5 PTLDEBUG=${PTLDEBUG:--1}
6 MOUNT_2=${MOUNT_2:-"yes"}
8 LUSTRE=${LUSTRE:-$(cd $(dirname $0)/..; echo $PWD)}
9 . $LUSTRE/tests/test-framework.sh
13 remote_mds_nodsh && skip "remote MDS with nodsh" && exit 0
15 ALWAYS_EXCEPT="$REPLAY_DUAL_EXCEPT "
16 # bug number for skipped test: LU-2012 LU-8333 LU-7372
17 ALWAYS_EXCEPT+=" 14b 21b 26 "
18 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
20 [[ "$mds1_FSTYPE" == zfs ]] &&
21 # bug number for skipped test: LU-2230
22 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 21b"
25 # bug number for skipped tests: LU-9795 LU-9795
26 ALWAYS_EXCEPT+=" 0a 0b "
30 [ "$SLOW" = "no" ] && EXCEPT_SLOW="21b "
33 check_and_setup_lustre
35 MOUNTED=$(mounted_lustre_filesystems)
36 if ! $(echo $MOUNTED' ' | grep -w -q $MOUNT2' '); then
37 zconf_mount $HOSTNAME $MOUNT2
42 rm -rf $DIR/[df][0-9]*
44 [ "$DAEMONFILE" ] && $LCTL debug_daemon start $DAEMONFILE $DAEMONSIZE
46 # if there is no CLIENT1 defined, some tests can be ran on localhost
47 CLIENT1=${CLIENT1:-$HOSTNAME}
48 # if CLIENT2 doesn't exist then use CLIENT1 instead
49 # All tests should use CLIENT2 with MOUNT2 only therefore it will work if
51 # Exception is the test which need two separate nodes
52 CLIENT2=${CLIENT2:-$CLIENT1}
54 # LU-482 Avert LVM and VM inability to flush caches in pre .33 kernels
55 if [ $LINUX_VERSION_CODE -lt $(version_code 2.6.33) ]; then
57 do_facet $SINGLEMDS "sync; sleep 10; sync; sleep 10; sync"
60 LU482_FAILED=$(mktemp -u $TMP/$TESTSUITE.lu482.XXXXXX)
62 echo "Check file is LU482_FAILED=$LU482_FAILED"
63 touch $MOUNT2/$tfile-A # force sync FLD/SEQ update before barrier
64 replay_barrier $SINGLEMDS
65 #define OBD_FAIL_PTLRPC_FINISH_REPLAY | OBD_FAIL_ONCE
67 createmany -o $MOUNT1/$tfile- 50
68 $LCTL set_param fail_loc=0x80000514
69 facet_failover $SINGLEMDS
70 [ -f "$LU482_FAILED" ] && skip "LU-482 failure" && return 0
74 zconf_mount `hostname` $MOUNT2 || error "mount2 fais"
75 unlinkmany $MOUNT1/$tfile- 50 || return 2
76 rm $MOUNT2/$tfile || return 3
77 rm $MOUNT2/$tfile-A || return 4
79 run_test 0a "expired recovery with lost client"
81 if [ -f "$LU482_FAILED" ]; then
82 log "Found check file $LU482_FAILED, aborting test script"
83 rm -vf "$LU482_FAILED"
85 do_nodes $CLIENTS umount -f $MOUNT2 || true
86 do_nodes $CLIENTS umount -f $MOUNT || true
87 # copied from stopall, but avoid the MDS recovery
88 for num in `seq $OSTCOUNT`; do
90 rm -f $TMP/ost${num}active
92 if ! combined_mgs_mds ; then
100 replay_barrier $SINGLEMDS
102 touch $MOUNT1/$tfile-2
104 facet_failover $SINGLEMDS
106 zconf_mount `hostname` $MOUNT1 || error "mount1 fais"
107 zconf_mount `hostname` $MOUNT2 || error "mount2 fais"
108 # it is uncertain if file-2 exists or not, remove it if it does
109 checkstat $MOUNT1/$tfile-2 && rm $MOUNT1/$tfile-2
110 checkstat $MOUNT2/$tfile && return 2
113 run_test 0b "lost client during waiting for next transno"
117 replay_barrier $SINGLEMDS
121 checkstat $MOUNT2/a || return 1
122 checkstat $MOUNT1/b || return 2
123 rm $MOUNT2/a $MOUNT1/b
124 checkstat $MOUNT1/a && return 3
125 checkstat $MOUNT2/b && return 4
