X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Ftests%2Freplay-ost-single.sh;h=83f790f60453f19749f43dae5867a1c22ec1191d;hb=a9257f57ad27a1b1219cfd8ce29fbfabdca1df84;hp=512bfe62a59d8c3c38e8e696800b9504423f77c9;hpb=85027fe58ce73993d909f1390ff76c8f3085f611;p=fs%2Flustre-release.git diff --git a/lustre/tests/replay-ost-single.sh b/lustre/tests/replay-ost-single.sh index 512bfe6..83f790f 100755 --- a/lustre/tests/replay-ost-single.sh +++ b/lustre/tests/replay-ost-single.sh @@ -1,26 +1,23 @@ #!/bin/bash +# -*- mode: Bash; tab-width: 4; indent-tabs-mode: t; -*- +# vim:shiftwidth=4:softtabstop=4:tabstop=4: set -e -PTLDEBUG=${PTLDEBUG:--1} LUSTRE=${LUSTRE:-$(cd $(dirname $0)/..; echo $PWD)} SETUP=${SETUP:-""} CLEANUP=${CLEANUP:-""} . $LUSTRE/tests/test-framework.sh init_test_env $@ . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh} +init_logging # While we do not use OSTCOUNT=1 setup anymore, # ost1failover_HOST is used #ostfailover_HOST=${ostfailover_HOST:-$ost_HOST} #failover= must be defined in OST_MKFS_OPTIONS if ostfailover_HOST != ost_HOST -remote_ost_nodsh && skip "remote OST with nodsh" && exit 0 - -if [ "$FAILURE_MODE" = "HARD" ] && mixed_ost_devs; then - skip "$0: Several ost services on one ost node are used with FAILURE_MODE=$FAILURE_MODE. " - exit 0 -fi +require_dsh_ost || exit 0 # Tests that fail on uml CPU=`awk '/model/ {print $4}' /proc/cpuinfo` @@ -125,7 +122,7 @@ iozone_bg () { # need to check iozone output on "complete" local iozonelog=$TMP/${TESTSUITE}.iozone.log rm -f $iozonelog - cat $tmppipe | while read line ; do + cat $tmppipe | while read line ; do echo "$line" echo "$line" >>$iozonelog done; @@ -139,11 +136,11 @@ iozone_bg () { fi rm -f $tmppipe rm -f $iozonelog - return $rc + return $rc } test_5() { - [ -z "`which iozone 2> /dev/null`" ] && skip "iozone missing" && return 0 + [ -z "`which iozone 2> /dev/null`" ] && skip_env "iozone missing" && return 0 # striping is -c 1, get min of available local minavail=$(lctl get_param -n osc.*[oO][sS][cC][-_]*.kbytesavail | sort -n | head -1) @@ -159,7 +156,7 @@ test_5() { local pid=$! echo iozone bg pid=$pid - + sleep 8 fail ost1 local rc=0 @@ -181,29 +178,39 @@ test_6() { f=$TDIR/$tfile rm -f $f - sync && sleep 2 && sync # wait for delete thread + sync && sleep 2 && sync # wait for delete thread + + # wait till space is returned, following + # (( $before > $after_dd)) test counting on that + wait_mds_ost_sync || return 4 + wait_destroy_complete || return 5 + before=`kbytesfree` dd if=/dev/urandom bs=4096 count=1280 of=$f || return 28 lfs getstripe $f - get_stripe_info client $f + stripe_index=$(lfs getstripe -i $f) sync - sleep 2 # ensure we have a fresh statfs + sleep 2 # ensure we have a fresh statfs sync -#define OBD_FAIL_MDS_REINT_NET_REP 0x119 - do_facet mds "lctl set_param fail_loc=0x80000119" + + #define OBD_FAIL_MDS_REINT_NET_REP 0x119 + do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000119" after_dd=`kbytesfree` log "before: $before after_dd: $after_dd" (( $before > $after_dd )) || return 1 rm -f $f fail ost$((stripe_index + 1)) + wait_recovery_complete ost$((stripe_index + 1)) || + error "OST$((stripe_index + 1)) recovery not completed" $CHECKSTAT -t file $f && return 2 || true sync # let the delete happen - sleep 5 + wait_mds_ost_sync || return 4 + wait_delete_completed || return 5 after=`kbytesfree` log "before: $before after: $after" - (( $before <= $after + 40 )) || return 3 # take OST logs into account + (( $before <= $after + 40 )) || return 3 # take OST logs into account } run_test 6 "Fail OST before obd_destroy" @@ -211,10 +218,16 @@ test_7() { f=$TDIR/$tfile rm -f $f sync && sleep 5 && sync # wait for delete thread + + # wait till space is returned, following + # (( $before > $after_dd)) test counting on that + wait_mds_ost_sync || return 4 + wait_destroy_complete || return 5 + before=`kbytesfree` dd if=/dev/urandom bs=4096 count=1280 of=$f || return 4 sync - sleep 2 # ensure we have a fresh statfs + sleep 2 # ensure we have a fresh statfs sync after_dd=`kbytesfree` log "before: $before after_dd: $after_dd" @@ -222,16 +235,128 @@ test_7() { replay_barrier ost1 rm -f $f fail ost1 + wait_recovery_complete ost1 || error "OST recovery not done" $CHECKSTAT -t file $f && return 2 || true sync # let the delete happen - sleep 5 + wait_mds_ost_sync || return 4 + wait_delete_completed || return 5 after=`kbytesfree` log "before: $before after: $after" (( $before <= $after + 40 )) || return 3 # take OST logs into account } run_test 7 "Fail OST before obd_destroy" -equals_msg `basename $0`: test complete, cleaning up +test_8a() { + verify=$ROOT/tmp/verify-$$ + dd if=/dev/urandom of=$verify bs=4096 count=1280 || + error "Create verify file failed" +#define OBD_FAIL_OST_DQACQ_NET 0x230 + do_facet ost1 "lctl set_param fail_loc=0x230" + dd if=$verify of=$TDIR/$tfile bs=4096 count=1280 oflag=sync & + ddpid=$! + sleep $TIMEOUT # wait for the io to become redo io + if ! ps -p $ddpid > /dev/null 2>&1; then + error "redo io finished incorrectly" + return 1 + fi + do_facet ost1 "lctl set_param fail_loc=0" + wait $ddpid || true + cancel_lru_locks osc + cmp $verify $TDIR/$tfile || return 2 + rm -f $verify $TDIR/$tfile + message=`dmesg | grep "redo for recoverable error -115"` + [ -z "$message" ] || error "redo error messages found in dmesg" +} +run_test 8a "Verify redo io: redo io when get -EINPROGRESS error" + +test_8b() { + verify=$ROOT/tmp/verify-$$ + dd if=/dev/urandom of=$verify bs=4096 count=1280 || + error "Create verify file failed" +#define OBD_FAIL_OST_DQACQ_NET 0x230 + do_facet ost1 "lctl set_param fail_loc=0x230" + dd if=$verify of=$TDIR/$tfile bs=4096 count=1280 oflag=sync & + ddpid=$! + sleep $TIMEOUT # wait for the io to become redo io + fail ost1 + do_facet ost1 "lctl set_param fail_loc=0" + wait $ddpid || return 1 + cancel_lru_locks osc + cmp $verify $TDIR/$tfile || return 2 + rm -f $verify $TDIR/$tfile +} +run_test 8b "Verify redo io: redo io should success after recovery" + +test_8c() { + verify=$ROOT/tmp/verify-$$ + dd if=/dev/urandom of=$verify bs=4096 count=1280 || + error "Create verify file failed" +#define OBD_FAIL_OST_DQACQ_NET 0x230 + do_facet ost1 "lctl set_param fail_loc=0x230" + dd if=$verify of=$TDIR/$tfile bs=4096 count=1280 oflag=sync & + ddpid=$! + sleep $TIMEOUT # wait for the io to become redo io + ost_evict_client + # allow recovery to complete + sleep $((TIMEOUT + 2)) + do_facet ost1 "lctl set_param fail_loc=0" + wait $ddpid + cancel_lru_locks osc + cmp $verify $TDIR/$tfile && return 2 + rm -f $verify $TDIR/$tfile +} +run_test 8c "Verify redo io: redo io should fail after eviction" + +test_8d() { +#define OBD_FAIL_MDS_DQACQ_NET 0x187 + do_facet $SINGLEMDS "lctl set_param fail_loc=0x187" + # test the non-intent create path + mcreate $TDIR/$tfile & + cpid=$! + sleep $TIMEOUT + if ! ps -p $cpid > /dev/null 2>&1; then + error "mknod finished incorrectly" + return 1 + fi + do_facet $SINGLEMDS "lctl set_param fail_loc=0" + wait $cpid || return 2 + stat $TDIR/$tfile || error "mknod failed" + + rm $TDIR/$tfile + +#define OBD_FAIL_MDS_DQACQ_NET 0x187 + do_facet $SINGLEMDS "lctl set_param fail_loc=0x187" + # test the intent create path + openfile -f O_RDWR:O_CREAT $TDIR/$tfile & + cpid=$! + sleep $TIMEOUT + if ! ps -p $cpid > /dev/null 2>&1; then + error "open finished incorrectly" + return 3 + fi + do_facet $SINGLEMDS "lctl set_param fail_loc=0" + wait $cpid || return 4 + stat $TDIR/$tfile || error "open failed" +} +run_test 8d "Verify redo creation on -EINPROGRESS" + +test_8e() { + sleep 1 # ensure we have a fresh statfs +#define OBD_FAIL_OST_STATFS_EINPROGRESS 0x231 + do_facet ost1 "lctl set_param fail_loc=0x231" + df $MOUNT & + dfpid=$! + sleep $TIMEOUT + if ! ps -p $dfpid > /dev/null 2>&1; then + do_facet ost1 "lctl set_param fail_loc=0" + error "df shouldn't have completed!" + return 1 + fi + do_facet ost1 "lctl set_param fail_loc=0" +} +run_test 8e "Verify that ptlrpc resends request on -EINPROGRESS" + +complete $(basename $0) $SECONDS check_and_cleanup_lustre -[ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG && grep -q FAIL $TESTSUITELOG && exit 1 || true +exit_status