X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Fconf-sanity.sh;h=15e81ee50be97b8922fda61d3303cd760a815cdf;hp=005a2e532810778ae9b5996ec4c728ef027972e7;hb=3e4061862e751dbbbbba273403b56201e705a830;hpb=e4cf956f93a4384d19ea73e601a6651710703492

diff --git a/lustre/tests/conf-sanity.sh b/lustre/tests/conf-sanity.sh
index 005a2e5..15e81ee 100644
--- a/lustre/tests/conf-sanity.sh
+++ b/lustre/tests/conf-sanity.sh
@@ -13,9 +13,15 @@ ONLY=${ONLY:-"$*"}
 
 # bug number for skipped test:
 #               15977
-ALWAYS_EXCEPT=" 39    $CONF_SANITY_EXCEPT"
+ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT"
 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
 
+if [ "$FAILURE_MODE" = "HARD" ]; then
+	CONFIG_EXCEPTIONS="24a " && \
+	echo "Except the tests: $CONFIG_EXCEPTIONS for FAILURE_MODE=$FAILURE_MODE, bug 23573" && \
+	ALWAYS_EXCEPT="$ALWAYS_EXCEPT $CONFIG_EXCEPTIONS"
+fi
+
 SRCDIR=`dirname $0`
 PATH=$PWD/$SRCDIR:$SRCDIR:$SRCDIR/../utils:$PATH
 
@@ -26,20 +32,41 @@ RLUSTRE=${RLUSTRE:-$LUSTRE}
 
 . $LUSTRE/tests/test-framework.sh
 init_test_env $@
-init_logging
+
+# use small MDS + OST size to speed formatting time
+# do not use too small MDSSIZE/OSTSIZE, which affect the default jouranl size
+MDSSIZE=200000
+OSTSIZE=200000
+. ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
+
+if ! combined_mgs_mds; then
+    # bug number for skipped test:    23954
+    ALWAYS_EXCEPT="$ALWAYS_EXCEPT       24b"
+fi
+
 # STORED_MDSSIZE is used in test_18
 if [ -n "$MDSSIZE" ]; then
     STORED_MDSSIZE=$MDSSIZE
 fi
-# use small MDS + OST size to speed formatting time
-MDSSIZE=40000
-OSTSIZE=40000
-. ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
 
+# pass "-E lazy_itable_init" to mke2fs to speed up the formatting time
+for facet in MGS MDS OST; do
+    opts=${facet}_MKFS_OPTS
+    if [[ ${!opts} != *lazy_itable_init* ]]; then
+        eval SAVED_${facet}_MKFS_OPTS=\"${!opts}\"
+        eval ${facet}_MKFS_OPTS=\"${!opts} \
+--mkfsoptions='\\\"-E lazy_itable_init\\\"'\"
+    fi
+done
+
+init_logging
+
+#
 require_dsh_mds || exit 0
 require_dsh_ost || exit 0
+#
+[ "$SLOW" = "no" ] && EXCEPT_SLOW="30a 31 45"
 
-[ "$SLOW" = "no" ] && EXCEPT_SLOW="0 1 2 3 6 7 15 18 24b 25 30 31 32 33 34a 45"
 
 assert_DIR
 
@@ -47,16 +74,25 @@ reformat() {
         formatall
 }
 
-writeconf() {
-	local facet=$SINGLEMDS
-	local dev=${facet}_dev
-	shift
+writeconf1() {
+	local facet=$1
+	local dev=$2
+
 	stop ${facet} -f
 	rm -f ${facet}active
 	# who knows if/where $TUNEFS is installed?  Better reformat if it fails...
-	do_facet ${facet} "$TUNEFS --writeconf ${!dev}" || echo "tunefs failed, reformatting instead" && reformat
+	do_facet ${facet} "$TUNEFS --quiet --writeconf $dev" ||
+		{ echo "tunefs failed, reformatting instead" && reformat_and_config && return 1; }
+	return 0
+}
 
-	gen_config
+writeconf() {
+	# we need ldiskfs
+	load_modules
+	# if writeconf fails anywhere, we reformat everything
+	writeconf1 mds `mdsdevname 1` || return 0
+	writeconf1 ost1 `ostdevname 1` || return 0
+	writeconf1 ost2 `ostdevname 2` || return 0
 }
 
 gen_config() {
@@ -71,16 +107,31 @@ gen_config() {
 
 reformat_and_config() {
 	reformat
+	if ! combined_mgs_mds ; then
+		start_mgs
+	fi
 	gen_config
 }
 
+start_mgs () {
+	echo "start mgs"
+	start mgs $MGSDEV $MGS_MOUNT_OPTS
+}
+
 start_mds() {
 	local facet=$SINGLEMDS
 	# we can not use MDSDEV1 here because SINGLEMDS could be set not to mds1 only
 	local num=$(echo $facet | tr -d "mds")
 	local dev=$(mdsdevname $num)
 	echo "start mds service on `facet_active_host $facet`"
-	start $facet ${dev} $MDS_MOUNT_OPTS || return 94
+	start $facet ${dev} $MDS_MOUNT_OPTS $@ || return 94
+}
+
+start_mgsmds() {
+	if ! combined_mgs_mds ; then
+		start_mgs
+	fi
+	start_mds $@
 }
 
 stop_mds() {
@@ -89,9 +140,15 @@ stop_mds() {
 	stop $SINGLEMDS -f  || return 97
 }
 
