6 ALWAYS_EXCEPT="52 60 $RECOVERY_SMALL_EXCEPT"
8 export MULTIOP=${MULTIOP:-multiop}
9 PTLDEBUG=${PTLDEBUG:--1}
10 LUSTRE=${LUSTRE:-`dirname $0`/..}
11 . $LUSTRE/tests/test-framework.sh
13 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
16 require_dsh_mds || exit 0
18 # also long tests: 19, 21a, 21e, 21f, 23, 27
19 # 1 2.5 2.5 4 4 (min)"
20 [ "$SLOW" = "no" ] && EXCEPT_SLOW="17 26a 26b 50 51 57"
24 # Allow us to override the setup if we already have a mounted system by
25 # setting SETUP=" " and CLEANUP=" "
27 CLEANUP=${CLEANUP:-""}
29 check_and_setup_lustre
32 rm -rf $DIR/d[0-9]* $DIR/f.${TESTSUITE}*
35 local f1="$DIR/$tfile"
36 local f2="$DIR/$tfile.2"
38 drop_request "mcreate $f1" ||
39 error_noexit "create '$f1': drop req"
41 drop_reint_reply "mcreate $f2" ||
42 error_noexit "create '$f2': drop rep"
44 drop_request "tchmod 111 $f2" ||
45 error_noexit "chmod '$f2': drop req"
47 drop_reint_reply "tchmod 666 $f2" ||
48 error_noexit "chmod '$f2': drop rep"
50 drop_request "statone $f2" ||
51 error_noexit "stat '$f2': drop req"
53 drop_reply "statone $f2" ||
54 error_noexit "stat '$f2': drop rep"
56 run_test 1 "create, chmod, stat: drop req, drop rep"
60 do_facet_create_file client $t 10K ||
61 error_noexit "Create file $t"
63 drop_request "cat $t > /dev/null" ||
64 error_noexit "Open request for $t file"
66 drop_reply "cat $t > /dev/null" ||
67 error_noexit "Open replay for $t file"
69 run_test 4 "open: drop req, drop rep"
74 local RR="$T-renamed-again"
75 do_facet_create_file client $T 10K ||
76 error_noexit "Create file $T"
78 drop_request "mv $T $R" ||
79 error_noexit "Rename $T"
81 drop_reint_reply "mv $R $RR" ||
82 error_noexit "Failed rename replay on $R"
84 do_facet client "checkstat -v $RR" ||
85 error_noexit "checkstat error on $RR"
87 do_facet client "rm $RR" ||
88 error_noexit "Can't remove file $RR"
90 run_test 5 "rename: drop req, drop rep"
94 local LINK1=$DIR/$tfile.link1
95 local LINK2=$DIR/$tfile.link2
97 do_facet_create_file client $T 10K ||
98 error_noexit "Create file $T"
100 drop_request "mlink $T $LINK1" ||
101 error_noexit "mlink request for $T"
103 drop_reint_reply "mlink $T $LINK2" ||
104 error_noexit "mlink reply for $T"
106 drop_request "munlink $LINK1" ||
107 error_noexit "munlink request for $T"
109 drop_reint_reply "munlink $LINK2" ||
110 error_noexit "munlink reply for $T"
112 do_facet client "rm $T" ||
113 error_noexit "Can't remove file $T"
115 run_test 6 "link, unlink: drop req, drop rep"
119 drop_reint_reply "touch $DIR/$tfile" || return 1
121 run_test 8 "touch: drop rep (bug 1423)"
125 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
129 do_facet_random_file client $TMP/$tfile 1K ||
130 error_noexit "Create random file $TMP/$tfile"
131 # make this big, else test 9 doesn't wait for bulk -- bz 5595
132 do_facet_create_file client $TMP/$t1 4M ||
133 error_noexit "Create file $TMP/$t1"
134 do_facet client "cp $TMP/$t1 $DIR/$t1" ||
135 error_noexit "Can't copy to $DIR/$t1 file"
136 pause_bulk "cp $TMP/$tfile $DIR/$tfile" ||
137 error_noexit "Can't pause_bulk copy"
138 do_facet client "cp $TMP/$t1 $DIR/$t2" ||
139 error_noexit "Can't copy file"
140 do_facet client "sync"
141 do_facet client "rm $DIR/$tfile $DIR/$t2 $DIR/$t1" ||
142 error_noexit "Can't remove files"
143 do_facet client "rm $TMP/$t1 $TMP/$tfile"
145 run_test 9 "pause bulk on OST (bug 1420)"
149 do_facet client mcreate $DIR/$tfile ||
150 { error "mcreate failed: $?"; return 1; }
151 drop_bl_callback "chmod 0777 $DIR/$tfile" || echo "evicted as expected"
152 # wait for the mds to evict the client
153 #echo "sleep $(($TIMEOUT*2))"
154 #sleep $(($TIMEOUT*2))
155 do_facet client touch $DIR/$tfile || echo "touch failed, evicted"
156 do_facet client checkstat -v -p 0777 $DIR/$tfile ||
157 { error "client checkstat failed: $?"; return 3; }
158 do_facet client "munlink $DIR/$tfile"
159 # allow recovery to complete
160 client_up || client_up || sleep $TIMEOUT
162 run_test 10 "finish request on server after client eviction (bug 1521)"
165 # wake up a thread waiting for completion after eviction
167 do_facet client $MULTIOP $DIR/$tfile Ow ||
168 { error "multiop write failed: $?"; return 1; }
169 do_facet client $MULTIOP $DIR/$tfile or ||
170 { error "multiop read failed: $?"; return 2; }
174 do_facet client $MULTIOP $DIR/$tfile or ||
175 { error "multiop read failed: $?"; return 3; }
176 drop_bl_callback $MULTIOP $DIR/$tfile Ow || echo "evicted as expected"
178 do_facet client munlink $DIR/$tfile ||
179 { error "munlink failed: $?"; return 4; }
180 # allow recovery to complete
181 client_up || client_up || sleep $TIMEOUT
183 run_test 11 "wake up a thread waiting for completion after eviction (b=2460)"
187 $LCTL mark $MULTIOP $DIR/$tfile OS_c
188 do_facet $SINGLEMDS "lctl set_param fail_loc=0x115"
189 clear_failloc $SINGLEMDS $((TIMEOUT * 2)) &
190 multiop_bg_pause $DIR/$tfile OS_c ||
191 { error "multiop failed: $?"; return 1; }
193 #define OBD_FAIL_MDS_CLOSE_NET 0x115
195 echo "waiting for multiop $PID"
196 wait $PID || { error "wait for multiop faile: $?"; return 2; }
197 do_facet client munlink $DIR/$tfile ||
198 { error "client munlink failed: $?"; return 3; }
199 # allow recovery to complete
200 client_up || client_up || sleep $TIMEOUT
202 run_test 12 "recover from timed out resend in ptlrpcd (b=2494)"
204 # Bug 113, check that readdir lost recv timeout works.
