init_test_env $@
-. ${CONFIG:=$LUSTRE/tests/cfg/insanity-lmv.sh}
+. ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
+#
+ALWAYS_EXCEPT="10 $INSANITY_EXCEPT"
+
+if [ "$FAILURE_MODE" = "HARD" ]; then
+ mixed_ost_devs && CONFIG_EXCEPTIONS="0 2 4 5 6 8" && \
+ echo -n "Several ost services on one ost node are used with FAILURE_MODE=$FAILURE_MODE. " && \
+ echo "Except the tests: $CONFIG_EXCEPTIONS" && \
+ ALWAYS_EXCEPT="$ALWAYS_EXCEPT $CONFIG_EXCEPTIONS"
+fi
+
+#
+[ "$SLOW" = "no" ] && EXCEPT_SLOW=""
-ALWAYS_EXCEPT="10"
+SETUP=${SETUP:-""}
+CLEANUP=${CLEANUP:-""}
build_test_filter
-assert_env MDSCOUNT mds1_HOST ost1_HOST ost2_HOST client_HOST LIVE_CLIENT
+SINGLECLIENT=${SINGLECLIENT:-$HOSTNAME}
+LIVE_CLIENT=${LIVE_CLIENT:-$SINGLECLIENT}
+FAIL_CLIENTS=${FAIL_CLIENTS:-$RCLIENTS}
-####
-# Initialize all the ostN_HOST
-NUMOST=2
-if [ "$EXTRA_OSTS" ]; then
- for host in $EXTRA_OSTS; do
- NUMOST=$((NUMOST + 1))
- OST=ost$NUMOST
- eval ${OST}_HOST=$host
- done
-fi
+assert_env mds_HOST MDS_MKFS_OPTS
+assert_env ost_HOST OST_MKFS_OPTS OSTCOUNT
+assert_env LIVE_CLIENT FSNAME
+
+remote_mds_nodsh && skip "remote MDS with nodsh" && exit 0
+remote_ost_nodsh && skip "remote OST with nodsh" && exit 0
-# This can be a regexp, to allow more clients
-CLIENTS=${CLIENTS:-"`comma_list $LIVE_CLIENT $FAIL_CLIENTS $EXTRA_CLIENTS`"}
+# FAIL_CLIENTS list should not contain the LIVE_CLIENT
+FAIL_CLIENTS=$(echo " $FAIL_CLIENTS " | sed -re "s/\s+$LIVE_CLIENT\s+/ /g")
DIR=${DIR:-$MOUNT}
fi
}
-reboot_node() {
- NODE=$1
- if [ "$FAILURE_MODE" = HARD ]; then
- $POWER_UP $NODE
- fi
-}
-
fail_clients() {
num=$1
+
+ log "Request clients to fail: ${num}. Num of clients to fail: ${FAIL_NUM}, already failed: $DOWN_NUM"
if [ -z "$num" ] || [ "$num" -gt $((FAIL_NUM - DOWN_NUM)) ]; then
num=$((FAIL_NUM - DOWN_NUM))
fi
if [ -z "$num" ] || [ "$num" -le 0 ]; then
+ log "No clients failed!"
return
fi
echo "down clients: $DOWN_CLIENTS"
for client in $DOWN_CLIENTS; do
- reboot_node $client
+ boot_node $client
done
DOWN_NUM=`echo $DOWN_CLIENTS | wc -w`
client_rmdirs
DOWN_NUM=0
}
-gen_config() {
- rm -f $XMLCONFIG
- if [ "$MDSCOUNT" -gt 1 ]; then
- add_lmv lmv1
- for mds in `mds_list`; do
- MDSDEV=$TMP/${mds}-`hostname`
- add_mds $mds --dev $MDSDEV --size $MDSSIZE --lmv lmv1
- done
- MDS=lmv1
- add_lov_to_lmv lov1 lmv1 --stripe_sz $STRIPE_BYTES \
- --stripe_cnt $STRIPES_PER_OBJ --stripe_pattern 0
- else
- add_mds mds1 --dev $MDSDEV --size $MDSSIZE
- if [ ! -z "$mds1failover_HOST" ]; then
- add_mdsfailover mds1 --dev $MDSDEV --size $MDSSIZE
- fi
- add_lov lov1 mds1 --stripe_sz $STRIPE_BYTES \
- --stripe_cnt $STRIPES_PER_OBJ --stripe_pattern 0
- MDS=mds1_svc
- fi
-
- for i in `seq $NUMOST`; do
- dev=`printf $OSTDEV $i`
- add_ost ost$i --lov lov1 --dev $dev --size $OSTSIZE \
- --journal-size $OSTJOURNALSIZE
- done
-
- add_client client --mds $MDS --lov lov1 --path $MOUNT
-}
-
-setup() {
- rm -rf logs/*
- for i in `seq $NUMOST`; do
- wait_for ost$i
- start ost$i ${REFORMAT} $OSTLCONFARGS
- done
- [ "$DAEMONFILE" ] && $LCTL debug_daemon start $DAEMONFILE $DAEMONSIZE
- for mds in `mds_list`; do
- wait_for $mds
- start $mds $MDSLCONFARGS ${REFORMAT}
- done
- while ! do_node $CLIENTS "ls -d $LUSTRE" > /dev/null; do sleep 5; done
- grep " $MOUNT " /proc/mounts || zconf_mount $CLIENTS $MOUNT
