From 50e3ab3c73cd70f6616a5baf4fa6a4c1a84a7e56 Mon Sep 17 00:00:00 2001
From: rread <rread>
Date: Mon, 12 Jul 2004 22:19:10 +0000
Subject: [PATCH] b=3405
- Fix insanity.sh, still using the insanity-local.sh config by default.
- More cleanups in replay-single.sh. We need to generalize failures a
bit more, so we can choose a different mds each time, if more than one
is available.
---
lustre/tests/cfg/insanity-local.sh | 5 +-
lustre/tests/insanity.sh | 68 +++++++++++--------
lustre/tests/replay-single.sh | 104 ++++++++++++++---------------
3 files changed, 96 insertions(+), 81 deletions(-)
diff --git a/lustre/tests/cfg/insanity-local.sh b/lustre/tests/cfg/insanity-local.sh
index 1e0c6e974b..a4839c80a1 100644
--- a/lustre/tests/cfg/insanity-local.sh
+++ b/lustre/tests/cfg/insanity-local.sh
@@ -1,5 +1,6 @@
-mds_HOST=${mds_HOST:-`hostname`}
-mdsfailover_HOST=${mdsfailover_HOST:-""}
+MDSCOUNT=${MDSCOUNT:-1}
+mds1_HOST=${mds1_HOST:-`hostname`}
+mds1failover_HOST=${mds1failover_HOST:-""}
ost1_HOST=${ost1_HOST:-"`hostname`"}
ost2_HOST=${ost2_HOST:-"`hostname`"}
EXTRA_OSTS=${EXTRA_OSTS:-"`hostname`"}
diff --git a/lustre/tests/insanity.sh b/lustre/tests/insanity.sh
index 68d0ff9282..9e392c774f 100755
--- a/lustre/tests/insanity.sh
+++ b/lustre/tests/insanity.sh
@@ -14,7 +14,7 @@ ALWAYS_EXCEPT="10"
build_test_filter
-assert_env mds_HOST ost1_HOST ost2_HOST client_HOST LIVE_CLIENT
+assert_env MDSCOUNT mds1_HOST ost1_HOST ost2_HOST client_HOST LIVE_CLIENT
####
# Initialize all the ostN_HOST
@@ -109,22 +109,32 @@ reintegrate_clients() {
gen_config() {
rm -f $XMLCONFIG
- add_mds mds --dev $MDSDEV --size $MDSSIZE --journal-size $MDSJOURNALSIZE
-
- if [ ! -z "$mdsfailover_HOST" ]; then
- add_mdsfailover mds --dev $MDSDEV --size $MDSSIZE
+ if [ "$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
- add_lov lov1 mds --stripe_sz $STRIPE_BYTES\
- --stripe_cnt $STRIPES_PER_OBJ --stripe_pattern 0
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 --lov lov1 --path $MOUNT
+ add_client client --mds $MDS --lov lov1 --path $MOUNT
}
setup() {
@@ -134,8 +144,10 @@ setup() {
start ost$i ${REFORMAT} $OSTLCONFARGS
done
[ "$DAEMONFILE" ] && $LCTL debug_daemon start $DAEMONFILE $DAEMONSIZE
- wait_for mds
- start mds $MDSLCONFARGS ${REFORMAT}
+ 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
@@ -144,7 +156,9 @@ setup() {
cleanup() {
zconf_umount $CLIENTS $MOUNT
- stop mds ${FORCE} $MDSLCONFARGS || :
+ for mds in `mds_list`; do
+ stop $mds ${FORCE} $MDSLCONFARGS || :
+ done
for i in `seq $NUMOST`; do
stop ost$i ${REFORMAT} ${FORCE} $OSTLCONFARGS || :
done
@@ -228,7 +242,7 @@ echo "Starting Test 17 at `date`"
test_0() {
echo "Failover MDS"
- facet_failover mds
+ facet_failover mds1
echo "Waiting for df pid: $DFPID"
wait $DFPID || return 1
@@ -259,12 +273,12 @@ test_2() {
client_df
echo "Failing MDS"
- shutdown_facet mds
- reboot_facet mds
+ shutdown_facet mds1
+ reboot_facet mds1
# prepare for MDS failover
- change_active mds
- reboot_facet mds
+ change_active mds1
+ reboot_facet mds1
client_df &
DFPID=$!
