2 # Test multiple failures, AKA Test 17
6 LUSTRE=${LUSTRE:-`dirname $0`/..}
7 . $LUSTRE/tests/test-framework.sh
11 . ${CONFIG:=$LUSTRE/tests/cfg/insanity-local.sh}
13 ALWAYS_EXCEPT="2 4 10 $INSANITY_EXCEPT"
16 [ "$SLOW" = "no" ] && EXCEPT_SLOW=""
19 CLEANUP=${CLEANUP:-""}
23 assert_env mds_HOST MDS_MKFS_OPTS
24 assert_env ost_HOST OST_MKFS_OPTS OSTCOUNT
25 assert_env LIVE_CLIENT FSNAME
28 # This can be a regexp, to allow more clients
29 CLIENTS=${CLIENTS:-"`comma_list $LIVE_CLIENT $FAIL_CLIENTS $EXTRA_CLIENTS`"}
34 # fail clients round robin
36 # list of failable clients
37 FAIL_LIST=($FAIL_CLIENTS)
38 FAIL_NUM=${#FAIL_LIST[*]}
41 DOWN_NUM=0 # number of nodes currently down
43 # set next client to fail
45 FAIL_CLIENT=${FAIL_LIST[$FAIL_NEXT]}
46 FAIL_NEXT=$(( (FAIL_NEXT+1) % FAIL_NUM ))
47 echo "fail $FAIL_CLIENT, next is $FAIL_NEXT"
52 if [ "$FAILURE_MODE" = HARD ]; then
54 while ping -w 3 -c 1 $client > /dev/null 2>&1; do
55 echo "waiting for node $client to fail"
58 elif [ "$FAILURE_MODE" = SOFT ]; then
59 zconf_umount $client $MOUNT -f
65 if [ "$FAILURE_MODE" = HARD ]; then
72 if [ -z "$num" ] || [ "$num" -gt $((FAIL_NUM - DOWN_NUM)) ]; then
73 num=$((FAIL_NUM - DOWN_NUM))
76 if [ -z "$num" ] || [ "$num" -le 0 ]; then
82 for i in `seq $num`; do
85 DOWN_CLIENTS="$DOWN_CLIENTS $client"
86 shutdown_client $client
89 echo "down clients: $DOWN_CLIENTS"
91 for client in $DOWN_CLIENTS; do
94 DOWN_NUM=`echo $DOWN_CLIENTS | wc -w`
98 reintegrate_clients() {
99 for client in $DOWN_CLIENTS; do
100 wait_for_host $client
101 echo "Restarting $client"
102 zconf_mount $client $MOUNT || return 1
109 start ost$1 `ostdevname $1` $OST_MOUNT_OPTS
116 for c in $LIVE_CLIENT $FAIL_CLIENTS; do
117 if echo $DOWN_CLIENTS | grep -q $c; then continue; fi
118 $PDSH $c touch $MOUNT/${c}_$file || return 1
124 for c in $LIVE_CLIENT $FAIL_CLIENTS; do
125 $PDSH $c rm $MOUNT/${c}_$file
130 for c in $LIVE_CLIENT $FAIL_CLIENTS; do
131 echo "$c mkdir $MOUNT/$c"
132 $PDSH $c "mkdir $MOUNT/$c"
133 $PDSH $c "ls -l $MOUNT/$c"
138 for c in $LIVE_CLIENT $FAIL_CLIENTS; do
139 echo "rmdir $MOUNT/$c"
140 $PDSH $LIVE_CLIENT "rmdir $MOUNT/$c"
144 clients_recover_osts() {
146 # do_node $CLIENTS "$LCTL "'--device %OSC_`hostname`_'"${facet}_svc_MNT_client_facet recover"
149 cleanup_and_setup_lustre
151 # 9 Different Failure Modes Combinations
152 echo "Starting Test 17 at `date`"
155 facet_failover $SINGLEMDS
156 echo "Waiting for df pid: $DFPID"
157 wait $DFPID || { echo "df returned $?" && return 1; }
159 facet_failover ost1 || return 4
160 echo "Waiting for df pid: $DFPID"
161 wait $DFPID || { echo "df returned $?" && return 2; }
163 facet_failover ost2 || return 5
164 echo "Waiting for df pid: $DFPID"
165 wait $DFPID || { echo "df returned $?" && return 3; }
168 run_test 0 "Fail all nodes, independently"
170 ############### First Failure Mode ###############
172 echo "Don't do a MDS - MDS Failure Case"
173 echo "This makes no sense"
175 run_test 1 "MDS/MDS failure"
176 ###################################################
178 ############### Second Failure Mode ###############
180 echo "Verify Lustre filesystem is up and running"
181 [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
185 shutdown_facet $SINGLEMDS
186 reboot_facet $SINGLEMDS
188 # prepare for MDS failover
189 change_active $SINGLEMDS
190 reboot_facet $SINGLEMDS
198 echo "Reintegrating OST"
201 start_ost 1 || return 2
204 start $SINGLEMDS `mdsdevname 1` $MDS_MOUNT_OPTS || return $?