-
-}
-
-cleanup() {
- zconf_umount $CLIENTS $MOUNT
-
- for mds in `mds_list`; do
- stop $mds ${FORCE} $MDSLCONFARGS || :
- done
- for i in `seq $NUMOST`; do
- stop ost$i ${REFORMAT} ${FORCE} $OSTLCONFARGS || :
- done
+start_ost() {
+ start ost$1 `ostdevname $1` $OST_MOUNT_OPTS
}
trap exit INT
# do_node $CLIENTS "$LCTL "'--device %OSC_`hostname`_'"${facet}_svc_MNT_client_facet recover"
}
-node_to_ost() {
- node=$1
- retvar=$2
- for i in `seq $NUMOST`; do
- ostvar="ost${i}_HOST"
- if [ "${!ostvar}" == $node ]; then
- eval $retvar=ost${i}
- return 0
- fi
- done
- echo "No ost found for node; $node"
- return 1
-
-}
-
-
-
-if [ "$ONLY" == "cleanup" ]; then
- cleanup
- exit
-fi
-
-if [ -z "$NOSETUP" ]; then
- gen_config
- setup
-fi
-
-if [ ! -z "$EVAL" ]; then
- eval "$EVAL"
- exit $?
-fi
-
-if [ "$ONLY" == "setup" ]; then
- exit 0
-fi
+check_and_setup_lustre
# 9 Different Failure Modes Combinations
echo "Starting Test 17 at `date`"
test_0() {
- echo "Failover MDS"
- facet_failover mds1
+ facet_failover $SINGLEMDS
echo "Waiting for df pid: $DFPID"
- wait $DFPID || return 1
+ wait $DFPID || { echo "df returned $?" && return 1; }
- echo "Failing OST1"
- facet_failover ost1
- echo "Waiting for df pid: $DFPID"
- wait $DFPID || return 2
-
- echo "Failing OST2"
- facet_failover ost2
- echo "Waiting for df pid: $DFPID"
- wait $DFPID || return 3
+ for i in $(seq $OSTCOUNT) ; do
+ facet_failover ost$i || return 4
+ echo "Waiting for df pid: $DFPID"
+ wait $DFPID || { echo "df returned $?" && return 3; }
+ done
return 0
}
run_test 0 "Fail all nodes, independently"
############### Second Failure Mode ###############
test_2() {
echo "Verify Lustre filesystem is up and running"
+ [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
+
client_df
- echo "Failing MDS"
- shutdown_facet mds1
- reboot_facet mds1
+ shutdown_facet $SINGLEMDS
+ reboot_facet $SINGLEMDS
# prepare for MDS failover
- change_active mds1
- reboot_facet mds1
+ change_active $SINGLEMDS
+ reboot_facet $SINGLEMDS
client_df &
DFPID=$!
sleep 5
- echo "Failing OST"
shutdown_facet ost1
echo "Reintegrating OST"
reboot_facet ost1
wait_for ost1
- start ost1
+ start_ost 1 || return 2
- echo "Failover MDS"
- wait_for mds1
- start mds1
+ wait_for $SINGLEMDS
+ start $SINGLEMDS `mdsdevname 1` $MDS_MOUNT_OPTS || return $?
#Check FS
wait $DFPID
test_3() {
#Create files
echo "Verify Lustre filesystem is up and running"
+ [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
#MDS Portion
- facet_failover mds1
+ facet_failover $SINGLEMDS
wait $DFPID || echo df failed: $?
#Check FS
reintegrate_clients || return 1
client_df || return 3
+ sleep 2 # give it a little time for fully recovered before next test
}
run_test 3 "Thirdb Failure Mode: MDS/CLIENT `date`"
###################################################
echo "Fourth Failure Mode: OST/MDS `date`"
#OST Portion
- echo "Failing OST ost1"
shutdown_facet ost1
#Check FS
echo "Test Lustre stability after OST failure"
- client_df
+ client_df &
+ DFPIDA=$!
+ sleep 5
#MDS Portion
- echo "Failing MDS"
- shutdown_facet mds1
- reboot_facet mds1
+ shutdown_facet $SINGLEMDS
+ reboot_facet $SINGLEMDS
# prepare for MDS failover
- change_active mds1
- reboot_facet mds1
+ change_active $SINGLEMDS
+ reboot_facet $SINGLEMDS
client_df &
- DFPID=$!