@@ -279,8 +293,8 @@ test_2() {
start ost1
echo "Failover MDS"
- wait_for mds
- start mds
+ wait_for mds1
+ start mds1
#Check FS
wait $DFPID
@@ -299,7 +313,7 @@ test_3() {
echo "Verify Lustre filesystem is up and running"
#MDS Portion
- facet_failover mds
+ facet_failover mds1
wait $DFPID || echo df failed: $?
#Check FS
@@ -337,12 +351,12 @@ test_4() {
#MDS Portion
echo "Failing MDS"
- shutdown_facet mds
- reboot_facet mds
+ shutdown_facet mds1
+ reboot_facet mds1
# prepare for MDS failover
- change_active mds
- reboot_facet mds
+ change_active mds1
+ reboot_facet mds1
client_df &
DFPID=$!
@@ -355,8 +369,8 @@ test_4() {
start ost1
echo "Failover MDS"
- wait_for mds
- start mds
+ wait_for mds1
+ start mds1
#Check FS
wait $DFPID
@@ -479,7 +493,7 @@ test_7() {
#MDS Portion
echo "Failing MDS"
- facet_failover mds
+ facet_failover mds1
#Check FS
echo "Test Lustre stability after MDS failover"
diff --git a/lustre/tests/replay-single.sh b/lustre/tests/replay-single.sh
index e3da13baec..2a6cddb303 100755
--- a/lustre/tests/replay-single.sh
+++ b/lustre/tests/replay-single.sh
@@ -34,7 +34,7 @@ gen_config() {
--stripe_cnt $STRIPES_PER_OBJ --stripe_pattern 0
else
add_mds mds1 --dev $MDSDEV --size $MDSSIZE
- if [ ! -z "$mdsfailover_HOST" ]; then
+ if [ ! -z "$mds1failover_HOST" ]; then
add_mdsfailover mds1 --dev $MDSDEV --size $MDSSIZE
fi
add_lov lov1 mds1 --stripe_sz $STRIPE_BYTES \
@@ -110,16 +110,16 @@ test_0b() {
run_test 0b "ensure object created after recover exists. (3284)"
test_1() {
- replay_barrier mds2
+ replay_barrier mds1
mcreate $DIR/$tfile
- fail mds2
+ fail mds1
$CHECKSTAT -t file $DIR/$tfile || return 1
rm $DIR/$tfile
}
run_test 1 "simple create"
test_2a() {
- replay_barrier mds
+ replay_barrier mds1
touch $DIR/$tfile
fail mds1
$CHECKSTAT -t file $DIR/$tfile || return 1
@@ -129,7 +129,7 @@ run_test 2a "touch"
test_2b() {
./mcreate $DIR/$tfile
- replay_barrier mds
+ replay_barrier mds1
touch $DIR/$tfile
fail mds1
$CHECKSTAT -t file $DIR/$tfile || return 1
@@ -138,7 +138,7 @@ test_2b() {
run_test 2b "touch"
test_3a() {
- replay_barrier mds
+ replay_barrier mds1
mcreate $DIR/$tfile
o_directory $DIR/$tfile
fail mds1
@@ -148,7 +148,7 @@ test_3a() {
run_test 3a "replay failed open(O_DIRECTORY)"
test_3b() {
- replay_barrier mds
+ replay_barrier mds1
#define OBD_FAIL_MDS_OPEN_PACK | OBD_FAIL_ONCE
do_facet mds "sysctl -w lustre.fail_loc=0x80000114"
touch $DIR/$tfile
@@ -160,7 +160,7 @@ test_3b() {
run_test 3b "replay failed open -ENOMEM"
test_3c() {
- replay_barrier mds
+ replay_barrier mds1
#define OBD_FAIL_MDS_ALLOC_OBDO | OBD_FAIL_ONCE
do_facet mds "sysctl -w lustre.fail_loc=0x80000128"
touch $DIR/$tfile
@@ -173,7 +173,7 @@ test_3c() {
run_test 3c "replay failed open -ENOMEM"
test_4() {
- replay_barrier mds
+ replay_barrier mds1
for i in `seq 10`; do
echo "tag-$i" > $DIR/$tfile-$i
done
@@ -185,7 +185,7 @@ test_4() {
run_test 4 "|x| 10 open(O_CREAT)s"
test_4b() {
- replay_barrier mds
+ replay_barrier mds1
rm -rf $DIR/$tfile-*
fail mds1
$CHECKSTAT -t file $DIR/$tfile-* && return 1 || true
@@ -195,7 +195,7 @@ run_test 4b "|x| rm 10 files"