129 run_test 1 "|X| simple create"
133 replay_barrier $SINGLEMDS
137 checkstat $MOUNT2/adir || return 1
139 checkstat $MOUNT2/adir && return 2
142 run_test 2 "|X| mkdir adir"
145 replay_barrier $SINGLEMDS
147 mkdir $MOUNT2/adir/bdir
150 checkstat $MOUNT2/adir || return 1
151 checkstat $MOUNT1/adir/bdir || return 2
152 rmdir $MOUNT2/adir/bdir $MOUNT1/adir
153 checkstat $MOUNT1/adir && return 3
154 checkstat $MOUNT2/adir/bdir && return 4
157 run_test 3 "|X| mkdir adir, mkdir adir/bdir "
161 replay_barrier $SINGLEMDS
162 mkdir $MOUNT1/adir && return 1
163 mkdir $MOUNT2/adir/bdir
166 checkstat $MOUNT2/adir || return 2
167 checkstat $MOUNT1/adir/bdir || return 3
169 rmdir $MOUNT2/adir/bdir $MOUNT1/adir
170 checkstat $MOUNT1/adir && return 4
171 checkstat $MOUNT2/adir/bdir && return 5
174 run_test 4 "|X| mkdir adir (-EEXIST), mkdir adir/bdir "
178 # multiclient version of replay_single.sh/test_8
180 multiop_bg_pause $MOUNT2/a o_tSc || return 1
183 replay_barrier $SINGLEMDS
185 wait $pid || return 1
188 [ -e $MOUNT2/a ] && return 2
191 run_test 5 "open, unlink |X| close"
196 multiop_bg_pause $MOUNT2/a o_c || return 1
198 multiop_bg_pause $MOUNT1/a o_c || return 1
201 replay_barrier $SINGLEMDS
203 wait $pid1 || return 1
207 wait $pid2 || return 1
208 [ -e $MOUNT2/a ] && return 2
211 run_test 6 "open1, open2, unlink |X| close1 [fail $SINGLEMDS] close2"
214 replay_barrier $SINGLEMDS
215 drop_reint_reply "mcreate $MOUNT1/$tfile" || return 1
217 checkstat $MOUNT2/$tfile || return 2
218 rm $MOUNT1/$tfile || return 3
222 run_test 8 "replay of resent request"
225 replay_barrier $SINGLEMDS
226 mcreate $MOUNT1/$tfile-1
227 mcreate $MOUNT2/$tfile-2
228 # drop first reint reply
229 do_facet $SINGLEMDS lctl set_param fail_loc=0x80000119
231 do_facet $SINGLEMDS lctl set_param fail_loc=0
233 rm $MOUNT1/$tfile-[1,2] || return 1
237 run_test 9 "resending a replayed create"
240 mcreate $MOUNT1/$tfile-1
241 replay_barrier $SINGLEMDS
242 munlink $MOUNT1/$tfile-1
243 mcreate $MOUNT2/$tfile-2
244 # drop first reint reply
245 do_facet $SINGLEMDS lctl set_param fail_loc=0x80000119
247 do_facet $SINGLEMDS lctl set_param fail_loc=0
249 checkstat $MOUNT1/$tfile-1 && return 1
250 checkstat $MOUNT1/$tfile-2 || return 2
255 run_test 10 "resending a replayed unlink"
258 replay_barrier $SINGLEMDS
259 mcreate $DIR1/$tfile-1
260 mcreate $DIR2/$tfile-2
261 mcreate $DIR1/$tfile-3
262 mcreate $DIR2/$tfile-4
263 mcreate $DIR1/$tfile-5
264 # drop all reint replies for a while
265 do_facet $SINGLEMDS $LCTL set_param fail_loc=0x0119
266 # note that with this fail_loc set, facet_failover df will fail
267 facet_failover $SINGLEMDS
269 local clients=${CLIENTS:-$HOSTNAME}
270 wait_clients_import_state "$clients" $SINGLEMDS FULL
272 do_facet $SINGLEMDS $LCTL set_param fail_loc=0
274 rm $DIR1/$tfile-[1-5] || return 1
278 run_test 11 "both clients timeout during replay"
281 replay_barrier $SINGLEMDS
283 multiop_bg_pause $DIR/$tfile mo_c || return 1
286 #define OBD_FAIL_LDLM_ENQUEUE_NET 0x302
287 do_facet $SINGLEMDS lctl set_param fail_loc=0x80000302
288 facet_failover $SINGLEMDS
