5 PTLDEBUG=${PTLDEBUG:--1}
6 LUSTRE=${LUSTRE:-$(dirname $0)/..}
7 . $LUSTRE/tests/test-framework.sh
11 ALWAYS_EXCEPT="$RECOVERY_SMALL_EXCEPT "
13 # bug number for skipped test: LU-12896
14 ALWAYS_EXCEPT+=" 110k"
15 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
20 require_dsh_mds || exit 0
22 # Allow us to override the setup if we already have a mounted system by
23 # setting SETUP=" " and CLEANUP=" "
25 CLEANUP=${CLEANUP:-""}
27 check_and_setup_lustre
30 rm -rf $DIR/d[0-9]* $DIR/f.${TESTSUITE}*
33 local f1="$DIR/$tfile"
34 local f2="$DIR/$tfile.2"
36 drop_request "mcreate $f1" ||
37 error_noexit "create '$f1': drop req"
39 drop_reint_reply "mcreate $f2" ||
40 error_noexit "create '$f2': drop rep"
42 drop_request "tchmod 111 $f2" ||
43 error_noexit "chmod '$f2': drop req"
45 drop_reint_reply "tchmod 666 $f2" ||
46 error_noexit "chmod '$f2': drop rep"
48 drop_request "statone $f2" ||
49 error_noexit "stat '$f2': drop req"
51 drop_reply "statone $f2" ||
52 error_noexit "stat '$f2': drop rep"
54 run_test 1 "create, chmod, stat: drop req, drop rep"
58 do_facet_create_file client $t 10K ||
59 error_noexit "Create file $t"
61 drop_request "cat $t > /dev/null" ||
62 error_noexit "Open request for $t file"
64 drop_reply "cat $t > /dev/null" ||
65 error_noexit "Open replay for $t file"
67 run_test 4 "open: drop req, drop rep"
72 local RR="$T-renamed-again"
73 do_facet_create_file client $T 10K ||
74 error_noexit "Create file $T"
76 drop_request "mv $T $R" ||
77 error_noexit "Rename $T"
79 drop_reint_reply "mv $R $RR" ||
80 error_noexit "Failed rename replay on $R"
82 do_facet client "checkstat -v $RR" ||
83 error_noexit "checkstat error on $RR"
85 do_facet client "rm $RR" ||
86 error_noexit "Can't remove file $RR"
88 run_test 5 "rename: drop req, drop rep"
92 local LINK1=$DIR/$tfile.link1
93 local LINK2=$DIR/$tfile.link2
95 do_facet_create_file client $T 10K ||
96 error_noexit "Create file $T"
98 drop_request "mlink $T $LINK1" ||
99 error_noexit "mlink request for $T"
101 drop_reint_reply "mlink $T $LINK2" ||
102 error_noexit "mlink reply for $T"
104 drop_request "munlink $LINK1" ||
105 error_noexit "munlink request for $T"
107 drop_reint_reply "munlink $LINK2" ||
108 error_noexit "munlink reply for $T"
110 do_facet client "rm $T" ||
111 error_noexit "Can't remove file $T"
113 run_test 6 "link, unlink: drop req, drop rep"
117 drop_reint_reply "touch $DIR/$tfile" || return 1
119 run_test 8 "touch: drop rep (bug 1423)"
123 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
127 do_facet_random_file client $TMP/$tfile 1K ||
128 error_noexit "Create random file $TMP/$tfile"
129 # make this big, else test 9 doesn't wait for bulk -- bz 5595
130 do_facet_create_file client $TMP/$t1 4M ||
131 error_noexit "Create file $TMP/$t1"
132 do_facet client "cp $TMP/$t1 $DIR/$t1" ||
133 error_noexit "Can't copy to $DIR/$t1 file"
134 pause_bulk "cp $TMP/$tfile $DIR/$tfile" ||
135 error_noexit "Can't pause_bulk copy"
136 do_facet client "cp $TMP/$t1 $DIR/$t2" ||
137 error_noexit "Can't copy file"
138 do_facet client "sync"
139 do_facet client "rm $DIR/$tfile $DIR/$t2 $DIR/$t1" ||
140 error_noexit "Can't remove files"
141 do_facet client "rm $TMP/$t1 $TMP/$tfile"
143 run_test 9 "pause bulk on OST (bug 1420)"
147 local before=$(date +%s)
150 do_facet client "stat $DIR > /dev/null" ||
151 error "failed to stat $DIR: $?"
152 drop_bl_callback "chmod 0777 $DIR" ||
153 error "failed to chmod $DIR: $?"
155 # let the client reconnect
157 evict=$(do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state |
158 awk -F"[ [,]" '/EVICTED ]$/ { if (mx<$5) {mx=$5;} } END { print mx }')
159 [ ! -z "$evict" ] && [[ $evict -gt $before ]] ||
160 (do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state;
161 error "no eviction: $evict before:$before")
163 do_facet client checkstat -v -p 0777 $DIR ||
164 error "client checkstat failed: $?"
166 run_test 10a "finish request on server after client eviction (bug 1521)"
169 local before=$(date +%s)
172 [[ "$MDS1_VERSION" -lt $(version_code 2.6.53) ]] &&
173 skip "Need MDS version at least 2.6.53"
174 do_facet client "stat $DIR > /dev/null" ||
175 error "failed to stat $DIR: $?"
176 drop_bl_callback_once "chmod 0777 $DIR" ||
177 error "failed to chmod $DIR: $?"
179 # let the client reconnect
181 evict=$(do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state |
182 awk -F"[ [,]" '/EVICTED ]$/ { if (mx<$5) {mx=$5;} } END { print mx }')
184 [ -z "$evict" ] || [[ $evict -le $before ]] ||
185 (do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state;
186 error "eviction happened: $evict before:$before")
188 do_facet client checkstat -v -p 0777 $DIR ||
189 error "client checkstat failed: $?"
191 run_test 10b "re-send BL AST"
194 local before=$(date +%s)
203 workdir="${DIR}/${tdir}"
204 mkdir -p ${workdir} || error "can't create workdir $?"
205 stat ${workdir} > /dev/null ||
206 error "failed to stat ${workdir}: $?"
207 mdtidx=$($LFS getdirstripe -i ${workdir})
208 mdtname=$($LFS mdts ${workdir} | grep -e "^$mdtidx:" |
209 awk '{sub("_UUID", "", $2); print $2;}')
211 mdccli=$($LCTL dl | grep "${mdtname}-mdc" | awk '{print $4;}')
212 conn_uuid=$($LCTL get_param -n mdc.${mdccli}.conn_uuid)
213 mdcpath="mdc.${mdccli}.import=connection=${conn_uuid}"
215 drop_bl_callback_once "chmod 0777 ${workdir}" &
220 # force client reconnect
221 $LCTL set_param "${mdcpath}"
223 # wait client reconnect
227 evict=$($LCTL get_param mdc.${mdccli}.state |
228 awk -F"[ [,]" '/EVICTED]$/ { if (t<$4) {t=$4;} } END { print t }')
230 [[ $evict -le $before ]] ||
231 ( $LCTL get_param mdc.$FSNAME-MDT*.state;
232 error "eviction happened: $EVICT before:$BEFORE" )
234 [ $rc -eq 0 ] || error "chmod must finished OK"
235 checkstat -v -p 0777 "${workdir}" ||
236 error "client checkstat failed: $?"
238 run_test 10c "re-send BL AST vs reconnect race (LU-5569)"
241 local before=$(date +%s)
244 [[ "$MDS1_VERSION" -lt $(version_code 2.6.90) ]] &&
245 skip "Need MDS version at least 2.6.90"
247 # sleep 1 is to make sure that BEFORE is not equal to EVICTED below
250 echo -n ", world" | dd of=$TMP/$tfile bs=1c seek=5
252 remount_client $MOUNT
256 $LFS setstripe -i 0 -c 1 $DIR1/$tfile
257 echo -n hello | dd of=$DIR1/$tfile bs=5
259 stat $DIR2/$tfile >& /dev/null
260 $LCTL set_param fail_err=71
261 drop_bl_callback "echo -n \\\", world\\\" >> $DIR2/$tfile"
266 cmp -l $DIR1/$tfile $DIR2/$tfile || error "file contents differ"
267 cmp -l $DIR1/$tfile $TMP/$tfile || error "wrong content found"
269 evict=$(do_facet client $LCTL get_param osc.$FSNAME-OST0000*.state | \
271 awk -F"[ [,]" '/EVICTED$/ { if (mx<$1) {mx=$1;} } END { print mx }')
273 [[ $evict -gt $before ]] ||
274 (do_facet client $LCTL get_param osc.$FSNAME-OST0000*.state;
275 error "no eviction: $evict before:$before")
277 $LCTL set_param fail_err=0
279 umount_client $MOUNT2
281 run_test 10d "test failed blocking ast"
285 [[ "$OST1_VERSION" -le $(version_code 2.8.58) ]] &&
286 skip "Need OST version at least 2.8.59"
287 [ $CLIENTCOUNT -lt 2 ] && skip "need two clients"
288 [ $(facet_host client) == $(facet_host ost1) ] &&
289 skip "need ost1 and client on different nodes"
290 local -a clients=(${CLIENTS//,/ })
291 local client1=${clients[0]}
292 local client2=${clients[1]}
294 $LFS setstripe -c 1 -i 0 $DIR/$tfile-1 $DIR/$tfile-2
295 $MULTIOP $DIR/$tfile-1 Ow1048576c
297 #define OBD_FAIL_LDLM_BL_CALLBACK_NET 0x305
298 $LCTL set_param fail_loc=0x80000305
300 #define OBD_FAIL_LDLM_ENQUEUE_OLD_EXPORT 0x30e
301 do_facet ost1 "$LCTL set_param fail_loc=0x1000030e"
302 # hit OBD_FAIL_LDLM_ENQUEUE_OLD_EXPORT twice:
303 # 1. to return ENOTCONN from ldlm_handle_enqueue0
304 # 2. to pause reconnect handling between resend and setting
305 # import to LUSTRE_IMP_FULL state
306 do_facet ost1 "$LCTL set_param fail_val=3"
308 # client1 fails ro respond to bl ast
309 do_node $client2 "$MULTIOP $DIR/$tfile-1 Ow1048576c" &
312 # ost1 returns error on enqueue, which causes client1 to reconnect
313 do_node $client1 "$MULTIOP $DIR/$tfile-2 Ow1048576c" ||
314 error "multiop failed"
317 do_facet ost1 "$LCTL set_param fail_loc=0"
318 do_facet ost1 "$LCTL set_param fail_val=0"
320 run_test 10e "re-send BL AST vs reconnect race 2"
323 # wake up a thread waiting for completion after eviction
325 do_facet client $MULTIOP $DIR/$tfile Ow ||
326 { error "multiop write failed: $?"; return 1; }
327 do_facet client $MULTIOP $DIR/$tfile or ||
328 { error "multiop read failed: $?"; return 2; }
332 do_facet client $MULTIOP $DIR/$tfile or ||
333 { error "multiop read failed: $?"; return 3; }
334 drop_bl_callback_once $MULTIOP $DIR/$tfile Ow ||
335 echo "evicted as expected"
337 do_facet client munlink $DIR/$tfile ||
338 { error "munlink failed: $?"; return 4; }
339 # allow recovery to complete
340 client_up || client_up || sleep $TIMEOUT
342 run_test 11 "wake up a thread waiting for completion after eviction (b=2460)"
346 $LCTL mark $MULTIOP $DIR/$tfile OS_c
347 do_facet $SINGLEMDS "lctl set_param fail_loc=0x115"
348 clear_failloc $SINGLEMDS $((TIMEOUT * 2)) &
349 multiop_bg_pause $DIR/$tfile OS_c ||
350 { error "multiop failed: $?"; return 1; }
352 #define OBD_FAIL_MDS_CLOSE_NET 0x115
354 echo "waiting for multiop $PID"
355 wait $PID || { error "wait for multiop faile: $?"; return 2; }
356 do_facet client munlink $DIR/$tfile ||
357 { error "client munlink failed: $?"; return 3; }
358 # allow recovery to complete
359 client_up || client_up || sleep $TIMEOUT
361 run_test 12 "recover from timed out resend in ptlrpcd (b=2494)"
363 # Bug 113, check that readdir lost recv timeout works.