+stop_mgs() {
+       echo "stop mgs service on `facet_active_host mgs`"
+       # These tests all use non-failover stop
+       stop mgs -f  || return 97
+}
+
 start_ost() {
 	echo "start ost1 service on `facet_active_host ost1`"
-	start ost1 `ostdevname 1` $OST_MOUNT_OPTS || return 95
+	start ost1 `ostdevname 1` $OST_MOUNT_OPTS $@ || return 95
 }
 
 stop_ost() {
@@ -102,7 +159,7 @@ stop_ost() {
 
 start_ost2() {
 	echo "start ost2 service on `facet_active_host ost2`"
-	start ost2 `ostdevname 2` $OST_MOUNT_OPTS || return 92
+	start ost2 `ostdevname 2` $OST_MOUNT_OPTS $@ || return 92
 }
 
 stop_ost2() {
@@ -118,12 +175,10 @@ mount_client() {
 }
 
 remount_client() {
-	local SAVEMOUNTOPT=$MOUNTOPT
-	MOUNTOPT="remount,$1"
+	local mountopt="-o remount,$1"
 	local MOUNTPATH=$2
 	echo "remount '$1' lustre on ${MOUNTPATH}....."
-	zconf_mount `hostname`  $MOUNTPATH  || return 96
-	MOUNTOPT=$SAVEMOUNTOPT
+	zconf_mount `hostname`  $MOUNTPATH "$mountopt"  || return 96
 }
 
 umount_client() {
@@ -142,25 +197,36 @@ manual_umount_client(){
 }
 
 setup() {
-	start_ost
-	start_mds
-	mount_client $MOUNT
+	start_mds || error "MDT start failed"
+	start_ost || error "OST start failed"
+	mount_client $MOUNT || error "client start failed"
+	client_up || error "client_up failed"
 }
 
 setup_noconfig() {
+	if ! combined_mgs_mds ; then
+		start_mgs
+	fi
+
 	start_mds
 	start_ost
 	mount_client $MOUNT
 }
 
+unload_modules_conf () {
+	if combined_mgs_mds || ! local_mode; then
+		unload_modules || return 1
+	fi
+}
+
 cleanup_nocli() {
-	stop_mds || return 201
 	stop_ost || return 202
-	unload_modules || return 203
+	stop_mds || return 201
+	unload_modules_conf || return 203
 }
 
 cleanup() {
- 	umount_client $MOUNT || return 200
+	umount_client $MOUNT || return 200
 	cleanup_nocli || return $?
 }
 
@@ -207,19 +273,21 @@ test_0() {
 run_test 0 "single mount setup"
 
 test_1() {
+	start_mds || error "MDT start failed"
 	start_ost
 	echo "start ost second time..."
-	setup
+	start_ost && error "2nd OST start should fail"
+	mount_client $MOUNT || error "client start failed"
 	check_mount || return 42
 	cleanup || return $?
 }
 run_test 1 "start up ost twice (should return errors)"
 
 test_2() {
-	start_ost
 	start_mds
 	echo "start mds second time.."
-	start_mds
+	start_mds && error "2nd MDT start should fail"
+	start_ost
 	mount_client $MOUNT
 	check_mount || return 43
 	cleanup || return $?
@@ -229,7 +297,7 @@ run_test 2 "start up mds twice (should return err)"
 test_3() {
 	setup
 	#mount.lustre returns an error if already in mtab
-	mount_client $MOUNT && return $?
+	mount_client $MOUNT && error "2nd client mount should fail"
 	check_mount || return 44
 	cleanup || return $?
 }
@@ -259,7 +327,7 @@ test_5a() {	# was test_5
 	# cleanup may return an error from the failed
 	# disconnects; for now I'll consider this successful
 	# if all the modules have unloaded.
- 	umount -d $MOUNT &
+	umount -d $MOUNT &
 	UMOUNT_PID=$!
 	sleep 6
 	echo "killing umount"
@@ -290,7 +358,7 @@ test_5a() {	# was test_5
 	while [ "$WAIT" -ne "$MAX_WAIT" ]; do
 		sleep $sleep
 		grep -q $MOUNT" " /etc/mtab || break
-        	echo "Waiting /etc/mtab updated ... "
+		echo "Waiting /etc/mtab updated ... "
 		WAIT=$(( WAIT + sleep))
 	done
 	[ "$WAIT" -eq "$MAX_WAIT" ] && error "/etc/mtab is not updated in $WAIT secs"
@@ -298,60 +366,132 @@ test_5a() {	# was test_5
 }
 run_test 5a "force cleanup mds, then cleanup"
 
+cleanup_5b () {
+	trap 0
+	start_mgs
+}
+
 test_5b() {
+	grep " $MOUNT " /etc/mtab && \
+		error false "unexpected entry in mtab before mount" && return 10
+
+	local rc=0
 	start_ost
+	if ! combined_mgs_mds ; then
+		trap cleanup_5b EXIT ERR
+		start_mds
+		stop mgs
+	fi
+
 	[ -d $MOUNT ] || mkdir -p $MOUNT
-	grep " $MOUNT " /etc/mtab && echo "test 5b: mtab before mount" && return 10
-	mount_client $MOUNT && return 1
-	grep " $MOUNT " /etc/mtab && echo "test 5b: mtab after failed mount" && return 11
+	mount_client $MOUNT && rc=1
+	grep " $MOUNT " /etc/mtab && \
+		error "$MOUNT entry in mtab after failed mount" && rc=11
 	umount_client $MOUNT
 	# stop_mds is a no-op here, and should not fail
-	cleanup_nocli || return $?
-	return 0
+	cleanup_nocli || rc=$?
+	if ! combined_mgs_mds ; then
+		cleanup_5b
+	fi
+	return $rc
 }
-run_test 5b "mds down, cleanup after failed mount (bug 2712) (should return errs)"
+run_test 5b "Try to start a client with no MGS (should return errs)"
 
 test_5c() {
-	start_ost
+	grep " $MOUNT " /etc/mtab && \
+		error false "unexpected entry in mtab before mount" && return 10
+
+	local rc=0
 	start_mds
+	start_ost
 	[ -d $MOUNT ] || mkdir -p $MOUNT
-	grep " $MOUNT " /etc/mtab && echo "test 5c: mtab before mount" && return 10
 	local oldfs="${FSNAME}"
 	FSNAME="wrong.${FSNAME}"
 	mount_client $MOUNT || :
 	FSNAME=${oldfs}
-	grep " $MOUNT " /etc/mtab && echo "test 5c: mtab after failed mount" && return 11
+	grep " $MOUNT " /etc/mtab && \
+		error "$MOUNT entry in mtab after failed mount" && rc=11
 	umount_client $MOUNT
-	cleanup_nocli  || return $?
+	cleanup_nocli  || rc=$?
+	return $rc
 }
 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
 