289 do_facet $SINGLEMDS lctl set_param fail_loc=0
290 clients_up || return 1
293 kill -USR1 $MULTIPID || return 3
294 wait $MULTIPID || return 4
295 $CHECKSTAT -t file $DIR/$tfile || return 2
300 run_test 12 "open resend timeout"
303 multiop_bg_pause $DIR/$tfile mo_c || return 1
306 replay_barrier $SINGLEMDS
308 kill -USR1 $MULTIPID || return 3
309 wait $MULTIPID || return 4
312 do_facet $SINGLEMDS lctl set_param fail_loc=0x80000115
313 facet_failover $SINGLEMDS
314 do_facet $SINGLEMDS lctl set_param fail_loc=0
315 clients_up || return 1
318 $CHECKSTAT -t file $DIR/$tfile || return 2
323 run_test 13 "close resend timeout"
325 # test 14a removed after 18143 because it shouldn't fail anymore and do the same
330 wait_delete_completed
332 local beforeused=$(df -P $DIR | tail -1 | awk '{ print $3 }')
334 mkdir -p $MOUNT1/$tdir
335 $LFS setstripe -i 0 $MOUNT1/$tdir
336 replay_barrier $SINGLEMDS
337 createmany -o $MOUNT1/$tdir/$tfile- 5
339 $LFS setstripe -i 0 $MOUNT2/$tfile-2
340 dd if=/dev/zero of=$MOUNT2/$tfile-2 bs=1M count=5
341 createmany -o $MOUNT1/$tdir/$tfile-3- 5
345 wait_recovery_complete $SINGLEMDS || error "MDS recovery not done"
347 # first set of files should have been replayed
348 unlinkmany $MOUNT1/$tdir/$tfile- 5 || error "first unlinks failed"
349 unlinkmany $MOUNT1/$tdir/$tfile-3- 5 || error "second unlinks failed"
351 zconf_mount $HOSTNAME $MOUNT2 || error "mount $MOUNT2 failed"
352 [ -f $MOUNT2/$tfile-2 ] && error "$MOUNT2/$tfile-2 exists!"
354 wait_mds_ost_sync || error "wait_mds_ost_sync failed"
355 wait_delete_completed || error "wait_delete_complete failed"
357 local afterused=$(df -P $DIR | tail -1 | awk '{ print $3 }')
358 log "before $beforeused, after $afterused"
359 # leave some margin for some files/dirs to be modified (OI, llog, etc)
360 [ $afterused -le $((beforeused + $(fs_log_size))) ] ||
361 error "after $afterused > before $beforeused"
363 run_test 14b "delete ost orphans if gap occured in objids due to VBR"
365 test_15a() { # was test_15
366 replay_barrier $SINGLEMDS
367 createmany -o $MOUNT1/$tfile- 25
368 createmany -o $MOUNT2/$tfile-2- 1
373 unlinkmany $MOUNT1/$tfile- 25 || return 2
374 [ -e $MOUNT1/$tfile-2-0 ] && error "$tfile-2-0 exists"
376 zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
379 run_test 15a "timeout waiting for lost client during replay, 1 client completes"
382 replay_barrier $SINGLEMDS
383 for ((i = 0; i < 2000; i++)); do
384 echo "data" > "$MOUNT2/${tfile}-$i" || error "create ${tfile}-$i failed"
390 zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
393 run_test 15c "remove multiple OST orphans"
396 replay_barrier $SINGLEMDS
397 createmany -o $MOUNT1/$tfile- 25
398 createmany -o $MOUNT2/$tfile-2- 1
401 facet_failover $SINGLEMDS
405 unlinkmany $MOUNT1/$tfile- 25 || return 2
407 zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
411 run_test 16 "fail MDS during recovery (3571)"
414 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
416 createmany -o $MOUNT1/$tfile- 25
417 createmany -o $MOUNT2/$tfile-2- 1
419 # Make sure the disconnect is lost
427 unlinkmany $MOUNT1/$tfile- 25 || return 2
