7 # This test needs to be run on the client
10 export MULTIOP=${MULTIOP:-multiop}
11 LUSTRE=${LUSTRE:-$(cd $(dirname $0)/..; echo $PWD)}
14 . $LUSTRE/tests/test-framework.sh
16 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
18 CHECK_GRANT=${CHECK_GRANT:-"yes"}
19 GRANT_CHECK_LIST=${GRANT_CHECK_LIST:-""}
21 require_dsh_mds || exit 0
24 # bug number for skipped tests:
25 # b=17466/LU-472 : 61d
27 ALWAYS_EXCEPT="61d 53a 53d $REPLAY_SINGLE_EXCEPT"
28 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
30 case "$(lsb_release -sr)" in # only disable tests for el7
31 7*) # bug number: LU-6455-----
32 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 28"
36 # 63 min 7 min AT AT AT AT"
37 [ "$SLOW" = "no" ] && EXCEPT_SLOW="1 2 3 4 6 12 16 44a 44b 65 66 67 68"
39 [ $(facet_fstype $SINGLEMDS) = "zfs" ] &&
40 # bug number for skipped test: LU-1867 LU-3127
41 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 89 73b"
45 check_and_setup_lustre
50 rm -rf $DIR/[df][0-9]* $DIR/f.$TESTSUITE.*
52 # LU-482 Avert LVM and VM inability to flush caches in pre .33 kernels
53 if [ $LINUX_VERSION_CODE -lt $(version_code 2.6.33) ]; then
55 do_facet $SINGLEMDS sync
58 test_0a() { # was test_0
59 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
60 replay_barrier $SINGLEMDS
64 run_test 0a "empty replay"
67 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
69 # this test attempts to trigger a race in the precreation code,
70 # and must run before any other objects are created on the filesystem
72 createmany -o $DIR/$tfile 20 || error "createmany -o $DIR/$tfile failed"
73 unlinkmany $DIR/$tfile 20 || error "unlinkmany $DIR/$tfile failed"
75 run_test 0b "ensure object created after recover exists. (3284)"
78 replay_barrier $SINGLEMDS
81 facet_failover $SINGLEMDS
82 zconf_mount $(hostname) $MOUNT || error "mount fails"
83 client_up || error "post-failover df failed"
84 # file shouldn't exist if replay-barrier works as expected
85 rm $DIR/$tfile && error "File exists and it shouldn't"
88 run_test 0c "check replay-barrier"
91 replay_barrier $SINGLEMDS
93 facet_failover $SINGLEMDS
94 zconf_mount $(hostname) $MOUNT || error "mount fails"
95 client_up || error "post-failover df failed"
97 run_test 0d "expired recovery with no clients"
100 replay_barrier $SINGLEMDS
103 $CHECKSTAT -t file $DIR/$tfile ||
104 error "$CHECKSTAT $DIR/$tfile attribute check failed"
107 run_test 1 "simple create"
110 replay_barrier $SINGLEMDS
113 $CHECKSTAT -t file $DIR/$tfile ||
114 error "$CHECKSTAT $DIR/$tfile attribute check failed"
120 mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed"
121 replay_barrier $SINGLEMDS
124 $CHECKSTAT -t file $DIR/$tfile ||
125 error "$CHECKSTAT $DIR/$tfile attribute check failed"
131 local file=$DIR/$tfile
132 replay_barrier $SINGLEMDS
134 openfile -f O_DIRECTORY $file
136 $CHECKSTAT -t file $file ||
137 error "$CHECKSTAT $file attribute check failed"
140 run_test 3a "replay failed open(O_DIRECTORY)"
143 replay_barrier $SINGLEMDS
144 #define OBD_FAIL_MDS_OPEN_PACK | OBD_FAIL_ONCE
145 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000114"
147 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
149 $CHECKSTAT -t file $DIR/$tfile &&
150 error "$CHECKSTAT $DIR/$tfile attribute check should fail"
153 run_test 3b "replay failed open -ENOMEM"
156 replay_barrier $SINGLEMDS
157 #define OBD_FAIL_MDS_ALLOC_OBDO | OBD_FAIL_ONCE
158 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000128"
160 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
163 $CHECKSTAT -t file $DIR/$tfile &&
164 error "$CHECKSTAT $DIR/$tfile attribute check should fail"
167 run_test 3c "replay failed open -ENOMEM"
169 test_4a() { # was test_4
170 replay_barrier $SINGLEMDS
171 for i in $(seq 10); do
172 echo "tag-$i" > $DIR/$tfile-$i
175 for i in $(seq 10); do
176 grep -q "tag-$i" $DIR/$tfile-$i || error "$tfile-$i"
179 run_test 4a "|x| 10 open(O_CREAT)s"
182 for i in $(seq 10); do
183 echo "tag-$i" > $DIR/$tfile-$i
185 replay_barrier $SINGLEMDS
188 $CHECKSTAT -t file $DIR/$tfile-* &&
189 error "$CHECKSTAT $DIR/$tfile-* attribute check should fail" ||
192 run_test 4b "|x| rm 10 files"
194 # The idea is to get past the first block of precreated files on both
195 # osts, and then replay.
197 replay_barrier $SINGLEMDS
198 for i in $(seq 220); do
199 echo "tag-$i" > $DIR/$tfile-$i
202 for i in $(seq 220); do
203 grep -q "tag-$i" $DIR/$tfile-$i || error "$tfile-$i"
207 # waiting for commitment of removal
209 run_test 5 "|x| 220 open(O_CREAT)"
211 test_6a() { # was test_6
212 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
213 replay_barrier $SINGLEMDS
214 mcreate $DIR/$tdir/$tfile
216 $CHECKSTAT -t dir $DIR/$tdir ||
217 error "$CHECKSTAT $DIR/$tdir attribute check failed"
218 $CHECKSTAT -t file $DIR/$tdir/$tfile ||
219 error "$CHECKSTAT $DIR/$tdir/$tfile attribute check failed"
221 # waiting for log process thread
223 run_test 6a "mkdir + contained create"
226 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
227 replay_barrier $SINGLEMDS
230 $CHECKSTAT -t dir $DIR/$tdir &&
231 error "$CHECKSTAT $DIR/$tdir attribute check should fail" ||
234 run_test 6b "|X| rmdir"
237 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
238 replay_barrier $SINGLEMDS
239 mcreate $DIR/$tdir/$tfile
241 $CHECKSTAT -t dir $DIR/$tdir ||
242 error "$CHECKSTAT $DIR/$tdir attribute check failed"
243 $CHECKSTAT -t file $DIR/$tdir/$tfile ||
244 error "$CHECKSTAT $DIR/$tdir/$tfile attribute check failed"
247 run_test 7 "mkdir |X| contained create"
250 replay_barrier $SINGLEMDS
251 multiop_bg_pause $DIR/$tfile mo_c ||
252 error "multiop mknod $DIR/$tfile failed"
256 $CHECKSTAT -t file $DIR/$tfile ||
257 error "$CHECKSTAT $DIR/$tfile attribute check failed"
258 kill -USR1 $MULTIPID || error "multiop mknod $MULTIPID not running"
259 wait $MULTIPID || error "multiop mknod $MULTIPID failed"
262 run_test 8 "creat open |X| close"
265 replay_barrier $SINGLEMDS
267 local old_inum=$(ls -i $DIR/$tfile | awk '{print $1}')
269 local new_inum=$(ls -i $DIR/$tfile | awk '{print $1}')
271 echo " old_inum == $old_inum, new_inum == $new_inum"
272 if [ $old_inum -eq $new_inum ] ;
274 echo "old_inum and new_inum match"
276 echo " old_inum and new_inum do not match"
277 error "old index($old_inum) does not match new index($new_inum)"
281 run_test 9 "|X| create (same inum/gen)"
284 mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed"
285 replay_barrier $SINGLEMDS
286 mv $DIR/$tfile $DIR/$tfile-2
289 $CHECKSTAT $DIR/$tfile &&
290 error "$CHECKSTAT $DIR/$tfile attribute check should fail"
291 $CHECKSTAT $DIR/$tfile-2 ||
292 error "$CHECKSTAT $DIR/$tfile-2 attribute check failed"
296 run_test 10 "create |X| rename unlink"
299 mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed"
300 echo "old" > $DIR/$tfile
301 mv $DIR/$tfile $DIR/$tfile-2
302 replay_barrier $SINGLEMDS
303 echo "new" > $DIR/$tfile
305 grep old $DIR/$tfile-2
307 grep new $DIR/$tfile || error "grep $DIR/$tfile failed"
308 grep old $DIR/$tfile-2 || error "grep $DIR/$tfile-2 failed"
310 run_test 11 "create open write rename |X| create-old-name read"
313 mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed"
314 multiop_bg_pause $DIR/$tfile o_tSc ||
315 error "multiop_bg_pause $DIR/$tfile failed"
318 replay_barrier $SINGLEMDS
319 kill -USR1 $pid || error "multiop $pid not running"
320 wait $pid || error "multiop $pid failed"
323 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
326 run_test 12 "open, unlink |X| close"
328 # 1777 - replay open after committed chmod that would make
329 # a regular open a failure
331 mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed"
332 multiop_bg_pause $DIR/$tfile O_wc ||
333 error "multiop_bg_pause $DIR/$tfile failed"
336 $CHECKSTAT -p 0 $DIR/$tfile ||
337 error "$CHECKSTAT $DIR/$tfile attribute check failed"
338 replay_barrier $SINGLEMDS
340 kill -USR1 $pid || error "multiop $pid not running"
341 wait $pid || error "multiop $pid failed"
343 $CHECKSTAT -s 1 -p 0 $DIR/$tfile ||
344 error "second $CHECKSTAT $DIR/$tfile attribute check failed"
345 rm $DIR/$tfile || error "rm $DIR/$tfile failed"
348 run_test 13 "open chmod 0 |x| write close"
351 multiop_bg_pause $DIR/$tfile O_tSc ||
352 error "multiop_bg_pause $DIR/$tfile failed"
355 replay_barrier $SINGLEMDS
356 kill -USR1 $pid || error "multiop $pid not running"
357 wait $pid || error "multiop $pid failed"
360 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
363 run_test 14 "open(O_CREAT), unlink |X| close"
366 multiop_bg_pause $DIR/$tfile O_tSc ||
367 error "multiop_bg_pause $DIR/$tfile failed"
370 replay_barrier $SINGLEMDS
371 touch $DIR/$tfile-1 || error "touch $DIR/$tfile-1 failed"
372 kill -USR1 $pid || error "multiop $pid not running"
373 wait $pid || error "multiop $pid failed"
376 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
377 touch $DIR/$tfile-2 || error "touch $DIR/$tfile-2 failed"
380 run_test 15 "open(O_CREAT), unlink |X| touch new, close"
383 replay_barrier $SINGLEMDS
386 mcreate $DIR/$tfile-2
388 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
389 [ -e $DIR/$tfile-2 ] || error "file $DIR/$tfile-2 does not exist"
390 munlink $DIR/$tfile-2 || error "munlink $DIR/$tfile-2 failed"
392 run_test 16 "|X| open(O_CREAT), unlink, touch new, unlink new"
395 replay_barrier $SINGLEMDS
396 multiop_bg_pause $DIR/$tfile O_c ||
397 error "multiop_bg_pause $DIR/$tfile failed"
400 kill -USR1 $pid || error "multiop $pid not running"
401 wait $pid || error "multiop $pid failed"
402 $CHECKSTAT -t file $DIR/$tfile ||
403 error "$CHECKSTAT $DIR/$tfile attribute check failed"
406 run_test 17 "|X| open(O_CREAT), |replay| close"
409 replay_barrier $SINGLEMDS
410 multiop_bg_pause $DIR/$tfile O_tSc ||
411 error "multiop_bg_pause $DIR/$tfile failed"
414 touch $DIR/$tfile-2 || error "touch $DIR/$tfile-2 failed"
415 echo "pid: $pid will close"
416 kill -USR1 $pid || error "multiop $pid not running"
417 wait $pid || error "multiop $pid failed"
420 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
421 [ -e $DIR/$tfile-2 ] || error "file $DIR/$tfile-2 does not exist"
422 # this touch frequently fails
423 touch $DIR/$tfile-3 || error "touch $DIR/$tfile-3 failed"
424 munlink $DIR/$tfile-2 || error "munlink $DIR/$tfile-2 failed"
425 munlink $DIR/$tfile-3 || error "munlink $DIR/$tfile-3 failed"
428 run_test 18 "open(O_CREAT), unlink, touch new, close, touch, unlink"
430 # bug 1855 (a simpler form of test_11 above)
432 replay_barrier $SINGLEMDS
434 echo "old" > $DIR/$tfile
435 mv $DIR/$tfile $DIR/$tfile-2
436 grep old $DIR/$tfile-2
438 grep old $DIR/$tfile-2 || error "grep $DIR/$tfile-2 failed"
440 run_test 19 "mcreate, open, write, rename "
442 test_20a() { # was test_20
443 replay_barrier $SINGLEMDS
444 multiop_bg_pause $DIR/$tfile O_tSc ||
445 error "multiop_bg_pause $DIR/$tfile failed"
450 kill -USR1 $pid || error "multiop $pid not running"
451 wait $pid || error "multiop $pid failed"
452 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
455 run_test 20a "|X| open(O_CREAT), unlink, replay, close (test mds_cleanup_orphans)"
457 test_20b() { # bug 10480
458 local wait_timeout=$((TIMEOUT * 4))
462 BEFOREUSED=$(df -P $DIR | tail -1 | awk '{ print $3 }')
463 dd if=/dev/zero of=$DIR/$tfile bs=4k count=10000 &
464 while [ ! -e $DIR/$tfile ] ; do
465 usleep 60 # give dd a chance to start
468 $GETSTRIPE $DIR/$tfile || error "$GETSTRIPE $DIR/$tfile failed"
470 rm -f $DIR/$tfile || error "rm -f $DIR/$tfile failed"
472 client_up || client_up || true # reconnect
474 do_facet $SINGLEMDS "lctl set_param -n osd*.*MDT*.force_sync 1"
476 fail $SINGLEMDS # start orphan recovery
477 wait_recovery_complete $SINGLEMDS || error "MDS recovery not done"
478 wait_delete_completed_mds $wait_timeout ||
479 error "delete did not complete"
481 AFTERUSED=$(df -P $DIR | tail -1 | awk '{ print $3 }')
482 log "before $BEFOREUSED, after $AFTERUSED"
483 (( $AFTERUSED > $BEFOREUSED + $(fs_log_size) )) &&
484 error "after $AFTERUSED > before $BEFOREUSED"
487 run_test 20b "write, unlink, eviction, replay, (test mds_cleanup_orphans)"
489 test_20c() { # bug 10480
490 multiop_bg_pause $DIR/$tfile Ow_c ||
491 error "multiop_bg_pause $DIR/$tfile failed"
497 client_up || client_up || true # reconnect