365 mkdir -p $DIR/$tdir || { error "mkdir failed: $?"; return 1; }
366 touch $DIR/$tdir/newentry || { error "touch failed: $?"; return 2; }
367 # OBD_FAIL_MDS_READPAGE_NET|OBD_FAIL_ONCE
368 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000104"
369 ls $DIR/$tdir || { error "ls failed: $?"; return 3; }
370 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
371 rm -rf $DIR/$tdir || { error "remove test dir failed: $?"; return 4; }
373 run_test 13 "mdc_readpage restart test (bug 1138)"
375 # Bug 113, check that readdir lost send timeout works.
378 touch $DIR/$tdir/newentry
379 # OBD_FAIL_MDS_SENDPAGE|OBD_FAIL_ONCE
380 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000106"
381 ls $DIR/$tdir || return 1
382 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
384 run_test 14 "mdc_readpage resend test (bug 1138)"
387 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000128"
388 touch $DIR/$tfile && return 1
391 run_test 15 "failed open (-ENOMEM)"
393 READ_AHEAD=`lctl get_param -n llite.*.max_read_ahead_mb | head -n 1`
395 lctl set_param -n llite.*.max_read_ahead_mb 0
399 lctl set_param -n llite.*.max_read_ahead_mb $READ_AHEAD
403 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
405 do_facet_random_file client $TMP/$tfile 100K ||
406 { error_noexit "Create random file $TMP/$T" ; return 0; }
407 do_facet client "cp $TMP/$tfile $DIR/$tfile" ||
408 { error_noexit "Copy to $DIR/$tfile file" ; return 0; }
412 #define OBD_FAIL_PTLRPC_BULK_PUT_NET 0x504 | OBD_FAIL_ONCE
413 do_facet ost1 "lctl set_param fail_loc=0x80000504"
415 # OST bulk will time out here, client resends
416 do_facet client "cmp $TMP/$tfile $DIR/$tfile" || return 1
417 do_facet ost1 lctl set_param fail_loc=0
418 # give recovery a chance to finish (shouldn't take long)
420 do_facet client "cmp $TMP/$tfile $DIR/$tfile" || return 2
424 run_test 16 "timeout bulk put, don't evict client (2732)"
429 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
431 local SAMPLE_FILE=$TMP/$tfile
432 do_facet_random_file client $SAMPLE_FILE 20K ||
433 { error_noexit "Create random file $SAMPLE_FILE" ; return 0; }
435 # With adaptive timeouts, bulk_get won't expire until adaptive_timeout_max
436 if at_is_enabled; then
437 at_max_saved=$(at_max_get ost1)
438 at_max_set $TIMEOUT ost1
441 # OBD_FAIL_PTLRPC_BULK_GET_NET 0x0503 | OBD_FAIL_ONCE
442 # OST bulk will time out here, client retries
443 do_facet ost1 lctl set_param fail_loc=0x80000503
444 # need to ensure we send an RPC
445 do_facet client cp $SAMPLE_FILE $DIR/$tfile
448 # with AT, client will wait adaptive_max*factor+net_latency before
449 # expiring the req, hopefully timeout*2 is enough
450 sleep $(($TIMEOUT*2))
452 do_facet ost1 lctl set_param fail_loc=0
453 do_facet client "df $DIR"
454 # expect cmp to succeed, client resent bulk
455 do_facet client "cmp $SAMPLE_FILE $DIR/$tfile" || return 3
456 do_facet client "rm $DIR/$tfile" || return 4
457 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved ost1
460 run_test 17a "timeout bulk get, don't evict client (2732)"
463 [ -z "$RCLIENTS" ] && skip "Needs multiple clients" && return 0
465 # get one of the clients from client list
466 local rcli=$(echo $RCLIENTS | cut -d ' ' -f 1)
467 local p="$TMP/$TESTSUITE-$TESTNAME.parameters"
468 local ldlm_enqueue_min=$(do_facet ost1 find /sys -name ldlm_enqueue_min)
469 [ -z "$ldlm_enqueue_min" ] &&
470 skip "missing /sys/.../ldlm_enqueue_min" && return 0
472 $LFS setstripe -i 0 -c 1 -S 1048576 $DIR/$tfile ||
473 error "setstripe failed"
474 $LFS setstripe -i 0 -c 1 -S 1048576 $DIR/${tfile}2 ||
475 error "setstripe 2 failed"
477 save_lustre_params ost1 "at_history" > $p
478 save_lustre_params ost1 "bulk_timeout" >> $p
479 local dev="${FSNAME}-OST0000"
480 save_lustre_params ost1 "obdfilter.$dev.brw_size" >> $p
481 local ldlm_enqueue_min_save=$(do_facet ost1 cat $ldlm_enqueue_min)
483 local new_at_history=15
485 do_facet ost1 "$LCTL set_param at_history=$new_at_history"
486 do_facet ost1 "$LCTL set_param bulk_timeout=30"
487 do_facet ost1 "echo 30 > /sys/module/ptlrpc/parameters/ldlm_enqueue_min"
489 # brw_size is required to be 4m so that bulk transfer timeout
490 # could be simulated with OBD_FAIL_PTLRPC_CLIENT_BULK_CB3
491 local brw_size=$($LCTL get_param -n \
492 osc.$FSNAME-OST0000-osc-[^M]*.import |
493 awk '/max_brw_size/{print $2}')
494 if [ $brw_size -ne 4194304 ]
496 save_lustre_params ost1 "obdfilter.$dev.brw_size" >> $p
497 do_facet ost1 "$LCTL set_param obdfilter.$dev.brw_size=4"
498 remount_client $MOUNT
501 # get service estimate expanded
502 #define OBD_FAIL_OST_BRW_PAUSE_PACK 0x224
503 do_facet ost1 "$LCTL set_param fail_loc=0x80000224 fail_val=35"
504 echo "delay rpc servicing by 35 seconds"
505 dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 conv=fdatasync
508 estimate=$($LCTL get_param -n osc.$dev-osc-*.timeouts |
509 awk '/portal 6/ {print $5}')
510 echo "service estimates increased to $estimate"
512 # let current worst service estimate to get closer to obliteration
513 sleep $((new_at_history / 3))
515 # start i/o and simulate bulk transfer loss
516 #define OBD_FAIL_PTLRPC_CLIENT_BULK_CB3 0x520
517 do_facet ost1 "$LCTL set_param fail_loc=0xa0000520 fail_val=1"
518 dd if=/dev/zero of=$DIR/$tfile bs=2M count=1 conv=fdatasync,notrunc &
521 # start lock conflict handling
522 sleep $((new_at_history / 3))
523 do_node $rcli "dd if=$DIR/$tfile of=/dev/null bs=1M count=1" &
526 # obliterate the worst service estimate
527 sleep $((new_at_history / 3 + 1))
528 dd if=/dev/zero of=$DIR/${tfile}2 bs=1M count=1
530 estimate=$($LCTL get_param -n osc.$dev-osc-*.timeouts |
531 awk '/portal 6/ {print $5}')
532 echo "service estimate dropped to $estimate"
535 [[ $? == 0 ]] || error "write failed"
537 [[ $? == 0 ]] || error "read failed"
539 restore_lustre_params <$p
540 if [ $brw_size -ne 4194304 ]
542 remount_client $MOUNT || error "remount_client failed"
544 do_facet ost1 "echo $ldlm_enqueue_min_save > $ldlm_enqueue_min"
546 run_test 17b "timeout bulk get, dont evict client (3582)"
549 [ -z ${ost2_svc} ] && skip_env "needs 2 osts" && return 0
551 do_facet_create_file client $TMP/$tfile 20K ||
552 { error_noexit "Create file $TMP/$tfile" ; return 0; }
554 do_facet client mkdir -p $DIR/$tdir
558 pgcache_empty || return 1
561 $LFS setstripe -i 1 -c 1 $f
562 stripe_index=$($LFS getstripe -i $f)
563 if [ $stripe_index -ne 1 ]; then
565 error "$f: stripe_index $stripe_index != 1" && return
568 do_facet client cp $TMP/$tfile $f
570 local osc2dev=`lctl get_param -n devices | grep ${ost2_svc}-osc- | egrep -v 'MDT' | awk '{print $1}'`
571 $LCTL --device $osc2dev deactivate || return 3
572 # my understanding is that there should be nothing in the page
573 # cache after the client reconnects?
575 pgcache_empty || rc=2
576 $LCTL --device $osc2dev activate
580 run_test 18a "manual ost invalidate clears page cache immediately"
583 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
585 do_facet_create_file client $TMP/$tfile 20K ||
586 { error_noexit "Create file $TMP/$tfile" ; return 0; }
588 do_facet client mkdir -p $DIR/$tdir
592 pgcache_empty || return 1
594 $LFS setstripe -i 0 -c 1 $f
595 stripe_index=$($LFS getstripe -i $f)
596 if [ $stripe_index -ne 0 ]; then
598 error "$f: stripe_index $stripe_index != 0" && return
601 do_facet client cp $TMP/$tfile $f
604 # allow recovery to complete
605 sleep $((TIMEOUT + 2))
606 # my understanding is that there should be nothing in the page
607 # cache after the client reconnects?
609 pgcache_empty || rc=2
613 run_test 18b "eviction and reconnect clears page cache (2766)"
616 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
618 do_facet_create_file client $TMP/$tfile 20K ||
619 { error_noexit "Create file $TMP/$tfile" ; return 0; }
621 do_facet client mkdir -p $DIR/$tdir
625 pgcache_empty || return 1
627 $LFS setstripe -i 0 -c 1 $f
628 stripe_index=$($LFS getstripe -i $f)
629 if [ $stripe_index -ne 0 ]; then
631 error "$f: stripe_index $stripe_index != 0" && return
634 do_facet client cp $TMP/$tfile $f
638 # OBD_FAIL_OST_CONNECT_NET2
639 # lost reply to connect request
640 do_facet ost1 lctl set_param fail_loc=0x80000225
643 $LFS df $MOUNT > /dev/null 2>&1
645 # my understanding is that there should be nothing in the page
646 # cache after the client reconnects?
648 pgcache_empty || rc=2
652 run_test 18c "Dropped connect reply after eviction handing (14755)"
655 local BEFORE=`date +%s`
658 mount_client $DIR2 || error "failed to mount $DIR2"
660 # cancel cached locks from OST to avoid eviction from it
663 do_facet client "stat $DIR > /dev/null" ||
664 error "failed to stat $DIR: $?"
665 drop_ldlm_cancel "chmod 0777 $DIR2" ||
666 error "failed to chmod $DIR2"
670 # let the client reconnect
672 EVICT=$(do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state |
673 awk -F"[ [,]" '/EVICTED ]$/ \
674 { if (mx<$5) {mx=$5;} } END { print mx }')
676 [ ! -z "$EVICT" ] && [[ $EVICT -gt $BEFORE ]] ||
677 (do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state;
678 error "no eviction: $EVICT before:$BEFORE")
680 run_test 19a "test expired_lock_main on mds (2867)"
683 local BEFORE=`date +%s`
686 mount_client $DIR2 || error "failed to mount $DIR2: $?"
688 # cancel cached locks from MDT to avoid eviction from it
691 do_facet client $MULTIOP $DIR/$tfile Ow ||
692 error "failed to run multiop: $?"
693 drop_ldlm_cancel $MULTIOP $DIR2/$tfile Ow ||
694 error "failed to ldlm_cancel: $?"
696 umount_client $DIR2 || error "failed to unmount $DIR2: $?"
697 do_facet client munlink $DIR/$tfile ||
698 error "failed to unlink $DIR/$tfile: $?"
