require_dsh_mds || exit 0
# Skip these tests
-# bug number: 17466 18857 LU1867
-ALWAYS_EXCEPT="61d 33a 33b 89 $REPLAY_SINGLE_EXCEPT"
+# bug number: 17466 18857 LU-1867 LU-1473
+ALWAYS_EXCEPT="61d 33a 33b 89 62 $REPLAY_SINGLE_EXCEPT"
[ $(facet_fstype $SINGLEMDS) = "zfs" ] &&
-# bug number for skipped test: LU-2342 LU-951
- ALWAYS_EXCEPT="$ALWAYS_EXCEPT 20b 70a 73a"
+# bug number for skipped test: LU-951
+ ALWAYS_EXCEPT="$ALWAYS_EXCEPT 73a"
# 63 min 7 min AT AT AT AT"
[ "$SLOW" = "no" ] && EXCEPT_SLOW="1 2 3 4 6 12 16 44a 44b 65 66 67 68"
+[ $(facet_fstype $SINGLEMDS) = "zfs" ] &&
+# bug number for skipped test: LU-3127
+ ALWAYS_EXCEPT="$ALWAYS_EXCEPT 73b"
+
build_test_filter
check_and_setup_lustre
run_test 20a "|X| open(O_CREAT), unlink, replay, close (test mds_cleanup_orphans)"
test_20b() { # bug 10480
- BEFOREUSED=`df -P $DIR | tail -1 | awk '{ print $3 }'`
-
- dd if=/dev/zero of=$DIR/$tfile bs=4k count=10000 &
- pid=$!
- while [ ! -e $DIR/$tfile ] ; do
- usleep 60 # give dd a chance to start
- done
+ local wait_timeout=$((TIMEOUT * 4))
+ local BEFOREUSED
+ local AFTERUSED
- $GETSTRIPE $DIR/$tfile || return 1
- rm -f $DIR/$tfile || return 2 # make it an orphan
- mds_evict_client
- client_up || client_up || true # reconnect
+ BEFOREUSED=`df -P $DIR | tail -1 | awk '{ print $3 }'`
+ dd if=/dev/zero of=$DIR/$tfile bs=4k count=10000 &
+ pid=$!
+ while [ ! -e $DIR/$tfile ] ; do
+ usleep 60 # give dd a chance to start
+ done
- fail $SINGLEMDS # start orphan recovery
- wait_recovery_complete $SINGLEMDS || error "MDS recovery not done"
- wait_mds_ost_sync || return 3
- AFTERUSED=`df -P $DIR | tail -1 | awk '{ print $3 }'`
- log "before $BEFOREUSED, after $AFTERUSED"
- [ $AFTERUSED -gt $((BEFOREUSED + 20)) ] && \
- error "after $AFTERUSED > before $BEFOREUSED"
- return 0
+ $GETSTRIPE $DIR/$tfile || return 1
+ rm -f $DIR/$tfile || return 2 # make it an orphan
+ mds_evict_client
+ client_up || client_up || true # reconnect
+
+ fail $SINGLEMDS # start orphan recovery
+ wait_recovery_complete $SINGLEMDS || error "MDS recovery not done"
+ wait_delete_completed_mds $wait_timeout || return 3
+ AFTERUSED=$(df -P $DIR | tail -1 | awk '{ print $3 }')
+ log "before $BEFOREUSED, after $AFTERUSED"
+ (( $AFTERUSED > $BEFOREUSED + $(fs_log_size) )) &&
+ error "after $AFTERUSED > before $BEFOREUSED"
+ return 0
}
run_test 20b "write, unlink, eviction, replay, (test mds_cleanup_orphans)"
#b=2477,2532
test_40(){
- $LCTL mark multiop $MOUNT/$tfile OS_c
- multiop $MOUNT/$tfile OS_c &
- PID=$!
- writeme -s $MOUNT/${tfile}-2 &
- WRITE_PID=$!
- sleep 1
- facet_failover $SINGLEMDS
+ # always need connection to MDS to verify layout during IO. LU-2628.
+ lctl get_param mdc.*.connect_flags | grep -q layout_lock &&
+ skip "layout_lock needs MDS connection for IO" && return 0
+
+ $LCTL mark multiop $MOUNT/$tfile OS_c
+ multiop $MOUNT/$tfile OS_c &
+ PID=$!
