && exit 0
23 [ "$SLOW" = "no" ] && EXCEPT_SLOW="21b"
25 [[ $(facet_fstype $SINGLEMDS) == zfs ]] &&
26 # bug number for skipped test: LU-2230
27 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 21b"
31 check_and_setup_lustre
32 MOUNTED=$(mounted_lustre_filesystems)
33 if ! $(echo $MOUNTED' ' | grep -w -q $MOUNT2' '); then
34 zconf_mount $HOSTNAME $MOUNT2
39 rm -rf $DIR/[df][0-9]*
41 [ "$DAEMONFILE" ] && $LCTL debug_daemon start $DAEMONFILE $DAEMONSIZE
43 # if there is no CLIENT1 defined, some tests can be ran on localhost
44 CLIENT1=${CLIENT1:-$HOSTNAME}
45 # if CLIENT2 doesn't exist then use CLIENT1 instead
46 # All tests should use CLIENT2 with MOUNT2 only therefore it will work if
48 # Exception is the test which need two separate nodes
49 CLIENT2=${CLIENT2:-$CLIENT1}
51 # LU-482 Avert LVM and VM inability to flush caches in pre .33 kernels
52 if [ $LINUX_VERSION_CODE -lt $(version_code 2.6.33) ]; then
54 do_facet $SINGLEMDS "sync; sleep 10; sync; sleep 10; sync"
57 LU482_FAILED=$(mktemp -u $TMP/$TESTSUITE.lu482.XXXXXX)
59 echo "Check file is LU482_FAILED=$LU482_FAILED"
60 touch $MOUNT2/$tfile-A # force sync FLD/SEQ update before barrier
61 replay_barrier $SINGLEMDS
62 #define OBD_FAIL_PTLRPC_FINISH_REPLAY | OBD_FAIL_ONCE
64 createmany -o $MOUNT1/$tfile- 50
65 $LCTL set_param fail_loc=0x80000514
66 facet_failover $SINGLEMDS
67 [ -f "$LU482_FAILED" ] && skip "LU-482 failure" && return 0
71 zconf_mount `hostname` $MOUNT2 || error "mount2 fais"
72 unlinkmany $MOUNT1/$tfile- 50 || return 2
73 rm $MOUNT2/$tfile || return 3
74 rm $MOUNT2/$tfile-A || return 4
76 run_test 0a "expired recovery with lost client"
78 if [ -f "$LU482_FAILED" ]; then
79 log "Found check file $LU482_FAILED, aborting test script"
80 rm -vf "$LU482_FAILED"
82 do_nodes $CLIENTS umount -f $MOUNT2 || true
83 do_nodes $CLIENTS umount -f $MOUNT || true
84 # copied from stopall, but avoid the MDS recovery
85 for num in `seq $OSTCOUNT`; do
87 rm -f $TMP/ost${num}active
89 if ! combined_mgs_mds ; then
97 replay_barrier $SINGLEMDS
99 touch $MOUNT1/$tfile-2
101 facet_failover $SINGLEMDS
103 zconf_mount `hostname` $MOUNT1 || error "mount1 fais"
104 zconf_mount `hostname` $MOUNT2 || error "mount2 fais"
105 # it is uncertain if file-2 exists or not, remove it if it does
106 checkstat $MOUNT1/$tfile-2 && rm $MOUNT1/$tfile-2
107 checkstat $MOUNT2/$tfile && return 2
110 run_test 0b "lost client during waiting for next transno"
114 replay_barrier $SINGLEMDS
118 checkstat $MOUNT2/a || return 1
119 checkstat $MOUNT1/b || return 2
120 rm $MOUNT2/a $MOUNT1/b
121 checkstat $MOUNT1/a && return 3
122 checkstat $MOUNT2/b && return 4
126 run_test 1 "|X| simple create"
130 replay_barrier $SINGLEMDS
134 checkstat $MOUNT2/adir || return 1
136 checkstat $MOUNT2/adir && return 2
139 run_test 2 "|X| mkdir adir"
142 replay_barrier $SINGLEMDS
144 mkdir $MOUNT2/adir/bdir
147 checkstat $MOUNT2/adir || return 1
148 checkstat $MOUNT1/adir/bdir || return 2
149 rmdir $MOUNT2/adir/bdir $MOUNT1/adir
150 checkstat $MOUNT1/adir && return 3
151 checkstat $MOUNT2/adir/bdir && return 4