429 zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
433 run_test 17 "fail OST during recovery (3571)"
435 # cleanup with blocked enqueue fails until timer elapses (MDS busy), wait for it
438 test_18() { # bug 3822 - evicting client with enqueued lock
440 local DLMTRACE=$(do_facet $SINGLEMDS lctl get_param debug)
441 do_facet $SINGLEMDS lctl set_param debug=+dlmtrace
442 mkdir -p $MOUNT1/$tdir || error "mkdir $MOUNT1/$tdir failed"
443 touch $MOUNT1/$tdir/${tfile}0 || error "touch file failed"
444 statmany -s $MOUNT1/$tdir/$tfile 1 500 &
447 #define OBD_FAIL_LDLM_ENQUEUE_BLOCKED 0x30b
448 do_facet $SINGLEMDS lctl set_param fail_loc=0x8000030b # hold enqueue
450 #define OBD_FAIL_LDLM_BL_CALLBACK_NET 0x305
451 do_facet client lctl set_param ldlm.namespaces.*.early_lock_cancel=0
452 do_facet client lctl set_param fail_loc=0x80000305 # drop cb, evict
454 usleep 500 # wait to ensure first client is one that will be evicted
455 openfile -f O_RDONLY $MOUNT2/$tdir/$tfile
457 do_facet client lctl set_param ldlm.namespaces.*.early_lock_cancel=1
458 do_facet $SINGLEMDS lctl debug_kernel |
459 grep "not entering recovery" && error "client not evicted"
460 do_facet client "lctl set_param fail_loc=0"
461 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
463 run_test 18 "ldlm_handle_enqueue succeeds on evicted export (3822)"
465 test_19() { # Bug 10991 - resend of open request does not fail assertion.
466 replay_barrier $SINGLEMDS
467 drop_mdt_ldlm_reply "createmany -o $DIR/$tfile 1" || return 1
469 checkstat $DIR2/${tfile}0 || return 2
470 rm $DIR/${tfile}0 || return 3
474 run_test 19 "resend of open request"
477 local before=$SECONDS
478 replay_barrier $SINGLEMDS
484 zconf_mount $HOSTNAME $DIR2 || error "mount $DIR2 fail"
485 local tier1=$((SECONDS - before))
488 replay_barrier $SINGLEMDS
494 zconf_mount $HOSTNAME $DIR2 || error "mount $DIR2 fail"
495 local tier2=$((SECONDS - before))
497 # timeout is more than 1.5x original timeout
498 ((tier2 < tier1 * 6 / 4)) ||
499 error "recovery time $tier2 >= 1.5x original time $tier1"
501 run_test 20 "recovery time is not increasing"
503 # commit on sharing tests
505 local param_file=$TMP/$tfile-params
507 save_lustre_params $SINGLEMDS "mdt.*.commit_on_sharing" > $param_file
508 do_facet $SINGLEMDS lctl set_param mdt.*.commit_on_sharing=1
509 touch $MOUNT1/$tfile-1
510 mv $MOUNT2/$tfile-1 $MOUNT2/$tfile-2
511 mv $MOUNT1/$tfile-2 $MOUNT1/$tfile-3
512 replay_barrier_nosync $SINGLEMDS
515 facet_failover $SINGLEMDS
517 # all renames are replayed
518 unlink $MOUNT1/$tfile-3 || return 2
520 zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
522 do_facet $SINGLEMDS lctl set_param mdt.*.commit_on_sharing=0
523 rm -rf $MOUNT1/$tfile-*
524 restore_lustre_params < $param_file
528 run_test 21a "commit on sharing"
532 do_node $CLIENT1 rm -f $MOUNT1/$tfile-*
535 do_node $CLIENT1 touch $MOUNT1/$tfile-1
536 do_node $CLIENT2 mv $MOUNT1/$tfile-1 $MOUNT1/$tfile-2
537 do_node $CLIENT1 mv $MOUNT1/$tfile-2 $MOUNT1/$tfile-3
539 replay_barrier_nosync $mds
540 shutdown_client $CLIENT2 $MOUNT1
544 # were renames replayed?