+ writeme -s $MOUNT/${tfile}-2 &
+ WRITE_PID=$!
+ sleep 1
+ facet_failover $SINGLEMDS
#define OBD_FAIL_MDS_CONNECT_NET 0x117
- do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000117"
- kill -USR1 $PID
- stat1=`count_ost_writes`
- sleep $TIMEOUT
- stat2=`count_ost_writes`
- echo "$stat1, $stat2"
- if [ $stat1 -lt $stat2 ]; then
- echo "writes continuing during recovery"
- RC=0
- else
- echo "writes not continuing during recovery, bug 2477"
- RC=4
- fi
- echo "waiting for writeme $WRITE_PID"
- kill $WRITE_PID
- wait $WRITE_PID
+ do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000117"
+ kill -USR1 $PID
+ stat1=`count_ost_writes`
+ sleep $TIMEOUT
+ stat2=`count_ost_writes`
+ echo "$stat1, $stat2"
+ if [ $stat1 -lt $stat2 ]; then
+ echo "writes continuing during recovery"
+ RC=0
+ else
+ echo "writes not continuing during recovery, bug 2477"
+ RC=4
+ fi
+ echo "waiting for writeme $WRITE_PID"
+ kill $WRITE_PID
+ wait $WRITE_PID
- echo "waiting for multiop $PID"
- wait $PID || return 2
- do_facet client munlink $MOUNT/$tfile || return 3
- do_facet client munlink $MOUNT/${tfile}-2 || return 3
- return $RC
+ echo "waiting for multiop $PID"
+ wait $PID || return 2
+ do_facet client munlink $MOUNT/$tfile || return 3
+ do_facet client munlink $MOUNT/${tfile}-2 || return 3
+ return $RC
}
run_test 40 "cause recovery in ptlrpc, ensure IO continues"
test_44a() { # was test_44
local at_max_saved=0
- local mdcdev=$($LCTL get_param -n devices |
- awk "/ ${FSNAME}-MDT0000-mdc-/ {print \$1}")
+ local mdcdev=$($LCTL dl |
+ awk "/${FSNAME}-MDT0000-mdc-/ {if (\$2 == \"UP\") {print \$1}}")
[ "$mdcdev" ] || return 2
[ $(echo $mdcdev | wc -w) -eq 1 ] ||
{ echo mdcdev=$mdcdev; $LCTL dl; return 3; }
for i in `seq 1 10`; do
echo "$i of 10 ($(date +%s))"
- do_facet $SINGLEMDS "lctl get_param -n mdt.*.mdt.timeouts | grep service"
+ do_facet $SINGLEMDS \
+ "lctl get_param -n md[ts].*.mdt.timeouts | grep service"
#define OBD_FAIL_TGT_CONN_RACE 0x701
do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000701"
# lctl below may fail, it is valid case
run_test 44a "race in target handle connect"
test_44b() {
- local mdcdev=$($LCTL get_param -n devices |
- awk "/ ${FSNAME}-MDT0000-mdc-/ {print \$1}")
+ local mdcdev=$($LCTL dl |
+ awk "/${FSNAME}-MDT0000-mdc-/ {if (\$2 == \"UP\") {print \$1}}")
[ "$mdcdev" ] || return 2
[ $(echo $mdcdev | wc -w) -eq 1 ] ||
{ echo mdcdev=$mdcdev; $LCTL dl; return 3; }
for i in `seq 1 10`; do
echo "$i of 10 ($(date +%s))"
- do_facet $SINGLEMDS "lctl get_param -n mdt.*.mdt.timeouts | grep service"
+ do_facet $SINGLEMDS \
+ "lctl get_param -n md[ts].*.mdt.timeouts | grep service"
#define OBD_FAIL_TGT_DELAY_RECONNECT 0x704
do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000704"
# lctl below may fail, it is valid case
}
run_test 57 "test recovery from llog for setattr op"
+cleanup_58() {
+ zconf_umount `hostname` $MOUNT2
+ trap - EXIT
+}
+
#recovery many mds-ost setattr from llog
test_58a() {
mkdir -p $DIR/$tdir
local orig
local new
+ trap cleanup_58 EXIT
+
large_xattr_enabled &&
orig="$(generate_string $(max_xattr_size))" || orig="bar"
[[ "$new" = "$orig" ]] || return 1
rm -f $DIR/$tdir/$tfile
rmdir $DIR/$tdir
