6 ALWAYS_EXCEPT="24 27 52 $RECOVERY_SMALL_EXCEPT"
8 PTLDEBUG=${PTLDEBUG:--1}
9 LUSTRE=${LUSTRE:-`dirname $0`/..}
10 . $LUSTRE/tests/test-framework.sh
12 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
14 if [ "$FAILURE_MODE" = "HARD" ] && mixed_ost_devs; then
15 CONFIG_EXCEPTIONS="52"
16 echo -n "Several ost services on one ost node are used with FAILURE_MODE=$FAILURE_MODE. "
17 echo "Except the tests: $CONFIG_EXCEPTIONS"
18 ALWAYS_EXCEPT="$ALWAYS_EXCEPT $CONFIG_EXCEPTIONS"
21 remote_mds_nodsh && skip "remote MDS with nodsh" && exit 0
23 # also long tests: 19, 21a, 21e, 21f, 23, 27
24 # 1 2.5 2.5 4 4 (min)"
25 [ "$SLOW" = "no" ] && EXCEPT_SLOW="17 26a 26b 50 51 57"
29 # Allow us to override the setup if we already have a mounted system by
30 # setting SETUP=" " and CLEANUP=" "
32 CLEANUP=${CLEANUP:-""}
34 check_and_setup_lustre
37 rm -rf $DIR/[df][0-9]*
40 drop_request "mcreate $DIR/f1" || return 1
41 drop_reint_reply "mcreate $DIR/f2" || return 2
43 run_test 1 "mcreate: drop req, drop rep"
46 drop_request "tchmod 111 $DIR/f2" || return 1
47 drop_reint_reply "tchmod 666 $DIR/f2" || return 2
49 run_test 2 "chmod: drop req, drop rep"
52 drop_request "statone $DIR/f2" || return 1
53 drop_reply "statone $DIR/f2" || return 2
55 run_test 3 "stat: drop req, drop rep"
57 SAMPLE_NAME=f0.recovery-small.junk
58 SAMPLE_FILE=$TMP/$SAMPLE_NAME
59 # make this big, else test 9 doesn't wait for bulk -- bz 5595
60 dd if=/dev/urandom of=$SAMPLE_FILE bs=1M count=4
63 do_facet client "cp $SAMPLE_FILE $DIR/$SAMPLE_NAME" || return 1
64 drop_request "cat $DIR/$SAMPLE_NAME > /dev/null" || return 2
65 drop_reply "cat $DIR/$SAMPLE_NAME > /dev/null" || return 3
67 run_test 4 "open: drop req, drop rep"
69 RENAMED_AGAIN=$DIR/f0.renamed-again
72 drop_request "mv $DIR/$SAMPLE_NAME $DIR/$tfile-renamed" || return 1
73 drop_reint_reply "mv $DIR/$tfile-renamed $RENAMED_AGAIN" || return 2
74 do_facet client "checkstat -v $RENAMED_AGAIN" || return 3
76 run_test 5 "rename: drop req, drop rep"
78 [ ! -e $RENAMED_AGAIN ] && cp $SAMPLE_FILE $RENAMED_AGAIN
83 drop_request "mlink $RENAMED_AGAIN $LINK1" || return 1
84 drop_reint_reply "mlink $RENAMED_AGAIN $LINK2" || return 2
86 run_test 6 "link: drop req, drop rep"
88 [ ! -e $LINK1 ] && mlink $RENAMED_AGAIN $LINK1
89 [ ! -e $LINK2 ] && mlink $RENAMED_AGAIN $LINK2
91 drop_request "munlink $LINK1" || return 1
92 drop_reint_reply "munlink $LINK2" || return 2
94 run_test 7 "unlink: drop req, drop rep"
98 drop_reint_reply "touch $DIR/$tfile" || return 1
100 run_test 8 "touch: drop rep (bug 1423)"
104 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
106 pause_bulk "cp /etc/profile $DIR/$tfile" || return 1
107 do_facet client "cp $SAMPLE_FILE $DIR/${tfile}.2" || return 2
108 do_facet client "sync"
109 do_facet client "rm $DIR/$tfile $DIR/${tfile}.2" || return 3
111 run_test 9 "pause bulk on OST (bug 1420)"
115 do_facet client mcreate $DIR/$tfile || return 1
116 drop_bl_callback "chmod 0777 $DIR/$tfile" || echo "evicted as expected"