# The idea is to get past the first block of precreated files on both
# osts, and then replay.
test_5() {
- replay_barrier mds
+ replay_barrier mds1
for i in `seq 220`; do
echo "tag-$i" > $DIR/$tfile-$i
done
@@ -211,7 +211,7 @@ run_test 5 "|x| 220 open(O_CREAT)"
test_6() {
- replay_barrier mds
+ replay_barrier mds1
mkdir $DIR/$tdir
mcreate $DIR/$tdir/$tfile
fail mds1
@@ -223,7 +223,7 @@ test_6() {
run_test 6 "mkdir + contained create"
test_6b() {
- replay_barrier mds
+ replay_barrier mds1
rm -rf $DIR/$tdir
fail mds1
$CHECKSTAT -t dir $DIR/$tdir && return 1 || true
@@ -232,7 +232,7 @@ run_test 6b "|X| rmdir"
test_7() {
mkdir $DIR/$tdir
- replay_barrier mds
+ replay_barrier mds1
mcreate $DIR/$tdir/$tfile
fail mds1
$CHECKSTAT -t dir $DIR/$tdir || return 1
@@ -242,7 +242,7 @@ test_7() {
run_test 7 "mkdir |X| contained create"
test_8() {
- replay_barrier mds
+ replay_barrier mds1
multiop $DIR/$tfile mo_c &
MULTIPID=$!
sleep 1
@@ -256,7 +256,7 @@ test_8() {
run_test 8 "creat open |X| close"
test_9() {
- replay_barrier mds
+ replay_barrier mds1
mcreate $DIR/$tfile
local old_inum=`ls -i $DIR/$tfile | awk '{print $1}'`
fail mds1
@@ -276,7 +276,7 @@ run_test 9 "|X| create (same inum/gen)"
test_10() {
mcreate $DIR/$tfile
- replay_barrier mds
+ replay_barrier mds1
mv $DIR/$tfile $DIR/$tfile-2
rm -f $DIR/$tfile
fail mds1
@@ -291,7 +291,7 @@ test_11() {
mcreate $DIR/$tfile
echo "old" > $DIR/$tfile
mv $DIR/$tfile $DIR/$tfile-2
- replay_barrier mds
+ replay_barrier mds1
echo "new" > $DIR/$tfile
grep new $DIR/$tfile
grep old $DIR/$tfile-2
@@ -308,7 +308,7 @@ test_12() {
# give multiop a chance to open
sleep 1
rm -f $DIR/$tfile
- replay_barrier mds
+ replay_barrier mds1
kill -USR1 $pid
wait $pid || return 1
@@ -329,7 +329,7 @@ test_13() {
sleep 1
chmod 0 $DIR/$tfile
$CHECKSTAT -p 0 $DIR/$tfile
- replay_barrier mds
+ replay_barrier mds1
fail mds1
kill -USR1 $pid
wait $pid || return 1
@@ -345,7 +345,7 @@ test_14() {
# give multiop a chance to open
sleep 1
rm -f $DIR/$tfile
- replay_barrier mds
+ replay_barrier mds1
kill -USR1 $pid || return 1
wait $pid || return 2
@@ -361,7 +361,7 @@ test_15() {
# give multiop a chance to open
sleep 1
rm -f $DIR/$tfile
- replay_barrier mds
+ replay_barrier mds1
touch $DIR/g11 || return 1
kill -USR1 $pid
wait $pid || return 2
@@ -375,7 +375,7 @@ run_test 15 "open(O_CREAT), unlink |X| touch new, close"
test_16() {
- replay_barrier mds
+ replay_barrier mds1
mcreate $DIR/$tfile
munlink $DIR/$tfile
mcreate $DIR/$tfile-2
@@ -387,7 +387,7 @@ test_16() {
run_test 16 "|X| open(O_CREAT), unlink, touch new, unlink new"
test_17() {
- replay_barrier mds
+ replay_barrier mds1
multiop $DIR/$tfile O_c &
pid=$!