206 mkdir -p $DIR/$tdir || { error "mkdir failed: $?"; return 1; }
207 touch $DIR/$tdir/newentry || { error "touch failed: $?"; return 2; }
208 # OBD_FAIL_MDS_READPAGE_NET|OBD_FAIL_ONCE
209 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000104"
210 ls $DIR/$tdir || { error "ls failed: $?"; return 3; }
211 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
212 rm -rf $DIR/$tdir || { error "remove test dir failed: $?"; return 4; }
214 run_test 13 "mdc_readpage restart test (bug 1138)"
216 # Bug 113, check that readdir lost send timeout works.
219 touch $DIR/$tdir/newentry
220 # OBD_FAIL_MDS_SENDPAGE|OBD_FAIL_ONCE
221 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000106"
222 ls $DIR/$tdir || return 1
223 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
225 run_test 14 "mdc_readpage resend test (bug 1138)"
228 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000128"
229 touch $DIR/$tfile && return 1
232 run_test 15 "failed open (-ENOMEM)"
234 READ_AHEAD=`lctl get_param -n llite.*.max_read_ahead_mb | head -n 1`
236 lctl set_param -n llite.*.max_read_ahead_mb 0
240 lctl set_param -n llite.*.max_read_ahead_mb $READ_AHEAD
244 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
246 do_facet_random_file client $TMP/$tfile 100K ||
247 { error_noexit "Create random file $TMP/$T" ; return 0; }
248 do_facet client "cp $TMP/$tfile $DIR/$tfile" ||
249 { error_noexit "Copy to $DIR/$tfile file" ; return 0; }
253 #define OBD_FAIL_PTLRPC_BULK_PUT_NET 0x504 | OBD_FAIL_ONCE
254 do_facet ost1 "lctl set_param fail_loc=0x80000504"
256 # OST bulk will time out here, client resends
257 do_facet client "cmp $TMP/$tfile $DIR/$tfile" || return 1
258 do_facet ost1 lctl set_param fail_loc=0
259 # give recovery a chance to finish (shouldn't take long)
261 do_facet client "cmp $TMP/$tfile $DIR/$tfile" || return 2
264 run_test 16 "timeout bulk put, don't evict client (2732)"
269 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
271 local SAMPLE_FILE=$TMP/$tfile
272 do_facet_random_file client $SAMPLE_FILE 20K ||
273 { error_noexit "Create random file $SAMPLE_FILE" ; return 0; }
275 # With adaptive timeouts, bulk_get won't expire until adaptive_timeout_max
276 if at_is_enabled; then
277 at_max_saved=$(at_max_get ost1)
278 at_max_set $TIMEOUT ost1
281 # OBD_FAIL_PTLRPC_BULK_GET_NET 0x0503 | OBD_FAIL_ONCE
282 # OST bulk will time out here, client retries
283 do_facet ost1 lctl set_param fail_loc=0x80000503
284 # need to ensure we send an RPC
285 do_facet client cp $SAMPLE_FILE $DIR/$tfile
288 # with AT, client will wait adaptive_max*factor+net_latency before
289 # expiring the req, hopefully timeout*2 is enough
290 sleep $(($TIMEOUT*2))
292 do_facet ost1 lctl set_param fail_loc=0
293 do_facet client "df $DIR"
294 # expect cmp to succeed, client resent bulk
295 do_facet client "cmp $SAMPLE_FILE $DIR/$tfile" || return 3
296 do_facet client "rm $DIR/$tfile" || return 4
297 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved ost1
300 run_test 17 "timeout bulk get, don't evict client (2732)"
303 [ -z ${ost2_svc} ] && skip_env "needs 2 osts" && return 0
305 do_facet_create_file client $TMP/$tfile 20K ||
306 { error_noexit "Create file $TMP/$tfile" ; return 0; }
308 do_facet client mkdir -p $DIR/$tdir
312 pgcache_empty || return 1
315 $LFS setstripe -i 1 -c 1 $f
316 stripe_index=$($LFS getstripe -i $f)
317 if [ $stripe_index -ne 1 ]; then
319 error "$f: stripe_index $stripe_index != 1" && return
322 do_facet client cp $TMP/$tfile $f
324 local osc2dev=`lctl get_param -n devices | grep ${ost2_svc}-osc- | egrep -v 'MDT' | awk '{print $1}'`
325 $LCTL --device $osc2dev deactivate || return 3
326 # my understanding is that there should be nothing in the page
327 # cache after the client reconnects?
329 pgcache_empty || rc=2
330 $LCTL --device $osc2dev activate
334 run_test 18a "manual ost invalidate clears page cache immediately"
337 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
339 do_facet_create_file client $TMP/$tfile 20K ||
340 { error_noexit "Create file $TMP/$tfile" ; return 0; }
342 do_facet client mkdir -p $DIR/$tdir
346 pgcache_empty || return 1
348 $LFS setstripe -i 0 -c 1 $f
349 stripe_index=$($LFS getstripe -i $f)
350 if [ $stripe_index -ne 0 ]; then
352 error "$f: stripe_index $stripe_index != 0" && return
355 do_facet client cp $TMP/$tfile $f
358 # allow recovery to complete
359 sleep $((TIMEOUT + 2))
360 # my understanding is that there should be nothing in the page
361 # cache after the client reconnects?