208 clients_recover_osts ost1
209 echo "Verify reintegration"
210 client_df || return 1
213 run_test 2 "Second Failure Mode: MDS/OST `date`"
214 ###################################################
217 ############### Third Failure Mode ###############
220 echo "Verify Lustre filesystem is up and running"
221 [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
224 facet_failover $SINGLEMDS
225 wait $DFPID || echo df failed: $?
228 echo "Test Lustre stability after MDS failover"
232 echo "Failing 2 CLIENTS"
236 echo "Test Lustre stability after CLIENT failure"
240 echo "Reintegrating CLIENTS"
241 reintegrate_clients || return 1
243 client_df || return 3
244 sleep 2 # give it a little time for fully recovered before next test
246 run_test 3 "Thirdb Failure Mode: MDS/CLIENT `date`"
247 ###################################################
249 ############### Fourth Failure Mode ###############
251 echo "Fourth Failure Mode: OST/MDS `date`"
257 echo "Test Lustre stability after OST failure"
263 shutdown_facet $SINGLEMDS
264 reboot_facet $SINGLEMDS
266 # prepare for MDS failover
267 change_active $SINGLEMDS
268 reboot_facet $SINGLEMDS
275 echo "Reintegrating OST"
281 start $SINGLEMDS `mdsdevname 1` $MDS_MOUNT_OPTS
286 clients_recover_osts ost1
287 echo "Test Lustre stability after MDS failover"
288 client_df || return 1
290 run_test 4 "Fourth Failure Mode: OST/MDS `date`"
291 ###################################################
293 ############### Fifth Failure Mode ###############
295 echo "Fifth Failure Mode: OST/OST `date`"
298 echo "Verify Lustre filesystem is up and running"
299 [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
308 echo "Test Lustre stability after OST failure"
318 echo "Test Lustre stability after OST failure"
324 echo "Reintegrating OSTs"
330 clients_recover_osts ost1
331 clients_recover_osts ost2
336 client_df || return 2
338 run_test 5 "Fifth Failure Mode: OST/OST `date`"
339 ###################################################
341 ############### Sixth Failure Mode ###############
343 echo "Sixth Failure Mode: OST/CLIENT `date`"
346 echo "Verify Lustre filesystem is up and running"
347 [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
350 client_touch testfile || return 2
357 echo "Test Lustre stability after OST failure"
363 echo "Failing CLIENTs"
367 echo "Test Lustre stability after CLIENTs failure"
373 echo "Reintegrating OST/CLIENTs"
381 echo "Verifying mount"
382 [ -z "$(mounted_lustre_filesystems)" ] && return 3
385 run_test 6 "Sixth Failure Mode: OST/CLIENT `date`"
386 ###################################################
389 ############### Seventh Failure Mode ###############
391 echo "Seventh Failure Mode: CLIENT/MDS `date`"
394 echo "Verify Lustre filesystem is up and running"
395 [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
398 client_touch testfile || return 1
401 echo "Part 1: Failing CLIENT"
405 echo "Test Lustre stability after CLIENTs failure"
407 $PDSH $LIVE_CLIENT "ls -l $MOUNT"
408 $PDSH $LIVE_CLIENT "rm -f $MOUNT/*_testfile"
