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 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"
475 replay_barrier $SINGLEMDS
481 zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
482 TIER1=$((`date +%s` - BEFORE))
484 replay_barrier $SINGLEMDS
490 zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
491 TIER2=$((`date +%s` - BEFORE))
492 [ $TIER2 -ge $((TIER1 * 2)) ] && \
493 error "recovery time is growing $TIER2 > $TIER1"
496 run_test 20 "recovery time is not increasing"
498 # commit on sharing tests
500 local param_file=$TMP/$tfile-params
502 save_lustre_params $SINGLEMDS "mdt.*.commit_on_sharing" > $param_file
503 do_facet $SINGLEMDS lctl set_param mdt.*.commit_on_sharing=1
504 touch $MOUNT1/$tfile-1
505 mv $MOUNT2/$tfile-1 $MOUNT2/$tfile-2
506 mv $MOUNT1/$tfile-2 $MOUNT1/$tfile-3
507 replay_barrier_nosync $SINGLEMDS
510 facet_failover $SINGLEMDS
512 # all renames are replayed
513 unlink $MOUNT1/$tfile-3 || return 2
515 zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
517 do_facet $SINGLEMDS lctl set_param mdt.*.commit_on_sharing=0
518 rm -rf $MOUNT1/$tfile-*
519 restore_lustre_params < $param_file
523 run_test 21a "commit on sharing"
527 do_node $CLIENT1 rm -f $MOUNT1/$tfile-*
530 do_node $CLIENT1 touch $MOUNT1/$tfile-1
531 do_node $CLIENT2 mv $MOUNT1/$tfile-1 $MOUNT1/$tfile-2
532 do_node $CLIENT1 mv $MOUNT1/$tfile-2 $MOUNT1/$tfile-3
534 replay_barrier_nosync $mds
535 shutdown_client $CLIENT2 $MOUNT1
539 # were renames replayed?
541 echo UNLINK $MOUNT1/$tfile-3
542 do_node $CLIENT1 unlink $MOUNT1/$tfile-3 || \
543 { echo "unlink $tfile-3 fail!" && rc=1; }
546 zconf_mount_clients $CLIENT2 $MOUNT1 || error "mount $CLIENT2 $MOUNT1 fail"
552 [ -z "$CLIENTS" ] && skip "Need two or more clients." && return
553 [ $CLIENTCOUNT -lt 2 ] && \
554 { skip "Need two or more clients, have $CLIENTCOUNT" && return; }
556 if [ "$FAILURE_MODE" = "HARD" ] && mixed_mdt_devs; then
557 skip "Several mdt services on one mds node are used with FAILURE_MODE=$FAILURE_MODE. "
562 zconf_umount_clients $CLIENTS $MOUNT2
563 zconf_mount_clients $CLIENTS $MOUNT1
565 local param_file=$TMP/$tfile-params
567 local num=$(get_mds_dir $MOUNT1)
569 save_lustre_params mds$num "mdt.*.commit_on_sharing" > $param_file
573 do_facet mds$num lctl set_param mdt.*.commit_on_sharing=$COS
575 test_21b_sub mds$num || error "Not all renames are replayed. COS=$COS"
577 # COS disabled (should fail)
579 do_facet mds$num lctl set_param mdt.*.commit_on_sharing=$COS
581 # there is still a window when transactions may be written to disk before
582 # the mds device is set R/O. To avoid such a rare test failure, the check
583 # is repeated several times.
586 test_21b_sub mds$num || break;
587 let n_attempts=n_attempts+1
588 [ $n_attempts -gt 3 ] &&
589 error "The test cannot check whether COS works or not: all renames are replied w/o COS"
591 zconf_mount_clients $CLIENTS $MOUNT2
592 restore_lustre_params < $param_file
596 run_test 21b "commit on sharing, two clients"
599 checkstat $MOUNT1/$remote_dir || return 1
600 checkstat $MOUNT1/$remote_dir/dir || return 2
601 checkstat $MOUNT1/$remote_dir/$tfile-1 || return 3
602 checkstat $MOUNT1/$remote_dir/dir/$tfile-1 || return 4
606 create_remote_dir_files_22() {
607 do_node $CLIENT2 mkdir ${MOUNT2}/$remote_dir/dir || return 1
608 do_node $CLIENT1 createmany -o $MOUNT1/$remote_dir/dir/$tfile- 2 ||
610 do_node $CLIENT2 createmany -o $MOUNT2/$remote_dir/$tfile- 2 ||
616 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
617 ([ $FAILURE_MODE == "HARD" ] &&
618 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
619 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
623 local remote_dir=${tdir}/remote_dir
625 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
627 # OBD_FAIL_MDS_REINT_NET_REP 0x119
628 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
629 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir &
632 fail mds$((MDTIDX + 1))
633 wait $CLIENT_PID || error "lfs mkdir failed"
635 replay_barrier mds$MDTIDX
636 create_remote_dir_files_22 || error "Remote creation failed $?"