- zconf_umount `hostname` $MOUNT2
+ cleanup_58
}
run_test 58b "test replay of setxattr op"
local orig1
local new
+ trap cleanup_58 EXIT
+
if large_xattr_enabled; then
local xattr_size=$(max_xattr_size)
orig="$(generate_string $((xattr_size / 2)))"
[[ "$new" = "$orig1" ]] || return 4
rm -f $DIR/$tdir/$tfile
rmdir $DIR/$tdir
- zconf_umount $HOSTNAME $MOUNT2
+ cleanup_58
}
run_test 58c "resend/reconstruct setxattr op"
# set duration to 900 because it takes some time to boot node
[ "$FAILURE_MODE" = HARD ] && duration=900
+ local elapsed
local start_ts=$(date +%s)
local cmd="rundbench 1 -t $duration"
local pid=""
DBENCH_LIB=$DBENCH_LIB TESTSUITE=$TESTSUITE TESTNAME=$TESTNAME \
MOUNT=$MOUNT DIR=$DIR/$tdir/\\\$(hostname) LCTL=$LCTL $cmd" &
pid=$!
+
+ #LU-1897 wait for all dbench copies to start
+ while ! check_for_process $clients dbench; do
+ elapsed=$(($(date +%s) - start_ts))
+ if [ $elapsed -gt $duration ]; then
+ killall_process $clients dbench
+ error "dbench failed to start on $clients!"
+ fi
+ sleep 1
+ done
+
log "Started rundbench load pid=$pid ..."
- # give rundbench a chance to start, bug 24118
- sleep 12
- local elapsed=$(($(date +%s) - start_ts))
+ elapsed=$(($(date +%s) - start_ts))
local num_failovers=0
while [ $elapsed -lt $duration ]; do
if ! check_for_process $clients dbench; then
- error_noexit "dbench not running on some of $clients!"
+ error_noexit "dbench stopped on some of $clients!"
killall_process $clients dbench
break
fi
local remote_dir=$DIR/$tdir/remote_dir
mkdir -p $DIR/$tdir
- # OBD_FAIL_UPDATE_OBJ_NET 0x1500
- do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x1500
+ #define OBD_FAIL_UPDATE_OBJ_NET_REP 0x1701
+ do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x1701
$LFS mkdir -i $MDTIDX $remote_dir &
local CLIENT_PID=$!
- do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0
fail mds$((MDTIDX + 1))
local remote_dir=$DIR/$tdir/remote_dir
mkdir -p $DIR/$tdir
- # OBD_FAIL_UPDATE_OBJ_NET 0x1500
- do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x1500
+ #define OBD_FAIL_UPDATE_OBJ_NET_REP 0x1701
+ do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x1701
$LFS mkdir -i $MDTIDX $remote_dir &
local CLIENT_PID=$!
- do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0
fail mds${MDTIDX}
local remote_dir=$DIR/$tdir/remote_dir
mkdir -p $DIR/$tdir
- # OBD_FAIL_UPDATE_OBJ_NET 0x1500
- do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x1500
+ #define OBD_FAIL_UPDATE_OBJ_NET_REP 0x1701
+ do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x1701
$LFS mkdir -i $MDTIDX $remote_dir &
local CLIENT_PID=$!
- do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0
fail mds${MDTIDX}
fail mds$((MDTIDX + 1))
local remote_dir=$DIR/$tdir/remote_dir
mkdir -p $DIR/$tdir
- # OBD_FAIL_UPDATE_OBJ_NET 0x1500
- do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x1500
+ #define OBD_FAIL_UPDATE_OBJ_NET_REP 0x1701
+ do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x1701
$LFS mkdir -i $MDTIDX $remote_dir &
local CLIENT_PID=$!
- do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0
fail mds${MDTIDX},mds$((MDTIDX + 1))
do_facet mds${MDTIDX} lctl set_param fail_loc=0x119
$LFS mkdir -i $MDTIDX $remote_dir &
local CLIENT_PID=$!