700 # let the client reconnect
702 EVICT=$(do_facet client $LCTL get_param osc.$FSNAME-OST*.state |
703 awk -F"[ [,]" '/EVICTED ]$/ \
704 { if (mx < $5) {mx = $5;} } END { print mx }')
706 [ ! -z "$EVICT" ] && [[ $EVICT -gt $BEFORE ]] ||
707 (do_facet client $LCTL get_param osc.$FSNAME-OST*.state;
708 error "no eviction: $EVICT before:$BEFORE")
710 run_test 19b "test expired_lock_main on ost (2867)"
713 local BEFORE=`date +%s`
716 $LCTL set_param ldlm.namespaces.*.early_lock_cancel=0
718 mkdir -p $DIR1/$tfile
721 #define OBD_FAIL_PTLRPC_CANCEL_RESEND 0x516
722 do_facet mds $LCTL set_param fail_loc=0x80000516
724 touch $DIR2/$tfile/file1 &
726 # let touch to get blocked on the server
730 $LCTL set_param ldlm.namespaces.*.early_lock_cancel=1
733 # let the client reconnect
735 EVICT=$(do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state |
736 awk -F"[ [,]" '/EVICTED ]$/ { if (mx<$5) {mx=$5;} } END { print mx }')
738 [ -z "$EVICT" ] || [[ $EVICT -le $BEFORE ]] || error "eviction happened"
740 run_test 19c "check reconnect and lock resend do not trigger expired_lock_main"
742 test_20a() { # bug 2983 - ldlm_handle_enqueue cleanup
743 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
746 $LFS setstripe -i 0 -c 1 $DIR/$tdir/${tfile}
747 multiop_bg_pause $DIR/$tdir/${tfile} O_wc || return 1
750 #define OBD_FAIL_LDLM_ENQUEUE_EXTENT_ERR 0x308
751 do_facet ost1 lctl set_param fail_loc=0x80000308
752 kill -USR1 $MULTI_PID
755 [ $rc -eq 0 ] && error "multiop didn't fail enqueue: rc $rc" || true
757 run_test 20a "ldlm_handle_enqueue error (should return error)"
759 test_20b() { # bug 2986 - ldlm_handle_enqueue error during open
760 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
763 $LFS setstripe -i 0 -c 1 $DIR/$tdir/${tfile}
765 #define OBD_FAIL_LDLM_ENQUEUE_EXTENT_ERR 0x308
766 do_facet ost1 lctl set_param fail_loc=0x80000308
767 dd if=/etc/hosts of=$DIR/$tdir/$tfile && \
768 error "didn't fail open enqueue" || true
770 run_test 20b "ldlm_handle_enqueue error (should return error)"
773 mkdir -p $DIR/$tdir-1
774 mkdir -p $DIR/$tdir-2
775 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
778 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000129"
779 $MULTIOP $DIR/$tdir-2/f Oc &
782 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
784 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
785 kill -USR1 $close_pid
787 wait $close_pid || return 1
788 wait $open_pid || return 2
789 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
791 $CHECKSTAT -t file $DIR/$tdir-1/f || return 3
792 $CHECKSTAT -t file $DIR/$tdir-2/f || return 4
796 run_test 21a "drop close request while close and open are both in flight"
799 mkdir -p $DIR/$tdir-1
800 mkdir -p $DIR/$tdir-2
801 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
804 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000107"
805 mcreate $DIR/$tdir-2/f &
808 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
810 kill -USR1 $close_pid
812 wait $close_pid || return 1
813 wait $open_pid || return 3
815 $CHECKSTAT -t file $DIR/$tdir-1/f || return 4
816 $CHECKSTAT -t file $DIR/$tdir-2/f || return 5
819 run_test 21b "drop open request while close and open are both in flight"
822 mkdir -p $DIR/$tdir-1
823 mkdir -p $DIR/$tdir-2
824 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
827 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000107"
828 mcreate $DIR/$tdir-2/f &
831 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
833 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
834 kill -USR1 $close_pid
836 wait $close_pid || return 1
837 wait $open_pid || return 2
839 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
841 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
842 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
845 run_test 21c "drop both request while close and open are both in flight"
848 mkdir -p $DIR/$tdir-1
849 mkdir -p $DIR/$tdir-2
850 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
853 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000129"
854 $MULTIOP $DIR/$tdir-2/f Oc &
856 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
858 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000122"
861 wait $pid || return 1
862 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
864 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
865 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
869 run_test 21d "drop close reply while close and open are both in flight"
872 mkdir -p $DIR/$tdir-1
873 mkdir -p $DIR/$tdir-2
874 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
877 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000119"
878 touch $DIR/$tdir-2/f &
880 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
884 wait $pid || return 1
887 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
888 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
891 run_test 21e "drop open reply while close and open are both in flight"
894 mkdir -p $DIR/$tdir-1
895 mkdir -p $DIR/$tdir-2
896 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
899 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000119"
900 touch $DIR/$tdir-2/f &
902 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
904 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000122"
907 wait $pid || return 1
908 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
910 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
911 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
914 run_test 21f "drop both reply while close and open are both in flight"
917 mkdir -p $DIR/$tdir-1
918 mkdir -p $DIR/$tdir-2
919 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
922 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000119"
923 touch $DIR/$tdir-2/f &
925 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
927 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
930 wait $pid || return 1
931 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
933 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
934 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
937 run_test 21g "drop open reply and close request while close and open are both in flight"
940 mkdir -p $DIR/$tdir-1
941 mkdir -p $DIR/$tdir-2
942 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
945 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000107"
946 touch $DIR/$tdir-2/f &
949 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
951 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000122"
954 wait $pid || return 1
955 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
957 wait $touch_pid || return 2
959 $CHECKSTAT -t file $DIR/$tdir-1/f || return 3
960 $CHECKSTAT -t file $DIR/$tdir-2/f || return 4
963 run_test 21h "drop open request and close reply while close and open are both in flight"
965 # bug 3462 - multiple MDC requests
970 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
975 $MULTIOP $f1 msu || return 1
978 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
980 wait $close_pid || return 2
981 rm -rf $f2 || return 4
983 run_test 22 "drop close request and do mknod"
986 multiop_bg_pause $DIR/$tfile O_c || return 1
988 # give a chance for open
992 drop_request "kill -USR1 $pid"
995 wait $pid || return 1
998 run_test 23 "client hang when close a file after mds crash"
1000 test_24a() { # bug 11710 details correct fsync() behavior
1001 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1004 $LFS setstripe -i 0 -c 1 $DIR/$tdir
1005 cancel_lru_locks osc
1006 multiop_bg_pause $DIR/$tdir/$tfile Owy_wyc || return 1
1009 kill -USR1 $MULTI_PID
1012 lctl set_param fail_loc=0x0
1015 error_ignore bz5494 "multiop didn't fail fsync: rc $rc" || true
1017 run_test 24a "fsync error (should return error)"
1019 wait_client_evicted () {
1022 local varsvc=${facet}_svc
1024 wait_update $(facet_active_host $facet) \
1025 "lctl get_param -n *.${!varsvc}.num_exports | cut -d' ' -f2" \
1030 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1032 dmesg -c > /dev/null
1034 lfs setstripe $DIR/$tdir -S 0 -i 0 -c 1 ||
1035 error "$LFS setstripe failed"
1036 cancel_lru_locks osc
1037 multiop_bg_pause $DIR/$tdir/$tfile-1 Ow8192_yc ||
1038 error "mulitop Ow8192_yc failed"
1041 multiop_bg_pause $DIR/$tdir/$tfile-2 Ow8192_c ||
1042 error "mulitop Ow8192_c failed"
1047 kill -USR1 $MULTI_PID1
1050 kill -USR1 $MULTI_PID2
1053 lctl set_param fail_loc=0x0
1055 [ $rc1 -eq 0 -o $rc2 -eq 0 ] &&
1056 error_ignore bz5494 "multiop didn't fail fsync: $rc1 or close: $rc2" ||
1059 dmesg | grep "dirty page discard:" ||
1060 error "no discarded dirty page found!"
1062 run_test 24b "test dirty page discard due to client eviction"
1064 test_26a() { # was test_26 bug 5921 - evict dead exports by pinger
1065 # this test can only run from a client on a separate node.
1066 remote_ost || { skip "local OST" && return 0; }
1067 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1068 remote_mds || { skip "local MDS" && return 0; }
1070 if [ $(facet_host mgs) = $(facet_host ost1) ]; then
1071 skip "msg and ost1 are at the same node"
1075 check_timeout || return 1
1077 # OBD_FAIL_PTLRPC_DROP_RPC 0x505
1078 do_facet client lctl set_param fail_loc=0x505
1079 local before=$(date +%s)
1082 # evictor takes PING_EVICT_TIMEOUT + 3 * PING_INTERVAL to evict.
1083 # But if there's a race to start the evictor from various obds,
1084 # the loser might have to wait for the next ping.
1085 sleep $((TIMEOUT * 2 + TIMEOUT * 3 / 4))
1086 do_facet client lctl set_param fail_loc=0x0
1087 do_facet client df > /dev/null
1089 local oscs=$(lctl dl | awk '/-osc-/ {print $4}')
1090 check_clients_evicted $before ${oscs[@]}
1091 check_clients_full 10 ${oscs[@]}
1093 run_test 26a "evict dead exports"
1095 test_26b() { # bug 10140 - evict dead exports by pinger
1096 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1098 if [ $(facet_host mgs) = $(facet_host ost1) ]; then
1099 skip "msg and ost1 are at the same node"
1103 check_timeout || return 1
1105 zconf_mount `hostname` $MOUNT2 ||
1106 { error "Failed to mount $MOUNT2"; return 2; }
1107 sleep 1 # wait connections being established
1109 local MDS_NEXP=$(do_facet $SINGLEMDS lctl get_param -n mdt.${mds1_svc}.num_exports | cut -d' ' -f2)
1110 local OST_NEXP=$(do_facet ost1 lctl get_param -n obdfilter.${ost1_svc}.num_exports | cut -d' ' -f2)
1112 echo starting with $OST_NEXP OST and $MDS_NEXP MDS exports
1114 zconf_umount `hostname` $MOUNT2 -f
1116 # PING_INTERVAL max(obd_timeout / 4, 1U)
1117 # PING_EVICT_TIMEOUT (PING_INTERVAL * 6)
1119 # evictor takes PING_EVICT_TIMEOUT + 3 * PING_INTERVAL to evict.
1120 # But if there's a race to start the evictor from various obds,
1121 # the loser might have to wait for the next ping.
1122 # = 9 * PING_INTERVAL + PING_INTERVAL
1123 # = 10 PING_INTERVAL = 10 obd_timeout / 4 = 2.5 obd_timeout
1124 # let's wait $((TIMEOUT * 3)) # bug 19887
1126 wait_client_evicted ost1 $OST_NEXP $((TIMEOUT * 3)) || \
1127 error "Client was not evicted by ost" rc=1
1128 wait_client_evicted $SINGLEMDS $MDS_NEXP $((TIMEOUT * 3)) || \
1129 error "Client was not evicted by mds"
1131 run_test 26b "evict dead exports"
1135 writemany -q -a $DIR/$tdir/$tfile 0 5 &
1138 local save_FAILURE_MODE=$FAILURE_MODE
1140 facet_failover $SINGLEMDS
1141 #define OBD_FAIL_OSC_SHUTDOWN 0x407
1142 do_facet $SINGLEMDS lctl set_param fail_loc=0x80000407
1143 # need to wait for reconnect
1144 echo waiting for fail_loc
1145 wait_update_facet $SINGLEMDS "lctl get_param -n fail_loc" "-2147482617"
1146 facet_failover $SINGLEMDS
1147 #no crashes allowed!