154 run_test 3 "|X| mkdir adir, mkdir adir/bdir "
158 replay_barrier $SINGLEMDS
159 mkdir $MOUNT1/adir && return 1
160 mkdir $MOUNT2/adir/bdir
163 checkstat $MOUNT2/adir || return 2
164 checkstat $MOUNT1/adir/bdir || return 3
166 rmdir $MOUNT2/adir/bdir $MOUNT1/adir
167 checkstat $MOUNT1/adir && return 4
168 checkstat $MOUNT2/adir/bdir && return 5
171 run_test 4 "|X| mkdir adir (-EEXIST), mkdir adir/bdir "
175 # multiclient version of replay_single.sh/test_8
177 multiop_bg_pause $MOUNT2/a o_tSc || return 1
180 replay_barrier $SINGLEMDS
182 wait $pid || return 1
185 [ -e $MOUNT2/a ] && return 2
188 run_test 5 "open, unlink |X| close"
193 multiop_bg_pause $MOUNT2/a o_c || return 1
195 multiop_bg_pause $MOUNT1/a o_c || return 1
198 replay_barrier $SINGLEMDS
200 wait $pid1 || return 1
204 wait $pid2 || return 1
205 [ -e $MOUNT2/a ] && return 2
208 run_test 6 "open1, open2, unlink |X| close1 [fail $SINGLEMDS] close2"
211 replay_barrier $SINGLEMDS
212 drop_reint_reply "mcreate $MOUNT1/$tfile" || return 1
214 checkstat $MOUNT2/$tfile || return 2
215 rm $MOUNT1/$tfile || return 3
219 run_test 8 "replay of resent request"
222 replay_barrier $SINGLEMDS
223 mcreate $MOUNT1/$tfile-1
224 mcreate $MOUNT2/$tfile-2
225 # drop first reint reply
226 do_facet $SINGLEMDS lctl set_param fail_loc=0x80000119
228 do_facet $SINGLEMDS lctl set_param fail_loc=0
230 rm $MOUNT1/$tfile-[1,2] || return 1
234 run_test 9 "resending a replayed create"
237 mcreate $MOUNT1/$tfile-1
238 replay_barrier $SINGLEMDS
239 munlink $MOUNT1/$tfile-1
240 mcreate $MOUNT2/$tfile-2
241 # drop first reint reply
242 do_facet $SINGLEMDS lctl set_param fail_loc=0x80000119
244 do_facet $SINGLEMDS lctl set_param fail_loc=0
246 checkstat $MOUNT1/$tfile-1 && return 1
247 checkstat $MOUNT1/$tfile-2 || return 2
252 run_test 10 "resending a replayed unlink"
255 replay_barrier $SINGLEMDS
256 mcreate $DIR1/$tfile-1
257 mcreate $DIR2/$tfile-2
258 mcreate $DIR1/$tfile-3
259 mcreate $DIR2/$tfile-4
260 mcreate $DIR1/$tfile-5
261 # drop all reint replies for a while
262 do_facet $SINGLEMDS $LCTL set_param fail_loc=0x0119
263 # note that with this fail_loc set, facet_failover df will fail
264 facet_failover $SINGLEMDS
266 local clients=${CLIENTS:-$HOSTNAME}
267 wait_clients_import_state "$clients" $SINGLEMDS FULL
269 do_facet $SINGLEMDS $LCTL set_param fail_loc=0
271 rm $DIR1/$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 ldlm.namespaces.*.early_lock_cancel=0
449 do_facet client lctl set_param fail_loc=0x80000305 # drop cb, evict
451 usleep 500 # wait to ensure first client is one that will be evicted
452 openfile -f O_RDONLY $MOUNT2/$tdir/$tfile
454 do_facet client lctl set_param ldlm.namespaces.*.early_lock_cancel=1
455 do_facet $SINGLEMDS lctl debug_kernel |
456 grep "not entering recovery" && error "client not evicted"
457 do_facet client "lctl set_param fail_loc=0"
458 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
460 run_test 18 "ldlm_handle_enqueue succeeds on evicted export (3822)"
462 test_19() { # Bug 10991 - resend of open request does not fail assertion.