639 checkstat_22 || error "check stat failed $?"
641 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
644 run_test 22a "c1 lfs mkdir -i 1 dir1, M1 drop reply & fail, c2 mkdir dir1/dir"
647 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
649 local remote_dir=$tdir/remote_dir
651 # OBD_FAIL_MDS_REINT_NET_REP 0x119
652 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
654 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
655 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir &
658 fail mds${MDTIDX},mds$((MDTIDX + 1))
659 wait $CLIENT_PID || error "lfs mkdir failed"
661 replay_barrier mds$MDTIDX
662 create_remote_dir_files_22 || error "Remote creation failed $?"
665 checkstat_22 || error "check stat failed $?"
667 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
670 run_test 22b "c1 lfs mkdir -i 1 d1, M1 drop reply & fail M0/M1, c2 mkdir d1/dir"
673 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
674 ([ $FAILURE_MODE == "HARD" ] &&
675 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
676 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
679 local remote_dir=${tdir}/remote_dir
681 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
683 # OBD_FAIL_UPDATE_OBJ_NET_REP 0x1701
684 do_facet mds$MDTIDX lctl set_param fail_loc=0x1701
685 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir &
687 do_facet mds$MDTIDX lctl set_param fail_loc=0
690 wait $CLIENT_PID || error "lfs mkdir failed"
692 replay_barrier mds$MDTIDX
693 create_remote_dir_files_22 || error "Remote creation failed $?"
696 checkstat_22 || error "check stat failed $?"
698 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
701 run_test 22c "c1 lfs mkdir -i 1 d1, M1 drop update & fail M1, c2 mkdir d1/dir"
704 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
706 local remote_dir=${tdir}/remote_dir
708 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
710 # OBD_FAIL_UPDATE_OBJ_NET_REP 0x1701
711 do_facet mds$MDTIDX lctl set_param fail_loc=0x1701
712 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir &
714 do_facet mds$MDTIDX lctl set_param fail_loc=0
716 fail mds${MDTIDX},mds$((MDTIDX + 1))
717 wait $CLIENT_PID || error "lfs mkdir failed"
719 replay_barrier mds$MDTIDX
720 create_remote_dir_files_22 || error "Remote creation failed $?"
723 checkstat_22 || error "check stat failed $?"
725 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
728 run_test 22d "c1 lfs mkdir -i 1 d1, M1 drop update & fail M0/M1,c2 mkdir d1/dir"
731 checkstat $MOUNT1/$remote_dir || return 1
732 checkstat $MOUNT1/$remote_dir/$tfile-1 || return 2
736 create_remote_dir_files_23() {
737 do_node $CLIENT2 mkdir ${MOUNT2}/$remote_dir || return 1
738 do_node $CLIENT2 createmany -o $MOUNT2/$remote_dir/$tfile- 2 || return 2
743 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
744 ([ $FAILURE_MODE == "HARD" ] &&
745 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
746 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
749 local remote_dir=$tdir/remote_dir
751 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
752 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir ||
753 error "lfs mkdir failed"
754 # OBD_FAIL_MDS_REINT_NET_REP 0x119
755 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
756 do_node $CLIENT1 rmdir $MOUNT1/$remote_dir &
758 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0
760 fail mds$((MDTIDX + 1))
761 wait $CLIENT_PID || error "rmdir remote dir failed"
763 replay_barrier mds${MDTIDX}
764 create_remote_dir_files_23 || error "Remote creation failed $?"
767 checkstat_23 || error "check stat failed $?"
