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=${GRANT_CHECK_LIST:-""}
21 require_dsh_mds || exit 0
24 # bug number for skipped tests:
26 ALWAYS_EXCEPT="$REPLAY_SINGLE_EXCEPT 61d"
27 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
30 [ "$SLOW" = "no" ] && EXCEPT_SLOW="44b"
32 [ $(facet_fstype $SINGLEMDS) = "zfs" ] &&
33 # bug number for skipped test: LU-1867 LU-3127
34 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 89 73b"
38 check_and_setup_lustre
43 rm -rf $DIR/[df][0-9]* $DIR/f.$TESTSUITE.*
45 # LU-482 Avert LVM and VM inability to flush caches in pre .33 kernels
46 if [ $LINUX_VERSION_CODE -lt $(version_code 2.6.33) ]; then
48 do_facet $SINGLEMDS sync
51 test_0a() { # was test_0
52 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
53 replay_barrier $SINGLEMDS
57 run_test 0a "empty replay"
60 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
62 # this test attempts to trigger a race in the precreation code,
63 # and must run before any other objects are created on the filesystem
65 createmany -o $DIR/$tfile 20 || error "createmany -o $DIR/$tfile failed"
66 unlinkmany $DIR/$tfile 20 || error "unlinkmany $DIR/$tfile failed"
68 run_test 0b "ensure object created after recover exists. (3284)"
71 replay_barrier $SINGLEMDS
74 facet_failover $SINGLEMDS
75 zconf_mount $(hostname) $MOUNT || error "mount fails"
76 client_up || error "post-failover df failed"
77 # file shouldn't exist if replay-barrier works as expected
78 rm $DIR/$tfile && error "File exists and it shouldn't"
81 run_test 0c "check replay-barrier"
84 replay_barrier $SINGLEMDS
86 facet_failover $SINGLEMDS
87 zconf_mount $(hostname) $MOUNT || error "mount fails"
88 client_up || error "post-failover df failed"
90 run_test 0d "expired recovery with no clients"
93 replay_barrier $SINGLEMDS
96 $CHECKSTAT -t file $DIR/$tfile ||
97 error "$CHECKSTAT $DIR/$tfile attribute check failed"
100 run_test 1 "simple create"
103 replay_barrier $SINGLEMDS
106 $CHECKSTAT -t file $DIR/$tfile ||
107 error "$CHECKSTAT $DIR/$tfile attribute check failed"
113 mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed"
114 replay_barrier $SINGLEMDS
117 $CHECKSTAT -t file $DIR/$tfile ||
118 error "$CHECKSTAT $DIR/$tfile attribute check failed"
124 local file=$DIR/$tfile
125 replay_barrier $SINGLEMDS
127 openfile -f O_DIRECTORY $file
129 $CHECKSTAT -t file $file ||
130 error "$CHECKSTAT $file attribute check failed"
133 run_test 3a "replay failed open(O_DIRECTORY)"
136 replay_barrier $SINGLEMDS
137 #define OBD_FAIL_MDS_OPEN_PACK | OBD_FAIL_ONCE
138 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000114"
140 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
142 $CHECKSTAT -t file $DIR/$tfile &&
143 error "$CHECKSTAT $DIR/$tfile attribute check should fail"
146 run_test 3b "replay failed open -ENOMEM"
149 replay_barrier $SINGLEMDS
150 #define OBD_FAIL_MDS_ALLOC_OBDO | OBD_FAIL_ONCE
151 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000128"
153 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
156 $CHECKSTAT -t file $DIR/$tfile &&
157 error "$CHECKSTAT $DIR/$tfile attribute check should fail"
160 run_test 3c "replay failed open -ENOMEM"
162 test_4a() { # was test_4
163 replay_barrier $SINGLEMDS
164 for i in $(seq 10); do
165 echo "tag-$i" > $DIR/$tfile-$i
168 for i in $(seq 10); do
169 grep -q "tag-$i" $DIR/$tfile-$i || error "$tfile-$i"
172 run_test 4a "|x| 10 open(O_CREAT)s"
175 for i in $(seq 10); do
176 echo "tag-$i" > $DIR/$tfile-$i
178 replay_barrier $SINGLEMDS
181 $CHECKSTAT -t file $DIR/$tfile-* &&
182 error "$CHECKSTAT $DIR/$tfile-* attribute check should fail" ||
185 run_test 4b "|x| rm 10 files"
187 # The idea is to get past the first block of precreated files on both
188 # osts, and then replay.
190 replay_barrier $SINGLEMDS
191 for i in $(seq 220); do
192 echo "tag-$i" > $DIR/$tfile-$i
195 for i in $(seq 220); do
196 grep -q "tag-$i" $DIR/$tfile-$i || error "$tfile-$i"
200 # waiting for commitment of removal
202 run_test 5 "|x| 220 open(O_CREAT)"
204 test_6a() { # was test_6
205 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
206 replay_barrier $SINGLEMDS
207 mcreate $DIR/$tdir/$tfile
209 $CHECKSTAT -t dir $DIR/$tdir ||
210 error "$CHECKSTAT $DIR/$tdir attribute check failed"
211 $CHECKSTAT -t file $DIR/$tdir/$tfile ||
212 error "$CHECKSTAT $DIR/$tdir/$tfile attribute check failed"
214 # waiting for log process thread
216 run_test 6a "mkdir + contained create"
219 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
220 replay_barrier $SINGLEMDS
223 $CHECKSTAT -t dir $DIR/$tdir &&
224 error "$CHECKSTAT $DIR/$tdir attribute check should fail" ||
227 run_test 6b "|X| rmdir"
230 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
231 replay_barrier $SINGLEMDS
232 mcreate $DIR/$tdir/$tfile
234 $CHECKSTAT -t dir $DIR/$tdir ||
235 error "$CHECKSTAT $DIR/$tdir attribute check failed"
236 $CHECKSTAT -t file $DIR/$tdir/$tfile ||
237 error "$CHECKSTAT $DIR/$tdir/$tfile attribute check failed"
240 run_test 7 "mkdir |X| contained create"
243 replay_barrier $SINGLEMDS
244 multiop_bg_pause $DIR/$tfile mo_c ||
245 error "multiop mknod $DIR/$tfile failed"
249 $CHECKSTAT -t file $DIR/$tfile ||
250 error "$CHECKSTAT $DIR/$tfile attribute check failed"
251 kill -USR1 $MULTIPID || error "multiop mknod $MULTIPID not running"
252 wait $MULTIPID || error "multiop mknod $MULTIPID failed"
255 run_test 8 "creat open |X| close"
258 replay_barrier $SINGLEMDS
260 local old_inum=$(ls -i $DIR/$tfile | awk '{print $1}')
262 local new_inum=$(ls -i $DIR/$tfile | awk '{print $1}')
264 echo " old_inum == $old_inum, new_inum == $new_inum"
265 if [ $old_inum -eq $new_inum ] ;
267 echo "old_inum and new_inum match"
269 echo " old_inum and new_inum do not match"
270 error "old index($old_inum) does not match new index($new_inum)"
274 run_test 9 "|X| create (same inum/gen)"
277 mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed"
278 replay_barrier $SINGLEMDS
279 mv $DIR/$tfile $DIR/$tfile-2
282 $CHECKSTAT $DIR/$tfile &&
283 error "$CHECKSTAT $DIR/$tfile attribute check should fail"
284 $CHECKSTAT $DIR/$tfile-2 ||
285 error "$CHECKSTAT $DIR/$tfile-2 attribute check failed"
289 run_test 10 "create |X| rename unlink"
292 mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed"
293 echo "old" > $DIR/$tfile
294 mv $DIR/$tfile $DIR/$tfile-2
295 replay_barrier $SINGLEMDS
296 echo "new" > $DIR/$tfile
298 grep old $DIR/$tfile-2
300 grep new $DIR/$tfile || error "grep $DIR/$tfile failed"
301 grep old $DIR/$tfile-2 || error "grep $DIR/$tfile-2 failed"
303 run_test 11 "create open write rename |X| create-old-name read"
306 mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed"
307 multiop_bg_pause $DIR/$tfile o_tSc ||
308 error "multiop_bg_pause $DIR/$tfile failed"
311 replay_barrier $SINGLEMDS
312 kill -USR1 $pid || error "multiop $pid not running"
313 wait $pid || error "multiop $pid failed"
316 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
319 run_test 12 "open, unlink |X| close"
321 # 1777 - replay open after committed chmod that would make
322 # a regular open a failure
324 mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed"
325 multiop_bg_pause $DIR/$tfile O_wc ||
326 error "multiop_bg_pause $DIR/$tfile failed"
329 $CHECKSTAT -p 0 $DIR/$tfile ||
330 error "$CHECKSTAT $DIR/$tfile attribute check failed"
331 replay_barrier $SINGLEMDS
333 kill -USR1 $pid || error "multiop $pid not running"
334 wait $pid || error "multiop $pid failed"
336 $CHECKSTAT -s 1 -p 0 $DIR/$tfile ||
337 error "second $CHECKSTAT $DIR/$tfile attribute check failed"
338 rm $DIR/$tfile || error "rm $DIR/$tfile failed"
341 run_test 13 "open chmod 0 |x| write close"
344 multiop_bg_pause $DIR/$tfile O_tSc ||
345 error "multiop_bg_pause $DIR/$tfile failed"
348 replay_barrier $SINGLEMDS
349 kill -USR1 $pid || error "multiop $pid not running"
350 wait $pid || error "multiop $pid failed"
353 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
356 run_test 14 "open(O_CREAT), unlink |X| close"
359 multiop_bg_pause $DIR/$tfile O_tSc ||
360 error "multiop_bg_pause $DIR/$tfile failed"
363 replay_barrier $SINGLEMDS
364 touch $DIR/$tfile-1 || error "touch $DIR/$tfile-1 failed"
365 kill -USR1 $pid || error "multiop $pid not running"
366 wait $pid || error "multiop $pid failed"
369 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
370 touch $DIR/$tfile-2 || error "touch $DIR/$tfile-2 failed"
373 run_test 15 "open(O_CREAT), unlink |X| touch new, close"
376 replay_barrier $SINGLEMDS
379 mcreate $DIR/$tfile-2
381 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
382 [ -e $DIR/$tfile-2 ] || error "file $DIR/$tfile-2 does not exist"
383 munlink $DIR/$tfile-2 || error "munlink $DIR/$tfile-2 failed"
385 run_test 16 "|X| open(O_CREAT), unlink, touch new, unlink new"
388 replay_barrier $SINGLEMDS
389 multiop_bg_pause $DIR/$tfile O_c ||
390 error "multiop_bg_pause $DIR/$tfile failed"
393 kill -USR1 $pid || error "multiop $pid not running"
394 wait $pid || error "multiop $pid failed"
395 $CHECKSTAT -t file $DIR/$tfile ||
396 error "$CHECKSTAT $DIR/$tfile attribute check failed"
399 run_test 17 "|X| open(O_CREAT), |replay| close"
402 replay_barrier $SINGLEMDS
403 multiop_bg_pause $DIR/$tfile O_tSc ||
404 error "multiop_bg_pause $DIR/$tfile failed"
407 touch $DIR/$tfile-2 || error "touch $DIR/$tfile-2 failed"
408 echo "pid: $pid will close"
409 kill -USR1 $pid || error "multiop $pid not running"
410 wait $pid || error "multiop $pid failed"
413 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
414 [ -e $DIR/$tfile-2 ] || error "file $DIR/$tfile-2 does not exist"
415 # this touch frequently fails
416 touch $DIR/$tfile-3 || error "touch $DIR/$tfile-3 failed"
417 munlink $DIR/$tfile-2 || error "munlink $DIR/$tfile-2 failed"
418 munlink $DIR/$tfile-3 || error "munlink $DIR/$tfile-3 failed"
421 run_test 18 "open(O_CREAT), unlink, touch new, close, touch, unlink"
423 # bug 1855 (a simpler form of test_11 above)
425 replay_barrier $SINGLEMDS
427 echo "old" > $DIR/$tfile
428 mv $DIR/$tfile $DIR/$tfile-2
429 grep old $DIR/$tfile-2
431 grep old $DIR/$tfile-2 || error "grep $DIR/$tfile-2 failed"
433 run_test 19 "mcreate, open, write, rename "
435 test_20a() { # was test_20
436 replay_barrier $SINGLEMDS
437 multiop_bg_pause $DIR/$tfile O_tSc ||
438 error "multiop_bg_pause $DIR/$tfile failed"
443 kill -USR1 $pid || error "multiop $pid not running"
444 wait $pid || error "multiop $pid failed"
445 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
448 run_test 20a "|X| open(O_CREAT), unlink, replay, close (test mds_cleanup_orphans)"
450 test_20b() { # bug 10480
451 local wait_timeout=$((TIMEOUT * 4))
455 BEFOREUSED=$(df -P $DIR | tail -1 | awk '{ print $3 }')
456 dd if=/dev/zero of=$DIR/$tfile bs=4k count=10000 &
457 while [ ! -e $DIR/$tfile ] ; do
458 usleep 60 # give dd a chance to start
461 $GETSTRIPE $DIR/$tfile || error "$GETSTRIPE $DIR/$tfile failed"
463 rm -f $DIR/$tfile || error "rm -f $DIR/$tfile failed"
465 client_up || client_up || true # reconnect
467 do_facet $SINGLEMDS "lctl set_param -n osd*.*MDT*.force_sync 1"
469 fail $SINGLEMDS # start orphan recovery
470 wait_recovery_complete $SINGLEMDS || error "MDS recovery not done"
471 wait_delete_completed_mds $wait_timeout ||
472 error "delete did not complete"
474 AFTERUSED=$(df -P $DIR | tail -1 | awk '{ print $3 }')
475 log "before $BEFOREUSED, after $AFTERUSED"
476 (( $AFTERUSED > $BEFOREUSED + $(fs_log_size) )) &&
477 error "after $AFTERUSED > before $BEFOREUSED"
480 run_test 20b "write, unlink, eviction, replay, (test mds_cleanup_orphans)"
482 test_20c() { # bug 10480
483 multiop_bg_pause $DIR/$tfile Ow_c ||
484 error "multiop_bg_pause $DIR/$tfile failed"
490 client_up || client_up || true # reconnect
492 kill -USR1 $pid || error "multiop $pid not running"
493 wait $pid || error "multiop $pid failed"
494 [ -s $DIR/$tfile ] || error "File was truncated"
498 run_test 20c "check that client eviction does not affect file content"
501 replay_barrier $SINGLEMDS
502 multiop_bg_pause $DIR/$tfile O_tSc ||
503 error "multiop_bg_pause $DIR/$tfile failed"
506 touch $DIR/$tfile-1 || error "touch $DIR/$tfile-1 failed"
509 kill -USR1 $pid || error "multiop $pid not running"
510 wait $pid || error "multiop $pid failed"
511 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
512 touch $DIR/$tfile-2 || error "touch $DIR/$tfile-2 failed"
515 run_test 21 "|X| open(O_CREAT), unlink touch new, replay, close (test mds_cleanup_orphans)"
518 multiop_bg_pause $DIR/$tfile O_tSc ||
519 error "multiop_bg_pause $DIR/$tfile failed"
522 replay_barrier $SINGLEMDS
526 kill -USR1 $pid || error "multiop $pid not running"
527 wait $pid || error "multiop $pid failed"
528 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
531 run_test 22 "open(O_CREAT), |X| unlink, replay, close (test mds_cleanup_orphans)"
534 multiop_bg_pause $DIR/$tfile O_tSc ||
535 error "multiop_bg_pause $DIR/$tfile failed"
538 replay_barrier $SINGLEMDS
540 touch $DIR/$tfile-1 || error "touch $DIR/$tfile-1 failed"
543 kill -USR1 $pid || error "multiop $pid not running"
544 wait $pid || error "multiop $pid failed"
545 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
546 touch $DIR/$tfile-2 || error "touch $DIR/$tfile-2 failed"
549 run_test 23 "open(O_CREAT), |X| unlink touch new, replay, close (test mds_cleanup_orphans)"
552 multiop_bg_pause $DIR/$tfile O_tSc ||
553 error "multiop_bg_pause $DIR/$tfile failed"
556 replay_barrier $SINGLEMDS
559 kill -USR1 $pid || error "multiop $pid not running"
560 wait $pid || error "multiop $pid failed"
561 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
564 run_test 24 "open(O_CREAT), replay, unlink, close (test mds_cleanup_orphans)"
567 multiop_bg_pause $DIR/$tfile O_tSc ||
568 error "multiop_bg_pause $DIR/$tfile failed"
572 replay_barrier $SINGLEMDS
574 kill -USR1 $pid || error "multiop $pid not running"
575 wait $pid || error "multiop $pid failed"
576 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
579 run_test 25 "open(O_CREAT), unlink, replay, close (test mds_cleanup_orphans)"
582 replay_barrier $SINGLEMDS
583 multiop_bg_pause $DIR/$tfile-1 O_tSc ||
584 error "multiop_bg_pause $DIR/$tfile-1 failed"
586 multiop_bg_pause $DIR/$tfile-2 O_tSc ||
587 error "multiop_bg_pause $DIR/$tfile-2 failed"
591 kill -USR1 $pid2 || error "second multiop $pid2 not running"
592 wait $pid2 || error "second multiop $pid2 failed"
595 kill -USR1 $pid1 || error "multiop $pid1 not running"
596 wait $pid1 || error "multiop $pid1 failed"
597 [ -e $DIR/$tfile-1 ] && error "file $DIR/$tfile-1 should not exist"
598 [ -e $DIR/$tfile-2 ] && error "file $DIR/$tfile-2 should not exist"
601 run_test 26 "|X| open(O_CREAT), unlink two, close one, replay, close one (test mds_cleanup_orphans)"
604 replay_barrier $SINGLEMDS
605 multiop_bg_pause $DIR/$tfile-1 O_tSc ||
606 error "multiop_bg_pause $DIR/$tfile-1 failed"
608 multiop_bg_pause $DIR/$tfile-2 O_tSc ||
609 error "multiop_bg_pause $DIR/$tfile-2 failed"
615 kill -USR1 $pid1 || error "multiop $pid1 not running"
616 wait $pid1 || error "multiop $pid1 failed"
617 kill -USR1 $pid2 || error "second multiop $pid2 not running"
618 wait $pid2 || error "second multiop $pid2 failed"
619 [ -e $DIR/$tfile-1 ] && error "file $DIR/$tfile-1 should not exist"
620 [ -e $DIR/$tfile-2 ] && error "file $DIR/$tfile-2 should not exist"
623 run_test 27 "|X| open(O_CREAT), unlink two, replay, close two (test mds_cleanup_orphans)"
626 multiop_bg_pause $DIR/$tfile-1 O_tSc ||
627 error "multiop_bg_pause $DIR/$tfile-1 failed"
629 multiop_bg_pause $DIR/$tfile-2 O_tSc ||
630 error "multiop_bg_pause $DIR/$tfile-2 failed"
632 replay_barrier $SINGLEMDS
635 kill -USR1 $pid2 || error "second multiop $pid2 not running"
636 wait $pid2 || error "second multiop $pid2 failed"
639 kill -USR1 $pid1 || error "multiop $pid1 not running"
640 wait $pid1 || error "multiop $pid1 failed"
641 [ -e $DIR/$tfile-1 ] && error "file $DIR/$tfile-1 should not exist"
642 [ -e $DIR/$tfile-2 ] && error "file $DIR/$tfile-2 should not exist"
645 run_test 28 "open(O_CREAT), |X| unlink two, close one, replay, close one (test mds_cleanup_orphans)"
648 multiop_bg_pause $DIR/$tfile-1 O_tSc ||
649 error "multiop_bg_pause $DIR/$tfile-1 failed"
651 multiop_bg_pause $DIR/$tfile-2 O_tSc ||
652 error "multiop_bg_pause $DIR/$tfile-2 failed"
654 replay_barrier $SINGLEMDS
659 kill -USR1 $pid1 || error "multiop $pid1 not running"
660 wait $pid1 || error "multiop $pid1 failed"
661 kill -USR1 $pid2 || error "second multiop $pid2 not running"
662 wait $pid2 || error "second multiop $pid2 failed"
663 [ -e $DIR/$tfile-1 ] && error "file $DIR/$tfile-1 should not exist"
664 [ -e $DIR/$tfile-2 ] && error "file $DIR/$tfile-2 should not exist"
667 run_test 29 "open(O_CREAT), |X| unlink two, replay, close two (test mds_cleanup_orphans)"
670 multiop_bg_pause $DIR/$tfile-1 O_tSc ||
671 error "multiop_bg_pause $DIR/$tfile-1 failed"
673 multiop_bg_pause $DIR/$tfile-2 O_tSc ||
674 error "multiop_bg_pause $DIR/$tfile-2 failed"
679 replay_barrier $SINGLEMDS
681 kill -USR1 $pid1 || error "multiop $pid1 not running"
682 wait $pid1 || error "multiop $pid1 failed"
683 kill -USR1 $pid2 || error "second multiop $pid2 not running"
684 wait $pid2 || error "second multiop $pid2 failed"
685 [ -e $DIR/$tfile-1 ] && error "file $DIR/$tfile-1 should not exist"
686 [ -e $DIR/$tfile-2 ] && error "file $DIR/$tfile-2 should not exist"
689 run_test 30 "open(O_CREAT) two, unlink two, replay, close two (test mds_cleanup_orphans)"
692 multiop_bg_pause $DIR/$tfile-1 O_tSc ||
693 error "multiop_bg_pause $DIR/$tfile-1 failed"
695 multiop_bg_pause $DIR/$tfile-2 O_tSc ||
696 error "multiop_bg_pause $DIR/$tfile-2 failed"
700 replay_barrier $SINGLEMDS
703 kill -USR1 $pid1 || error "multiop $pid1 not running"
704 wait $pid1 || error "multiop $pid1 failed"
705 kill -USR1 $pid2 || error "second multiop $pid2 not running"
706 wait $pid2 || error "second multiop $pid2 failed"
707 [ -e $DIR/$tfile-1 ] && error "file $DIR/$tfile-1 should not exist"
708 [ -e $DIR/$tfile-2 ] && error "file $DIR/$tfile-2 should not exist"
711 run_test 31 "open(O_CREAT) two, unlink one, |X| unlink one, close two (test mds_cleanup_orphans)"
713 # tests for bug 2104; completion without crashing is success. The close is
714 # stale, but we always return 0 for close, so the app never sees it.