411 echo "Wait 1 minutes"
415 echo "Verify Lustre filesystem is up and running"
416 [ -z "$(mounted_lustre_filesystems)" ] && return 2
422 facet_failover $SINGLEMDS
425 echo "Test Lustre stability after MDS failover"
426 wait $DFPID || echo "df on down clients fails " || return 1
427 $PDSH $LIVE_CLIENT "ls -l $MOUNT"
428 $PDSH $LIVE_CLIENT "rm -f $MOUNT/*_testfile"
431 echo "Reintegrating CLIENTs"
432 reintegrate_clients || return 2
436 echo "wait 1 minutes"
439 run_test 7 "Seventh Failure Mode: CLIENT/MDS `date`"
440 ###################################################
443 ############### Eighth Failure Mode ###############
445 echo "Eighth Failure Mode: CLIENT/OST `date`"
448 echo "Verify Lustre filesystem is up and running"
449 [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
452 client_touch testfile
455 echo "Failing CLIENTs"
459 echo "Test Lustre stability after CLIENTs failure"
461 $PDSH $LIVE_CLIENT "ls -l $MOUNT"
462 $PDSH $LIVE_CLIENT "rm -f $MOUNT/*_testfile"
465 echo "Wait 1 minutes"
469 echo "Verify Lustre filesystem is up and running"
470 [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
473 client_touch testfile
481 echo "Test Lustre stability after OST failure"
485 #non-failout hangs forever here
486 #$PDSH $LIVE_CLIENT "ls -l $MOUNT"
487 #$PDSH $LIVE_CLIENT "rm -f $MOUNT/*_testfile"
490 echo "Reintegrating CLIENTs/OST"
491 reintegrate_clients || return 3
495 client_df || return 1
496 client_touch testfile2 || return 2
499 echo "Wait 1 minutes"
502 run_test 8 "Eighth Failure Mode: CLIENT/OST `date`"
503 ###################################################
506 ############### Ninth Failure Mode ###############
511 echo "Verify Lustre filesystem is up and running"
512 [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
515 client_touch testfile || return 1
518 echo "Failing CLIENTs"
522 echo "Test Lustre stability after CLIENTs failure"
524 $PDSH $LIVE_CLIENT "ls -l $MOUNT" || return 1
525 $PDSH $LIVE_CLIENT "rm -f $MOUNT/*_testfile" || return 2
528 echo "Wait 1 minutes"
532 echo "Verify Lustre filesystem is up and running"
533 $PDSH $LIVE_CLIENT df $MOUNT || return 3
534 client_touch testfile || return 4
537 echo "Failing CLIENTs"
541 echo "Test Lustre stability after CLIENTs failure"
543 $PDSH $LIVE_CLIENT "ls -l $MOUNT" || return 5
544 $PDSH $LIVE_CLIENT "rm -f $MOUNT/*_testfile" || return 6
547 echo "Reintegrating CLIENTs/CLIENTs"
548 reintegrate_clients || return 7
552 echo "Wait 1 minutes"
555 run_test 9 "Ninth Failure Mode: CLIENT/CLIENT `date`"
556 ###################################################
559 #Run availability after all failures
560 DURATION=${DURATION:-$((2 * 60 * 60))} # 6 hours default
561 LOADTEST=${LOADTEST:-metadata-load.py}
562 $PWD/availability.sh $CONFIG $DURATION $CLIENTS || return 1
564 run_test 10 "Running Availability for 6 hours..."
566 equals_msg `basename $0`: test complete, cleaning up
567 check_and_cleanup_lustre
568 [ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG || true