X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Ftests%2Freplay-dual.sh;h=c47082e07ea59531d57b35a045b30d0aca115f76;hb=c7a42267f5740d587f726e610695f1c72818a21e;hp=c0a9fe45eb6ae67232c31e28905d5162455dfb58;hpb=a7fb215903f6fc9ea93e85b89d1ef9abcff0cdb7;p=fs%2Flustre-release.git diff --git a/lustre/tests/replay-dual.sh b/lustre/tests/replay-dual.sh index c0a9fe4..c47082e 100755 --- a/lustre/tests/replay-dual.sh +++ b/lustre/tests/replay-dual.sh @@ -2,8 +2,8 @@ set -e -# bug number: 13129 13129 10124 -ALWAYS_EXCEPT="2 3 15c $REPLAY_DUAL_EXCEPT" +# bug number: 10124 +ALWAYS_EXCEPT="15c $REPLAY_DUAL_EXCEPT" SAVE_PWD=$PWD PTLDEBUG=${PTLDEBUG:--1} @@ -26,11 +26,17 @@ init_test_env $@ remote_mds_nodsh && skip "remote MDS with nodsh" && exit 0 -[ "$SLOW" = "no" ] && EXCEPT_SLOW="1 2 3 4 5 14" +[ "$SLOW" = "no" ] && EXCEPT_SLOW="21b" build_test_filter -cleanup_and_setup_lustre +check_and_setup_lustre +MOUNTED=$(mounted_lustre_filesystems) +if ! $(echo $MOUNTED | grep -w -q $MOUNT2); then + zconf_mount $HOSTNAME $MOUNT2 + MOUNTED2=yes +fi + assert_DIR rm -rf $DIR/[df][0-9]* @@ -244,7 +250,7 @@ test_13() { } run_test 13 "close resend timeout" -test_14() { +test_14a() { replay_barrier $SINGLEMDS createmany -o $MOUNT1/$tfile- 25 createmany -o $MOUNT2/$tfile-2- 1 @@ -262,7 +268,35 @@ test_14() { zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail" return 0 } -run_test 14 "timeouts waiting for lost client during replay" +run_test 14a "timeouts waiting for lost client during replay" + +test_14b() { + BEFOREUSED=`df -P $DIR | tail -1 | awk '{ print $3 }'` + mkdir -p $MOUNT1/$tdir + replay_barrier $SINGLEMDS + createmany -o $MOUNT1/$tfile- 5 + echo "data" > $MOUNT2/$tdir/$tfile-2 + createmany -o $MOUNT1/$tfile-3- 5 + umount $MOUNT2 + + facet_failover $SINGLEMDS + # expect recovery don't fail due to VBR + df $MOUNT1 || return 1 + + # first 25 files should have been replayed + unlinkmany $MOUNT1/$tfile- 5 || return 2 + unlinkmany $MOUNT1/$tfile-3- 5 || return 3 + + zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail" + # give ost time to process llogs + sleep 3 + AFTERUSED=`df -P $DIR | tail -1 | awk '{ print $3 }'` + log "before $BEFOREUSED, after $AFTERUSED" + [ $AFTERUSED -ne $BEFOREUSED ] && \ + error "after $AFTERUSED > before $BEFOREUSED" && return 4 + return 0 +} +run_test 14b "delete ost orphans if gap occured in objids due to VBR" test_15a() { # was test_15 replay_barrier $SINGLEMDS @@ -376,9 +410,129 @@ test_19() { # Bug 10991 - resend of open request does not fail assertion. } run_test 19 "resend of open request" +test_20() { #16389 + BEFORE=`date +%s` + replay_barrier $SINGLEMDS + touch $MOUNT1/a + touch $MOUNT2/b + umount $MOUNT2 + facet_failover $SINGLEMDS + df $MOUNT1 || return 1 + rm $MOUNT1/a + zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail" + TIER1=$((`date +%s` - BEFORE)) + BEFORE=`date +%s` + replay_barrier $SINGLEMDS + touch $MOUNT1/a + touch $MOUNT2/b + umount $MOUNT2 + facet_failover $SINGLEMDS + df $MOUNT1 || return 1 + rm $MOUNT1/a + zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail" + TIER2=$((`date +%s` - BEFORE)) + [ $TIER2 -ge $((TIER1 * 2)) ] && \ + error "recovery time is growing $TIER2 > $TIER1" + return 0 +} +run_test 20 "recovery time is not increasing" + +# commit on sharing tests +test_21a() { + local param_file=$TMP/$tfile-params + + save_lustre_params $(facet_active_host $SINGLEMDS) "mdt.