769 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
772 run_test 23a "c1 rmdir d1, M1 drop reply and fail, client2 mkdir d1"
775 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
777 local remote_dir=$tdir/remote_dir
779 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
780 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir ||
781 error "lfs mkdir failed"
783 # OBD_FAIL_MDS_REINT_NET_REP 0x119
784 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
785 do_node $CLIENT1 rmdir $MOUNT1/$remote_dir &
787 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0
789 fail mds${MDTIDX},mds$((MDTIDX + 1))
790 wait $CLIENT_PID || error "rmdir remote dir failed"
792 replay_barrier mds${MDTIDX}
793 create_remote_dir_files_23 || error "Remote creation failed $?"
796 checkstat_23 || error "check stat failed $?"
798 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
801 run_test 23b "c1 rmdir d1, M1 drop reply and fail M0/M1, c2 mkdir d1"
804 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
806 ([ $FAILURE_MODE == "HARD" ] &&
807 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
808 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
811 local remote_dir=$tdir/remote_dir
813 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
814 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir ||
815 error "lfs mkdir failed"
817 # OBD_FAIL_UPDATE_OBJ_NET_REP 0x1701
818 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
819 do_node $CLIENT1 rmdir $MOUNT1/$remote_dir &
821 do_facet mds${MDTIDX} lctl set_param fail_loc=0
824 wait $CLIENT_PID || error "rmdir remote dir failed"
826 replay_barrier mds${MDTIDX}
827 create_remote_dir_files_23 || error "Remote creation failed $?"
830 checkstat_23 || error "check stat failed $?"
832 rm -rf $MOUNT1/$tdir || return 6
835 run_test 23c "c1 rmdir d1, M0 drop update reply and fail M0, c2 mkdir d1"
838 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
840 local remote_dir=$tdir/remote_dir
842 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
843 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir ||
844 error "lfs mkdir failed"
846 # OBD_FAIL_UPDATE_OBJ_NET 0x1701
847 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
848 do_node $CLIENT1 rmdir $MOUNT1/$remote_dir &
850 do_facet mds${MDTIDX} lctl set_param fail_loc=0
852 fail mds${MDTIDX},mds$((MDTIDX + 1))
853 wait $CLIENT_PID || error "rmdir remote dir failed"
855 replay_barrier mds${MDTIDX}
856 create_remote_dir_files_23 || error "Remote creation failed $?"
859 checkstat_23 || error "check stat failed $?"
861 rm -rf $MOUNT1/$tdir || return 6
864 run_test 23d "c1 rmdir d1, M0 drop update reply and fail M0/M1, c2 mkdir d1"
867 [[ $(lustre_version_code $SINGLEMDS) -gt $(version_code 2.5.2) ]] ||
868 { skip "Need MDS version newer than 2.5.2"; return 0; }
871 stat $MOUNT/$tfile >&/dev/null
872 # OBD_FAIL_MDS_REINT_NET_REP
873 do_facet $SINGLEMDS $LCTL set_param fail_loc=0x119
874 $TRUNCATE $MOUNT/$tfile 100 &
877 do_facet $SINGLEMDS lctl set_param fail_loc=0
878 # sync to release rep-ack lock quickly
879 do_nodes $(comma_list $(mdts_nodes)) \
880 "lctl set_param -n osd*.*MDT*.force_sync 1"
884 run_test 24 "reconstruct on non-existing object"
886 # end commit on sharing tests
891 $SETSTRIPE -i 0 -c 1 $DIR/$tfile
893 # get lock for the 1st client
894 dd if=/dev/zero of=$DIR/$tfile count=1 >/dev/null ||
895 error "failed to write data"
897 # get waiting locks for the 2nd client
898 drop_ldlm_cancel "multiop $DIR2/$tfile Ow512" &
901 #define OBD_FAIL_OST_LDLM_REPLY_NET 0x213
902 # failover, replay and resend replayed waiting locks
903 do_facet ost1 lctl set_param fail_loc=0x80000213
906 # multiop does not finish because CP AST is skipped;
907 # it is ok to kill it in the test, because CP AST is already re-sent
908 # and it does not hung forever in real life
912 run_test 25 "replay|resend"
915 SLEEP=$((`date +%s` - $NOW))
916 [ $SLEEP -lt $TIMEOUT ] && sleep $SLEEP
917 [ "$MOUNTED2" = yes ] && zconf_umount $HOSTNAME $MOUNT2 || true
918 check_and_cleanup_lustre