117 # wait for the mds to evict the client
118 #echo "sleep $(($TIMEOUT*2))"
119 #sleep $(($TIMEOUT*2))
120 do_facet client touch $DIR/$tfile || echo "touch failed, evicted"
121 do_facet client checkstat -v -p 0777 $DIR/$tfile || return 3
122 do_facet client "munlink $DIR/$tfile"
124 run_test 10 "finish request on server after client eviction (bug 1521)"
127 # wake up a thread waiting for completion after eviction
129 do_facet client multiop $DIR/$tfile Ow || return 1
130 do_facet client multiop $DIR/$tfile or || return 2
134 do_facet client multiop $DIR/$tfile or || return 3
135 drop_bl_callback multiop $DIR/$tfile Ow || echo "evicted as expected"
137 do_facet client munlink $DIR/$tfile || return 4
139 run_test 11 "wake up a thread waiting for completion after eviction (b=2460)"
143 $LCTL mark multiop $DIR/$tfile OS_c
144 do_facet $SINGLEMDS "lctl set_param fail_loc=0x115"
145 clear_failloc $SINGLEMDS $((TIMEOUT * 2)) &
146 multiop_bg_pause $DIR/$tfile OS_c || return 1
148 #define OBD_FAIL_MDS_CLOSE_NET 0x115
150 echo "waiting for multiop $PID"
151 wait $PID || return 2
152 do_facet client munlink $DIR/$tfile || return 3
154 run_test 12 "recover from timed out resend in ptlrpcd (b=2494)"
156 # Bug 113, check that readdir lost recv timeout works.
158 mkdir -p $DIR/$tdir || return 1
159 touch $DIR/$tdir/newentry || return
160 # OBD_FAIL_MDS_READPAGE_NET|OBD_FAIL_ONCE
161 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000104"
162 ls $DIR/$tdir || return 3
163 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
164 rm -rf $DIR/$tdir || return 4
166 run_test 13 "mdc_readpage restart test (bug 1138)"
168 # Bug 113, check that readdir lost send timeout works.
171 touch $DIR/$tdir/newentry
172 # OBD_FAIL_MDS_SENDPAGE|OBD_FAIL_ONCE
173 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000106"
174 ls $DIR/$tdir || return 1
175 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
177 run_test 14 "mdc_readpage resend test (bug 1138)"
180 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000128"
181 touch $DIR/$tfile && return 1
184 run_test 15 "failed open (-ENOMEM)"
186 READ_AHEAD=`lctl get_param -n llite.*.max_read_ahead_mb | head -n 1`
188 lctl set_param -n llite.*.max_read_ahead_mb 0
192 lctl set_param -n llite.*.max_read_ahead_mb $READ_AHEAD
196 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
198 do_facet client cp $SAMPLE_FILE $DIR
202 #define OBD_FAIL_PTLRPC_BULK_PUT_NET 0x504 | OBD_FAIL_ONCE
203 do_facet ost1 "lctl set_param fail_loc=0x80000504"
205 # OST bulk will time out here, client resends
206 do_facet client "cmp $SAMPLE_FILE $DIR/${SAMPLE_FILE##*/}" || return 1
207 do_facet ost1 lctl set_param fail_loc=0
208 # give recovery a chance to finish (shouldn't take long)
210 do_facet client "cmp $SAMPLE_FILE $DIR/${SAMPLE_FILE##*/}" || return 2
213 run_test 16 "timeout bulk put, don't evict client (2732)"
218 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