499 kill -USR1 $pid || error "multiop $pid not running"
500 wait $pid || error "multiop $pid failed"
501 [ -s $DIR/$tfile ] || error "File was truncated"
505 run_test 20c "check that client eviction does not affect file content"
508 replay_barrier $SINGLEMDS
509 multiop_bg_pause $DIR/$tfile O_tSc ||
510 error "multiop_bg_pause $DIR/$tfile failed"
513 touch $DIR/$tfile-1 || error "touch $DIR/$tfile-1 failed"
516 kill -USR1 $pid || error "multiop $pid not running"
517 wait $pid || error "multiop $pid failed"
518 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
519 touch $DIR/$tfile-2 || error "touch $DIR/$tfile-2 failed"
522 run_test 21 "|X| open(O_CREAT), unlink touch new, replay, close (test mds_cleanup_orphans)"
525 multiop_bg_pause $DIR/$tfile O_tSc ||
526 error "multiop_bg_pause $DIR/$tfile failed"
529 replay_barrier $SINGLEMDS
533 kill -USR1 $pid || error "multiop $pid not running"
534 wait $pid || error "multiop $pid failed"
535 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
538 run_test 22 "open(O_CREAT), |X| unlink, replay, close (test mds_cleanup_orphans)"
541 multiop_bg_pause $DIR/$tfile O_tSc ||
542 error "multiop_bg_pause $DIR/$tfile failed"
545 replay_barrier $SINGLEMDS
547 touch $DIR/$tfile-1 || error "touch $DIR/$tfile-1 failed"
550 kill -USR1 $pid || error "multiop $pid not running"
551 wait $pid || error "multiop $pid failed"
552 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
553 touch $DIR/$tfile-2 || error "touch $DIR/$tfile-2 failed"
556 run_test 23 "open(O_CREAT), |X| unlink touch new, replay, close (test mds_cleanup_orphans)"
559 multiop_bg_pause $DIR/$tfile O_tSc ||
560 error "multiop_bg_pause $DIR/$tfile failed"
563 replay_barrier $SINGLEMDS
566 kill -USR1 $pid || error "multiop $pid not running"
567 wait $pid || error "multiop $pid failed"
568 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
571 run_test 24 "open(O_CREAT), replay, unlink, close (test mds_cleanup_orphans)"
574 multiop_bg_pause $DIR/$tfile O_tSc ||
575 error "multiop_bg_pause $DIR/$tfile failed"
579 replay_barrier $SINGLEMDS
581 kill -USR1 $pid || error "multiop $pid not running"
582 wait $pid || error "multiop $pid failed"
583 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
586 run_test 25 "open(O_CREAT), unlink, replay, close (test mds_cleanup_orphans)"
589 replay_barrier $SINGLEMDS
590 multiop_bg_pause $DIR/$tfile-1 O_tSc ||
591 error "multiop_bg_pause $DIR/$tfile-1 failed"
593 multiop_bg_pause $DIR/$tfile-2 O_tSc ||
594 error "multiop_bg_pause $DIR/$tfile-2 failed"
598 kill -USR1 $pid2 || error "second multiop $pid2 not running"
599 wait $pid2 || error "second multiop $pid2 failed"
602 kill -USR1 $pid1 || error "multiop $pid1 not running"
603 wait $pid1 || error "multiop $pid1 failed"
604 [ -e $DIR/$tfile-1 ] && error "file $DIR/$tfile-1 should not exist"
605 [ -e $DIR/$tfile-2 ] && error "file $DIR/$tfile-2 should not exist"
608 run_test 26 "|X| open(O_CREAT), unlink two, close one, replay, close one (test mds_cleanup_orphans)"
611 replay_barrier $SINGLEMDS
612 multiop_bg_pause $DIR/$tfile-1 O_tSc ||
613 error "multiop_bg_pause $DIR/$tfile-1 failed"
615 multiop_bg_pause $DIR/$tfile-2 O_tSc ||
616 error "multiop_bg_pause $DIR/$tfile-2 failed"
622 kill -USR1 $pid1 || error "multiop $pid1 not running"
623 wait $pid1 || error "multiop $pid1 failed"
624 kill -USR1 $pid2 || error "second multiop $pid2 not running"
625 wait $pid2 || error "second multiop $pid2 failed"
626 [ -e $DIR/$tfile-1 ] && error "file $DIR/$tfile-1 should not exist"
627 [ -e $DIR/$tfile-2 ] && error "file $DIR/$tfile-2 should not exist"
630 run_test 27 "|X| open(O_CREAT), unlink two, replay, close two (test mds_cleanup_orphans)"
633 multiop_bg_pause $DIR/$tfile-1 O_tSc ||
634 error "multiop_bg_pause $DIR/$tfile-1 failed"
636 multiop_bg_pause $DIR/$tfile-2 O_tSc ||
637 error "multiop_bg_pause $DIR/$tfile-2 failed"
639 replay_barrier $SINGLEMDS
642 kill -USR1 $pid2 || error "second multiop $pid2 not running"
643 wait $pid2 || error "second multiop $pid2 failed"
646 kill -USR1 $pid1 || error "multiop $pid1 not running"
647 wait $pid1 || error "multiop $pid1 failed"
648 [ -e $DIR/$tfile-1 ] && error "file $DIR/$tfile-1 should not exist"
649 [ -e $DIR/$tfile-2 ] && error "file $DIR/$tfile-2 should not exist"
652 run_test 28 "open(O_CREAT), |X| unlink two, close one, replay, close one (test mds_cleanup_orphans)"
655 multiop_bg_pause $DIR/$tfile-1 O_tSc ||
656 error "multiop_bg_pause $DIR/$tfile-1 failed"
658 multiop_bg_pause $DIR/$tfile-2 O_tSc ||
659 error "multiop_bg_pause $DIR/$tfile-2 failed"
661 replay_barrier $SINGLEMDS
666 kill -USR1 $pid1 || error "multiop $pid1 not running"
667 wait $pid1 || error "multiop $pid1 failed"
668 kill -USR1 $pid2 || error "second multiop $pid2 not running"
669 wait $pid2 || error "second multiop $pid2 failed"
670 [ -e $DIR/$tfile-1 ] && error "file $DIR/$tfile-1 should not exist"
671 [ -e $DIR/$tfile-2 ] && error "file $DIR/$tfile-2 should not exist"
674 run_test 29 "open(O_CREAT), |X| unlink two, replay, close two (test mds_cleanup_orphans)"
677 multiop_bg_pause $DIR/$tfile-1 O_tSc ||
678 error "multiop_bg_pause $DIR/$tfile-1 failed"
680 multiop_bg_pause $DIR/$tfile-2 O_tSc ||
681 error "multiop_bg_pause $DIR/$tfile-2 failed"
686 replay_barrier $SINGLEMDS
688 kill -USR1 $pid1 || error "multiop $pid1 not running"
689 wait $pid1 || error "multiop $pid1 failed"
690 kill -USR1 $pid2 || error "second multiop $pid2 not running"
691 wait $pid2 || error "second multiop $pid2 failed"
692 [ -e $DIR/$tfile-1 ] && error "file $DIR/$tfile-1 should not exist"
693 [ -e $DIR/$tfile-2 ] && error "file $DIR/$tfile-2 should not exist"
696 run_test 30 "open(O_CREAT) two, unlink two, replay, close two (test mds_cleanup_orphans)"
699 multiop_bg_pause $DIR/$tfile-1 O_tSc ||
700 error "multiop_bg_pause $DIR/$tfile-1 failed"
702 multiop_bg_pause $DIR/$tfile-2 O_tSc ||
703 error "multiop_bg_pause $DIR/$tfile-2 failed"
707 replay_barrier $SINGLEMDS
710 kill -USR1 $pid1 || error "multiop $pid1 not running"
711 wait $pid1 || error "multiop $pid1 failed"
712 kill -USR1 $pid2 || error "second multiop $pid2 not running"
713 wait $pid2 || error "second multiop $pid2 failed"
714 [ -e $DIR/$tfile-1 ] && error "file $DIR/$tfile-1 should not exist"
715 [ -e $DIR/$tfile-2 ] && error "file $DIR/$tfile-2 should not exist"
718 run_test 31 "open(O_CREAT) two, unlink one, |X| unlink one, close two (test mds_cleanup_orphans)"
720 # tests for bug 2104; completion without crashing is success. The close is
721 # stale, but we always return 0 for close, so the app never sees it.
723 multiop_bg_pause $DIR/$tfile O_c ||
724 error "multiop_bg_pause $DIR/$tfile failed"
726 multiop_bg_pause $DIR/$tfile O_c ||
727 error "second multiop_bg_pause $DIR/$tfile failed"
730 client_up || client_up || error "client_up failed"
731 kill -USR1 $pid1 || error "multiop $pid1 not running"
732 kill -USR1 $pid2 || error "second multiop $pid2 not running"
733 wait $pid1 || error "multiop $pid1 failed"
734 wait $pid2 || error "second multiop $pid2 failed"
737 run_test 32 "close() notices client eviction; close() after client eviction"
740 createmany -o $DIR/$tfile-%d 10 ||
741 error "createmany create $DIR/$tfile failed"
742 replay_barrier_nosync $SINGLEMDS
743 fail_abort $SINGLEMDS
744 # recreate shouldn't fail
745 createmany -o $DIR/$tfile--%d 10 ||
746 error "createmany recreate $DIR/$tfile failed"
750 run_test 33a "fid seq shouldn't be reused after abort recovery"
753 #define OBD_FAIL_SEQ_ALLOC 0x1311
754 do_facet $SINGLEMDS "lctl set_param fail_loc=0x1311"
756 createmany -o $DIR/$tfile-%d 10
757 replay_barrier_nosync $SINGLEMDS
758 fail_abort $SINGLEMDS
759 # recreate shouldn't fail
760 createmany -o $DIR/$tfile--%d 10 ||
761 error "createmany recreate $DIR/$tfile failed"
765 run_test 33b "test fid seq allocation"
768 multiop_bg_pause $DIR/$tfile O_c ||
769 error "multiop_bg_pause $DIR/$tfile failed"
773 replay_barrier $SINGLEMDS
774 fail_abort $SINGLEMDS
775 kill -USR1 $pid || error "multiop $pid not running"
776 wait $pid || error "multiop $pid failed"
777 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
781 run_test 34 "abort recovery before client does replay (test mds_cleanup_orphans)"
783 # bug 2278 - generate one orphan on OST, then destroy it during recovery from llog
785 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
787 #define OBD_FAIL_MDS_REINT_NET_REP 0x119
788 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000119"
793 # give a chance to remove from MDS
794 fail_abort $SINGLEMDS
795 $CHECKSTAT -t file $DIR/$tfile &&
796 error "$CHECKSTAT $DIR/$tfile attribute check should fail" ||
799 run_test 35 "test recovery from llog for unlink op"
801 # b=2432 resent cancel after replay uses wrong cookie,
802 # so don't resend cancels
804 replay_barrier $SINGLEMDS
806 checkstat $DIR/$tfile
807 facet_failover $SINGLEMDS
809 if dmesg | grep "unknown lock cookie"; then
810 error "cancel after replay failed"
813 run_test 36 "don't resend cancel"
816 # directory orphans can't be unlinked from PENDING directory
818 rmdir $DIR/$tfile 2>/dev/null
819 multiop_bg_pause $DIR/$tfile dD_c ||
820 error "multiop_bg_pause $DIR/$tfile failed"
824 replay_barrier $SINGLEMDS
825 # clear the dmesg buffer so we only see errors from this recovery
826 do_facet $SINGLEMDS dmesg -c >/dev/null
827 fail_abort $SINGLEMDS
828 kill -USR1 $pid || error "multiop $pid not running"
829 do_facet $SINGLEMDS dmesg | grep "error .* unlinking .* from PENDING" &&
830 error "error unlinking files"
831 wait $pid || error "multiop $pid failed"
835 run_test 37 "abort recovery before client does replay (test mds_cleanup_orphans for directories)"
838 createmany -o $DIR/$tfile-%d 800 ||
839 error "createmany -o $DIR/$tfile failed"
840 unlinkmany $DIR/$tfile-%d 0 400 || error "unlinkmany $DIR/$tfile failed"
841 replay_barrier $SINGLEMDS
843 unlinkmany $DIR/$tfile-%d 400 400 ||
844 error "unlinkmany $DIR/$tfile 400 failed"
846 $CHECKSTAT -t file $DIR/$tfile-* &&
847 error "$CHECKSTAT $DIR/$tfile-* attribute check should fail" ||
850 run_test 38 "test recovery from unlink llog (test llog_gen_rec) "
852 test_39() { # bug 4176
853 createmany -o $DIR/$tfile-%d 800 ||
854 error "createmany -o $DIR/$tfile failed"
855 replay_barrier $SINGLEMDS
856 unlinkmany $DIR/$tfile-%d 0 400
858 unlinkmany $DIR/$tfile-%d 400 400 ||
859 error "unlinkmany $DIR/$tfile 400 failed"
861 $CHECKSTAT -t file $DIR/$tfile-* &&
862 error "$CHECKSTAT $DIR/$tfile-* attribute check should fail" ||
865 run_test 39 "test recovery from unlink llog (test llog_gen_rec) "
868 lctl get_param -n osc.*.stats | awk -vwrites=0 '/ost_write/ { writes += $2 } END { print writes; }'
873 # always need connection to MDS to verify layout during IO. LU-2628.
874 lctl get_param mdc.*.connect_flags | grep -q layout_lock &&
875 skip "layout_lock needs MDS connection for IO" && return 0
877 $LCTL mark multiop $MOUNT/$tfile OS_c
878 multiop $MOUNT/$tfile OS_c &
880 writeme -s $MOUNT/${tfile}-2 &
883 facet_failover $SINGLEMDS
884 #define OBD_FAIL_MDS_CONNECT_NET 0x117
885 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000117"
887 stat1=$(count_ost_writes)
889 stat2=$(count_ost_writes)
890 echo "$stat1, $stat2"
891 if [ $stat1 -lt $stat2 ]; then
892 echo "writes continuing during recovery"
895 echo "writes not continuing during recovery, bug 2477"
898 echo "waiting for writeme $WRITE_PID"
902 echo "waiting for multiop $PID"
903 wait $PID || error "multiop $PID failed"
904 do_facet client munlink $MOUNT/$tfile ||
905 error "munlink $MOUNT/$tfile failed"
906 do_facet client munlink $MOUNT/${tfile}-2 ||
907 error "munlink $MOUNT/$tfile-2 failed"
910 run_test 40 "cause recovery in ptlrpc, ensure IO continues"
913 # make sure that a read to one osc doesn't try to double-unlock its page just
914 # because another osc is invalid. trigger_group_io used to mistakenly return
915 # an error if any oscs were invalid even after having successfully put rpcs
916 # on valid oscs. This was fatal if the caller was ll_readpage who unlocked
917 # the page, guarnateeing that the unlock from the RPC completion would
918 # assert on trying to unlock the unlocked page.
920 [ $OSTCOUNT -lt 2 ] &&
921 skip_env "skipping test 41: we don't have a second OST to test with" &&
924 local f=$MOUNT/$tfile
925 # make sure the start of the file is ost1
926 $SETSTRIPE -S $((128 * 1024)) -i 0 $f
927 do_facet client dd if=/dev/zero of=$f bs=4k count=1 ||
928 error "dd on client failed"
930 # fail ost2 and read from ost1
931 local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS $ost2_svc)
932 local osc2dev=$(do_facet $SINGLEMDS "lctl get_param -n devices" |
933 grep $mdtosc | awk '{print $1}')
934 [ -z "$osc2dev" ] && echo "OST: $ost2_svc" &&
935 lctl get_param -n devices &&
936 error "OST 2 $osc2dev does not exist"
937 do_facet $SINGLEMDS $LCTL --device $osc2dev deactivate ||
938 error "deactive device on $SINGLEMDS failed"
939 do_facet client dd if=$f of=/dev/null bs=4k count=1 ||
940 error "second dd on client failed"
941 do_facet $SINGLEMDS $LCTL --device $osc2dev activate ||
942 error "active device on $SINGLEMDS failed"
945 run_test 41 "read from a valid osc while other oscs are invalid"
947 # test MDS recovery after ost failure
949 blocks=$(df -P $MOUNT | tail -n 1 | awk '{ print $2 }')
950 createmany -o $DIR/$tfile-%d 800 ||
951 error "createmany -o $DIR/$tfile failed"
953 unlinkmany $DIR/$tfile-%d 0 400
955 lctl set_param debug=-1
958 # osc is evicted, fs is smaller (but only with failout OSTs (bug 7287)
959 #blocks_after=`df -P $MOUNT | tail -n 1 | awk '{ print $2 }'`
960 #[ $blocks_after -lt $blocks ] || return 1
961 echo "wait for MDS to timeout and recover"
962 sleep $((TIMEOUT * 2))
964 unlinkmany $DIR/$tfile-%d 400 400 ||
965 error "unlinkmany $DIR/$tfile 400 failed"
966 $CHECKSTAT -t file $DIR/$tfile-* &&
967 error "$CHECKSTAT $DIR/$tfile-* attribute check should fail" ||
970 run_test 42 "recovery after ost failure"
972 # timeout in MDS/OST recovery RPC will LBUG MDS
973 test_43() { # bug 2530
974 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
976 replay_barrier $SINGLEMDS
978 # OBD_FAIL_OST_CREATE_NET 0x204
979 do_facet ost1 "lctl set_param fail_loc=0x80000204"
985 run_test 43 "mds osc import failure during recovery; don't LBUG"
987 test_44a() { # was test_44
990 local mdcdev=$($LCTL dl |
991 awk "/${FSNAME}-MDT0000-mdc-/ {if (\$2 == \"UP\") {print \$1}}")
992 [ "$mdcdev" ] || error "${FSNAME}-MDT0000-mdc- not UP"
993 [ $(echo $mdcdev | wc -w) -eq 1 ] ||
994 { $LCTL dl; error "looking for mdcdev=$mdcdev"; }
996 # adaptive timeouts slow this way down
997 if at_is_enabled; then
998 at_max_saved=$(at_max_get mds)
1002 for i in $(seq 1 10); do
1003 echo "$i of 10 ($(date +%s))"
1004 do_facet $SINGLEMDS \
1005 "lctl get_param -n md[ts].*.mdt.timeouts | grep service"
1006 #define OBD_FAIL_TGT_CONN_RACE 0x701
1007 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000701"
1008 # lctl below may fail, it is valid case
1009 $LCTL --device $mdcdev recover
1012 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1013 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds
1016 run_test 44a "race in target handle connect"
1019 local mdcdev=$($LCTL dl |
1020 awk "/${FSNAME}-MDT0000-mdc-/ {if (\$2 == \"UP\") {print \$1}}")
1021 [ "$mdcdev" ] || error "${FSNAME}-MDT0000-mdc not up"
1022 [ $(echo $mdcdev | wc -w) -eq 1 ] ||
1023 { echo mdcdev=$mdcdev; $LCTL dl;
1024 error "more than one ${FSNAME}-MDT0000-mdc"; }
1026 for i in $(seq 1 10); do
1027 echo "$i of 10 ($(date +%s))"
1028 do_facet $SINGLEMDS \
1029 "lctl get_param -n md[ts].*.mdt.timeouts | grep service"
1030 #define OBD_FAIL_TGT_DELAY_RECONNECT 0x704
1031 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000704"
1032 # lctl below may fail, it is valid case
1033 $LCTL --device $mdcdev recover
1038 run_test 44b "race in target handle connect"
1041 replay_barrier $SINGLEMDS
1042 createmany -m $DIR/$tfile-%d 100 || error "failed to create directories"
1043 #define OBD_FAIL_TGT_RCVG_FLAG 0x712
1044 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000712"
1045 fail_abort $SINGLEMDS
1046 unlinkmany $DIR/$tfile-%d 100 && error "unliked after fail abort"
1048 unlinkmany $DIR/$tfile-%d 100 && error "unliked after fail"
1051 run_test 44c "race in target handle connect"
1053 # Handle failed close
1055 local mdcdev=$($LCTL get_param -n devices |
1056 awk "/ ${FSNAME}-MDT0000-mdc-/ {print \$1}")
1057 [ "$mdcdev" ] || error "${FSNAME}-MDT0000-mdc not up"
1058 [ $(echo $mdcdev | wc -w) -eq 1 ] ||
1059 { echo mdcdev=$mdcdev; $LCTL dl;
1060 error "more than one ${FSNAME}-MDT0000-mdc"; }
1062 $LCTL --device $mdcdev recover ||
1063 error "$LCTL --device $mdcdev recover failed"
1065 multiop_bg_pause $DIR/$tfile O_c ||
1066 error "multiop_bg_pause $DIR/$tfile failed"
1069 # This will cause the CLOSE to fail before even
1070 # allocating a reply buffer
1071 $LCTL --device $mdcdev deactivate ||
1072 error "$LCTL --device $mdcdev deactivate failed"
1075 kill -USR1 $pid || error "multiop $pid not running"
1076 wait $pid || error "multiop $pid failed"
1078 $LCTL --device $mdcdev activate ||
1079 error "$LCTL --device $mdcdev activate failed"
1082 $CHECKSTAT -t file $DIR/$tfile ||
1083 error "$CHECKSTAT $DIR/$tfile attribute check failed"
1086 run_test 45 "Handle failed close"
1090 drop_reply "touch $DIR/$tfile"
1092 # ironically, the previous test, 45, will cause a real forced close,
1093 # so just look for one for this test
1094 dmesg | grep -i "force closing client file handle for $tfile" &&
1095 error "found force closing in dmesg"
1098 run_test 46 "Don't leak file handle after open resend (3325)"
1100 test_47() { # bug 2824
1101 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1103 # create some files to make sure precreate has been done on all
1104 # OSTs. (just in case this test is run independently)
1105 createmany -o $DIR/$tfile 20 ||
1106 error "createmany create $DIR/$tfile failed"
1108 # OBD_FAIL_OST_CREATE_NET 0x204
1110 do_facet ost1 "lctl set_param fail_loc=0x80000204"
1111 client_up || error "client_up failed"
1113 # let the MDS discover the OST failure, attempt to recover, fail
1114 # and recover again.
1115 sleep $((3 * TIMEOUT))
1117 # Without 2824, this createmany would hang
1118 createmany -o $DIR/$tfile 20 ||
1119 error "createmany recraete $DIR/$tfile failed"
1120 unlinkmany $DIR/$tfile 20 || error "unlinkmany $DIR/$tfile failed"
1124 run_test 47 "MDS->OSC failure during precreate cleanup (2824)"
1127 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1128 [ "$OSTCOUNT" -lt "2" ] &&
1129 skip_env "$OSTCOUNT < 2 OSTs -- skipping" && return
1131 replay_barrier $SINGLEMDS
1132 createmany -o $DIR/$tfile 20 ||
1133 error "createmany -o $DIR/$tfile failed"
1134 # OBD_FAIL_OST_EROFS 0x216
1135 facet_failover $SINGLEMDS
1136 do_facet ost1 "lctl set_param fail_loc=0x80000216"
1137 client_up || error "client_up failed"
1139 createmany -o $DIR/$tfile 20 20 ||
1140 error "createmany recraete $DIR/$tfile failed"
1141 unlinkmany $DIR/$tfile 40 || error "unlinkmany $DIR/$tfile failed"
1144 run_test 48 "MDS->OSC failure during precreate cleanup (2824)"
1147 local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS $ost1_svc)
1148 local oscdev=$(do_facet $SINGLEMDS "lctl get_param -n devices" |
1149 grep $mdtosc | awk '{print $1}')
1150 [ "$oscdev" ] || error "could not find OSC device on MDS"
1151 do_facet $SINGLEMDS $LCTL --device $oscdev recover ||
1152 error "OSC device $oscdev recovery failed"
1153 do_facet $SINGLEMDS $LCTL --device $oscdev recover ||
1154 error "second OSC device $oscdev recovery failed"
1155 # give the mds_lov_sync threads a chance to run
1158 run_test 50 "Double OSC recovery, don't LASSERT (3812)"
1160 # b3764 timed out lock replay
1162 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.6.90) ] &&
1163 skip "MDS prior to 2.6.90 handle LDLM_REPLY_NET incorrectly" &&
1166 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
1167 cancel_lru_locks mdc
1169 multiop_bg_pause $DIR/$tfile s_s || error "multiop $DIR/$tfile failed"
1172 #define OBD_FAIL_MDS_LDLM_REPLY_NET 0x157
1173 lctl set_param -n ldlm.cancel_unused_locks_before_replay "0"
1174 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000157"
1176 fail $SINGLEMDS || error "fail $SINGLEMDS failed"
1178 wait $mpid || error "multiop_bg_pause pid failed"
1180 do_facet $SINGLEMDS "lctl set_param fail_loc=0x0"
1181 lctl set_param fail_loc=0x0
1182 lctl set_param -n ldlm.cancel_unused_locks_before_replay "1"
1185 run_test 52 "time out lock replay (3764)"
1187 # bug 3462 - simultaneous MDC requests
1189 cancel_lru_locks mdc # cleanup locks from former test cases
1190 mkdir $DIR/${tdir}-1 || error "mkdir $DIR/${tdir}-1 failed"
1191 mkdir $DIR/${tdir}-2 || error "mkdir $DIR/${tdir}-2 failed"
1192 multiop $DIR/${tdir}-1/f O_c &
1194 # give multiop a change to open
1197 #define OBD_FAIL_MDS_CLOSE_NET 0x115
1198 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
1199 kill -USR1 $close_pid
1200 cancel_lru_locks mdc # force the close
1201 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1203 mcreate $DIR/${tdir}-2/f || error "mcreate $DIR/${tdir}-2/f failed"
1205 # close should still be here
1206 [ -d /proc/$close_pid ] || error "close_pid doesn't exist"
1208 replay_barrier_nodf $SINGLEMDS
1210 wait $close_pid || error "close_pid $close_pid failed"
1212 $CHECKSTAT -t file $DIR/${tdir}-1/f ||
1213 error "$CHECKSTAT $DIR/${tdir}-1/f attribute check failed"
1214 $CHECKSTAT -t file $DIR/${tdir}-2/f ||
1215 error "$CHECKSTAT $DIR/${tdir}-2/f attribute check failed"
1216 rm -rf $DIR/${tdir}-*
1218 run_test 53a "|X| close request while two MDC requests in flight"
1221 cancel_lru_locks mdc # cleanup locks from former test cases
1223 mkdir $DIR/${tdir}-1 || error "mkdir $DIR/${tdir}-1 failed"
1224 mkdir $DIR/${tdir}-2 || error "mkdir $DIR/${tdir}-2 failed"
1225 multiop_bg_pause $DIR/${tdir}-1/f O_c ||
1226 error "multiop_bg_pause $DIR/${tdir}-1/f failed"
1229 #define OBD_FAIL_MDS_REINT_NET 0x107
1230 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000107"
1231 mcreate $DIR/${tdir}-2/f &
1235 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1236 kill -USR1 $close_pid
1237 cancel_lru_locks mdc # force the close
1238 wait $close_pid || error "close_pid $close_pid failed"
1239 # open should still be here
1240 [ -d /proc/$open_pid ] || error "open_pid doesn't exist"
1242 replay_barrier_nodf $SINGLEMDS
1244 wait $open_pid || error "open_pid failed"
1246 $CHECKSTAT -t file $DIR/${tdir}-1/f ||
1247 error "$CHECKSTAT $DIR/${tdir}-1/f attribute check failed"
1248 $CHECKSTAT -t file $DIR/${tdir}-2/f ||
1249 error "$CHECKSTAT $DIR/${tdir}-2/f attribute check failed"
1250 rm -rf $DIR/${tdir}-*
1252 run_test 53b "|X| open request while two MDC requests in flight"
1255 cancel_lru_locks mdc # cleanup locks from former test cases
1257 mkdir $DIR/${tdir}-1 || error "mkdir $DIR/${tdir}-1 failed"
1258 mkdir $DIR/${tdir}-2 || error "mkdir $DIR/${tdir}-2 failed"
1259 multiop $DIR/${tdir}-1/f O_c &
1262 #define OBD_FAIL_MDS_REINT_NET 0x107
1263 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000107"
1264 mcreate $DIR/${tdir}-2/f &
1268 #define OBD_FAIL_MDS_CLOSE_NET 0x115
1269 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
1270 kill -USR1 $close_pid
1271 cancel_lru_locks mdc # force the close
1273 #bz20647: make sure all pids exist before failover
1274 [ -d /proc/$close_pid ] || error "close_pid doesn't exist"
1275 [ -d /proc/$open_pid ] || error "open_pid doesn't exists"
1276 replay_barrier_nodf $SINGLEMDS
1277 fail_nodf $SINGLEMDS
1278 wait $open_pid || error "open_pid failed"
1280 # close should be gone
1281 [ -d /proc/$close_pid ] && error "close_pid should not exist"
1282 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1284 $CHECKSTAT -t file $DIR/${tdir}-1/f ||
1285 error "$CHECKSTAT $DIR/${tdir}-1/f attribute check failed"
1286 $CHECKSTAT -t file $DIR/${tdir}-2/f ||
1287 error "$CHECKSTAT $DIR/${tdir}-2/f attribute check failed"
1288 rm -rf $DIR/${tdir}-*
1290 run_test 53c "|X| open request and close request while two MDC requests in flight"
1293 cancel_lru_locks mdc # cleanup locks from former test cases
1295 mkdir $DIR/${tdir}-1 || error "mkdir $DIR/${tdir}-1 failed"
1296 mkdir $DIR/${tdir}-2 || error "mkdir $DIR/${tdir}-2 failed"
1297 multiop $DIR/${tdir}-1/f O_c &
1299 # give multiop a chance to open
1302 #define OBD_FAIL_MDS_CLOSE_NET_REP 0x13b
1303 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000013b"
1304 kill -USR1 $close_pid
1305 cancel_lru_locks mdc # force the close
1306 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1307 mcreate $DIR/${tdir}-2/f || error "mcreate $DIR/${tdir}-2/f failed"
1309 # close should still be here
1310 [ -d /proc/$close_pid ] || error "close_pid doesn't exist"
1312 wait $close_pid || error "close_pid failed"
1314 $CHECKSTAT -t file $DIR/${tdir}-1/f ||
1315 error "$CHECKSTAT $DIR/${tdir}-1/f attribute check failed"
1316 $CHECKSTAT -t file $DIR/${tdir}-2/f ||
1317 error "$CHECKSTAT $DIR/${tdir}-2/f attribute check failed"
1318 rm -rf $DIR/${tdir}-*
1320 run_test 53d "close reply while two MDC requests in flight"
1323 cancel_lru_locks mdc # cleanup locks from former test cases
1325 mkdir $DIR/${tdir}-1 || error "mkdir $DIR/${tdir}-1 failed"
1326 mkdir $DIR/${tdir}-2 || error "mkdir $DIR/${tdir}-2 failed"
1327 multiop $DIR/${tdir}-1/f O_c &
1330 #define OBD_FAIL_MDS_REINT_NET_REP 0x119
1331 do_facet $SINGLEMDS "lctl set_param fail_loc=0x119"
1332 mcreate $DIR/${tdir}-2/f &
1336 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1337 kill -USR1 $close_pid
1338 cancel_lru_locks mdc # force the close
1339 wait $close_pid || error "close_pid failed"
1340 # open should still be here
1341 [ -d /proc/$open_pid ] || error "open_pid doesn't exists"
1343 replay_barrier_nodf $SINGLEMDS
1345 wait $open_pid || error "open_pid failed"
1347 $CHECKSTAT -t file $DIR/${tdir}-1/f ||
1348 error "$CHECKSTAT $DIR/${tdir}-1/f attribute check failed"
1349 $CHECKSTAT -t file $DIR/${tdir}-2/f ||
1350 error "$CHECKSTAT $DIR/${tdir}-2/f attribute check failed"
1351 rm -rf $DIR/${tdir}-*
1353 run_test 53e "|X| open reply while two MDC requests in flight"
1356 cancel_lru_locks mdc # cleanup locks from former test cases
1358 mkdir $DIR/${tdir}-1 || error "mkdir $DIR/${tdir}-1 failed"
1359 mkdir $DIR/${tdir}-2 || error "mkdir $DIR/${tdir}-2 failed"
1360 multiop $DIR/${tdir}-1/f O_c &
1363 #define OBD_FAIL_MDS_REINT_NET_REP 0x119
1364 do_facet $SINGLEMDS "lctl set_param fail_loc=0x119"
1365 mcreate $DIR/${tdir}-2/f &
1369 #define OBD_FAIL_MDS_CLOSE_NET_REP 0x13b
1370 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000013b"
1371 kill -USR1 $close_pid
1372 cancel_lru_locks mdc # force the close
1374 #bz20647: make sure all pids are exists before failover
1375 [ -d /proc/$close_pid ] || error "close_pid doesn't exist"
1376 [ -d /proc/$open_pid ] || error "open_pid doesn't exists"
1377 replay_barrier_nodf $SINGLEMDS
1378 fail_nodf $SINGLEMDS
1379 wait $open_pid || error "open_pid failed"
1381 # close should be gone
1382 [ -d /proc/$close_pid ] && error "close_pid should not exist"
1383 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1385 $CHECKSTAT -t file $DIR/${tdir}-1/f ||
1386 error "$CHECKSTAT $DIR/${tdir}-1/f attribute check failed"
1387 $CHECKSTAT -t file $DIR/${tdir}-2/f ||
1388 error "$CHECKSTAT $DIR/${tdir}-2/f attribute check failed"
1389 rm -rf $DIR/${tdir}-*
1391 run_test 53f "|X| open reply and close reply while two MDC requests in flight"
1394 cancel_lru_locks mdc # cleanup locks from former test cases
1396 mkdir $DIR/${tdir}-1 || error "mkdir $DIR/${tdir}-1 failed"
1397 mkdir $DIR/${tdir}-2 || error "mkdir $DIR/${tdir}-2 failed"
1398 multiop $DIR/${tdir}-1/f O_c &
1401 #define OBD_FAIL_MDS_REINT_NET_REP 0x119
1402 do_facet $SINGLEMDS "lctl set_param fail_loc=0x119"
1403 mcreate $DIR/${tdir}-2/f &
1407 #define OBD_FAIL_MDS_CLOSE_NET 0x115
1408 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
1409 kill -USR1 $close_pid
1410 cancel_lru_locks mdc # force the close
1411 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1413 #bz20647: make sure all pids are exists before failover
1414 [ -d /proc/$close_pid ] || error "close_pid doesn't exist"
1415 [ -d /proc/$open_pid ] || error "open_pid doesn't exists"
1416 replay_barrier_nodf $SINGLEMDS
1417 fail_nodf $SINGLEMDS
1418 wait $open_pid || error "open_pid failed"
1420 # close should be gone
1421 [ -d /proc/$close_pid ] && error "close_pid should not exist"
1423 $CHECKSTAT -t file $DIR/${tdir}-1/f ||
1424 error "$CHECKSTAT $DIR/${tdir}-1/f attribute check failed"
1425 $CHECKSTAT -t file $DIR/${tdir}-2/f ||
1426 error "$CHECKSTAT $DIR/${tdir}-2/f attribute check failed"
1427 rm -rf $DIR/${tdir}-*
1429 run_test 53g "|X| drop open reply and close request while close and open are both in flight"
1432 cancel_lru_locks mdc # cleanup locks from former test cases
1434 mkdir $DIR/${tdir}-1 || error "mkdir $DIR/${tdir}-1 failed"
1435 mkdir $DIR/${tdir}-2 || error "mkdir $DIR/${tdir}-2 failed"
1436 multiop $DIR/${tdir}-1/f O_c &
1439 #define OBD_FAIL_MDS_REINT_NET 0x107
1440 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000107"
1441 mcreate $DIR/${tdir}-2/f &
1445 #define OBD_FAIL_MDS_CLOSE_NET_REP 0x13b
1446 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000013b"
1447 kill -USR1 $close_pid
1448 cancel_lru_locks mdc # force the close
1451 #bz20647: make sure all pids are exists before failover
1452 [ -d /proc/$close_pid ] || error "close_pid doesn't exist"
1453 [ -d /proc/$open_pid ] || error "open_pid doesn't exists"
1454 replay_barrier_nodf $SINGLEMDS
1455 fail_nodf $SINGLEMDS
1456 wait $open_pid || error "open_pid failed"
1458 # close should be gone
1459 [ -d /proc/$close_pid ] && error "close_pid should not exist"
1460 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1462 $CHECKSTAT -t file $DIR/${tdir}-1/f ||
1463 error "$CHECKSTAT $DIR/${tdir}-1/f attribute check failed"
1464 $CHECKSTAT -t file $DIR/${tdir}-2/f ||
1465 error "$CHECKSTAT $DIR/${tdir}-2/f attribute check failed"
1466 rm -rf $DIR/${tdir}-*
1468 run_test 53h "open request and close reply while two MDC requests in flight"
1470 #b3761 ASSERTION(hash != 0) failed
1472 # OBD_FAIL_MDS_OPEN_CREATE | OBD_FAIL_ONCE
1473 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000012b"
1475 # give touch a chance to run
1477 do_facet $SINGLEMDS "lctl set_param fail_loc=0x0"
1481 run_test 55 "let MDS_CHECK_RESENT return the original return code instead of 0"
1483 #b3440 ASSERTION(rec->ur_fid2->id) failed
1485 ln -s foo $DIR/$tfile
1486 replay_barrier $SINGLEMDS
1487 #drop_reply "cat $DIR/$tfile"
1491 run_test 56 "don't replay a symlink open request (3440)"
1493 #recovery one mds-ost setattr from llog
1495 #define OBD_FAIL_MDS_OST_SETATTR 0x12c
1496 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000012c"
1497 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
1498 replay_barrier $SINGLEMDS
1501 $CHECKSTAT -t file $DIR/$tfile ||
1502 error "$CHECKSTAT $DIR/$tfile attribute check failed"
1503 do_facet $SINGLEMDS "lctl set_param fail_loc=0x0"
1506 run_test 57 "test recovery from llog for setattr op"
1509 zconf_umount $(hostname) $MOUNT2
1513 #recovery many mds-ost setattr from llog
1515 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
1516 #define OBD_FAIL_MDS_OST_SETATTR 0x12c
1517 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000012c"
1518 createmany -o $DIR/$tdir/$tfile-%d 2500
1519 replay_barrier $SINGLEMDS
1522 $CHECKSTAT -t file $DIR/$tdir/$tfile-* >/dev/null ||
1523 error "$CHECKSTAT $DIR/$tfile-* attribute check failed"
1524 do_facet $SINGLEMDS "lctl set_param fail_loc=0x0"
1525 unlinkmany $DIR/$tdir/$tfile-%d 2500 ||
1526 error "unlinkmany $DIR/$tfile failed"
1529 run_test 58a "test recovery from llog for setattr op (test llog_gen_rec)"
1535 trap cleanup_58 EXIT
1537 large_xattr_enabled &&
1538 orig="$(generate_string $(max_xattr_size))" || orig="bar"
1539 # Original extended attribute can be long. Print a small version of
1540 # attribute if an error occurs
1541 local sm_msg=$(printf "%.9s" $orig)
1543 mount_client $MOUNT2 || error "mount_client on $MOUNT2 failed"
1544 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
1545 touch $DIR/$tdir/$tfile || error "touch $DIR/$tdir/$tfile failed"
1546 replay_barrier $SINGLEMDS
1547 setfattr -n trusted.foo -v $orig $DIR/$tdir/$tfile
1549 new=$(get_xattr_value trusted.foo $MOUNT2/$tdir/$tfile)
1550 [[ "$new" = "$orig" ]] ||
1551 error "xattr set ($sm_msg...) differs from xattr get ($new)"
1552 rm -f $DIR/$tdir/$tfile
1555 wait_clients_import_state ${CLIENTS:-$HOSTNAME} "mgs" FULL
1557 run_test 58b "test replay of setxattr op"
1559 test_58c() { # bug 16570
1564 trap cleanup_58 EXIT
1566 if large_xattr_enabled; then
1567 local xattr_size=$(max_xattr_size)
1568 orig="$(generate_string $((xattr_size / 2)))"
1569 orig1="$(generate_string $xattr_size)"
1575 # PING_INTERVAL max(obd_timeout / 4, 1U)
1576 sleep $((TIMEOUT / 4))
1578 # Original extended attribute can be long. Print a small version of
1579 # attribute if an error occurs
1580 local sm_msg=$(printf "%.9s" $orig)
1581 local sm_msg1=$(printf "%.9s" $orig1)
1583 mount_client $MOUNT2 || error "mount_client on $MOUNT2 failed"
1584 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
1585 touch $DIR/$tdir/$tfile || error "touch $DIR/$tdir/$tfile failed"
1586 drop_request "setfattr -n trusted.foo -v $orig $DIR/$tdir/$tfile" ||
1587 error "drop_request for setfattr failed"
1588 new=$(get_xattr_value trusted.foo $MOUNT2/$tdir/$tfile)
1589 [[ "$new" = "$orig" ]] ||
1590 error "xattr set ($sm_msg...) differs from xattr get ($new)"
1591 drop_reint_reply "setfattr -n trusted.foo1 \
1592 -v $orig1 $DIR/$tdir/$tfile" ||
1593 error "drop_reint_reply for setfattr failed"
1594 new=$(get_xattr_value trusted.foo1 $MOUNT2/$tdir/$tfile)
1595 [[ "$new" = "$orig1" ]] ||
1596 error "second xattr set ($sm_msg1...) differs xattr get ($new)"
1597 rm -f $DIR/$tdir/$tfile
1601 run_test 58c "resend/reconstruct setxattr op"
1603 # log_commit_thread vs filter_destroy race used to lead to import use after free
1606 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1608 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
1609 createmany -o $DIR/$tdir/$tfile-%d 200 ||
1610 error "createmany create files failed"
1612 unlinkmany $DIR/$tdir/$tfile-%d 200 ||
1613 error "unlinkmany $DIR/$tdir/$tfile failed"
1614 #define OBD_FAIL_PTLRPC_DELAY_RECOV 0x507
1615 do_facet ost1 "lctl set_param fail_loc=0x507"
1618 do_facet ost1 "lctl set_param fail_loc=0x0"
1622 run_test 59 "test log_commit_thread vs filter_destroy race"
1624 # race between add unlink llog vs cat log init in post_recovery (only for b1_6)
1625 # bug 12086: should no oops and No ctxt error for this test
1627 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
1628 createmany -o $DIR/$tdir/$tfile-%d 200 ||
1629 error "createmany create files failed"
1630 replay_barrier $SINGLEMDS
1631 unlinkmany $DIR/$tdir/$tfile-%d 0 100
1633 unlinkmany $DIR/$tdir/$tfile-%d 100 100
1634 local no_ctxt=$(dmesg | grep "No ctxt")
1635 [ -z "$no_ctxt" ] || error "ctxt is not initialized in recovery"
1637 run_test 60 "test llog post recovery init vs llog unlink"
1639 #test race llog recovery thread vs llog cleanup
1640 test_61a() { # was test_61
1641 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1643 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
1644 createmany -o $DIR/$tdir/$tfile-%d 800 ||
1645 error "createmany create files failed"
1647 unlinkmany $DIR/$tdir/$tfile-%d 800
1648 # OBD_FAIL_OST_LLOG_RECOVERY_TIMEOUT 0x221
1649 set_nodes_failloc "$(osts_nodes)" 0x80000221
1654 set_nodes_failloc "$(osts_nodes)" 0x0
1656 $CHECKSTAT -t file $DIR/$tdir/$tfile-* &&
1657 error "$CHECKSTAT $DIR/$tdir/$tfile attribute check should fail"
1660 run_test 61a "test race llog recovery vs llog cleanup"
1662 #test race mds llog sync vs llog cleanup
1664 # OBD_FAIL_MDS_LLOG_SYNC_TIMEOUT 0x13a
1665 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000013a"
1666 facet_failover $SINGLEMDS
1669 do_facet client dd if=/dev/zero of=$DIR/$tfile bs=4k count=1 ||
1672 run_test 61b "test race mds llog sync vs llog cleanup"
1674 #test race cancel cookie cb vs llog cleanup
1676 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1678 # OBD_FAIL_OST_CANCEL_COOKIE_TIMEOUT 0x222
1679 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
1680 set_nodes_failloc "$(osts_nodes)" 0x80000222
1684 set_nodes_failloc "$(osts_nodes)" 0x0
1686 run_test 61c "test race mds llog sync vs llog cleanup"
1688 test_61d() { # bug 16002 # bug 17466 # bug 22137
1689 # OBD_FAIL_OBD_LLOG_SETUP 0x605
1691 do_facet mgs "lctl set_param fail_loc=0x80000605"
1692 start mgs $(mgsdevname) $MGS_MOUNT_OPTS &&
1693 error "mgs start should have failed"
1694 do_facet mgs "lctl set_param fail_loc=0"
1695 start mgs $(mgsdevname) $MGS_MOUNT_OPTS || error "cannot restart mgs"
1697 run_test 61d "error in llog_setup should cleanup the llog context correctly"
1699 test_62() { # Bug 15756 - don't mis-drop resent replay
1700 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
1701 replay_barrier $SINGLEMDS
1702 createmany -o $DIR/$tdir/$tfile- 25 ||
1703 error "createmany create files failed"
1704 #define OBD_FAIL_TGT_REPLAY_DROP 0x707
1705 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000707"
1707 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1708 unlinkmany $DIR/$tdir/$tfile- 25 ||
1709 error "unlinkmany $DIR/$tdir/$tfile failed"
1712 run_test 62 "don't mis-drop resent replay"
1714 #Adaptive Timeouts (bug 3055)
1722 echo "Cleaning up AT ..."
1723 if [ -n "$ATOLDBASE" ]; then
1724 local at_history=$($LCTL get_param -n at_history)
1725 do_facet $SINGLEMDS "lctl set_param at_history=$at_history" || true
1726 do_facet ost1 "lctl set_param at_history=$at_history" || true
1729 if [ $AT_MAX_SET -ne 0 ]; then
1730 for facet in mds client ost; do
1731 var=AT_MAX_SAVE_${facet}
1732 echo restore AT on $facet to saved value ${!var}
1733 at_max_set ${!var} $facet
1734 at_new=$(at_max_get $facet)
1735 echo Restored AT value on $facet $at_new
1736 [ $at_new -eq ${!var} ] ||
1737 error "AT value not restored SAVED ${!var} NEW $at_new"
1744 local at_max_new=600
1746 # Save at_max original values
1748 if [ $AT_MAX_SET -eq 0 ]; then
1749 # Suppose that all osts have the same at_max
1750 for facet in mds client ost; do
1751 eval AT_MAX_SAVE_${facet}=$(at_max_get $facet)
1755 for facet in mds client ost; do
1756 at_max=$(at_max_get $facet)
1757 if [ $at_max -ne $at_max_new ]; then
1758 echo "AT value on $facet is $at_max, set it by force temporarily to $at_max_new"
1759 at_max_set $at_max_new $facet
1764 if [ -z "$ATOLDBASE" ]; then
1765 ATOLDBASE=$(do_facet $SINGLEMDS "lctl get_param -n at_history")
1766 # speed up the timebase so we can check decreasing AT
1767 do_facet $SINGLEMDS "lctl set_param at_history=8" || true
1768 do_facet ost1 "lctl set_param at_history=8" || true
1770 # sleep for a while to cool down, should be > 8s and also allow
1771 # at least one ping to be sent. simply use TIMEOUT to be safe.
1776 test_65a() #bug 3055
1778 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1780 at_start || return 0
1781 $LCTL dk > /dev/null
1783 $LCTL set_param debug="other"
1784 # Slow down a request to the current service time, this is critical
1785 # because previous tests may have caused this value to increase.
1786 REQ_DELAY=`lctl get_param -n mdc.${FSNAME}-MDT0000-mdc-*.timeouts |
1787 awk '/portal 12/ {print $5}'`
1788 REQ_DELAY=$((${REQ_DELAY} + ${REQ_DELAY} / 4 + 5))
1790 do_facet $SINGLEMDS lctl set_param fail_val=$((${REQ_DELAY} * 1000))
1791 #define OBD_FAIL_PTLRPC_PAUSE_REQ 0x50a
1792 do_facet $SINGLEMDS $LCTL set_param fail_loc=0x8000050a
1793 createmany -o $DIR/$tfile 10 > /dev/null
1794 unlinkmany $DIR/$tfile 10 > /dev/null
1795 # check for log message
1796 $LCTL dk | grep "Early reply #" || error "No early reply"
1798 # client should show REQ_DELAY estimates
1799 lctl get_param -n mdc.${FSNAME}-MDT0000-mdc-*.timeouts | grep portal
1801 lctl get_param -n mdc.${FSNAME}-MDT0000-mdc-*.timeouts | grep portal
1803 run_test 65a "AT: verify early replies"
1805 test_65b() #bug 3055
1807 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1809 at_start || return 0
1812 $LCTL set_param debug="other trace"
1813 $LCTL dk > /dev/null
1814 # Slow down a request to the current service time, this is critical
1815 # because previous tests may have caused this value to increase.
1816 $SETSTRIPE --stripe-index=0 --count=1 $DIR/$tfile
1817 multiop $DIR/$tfile Ow1yc
1818 REQ_DELAY=`lctl get_param -n osc.${FSNAME}-OST0000-osc-*.timeouts |
1819 awk '/portal 6/ {print $5}'`
1820 REQ_DELAY=$((${REQ_DELAY} + ${REQ_DELAY} / 4 + 5))
1822 do_facet ost1 lctl set_param fail_val=${REQ_DELAY}
1823 #define OBD_FAIL_OST_BRW_PAUSE_PACK 0x224
1824 do_facet ost1 $LCTL set_param fail_loc=0x224
1827 $SETSTRIPE --stripe-index=0 --count=1 $DIR/$tfile
1828 # force some real bulk transfer
1829 multiop $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c
1831 do_facet ost1 $LCTL set_param fail_loc=0
1832 # check for log message
1833 $LCTL dk | grep "Early reply #" || error "No early reply"
1835 # client should show REQ_DELAY estimates
1836 lctl get_param -n osc.${FSNAME}-OST0000-osc-*.timeouts | grep portal
1838 run_test 65b "AT: verify early replies on packed reply / bulk"
1840 test_66a() #bug 3055
1842 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1844 at_start || return 0
1845 lctl get_param -n mdc.${FSNAME}-MDT0000-mdc-*.timeouts | grep "portal 12"
1846 # adjust 5s at a time so no early reply is sent (within deadline)
1847 do_facet $SINGLEMDS "$LCTL set_param fail_val=5000"
1848 #define OBD_FAIL_PTLRPC_PAUSE_REQ 0x50a
1849 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x8000050a"
1850 createmany -o $DIR/$tfile 20 > /dev/null
1851 unlinkmany $DIR/$tfile 20 > /dev/null
1852 lctl get_param -n mdc.${FSNAME}-MDT0000-mdc-*.timeouts | grep "portal 12"
1853 do_facet $SINGLEMDS "$LCTL set_param fail_val=10000"
1854 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x8000050a"
1855 createmany -o $DIR/$tfile 20 > /dev/null
1856 unlinkmany $DIR/$tfile 20 > /dev/null
1857 lctl get_param -n mdc.${FSNAME}-MDT0000-mdc-*.timeouts | grep "portal 12"
1858 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0"
1860 createmany -o $DIR/$tfile 20 > /dev/null
1861 unlinkmany $DIR/$tfile 20 > /dev/null
1862 lctl get_param -n mdc.${FSNAME}-MDT0000-mdc-*.timeouts | grep "portal 12"
1863 CUR=$(lctl get_param -n mdc.${FSNAME}-MDT0000-mdc-*.timeouts | awk '/portal 12/ {print $5}')
1864 WORST=$(lctl get_param -n mdc.${FSNAME}-MDT0000-mdc-*.timeouts | awk '/portal 12/ {print $7}')
1865 echo "Current MDT timeout $CUR, worst $WORST"
1866 [ $CUR -lt $WORST ] || error "Current $CUR should be less than worst $WORST"
1868 run_test 66a "AT: verify MDT service time adjusts with no early replies"
1870 test_66b() #bug 3055
1872 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1874 at_start || return 0
1875 ORIG=$(lctl get_param -n mdc.${FSNAME}-MDT0000*.timeouts |
1876 awk '/network/ {print $4}')
1877 $LCTL set_param fail_val=$(($ORIG + 5))
1878 #define OBD_FAIL_PTLRPC_PAUSE_REP 0x50c
1879 $LCTL set_param fail_loc=0x50c
1880 ls $DIR/$tfile > /dev/null 2>&1
1881 $LCTL set_param fail_loc=0
1882 CUR=$(lctl get_param -n mdc.${FSNAME}-MDT0000*.timeouts |
1883 awk '/network/ {print $4}')
1884 WORST=$(lctl get_param -n mdc.${FSNAME}-MDT0000*.timeouts |
1885 awk '/network/ {print $6}')
1886 echo "network timeout orig $ORIG, cur $CUR, worst $WORST"
1887 [ $WORST -gt $ORIG ] ||
1888 error "Worst $WORST should be worse than orig $ORIG"
1890 run_test 66b "AT: verify net latency adjusts"
1892 test_67a() #bug 3055
1894 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1896 at_start || return 0
1897 CONN1=$(lctl get_param -n osc.*.stats | awk '/_connect/ {total+=$2} END {print total}')
1898 # sleeping threads may drive values above this
1899 do_facet ost1 "$LCTL set_param fail_val=400"
1900 #define OBD_FAIL_PTLRPC_PAUSE_REQ 0x50a
1901 do_facet ost1 "$LCTL set_param fail_loc=0x50a"
1902 createmany -o $DIR/$tfile 20 > /dev/null
1903 unlinkmany $DIR/$tfile 20 > /dev/null
1904 do_facet ost1 "$LCTL set_param fail_loc=0"
1905 CONN2=$(lctl get_param -n osc.*.stats | awk '/_connect/ {total+=$2} END {print total}')
1906 ATTEMPTS=$(($CONN2 - $CONN1))
1907 echo "$ATTEMPTS osc reconnect attempts on gradual slow"
1908 [ $ATTEMPTS -gt 0 ] &&
1909 error_ignore bz13721 "AT should have prevented reconnect"
1912 run_test 67a "AT: verify slow request processing doesn't induce reconnects"
1914 test_67b() #bug 3055
1916 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1918 at_start || return 0
1919 CONN1=$(lctl get_param -n osc.*.stats | awk '/_connect/ {total+=$2} END {print total}')
1921 # exhaust precreations on ost1
1922 local OST=$(ostname_from_index 0)
1923 local mdtosc=$(get_mdtosc_proc_path mds $OST)
1924 local last_id=$(do_facet $SINGLEMDS lctl get_param -n \
1925 osc.$mdtosc.prealloc_last_id)
1926 local next_id=$(do_facet $SINGLEMDS lctl get_param -n \
1927 osc.$mdtosc.prealloc_next_id)
1929 mkdir -p $DIR/$tdir/${OST} || error "mkdir $DIR/$tdir/${OST} failed"
1930 $SETSTRIPE -i 0 -c 1 $DIR/$tdir/${OST} || error "$SETSTRIPE failed"
1931 echo "Creating to objid $last_id on ost $OST..."
1932 #define OBD_FAIL_OST_PAUSE_CREATE 0x223
1933 do_facet ost1 "$LCTL set_param fail_val=20000"
1934 do_facet ost1 "$LCTL set_param fail_loc=0x80000223"
1935 createmany -o $DIR/$tdir/${OST}/f $next_id $((last_id - next_id + 2))
1938 do_facet ost1 "lctl get_param -n ost.OSS.ost_create.timeouts"
1940 CONN2=$(lctl get_param -n osc.*.stats | awk '/_connect/ {total+=$2} END {print total}')
1941 ATTEMPTS=$(($CONN2 - $CONN1))
1942 echo "$ATTEMPTS osc reconnect attempts on instant slow"
1943 # do it again; should not timeout
1944 do_facet ost1 "$LCTL set_param fail_loc=0x80000223"
1945 cp /etc/profile $DIR/$tfile || error "cp failed"
1946 do_facet ost1 "$LCTL set_param fail_loc=0"
1948 do_facet ost1 "lctl get_param -n ost.OSS.ost_create.timeouts"
1949 CONN3=$(lctl get_param -n osc.*.stats | awk '/_connect/ {total+=$2} END {print total}')
1950 ATTEMPTS=$(($CONN3 - $CONN2))
1951 echo "$ATTEMPTS osc reconnect attempts on 2nd slow"
1952 [ $ATTEMPTS -gt 0 ] && error "AT should have prevented reconnect"
1955 run_test 67b "AT: verify instant slowdown doesn't induce reconnects"
1957 test_68 () #bug 13813
1959 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1961 at_start || return 0
1962 local ldlm_enqueue_min=$(find /sys -name ldlm_enqueue_min)
1963 [ -z "$ldlm_enqueue_min" ] && skip "missing /sys/.../ldlm_enqueue_min" && return 0
1964 local ldlm_enqueue_min_r=$(do_facet ost1 "find /sys -name ldlm_enqueue_min")
1965 [ -z "$ldlm_enqueue_min_r" ] && skip "missing /sys/.../ldlm_enqueue_min in the ost1" && return 0
1966 local ENQ_MIN=$(cat $ldlm_enqueue_min)
1967 local ENQ_MIN_R=$(do_facet ost1 "cat $ldlm_enqueue_min_r")
1968 echo $TIMEOUT >> $ldlm_enqueue_min
1969 do_facet ost1 "echo $TIMEOUT >> $ldlm_enqueue_min_r"
1971 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
1972 $SETSTRIPE --stripe-index=0 --count=1 $DIR/$tdir
1973 #define OBD_FAIL_LDLM_PAUSE_CANCEL 0x312
1974 $LCTL set_param fail_val=$(($TIMEOUT - 1))
1975 $LCTL set_param fail_loc=0x80000312
1976 cp /etc/profile $DIR/$tdir/${tfile}_1 || error "1st cp failed $?"
1977 $LCTL set_param fail_val=$((TIMEOUT * 5 / 4))
1978 $LCTL set_param fail_loc=0x80000312
1979 cp /etc/profile $DIR/$tdir/${tfile}_2 || error "2nd cp failed $?"
1980 $LCTL set_param fail_loc=0
1982 echo $ENQ_MIN >> $ldlm_enqueue_min
1983 do_facet ost1 "echo $ENQ_MIN_R >> $ldlm_enqueue_min_r"
1987 run_test 68 "AT: verify slowing locks"
1990 # end of AT tests includes above lines
1992 # start multi-client tests
1994 [ -z "$CLIENTS" ] &&
1995 { skip "Need two or more clients." && return; }
1996 [ $CLIENTCOUNT -lt 2 ] &&
1997 { skip "Need two or more clients, have $CLIENTCOUNT" && return; }
1999 echo "mount clients $CLIENTS ..."
2000 zconf_mount_clients $CLIENTS $MOUNT
2002 local clients=${CLIENTS//,/ }
2003 echo "Write/read files on $DIR ; clients $CLIENTS ... "
2004 for CLIENT in $clients; do
2005 do_node $CLIENT dd bs=1M count=10 if=/dev/zero \
2006 of=$DIR/${tfile}_${CLIENT} 2>/dev/null ||
2007 error "dd failed on $CLIENT"
2010 local prev_client=$(echo $clients | sed 's/^.* \(.\+\)$/\1/')
2011 for C in ${CLIENTS//,/ }; do
2012 do_node $prev_client dd if=$DIR/${tfile}_${C} \
2013 of=/dev/null 2>/dev/null ||
2014 error "dd if=$DIR/${tfile}_${C} failed on $prev_client"
2020 run_test 70a "check multi client t-f"
2022 check_for_process () {
2027 killall_process $clients "$prog" -0
2030 killall_process () {
2031 local clients=${1:-$(hostname)}
2036 do_nodes $clients "killall $signal $name"
2040 local clients=${CLIENTS:-$HOSTNAME}
2042 zconf_mount_clients $clients $MOUNT
2045 [ "$SLOW" = "no" ] && duration=120
2046 # set duration to 900 because it takes some time to boot node
2047 [ "$FAILURE_MODE" = HARD ] && duration=900
2050 local start_ts=$(date +%s)
2051 local cmd="rundbench 1 -t $duration"
2053 if [ $MDSCOUNT -ge 2 ]; then
2054 test_mkdir -p -c$MDSCOUNT $DIR/$tdir
2055 $LFS setdirstripe -D -c$MDSCOUNT $DIR/$tdir
2057 do_nodesv $clients "set -x; MISSING_DBENCH_OK=$MISSING_DBENCH_OK \
2058 PATH=\$PATH:$LUSTRE/utils:$LUSTRE/tests/:$DBENCH_LIB \
2059 DBENCH_LIB=$DBENCH_LIB TESTSUITE=$TESTSUITE TESTNAME=$TESTNAME \
2060 MOUNT=$MOUNT DIR=$DIR/$tdir/\\\$(hostname) LCTL=$LCTL $cmd" &
2063 #LU-1897 wait for all dbench copies to start
2064 while ! check_for_process $clients dbench; do
2065 elapsed=$(($(date +%s) - start_ts))
2066 if [ $elapsed -gt $duration ]; then
2067 killall_process $clients dbench
2068 error "dbench failed to start on $clients!"
2073 log "Started rundbench load pid=$pid ..."
2075 elapsed=$(($(date +%s) - start_ts))
2076 local num_failovers=0
2078 while [ $elapsed -lt $duration ]; do
2079 if ! check_for_process $clients dbench; then
2080 error_noexit "dbench stopped on some of $clients!"
2081 killall_process $clients dbench
2085 replay_barrier mds$fail_index
2086 sleep 1 # give clients a time to do operations
2087 # Increment the number of failovers
2088 num_failovers=$((num_failovers+1))
2089 log "$TESTNAME fail mds$fail_index $num_failovers times"
2091 elapsed=$(($(date +%s) - start_ts))
2092 if [ $fail_index -ge $MDSCOUNT ]; then
2095 fail_index=$((fail_index+1))
2099 wait $pid || error "rundbench load on $clients failed!"
2101 run_test 70b "dbench ${MDSCOUNT}mdts recovery; $CLIENTCOUNT clients"
2102 # end multi-client tests
2105 multiop_bg_pause $DIR/$tfile O_tSc ||
2106 error "multiop_bg_pause $DIR/$tfile failed"
2110 replay_barrier $SINGLEMDS
2111 #define OBD_FAIL_LDLM_ENQUEUE_NET 0x302
2112 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000302"
2115 wait $pid || error "multiop pid failed"
2116 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
2119 run_test 73a "open(O_CREAT), unlink, replay, reconnect before open replay, close"
2122 multiop_bg_pause $DIR/$tfile O_tSc ||
2123 error "multiop_bg_pause $DIR/$tfile failed"
2127 replay_barrier $SINGLEMDS
2128 #define OBD_FAIL_LDLM_REPLY 0x30c
2129 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000030c"
2132 wait $pid || error "multiop pid failed"
2133 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
2136 run_test 73b "open(O_CREAT), unlink, replay, reconnect at open_replay reply, close"
2140 local clients=${CLIENTS:-$HOSTNAME}
2142 zconf_umount_clients $clients $MOUNT
2144 facet_failover $SINGLEMDS
2145 zconf_mount_clients $clients $MOUNT
2147 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2148 rm $DIR/$tfile || error "rm $DIR/$tfile failed"
2149 clients_up || error "client evicted: $?"
2152 run_test 74 "Ensure applications don't fail waiting for OST recovery"
2154 remote_dir_check_80() {
2159 diridx=$($GETSTRIPE -M $remote_dir) ||
2160 error "$GETSTRIPE -M $remote_dir failed"
2161 [ $diridx -eq $MDTIDX ] || error "$diridx != $MDTIDX"
2163 createmany -o $remote_dir/f-%d 20 || error "creation failed"
2164 fileidx=$($GETSTRIPE -M $remote_dir/f-1) ||
2165 error "$GETSTRIPE -M $remote_dir/f-1 failed"
2166 [ $fileidx -eq $MDTIDX ] || error "$fileidx != $MDTIDX"
2172 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2173 ([ $FAILURE_MODE == "HARD" ] &&
2174 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2175 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2179 local remote_dir=$DIR/$tdir/remote_dir
2181 mkdir -p $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2182 #define OBD_FAIL_OUT_UPDATE_NET_REP 0x1701
2183 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
2184 $LFS mkdir -i $MDTIDX $remote_dir &
2190 wait $CLIENT_PID || error "remote creation failed"
2192 remote_dir_check_80 || error "remote dir check failed"
2193 rm -rf $DIR/$tdir || error "rmdir failed"
2197 run_test 80a "DNE: create remote dir, drop update rep from MDT0, fail MDT0"
2200 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2201 ([ $FAILURE_MODE == "HARD" ] &&
2202 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2203 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2207 local remote_dir=$DIR/$tdir/remote_dir
2209 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2210 #define OBD_FAIL_UPDATE_OBJ_NET_REP 0x1701
2211 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
2212 $LFS mkdir -i $MDTIDX $remote_dir &
2217 fail mds$((MDTIDX + 1))
2219 wait $CLIENT_PID || error "remote creation failed"
2221 remote_dir_check_80 || error "remote dir check failed"
2222 rm -rf $DIR/$tdir || error "rmdir failed"
2226 run_test 80b "DNE: create remote dir, drop update rep from MDT0, fail MDT1"
2229 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2230 ([ $FAILURE_MODE == "HARD" ] &&
2231 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2232 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2236 local remote_dir=$DIR/$tdir/remote_dir
2238 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2239 #define OBD_FAIL_UPDATE_OBJ_NET_REP 0x1701
2240 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
2241 $LFS mkdir -i $MDTIDX $remote_dir &
2247 fail mds$((MDTIDX + 1))
2249 wait $CLIENT_PID || error "remote creation failed"
2251 remote_dir_check_80 || error "remote dir check failed"
2252 rm -rf $DIR/$tdir || error "rmdir failed"
2256 run_test 80c "DNE: create remote dir, drop update rep from MDT1, fail MDT[0,1]"
2259 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2261 local remote_dir=$DIR/$tdir/remote_dir
2263 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2264 #define OBD_FAIL_UPDATE_OBJ_NET_REP 0x1701
2265 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
2266 $LFS mkdir -i $MDTIDX $remote_dir &
2269 # sleep 3 seconds to make sure MDTs are failed after
2270 # lfs mkdir -i has finished on all of MDTs.
2275 fail mds${MDTIDX},mds$((MDTIDX + 1))
2277 wait $CLIENT_PID || error "remote creation failed"
2279 remote_dir_check_80 || error "remote dir check failed"
2280 rm -rf $DIR/$tdir || error "rmdir failed"
2284 run_test 80d "DNE: create remote dir, drop update rep from MDT1, fail 2 MDTs"
2287 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2288 ([ $FAILURE_MODE == "HARD" ] &&
2289 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2290 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2294 local remote_dir=$DIR/$tdir/remote_dir
2296 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2297 # OBD_FAIL_MDS_REINT_NET_REP 0x119
2298 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
2299 $LFS mkdir -i $MDTIDX $remote_dir &
2302 # sleep 3 seconds to make sure MDTs are failed after
2303 # lfs mkdir -i has finished on all of MDTs.
2309 wait $CLIENT_PID || error "remote creation failed"
2311 remote_dir_check_80 || error "remote dir check failed"
2312 rm -rf $DIR/$tdir || error "rmdir failed"
2316 run_test 80e "DNE: create remote dir, drop MDT1 rep, fail MDT0"
2319 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2320 ([ $FAILURE_MODE == "HARD" ] &&
2321 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2322 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2325 local remote_dir=$DIR/$tdir/remote_dir
2327 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2328 # OBD_FAIL_MDS_REINT_NET_REP 0x119
2329 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
2330 $LFS mkdir -i $MDTIDX $remote_dir &
2334 fail mds$((MDTIDX + 1))
2336 wait $CLIENT_PID || error "remote creation failed"
2338 remote_dir_check_80 || error "remote dir check failed"
2339 rm -rf $DIR/$tdir || error "rmdir failed"
2343 run_test 80f "DNE: create remote dir, drop MDT1 rep, fail MDT1"
2346 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2347 ([ $FAILURE_MODE == "HARD" ] &&
2348 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2349 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2353 local remote_dir=$DIR/$tdir/remote_dir
2355 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2356 # OBD_FAIL_MDS_REINT_NET_REP 0x119
2357 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
2358 $LFS mkdir -i $MDTIDX $remote_dir &
2361 # sleep 3 seconds to make sure MDTs are failed after
2362 # lfs mkdir -i has finished on all of MDTs.
2368 fail mds$((MDTIDX + 1))
2370 wait $CLIENT_PID || error "remote creation failed"
2372 remote_dir_check_80 || error "remote dir check failed"
2373 rm -rf $DIR/$tdir || error "rmdir failed"
2377 run_test 80g "DNE: create remote dir, drop MDT1 rep, fail MDT0, then MDT1"
2380 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2382 local remote_dir=$DIR/$tdir/remote_dir
2384 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2385 # OBD_FAIL_MDS_REINT_NET_REP 0x119
2386 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
2387 $LFS mkdir -i $MDTIDX $remote_dir &
2390 # sleep 3 seconds to make sure MDTs are failed after
2391 # lfs mkdir -i has finished on all of MDTs.
2396 fail mds${MDTIDX},mds$((MDTIDX + 1))
2398 wait $CLIENT_PID || error "remote dir creation failed"
2400 remote_dir_check_80 || error "remote dir check failed"
2401 rm -rf $DIR/$tdir || error "rmdir failed"
2405 run_test 80h "DNE: create remote dir, drop MDT1 rep, fail 2 MDTs"
2408 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2409 ([ $FAILURE_MODE == "HARD" ] &&
2410 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2411 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2415 local remote_dir=$DIR/$tdir/remote_dir
2417 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2418 $LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
2420 touch $remote_dir || error "touch $remote_dir failed"
2421 # OBD_FAIL_OBJ_UPDATE_NET_REP 0x1701
2422 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
2427 fail mds$((MDTIDX + 1))
2429 wait $CLIENT_PID || error "rm remote dir failed"
2431 stat $remote_dir 2&>/dev/null && error "$remote_dir still exist!"
2433 rm -rf $DIR/$tdir || error "rmdir failed"
2437 run_test 81a "DNE: unlink remote dir, drop MDT0 update rep, fail MDT1"
2440 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2441 ([ $FAILURE_MODE == "HARD" ] &&
2442 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2443 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2446 local remote_dir=$DIR/$tdir/remote_dir
2448 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2449 $LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
2451 # OBD_FAIL_OBJ_UPDATE_NET_REP 0x1701
2452 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
2459 wait $CLIENT_PID || error "rm remote dir failed"
2461 stat $remote_dir 2&>/dev/null && error "$remote_dir still exist!"
2463 rm -rf $DIR/$tdir || error "rmdir failed"
2467 run_test 81b "DNE: unlink remote dir, drop MDT0 update reply, fail MDT0"
2470 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2471 ([ $FAILURE_MODE == "HARD" ] &&
2472 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2473 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2477 local remote_dir=$DIR/$tdir/remote_dir
2479 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2480 $LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
2482 # OBD_FAIL_OBJ_UPDATE_NET_REP 0x1701
2483 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
2490 fail mds$((MDTIDX + 1))
2492 wait $CLIENT_PID || error "rm remote dir failed"
2494 stat $remote_dir 2&>/dev/null && error "$remote_dir still exist!"
2496 rm -rf $DIR/$tdir || error "rmdir failed"
2500 run_test 81c "DNE: unlink remote dir, drop MDT0 update reply, fail MDT0,MDT1"
2503 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2505 local remote_dir=$DIR/$tdir/remote_dir
2507 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2508 $LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
2510 # OBD_FAIL_OBJ_UPDATE_NET_REP 0x1701
2511 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
2517 fail mds${MDTIDX},mds$((MDTIDX + 1))
2519 wait $CLIENT_PID || error "rm remote dir failed"
2521 stat $remote_dir 2&>/dev/null && error "$remote_dir still exist!"
2523 rm -rf $DIR/$tdir || error "rmdir failed"
2527 run_test 81d "DNE: unlink remote dir, drop MDT0 update reply, fail 2 MDTs"
2530 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2531 ([ $FAILURE_MODE == "HARD" ] &&
2532 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2533 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2537 local remote_dir=$DIR/$tdir/remote_dir
2539 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2540 $LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
2542 # OBD_FAIL_MDS_REINT_NET_REP 0x119
2543 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
2546 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0
2551 wait $CLIENT_PID || error "rm remote dir failed"
2553 stat $remote_dir 2&>/dev/null && error "$remote_dir still exist!"
2555 rm -rf $DIR/$tdir || error "rmdir failed"
2559 run_test 81e "DNE: unlink remote dir, drop MDT1 req reply, fail MDT0"
2562 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2563 ([ $FAILURE_MODE == "HARD" ] &&
2564 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2565 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2569 local remote_dir=$DIR/$tdir/remote_dir
2571 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2572 $LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
2574 # OBD_FAIL_MDS_REINT_NET_REP 0x119
2575 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
2580 fail mds$((MDTIDX + 1))
2582 wait $CLIENT_PID || error "rm remote dir failed"
2584 stat $remote_dir 2&>/dev/null && error "$remote_dir still exist!"
2586 rm -rf $DIR/$tdir || error "rmdir failed"
2590 run_test 81f "DNE: unlink remote dir, drop MDT1 req reply, fail MDT1"
2593 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2594 ([ $FAILURE_MODE == "HARD" ] &&
2595 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2596 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2600 local remote_dir=$DIR/$tdir/remote_dir
2602 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2603 $LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
2605 # OBD_FAIL_MDS_REINT_NET_REP 0x119
2606 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
2613 fail mds$((MDTIDX + 1))
2615 wait $CLIENT_PID || error "rm remote dir failed"
2617 stat $remote_dir 2&>/dev/null && error "$remote_dir still exist!"
2619 rm -rf $DIR/$tdir || error "rmdir failed"
2623 run_test 81g "DNE: unlink remote dir, drop req reply, fail M0, then M1"
2626 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2628 local remote_dir=$DIR/$tdir/remote_dir
2630 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2631 $LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
2633 # OBD_FAIL_MDS_REINT_NET_REP 0x119
2634 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
2640 fail mds${MDTIDX},mds$((MDTIDX + 1))
2642 wait $CLIENT_PID || error "rm remote dir failed"
2644 stat $remote_dir 2&>/dev/null && error "$remote_dir still exist!"