716 multiop_bg_pause $DIR/$tfile O_c ||
717 error "multiop_bg_pause $DIR/$tfile failed"
719 multiop_bg_pause $DIR/$tfile O_c ||
720 error "second multiop_bg_pause $DIR/$tfile failed"
723 client_up || client_up || error "client_up failed"
724 kill -USR1 $pid1 || error "multiop $pid1 not running"
725 kill -USR1 $pid2 || error "second multiop $pid2 not running"
726 wait $pid1 || error "multiop $pid1 failed"
727 wait $pid2 || error "second multiop $pid2 failed"
730 run_test 32 "close() notices client eviction; close() after client eviction"
733 createmany -o $DIR/$tfile-%d 10 ||
734 error "createmany create $DIR/$tfile failed"
735 replay_barrier_nosync $SINGLEMDS
736 fail_abort $SINGLEMDS
737 # recreate shouldn't fail
738 createmany -o $DIR/$tfile--%d 10 ||
739 error "createmany recreate $DIR/$tfile failed"
743 run_test 33a "fid seq shouldn't be reused after abort recovery"
746 #define OBD_FAIL_SEQ_ALLOC 0x1311
747 do_facet $SINGLEMDS "lctl set_param fail_loc=0x1311"
749 createmany -o $DIR/$tfile-%d 10
750 replay_barrier_nosync $SINGLEMDS
751 fail_abort $SINGLEMDS
752 # recreate shouldn't fail
753 createmany -o $DIR/$tfile--%d 10 ||
754 error "createmany recreate $DIR/$tfile failed"
758 run_test 33b "test fid seq allocation"
761 multiop_bg_pause $DIR/$tfile O_c ||
762 error "multiop_bg_pause $DIR/$tfile failed"
766 replay_barrier $SINGLEMDS
767 fail_abort $SINGLEMDS
768 kill -USR1 $pid || error "multiop $pid not running"
769 wait $pid || error "multiop $pid failed"
770 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
774 run_test 34 "abort recovery before client does replay (test mds_cleanup_orphans)"
776 # bug 2278 - generate one orphan on OST, then destroy it during recovery from llog
778 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
780 #define OBD_FAIL_MDS_REINT_NET_REP 0x119
781 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000119"
786 # give a chance to remove from MDS
787 fail_abort $SINGLEMDS
788 $CHECKSTAT -t file $DIR/$tfile &&
789 error "$CHECKSTAT $DIR/$tfile attribute check should fail" ||
792 run_test 35 "test recovery from llog for unlink op"
794 # b=2432 resent cancel after replay uses wrong cookie,
795 # so don't resend cancels
797 replay_barrier $SINGLEMDS
799 checkstat $DIR/$tfile
800 facet_failover $SINGLEMDS
802 if dmesg | grep "unknown lock cookie"; then
803 error "cancel after replay failed"
806 run_test 36 "don't resend cancel"
809 # directory orphans can't be unlinked from PENDING directory
811 rmdir $DIR/$tfile 2>/dev/null
812 multiop_bg_pause $DIR/$tfile dD_c ||
813 error "multiop_bg_pause $DIR/$tfile failed"
817 replay_barrier $SINGLEMDS
818 # clear the dmesg buffer so we only see errors from this recovery
819 do_facet $SINGLEMDS dmesg -c >/dev/null
820 fail_abort $SINGLEMDS
821 kill -USR1 $pid || error "multiop $pid not running"
822 do_facet $SINGLEMDS dmesg | grep "error .* unlinking .* from PENDING" &&
823 error "error unlinking files"
824 wait $pid || error "multiop $pid failed"
828 run_test 37 "abort recovery before client does replay (test mds_cleanup_orphans for directories)"
831 createmany -o $DIR/$tfile-%d 800 ||
832 error "createmany -o $DIR/$tfile failed"
833 unlinkmany $DIR/$tfile-%d 0 400 || error "unlinkmany $DIR/$tfile failed"
834 replay_barrier $SINGLEMDS
836 unlinkmany $DIR/$tfile-%d 400 400 ||
837 error "unlinkmany $DIR/$tfile 400 failed"
839 $CHECKSTAT -t file $DIR/$tfile-* &&
840 error "$CHECKSTAT $DIR/$tfile-* attribute check should fail" ||
843 run_test 38 "test recovery from unlink llog (test llog_gen_rec) "
845 test_39() { # bug 4176
846 createmany -o $DIR/$tfile-%d 800 ||
847 error "createmany -o $DIR/$tfile failed"
848 replay_barrier $SINGLEMDS
849 unlinkmany $DIR/$tfile-%d 0 400
851 unlinkmany $DIR/$tfile-%d 400 400 ||
852 error "unlinkmany $DIR/$tfile 400 failed"
854 $CHECKSTAT -t file $DIR/$tfile-* &&
855 error "$CHECKSTAT $DIR/$tfile-* attribute check should fail" ||
858 run_test 39 "test recovery from unlink llog (test llog_gen_rec) "
861 lctl get_param -n osc.*.stats | awk -vwrites=0 '/ost_write/ { writes += $2 } END { print writes; }'
866 # always need connection to MDS to verify layout during IO. LU-2628.
867 lctl get_param mdc.*.connect_flags | grep -q layout_lock &&
868 skip "layout_lock needs MDS connection for IO" && return 0
870 $LCTL mark multiop $MOUNT/$tfile OS_c
871 multiop $MOUNT/$tfile OS_c &
873 writeme -s $MOUNT/${tfile}-2 &
876 facet_failover $SINGLEMDS
877 #define OBD_FAIL_MDS_CONNECT_NET 0x117
878 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000117"
880 stat1=$(count_ost_writes)
882 stat2=$(count_ost_writes)
883 echo "$stat1, $stat2"
884 if [ $stat1 -lt $stat2 ]; then
885 echo "writes continuing during recovery"
888 echo "writes not continuing during recovery, bug 2477"
891 echo "waiting for writeme $WRITE_PID"
895 echo "waiting for multiop $PID"
896 wait $PID || error "multiop $PID failed"
897 do_facet client munlink $MOUNT/$tfile ||
898 error "munlink $MOUNT/$tfile failed"
899 do_facet client munlink $MOUNT/${tfile}-2 ||
900 error "munlink $MOUNT/$tfile-2 failed"
903 run_test 40 "cause recovery in ptlrpc, ensure IO continues"
906 # make sure that a read to one osc doesn't try to double-unlock its page just
907 # because another osc is invalid. trigger_group_io used to mistakenly return
908 # an error if any oscs were invalid even after having successfully put rpcs
909 # on valid oscs. This was fatal if the caller was ll_readpage who unlocked
910 # the page, guarnateeing that the unlock from the RPC completion would
911 # assert on trying to unlock the unlocked page.
913 [ $OSTCOUNT -lt 2 ] &&
914 skip_env "skipping test 41: we don't have a second OST to test with" &&
917 local f=$MOUNT/$tfile
918 # make sure the start of the file is ost1
919 $SETSTRIPE -S $((128 * 1024)) -i 0 $f
920 do_facet client dd if=/dev/zero of=$f bs=4k count=1 ||
921 error "dd on client failed"
923 # fail ost2 and read from ost1
924 local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS $ost2_svc)
925 local osc2dev=$(do_facet $SINGLEMDS "lctl get_param -n devices" |
926 grep $mdtosc | awk '{print $1}')
927 [ -z "$osc2dev" ] && echo "OST: $ost2_svc" &&
928 lctl get_param -n devices &&
929 error "OST 2 $osc2dev does not exist"
930 do_facet $SINGLEMDS $LCTL --device $osc2dev deactivate ||
931 error "deactive device on $SINGLEMDS failed"
932 do_facet client dd if=$f of=/dev/null bs=4k count=1 ||
933 error "second dd on client failed"
934 do_facet $SINGLEMDS $LCTL --device $osc2dev activate ||
935 error "active device on $SINGLEMDS failed"
938 run_test 41 "read from a valid osc while other oscs are invalid"
940 # test MDS recovery after ost failure
942 blocks=$(df -P $MOUNT | tail -n 1 | awk '{ print $2 }')
943 createmany -o $DIR/$tfile-%d 800 ||
944 error "createmany -o $DIR/$tfile failed"
946 unlinkmany $DIR/$tfile-%d 0 400
948 lctl set_param debug=-1
951 # osc is evicted, fs is smaller (but only with failout OSTs (bug 7287)
952 #blocks_after=`df -P $MOUNT | tail -n 1 | awk '{ print $2 }'`
953 #[ $blocks_after -lt $blocks ] || return 1
954 echo "wait for MDS to timeout and recover"
955 sleep $((TIMEOUT * 2))
957 unlinkmany $DIR/$tfile-%d 400 400 ||
958 error "unlinkmany $DIR/$tfile 400 failed"
959 $CHECKSTAT -t file $DIR/$tfile-* &&
960 error "$CHECKSTAT $DIR/$tfile-* attribute check should fail" ||
963 run_test 42 "recovery after ost failure"
965 # timeout in MDS/OST recovery RPC will LBUG MDS
966 test_43() { # bug 2530
967 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
969 replay_barrier $SINGLEMDS
971 # OBD_FAIL_OST_CREATE_NET 0x204
972 do_facet ost1 "lctl set_param fail_loc=0x80000204"
978 run_test 43 "mds osc import failure during recovery; don't LBUG"
980 test_44a() { # was test_44
983 local mdcdev=$($LCTL dl |
984 awk "/${FSNAME}-MDT0000-mdc-/ {if (\$2 == \"UP\") {print \$1}}")
985 [ "$mdcdev" ] || error "${FSNAME}-MDT0000-mdc- not UP"
986 [ $(echo $mdcdev | wc -w) -eq 1 ] ||
987 { $LCTL dl; error "looking for mdcdev=$mdcdev"; }
989 # adaptive timeouts slow this way down
990 if at_is_enabled; then
991 at_max_saved=$(at_max_get mds)
995 for i in $(seq 1 10); do
996 echo "$i of 10 ($(date +%s))"
997 do_facet $SINGLEMDS \
998 "lctl get_param -n md[ts].*.mdt.timeouts | grep service"
999 #define OBD_FAIL_TGT_CONN_RACE 0x701
1000 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000701"
1001 # lctl below may fail, it is valid case
1002 $LCTL --device $mdcdev recover
1005 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1006 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds
1009 run_test 44a "race in target handle connect"
1012 local mdcdev=$($LCTL dl |
1013 awk "/${FSNAME}-MDT0000-mdc-/ {if (\$2 == \"UP\") {print \$1}}")
1014 [ "$mdcdev" ] || error "${FSNAME}-MDT0000-mdc not up"
1015 [ $(echo $mdcdev | wc -w) -eq 1 ] ||
1016 { echo mdcdev=$mdcdev; $LCTL dl;
1017 error "more than one ${FSNAME}-MDT0000-mdc"; }
1019 for i in $(seq 1 10); do
1020 echo "$i of 10 ($(date +%s))"
1021 do_facet $SINGLEMDS \
1022 "lctl get_param -n md[ts].*.mdt.timeouts | grep service"
1023 #define OBD_FAIL_TGT_DELAY_RECONNECT 0x704
1024 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000704"
1025 # lctl below may fail, it is valid case
1026 $LCTL --device $mdcdev recover
1031 run_test 44b "race in target handle connect"
1034 replay_barrier $SINGLEMDS
1035 createmany -m $DIR/$tfile-%d 100 || error "failed to create directories"
1036 #define OBD_FAIL_TGT_RCVG_FLAG 0x712
1037 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000712"
1038 fail_abort $SINGLEMDS
1039 unlinkmany $DIR/$tfile-%d 100 && error "unliked after fail abort"
1041 unlinkmany $DIR/$tfile-%d 100 && error "unliked after fail"
1044 run_test 44c "race in target handle connect"
1046 # Handle failed close
1048 local mdcdev=$($LCTL get_param -n devices |
1049 awk "/ ${FSNAME}-MDT0000-mdc-/ {print \$1}")
1050 [ "$mdcdev" ] || error "${FSNAME}-MDT0000-mdc not up"
1051 [ $(echo $mdcdev | wc -w) -eq 1 ] ||
1052 { echo mdcdev=$mdcdev; $LCTL dl;
1053 error "more than one ${FSNAME}-MDT0000-mdc"; }
1055 $LCTL --device $mdcdev recover ||
1056 error "$LCTL --device $mdcdev recover failed"
1058 multiop_bg_pause $DIR/$tfile O_c ||
1059 error "multiop_bg_pause $DIR/$tfile failed"
1062 # This will cause the CLOSE to fail before even
1063 # allocating a reply buffer
1064 $LCTL --device $mdcdev deactivate ||
1065 error "$LCTL --device $mdcdev deactivate failed"
1068 kill -USR1 $pid || error "multiop $pid not running"
1069 wait $pid || error "multiop $pid failed"
1071 $LCTL --device $mdcdev activate ||
1072 error "$LCTL --device $mdcdev activate failed"
1075 $CHECKSTAT -t file $DIR/$tfile ||
1076 error "$CHECKSTAT $DIR/$tfile attribute check failed"
1079 run_test 45 "Handle failed close"
1083 drop_reply "touch $DIR/$tfile"
1085 # ironically, the previous test, 45, will cause a real forced close,
1086 # so just look for one for this test
1087 dmesg | grep -i "force closing client file handle for $tfile" &&
1088 error "found force closing in dmesg"
1091 run_test 46 "Don't leak file handle after open resend (3325)"
1093 test_47() { # bug 2824
1094 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1096 # create some files to make sure precreate has been done on all
1097 # OSTs. (just in case this test is run independently)
1098 createmany -o $DIR/$tfile 20 ||
1099 error "createmany create $DIR/$tfile failed"
1101 # OBD_FAIL_OST_CREATE_NET 0x204
1103 do_facet ost1 "lctl set_param fail_loc=0x80000204"
1104 client_up || error "client_up failed"
1106 # let the MDS discover the OST failure, attempt to recover, fail
1107 # and recover again.