220 # With adaptive timeouts, bulk_get won't expire until adaptive_timeout_max
221 if at_is_enabled; then
222 at_max_saved=$(at_max_get ost1)
223 at_max_set $TIMEOUT ost1
226 # OBD_FAIL_PTLRPC_BULK_GET_NET 0x0503 | OBD_FAIL_ONCE
227 # OST bulk will time out here, client retries
228 do_facet ost1 lctl set_param fail_loc=0x80000503
229 # need to ensure we send an RPC
230 do_facet client cp /etc/termcap $DIR/$tfile
233 # with AT, client will wait adaptive_max*factor+net_latency before
234 # expiring the req, hopefully timeout*2 is enough
235 sleep $(($TIMEOUT*2))
237 do_facet ost1 lctl set_param fail_loc=0
238 do_facet client "df $DIR"
239 # expect cmp to succeed, client resent bulk
240 do_facet client "cmp /etc/termcap $DIR/$tfile" || return 3
241 do_facet client "rm $DIR/$tfile" || return 4
242 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved ost1
245 run_test 17 "timeout bulk get, don't evict client (2732)"
248 [ -z ${ost2_svc} ] && skip_env "needs 2 osts" && return 0
250 do_facet client mkdir -p $DIR/$tdir
254 pgcache_empty || return 1
257 lfs setstripe $f -s $((128 * 1024)) -i 1 -c 1
258 get_stripe_info client $f
259 if [ $stripe_index -ne 1 ]; then
261 error "$f: different stripe offset ($stripe_index)" && return
264 do_facet client cp $SAMPLE_FILE $f
266 local osc2dev=`lctl get_param -n devices | grep ${ost2_svc}-osc- | egrep -v 'MDT' | awk '{print $1}'`
267 $LCTL --device $osc2dev deactivate || return 3
268 # my understanding is that there should be nothing in the page
269 # cache after the client reconnects?
271 pgcache_empty || rc=2
272 $LCTL --device $osc2dev activate
276 run_test 18a "manual ost invalidate clears page cache immediately"
279 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
281 do_facet client mkdir -p $DIR/$tdir
285 pgcache_empty || return 1
287 # shouldn't have to set stripe size of count==1
288 lfs setstripe $f -s $((128 * 1024)) -i 0 -c 1
289 get_stripe_info client $f
290 if [ $stripe_index -ne 0 ]; then
292 error "$f: different stripe offset ($stripe_index)" && return
295 do_facet client cp $SAMPLE_FILE $f
298 # allow recovery to complete
299 sleep $((TIMEOUT + 2))
300 # my understanding is that there should be nothing in the page
301 # cache after the client reconnects?
303 pgcache_empty || rc=2
307 run_test 18b "eviction and reconnect clears page cache (2766)"
310 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
312 do_facet client mkdir -p $DIR/$tdir
316 pgcache_empty || return 1
318 # shouldn't have to set stripe size of count==1
319 lfs setstripe $f -s $((128 * 1024)) -i 0 -c 1
320 get_stripe_info client $f
321 if [ $stripe_index -ne 0 ]; then
323 error "$f: different stripe offset ($stripe_index)" && return
326 do_facet client cp $SAMPLE_FILE $f
330 # OBD_FAIL_OST_CONNECT_NET2
331 # lost reply to connect request
332 do_facet ost1 lctl set_param fail_loc=0x80000225
334 df $MOUNT > /dev/null 2>&1
336 # my understanding is that there should be nothing in the page
337 # cache after the client reconnects?