363 pgcache_empty || rc=2
367 run_test 18b "eviction and reconnect clears page cache (2766)"
370 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
372 do_facet_create_file client $TMP/$tfile 20K ||
373 { error_noexit "Create file $TMP/$tfile" ; return 0; }
375 do_facet client mkdir -p $DIR/$tdir
379 pgcache_empty || return 1
381 $LFS setstripe -i 0 -c 1 $f
382 stripe_index=$($LFS getstripe -i $f)
383 if [ $stripe_index -ne 0 ]; then
385 error "$f: stripe_index $stripe_index != 0" && return
388 do_facet client cp $TMP/$tfile $f
392 # OBD_FAIL_OST_CONNECT_NET2
393 # lost reply to connect request
394 do_facet ost1 lctl set_param fail_loc=0x80000225
397 df $MOUNT > /dev/null 2>&1
399 # my understanding is that there should be nothing in the page
400 # cache after the client reconnects?
402 pgcache_empty || rc=2
406 run_test 18c "Dropped connect reply after eviction handing (14755)"
409 local BEFORE=`date +%s`
414 do_facet client mcreate $DIR/$tfile || return 1
415 drop_ldlm_cancel "chmod 0777 $DIR2"
418 do_facet client "munlink $DIR/$tfile"
420 # let the client reconnect
422 EVICT=$(do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state | \
423 awk -F"[ [,]" '/EVICTED]$/ { if (mx<$4) {mx=$4;} } END { print mx }')
425 [ ! -z "$EVICT" ] && [[ $EVICT -gt $BEFORE ]] || error "no eviction"
427 run_test 19a "test expired_lock_main on mds (2867)"
430 local BEFORE=`date +%s`
435 do_facet client $MULTIOP $DIR/$tfile Ow || return 1
436 drop_ldlm_cancel $MULTIOP $DIR2/$tfile Ow
438 do_facet client munlink $DIR/$tfile
440 # let the client reconnect
442 EVICT=$(do_facet client $LCTL get_param osc.$FSNAME-OST*.state | \
443 awk -F"[ [,]" '/EVICTED]$/ { if (mx<$4) {mx=$4;} } END { print mx }')
445 [ ! -z "$EVICT" ] && [[ $EVICT -gt $BEFORE ]] || error "no eviction"
447 run_test 19b "test expired_lock_main on ost (2867)"
449 test_20a() { # bug 2983 - ldlm_handle_enqueue cleanup
450 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
453 $LFS setstripe -i 0 -c 1 $DIR/$tdir/${tfile}
454 multiop_bg_pause $DIR/$tdir/${tfile} O_wc || return 1
457 #define OBD_FAIL_LDLM_ENQUEUE_EXTENT_ERR 0x308
458 do_facet ost1 lctl set_param fail_loc=0x80000308
459 kill -USR1 $MULTI_PID
462 [ $rc -eq 0 ] && error "multiop didn't fail enqueue: rc $rc" || true
464 run_test 20a "ldlm_handle_enqueue error (should return error)"
466 test_20b() { # bug 2986 - ldlm_handle_enqueue error during open
467 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
470 $LFS setstripe -i 0 -c 1 $DIR/$tdir/${tfile}
472 #define OBD_FAIL_LDLM_ENQUEUE_EXTENT_ERR 0x308
473 do_facet ost1 lctl set_param fail_loc=0x80000308
474 dd if=/etc/hosts of=$DIR/$tdir/$tfile && \
475 error "didn't fail open enqueue" || true
477 run_test 20b "ldlm_handle_enqueue error (should return error)"
480 mkdir -p $DIR/$tdir-1
481 mkdir -p $DIR/$tdir-2
482 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
485 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000129"
486 $MULTIOP $DIR/$tdir-2/f Oc &
489 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
491 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
492 kill -USR1 $close_pid
494 wait $close_pid || return 1
495 wait $open_pid || return 2
496 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
498 $CHECKSTAT -t file $DIR/$tdir-1/f || return 3
499 $CHECKSTAT -t file $DIR/$tdir-2/f || return 4
503 run_test 21a "drop close request while close and open are both in flight"
506 mkdir -p $DIR/$tdir-1
507 mkdir -p $DIR/$tdir-2
508 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
511 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000107"
512 mcreate $DIR/$tdir-2/f &
515 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
517 kill -USR1 $close_pid
519 wait $close_pid || return 1
520 wait $open_pid || return 3
522 $CHECKSTAT -t file $DIR/$tdir-1/f || return 4
523 $CHECKSTAT -t file $DIR/$tdir-2/f || return 5
526 run_test 21b "drop open request while close and open are both in flight"
529 mkdir -p $DIR/$tdir-1
530 mkdir -p $DIR/$tdir-2
531 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
534 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000107"
535 mcreate $DIR/$tdir-2/f &
538 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
540 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
541 kill -USR1 $close_pid
543 wait $close_pid || return 1
544 wait $open_pid || return 2
546 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
548 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
549 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
552 run_test 21c "drop both request while close and open are both in flight"
555 mkdir -p $DIR/$tdir-1
556 mkdir -p $DIR/$tdir-2
557 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
560 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000129"
561 $MULTIOP $DIR/$tdir-2/f Oc &
563 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
565 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000122"
568 wait $pid || return 1
569 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
571 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
572 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
576 run_test 21d "drop close reply while close and open are both in flight"
579 mkdir -p $DIR/$tdir-1
580 mkdir -p $DIR/$tdir-2
581 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
584 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000119"
585 touch $DIR/$tdir-2/f &
587 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
591 wait $pid || return 1
594 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
595 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
