X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Ftests%2Freplay-ost-single.sh;h=6a3b9037d92b20d48386c461178d38764bf63cc0;hb=ea1f4f5d0e4496c33a307cc42ee01cdcdf2661fc;hp=398386d27fd8a898d0de8b3c3b156015d708555a;hpb=fbb7ead129258897f5a5d5c9ce28d31fbbe5bca2;p=fs%2Flustre-release.git diff --git a/lustre/tests/replay-ost-single.sh b/lustre/tests/replay-ost-single.sh index 398386d..6a3b903 100755 --- a/lustre/tests/replay-ost-single.sh +++ b/lustre/tests/replay-ost-single.sh @@ -1,62 +1,36 @@ -#!/bin/sh +#!/bin/bash set -e +PTLDEBUG=${PTLDEBUG:--1} LUSTRE=${LUSTRE:-`dirname $0`/..} . $LUSTRE/tests/test-framework.sh - init_test_env $@ - -. ${CONFIG:=$LUSTRE/tests/cfg/lmv.sh} +. ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh} ostfailover_HOST=${ostfailover_HOST:-$ost_HOST} +#failover= must be defined in OST_MKFS_OPTIONS if ostfailover_HOST != ost_HOST -# Skip these tests -ALWAYS_EXCEPT="5" -# test 5 needs a larger fs than what local normally has +# Tests that fail on uml +CPU=`awk '/model/ {print $4}' /proc/cpuinfo` +[ "$CPU" = "UML" ] && EXCEPT="$EXCEPT 6" -gen_config() { - rm -f $XMLCONFIG - if [ "$MDSCOUNT" -gt 1 ]; then - add_lmv lmv1_svc - for mds in `mds_list`; do - MDSDEV=$TMP/${mds}-`hostname` - add_mds $mds --dev $MDSDEV --size $MDSSIZE --lmv lmv1_svc - done - add_lov_to_lmv lov1 lmv1_svc --stripe_sz $STRIPE_BYTES \ - --stripe_cnt $STRIPES_PER_OBJ --stripe_pattern 0 - MDS=lmv1 - else - add_mds mds1 --dev $MDSDEV --size $MDSSIZE - add_lov lov1 mds1 --stripe_sz $STRIPE_BYTES \ - --stripe_cnt $STRIPES_PER_OBJ --stripe_pattern 0 - MDS=mds1_svc +# Skip these tests +# BUG NUMBER: +ALWAYS_EXCEPT="$REPLAY_OST_SINGLE_EXCEPT" - fi +# It is replay-ost-single, after all +OSTCOUNT=1 - add_ost ost --lov lov1 --dev $OSTDEV --size $OSTSIZE --failover - if [ ! -z "$ostfailover_HOST" ]; then - add_ostfailover ost --dev $OSTDEV --size $OSTSIZE - fi - add_client client $MDS --lov lov1 --path $MOUNT +gen_config() { + formatall } cleanup() { - # make sure we are using the primary MDS, so the config log will - # be able to clean up properly. - activeost=`facet_active ost` - if [ $activeost != "ost" ]; then - fail ost - fi - zconf_umount `hostname` $MOUNT - for mds in `mds_list`; do - stop $mds ${FORCE} $MDSLCONFARGS - done - stop ost ${FORCE} --dump cleanup.log + cleanupall } if [ "$ONLY" == "cleanup" ]; then - sysctl -w portals.debug=0 cleanup exit fi @@ -66,33 +40,47 @@ build_test_filter SETUP=${SETUP:-"setup"} CLEANUP=${CLEANUP:-"cleanup"} +test_0a() { + # needs to run during initial client->OST connection + #define OBD_FAIL_OST_ALL_REPLY_NET 0x211 + do_facet ost "sysctl -w lustre.fail_loc=0x80000211" + zconf_mount `hostname` $MOUNT && df $MOUNT || error "0a mount fail" +} + setup() { gen_config - - start ost --reformat $OSTLCONFARGS + start mds1 `mdsdevname 1` $MDS_MOUNT_OPTS + start ost1 `ostdevname 1` $OST_MOUNT_OPTS [ "$DAEMONFILE" ] && $LCTL debug_daemon start $DAEMONFILE $DAEMONSIZE - for mds in `mds_list`; do - start $mds --reformat $MDSLCONFARGS - done - grep " $MOUNT " /proc/mounts || zconf_mount `hostname` $MOUNT + + # this might not mount if we aren't running test 0a + [ -z "`grep " $MOUNT " /proc/mounts`" ] && \ + run_test 0a "target handle mismatch (bug 5317) `date +%H:%M:%S`" + + if [ -z "`grep " $MOUNT " /proc/mounts`" ]; then + zconf_mount `hostname` $MOUNT || error "mount fail" + fi + sleep 5 + + do_facet ost1 "sysctl -w lustre.fail_loc=0" } mkdir -p $DIR $SETUP -test_0() { - fail ost +test_0b() { + fail ost1 cp /etc/profile $DIR/$tfile sync diff /etc/profile $DIR/$tfile rm -f $DIR/$tfile } -run_test 0 "empty replay" +run_test 0b "empty replay" test_1() { date > $DIR/$tfile - fail ost + fail ost1 $CHECKSTAT -t file $DIR/$tfile || return 1 rm -f $DIR/$tfile } @@ -102,7 +90,7 @@ test_2() { for i in `seq 10`; do echo "tag-$i" > $DIR/$tfile-$i done - fail ost + fail ost1 for i in `seq 10`; do grep -q "tag-$i" $DIR/$tfile-$i || error "f2-$i" done @@ -115,7 +103,7 @@ test_3() { dd if=/dev/urandom bs=4096 count=1280 | tee $verify > $DIR/$tfile & ddpid=$! sync & - fail ost + fail ost1 wait $ddpid || return 1 cmp $verify $DIR/$tfile || return 2 rm -f $verify $DIR/$tfile @@ -126,18 +114,19 @@ test_4() { verify=$ROOT/tmp/verify-$$ dd if=/dev/urandom bs=4096 count=1280 | tee $verify > $DIR/$tfile # invalidate cache, so that we're reading over the wire - for i in /proc/fs/lustre/ldlm/namespaces/OSC_*MNT*; do + for i in /proc/fs/lustre/ldlm/namespaces/*-osc-*; do echo -n clear > $i/lru_size done cmp $verify $DIR/$tfile & cmppid=$! - fail ost + fail ost1 wait $cmppid || return 1 rm -f $verify $DIR/$tfile } run_test 4 "Fail OST during read, with verification" test_5() { + [ -z "`which iozone 2> /dev/null`" ] && log "iozone missing" && return FREE=`df -P -h $DIR | tail -n 1 | awk '{ print $3 }'` case $FREE in *T|*G) FREE=1G;; @@ -147,14 +136,17 @@ test_5() { PID=$! sleep 8 - fail ost - wait $PID || return 1 + fail ost1 + wait $PID + RC=$? + log "iozone rc=$RC" rm -f $DIR/$tfile + [ $RC -ne 0 ] && return $RC || true } run_test 5 "Fail OST during iozone" kbytesfree() { - awk '{total+=$1} END {print total}' /proc/fs/lustre/osc/OSC_*MNT*/kbytesfree + awk '{total+=$1} END {print total}' /proc/fs/lustre/osc/*-osc-*/kbytesfree } test_6() { @@ -162,20 +154,22 @@ test_6() { rm -f $f sync && sleep 2 && sync # wait for delete thread before=`kbytesfree` - dd if=/dev/urandom bs=4096 count=1280 of=$f + dd if=/dev/urandom bs=4096 count=1280 of=$f || return 28 + lfs getstripe $f + sync + sleep 2 # ensure we have a fresh statfs + sync #define OBD_FAIL_MDS_REINT_NET_REP 0x119 do_facet mds "sysctl -w lustre.fail_loc=0x80000119" - sync - sleep 1 # ensure we have a fresh statfs after_dd=`kbytesfree` log "before: $before after_dd: $after_dd" (( $before > $after_dd )) || return 1 rm -f $f - fail ost + fail ost1 $CHECKSTAT -t file $f && return 2 || true sync # let the delete happen - sleep 2 + sleep 5 after=`kbytesfree` log "before: $before after: $after" (( $before <= $after + 40 )) || return 3 # take OST logs into account @@ -185,20 +179,22 @@ run_test 6 "Fail OST before obd_destroy" test_7() { f=$DIR/$tfile rm -f $f - sync && sleep 2 && sync # wait for delete thread + sync && sleep 5 && sync # wait for delete thread before=`kbytesfree` - dd if=/dev/urandom bs=4096 count=1280 of=$f + dd if=/dev/urandom bs=4096 count=1280 of=$f || return 4 + sync + sleep 2 # ensure we have a fresh statfs sync after_dd=`kbytesfree` log "before: $before after_dd: $after_dd" (( $before > $after_dd )) || return 1 - replay_barrier ost + replay_barrier ost1 rm -f $f - fail ost + fail ost1 $CHECKSTAT -t file $f && return 2 || true sync # let the delete happen - sleep 2 + sleep 5 after=`kbytesfree` log "before: $before after: $after" (( $before <= $after + 40 )) || return 3 # take OST logs into account