1148 kill -USR1 $CLIENT_PID
1151 FAILURE_MODE=$save_FAILURE_MODE
1153 run_test 27 "fail LOV while using OSC's"
1155 test_28() { # bug 6086 - error adding new clients
1156 do_facet client mcreate $DIR/$tfile || return 1
1157 drop_bl_callback_once "chmod 0777 $DIR/$tfile" ||
1158 echo "evicted as expected"
1159 #define OBD_FAIL_MDS_CLIENT_ADD 0x12f
1160 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000012f"
1161 # fail once (evicted), reconnect fail (fail_loc), ok
1162 client_up || (sleep 10; client_up) || (sleep 10; client_up) || error "reconnect failed"
1164 fail $SINGLEMDS # verify MDS last_rcvd can be loaded
1166 run_test 28 "handle error adding new clients (bug 6086)"
1168 test_29a() { # bug 22273 - error adding new clients
1169 #define OBD_FAIL_TGT_CLIENT_ADD 0x711
1170 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000711"
1171 # fail abort so client will be new again
1172 fail_abort $SINGLEMDS
1173 client_up || error "reconnect failed"
1174 wait_osc_import_state $SINGLEMDS ost FULL
1177 run_test 29a "error adding new clients doesn't cause LBUG (bug 22273)"
1179 test_29b() { # bug 22273 - error adding new clients
1180 #define OBD_FAIL_TGT_CLIENT_ADD 0x711
1181 do_facet ost1 "lctl set_param fail_loc=0x80000711"
1182 # fail abort so client will be new again
1184 client_up || error "reconnect failed"
1187 run_test 29b "error adding new clients doesn't cause LBUG (bug 22273)"
1191 # put a load of file creates/writes/deletes
1192 writemany -q $DIR/$tdir/$tfile 0 5 &
1194 echo writemany pid $CLIENT_PID
1198 # wait for client to reconnect to MDS
1203 # client process should see no problems even though MDS went down
1205 kill -USR1 $CLIENT_PID
1208 echo writemany returned $rc
1209 #these may fail because of eviction due to slow AST response.
1211 error_ignore bz13652 "writemany returned rc $rc" || true
1213 run_test 50 "failover MDS under load"
1216 #define OBD_FAIL_MDS_SYNC_CAPA_SL 0x1310
1217 do_facet ost1 lctl set_param fail_loc=0x00001310
1220 # put a load of file creates/writes/deletes
1221 writemany -q $DIR/$tdir/$tfile 0 5 &
1225 facet_failover $SINGLEMDS
1226 # failover at various points during recovery
1227 SEQ="1 5 10 $(seq $TIMEOUT 5 $(($TIMEOUT+10)))"
1228 echo will failover at $SEQ
1231 #echo failover in $i sec
1232 log "$TESTNAME: failover in $i sec"
1234 facet_failover $SINGLEMDS
1236 # client process should see no problems even though MDS went down
1237 # and recovery was interrupted
1239 kill -USR1 $CLIENT_PID
1242 echo writemany returned $rc
1244 error_ignore bz13652 "writemany returned rc $rc" || true
1246 run_test 51 "failover MDS during recovery"
1249 do_facet client "mkdir -p $DIR/$tdir"
1250 do_facet client "writemany -q -a $DIR/$tdir/$tfile 300 5" &
1252 echo writemany pid $CLIENT_PID
1257 wait $CLIENT_PID || rc=$?
1258 # active client process should see an EIO for down OST
1259 [ $rc -eq 5 ] && { echo "writemany correctly failed $rc" && return 0; }
1260 # but timing or failover setup may allow success
1261 [ $rc -eq 0 ] && { echo "writemany succeeded" && return 0; }
1262 echo "writemany returned $rc"
1267 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1272 [ $rc -ne 0 ] && { return $rc; }
1273 # wait for client to reconnect to OST
1277 [ $rc -ne 0 ] && { return $rc; }
1284 run_test 52 "failover OST under load"
1286 # test of open reconstruct
1289 drop_mdt_ldlm_reply "openfile -f O_RDWR:O_CREAT -m 0755 $DIR/$tfile" ||\
1292 run_test 53 "touch: drop rep"
1295 zconf_mount $(hostname) $MOUNT2
1297 touch $DIR2/$tfile.1
1299 cat $DIR2/$tfile.missing # save transno = 0, rc != 0 into last_rcvd
1302 ERROR=$(dmesg | egrep "(test 54|went back in time)" | tail -n1 |
1303 grep "went back in time")
1304 [ x"$ERROR" == x ] || error "back in time occured"
1306 run_test 54 "back in time"
1308 # bug 11330 - liblustre application death during I/O locks up OST
1310 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1314 # This test assumes relatively small max_dirty_mb setting
1315 # which we want to walk away from, so just for it we will
1316 # temporarily lower the value
1317 local max_dirty_mb=$(lctl get_param -n osc.*.max_dirty_mb | head -n 1)
1318 lctl set_param -n osc.*.max_dirty_mb=32
1319 stack_trap "lctl set_param osc.*.max_dirty_mb=$max_dirty_mb" EXIT
1321 # Minimum pass speed is 2MBps
1323 # LU-2887/LU-3089 - set min pass speed to 500KBps
1324 [ "$ost1_FSTYPE" = zfs ] && ddtimeout=256
1326 # first dd should be finished quickly
1327 $LFS setstripe -c 1 -i 0 $DIR/$tdir/$tfile-1
1328 dd if=/dev/zero of=$DIR/$tdir/$tfile-1 bs=32M count=4 &
1331 echo "step1: testing ......"
1332 while kill -0 $DDPID 2> /dev/null; do
1334 if [ $count -gt $ddtimeout ]; then
1335 error "dd should be finished!"
1339 echo "(dd_pid=$DDPID, time=$count)successful"
1341 $LFS setstripe -c 1 -i 0 $DIR/$tdir/$tfile-2
1342 #define OBD_FAIL_OST_DROP_REQ 0x21d
1343 do_facet ost1 lctl set_param fail_loc=0x0000021d
1344 # second dd will be never finished
1345 dd if=/dev/zero of=$DIR/$tdir/$tfile-2 bs=32M count=4 &
1348 echo "step2: testing ......"
1349 while [ $count -le $ddtimeout ]; do
1350 if ! kill -0 $DDPID 2> /dev/null; then
1352 error "dd shouldn't be finished! (time=$count)"
1357 echo "(dd_pid=$DDPID, time=$count)successful"
1359 #Recover fail_loc and dd will finish soon
1360 do_facet ost1 lctl set_param fail_loc=0
1362 echo "step3: testing ......"
1363 while kill -0 $DDPID 2> /dev/null; do
1365 if [ $count -gt $((ddtimeout + 440)) ]; then
1366 error "dd should be finished!"
1370 echo "(dd_pid=$DDPID, time=$count)successful"
1374 run_test 55 "ost_brw_read/write drops timed-out read/write request"
1376 test_56() { # b=11277
1377 #define OBD_FAIL_MDS_RESEND 0x136
1379 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000136"
1380 stat $DIR/$tfile || error "stat failed"
1381 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1384 run_test 56 "do not fail on getattr resend"
1387 # no oscs means no client or mdt
1388 while lctl get_param osc.*.* > /dev/null 2>&1; do
1389 : # loop until proc file is removed
1393 test_57() { # bug 10866
1397 #define OBD_FAIL_LPROC_REMOVE 0xB00
1398 lctl set_param fail_loc=0x80000B00
1399 zconf_umount `hostname` $DIR
1400 lctl set_param fail_loc=0x80000B00
1401 fail_abort $SINGLEMDS
1403 lctl set_param fail_loc=0
1405 do_facet client "df $DIR"
1407 run_test 57 "read procfs entries causes kernel crash"
1409 test_58() { # bug 11546
1410 #define OBD_FAIL_MDC_ENQUEUE_PAUSE 0x801
1413 lctl set_param fail_loc=0x80000801
1414 cp $DIR/$tfile /dev/null &
1417 lctl set_param fail_loc=0
1418 drop_bl_callback_once rm -f $DIR/$tfile
1420 # the first 'df' could tigger the eviction caused by
1421 # 'drop_bl_callback_once', and it's normal case.
1422 # but the next 'df' should return successfully.
1423 do_facet client "df $DIR" || do_facet client "df $DIR"
1425 run_test 58 "Eviction in the middle of open RPC reply processing"
1427 test_59() { # bug 10589
1428 zconf_mount `hostname` $MOUNT2 || error "Failed to mount $MOUNT2"
1429 echo $DIR2 | grep -q $MOUNT2 || error "DIR2 is not set properly: $DIR2"
1430 #define OBD_FAIL_LDLM_CANCEL_EVICT_RACE 0x311
1431 lctl set_param fail_loc=0x311
1432 writes=$(LANG=C dd if=/dev/zero of=$DIR2/$tfile count=1 2>&1)
1433 [ $? = 0 ] || error "dd write failed"
1434 writes=$(echo $writes | awk -F '+' '/out/ {print $1}')
1435 lctl set_param fail_loc=0
1437 zconf_umount `hostname` $MOUNT2 -f
1438 reads=$(LANG=C dd if=$DIR/$tfile of=/dev/null 2>&1)
1439 [ $? = 0 ] || error "dd read failed"
1440 reads=$(echo $reads | awk -F '+' '/in/ {print $1}')
1441 [ "$reads" -eq "$writes" ] || error "read" $reads "blocks, must be" $writes
1443 run_test 59 "Read cancel race on client eviction"
1446 local num_files=${COUNT:-5000}
1447 test_mkdir $DIR/$tdir
1449 # Register (and start) changelog
1450 changelog_register || error "changelog_register failed"
1452 # Generate a large number of changelog entries
1453 createmany -o $DIR/$tdir/$tfile $num_files
1457 # Unlink files in the background
1458 unlinkmany $DIR/$tdir/$tfile $num_files &
1462 # Failover the MDS while unlinks are happening
1463 facet_failover $SINGLEMDS
1465 # Wait for unlinkmany to finish
1468 # Check if all the create/unlink events were recorded
1470 local cl_count=$(changelog_dump | grep -c UNLNK)
1471 echo "$cl_count unlinks in changelog"
1473 changelog_deregister || error "changelog_deregister failed"
1474 if ! changelog_users $SINGLEMDS | grep -q "^cl"; then
1475 [ $cl_count -eq $num_files ] ||
1476 error "Recorded $cl_count of $num_files unlinks"
1477 # Also make sure we can clear large changelogs
1478 cl_count=$(changelog_dump | wc -l)
1479 [ $cl_count -le 2 ] ||
1480 error "Changelog not empty: $cl_count entries"
1481 else # If other users, there may be other unlinks in the log
1482 [ $cl_count -ge $num_files ] ||
1483 error "Recorded $cl_count of $num_files unlinks"
1484 changelog_users $SINGLEMDS
1485 echo "other changelog users; can't verify clear"
1488 run_test 60 "Add Changelog entries during MDS failover"
1492 local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS $FSNAME-OST0000)
1493 mdtosc=${mdtosc/-MDT*/-MDT\*}
1494 local cflags="osc.$mdtosc.connect_flags"
1495 do_facet $SINGLEMDS "lctl get_param -n $cflags" |grep -q skip_orphan
1496 [ $? -ne 0 ] && skip "don't have skip orphan feature" && return
1498 mkdir -p $DIR/$tdir || error "mkdir dir $DIR/$tdir failed"
1499 # Set the default stripe of $DIR/$tdir to put the files to ost1
1500 $LFS setstripe -c 1 -i 0 $DIR/$tdir
1502 replay_barrier $SINGLEMDS
1503 createmany -o $DIR/$tdir/$tfile-%d 10
1504 local oid=$(do_facet ost1 "lctl get_param -n \
1505 obdfilter.${ost1_svc}.last_id" | sed -e 's/.*://')
1507 fail_abort $SINGLEMDS
1509 touch $DIR/$tdir/$tfile
1510 local id=$($LFS getstripe $DIR/$tdir/$tfile |
1511 awk '$1 == 0 { print $2 }')
1512 [ $id -le $oid ] && error "the orphan objid was reused, failed"
1517 run_test 61 "Verify to not reuse orphan objects - bug 17025"
1519 # test_62 as seen it b2_1 please do not reuse test_62
1522 # zconf_umount `hostname` $DIR
1523 # #define OBD_FAIL_PTLRPC_DELAY_IMP_FULL 0x516
1524 # lctl set_param fail_loc=0x516
1527 #run_test 62 "Verify connection flags race - bug LU-1716"
1532 #grant lock1, export2
1533 $LFS setstripe -i -0 $DIR2/$tfile || error "setstripe failed"
1534 $MULTIOP $DIR2/$tfile Ow || error "multiop failed"
1536 #define OBD_FAIL_LDLM_BL_EVICT 0x31e
1537 do_facet ost $LCTL set_param fail_loc=0x31e
1538 #get waiting lock2, export1
1539 $MULTIOP $DIR/$tfile Ow &
1541 # let enqueue to get asleep
1545 $MULTIOP $DIR2/$tfile Ow &
1553 do_facet ost $LCTL set_param fail_loc=0
1560 run_test 65 "lock enqueue for destroyed export"
1564 [[ "$MDS1_VERSION" -ge $(version_code 2.7.51) ]] ||
1565 skip "Need MDS version at least 2.7.51"
1567 local list=$(comma_list $(osts_nodes))
1569 # modify dir so that next revalidate would not obtain UPDATE lock
1572 # drop 1 reply with UPDATE lock
1573 mcreate $DIR/$tfile || error "mcreate failed: $?"
