5 # bug number: LU-2012 LU-8333
6 ALWAYS_EXCEPT="14b 21b $REPLAY_DUAL_EXCEPT"
7 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
10 PTLDEBUG=${PTLDEBUG:--1}
11 LUSTRE=${LUSTRE:-$(cd $(dirname $0)/..; echo $PWD)}
13 CLEANUP=${CLEANUP:-""}
14 MOUNT_2=${MOUNT_2:-"yes"}
15 export MULTIOP=${MULTIOP:-multiop}
16 . $LUSTRE/tests/test-framework.sh
19 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
22 remote_mds_nodsh && skip "remote MDS with nodsh" && exit 0
25 [ "$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 $DIR1/$tfile-1
259 mcreate $DIR2/$tfile-2
260 mcreate $DIR1/$tfile-3
261 mcreate $DIR2/$tfile-4
262 mcreate $DIR1/$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 $SINGLEMDS
268 local clients=${CLIENTS:-$HOSTNAME}
269 wait_clients_import_state "$clients" $SINGLEMDS FULL
271 do_facet $SINGLEMDS $LCTL set_param fail_loc=0
273 rm $DIR1/$tfile-[1-5] || return 1
277 run_test 11 "both clients timeout during replay"
280 replay_barrier $SINGLEMDS
282 multiop_bg_pause $DIR/$tfile mo_c || return 1
285 #define OBD_FAIL_LDLM_ENQUEUE_NET 0x302
286 do_facet $SINGLEMDS lctl set_param fail_loc=0x80000302
287 facet_failover $SINGLEMDS
288 do_facet $SINGLEMDS lctl set_param fail_loc=0
289 clients_up || return 1
292 kill -USR1 $MULTIPID || return 3
293 wait $MULTIPID || return 4
294 $CHECKSTAT -t file $DIR/$tfile || return 2
299 run_test 12 "open resend timeout"
302 multiop_bg_pause $DIR/$tfile mo_c || return 1
305 replay_barrier $SINGLEMDS
307 kill -USR1 $MULTIPID || return 3
308 wait $MULTIPID || return 4
311 do_facet $SINGLEMDS lctl set_param fail_loc=0x80000115
312 facet_failover $SINGLEMDS
313 do_facet $SINGLEMDS lctl set_param fail_loc=0
314 clients_up || return 1
317 $CHECKSTAT -t file $DIR/$tfile || return 2
322 run_test 13 "close resend timeout"
324 # test 14a removed after 18143 because it shouldn't fail anymore and do the same
329 wait_delete_completed
331 local beforeused=$(df -P $DIR | tail -1 | awk '{ print $3 }')
333 mkdir -p $MOUNT1/$tdir
334 $SETSTRIPE -i 0 $MOUNT1/$tdir
335 replay_barrier $SINGLEMDS
336 createmany -o $MOUNT1/$tdir/$tfile- 5
338 $SETSTRIPE -i 0 $MOUNT2/$tfile-2
339 dd if=/dev/zero of=$MOUNT2/$tfile-2 bs=1M count=5
340 createmany -o $MOUNT1/$tdir/$tfile-3- 5
344 wait_recovery_complete $SINGLEMDS || error "MDS recovery not done"
346 # first set of files should have been replayed
347 unlinkmany $MOUNT1/$tdir/$tfile- 5 || error "first unlinks failed"
348 unlinkmany $MOUNT1/$tdir/$tfile-3- 5 || error "second unlinks failed"
350 zconf_mount $HOSTNAME $MOUNT2 || error "mount $MOUNT2 failed"
351 [ -f $MOUNT2/$tfile-2 ] && error "$MOUNT2/$tfile-2 exists!"