# give multiop a chance to open
@@ -401,7 +401,7 @@ test_17() {
run_test 17 "|X| open(O_CREAT), |replay| close"
test_18() {
- replay_barrier mds
+ replay_barrier mds1
multiop $DIR/$tfile O_tSc &
pid=$!
# give multiop a chance to open
@@ -425,7 +425,7 @@ run_test 18 "|X| open(O_CREAT), unlink, touch new, close, touch, unlink"
# bug 1855 (a simpler form of test_11 above)
test_19() {
- replay_barrier mds
+ replay_barrier mds1
mcreate $DIR/$tfile
echo "old" > $DIR/$tfile
mv $DIR/$tfile $DIR/$tfile-2
@@ -436,7 +436,7 @@ test_19() {
run_test 19 "|X| mcreate, open, write, rename "
test_20() {
- replay_barrier mds
+ replay_barrier mds1
multiop $DIR/$tfile O_tSc &
pid=$!
# give multiop a chance to open
@@ -452,7 +452,7 @@ test_20() {
run_test 20 "|X| open(O_CREAT), unlink, replay, close (test mds_cleanup_orphans)"
test_21() {
- replay_barrier mds
+ replay_barrier mds1
multiop $DIR/$tfile O_tSc &
pid=$!
# give multiop a chance to open
@@ -475,7 +475,7 @@ test_22() {
# give multiop a chance to open
sleep 1
- replay_barrier mds
+ replay_barrier mds1
rm -f $DIR/$tfile
fail mds1
@@ -492,7 +492,7 @@ test_23() {
# give multiop a chance to open
sleep 1
- replay_barrier mds
+ replay_barrier mds1
rm -f $DIR/$tfile
touch $DIR/g11 || return 1
@@ -511,7 +511,7 @@ test_24() {
# give multiop a chance to open
sleep 1
- replay_barrier mds
+ replay_barrier mds1
fail mds1
rm -f $DIR/$tfile
kill -USR1 $pid
@@ -528,7 +528,7 @@ test_25() {
sleep 1
rm -f $DIR/$tfile
- replay_barrier mds
+ replay_barrier mds1
fail mds1
kill -USR1 $pid
wait $pid || return 1
@@ -538,7 +538,7 @@ test_25() {
run_test 25 "open(O_CREAT), unlink, replay, close (test mds_cleanup_orphans)"
test_26() {
- replay_barrier mds
+ replay_barrier mds1
multiop $DIR/$tfile-1 O_tSc &
pid1=$!
multiop $DIR/$tfile-2 O_tSc &
@@ -560,7 +560,7 @@ test_26() {
run_test 26 "|X| open(O_CREAT), unlink two, close one, replay, close one (test mds_cleanup_orphans)"
test_27() {
- replay_barrier mds
+ replay_barrier mds1
multiop $DIR/$tfile-1 O_tSc &
pid1=$!
multiop $DIR/$tfile-2 O_tSc &
@@ -588,7 +588,7 @@ test_28() {
pid2=$!
# give multiop a chance to open
sleep 1
- replay_barrier mds
+ replay_barrier mds1
rm -f $DIR/$tfile-1
rm -f $DIR/$tfile-2
kill -USR1 $pid2
@@ -610,7 +610,7 @@ test_29() {
pid2=$!