1108 sleep $((3 * TIMEOUT))
1110 # Without 2824, this createmany would hang
1111 createmany -o $DIR/$tfile 20 ||
1112 error "createmany recraete $DIR/$tfile failed"
1113 unlinkmany $DIR/$tfile 20 || error "unlinkmany $DIR/$tfile failed"
1117 run_test 47 "MDS->OSC failure during precreate cleanup (2824)"
1120 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1121 [ "$OSTCOUNT" -lt "2" ] &&
1122 skip_env "$OSTCOUNT < 2 OSTs -- skipping" && return
1124 replay_barrier $SINGLEMDS
1125 createmany -o $DIR/$tfile 20 ||
1126 error "createmany -o $DIR/$tfile failed"
1127 # OBD_FAIL_OST_EROFS 0x216
1128 facet_failover $SINGLEMDS
1129 do_facet ost1 "lctl set_param fail_loc=0x80000216"
1130 client_up || error "client_up failed"
1132 createmany -o $DIR/$tfile 20 20 ||
1133 error "createmany recraete $DIR/$tfile failed"
1134 unlinkmany $DIR/$tfile 40 || error "unlinkmany $DIR/$tfile failed"
1137 run_test 48 "MDS->OSC failure during precreate cleanup (2824)"
1140 local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS $ost1_svc)
1141 local oscdev=$(do_facet $SINGLEMDS "lctl get_param -n devices" |
1142 grep $mdtosc | awk '{print $1}')
1143 [ "$oscdev" ] || error "could not find OSC device on MDS"
1144 do_facet $SINGLEMDS $LCTL --device $oscdev recover ||
1145 error "OSC device $oscdev recovery failed"
1146 do_facet $SINGLEMDS $LCTL --device $oscdev recover ||
1147 error "second OSC device $oscdev recovery failed"
1148 # give the mds_lov_sync threads a chance to run
1151 run_test 50 "Double OSC recovery, don't LASSERT (3812)"
1153 # b3764 timed out lock replay
1155 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.6.90) ] &&
1156 skip "MDS prior to 2.6.90 handle LDLM_REPLY_NET incorrectly" &&
1159 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
1160 cancel_lru_locks mdc
1162 multiop_bg_pause $DIR/$tfile s_s || error "multiop $DIR/$tfile failed"
1165 #define OBD_FAIL_MDS_LDLM_REPLY_NET 0x157
1166 lctl set_param -n ldlm.cancel_unused_locks_before_replay "0"
1167 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000157"
1169 fail $SINGLEMDS || error "fail $SINGLEMDS failed"
1171 wait $mpid || error "multiop_bg_pause pid failed"
1173 do_facet $SINGLEMDS "lctl set_param fail_loc=0x0"
1174 lctl set_param fail_loc=0x0
1175 lctl set_param -n ldlm.cancel_unused_locks_before_replay "1"
1178 run_test 52 "time out lock replay (3764)"
1180 # bug 3462 - simultaneous MDC requests
1182 [[ $(lctl get_param mdc.*.import |
1183 grep "connect_flags:.*multi_mod_rpc") ]] ||
1184 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
1186 cancel_lru_locks mdc # cleanup locks from former test cases
1187 mkdir $DIR/${tdir}-1 || error "mkdir $DIR/${tdir}-1 failed"
1188 mkdir $DIR/${tdir}-2 || error "mkdir $DIR/${tdir}-2 failed"
1189 multiop $DIR/${tdir}-1/f O_c &
1191 # give multiop a change to open
1194 #define OBD_FAIL_MDS_CLOSE_NET 0x115
1195 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
1196 kill -USR1 $close_pid
1197 cancel_lru_locks mdc # force the close
1198 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1200 mcreate $DIR/${tdir}-2/f || error "mcreate $DIR/${tdir}-2/f failed"
1202 # close should still be here
1203 [ -d /proc/$close_pid ] || error "close_pid doesn't exist"
1205 replay_barrier_nodf $SINGLEMDS
1207 wait $close_pid || error "close_pid $close_pid failed"
1209 $CHECKSTAT -t file $DIR/${tdir}-1/f ||
1210 error "$CHECKSTAT $DIR/${tdir}-1/f attribute check failed"
1211 $CHECKSTAT -t file $DIR/${tdir}-2/f ||
1212 error "$CHECKSTAT $DIR/${tdir}-2/f attribute check failed"
1213 rm -rf $DIR/${tdir}-*
1215 run_test 53a "|X| close request while two MDC requests in flight"
1218 cancel_lru_locks mdc # cleanup locks from former test cases
1220 mkdir $DIR/${tdir}-1 || error "mkdir $DIR/${tdir}-1 failed"
1221 mkdir $DIR/${tdir}-2 || error "mkdir $DIR/${tdir}-2 failed"
1222 multiop_bg_pause $DIR/${tdir}-1/f O_c ||
1223 error "multiop_bg_pause $DIR/${tdir}-1/f failed"
1226 #define OBD_FAIL_MDS_REINT_NET 0x107
1227 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000107"
1228 mcreate $DIR/${tdir}-2/f &
1232 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1233 kill -USR1 $close_pid
1234 cancel_lru_locks mdc # force the close
1235 wait $close_pid || error "close_pid $close_pid failed"
1236 # open should still be here
1237 [ -d /proc/$open_pid ] || error "open_pid doesn't exist"
1239 replay_barrier_nodf $SINGLEMDS
1241 wait $open_pid || error "open_pid failed"
1243 $CHECKSTAT -t file $DIR/${tdir}-1/f ||
1244 error "$CHECKSTAT $DIR/${tdir}-1/f attribute check failed"
1245 $CHECKSTAT -t file $DIR/${tdir}-2/f ||
1246 error "$CHECKSTAT $DIR/${tdir}-2/f attribute check failed"
1247 rm -rf $DIR/${tdir}-*
1249 run_test 53b "|X| open request while two MDC requests in flight"
1252 cancel_lru_locks mdc # cleanup locks from former test cases
1254 mkdir $DIR/${tdir}-1 || error "mkdir $DIR/${tdir}-1 failed"
1255 mkdir $DIR/${tdir}-2 || error "mkdir $DIR/${tdir}-2 failed"
1256 multiop $DIR/${tdir}-1/f O_c &
1259 #define OBD_FAIL_MDS_REINT_NET 0x107
1260 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000107"
1261 mcreate $DIR/${tdir}-2/f &
1265 #define OBD_FAIL_MDS_CLOSE_NET 0x115
1266 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
1267 kill -USR1 $close_pid
1268 cancel_lru_locks mdc # force the close
1270 #bz20647: make sure all pids exist before failover
1271 [ -d /proc/$close_pid ] || error "close_pid doesn't exist"
1272 [ -d /proc/$open_pid ] || error "open_pid doesn't exists"
1273 replay_barrier_nodf $SINGLEMDS
1274 fail_nodf $SINGLEMDS
1275 wait $open_pid || error "open_pid failed"
1277 # close should be gone
1278 [ -d /proc/$close_pid ] && error "close_pid should not exist"
1279 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1281 $CHECKSTAT -t file $DIR/${tdir}-1/f ||
1282 error "$CHECKSTAT $DIR/${tdir}-1/f attribute check failed"
1283 $CHECKSTAT -t file $DIR/${tdir}-2/f ||
1284 error "$CHECKSTAT $DIR/${tdir}-2/f attribute check failed"
1285 rm -rf $DIR/${tdir}-*
1287 run_test 53c "|X| open request and close request while two MDC requests in flight"
1290 [[ $(lctl get_param mdc.*.import |
1291 grep "connect_flags:.*multi_mod_rpc") ]] ||
1292 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
1294 cancel_lru_locks mdc # cleanup locks from former test cases
1296 mkdir $DIR/${tdir}-1 || error "mkdir $DIR/${tdir}-1 failed"
1297 mkdir $DIR/${tdir}-2 || error "mkdir $DIR/${tdir}-2 failed"
1298 multiop $DIR/${tdir}-1/f O_c &
1300 # give multiop a chance to open
1303 #define OBD_FAIL_MDS_CLOSE_NET_REP 0x13b
1304 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000013b"
1305 kill -USR1 $close_pid
1306 cancel_lru_locks mdc # force the close
1307 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1308 mcreate $DIR/${tdir}-2/f || error "mcreate $DIR/${tdir}-2/f failed"
1310 # close should still be here
1311 [ -d /proc/$close_pid ] || error "close_pid doesn't exist"
1313 wait $close_pid || error "close_pid failed"
1315 $CHECKSTAT -t file $DIR/${tdir}-1/f ||
1316 error "$CHECKSTAT $DIR/${tdir}-1/f attribute check failed"
1317 $CHECKSTAT -t file $DIR/${tdir}-2/f ||
1318 error "$CHECKSTAT $DIR/${tdir}-2/f attribute check failed"
1319 rm -rf $DIR/${tdir}-*
1321 run_test 53d "close reply while two MDC requests in flight"
1324 cancel_lru_locks mdc # cleanup locks from former test cases
1326 mkdir $DIR/${tdir}-1 || error "mkdir $DIR/${tdir}-1 failed"
1327 mkdir $DIR/${tdir}-2 || error "mkdir $DIR/${tdir}-2 failed"
1328 multiop $DIR/${tdir}-1/f O_c &
1331 #define OBD_FAIL_MDS_REINT_NET_REP 0x119
1332 do_facet $SINGLEMDS "lctl set_param fail_loc=0x119"
1333 mcreate $DIR/${tdir}-2/f &
1337 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1338 kill -USR1 $close_pid
1339 cancel_lru_locks mdc # force the close
1340 wait $close_pid || error "close_pid failed"
1341 # open should still be here
1342 [ -d /proc/$open_pid ] || error "open_pid doesn't exists"
1344 replay_barrier_nodf $SINGLEMDS
1346 wait $open_pid || error "open_pid failed"
1348 $CHECKSTAT -t file $DIR/${tdir}-1/f ||
1349 error "$CHECKSTAT $DIR/${tdir}-1/f attribute check failed"
1350 $CHECKSTAT -t file $DIR/${tdir}-2/f ||
1351 error "$CHECKSTAT $DIR/${tdir}-2/f attribute check failed"
1352 rm -rf $DIR/${tdir}-*
1354 run_test 53e "|X| open reply while two MDC requests in flight"
1357 cancel_lru_locks mdc # cleanup locks from former test cases
1359 mkdir $DIR/${tdir}-1 || error "mkdir $DIR/${tdir}-1 failed"
1360 mkdir $DIR/${tdir}-2 || error "mkdir $DIR/${tdir}-2 failed"
1361 multiop $DIR/${tdir}-1/f O_c &
1364 #define OBD_FAIL_MDS_REINT_NET_REP 0x119
1365 do_facet $SINGLEMDS "lctl set_param fail_loc=0x119"
1366 mcreate $DIR/${tdir}-2/f &
1370 #define OBD_FAIL_MDS_CLOSE_NET_REP 0x13b
1371 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000013b"
1372 kill -USR1 $close_pid
1373 cancel_lru_locks mdc # force the close
1375 #bz20647: make sure all pids are exists before failover
1376 [ -d /proc/$close_pid ] || error "close_pid doesn't exist"
1377 [ -d /proc/$open_pid ] || error "open_pid doesn't exists"
1378 replay_barrier_nodf $SINGLEMDS
1379 fail_nodf $SINGLEMDS
1380 wait $open_pid || error "open_pid failed"
1382 # close should be gone
1383 [ -d /proc/$close_pid ] && error "close_pid should not exist"
1384 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1386 $CHECKSTAT -t file $DIR/${tdir}-1/f ||
1387 error "$CHECKSTAT $DIR/${tdir}-1/f attribute check failed"
1388 $CHECKSTAT -t file $DIR/${tdir}-2/f ||
1389 error "$CHECKSTAT $DIR/${tdir}-2/f attribute check failed"
1390 rm -rf $DIR/${tdir}-*
1392 run_test 53f "|X| open reply and close reply while two MDC requests in flight"
1395 cancel_lru_locks mdc # cleanup locks from former test cases
1397 mkdir $DIR/${tdir}-1 || error "mkdir $DIR/${tdir}-1 failed"
1398 mkdir $DIR/${tdir}-2 || error "mkdir $DIR/${tdir}-2 failed"
1399 multiop $DIR/${tdir}-1/f O_c &
1402 #define OBD_FAIL_MDS_REINT_NET_REP 0x119
1403 do_facet $SINGLEMDS "lctl set_param fail_loc=0x119"
1404 mcreate $DIR/${tdir}-2/f &
1408 #define OBD_FAIL_MDS_CLOSE_NET 0x115
1409 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
1410 kill -USR1 $close_pid
1411 cancel_lru_locks mdc # force the close
1412 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1414 #bz20647: make sure all pids are exists before failover
1415 [ -d /proc/$close_pid ] || error "close_pid doesn't exist"
1416 [ -d /proc/$open_pid ] || error "open_pid doesn't exists"
1417 replay_barrier_nodf $SINGLEMDS
1418 fail_nodf $SINGLEMDS
1419 wait $open_pid || error "open_pid failed"
1421 # close should be gone
1422 [ -d /proc/$close_pid ] && error "close_pid should not exist"
1424 $CHECKSTAT -t file $DIR/${tdir}-1/f ||
1425 error "$CHECKSTAT $DIR/${tdir}-1/f attribute check failed"
1426 $CHECKSTAT -t file $DIR/${tdir}-2/f ||
1427 error "$CHECKSTAT $DIR/${tdir}-2/f attribute check failed"
1428 rm -rf $DIR/${tdir}-*
1430 run_test 53g "|X| drop open reply and close request while close and open are both in flight"
1433 cancel_lru_locks mdc # cleanup locks from former test cases
1435 mkdir $DIR/${tdir}-1 || error "mkdir $DIR/${tdir}-1 failed"
1436 mkdir $DIR/${tdir}-2 || error "mkdir $DIR/${tdir}-2 failed"
1437 multiop $DIR/${tdir}-1/f O_c &
1440 #define OBD_FAIL_MDS_REINT_NET 0x107
1441 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000107"
1442 mcreate $DIR/${tdir}-2/f &
1446 #define OBD_FAIL_MDS_CLOSE_NET_REP 0x13b
1447 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000013b"
1448 kill -USR1 $close_pid
1449 cancel_lru_locks mdc # force the close
1452 #bz20647: make sure all pids are exists before failover
1453 [ -d /proc/$close_pid ] || error "close_pid doesn't exist"
1454 [ -d /proc/$open_pid ] || error "open_pid doesn't exists"
1455 replay_barrier_nodf $SINGLEMDS
1456 fail_nodf $SINGLEMDS
1457 wait $open_pid || error "open_pid failed"
1459 # close should be gone
1460 [ -d /proc/$close_pid ] && error "close_pid should not exist"
1461 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1463 $CHECKSTAT -t file $DIR/${tdir}-1/f ||
1464 error "$CHECKSTAT $DIR/${tdir}-1/f attribute check failed"
1465 $CHECKSTAT -t file $DIR/${tdir}-2/f ||
1466 error "$CHECKSTAT $DIR/${tdir}-2/f attribute check failed"
1467 rm -rf $DIR/${tdir}-*
1469 run_test 53h "open request and close reply while two MDC requests in flight"
1471 #b3761 ASSERTION(hash != 0) failed
1473 # OBD_FAIL_MDS_OPEN_CREATE | OBD_FAIL_ONCE
1474 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000012b"
1476 # give touch a chance to run
1478 do_facet $SINGLEMDS "lctl set_param fail_loc=0x0"
1482 run_test 55 "let MDS_CHECK_RESENT return the original return code instead of 0"
1484 #b3440 ASSERTION(rec->ur_fid2->id) failed
1486 ln -s foo $DIR/$tfile
1487 replay_barrier $SINGLEMDS
1488 #drop_reply "cat $DIR/$tfile"
1492 run_test 56 "don't replay a symlink open request (3440)"
1494 #recovery one mds-ost setattr from llog
1496 #define OBD_FAIL_MDS_OST_SETATTR 0x12c
1497 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000012c"
1498 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
1499 replay_barrier $SINGLEMDS
1501 wait_recovery_complete $SINGLEMDS || error "MDS recovery is not done"
1502 wait_mds_ost_sync || error "wait_mds_ost_sync failed"
1503 $CHECKSTAT -t file $DIR/$tfile ||
1504 error "$CHECKSTAT $DIR/$tfile attribute check failed"
1505 do_facet $SINGLEMDS "lctl set_param fail_loc=0x0"
1508 run_test 57 "test recovery from llog for setattr op"
1511 zconf_umount $(hostname) $MOUNT2
1515 #recovery many mds-ost setattr from llog
1517 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
1518 #define OBD_FAIL_MDS_OST_SETATTR 0x12c
1519 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000012c"
1520 createmany -o $DIR/$tdir/$tfile-%d 2500
1521 replay_barrier $SINGLEMDS
1524 $CHECKSTAT -t file $DIR/$tdir/$tfile-* >/dev/null ||
1525 error "$CHECKSTAT $DIR/$tfile-* attribute check failed"
1526 do_facet $SINGLEMDS "lctl set_param fail_loc=0x0"
1527 unlinkmany $DIR/$tdir/$tfile-%d 2500 ||
1528 error "unlinkmany $DIR/$tfile failed"
1531 run_test 58a "test recovery from llog for setattr op (test llog_gen_rec)"
1537 trap cleanup_58 EXIT
1539 large_xattr_enabled &&
1540 orig="$(generate_string $(max_xattr_size))" || orig="bar"
1541 # Original extended attribute can be long. Print a small version of
1542 # attribute if an error occurs
1543 local sm_msg=$(printf "%.9s" $orig)
1545 mount_client $MOUNT2 || error "mount_client on $MOUNT2 failed"
1546 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
1547 touch $DIR/$tdir/$tfile || error "touch $DIR/$tdir/$tfile failed"
1548 replay_barrier $SINGLEMDS
1549 setfattr -n trusted.foo -v $orig $DIR/$tdir/$tfile
1551 new=$(get_xattr_value trusted.foo $MOUNT2/$tdir/$tfile)
1552 [[ "$new" = "$orig" ]] ||