353 wait_mds_ost_sync || error "wait_mds_ost_sync failed"
354 wait_delete_completed || error "wait_delete_complete failed"
356 local afterused=$(df -P $DIR | tail -1 | awk '{ print $3 }')
357 log "before $beforeused, after $afterused"
358 # leave some margin for some files/dirs to be modified (OI, llog, etc)
359 [ $afterused -le $((beforeused + $(fs_log_size))) ] ||
360 error "after $afterused > before $beforeused"
362 run_test 14b "delete ost orphans if gap occured in objids due to VBR"
364 test_15a() { # was test_15
365 replay_barrier $SINGLEMDS
366 createmany -o $MOUNT1/$tfile- 25
367 createmany -o $MOUNT2/$tfile-2- 1
372 unlinkmany $MOUNT1/$tfile- 25 || return 2
373 [ -e $MOUNT1/$tfile-2-0 ] && error "$tfile-2-0 exists"
375 zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
378 run_test 15a "timeout waiting for lost client during replay, 1 client completes"
381 replay_barrier $SINGLEMDS
382 for ((i = 0; i < 2000; i++)); do
383 echo "data" > "$MOUNT2/${tfile}-$i" || error "create ${tfile}-$i failed"
389 zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
392 run_test 15c "remove multiple OST orphans"
395 replay_barrier $SINGLEMDS
396 createmany -o $MOUNT1/$tfile- 25
397 createmany -o $MOUNT2/$tfile-2- 1
400 facet_failover $SINGLEMDS
404 unlinkmany $MOUNT1/$tfile- 25 || return 2
406 zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
410 run_test 16 "fail MDS during recovery (3571)"
413 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
415 createmany -o $MOUNT1/$tfile- 25
416 createmany -o $MOUNT2/$tfile-2- 1
418 # Make sure the disconnect is lost
426 unlinkmany $MOUNT1/$tfile- 25 || return 2
428 zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
432 run_test 17 "fail OST during recovery (3571)"
434 # cleanup with blocked enqueue fails until timer elapses (MDS busy), wait for it
437 test_18() { # bug 3822 - evicting client with enqueued lock
439 local DLMTRACE=$(do_facet $SINGLEMDS lctl get_param debug)
440 do_facet $SINGLEMDS lctl set_param debug=+dlmtrace
441 mkdir -p $MOUNT1/$tdir || error "mkdir $MOUNT1/$tdir failed"
442 touch $MOUNT1/$tdir/${tfile}0 || error "touch file failed"
443 statmany -s $MOUNT1/$tdir/$tfile 1 500 &
446 #define OBD_FAIL_LDLM_ENQUEUE_BLOCKED 0x30b
447 do_facet $SINGLEMDS lctl set_param fail_loc=0x8000030b # hold enqueue
449 #define OBD_FAIL_LDLM_BL_CALLBACK_NET 0x305
450 do_facet client lctl set_param ldlm.namespaces.*.early_lock_cancel=0
451 do_facet client lctl set_param fail_loc=0x80000305 # drop cb, evict
453 usleep 500 # wait to ensure first client is one that will be evicted
454 openfile -f O_RDONLY $MOUNT2/$tdir/$tfile
456 do_facet client lctl set_param ldlm.namespaces.*.early_lock_cancel=1
457 do_facet $SINGLEMDS lctl debug_kernel |
458 grep "not entering recovery" && error "client not evicted"
459 do_facet client "lctl set_param fail_loc=0"
460 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
462 run_test 18 "ldlm_handle_enqueue succeeds on evicted export (3822)"
464 test_19() { # Bug 10991 - resend of open request does not fail assertion.
465 replay_barrier $SINGLEMDS
466 drop_ldlm_reply "createmany -o $DIR/$tfile 1" || return 1
468 checkstat $DIR2/${tfile}0 || return 2
469 rm $DIR/${tfile}0 || return 3
473 run_test 19 "resend of open request"
476 local before=$SECONDS
477 replay_barrier $SINGLEMDS
483 zconf_mount $HOSTNAME $DIR2 || error "mount $DIR2 fail"
484 local tier1=$((SECONDS - before))
487 replay_barrier $SINGLEMDS
493 zconf_mount $HOSTNAME $DIR2 || error "mount $DIR2 fail"
494 local tier2=$((SECONDS - before))
496 # timeout is more than 1.5x original timeout
497 ((tier2 < tier1 * 6 / 4)) ||
498 error "recovery time $tier2 >= 1.5x original time $tier1"
500 run_test 20 "recovery time is not increasing"
502 # commit on sharing tests
504 local param_file=$TMP/$tfile-params
506 save_lustre_params $SINGLEMDS "mdt.*.commit_on_sharing" > $param_file
507 do_facet $SINGLEMDS lctl set_param mdt.*.commit_on_sharing=1
508 touch $MOUNT1/$tfile-1
509 mv $MOUNT2/$tfile-1 $MOUNT2/$tfile-2
510 mv $MOUNT1/$tfile-2 $MOUNT1/$tfile-3
511 replay_barrier_nosync $SINGLEMDS
514 facet_failover $SINGLEMDS
516 # all renames are replayed
517 unlink $MOUNT1/$tfile-3 || return 2
519 zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
521 do_facet $SINGLEMDS lctl set_param mdt.*.commit_on_sharing=0
522 rm -rf $MOUNT1/$tfile-*
523 restore_lustre_params < $param_file
527 run_test 21a "commit on sharing"
531 do_node $CLIENT1 rm -f $MOUNT1/$tfile-*
534 do_node $CLIENT1 touch $MOUNT1/$tfile-1
535 do_node $CLIENT2 mv $MOUNT1/$tfile-1 $MOUNT1/$tfile-2
536 do_node $CLIENT1 mv $MOUNT1/$tfile-2 $MOUNT1/$tfile-3
538 replay_barrier_nosync $mds
539 shutdown_client $CLIENT2 $MOUNT1
543 # were renames replayed?
545 echo UNLINK $MOUNT1/$tfile-3
546 do_node $CLIENT1 unlink $MOUNT1/$tfile-3 ||
547 { echo "unlink $tfile-3 fail!" && rc=1; }
550 zconf_mount_clients $CLIENT2 $MOUNT1 ||
551 error "mount $CLIENT2 $MOUNT1 fail"
557 [ -z "$CLIENTS" ] && skip "Need two or more clients" && return
558 [ $CLIENTCOUNT -lt 2 ] &&
559 { skip "Need 2+ clients, have $CLIENTCOUNT" && return; }
561 if [ "$FAILURE_MODE" = "HARD" ] && mixed_mdt_devs; then
562 skip "Several MDTs on one MDS with FAILURE_MODE=$FAILURE_MODE"
566 zconf_umount_clients $CLIENTS $MOUNT2
567 zconf_mount_clients $CLIENTS $MOUNT1
569 local param_file=$TMP/$tfile-params
571 local mdtidx=$($LFS getstripe -m $MOUNT1)
572 local facet=mds$((mdtidx + 1))
574 save_lustre_params $facet "mdt.*.commit_on_sharing" > $param_file
578 do_facet $facet lctl set_param mdt.*.commit_on_sharing=$COS
580 test_21b_sub $facet || error "Not all renames are replayed. COS=$COS"
582 # there is still a window when transactions may be written to disk
583 # before the mds device is set R/O. To avoid such a rare test failure,
584 # the check is repeated several times.
588 # COS disabled (should fail)
589 do_facet $facet lctl set_param mdt.*.commit_on_sharing=$COS
591 test_21b_sub $facet || break
592 n_attempts=$((n_attempts + 1))
593 [ $n_attempts -gt 3 ] &&
594 error "can't check if COS works: rename replied w/o COS"
596 zconf_mount_clients $CLIENTS $MOUNT2
597 restore_lustre_params < $param_file
601 run_test 21b "commit on sharing, two clients"
604 checkstat $MOUNT1/$remote_dir || return 1
605 checkstat $MOUNT1/$remote_dir/dir || return 2
606 checkstat $MOUNT1/$remote_dir/$tfile-1 || return 3
607 checkstat $MOUNT1/$remote_dir/dir/$tfile-1 || return 4
611 create_remote_dir_files_22() {
612 do_node $CLIENT2 mkdir ${MOUNT2}/$remote_dir/dir || return 1
613 do_node $CLIENT1 createmany -o $MOUNT1/$remote_dir/dir/$tfile- 2 ||
615 do_node $CLIENT2 createmany -o $MOUNT2/$remote_dir/$tfile- 2 ||
621 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
622 ([ $FAILURE_MODE == "HARD" ] &&
623 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
624 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
628 local remote_dir=${tdir}/remote_dir
630 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
632 # OBD_FAIL_MDS_REINT_NET_REP 0x119
633 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
634 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir &
637 fail mds$((MDTIDX + 1))
638 wait $CLIENT_PID || error "lfs mkdir failed"
640 replay_barrier mds$MDTIDX
641 create_remote_dir_files_22 || error "Remote creation failed $?"