+ DFPIDB=$!
sleep 5
#Reintegration
echo "Reintegrating OST"
reboot_facet ost1
wait_for ost1
- start ost1
+ start_ost 1
- echo "Failover MDS"
- wait_for mds1
- start mds1
+ wait_for $SINGLEMDS
+ start $SINGLEMDS `mdsdevname 1` $MDS_MOUNT_OPTS
#Check FS
- wait $DFPID
+ wait $DFPIDA
+ wait $DFPIDB
clients_recover_osts ost1
echo "Test Lustre stability after MDS failover"
client_df || return 1
############### Fifth Failure Mode ###############
test_5() {
+ [ $OSTCOUNT -lt 2 ] && skip "$OSTCOUNT < 2, not enough OSTs" && return 0
+
echo "Fifth Failure Mode: OST/OST `date`"
#Create files
echo "Verify Lustre filesystem is up and running"
+ [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
+
client_df
#OST Portion
- echo "Failing OST"
shutdown_facet ost1
reboot_facet ost1
#Check FS
echo "Test Lustre stability after OST failure"
- client_df
+ client_df &
+ DFPIDA=$!
+ sleep 5
#OST Portion
- echo "Failing OST"
shutdown_facet ost2
reboot_facet ost2
#Check FS
echo "Test Lustre stability after OST failure"
- client_df
+ client_df &
+ DFPIDB=$!
+ sleep 5
#Reintegration
echo "Reintegrating OSTs"
wait_for ost1
- start ost1
+ start_ost 1
wait_for ost2
- start ost2
+ start_ost 2
clients_recover_osts ost1
clients_recover_osts ost2
sleep $TIMEOUT
+ wait $DFPIDA
+ wait $DFPIDB
client_df || return 2
}
run_test 5 "Fifth Failure Mode: OST/OST `date`"
#Create files
echo "Verify Lustre filesystem is up and running"
- client_df || return 1
+ [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
+
+ client_df
client_touch testfile || return 2
#OST Portion
- echo "Failing OST"
shutdown_facet ost1
reboot_facet ost1
#Check FS
echo "Test Lustre stability after OST failure"
- client_df
+ client_df &
+ DFPIDA=$!
+ echo DFPIDA=$DFPIDA
+ sleep 5
#CLIENT Portion
echo "Failing CLIENTs"
#Check FS
echo "Test Lustre stability after CLIENTs failure"
- client_df
+ client_df &
+ DFPIDB=$!
+ echo DFPIDB=$DFPIDB
+ sleep 5
#Reintegration
echo "Reintegrating OST/CLIENTs"
wait_for ost1
- start ost1
- reintegrate_clients
+ start_ost 1
+ reintegrate_clients || return 1
sleep 5
+ wait_remote_prog df $((TIMEOUT * 3 + 10))
+ wait $DFPIDA
+ wait $DFPIDB
+
echo "Verifying mount"
- client_df || return 3
+ [ -z "$(mounted_lustre_filesystems)" ] && return 3
+ client_df
}
run_test 6 "Sixth Failure Mode: OST/CLIENT `date`"
###################################################
#Create files
echo "Verify Lustre filesystem is up and running"
+ [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
+
client_df
client_touch testfile || return 1
#Create files
echo "Verify Lustre filesystem is up and running"
+ [ -z "$(mounted_lustre_filesystems)" ] && return 2
+
client_df
client_rm testfile
#MDS Portion
- echo "Failing MDS"
- facet_failover mds1
+ facet_failover $SINGLEMDS
#Check FS
echo "Test Lustre stability after MDS failover"
#Reintegration
echo "Reintegrating CLIENTs"
- reintegrate_clients
- client_df || return 2
+ reintegrate_clients || return 2
+ client_df
#Sleep
echo "wait 1 minutes"
#Create files
echo "Verify Lustre filesystem is up and running"
+ [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
+
client_df
client_touch testfile
#Create files
echo "Verify Lustre filesystem is up and running"
+ [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
+
client_df
client_touch testfile
#OST Portion
- echo "Failing OST"
shutdown_facet ost1
reboot_facet ost1
#Check FS
echo "Test Lustre stability after OST failure"
- client_df
- $PDSH $LIVE_CLIENT "ls -l $MOUNT"
- $PDSH $LIVE_CLIENT "rm -f $MOUNT/*_testfile"
+ client_df &
+ DFPID=$!
+ sleep 5
+ #non-failout hangs forever here
+ #$PDSH $LIVE_CLIENT "ls -l $MOUNT"
+ #$PDSH $LIVE_CLIENT "rm -f $MOUNT/*_testfile"
#Reintegration
echo "Reintegrating CLIENTs/OST"
- reintegrate_clients
+ reintegrate_clients || return 3
wait_for ost1
- start ost1
+ start_ost 1
+ wait $DFPID
client_df || return 1
client_touch testfile2 || return 2
#Create files
echo "Verify Lustre filesystem is up and running"
+ [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
+
client_df
client_touch testfile || return 1
#Reintegration
echo "Reintegrating CLIENTs/CLIENTs"
- reintegrate_clients
- client_df || return 7
+ reintegrate_clients || return 7
+ client_df
#Sleep
echo "Wait 1 minutes"
}
run_test 10 "Running Availability for 6 hours..."
-equals_msg "Done, cleaning up"
-cleanup
+equals_msg `basename $0`: test complete, cleaning up
+check_and_cleanup_lustre
+[ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG || true
git://git.whamcloud.com / fs / lustre-release.git / blobdiff