 test_5d() {
+	grep " $MOUNT " /etc/mtab && \
+		error false "unexpected entry in mtab before mount" && return 10
+
+	local rc=0
 	start_ost
 	start_mds
 	stop_ost -f
-	grep " $MOUNT " /etc/mtab && echo "test 5d: mtab before mount" && return 10
-	mount_client $MOUNT || return 1
-	cleanup  || return $?
-	grep " $MOUNT " /etc/mtab && echo "test 5d: mtab after unmount" && return 11
-	return 0
+	mount_client $MOUNT || rc=1
+	cleanup  || rc=$?
+	grep " $MOUNT " /etc/mtab && \
+		error "$MOUNT entry in mtab after unmount" && rc=11
+	return $rc
 }
 run_test 5d "mount with ost down"
 
 test_5e() {
-	start_ost
+	grep " $MOUNT " /etc/mtab && \
+		error false "unexpected entry in mtab before mount" && return 10
+
+	local rc=0
 	start_mds
+	start_ost
 
 #define OBD_FAIL_PTLRPC_DELAY_SEND       0x506
 	do_facet client "lctl set_param fail_loc=0x80000506"
-	grep " $MOUNT " /etc/mtab && echo "test 5e: mtab before mount" && return 10
 	mount_client $MOUNT || echo "mount failed (not fatal)"
-	cleanup  || return $?
-	grep " $MOUNT " /etc/mtab && echo "test 5e: mtab after unmount" && return 11
-	return 0
+	cleanup  || rc=$?
+	grep " $MOUNT " /etc/mtab && \
+		error "$MOUNT entry in mtab after unmount" && rc=11
+	return $rc
 }
 run_test 5e "delayed connect, don't crash (bug 10268)"
 
+test_5f() {
+	if combined_mgs_mds ; then
+		skip "combined mgs and mds"
+		return 0
+	fi
+
+	grep " $MOUNT " /etc/mtab && \
+		error false "unexpected entry in mtab before mount" && return 10
+
+	local rc=0
+	start_ost
+	[ -d $MOUNT ] || mkdir -p $MOUNT
+	mount_client $MOUNT &
+	local pid=$!
+	echo client_mount pid is $pid
+
+	sleep 5
+
+	if ! ps -f -p $pid >/dev/null; then
+		wait $pid
+		rc=$?
+		grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
+		error "mount returns $rc, expected to hang"
+		rc=11
+		cleanup || rc=$?
+		return $rc
+	fi
+
+	# start mds
+	start_mds
+
+	# mount should succeed after start mds
+	wait $pid
+	rc=$?
+	[ $rc -eq 0 ] || error "mount returned $rc"
+	grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
+	cleanup || return $?
+	return $rc
+}
+run_test 5f "mds down, cleanup after failed mount (bug 2712)"
+
 test_6() {
 	setup
 	manual_umount_client
@@ -401,70 +541,55 @@ test_9() {
         fi
         stop_ost || return $?
 }
-
 run_test 9 "test ptldebug and subsystem for mkfs"
 
-# LOGS/PENDING do not exist anymore since CMD3
-test_16() {
-        local TMPMTPT="${TMP}/conf16"
-        local dev=${SINGLEMDS}_dev
-        local MDSDEV=${!dev}
-        if [ ! -e "$MDSDEV" ]; then
-            log "no $MDSDEV existing, so mount Lustre to create one"
-	    setup
-            check_mount || return 41
-            cleanup || return $?
-        fi
+is_blkdev () {
+        local facet=$1
+        local dev=$2
+        local size=${3:-""}
 
-        [ -f "$MDSDEV" ] && LOOPOPT="-o loop"
+        local rc=0
+        do_facet $facet "test -b $dev" || rc=1
+        if [[ "$size" ]]; then
+                local in=$(do_facet $facet "dd if=$dev of=/dev/null bs=1k count=1 skip=$size 2>&1" |\
+                        awk '($3 == "in") { print $1 }')
+                [[ $in  = "1+0" ]] || rc=1
+        fi
+        return $rc
+}
 
-        log "change the mode of $MDSDEV/OBJECTS to 555"
-        do_facet $SINGLEMDS "mkdir -p $TMPMTPT &&
-                      mount $LOOPOPT -t $FSTYPE $MDSDEV $TMPMTPT &&
-                      chmod 555 $TMPMTPT/OBJECTS &&
-                      umount $TMPMTPT" || return $?
+#
+# Test 16 was to "verify that lustre will correct the mode of OBJECTS".
+# But with new MDS stack we don't care about the mode of local objects
+# anymore, so this test is removed. See bug 22944 for more details.
+#
 
-        log "mount Lustre to change the mode of OBJECTS, then umount Lustre"
-	setup
+test_17() {
+        setup
         check_mount || return 41
         cleanup || return $?
 
-        log "read the mode of OBJECTS and check if they has been changed properly"
-        EXPECTEDOBJECTSMODE=`do_facet $SINGLEMDS "$DEBUGFS -R 'stat OBJECTS' $MDSDEV 2> /dev/null" | grep 'Mode: ' | sed -e "s/.*Mode: *//" -e "s/ *Flags:.*//"`
-
-        if [ "$EXPECTEDOBJECTSMODE" = "0777" ]; then
-                log "Success:Lustre change the mode of OBJECTS correctly"
-        else
-                error "Lustre does not change mode of OBJECTS properly"
+        echo "Remove mds config log"
+        if ! combined_mgs_mds ; then
+                stop mgs
         fi
-}
-run_test 16 "verify that lustre will correct the mode of OBJECTS"
 
-test_17() {
-        local dev=${SINGLEMDS}_dev
-        local MDSDEV=${!dev}
-
-        if [ ! -e "$MDSDEV" ]; then
-            echo "no $MDSDEV existing, so mount Lustre to create one"
-	    setup
-            check_mount || return 41
-            cleanup || return $?
-        fi
+        do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' $MGSDEV || return \$?" || return $?
 
-        echo "Remove mds config log"
-        do_facet $SINGLEMDS "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' $MDSDEV || return \$?" || return $?
+        if ! combined_mgs_mds ; then
+                start_mgs
+        fi
 
         start_ost
-	start_mds && return 42
-	reformat_and_config
+        start_mds && return 42
+        reformat_and_config
 }
 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
 
 test_18() {
         [ "$FSTYPE" != "ldiskfs" ] && skip "not needed for FSTYPE=$FSTYPE" && return
 
-        local dev=${SINGLEMDS}_dev
-        local MDSDEV=${!dev}
+        local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
 
         local MIN=2000000
 
@@ -479,17 +604,18 @@ test_18() {
                 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
 
         # check if the block device is large enough
-        [ -z "$OK" -a -b $MDSDEV ] && \
-                [ "$(dd if=$MDSDEV of=/dev/null bs=1k count=1 skip=$MIN 2>&1 |
-                     awk '($3 == "in") { print $1 }')" = "1+0" ] && OK=1 && \
+        [ -z "$OK" ] && $(is_blkdev $SINGLEMDS $MDSDEV $MIN) && OK=1 &&
                 myMDSSIZE=$MIN && log "use device $MDSDEV with MIN=$MIN"
 
         # check if a loopback device has enough space for fs metadata (5%)
-        [ -z "$OK" ] && [ -f $MDSDEV -o ! -e $MDSDEV ] &&
-                SPACE=$(df -P $(dirname $MDSDEV) |
-                        awk '($1 != "Filesystem") {print $4}') &&
-                [ $SPACE -gt $((MIN / 20)) ] && OK=1 && myMDSSIZE=$MIN && \
+
+        if [ -z "$OK" ]; then
+                local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
+                        awk '($1 != "Filesystem") {print $4}')
+                ! [ -z "$SPACE" ]  &&  [ $SPACE -gt $((MIN / 20)) ] && \
+                        OK=1 && myMDSSIZE=$MIN && \
                         log "use file $MDSDEV with MIN=$MIN"
+        fi
 
         [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS" && return
 
@@ -497,15 +623,21 @@ test_18() {
         echo "mount mds with large journal..."
         local OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS
 
-        MDS_MKFS_OPTS="--mgs --mdt --fsname=$FSNAME --device-size=$myMDSSIZE --param sys.timeout=$TIMEOUT $MDSOPT"
+        local opts="--mdt --fsname=$FSNAME --device-size=$myMDSSIZE --param sys.timeout=$TIMEOUT $MDSOPT"
+
+        if combined_mgs_mds ; then
+            MDS_MKFS_OPTS="--mgs $opts"
+        else
+            MDS_MKFS_OPTS="--mgsnode=$MGSNID $opts"
+        fi
 
         reformat_and_config
         echo "mount lustre system..."
-	setup
+        setup
         check_mount || return 41
 
         echo "check journal size..."
-        local FOUNDSIZE=`do_facet mds "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}'`
+        local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
         if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
                 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
         else
@@ -533,8 +665,8 @@ run_test 19b "start/stop OSTs without MDS"
 
 test_20() {
 	# first format the ost/mdt
-	start_ost
 	start_mds
+	start_ost
 	mount_client $MOUNT
 	check_mount || return 43
 	rm -f $DIR/$tfile
@@ -583,6 +715,30 @@ test_21c() {
 }
 run_test 21c "start mds between two osts, stop mds last"
 
+test_21d() {
+        if combined_mgs_mds ; then
+                skip "need separate mgs device" && return 0
+        fi
+        stopall
+
+        reformat
+
+        start_mgs
+        start_ost
+        start_ost2
+        start_mds
+        wait_osc_import_state mds ost2 FULL
+
+        stop_ost
+        stop_ost2
+        stop_mds
+        stop_mgs
+        #writeconf to remove all ost2 traces for subsequent tests
+        writeconf
+        start_mgs
+}
+run_test 21d "start mgs then ost and then mds"
+
 test_22() {
 	start_mds
 
@@ -609,6 +765,8 @@ test_22() {
 		sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
 	fi
 	mount_client $MOUNT
+        wait_osc_import_state mds ost FULL
+        wait_osc_import_state client ost FULL
 	check_mount || return 41
 	pass
 
@@ -617,15 +775,15 @@ test_22() {
 run_test 22 "start a client before osts (should return errs)"
 
 test_23a() {	# was test_23
-        setup
-        # fail mds
+	setup
+	# fail mds
 	stop $SINGLEMDS
 	# force down client so that recovering mds waits for reconnect
 	local running=$(grep -c $MOUNT /proc/mounts) || true
-    	if [ $running -ne 0 ]; then
-        	echo "Stopping client $MOUNT (opts: -f)"
-        	umount -f $MOUNT
-    	fi
+	if [ $running -ne 0 ]; then
+		echo "Stopping client $MOUNT (opts: -f)"
+		umount -f $MOUNT
+	fi
 
 	# enter recovery on mds
 	start_mds
@@ -637,15 +795,17 @@ test_23a() {	# was test_23
 	echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
 	ps --ppid $MOUNT_PID
 	ps --ppid $MOUNT_LUSTRE_PID
-	# FIXME why o why can't I kill these? Manual "ctrl-c" works...
-	kill -TERM $MOUNT_LUSTRE_PID
 	echo "waiting for mount to finish"
 	ps -ef | grep mount
+	# "ctrl-c" sends SIGINT but it usually (in script) does not work on child process
+	# SIGTERM works but it does not spread to offspring processses
+	kill -s TERM $MOUNT_PID
+	kill -s TERM $MOUNT_LUSTRE_PID
 	# we can not wait $MOUNT_PID because it is not a child of this shell
 	local PID1
 	local PID2
 	local WAIT=0
-	local MAX_WAIT=20
+	local MAX_WAIT=30
 	local sleep=1
 	while [ "$WAIT" -lt "$MAX_WAIT" ]; do
 		sleep $sleep
@@ -657,9 +817,11 @@ test_23a() {	# was test_23
 		echo "waiting for mount to finish ... "
 		WAIT=$(( WAIT + sleep))
 	done
-	[ "$WAIT" -eq "$MAX_WAIT" ] && error "MOUNT_PID $MOUNT_PID and "\
+	if [ "$WAIT" -eq "$MAX_WAIT" ]; then
+		error "MOUNT_PID $MOUNT_PID and "\
 		"MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
-	ps -ef | grep mount
+		ps -ef | grep mount
+	fi
 	stop_mds || error
 	stop_ost || error
 }
@@ -669,8 +831,8 @@ umount_client $MOUNT
 cleanup_nocli
 
 test_23b() {    # was test_23
-	start_ost
 	start_mds
+	start_ost
 	# Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
 	lctl set_param fail_loc=0x80000313
 	mount_client $MOUNT
@@ -692,15 +854,15 @@ cleanup_24a() {
 }
 
 test_24a() {
-	#set up fs1
-	gen_config
-	#set up fs2
-	[ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
+	local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
+
 	if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
-		do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
+		is_blkdev $SINGLEMDS $MDSDEV && \
 		skip_env "mixed loopback and real device not working" && return
 	fi
 
+	[ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
+
 	local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
 	local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
 
@@ -731,7 +893,7 @@ test_24a() {
 	facet_failover fs2mds
 	facet_failover fs2ost
 	df
- 	umount_client $MOUNT
+	umount_client $MOUNT
 	# the MDS must remain up until last MDT
 	stop_mds
 	MDS=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
@@ -742,8 +904,12 @@ test_24a() {
 run_test 24a "Multiple MDTs on a single node"
 
 test_24b() {
+	local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
+
 	if [ -z "$fs2mds_DEV" ]; then
-		do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
+		local dev=${SINGLEMDS}_dev
+		local MDSDEV=${!dev}
+		is_blkdev $SINGLEMDS $MDSDEV && \
 		skip_env "mixed loopback and real device not working" && return
 	fi
 
@@ -773,9 +939,9 @@ test_26() {
     do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000135"
     start_mds && echo MDS started && return 1
     lctl get_param -n devices
-    DEVS=$(lctl get_param -n devices | wc -l)
+    DEVS=$(lctl get_param -n devices | egrep -v MG | wc -l)
     [ $DEVS -gt 0 ] && return 2
-    unload_modules || return 203
+    unload_modules_conf || return $?
 }
 run_test 26 "MDT startup failure cleans LOV (should return errs)"
 
@@ -828,7 +994,7 @@ test_28() {
 	set_and_check client "$TEST" "$PARAM" $FINAL || return 3
 	FINAL=$(($FINAL + 1))
 	set_and_check client "$TEST" "$PARAM" $FINAL || return 4
- 	umount_client $MOUNT || return 200
+	umount_client $MOUNT || return 200
 	mount_client $MOUNT
 	RESULT=$($TEST)
 	if [ $RESULT -ne $FINAL ]; then
@@ -865,7 +1031,9 @@ test_29() {
 	fi
 
 	# check MDT too
-	local MPROC="osc.$FSNAME-OST0001-osc-[M]*.active"
+	local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS $FSNAME-OST0001)
+	mdtosc=${mdtosc/-MDT*/-MDT\*}
+	local MPROC="osc.$mdtosc.active"
 	local MAX=30
 	local WAIT=0
 	while [ 1 ]; do
@@ -884,8 +1052,11 @@ test_29() {
 	    echo "Waiting $(($MAX - $WAIT)) secs for MDT deactivated"
 	done
 
+        # quotacheck should not fail immediately after deactivate
+	[ -n "$ENABLE_QUOTA" ] && { $LFS quotacheck -ug $MOUNT || error "quotacheck has failed" ; }
+
         # test new client starts deactivated
- 	umount_client $MOUNT || return 200
+	umount_client $MOUNT || return 200
 	mount_client $MOUNT
 	RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
 	if [ -z "$RESULT" ]; then
@@ -895,10 +1066,13 @@ test_29() {
 	    echo "New client success: got $RESULT"
 	fi
 
+        # quotacheck should not fail after umount/mount operation
+	[ -n "$ENABLE_QUOTA" ] && { $LFS quotacheck -ug $MOUNT || error "quotacheck has failed" ; }
+
 	# make sure it reactivates
 	set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $ACTV || return 6
 
- 	umount_client $MOUNT
+	umount_client $MOUNT
 	stop_ost2
 	cleanup_nocli
 	#writeconf to remove all ost2 traces for subsequent tests
@@ -906,9 +1080,10 @@ test_29() {
 }
 run_test 29 "permanently remove an OST"
 
-test_30() {
+test_30a() {
 	setup
 
+	echo Big config llog
 	TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
 	ORIG=$($TEST)
 	LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
@@ -916,13 +1091,56 @@ test_30() {
 	    set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3
 	done
 	# make sure client restart still works
- 	umount_client $MOUNT
+	umount_client $MOUNT
 	mount_client $MOUNT || return 4
-	[ "$($TEST)" -ne "$i" ] && return 5
-	set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" $ORIG || return 6
+	[ "$($TEST)" -ne "$i" ] && error "Param didn't stick across restart $($TEST) != $i"
+	pass
+
+	echo Erase parameter setting
+	do_facet mgs "$LCTL conf_param -d $FSNAME.llite.max_read_ahead_whole_mb" || return 6
+	umount_client $MOUNT
+	mount_client $MOUNT || return 6
+	FINAL=$($TEST)
+	echo "deleted (default) value=$FINAL, orig=$ORIG"
+	# assumes this parameter started at the default value
+	[ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
+
+	cleanup
+}
+run_test 30a "Big config llog and conf_param deletion"
+
+test_30b() {
+	setup
+
+	# Make a fake nid.  Use the OST nid, and add 20 to the least significant
+	# numerical part of it. Hopefully that's not already a failover address for
+	# the server.
+	OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
+	ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
+	NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
+	NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
+	echo "Using fake nid $NEW"
+
+	TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
+	set_and_check client "$TEST" "$FSNAME-OST0000.failover.node" $NEW || error "didn't add failover nid $NEW"
+	NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
+	echo $NIDS
+	NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1))
+	echo "should have 2 failover nids: $NIDCOUNT"
+	[ $NIDCOUNT -eq 2 ] || error "Failover nid not added"
+	do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" || error "conf_param delete failed"
+	umount_client $MOUNT
+	mount_client $MOUNT || return 3
+
+	NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
+	echo $NIDS
+	NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1))
+	echo "only 1 final nid should remain: $NIDCOUNT"
+	[ $NIDCOUNT -eq 1 ] || error "Failover nids not removed"
+
 	cleanup
 }
-run_test 30 "Big config llog"
+run_test 30b "Remove failover nids"
 
 test_31() { # bug 10734
         # ipaddr must not exist
@@ -971,7 +1189,8 @@ cleanup_32() {
 	umount -f $MOUNT || true
 	echo "Cleanup local mds ost1 ..."
 	cleanup_nocli32
-	unload_modules
+	combined_mgs_mds || start_mgs
+	unload_modules_conf
 }
 
 test_32a() {
@@ -991,6 +1210,8 @@ test_32a() {
 
 	$TUNEFS $tmpdir/mds || error "tunefs failed"
 
+	combined_mgs_mds || stop mgs
+
 	# nids are wrong, so client wont work, but server should start
 	start32 mds1 $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \
 		trap cleanup_32 EXIT INT || return 3
@@ -1027,6 +1248,9 @@ test_32a() {
 	# mount a second time to make sure we didnt leave upgrade flag on
 	load_modules
 	$TUNEFS --dryrun $tmpdir/mds || error "tunefs failed"
+
+	combined_mgs_mds || stop mgs
+
 	start32 mds1 $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \
 		trap cleanup_32 EXIT INT || return 12
 
@@ -1054,6 +1278,8 @@ test_32b() {
 
 	# writeconf will cause servers to register with their current nids
 	$TUNEFS --writeconf --fsname=$NEWNAME $tmpdir/mds || error "tunefs failed"
+	combined_mgs_mds || stop mgs
+
 	start32 mds1 $tmpdir/mds "-o loop" && \
 		trap cleanup_32 EXIT INT || return 3
 
@@ -1110,13 +1336,19 @@ run_test 32b "Upgrade from 1.8 with writeconf"
 test_33a() { # bug 12333, was test_33
         local rc=0
         local FSNAME2=test-123
+        local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
+
         [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
 
         if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
-                do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
+                local dev=${SINGLEMDS}_dev
+                local MDSDEV=${!dev}
+                is_blkdev $SINGLEMDS $MDSDEV && \
                 skip_env "mixed loopback and real device not working" && return
         fi
 
+        combined_mgs_mds || mkfs_opts="$mkfs_opts --nomgs"
+
         local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
         local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
         add fs2mds $MDS_MKFS_OPTS --mkfsoptions='\"-J size=8\"' --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
@@ -1174,7 +1406,7 @@ test_34b() {
 	touch $DIR/$tfile || return 1
 	stop_mds --force || return 2
 
- 	manual_umount_client --force
+	manual_umount_client --force
 	rc=$?
 	if [ $rc -ne 0 ]; then
 		error "mtab after failed umount - rc $rc"
@@ -1190,7 +1422,7 @@ test_34c() {
 	touch $DIR/$tfile || return 1
 	stop_ost --force || return 2
 
- 	manual_umount_client --force
+	manual_umount_client --force
 	rc=$?
 	if [ $rc -ne 0 ]; then
 		error "mtab after failed umount - rc $rc"
@@ -1245,6 +1477,8 @@ test_35a() { # bug 12459
 		       }" $TMP/lustre-log-$TESTNAME.log`
 	[ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7
 	cleanup
+	# remove nid settings
+	writeconf
 }
 run_test 35a "Reconnect to the last active server first"
 
@@ -1260,31 +1494,33 @@ test_35b() { # bug 18674
 
 	log "Set up a fake failnode for the MDS"
 	FAKENID="127.0.0.2"
-	local device=$(do_facet mds "$LCTL get_param -n devices" | \
+	local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" | \
 			awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
-	do_facet mds "$LCTL conf_param ${device}.failover.node=$FAKENID" || \
+	do_facet $SINGLEMDS "$LCTL conf_param ${device}.failover.node=$FAKENID" || \
 		return 1
 
 	local at_max_saved=0
-	# adaptive timeouts may prevent seeing the issue 
+	# adaptive timeouts may prevent seeing the issue
 	if at_is_enabled; then
 		at_max_saved=$(at_max_get mds)
 		at_max_set 0 mds client
 	fi
 
-	mkdir -p $MOUNT/testdir
-	touch $MOUNT/testdir/test
+	mkdir -p $MOUNT/$tdir
 
 	log "Injecting EBUSY on MDS"
 	# Setting OBD_FAIL_MDS_RESEND=0x136
-	do_facet mds "$LCTL set_param fail_loc=0x80000136" || return 2
+	do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" || return 2
+
+	$LCTL set_param mdc.${FSNAME}*.stats=clear
 
-	log "Stat on a test file"
-	stat $MOUNT/testdir/test
+	log "Creating a test file and stat it"
+	touch $MOUNT/$tdir/$tfile
+	stat $MOUNT/$tdir/$tfile
 
 	log "Stop injecting EBUSY on MDS"
-	do_facet mds "$LCTL set_param fail_loc=0" || return 3
-	rm -f $MOUNT/testdir/test
+	do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" || return 3
+	rm -f $MOUNT/$tdir/$tfile
 
 	log "done"
 	# restore adaptive timeout
@@ -1292,6 +1528,8 @@ test_35b() { # bug 18674
 
