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 [ $OSTCOUNT -lt 2 ] && skip_env "needs >= 2 OSTs" && return
915 local f=$MOUNT/$tfile
916 # make sure the start of the file is ost1
917 $SETSTRIPE -S $((128 * 1024)) -i 0 $f
918 do_facet client dd if=/dev/zero of=$f bs=4k count=1 ||
919 error "dd on client failed"
921 # fail ost2 and read from ost1
922 local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS $ost2_svc)
923 local osc2dev=$(do_facet $SINGLEMDS "lctl get_param -n devices" |
924 grep $mdtosc | awk '{print $1}')
925 [ -z "$osc2dev" ] && echo "OST: $ost2_svc" &&
926 lctl get_param -n devices &&
927 error "OST 2 $osc2dev does not exist"
928 do_facet $SINGLEMDS $LCTL --device $osc2dev deactivate ||
929 error "deactive device on $SINGLEMDS failed"
930 do_facet client dd if=$f of=/dev/null bs=4k count=1 ||
931 error "second dd on client failed"
932 do_facet $SINGLEMDS $LCTL --device $osc2dev activate ||
933 error "active device on $SINGLEMDS failed"
936 run_test 41 "read from a valid osc while other oscs are invalid"
938 # test MDS recovery after ost failure
940 blocks=$(df -P $MOUNT | tail -n 1 | awk '{ print $2 }')
941 createmany -o $DIR/$tfile-%d 800 ||
942 error "createmany -o $DIR/$tfile failed"
944 unlinkmany $DIR/$tfile-%d 0 400
946 lctl set_param debug=-1
949 # osc is evicted, fs is smaller (but only with failout OSTs (bug 7287)
950 #blocks_after=`df -P $MOUNT | tail -n 1 | awk '{ print $2 }'`
951 #[ $blocks_after -lt $blocks ] || return 1
952 echo "wait for MDS to timeout and recover"
953 sleep $((TIMEOUT * 2))
955 unlinkmany $DIR/$tfile-%d 400 400 ||
956 error "unlinkmany $DIR/$tfile 400 failed"
957 $CHECKSTAT -t file $DIR/$tfile-* &&
958 error "$CHECKSTAT $DIR/$tfile-* attribute check should fail" ||
961 run_test 42 "recovery after ost failure"
963 # timeout in MDS/OST recovery RPC will LBUG MDS
964 test_43() { # bug 2530
965 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
967 replay_barrier $SINGLEMDS
969 # OBD_FAIL_OST_CREATE_NET 0x204
970 do_facet ost1 "lctl set_param fail_loc=0x80000204"
976 run_test 43 "mds osc import failure during recovery; don't LBUG"
978 test_44a() { # was test_44
981 local mdcdev=$($LCTL dl |
982 awk "/${FSNAME}-MDT0000-mdc-/ {if (\$2 == \"UP\") {print \$1}}")
983 [ "$mdcdev" ] || error "${FSNAME}-MDT0000-mdc- not UP"
984 [ $(echo $mdcdev | wc -w) -eq 1 ] ||
985 { $LCTL dl; error "looking for mdcdev=$mdcdev"; }
987 # adaptive timeouts slow this way down
988 if at_is_enabled; then
989 at_max_saved=$(at_max_get mds)
993 for i in $(seq 1 10); do
994 echo "$i of 10 ($(date +%s))"
995 do_facet $SINGLEMDS \
996 "lctl get_param -n md[ts].*.mdt.timeouts | grep service"
997 #define OBD_FAIL_TGT_CONN_RACE 0x701
998 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000701"
999 # lctl below may fail, it is valid case
1000 $LCTL --device $mdcdev recover
1003 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1004 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds
1007 run_test 44a "race in target handle connect"
1010 local mdcdev=$($LCTL dl |
1011 awk "/${FSNAME}-MDT0000-mdc-/ {if (\$2 == \"UP\") {print \$1}}")
1012 [ "$mdcdev" ] || error "${FSNAME}-MDT0000-mdc not up"
1013 [ $(echo $mdcdev | wc -w) -eq 1 ] ||
1014 { echo mdcdev=$mdcdev; $LCTL dl;
1015 error "more than one ${FSNAME}-MDT0000-mdc"; }
1017 for i in $(seq 1 10); do
1018 echo "$i of 10 ($(date +%s))"
1019 do_facet $SINGLEMDS \
1020 "lctl get_param -n md[ts].*.mdt.timeouts | grep service"
1021 #define OBD_FAIL_TGT_DELAY_RECONNECT 0x704
1022 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000704"
1023 # lctl below may fail, it is valid case
1024 $LCTL --device $mdcdev recover
1029 run_test 44b "race in target handle connect"
1032 replay_barrier $SINGLEMDS
1033 createmany -m $DIR/$tfile-%d 100 || error "failed to create directories"
1034 #define OBD_FAIL_TGT_RCVG_FLAG 0x712
1035 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000712"
1036 fail_abort $SINGLEMDS
1037 unlinkmany $DIR/$tfile-%d 100 && error "unliked after fail abort"
1039 unlinkmany $DIR/$tfile-%d 100 && error "unliked after fail"
1042 run_test 44c "race in target handle connect"
1044 # Handle failed close
1046 local mdcdev=$($LCTL get_param -n devices |
1047 awk "/ ${FSNAME}-MDT0000-mdc-/ {print \$1}")
1048 [ "$mdcdev" ] || error "${FSNAME}-MDT0000-mdc not up"
1049 [ $(echo $mdcdev | wc -w) -eq 1 ] ||
1050 { echo mdcdev=$mdcdev; $LCTL dl;
1051 error "more than one ${FSNAME}-MDT0000-mdc"; }
1053 $LCTL --device $mdcdev recover ||
1054 error "$LCTL --device $mdcdev recover failed"
1056 multiop_bg_pause $DIR/$tfile O_c ||
1057 error "multiop_bg_pause $DIR/$tfile failed"
1060 # This will cause the CLOSE to fail before even
1061 # allocating a reply buffer
1062 $LCTL --device $mdcdev deactivate ||
1063 error "$LCTL --device $mdcdev deactivate failed"
1066 kill -USR1 $pid || error "multiop $pid not running"
1067 wait $pid || error "multiop $pid failed"
1069 $LCTL --device $mdcdev activate ||
1070 error "$LCTL --device $mdcdev activate failed"
1073 $CHECKSTAT -t file $DIR/$tfile ||
1074 error "$CHECKSTAT $DIR/$tfile attribute check failed"
1077 run_test 45 "Handle failed close"
1081 drop_reply "touch $DIR/$tfile"
1083 # ironically, the previous test, 45, will cause a real forced close,
1084 # so just look for one for this test
1085 dmesg | grep -i "force closing client file handle for $tfile" &&
1086 error "found force closing in dmesg"
1089 run_test 46 "Don't leak file handle after open resend (3325)"
1091 test_47() { # bug 2824
1092 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1094 # create some files to make sure precreate has been done on all
1095 # OSTs. (just in case this test is run independently)
1096 createmany -o $DIR/$tfile 20 ||
1097 error "createmany create $DIR/$tfile failed"
1099 # OBD_FAIL_OST_CREATE_NET 0x204
1101 do_facet ost1 "lctl set_param fail_loc=0x80000204"
1102 client_up || error "client_up failed"
1104 # let the MDS discover the OST failure, attempt to recover, fail
1105 # and recover again.
1106 sleep $((3 * TIMEOUT))
1108 # Without 2824, this createmany would hang
1109 createmany -o $DIR/$tfile 20 ||
1110 error "createmany recraete $DIR/$tfile failed"
1111 unlinkmany $DIR/$tfile 20 || error "unlinkmany $DIR/$tfile failed"
1115 run_test 47 "MDS->OSC failure during precreate cleanup (2824)"
1118 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1119 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >= 2 OSTs" && return
1121 replay_barrier $SINGLEMDS
1122 createmany -o $DIR/$tfile 20 ||
1123 error "createmany -o $DIR/$tfile failed"
1124 # OBD_FAIL_OST_EROFS 0x216
1125 facet_failover $SINGLEMDS
1126 do_facet ost1 "lctl set_param fail_loc=0x80000216"
1127 client_up || error "client_up failed"
1129 createmany -o $DIR/$tfile 20 20 ||
1130 error "createmany recraete $DIR/$tfile failed"
1131 unlinkmany $DIR/$tfile 40 || error "unlinkmany $DIR/$tfile failed"
1134 run_test 48 "MDS->OSC failure during precreate cleanup (2824)"
1137 local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS $ost1_svc)
1138 local oscdev=$(do_facet $SINGLEMDS "lctl get_param -n devices" |
1139 grep $mdtosc | awk '{print $1}')
1140 [ "$oscdev" ] || error "could not find OSC device on MDS"
1141 do_facet $SINGLEMDS $LCTL --device $oscdev recover ||
1142 error "OSC device $oscdev recovery failed"
1143 do_facet $SINGLEMDS $LCTL --device $oscdev recover ||
1144 error "second OSC device $oscdev recovery failed"
1145 # give the mds_lov_sync threads a chance to run
1148 run_test 50 "Double OSC recovery, don't LASSERT (3812)"
1150 # b3764 timed out lock replay
1152 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.6.90) ] &&
1153 skip "MDS prior to 2.6.90 handle LDLM_REPLY_NET incorrectly" &&
1156 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
1157 cancel_lru_locks mdc
1159 multiop_bg_pause $DIR/$tfile s_s || error "multiop $DIR/$tfile failed"
1162 #define OBD_FAIL_MDS_LDLM_REPLY_NET 0x157
1163 lctl set_param -n ldlm.cancel_unused_locks_before_replay "0"
1164 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000157"
1166 fail $SINGLEMDS || error "fail $SINGLEMDS failed"
1168 wait $mpid || error "multiop_bg_pause pid failed"
1170 do_facet $SINGLEMDS "lctl set_param fail_loc=0x0"
1171 lctl set_param fail_loc=0x0
1172 lctl set_param -n ldlm.cancel_unused_locks_before_replay "1"
1175 run_test 52 "time out lock replay (3764)"
1177 # bug 3462 - simultaneous MDC requests
1179 [[ $(lctl get_param mdc.*.import |
1180 grep "connect_flags:.*multi_mod_rpc") ]] ||
1181 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
1183 cancel_lru_locks mdc # cleanup locks from former test cases
1184 mkdir $DIR/${tdir}-1 || error "mkdir $DIR/${tdir}-1 failed"
1185 mkdir $DIR/${tdir}-2 || error "mkdir $DIR/${tdir}-2 failed"
1186 multiop $DIR/${tdir}-1/f O_c &
1188 # give multiop a change to open
1191 #define OBD_FAIL_MDS_CLOSE_NET 0x115
1192 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
1193 kill -USR1 $close_pid
1194 cancel_lru_locks mdc # force the close
1195 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1197 mcreate $DIR/${tdir}-2/f || error "mcreate $DIR/${tdir}-2/f failed"
1199 # close should still be here
1200 [ -d /proc/$close_pid ] || error "close_pid doesn't exist"
1202 replay_barrier_nodf $SINGLEMDS
1204 wait $close_pid || error "close_pid $close_pid failed"
1206 $CHECKSTAT -t file $DIR/${tdir}-1/f ||
1207 error "$CHECKSTAT $DIR/${tdir}-1/f attribute check failed"
1208 $CHECKSTAT -t file $DIR/${tdir}-2/f ||
1209 error "$CHECKSTAT $DIR/${tdir}-2/f attribute check failed"
1210 rm -rf $DIR/${tdir}-*
1212 run_test 53a "|X| close request while two MDC requests in flight"
1215 cancel_lru_locks mdc # cleanup locks from former test cases
1217 mkdir $DIR/${tdir}-1 || error "mkdir $DIR/${tdir}-1 failed"
1218 mkdir $DIR/${tdir}-2 || error "mkdir $DIR/${tdir}-2 failed"
1219 multiop_bg_pause $DIR/${tdir}-1/f O_c ||
1220 error "multiop_bg_pause $DIR/${tdir}-1/f failed"
1223 #define OBD_FAIL_MDS_REINT_NET 0x107
1224 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000107"
1225 mcreate $DIR/${tdir}-2/f &
1229 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1230 kill -USR1 $close_pid
1231 cancel_lru_locks mdc # force the close
1232 wait $close_pid || error "close_pid $close_pid failed"
1233 # open should still be here
1234 [ -d /proc/$open_pid ] || error "open_pid doesn't exist"
1236 replay_barrier_nodf $SINGLEMDS
1238 wait $open_pid || error "open_pid failed"
1240 $CHECKSTAT -t file $DIR/${tdir}-1/f ||
1241 error "$CHECKSTAT $DIR/${tdir}-1/f attribute check failed"
1242 $CHECKSTAT -t file $DIR/${tdir}-2/f ||
1243 error "$CHECKSTAT $DIR/${tdir}-2/f attribute check failed"
1244 rm -rf $DIR/${tdir}-*
1246 run_test 53b "|X| open request while two MDC requests in flight"
1249 cancel_lru_locks mdc # cleanup locks from former test cases
1251 mkdir $DIR/${tdir}-1 || error "mkdir $DIR/${tdir}-1 failed"
1252 mkdir $DIR/${tdir}-2 || error "mkdir $DIR/${tdir}-2 failed"
1253 multiop $DIR/${tdir}-1/f O_c &
1256 #define OBD_FAIL_MDS_REINT_NET 0x107
1257 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000107"
1258 mcreate $DIR/${tdir}-2/f &
1262 #define OBD_FAIL_MDS_CLOSE_NET 0x115
1263 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
1264 kill -USR1 $close_pid
1265 cancel_lru_locks mdc # force the close
1267 #bz20647: make sure all pids exist before failover
1268 [ -d /proc/$close_pid ] || error "close_pid doesn't exist"
1269 [ -d /proc/$open_pid ] || error "open_pid doesn't exists"
1270 replay_barrier_nodf $SINGLEMDS
1271 fail_nodf $SINGLEMDS
1272 wait $open_pid || error "open_pid failed"
1274 # close should be gone
1275 [ -d /proc/$close_pid ] && error "close_pid should not exist"
1276 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1278 $CHECKSTAT -t file $DIR/${tdir}-1/f ||
1279 error "$CHECKSTAT $DIR/${tdir}-1/f attribute check failed"
1280 $CHECKSTAT -t file $DIR/${tdir}-2/f ||
1281 error "$CHECKSTAT $DIR/${tdir}-2/f attribute check failed"
1282 rm -rf $DIR/${tdir}-*
1284 run_test 53c "|X| open request and close request while two MDC requests in flight"
1287 [[ $(lctl get_param mdc.*.import |
1288 grep "connect_flags:.*multi_mod_rpc") ]] ||
1289 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
1291 cancel_lru_locks mdc # cleanup locks from former test cases
1293 mkdir $DIR/${tdir}-1 || error "mkdir $DIR/${tdir}-1 failed"
1294 mkdir $DIR/${tdir}-2 || error "mkdir $DIR/${tdir}-2 failed"
1295 multiop $DIR/${tdir}-1/f O_c &
1297 # give multiop a chance to open
1300 #define OBD_FAIL_MDS_CLOSE_NET_REP 0x13b
1301 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000013b"
1302 kill -USR1 $close_pid
1303 cancel_lru_locks mdc # force the close
1304 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1305 mcreate $DIR/${tdir}-2/f || error "mcreate $DIR/${tdir}-2/f failed"
1307 # close should still be here
1308 [ -d /proc/$close_pid ] || error "close_pid doesn't exist"