339 pgcache_empty || rc=2
343 run_test 18c "Dropped connect reply after eviction handing (14755)"
347 do_facet client mcreate $f || return 1
348 drop_ldlm_cancel "chmod 0777 $f" || echo "evicted as expected"
350 do_facet client checkstat -v -p 0777 $f || echo evicted
351 # let the client reconnect
353 do_facet client "munlink $f"
355 run_test 19a "test expired_lock_main on mds (2867)"
359 do_facet client multiop $f Ow || return 1
360 do_facet client multiop $f or || return 2
364 do_facet client multiop $f or || return 3
365 drop_ldlm_cancel multiop $f Ow || echo "client evicted, as expected"
367 do_facet client munlink $f || return 4
369 run_test 19b "test expired_lock_main on ost (2867)"
371 test_20a() { # bug 2983 - ldlm_handle_enqueue cleanup
372 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
375 lfs setstripe $DIR/$tdir/${tfile} -i 0 -c 1
376 multiop_bg_pause $DIR/$tdir/${tfile} O_wc || return 1
379 #define OBD_FAIL_LDLM_ENQUEUE_EXTENT_ERR 0x308
380 do_facet ost1 lctl set_param fail_loc=0x80000308
381 kill -USR1 $MULTI_PID
384 [ $rc -eq 0 ] && error "multiop didn't fail enqueue: rc $rc" || true
386 run_test 20a "ldlm_handle_enqueue error (should return error)"
388 test_20b() { # bug 2986 - ldlm_handle_enqueue error during open
389 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
392 lfs setstripe $DIR/$tdir/${tfile} -i 0 -c 1
394 #define OBD_FAIL_LDLM_ENQUEUE_EXTENT_ERR 0x308
395 do_facet ost1 lctl set_param fail_loc=0x80000308
396 dd if=/etc/hosts of=$DIR/$tdir/$tfile && \
397 error "didn't fail open enqueue" || true
399 run_test 20b "ldlm_handle_enqueue error (should return error)"
402 mkdir -p $DIR/$tdir-1
403 mkdir -p $DIR/$tdir-2
404 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
407 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000129"
408 multiop $DIR/$tdir-2/f Oc &
411 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
413 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
414 kill -USR1 $close_pid
416 wait $close_pid || return 1
417 wait $open_pid || return 2
418 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
420 $CHECKSTAT -t file $DIR/$tdir-1/f || return 3
421 $CHECKSTAT -t file $DIR/$tdir-2/f || return 4
425 run_test 21a "drop close request while close and open are both in flight"
428 mkdir -p $DIR/$tdir-1
429 mkdir -p $DIR/$tdir-2
430 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
433 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000107"
434 mcreate $DIR/$tdir-2/f &
437 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
439 kill -USR1 $close_pid
441 wait $close_pid || return 1
442 wait $open_pid || return 3
444 $CHECKSTAT -t file $DIR/$tdir-1/f || return 4
445 $CHECKSTAT -t file $DIR/$tdir-2/f || return 5
448 run_test 21b "drop open request while close and open are both in flight"
451 mkdir -p $DIR/$tdir-1
452 mkdir -p $DIR/$tdir-2
453 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
456 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000107"
457 mcreate $DIR/$tdir-2/f &
460 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
462 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
463 kill -USR1 $close_pid
465 wait $close_pid || return 1
466 wait $open_pid || return 2
468 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
470 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
471 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
474 run_test 21c "drop both request while close and open are both in flight"
477 mkdir -p $DIR/$tdir-1
478 mkdir -p $DIR/$tdir-2
479 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
482 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000129"
483 multiop $DIR/$tdir-2/f Oc &
485 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
487 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000122"
490 wait $pid || return 1
491 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
493 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
494 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
498 run_test 21d "drop close reply while close and open are both in flight"
501 mkdir -p $DIR/$tdir-1
502 mkdir -p $DIR/$tdir-2
503 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
506 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000119"
507 touch $DIR/$tdir-2/f &
509 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
513 wait $pid || return 1
516 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
517 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
520 run_test 21e "drop open reply while close and open are both in flight"
523 mkdir -p $DIR/$tdir-1
524 mkdir -p $DIR/$tdir-2
525 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
528 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000119"
529 touch $DIR/$tdir-2/f &