598 run_test 21e "drop open reply while close and open are both in flight"
601 mkdir -p $DIR/$tdir-1
602 mkdir -p $DIR/$tdir-2
603 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
606 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000119"
607 touch $DIR/$tdir-2/f &
609 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
611 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000122"
614 wait $pid || return 1
615 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
617 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
618 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
621 run_test 21f "drop both reply while close and open are both in flight"
624 mkdir -p $DIR/$tdir-1
625 mkdir -p $DIR/$tdir-2
626 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
629 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000119"
630 touch $DIR/$tdir-2/f &
632 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
634 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
637 wait $pid || return 1
638 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
640 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
641 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
644 run_test 21g "drop open reply and close request while close and open are both in flight"
647 mkdir -p $DIR/$tdir-1
648 mkdir -p $DIR/$tdir-2
649 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
652 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000107"
653 touch $DIR/$tdir-2/f &
656 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
658 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000122"
661 wait $pid || return 1
662 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
664 wait $touch_pid || return 2
666 $CHECKSTAT -t file $DIR/$tdir-1/f || return 3
667 $CHECKSTAT -t file $DIR/$tdir-2/f || return 4
670 run_test 21h "drop open request and close reply while close and open are both in flight"
672 # bug 3462 - multiple MDC requests
677 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
682 $MULTIOP $f1 msu || return 1
685 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
687 wait $close_pid || return 2
688 rm -rf $f2 || return 4
690 run_test 22 "drop close request and do mknod"
693 multiop_bg_pause $DIR/$tfile O_c || return 1
695 # give a chance for open
699 drop_request "kill -USR1 $pid"
702 wait $pid || return 1
705 run_test 23 "client hang when close a file after mds crash"
707 test_24a() { # bug 11710 details correct fsync() behavior
708 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
711 $LFS setstripe -i 0 -c 1 $DIR/$tdir
713 multiop_bg_pause $DIR/$tdir/$tfile Owy_wyc || return 1
716 kill -USR1 $MULTI_PID
719 lctl set_param fail_loc=0x0
721 [ $rc -eq 0 ] && error_ignore 5494 "multiop didn't fail fsync: rc $rc" || true
723 run_test 24a "fsync error (should return error)"
725 wait_client_evicted () {
728 local varsvc=${facet}_svc
730 wait_update $(facet_active_host $facet) \
731 "lctl get_param -n *.${!varsvc}.num_exports | cut -d' ' -f2" \
736 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
740 lfs setstripe $DIR/$tdir -s 0 -i 0 -c 1
742 multiop_bg_pause $DIR/$tdir/$tfile-1 Ow8192_yc ||
743 error "mulitop Ow8192_yc failed"
746 multiop_bg_pause $DIR/$tdir/$tfile-2 Ow8192_c ||
747 error "mulitop Ow8192_c failed"
752 kill -USR1 $MULTI_PID1
755 kill -USR1 $MULTI_PID2
758 lctl set_param fail_loc=0x0
760 [ $rc1 -eq 0 -o $rc2 -eq 0 ] &&
761 error_ignore "multiop didn't fail fsync: $rc1 or close: $rc2" || true
763 dmesg | grep "dirty page discard:" || \
764 error "no discarded dirty page found!"
766 run_test 24b "test dirty page discard due to client eviction"
768 test_26a() { # was test_26 bug 5921 - evict dead exports by pinger
769 # this test can only run from a client on a separate node.
770 remote_ost || { skip "local OST" && return 0; }
771 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
772 remote_mds || { skip "local MDS" && return 0; }
774 if [ $(facet_host mgs) = $(facet_host ost1) ]; then
775 skip "msg and ost1 are at the same node"
779 check_timeout || return 1
781 local OST_NEXP=$(do_facet ost1 lctl get_param -n obdfilter.${ost1_svc}.num_exports | cut -d' ' -f2)
783 echo starting with $OST_NEXP OST exports
784 # OBD_FAIL_PTLRPC_DROP_RPC 0x505
785 do_facet client lctl set_param fail_loc=0x505
786 # evictor takes PING_EVICT_TIMEOUT + 3 * PING_INTERVAL to evict.
787 # But if there's a race to start the evictor from various obds,
788 # the loser might have to wait for the next ping.
791 wait_client_evicted ost1 $OST_NEXP $((TIMEOUT * 2 + TIMEOUT * 3 / 4))
793 do_facet client lctl set_param fail_loc=0x0
794 [ $rc -eq 0 ] || error "client not evicted from OST"
796 run_test 26a "evict dead exports"
798 test_26b() { # bug 10140 - evict dead exports by pinger
799 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
801 if [ $(facet_host mgs) = $(facet_host ost1) ]; then
802 skip "msg and ost1 are at the same node"
806 check_timeout || return 1
808 zconf_mount `hostname` $MOUNT2 ||
809 { error "Failed to mount $MOUNT2"; return 2; }
810 sleep 1 # wait connections being established
812 local MDS_NEXP=$(do_facet $SINGLEMDS lctl get_param -n mdt.${mds1_svc}.num_exports | cut -d' ' -f2)
813 local OST_NEXP=$(do_facet ost1 lctl get_param -n obdfilter.${ost1_svc}.num_exports | cut -d' ' -f2)
815 echo starting with $OST_NEXP OST and $MDS_NEXP MDS exports
817 zconf_umount `hostname` $MOUNT2 -f
819 # PING_INTERVAL max(obd_timeout / 4, 1U)
820 # PING_EVICT_TIMEOUT (PING_INTERVAL * 6)
822 # evictor takes PING_EVICT_TIMEOUT + 3 * PING_INTERVAL to evict.
823 # But if there's a race to start the evictor from various obds,
824 # the loser might have to wait for the next ping.