1310 wait $close_pid || error "close_pid failed"
1312 $CHECKSTAT -t file $DIR/${tdir}-1/f ||
1313 error "$CHECKSTAT $DIR/${tdir}-1/f attribute check failed"
1314 $CHECKSTAT -t file $DIR/${tdir}-2/f ||
1315 error "$CHECKSTAT $DIR/${tdir}-2/f attribute check failed"
1316 rm -rf $DIR/${tdir}-*
1318 run_test 53d "close reply while two MDC requests in flight"
1321 cancel_lru_locks mdc # cleanup locks from former test cases
1323 mkdir $DIR/${tdir}-1 || error "mkdir $DIR/${tdir}-1 failed"
1324 mkdir $DIR/${tdir}-2 || error "mkdir $DIR/${tdir}-2 failed"
1325 multiop $DIR/${tdir}-1/f O_c &
1328 #define OBD_FAIL_MDS_REINT_NET_REP 0x119
1329 do_facet $SINGLEMDS "lctl set_param fail_loc=0x119"
1330 mcreate $DIR/${tdir}-2/f &
1334 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1335 kill -USR1 $close_pid
1336 cancel_lru_locks mdc # force the close
1337 wait $close_pid || error "close_pid failed"
1338 # open should still be here
1339 [ -d /proc/$open_pid ] || error "open_pid doesn't exists"
1341 replay_barrier_nodf $SINGLEMDS
1343 wait $open_pid || error "open_pid failed"
1345 $CHECKSTAT -t file $DIR/${tdir}-1/f ||
1346 error "$CHECKSTAT $DIR/${tdir}-1/f attribute check failed"
1347 $CHECKSTAT -t file $DIR/${tdir}-2/f ||
1348 error "$CHECKSTAT $DIR/${tdir}-2/f attribute check failed"
1349 rm -rf $DIR/${tdir}-*
1351 run_test 53e "|X| open reply while two MDC requests in flight"
1354 cancel_lru_locks mdc # cleanup locks from former test cases
1356 mkdir $DIR/${tdir}-1 || error "mkdir $DIR/${tdir}-1 failed"
1357 mkdir $DIR/${tdir}-2 || error "mkdir $DIR/${tdir}-2 failed"
1358 multiop $DIR/${tdir}-1/f O_c &
1361 #define OBD_FAIL_MDS_REINT_NET_REP 0x119
1362 do_facet $SINGLEMDS "lctl set_param fail_loc=0x119"
1363 mcreate $DIR/${tdir}-2/f &
1367 #define OBD_FAIL_MDS_CLOSE_NET_REP 0x13b
1368 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000013b"
1369 kill -USR1 $close_pid
1370 cancel_lru_locks mdc # force the close
1372 #bz20647: make sure all pids are exists before failover
1373 [ -d /proc/$close_pid ] || error "close_pid doesn't exist"
1374 [ -d /proc/$open_pid ] || error "open_pid doesn't exists"
1375 replay_barrier_nodf $SINGLEMDS
1376 fail_nodf $SINGLEMDS
1377 wait $open_pid || error "open_pid failed"
1379 # close should be gone
1380 [ -d /proc/$close_pid ] && error "close_pid should not exist"
1381 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1383 $CHECKSTAT -t file $DIR/${tdir}-1/f ||
1384 error "$CHECKSTAT $DIR/${tdir}-1/f attribute check failed"
1385 $CHECKSTAT -t file $DIR/${tdir}-2/f ||
1386 error "$CHECKSTAT $DIR/${tdir}-2/f attribute check failed"
1387 rm -rf $DIR/${tdir}-*
1389 run_test 53f "|X| open reply and close reply while two MDC requests in flight"
1392 cancel_lru_locks mdc # cleanup locks from former test cases
1394 mkdir $DIR/${tdir}-1 || error "mkdir $DIR/${tdir}-1 failed"
1395 mkdir $DIR/${tdir}-2 || error "mkdir $DIR/${tdir}-2 failed"
1396 multiop $DIR/${tdir}-1/f O_c &
1399 #define OBD_FAIL_MDS_REINT_NET_REP 0x119
1400 do_facet $SINGLEMDS "lctl set_param fail_loc=0x119"
1401 mcreate $DIR/${tdir}-2/f &
1405 #define OBD_FAIL_MDS_CLOSE_NET 0x115
1406 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
1407 kill -USR1 $close_pid
1408 cancel_lru_locks mdc # force the close
1409 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1411 #bz20647: make sure all pids are exists before failover
1412 [ -d /proc/$close_pid ] || error "close_pid doesn't exist"
1413 [ -d /proc/$open_pid ] || error "open_pid doesn't exists"
1414 replay_barrier_nodf $SINGLEMDS
1415 fail_nodf $SINGLEMDS
1416 wait $open_pid || error "open_pid failed"
1418 # close should be gone
1419 [ -d /proc/$close_pid ] && error "close_pid should not exist"
1421 $CHECKSTAT -t file $DIR/${tdir}-1/f ||
1422 error "$CHECKSTAT $DIR/${tdir}-1/f attribute check failed"
1423 $CHECKSTAT -t file $DIR/${tdir}-2/f ||
1424 error "$CHECKSTAT $DIR/${tdir}-2/f attribute check failed"
1425 rm -rf $DIR/${tdir}-*
1427 run_test 53g "|X| drop open reply and close request while close and open are both in flight"
1430 cancel_lru_locks mdc # cleanup locks from former test cases
1432 mkdir $DIR/${tdir}-1 || error "mkdir $DIR/${tdir}-1 failed"
1433 mkdir $DIR/${tdir}-2 || error "mkdir $DIR/${tdir}-2 failed"
1434 multiop $DIR/${tdir}-1/f O_c &
1437 #define OBD_FAIL_MDS_REINT_NET 0x107
1438 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000107"
1439 mcreate $DIR/${tdir}-2/f &
1443 #define OBD_FAIL_MDS_CLOSE_NET_REP 0x13b
1444 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000013b"
1445 kill -USR1 $close_pid
1446 cancel_lru_locks mdc # force the close
1449 #bz20647: make sure all pids are exists before failover
1450 [ -d /proc/$close_pid ] || error "close_pid doesn't exist"
1451 [ -d /proc/$open_pid ] || error "open_pid doesn't exists"
1452 replay_barrier_nodf $SINGLEMDS
1453 fail_nodf $SINGLEMDS
1454 wait $open_pid || error "open_pid failed"
1456 # close should be gone
1457 [ -d /proc/$close_pid ] && error "close_pid should not exist"
1458 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1460 $CHECKSTAT -t file $DIR/${tdir}-1/f ||
1461 error "$CHECKSTAT $DIR/${tdir}-1/f attribute check failed"
1462 $CHECKSTAT -t file $DIR/${tdir}-2/f ||
1463 error "$CHECKSTAT $DIR/${tdir}-2/f attribute check failed"
1464 rm -rf $DIR/${tdir}-*
1466 run_test 53h "open request and close reply while two MDC requests in flight"
1468 #b3761 ASSERTION(hash != 0) failed
1470 # OBD_FAIL_MDS_OPEN_CREATE | OBD_FAIL_ONCE
1471 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000012b"
1473 # give touch a chance to run
1475 do_facet $SINGLEMDS "lctl set_param fail_loc=0x0"
1479 run_test 55 "let MDS_CHECK_RESENT return the original return code instead of 0"
1481 #b3440 ASSERTION(rec->ur_fid2->id) failed
1483 ln -s foo $DIR/$tfile
1484 replay_barrier $SINGLEMDS
1485 #drop_reply "cat $DIR/$tfile"
1489 run_test 56 "don't replay a symlink open request (3440)"
1491 #recovery one mds-ost setattr from llog
1493 #define OBD_FAIL_MDS_OST_SETATTR 0x12c
1494 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000012c"
1495 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
1496 replay_barrier $SINGLEMDS
1498 wait_recovery_complete $SINGLEMDS || error "MDS recovery is not done"
1499 wait_mds_ost_sync || error "wait_mds_ost_sync failed"
1500 $CHECKSTAT -t file $DIR/$tfile ||
1501 error "$CHECKSTAT $DIR/$tfile attribute check failed"
1502 do_facet $SINGLEMDS "lctl set_param fail_loc=0x0"
1505 run_test 57 "test recovery from llog for setattr op"
1508 zconf_umount $(hostname) $MOUNT2
1512 #recovery many mds-ost setattr from llog
1514 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
1515 #define OBD_FAIL_MDS_OST_SETATTR 0x12c
1516 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000012c"
1517 createmany -o $DIR/$tdir/$tfile-%d 2500
1518 replay_barrier $SINGLEMDS
1521 $CHECKSTAT -t file $DIR/$tdir/$tfile-* >/dev/null ||
1522 error "$CHECKSTAT $DIR/$tfile-* attribute check failed"
1523 do_facet $SINGLEMDS "lctl set_param fail_loc=0x0"
1524 unlinkmany $DIR/$tdir/$tfile-%d 2500 ||
1525 error "unlinkmany $DIR/$tfile failed"
1528 run_test 58a "test recovery from llog for setattr op (test llog_gen_rec)"
1534 trap cleanup_58 EXIT
1536 large_xattr_enabled &&
1537 orig="$(generate_string $(max_xattr_size))" || orig="bar"
1538 # Original extended attribute can be long. Print a small version of
1539 # attribute if an error occurs
1540 local sm_msg=$(printf "%.9s" $orig)
1542 mount_client $MOUNT2 || error "mount_client on $MOUNT2 failed"
1543 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
1544 touch $DIR/$tdir/$tfile || error "touch $DIR/$tdir/$tfile failed"
1545 replay_barrier $SINGLEMDS
1546 setfattr -n trusted.foo -v $orig $DIR/$tdir/$tfile
1548 new=$(get_xattr_value trusted.foo $MOUNT2/$tdir/$tfile)
1549 [[ "$new" = "$orig" ]] ||
1550 error "xattr set ($sm_msg...) differs from xattr get ($new)"
1551 rm -f $DIR/$tdir/$tfile
1554 wait_clients_import_state ${CLIENTS:-$HOSTNAME} "mgs" FULL
1556 run_test 58b "test replay of setxattr op"
1558 test_58c() { # bug 16570
1563 trap cleanup_58 EXIT
1565 if large_xattr_enabled; then
1566 local xattr_size=$(max_xattr_size)
1567 orig="$(generate_string $((xattr_size / 2)))"
1568 orig1="$(generate_string $xattr_size)"
1574 # PING_INTERVAL max(obd_timeout / 4, 1U)
1575 sleep $((TIMEOUT / 4))
1577 # Original extended attribute can be long. Print a small version of
1578 # attribute if an error occurs
1579 local sm_msg=$(printf "%.9s" $orig)
1580 local sm_msg1=$(printf "%.9s" $orig1)
1582 mount_client $MOUNT2 || error "mount_client on $MOUNT2 failed"
1583 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
1584 touch $DIR/$tdir/$tfile || error "touch $DIR/$tdir/$tfile failed"
1585 drop_request "setfattr -n trusted.foo -v $orig $DIR/$tdir/$tfile" ||
1586 error "drop_request for setfattr failed"
1587 new=$(get_xattr_value trusted.foo $MOUNT2/$tdir/$tfile)
1588 [[ "$new" = "$orig" ]] ||
1589 error "xattr set ($sm_msg...) differs from xattr get ($new)"
1590 drop_reint_reply "setfattr -n trusted.foo1 \
1591 -v $orig1 $DIR/$tdir/$tfile" ||
1592 error "drop_reint_reply for setfattr failed"
1593 new=$(get_xattr_value trusted.foo1 $MOUNT2/$tdir/$tfile)
1594 [[ "$new" = "$orig1" ]] ||
1595 error "second xattr set ($sm_msg1...) differs xattr get ($new)"
1596 rm -f $DIR/$tdir/$tfile
1600 run_test 58c "resend/reconstruct setxattr op"
1602 # log_commit_thread vs filter_destroy race used to lead to import use after free
1605 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1607 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
1608 createmany -o $DIR/$tdir/$tfile-%d 200 ||
1609 error "createmany create files failed"
1611 unlinkmany $DIR/$tdir/$tfile-%d 200 ||
1612 error "unlinkmany $DIR/$tdir/$tfile failed"
1613 #define OBD_FAIL_PTLRPC_DELAY_RECOV 0x507
1614 do_facet ost1 "lctl set_param fail_loc=0x507"
1617 do_facet ost1 "lctl set_param fail_loc=0x0"
1621 run_test 59 "test log_commit_thread vs filter_destroy race"
1623 # race between add unlink llog vs cat log init in post_recovery (only for b1_6)
1624 # bug 12086: should no oops and No ctxt error for this test
1626 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
1627 createmany -o $DIR/$tdir/$tfile-%d 200 ||
1628 error "createmany create files failed"
1629 replay_barrier $SINGLEMDS
1630 unlinkmany $DIR/$tdir/$tfile-%d 0 100
1632 unlinkmany $DIR/$tdir/$tfile-%d 100 100
1633 local no_ctxt=$(dmesg | grep "No ctxt")
1634 [ -z "$no_ctxt" ] || error "ctxt is not initialized in recovery"
1636 run_test 60 "test llog post recovery init vs llog unlink"
1638 #test race llog recovery thread vs llog cleanup
1639 test_61a() { # was test_61
1640 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1642 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
1643 createmany -o $DIR/$tdir/$tfile-%d 800 ||
1644 error "createmany create files failed"
1646 unlinkmany $DIR/$tdir/$tfile-%d 800
1647 # OBD_FAIL_OST_LLOG_RECOVERY_TIMEOUT 0x221
1648 set_nodes_failloc "$(osts_nodes)" 0x80000221
1653 set_nodes_failloc "$(osts_nodes)" 0x0
1655 $CHECKSTAT -t file $DIR/$tdir/$tfile-* &&
1656 error "$CHECKSTAT $DIR/$tdir/$tfile attribute check should fail"
1659 run_test 61a "test race llog recovery vs llog cleanup"
1661 #test race mds llog sync vs llog cleanup
1663 # OBD_FAIL_MDS_LLOG_SYNC_TIMEOUT 0x13a
1664 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000013a"
1665 facet_failover $SINGLEMDS
1668 do_facet client dd if=/dev/zero of=$DIR/$tfile bs=4k count=1 ||
1671 run_test 61b "test race mds llog sync vs llog cleanup"
1673 #test race cancel cookie cb vs llog cleanup
1675 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1677 # OBD_FAIL_OST_CANCEL_COOKIE_TIMEOUT 0x222
1678 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
1679 set_nodes_failloc "$(osts_nodes)" 0x80000222
1683 set_nodes_failloc "$(osts_nodes)" 0x0
1685 run_test 61c "test race mds llog sync vs llog cleanup"
1687 test_61d() { # bug 16002 # bug 17466 # bug 22137
1688 # OBD_FAIL_OBD_LLOG_SETUP 0x605
1690 do_facet mgs "lctl set_param fail_loc=0x80000605"
1691 start mgs $(mgsdevname) $MGS_MOUNT_OPTS &&
1692 error "mgs start should have failed"
1693 do_facet mgs "lctl set_param fail_loc=0"
1694 start mgs $(mgsdevname) $MGS_MOUNT_OPTS || error "cannot restart mgs"
1696 run_test 61d "error in llog_setup should cleanup the llog context correctly"
1698 test_62() { # Bug 15756 - don't mis-drop resent replay
1699 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
1700 replay_barrier $SINGLEMDS
1701 createmany -o $DIR/$tdir/$tfile- 25 ||
1702 error "createmany create files failed"
1703 #define OBD_FAIL_TGT_REPLAY_DROP 0x707
1704 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000707"
1706 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1707 unlinkmany $DIR/$tdir/$tfile- 25 ||
1708 error "unlinkmany $DIR/$tdir/$tfile failed"
1711 run_test 62 "don't mis-drop resent replay"
1713 #Adaptive Timeouts (bug 3055)
1721 echo "Cleaning up AT ..."