1574 drop_mdt_ldlm_reply_once "stat $DIR/$tfile" &
1577 # make the re-sent lock to sleep
1578 #define OBD_FAIL_MDS_RESEND 0x136
1579 do_nodes $list $LCTL set_param fail_loc=0x80000136
1581 #initiate the re-connect & re-send
1582 local mdtname="MDT0000"
1583 local mdccli=$($LCTL dl | grep "${mdtname}-mdc" | awk '{print $4;}')
1584 local conn_uuid=$($LCTL get_param -n mdc.${mdccli}.conn_uuid)
1585 $LCTL set_param "mdc.${mdccli}.import=connection=${conn_uuid}"
1588 #initiate the client eviction while enqueue re-send is in progress
1594 run_test 66 "lock enqueue re-send vs client eviction"
1598 #define OBD_FAIL_PTLRPC_CONNECT_RACE 0x531
1599 $LCTL set_param fail_loc=0x80000531
1601 local mdtname="MDT0000"
1602 local mdccli=$($LCTL dl | grep "${mdtname}-mdc" | awk '{print $4;}')
1603 local conn_uuid=$($LCTL get_param -n mdc.${mdccli}.mds_conn_uuid)
1604 $LCTL set_param "mdc.${mdccli}.import=connection=${conn_uuid}" &
1613 run_test 67 "connect vs import invalidate race"
1615 check_cli_ir_state()
1617 local NODE=${1:-$HOSTNAME}
1619 st=$(do_node $NODE "lctl get_param mgc.*.ir_state |
1620 awk '/imperative_recovery:/ { print \\\$2}'")
1621 [ $st != ON -o $st != OFF -o $st != ENABLED -o $st != DISABLED ] ||
1622 error "Error state $st, must be ENABLED or DISABLED"
1626 check_target_ir_state()
1629 local name=${target}_svc
1630 local recovery_proc=obdfilter.${!name}.recovery_status
1634 st=$(do_facet $target "$LCTL get_param -n $recovery_proc |
1635 awk '/status:/{ print \\\$2}'")
1636 [ x$st = xRECOVERING ] || break
1638 st=$(do_facet $target "lctl get_param -n $recovery_proc |
1639 awk '/IR:/{ print \\\$2}'")
1640 [ $st != ON -o $st != OFF -o $st != ENABLED -o $st != DISABLED ] ||
1641 error "Error state $st, must be ENABLED or DISABLED"
1647 do_facet mgs lctl set_param -n mgs.MGS.live.$FSNAME="state=$1"
1652 local state=$(do_facet mgs "lctl get_param -n mgs.MGS.live.$FSNAME |
1653 awk '/state:/{ print \\\$2 }'")
1657 nidtbl_version_mgs()
1659 local ver=$(do_facet mgs "lctl get_param -n mgs.MGS.live.$FSNAME |
1660 awk '/nidtbl_version:/{ print \\\$2 }'")
1664 # nidtbl_version_client <mds1|client> [node]
1665 nidtbl_version_client()
1668 local node=${2:-$HOSTNAME}
1670 if [ X$cli = Xclient ]; then
1673 local obdtype=${cli/%[0-9]*/}
1674 [ $obdtype != mds ] && error "wrong parameters $cli"
1676 node=$(facet_active_host $cli)
1681 local vers=$(do_node $node "lctl get_param -n mgc.*.ir_state" |
1682 awk "/$cli/{print \$6}" |sort -u)
1684 # in case there are multiple mounts on the client node
1686 [ ${#arr[@]} -ne 1 ] && error "versions on client node mismatch"
1690 nidtbl_versions_match()
1692 [ $(nidtbl_version_mgs) -eq $(nidtbl_version_client ${1:-client}) ]
1695 target_instance_match()
1701 obdtype=${srv/%[0-9]*/}
1712 error "invalid target type" $srv
1717 local target=${srv}_svc
1718 local si=$(do_facet $srv lctl get_param -n $obdname.${!target}.instance)
1719 local ci=$(lctl get_param -n $cliname.${!target}-${cliname}-*.import |
1720 awk '/instance/{ print $2 }' | head -n1)
1722 return $([ $si -eq $ci ])
1727 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1728 { skip "MGS without IR support"; return 0; }
1730 # MDT was just restarted in the previous test, make sure everything
1733 while [ $cnt -gt 0 ]; do
1734 nidtbl_versions_match && break
1740 set_ir_status disabled
1742 local prev_ver=$(nidtbl_version_client client)
1744 local saved_FAILURE_MODE=$FAILURE_MODE
1745 [ $(facet_host mgs) = $(facet_host ost1) ] && FAILURE_MODE="SOFT"
1749 [ $(nidtbl_version_client client) -eq $prev_ver ] ||
1750 error "version must not change due to IR disabled"
1751 target_instance_match ost1 || error "instance mismatch"
1755 FAILURE_MODE=$saved_FAILURE_MODE
1757 run_test 100 "IR: Make sure normal recovery still works w/o IR"
1761 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1762 skip "MGS without IR support"
1766 local ost1_imp=$(get_osc_import_name client ost1)
1768 # disable pinger recovery
1769 lctl set_param -n osc.$ost1_imp.pinger_recov=0
1770 stack_trap "$LCTL set_param -n osc.$ost1_imp.pinger_recov=1" EXIT
1774 target_instance_match ost1 || error "instance mismatch"
1775 nidtbl_versions_match || error "version must match"
1777 run_test 101a "IR: Make sure IR works w/o normal recovery"
1781 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1782 skip "MGS without IR support"
1786 local ost1_imp=$(get_osc_import_name client ost1)
1788 #define OBD_FAIL_OST_PREPARE_DELAY 0x247
1789 do_facet ost1 "$LCTL set_param fail_loc=0x247"
1790 # disable pinger recovery
1791 $LCTL set_param -n osc.$ost1_imp.pinger_recov=0
1792 stack_trap "$LCTL set_param -n osc.$ost1_imp.pinger_recov=1" EXIT
1794 #OST may return EAGAIN if it is not configured yet
1797 run_test 101b "IR: Make sure IR works w/o normal recovery and proceed EAGAIN"
1801 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1802 { skip "MGS without IR support"; return 0; }
1804 local clients=${CLIENTS:-$HOSTNAME}
1811 # let's have a new nidtbl version
1814 # sleep for a while so that clients can see the failure of ost
1815 # it must be MGC_TIMEOUT_MIN_SECONDS + MGC_TIMEOUT_RAND_CENTISEC.
1816 # int mgc_request.c:
1817 # define MGC_TIMEOUT_MIN_SECONDS 5
1818 # define MGC_TIMEOUT_RAND_CENTISEC 0x1ff /* ~500 *
1819 local count=30 # 20 seconds at most
1820 while [ $count -gt 0 ]; do
1821 nidtbl_versions_match && break
1826 nidtbl_versions_match || error "nidtbl mismatch"
1829 old_version=$(nidtbl_version_client client)
1831 zconf_umount_clients $clients $MOUNT || error "Cannot umount client"
1834 combined_mgs_mds && mgsdev=mds1
1835 remount_facet $mgsdev
1838 zconf_mount_clients $clients $MOUNT || error "Cannot mount client"
1841 new_version=$(nidtbl_version_client client)
1842 [ $new_version -lt $old_version ] &&
1843 error "nidtbl version wrong after mgs restarts"
1846 run_test 102 "IR: New client gets updated nidtbl after MGS restart"
1850 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1851 { skip "MGS without IR support"; return 0; }
1853 combined_mgs_mds && skip "needs separate mgs and mds" && return 0
1855 # workaround solution to generate config log on the mds
1861 # We need this test because mds is like a client in IR context.
1862 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS ||
1863 error "MDS should start w/o mgs"
1865 # start mgs and remount mds w/ ir
1866 start mgs $(mgsdevname) $MGS_MOUNT_OPTS
1869 # remount client so that fsdb will be created on the MGS
1870 umount_client $MOUNT || error "umount failed"
1871 mount_client $MOUNT || error "mount failed"
1873 # sleep 30 seconds so the MDS has a chance to detect MGS restarting
1875 while [ $count -gt 0 ]; do
1876 [ $(nidtbl_version_client mds1) -ne 0 ] && break
1881 # after a while, mds should be able to reconnect to mgs and fetch
1882 # up-to-date nidtbl version
1883 nidtbl_versions_match mds1 || error "mds nidtbl mismatch"
1888 run_test 103 "IR: MDS can start w/o MGS and get updated nidtbl later"
1892 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1893 { skip "MGS without IR support"; return 0; }
1898 start ost1 $(ostdevname 1) "$OST_MOUNT_OPTS -onoir" ||
1899 error "OST1 cannot start"
1902 local ir_state=$(check_target_ir_state ost1)
1903 [ $ir_state = "DISABLED" -o $ir_state = "OFF" ] ||
1904 error "ir status on ost1 should be DISABLED"
1906 run_test 104 "IR: ost can disable IR voluntarily"
1910 [ -z "$RCLIENTS" ] && skip "Needs multiple clients" && return 0
1911 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1912 { skip "MGS without IR support"; return 0; }
1916 # get one of the clients from client list
1917 local rcli=$(echo $RCLIENTS |cut -d' ' -f 1)
1919 local mount_opts=${MOUNT_OPTS:+$MOUNT_OPTS,}noir
1920 zconf_umount $rcli $MOUNT || error "umount failed"
1921 zconf_mount $rcli $MOUNT $mount_opts || error "mount failed"
1923 # make sure lustre mount at $rcli disabling IR
1924 local ir_state=$(check_cli_ir_state $rcli)
1925 [ $ir_state = "DISABLED" -o $ir_state = "OFF" ] ||
1926 error "IR state must be DISABLED at $rcli"
1928 # Since the client just mounted, its last_rcvd entry is not on disk.