825 # = 9 * PING_INTERVAL + PING_INTERVAL
826 # = 10 PING_INTERVAL = 10 obd_timeout / 4 = 2.5 obd_timeout
827 # let's wait $((TIMEOUT * 3)) # bug 19887
829 wait_client_evicted ost1 $OST_NEXP $((TIMEOUT * 3)) || \
830 error "Client was not evicted by ost" rc=1
831 wait_client_evicted $SINGLEMDS $MDS_NEXP $((TIMEOUT * 3)) || \
832 error "Client was not evicted by mds"
834 run_test 26b "evict dead exports"
838 writemany -q -a $DIR/$tdir/$tfile 0 5 &
841 local save_FAILURE_MODE=$FAILURE_MODE
843 facet_failover $SINGLEMDS
844 #define OBD_FAIL_OSC_SHUTDOWN 0x407
845 do_facet $SINGLEMDS lctl set_param fail_loc=0x80000407
846 # need to wait for reconnect
847 echo waiting for fail_loc
848 wait_update_facet $SINGLEMDS "lctl get_param -n fail_loc" "-2147482617"
849 facet_failover $SINGLEMDS
851 kill -USR1 $CLIENT_PID
854 FAILURE_MODE=$save_FAILURE_MODE
856 run_test 27 "fail LOV while using OSC's"
858 test_28() { # bug 6086 - error adding new clients
859 do_facet client mcreate $DIR/$tfile || return 1
860 drop_bl_callback "chmod 0777 $DIR/$tfile" ||echo "evicted as expected"
861 #define OBD_FAIL_MDS_CLIENT_ADD 0x12f
862 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000012f"
863 # fail once (evicted), reconnect fail (fail_loc), ok
864 client_up || (sleep 10; client_up) || (sleep 10; client_up) || error "reconnect failed"
866 fail $SINGLEMDS # verify MDS last_rcvd can be loaded
868 run_test 28 "handle error adding new clients (bug 6086)"
870 test_29a() { # bug 22273 - error adding new clients
871 #define OBD_FAIL_TGT_CLIENT_ADD 0x711
872 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000711"
873 # fail abort so client will be new again
874 fail_abort $SINGLEMDS
875 client_up || error "reconnect failed"
878 run_test 29a "error adding new clients doesn't cause LBUG (bug 22273)"
880 test_29b() { # bug 22273 - error adding new clients
881 #define OBD_FAIL_TGT_CLIENT_ADD 0x711
882 do_facet ost1 "lctl set_param fail_loc=0x80000711"
883 # fail abort so client will be new again
885 client_up || error "reconnect failed"
888 run_test 29b "error adding new clients doesn't cause LBUG (bug 22273)"
892 # put a load of file creates/writes/deletes
893 writemany -q $DIR/$tdir/$tfile 0 5 &
895 echo writemany pid $CLIENT_PID
899 # wait for client to reconnect to MDS
904 # client process should see no problems even though MDS went down
906 kill -USR1 $CLIENT_PID
909 echo writemany returned $rc
910 #these may fail because of eviction due to slow AST response.
911 [ $rc -eq 0 ] || error_ignore 13652 "writemany returned rc $rc" || true
913 run_test 50 "failover MDS under load"
916 #define OBD_FAIL_MDS_SYNC_CAPA_SL 0x1310
917 do_facet ost1 lctl set_param fail_loc=0x00001310
920 # put a load of file creates/writes/deletes
921 writemany -q $DIR/$tdir/$tfile 0 5 &
925 facet_failover $SINGLEMDS
926 # failover at various points during recovery
927 SEQ="1 5 10 $(seq $TIMEOUT 5 $(($TIMEOUT+10)))"
928 echo will failover at $SEQ
931 #echo failover in $i sec
932 log "test_$testnum: failover in $i sec"
934 facet_failover $SINGLEMDS
936 # client process should see no problems even though MDS went down
937 # and recovery was interrupted
939 kill -USR1 $CLIENT_PID
942 echo writemany returned $rc
943 [ $rc -eq 0 ] || error_ignore 13652 "writemany returned rc $rc" || true
945 run_test 51 "failover MDS during recovery"
948 do_facet client "mkdir -p $DIR/$tdir"
949 do_facet client "writemany -q -a $DIR/$tdir/$tfile 300 5" &
951 echo writemany pid $CLIENT_PID
956 wait $CLIENT_PID || rc=$?
957 # active client process should see an EIO for down OST
958 [ $rc -eq 5 ] && { echo "writemany correctly failed $rc" && return 0; }
959 # but timing or failover setup may allow success
960 [ $rc -eq 0 ] && { echo "writemany succeeded" && return 0; }
961 echo "writemany returned $rc"
966 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
971 [ $rc -ne 0 ] && { return $rc; }
972 # wait for client to reconnect to OST
976 [ $rc -ne 0 ] && { return $rc; }
983 run_test 52 "failover OST under load"
985 # test of open reconstruct
988 drop_ldlm_reply "openfile -f O_RDWR:O_CREAT -m 0755 $DIR/$tfile" ||\
991 run_test 53 "touch: drop rep"
994 zconf_mount `hostname` $MOUNT2
998 cat $DIR2/$tfile.missing # save transno = 0, rc != 0 into last_rcvd
1001 ERROR=`dmesg | egrep "(test 54|went back in time)" | tail -n1 | grep "went back in time"`
1002 [ x"$ERROR" == x ] || error "back in time occured"
1004 run_test 54 "back in time"
1006 # bug 11330 - liblustre application death during I/O locks up OST
1008 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1012 # first dd should be finished quickly
1013 $LFS setstripe -c 1 -i 0 $DIR/$tdir/$tfile-1
1014 dd if=/dev/zero of=$DIR/$tdir/$tfile-1 bs=32M count=4 &
1017 echo "step1: testing ......"
1019 if [ -z `ps x | awk '$1 == '$DDPID' { print $5 }'` ]; then break; fi
1021 if [ $count -gt 64 ]; then
1022 error "dd should be finished!"
1026 echo "(dd_pid=$DDPID, time=$count)successful"
1028 $LFS setstripe -c 1 -i 0 $DIR/$tdir/$tfile-2
1029 #define OBD_FAIL_OST_DROP_REQ 0x21d
1030 do_facet ost1 lctl set_param fail_loc=0x0000021d
1031 # second dd will be never finished
1032 dd if=/dev/zero of=$DIR/$tdir/$tfile-2 bs=32M count=4 &
1035 echo "step2: testing ......"