1722 if [ -n "$ATOLDBASE" ]; then
1723 local at_history=$($LCTL get_param -n at_history)
1724 do_facet $SINGLEMDS "lctl set_param at_history=$at_history" || true
1725 do_facet ost1 "lctl set_param at_history=$at_history" || true
1728 if [ $AT_MAX_SET -ne 0 ]; then
1729 for facet in mds client ost; do
1730 var=AT_MAX_SAVE_${facet}
1731 echo restore AT on $facet to saved value ${!var}
1732 at_max_set ${!var} $facet
1733 at_new=$(at_max_get $facet)
1734 echo Restored AT value on $facet $at_new
1735 [ $at_new -eq ${!var} ] ||
1736 error "AT value not restored SAVED ${!var} NEW $at_new"
1743 local at_max_new=600
1745 # Save at_max original values
1747 if [ $AT_MAX_SET -eq 0 ]; then
1748 # Suppose that all osts have the same at_max
1749 for facet in mds client ost; do
1750 eval AT_MAX_SAVE_${facet}=$(at_max_get $facet)
1754 for facet in mds client ost; do
1755 at_max=$(at_max_get $facet)
1756 if [ $at_max -ne $at_max_new ]; then
1757 echo "AT value on $facet is $at_max, set it by force temporarily to $at_max_new"
1758 at_max_set $at_max_new $facet
1763 if [ -z "$ATOLDBASE" ]; then
1764 ATOLDBASE=$(do_facet $SINGLEMDS "lctl get_param -n at_history")
1765 # speed up the timebase so we can check decreasing AT
1766 do_facet $SINGLEMDS "lctl set_param at_history=8" || true
1767 do_facet ost1 "lctl set_param at_history=8" || true
1769 # sleep for a while to cool down, should be > 8s and also allow
1770 # at least one ping to be sent. simply use TIMEOUT to be safe.
1775 test_65a() #bug 3055
1777 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1779 at_start || return 0
1780 $LCTL dk > /dev/null
1782 $LCTL set_param debug="other"
1783 # Slow down a request to the current service time, this is critical
1784 # because previous tests may have caused this value to increase.
1785 REQ_DELAY=`lctl get_param -n mdc.${FSNAME}-MDT0000-mdc-*.timeouts |
1786 awk '/portal 12/ {print $5}'`
1787 REQ_DELAY=$((${REQ_DELAY} + ${REQ_DELAY} / 4 + 5))
1789 do_facet $SINGLEMDS lctl set_param fail_val=$((${REQ_DELAY} * 1000))
1790 #define OBD_FAIL_PTLRPC_PAUSE_REQ 0x50a
1791 do_facet $SINGLEMDS $LCTL set_param fail_loc=0x8000050a
1792 createmany -o $DIR/$tfile 10 > /dev/null
1793 unlinkmany $DIR/$tfile 10 > /dev/null
1794 # check for log message
1795 $LCTL dk | grep "Early reply #" || error "No early reply"
1797 # client should show REQ_DELAY estimates
1798 lctl get_param -n mdc.${FSNAME}-MDT0000-mdc-*.timeouts | grep portal
1800 lctl get_param -n mdc.${FSNAME}-MDT0000-mdc-*.timeouts | grep portal
1802 run_test 65a "AT: verify early replies"
1804 test_65b() #bug 3055
1806 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1808 at_start || return 0
1811 $LCTL set_param debug="other trace"
1812 $LCTL dk > /dev/null
1813 # Slow down a request to the current service time, this is critical
1814 # because previous tests may have caused this value to increase.
1815 $SETSTRIPE --stripe-index=0 --stripe-count=1 $DIR/$tfile ||
1816 error "$SETSTRIPE failed for $DIR/$tfile"
1818 multiop $DIR/$tfile Ow1yc
1819 REQ_DELAY=`lctl get_param -n osc.${FSNAME}-OST0000-osc-*.timeouts |
1820 awk '/portal 6/ {print $5}'`
1821 REQ_DELAY=$((${REQ_DELAY} + ${REQ_DELAY} / 4 + 5))
1823 do_facet ost1 lctl set_param fail_val=${REQ_DELAY}
1824 #define OBD_FAIL_OST_BRW_PAUSE_PACK 0x224
1825 do_facet ost1 $LCTL set_param fail_loc=0x224
1828 $SETSTRIPE --stripe-index=0 --stripe-count=1 $DIR/$tfile ||
1829 error "$SETSTRIPE failed"
1830 # force some real bulk transfer
1831 multiop $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c
1833 do_facet ost1 $LCTL set_param fail_loc=0
1834 # check for log message
1835 $LCTL dk | grep "Early reply #" || error "No early reply"
1837 # client should show REQ_DELAY estimates
1838 lctl get_param -n osc.${FSNAME}-OST0000-osc-*.timeouts | grep portal
1840 run_test 65b "AT: verify early replies on packed reply / bulk"
1842 test_66a() #bug 3055
1844 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1846 at_start || return 0
1847 lctl get_param -n mdc.${FSNAME}-MDT0000-mdc-*.timeouts | grep "portal 12"
1848 # adjust 5s at a time so no early reply is sent (within deadline)
1849 do_facet $SINGLEMDS "$LCTL set_param fail_val=5000"
1850 #define OBD_FAIL_PTLRPC_PAUSE_REQ 0x50a
1851 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x8000050a"
1852 createmany -o $DIR/$tfile 20 > /dev/null
1853 unlinkmany $DIR/$tfile 20 > /dev/null
1854 lctl get_param -n mdc.${FSNAME}-MDT0000-mdc-*.timeouts | grep "portal 12"
1855 do_facet $SINGLEMDS "$LCTL set_param fail_val=10000"
1856 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x8000050a"
1857 createmany -o $DIR/$tfile 20 > /dev/null
1858 unlinkmany $DIR/$tfile 20 > /dev/null
1859 lctl get_param -n mdc.${FSNAME}-MDT0000-mdc-*.timeouts | grep "portal 12"
1860 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0"
1862 createmany -o $DIR/$tfile 20 > /dev/null
1863 unlinkmany $DIR/$tfile 20 > /dev/null
1864 lctl get_param -n mdc.${FSNAME}-MDT0000-mdc-*.timeouts | grep "portal 12"
1865 CUR=$(lctl get_param -n mdc.${FSNAME}-MDT0000-mdc-*.timeouts | awk '/portal 12/ {print $5}')
1866 WORST=$(lctl get_param -n mdc.${FSNAME}-MDT0000-mdc-*.timeouts | awk '/portal 12/ {print $7}')
1867 echo "Current MDT timeout $CUR, worst $WORST"
1868 [ $CUR -lt $WORST ] || error "Current $CUR should be less than worst $WORST"
1870 run_test 66a "AT: verify MDT service time adjusts with no early replies"
1872 test_66b() #bug 3055
1874 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1876 at_start || return 0
1877 ORIG=$(lctl get_param -n mdc.${FSNAME}-MDT0000*.timeouts |
1878 awk '/network/ {print $4}')
1879 $LCTL set_param fail_val=$(($ORIG + 5))
1880 #define OBD_FAIL_PTLRPC_PAUSE_REP 0x50c
1881 $LCTL set_param fail_loc=0x50c
1882 ls $DIR/$tfile > /dev/null 2>&1
1883 $LCTL set_param fail_loc=0
1884 CUR=$(lctl get_param -n mdc.${FSNAME}-MDT0000*.timeouts |
1885 awk '/network/ {print $4}')
1886 WORST=$(lctl get_param -n mdc.${FSNAME}-MDT0000*.timeouts |
1887 awk '/network/ {print $6}')
1888 echo "network timeout orig $ORIG, cur $CUR, worst $WORST"
1889 [ $WORST -gt $ORIG ] ||
1890 error "Worst $WORST should be worse than orig $ORIG"
1892 run_test 66b "AT: verify net latency adjusts"
1894 test_67a() #bug 3055
1896 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1898 at_start || return 0
1899 CONN1=$(lctl get_param -n osc.*.stats | awk '/_connect/ {total+=$2} END {print total}')
1900 # sleeping threads may drive values above this
1901 do_facet ost1 "$LCTL set_param fail_val=400"
1902 #define OBD_FAIL_PTLRPC_PAUSE_REQ 0x50a
1903 do_facet ost1 "$LCTL set_param fail_loc=0x50a"
1904 createmany -o $DIR/$tfile 20 > /dev/null
1905 unlinkmany $DIR/$tfile 20 > /dev/null
1906 do_facet ost1 "$LCTL set_param fail_loc=0"
1907 CONN2=$(lctl get_param -n osc.*.stats | awk '/_connect/ {total+=$2} END {print total}')
1908 ATTEMPTS=$(($CONN2 - $CONN1))
1909 echo "$ATTEMPTS osc reconnect attempts on gradual slow"
1910 [ $ATTEMPTS -gt 0 ] &&
1911 error_ignore bz13721 "AT should have prevented reconnect"
1914 run_test 67a "AT: verify slow request processing doesn't induce reconnects"
1916 test_67b() #bug 3055
1918 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1920 at_start || return 0
1921 CONN1=$(lctl get_param -n osc.*.stats | awk '/_connect/ {total+=$2} END {print total}')
1923 # exhaust precreations on ost1
1924 local OST=$(ostname_from_index 0)
1925 local mdtosc=$(get_mdtosc_proc_path mds $OST)
1926 local last_id=$(do_facet $SINGLEMDS lctl get_param -n \
1927 osc.$mdtosc.prealloc_last_id)
1928 local next_id=$(do_facet $SINGLEMDS lctl get_param -n \
1929 osc.$mdtosc.prealloc_next_id)
1931 mkdir -p $DIR/$tdir/${OST} || error "mkdir $DIR/$tdir/${OST} failed"
1932 $SETSTRIPE -i 0 -c 1 $DIR/$tdir/${OST} || error "$SETSTRIPE failed"
1933 echo "Creating to objid $last_id on ost $OST..."
1934 #define OBD_FAIL_OST_PAUSE_CREATE 0x223
1935 do_facet ost1 "$LCTL set_param fail_val=20000"
1936 do_facet ost1 "$LCTL set_param fail_loc=0x80000223"
1937 createmany -o $DIR/$tdir/${OST}/f $next_id $((last_id - next_id + 2))
1940 do_facet ost1 "lctl get_param -n ost.OSS.ost_create.timeouts"
1942 CONN2=$(lctl get_param -n osc.*.stats | awk '/_connect/ {total+=$2} END {print total}')
1943 ATTEMPTS=$(($CONN2 - $CONN1))
1944 echo "$ATTEMPTS osc reconnect attempts on instant slow"
1945 # do it again; should not timeout
1946 do_facet ost1 "$LCTL set_param fail_loc=0x80000223"
1947 cp /etc/profile $DIR/$tfile || error "cp failed"
1948 do_facet ost1 "$LCTL set_param fail_loc=0"
1950 do_facet ost1 "lctl get_param -n ost.OSS.ost_create.timeouts"
1951 CONN3=$(lctl get_param -n osc.*.stats | awk '/_connect/ {total+=$2} END {print total}')
1952 ATTEMPTS=$(($CONN3 - $CONN2))
1953 echo "$ATTEMPTS osc reconnect attempts on 2nd slow"
1954 [ $ATTEMPTS -gt 0 ] && error "AT should have prevented reconnect"
1957 run_test 67b "AT: verify instant slowdown doesn't induce reconnects"
1959 test_68 () #bug 13813
1961 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1963 at_start || return 0
1964 local ldlm_enqueue_min=$(find /sys -name ldlm_enqueue_min)
1965 [ -z "$ldlm_enqueue_min" ] && skip "missing /sys/.../ldlm_enqueue_min" && return 0
1966 local ldlm_enqueue_min_r=$(do_facet ost1 "find /sys -name ldlm_enqueue_min")
1967 [ -z "$ldlm_enqueue_min_r" ] && skip "missing /sys/.../ldlm_enqueue_min in the ost1" && return 0
1968 local ENQ_MIN=$(cat $ldlm_enqueue_min)
1969 local ENQ_MIN_R=$(do_facet ost1 "cat $ldlm_enqueue_min_r")
1970 echo $TIMEOUT >> $ldlm_enqueue_min
1971 do_facet ost1 "echo $TIMEOUT >> $ldlm_enqueue_min_r"
1973 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
1974 $SETSTRIPE --stripe-index=0 -c 1 $DIR/$tdir ||
1975 error "$SETSTRIPE failed for $DIR/$tdir"
1976 #define OBD_FAIL_LDLM_PAUSE_CANCEL 0x312
1977 $LCTL set_param fail_val=$(($TIMEOUT - 1))
1978 $LCTL set_param fail_loc=0x80000312
1979 cp /etc/profile $DIR/$tdir/${tfile}_1 || error "1st cp failed $?"