1929 # Send an RPC so exp_need_sync forces last_rcvd to commit this export
1930 # so the client can reconnect during OST recovery (LU-924, LU-1582)
1931 $LFS setstripe -i 0 $DIR/$tfile
1932 dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 conv=sync
1934 # make sure MGS's state is Partial
1935 [ $(get_ir_status) = "partial" ] || error "MGS IR state must be partial"
1938 # make sure IR on ost1 is DISABLED
1939 local ir_state=$(check_target_ir_state ost1)
1940 [ $ir_state = "DISABLED" -o $ir_state = "OFF" ] ||
1941 error "IR status on ost1 should be DISABLED"
1943 # remount with the default MOUNT_OPTS
1944 zconf_umount $rcli $MOUNT || error "umount failed"
1945 zconf_mount $rcli $MOUNT || error "mount failed"
1947 # make sure MGS's state is full
1948 [ $(get_ir_status) = "full" ] || error "MGS IR status must be full"
1951 # make sure IR on ost1 is ENABLED
1952 local ir_state=$(check_target_ir_state ost1)
1953 [ $ir_state = "ENABLED" -o $ir_state = "ON" ] ||
1954 error "IR status on ost1 should be ENABLED"
1958 run_test 105 "IR: NON IR clients support"
1966 test_106() { # LU-1789
1967 [[ "$MDS1_VERSION" -ge $(version_code 2.3.50) ]] ||
1968 skip "Need MDS version at least 2.3.50"
1970 #define OBD_FAIL_MDC_LIGHTWEIGHT 0x805
1971 $LCTL set_param fail_loc=0x805
1974 trap cleanup_106 EXIT
1976 # enable lightweight flag on mdc connection
1979 local MDS_NEXP=$(do_facet $SINGLEMDS \
1980 lctl get_param -n mdt.${mds1_svc}.num_exports |
1982 $LCTL set_param fail_loc=0
1984 touch $DIR2/$tfile || error "failed to create empty file"
1985 replay_barrier $SINGLEMDS
1987 $LCTL set_param debug=ha
1989 facet_failover $SINGLEMDS
1991 # lightweight goes through LUSTRE_IMP_RECOVER during failover
1992 touch -c $DIR2/$tfile || true
1993 $LCTL dk $TMP/lustre-log-$TESTNAME.log
1994 recovered=$(awk '/MDT0000-mdc-[0-9a-f]*. lwp recover/ { print }' \
1995 $TMP/lustre-log-$TESTNAME.log)
1996 [ -z "$recovered" ] && error "lightweight client was not recovered"
1998 # and all operations performed by lightweight client should be
1999 # synchronous, so the file created before mds restart should be there
2000 $CHECKSTAT -t file $DIR/$tfile || error "file not present"
2005 run_test 106 "lightweight connection support"
2012 # OBD_FAIL_MDS_REINT_NET_REP 0x119
2013 do_facet $SINGLEMDS lctl set_param fail_loc=0x119
2014 multiop $DIR/$tdir D_c &
2016 mkdir $DIR/$tdir/dir_106 &
2018 do_facet $SINGLEMDS lctl set_param fail_loc=0
2021 wait $CLIENT_PID || rc=$?
2022 checkstat -t dir $DIR/$tdir/dir_106 || return 1
2024 kill -USR1 $close_pid
2025 wait $close_pid || return 2
2029 run_test 107 "drop reint reply, then restart MDT"
2033 $LFS setstripe -c 1 -i 0 $DIR/$tdir
2035 dd if=/dev/zero of=$DIR/$tdir/$tfile bs=1M count=256 &
2043 client_up || error "reconnect failed"
2044 rm -f $DIR/$tdir/$tfile
2046 run_test 108 "client eviction don't crash"
2049 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2050 local remote_dir=$DIR/$tdir/remote_dir
2054 #prepare for 110 test, which need set striped dir on remote MDT.
2055 for num in $(seq $MDSCOUNT); do
2057 lctl set_param -n mdt.$FSNAME*.enable_remote_dir=1 \
2062 drop_request "$LFS mkdir -i $mdtidx -c2 $remote_dir" ||
2063 error "lfs mkdir failed"
2064 local diridx=$($LFS getstripe -m $remote_dir)
2065 [ $diridx -eq $mdtidx ] || error "$diridx != $mdtidx"
2067 rm -rf $DIR/$tdir || error "rmdir failed"
2069 run_test 110a "create remote directory: drop client req"
2072 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2073 local remote_dir=$DIR/$tdir/remote_dir
2077 drop_reint_reply "$LFS mkdir -i $mdtidx -c2 $remote_dir" ||
2078 error "lfs mkdir failed"
2080 diridx=$($LFS getstripe -m $remote_dir)
2081 [ $diridx -eq $mdtidx ] || error "$diridx != $mdtidx"
2083 rm -rf $DIR/$tdir || error "rmdir failed"
2085 run_test 110b "create remote directory: drop Master rep"
2088 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2089 local remote_dir=$DIR/$tdir/remote_dir
2093 drop_update_reply $mdtidx "$LFS mkdir -i $mdtidx -c2 $remote_dir" ||
2094 error "lfs mkdir failed"
2096 diridx=$($LFS getstripe -m $remote_dir)
2097 [ $diridx -eq $mdtidx ] || error "$diridx != $mdtidx"
2099 rm -rf $DIR/$tdir || error "rmdir failed"
2101 run_test 110c "create remote directory: drop update rep on slave MDT"
2104 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2105 local remote_dir=$DIR/$tdir/remote_dir
2109 $LFS mkdir -i $MDTIDX -c2 $remote_dir || error "lfs mkdir failed"
2111 drop_request "rm -rf $remote_dir" || error "rm remote dir failed"
2113 rm -rf $DIR/$tdir || error "rmdir failed"
2117 run_test 110d "remove remote directory: drop client req"
2120 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2121 local remote_dir=$DIR/$tdir/remote_dir
2125 $LFS mkdir -i $MDTIDX -c2 $remote_dir || error "lfs mkdir failed"
2126 drop_reint_reply "rm -rf $remote_dir" || error "rm remote dir failed"
2128 rm -rf $DIR/$tdir || error "rmdir failed"
2132 run_test 110e "remove remote directory: drop master rep"
2135 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2136 local remote_dir=$DIR/$tdir/remote_dir
2140 $LFS mkdir -i $MDTIDX -c2 $remote_dir || error "lfs mkdir failed"
2141 drop_update_reply $MDTIDX "rm -rf $remote_dir" ||
2142 error "rm remote dir failed"
2144 rm -rf $DIR/$tdir || error "rmdir failed"
2146 run_test 110f "remove remote directory: drop slave rep"
2149 [[ "$MDS1_VERSION" -ge $(version_code 2.11.0) ]] ||
2150 skip "Need MDS version at least 2.11.0"
2152 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
2155 touch $DIR/$tdir/$tfile
2157 # OBD_FAIL_MDS_REINT_NET_REP 0x119
2158 do_facet mds1 $LCTL set_param fail_loc=0x119
2159 $LFS migrate -m 1 $DIR/$tdir &
2162 do_facet mds1 $LCTL set_param fail_loc=0
2166 mdt_index=$($LFS getstripe -m $DIR/$tdir)
2167 [ $mdt_index == 1 ] || error "$tdir is not on MDT1"
2168 mdt_index=$($LFS getstripe -m $DIR/$tdir/$tfile)
2169 [ $mdt_index == 1 ] || error "$tfile is not on MDT1"
2171 rm -rf $DIR/$tdir || error "rmdir failed"
2173 run_test 110g "drop reply during migration"
2176 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
2177 [[ "$MDS1_VERSION" -ge $(version_code 2.7.56) ]] ||
2178 skip "Need MDS version at least 2.7.56"
2180 local src_dir=$DIR/$tdir/source_dir
2181 local tgt_dir=$DIR/$tdir/target_dir
2185 $LFS mkdir -i $MDTIDX $tgt_dir
2187 dd if=/etc/hosts of=$src_dir/src_file
2188 touch $tgt_dir/tgt_file
2189 drop_update_reply $MDTIDX \
2190 "mrename $src_dir/src_file $tgt_dir/tgt_file" ||
2191 error "mrename failed"
2193 $CHECKSTAT -t file $src_dir/src_file &&
2194 error "src_file present after rename"
2196 diff /etc/hosts $tgt_dir/tgt_file ||
2197 error "file changed after rename"
2200 run_test 110h "drop update reply during cross-MDT file rename"
2203 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
2204 [[ "$MDS1_VERSION" -ge $(version_code 2.7.56) ]] ||
2205 skip "Need MDS version at least 2.7.56"
2207 local src_dir=$DIR/$tdir/source_dir
2208 local tgt_dir=$DIR/$tdir/target_dir
2212 $LFS mkdir -i $MDTIDX $tgt_dir
2214 mkdir $src_dir/src_dir
2215 touch $src_dir/src_dir/a
2216 mkdir $tgt_dir/tgt_dir
2217 drop_update_reply $MDTIDX \
2218 "mrename $src_dir/src_dir $tgt_dir/tgt_dir" ||
2219 error "mrename failed"
2221 $CHECKSTAT -t dir $src_dir/src_dir &&
2222 error "src_dir present after rename"
2224 $CHECKSTAT -t dir $tgt_dir/tgt_dir ||
2225 error "tgt_dir not present after rename"
2227 $CHECKSTAT -t file $tgt_dir/tgt_dir/a ||
2228 error "a not present after rename"
2230 run_test 110i "drop update reply during cross-MDT dir rename"
2233 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
2234 [[ "$MDS1_VERSION" -ge $(version_code 2.7.56) ]] ||
2235 skip "Need MDS version at least 2.7.56"
2237 local remote_dir=$DIR/$tdir/remote_dir
2238 local local_dir=$DIR/$tdir/local_dir
2242 mkdir $DIR/$tdir/local_dir
2243 $LFS mkdir -i $MDTIDX $remote_dir
2245 touch $local_dir/local_file
2246 drop_update_reply $MDTIDX \
2247 "ln $local_dir/local_file $remote_dir/remote_file" ||
2250 $CHECKSTAT -t file $remote_dir/remote_file ||
2251 error "remote not present after ln"
2253 run_test 110j "drop update reply during cross-MDT ln"
2256 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTS"
2257 [[ "$MDS1_VERSION" -ge $(version_code 2.12.55) ]] ||
2258 skip "Need MDS version at least 2.12.55"
2260 stop mds2 || error "stop mds2 failed"
2263 #define OBD_FAIL_FLD_QUERY_REQ 0x1103
2264 do_facet mds2 lctl set_param fail_loc=0x1103
2265 local OPTS="$MDS_MOUNT_OPTS -o abort_recovery"
2266 start mds2 $(mdsdevname 2) $OPTS ||
2267 error "start MDS with abort_recovery should succeed"
2268 do_facet mds2 lctl set_param fail_loc=0
2271 stop mds2 || error "cleanup: stop mds2 failed"
2272 start mds2 $(mdsdevname 2) $MDS_MOUNT_OPTS ||
2273 error "cleanup: start mds2 failed"
2274 zconf_mount $(hostname) $MOUNT || error "cleanup: mount failed"
2275 client_up || error "post-failover df failed"
2277 run_test 110k "FID_QUERY failed during recovery"
2279 # LU-2844 mdt prepare fail should not cause umount oops