- do_facet mds${MDTIDX} lctl set_param fail_loc=0
fail mds${MDTIDX}
do_facet mds${MDTIDX} lctl set_param fail_loc=0x119
$LFS mkdir -i $MDTIDX $remote_dir &
local CLIENT_PID=$!
- do_facet mds${MDTIDX} lctl set_param fail_loc=0
fail mds$((MDTIDX + 1))
do_facet mds${MDTIDX} lctl set_param fail_loc=0x119
$LFS mkdir -i $MDTIDX $remote_dir &
local CLIENT_PID=$!
- do_facet mds${MDTIDX} lctl set_param fail_loc=0
fail mds${MDTIDX}
fail mds$((MDTIDX + 1))
do_facet mds${MDTIDX} lctl set_param fail_loc=0x119
$LFS mkdir -i $MDTIDX $remote_dir &
local CLIENT_PID=$!
- do_facet mds${MDTIDX} lctl set_param fail_loc=0
fail mds${MDTIDX},mds$((MDTIDX + 1))
mkdir -p $DIR/$tdir
$LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
- # OBD_FAIL_UPDATE_OBJ_NET 0x1500
- do_facet mds${MDTIDX} lctl set_param fail_loc=0x1500
+ touch $remote_dir
+ # OBD_FAIL_OBJ_UPDATE_NET_REP 0x1701
+ do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
rmdir $remote_dir &
local CLIENT_PID=$!
- do_facet mds${MDTIDX} lctl set_param fail_loc=0
fail mds$((MDTIDX + 1))
mkdir -p $DIR/$tdir
$LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
- # OBD_FAIL_UPDATE_OBJ_NET 0x1500
- do_facet mds${MDTIDX} lctl set_param fail_loc=0x1500
+ # OBD_FAIL_OBJ_UPDATE_NET_REP 0x1701
+ do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
rmdir $remote_dir &
local CLIENT_PID=$!
- do_facet mds${MDTIDX} lctl set_param fail_loc=0
fail mds${MDTIDX}
mkdir -p $DIR/$tdir
$LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
- # OBD_FAIL_UPDATE_OBJ_NET 0x1500
- do_facet mds${MDTIDX} lctl set_param fail_loc=0x1500
+ # OBD_FAIL_OBJ_UPDATE_NET_REP 0x1701
+ do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
rmdir $remote_dir &
local CLIENT_PID=$!
- do_facet mds${MDTIDX} lctl set_param fail_loc=0
fail mds${MDTIDX}
fail mds$((MDTIDX + 1))
mkdir -p $DIR/$tdir
$LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
- # OBD_FAIL_UPDATE_OBJ_NET 0x1500
- do_facet mds${MDTIDX} lctl set_param fail_loc=0x1500
+ # OBD_FAIL_OBJ_UPDATE_NET_REP 0x1701
+ do_facet mds${MDTIDX} lctl set_param fail_loc=0x1701
rmdir $remote_dir &
local CLIENT_PID=$!
- do_facet mds${MDTIDX} lctl set_param fail_loc=0
fail mds${MDTIDX},mds$((MDTIDX + 1))
do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
rmdir $remote_dir &
local CLIENT_PID=$!
- do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0
fail mds$((MDTIDX + 1))
do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
rmdir $remote_dir &
local CLIENT_PID=$!
- do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0
fail mds${MDTIDX}
fail mds$((MDTIDX + 1))
do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0x119
rmdir $remote_dir &
local CLIENT_PID=$!
- do_facet mds$((MDTIDX + 1)) lctl set_param fail_loc=0
fail mds${MDTIDX},mds$((MDTIDX + 1))
local uuid=$(ostuuid_from_index $i)
for file in f$i all; do
if [[ $dir/$file != $($LFS find --obd $uuid --name $file $dir) ]]; then
- $GETSTRIPE $dir/file
- error wrong stripe: $file, uuid: $uuid
+ $GETSTRIPE $dir/$file
+ error wrong stripe: $file, uuid: $uuid
fi
done
done
# Before failing an OST, get its obd name and index
local varsvc=${ostfail}_svc
local obd=$(do_facet $ostfail lctl get_param -n obdfilter.${!varsvc}.uuid)
- local index=${obd:(-6):1}
+ local index=$(($(facet_number $ostfail) - 1))
echo "Fail $ostfail $obd, display the list of affected files"
shutdown_facet $ostfail || return 2