1980 $LCTL set_param fail_val=$((TIMEOUT * 5 / 4))
1981 $LCTL set_param fail_loc=0x80000312
1982 cp /etc/profile $DIR/$tdir/${tfile}_2 || error "2nd cp failed $?"
1983 $LCTL set_param fail_loc=0
1985 echo $ENQ_MIN >> $ldlm_enqueue_min
1986 do_facet ost1 "echo $ENQ_MIN_R >> $ldlm_enqueue_min_r"
1990 run_test 68 "AT: verify slowing locks"
1993 # end of AT tests includes above lines
1995 # start multi-client tests
1997 [ -z "$CLIENTS" ] &&
1998 { skip "Need two or more clients." && return; }
1999 [ $CLIENTCOUNT -lt 2 ] &&
2000 { skip "Need two or more clients, have $CLIENTCOUNT" && return; }
2002 echo "mount clients $CLIENTS ..."
2003 zconf_mount_clients $CLIENTS $MOUNT
2005 local clients=${CLIENTS//,/ }
2006 echo "Write/read files on $DIR ; clients $CLIENTS ... "
2007 for CLIENT in $clients; do
2008 do_node $CLIENT dd bs=1M count=10 if=/dev/zero \
2009 of=$DIR/${tfile}_${CLIENT} 2>/dev/null ||
2010 error "dd failed on $CLIENT"
2013 local prev_client=$(echo $clients | sed 's/^.* \(.\+\)$/\1/')
2014 for C in ${CLIENTS//,/ }; do
2015 do_node $prev_client dd if=$DIR/${tfile}_${C} \
2016 of=/dev/null 2>/dev/null ||
2017 error "dd if=$DIR/${tfile}_${C} failed on $prev_client"
2023 run_test 70a "check multi client t-f"
2025 check_for_process () {
2030 killall_process $clients "$prog" -0
2034 local clients=${CLIENTS:-$HOSTNAME}
2036 zconf_mount_clients $clients $MOUNT
2039 [ "$SLOW" = "no" ] && duration=120
2040 # set duration to 900 because it takes some time to boot node
2041 [ "$FAILURE_MODE" = HARD ] && duration=900
2044 local start_ts=$(date +%s)
2045 local cmd="rundbench 1 -t $duration"
2047 if [ $MDSCOUNT -ge 2 ]; then
2048 test_mkdir -p -c$MDSCOUNT $DIR/$tdir
2049 $LFS setdirstripe -D -c$MDSCOUNT $DIR/$tdir
2051 do_nodesv $clients "set -x; MISSING_DBENCH_OK=$MISSING_DBENCH_OK \
2052 PATH=\$PATH:$LUSTRE/utils:$LUSTRE/tests/:$DBENCH_LIB \
2053 DBENCH_LIB=$DBENCH_LIB TESTSUITE=$TESTSUITE TESTNAME=$TESTNAME \
2054 MOUNT=$MOUNT DIR=$DIR/$tdir/\\\$(hostname) LCTL=$LCTL $cmd" &
2057 #LU-1897 wait for all dbench copies to start
2058 while ! check_for_process $clients dbench; do
2059 elapsed=$(($(date +%s) - start_ts))
2060 if [ $elapsed -gt $duration ]; then
2061 killall_process $clients dbench
2062 error "dbench failed to start on $clients!"
2067 log "Started rundbench load pid=$pid ..."
2069 elapsed=$(($(date +%s) - start_ts))
2070 local num_failovers=0
2072 while [ $elapsed -lt $duration ]; do
2073 if ! check_for_process $clients dbench; then
2074 error_noexit "dbench stopped on some of $clients!"
2075 killall_process $clients dbench
2079 replay_barrier mds$fail_index
2080 sleep 1 # give clients a time to do operations
2081 # Increment the number of failovers
2082 num_failovers=$((num_failovers+1))
2083 log "$TESTNAME fail mds$fail_index $num_failovers times"
2085 elapsed=$(($(date +%s) - start_ts))
2086 if [ $fail_index -ge $MDSCOUNT ]; then
2089 fail_index=$((fail_index+1))
2093 wait $pid || error "rundbench load on $clients failed!"
2095 run_test 70b "dbench ${MDSCOUNT}mdts recovery; $CLIENTCOUNT clients"
2096 # end multi-client tests
2101 local monitor_pid=$3
2103 local start_ts=$(date +%s)
2104 local num_failovers=0
2107 elapsed=$(($(date +%s) - start_ts))
2108 while [ $elapsed -lt $duration ]; do
2109 fail_index=$((RANDOM%max_index+1))
2110 kill -0 $monitor_pid ||
2111 error "$monitor_pid stopped"
2113 replay_barrier mds$fail_index
2115 # Increment the number of failovers
2116 num_failovers=$((num_failovers+1))
2117 log "$TESTNAME fail mds$fail_index $num_failovers times"
2119 elapsed=$(($(date +%s) - start_ts))
2125 rm -f $DIR/replay-single.70c.lck
2130 local clients=${CLIENTS:-$HOSTNAME}
2133 zconf_mount_clients $clients $MOUNT
2136 [ "$SLOW" = "no" ] && duration=180
2137 # set duration to 900 because it takes some time to boot node
2138 [ "$FAILURE_MODE" = HARD ] && duration=600
2141 local start_ts=$(date +%s)
2143 trap cleanup_70c EXIT
2145 while [ ! -e $DIR/replay-single.70c.lck ]; do
2146 test_mkdir -p -c$MDSCOUNT $DIR/$tdir || break
2147 if [ $MDSCOUNT -ge 2 ]; then
2148 $LFS setdirstripe -D -c$MDSCOUNT $DIR/$tdir ||
2149 error "set default dirstripe failed"
2151 cd $DIR/$tdir || break
2152 tar cf - /etc | tar xf - || error "tar failed in loop"
2156 echo "Started tar $tar_70c_pid"
2158 random_fail_mdt $MDSCOUNT $duration $tar_70c_pid
2159 kill -0 $tar_70c_pid || error "tar $tar_70c_pid stopped"
2161 touch $DIR/replay-single.70c.lck
2162 wait $tar_70c_pid || error "$?: tar failed"
2167 run_test 70c "tar ${MDSCOUNT}mdts recovery"
2171 kill -9 $mkdir_70d_pid
2175 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2176 local clients=${CLIENTS:-$HOSTNAME}
2179 zconf_mount_clients $clients $MOUNT
2182 [ "$SLOW" = "no" ] && duration=180
2183 # set duration to 900 because it takes some time to boot node
2184 [ "$FAILURE_MODE" = HARD ] && duration=900
2189 local start_ts=$(date +%s)
2191 trap cleanup_70d EXIT
2194 $LFS mkdir -i0 -c2 $DIR/$tdir/test || {
2198 $LFS mkdir -i1 -c2 $DIR/$tdir/test1 || {
2203 touch $DIR/$tdir/test/a || {
2207 mkdir $DIR/$tdir/test/b || {
2211 rm -rf $DIR/$tdir/test || {
2216 touch $DIR/$tdir/test1/a || {
2220 mkdir $DIR/$tdir/test1/b || {
2225 rm -rf $DIR/$tdir/test1 || {
2232 echo "Started $mkdir_70d_pid"
2234 random_fail_mdt $MDSCOUNT $duration $mkdir_70d_pid
2235 kill -0 $mkdir_70d_pid || error "mkdir/rmdir $mkdir_70d_pid stopped"
2240 run_test 70d "mkdir/rmdir striped dir ${MDSCOUNT}mdts recovery"
2244 kill -9 $rename_70e_pid
2248 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2249 local clients=${CLIENTS:-$HOSTNAME}
2252 echo ha > /proc/sys/lnet/debug
2253 zconf_mount_clients $clients $MOUNT
2256 [ "$SLOW" = "no" ] && duration=180
2257 # set duration to 900 because it takes some time to boot node
2258 [ "$FAILURE_MODE" = HARD ] && duration=900
2261 $LFS mkdir -i0 $DIR/$tdir/test_0
2262 $LFS mkdir -i0 $DIR/$tdir/test_1
2263 touch $DIR/$tdir/test_0/a
2264 touch $DIR/$tdir/test_1/b
2265 trap cleanup_70e EXIT
2268 mrename $DIR/$tdir/test_0/a $DIR/$tdir/test_1/b > \
2274 checkstat $DIR/$tdir/test_0/a && {
2275 echo "a still exists"
2279 checkstat $DIR/$tdir/test_1/b || {
2280 echo "b still exists"
2284 touch $DIR/$tdir/test_0/a || {
2285 echo "touch a fails"
2289 mrename $DIR/$tdir/test_1/b $DIR/$tdir/test_0/a > \
2297 echo "Started $rename_70e_pid"
2299 random_fail_mdt 2 $duration $rename_70e_pid
2300 kill -0 $rename_70e_pid || error "rename $rename_70e_pid stopped"
2305 run_test 70e "rename cross-MDT with random fails"
2307 test_70f_write_and_read(){
2312 echo "Write/read files in: '$DIR/$tdir', clients: '$CLIENTS' ..."
2313 for client in ${CLIENTS//,/ }; do
2314 [ -f $stopflag ] || return
2316 local tgtfile=$DIR/$tdir/$tfile.$client
2317 do_node $client dd $DD_OPTS bs=1M count=10 if=$srcfile \
2318 of=$tgtfile 2>/dev/null ||
2319 error "dd $DD_OPTS bs=1M count=10 if=$srcfile " \
2320 "of=$tgtfile failed on $client, rc=$?"
2323 local prev_client=$(echo ${CLIENTS//,/ } | awk '{ print $NF }')
2326 for client in ${CLIENTS//,/ }; do
2327 [ -f $stopflag ] || return
2329 # flush client cache in case test is running on only one client
2330 # do_node $client cancel_lru_locks osc
2331 do_node $client $LCTL set_param ldlm.namespaces.*.lru_size=clear
2333 tgtfile=$DIR/$tdir/$tfile.$client
2334 local md5=$(do_node $prev_client "md5sum $tgtfile")
2335 [ ${checksum[$index]// */} = ${md5// */} ] ||
2336 error "$tgtfile: checksum doesn't match on $prev_client"
2337 index=$((index + 1))
2347 mkdir -p $DIR/$tdir || error "cannot create $DIR/$tdir directory"
2348 $SETSTRIPE -c -1 $DIR/$tdir || error "cannot $SETSTRIPE $DIR/$tdir"
2351 while [ -f $stopflag ]; do
2352 test_70f_write_and_read $srcfile $stopflag
2353 # use direct IO and buffer cache in turns if loop
2354 [ -n "$DD_OPTS" ] && DD_OPTS="" || DD_OPTS="oflag=direct"
2358 test_70f_cleanup() {
2360 rm -f $TMP/$tfile.stop
2361 do_nodes $CLIENTS rm -f $TMP/$tfile
2362 rm -f $DIR/$tdir/$tfile.*
2366 # [ x$ost1failover_HOST = x$ost_HOST ] &&
2367 # { skip "Failover host not defined" && return; }
2368 # [ -z "$CLIENTS" ] &&
2369 # { skip "CLIENTS are not specified." && return; }
2370 # [ $CLIENTCOUNT -lt 2 ] &&
2371 # { skip "Need 2 or more clients, have $CLIENTCOUNT" && return; }
2373 [[ $(lustre_version_code ost1) -lt $(version_code 2.9.53) ]] &&
2374 skip "Need server version at least 2.9.53" && return
2376 echo "mount clients $CLIENTS ..."