463 replay_barrier $SINGLEMDS
464 drop_ldlm_reply "createmany -o $DIR/$tfile 1" || return 1
466 checkstat $DIR2/${tfile}0 || return 2
467 rm $DIR/${tfile}0 || return 3
471 run_test 19 "resend of open request"
474 local before=$SECONDS
475 replay_barrier $SINGLEMDS
481 zconf_mount $HOSTNAME $DIR2 || error "mount $DIR2 fail"
482 local tier1=$((SECONDS - before))
485 replay_barrier $SINGLEMDS
491 zconf_mount $HOSTNAME $DIR2 || error "mount $DIR2 fail"
492 local tier2=$((SECONDS - before))
494 # timeout is more than 1.5x original timeout
495 ((tier2 < tier1 * 6 / 4)) ||
496 error "recovery time $tier2 >= 1.5x original time $tier1"
498 run_test 20 "recovery time is not increasing"
500 # commit on sharing tests
502 local param_file=$TMP/$tfile-params
504 save_lustre_params $SINGLEMDS "mdt.*.commit_on_sharing" > $param_file
505 do_facet $SINGLEMDS lctl set_param mdt.*.commit_on_sharing=1
506 touch $MOUNT1/$tfile-1
507 mv $MOUNT2/$tfile-1 $MOUNT2/$tfile-2
508 mv $MOUNT1/$tfile-2 $MOUNT1/$tfile-3
509 replay_barrier_nosync $SINGLEMDS
512 facet_failover $SINGLEMDS
514 # all renames are replayed
515 unlink $MOUNT1/$tfile-3 || return 2
517 zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
519 do_facet $SINGLEMDS lctl set_param mdt.*.commit_on_sharing=0
520 rm -rf $MOUNT1/$tfile-*
521 restore_lustre_params < $param_file
525 run_test 21a "commit on sharing"
529 do_node $CLIENT1 rm -f $MOUNT1/$tfile-*
532 do_node $CLIENT1 touch $MOUNT1/$tfile-1
533 do_node $CLIENT2 mv $MOUNT1/$tfile-1 $MOUNT1/$tfile-2
534 do_node $CLIENT1 mv $MOUNT1/$tfile-2 $MOUNT1/$tfile-3
536 replay_barrier_nosync $mds
537 shutdown_client $CLIENT2 $MOUNT1
541 # were renames replayed?
543 echo UNLINK $MOUNT1/$tfile-3
544 do_node $CLIENT1 unlink $MOUNT1/$tfile-3 ||
545 { echo "unlink $tfile-3 fail!" && rc=1; }
548 zconf_mount_clients $CLIENT2 $MOUNT1 ||
549 error "mount $CLIENT2 $MOUNT1 fail"
555 [ -z "$CLIENTS" ] && skip "Need two or more clients" && return
556 [ $CLIENTCOUNT -lt 2 ] &&
557 { skip "Need 2+ clients, have $CLIENTCOUNT" && return; }
559 if [ "$FAILURE_MODE" = "HARD" ] && mixed_mdt_devs; then
560 skip "Several MDTs on one MDS with FAILURE_MODE=$FAILURE_MODE"
564 zconf_umount_clients $CLIENTS $MOUNT2
565 zconf_mount_clients $CLIENTS $MOUNT1
567 local param_file=$TMP/$tfile-params
569 local mdtidx=$($LFS getstripe -M $MOUNT1)
570 local facet=mds$((mdtidx + 1))
572 save_lustre_params $facet "mdt.*.commit_on_sharing" > $param_file
576 do_facet $facet lctl set_param mdt.*.commit_on_sharing=$COS
578 test_21b_sub $facet || error "Not all renames are replayed. COS=$COS"
580 # COS disabled (should fail)
582 do_facet $facet lctl set_param mdt.*.commit_on_sharing=$COS
584 # there is still a window when transactions may be written to disk
585 # before the mds device is set R/O. To avoid such a rare test failure,
586 # the check is repeated several times.
589 test_21b_sub $facet || break
590 n_attempts=$((n_attempts + 1))
591 [ $n_attempts -gt 3 ] &&
592 error "can't check if COS works: rename replied w/o COS"
594 zconf_mount_clients $CLIENTS $MOUNT2
595 restore_lustre_params < $param_file
599 run_test 21b "commit on sharing, two clients"
602 checkstat $MOUNT1/$remote_dir || return 1
603 checkstat $MOUNT1/$remote_dir/dir || return 2
604 checkstat $MOUNT1/$remote_dir/$tfile-1 || return 3
605 checkstat $MOUNT1/$remote_dir/dir/$tfile-1 || return 4
609 create_remote_dir_files_22() {
610 do_node $CLIENT2 mkdir ${MOUNT2}/$remote_dir/dir || return 1
611 do_node $CLIENT1 createmany -o $MOUNT1/$remote_dir/dir/$tfile- 2 ||
613 do_node $CLIENT2 createmany -o $MOUNT2/$remote_dir/$tfile- 2 ||
619 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
620 ([ $FAILURE_MODE == "HARD" ] &&
621 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
622 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
626 local remote_dir=${tdir}/remote_dir
628 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
630 # OBD_FAIL_MDS_REINT_NET_REP 0x119
631 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
632 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir &
635 fail mds$((MDTIDX + 1))
636 wait $CLIENT_PID || error "lfs mkdir failed"
638 replay_barrier mds$MDTIDX
639 create_remote_dir_files_22 || error "Remote creation failed $?"