2282 [[ "$MDS1_VERSION" -ge $(version_code 2.3.62) ]] ||
2283 skip "Need MDS version at least 2.3.62"
2285 #define OBD_FAIL_MDS_CHANGELOG_INIT 0x151
2286 do_facet $SINGLEMDS lctl set_param fail_loc=0x151
2287 stop $SINGLEMDS || error "stop MDS failed"
2288 start $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) $MDS_MOUNT_OPTS &&
2289 error "start MDS should fail"
2290 do_facet $SINGLEMDS lctl set_param fail_loc=0
2291 start $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) $MDS_MOUNT_OPTS ||
2292 error "start MDS failed"
2294 run_test 111 "mdd setup fail should not cause umount oops"
2298 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
2300 do_facet_random_file client $TMP/$tfile 100K ||
2301 error_noexit "Create random file $TMP/$tfile"
2303 pause_bulk "cp $TMP/$tfile $DIR/$tfile" $TIMEOUT ||
2304 error_noexit "Can't pause_bulk copy"
2307 # expect cmp to succeed, client resent bulk
2308 cmp $TMP/$tfile $DIR/$tfile ||
2309 error_noexit "Wrong data has been written"
2311 error_noexit "Can't remove file"
2314 run_test 112a "bulk resend while orignal request is in progress"
2321 dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1
2322 cancel_lru_locks osc
2325 $LCTL set_param fail_loc=$fail1 fail_val=3
2326 dd of=/dev/null if=$DIR/$tfile bs=4096 count=1 &
2330 set_nodes_failloc "$(osts_nodes)" $fail2
2332 wait $pid || error "dd failed"
2340 local fail_val2=${4:-0}
2346 $LCTL set_param fail_loc=$fail1 fail_val=4
2347 dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 oflag=dsync &
2352 set_nodes_failloc "$(osts_nodes)" $fail2 $fail_val2
2356 [ $error -eq 0 ] && [ $rc -ne 0 ] && error "dd error ($rc)"
2357 [ $error -ne 0 ] && [ $rc -eq 0 ] && error "dd success"
2362 [ "$OST1_VERSION" -lt $(version_code 2.8.50) ] &&
2363 skip "need at least 2.8.50 on OST"
2365 #define OBD_FAIL_PTLRPC_LONG_REQ_UNLINK 0x51b
2366 #define OBD_FAIL_PTLRPC_DROP_BULK 0x51a
2367 test_115_read 0x8000051b 0x8000051a
2369 run_test 115a "read: late REQ MDunlink and no bulk"
2372 [ "$OST1_VERSION" -lt $(version_code 2.8.50) ] &&
2373 skip "need at least 2.8.50 on OST"
2375 #define OBD_FAIL_PTLRPC_LONG_REQ_UNLINK 0x51b
2376 #define OBD_FAIL_OST_ENOSPC 0x215
2378 # pass $OSTCOUNT for the fail_loc to be caught
2379 # appropriately by the IO thread
2380 test_115_write 0x8000051b 0x80000215 1 $OSTCOUNT
2382 run_test 115b "write: late REQ MDunlink and no bulk"
2385 [ "$OST1_VERSION" -lt $(version_code 2.8.50) ] &&
2386 skip "need at least 2.8.50 on OST"
2388 #define OBD_FAIL_PTLRPC_LONG_REPL_UNLINK 0x50f
2389 #define OBD_FAIL_PTLRPC_DROP_BULK 0x51a
2390 test_115_read 0x8000050f 0x8000051a
2392 run_test 115c "read: late Reply MDunlink and no bulk"
2395 [ "$OST1_VERSION" -lt $(version_code 2.8.50) ] &&
2396 skip "need at least 2.8.50 on OST"
2398 #define OBD_FAIL_PTLRPC_LONG_REPL_UNLINK 0x50f
2399 #define OBD_FAIL_OST_ENOSPC 0x215
2400 test_115_write 0x8000050f 0x80000215 0
2402 run_test 115d "write: late Reply MDunlink and no bulk"
2405 [ "$OST1_VERSION" -lt $(version_code 2.8.50) ] &&
2406 skip "need at least 2.8.50 on OST"
2408 #define OBD_FAIL_PTLRPC_LONG_BULK_UNLINK 0x510
2409 #define OBD_FAIL_OST_ALL_REPLY_NET 0x211
2410 test_115_read 0x80000510 0x80000211
2412 run_test 115e "read: late Bulk MDunlink and no reply"
2415 [ "$OST1_VERSION" -lt $(version_code 2.8.50) ] &&
2416 skip "need at least 2.8.50 on OST"
2418 #define OBD_FAIL_PTLRPC_LONG_REQ_UNLINK 0x51b
2419 #define OBD_FAIL_OST_ALL_REPLY_NET 0x211
2420 test_115_read 0x8000051b 0x80000211
2422 run_test 115f "read: late REQ MDunlink and no reply"
2425 [ "$OST1_VERSION" -lt $(version_code 2.8.50) ] &&
2426 skip "need at least 2.8.50 on OST"
2428 #define OBD_FAIL_PTLRPC_LONG_BOTH_UNLINK 0x51c
2429 test_115_read 0x8000051c 0
2431 run_test 115g "read: late REQ MDunlink and Reply MDunlink"
2433 # parameters: fail_loc CMD RC
2440 #define OBD_FAIL_LDLM_CP_CB_WAIT2 0x320
2441 #define OBD_FAIL_LDLM_CP_CB_WAIT3 0x321
2442 #define OBD_FAIL_LDLM_CP_CB_WAIT4 0x322
2443 #define OBD_FAIL_LDLM_CP_CB_WAIT5 0x323
2446 echo -n "** FLOCK REPLY vs. EVICTION race, lock $2"
2447 [ "$1" = "CLEANUP" ] &&
2448 fail=0x80000320 && echo ", $1 cp first"
2449 [ "$1" = "REPLY" ] &&
2450 fail=0x80000321 && echo ", $1 cp first"
2451 [ "$1" = "DEADLOCK CLEANUP" ] &&
2452 fail=0x80000322 && echo " DEADLOCK, CLEANUP cp first"
2453 [ "$1" = "DEADLOCK REPLY" ] &&
2454 fail=0x80000323 && echo " DEADLOCK, REPLY cp first"
2456 if [ x"$2" = x"get" ]; then
2457 #for TEST lock, take a conflict in advance
2458 # sleep longer than evictor to not confuse fail_loc: 2+2+4
2459 echo "** Taking conflict **"
2460 flocks_test 5 set read sleep 10 $DIR/$tfile &
2466 $LCTL set_param fail_loc=$fail
2468 flocks_test 5 $2 write $DIR/$tfile &
2472 echo "** Evicting and re-connecting client **"
2477 if [ x"$2" = x"get" ]; then
2484 # check if the return value is allowed
2485 [ $rc -eq $3 ] && rc=0
2487 $LCTL set_param fail_loc=0
2491 # a lock is taken, unlock vs. cleanup_resource() race for destroying
2492 # the ORIGINAL lock.
2497 flocks_test 5 set write sleep 4 $DIR/$tfile &
2501 # let unlock to sleep in CP CB
2502 $LCTL set_param fail_loc=$1
2505 # let cleanup to cleep in CP CB
2513 $LCTL set_param fail_loc=0
2518 flock_is_enabled || { skip "mount w/o flock enabled" && return; }
2521 test_120_reply "CLEANUP" set 5 || error "SET race failed"
2522 test_120_reply "CLEANUP" get 5 || error "GET race failed"
2523 test_120_reply "CLEANUP" unlock 5 || error "UNLOCK race failed"
2525 test_120_reply "REPLY" set 5 || error "SET race failed"
2526 test_120_reply "REPLY" get 5 || error "GET race failed"
2527 test_120_reply "REPLY" unlock 5 || error "UNLOCK race failed"
2530 test_120_reply "DEADLOCK CLEANUP" set 5 || error "DEADLOCK race failed"
2531 test_120_reply "DEADLOCK REPLY" set 35 || error "DEADLOCK race failed"
2533 test_120_destroy 0x320 || error "unlock-cleanup race failed"
2535 run_test 120 "flock race: completion vs. evict"
2538 local BEFORE=$(date +%s)
2541 # modify dir so that next revalidate would not obtain UPDATE lock
2544 # drop 1 reply with UPDATE lock,
2545 # resend should not create 2nd lock on server
2546 mcreate $DIR/$tfile || error "mcreate failed: $?"
2547 drop_mdt_ldlm_reply_once "stat $DIR/$tfile" || error "stat failed: $?"
2549 # 2 BL AST will be sent to client, both must find the same lock,
2550 # race them to not get EINVAL for 2nd BL AST
2551 #define OBD_FAIL_LDLM_PAUSE_CANCEL2 0x31f
2552 $LCTL set_param fail_loc=0x8000031f
2554 $LCTL set_param ldlm.namespaces.*.early_lock_cancel=0 > /dev/null
2555 chmod 0777 $DIR/$tfile || error "chmod failed: $?"
2556 $LCTL set_param ldlm.namespaces.*.early_lock_cancel=1 > /dev/null
2558 # let the client reconnect
2560 EVICT=$($LCTL get_param mdc.$FSNAME-MDT*.state |
2561 awk -F"[ [,]" '/EVICTED ]$/ { if (mx<$5) {mx=$5;} } END { print mx }')
2563 [ -z "$EVICT" ] || [[ $EVICT -le $BEFORE ]] || error "eviction happened"
2565 run_test 113 "ldlm enqueue dropped reply should not cause deadlocks"
2569 test_mkdir -p $DIR/$tdir
2571 # Prevent interference from layout intent RPCs due to
2572 # asynchronous writeback. These will be tested in 130c below.
2573 do_nodes ${CLIENTS:-$HOSTNAME} sync
2575 # get only LOOKUP lock on $tdir
2576 cancel_lru_locks mdc
2577 ls $DIR/$tdir/$tfile 2>/dev/null
2579 # get getattr by fid on $tdir
2581 # we need to race with unlink, unlink must complete before we will
2582 # take a DLM lock, otherwise unlink will wait until getattr will
2583 # complete; but later than getattr starts so that getattr found
2585 #define OBD_FAIL_MDS_INTENT_DELAY 0x160
2586 set_nodes_failloc "$(mdts_nodes)" 0x80000160
2593 # drop the reply so that resend happens on an unlinked file.
2594 #define OBD_FAIL_MDS_LDLM_REPLY_NET 0x157
2595 set_nodes_failloc "$(mdts_nodes)" 0x80000157
2599 remote_mds_nodsh && skip "remote MDS with nodsh"
2600 [[ "$MDS1_VERSION" -ge $(version_code 2.7.2) ]] ||
2601 skip "Need server version newer than 2.7.1"
2605 wait $T130_PID || [ $? -eq 0 ] && error "stat should fail"
2608 run_test 130a "enqueue resend on not existing file"
2611 remote_mds_nodsh && skip "remote MDS with nodsh"
2612 [[ "$MDS1_VERSION" -ge $(version_code 2.7.2) ]] ||
2613 skip "Need server version newer than 2.7.1"
2616 # let the reply to be dropped
2619 #define OBD_FAIL_SRV_ENOENT 0x217
2620 set_nodes_failloc "$(mdts_nodes)" 0x80000217
2622 wait $T130_PID || [ $? -eq 0 ] && error "stat should fail"
2625 run_test 130b "enqueue resend on a stale inode"
2628 remote_mds_nodsh && skip "remote MDS with nodsh" && return
2630 do_nodes ${CLIENTS:-$HOSTNAME} sync
2631 echo XXX > $DIR/$tfile
2633 cancel_lru_locks mdc
2635 # Trigger writeback on $tfile.
2637 # we need to race with unlink, unlink must complete before we will
2638 # take a DLM lock, otherwise unlink will wait until intent will
2639 # complete; but later than intent starts so that intent found
2641 #define OBD_FAIL_MDS_INTENT_DELAY 0x160
2642 set_nodes_failloc "$(mdts_nodes)" 0x80000160
2649 # drop the reply so that resend happens on an unlinked file.
2650 #define OBD_FAIL_MDS_LDLM_REPLY_NET 0x157
2651 set_nodes_failloc "$(mdts_nodes)" 0x80000157
2653 # let the reply to be dropped
2656 #define OBD_FAIL_SRV_ENOENT 0x217
2657 set_nodes_failloc "$(mdts_nodes)" 0x80000217
2663 run_test 130c "layout intent resend on a stale inode"
2666 local before=$(date +%s)
2669 mount_client $MOUNT2 || error "mount filed"
2672 # get a lock on client so that export would reach the stale list
2673 $LFS setstripe -i 0 $DIR/$tfile || error "setstripe failed"
2674 dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 conv=fsync ||
2677 #define OBD_FAIL_OST_PAUSE_PUNCH 0x236
2678 do_facet ost1 $LCTL set_param fail_val=120 fail_loc=0x80000236
2680 $TRUNCATE $DIR/$tfile 100 &
2683 dd if=/dev/zero of=$DIR2/$tfile bs=4096 count=1 conv=notrunc ||
2687 umount_client $MOUNT2
2689 evict=$(do_facet client $LCTL get_param \
2690 osc.$FSNAME-OST0000-osc-*/state |
2691 awk -F"[ [,]" '/EVICTED ]$/ { if (t<$5) {t=$5;} } END { print t }')
2693 [ -z "$evict" ] || [[ $evict -le $before ]] ||
2694 (do_facet client $LCTL get_param \
2695 osc.$FSNAME-OST0000-osc-*/state;
2696 error "eviction happened: $evict before:$before")
2698 run_test 132 "long punch"
2701 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
2704 # get a lock on client so that export would reach the stale list
2705 $LFS setstripe -i 0 $DIR/$tfile || error "setstripe failed"
2706 dd if=/dev/zero of=$DIR/$tfile count=1 || error "dd failed"
2708 # another IO under the same lock
2709 #define OBD_FAIL_OSC_DELAY_IO 0x414
2710 $LCTL set_param fail_loc=0x80000414
2711 $LCTL set_param fail_val=4 fail_loc=0x80000414
2712 dd if=/dev/zero of=$DIR/$tfile count=1 conv=notrunc oflag=dsync &
2716 #define OBD_FAIL_LDLM_BL_EVICT 0x31e
2717 set_nodes_failloc "$(osts_nodes)" 0x8000031e
2721 wait $pid && error "dd succeeded"
2724 run_test 131 "IO vs evict results to IO under staled lock"
2727 local list=$(comma_list $(mdts_nodes))
2729 local t=$((TIMEOUT * 2))
2732 flock $DIR/$tfile -c "echo bl lock;sleep $t;echo bl flock unlocked" &
2734 multiop_bg_pause $DIR/$tfile O_jc || return 1
2737 #define OBD_FAIL_MDS_LDLM_REPLY_NET 0x157
2738 do_nodes $list $LCTL set_param fail_loc=0x80000157
2740 echo "waiting for multiop $PID"
2741 wait $PID || return 2
2747 run_test 133 "don't fail on flock resend"
2750 [ -z "$CLIENTS" ] && skip "Need two or more clients" && return
2751 [ $CLIENTCOUNT -lt 2 ] &&
2752 { skip "Need 2+ clients, have $CLIENTCOUNT" && return; }
2754 mkdir -p $MOUNT/$tdir/1 $MOUNT/$tdir/2 || error "mkdir failed"
2755 touch $MOUNT/$tdir/1/$tfile $MOUNT/$tdir/2/$tfile ||
2756 error "touch failed"
2757 zconf_umount_clients $CLIENTS $MOUNT
2758 zconf_mount_clients $CLIENTS $MOUNT
2760 #define OBD_FAIL_TGT_REPLY_DATA_RACE 0x722
2761 # assume commit interval is 5
2762 do_facet mds1 "$LCTL set_param fail_loc=0x722 fail_val=5"
2764 local -a clients=(${CLIENTS//,/ })
2765 local client1=${clients[0]}
2766 local client2=${clients[1]}
2768 do_node $client1 rm $MOUNT/$tdir/1/$tfile &
2770 do_node $client2 mv $MOUNT/$tdir/2/$tfile $MOUNT/$tdir/2/${tfile}_2 &
2773 wait $rmpid || error "rm failed"
2774 wait $mvpid || error "mv failed"
2777 run_test 134 "race between failover and search for reply data free slot"
2780 [ "$MDS1_VERSION" -lt $(version_code 2.12.51) ] &&
2781 skip "Need MDS version at least 2.12.51"
2784 $LFS setstripe -E 1M -L mdt $DIR/$tdir
2785 # to have parent dir write lock before open/resend
2786 touch $DIR/$tdir/$tfile
2787 #define OBD_FAIL_MDS_LDLM_REPLY_NET 0x157
2788 do_nodes $(comma_list $(mdts_nodes)) $LCTL set_param fail_loc=0x80000157
2789 openfile -f O_RDWR:O_CREAT -m 0755 $DIR/$tdir/$tfile ||
2790 error "Failed to open DOM file"
2792 run_test 135 "DOM: open/create resend to return size"
2795 remote_mds_nodsh && skip "remote MDS with nodsh"
2796 [[ "$MDS1_VERSION" -ge $(version_code 2.12.52) ]] ||
2797 skip "Need MDS version at least 2.12.52"
2799 local mdts=$(comma_list $(mdts_nodes))
2800 local MDT0=$(facet_svc $SINGLEMDS)
2802 local clog=$(do_facet mds1 $LCTL --device $MDT0 changelog_register -n)
2803 [ -n "$clog" ] || error "changelog_register failed"
2804 cl_mask=$(do_facet mds1 $LCTL get_param \
2805 mdd.$MDT0.changelog_mask -n)
2806 changelog_chmask "ALL"
2808 # generate some changelog records to accumulate
2809 test_mkdir -i 0 -c 0 $DIR/$tdir || error "mkdir $tdir failed"
2810 createmany -m $DIR/$tdir/$tfile 10000 ||
2811 error "create $DIR/$tdir/$tfile failed"
2813 local size1=$(do_facet $SINGLEMDS \
2814 $LCTL get_param -n mdd.$MDT0.changelog_size)
2815 echo "Changelog size $size1"
2817 #define OBD_FAIL_LLOG_PURGE_DELAY 0x1318
2818 do_nodes $mdts $LCTL set_param fail_loc=0x1318 fail_val=30
2820 # launch changelog_deregister in background on MDS
2821 do_facet mds1 "nohup $LCTL --device $MDT0 changelog_deregister $clog \
2822 > foo.out 2> foo.err < /dev/null &"
2823 # give time to reach fail_loc
2826 # fail_loc will make MDS sleep in the middle of changelog_deregister
2827 # take this opportunity to abruptly kill MDS
2828 FAILURE_MODE_save=$FAILURE_MODE
2831 FAILURE_MODE=$FAILURE_MODE_save
2833 do_nodes $mdts $LCTL set_param fail_loc=0x0 fail_val=0
2835 local size2=$(do_facet $SINGLEMDS \
2836 $LCTL get_param -n mdd.$MDT0.changelog_size)
2837 echo "Changelog size $size2"
2838 local clog2=$(do_facet $SINGLEMDS "$LCTL get_param -n \
2839 mdd.$MDT0.changelog_users | grep $clog")
2840 echo "After crash, changelog user $clog2"
2842 [ -n "$clog2" -o $size2 -lt $size1 ] ||
2843 error "changelog record count unchanged"
2845 do_facet mds1 $LCTL set_param mdd.$MDT0.changelog_mask=\'$cl_mask\' -n
2847 run_test 136 "changelog_deregister leaving pending records"
2853 dd if=/dev/zero of=$DIR/d1/$tfile bs=4096 count=1
2854 dd if=/dev/zero of=$DIR/d2/$tfile bs=4096 count=1
2855 cancel_lru_locks osc
2857 #define OBD_FAIL_PTLRPC_RESEND_RACE 0x525
2858 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000525"
2860 # RPC1: any reply is to be delayed to disable last_xid logic
2861 ln $DIR/d1/$tfile $DIR/d1/f2 &
2864 # RPC2: setattr1 reply is delayed & resent
2865 # original reply comes to client; the resend get asleep
2866 chmod 666 $DIR/d2/$tfile
2868 # RPC3: setattr2 on the same file; run ahead of RPC2 resend
2869 chmod 777 $DIR/d2/$tfile
2871 # RPC2 resend wakes up
2873 [ $(stat -c "%a" $DIR/d2/$tfile) == 777 ] || error "resend got applied"
2875 run_test 137 "late resend must be skipped if already applied"
2878 remote_mds_nodsh && skip "remote MDS with nodsh"
2879 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2880 [[ "$MDS1_VERSION" -ge $(version_code 2.12.59) ]] ||
2881 skip "Need server version newer than 2.12.59"
2883 zconf_umount_clients $CLIENTS $MOUNT
2885 #define OBD_FAIL_TGT_RECOVERY_CONNECT 0x724
2886 #delay a first step of recovey when MDS waiting clients
2887 #and failing to get osp logs
2888 do_facet $SINGLEMDS $LCTL set_param fail_loc=0x724 fail_val=5
2890 facet_failover $SINGLEMDS
2892 #waiting failover and recovery timer
2893 #the valuse is based on target_recovery_overseer() wait_event timeout
2895 stop $SINGLEMDS || error "stop MDS failed"
2896 do_facet $SINGLEMDS $LCTL set_param fail_loc=0
2897 start $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) ||
2898 error "start MDS failed"
2899 zconf_mount_clients $CLIENTS $MOUNT
2901 run_test 138 "Umount MDT during recovery"
2904 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2905 [ $MDS1_VERSION -lt $(version_code 2.13.50) ] &&
2906 skip "Need MDS version at least 2.13.50"
2908 mdt_dev=$(mdsdevname 1)
2910 stop $SINGLEMDS || error "stop $SINGLEMDS failed"
2912 #define OBD_FAIL_OSP_INVALID_LOGID 0x2106
2913 do_facet $SINGLEMDS $LCTL set_param fail_val=0x68 fail_loc=0x80002106
2914 start $SINGLEMDS $mdt_dev $MDS_MOUNT_OPTS || error "Fail to start MDT"
2916 run_test 139 "corrupted catid won't cause crash"
2919 [ $MDS1_VERSION -lt $(version_code 2.12.58) ] &&
2920 skip "Need MDS version at least 2.13.50"
2922 [ "$SHARED_KEY" = true ] &&
2923 skip "server local client incompatible with SSK keys installed"
2925 slr=$(do_facet mds1 \
2926 $LCTL get_param -n mdt.$FSNAME-MDT0000.local_recovery)
2927 stack_trap "do_facet mds1 $LCTL set_param \
2928 mdt.*.local_recovery=$slr" EXIT
2930 # disable recovery for local clients
2931 # so local clients should be marked with no_recovery flag
2932 do_facet mds1 $LCTL set_param mdt.*.local_recovery=0
2936 cnt=$(do_facet mds1 $LCTL get_param "mdt.*.exports.*.export" |
2937 grep export_flags.*no_recovery | wc -l)
2938 echo "$cnt clients with recovery disabled"
2940 [ $cnt -eq 0 ] && error "no clients with recovery disabled"
2942 # enable recovery for local clients
2943 # so no local clients should be marked with no_recovery flag
2944 do_facet mds1 $LCTL set_param mdt.*.local_recovery=1
2947 cnt=$(do_facet mds1 $LCTL get_param "mdt.*.exports.*.export" |
2948 grep export_flags.*no_recovery | wc -l)
2949 echo "$cnt clients with recovery disabled"
2951 [ $cnt -eq 0 ] || error "$cnt clients with recovery disabled"
2953 run_test 140a "local mount is flagged properly"
2956 [ $MDS1_VERSION -lt $(version_code 2.12.58) ] &&
2957 skip "Need MDS version at least 2.13.50"
2959 [ "$SHARED_KEY" = true ] &&
2960 skip "server local client incompatible with SSK keys installed"
2962 slr=$(do_facet mds1 \
2963 $LCTL get_param -n mdt.$FSNAME-MDT0000.local_recovery)
2964 stack_trap "do_facet mds1 $LCTL set_param \
2965 mdt.*.local_recovery=$slr" EXIT
2967 # disable recovery for local clients
2968 do_facet mds1 $LCTL set_param mdt.*.local_recovery=0
2973 local before=$SECONDS
2975 local after=$SECONDS
2976 (( $after-$before < $TIMEOUT*2 )) ||
2977 error "recovery took too long" $((after-bsfore)) $TIMEOUT
2979 run_test 140b "local mount is excluded from recovery"
2985 [ $PARALLEL == "yes" ] && skip "skip parallel run"
2986 combined_mgs_mds || skip "needs combined MGS/MDT"
2987 ( local_mode || from_build_tree ) &&
2988 skip "cannot run in local mode or from build tree"
2990 # some get_param have a bug to handle dot in param name
2991 do_rpc_nodes $(facet_active_host $SINGLEMDS) cancel_lru_locks MGC
2992 oldc=$(do_facet $SINGLEMDS $LCTL get_param -n \
2993 'ldlm.namespaces.MGC*.lock_count')
2995 do_rpc_nodes $(facet_active_host $SINGLEMDS) cancel_lru_locks MGC
2996 newc=$(do_facet $SINGLEMDS $LCTL get_param -n \
2997 'ldlm.namespaces.MGC*.lock_count')
2999 [ $oldc -eq $newc ] || error "mgc lost locks ($oldc != $newc)"
3002 run_test 141 "do not lose locks on MGS restart"
3005 check_and_cleanup_lustre