644 checkstat_22 || error "check stat failed $?"
646 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
649 run_test 22a "c1 lfs mkdir -i 1 dir1, M1 drop reply & fail, c2 mkdir dir1/dir"
652 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
654 local remote_dir=$tdir/remote_dir
656 # OBD_FAIL_MDS_REINT_NET_REP 0x119
657 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
659 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
660 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir &
663 fail mds${MDTIDX},mds$((MDTIDX + 1))
664 wait $CLIENT_PID || error "lfs mkdir failed"
666 replay_barrier mds$MDTIDX
667 create_remote_dir_files_22 || error "Remote creation failed $?"
670 checkstat_22 || error "check stat failed $?"
672 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
675 run_test 22b "c1 lfs mkdir -i 1 d1, M1 drop reply & fail M0/M1, c2 mkdir d1/dir"
678 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
679 ([ $FAILURE_MODE == "HARD" ] &&
680 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
681 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
684 local remote_dir=${tdir}/remote_dir
686 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
688 # OBD_FAIL_UPDATE_OBJ_NET_REP 0x1701
689 do_facet mds$MDTIDX lctl set_param fail_loc=0x1701
690 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir &
692 do_facet mds$MDTIDX lctl set_param fail_loc=0
695 wait $CLIENT_PID || error "lfs mkdir failed"
697 replay_barrier mds$MDTIDX
698 create_remote_dir_files_22 || error "Remote creation failed $?"
701 checkstat_22 || error "check stat failed $?"
703 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
706 run_test 22c "c1 lfs mkdir -i 1 d1, M1 drop update & fail M1, c2 mkdir d1/dir"
709 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
711 local remote_dir=${tdir}/remote_dir
713 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
715 # OBD_FAIL_UPDATE_OBJ_NET_REP 0x1701
716 do_facet mds$MDTIDX lctl set_param fail_loc=0x1701
717 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir &
719 do_facet mds$MDTIDX lctl set_param fail_loc=0
721 fail mds${MDTIDX},mds$((MDTIDX + 1))
722 wait $CLIENT_PID || error "lfs mkdir failed"
724 replay_barrier mds$MDTIDX
725 create_remote_dir_files_22 || error "Remote creation failed $?"
728 checkstat_22 || error "check stat failed $?"
730 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
733 run_test 22d "c1 lfs mkdir -i 1 d1, M1 drop update & fail M0/M1,c2 mkdir d1/dir"
736 checkstat $MOUNT1/$remote_dir || return 1
737 checkstat $MOUNT1/$remote_dir/$tfile-1 || return 2
741 create_remote_dir_files_23() {
742 do_node $CLIENT2 mkdir ${MOUNT2}/$remote_dir || return 1
743 do_node $CLIENT2 createmany -o $MOUNT2/$remote_dir/$tfile- 2 || return 2
748 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
749 ([ $FAILURE_MODE == "HARD" ] &&
750 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
751 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
754 local remote_dir=$tdir/remote_dir
756 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
757 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir ||
758 error "lfs mkdir failed"
759 # OBD_FAIL_MDS_REINT_NET_REP 0x119
760 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
761 do_node $CLIENT1 rmdir $MOUNT1/$remote_dir &
763 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0
765 fail mds$((MDTIDX + 1))
766 wait $CLIENT_PID || error "rmdir remote dir failed"
768 replay_barrier mds${MDTIDX}
769 create_remote_dir_files_23 || error "Remote creation failed $?"
772 checkstat_23 || error "check stat failed $?"
774 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
777 run_test 23a "c1 rmdir d1, M1 drop reply and fail, client2 mkdir d1"
780 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
782 local remote_dir=$tdir/remote_dir
784 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
785 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir ||
786 error "lfs mkdir failed"
788 # OBD_FAIL_MDS_REINT_NET_REP 0x119
789 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
790 do_node $CLIENT1 rmdir $MOUNT1/$remote_dir &
792 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0
794 fail mds${MDTIDX},mds$((MDTIDX + 1))
795 wait $CLIENT_PID || error "rmdir remote dir failed"
797 replay_barrier mds${MDTIDX}
798 create_remote_dir_files_23 || error "Remote creation failed $?"
801 checkstat_23 || error "check stat failed $?"
803 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
806 run_test 23b "c1 rmdir d1, M1 drop reply and fail M0/M1, c2 mkdir d1"
809 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
811 ([ $FAILURE_MODE == "HARD" ] &&
812 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
813 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
816 local remote_dir=$tdir/remote_dir
818 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
819 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir ||
820 error "lfs mkdir failed"
822 # OBD_FAIL_UPDATE_OBJ_NET_REP 0x1701
823 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
824 do_node $CLIENT1 rmdir $MOUNT1/$remote_dir &
826 do_facet mds${MDTIDX} lctl set_param fail_loc=0
829 wait $CLIENT_PID || error "rmdir remote dir failed"
831 replay_barrier mds${MDTIDX}
832 create_remote_dir_files_23 || error "Remote creation failed $?"
835 checkstat_23 || error "check stat failed $?"
837 rm -rf $MOUNT1/$tdir || return 6
840 run_test 23c "c1 rmdir d1, M0 drop update reply and fail M0, c2 mkdir d1"
843 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
845 local remote_dir=$tdir/remote_dir
847 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
848 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir ||
849 error "lfs mkdir failed"
851 # OBD_FAIL_UPDATE_OBJ_NET 0x1701
852 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
853 do_node $CLIENT1 rmdir $MOUNT1/$remote_dir &
855 do_facet mds${MDTIDX} lctl set_param fail_loc=0
857 fail mds${MDTIDX},mds$((MDTIDX + 1))
858 wait $CLIENT_PID || error "rmdir remote dir failed"
860 replay_barrier mds${MDTIDX}
861 create_remote_dir_files_23 || error "Remote creation failed $?"
864 checkstat_23 || error "check stat failed $?"
866 rm -rf $MOUNT1/$tdir || return 6
869 run_test 23d "c1 rmdir d1, M0 drop update reply and fail M0/M1, c2 mkdir d1"
872 [[ $(lustre_version_code $SINGLEMDS) -gt $(version_code 2.5.2) ]] ||
873 { skip "Need MDS version newer than 2.5.2"; return 0; }
876 stat $MOUNT/$tfile >&/dev/null
877 # OBD_FAIL_MDS_REINT_NET_REP
878 do_facet $SINGLEMDS $LCTL set_param fail_loc=0x119
879 $TRUNCATE $MOUNT/$tfile 100 &
882 do_facet $SINGLEMDS lctl set_param fail_loc=0
883 # sync to release rep-ack lock quickly
884 do_nodes $(comma_list $(mdts_nodes)) \
885 "lctl set_param -n osd*.*MDT*.force_sync 1"
889 run_test 24 "reconstruct on non-existing object"