531 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
533 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000122"
536 wait $pid || return 1
537 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
539 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
540 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
543 run_test 21f "drop both reply while close and open are both in flight"
546 mkdir -p $DIR/$tdir-1
547 mkdir -p $DIR/$tdir-2
548 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
551 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000119"
552 touch $DIR/$tdir-2/f &
554 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
556 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
559 wait $pid || return 1
560 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
562 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
563 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
566 run_test 21g "drop open reply and close request while close and open are both in flight"
569 mkdir -p $DIR/$tdir-1
570 mkdir -p $DIR/$tdir-2
571 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
574 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000107"
575 touch $DIR/$tdir-2/f &
578 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
580 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000122"
583 wait $pid || return 1
584 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
586 wait $touch_pid || return 2
588 $CHECKSTAT -t file $DIR/$tdir-1/f || return 3
589 $CHECKSTAT -t file $DIR/$tdir-2/f || return 4
592 run_test 21h "drop open request and close reply while close and open are both in flight"
594 # bug 3462 - multiple MDC requests
599 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
604 multiop $f1 msu || return 1
607 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
609 wait $close_pid || return 2
610 rm -rf $f2 || return 4
612 run_test 22 "drop close request and do mknod"
615 multiop_bg_pause $DIR/$tfile O_c || return 1
617 # give a chance for open
621 drop_request "kill -USR1 $pid"
624 wait $pid || return 1
627 run_test 23 "client hang when close a file after mds crash"
629 test_24() { # bug 11710 details correct fsync() behavior
630 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
633 lfs setstripe $DIR/$tdir -s 0 -i 0 -c 1
635 multiop_bg_pause $DIR/$tdir/$tfile Owy_wyc || return 1
638 kill -USR1 $MULTI_PID
641 lctl set_param fail_loc=0x0
643 [ $rc -eq 0 ] && error_ignore 5494 "multiop didn't fail fsync: rc $rc" || true
645 run_test 24 "fsync error (should return error)"
647 wait_client_evicted () {
650 local varsvc=${facet}_svc
652 wait_update $(facet_host $facet) "lctl get_param -n *.${!varsvc}.num_exports | cut -d' ' -f2" $((exports - 1)) $3
655 test_26a() { # was test_26 bug 5921 - evict dead exports by pinger
656 # this test can only run from a client on a separate node.
657 remote_ost || { skip "local OST" && return 0; }
658 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
659 remote_mds || { skip "local MDS" && return 0; }
661 if [ $(facet_host mgs) = $(facet_host ost1) ]; then
662 skip "msg and ost1 are at the same node"
666 check_timeout || return 1
668 local OST_NEXP=$(do_facet ost1 lctl get_param -n obdfilter.${ost1_svc}.num_exports | cut -d' ' -f2)
670 echo starting with $OST_NEXP OST exports
671 # OBD_FAIL_PTLRPC_DROP_RPC 0x505
672 do_facet client lctl set_param fail_loc=0x505
673 # evictor takes PING_EVICT_TIMEOUT + 3 * PING_INTERVAL to evict.
674 # But if there's a race to start the evictor from various obds,
675 # the loser might have to wait for the next ping.
678 wait_client_evicted ost1 $OST_NEXP $((TIMEOUT * 2 + TIMEOUT * 3 / 4))
680 do_facet client lctl set_param fail_loc=0x0
681 [ $rc -eq 0 ] || error "client not evicted from OST"
683 run_test 26a "evict dead exports"
685 test_26b() { # bug 10140 - evict dead exports by pinger
686 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
688 if [ $(facet_host mgs) = $(facet_host ost1) ]; then
689 skip "msg and ost1 are at the same node"
693 check_timeout || return 1
695 zconf_mount `hostname` $MOUNT2 ||
696 { error "Failed to mount $MOUNT2"; return 2; }
697 sleep 1 # wait connections being established
699 local MDS_NEXP=$(do_facet $SINGLEMDS lctl get_param -n mdt.${mds1_svc}.num_exports | cut -d' ' -f2)
700 local OST_NEXP=$(do_facet ost1 lctl get_param -n obdfilter.${ost1_svc}.num_exports | cut -d' ' -f2)
702 echo starting with $OST_NEXP OST and $MDS_NEXP MDS exports
704 zconf_umount `hostname` $MOUNT2 -f
706 # PING_INTERVAL max(obd_timeout / 4, 1U)
707 # PING_EVICT_TIMEOUT (PING_INTERVAL * 6)
709 # evictor takes PING_EVICT_TIMEOUT + 3 * PING_INTERVAL to evict.
710 # But if there's a race to start the evictor from various obds,
711 # the loser might have to wait for the next ping.