2377 zconf_mount_clients $CLIENTS $MOUNT
2379 local srcfile=$TMP/$tfile
2383 trap test_70f_cleanup EXIT
2384 # create a different source file local to each client node so we can
2385 # detect if the file wasn't written out properly after failover
2386 do_nodes $CLIENTS dd bs=1M count=10 if=/dev/urandom of=$srcfile \
2387 2>/dev/null || error "can't create $srcfile on $CLIENTS"
2388 for client in ${CLIENTS//,/ }; do
2389 checksum[$index]=$(do_node $client "md5sum $srcfile")
2390 index=$((index + 1))
2394 [ "$SLOW" = "no" ] && duration=60
2395 # set duration to 900 because it takes some time to boot node
2396 [ "$FAILURE_MODE" = HARD ] && duration=900
2398 local stopflag=$TMP/$tfile.stop
2399 test_70f_loop $srcfile $stopflag &
2403 local num_failovers=0
2404 local start_ts=$SECONDS
2405 while [ $elapsed -lt $duration ]; do
2409 num_failovers=$((num_failovers + 1))
2410 log "$TESTNAME failing OST $num_failovers times"
2413 elapsed=$((SECONDS - start_ts))
2420 run_test 70f "OSS O_DIRECT recovery with $CLIENTCOUNT clients"
2424 kill -9 $mkdir_71a_pid
2427 random_double_fail_mdt() {
2430 local monitor_pid=$3
2432 local start_ts=$(date +%s)
2433 local num_failovers=0
2437 elapsed=$(($(date +%s) - start_ts))
2438 while [ $elapsed -lt $duration ]; do
2439 fail_index=$((RANDOM%max_index + 1))
2440 if [ $fail_index -eq $max_index ]; then
2443 second_index=$((fail_index + 1))
2445 kill -0 $monitor_pid ||
2446 error "$monitor_pid stopped"
2448 replay_barrier mds$fail_index
2449 replay_barrier mds$second_index
2451 # Increment the number of failovers
2452 num_failovers=$((num_failovers+1))
2453 log "fail mds$fail_index mds$second_index $num_failovers times"
2454 fail mds${fail_index},mds${second_index}
2455 elapsed=$(($(date +%s) - start_ts))
2460 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2461 local clients=${CLIENTS:-$HOSTNAME}
2464 zconf_mount_clients $clients $MOUNT
2467 [ "$SLOW" = "no" ] && duration=180
2468 # set duration to 900 because it takes some time to boot node
2469 [ "$FAILURE_MODE" = HARD ] && duration=900
2474 local start_ts=$(date +%s)
2476 trap cleanup_71a EXIT
2479 $LFS mkdir -i0 -c2 $DIR/$tdir/test
2480 rmdir $DIR/$tdir/test
2484 echo "Started $mkdir_71a_pid"
2486 random_double_fail_mdt 2 $duration $mkdir_71a_pid
2487 kill -0 $mkdir_71a_pid || error "mkdir/rmdir $mkdir_71a_pid stopped"
2492 run_test 71a "mkdir/rmdir striped dir with 2 mdts recovery"
2495 multiop_bg_pause $DIR/$tfile O_tSc ||
2496 error "multiop_bg_pause $DIR/$tfile failed"
2500 replay_barrier $SINGLEMDS
2501 #define OBD_FAIL_LDLM_ENQUEUE_NET 0x302
2502 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000302"
2505 wait $pid || error "multiop pid failed"
2506 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
2509 run_test 73a "open(O_CREAT), unlink, replay, reconnect before open replay, close"
2512 multiop_bg_pause $DIR/$tfile O_tSc ||
2513 error "multiop_bg_pause $DIR/$tfile failed"
2517 replay_barrier $SINGLEMDS
2518 #define OBD_FAIL_LDLM_REPLY 0x30c
2519 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000030c"
2522 wait $pid || error "multiop pid failed"
2523 [ -e $DIR/$tfile ] && error "file $DIR/$tfile should not exist"
2526 run_test 73b "open(O_CREAT), unlink, replay, reconnect at open_replay reply, close"
2530 local clients=${CLIENTS:-$HOSTNAME}
2532 zconf_umount_clients $clients $MOUNT
2534 facet_failover $SINGLEMDS
2535 zconf_mount_clients $clients $MOUNT
2537 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2538 rm $DIR/$tfile || error "rm $DIR/$tfile failed"
2539 clients_up || error "client evicted: $?"
2542 run_test 74 "Ensure applications don't fail waiting for OST recovery"
2544 remote_dir_check_80() {
2549 diridx=$($GETSTRIPE -M $remote_dir) ||
2550 error "$GETSTRIPE -M $remote_dir failed"
2551 [ $diridx -eq $MDTIDX ] || error "$diridx != $MDTIDX"
2553 createmany -o $remote_dir/f-%d 20 || error "creation failed"
2554 fileidx=$($GETSTRIPE -M $remote_dir/f-1) ||
2555 error "$GETSTRIPE -M $remote_dir/f-1 failed"
2556 [ $fileidx -eq $MDTIDX ] || error "$fileidx != $MDTIDX"
2562 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2563 ([ $FAILURE_MODE == "HARD" ] &&
2564 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2565 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2569 local remote_dir=$DIR/$tdir/remote_dir
2571 mkdir -p $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2572 #define OBD_FAIL_OUT_UPDATE_NET_REP 0x1701
2573 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
2574 $LFS mkdir -i $MDTIDX $remote_dir &
2580 wait $CLIENT_PID || error "remote creation failed"
2582 remote_dir_check_80 || error "remote dir check failed"
2583 rm -rf $DIR/$tdir || error "rmdir failed"
2587 run_test 80a "DNE: create remote dir, drop update rep from MDT0, fail MDT0"
2590 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2591 ([ $FAILURE_MODE == "HARD" ] &&
2592 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2593 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2597 local remote_dir=$DIR/$tdir/remote_dir
2599 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2600 #define OBD_FAIL_UPDATE_OBJ_NET_REP 0x1701
2601 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
2602 $LFS mkdir -i $MDTIDX $remote_dir &
2607 fail mds$((MDTIDX + 1))
2609 wait $CLIENT_PID || error "remote creation failed"
2611 remote_dir_check_80 || error "remote dir check failed"
2612 rm -rf $DIR/$tdir || error "rmdir failed"
2616 run_test 80b "DNE: create remote dir, drop update rep from MDT0, fail MDT1"
2619 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2620 ([ $FAILURE_MODE == "HARD" ] &&
2621 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2622 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2626 local remote_dir=$DIR/$tdir/remote_dir
2628 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2629 #define OBD_FAIL_UPDATE_OBJ_NET_REP 0x1701
2630 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
2631 $LFS mkdir -i $MDTIDX $remote_dir &
2637 fail mds$((MDTIDX + 1))
2639 wait $CLIENT_PID || error "remote creation failed"
2641 remote_dir_check_80 || error "remote dir check failed"
2642 rm -rf $DIR/$tdir || error "rmdir failed"
2646 run_test 80c "DNE: create remote dir, drop update rep from MDT1, fail MDT[0,1]"
2649 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2651 local remote_dir=$DIR/$tdir/remote_dir
2653 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2654 #define OBD_FAIL_UPDATE_OBJ_NET_REP 0x1701
2655 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
2656 $LFS mkdir -i $MDTIDX $remote_dir &
2659 # sleep 3 seconds to make sure MDTs are failed after
2660 # lfs mkdir -i has finished on all of MDTs.
2665 fail mds${MDTIDX},mds$((MDTIDX + 1))
2667 wait $CLIENT_PID || error "remote creation failed"
2669 remote_dir_check_80 || error "remote dir check failed"
2670 rm -rf $DIR/$tdir || error "rmdir failed"
2674 run_test 80d "DNE: create remote dir, drop update rep from MDT1, fail 2 MDTs"
2677 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2678 ([ $FAILURE_MODE == "HARD" ] &&
2679 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2680 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2684 local remote_dir=$DIR/$tdir/remote_dir
2686 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2687 # OBD_FAIL_MDS_REINT_NET_REP 0x119
2688 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
2689 $LFS mkdir -i $MDTIDX $remote_dir &
2692 # sleep 3 seconds to make sure MDTs are failed after
2693 # lfs mkdir -i has finished on all of MDTs.
2699 wait $CLIENT_PID || error "remote creation failed"
2701 remote_dir_check_80 || error "remote dir check failed"
2702 rm -rf $DIR/$tdir || error "rmdir failed"
2706 run_test 80e "DNE: create remote dir, drop MDT1 rep, fail MDT0"
2709 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2710 ([ $FAILURE_MODE == "HARD" ] &&
2711 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2712 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2715 local remote_dir=$DIR/$tdir/remote_dir
2717 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2718 # OBD_FAIL_MDS_REINT_NET_REP 0x119
2719 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
2720 $LFS mkdir -i $MDTIDX $remote_dir &
2724 fail mds$((MDTIDX + 1))
2726 wait $CLIENT_PID || error "remote creation failed"
2728 remote_dir_check_80 || error "remote dir check failed"
2729 rm -rf $DIR/$tdir || error "rmdir failed"
2733 run_test 80f "DNE: create remote dir, drop MDT1 rep, fail MDT1"
2736 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2737 ([ $FAILURE_MODE == "HARD" ] &&
2738 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2739 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2743 local remote_dir=$DIR/$tdir/remote_dir
2745 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2746 # OBD_FAIL_MDS_REINT_NET_REP 0x119
2747 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
2748 $LFS mkdir -i $MDTIDX $remote_dir &
2751 # sleep 3 seconds to make sure MDTs are failed after
2752 # lfs mkdir -i has finished on all of MDTs.
2758 fail mds$((MDTIDX + 1))
2760 wait $CLIENT_PID || error "remote creation failed"
2762 remote_dir_check_80 || error "remote dir check failed"
2763 rm -rf $DIR/$tdir || error "rmdir failed"
2767 run_test 80g "DNE: create remote dir, drop MDT1 rep, fail MDT0, then MDT1"
2770 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2772 local remote_dir=$DIR/$tdir/remote_dir
2774 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2775 # OBD_FAIL_MDS_REINT_NET_REP 0x119
2776 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
2777 $LFS mkdir -i $MDTIDX $remote_dir &
2780 # sleep 3 seconds to make sure MDTs are failed after
2781 # lfs mkdir -i has finished on all of MDTs.
2786 fail mds${MDTIDX},mds$((MDTIDX + 1))
2788 wait $CLIENT_PID || error "remote dir creation failed"
2790 remote_dir_check_80 || error "remote dir check failed"
2791 rm -rf $DIR/$tdir || error "rmdir failed"
2795 run_test 80h "DNE: create remote dir, drop MDT1 rep, fail 2 MDTs"
2798 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2799 ([ $FAILURE_MODE == "HARD" ] &&
2800 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2801 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2805 local remote_dir=$DIR/$tdir/remote_dir
2807 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2808 $LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
2810 touch $remote_dir || error "touch $remote_dir failed"
2811 # OBD_FAIL_OBJ_UPDATE_NET_REP 0x1701
2812 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
2817 fail mds$((MDTIDX + 1))
2819 wait $CLIENT_PID || error "rm remote dir failed"
2821 stat $remote_dir &>/dev/null && error "$remote_dir still exist!"
2823 rm -rf $DIR/$tdir || error "rmdir failed"
2827 run_test 81a "DNE: unlink remote dir, drop MDT0 update rep, fail MDT1"
2830 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2831 ([ $FAILURE_MODE == "HARD" ] &&
2832 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2833 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2836 local remote_dir=$DIR/$tdir/remote_dir
2838 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2839 $LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
2841 # OBD_FAIL_OBJ_UPDATE_NET_REP 0x1701
2842 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
2849 wait $CLIENT_PID || error "rm remote dir failed"
2851 stat $remote_dir &>/dev/null && error "$remote_dir still exist!"
2853 rm -rf $DIR/$tdir || error "rmdir failed"
2857 run_test 81b "DNE: unlink remote dir, drop MDT0 update reply, fail MDT0"
2860 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2861 ([ $FAILURE_MODE == "HARD" ] &&
2862 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2863 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2867 local remote_dir=$DIR/$tdir/remote_dir
2869 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2870 $LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
2872 # OBD_FAIL_OBJ_UPDATE_NET_REP 0x1701
2873 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
2880 fail mds$((MDTIDX + 1))
2882 wait $CLIENT_PID || error "rm remote dir failed"
2884 stat $remote_dir &>/dev/null && error "$remote_dir still exist!"
2886 rm -rf $DIR/$tdir || error "rmdir failed"
2890 run_test 81c "DNE: unlink remote dir, drop MDT0 update reply, fail MDT0,MDT1"
2893 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2895 local remote_dir=$DIR/$tdir/remote_dir
2897 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2898 $LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
2900 # OBD_FAIL_OBJ_UPDATE_NET_REP 0x1701
2901 do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
2907 fail mds${MDTIDX},mds$((MDTIDX + 1))
2909 wait $CLIENT_PID || error "rm remote dir failed"
2911 stat $remote_dir &>/dev/null && error "$remote_dir still exist!"
2913 rm -rf $DIR/$tdir || error "rmdir failed"
2917 run_test 81d "DNE: unlink remote dir, drop MDT0 update reply, fail 2 MDTs"
2920 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2921 ([ $FAILURE_MODE == "HARD" ] &&
2922 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2923 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2927 local remote_dir=$DIR/$tdir/remote_dir
2929 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2930 $LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
2932 # OBD_FAIL_MDS_REINT_NET_REP 0x119
2933 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
2936 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0
2941 wait $CLIENT_PID || error "rm remote dir failed"
2943 stat $remote_dir &>/dev/null && error "$remote_dir still exist!"
2945 rm -rf $DIR/$tdir || error "rmdir failed"
2949 run_test 81e "DNE: unlink remote dir, drop MDT1 req reply, fail MDT0"
2952 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2953 ([ $FAILURE_MODE == "HARD" ] &&
2954 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2955 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2959 local remote_dir=$DIR/$tdir/remote_dir
2961 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2962 $LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
2964 # OBD_FAIL_MDS_REINT_NET_REP 0x119
2965 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
2970 fail mds$((MDTIDX + 1))
2972 wait $CLIENT_PID || error "rm remote dir failed"
2974 stat $remote_dir &>/dev/null && error "$remote_dir still exist!"
2976 rm -rf $DIR/$tdir || error "rmdir failed"
2980 run_test 81f "DNE: unlink remote dir, drop MDT1 req reply, fail MDT1"
2983 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2984 ([ $FAILURE_MODE == "HARD" ] &&
2985 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
2986 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
2990 local remote_dir=$DIR/$tdir/remote_dir
2992 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2993 $LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
2995 # OBD_FAIL_MDS_REINT_NET_REP 0x119
2996 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
3003 fail mds$((MDTIDX + 1))
3005 wait $CLIENT_PID || error "rm remote dir failed"
3007 stat $remote_dir &>/dev/null && error "$remote_dir still exist!"
3009 rm -rf $DIR/$tdir || error "rmdir failed"
3013 run_test 81g "DNE: unlink remote dir, drop req reply, fail M0, then M1"
3016 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3018 local remote_dir=$DIR/$tdir/remote_dir
3020 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3021 $LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
3023 # OBD_FAIL_MDS_REINT_NET_REP 0x119
3024 do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
3030 fail mds${MDTIDX},mds$((MDTIDX + 1))
3032 wait $CLIENT_PID || error "rm remote dir failed"
3034 stat $remote_dir &>/dev/null && error "$remote_dir still exist!"
3036 rm -rf $DIR/$tdir || error "rmdir failed"
3040 run_test 81h "DNE: unlink remote dir, drop request reply, fail 2 MDTs"
3043 #define OBD_FAIL_MDS_OPEN_WAIT_CREATE 0x144
3044 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000144"
3045 createmany -o $DIR/$tfile- 1 &
3049 client_up || client_up || true # reconnect
3051 run_test 84a "stale open during export disconnect"
3053 test_85a() { #bug 16774
3054 lctl set_param -n ldlm.cancel_unused_locks_before_replay "1"
3056 for i in $(seq 100); do
3057 echo "tag-$i" > $DIR/$tfile-$i
3058 grep -q "tag-$i" $DIR/$tfile-$i || error "f2-$i"
3061 lov_id=$(lctl dl | grep "clilov")
3062 addr=$(echo $lov_id | awk '{print $4}' | awk -F '-' '{print $NF}')
3063 count=$(lctl get_param -n \
3064 ldlm.namespaces.*MDT0000*$addr.lock_unused_count)
3065 echo "before recovery: unused locks count = $count"
3069 count2=$(lctl get_param -n \
3070 ldlm.namespaces.*MDT0000*$addr.lock_unused_count)
3071 echo "after recovery: unused locks count = $count2"
3073 if [ $count2 -ge $count ]; then
3074 error "unused locks are not canceled"
3077 run_test 85a "check the cancellation of unused locks during recovery(IBITS)"
3079 test_85b() { #bug 16774
3080 lctl set_param -n ldlm.cancel_unused_locks_before_replay "1"
3082 create_pool $FSNAME.$TESTNAME ||
3083 error "unable to create pool $TESTNAME"
3084 do_facet mgs $LCTL pool_add $FSNAME.$TESTNAME $FSNAME-OST0000 ||
3085 error "unable to add pool $TESTNAME"
3087 $SETSTRIPE -c 1 -p $FSNAME.$TESTNAME $DIR
3089 for i in $(seq 100); do
3090 dd if=/dev/urandom of=$DIR/$tfile-$i bs=4096 \
3091 count=32 >/dev/null 2>&1
3094 cancel_lru_locks osc
3096 for i in $(seq 100); do
3097 dd if=$DIR/$tfile-$i of=/dev/null bs=4096 \
3098 count=32 >/dev/null 2>&1
3101 lov_id=$(lctl dl | grep "clilov")
3102 addr=$(echo $lov_id | awk '{print $4}' | awk -F '-' '{print $NF}')
3103 count=$(lctl get_param -n \
3104 ldlm.namespaces.*OST0000*$addr.lock_unused_count)
3105 echo "before recovery: unused locks count = $count"
3106 [ $count -ne 0 ] || error "unused locks ($count) should be zero"
3110 count2=$(lctl get_param \
3111 -n ldlm.namespaces.*OST0000*$addr.lock_unused_count)
3112 echo "after recovery: unused locks count = $count2"
3114 do_facet mgs $LCTL pool_remove $FSNAME.$TESTNAME $FSNAME-OST0000 ||
3115 error "unable to remove pool $TESTNAME"
3116 do_facet mgs $LCTL pool_destroy $FSNAME.$TESTNAME ||
3117 error "unable to destroy the pool $TESTNAME"
3119 if [ $count2 -ge $count ]; then
3120 error "unused locks are not canceled"
3123 run_test 85b "check the cancellation of unused locks during recovery(EXTENT)"
3126 local clients=${CLIENTS:-$HOSTNAME}
3128 zconf_umount_clients $clients $MOUNT
3129 do_facet $SINGLEMDS lctl set_param mdt.${FSNAME}-MDT*.exports.clear=0
3130 remount_facet $SINGLEMDS
3131 zconf_mount_clients $clients $MOUNT
3133 run_test 86 "umount server after clear nid_stats should not hit LBUG"
3136 do_facet ost1 "lctl set_param -n obdfilter.${ost1_svc}.sync_journal 0"
3139 $SETSTRIPE -i 0 -c 1 $DIR/$tfile
3140 dd if=/dev/urandom of=$DIR/$tfile bs=1024k count=8 ||
3141 error "dd to $DIR/$tfile failed"
3142 cksum=$(md5sum $DIR/$tfile | awk '{print $1}')
3143 cancel_lru_locks osc
3145 dd if=$DIR/$tfile of=/dev/null bs=1024k count=8 || error "Cannot read"
3146 cksum2=$(md5sum $DIR/$tfile | awk '{print $1}')
3147 if [ $cksum != $cksum2 ] ; then
3148 error "New checksum $cksum2 does not match original $cksum"
3151 run_test 87a "write replay"
3154 do_facet ost1 "lctl set_param -n obdfilter.${ost1_svc}.sync_journal 0"
3157 $SETSTRIPE -i 0 -c 1 $DIR/$tfile
3158 dd if=/dev/urandom of=$DIR/$tfile bs=1024k count=8 ||
3159 error "dd to $DIR/$tfile failed"
3160 sleep 1 # Give it a chance to flush dirty data
3161 echo TESTTEST | dd of=$DIR/$tfile bs=1 count=8 seek=64
3162 cksum=$(md5sum $DIR/$tfile | awk '{print $1}')
3163 cancel_lru_locks osc
3165 dd if=$DIR/$tfile of=/dev/null bs=1024k count=8 || error "Cannot read"
3166 cksum2=$(md5sum $DIR/$tfile | awk '{print $1}')
3167 if [ $cksum != $cksum2 ] ; then
3168 error "New checksum $cksum2 does not match original $cksum"
3171 run_test 87b "write replay with changed data (checksum resend)"
3173 test_88() { #bug 17485
3174 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3175 mkdir -p $TMP/$tdir || error "mkdir $TMP/$tdir failed"
3177 $SETSTRIPE -i 0 -c 1 $DIR/$tdir || error "$SETSTRIPE"
3180 replay_barrier $SINGLEMDS
3182 # exhaust precreations on ost1
3183 local OST=$(ostname_from_index 0)
3184 local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS $OST)
3185 local last_id=$(do_facet $SINGLEMDS lctl get_param -n osc.$mdtosc.prealloc_last_id)
3186 local next_id=$(do_facet $SINGLEMDS lctl get_param -n osc.$mdtosc.prealloc_next_id)
3187 echo "before test: last_id = $last_id, next_id = $next_id"
3189 echo "Creating to objid $last_id on ost $OST..."
3190 createmany -o $DIR/$tdir/f-%d $next_id $((last_id - next_id + 2)) ||
3191 error "createmany create files to last_id failed"
3193 #create some files to use some uncommitted objids
3194 last_id=$(($last_id + 1))
3195 createmany -o $DIR/$tdir/f-%d $last_id 8 ||
3196 error "createmany create files with uncommitted objids failed"
3198 last_id2=$(do_facet $SINGLEMDS lctl get_param -n osc.$mdtosc.prealloc_last_id)
3199 next_id2=$(do_facet $SINGLEMDS lctl get_param -n osc.$mdtosc.prealloc_next_id)
3200 echo "before recovery: last_id = $last_id2, next_id = $next_id2"
3202 # if test uses shutdown_facet && reboot_facet instead of facet_failover ()
3203 # it has to take care about the affected facets, bug20407
3204 local affected_mds1=$(affected_facets mds1)
3205 local affected_ost1=$(affected_facets ost1)
3207 shutdown_facet $SINGLEMDS
3210 reboot_facet $SINGLEMDS
3211 change_active $affected_mds1
3212 wait_for_facet $affected_mds1
3213 mount_facets $affected_mds1 || error "Restart of mds failed"
3216 change_active $affected_ost1
3217 wait_for_facet $affected_ost1
3218 mount_facets $affected_ost1 || error "Restart of ost1 failed"
3222 last_id2=$(do_facet $SINGLEMDS lctl get_param -n osc.$mdtosc.prealloc_last_id)
3223 next_id2=$(do_facet $SINGLEMDS lctl get_param -n osc.$mdtosc.prealloc_next_id)
3224 echo "after recovery: last_id = $last_id2, next_id = $next_id2"
3226 # create new files, which should use new objids, and ensure the orphan
3227 # cleanup phase for ost1 is completed at the same time
3228 for i in $(seq 8); do
3229 file_id=$(($last_id + 10 + $i))
3230 dd if=/dev/urandom of=$DIR/$tdir/f-$file_id bs=4096 count=128
3233 # if the objids were not recreated, then "ls" will fail with -ENOENT
3234 ls -l $DIR/$tdir/* || error "can't get the status of precreated files"
3237 # write into previously created files
3238 for i in $(seq 8); do
3239 file_id=$(($last_id + $i))
3240 dd if=/dev/urandom of=$DIR/$tdir/f-$file_id bs=4096 count=128
3241 cp -f $DIR/$tdir/f-$file_id $TMP/$tdir/
3244 # compare the content
3245 for i in $(seq 8); do
3246 file_id=$(($last_id + $i))
3247 cmp $TMP/$tdir/f-$file_id $DIR/$tdir/f-$file_id ||
3248 error "the content of file is modified!"
3253 run_test 88 "MDS should not assign same objid to different files "
3256 cancel_lru_locks osc
3257 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3258 rm -f $DIR/$tdir/$tfile
3260 wait_delete_completed
3261 BLOCKS1=$(df -P $MOUNT | tail -n 1 | awk '{ print $3 }')
3262 $SETSTRIPE -i 0 -c 1 $DIR/$tdir/$tfile
3263 dd if=/dev/zero bs=1M count=10 of=$DIR/$tdir/$tfile
3266 facet_failover $SINGLEMDS
3267 rm $DIR/$tdir/$tfile
3270 zconf_mount $(hostname) $MOUNT || error "mount fails"
3271 client_up || error "client_up failed"
3273 wait_delete_completed
3274 BLOCKS2=$(df -P $MOUNT | tail -n 1 | awk '{ print $3 }')
3275 [ $((BLOCKS2 - BLOCKS1)) -le 4 ] ||
3276 error $((BLOCKS2 - BLOCKS1)) blocks leaked
3279 run_test 89 "no disk space leak on late ost connection"
3285 change_active $facet
3286 wait_for_facet $facet
3287 mount_facet $facet || error "Restart of $facet failed"
3291 test_90() { # bug 19494
3292 local dir=$DIR/$tdir
3293 local ostfail=$(get_random_entry $(get_facets OST))
3295 if [[ $FAILURE_MODE = HARD ]]; then
3296 local affected=$(affected_facets $ostfail);
3297 if [[ "$affected" != $ostfail ]]; then
3298 skip not functional with FAILURE_MODE=$FAILURE_MODE, affected: $affected
3302 # ensure all OSTs are active to allow allocations
3305 mkdir $dir || error "mkdir $dir failed"
3307 echo "Create the files"
3309 # file "f${index}" striped over 1 OST
3310 # file "all" striped over all OSTs
3312 $SETSTRIPE -c $OSTCOUNT $dir/all ||
3313 error "setstripe failed to create $dir/all"
3315 for (( i=0; i<$OSTCOUNT; i++ )); do
3317 $SETSTRIPE -i $i -c 1 $f || error "$SETSTRIPE failed to create $f"
3319 # confirm setstripe actually created the stripe on the requested OST
3320 local uuid=$(ostuuid_from_index $i)
3321 for file in f$i all; do
3322 if [[ $dir/$file != $($LFS find --obd $uuid --name $file $dir) ]]; then
3323 $GETSTRIPE $dir/$file
3324 error wrong stripe: $file, uuid: $uuid
3329 # Before failing an OST, get its obd name and index
3330 local varsvc=${ostfail}_svc
3331 local obd=$(do_facet $ostfail lctl get_param \
3332 -n obdfilter.${!varsvc}.uuid)
3333 local index=$(($(facet_number $ostfail) - 1))
3335 echo "Fail $ostfail $obd, display the list of affected files"
3336 shutdown_facet $ostfail || error "shutdown_facet $ostfail failed"
3338 trap "cleanup_90 $ostfail" EXIT INT
3339 echo "General Query: lfs find $dir"
3340 local list=$($LFS find $dir)
3342 for (( i=0; i<$OSTCOUNT; i++ )); do
3343 list_member "$list" $dir/f$i ||
3344 error_noexit "lfs find $dir: no file f$i"
3346 list_member "$list" $dir/all ||
3347 error_noexit "lfs find $dir: no file all"
3349 # focus on the missing OST,
3350 # we expect to see only two files affected: "f$(index)" and "all"
3352 echo "Querying files on shutdown $ostfail: lfs find --obd $obd"
3353 list=$($LFS find --obd $obd $dir)
3355 for file in all f$index; do
3356 list_member "$list" $dir/$file ||
3357 error_noexit "lfs find does not report the affected $obd for $file"
3360 [[ $(echo $list | wc -w) -eq 2 ]] ||
3361 error_noexit "lfs find reports the wrong list of affected files ${#list[@]}"
3363 echo "Check getstripe: $GETSTRIPE -r --obd $obd"
3364 list=$($GETSTRIPE -r --obd $obd $dir)
3366 for file in all f$index; do
3367 echo "$list" | grep $dir/$file ||
3368 error_noexit "lfs getsripe does not report the affected $obd for $file"
3373 run_test 90 "lfs find identifies the missing striped file segments"
3376 local server_version=$(lustre_version_code $SINGLEMDS)
3377 [[ $server_version -ge $(version_code 2.6.90) ]] ||
3378 [[ $server_version -ge $(version_code 2.5.4) &&
3379 $server_version -lt $(version_code 2.5.50) ]] ||
3380 { skip "Need MDS version 2.5.4+ or 2.6.90+"; return; }
3382 cancel_lru_locks osc
3384 $SETSTRIPE -i 0 -c 1 $DIR/$tfile ||
3385 error "$SETSTRIPE $DIR/$tfile failed"
3386 dd if=/dev/zero of=$DIR/$tfile bs=1024 count=1 ||
3387 error "dd to $DIR/$tfile failed"
3388 #define OBD_FAIL_TGT_REPLAY_RECONNECT 0x715
3389 # We need to emulate a state that OST is waiting for other clients
3390 # not completing the recovery. Final ping is queued, but reply will be
3391 # sent on the recovery completion. It is done by sleep before
3392 # processing final pings
3393 do_facet ost1 "$LCTL set_param fail_val=40"
3394 do_facet ost1 "$LCTL set_param fail_loc=0x715"
3397 run_test 93a "replay + reconnect"
3400 local server_version=$(lustre_version_code $SINGLEMDS)
3401 [[ $server_version -ge $(version_code 2.7.90) ]] ||
3402 { skip "Need MDS version 2.7.90+"; return; }
3404 cancel_lru_locks mdc
3406 createmany -o $DIR/$tfile 20 ||
3407 error "createmany -o $DIR/$tfile failed"
3409 #define OBD_FAIL_TGT_REPLAY_RECONNECT 0x715
3410 # We need to emulate a state that MDT is waiting for other clients
3411 # not completing the recovery. Final ping is queued, but reply will be
3412 # sent on the recovery completion. It is done by sleep before
3413 # processing final pings
3414 do_facet mds1 "$LCTL set_param fail_val=80"
3415 do_facet mds1 "$LCTL set_param fail_loc=0x715"
3418 run_test 93b "replay + reconnect on mds"
3420 striped_dir_check_100() {
3421 local striped_dir=$DIR/$tdir/striped_dir
3422 local stripe_count=$($LFS getdirstripe -c $striped_dir)
3424 $LFS getdirstripe $striped_dir
3425 [ $stripe_count -eq 2 ] || error "$stripe_count != 2"
3427 createmany -o $striped_dir/f-%d 20 ||
3428 error "creation failed under striped dir"
3432 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3433 ([ $FAILURE_MODE == "HARD" ] &&
3434 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3435 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3438 local striped_dir=$DIR/$tdir/striped_dir
3441 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3443 #To make sure MDT1 and MDT0 are connected
3444 #otherwise it may create single stripe dir here
3445 $LFS setdirstripe -i1 $DIR/$tdir/remote_dir
3447 #define OBD_FAIL_OUT_UPDATE_NET_REP 0x1701
3448 do_facet mds$((MDTIDX+1)) lctl set_param fail_loc=0x1701
3449 $LFS setdirstripe -i0 -c2 $striped_dir &
3452 fail mds$((MDTIDX + 1))
3454 wait $CLIENT_PID || error "striped dir creation failed"
3456 striped_dir_check_100 || error "striped dir check failed"
3457 rm -rf $DIR/$tdir || error "rmdir failed"
3459 run_test 100a "DNE: create striped dir, drop update rep from MDT1, fail MDT1"
3462 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3463 ([ $FAILURE_MODE == "HARD" ] &&
3464 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3465 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3468 local striped_dir=$DIR/$tdir/striped_dir
3471 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3473 #To make sure MDT1 and MDT0 are connected
3474 #otherwise it may create single stripe dir here
3475 $LFS setdirstripe -i1 $DIR/$tdir/remote_dir
3477 # OBD_FAIL_MDS_REINT_NET_REP 0x119
3478 do_facet mds$MDTIDX lctl set_param fail_loc=0x119
3479 $LFS mkdir -i0 -c2 $striped_dir &
3484 wait $CLIENT_PID || error "striped dir creation failed"
3486 striped_dir_check_100 || error "striped dir check failed"
3487 rm -rf $DIR/$tdir || error "rmdir failed"
3489 run_test 100b "DNE: create striped dir, fail MDT0"
3491 test_101() { #LU-5648
3492 mkdir -p $DIR/$tdir/d1
3493 mkdir -p $DIR/$tdir/d2
3494 touch $DIR/$tdir/file0
3497 replay_barrier $SINGLEMDS
3498 for i in $(seq $num) ; do
3499 echo test$i > $DIR/$tdir/d1/file$i
3502 fail_abort $SINGLEMDS
3503 for i in $(seq $num) ; do
3504 touch $DIR/$tdir/d2/file$i
3505 test -s $DIR/$tdir/d2/file$i &&
3506 ls -al $DIR/$tdir/d2/file$i && error "file$i's size > 0"
3511 run_test 101 "Shouldn't reassign precreated objs to other files after recovery"
3520 [[ $(lctl get_param mdc.*.import |
3521 grep "connect_flags:.*multi_mod_rpc") ]] ||
3522 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
3524 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3525 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
3526 facet="mds$((0x$idx + 1))"
3528 # get current value of max_mod_rcps_in_flight
3529 num=$($LCTL get_param -n \
3530 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
3531 # set default value if client does not support multi mod RPCs
3532 [ -z "$num" ] && num=1
3534 echo "creating $num files ..."