2646 rm -rf $DIR/$tdir || error "rmdir failed"
2650 run_test 81h "DNE: unlink remote dir, drop request reply, fail 2 MDTs"
2653 #define OBD_FAIL_MDS_OPEN_WAIT_CREATE 0x144
2654 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000144"
2655 createmany -o $DIR/$tfile- 1 &
2659 client_up || client_up || true # reconnect
2661 run_test 84a "stale open during export disconnect"
2663 test_85a() { #bug 16774
2664 lctl set_param -n ldlm.cancel_unused_locks_before_replay "1"
2666 for i in $(seq 100); do
2667 echo "tag-$i" > $DIR/$tfile-$i
2668 grep -q "tag-$i" $DIR/$tfile-$i || error "f2-$i"
2671 lov_id=$(lctl dl | grep "clilov")
2672 addr=$(echo $lov_id | awk '{print $4}' | awk -F '-' '{print $3}')
2673 count=$(lctl get_param -n \
2674 ldlm.namespaces.*MDT0000*$addr.lock_unused_count)
2675 echo "before recovery: unused locks count = $count"
2679 count2=$(lctl get_param -n \
2680 ldlm.namespaces.*MDT0000*$addr.lock_unused_count)
2681 echo "after recovery: unused locks count = $count2"
2683 if [ $count2 -ge $count ]; then
2684 error "unused locks are not canceled"
2687 run_test 85a "check the cancellation of unused locks during recovery(IBITS)"
2689 test_85b() { #bug 16774
2690 lctl set_param -n ldlm.cancel_unused_locks_before_replay "1"
2692 do_facet mgs $LCTL pool_new $FSNAME.$TESTNAME ||
2693 error "unable to create pool $TESTNAME"
2694 do_facet mgs $LCTL pool_add $FSNAME.$TESTNAME $FSNAME-OST0000 ||
2695 error "unable to add pool $TESTNAME"
2697 $SETSTRIPE -c 1 -p $FSNAME.$TESTNAME $DIR
2699 for i in $(seq 100); do
2700 dd if=/dev/urandom of=$DIR/$tfile-$i bs=4096 \
2701 count=32 >/dev/null 2>&1
2704 cancel_lru_locks osc
2706 for i in $(seq 100); do
2707 dd if=$DIR/$tfile-$i of=/dev/null bs=4096 \
2708 count=32 >/dev/null 2>&1
2711 lov_id=$(lctl dl | grep "clilov")
2712 addr=$(echo $lov_id | awk '{print $4}' | awk -F '-' '{print $3}')
2713 count=$(lctl get_param \
2714 -n ldlm.namespaces.*OST0000*$addr.lock_unused_count)
2715 echo "before recovery: unused locks count = $count"
2716 [ $count != 0 ] || error "unused locks ($count) should be zero"
2720 count2=$(lctl get_param \
2721 -n ldlm.namespaces.*OST0000*$addr.lock_unused_count)
2722 echo "after recovery: unused locks count = $count2"
2724 do_facet mgs $LCTL pool_remove $FSNAME.$TESTNAME $FSNAME-OST0000 ||
2725 error "unable to remove pool $TESTNAME"
2726 do_facet mgs $LCTL pool_destroy $FSNAME.$TESTNAME ||
2727 error "unable to destroy the pool $TESTNAME"
2729 if [ $count2 -ge $count ]; then
2730 error "unused locks are not canceled"
2733 run_test 85b "check the cancellation of unused locks during recovery(EXTENT)"
2736 local clients=${CLIENTS:-$HOSTNAME}
2738 zconf_umount_clients $clients $MOUNT
2739 do_facet $SINGLEMDS lctl set_param mdt.${FSNAME}-MDT*.exports.clear=0
2740 remount_facet $SINGLEMDS
2741 zconf_mount_clients $clients $MOUNT
2743 run_test 86 "umount server after clear nid_stats should not hit LBUG"
2746 do_facet ost1 "lctl set_param -n obdfilter.${ost1_svc}.sync_journal 0"
2749 $SETSTRIPE -i 0 -c 1 $DIR/$tfile
2750 dd if=/dev/urandom of=$DIR/$tfile bs=1024k count=8 ||
2751 error "dd to $DIR/$tfile failed"
2752 cksum=$(md5sum $DIR/$tfile | awk '{print $1}')
2753 cancel_lru_locks osc
2755 dd if=$DIR/$tfile of=/dev/null bs=1024k count=8 || error "Cannot read"
2756 cksum2=$(md5sum $DIR/$tfile | awk '{print $1}')
2757 if [ $cksum != $cksum2 ] ; then
2758 error "New checksum $cksum2 does not match original $cksum"
2761 run_test 87 "write replay"
2764 do_facet ost1 "lctl set_param -n obdfilter.${ost1_svc}.sync_journal 0"
2767 $SETSTRIPE -i 0 -c 1 $DIR/$tfile
2768 dd if=/dev/urandom of=$DIR/$tfile bs=1024k count=8 ||
2769 error "dd to $DIR/$tfile failed"
2770 sleep 1 # Give it a chance to flush dirty data
2771 echo TESTTEST | dd of=$DIR/$tfile bs=1 count=8 seek=64
2772 cksum=$(md5sum $DIR/$tfile | awk '{print $1}')
2773 cancel_lru_locks osc
2775 dd if=$DIR/$tfile of=/dev/null bs=1024k count=8 || error "Cannot read"
2776 cksum2=$(md5sum $DIR/$tfile | awk '{print $1}')
2777 if [ $cksum != $cksum2 ] ; then
2778 error "New checksum $cksum2 does not match original $cksum"
2781 run_test 87b "write replay with changed data (checksum resend)"
2783 test_88() { #bug 17485
2784 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2785 mkdir -p $TMP/$tdir || error "mkdir $TMP/$tdir failed"
2787 $SETSTRIPE -i 0 -c 1 $DIR/$tdir || error "$SETSTRIPE"
2790 replay_barrier $SINGLEMDS
2792 # exhaust precreations on ost1
2793 local OST=$(ostname_from_index 0)
2794 local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS $OST)
2795 local last_id=$(do_facet $SINGLEMDS lctl get_param -n osc.$mdtosc.prealloc_last_id)
2796 local next_id=$(do_facet $SINGLEMDS lctl get_param -n osc.$mdtosc.prealloc_next_id)
2797 echo "before test: last_id = $last_id, next_id = $next_id"
2799 echo "Creating to objid $last_id on ost $OST..."
2800 createmany -o $DIR/$tdir/f-%d $next_id $((last_id - next_id + 2)) ||
2801 error "createmany create files to last_id failed"
2803 #create some files to use some uncommitted objids
2804 last_id=$(($last_id + 1))
2805 createmany -o $DIR/$tdir/f-%d $last_id 8 ||
2806 error "createmany create files with uncommitted objids failed"
2808 last_id2=$(do_facet $SINGLEMDS lctl get_param -n osc.$mdtosc.prealloc_last_id)
2809 next_id2=$(do_facet $SINGLEMDS lctl get_param -n osc.$mdtosc.prealloc_next_id)
2810 echo "before recovery: last_id = $last_id2, next_id = $next_id2"
2812 # if test uses shutdown_facet && reboot_facet instead of facet_failover ()
2813 # it has to take care about the affected facets, bug20407
2814 local affected_mds1=$(affected_facets mds1)
2815 local affected_ost1=$(affected_facets ost1)
2817 shutdown_facet $SINGLEMDS
2820 reboot_facet $SINGLEMDS
2821 change_active $affected_mds1
2822 wait_for_facet $affected_mds1
2823 mount_facets $affected_mds1 || error "Restart of mds failed"
2826 change_active $affected_ost1
2827 wait_for_facet $affected_ost1
2828 mount_facets $affected_ost1 || error "Restart of ost1 failed"
2832 last_id2=$(do_facet $SINGLEMDS lctl get_param -n osc.$mdtosc.prealloc_last_id)
2833 next_id2=$(do_facet $SINGLEMDS lctl get_param -n osc.$mdtosc.prealloc_next_id)
2834 echo "after recovery: last_id = $last_id2, next_id = $next_id2"
2836 # create new files, which should use new objids, and ensure the orphan
2837 # cleanup phase for ost1 is completed at the same time
2838 for i in $(seq 8); do
2839 file_id=$(($last_id + 10 + $i))
2840 dd if=/dev/urandom of=$DIR/$tdir/f-$file_id bs=4096 count=128
2843 # if the objids were not recreated, then "ls" will fail with -ENOENT
2844 ls -l $DIR/$tdir/* || error "can't get the status of precreated files"
2847 # write into previously created files
2848 for i in $(seq 8); do
2849 file_id=$(($last_id + $i))
2850 dd if=/dev/urandom of=$DIR/$tdir/f-$file_id bs=4096 count=128
2851 cp -f $DIR/$tdir/f-$file_id $TMP/$tdir/
2854 # compare the content
2855 for i in $(seq 8); do
2856 file_id=$(($last_id + $i))
2857 cmp $TMP/$tdir/f-$file_id $DIR/$tdir/f-$file_id ||
2858 error "the content of file is modified!"
2863 run_test 88 "MDS should not assign same objid to different files "
2866 cancel_lru_locks osc
2867 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2868 rm -f $DIR/$tdir/$tfile
2870 wait_delete_completed
2871 BLOCKS1=$(df -P $MOUNT | tail -n 1 | awk '{ print $3 }')
2872 $SETSTRIPE -i 0 -c 1 $DIR/$tdir/$tfile
2873 dd if=/dev/zero bs=1M count=10 of=$DIR/$tdir/$tfile
2876 facet_failover $SINGLEMDS
2877 rm $DIR/$tdir/$tfile
2880 zconf_mount $(hostname) $MOUNT || error "mount fails"
2881 client_up || error "client_up failed"
2883 wait_delete_completed
2884 BLOCKS2=$(df -P $MOUNT | tail -n 1 | awk '{ print $3 }')
2885 [ $((BLOCKS2 - BLOCKS1)) -le 4 ] ||
2886 error $((BLOCKS2 - BLOCKS1)) blocks leaked
2889 run_test 89 "no disk space leak on late ost connection"
2895 change_active $facet
2896 wait_for_facet $facet
2897 mount_facet $facet || error "Restart of $facet failed"
2901 test_90() { # bug 19494
2902 local dir=$DIR/$tdir
2903 local ostfail=$(get_random_entry $(get_facets OST))
2905 if [[ $FAILURE_MODE = HARD ]]; then
2906 local affected=$(affected_facets $ostfail);
2907 if [[ "$affected" != $ostfail ]]; then
2908 skip not functional with FAILURE_MODE=$FAILURE_MODE, affected: $affected
2913 mkdir $dir || error "mkdir $dir failed"
2915 echo "Create the files"
2917 # file "f${index}" striped over 1 OST
2918 # file "all" striped over all OSTs
2920 $SETSTRIPE -c $OSTCOUNT $dir/all ||
2921 error "setstripe failed to create $dir/all"
2923 for (( i=0; i<$OSTCOUNT; i++ )); do
2925 $SETSTRIPE -i $i -c 1 $f || error "$SETSTRIPE failed to create $f"
2927 # confirm setstripe actually created the stripe on the requested OST
2928 local uuid=$(ostuuid_from_index $i)
2929 for file in f$i all; do
2930 if [[ $dir/$file != $($LFS find --obd $uuid --name $file $dir) ]]; then
2931 $GETSTRIPE $dir/$file
2932 error wrong stripe: $file, uuid: $uuid
2937 # Before failing an OST, get its obd name and index
2938 local varsvc=${ostfail}_svc
2939 local obd=$(do_facet $ostfail lctl get_param \
2940 -n obdfilter.${!varsvc}.uuid)
2941 local index=$(($(facet_number $ostfail) - 1))
2943 echo "Fail $ostfail $obd, display the list of affected files"
2944 shutdown_facet $ostfail || error "shutdown_facet $ostfail failed"
2946 trap "cleanup_90 $ostfail" EXIT INT
2947 echo "General Query: lfs find $dir"
2948 local list=$($LFS find $dir)
2950 for (( i=0; i<$OSTCOUNT; i++ )); do
2951 list_member "$list" $dir/f$i ||
2952 error_noexit "lfs find $dir: no file f$i"
2954 list_member "$list" $dir/all ||
2955 error_noexit "lfs find $dir: no file all"
2957 # focus on the missing OST,
2958 # we expect to see only two files affected: "f$(index)" and "all"
2960 echo "Querying files on shutdown $ostfail: lfs find --obd $obd"
2961 list=$($LFS find --obd $obd $dir)
2963 for file in all f$index; do
2964 list_member "$list" $dir/$file ||
2965 error_noexit "lfs find does not report the affected $obd for $file"
2968 [[ $(echo $list | wc -w) -eq 2 ]] ||
2969 error_noexit "lfs find reports the wrong list of affected files ${#list[@]}"
2971 echo "Check getstripe: $GETSTRIPE -r --obd $obd"
2972 list=$($GETSTRIPE -r --obd $obd $dir)
2974 for file in all f$index; do
2975 echo "$list" | grep $dir/$file ||
2976 error_noexit "lfs getsripe does not report the affected $obd for $file"
2981 run_test 90 "lfs find identifies the missing striped file segments"
2984 local server_version=$(lustre_version_code $SINGLEMDS)
2985 [[ $server_version -ge $(version_code 2.6.90) ]] ||
2986 [[ $server_version -ge $(version_code 2.5.4) &&
2987 $server_version -lt $(version_code 2.5.50) ]] ||
2988 { skip "Need MDS version 2.5.4+ or 2.6.90+"; return; }
2990 cancel_lru_locks osc
2992 $SETSTRIPE -i 0 -c 1 $DIR/$tfile ||
2993 error "$SETSTRIPE $DIR/$tfile failed"
2994 dd if=/dev/zero of=$DIR/$tfile bs=1024 count=1 ||
2995 error "dd to $DIR/$tfile failed"
2996 #define OBD_FAIL_TGT_REPLAY_RECONNECT 0x715
2997 # We need to emulate a state that OST is waiting for other clients
2998 # not completing the recovery. Final ping is queued, but reply will be
2999 # sent on the recovery completion. It is done by sleep before
3000 # processing final pings
3001 do_facet ost1 "$LCTL set_param fail_val=40"
3002 do_facet ost1 "$LCTL set_param fail_loc=0x715"
3005 run_test 93 "replay + reconnect"
3007 striped_dir_check_100() {
3008 local striped_dir=$DIR/$tdir/striped_dir
3009 local stripe_count=$($LFS getdirstripe -c $striped_dir)
3011 $LFS getdirstripe $striped_dir
3012 [ $stripe_count -eq 2 ] || error "$stripe_count != 2"
3014 createmany -o $striped_dir/f-%d 20 ||
3015 error "creation failed under striped dir"
3019 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3020 ([ $FAILURE_MODE == "HARD" ] &&
3021 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3022 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3025 local striped_dir=$DIR/$tdir/striped_dir
3028 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3030 #To make sure MDT1 and MDT0 are connected
3031 #otherwise it may create single stripe dir here
3032 $LFS setdirstripe -i1 $DIR/$tdir/remote_dir
3034 #define OBD_FAIL_OUT_UPDATE_NET_REP 0x1701
3035 do_facet mds$((MDTIDX+1)) lctl set_param fail_loc=0x1701
3036 $LFS setdirstripe -i0 -c2 $striped_dir &
3039 fail mds$((MDTIDX + 1))
3041 wait $CLIENT_PID || error "striped dir creation failed"
3043 striped_dir_check_100 || error "striped dir check failed"
3044 rm -rf $DIR/$tdir || error "rmdir failed"
3046 run_test 100a "DNE: create striped dir, drop update rep from MDT1, fail MDT1"
3049 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3050 ([ $FAILURE_MODE == "HARD" ] &&
3051 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3052 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3055 local striped_dir=$DIR/$tdir/striped_dir
3058 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3060 #To make sure MDT1 and MDT0 are connected
3061 #otherwise it may create single stripe dir here
3062 $LFS setdirstripe -i1 $DIR/$tdir/remote_dir
3064 # OBD_FAIL_MDS_REINT_NET_REP 0x119
3065 do_facet mds$MDTIDX lctl set_param fail_loc=0x119
3066 $LFS mkdir -i0 -c2 $striped_dir &
3071 wait $CLIENT_PID || error "striped dir creation failed"
3073 striped_dir_check_100 || error "striped dir check failed"
3074 rm -rf $DIR/$tdir || error "rmdir failed"
3076 run_test 100b "DNE: create striped dir, fail MDT0"
3078 test_101() { #LU-5648
3079 mkdir -p $DIR/$tdir/d1
3080 mkdir -p $DIR/$tdir/d2
3081 touch $DIR/$tdir/file0
3084 replay_barrier $SINGLEMDS
3085 for i in $(seq $num) ; do
3086 echo test$i > $DIR/$tdir/d1/file$i
3089 fail_abort $SINGLEMDS
3090 for i in $(seq $num) ; do
3091 touch $DIR/$tdir/d2/file$i
3092 test -s $DIR/$tdir/d2/file$i &&
3093 ls -al $DIR/$tdir/d2/file$i && error "file$i's size > 0"
3098 run_test 101 "Shouldn't reassign precreated objs to other files after recovery"
3100 check_striped_dir_110()
3102 $CHECKSTAT -t dir $DIR/$tdir/striped_dir ||
3103 error "create striped dir failed"
3104 local stripe_count=$($LFS getdirstripe -c $DIR/$tdir/striped_dir)
3105 [ $stripe_count -eq $MDSCOUNT ] ||
3106 error "$stripe_count != 2 after recovery"
3110 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3111 ([ $FAILURE_MODE == "HARD" ] &&
3112 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3113 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3118 $LFS mkdir -i1 -c$MDSCOUNT $DIR/$tdir/striped_dir
3121 check_striped_dir_110 || error "check striped_dir failed"
3122 rm -rf $DIR/$tdir || error "rmdir failed"
3126 run_test 110a "DNE: create striped dir, fail MDT1"
3129 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3130 ([ $FAILURE_MODE == "HARD" ] &&
3131 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3132 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3137 $LFS mkdir -i1 -c$MDSCOUNT $DIR/$tdir/striped_dir
3140 zconf_mount $(hostname) $MOUNT
3141 client_up || return 1