 	$LCTL dk $TMP/lustre-log-$TESTNAME.log
 
+	CONNCNT=`$LCTL get_param mdc.${FSNAME}*.stats | awk '/mds_connect/{print $2}'`
+
 	# retrieve from the log if the client has ever tried to
 	# contact the fake server after the loss of connection
 	FAILCONN=`awk "BEGIN {ret = 0;}
@@ -1311,26 +1549,36 @@ test_35b() { # bug 18674
 		log "ERROR: The client tried to reconnect to the failover server while the primary was busy" && \
 		return 5
 
-        cleanup
+	# LU-290
+	# When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
+	# Reconnects are supposed to be rate limited to one every 5s
+	[ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] && \
+		log "ERROR: Too many reconnects $CONNCNT" && \
+		return 6
+
+	cleanup
+	# remove nid settings
+	writeconf
 }
 run_test 35b "Continue reconnection retries, if the active server is busy"
 
 test_36() { # 12743
-        local rc
+        [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
+
+        [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
+		{ skip "remote OST" && return 0; }
+
+        local rc=0
         local FSNAME2=test1234
         local fs3ost_HOST=$ost_HOST
+        local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
 
         [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
-        rc=0
 
         if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
-		do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
+		is_blkdev $SINGLEMDS $MDSDEV && \
 		skip_env "mixed loopback and real device not working" && return
         fi
-        [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
-
-	[ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
-		{ skip "remote OST" && return 0; }
 
         local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
         local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
@@ -1383,34 +1631,38 @@ test_36() { # 12743
         stop fs2ost -f || return 201
         stop fs2mds -f || return 202
         rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev $fs3ostdev
-        unload_modules || return 203
+        unload_modules_conf || return 203
         return $rc
 }
 run_test 36 "df report consistency on OSTs with different block size"
 
 test_37() {
-	client_only && skip "client only testing" && return 0
-	LOCAL_MDSDEV="$TMP/mdt.img"
-	SYM_MDSDEV="$TMP/sym_mdt.img"
+	local mntpt=$(facet_mntpt $SINGLEMDS)
+	local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
+	local mdsdev_sym="$TMP/sym_mdt.img"
 
-	echo "MDS :     $LOCAL_MDSDEV"
-	echo "SYMLINK : $SYM_MDSDEV"
-	rm -f $LOCAL_MDSDEV
+	echo "MDS :     $mdsdev"
+	echo "SYMLINK : $mdsdev_sym"
+	do_facet $SINGLEMDS rm -f $mdsdev_sym
 
-	touch $LOCAL_MDSDEV
-	mkfs.lustre --reformat --fsname=lustre --mdt --mgs --device-size=9000 $LOCAL_MDSDEV ||
-		error "mkfs.lustre $LOCAL_MDSDEV failed"
-	ln -s $LOCAL_MDSDEV $SYM_MDSDEV
+	do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
 
-	echo "mount symlink device - $SYM_MDSDEV"
+	echo "mount symlink device - $mdsdev_sym"
 
-	mount_op=`mount -v -t lustre -o loop $SYM_MDSDEV ${MOUNT%/*}/mds 2>&1 | grep "unable to set tunable"`
-	umount -d ${MOUNT%/*}/mds
-	rm -f $LOCAL_MDSDEV $SYM_MDSDEV
+	local rc=0
+	mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $MDS_MOUNT_OPTS  $mdsdev_sym $mntpt 2>&1 )
+	rc=${PIPESTATUS[0]}
+
+	echo mount_op=$mount_op
 
-	if [ -n "$mount_op" ]; then
-		error "**** FAIL: set tunables failed for symlink device"
+	do_facet $SINGLEMDS "umount -d $mntpt && rm -f $mdsdev_sym"
+
+	if $(echo $mount_op | grep -q "unable to set tunable"); then
+		error "set tunables failed for symlink device"
 	fi
+
+	[ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
+
 	return 0
 }
 run_test 37 "verify set tunables works for symlink device"
@@ -1431,8 +1683,7 @@ test_38() { # bug 14222
 	log "rename lov_objid file on MDS"
 	rm -f $TMP/lov_objid.orig
 
-	local dev=${SINGLEMDS}_dev
-	local MDSDEV=${!dev}
+	local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
 	do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.orig\\\" $MDSDEV"
 	do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
 
@@ -1485,16 +1736,17 @@ run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
 test_40() { # bug 15759
 	start_ost
 	#define OBD_FAIL_TGT_TOOMANY_THREADS     0x706
-	do_facet $SINGLEMDS "sysctl -w lustre.fail_loc=0x80000706"
+	do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
 	start_mds
 	cleanup
 }
 run_test 40 "race during service thread startup"
 
-test_41() { #bug 14134
+test_41a() { #bug 14134
+        echo $MDS_MOUNT_OPTS | grep "loop" && skip " loop devices does not work with nosvc option" && return
+
         local rc
-        local dev=${SINGLEMDS}_dev
-        local MDSDEV=${!dev}
+        local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
 
         start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
         start ost1 `ostdevname 1` $OST_MOUNT_OPTS
@@ -1510,15 +1762,42 @@ test_41() { #bug 14134
         stop ost1 -f || return 201
         stop_mds -f || return 202
         stop_mds -f || return 203
-        unload_modules || return 204
+        unload_modules_conf || return 204
         return $rc
 }
-run_test 41 "mount mds with --nosvc and --nomgs"
+run_test 41a "mount mds with --nosvc and --nomgs"
+
+test_41b() {
+        echo $MDS_MOUNT_OPTS | grep "loop" && skip " loop devices does not work with nosvc option" && return
+
+        ! combined_mgs_mds && skip "needs combined mgs device" && return 0
+
+        stopall
+        reformat
+        local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
+
+        start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
+        start_ost
+        start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
+        mkdir -p $MOUNT
+        mount_client $MOUNT || return 1
+        sleep 5
+
+        echo "blah blah" > $MOUNT/$tfile
+        cat $MOUNT/$tfile || return 200
+
+        umount_client $MOUNT
+        stop_ost || return 201
+        stop_mds -f || return 202
+        stop_mds -f || return 203
+
+}
+run_test 41b "mount mds with --nosvc and --nomgs on first mount"
 