3536 for i in $(seq $num); do
3537 touch $DIR/$tdir/file-$i
3540 # drop request on MDT to force resend
3541 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
3542 do_facet $facet "$LCTL set_param fail_loc=0x159"
3543 echo "launch $num chmod in parallel ($(date +%H:%M:%S)) ..."
3544 for i in $(seq $num); do
3545 chmod 0600 $DIR/$tdir/file-$i &
3549 do_facet $facet "$LCTL set_param fail_loc=0"
3550 for pid in $pids; do
3551 wait $pid || error "chmod failed"
3553 echo "done ($(date +%H:%M:%S))"
3555 # check chmod succeed
3556 for i in $(seq $num); do
3557 checkstat -vp 0600 $DIR/$tdir/file-$i
3562 run_test 102a "check resend (request lost) with multiple modify RPCs in flight"
3571 [[ $(lctl get_param mdc.*.import |
3572 grep "connect_flags:.*multi_mod_rpc") ]] ||
3573 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
3575 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3576 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
3577 facet="mds$((0x$idx + 1))"
3579 # get current value of max_mod_rcps_in_flight
3580 num=$($LCTL get_param -n \
3581 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
3582 # set default value if client does not support multi mod RPCs
3583 [ -z "$num" ] && num=1
3585 echo "creating $num files ..."
3587 for i in $(seq $num); do
3588 touch $DIR/$tdir/file-$i
3591 # drop reply on MDT to force reconstruction
3592 #define OBD_FAIL_MDS_REINT_MULTI_NET_REP 0x15a
3593 do_facet $facet "$LCTL set_param fail_loc=0x15a"
3594 echo "launch $num chmod in parallel ($(date +%H:%M:%S)) ..."
3595 for i in $(seq $num); do
3596 chmod 0600 $DIR/$tdir/file-$i &
3600 do_facet $facet "$LCTL set_param fail_loc=0"
3601 for pid in $pids; do
3602 wait $pid || error "chmod failed"
3604 echo "done ($(date +%H:%M:%S))"
3606 # check chmod succeed
3607 for i in $(seq $num); do
3608 checkstat -vp 0600 $DIR/$tdir/file-$i
3613 run_test 102b "check resend (reply lost) with multiple modify RPCs in flight"
3622 [[ $(lctl get_param mdc.*.import |
3623 grep "connect_flags:.*multi_mod_rpc") ]] ||
3624 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
3626 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3627 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
3628 facet="mds$((0x$idx + 1))"
3630 # get current value of max_mod_rcps_in_flight
3631 num=$($LCTL get_param -n \
3632 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
3633 # set default value if client does not support multi mod RPCs
3634 [ -z "$num" ] && num=1
3636 echo "creating $num files ..."
3638 for i in $(seq $num); do
3639 touch $DIR/$tdir/file-$i
3642 replay_barrier $facet
3645 #define OBD_FAIL_MDS_REINT_MULTI_NET_REP 0x15a
3646 do_facet $facet "$LCTL set_param fail_loc=0x15a"
3647 echo "launch $num chmod in parallel ($(date +%H:%M:%S)) ..."
3648 for i in $(seq $num); do
3649 chmod 0600 $DIR/$tdir/file-$i &
3653 do_facet $facet "$LCTL set_param fail_loc=0"
3658 for pid in $pids; do
3659 wait $pid || error "chmod failed"
3661 echo "done ($(date +%H:%M:%S))"
3663 # check chmod succeed
3664 for i in $(seq $num); do
3665 checkstat -vp 0600 $DIR/$tdir/file-$i
3670 run_test 102c "check replay w/o reconstruction with multiple mod RPCs in flight"
3679 [[ $(lctl get_param mdc.*.import |
3680 grep "connect_flags:.*multi_mod_rpc") ]] ||
3681 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
3683 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3684 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
3685 facet="mds$((0x$idx + 1))"
3687 # get current value of max_mod_rcps_in_flight
3688 num=$($LCTL get_param -n \
3689 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
3690 # set default value if client does not support multi mod RPCs
3691 [ -z "$num" ] && num=1
3693 echo "creating $num files ..."
3695 for i in $(seq $num); do
3696 touch $DIR/$tdir/file-$i
3700 #define OBD_FAIL_MDS_REINT_MULTI_NET_REP 0x15a
3701 do_facet $facet "$LCTL set_param fail_loc=0x15a"
3702 echo "launch $num chmod in parallel ($(date +%H:%M:%S)) ..."
3703 for i in $(seq $num); do
3704 chmod 0600 $DIR/$tdir/file-$i &
3709 # write MDT transactions to disk
3710 do_facet $facet "sync; sync; sync"
3712 do_facet $facet "$LCTL set_param fail_loc=0"
3717 for pid in $pids; do
3718 wait $pid || error "chmod failed"
3720 echo "done ($(date +%H:%M:%S))"
3722 # check chmod succeed
3723 for i in $(seq $num); do
3724 checkstat -vp 0600 $DIR/$tdir/file-$i
3729 run_test 102d "check replay & reconstruction with multiple mod RPCs in flight"
3732 remote_mds_nodsh && skip "remote MDS with nodsh" && return
3733 #define OBD_FAIL_MDS_TRACK_OVERFLOW 0x162
3734 do_facet mds1 $LCTL set_param fail_loc=0x80000162
3737 createmany -o $DIR/$tdir/t- 30 ||
3738 error "create files on remote directory failed"
3740 rm -rf $DIR/$tdir/t-*
3742 #MDS should crash with tr->otr_next_id overflow
3745 run_test 103 "Check otr_next_id overflow"
3748 check_striped_dir_110()
3750 $CHECKSTAT -t dir $DIR/$tdir/striped_dir ||
3751 error "create striped dir failed"
3752 local stripe_count=$($LFS getdirstripe -c $DIR/$tdir/striped_dir)
3753 [ $stripe_count -eq $MDSCOUNT ] ||
3754 error "$stripe_count != 2 after recovery"
3758 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3759 ([ $FAILURE_MODE == "HARD" ] &&
3760 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3761 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3766 $LFS mkdir -i1 -c$MDSCOUNT $DIR/$tdir/striped_dir
3769 check_striped_dir_110 || error "check striped_dir failed"
3770 rm -rf $DIR/$tdir || error "rmdir failed"
3774 run_test 110a "DNE: create striped dir, fail MDT1"
3777 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3778 ([ $FAILURE_MODE == "HARD" ] &&
3779 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3780 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3785 $LFS mkdir -i1 -c$MDSCOUNT $DIR/$tdir/striped_dir
3788 zconf_mount $(hostname) $MOUNT
3789 client_up || return 1
3791 check_striped_dir_110 || error "check striped_dir failed"
3793 rm -rf $DIR/$tdir || error "rmdir failed"
3797 run_test 110b "DNE: create striped dir, fail MDT1 and client"
3800 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3801 ([ $FAILURE_MODE == "HARD" ] &&
3802 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3803 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3808 $LFS mkdir -i1 -c$MDSCOUNT $DIR/$tdir/striped_dir
3811 check_striped_dir_110 || error "check striped_dir failed"
3813 rm -rf $DIR/$tdir || error "rmdir failed"
3817 run_test 110c "DNE: create striped dir, fail MDT2"
3820 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3821 ([ $FAILURE_MODE == "HARD" ] &&
3822 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3823 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3828 $LFS mkdir -i1 -c$MDSCOUNT $DIR/$tdir/striped_dir
3831 zconf_mount $(hostname) $MOUNT
3832 client_up || return 1
3834 check_striped_dir_110 || error "check striped_dir failed"
3836 rm -rf $DIR/$tdir || error "rmdir failed"
3840 run_test 110d "DNE: create striped dir, fail MDT2 and client"
3843 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3844 ([ $FAILURE_MODE == "HARD" ] &&
3845 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3846 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3851 $LFS mkdir -i1 -c$MDSCOUNT $DIR/$tdir/striped_dir
3855 zconf_mount $(hostname) $MOUNT
3856 client_up || return 1
3858 check_striped_dir_110 || error "check striped_dir failed"
3860 rm -rf $DIR/$tdir || error "rmdir failed"
3864 run_test 110e "DNE: create striped dir, uncommit on MDT2, fail client/MDT1/MDT2"
3867 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3868 ([ $FAILURE_MODE == "HARD" ] &&
3869 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3870 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3876 $LFS mkdir -i1 -c$MDSCOUNT $DIR/$tdir/striped_dir
3879 check_striped_dir_110 || error "check striped_dir failed"
3881 rm -rf $DIR/$tdir || error "rmdir failed"
3885 run_test 110f "DNE: create striped dir, fail MDT1/MDT2"
3888 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3889 ([ $FAILURE_MODE == "HARD" ] &&
3890 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3891 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3896 $LFS mkdir -i1 -c$MDSCOUNT $DIR/$tdir/striped_dir
3900 zconf_mount $(hostname) $MOUNT
3901 client_up || return 1
3903 check_striped_dir_110 || error "check striped_dir failed"
3905 rm -rf $DIR/$tdir || error "rmdir failed"
3909 run_test 110g "DNE: create striped dir, uncommit on MDT1, fail client/MDT1/MDT2"
3912 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3913 ([ $FAILURE_MODE == "HARD" ] &&
3914 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3915 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3919 $LFS mkdir -i1 -c2 $DIR/$tdir/striped_dir
3921 rm -rf $DIR/$tdir/striped_dir
3924 $CHECKSTAT -t dir $DIR/$tdir/striped_dir &&
3925 error "striped dir still exists"
3928 run_test 111a "DNE: unlink striped dir, fail MDT1"
3931 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3932 ([ $FAILURE_MODE == "HARD" ] &&
3933 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3934 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3938 $LFS mkdir -i1 -c2 $DIR/$tdir/striped_dir
3940 rm -rf $DIR/$tdir/striped_dir
3943 zconf_mount $(hostname) $MOUNT
3944 client_up || return 1
3946 $CHECKSTAT -t dir $DIR/$tdir/striped_dir &&
3947 error "striped dir still exists"
3950 run_test 111b "DNE: unlink striped dir, fail MDT2"
3953 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3954 ([ $FAILURE_MODE == "HARD" ] &&
3955 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3956 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3960 $LFS mkdir -i1 -c2 $DIR/$tdir/striped_dir
3962 rm -rf $DIR/$tdir/striped_dir
3966 zconf_mount $(hostname) $MOUNT
3967 client_up || return 1
3968 $CHECKSTAT -t dir $DIR/$tdir/striped_dir &&
3969 error "striped dir still exists"
3972 run_test 111c "DNE: unlink striped dir, uncommit on MDT1, fail client/MDT1/MDT2"
3975 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3976 ([ $FAILURE_MODE == "HARD" ] &&
3977 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
3978 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
3982 $LFS mkdir -i1 -c2 $DIR/$tdir/striped_dir
3984 rm -rf $DIR/$tdir/striped_dir
3988 zconf_mount $(hostname) $MOUNT
3989 client_up || return 1
3990 $CHECKSTAT -t dir $DIR/$tdir/striped_dir &&
3991 error "striped dir still exists"
3995 run_test 111d "DNE: unlink striped dir, uncommit on MDT2, fail client/MDT1/MDT2"
3998 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
3999 ([ $FAILURE_MODE == "HARD" ] &&
4000 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4001 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4005 $LFS mkdir -i1 -c2 $DIR/$tdir/striped_dir
4007 rm -rf $DIR/$tdir/striped_dir
4010 $CHECKSTAT -t dir $DIR/$tdir/striped_dir &&
4011 error "striped dir still exists"
4014 run_test 111e "DNE: unlink striped dir, uncommit on MDT2, fail MDT1/MDT2"
4017 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
4018 ([ $FAILURE_MODE == "HARD" ] &&
4019 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4020 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4024 $LFS mkdir -i1 -c2 $DIR/$tdir/striped_dir
4026 rm -rf $DIR/$tdir/striped_dir
4029 $CHECKSTAT -t dir $DIR/$tdir/striped_dir &&
4030 error "striped dir still exists"
4033 run_test 111f "DNE: unlink striped dir, uncommit on MDT1, fail MDT1/MDT2"
4036 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
4037 ([ $FAILURE_MODE == "HARD" ] &&
4038 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4039 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4043 $LFS mkdir -i1 -c2 $DIR/$tdir/striped_dir
4046 rm -rf $DIR/$tdir/striped_dir
4048 $CHECKSTAT -t dir $DIR/$tdir/striped_dir &&
4049 error "striped dir still exists"
4052 run_test 111g "DNE: unlink striped dir, fail MDT1/MDT2"
4054 test_112_rename_prepare() {
4055 mkdir -p $DIR/$tdir/src_dir
4056 $LFS mkdir -i 1 $DIR/$tdir/src_dir/src_child ||
4057 error "create remote source failed"
4059 touch $DIR/$tdir/src_dir/src_child/a
4061 $LFS mkdir -i 2 $DIR/$tdir/tgt_dir ||
4062 error "create remote target dir failed"
4064 $LFS mkdir -i 3 $DIR/$tdir/tgt_dir/tgt_child ||
4065 error "create remote target child failed"
4070 $CHECKSTAT -t dir $DIR/$tdir/src_dir/src_child &&
4071 error "src_child still exists after rename"
4073 $CHECKSTAT -t file $DIR/$tdir/tgt_dir/tgt_child/a ||
4074 error "missing file(a) after rename"
4078 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4079 ([ $FAILURE_MODE == "HARD" ] &&
4080 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4081 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4084 test_112_rename_prepare
4087 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
4088 error "rename dir cross MDT failed!"