# give multiop a chance to open
sleep 1
- replay_barrier mds
+ replay_barrier mds1
rm -f $DIR/$tfile-1
rm -f $DIR/$tfile-2
@@ -635,7 +635,7 @@ test_30() {
rm -f $DIR/$tfile-1
rm -f $DIR/$tfile-2
- replay_barrier mds
+ replay_barrier mds1
fail mds1
kill -USR1 $pid1
wait $pid1 || return 1
@@ -656,7 +656,7 @@ test_31() {
sleep 1
rm -f $DIR/$tfile-1
- replay_barrier mds
+ replay_barrier mds1
rm -f $DIR/$tfile-2
fail mds1
kill -USR1 $pid1
@@ -689,9 +689,9 @@ run_test 32 "close() notices client eviction; close() after client eviction"
# Abort recovery before client complete
test_33() {
- replay_barrier mds
+ replay_barrier mds1
touch $DIR/$tfile
- fail_abort mds
+ fail_abort mds1
# this file should be gone, because the replay was aborted
$CHECKSTAT -t file $DIR/$tfile && return 1
return 0
@@ -705,8 +705,8 @@ test_34() {
sleep 1
rm -f $DIR/$tfile
- replay_barrier mds
- fail_abort mds
+ replay_barrier mds1
+ fail_abort mds1
kill -USR1 $pid
[ -e $DIR/$tfile ] && return 1
sync
@@ -725,7 +725,7 @@ test_35() {
sync
sleep 1
# give a chance to remove from MDS
- fail_abort mds
+ fail_abort mds1
$CHECKSTAT -t file $DIR/$tfile && return 1 || true
}
run_test 35 "test recovery from llog for unlink op"
@@ -733,10 +733,10 @@ run_test 35 "test recovery from llog for unlink op"
# b=2432 resent cancel after replay uses wrong cookie,
# so don't resend cancels
test_36() {
- replay_barrier mds
+ replay_barrier mds1
touch $DIR/$tfile
checkstat $DIR/$tfile
- facet_failover mds
+ facet_failover mds1
cancel_lru_locks MDC
if dmesg | grep "unknown lock cookie"; then
echo "cancel after replay failed"
@@ -755,10 +755,10 @@ test_37() {
sleep 1
rmdir $DIR/$tfile
- replay_barrier mds
+ replay_barrier mds1
# clear the dmesg buffer so we only see errors from this recovery
dmesg -c >/dev/null
- fail_abort mds
+ fail_abort mds1
kill -USR1 $pid
dmesg | grep "mds_unlink_orphan.*error .* unlinking orphan" && return 1
sync
@@ -769,7 +769,7 @@ run_test 37 "abort recovery before client does replay (test mds_cleanup_orphans
test_38() {
createmany -o $DIR/$tfile-%d 800
unlinkmany $DIR/$tfile-%d 0 400
- replay_barrier mds
+ replay_barrier mds1
fail mds1
unlinkmany $DIR/$tfile-%d 400 400
sleep 2
@@ -779,7 +779,7 @@ run_test 38 "test recovery from unlink llog (test llog_gen_rec) "
test_39() {
createmany -o $DIR/$tfile-%d 800
- replay_barrier mds
+ replay_barrier mds1
unlinkmany $DIR/$tfile-%d 0 400
fail mds1
unlinkmany $DIR/$tfile-%d 400 400
@@ -801,7 +801,7 @@ test_40(){
writeme -s $MOUNT/${tfile}-2 &
WRITE_PID=$!
sleep 1
- facet_failover mds
+ facet_failover mds1
#define OBD_FAIL_MDS_CONNECT_NET 0x117
do_facet mds "sysctl -w lustre.fail_loc=0x80000117"
kill -USR1 $PID
@@ -873,11 +873,11 @@ run_test 42 "recovery after ost failure"
# b=2530
# timeout in MDS/OST recovery RPC will LBUG MDS
test_43() {
- replay_barrier mds
+ replay_barrier mds1
# OBD_FAIL_OST_CREATE_NET 0x204
do_facet ost "sysctl -w lustre.fail_loc=0x80000204"
- facet_failover mds
+ facet_failover mds1
df $MOUNT || return 1
sleep 10
do_facet ost "sysctl -w lustre.fail_loc=0"
--
GitLab