1036 while [ $count -le 64 ]; do
1037 dd_name="`ps x | awk '$1 == '$DDPID' { print $5 }'`"
1038 if [ -z $dd_name ]; then
1040 echo "debug: (dd_name=$dd_name, dd_pid=$DDPID, time=$count)"
1041 error "dd shouldn't be finished!"
1046 echo "(dd_pid=$DDPID, time=$count)successful"
1048 #Recover fail_loc and dd will finish soon
1049 do_facet ost1 lctl set_param fail_loc=0
1051 echo "step3: testing ......"
1053 if [ -z `ps x | awk '$1 == '$DDPID' { print $5 }'` ]; then break; fi
1055 if [ $count -gt 500 ]; then
1056 error "dd should be finished!"
1060 echo "(dd_pid=$DDPID, time=$count)successful"
1064 run_test 55 "ost_brw_read/write drops timed-out read/write request"
1066 test_56() { # b=11277
1067 #define OBD_FAIL_MDS_RESEND 0x136
1069 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000136"
1071 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1074 run_test 56 "do not allow reconnect to busy exports"
1077 # no oscs means no client or mdt
1078 while lctl get_param osc.*.* > /dev/null 2>&1; do
1079 : # loop until proc file is removed
1083 test_57() { # bug 10866
1087 #define OBD_FAIL_LPROC_REMOVE 0xB00
1088 lctl set_param fail_loc=0x80000B00
1089 zconf_umount `hostname` $DIR
1090 lctl set_param fail_loc=0x80000B00
1091 fail_abort $SINGLEMDS
1093 lctl set_param fail_loc=0
1095 do_facet client "df $DIR"
1097 run_test 57 "read procfs entries causes kernel crash"
1099 test_58() { # bug 11546
1100 #define OBD_FAIL_MDC_ENQUEUE_PAUSE 0x801
1103 lctl set_param fail_loc=0x80000801
1104 cp $DIR/$tfile /dev/null &
1107 lctl set_param fail_loc=0
1108 drop_bl_callback rm -f $DIR/$tfile
1110 # the first 'df' could tigger the eviction caused by
1111 # 'drop_bl_callback', and it's normal case.
1112 # but the next 'df' should return successfully.
1113 do_facet client "df $DIR" || do_facet client "df $DIR"
1115 run_test 58 "Eviction in the middle of open RPC reply processing"
1117 test_59() { # bug 10589
1118 zconf_mount `hostname` $MOUNT2 || error "Failed to mount $MOUNT2"
1119 echo $DIR2 | grep -q $MOUNT2 || error "DIR2 is not set properly: $DIR2"
1120 #define OBD_FAIL_LDLM_CANCEL_EVICT_RACE 0x311
1121 lctl set_param fail_loc=0x311
1122 writes=$(LANG=C dd if=/dev/zero of=$DIR2/$tfile count=1 2>&1)
1123 [ $? = 0 ] || error "dd write failed"
1124 writes=$(echo $writes | awk -F '+' '/out/ {print $1}')
1125 lctl set_param fail_loc=0
1127 zconf_umount `hostname` $MOUNT2 -f
1128 reads=$(LANG=C dd if=$DIR/$tfile of=/dev/null 2>&1)
1129 [ $? = 0 ] || error "dd read failed"
1130 reads=$(echo $reads | awk -F '+' '/in/ {print $1}')
1131 [ "$reads" -eq "$writes" ] || error "read" $reads "blocks, must be" $writes
1133 run_test 59 "Read cancel race on client eviction"
1136 # we assume that all md changes are in the MDT0 changelog
1137 if [ $MDSCOUNT -gt 1 ]; then
1138 error_ignore 17935 $*
1145 MDT0=$($LCTL get_param -n mdc.*.mds_server_uuid | \
1146 awk '{gsub(/_UUID/,""); print $1}' | head -1)
1151 # Register (and start) changelog
1152 USER=$(do_facet $SINGLEMDS lctl --device $MDT0 changelog_register -n)
1153 echo "Registered as $MDT0 changelog user $USER"
1155 # Generate a large number of changelog entries
1156 createmany -o $DIR/$tdir/$tfile $NUM_FILES
1160 # Unlink files in the background
1161 unlinkmany $DIR/$tdir/$tfile $NUM_FILES &
1165 # Failover the MDS while unlinks are happening
1166 facet_failover $SINGLEMDS
1168 # Wait for unlinkmany to finish
1171 # Check if all the create/unlink events were recorded
1173 $LFS changelog $MDT0 >> $DIR/$tdir/changelog
1174 local cl_count=$(grep UNLNK $DIR/$tdir/changelog | wc -l)
1175 echo "$cl_count unlinks in $MDT0 changelog"
1177 do_facet $SINGLEMDS lctl --device $MDT0 changelog_deregister $USER
1178 USERS=$(( $(do_facet $SINGLEMDS lctl get_param -n \
1179 mdd.$MDT0.changelog_users | wc -l) - 2 ))
1180 if [ $USERS -eq 0 ]; then
1181 [ $cl_count -eq $NUM_FILES ] || \
1182 err17935 "Recorded ${cl_count} unlinks out of $NUM_FILES"
1183 # Also make sure we can clear large changelogs
1184 cl_count=$($LFS changelog $FSNAME | wc -l)
1185 [ $cl_count -le 2 ] || \
1186 error "Changelog not empty: $cl_count entries"
1188 # If there are other users, there may be other unlinks in the log
1189 [ $cl_count -ge $NUM_FILES ] || \
1190 err17935 "Recorded ${cl_count} unlinks out of $NUM_FILES"
1191 echo "$USERS other changelog users; can't verify clear"
1194 run_test 60 "Add Changelog entries during MDS failover"
1198 local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS $FSNAME-OST0000)
1199 mdtosc=${mdtosc/-MDT*/-MDT\*}
1200 local cflags="osc.$mdtosc.connect_flags"
1201 do_facet $SINGLEMDS "lctl get_param -n $cflags" |grep -q skip_orphan
1202 [ $? -ne 0 ] && skip "don't have skip orphan feature" && return
1204 mkdir -p $DIR/$tdir || error "mkdir dir $DIR/$tdir failed"
1205 # Set the default stripe of $DIR/$tdir to put the files to ost1
1206 $LFS setstripe -c 1 -i 0 $DIR/$tdir
1208 replay_barrier $SINGLEMDS