1553 error "xattr set ($sm_msg...) differs from xattr get ($new)"
1554 rm -f $DIR/$tdir/$tfile
1557 wait_clients_import_state ${CLIENTS:-$HOSTNAME} "mgs" FULL
1559 run_test 58b "test replay of setxattr op"
1561 test_58c() { # bug 16570
1566 trap cleanup_58 EXIT
1568 if large_xattr_enabled; then
1569 local xattr_size=$(max_xattr_size)
1570 orig="$(generate_string $((xattr_size / 2)))"
1571 orig1="$(generate_string $xattr_size)"
1577 # PING_INTERVAL max(obd_timeout / 4, 1U)
1578 sleep $((TIMEOUT / 4))
1580 # Original extended attribute can be long. Print a small version of
1581 # attribute if an error occurs
1582 local sm_msg=$(printf "%.9s" $orig)
1583 local sm_msg1=$(printf "%.9s" $orig1)
1585 mount_client $MOUNT2 || error "mount_client on $MOUNT2 failed"
1586 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
1587 touch $DIR/$tdir/$tfile || error "touch $DIR/$tdir/$tfile failed"
1588 drop_request "setfattr -n trusted.foo -v $orig $DIR/$tdir/$tfile" ||
1589 error "drop_request for setfattr failed"
1590 new=$(get_xattr_value trusted.foo $MOUNT2/$tdir/$tfile)
1591 [[ "$new" = "$orig" ]] ||
1592 error "xattr set ($sm_msg...) differs from xattr get ($new)"
1593 drop_reint_reply "setfattr -n trusted.foo1 \
1594 -v $orig1 $DIR/$tdir/$tfile" ||
1595 error "drop_reint_reply for setfattr failed"
1596 new=$(get_xattr_value trusted.foo1 $MOUNT2/$tdir/$tfile)
1597 [[ "$new" = "$orig1" ]] ||
1598 error "second xattr set ($sm_msg1...) differs xattr get ($new)"
1599 rm -f $DIR/$tdir/$tfile
1603 run_test 58c "resend/reconstruct setxattr op"
1605 # log_commit_thread vs filter_destroy race used to lead to import use after free
1608 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1610 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
1611 createmany -o $DIR/$tdir/$tfile-%d 200 ||
1612 error "createmany create files failed"
1614 unlinkmany $DIR/$tdir/$tfile-%d 200 ||
1615 error "unlinkmany $DIR/$tdir/$tfile failed"
1616 #define OBD_FAIL_PTLRPC_DELAY_RECOV 0x507
1617 do_facet ost1 "lctl set_param fail_loc=0x507"
1620 do_facet ost1 "lctl set_param fail_loc=0x0"
1624 run_test 59 "test log_commit_thread vs filter_destroy race"
1626 # race between add unlink llog vs cat log init in post_recovery (only for b1_6)
1627 # bug 12086: should no oops and No ctxt error for this test
1629 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
1630 createmany -o $DIR/$tdir/$tfile-%d 200 ||
1631 error "createmany create files failed"
1632 replay_barrier $SINGLEMDS
1633 unlinkmany $DIR/$tdir/$tfile-%d 0 100
1635 unlinkmany $DIR/$tdir/$tfile-%d 100 100
1636 local no_ctxt=$(dmesg | grep "No ctxt")
1637 [ -z "$no_ctxt" ] || error "ctxt is not initialized in recovery"
1639 run_test 60 "test llog post recovery init vs llog unlink"
1641 #test race llog recovery thread vs llog cleanup
1642 test_61a() { # was test_61
1643 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1645 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
1646 createmany -o $DIR/$tdir/$tfile-%d 800 ||
1647 error "createmany create files failed"
1649 unlinkmany $DIR/$tdir/$tfile-%d 800
1650 # OBD_FAIL_OST_LLOG_RECOVERY_TIMEOUT 0x221
1651 set_nodes_failloc "$(osts_nodes)" 0x80000221
1656 set_nodes_failloc "$(osts_nodes)" 0x0
1658 $CHECKSTAT -t file $DIR/$tdir/$tfile-* &&
1659 error "$CHECKSTAT $DIR/$tdir/$tfile attribute check should fail"
1662 run_test 61a "test race llog recovery vs llog cleanup"
1664 #test race mds llog sync vs llog cleanup
1666 # OBD_FAIL_MDS_LLOG_SYNC_TIMEOUT 0x13a
1667 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000013a"
1668 facet_failover $SINGLEMDS
1671 do_facet client dd if=/dev/zero of=$DIR/$tfile bs=4k count=1 ||
1674 run_test 61b "test race mds llog sync vs llog cleanup"
1676 #test race cancel cookie cb vs llog cleanup
1678 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1680 # OBD_FAIL_OST_CANCEL_COOKIE_TIMEOUT 0x222
1681 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
1682 set_nodes_failloc "$(osts_nodes)" 0x80000222
1686 set_nodes_failloc "$(osts_nodes)" 0x0
1688 run_test 61c "test race mds llog sync vs llog cleanup"
1690 test_61d() { # bug 16002 # bug 17466 # bug 22137
1691 # OBD_FAIL_OBD_LLOG_SETUP 0x605
1693 do_facet mgs "lctl set_param fail_loc=0x80000605"
1694 start mgs $(mgsdevname) $MGS_MOUNT_OPTS &&
1695 error "mgs start should have failed"
1696 do_facet mgs "lctl set_param fail_loc=0"
1697 start mgs $(mgsdevname) $MGS_MOUNT_OPTS || error "cannot restart mgs"
1699 run_test 61d "error in llog_setup should cleanup the llog context correctly"
1701 test_62() { # Bug 15756 - don't mis-drop resent replay
1702 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
1703 replay_barrier $SINGLEMDS
1704 createmany -o $DIR/$tdir/$tfile- 25 ||
1705 error "createmany create files failed"
1706 #define OBD_FAIL_TGT_REPLAY_DROP 0x707
1707 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000707"
1709 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1710 unlinkmany $DIR/$tdir/$tfile- 25 ||
1711 error "unlinkmany $DIR/$tdir/$tfile failed"
1714 run_test 62 "don't mis-drop resent replay"
1716 #Adaptive Timeouts (bug 3055)
1724 echo "Cleaning up AT ..."
1725 if [ -n "$ATOLDBASE" ]; then
1726 local at_history=$($LCTL get_param -n at_history)
1727 do_facet $SINGLEMDS "lctl set_param at_history=$at_history" || true
1728 do_facet ost1 "lctl set_param at_history=$at_history" || true
1731 if [ $AT_MAX_SET -ne 0 ]; then
1732 for facet in mds client ost; do
1733 var=AT_MAX_SAVE_${facet}
1734 echo restore AT on $facet to saved value ${!var}
1735 at_max_set ${!var} $facet
1736 at_new=$(at_max_get $facet)
1737 echo Restored AT value on $facet $at_new
1738 [ $at_new -eq ${!var} ] ||
1739 error "AT value not restored SAVED ${!var} NEW $at_new"
1746 local at_max_new=600
1748 # Save at_max original values
1750 if [ $AT_MAX_SET -eq 0 ]; then
1751 # Suppose that all osts have the same at_max
1752 for facet in mds client ost; do
1753 eval AT_MAX_SAVE_${facet}=$(at_max_get $facet)
1757 for facet in mds client ost; do
1758 at_max=$(at_max_get $facet)
1759 if [ $at_max -ne $at_max_new ]; then
1760 echo "AT value on $facet is $at_max, set it by force temporarily to $at_max_new"
1761 at_max_set $at_max_new $facet
1766 if [ -z "$ATOLDBASE" ]; then
1767 ATOLDBASE=$(do_facet $SINGLEMDS "lctl get_param -n at_history")
1768 # speed up the timebase so we can check decreasing AT
1769 do_facet $SINGLEMDS "lctl set_param at_history=8" || true
1770 do_facet ost1 "lctl set_param at_history=8" || true
1772 # sleep for a while to cool down, should be > 8s and also allow
1773 # at least one ping to be sent. simply use TIMEOUT to be safe.
1778 test_65a() #bug 3055
1780 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1782 at_start || return 0
1783 $LCTL dk > /dev/null
1785 $LCTL set_param debug="other"
1786 # Slow down a request to the current service time, this is critical
1787 # because previous tests may have caused this value to increase.
1788 REQ_DELAY=`lctl get_param -n mdc.${FSNAME}-MDT0000-mdc-*.timeouts |
1789 awk '/portal 12/ {print $5}'`
1790 REQ_DELAY=$((${REQ_DELAY} + ${REQ_DELAY} / 4 + 5))
1792 do_facet $SINGLEMDS lctl set_param fail_val=$((${REQ_DELAY} * 1000))
1793 #define OBD_FAIL_PTLRPC_PAUSE_REQ 0x50a
1794 do_facet $SINGLEMDS $LCTL set_param fail_loc=0x8000050a
1795 createmany -o $DIR/$tfile 10 > /dev/null
1796 unlinkmany $DIR/$tfile 10 > /dev/null
1797 # check for log message
1798 $LCTL dk | grep "Early reply #" || error "No early reply"
1800 # client should show REQ_DELAY estimates
1801 lctl get_param -n mdc.${FSNAME}-MDT0000-mdc-*.timeouts | grep portal
1803 lctl get_param -n mdc.${FSNAME}-MDT0000-mdc-*.timeouts | grep portal
1805 run_test 65a "AT: verify early replies"
1807 test_65b() #bug 3055
1809 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1811 at_start || return 0
1814 $LCTL set_param debug="other trace"
1815 $LCTL dk > /dev/null
1816 # Slow down a request to the current service time, this is critical
1817 # because previous tests may have caused this value to increase.
1818 $SETSTRIPE --stripe-index=0 --stripe-count=1 $DIR/$tfile ||
1819 error "$SETSTRIPE failed for $DIR/$tfile"
1821 multiop $DIR/$tfile Ow1yc
1822 REQ_DELAY=`lctl get_param -n osc.${FSNAME}-OST0000-osc-*.timeouts |
1823 awk '/portal 6/ {print $5}'`
1824 REQ_DELAY=$((${REQ_DELAY} + ${REQ_DELAY} / 4 + 5))
1826 do_facet ost1 lctl set_param fail_val=${REQ_DELAY}
1827 #define OBD_FAIL_OST_BRW_PAUSE_PACK 0x224
1828 do_facet ost1 $LCTL set_param fail_loc=0x224
1831 $SETSTRIPE --stripe-index=0 --stripe-count=1 $DIR/$tfile ||
1832 error "$SETSTRIPE failed"
1833 # force some real bulk transfer
1834 multiop $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c
1836 do_facet ost1 $LCTL set_param fail_loc=0
1837 # check for log message
1838 $LCTL dk | grep "Early reply #" || error "No early reply"
1840 # client should show REQ_DELAY estimates
1841 lctl get_param -n osc.${FSNAME}-OST0000-osc-*.timeouts | grep portal
1843 run_test 65b "AT: verify early replies on packed reply / bulk"
1845 test_66a() #bug 3055
1847 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1849 at_start || return 0
1850 lctl get_param -n mdc.${FSNAME}-MDT0000-mdc-*.timeouts | grep "portal 12"
1851 # adjust 5s at a time so no early reply is sent (within deadline)
1852 do_facet $SINGLEMDS "$LCTL set_param fail_val=5000"
1853 #define OBD_FAIL_PTLRPC_PAUSE_REQ 0x50a
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_val=10000"
1859 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x8000050a"
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 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0"
1865 createmany -o $DIR/$tfile 20 > /dev/null
1866 unlinkmany $DIR/$tfile 20 > /dev/null
1867 lctl get_param -n mdc.${FSNAME}-MDT0000-mdc-*.timeouts | grep "portal 12"
1868 CUR=$(lctl get_param -n mdc.${FSNAME}-MDT0000-mdc-*.timeouts | awk '/portal 12/ {print $5}')
1869 WORST=$(lctl get_param -n mdc.${FSNAME}-MDT0000-mdc-*.timeouts | awk '/portal 12/ {print $7}')
1870 echo "Current MDT timeout $CUR, worst $WORST"
1871 [ $CUR -lt $WORST ] || error "Current $CUR should be less than worst $WORST"
1873 run_test 66a "AT: verify MDT service time adjusts with no early replies"
1875 test_66b() #bug 3055
1877 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1879 at_start || return 0
1880 ORIG=$(lctl get_param -n mdc.${FSNAME}-MDT0000*.timeouts |
1881 awk '/network/ {print $4}')
1882 $LCTL set_param fail_val=$(($ORIG + 5))
1883 #define OBD_FAIL_PTLRPC_PAUSE_REP 0x50c
1884 $LCTL set_param fail_loc=0x50c
1885 ls $DIR/$tfile > /dev/null 2>&1
1886 $LCTL set_param fail_loc=0
1887 CUR=$(lctl get_param -n mdc.${FSNAME}-MDT0000*.timeouts |
1888 awk '/network/ {print $4}')
1889 WORST=$(lctl get_param -n mdc.${FSNAME}-MDT0000*.timeouts |
1890 awk '/network/ {print $6}')
1891 echo "network timeout orig $ORIG, cur $CUR, worst $WORST"
1892 [ $WORST -gt $ORIG ] ||
1893 error "Worst $WORST should be worse than orig $ORIG"
1895 run_test 66b "AT: verify net latency adjusts"
1897 test_67a() #bug 3055
1899 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1901 at_start || return 0
1902 CONN1=$(lctl get_param -n osc.*.stats | awk '/_connect/ {total+=$2} END {print total}')
1903 # sleeping threads may drive values above this
1904 do_facet ost1 "$LCTL set_param fail_val=400"
1905 #define OBD_FAIL_PTLRPC_PAUSE_REQ 0x50a
1906 do_facet ost1 "$LCTL set_param fail_loc=0x50a"
1907 createmany -o $DIR/$tfile 20 > /dev/null
1908 unlinkmany $DIR/$tfile 20 > /dev/null
1909 do_facet ost1 "$LCTL set_param fail_loc=0"
1910 CONN2=$(lctl get_param -n osc.*.stats | awk '/_connect/ {total+=$2} END {print total}')
1911 ATTEMPTS=$(($CONN2 - $CONN1))
1912 echo "$ATTEMPTS osc reconnect attempts on gradual slow"
1913 [ $ATTEMPTS -gt 0 ] &&
1914 error_ignore bz13721 "AT should have prevented reconnect"
1917 run_test 67a "AT: verify slow request processing doesn't induce reconnects"
1919 test_67b() #bug 3055
1921 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1923 at_start || return 0
1924 CONN1=$(lctl get_param -n osc.*.stats | awk '/_connect/ {total+=$2} END {print total}')
1926 # exhaust precreations on ost1
1927 local OST=$(ostname_from_index 0)
1928 local mdtosc=$(get_mdtosc_proc_path mds $OST)
1929 local last_id=$(do_facet $SINGLEMDS lctl get_param -n \
1930 osc.$mdtosc.prealloc_last_id)
1931 local next_id=$(do_facet $SINGLEMDS lctl get_param -n \
1932 osc.$mdtosc.prealloc_next_id)
1934 mkdir -p $DIR/$tdir/${OST} || error "mkdir $DIR/$tdir/${OST} failed"
1935 $SETSTRIPE -i 0 -c 1 $DIR/$tdir/${OST} || error "$SETSTRIPE failed"
1936 echo "Creating to objid $last_id on ost $OST..."
1937 #define OBD_FAIL_OST_PAUSE_CREATE 0x223
1938 do_facet ost1 "$LCTL set_param fail_val=20000"
1939 do_facet ost1 "$LCTL set_param fail_loc=0x80000223"
1940 createmany -o $DIR/$tdir/${OST}/f $next_id $((last_id - next_id + 2))
1943 do_facet ost1 "lctl get_param -n ost.OSS.ost_create.timeouts"
1945 CONN2=$(lctl get_param -n osc.*.stats | awk '/_connect/ {total+=$2} END {print total}')
1946 ATTEMPTS=$(($CONN2 - $CONN1))
1947 echo "$ATTEMPTS osc reconnect attempts on instant slow"
1948 # do it again; should not timeout
1949 do_facet ost1 "$LCTL set_param fail_loc=0x80000223"
1950 cp /etc/profile $DIR/$tfile || error "cp failed"
1951 do_facet ost1 "$LCTL set_param fail_loc=0"
1953 do_facet ost1 "lctl get_param -n ost.OSS.ost_create.timeouts"
1954 CONN3=$(lctl get_param -n osc.*.stats | awk '/_connect/ {total+=$2} END {print total}')
1955 ATTEMPTS=$(($CONN3 - $CONN2))
1956 echo "$ATTEMPTS osc reconnect attempts on 2nd slow"
1957 [ $ATTEMPTS -gt 0 ] && error "AT should have prevented reconnect"
1960 run_test 67b "AT: verify instant slowdown doesn't induce reconnects"
1962 test_68 () #bug 13813
1964 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1966 at_start || return 0
1967 local ldlm_enqueue_min=$(find /sys -name ldlm_enqueue_min)
1968 [ -z "$ldlm_enqueue_min" ] && skip "missing /sys/.../ldlm_enqueue_min" && return 0
1969 local ldlm_enqueue_min_r=$(do_facet ost1 "find /sys -name ldlm_enqueue_min")
1970 [ -z "$ldlm_enqueue_min_r" ] && skip "missing /sys/.../ldlm_enqueue_min in the ost1" && return 0
1971 local ENQ_MIN=$(cat $ldlm_enqueue_min)
1972 local ENQ_MIN_R=$(do_facet ost1 "cat $ldlm_enqueue_min_r")
1973 echo $TIMEOUT >> $ldlm_enqueue_min
1974 do_facet ost1 "echo $TIMEOUT >> $ldlm_enqueue_min_r"
1976 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
1977 $SETSTRIPE --stripe-index=0 -c 1 $DIR/$tdir ||
1978 error "$SETSTRIPE failed for $DIR/$tdir"
1979 #define OBD_FAIL_LDLM_PAUSE_CANCEL 0x312
1980 $LCTL set_param fail_val=$(($TIMEOUT - 1))
1981 $LCTL set_param fail_loc=0x80000312
1982 cp /etc/profile $DIR/$tdir/${tfile}_1 || error "1st cp failed $?"