642 checkstat_22 || error "check stat failed $?"
644 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
647 run_test 22a "c1 lfs mkdir -i 1 dir1, M1 drop reply & fail, c2 mkdir dir1/dir"
650 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
652 local remote_dir=$tdir/remote_dir
654 # OBD_FAIL_MDS_REINT_NET_REP 0x119
655 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
657 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
658 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir &
661 fail mds${MDTIDX},mds$((MDTIDX + 1))
662 wait $CLIENT_PID || error "lfs mkdir failed"
664 replay_barrier mds$MDTIDX
665 create_remote_dir_files_22 || error "Remote creation failed $?"
668 checkstat_22 || error "check stat failed $?"
670 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
673 run_test 22b "c1 lfs mkdir -i 1 d1, M1 drop reply & fail M0/M1, c2 mkdir d1/dir"
676 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
677 ([ $FAILURE_MODE == "HARD" ] &&
678 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
679 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
682 local remote_dir=${tdir}/remote_dir
684 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
686 # OBD_FAIL_UPDATE_OBJ_NET_REP 0x1701
687 do_facet mds$MDTIDX lctl set_param fail_loc=0x1701
688 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir &
690 do_facet mds$MDTIDX lctl set_param fail_loc=0
693 wait $CLIENT_PID || error "lfs mkdir failed"
695 replay_barrier mds$MDTIDX
696 create_remote_dir_files_22 || error "Remote creation failed $?"
699 checkstat_22 || error "check stat failed $?"
701 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
704 run_test 22c "c1 lfs mkdir -i 1 d1, M1 drop update & fail M1, c2 mkdir d1/dir"
707 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
709 local remote_dir=${tdir}/remote_dir
711 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
713 # OBD_FAIL_UPDATE_OBJ_NET_REP 0x1701
714 do_facet mds$MDTIDX lctl set_param fail_loc=0x1701
715 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir &
717 do_facet mds$MDTIDX lctl set_param fail_loc=0
719 fail mds${MDTIDX},mds$((MDTIDX + 1))
720 wait $CLIENT_PID || error "lfs mkdir failed"
722 replay_barrier mds$MDTIDX
723 create_remote_dir_files_22 || error "Remote creation failed $?"
726 checkstat_22 || error "check stat failed $?"
728 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
731 run_test 22d "c1 lfs mkdir -i 1 d1, M1 drop update & fail M0/M1,c2 mkdir d1/dir"
734 checkstat $MOUNT1/$remote_dir || return 1
735 checkstat $MOUNT1/$remote_dir/$tfile-1 || return 2
739 create_remote_dir_files_23() {
740 do_node $CLIENT2 mkdir ${MOUNT2}/$remote_dir || return 1
741 do_node $CLIENT2 createmany -o $MOUNT2/$remote_dir/$tfile- 2 || return 2
746 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
747 ([ $FAILURE_MODE == "HARD" ] &&
748 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
749 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
752 local remote_dir=$tdir/remote_dir
754 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
755 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir ||
756 error "lfs mkdir failed"
757 # OBD_FAIL_MDS_REINT_NET_REP 0x119
758 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
759 do_node $CLIENT1 rmdir $MOUNT1/$remote_dir &
761 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0
763 fail mds$((MDTIDX + 1))
764 wait $CLIENT_PID || error "rmdir remote dir failed"
766 replay_barrier mds${MDTIDX}
767 create_remote_dir_files_23 || error "Remote creation failed $?"
770 checkstat_23 || error "check stat failed $?"