1209 createmany -o $DIR/$tdir/$tfile-%d 10
1210 local oid=`do_facet ost1 "lctl get_param -n obdfilter.${ost1_svc}.last_id"`
1212 fail_abort $SINGLEMDS
1214 touch $DIR/$tdir/$tfile
1215 local id=`$LFS getstripe $DIR/$tdir/$tfile | awk '$1 == 0 { print $2 }'`
1216 [ $id -le $oid ] && error "the orphan objid was reused, failed"
1221 run_test 61 "Verify to not reuse orphan objects - bug 17025"
1223 check_cli_ir_state()
1225 local NODE=${1:-$HOSTNAME}
1227 st=$(do_node $NODE "lctl get_param mgc.*.ir_state |
1228 awk '/imperative_recovery:/ { print \\\$2}'")
1229 [ $st != ON -o $st != OFF -o $st != ENABLED -o $st != DISABLED ] ||
1230 error "Error state $st, must be ENABLED or DISABLED"
1234 check_target_ir_state()
1237 local name=${target}_svc
1238 local recovery_proc=obdfilter.${!name}.recovery_status
1241 st=$(do_facet $target "lctl get_param -n $recovery_proc |
1242 awk '/IR:/{ print \\\$2}'")
1243 [ $st != ON -o $st != OFF -o $st != ENABLED -o $st != DISABLED ] ||
1244 error "Error state $st, must be ENABLED or DISABLED"
1250 do_facet mgs lctl set_param -n mgs.MGS.live.$FSNAME="state=$1"
1255 local state=$(do_facet mgs "lctl get_param -n mgs.MGS.live.$FSNAME |
1256 awk '/state:/{ print \\\$2 }'")
1260 nidtbl_version_mgs()
1262 local ver=$(do_facet mgs "lctl get_param -n mgs.MGS.live.$FSNAME |
1263 awk '/nidtbl_version:/{ print \\\$2 }'")
1267 # nidtbl_version_client <mds1|client> [node]
1268 nidtbl_version_client()
1271 local node=${2:-$HOSTNAME}
1273 if [ X$cli = Xclient ]; then
1276 local obdtype=${cli/%[0-9]*/}
1277 [ $obdtype != mds ] && error "wrong parameters $cli"
1279 node=$(facet_active_host $cli)
1284 local vers=$(do_node $node "lctl get_param -n mgc.*.ir_state" |
1285 awk "/$cli/{print \$6}" |sort -u)
1287 # in case there are multiple mounts on the client node
1289 [ ${#arr[@]} -ne 1 ] && error "versions on client node mismatch"
1293 nidtbl_versions_match()
1295 [ $(nidtbl_version_mgs) -eq $(nidtbl_version_client ${1:-client}) ]
1298 target_instance_match()
1304 obdtype=${srv/%[0-9]*/}
1315 error "invalid target type" $srv
1320 local target=${srv}_svc
1321 local si=$(do_facet $srv lctl get_param -n $obdname.${!target}.instance)
1322 local ci=$(lctl get_param -n $cliname.${!target}-${cliname}-*.import | \
1323 awk '/instance/{ print $2 }' |head -1)
1325 return $([ $si -eq $ci ])
1330 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1331 { skip "MGS without IR support"; return 0; }
1333 # MDT was just restarted in the previous test, make sure everything
1336 while [ $cnt -gt 0 ]; do
1337 nidtbl_versions_match && break
1343 set_ir_status disabled
1345 local prev_ver=$(nidtbl_version_client client)
1347 local saved_FAILURE_MODE=$FAILURE_MODE
1348 [ $(facet_host mgs) = $(facet_host ost1) ] && FAILURE_MODE="SOFT"
1352 [ $(nidtbl_version_client client) -eq $prev_ver ] ||
1353 error "version must not change due to IR disabled"
1354 target_instance_match ost1 || error "instance mismatch"
1358 FAILURE_MODE=$saved_FAILURE_MODE
1360 run_test 100 "IR: Make sure normal recovery still works w/o IR"
1364 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1365 { skip "MGS without IR support"; return 0; }
1369 local OST1_IMP=$(get_osc_import_name client ost1)
1371 # disable pinger recovery
1372 lctl set_param -n osc.$OST1_IMP.pinger_recov=0
1376 target_instance_match ost1 || error "instance mismatch"
1377 nidtbl_versions_match || error "version must match"
1379 lctl set_param -n osc.$OST1_IMP.pinger_recov=1
1381 run_test 101 "IR: Make sure IR works w/o normal recovery"
1385 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1386 { skip "MGS without IR support"; return 0; }
1388 local clients=${CLIENTS:-$HOSTNAME}
1395 # let's have a new nidtbl version
1398 # sleep for a while so that clients can see the failure of ost
1399 # it must be MGC_TIMEOUT_MIN_SECONDS + MGC_TIMEOUT_RAND_CENTISEC.
1400 # int mgc_request.c:
1401 # define MGC_TIMEOUT_MIN_SECONDS 5
1402 # define MGC_TIMEOUT_RAND_CENTISEC 0x1ff /* ~500 *
1403 local count=30 # 20 seconds at most
1404 while [ $count -gt 0 ]; do
1405 nidtbl_versions_match && break
1410 nidtbl_versions_match || error "nidtbl mismatch"
1413 old_version=$(nidtbl_version_client client)
1415 zconf_umount_clients $clients $MOUNT || error "Cannot umount client"
1418 combined_mgs_mds && mgsdev=mds1
1419 remount_facet $mgsdev
1422 zconf_mount_clients $clients $MOUNT || error "Cannot mount client"
1425 new_version=$(nidtbl_version_client client)
1426 [ $new_version -lt $old_version ] &&
1427 error "nidtbl version wrong after mgs restarts"
1430 run_test 102 "IR: New client gets updated nidtbl after MGS restart"
1434 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1435 { skip "MGS without IR support"; return 0; }
1437 combined_mgs_mds && skip "mgs and mds on the same target" && return 0
1439 # workaround solution to generate config log on the mds
1445 # We need this test because mds is like a client in IR context.