1983 $LCTL set_param fail_val=$((TIMEOUT * 5 / 4))
1984 $LCTL set_param fail_loc=0x80000312
1985 cp /etc/profile $DIR/$tdir/${tfile}_2 || error "2nd cp failed $?"
1986 $LCTL set_param fail_loc=0
1988 echo $ENQ_MIN >> $ldlm_enqueue_min
1989 do_facet ost1 "echo $ENQ_MIN_R >> $ldlm_enqueue_min_r"
1993 run_test 68 "AT: verify slowing locks"
1996 # end of AT tests includes above lines
1998 # start multi-client tests
2000 [ -z "$CLIENTS" ] &&
2001 { skip "Need two or more clients." && return; }
2002 [ $CLIENTCOUNT -lt 2 ] &&
2003 { skip "Need two or more clients, have $CLIENTCOUNT" && return; }
2005 echo "mount clients $CLIENTS ..."
2006 zconf_mount_clients $CLIENTS $MOUNT
2008 local clients=${CLIENTS//,/ }
2009 echo "Write/read files on $DIR ; clients $CLIENTS ... "
2010 for CLIENT in $clients; do
2011 do_node $CLIENT dd bs=1M count=10 if=/dev/zero \
2012 of=$DIR/${tfile}_${CLIENT} 2>/dev/null ||
2013 error "dd failed on $CLIENT"
2016 local prev_client=$(echo $clients | sed 's/^.* \(.\+\)$/\1/')
2017 for C in ${CLIENTS//,/ }; do
2018 do_node $prev_client dd if=$DIR/${tfile}_${C} \
2019 of=/dev/null 2>/dev/null ||
2020 error "dd if=$DIR/${tfile}_${C} failed on $prev_client"
2026 run_test 70a "check multi client t-f"
2028 check_for_process () {
2033 killall_process $clients "$prog" -0
2037 local clients=${CLIENTS:-$HOSTNAME}
2039 zconf_mount_clients $clients $MOUNT
2042 [ "$SLOW" = "no" ] && duration=120
2043 # set duration to 900 because it takes some time to boot node
2044 [ "$FAILURE_MODE" = HARD ] && duration=900
2047 local start_ts=$(date +%s)
2048 local cmd="rundbench 1 -t $duration"
2050 if [ $MDSCOUNT -ge 2 ]; then
2051 test_mkdir -p -c$MDSCOUNT $DIR/$tdir
2052 $LFS setdirstripe -D -c$MDSCOUNT $DIR/$tdir
2054 do_nodesv $clients "set -x; MISSING_DBENCH_OK=$MISSING_DBENCH_OK \
2055 PATH=\$PATH:$LUSTRE/utils:$LUSTRE/tests/:$DBENCH_LIB \
2056 DBENCH_LIB=$DBENCH_LIB TESTSUITE=$TESTSUITE TESTNAME=$TESTNAME \
2057 MOUNT=$MOUNT DIR=$DIR/$tdir/\\\$(hostname) LCTL=$LCTL $cmd" &
2060 #LU-1897 wait for all dbench copies to start
2061 while ! check_for_process $clients dbench; do
2062 elapsed=$(($(date +%s) - start_ts))
2063 if [ $elapsed -gt $duration ]; then
2064 killall_process $clients dbench
2065 error "dbench failed to start on $clients!"
2070 log "Started rundbench load pid=$pid ..."
2072 elapsed=$(($(date +%s) - start_ts))
2073 local num_failovers=0
2075 while [ $elapsed -lt $duration ]; do
2076 if ! check_for_process $clients dbench; then
2077 error_noexit "dbench stopped on some of $clients!"
2078 killall_process $clients dbench
2082 replay_barrier mds$fail_index
2083 sleep 1 # give clients a time to do operations
2084 # Increment the number of failovers
2085 num_failovers=$((num_failovers+1))
2086 log "$TESTNAME fail mds$fail_index $num_failovers times"
2088 elapsed=$(($(date +%s) - start_ts))
2089 if [ $fail_index -ge $MDSCOUNT ]; then
2092 fail_index=$((fail_index+1))
2096 wait $pid || error "rundbench load on $clients failed!"
2098 run_test 70b "dbench ${MDSCOUNT}mdts recovery; $CLIENTCOUNT clients"
2099 # end multi-client tests
2104 local monitor_pid=$3
2106 local start_ts=$(date +%s)
2107 local num_failovers=0
2110 elapsed=$(($(date +%s) - start_ts))
2111 while [ $elapsed -lt $duration ]; do
2112 fail_index=$((RANDOM%max_index+1))
2113 kill -0 $monitor_pid ||
2114 error "$monitor_pid stopped"
2116 replay_barrier mds$fail_index
2118 # Increment the number of failovers
2119 num_failovers=$((num_failovers+1))
2120 log "$TESTNAME fail mds$fail_index $num_failovers times"
2122 elapsed=$(($(date +%s) - start_ts))
2128 rm -f $DIR/replay-single.70c.lck
2133 local clients=${CLIENTS:-$HOSTNAME}
2136 zconf_mount_clients $clients $MOUNT
2139 [ "$SLOW" = "no" ] && duration=180
2140 # set duration to 900 because it takes some time to boot node
2141 [ "$FAILURE_MODE" = HARD ] && duration=600
2144 local start_ts=$(date +%s)
2146 trap cleanup_70c EXIT
2148 while [ ! -e $DIR/replay-single.70c.lck ]; do
2149 test_mkdir -p -c$MDSCOUNT $DIR/$tdir || break
2150 if [ $MDSCOUNT -ge 2 ]; then
2151 $LFS setdirstripe -D -c$MDSCOUNT $DIR/$tdir ||
2152 error "set default dirstripe failed"
2154 cd $DIR/$tdir || break
2155 tar cf - /etc | tar xf - || error "tar failed in loop"
2159 echo "Started tar $tar_70c_pid"
2161 random_fail_mdt $MDSCOUNT $duration $tar_70c_pid
2162 kill -0 $tar_70c_pid || error "tar $tar_70c_pid stopped"
2164 touch $DIR/replay-single.70c.lck
2165 wait $tar_70c_pid || error "$?: tar failed"
2170 run_test 70c "tar ${MDSCOUNT}mdts recovery"
2174 kill -9 $mkdir_70d_pid
2178 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2179 local clients=${CLIENTS:-$HOSTNAME}
2182 zconf_mount_clients $clients $MOUNT
2185 [ "$SLOW" = "no" ] && duration=180
2186 # set duration to 900 because it takes some time to boot node
2187 [ "$FAILURE_MODE" = HARD ] && duration=900
2192 local start_ts=$(date +%s)
2194 trap cleanup_70d EXIT
2197 $LFS mkdir -i0 -c2 $DIR/$tdir/test || {
2201 $LFS mkdir -i1 -c2 $DIR/$tdir/test1 || {
2206 touch $DIR/$tdir/test/a || {
2210 mkdir $DIR/$tdir/test/b || {
2214 rm -rf $DIR/$tdir/test || {
2219 touch $DIR/$tdir/test1/a || {
2223 mkdir $DIR/$tdir/test1/b || {
2228 rm -rf $DIR/$tdir/test1 || {
2235 echo "Started $mkdir_70d_pid"
2237 random_fail_mdt $MDSCOUNT $duration $mkdir_70d_pid
2238 kill -0 $mkdir_70d_pid || error "mkdir/rmdir $mkdir_70d_pid stopped"
2243 run_test 70d "mkdir/rmdir striped dir ${MDSCOUNT}mdts recovery"
2247 kill -9 $rename_70e_pid
2251 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2252 local clients=${CLIENTS:-$HOSTNAME}
2255 echo ha > /proc/sys/lnet/debug
2256 zconf_mount_clients $clients $MOUNT
2259 [ "$SLOW" = "no" ] && duration=180
2260 # set duration to 900 because it takes some time to boot node
2261 [ "$FAILURE_MODE" = HARD ] && duration=900
2264 $LFS mkdir -i0 $DIR/$tdir/test_0
2265 $LFS mkdir -i0 $DIR/$tdir/test_1
2266 touch $DIR/$tdir/test_0/a
2267 touch $DIR/$tdir/test_1/b
2268 trap cleanup_70e EXIT
2271 mrename $DIR/$tdir/test_0/a $DIR/$tdir/test_1/b > \
2277 checkstat $DIR/$tdir/test_0/a && {
2278 echo "a still exists"
2282 checkstat $DIR/$tdir/test_1/b || {
2283 echo "b still exists"
2287 touch $DIR/$tdir/test_0/a || {
2288 echo "touch a fails"
2292 mrename $DIR/$tdir/test_1/b $DIR/$tdir/test_0/a > \
2300 echo "Started $rename_70e_pid"
2302 random_fail_mdt 2 $duration $rename_70e_pid
2303 kill -0 $rename_70e_pid || error "rename $rename_70e_pid stopped"
2308 run_test 70e "rename cross-MDT with random fails"
2310 test_70f_write_and_read(){
2315 echo "Write/read files in: '$DIR/$tdir', clients: '$CLIENTS' ..."
2316 for client in ${CLIENTS//,/ }; do
2317 [ -f $stopflag ] || return
2319 local tgtfile=$DIR/$tdir/$tfile.$client
2320 do_node $client dd $DD_OPTS bs=1M count=10 if=$srcfile \
2321 of=$tgtfile 2>/dev/null ||
2322 error "dd $DD_OPTS bs=1M count=10 if=$srcfile " \
2323 "of=$tgtfile failed on $client, rc=$?"
2326 local prev_client=$(echo ${CLIENTS//,/ } | awk '{ print $NF }')
2329 for client in ${CLIENTS//,/ }; do
2330 [ -f $stopflag ] || return
2332 # flush client cache in case test is running on only one client
2333 # do_node $client cancel_lru_locks osc
2334 do_node $client $LCTL set_param ldlm.namespaces.*.lru_size=clear
2336 tgtfile=$DIR/$tdir/$tfile.$client
2337 local md5=$(do_node $prev_client "md5sum $tgtfile")
2338 [ ${checksum[$index]// */} = ${md5// */} ] ||
2339 error "$tgtfile: checksum doesn't match on $prev_client"
2340 index=$((index + 1))
2350 mkdir -p $DIR/$tdir || error "cannot create $DIR/$tdir directory"
2351 $SETSTRIPE -c -1 $DIR/$tdir || error "cannot $SETSTRIPE $DIR/$tdir"
2354 while [ -f $stopflag ]; do
2355 test_70f_write_and_read $srcfile $stopflag
2356 # use direct IO and buffer cache in turns if loop
2357 [ -n "$DD_OPTS" ] && DD_OPTS="" || DD_OPTS="oflag=direct"
2361 test_70f_cleanup() {
2363 rm -f $TMP/$tfile.stop
2364 do_nodes $CLIENTS rm -f $TMP/$tfile
2365 rm -f $DIR/$tdir/$tfile.*
2369 # [ x$ost1failover_HOST = x$ost_HOST ] &&
2370 # { skip "Failover host not defined" && return; }
2371 # [ -z "$CLIENTS" ] &&
2372 # { skip "CLIENTS are not specified." && return; }
2373 # [ $CLIENTCOUNT -lt 2 ] &&
2374 # { skip "Need 2 or more clients, have $CLIENTCOUNT" && return; }
2376 [[ $(lustre_version_code ost1) -lt $(version_code 2.9.53) ]] &&
2377 skip "Need server version at least 2.9.53" && return
2379 echo "mount clients $CLIENTS ..."
2380 zconf_mount_clients $CLIENTS $MOUNT
2382 local srcfile=$TMP/$tfile
2386 trap test_70f_cleanup EXIT
2387 # create a different source file local to each client node so we can
2388 # detect if the file wasn't written out properly after failover
2389 do_nodes $CLIENTS dd bs=1M count=10 if=/dev/urandom of=$srcfile \
2390 2>/dev/null || error "can't create $srcfile on $CLIENTS"
2391 for client in ${CLIENTS//,/ }; do
2392 checksum[$index]=$(do_node $client "md5sum $srcfile")
2393 index=$((index + 1))
2397 [ "$SLOW" = "no" ] && duration=60
2398 # set duration to 900 because it takes some time to boot node
2399 [ "$FAILURE_MODE" = HARD ] && duration=900
2401 local stopflag=$TMP/$tfile.stop
2402 test_70f_loop $srcfile $stopflag &
2406 local num_failovers=0
2407 local start_ts=$SECONDS
2408 while [ $elapsed -lt $duration ]; do
2412 num_failovers=$((num_failovers + 1))
2413 log "$TESTNAME failing OST $num_failovers times"
2416 elapsed=$((SECONDS - start_ts))
2423 run_test 70f "OSS O_DIRECT recovery with $CLIENTCOUNT clients"
2427 kill -9 $mkdir_71a_pid
2430 random_double_fail_mdt() {
2433 local monitor_pid=$3
2435 local start_ts=$(date +%s)
2436 local num_failovers=0
2440 elapsed=$(($(date +%s) - start_ts))
2441 while [ $elapsed -lt $duration ]; do
2442 fail_index=$((RANDOM%max_index + 1))
2443 if [ $fail_index -eq $max_index ]; then
2446 second_index=$((fail_index + 1))
2448 kill -0 $monitor_pid ||
2449 error "$monitor_pid stopped"
2451 replay_barrier mds$fail_index
2452 replay_barrier mds$second_index
2454 # Increment the number of failovers
2455 num_failovers=$((num_failovers+1))
2456 log "fail mds$fail_index mds$second_index $num_failovers times"
2457 fail mds${fail_index},mds${second_index}
2458 elapsed=$(($(date +%s) - start_ts))
2463 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2464 local clients=${CLIENTS:-$HOSTNAME}
2467 zconf_mount_clients $clients $MOUNT
2470 [ "$SLOW" = "no" ] && duration=180
2471 # set duration to 900 because it takes some time to boot node
2472 [ "$FAILURE_MODE" = HARD ] && duration=900
2477 local start_ts=$(date +%s)
2479 trap cleanup_71a EXIT
2482 $LFS mkdir -i0 -c2 $DIR/$tdir/test
2483 rmdir $DIR/$tdir/test
2487 echo "Started $mkdir_71a_pid"
2489 random_double_fail_mdt 2 $duration $mkdir_71a_pid
2490 kill -0 $mkdir_71a_pid || error "mkdir/rmdir $mkdir_71a_pid stopped"
2495 run_test 71a "mkdir/rmdir striped dir with 2 mdts recovery"
2498 multiop_bg_pause $DIR/$tfile O_tSc ||
2499 error "multiop_bg_pause $DIR/$tfile failed"
2503 replay_barrier $SINGLEMDS
2504 #define OBD_FAIL_LDLM_ENQUEUE_NET 0x302
2505 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000302"
2508 wait $pid || error "multiop pid failed"
2509 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
2512 run_test 73a "open(O_CREAT), unlink, replay, reconnect before open replay, close"
2515 multiop_bg_pause $DIR/$tfile O_tSc ||
2516 error "multiop_bg_pause $DIR/$tfile failed"
2520 replay_barrier $SINGLEMDS
2521 #define OBD_FAIL_LDLM_REPLY 0x30c
2522 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000030c"
2525 wait $pid || error "multiop pid failed"
2526 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
2529 run_test 73b "open(O_CREAT), unlink, replay, reconnect at open_replay reply, close"
2533 local clients=${CLIENTS:-$HOSTNAME}
2535 zconf_umount_clients $clients $MOUNT
2537 facet_failover $SINGLEMDS
2538 zconf_mount_clients $clients $MOUNT
2540 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2541 rm $DIR/$tfile || error "rm $DIR/$tfile failed"
2542 clients_up || error "client evicted: $?"