891 # end commit on sharing tests
896 $SETSTRIPE -i 0 -c 1 $DIR/$tfile
898 # get lock for the 1st client
899 dd if=/dev/zero of=$DIR/$tfile count=1 >/dev/null ||
900 error "failed to write data"
902 # get waiting locks for the 2nd client
903 drop_ldlm_cancel "multiop $DIR2/$tfile Ow512" &
906 # failover, replay and resend replayed waiting locks
907 if [ $(lustre_version_code ost1) -ge $(version_code 2.6.90) ]; then
908 #define OBD_FAIL_LDLM_SRV_CP_AST 0x325
909 do_facet ost1 lctl set_param fail_loc=0x80000325
911 #define OBD_FAIL_OST_LDLM_REPLY_NET 0x213
912 do_facet ost1 lctl set_param fail_loc=0x80000213
917 # multiop does not finish because CP AST is skipped;
918 # it is ok to kill it in the test, because CP AST is already re-sent
919 # and it does not hung forever in real life
923 run_test 25 "replay|resend"
928 kill -9 $dbench_26_pid
933 local clients=${CLIENTS:-$HOSTNAME}
935 zconf_mount_clients $clients $MOUNT
938 [ "$SLOW" = "no" ] && duration=200
939 # set duration to 900 because it takes some time to boot node
940 [ "$FAILURE_MODE" = HARD ] && duration=900
942 local start_ts=$SECONDS
947 local tar_dir=$DIR/$tdir/run_tar
949 test_mkdir -p -c$MDSCOUNT $tar_dir || break
950 if [ $MDSCOUNT -ge 2 ]; then
951 $LFS setdirstripe -D -c$MDSCOUNT $tar_dir ||
952 error "set default dirstripe failed"
955 tar cf - /etc | tar xf - || error "tar failed"
956 cd $DIR/$tdir || break
957 rm -rf $tar_dir || break
961 echo "Started tar $tar_26_pid"
964 local dbench_dir=$DIR2/$tdir/run_dbench
966 test_mkdir -p -c$MDSCOUNT $dbench_dir || break
967 if [ $MDSCOUNT -ge 2 ]; then
968 $LFS setdirstripe -D -c$MDSCOUNT $dbench_dir ||
969 error "set default dirstripe failed"
971 cd $dbench_dir || break
972 rundbench 1 -D $dbench_dir -t 100 &>/dev/null || break
973 cd $DIR/$tdir || break
974 rm -rf $dbench_dir || break
978 echo "Started dbench $dbench_26_pid"
980 local num_failovers=0
982 while [ $((SECONDS - start_ts)) -lt $duration ]; do
983 kill -0 $tar_26_pid || error "tar $tar_26_pid missing"
984 kill -0 $dbench_26_pid || error "dbench $dbench_26_pid missing"
986 replay_barrier mds$fail_index
987 sleep 2 # give clients a time to do operations
988 # Increment the number of failovers
989 num_failovers=$((num_failovers + 1))
990 log "$TESTNAME fail mds$fail_index $num_failovers times"
992 if [ $fail_index -ge $MDSCOUNT ]; then
995 fail_index=$((fail_index + 1))
998 # stop the client loads
999 kill -0 $tar_26_pid || error "tar $tar_26_pid stopped"
1000 kill -0 $dbench_26_pid || error "dbench $dbench_26_pid stopped"
1003 run_test 26 "dbench and tar with mds failover"
1006 $SETSTRIPE -i 0 -c 1 $DIR2/$tfile
1007 dd if=/dev/zero of=$DIR2/$tfile bs=4096 count=1
1009 #define OBD_FAIL_LDLM_SRV_BL_AST 0x324
1010 do_facet ost1 $LCTL set_param fail_loc=0x80000324
1012 dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 &
1016 #define OBD_FAIL_LDLM_GRANT_CHECK 0x32a
1017 do_facet ost1 $LCTL set_param fail_loc=0x32a
1022 cancel_lru_locks OST0000-osc
1023 wait $pid || error "dd failed"
1025 run_test 28 "lock replay should be ordered: waiting after granted"
1028 SLEEP=$((SECONDS - $NOW))
1029 [ $SLEEP -lt $TIMEOUT ] && sleep $SLEEP
1030 [ "$MOUNTED2" = yes ] && zconf_umount $HOSTNAME $MOUNT2 || true
1031 check_and_cleanup_lustre