1446 start mds1 $MDSDEV1 || error "MDS should start w/o mgs"
1448 # start mgs and remount mds w/ ir
1452 # remount client so that fsdb will be created on the MGS
1453 umount_client $MOUNT || error "umount failed"
1454 mount_client $MOUNT || error "mount failed"
1456 # sleep 30 seconds so the MDS has a chance to detect MGS restarting
1458 while [ $count -gt 0 ]; do
1459 [ $(nidtbl_version_client mds1) -ne 0 ] && break
1464 # after a while, mds should be able to reconnect to mgs and fetch
1465 # up-to-date nidtbl version
1466 nidtbl_versions_match mds1 || error "mds nidtbl mismatch"
1471 run_test 103 "IR: MDS can start w/o MGS and get updated nidtbl later"
1475 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1476 { skip "MGS without IR support"; return 0; }
1481 start ost1 $(ostdevname 1) "$OST_MOUNT_OPTS -onoir" ||
1482 error "OST1 cannot start"
1485 local ir_state=$(check_target_ir_state ost1)
1486 [ $ir_state = "DISABLED" -o $ir_state = "OFF" ] ||
1487 error "ir status on ost1 should be DISABLED"
1489 run_test 104 "IR: ost can disable IR voluntarily"
1493 [ -z "$RCLIENTS" ] && skip "Needs multiple clients" && return 0
1494 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1495 { skip "MGS without IR support"; return 0; }
1499 # get one of the clients from client list
1500 local rcli=$(echo $RCLIENTS |cut -d' ' -f 1)
1502 local old_MOUNTOPT=$MOUNTOPT
1503 MOUNTOPT=${MOUNTOPT},noir
1504 zconf_umount $rcli $MOUNT || error "umount failed"
1505 zconf_mount $rcli $MOUNT || error "mount failed"
1507 # make sure lustre mount at $rcli disabling IR
1508 local ir_state=$(check_cli_ir_state $rcli)
1509 [ $ir_state = "DISABLED" -o $ir_state = "OFF" ] ||
1510 error "IR state must be DISABLED at $rcli"
1512 # Since the client just mounted, its last_rcvd entry is not on disk.
1513 # Send an RPC so exp_need_sync forces last_rcvd to commit this export
1514 # so the client can reconnect during OST recovery (LU-924, LU-1582)
1515 $SETSTRIPE -i 0 $DIR/$tfile
1516 dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 conv=sync
1518 # make sure MGS's state is Partial
1519 [ $(get_ir_status) = "partial" ] || error "MGS IR state must be partial"
1522 # make sure IR on ost1 is DISABLED
1523 local ir_state=$(check_target_ir_state ost1)
1524 [ $ir_state = "DISABLED" -o $ir_state = "OFF" ] ||
1525 error "IR status on ost1 should be DISABLED"
1528 MOUNTOPT=$old_MOUNTOPT
1529 zconf_umount $rcli $MOUNT || error "umount failed"
1530 zconf_mount $rcli $MOUNT || error "mount failed"
1532 # make sure MGS's state is full
1533 [ $(get_ir_status) = "full" ] || error "MGS IR status must be full"
1536 # make sure IR on ost1 is ENABLED
1537 local ir_state=$(check_target_ir_state ost1)
1538 [ $ir_state = "ENABLED" -o $ir_state = "ON" ] ||
1539 error "IR status on ost1 should be ENABLED"
1543 run_test 105 "IR: NON IR clients support"
1551 test_106() { # LU-1789
1552 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.3.50) ]] ||
1553 { skip "Need MDS version at least 2.3.50"; return 0; }
1555 #define OBD_FAIL_MDC_LIGHTWEIGHT 0x805
1556 $LCTL set_param fail_loc=0x805
1559 trap cleanup_106 EXIT
1561 # enable lightweight flag on mdc connection
1564 local MDS_NEXP=$(do_facet $SINGLEMDS \
1565 lctl get_param -n mdt.${mds1_svc}.num_exports |
1567 $LCTL set_param fail_loc=0
1569 touch $DIR2/$tfile || error "failed to create empty file"
1570 replay_barrier $SINGLEMDS
1572 $LCTL set_param debug=console
1574 facet_failover $SINGLEMDS
1576 # lightweight connection must be evicted
1577 touch -c $DIR2/$tfile || true
1578 $LCTL dk $TMP/lustre-log-$TESTNAME.log
1579 evicted=`awk '/This client was evicted by .*MDT0000/ {
1581 }' $TMP/lustre-log-$TESTNAME.log`
1582 [ -z "$evicted" ] && error "lightweight client not evicted by mds"
1584 # and all operations performed by lightweight client should be
1585 # synchronous, so the file created before mds restart should be there
1586 $CHECKSTAT -t file $DIR/$tfile || error "file not present"
1591 run_test 106 "lightweight connection support"
1598 # OBD_FAIL_MDS_REINT_NET_REP 0x119
1599 do_facet $SINGLEMDS lctl set_param fail_loc=0x119
1600 multiop $DIR/$tdir D_c &
1602 mkdir $DIR/$tdir/dir_106 &
1604 do_facet $SINGLEMDS lctl set_param fail_loc=0
1607 wait $CLIENT_PID || rc=$?
1608 checkstat -t dir $DIR/$tdir/dir_106 || return 1
1610 kill -USR1 $close_pid
1611 wait $close_pid || return 2
1615 run_test 107 "drop reint reply, then restart MDT"
1617 check_and_cleanup_lustre