712 # = 9 * PING_INTERVAL + PING_INTERVAL
713 # = 10 PING_INTERVAL = 10 obd_timeout / 4 = 2.5 obd_timeout
714 # let's wait $((TIMEOUT * 3)) # bug 19887
716 wait_client_evicted ost1 $OST_NEXP $((TIMEOUT * 3)) || \
717 error "Client was not evicted by ost" rc=1
718 wait_client_evicted $SINGLEMDS $MDS_NEXP $((TIMEOUT * 3)) || \
719 error "Client was not evicted by mds"
721 run_test 26b "evict dead exports"
725 writemany -q -a $DIR/$tdir/$tfile 0 5 &
728 local save_FAILURE_MODE=$FAILURE_MODE
730 facet_failover $SINGLEMDS
731 #define OBD_FAIL_OSC_SHUTDOWN 0x407
732 do_facet $SINGLEMDS lctl set_param fail_loc=0x80000407
733 # need to wait for reconnect
734 echo -n waiting for fail_loc
735 while [ $(do_facet $SINGLEMDS lctl get_param -n fail_loc) -eq -2147482617 ]; do
739 do_facet $SINGLEMDS lctl get_param -n fail_loc
740 facet_failover $SINGLEMDS
742 kill -USR1 $CLIENT_PID
745 FAILURE_MODE=$save_FAILURE_MODE
747 run_test 27 "fail LOV while using OSC's"
749 test_28() { # bug 6086 - error adding new clients
750 do_facet client mcreate $DIR/$tfile || return 1
751 drop_bl_callback "chmod 0777 $DIR/$tfile" ||echo "evicted as expected"
752 #define OBD_FAIL_MDS_CLIENT_ADD 0x12f
753 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000012f"
754 # fail once (evicted), reconnect fail (fail_loc), ok
755 client_up || (sleep 10; client_up) || (sleep 10; client_up) || error "reconnect failed"
757 fail $SINGLEMDS # verify MDS last_rcvd can be loaded
759 run_test 28 "handle error adding new clients (bug 6086)"
763 # put a load of file creates/writes/deletes
764 writemany -q $DIR/$tdir/$tfile 0 5 &
766 echo writemany pid $CLIENT_PID
770 # wait for client to reconnect to MDS
775 # client process should see no problems even though MDS went down
777 kill -USR1 $CLIENT_PID
780 echo writemany returned $rc
781 #these may fail because of eviction due to slow AST response.
782 [ $rc -eq 0 ] || error_ignore 13652 "writemany returned rc $rc" || true
784 run_test 50 "failover MDS under load"
788 # put a load of file creates/writes/deletes
789 writemany -q $DIR/$tdir/$tfile 0 5 &
793 facet_failover $SINGLEMDS
794 # failover at various points during recovery
795 SEQ="1 5 10 $(seq $TIMEOUT 5 $(($TIMEOUT+10)))"
796 echo will failover at $SEQ
799 echo failover in $i sec
801 facet_failover $SINGLEMDS
803 # client process should see no problems even though MDS went down
804 # and recovery was interrupted
806 kill -USR1 $CLIENT_PID
809 echo writemany returned $rc
810 [ $rc -eq 0 ] || error_ignore 13652 "writemany returned rc $rc" || true
812 run_test 51 "failover MDS during recovery"
815 do_facet client "mkdir -p $DIR/$tdir"
816 do_facet client "writemany -q -a $DIR/$tdir/$tfile 300 5" &
818 echo writemany pid $CLIENT_PID
823 wait $CLIENT_PID || rc=$?