772 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
775 run_test 23a "c1 rmdir d1, M1 drop reply and fail, client2 mkdir d1"
778 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
780 local remote_dir=$tdir/remote_dir
782 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
783 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir ||
784 error "lfs mkdir failed"
786 # OBD_FAIL_MDS_REINT_NET_REP 0x119
787 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
788 do_node $CLIENT1 rmdir $MOUNT1/$remote_dir &
790 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0
792 fail mds${MDTIDX},mds$((MDTIDX + 1))
793 wait $CLIENT_PID || error "rmdir remote dir failed"
795 replay_barrier mds${MDTIDX}
796 create_remote_dir_files_23 || error "Remote creation failed $?"
799 checkstat_23 || error "check stat failed $?"
801 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
804 run_test 23b "c1 rmdir d1, M1 drop reply and fail M0/M1, c2 mkdir d1"
807 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
809 ([ $FAILURE_MODE == "HARD" ] &&
810 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
811 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
814 local remote_dir=$tdir/remote_dir
816 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
817 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir ||
818 error "lfs mkdir failed"
820 # OBD_FAIL_UPDATE_OBJ_NET_REP 0x1701
821 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
822 do_node $CLIENT1 rmdir $MOUNT1/$remote_dir &
824 do_facet mds${MDTIDX} lctl set_param fail_loc=0
827 wait $CLIENT_PID || error "rmdir remote dir failed"
829 replay_barrier mds${MDTIDX}
830 create_remote_dir_files_23 || error "Remote creation failed $?"
833 checkstat_23 || error "check stat failed $?"
835 rm -rf $MOUNT1/$tdir || return 6
838 run_test 23c "c1 rmdir d1, M0 drop update reply and fail M0, c2 mkdir d1"
841 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
843 local remote_dir=$tdir/remote_dir
845 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
846 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir ||
847 error "lfs mkdir failed"
849 # OBD_FAIL_UPDATE_OBJ_NET 0x1701
850 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
851 do_node $CLIENT1 rmdir $MOUNT1/$remote_dir &
853 do_facet mds${MDTIDX} lctl set_param fail_loc=0
855 fail mds${MDTIDX},mds$((MDTIDX + 1))
856 wait $CLIENT_PID || error "rmdir remote dir failed"
858 replay_barrier mds${MDTIDX}
859 create_remote_dir_files_23 || error "Remote creation failed $?"
862 checkstat_23 || error "check stat failed $?"
864 rm -rf $MOUNT1/$tdir || return 6
867 run_test 23d "c1 rmdir d1, M0 drop update reply and fail M0/M1, c2 mkdir d1"
870 [[ $(lustre_version_code $SINGLEMDS) -gt $(version_code 2.5.2) ]] ||
871 { skip "Need MDS version newer than 2.5.2"; return 0; }
874 stat $MOUNT/$tfile >&/dev/null
875 # OBD_FAIL_MDS_REINT_NET_REP
876 do_facet $SINGLEMDS $LCTL set_param fail_loc=0x119
877 $TRUNCATE $MOUNT/$tfile 100 &
880 do_facet $SINGLEMDS lctl set_param fail_loc=0
881 # sync to release rep-ack lock quickly
882 do_nodes $(comma_list $(mdts_nodes)) \
883 "lctl set_param -n osd*.*MDT*.force_sync 1"
887 run_test 24 "reconstruct on non-existing object"
889 # end commit on sharing tests
894 $SETSTRIPE -i 0 -c 1 $DIR/$tfile
896 # get lock for the 1st client
897 dd if=/dev/zero of=$DIR/$tfile count=1 >/dev/null ||
898 error "failed to write data"
900 # get waiting locks for the 2nd client
901 drop_ldlm_cancel "multiop $DIR2/$tfile Ow512" &
904 # failover, replay and resend replayed waiting locks
905 if [ $(lustre_version_code ost1) -ge $(version_code 2.6.90) ]; then
906 #define OBD_FAIL_LDLM_SRV_CP_AST 0x325
907 do_facet ost1 lctl set_param fail_loc=0x80000325
909 #define OBD_FAIL_OST_LDLM_REPLY_NET 0x213
910 do_facet ost1 lctl set_param fail_loc=0x80000213
915 # multiop does not finish because CP AST is skipped;
916 # it is ok to kill it in the test, because CP AST is already re-sent
917 # and it does not hung forever in real life
921 run_test 25 "replay|resend"
924 SLEEP=$((`date +%s` - $NOW))
925 [ $SLEEP -lt $TIMEOUT ] && sleep $SLEEP
926 [ "$MOUNTED2" = yes ] && zconf_umount $HOSTNAME $MOUNT2 || true
927 check_and_cleanup_lustre