*.commit_on_sharing" > $param_file + do_facet $SINGLEMDS lctl set_param mdt.*.commit_on_sharing=1 + touch $MOUNT1/$tfile-1 + mv $MOUNT2/$tfile-1 $MOUNT2/$tfile-2 + mv $MOUNT1/$tfile-2 $MOUNT1/$tfile-3 + replay_barrier_nosync $SINGLEMDS + umount $MOUNT2 + + facet_failover $SINGLEMDS + + # all renames are replayed + unlink $MOUNT1/$tfile-3 || return 2 + + zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail" + + do_facet $SINGLEMDS lctl set_param mdt.*.commit_on_sharing=0 + rm -rf $MOUNT1/$tfile-* + restore_lustre_params < $param_file + rm -f $param_file + return 0 +} +run_test 21a "commit on sharing" + +test_21b_sub () { + local mds=$1 + do_node $CLIENT1 rm -f $MOUNT1/$tfile-* + + do_facet $mds sync + do_node $CLIENT1 touch $MOUNT1/$tfile-1 + do_node $CLIENT2 mv $MOUNT1/$tfile-1 $MOUNT1/$tfile-2 + do_node $CLIENT1 mv $MOUNT1/$tfile-2 $MOUNT1/$tfile-3 + + replay_barrier_nosync $mds + shutdown_client $CLIENT2 $MOUNT1 + + facet_failover $mds + + # were renames replayed? + local rc=0 + echo UNLINK $MOUNT1/$tfile-3 + do_node $CLIENT1 unlink $MOUNT1/$tfile-3 || \ + { echo "unlink $tfile-3 fail!" && rc=1; } + + boot_node $CLIENT2 + zconf_mount_clients $CLIENT2 $MOUNT1 || error "mount $CLIENT2 $MOUNT1 fail" + + return $rc +} + +test_21b() { + [ -z "$CLIENTS" ] && skip "Need two or more clients." && return + [ $CLIENTCOUNT -lt 2 ] && \ + { skip "Need two or more clients, have $CLIENTCOUNT" && return; } + + if [ "$FAILURE_MODE" = "HARD" ] && mixed_mdt_devs; then + skip "Several mdt services on one mds node are used with FAILURE_MODE=$FAILURE_MODE. " + return 0 + fi + + + zconf_umount_clients $CLIENTS $MOUNT2 + zconf_mount_clients $CLIENTS $MOUNT1 + + local param_file=$TMP/$tfile-params + + local num=$(get_mds_dir $MOUNT1) + + save_lustre_params $(facet_active_host mds$num) "mdt.*.commit_on_sharing" > $param_file + + # COS enabled + local COS=1 + do_facet mds$num lctl set_param mdt.*.commit_on_sharing=$COS + + test_21b_sub mds$num || error "Not all renames are replayed. COS=$COS" + + # COS disabled (should fail) + COS=0 + do_facet mds$num lctl set_param mdt.*.commit_on_sharing=$COS + + test_21b_sub mds$num && error "Not all renames are replayed. COS=$COS" + + restore_lustre_params < $param_file + rm -f $param_file + return 0 +} +run_test 21b "commit on sharing, two clients" + +# end commit on sharing tests + equals_msg `basename $0`: test complete, cleaning up SLEEP=$((`date +%s` - $NOW)) [ $SLEEP -lt $TIMEOUT ] && sleep $SLEEP +[ "$MOUNTED2" = yes ] && zconf_umount $HOSTNAME $MOUNT2 || true check_and_cleanup_lustre -[ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG || true - +[ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG && grep -q FAIL $TESTSUITELOG && exit 1 || true