2545 run_test 74 "Ensure applications don't fail waiting for OST recovery"
2547 remote_dir_check_80() {
2552 diridx=$($GETSTRIPE -M $remote_dir) ||
2553 error "$GETSTRIPE -M $remote_dir failed"
2554 [ $diridx -eq $MDTIDX ] || error "$diridx != $MDTIDX"
2556 createmany -o $remote_dir/f-%d 20 || error "creation failed"
2557 fileidx=$($GETSTRIPE -M $remote_dir/f-1) ||
2558 error "$GETSTRIPE -M $remote_dir/f-1 failed"
2559 [ $fileidx -eq $MDTIDX ] || error "$fileidx != $MDTIDX"
2565 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2566 ([ $FAILURE_MODE == "HARD" ] &&
2567 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2568 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2572 local remote_dir=$DIR/$tdir/remote_dir
2574 mkdir -p $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2575 #define OBD_FAIL_OUT_UPDATE_NET_REP 0x1701
2576 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
2577 $LFS mkdir -i $MDTIDX $remote_dir &
2583 wait $CLIENT_PID || error "remote creation failed"
2585 remote_dir_check_80 || error "remote dir check failed"
2586 rm -rf $DIR/$tdir || error "rmdir failed"
2590 run_test 80a "DNE: create remote dir, drop update rep from MDT0, fail MDT0"
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 #define OBD_FAIL_UPDATE_OBJ_NET_REP 0x1701
2604 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
2605 $LFS mkdir -i $MDTIDX $remote_dir &
2610 fail mds$((MDTIDX + 1))
2612 wait $CLIENT_PID || error "remote creation failed"
2614 remote_dir_check_80 || error "remote dir check failed"
2615 rm -rf $DIR/$tdir || error "rmdir failed"
2619 run_test 80b "DNE: create remote dir, drop update rep from MDT0, fail MDT1"
2622 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2623 ([ $FAILURE_MODE == "HARD" ] &&
2624 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2625 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2629 local remote_dir=$DIR/$tdir/remote_dir
2631 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2632 #define OBD_FAIL_UPDATE_OBJ_NET_REP 0x1701
2633 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
2634 $LFS mkdir -i $MDTIDX $remote_dir &
2640 fail mds$((MDTIDX + 1))
2642 wait $CLIENT_PID || error "remote creation failed"
2644 remote_dir_check_80 || error "remote dir check failed"
2645 rm -rf $DIR/$tdir || error "rmdir failed"
2649 run_test 80c "DNE: create remote dir, drop update rep from MDT1, fail MDT[0,1]"
2652 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2654 local remote_dir=$DIR/$tdir/remote_dir
2656 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2657 #define OBD_FAIL_UPDATE_OBJ_NET_REP 0x1701
2658 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
2659 $LFS mkdir -i $MDTIDX $remote_dir &
2662 # sleep 3 seconds to make sure MDTs are failed after
2663 # lfs mkdir -i has finished on all of MDTs.
2668 fail mds${MDTIDX},mds$((MDTIDX + 1))
2670 wait $CLIENT_PID || error "remote creation failed"
2672 remote_dir_check_80 || error "remote dir check failed"
2673 rm -rf $DIR/$tdir || error "rmdir failed"
2677 run_test 80d "DNE: create remote dir, drop update rep from MDT1, fail 2 MDTs"
2680 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2681 ([ $FAILURE_MODE == "HARD" ] &&
2682 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2683 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2687 local remote_dir=$DIR/$tdir/remote_dir
2689 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2690 # OBD_FAIL_MDS_REINT_NET_REP 0x119
2691 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
2692 $LFS mkdir -i $MDTIDX $remote_dir &
2695 # sleep 3 seconds to make sure MDTs are failed after
2696 # lfs mkdir -i has finished on all of MDTs.
2702 wait $CLIENT_PID || error "remote creation failed"
2704 remote_dir_check_80 || error "remote dir check failed"
2705 rm -rf $DIR/$tdir || error "rmdir failed"
2709 run_test 80e "DNE: create remote dir, drop MDT1 rep, fail MDT0"
2712 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2713 ([ $FAILURE_MODE == "HARD" ] &&
2714 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2715 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2718 local remote_dir=$DIR/$tdir/remote_dir
2720 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2721 # OBD_FAIL_MDS_REINT_NET_REP 0x119
2722 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
2723 $LFS mkdir -i $MDTIDX $remote_dir &
2727 fail mds$((MDTIDX + 1))
2729 wait $CLIENT_PID || error "remote creation failed"
2731 remote_dir_check_80 || error "remote dir check failed"
2732 rm -rf $DIR/$tdir || error "rmdir failed"
2736 run_test 80f "DNE: create remote dir, drop MDT1 rep, fail MDT1"
2739 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2740 ([ $FAILURE_MODE == "HARD" ] &&
2741 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2742 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2746 local remote_dir=$DIR/$tdir/remote_dir
2748 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2749 # OBD_FAIL_MDS_REINT_NET_REP 0x119
2750 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
2751 $LFS mkdir -i $MDTIDX $remote_dir &
2754 # sleep 3 seconds to make sure MDTs are failed after
2755 # lfs mkdir -i has finished on all of MDTs.
2761 fail mds$((MDTIDX + 1))
2763 wait $CLIENT_PID || error "remote creation failed"
2765 remote_dir_check_80 || error "remote dir check failed"
2766 rm -rf $DIR/$tdir || error "rmdir failed"
2770 run_test 80g "DNE: create remote dir, drop MDT1 rep, fail MDT0, then MDT1"
2773 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2775 local remote_dir=$DIR/$tdir/remote_dir
2777 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2778 # OBD_FAIL_MDS_REINT_NET_REP 0x119
2779 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
2780 $LFS mkdir -i $MDTIDX $remote_dir &
2783 # sleep 3 seconds to make sure MDTs are failed after
2784 # lfs mkdir -i has finished on all of MDTs.
2789 fail mds${MDTIDX},mds$((MDTIDX + 1))
2791 wait $CLIENT_PID || error "remote dir creation failed"
2793 remote_dir_check_80 || error "remote dir check failed"
2794 rm -rf $DIR/$tdir || error "rmdir failed"
2798 run_test 80h "DNE: create remote dir, drop MDT1 rep, fail 2 MDTs"
2801 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2802 ([ $FAILURE_MODE == "HARD" ] &&
2803 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2804 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2808 local remote_dir=$DIR/$tdir/remote_dir
2810 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2811 $LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
2813 touch $remote_dir || error "touch $remote_dir failed"
2814 # OBD_FAIL_OBJ_UPDATE_NET_REP 0x1701
2815 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
2820 fail mds$((MDTIDX + 1))
2822 wait $CLIENT_PID || error "rm remote dir failed"
2824 stat $remote_dir 2&>/dev/null && error "$remote_dir still exist!"
2826 rm -rf $DIR/$tdir || error "rmdir failed"
2830 run_test 81a "DNE: unlink remote dir, drop MDT0 update rep, fail MDT1"
2833 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2834 ([ $FAILURE_MODE == "HARD" ] &&
2835 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2836 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2839 local remote_dir=$DIR/$tdir/remote_dir
2841 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2842 $LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
2844 # OBD_FAIL_OBJ_UPDATE_NET_REP 0x1701
2845 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
2852 wait $CLIENT_PID || error "rm remote dir failed"
2854 stat $remote_dir 2&>/dev/null && error "$remote_dir still exist!"
2856 rm -rf $DIR/$tdir || error "rmdir failed"
2860 run_test 81b "DNE: unlink remote dir, drop MDT0 update reply, fail MDT0"
2863 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2864 ([ $FAILURE_MODE == "HARD" ] &&
2865 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2866 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2870 local remote_dir=$DIR/$tdir/remote_dir
2872 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2873 $LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
2875 # OBD_FAIL_OBJ_UPDATE_NET_REP 0x1701
2876 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
2883 fail mds$((MDTIDX + 1))
2885 wait $CLIENT_PID || error "rm remote dir failed"
2887 stat $remote_dir 2&>/dev/null && error "$remote_dir still exist!"
2889 rm -rf $DIR/$tdir || error "rmdir failed"
2893 run_test 81c "DNE: unlink remote dir, drop MDT0 update reply, fail MDT0,MDT1"
2896 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2898 local remote_dir=$DIR/$tdir/remote_dir
2900 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2901 $LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
2903 # OBD_FAIL_OBJ_UPDATE_NET_REP 0x1701
2904 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
2910 fail mds${MDTIDX},mds$((MDTIDX + 1))
2912 wait $CLIENT_PID || error "rm remote dir failed"
2914 stat $remote_dir 2&>/dev/null && error "$remote_dir still exist!"
2916 rm -rf $DIR/$tdir || error "rmdir failed"
2920 run_test 81d "DNE: unlink remote dir, drop MDT0 update reply, fail 2 MDTs"
2923 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2924 ([ $FAILURE_MODE == "HARD" ] &&
2925 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2926 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2930 local remote_dir=$DIR/$tdir/remote_dir
2932 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2933 $LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
2935 # OBD_FAIL_MDS_REINT_NET_REP 0x119
2936 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
2939 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0
2944 wait $CLIENT_PID || error "rm remote dir failed"
2946 stat $remote_dir 2&>/dev/null && error "$remote_dir still exist!"
2948 rm -rf $DIR/$tdir || error "rmdir failed"
2952 run_test 81e "DNE: unlink remote dir, drop MDT1 req reply, fail MDT0"
2955 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2956 ([ $FAILURE_MODE == "HARD" ] &&
2957 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2958 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2962 local remote_dir=$DIR/$tdir/remote_dir
2964 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2965 $LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
2967 # OBD_FAIL_MDS_REINT_NET_REP 0x119
2968 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
2973 fail mds$((MDTIDX + 1))
2975 wait $CLIENT_PID || error "rm remote dir failed"
2977 stat $remote_dir 2&>/dev/null && error "$remote_dir still exist!"
2979 rm -rf $DIR/$tdir || error "rmdir failed"
2983 run_test 81f "DNE: unlink remote dir, drop MDT1 req reply, fail MDT1"
2986 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2987 ([ $FAILURE_MODE == "HARD" ] &&
2988 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2989 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2993 local remote_dir=$DIR/$tdir/remote_dir
2995 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2996 $LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
2998 # OBD_FAIL_MDS_REINT_NET_REP 0x119
2999 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
3006 fail mds$((MDTIDX + 1))
3008 wait $CLIENT_PID || error "rm remote dir failed"
3010 stat $remote_dir 2&>/dev/null && error "$remote_dir still exist!"
3012 rm -rf $DIR/$tdir || error "rmdir failed"
3016 run_test 81g "DNE: unlink remote dir, drop req reply, fail M0, then M1"
3019 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3021 local remote_dir=$DIR/$tdir/remote_dir
3023 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3024 $LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
3026 # OBD_FAIL_MDS_REINT_NET_REP 0x119
3027 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
3033 fail mds${MDTIDX},mds$((MDTIDX + 1))
3035 wait $CLIENT_PID || error "rm remote dir failed"
3037 stat $remote_dir 2&>/dev/null && error "$remote_dir still exist!"
3039 rm -rf $DIR/$tdir || error "rmdir failed"
3043 run_test 81h "DNE: unlink remote dir, drop request reply, fail 2 MDTs"
3046 #define OBD_FAIL_MDS_OPEN_WAIT_CREATE 0x144
3047 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000144"
3048 createmany -o $DIR/$tfile- 1 &
3052 client_up || client_up || true # reconnect
3054 run_test 84a "stale open during export disconnect"
3056 test_85a() { #bug 16774
3057 lctl set_param -n ldlm.cancel_unused_locks_before_replay "1"
3059 for i in $(seq 100); do
3060 echo "tag-$i" > $DIR/$tfile-$i
3061 grep -q "tag-$i" $DIR/$tfile-$i || error "f2-$i"
3064 lov_id=$(lctl dl | grep "clilov")
3065 addr=$(echo $lov_id | awk '{print $4}' | awk -F '-' '{print $NF}')
3066 count=$(lctl get_param -n \
3067 ldlm.namespaces.*MDT0000*$addr.lock_unused_count)
3068 echo "before recovery: unused locks count = $count"
3072 count2=$(lctl get_param -n \
3073 ldlm.namespaces.*MDT0000*$addr.lock_unused_count)
3074 echo "after recovery: unused locks count = $count2"
3076 if [ $count2 -ge $count ]; then
3077 error "unused locks are not canceled"
3080 run_test 85a "check the cancellation of unused locks during recovery(IBITS)"
3082 test_85b() { #bug 16774
3083 lctl set_param -n ldlm.cancel_unused_locks_before_replay "1"
3085 create_pool $FSNAME.$TESTNAME ||
3086 error "unable to create pool $TESTNAME"
3087 do_facet mgs $LCTL pool_add $FSNAME.$TESTNAME $FSNAME-OST0000 ||
3088 error "unable to add pool $TESTNAME"
3090 $SETSTRIPE -c 1 -p $FSNAME.$TESTNAME $DIR
3092 for i in $(seq 100); do
3093 dd if=/dev/urandom of=$DIR/$tfile-$i bs=4096 \
3094 count=32 >/dev/null 2>&1
3097 cancel_lru_locks osc
3099 for i in $(seq 100); do
3100 dd if=$DIR/$tfile-$i of=/dev/null bs=4096 \
3101 count=32 >/dev/null 2>&1
3104 lov_id=$(lctl dl | grep "clilov")
3105 addr=$(echo $lov_id | awk '{print $4}' | awk -F '-' '{print $NF}')
3106 count=$(lctl get_param -n \
3107 ldlm.namespaces.*OST0000*$addr.lock_unused_count)
3108 echo "before recovery: unused locks count = $count"
3109 [ $count -ne 0 ] || error "unused locks ($count) should be zero"
3113 count2=$(lctl get_param \
3114 -n ldlm.namespaces.*OST0000*$addr.lock_unused_count)
3115 echo "after recovery: unused locks count = $count2"
3117 do_facet mgs $LCTL pool_remove $FSNAME.$TESTNAME $FSNAME-OST0000 ||
3118 error "unable to remove pool $TESTNAME"
3119 do_facet mgs $LCTL pool_destroy $FSNAME.$TESTNAME ||
3120 error "unable to destroy the pool $TESTNAME"
3122 if [ $count2 -ge $count ]; then
3123 error "unused locks are not canceled"
3126 run_test 85b "check the cancellation of unused locks during recovery(EXTENT)"
3129 local clients=${CLIENTS:-$HOSTNAME}
3131 zconf_umount_clients $clients $MOUNT
3132 do_facet $SINGLEMDS lctl set_param mdt.${FSNAME}-MDT*.exports.clear=0
3133 remount_facet $SINGLEMDS
3134 zconf_mount_clients $clients $MOUNT
3136 run_test 86 "umount server after clear nid_stats should not hit LBUG"
3139 do_facet ost1 "lctl set_param -n obdfilter.${ost1_svc}.sync_journal 0"
3142 $SETSTRIPE -i 0 -c 1 $DIR/$tfile
3143 dd if=/dev/urandom of=$DIR/$tfile bs=1024k count=8 ||
3144 error "dd to $DIR/$tfile failed"
3145 cksum=$(md5sum $DIR/$tfile | awk '{print $1}')
3146 cancel_lru_locks osc
3148 dd if=$DIR/$tfile of=/dev/null bs=1024k count=8 || error "Cannot read"
3149 cksum2=$(md5sum $DIR/$tfile | awk '{print $1}')
3150 if [ $cksum != $cksum2 ] ; then
3151 error "New checksum $cksum2 does not match original $cksum"
3154 run_test 87a "write replay"
3157 do_facet ost1 "lctl set_param -n obdfilter.${ost1_svc}.sync_journal 0"
3160 $SETSTRIPE -i 0 -c 1 $DIR/$tfile
3161 dd if=/dev/urandom of=$DIR/$tfile bs=1024k count=8 ||
3162 error "dd to $DIR/$tfile failed"
3163 sleep 1 # Give it a chance to flush dirty data
3164 echo TESTTEST | dd of=$DIR/$tfile bs=1 count=8 seek=64
3165 cksum=$(md5sum $DIR/$tfile | awk '{print $1}')
3166 cancel_lru_locks osc
3168 dd if=$DIR/$tfile of=/dev/null bs=1024k count=8 || error "Cannot read"
3169 cksum2=$(md5sum $DIR/$tfile | awk '{print $1}')
3170 if [ $cksum != $cksum2 ] ; then
3171 error "New checksum $cksum2 does not match original $cksum"
3174 run_test 87b "write replay with changed data (checksum resend)"
3176 test_88() { #bug 17485
3177 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3178 mkdir -p $TMP/$tdir || error "mkdir $TMP/$tdir failed"
3180 $SETSTRIPE -i 0 -c 1 $DIR/$tdir || error "$SETSTRIPE"
3183 replay_barrier $SINGLEMDS
3185 # exhaust precreations on ost1
3186 local OST=$(ostname_from_index 0)
3187 local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS $OST)
3188 local last_id=$(do_facet $SINGLEMDS lctl get_param -n osc.$mdtosc.prealloc_last_id)
3189 local next_id=$(do_facet $SINGLEMDS lctl get_param -n osc.$mdtosc.prealloc_next_id)
3190 echo "before test: last_id = $last_id, next_id = $next_id"
3192 echo "Creating to objid $last_id on ost $OST..."
3193 createmany -o $DIR/$tdir/f-%d $next_id $((last_id - next_id + 2)) ||
3194 error "createmany create files to last_id failed"
3196 #create some files to use some uncommitted objids
3197 last_id=$(($last_id + 1))
3198 createmany -o $DIR/$tdir/f-%d $last_id 8 ||
3199 error "createmany create files with uncommitted objids failed"
3201 last_id2=$(do_facet $SINGLEMDS lctl get_param -n osc.$mdtosc.prealloc_last_id)
3202 next_id2=$(do_facet $SINGLEMDS lctl get_param -n osc.$mdtosc.prealloc_next_id)
3203 echo "before recovery: last_id = $last_id2, next_id = $next_id2"
3205 # if test uses shutdown_facet && reboot_facet instead of facet_failover ()
3206 # it has to take care about the affected facets, bug20407
3207 local affected_mds1=$(affected_facets mds1)
3208 local affected_ost1=$(affected_facets ost1)
3210 shutdown_facet $SINGLEMDS
3213 reboot_facet $SINGLEMDS
3214 change_active $affected_mds1
3215 wait_for_facet $affected_mds1
3216 mount_facets $affected_mds1 || error "Restart of mds failed"
3219 change_active $affected_ost1
3220 wait_for_facet $affected_ost1
3221 mount_facets $affected_ost1 || error "Restart of ost1 failed"
3225 last_id2=$(do_facet $SINGLEMDS lctl get_param -n osc.$mdtosc.prealloc_last_id)
3226 next_id2=$(do_facet $SINGLEMDS lctl get_param -n osc.$mdtosc.prealloc_next_id)
3227 echo "after recovery: last_id = $last_id2, next_id = $next_id2"
3229 # create new files, which should use new objids, and ensure the orphan
3230 # cleanup phase for ost1 is completed at the same time
3231 for i in $(seq 8); do
3232 file_id=$(($last_id + 10 + $i))
3233 dd if=/dev/urandom of=$DIR/$tdir/f-$file_id bs=4096 count=128
3236 # if the objids were not recreated, then "ls" will fail with -ENOENT
3237 ls -l $DIR/$tdir/* || error "can't get the status of precreated files"
3240 # write into previously created files
3241 for i in $(seq 8); do
3242 file_id=$(($last_id + $i))
3243 dd if=/dev/urandom of=$DIR/$tdir/f-$file_id bs=4096 count=128
3244 cp -f $DIR/$tdir/f-$file_id $TMP/$tdir/
3247 # compare the content
3248 for i in $(seq 8); do
3249 file_id=$(($last_id + $i))
3250 cmp $TMP/$tdir/f-$file_id $DIR/$tdir/f-$file_id ||
3251 error "the content of file is modified!"