3143 check_striped_dir_110 || error "check striped_dir failed"
3145 rm -rf $DIR/$tdir || error "rmdir failed"
3149 run_test 110b "DNE: create striped dir, fail MDT1 and client"
3152 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3153 ([ $FAILURE_MODE == "HARD" ] &&
3154 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3155 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3160 $LFS mkdir -i1 -c$MDSCOUNT $DIR/$tdir/striped_dir
3163 check_striped_dir_110 || error "check striped_dir failed"
3165 rm -rf $DIR/$tdir || error "rmdir failed"
3169 run_test 110c "DNE: create striped dir, fail MDT2"
3172 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3173 ([ $FAILURE_MODE == "HARD" ] &&
3174 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3175 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3180 $LFS mkdir -i1 -c$MDSCOUNT $DIR/$tdir/striped_dir
3183 zconf_mount $(hostname) $MOUNT
3184 client_up || return 1
3186 check_striped_dir_110 || error "check striped_dir failed"
3188 rm -rf $DIR/$tdir || error "rmdir failed"
3192 run_test 110d "DNE: create striped dir, fail MDT2 and client"
3195 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3196 ([ $FAILURE_MODE == "HARD" ] &&
3197 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3198 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3203 $LFS mkdir -i1 -c$MDSCOUNT $DIR/$tdir/striped_dir
3207 zconf_mount $(hostname) $MOUNT
3208 client_up || return 1
3210 check_striped_dir_110 || error "check striped_dir failed"
3212 rm -rf $DIR/$tdir || error "rmdir failed"
3216 run_test 110e "DNE: create striped dir, uncommit on MDT2, fail client/MDT1/MDT2"
3219 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3220 ([ $FAILURE_MODE == "HARD" ] &&
3221 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3222 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3228 $LFS mkdir -i1 -c$MDSCOUNT $DIR/$tdir/striped_dir
3231 check_striped_dir_110 || error "check striped_dir failed"
3233 rm -rf $DIR/$tdir || error "rmdir failed"
3237 run_test 110f "DNE: create striped dir, fail MDT1/MDT2"
3240 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3241 ([ $FAILURE_MODE == "HARD" ] &&
3242 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3243 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3248 $LFS mkdir -i1 -c$MDSCOUNT $DIR/$tdir/striped_dir
3252 zconf_mount $(hostname) $MOUNT
3253 client_up || return 1
3255 check_striped_dir_110 || error "check striped_dir failed"
3257 rm -rf $DIR/$tdir || error "rmdir failed"
3261 run_test 110g "DNE: create striped dir, uncommit on MDT1, fail client/MDT1/MDT2"
3264 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3265 ([ $FAILURE_MODE == "HARD" ] &&
3266 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3267 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3271 $LFS mkdir -i1 -c2 $DIR/$tdir/striped_dir
3273 rm -rf $DIR/$tdir/striped_dir
3276 $CHECKSTAT -t dir $DIR/$tdir/striped_dir &&
3277 error "striped dir still exists"
3280 run_test 111a "DNE: unlink striped dir, fail MDT1"
3283 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3284 ([ $FAILURE_MODE == "HARD" ] &&
3285 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3286 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3290 $LFS mkdir -i1 -c2 $DIR/$tdir/striped_dir
3292 rm -rf $DIR/$tdir/striped_dir
3295 zconf_mount $(hostname) $MOUNT
3296 client_up || return 1
3298 $CHECKSTAT -t dir $DIR/$tdir/striped_dir &&
3299 error "striped dir still exists"
3302 run_test 111b "DNE: unlink striped dir, fail MDT2"
3305 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3306 ([ $FAILURE_MODE == "HARD" ] &&
3307 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3308 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3312 $LFS mkdir -i1 -c2 $DIR/$tdir/striped_dir
3314 rm -rf $DIR/$tdir/striped_dir
3318 zconf_mount $(hostname) $MOUNT
3319 client_up || return 1
3320 $CHECKSTAT -t dir $DIR/$tdir/striped_dir &&
3321 error "striped dir still exists"
3324 run_test 111c "DNE: unlink striped dir, uncommit on MDT1, fail client/MDT1/MDT2"
3327 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3328 ([ $FAILURE_MODE == "HARD" ] &&
3329 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3330 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3334 $LFS mkdir -i1 -c2 $DIR/$tdir/striped_dir
3336 rm -rf $DIR/$tdir/striped_dir
3340 zconf_mount $(hostname) $MOUNT
3341 client_up || return 1
3342 $CHECKSTAT -t dir $DIR/$tdir/striped_dir &&
3343 error "striped dir still exists"
3347 run_test 111d "DNE: unlink striped dir, uncommit on MDT2, fail client/MDT1/MDT2"
3350 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3351 ([ $FAILURE_MODE == "HARD" ] &&
3352 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3353 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3357 $LFS mkdir -i1 -c2 $DIR/$tdir/striped_dir
3359 rm -rf $DIR/$tdir/striped_dir
3362 $CHECKSTAT -t dir $DIR/$tdir/striped_dir &&
3363 error "striped dir still exists"
3366 run_test 111e "DNE: unlink striped dir, uncommit on MDT2, fail MDT1/MDT2"
3369 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3370 ([ $FAILURE_MODE == "HARD" ] &&
3371 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3372 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3376 $LFS mkdir -i1 -c2 $DIR/$tdir/striped_dir
3378 rm -rf $DIR/$tdir/striped_dir
3381 $CHECKSTAT -t dir $DIR/$tdir/striped_dir &&
3382 error "striped dir still exists"
3385 run_test 111f "DNE: unlink striped dir, uncommit on MDT1, fail MDT1/MDT2"
3388 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3389 ([ $FAILURE_MODE == "HARD" ] &&
3390 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3391 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3395 $LFS mkdir -i1 -c2 $DIR/$tdir/striped_dir
3398 rm -rf $DIR/$tdir/striped_dir
3400 $CHECKSTAT -t dir $DIR/$tdir/striped_dir &&
3401 error "striped dir still exists"
3404 run_test 111g "DNE: unlink striped dir, fail MDT1/MDT2"
3406 test_112_rename_prepare() {
3407 mkdir -p $DIR/$tdir/src_dir
3408 $LFS mkdir -i 1 $DIR/$tdir/src_dir/src_child ||
3409 error "create remote source failed"
3411 touch $DIR/$tdir/src_dir/src_child/a
3413 $LFS mkdir -i 2 $DIR/$tdir/tgt_dir ||
3414 error "create remote target dir failed"
3416 $LFS mkdir -i 3 $DIR/$tdir/tgt_dir/tgt_child ||
3417 error "create remote target child failed"
3422 $CHECKSTAT -t dir $DIR/$tdir/src_dir/src_child &&
3423 error "src_child still exists after rename"
3425 $CHECKSTAT -t file $DIR/$tdir/tgt_dir/tgt_child/a ||
3426 error "missing file(a) after rename"
3430 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
3431 ([ $FAILURE_MODE == "HARD" ] &&
3432 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3433 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3436 test_112_rename_prepare
3439 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
3440 error "rename dir cross MDT failed!"
3444 rm -rf $DIR/$tdir || error "rmdir failed"
3446 run_test 112a "DNE: cross MDT rename, fail MDT1"
3449 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
3450 ([ $FAILURE_MODE == "HARD" ] &&
3451 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3452 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3455 test_112_rename_prepare
3458 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
3459 error "rename dir cross MDT failed!"
3464 rm -rf $DIR/$tdir || error "rmdir failed"
3466 run_test 112b "DNE: cross MDT rename, fail MDT2"
3469 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
3470 ([ $FAILURE_MODE == "HARD" ] &&
3471 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3472 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3475 test_112_rename_prepare
3478 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
3479 error "rename dir cross MDT failed!"
3484 rm -rf $DIR/$tdir || error "rmdir failed"
3486 run_test 112c "DNE: cross MDT rename, fail MDT3"
3489 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
3490 ([ $FAILURE_MODE == "HARD" ] &&
3491 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3492 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3495 test_112_rename_prepare
3498 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
3499 error "rename dir cross MDT failed!"
3504 rm -rf $DIR/$tdir || error "rmdir failed"
3506 run_test 112d "DNE: cross MDT rename, fail MDT4"
3509 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
3510 ([ $FAILURE_MODE == "HARD" ] &&
3511 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3512 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3515 test_112_rename_prepare
3519 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
3520 error "rename dir cross MDT failed!"
3525 rm -rf $DIR/$tdir || error "rmdir failed"
3527 run_test 112e "DNE: cross MDT rename, fail MDT1 and MDT2"
3530 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
3531 ([ $FAILURE_MODE == "HARD" ] &&
3532 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3533 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3536 test_112_rename_prepare
3540 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
3541 error "rename dir cross MDT failed!"
3546 rm -rf $DIR/$tdir || error "rmdir failed"
3548 run_test 112f "DNE: cross MDT rename, fail MDT1 and MDT3"
3551 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
3552 ([ $FAILURE_MODE == "HARD" ] &&
3553 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3554 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3557 test_112_rename_prepare
3561 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
3562 error "rename dir cross MDT failed!"
3567 rm -rf $DIR/$tdir || error "rmdir failed"
3569 run_test 112g "DNE: cross MDT rename, fail MDT1 and MDT4"
3572 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
3573 ([ $FAILURE_MODE == "HARD" ] &&
3574 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3575 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3578 test_112_rename_prepare
3582 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
3583 error "rename dir cross MDT failed!"
3588 rm -rf $DIR/$tdir || error "rmdir failed"
3590 run_test 112h "DNE: cross MDT rename, fail MDT2 and MDT3"
3593 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
3594 ([ $FAILURE_MODE == "HARD" ] &&
3595 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3596 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3599 test_112_rename_prepare
3603 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
3604 error "rename dir cross MDT failed!"
3609 rm -rf $DIR/$tdir || error "rmdir failed"
3611 run_test 112i "DNE: cross MDT rename, fail MDT2 and MDT4"
3614 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
3615 ([ $FAILURE_MODE == "HARD" ] &&
3616 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3617 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3620 test_112_rename_prepare
3624 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
3625 error "rename dir cross MDT failed!"
3630 rm -rf $DIR/$tdir || error "rmdir failed"
3632 run_test 112j "DNE: cross MDT rename, fail MDT3 and MDT4"
3635 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
3636 ([ $FAILURE_MODE == "HARD" ] &&
3637 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3638 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3641 test_112_rename_prepare
3646 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
3647 error "rename dir cross MDT failed!"
3652 rm -rf $DIR/$tdir || error "rmdir failed"
3654 run_test 112k "DNE: cross MDT rename, fail MDT1,MDT2,MDT3"
3657 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
3658 ([ $FAILURE_MODE == "HARD" ] &&
3659 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3660 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3663 test_112_rename_prepare
3668 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
3669 error "rename dir cross MDT failed!"
3674 rm -rf $DIR/$tdir || error "rmdir failed"
3676 run_test 112l "DNE: cross MDT rename, fail MDT1,MDT2,MDT4"
3679 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
3680 ([ $FAILURE_MODE == "HARD" ] &&
3681 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3682 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3685 test_112_rename_prepare
3690 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
3691 error "rename dir cross MDT failed!"
3696 rm -rf $DIR/$tdir || error "rmdir failed"
3698 run_test 112m "DNE: cross MDT rename, fail MDT1,MDT3,MDT4"
3701 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
3702 ([ $FAILURE_MODE == "HARD" ] &&
3703 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3704 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3707 test_112_rename_prepare
3712 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
3713 error "rename dir cross MDT failed!"
3718 rm -rf $DIR/$tdir || error "rmdir failed"
3720 run_test 112n "DNE: cross MDT rename, fail MDT2,MDT3,MDT4"
3723 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3724 ([ $FAILURE_MODE == "HARD" ] &&
3725 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3726 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3734 for ((j=0;j<$((MDSCOUNT));j++)); do
3735 fail_index=$((fail_index+1))
3736 index=$((fail_index % MDSCOUNT))
3737 replay_barrier mds$((index + 1))
3738 for ((i=0;i<5;i++)); do
3739 test_mkdir -i$index -c$MDSCOUNT $DIR/$tdir/test_$i ||
3740 error "create striped dir $DIR/$tdir/test_$i"
3743 fail mds$((index + 1))
3744 for ((i=0;i<5;i++)); do
3745 checkstat -t dir $DIR/$tdir/test_$i ||
3746 error "$DIR/$tdir/test_$i does not exist!"
3748 rm -rf $DIR/$tdir/test_* ||
3752 run_test 115 "failover for create/unlink striped directory"
3755 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3756 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] &&
3757 skip "Do not support large update log before 2.7.55" &&
3759 ([ $FAILURE_MODE == "HARD" ] &&
3760 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3761 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3768 # OBD_FAIL_SPLIT_UPDATE_REC 0x1702
3769 do_facet mds1 "lctl set_param fail_loc=0x80001702"
3770 $LFS setdirstripe -c$MDSCOUNT $DIR/$tdir/striped_dir
3773 $CHECKSTAT -t dir $DIR/$tdir/striped_dir ||
3774 error "stried_dir does not exists"
3776 run_test 116a "large update log master MDT recovery"
3779 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3780 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] &&
3781 skip "Do not support large update log before 2.7.55" &&
3784 ([ $FAILURE_MODE == "HARD" ] &&
3785 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3786 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3793 # OBD_FAIL_SPLIT_UPDATE_REC 0x1702
3794 do_facet mds2 "lctl set_param fail_loc=0x80001702"
3795 $LFS setdirstripe -c$MDSCOUNT $DIR/$tdir/striped_dir
3798 $CHECKSTAT -t dir $DIR/$tdir/striped_dir ||
3799 error "stried_dir does not exists"
3801 run_test 116b "large update log slave MDT recovery"
3805 check_and_cleanup_lustre