824 # active client process should see an EIO for down OST
825 [ $rc -eq 5 ] && { echo "writemany correctly failed $rc" && return 0; }
826 # but timing or failover setup may allow success
827 [ $rc -eq 0 ] && { echo "writemany succeeded" && return 0; }
828 echo "writemany returned $rc"
833 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
838 [ $rc -ne 0 ] && { return $rc; }
839 # wait for client to reconnect to OST
843 [ $rc -ne 0 ] && { return $rc; }
850 run_test 52 "failover OST under load"
852 # test of open reconstruct
855 drop_ldlm_reply "openfile -f O_RDWR:O_CREAT -m 0755 $DIR/$tfile" ||\
858 run_test 53 "touch: drop rep"
861 zconf_mount `hostname` $MOUNT2
865 cat $DIR2/$tfile.missing # save transno = 0, rc != 0 into last_rcvd
868 ERROR=`dmesg | egrep "(test 54|went back in time)" | tail -n1 | grep "went back in time"`
869 [ x"$ERROR" == x ] || error "back in time occured"
871 run_test 54 "back in time"
873 # bug 11330 - liblustre application death during I/O locks up OST
875 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
879 # first dd should be finished quickly
880 lfs setstripe $DIR/$tdir/$tfile-1 -c 1 -i 0
881 dd if=/dev/zero of=$DIR/$tdir/$tfile-1 bs=32M count=4 &
884 echo "step1: testing ......"
886 if [ -z `ps x | awk '$1 == '$DDPID' { print $5 }'` ]; then break; fi
888 if [ $count -gt 64 ]; then
889 error "dd should be finished!"
893 echo "(dd_pid=$DDPID, time=$count)successful"
895 lfs setstripe $DIR/$tdir/$tfile-2 -c 1 -i 0
896 #define OBD_FAIL_OST_DROP_REQ 0x21d
897 do_facet ost1 lctl set_param fail_loc=0x0000021d
898 # second dd will be never finished
899 dd if=/dev/zero of=$DIR/$tdir/$tfile-2 bs=32M count=4 &
902 echo "step2: testing ......"
903 while [ $count -le 64 ]; do
904 dd_name="`ps x | awk '$1 == '$DDPID' { print $5 }'`"
905 if [ -z $dd_name ]; then
907 echo "debug: (dd_name=$dd_name, dd_pid=$DDPID, time=$count)"
908 error "dd shouldn't be finished!"
913 echo "(dd_pid=$DDPID, time=$count)successful"
915 #Recover fail_loc and dd will finish soon
916 do_facet ost1 lctl set_param fail_loc=0
918 echo "step3: testing ......"
920 if [ -z `ps x | awk '$1 == '$DDPID' { print $5 }'` ]; then break; fi
922 if [ $count -gt 500 ]; then
923 error "dd should be finished!"
927 echo "(dd_pid=$DDPID, time=$count)successful"
931 run_test 55 "ost_brw_read/write drops timed-out read/write request"
933 test_56() { # b=11277
934 #define OBD_FAIL_MDS_RESEND 0x136
936 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000136"
938 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
941 run_test 56 "do not allow reconnect to busy exports"
944 # no oscs means no client or mdt
945 while lctl get_param osc.*.* > /dev/null 2>&1; do
946 : # loop until proc file is removed
950 test_57() { # bug 10866
954 #define OBD_FAIL_LPROC_REMOVE 0xB00
955 lctl set_param fail_loc=0x80000B00
956 zconf_umount `hostname` $DIR
957 lctl set_param fail_loc=0x80000B00
958 fail_abort $SINGLEMDS
960 lctl set_param fail_loc=0
962 do_facet client "df $DIR"
964 run_test 57 "read procfs entries causes kernel crash"
966 test_58() { # bug 11546
967 #define OBD_FAIL_MDC_ENQUEUE_PAUSE 0x801
970 lctl set_param fail_loc=0x80000801
971 cp $DIR/$tfile /dev/null &
974 lctl set_param fail_loc=0
975 drop_bl_callback rm -f $DIR/$tfile
977 do_facet client "df $DIR"
979 run_test 58 "Eviction in the middle of open RPC reply processing"
981 test_59() { # bug 10589