546 echo UNLINK $MOUNT1/$tfile-3
547 do_node $CLIENT1 unlink $MOUNT1/$tfile-3 ||
548 { echo "unlink $tfile-3 fail!" && rc=1; }
551 zconf_mount_clients $CLIENT2 $MOUNT1 ||
552 error "mount $CLIENT2 $MOUNT1 fail"
558 [ -z "$CLIENTS" ] && skip "Need two or more clients" && return
559 [ $CLIENTCOUNT -lt 2 ] &&
560 { skip "Need 2+ clients, have $CLIENTCOUNT" && return; }
562 if [ "$FAILURE_MODE" = "HARD" ] && mixed_mdt_devs; then
563 skip "Several MDTs on one MDS with FAILURE_MODE=$FAILURE_MODE"
567 zconf_umount_clients $CLIENTS $MOUNT2
568 zconf_mount_clients $CLIENTS $MOUNT1
570 local param_file=$TMP/$tfile-params
572 local mdtidx=$($LFS getstripe -m $MOUNT1)
573 local facet=mds$((mdtidx + 1))
575 save_lustre_params $facet "mdt.*.commit_on_sharing" > $param_file
579 do_facet $facet lctl set_param mdt.*.commit_on_sharing=$COS
581 test_21b_sub $facet || error "Not all renames are replayed. COS=$COS"
583 # there is still a window when transactions may be written to disk
584 # before the mds device is set R/O. To avoid such a rare test failure,
585 # the check is repeated several times.
589 # COS disabled (should fail)
590 do_facet $facet lctl set_param mdt.*.commit_on_sharing=$COS
592 test_21b_sub $facet || break
593 n_attempts=$((n_attempts + 1))
594 [ $n_attempts -gt 3 ] &&
595 error "can't check if COS works: rename replied w/o COS"
597 zconf_mount_clients $CLIENTS $MOUNT2
598 restore_lustre_params < $param_file
602 run_test 21b "commit on sharing, two clients"
605 checkstat $MOUNT1/$remote_dir || return 1
606 checkstat $MOUNT1/$remote_dir/dir || return 2
607 checkstat $MOUNT1/$remote_dir/$tfile-1 || return 3
608 checkstat $MOUNT1/$remote_dir/dir/$tfile-1 || return 4
612 create_remote_dir_files_22() {
613 do_node $CLIENT2 mkdir ${MOUNT2}/$remote_dir/dir || return 1
614 do_node $CLIENT1 createmany -o $MOUNT1/$remote_dir/dir/$tfile- 2 ||
616 do_node $CLIENT2 createmany -o $MOUNT2/$remote_dir/$tfile- 2 ||
622 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
623 ([ $FAILURE_MODE == "HARD" ] &&
624 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
625 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
629 local remote_dir=${tdir}/remote_dir
631 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
633 # OBD_FAIL_MDS_REINT_NET_REP 0x119
634 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
635 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir &
638 fail mds$((MDTIDX + 1))
639 wait $CLIENT_PID || error "lfs mkdir failed"
641 replay_barrier mds$MDTIDX
642 create_remote_dir_files_22 || error "Remote creation failed $?"
645 checkstat_22 || error "check stat failed $?"
647 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
650 run_test 22a "c1 lfs mkdir -i 1 dir1, M1 drop reply & fail, c2 mkdir dir1/dir"
653 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
655 local remote_dir=$tdir/remote_dir
657 # OBD_FAIL_MDS_REINT_NET_REP 0x119
658 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
660 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
661 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir &
664 fail mds${MDTIDX},mds$((MDTIDX + 1))
665 wait $CLIENT_PID || error "lfs mkdir failed"
667 replay_barrier mds$MDTIDX
668 create_remote_dir_files_22 || error "Remote creation failed $?"
671 checkstat_22 || error "check stat failed $?"
673 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
676 run_test 22b "c1 lfs mkdir -i 1 d1, M1 drop reply & fail M0/M1, c2 mkdir d1/dir"
679 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
680 ([ $FAILURE_MODE == "HARD" ] &&
681 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
682 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
685 local remote_dir=${tdir}/remote_dir
687 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
689 # OBD_FAIL_OUT_UPDATE_NET_REP 0x1701
690 do_facet mds$MDTIDX lctl set_param fail_loc=0x1701
691 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir &
693 do_facet mds$MDTIDX lctl set_param fail_loc=0
696 wait $CLIENT_PID || error "lfs mkdir failed"
698 replay_barrier mds$MDTIDX
699 create_remote_dir_files_22 || error "Remote creation failed $?"
702 checkstat_22 || error "check stat failed $?"
704 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
707 run_test 22c "c1 lfs mkdir -i 1 d1, M1 drop update & fail M1, c2 mkdir d1/dir"
710 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
712 local remote_dir=${tdir}/remote_dir
714 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
716 # OBD_FAIL_OUT_UPDATE_NET_REP 0x1701
717 do_facet mds$MDTIDX lctl set_param fail_loc=0x1701
718 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir &
720 do_facet mds$MDTIDX lctl set_param fail_loc=0
722 fail mds${MDTIDX},mds$((MDTIDX + 1))
723 wait $CLIENT_PID || error "lfs mkdir failed"
725 replay_barrier mds$MDTIDX
726 create_remote_dir_files_22 || error "Remote creation failed $?"
729 checkstat_22 || error "check stat failed $?"
731 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
734 run_test 22d "c1 lfs mkdir -i 1 d1, M1 drop update & fail M0/M1,c2 mkdir d1/dir"
737 checkstat $MOUNT1/$remote_dir || return 1
738 checkstat $MOUNT1/$remote_dir/$tfile-1 || return 2
742 create_remote_dir_files_23() {
743 do_node $CLIENT2 mkdir ${MOUNT2}/$remote_dir || return 1
744 do_node $CLIENT2 createmany -o $MOUNT2/$remote_dir/$tfile- 2 || return 2
749 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
750 ([ $FAILURE_MODE == "HARD" ] &&
751 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
752 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
755 local remote_dir=$tdir/remote_dir
757 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
758 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir ||
759 error "lfs mkdir failed"
760 # OBD_FAIL_MDS_REINT_NET_REP 0x119
761 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
762 do_node $CLIENT1 rmdir $MOUNT1/$remote_dir &
764 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0
766 fail mds$((MDTIDX + 1))
767 wait $CLIENT_PID || error "rmdir remote dir failed"
769 replay_barrier mds${MDTIDX}
770 create_remote_dir_files_23 || error "Remote creation failed $?"
773 checkstat_23 || error "check stat failed $?"
775 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
778 run_test 23a "c1 rmdir d1, M1 drop reply and fail, client2 mkdir d1"
781 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
783 local remote_dir=$tdir/remote_dir
785 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
786 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir ||
787 error "lfs mkdir failed"
789 # OBD_FAIL_MDS_REINT_NET_REP 0x119
790 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
791 do_node $CLIENT1 rmdir $MOUNT1/$remote_dir &
793 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0
795 fail mds${MDTIDX},mds$((MDTIDX + 1))
796 wait $CLIENT_PID || error "rmdir remote dir failed"
798 replay_barrier mds${MDTIDX}
799 create_remote_dir_files_23 || error "Remote creation failed $?"
802 checkstat_23 || error "check stat failed $?"
804 rm -rf $MOUNT1/$tdir || error "rmdir remote_dir failed"
807 run_test 23b "c1 rmdir d1, M1 drop reply and fail M0/M1, c2 mkdir d1"
810 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
812 ([ $FAILURE_MODE == "HARD" ] &&
813 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
814 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
817 local remote_dir=$tdir/remote_dir
819 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
820 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir ||
821 error "lfs mkdir failed"
823 # OBD_FAIL_OUT_UPDATE_NET_REP 0x1701
824 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
825 do_node $CLIENT1 rmdir $MOUNT1/$remote_dir &
827 do_facet mds${MDTIDX} lctl set_param fail_loc=0
830 wait $CLIENT_PID || error "rmdir remote dir failed"
832 replay_barrier mds${MDTIDX}
833 create_remote_dir_files_23 || error "Remote creation failed $?"
836 checkstat_23 || error "check stat failed $?"
838 rm -rf $MOUNT1/$tdir || return 6
841 run_test 23c "c1 rmdir d1, M0 drop update reply and fail M0, c2 mkdir d1"
844 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
846 local remote_dir=$tdir/remote_dir
848 do_node $CLIENT1 mkdir -p $MOUNT1/${tdir}
849 do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir ||
850 error "lfs mkdir failed"
852 # OBD_FAIL_UPDATE_OBJ_NET 0x1701
853 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
854 do_node $CLIENT1 rmdir $MOUNT1/$remote_dir &
856 do_facet mds${MDTIDX} lctl set_param fail_loc=0
858 fail mds${MDTIDX},mds$((MDTIDX + 1))
859 wait $CLIENT_PID || error "rmdir remote dir failed"
861 replay_barrier mds${MDTIDX}
862 create_remote_dir_files_23 || error "Remote creation failed $?"
865 checkstat_23 || error "check stat failed $?"
867 rm -rf $MOUNT1/$tdir || return 6
870 run_test 23d "c1 rmdir d1, M0 drop update reply and fail M0/M1, c2 mkdir d1"
873 [[ "$MDS1_VERSION" -gt $(version_code 2.5.2) ]] ||
874 skip "Need MDS version newer than 2.5.2"
877 stat $MOUNT/$tfile >&/dev/null
878 # OBD_FAIL_MDS_REINT_NET_REP
879 do_facet $SINGLEMDS $LCTL set_param fail_loc=0x119
880 $TRUNCATE $MOUNT/$tfile 100 &
883 do_facet $SINGLEMDS lctl set_param fail_loc=0
884 # sync to release rep-ack lock quickly
885 do_nodes $(comma_list $(mdts_nodes)) \
886 "lctl set_param -n osd*.*MDT*.force_sync 1"
890 run_test 24 "reconstruct on non-existing object"
892 # end commit on sharing tests
897 $LFS setstripe -i 0 -c 1 $DIR/$tfile
899 # get lock for the 1st client
900 dd if=/dev/zero of=$DIR/$tfile count=1 >/dev/null ||
901 error "failed to write data"
903 # get waiting locks for the 2nd client
904 drop_ldlm_cancel "multiop $DIR2/$tfile Ow512" &
907 # failover, replay and resend replayed waiting locks
908 if [ "$OST1_VERSION" -ge $(version_code 2.6.90) ]; then
909 #define OBD_FAIL_LDLM_SRV_CP_AST 0x325
910 do_facet ost1 lctl set_param fail_loc=0x80000325
912 #define OBD_FAIL_OST_LDLM_REPLY_NET 0x213
913 do_facet ost1 lctl set_param fail_loc=0x80000213
918 # multiop does not finish because CP AST is skipped;
919 # it is ok to kill it in the test, because CP AST is already re-sent
920 # and it does not hung forever in real life
924 run_test 25 "replay|resend"
929 kill -9 $dbench_26_pid
934 local clients=${CLIENTS:-$HOSTNAME}