3256 run_test 88 "MDS should not assign same objid to different files "
3259 cancel_lru_locks osc
3260 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3261 rm -f $DIR/$tdir/$tfile
3263 wait_delete_completed
3264 BLOCKS1=$(df -P $MOUNT | tail -n 1 | awk '{ print $3 }')
3265 $SETSTRIPE -i 0 -c 1 $DIR/$tdir/$tfile
3266 dd if=/dev/zero bs=1M count=10 of=$DIR/$tdir/$tfile
3269 facet_failover $SINGLEMDS
3270 rm $DIR/$tdir/$tfile
3273 zconf_mount $(hostname) $MOUNT || error "mount fails"
3274 client_up || error "client_up failed"
3276 wait_delete_completed
3277 BLOCKS2=$(df -P $MOUNT | tail -n 1 | awk '{ print $3 }')
3278 [ $((BLOCKS2 - BLOCKS1)) -le 4 ] ||
3279 error $((BLOCKS2 - BLOCKS1)) blocks leaked
3282 run_test 89 "no disk space leak on late ost connection"
3288 change_active $facet
3289 wait_for_facet $facet
3290 mount_facet $facet || error "Restart of $facet failed"
3294 test_90() { # bug 19494
3295 local dir=$DIR/$tdir
3296 local ostfail=$(get_random_entry $(get_facets OST))
3298 if [[ $FAILURE_MODE = HARD ]]; then
3299 local affected=$(affected_facets $ostfail);
3300 if [[ "$affected" != $ostfail ]]; then
3301 skip not functional with FAILURE_MODE=$FAILURE_MODE, affected: $affected
3305 # ensure all OSTs are active to allow allocations
3308 mkdir $dir || error "mkdir $dir failed"
3310 echo "Create the files"
3312 # file "f${index}" striped over 1 OST
3313 # file "all" striped over all OSTs
3315 $SETSTRIPE -c $OSTCOUNT $dir/all ||
3316 error "setstripe failed to create $dir/all"
3318 for (( i=0; i<$OSTCOUNT; i++ )); do
3320 $SETSTRIPE -i $i -c 1 $f || error "$SETSTRIPE failed to create $f"
3322 # confirm setstripe actually created the stripe on the requested OST
3323 local uuid=$(ostuuid_from_index $i)
3324 for file in f$i all; do
3325 if [[ $dir/$file != $($LFS find --obd $uuid --name $file $dir) ]]; then
3326 $GETSTRIPE $dir/$file
3327 error wrong stripe: $file, uuid: $uuid
3332 # Before failing an OST, get its obd name and index
3333 local varsvc=${ostfail}_svc
3334 local obd=$(do_facet $ostfail lctl get_param \
3335 -n obdfilter.${!varsvc}.uuid)
3336 local index=$(($(facet_number $ostfail) - 1))
3338 echo "Fail $ostfail $obd, display the list of affected files"
3339 shutdown_facet $ostfail || error "shutdown_facet $ostfail failed"
3341 trap "cleanup_90 $ostfail" EXIT INT
3342 echo "General Query: lfs find $dir"
3343 local list=$($LFS find $dir)
3345 for (( i=0; i<$OSTCOUNT; i++ )); do
3346 list_member "$list" $dir/f$i ||
3347 error_noexit "lfs find $dir: no file f$i"
3349 list_member "$list" $dir/all ||
3350 error_noexit "lfs find $dir: no file all"
3352 # focus on the missing OST,
3353 # we expect to see only two files affected: "f$(index)" and "all"
3355 echo "Querying files on shutdown $ostfail: lfs find --obd $obd"
3356 list=$($LFS find --obd $obd $dir)
3358 for file in all f$index; do
3359 list_member "$list" $dir/$file ||
3360 error_noexit "lfs find does not report the affected $obd for $file"
3363 [[ $(echo $list | wc -w) -eq 2 ]] ||
3364 error_noexit "lfs find reports the wrong list of affected files ${#list[@]}"
3366 echo "Check getstripe: $GETSTRIPE -r --obd $obd"
3367 list=$($GETSTRIPE -r --obd $obd $dir)
3369 for file in all f$index; do
3370 echo "$list" | grep $dir/$file ||
3371 error_noexit "lfs getsripe does not report the affected $obd for $file"
3376 run_test 90 "lfs find identifies the missing striped file segments"
3379 local server_version=$(lustre_version_code $SINGLEMDS)
3380 [[ $server_version -ge $(version_code 2.6.90) ]] ||
3381 [[ $server_version -ge $(version_code 2.5.4) &&
3382 $server_version -lt $(version_code 2.5.50) ]] ||
3383 { skip "Need MDS version 2.5.4+ or 2.6.90+"; return; }
3385 cancel_lru_locks osc
3387 $SETSTRIPE -i 0 -c 1 $DIR/$tfile ||
3388 error "$SETSTRIPE $DIR/$tfile failed"
3389 dd if=/dev/zero of=$DIR/$tfile bs=1024 count=1 ||
3390 error "dd to $DIR/$tfile failed"
3391 #define OBD_FAIL_TGT_REPLAY_RECONNECT 0x715
3392 # We need to emulate a state that OST is waiting for other clients
3393 # not completing the recovery. Final ping is queued, but reply will be
3394 # sent on the recovery completion. It is done by sleep before
3395 # processing final pings
3396 do_facet ost1 "$LCTL set_param fail_val=40"
3397 do_facet ost1 "$LCTL set_param fail_loc=0x715"
3400 run_test 93a "replay + reconnect"
3403 local server_version=$(lustre_version_code $SINGLEMDS)
3404 [[ $server_version -ge $(version_code 2.7.90) ]] ||
3405 { skip "Need MDS version 2.7.90+"; return; }
3407 cancel_lru_locks mdc
3409 createmany -o $DIR/$tfile 20 ||
3410 error "createmany -o $DIR/$tfile failed"
3412 #define OBD_FAIL_TGT_REPLAY_RECONNECT 0x715
3413 # We need to emulate a state that MDT is waiting for other clients
3414 # not completing the recovery. Final ping is queued, but reply will be
3415 # sent on the recovery completion. It is done by sleep before
3416 # processing final pings
3417 do_facet mds1 "$LCTL set_param fail_val=80"
3418 do_facet mds1 "$LCTL set_param fail_loc=0x715"
3421 run_test 93b "replay + reconnect on mds"
3423 striped_dir_check_100() {
3424 local striped_dir=$DIR/$tdir/striped_dir
3425 local stripe_count=$($LFS getdirstripe -c $striped_dir)
3427 $LFS getdirstripe $striped_dir
3428 [ $stripe_count -eq 2 ] || error "$stripe_count != 2"
3430 createmany -o $striped_dir/f-%d 20 ||
3431 error "creation failed under striped dir"
3435 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3436 ([ $FAILURE_MODE == "HARD" ] &&
3437 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3438 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3441 local striped_dir=$DIR/$tdir/striped_dir
3444 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3446 #To make sure MDT1 and MDT0 are connected
3447 #otherwise it may create single stripe dir here
3448 $LFS setdirstripe -i1 $DIR/$tdir/remote_dir
3450 #define OBD_FAIL_OUT_UPDATE_NET_REP 0x1701
3451 do_facet mds$((MDTIDX+1)) lctl set_param fail_loc=0x1701
3452 $LFS setdirstripe -i0 -c2 $striped_dir &
3455 fail mds$((MDTIDX + 1))
3457 wait $CLIENT_PID || error "striped dir creation failed"
3459 striped_dir_check_100 || error "striped dir check failed"
3460 rm -rf $DIR/$tdir || error "rmdir failed"
3462 run_test 100a "DNE: create striped dir, drop update rep from MDT1, fail MDT1"
3465 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3466 ([ $FAILURE_MODE == "HARD" ] &&
3467 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3468 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3471 local striped_dir=$DIR/$tdir/striped_dir
3474 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3476 #To make sure MDT1 and MDT0 are connected
3477 #otherwise it may create single stripe dir here
3478 $LFS setdirstripe -i1 $DIR/$tdir/remote_dir
3480 # OBD_FAIL_MDS_REINT_NET_REP 0x119
3481 do_facet mds$MDTIDX lctl set_param fail_loc=0x119
3482 $LFS mkdir -i0 -c2 $striped_dir &
3487 wait $CLIENT_PID || error "striped dir creation failed"
3489 striped_dir_check_100 || error "striped dir check failed"
3490 rm -rf $DIR/$tdir || error "rmdir failed"
3492 run_test 100b "DNE: create striped dir, fail MDT0"
3494 test_101() { #LU-5648
3495 mkdir -p $DIR/$tdir/d1
3496 mkdir -p $DIR/$tdir/d2
3497 touch $DIR/$tdir/file0
3500 replay_barrier $SINGLEMDS
3501 for i in $(seq $num) ; do
3502 echo test$i > $DIR/$tdir/d1/file$i
3505 fail_abort $SINGLEMDS
3506 for i in $(seq $num) ; do
3507 touch $DIR/$tdir/d2/file$i
3508 test -s $DIR/$tdir/d2/file$i &&
3509 ls -al $DIR/$tdir/d2/file$i && error "file$i's size > 0"
3514 run_test 101 "Shouldn't reassign precreated objs to other files after recovery"
3523 [[ $(lctl get_param mdc.*.import |
3524 grep "connect_flags:.*multi_mod_rpc") ]] ||
3525 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
3527 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3528 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
3529 facet="mds$((0x$idx + 1))"
3531 # get current value of max_mod_rcps_in_flight
3532 num=$($LCTL get_param -n \
3533 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
3534 # set default value if client does not support multi mod RPCs
3535 [ -z "$num" ] && num=1
3537 echo "creating $num files ..."
3539 for i in $(seq $num); do
3540 touch $DIR/$tdir/file-$i
3543 # drop request on MDT to force resend
3544 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
3545 do_facet $facet "$LCTL set_param fail_loc=0x159"
3546 echo "launch $num chmod in parallel ($(date +%H:%M:%S)) ..."
3547 for i in $(seq $num); do
3548 chmod 0600 $DIR/$tdir/file-$i &
3552 do_facet $facet "$LCTL set_param fail_loc=0"
3553 for pid in $pids; do
3554 wait $pid || error "chmod failed"
3556 echo "done ($(date +%H:%M:%S))"
3558 # check chmod succeed
3559 for i in $(seq $num); do
3560 checkstat -vp 0600 $DIR/$tdir/file-$i
3565 run_test 102a "check resend (request lost) with multiple modify RPCs in flight"
3574 [[ $(lctl get_param mdc.*.import |
3575 grep "connect_flags:.*multi_mod_rpc") ]] ||
3576 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
3578 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3579 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
3580 facet="mds$((0x$idx + 1))"
3582 # get current value of max_mod_rcps_in_flight
3583 num=$($LCTL get_param -n \
3584 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
3585 # set default value if client does not support multi mod RPCs
3586 [ -z "$num" ] && num=1
3588 echo "creating $num files ..."
3590 for i in $(seq $num); do
3591 touch $DIR/$tdir/file-$i
3594 # drop reply on MDT to force reconstruction
3595 #define OBD_FAIL_MDS_REINT_MULTI_NET_REP 0x15a
3596 do_facet $facet "$LCTL set_param fail_loc=0x15a"
3597 echo "launch $num chmod in parallel ($(date +%H:%M:%S)) ..."
3598 for i in $(seq $num); do
3599 chmod 0600 $DIR/$tdir/file-$i &
3603 do_facet $facet "$LCTL set_param fail_loc=0"
3604 for pid in $pids; do
3605 wait $pid || error "chmod failed"
3607 echo "done ($(date +%H:%M:%S))"
3609 # check chmod succeed
3610 for i in $(seq $num); do
3611 checkstat -vp 0600 $DIR/$tdir/file-$i
3616 run_test 102b "check resend (reply lost) with multiple modify RPCs in flight"
3625 [[ $(lctl get_param mdc.*.import |
3626 grep "connect_flags:.*multi_mod_rpc") ]] ||
3627 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
3629 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3630 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
3631 facet="mds$((0x$idx + 1))"
3633 # get current value of max_mod_rcps_in_flight
3634 num=$($LCTL get_param -n \
3635 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
3636 # set default value if client does not support multi mod RPCs
3637 [ -z "$num" ] && num=1
3639 echo "creating $num files ..."
3641 for i in $(seq $num); do
3642 touch $DIR/$tdir/file-$i
3645 replay_barrier $facet
3648 #define OBD_FAIL_MDS_REINT_MULTI_NET_REP 0x15a
3649 do_facet $facet "$LCTL set_param fail_loc=0x15a"
3650 echo "launch $num chmod in parallel ($(date +%H:%M:%S)) ..."
3651 for i in $(seq $num); do
3652 chmod 0600 $DIR/$tdir/file-$i &
3656 do_facet $facet "$LCTL set_param fail_loc=0"
3661 for pid in $pids; do
3662 wait $pid || error "chmod failed"
3664 echo "done ($(date +%H:%M:%S))"
3666 # check chmod succeed
3667 for i in $(seq $num); do
3668 checkstat -vp 0600 $DIR/$tdir/file-$i
3673 run_test 102c "check replay w/o reconstruction with multiple mod RPCs in flight"
3682 [[ $(lctl get_param mdc.*.import |
3683 grep "connect_flags:.*multi_mod_rpc") ]] ||
3684 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
3686 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3687 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
3688 facet="mds$((0x$idx + 1))"
3690 # get current value of max_mod_rcps_in_flight
3691 num=$($LCTL get_param -n \
3692 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
3693 # set default value if client does not support multi mod RPCs
3694 [ -z "$num" ] && num=1
3696 echo "creating $num files ..."
3698 for i in $(seq $num); do
3699 touch $DIR/$tdir/file-$i
3703 #define OBD_FAIL_MDS_REINT_MULTI_NET_REP 0x15a
3704 do_facet $facet "$LCTL set_param fail_loc=0x15a"
3705 echo "launch $num chmod in parallel ($(date +%H:%M:%S)) ..."