4092 rm -rf $DIR/$tdir || error "rmdir failed"
4094 run_test 112a "DNE: cross MDT rename, fail MDT1"
4097 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4098 ([ $FAILURE_MODE == "HARD" ] &&
4099 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4100 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4103 test_112_rename_prepare
4106 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
4107 error "rename dir cross MDT failed!"
4112 rm -rf $DIR/$tdir || error "rmdir failed"
4114 run_test 112b "DNE: cross MDT rename, fail MDT2"
4117 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4118 ([ $FAILURE_MODE == "HARD" ] &&
4119 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4120 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4123 test_112_rename_prepare
4126 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
4127 error "rename dir cross MDT failed!"
4132 rm -rf $DIR/$tdir || error "rmdir failed"
4134 run_test 112c "DNE: cross MDT rename, fail MDT3"
4137 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4138 ([ $FAILURE_MODE == "HARD" ] &&
4139 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4140 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4143 test_112_rename_prepare
4146 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
4147 error "rename dir cross MDT failed!"
4152 rm -rf $DIR/$tdir || error "rmdir failed"
4154 run_test 112d "DNE: cross MDT rename, fail MDT4"
4157 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4158 ([ $FAILURE_MODE == "HARD" ] &&
4159 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4160 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4163 test_112_rename_prepare
4167 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
4168 error "rename dir cross MDT failed!"
4173 rm -rf $DIR/$tdir || error "rmdir failed"
4175 run_test 112e "DNE: cross MDT rename, fail MDT1 and MDT2"
4178 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4179 ([ $FAILURE_MODE == "HARD" ] &&
4180 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4181 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4184 test_112_rename_prepare
4188 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
4189 error "rename dir cross MDT failed!"
4194 rm -rf $DIR/$tdir || error "rmdir failed"
4196 run_test 112f "DNE: cross MDT rename, fail MDT1 and MDT3"
4199 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4200 ([ $FAILURE_MODE == "HARD" ] &&
4201 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4202 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4205 test_112_rename_prepare
4209 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
4210 error "rename dir cross MDT failed!"
4215 rm -rf $DIR/$tdir || error "rmdir failed"
4217 run_test 112g "DNE: cross MDT rename, fail MDT1 and MDT4"
4220 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4221 ([ $FAILURE_MODE == "HARD" ] &&
4222 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4223 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4226 test_112_rename_prepare
4230 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
4231 error "rename dir cross MDT failed!"
4236 rm -rf $DIR/$tdir || error "rmdir failed"
4238 run_test 112h "DNE: cross MDT rename, fail MDT2 and MDT3"
4241 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4242 ([ $FAILURE_MODE == "HARD" ] &&
4243 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4244 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4247 test_112_rename_prepare
4251 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
4252 error "rename dir cross MDT failed!"
4257 rm -rf $DIR/$tdir || error "rmdir failed"
4259 run_test 112i "DNE: cross MDT rename, fail MDT2 and MDT4"
4262 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4263 ([ $FAILURE_MODE == "HARD" ] &&
4264 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4265 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4268 test_112_rename_prepare
4272 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
4273 error "rename dir cross MDT failed!"
4278 rm -rf $DIR/$tdir || error "rmdir failed"
4280 run_test 112j "DNE: cross MDT rename, fail MDT3 and MDT4"
4283 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4284 ([ $FAILURE_MODE == "HARD" ] &&
4285 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4286 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4289 test_112_rename_prepare
4294 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
4295 error "rename dir cross MDT failed!"
4300 rm -rf $DIR/$tdir || error "rmdir failed"
4302 run_test 112k "DNE: cross MDT rename, fail MDT1,MDT2,MDT3"
4305 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4306 ([ $FAILURE_MODE == "HARD" ] &&
4307 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4308 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4311 test_112_rename_prepare
4316 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
4317 error "rename dir cross MDT failed!"
4322 rm -rf $DIR/$tdir || error "rmdir failed"
4324 run_test 112l "DNE: cross MDT rename, fail MDT1,MDT2,MDT4"
4327 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4328 ([ $FAILURE_MODE == "HARD" ] &&
4329 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4330 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4333 test_112_rename_prepare
4338 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
4339 error "rename dir cross MDT failed!"
4344 rm -rf $DIR/$tdir || error "rmdir failed"
4346 run_test 112m "DNE: cross MDT rename, fail MDT1,MDT3,MDT4"
4349 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4350 ([ $FAILURE_MODE == "HARD" ] &&
4351 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4352 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4355 test_112_rename_prepare
4360 mrename $DIR/$tdir/src_dir/src_child $DIR/$tdir/tgt_dir/tgt_child ||
4361 error "rename dir cross MDT failed!"
4366 rm -rf $DIR/$tdir || error "rmdir failed"
4368 run_test 112n "DNE: cross MDT rename, fail MDT2,MDT3,MDT4"
4371 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
4372 ([ $FAILURE_MODE == "HARD" ] &&
4373 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4374 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4382 for ((j=0;j<$((MDSCOUNT));j++)); do
4383 fail_index=$((fail_index+1))
4384 index=$((fail_index % MDSCOUNT))
4385 replay_barrier mds$((index + 1))
4386 for ((i=0;i<5;i++)); do
4387 test_mkdir -i$index -c$MDSCOUNT $DIR/$tdir/test_$i ||
4388 error "create striped dir $DIR/$tdir/test_$i"
4391 fail mds$((index + 1))
4392 for ((i=0;i<5;i++)); do
4393 checkstat -t dir $DIR/$tdir/test_$i ||
4394 error "$DIR/$tdir/test_$i does not exist!"
4396 rm -rf $DIR/$tdir/test_* ||
4400 run_test 115 "failover for create/unlink striped directory"
4403 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
4404 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] &&
4405 skip "Do not support large update log before 2.7.55" &&
4407 ([ $FAILURE_MODE == "HARD" ] &&
4408 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4409 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4416 # OBD_FAIL_SPLIT_UPDATE_REC 0x1702
4417 do_facet mds1 "lctl set_param fail_loc=0x80001702"
4418 $LFS setdirstripe -c$MDSCOUNT $DIR/$tdir/striped_dir
4421 $CHECKSTAT -t dir $DIR/$tdir/striped_dir ||
4422 error "stried_dir does not exists"
4424 run_test 116a "large update log master MDT recovery"
4427 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
4428 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] &&
4429 skip "Do not support large update log before 2.7.55" &&
4432 ([ $FAILURE_MODE == "HARD" ] &&
4433 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4434 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4441 # OBD_FAIL_SPLIT_UPDATE_REC 0x1702
4442 do_facet mds2 "lctl set_param fail_loc=0x80001702"
4443 $LFS setdirstripe -c$MDSCOUNT $DIR/$tdir/striped_dir
4446 $CHECKSTAT -t dir $DIR/$tdir/striped_dir ||
4447 error "stried_dir does not exists"
4449 run_test 116b "large update log slave MDT recovery"
4452 [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" && return 0
4453 ([ $FAILURE_MODE == "HARD" ] &&
4454 [ "$(facet_host mds1)" == "$(facet_host mds2)" ]) &&
4455 skip "MDTs needs to be on diff hosts for HARD fail mode" &&
4461 $LFS setdirstripe -i0 -c$MDSCOUNT $DIR/$tdir/remote_dir
4462 $LFS setdirstripe -i1 -c$MDSCOUNT $DIR/$tdir/remote_dir_1
4465 # Let's set rdonly on all MDTs, so client will send
4466 # replay requests on all MDTs and replay these requests
4467 # at the same time. This test will verify the recovery
4468 # will not be deadlock in this case, LU-7531.
4469 for ((index = 0; index < $((MDSCOUNT)); index++)); do
4470 replay_barrier mds$((index + 1))
4471 if [ -z $mds_indexs ]; then
4472 mds_indexs="${mds_indexs}mds$((index+1))"
4474 mds_indexs="${mds_indexs},mds$((index+1))"
4478 rm -rf $DIR/$tdir/remote_dir
4479 rm -rf $DIR/$tdir/remote_dir_1
4483 rm -rf $DIR/$tdir || error "rmdir failed"
4485 run_test 117 "DNE: cross MDT unlink, fail MDT1 and MDT2"
4488 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
4489 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.64) ] &&
4490 skip "Do not support large update log before 2.7.64" &&
4495 $LFS setdirstripe -c2 $DIR/$tdir/striped_dir ||
4496 error "setdirstripe fails"
4497 $LFS setdirstripe -c2 $DIR/$tdir/striped_dir1 ||
4498 error "setdirstripe fails 1"
4499 rm -rf $DIR/$tdir/striped_dir* || error "rmdir fails"
4501 # OBD_FAIL_INVALIDATE_UPDATE 0x1705
4502 do_facet mds1 "lctl set_param fail_loc=0x1705"
4503 $LFS setdirstripe -c2 $DIR/$tdir/striped_dir
4504 $LFS setdirstripe -c2 $DIR/$tdir/striped_dir1
4505 do_facet mds1 "lctl set_param fail_loc=0x0"
4508 $LFS setdirstripe -c2 $DIR/$tdir/striped_dir
4509 $LFS setdirstripe -c2 $DIR/$tdir/striped_dir1
4514 run_test 118 "invalidate osp update will not cause update log corruption"
4517 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
4518 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.64) ] &&
4519 skip "Do not support large update log before 2.7.64" &&
4522 local hard_timeout=$(do_facet mds1 \
4523 "lctl get_param -n mdt.$FSNAME-MDT0000.recovery_time_hard")
4525 local clients=${CLIENTS:-$HOSTNAME}
4526 local time_min=$(recovery_time_min)
4529 mkdir $DIR/$tdir/tmp
4530 rmdir $DIR/$tdir/tmp
4533 mkdir $DIR/$tdir/dir_1
4534 for ((i = 0; i < 20; i++)); do
4535 $LFS setdirstripe -c2 $DIR/$tdir/stripe_dir-$i
4542 #define OBD_FAIL_TGT_REPLAY_DELAY 0x714
4543 do_facet mds1 $LCTL set_param fail_loc=0x80000714
4544 #sleep (timeout + 5), so mds will evict the client exports,
4545 #but DNE update recovery will keep going.
4546 do_facet mds1 $LCTL set_param fail_val=$((time_min + 5))
4548 mount_facet mds1 "-o recovery_time_hard=$time_min"
4550 wait_clients_import_state "$clients" mds1 FULL
4552 clients_up || clients_up || error "failover df: $?"
4554 #revert back the hard timeout
4555 do_facet mds1 $LCTL set_param \
4556 mdt.$FSNAME-MDT0000.recovery_time_hard=$hard_timeout
4558 for ((i = 0; i < 20; i++)); do
4559 stripe_count=$($LFS getdirstripe -c $DIR/$tdir/stripe_dir-$i)
4560 [ $stripe_count == 2 ] || {
4561 error "stripe_dir-$i creation replay fails"
4566 run_test 119 "timeout of normal replay does not cause DNE replay fails "
4569 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
4570 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.64) ] &&
4571 skip "Do not support large update log before 2.7.64" &&
4575 replay_barrier_nosync mds1
4576 for ((i = 0; i < 20; i++)); do
4577 mkdir $DIR/$tdir/dir-$i || {
4578 error "create dir-$i fails"
4581 $LFS setdirstripe -c2 $DIR/$tdir/stripe_dir-$i || {
4582 error "create stripe_dir-$i fails"
4589 for ((i = 0; i < 20; i++)); do
4590 [ ! -e "$DIR/$tdir/dir-$i" ] || {
4591 error "dir-$i still exists"
4594 [ ! -e "$DIR/$tdir/stripe_dir-$i" ] || {
4595 error "stripe_dir-$i still exists"
4600 run_test 120 "DNE fail abort should stop both normal and DNE replay"
4603 check_and_cleanup_lustre