936 zconf_mount_clients $clients $MOUNT
939 [ "$SLOW" = "no" ] && duration=200
940 # set duration to 900 because it takes some time to boot node
941 [ "$FAILURE_MODE" = HARD ] && duration=900
943 local start_ts=$SECONDS
948 local tar_dir=$DIR/$tdir/run_tar
950 test_mkdir -p -c$MDSCOUNT $tar_dir || break
951 if [ $MDSCOUNT -ge 2 ]; then
952 $LFS setdirstripe -D -c$MDSCOUNT $tar_dir ||
953 error "set default dirstripe failed"
956 tar cf - /etc | tar xf - || error "tar failed"
957 cd $DIR/$tdir || break
958 rm -rf $tar_dir || break
962 echo "Started tar $tar_26_pid"
965 local dbench_dir=$DIR2/$tdir/run_dbench
967 test_mkdir -p -c$MDSCOUNT $dbench_dir || break
968 if [ $MDSCOUNT -ge 2 ]; then
969 $LFS setdirstripe -D -c$MDSCOUNT $dbench_dir ||
970 error "set default dirstripe failed"
972 cd $dbench_dir || break
973 rundbench 1 -D $dbench_dir -t 100 &>/dev/null || break
974 cd $DIR/$tdir || break
975 rm -rf $dbench_dir || break
979 echo "Started dbench $dbench_26_pid"
981 local num_failovers=0
983 while [ $((SECONDS - start_ts)) -lt $duration ]; do
984 kill -0 $tar_26_pid || error "tar $tar_26_pid missing"
985 kill -0 $dbench_26_pid || error "dbench $dbench_26_pid missing"
987 replay_barrier mds$fail_index
988 sleep 2 # give clients a time to do operations
989 # Increment the number of failovers
990 num_failovers=$((num_failovers + 1))
991 log "$TESTNAME fail mds$fail_index $num_failovers times"
993 if [ $fail_index -ge $MDSCOUNT ]; then
996 fail_index=$((fail_index + 1))
999 # stop the client loads
1000 kill -0 $tar_26_pid || error "tar $tar_26_pid stopped"
1001 kill -0 $dbench_26_pid || error "dbench $dbench_26_pid stopped"
1004 run_test 26 "dbench and tar with mds failover"
1007 $LFS setstripe -i 0 -c 1 $DIR2/$tfile
1008 dd if=/dev/zero of=$DIR2/$tfile bs=4096 count=1
1010 #define OBD_FAIL_LDLM_SRV_BL_AST 0x324
1011 do_facet ost1 $LCTL set_param fail_loc=0x80000324
1013 dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 &
1017 #define OBD_FAIL_LDLM_GRANT_CHECK 0x32a
1018 do_facet ost1 $LCTL set_param fail_loc=0x32a
1023 cancel_lru_locks OST0000-osc
1024 wait $pid || error "dd failed"
1026 run_test 28 "lock replay should be ordered: waiting after granted"
1029 local dir0=$DIR/$tdir/d0
1030 local dir1=$DIR/$tdir/d1
1032 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
1033 [ $CLIENTCOUNT -lt 2 ] && skip "needs >= 2 clients" && return 0
1034 [ "$CLIENT1" == "$CLIENT2" ] &&
1035 skip "clients must be on different nodes" && return 0
1038 $LFS mkdir -i0 $dir0
1039 $LFS mkdir -i1 $dir1
1043 # create a remote dir, drop reply
1044 #define OBD_FAIL_PTLRPC_ROUND_XID 0x530
1045 $LCTL set_param fail_loc=0x530 fail_val=36
1046 #define OBD_FAIL_MDS_REINT_MULTI_NET_REP 0x15a
1047 do_facet mds2 $LCTL set_param fail_loc=0x8000015a
1048 echo make remote dir d0 for $dir0
1049 $LFS mkdir -i1 -c1 $dir0/d3 &
1052 echo make local dir d1 for $dir1
1053 do_node $CLIENT2 $LCTL set_param fail_loc=0x530 fail_val=36
1054 do_node $CLIENT2 mkdir $dir1/d4
1058 run_test 29 "replay vs update with the same xid"
1061 SLEEP=$((SECONDS - $NOW))
1062 [ $SLEEP -lt $TIMEOUT ] && sleep $SLEEP
1063 [ "$MOUNTED2" = yes ] && zconf_umount $HOSTNAME $MOUNT2 || true
1064 check_and_cleanup_lustre