982 zconf_mount `hostname` $MOUNT2 || error "Failed to mount $MOUNT2"
983 echo $DIR2 | grep -q $MOUNT2 || error "DIR2 is not set properly: $DIR2"
984 #define OBD_FAIL_LDLM_CANCEL_EVICT_RACE 0x311
985 lctl set_param fail_loc=0x311
986 writes=$(LANG=C dd if=/dev/zero of=$DIR2/$tfile count=1 2>&1)
987 [ $? = 0 ] || error "dd write failed"
988 writes=$(echo $writes | awk -F '+' '/out/ {print $1}')
989 lctl set_param fail_loc=0
991 zconf_umount `hostname` $MOUNT2 -f
992 reads=$(LANG=C dd if=$DIR/$tfile of=/dev/null 2>&1)
993 [ $? = 0 ] || error "dd read failed"
994 reads=$(echo $reads | awk -F '+' '/in/ {print $1}')
995 [ "$reads" -eq "$writes" ] || error "read" $reads "blocks, must be" $writes
997 run_test 59 "Read cancel race on client eviction"
1000 # we assume that all md changes are in the MDT0 changelog
1001 if [ $MDSCOUNT -gt 1 ]; then
1002 error_ignore 17935 $*
1009 MDT0=$($LCTL get_param -n mdc.*.mds_server_uuid | \
1010 awk '{gsub(/_UUID/,""); print $1}' | head -1)
1015 # Register (and start) changelog
1016 USER=$(do_facet $SINGLEMDS lctl --device $MDT0 changelog_register -n)
1017 echo "Registered as $MDT0 changelog user $USER"
1019 # Generate a large number of changelog entries
1020 createmany -o $DIR/$tdir/$tfile $NUM_FILES
1024 # Unlink files in the background
1025 unlinkmany $DIR/$tdir/$tfile $NUM_FILES &
1029 # Failover the MDS while unlinks are happening
1030 facet_failover $SINGLEMDS
1032 # Wait for unlinkmany to finish
1035 # Check if all the create/unlink events were recorded
1037 $LFS changelog $MDT0 >> $DIR/$tdir/changelog
1038 local cl_count=$(grep UNLNK $DIR/$tdir/changelog | wc -l)
1039 echo "$cl_count unlinks in $MDT0 changelog"
1041 do_facet $SINGLEMDS lctl --device $MDT0 changelog_deregister $USER
1042 USERS=$(( $(do_facet $SINGLEMDS lctl get_param -n \
1043 mdd.$MDT0.changelog_users | wc -l) - 2 ))
1044 if [ $USERS -eq 0 ]; then
1045 [ $cl_count -eq $NUM_FILES ] || \
1046 err17935 "Recorded ${cl_count} unlinks out of $NUM_FILES"
1047 # Also make sure we can clear large changelogs
1048 cl_count=$($LFS changelog $FSNAME | wc -l)
1049 [ $cl_count -le 2 ] || \
1050 error "Changelog not empty: $cl_count entries"
1052 # If there are other users, there may be other unlinks in the log
1053 [ $cl_count -ge $NUM_FILES ] || \
1054 err17935 "Recorded ${cl_count} unlinks out of $NUM_FILES"
1055 echo "$USERS other changelog users; can't verify clear"
1058 run_test 60 "Add Changelog entries during MDS failover"
1062 local cflags='osc.*-OST0000-osc-MDT*.connect_flags'
1063 do_facet $SINGLEMDS "lctl get_param -n $cflags" |grep -q skip_orphan
1064 [ $? -ne 0 ] && skip "don't have skip orphan feature" && return
1066 mkdir -p $DIR/$tdir || error "mkdir dir $DIR/$tdir failed"
1067 # Set the default stripe of $DIR/$tdir to put the files to ost1
1068 $LFS setstripe -c 1 --index 0 $DIR/$tdir
1070 replay_barrier $SINGLEMDS
1071 createmany -o $DIR/$tdir/$tfile-%d 10
1072 local oid=`do_facet ost1 "lctl get_param -n obdfilter.*OST0000.last_id"`
1074 fail_abort $SINGLEMDS
1076 touch $DIR/$tdir/$tfile
1077 local id=`$LFS getstripe $DIR/$tdir/$tfile |awk '$2 ~ /^[1-9]+/ {print $2}'`
1078 [ $id -le $oid ] && error "the orphan objid was reused, failed"
1083 run_test 61 "Verify to not reuse orphan objects - bug 17025"
1085 equals_msg `basename $0`: test complete, cleaning up
1086 check_and_cleanup_lustre
1087 [ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG && grep -q FAIL $TESTSUITELOG && exit 1 || true