3706 for i in $(seq $num); do
3707 chmod 0600 $DIR/$tdir/file-$i &
3712 # write MDT transactions to disk
3713 do_facet $facet "sync; sync; sync"
3715 do_facet $facet "$LCTL set_param fail_loc=0"
3720 for pid in $pids; do
3721 wait $pid || error "chmod failed"
3723 echo "done ($(date +%H:%M:%S))"
3725 # check chmod succeed
3726 for i in $(seq $num); do
3727 checkstat -vp 0600 $DIR/$tdir/file-$i
3732 run_test 102d "check replay & reconstruction with multiple mod RPCs in flight"
3735 remote_mds_nodsh && skip "remote MDS with nodsh" && return
3736 #define OBD_FAIL_MDS_TRACK_OVERFLOW 0x162
3737 do_facet mds1 $LCTL set_param fail_loc=0x80000162
3740 createmany -o $DIR/$tdir/t- 30 ||
3741 error "create files on remote directory failed"
3743 rm -rf $DIR/$tdir/t-*
3745 #MDS should crash with tr->otr_next_id overflow
3748 run_test 103 "Check otr_next_id overflow"
3751 check_striped_dir_110()
3753 $CHECKSTAT -t dir $DIR/$tdir/striped_dir ||
3754 error "create striped dir failed"
3755 local stripe_count=$($LFS getdirstripe -c $DIR/$tdir/striped_dir)
3756 [ $stripe_count -eq $MDSCOUNT ] ||
3757 error "$stripe_count != 2 after recovery"
3761 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3762 ([ $FAILURE_MODE == "HARD" ] &&
3763 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3764 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3769 $LFS mkdir -i1 -c$MDSCOUNT $DIR/$tdir/striped_dir
3772 check_striped_dir_110 || error "check striped_dir failed"
3773 rm -rf $DIR/$tdir || error "rmdir failed"
3777 run_test 110a "DNE: create striped dir, fail MDT1"
3780 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3781 ([ $FAILURE_MODE == "HARD" ] &&
3782 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3783 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3788 $LFS mkdir -i1 -c$MDSCOUNT $DIR/$tdir/striped_dir
3791 zconf_mount $(hostname) $MOUNT
3792 client_up || return 1
3794 check_striped_dir_110 || error "check striped_dir failed"
3796 rm -rf $DIR/$tdir || error "rmdir failed"
3800 run_test 110b "DNE: create striped dir, fail MDT1 and client"
3803 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3804 ([ $FAILURE_MODE == "HARD" ] &&
3805 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3806 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3811 $LFS mkdir -i1 -c$MDSCOUNT $DIR/$tdir/striped_dir
3814 check_striped_dir_110 || error "check striped_dir failed"
3816 rm -rf $DIR/$tdir || error "rmdir failed"
3820 run_test 110c "DNE: create striped dir, fail MDT2"
3823 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3824 ([ $FAILURE_MODE == "HARD" ] &&
3825 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3826 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3831 $LFS mkdir -i1 -c$MDSCOUNT $DIR/$tdir/striped_dir
3834 zconf_mount $(hostname) $MOUNT
3835 client_up || return 1
3837 check_striped_dir_110 || error "check striped_dir failed"
3839 rm -rf $DIR/$tdir || error "rmdir failed"
3843 run_test 110d "DNE: create striped dir, fail MDT2 and client"
3846 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3847 ([ $FAILURE_MODE == "HARD" ] &&
3848 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3849 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3854 $LFS mkdir -i1 -c$MDSCOUNT $DIR/$tdir/striped_dir
3858 zconf_mount $(hostname) $MOUNT
3859 client_up || return 1
3861 check_striped_dir_110 || error "check striped_dir failed"
3863 rm -rf $DIR/$tdir || error "rmdir failed"
3867 run_test 110e "DNE: create striped dir, uncommit on MDT2, fail client/MDT1/MDT2"
3870 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3871 ([ $FAILURE_MODE == "HARD" ] &&
3872 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3873 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3879 $LFS mkdir -i1 -c$MDSCOUNT $DIR/$tdir/striped_dir
3882 check_striped_dir_110 || error "check striped_dir failed"
3884 rm -rf $DIR/$tdir || error "rmdir failed"
3888 run_test 110f "DNE: create striped dir, fail MDT1/MDT2"
3891 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3892 ([ $FAILURE_MODE == "HARD" ] &&
3893 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3894 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3899 $LFS mkdir -i1 -c$MDSCOUNT $DIR/$tdir/striped_dir
3903 zconf_mount $(hostname) $MOUNT
3904 client_up || return 1
3906 check_striped_dir_110 || error "check striped_dir failed"
3908 rm -rf $DIR/$tdir || error "rmdir failed"
3912 run_test 110g "DNE: create striped dir, uncommit on MDT1, fail client/MDT1/MDT2"
3915 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3916 ([ $FAILURE_MODE == "HARD" ] &&
3917 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3918 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3922 $LFS mkdir -i1 -c2 $DIR/$tdir/striped_dir
3924 rm -rf $DIR/$tdir/striped_dir
3927 $CHECKSTAT -t dir $DIR/$tdir/striped_dir &&
3928 error "striped dir still exists"
3931 run_test 111a "DNE: unlink striped dir, fail MDT1"
3934 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3935 ([ $FAILURE_MODE == "HARD" ] &&
3936 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3937 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3941 $LFS mkdir -i1 -c2 $DIR/$tdir/striped_dir
3943 rm -rf $DIR/$tdir/striped_dir
3946 zconf_mount $(hostname) $MOUNT
3947 client_up || return 1
3949 $CHECKSTAT -t dir $DIR/$tdir/striped_dir &&
3950 error "striped dir still exists"
3953 run_test 111b "DNE: unlink striped dir, fail MDT2"
3956 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3957 ([ $FAILURE_MODE == "HARD" ] &&
3958 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3959 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3963 $LFS mkdir -i1 -c2 $DIR/$tdir/striped_dir
3965 rm -rf $DIR/$tdir/striped_dir
3969 zconf_mount $(hostname) $MOUNT
3970 client_up || return 1
3971 $CHECKSTAT -t dir $DIR/$tdir/striped_dir &&
3972 error "striped dir still exists"
3975 run_test 111c "DNE: unlink striped dir, uncommit on MDT1, fail client/MDT1/MDT2"
3978 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3979 ([ $FAILURE_MODE == "HARD" ] &&
3980 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3981 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3985 $LFS mkdir -i1 -c2 $DIR/$tdir/striped_dir
3987 rm -rf $DIR/$tdir/striped_dir
3991 zconf_mount $(hostname) $MOUNT
3992 client_up || return 1
3993 $CHECKSTAT -t dir $DIR/$tdir/striped_dir &&
3994 error "striped dir still exists"
3998 run_test 111d "DNE: unlink striped dir, uncommit on MDT2, fail client/MDT1/MDT2"
4001 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
4002 ([ $FAILURE_MODE == "HARD" ] &&
4003 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4004 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4008 $LFS mkdir -i1 -c2 $DIR/$tdir/striped_dir
4010 rm -rf $DIR/$tdir/striped_dir
4013 $CHECKSTAT -t dir $DIR/$tdir/striped_dir &&
4014 error "striped dir still exists"
4017 run_test 111e "DNE: unlink striped dir, uncommit on MDT2, fail MDT1/MDT2"
4020 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
4021 ([ $FAILURE_MODE == "HARD" ] &&
4022 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4023 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4027 $LFS mkdir -i1 -c2 $DIR/$tdir/striped_dir
4029 rm -rf $DIR/$tdir/striped_dir
4032 $CHECKSTAT -t dir $DIR/$tdir/striped_dir &&
4033 error "striped dir still exists"
4036 run_test 111f "DNE: unlink striped dir, uncommit on MDT1, fail MDT1/MDT2"
4039 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
4040 ([ $FAILURE_MODE == "HARD" ] &&
4041 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4042 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4046 $LFS mkdir -i1 -c2 $DIR/$tdir/striped_dir
4049 rm -rf $DIR/$tdir/striped_dir
4051 $CHECKSTAT -t dir $DIR/$tdir/striped_dir &&
4052 error "striped dir still exists"
4055 run_test 111g "DNE: unlink striped dir, fail MDT1/MDT2"
4057 test_112_rename_prepare() {
4058 mkdir -p $DIR/$tdir/src_dir
4059 $LFS mkdir -i 1 $DIR/$tdir/src_dir/src_child ||
4060 error "create remote source failed"
4062 touch $DIR/$tdir/src_dir/src_child/a
4064 $LFS mkdir -i 2 $DIR/$tdir/tgt_dir ||
4065 error "create remote target dir failed"
4067 $LFS mkdir -i 3 $DIR/$tdir/tgt_dir/tgt_child ||
4068 error "create remote target child failed"
4073 $CHECKSTAT -t dir $DIR/$tdir/src_dir/src_child &&
4074 error "src_child still exists after rename"
4076 $CHECKSTAT -t file $DIR/$tdir/tgt_dir/tgt_child/a ||
4077 error "missing file(a) after rename"
4081 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4082 ([ $FAILURE_MODE == "HARD" ] &&
4083 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4084 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4087 test_112_rename_prepare
4090 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
4091 error "rename dir cross MDT failed!"
4095 rm -rf $DIR/$tdir || error "rmdir failed"
4097 run_test 112a "DNE: cross MDT rename, fail MDT1"
4100 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4101 ([ $FAILURE_MODE == "HARD" ] &&
4102 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4103 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4106 test_112_rename_prepare
4109 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
4110 error "rename dir cross MDT failed!"
4115 rm -rf $DIR/$tdir || error "rmdir failed"
4117 run_test 112b "DNE: cross MDT rename, fail MDT2"
4120 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4121 ([ $FAILURE_MODE == "HARD" ] &&
4122 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4123 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4126 test_112_rename_prepare
4129 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
4130 error "rename dir cross MDT failed!"
4135 rm -rf $DIR/$tdir || error "rmdir failed"
4137 run_test 112c "DNE: cross MDT rename, fail MDT3"
4140 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4141 ([ $FAILURE_MODE == "HARD" ] &&
4142 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4143 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4146 test_112_rename_prepare
4149 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
4150 error "rename dir cross MDT failed!"
4155 rm -rf $DIR/$tdir || error "rmdir failed"
4157 run_test 112d "DNE: cross MDT rename, fail MDT4"
4160 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4161 ([ $FAILURE_MODE == "HARD" ] &&
4162 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4163 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4166 test_112_rename_prepare
4170 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
4171 error "rename dir cross MDT failed!"
4176 rm -rf $DIR/$tdir || error "rmdir failed"
4178 run_test 112e "DNE: cross MDT rename, fail MDT1 and MDT2"
4181 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4182 ([ $FAILURE_MODE == "HARD" ] &&
4183 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4184 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4187 test_112_rename_prepare
4191 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
4192 error "rename dir cross MDT failed!"
4197 rm -rf $DIR/$tdir || error "rmdir failed"
4199 run_test 112f "DNE: cross MDT rename, fail MDT1 and MDT3"
4202 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4203 ([ $FAILURE_MODE == "HARD" ] &&
4204 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4205 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4208 test_112_rename_prepare
4212 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
4213 error "rename dir cross MDT failed!"
4218 rm -rf $DIR/$tdir || error "rmdir failed"
4220 run_test 112g "DNE: cross MDT rename, fail MDT1 and MDT4"
4223 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4224 ([ $FAILURE_MODE == "HARD" ] &&
4225 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4226 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4229 test_112_rename_prepare
4233 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
4234 error "rename dir cross MDT failed!"
4239 rm -rf $DIR/$tdir || error "rmdir failed"
4241 run_test 112h "DNE: cross MDT rename, fail MDT2 and MDT3"
4244 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4245 ([ $FAILURE_MODE == "HARD" ] &&
4246 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4247 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4250 test_112_rename_prepare
4254 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
4255 error "rename dir cross MDT failed!"
4260 rm -rf $DIR/$tdir || error "rmdir failed"
4262 run_test 112i "DNE: cross MDT rename, fail MDT2 and MDT4"
4265 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4266 ([ $FAILURE_MODE == "HARD" ] &&
4267 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4268 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4271 test_112_rename_prepare
4275 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
4276 error "rename dir cross MDT failed!"
4281 rm -rf $DIR/$tdir || error "rmdir failed"
4283 run_test 112j "DNE: cross MDT rename, fail MDT3 and MDT4"
4286 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4287 ([ $FAILURE_MODE == "HARD" ] &&
4288 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4289 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4292 test_112_rename_prepare
4297 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
4298 error "rename dir cross MDT failed!"
4303 rm -rf $DIR/$tdir || error "rmdir failed"
4305 run_test 112k "DNE: cross MDT rename, fail MDT1,MDT2,MDT3"
4308 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4309 ([ $FAILURE_MODE == "HARD" ] &&
4310 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4311 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4314 test_112_rename_prepare
4319 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
4320 error "rename dir cross MDT failed!"
4325 rm -rf $DIR/$tdir || error "rmdir failed"
4327 run_test 112l "DNE: cross MDT rename, fail MDT1,MDT2,MDT4"
4330 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4331 ([ $FAILURE_MODE == "HARD" ] &&
4332 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4333 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4336 test_112_rename_prepare
4341 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
4342 error "rename dir cross MDT failed!"
4347 rm -rf $DIR/$tdir || error "rmdir failed"
4349 run_test 112m "DNE: cross MDT rename, fail MDT1,MDT3,MDT4"
4352 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4353 ([ $FAILURE_MODE == "HARD" ] &&
4354 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4355 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4358 test_112_rename_prepare
4363 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
4364 error "rename dir cross MDT failed!"
4369 rm -rf $DIR/$tdir || error "rmdir failed"
4371 run_test 112n "DNE: cross MDT rename, fail MDT2,MDT3,MDT4"
4374 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
4375 ([ $FAILURE_MODE == "HARD" ] &&
4376 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4377 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4385 for ((j=0;j<$((MDSCOUNT));j++)); do
4386 fail_index=$((fail_index+1))
4387 index=$((fail_index % MDSCOUNT))
4388 replay_barrier mds$((index + 1))
4389 for ((i=0;i<5;i++)); do
4390 test_mkdir -i$index -c$MDSCOUNT $DIR/$tdir/test_$i ||
4391 error "create striped dir $DIR/$tdir/test_$i"
4394 fail mds$((index + 1))
4395 for ((i=0;i<5;i++)); do
4396 checkstat -t dir $DIR/$tdir/test_$i ||
4397 error "$DIR/$tdir/test_$i does not exist!"
4399 rm -rf $DIR/$tdir/test_* ||
4403 run_test 115 "failover for create/unlink striped directory"
4406 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
4407 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] &&
4408 skip "Do not support large update log before 2.7.55" &&
4410 ([ $FAILURE_MODE == "HARD" ] &&
4411 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4412 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4419 # OBD_FAIL_SPLIT_UPDATE_REC 0x1702
4420 do_facet mds1 "lctl set_param fail_loc=0x80001702"
4421 $LFS setdirstripe -c$MDSCOUNT $DIR/$tdir/striped_dir
4424 $CHECKSTAT -t dir $DIR/$tdir/striped_dir ||
4425 error "stried_dir does not exists"
4427 run_test 116a "large update log master MDT recovery"
4430 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
4431 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] &&
4432 skip "Do not support large update log before 2.7.55" &&
4435 ([ $FAILURE_MODE == "HARD" ] &&
4436 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4437 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4444 # OBD_FAIL_SPLIT_UPDATE_REC 0x1702
4445 do_facet mds2 "lctl set_param fail_loc=0x80001702"
4446 $LFS setdirstripe -c$MDSCOUNT $DIR/$tdir/striped_dir
4449 $CHECKSTAT -t dir $DIR/$tdir/striped_dir ||
4450 error "stried_dir does not exists"
4452 run_test 116b "large update log slave MDT recovery"
4455 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4456 ([ $FAILURE_MODE == "HARD" ] &&
4457 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4458 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4464 $LFS setdirstripe -i0 -c$MDSCOUNT $DIR/$tdir/remote_dir
4465 $LFS setdirstripe -i1 -c$MDSCOUNT $DIR/$tdir/remote_dir_1
4468 # Let's set rdonly on all MDTs, so client will send
4469 # replay requests on all MDTs and replay these requests
4470 # at the same time. This test will verify the recovery
4471 # will not be deadlock in this case, LU-7531.
4472 for ((index = 0; index < $((MDSCOUNT)); index++)); do
4473 replay_barrier mds$((index + 1))
4474 if [ -z $mds_indexs ]; then
4475 mds_indexs="${mds_indexs}mds$((index+1))"
4477 mds_indexs="${mds_indexs},mds$((index+1))"
4481 rm -rf $DIR/$tdir/remote_dir
4482 rm -rf $DIR/$tdir/remote_dir_1
4486 rm -rf $DIR/$tdir || error "rmdir failed"
4488 run_test 117 "DNE: cross MDT unlink, fail MDT1 and MDT2"
4491 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
4492 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.64) ] &&
4493 skip "Do not support large update log before 2.7.64" &&
4498 $LFS setdirstripe -c2 $DIR/$tdir/striped_dir ||
4499 error "setdirstripe fails"
4500 $LFS setdirstripe -c2 $DIR/$tdir/striped_dir1 ||
4501 error "setdirstripe fails 1"
4502 rm -rf $DIR/$tdir/striped_dir* || error "rmdir fails"
4504 # OBD_FAIL_INVALIDATE_UPDATE 0x1705
4505 do_facet mds1 "lctl set_param fail_loc=0x1705"
4506 $LFS setdirstripe -c2 $DIR/$tdir/striped_dir
4507 $LFS setdirstripe -c2 $DIR/$tdir/striped_dir1
4508 do_facet mds1 "lctl set_param fail_loc=0x0"
4511 $LFS setdirstripe -c2 $DIR/$tdir/striped_dir
4512 $LFS setdirstripe -c2 $DIR/$tdir/striped_dir1
4517 run_test 118 "invalidate osp update will not cause update log corruption"
4520 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
4521 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.64) ] &&
4522 skip "Do not support large update log before 2.7.64" &&
4525 local hard_timeout=$(do_facet mds1 \
4526 "lctl get_param -n mdt.$FSNAME-MDT0000.recovery_time_hard")
4528 local clients=${CLIENTS:-$HOSTNAME}
4529 local time_min=$(recovery_time_min)
4532 mkdir $DIR/$tdir/tmp
4533 rmdir $DIR/$tdir/tmp
4536 mkdir $DIR/$tdir/dir_1
4537 for ((i = 0; i < 20; i++)); do
4538 $LFS setdirstripe -c2 $DIR/$tdir/stripe_dir-$i
4545 #define OBD_FAIL_TGT_REPLAY_DELAY 0x714
4546 do_facet mds1 $LCTL set_param fail_loc=0x80000714
4547 #sleep (timeout + 5), so mds will evict the client exports,
4548 #but DNE update recovery will keep going.
4549 do_facet mds1 $LCTL set_param fail_val=$((time_min + 5))
4551 mount_facet mds1 "-o recovery_time_hard=$time_min"
4553 wait_clients_import_state "$clients" mds1 FULL
4555 clients_up || clients_up || error "failover df: $?"
4557 #revert back the hard timeout
4558 do_facet mds1 $LCTL set_param \
4559 mdt.$FSNAME-MDT0000.recovery_time_hard=$hard_timeout
4561 for ((i = 0; i < 20; i++)); do
4562 stripe_count=$($LFS getdirstripe -c $DIR/$tdir/stripe_dir-$i)
4563 [ $stripe_count == 2 ] || {
4564 error "stripe_dir-$i creation replay fails"
4569 run_test 119 "timeout of normal replay does not cause DNE replay fails "
4572 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
4573 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.64) ] &&
4574 skip "Do not support large update log before 2.7.64" &&
4578 replay_barrier_nosync mds1
4579 for ((i = 0; i < 20; i++)); do
4580 mkdir $DIR/$tdir/dir-$i || {
4581 error "create dir-$i fails"
4584 $LFS setdirstripe -c2 $DIR/$tdir/stripe_dir-$i || {
4585 error "create stripe_dir-$i fails"
4592 for ((i = 0; i < 20; i++)); do
4593 [ ! -e "$DIR/$tdir/dir-$i" ] || {
4594 error "dir-$i still exists"
4597 [ ! -e "$DIR/$tdir/stripe_dir-$i" ] || {
4598 error "stripe_dir-$i still exists"
4603 run_test 120 "DNE fail abort should stop both normal and DNE replay"
4606 check_and_cleanup_lustre