 test_42() { #bug 14693
         setup
         check_mount || return 2
-        do_facet client lctl conf_param lustre.llite.some_wrong_param=10
+        do_facet mgs $LCTL conf_param lustre.llite.some_wrong_param=10
         umount_client $MOUNT
         mount_client $MOUNT || return 1
         cleanup
@@ -1635,7 +1914,7 @@ test_44() { # 16317
         check_mount || return 2
         UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
         STATS_FOUND=no
-        UUIDS=$(do_facet mds "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
+        UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
         for VAL in $UUIDS; do
                 NID=$(echo $VAL | cut -d= -f1)
                 CLUUID=$(echo $VAL | cut -d= -f2)
@@ -1678,17 +1957,16 @@ cleanup_46a() {
 		stop ost${count} -f || rc=$?
 		let count=count-1
 	done	
-	stop_mds || rc=$? 
-	# writeconf is needed after the test, otherwise,
-	# we might end up with extra OSTs
-	writeconf || rc=$?
+	stop_mds || rc=$?
 	cleanup_nocli || rc=$?
+	#writeconf to remove all ost2 traces for subsequent tests
+	writeconf
 	return $rc
 }
 
 test_46a() {
 	echo "Testing with $OSTCOUNT OSTs"
-	reformat
+	reformat_and_config
 	start_mds || return 1
 	#first client should see only one ost
 	start_ost || return 2
@@ -1697,7 +1975,7 @@ test_46a() {
 	mount_client $MOUNT || return 3
 	trap "cleanup_46a $OSTCOUNT" EXIT ERR
 
-	local i 
+	local i
 	for (( i=2; i<=$OSTCOUNT; i++ )); do
 	    start ost$i `ostdevname $i` $OST_MOUNT_OPTS || return $((i+2))
 	done
@@ -1705,9 +1983,9 @@ test_46a() {
 	# wait until osts in sync
 	for (( i=2; i<=$OSTCOUNT; i++ )); do
 	    wait_osc_import_state mds ost$i FULL
+	    wait_osc_import_state client ost$i FULL
 	done
 
-
 	#second client see all ost's
 
 	mount_client $MOUNT2 || return 8
@@ -1811,13 +2089,11 @@ test_49() { # bug 17710
 	OST_MKFS_OPTS="--ost --fsname=$FSNAME --device-size=$OSTSIZE --mgsnode=$MGSNID --param sys.timeout=$LOCAL_TIMEOUT --param sys.ldlm_timeout=$LOCAL_TIMEOUT $MKFSOPT $OSTOPT"
 
 	reformat
-	start_mds
-	start_ost
-	mount_client $MOUNT
+	setup_noconfig
 	check_mount || return 1
 
 	echo "check ldlm_timout..."
-	LDLM_MDS="`do_facet mds lctl get_param -n ldlm_timeout`"
+	LDLM_MDS="`do_facet $SINGLEMDS lctl get_param -n ldlm_timeout`"
 	LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
 	LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
 
@@ -1836,12 +2112,10 @@ test_49() { # bug 17710
 	OST_MKFS_OPTS="--ost --fsname=$FSNAME --device-size=$OSTSIZE --mgsnode=$MGSNID --param sys.timeout=$LOCAL_TIMEOUT --param sys.ldlm_timeout=$((LOCAL_TIMEOUT - 1)) $MKFSOPT $OSTOPT"
 
 	reformat
-	start_mds || return 4
-	start_ost || return 5
-	mount_client $MOUNT || return 6
+	setup_noconfig
 	check_mount || return 7
 
-	LDLM_MDS="`do_facet mds lctl get_param -n ldlm_timeout`"
+	LDLM_MDS="`do_facet $SINGLEMDS lctl get_param -n ldlm_timeout`"
 	LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
 	LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
 
@@ -1924,9 +2198,11 @@ test_50c() {
         wait_osc_import_state mds ost DISCONN
 	lazystatfs $MOUNT || error "lazystatfs failed with one down server"
 
- 	umount_client $MOUNT || error "Unable to unmount client"
+	umount_client $MOUNT || error "Unable to unmount client"
 	stop_ost2 || error "Unable to stop OST2"
 	stop_mds || error "Unable to stop MDS"
+	#writeconf to remove all ost2 traces for subsequent tests
+	writeconf
 }
 run_test 50c "lazystatfs one server down =========================="
 
@@ -1944,9 +2220,11 @@ test_50d() {
 	stop_ost || error "Unable to stop OST1"
 	lazystatfs $MOUNT || error "lazystatfs failed with one down server"
 
- 	umount_client $MOUNT || error "Unable to unmount client"
+	umount_client $MOUNT || error "Unable to unmount client"
 	stop_ost2 || error "Unable to stop OST2"
 	stop_mds || error "Unable to stop MDS"
+	#writeconf to remove all ost2 traces for subsequent tests
+	writeconf
 }
 run_test 50d "lazystatfs client/server conn race =========================="
 
@@ -2027,6 +2305,7 @@ test_50f() {
 	umount_client $MOUNT || error "Unable to unmount client"
 	stop_ost || error "Unable to stop OST1"
 	stop_mds || error "Unable to stop MDS"
+	#writeconf to remove all ost2 traces for subsequent tests
 	writeconf
 }
 run_test 50f "normal statfs one server in down =========================="
@@ -2035,6 +2314,8 @@ test_50g() {
 	[ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
 	setup
 	start_ost2 || error "Unable to start OST2"
+        wait_osc_import_state mds ost2 FULL
+        wait_osc_import_state client ost2 FULL
 
 	local PARAM="${FSNAME}-OST0001.osc.active"
 
@@ -2052,6 +2333,7 @@ test_50g() {
 	stop_ost2 || error "Unable to stop OST2"
 	stop_ost || error "Unable to stop OST1"
 	stop_mds || error "Unable to stop MDS"
+	#writeconf to remove all ost2 traces for subsequent tests
 	writeconf
 }
 run_test 50g "deactivated OST should not cause panic====================="
@@ -2060,9 +2342,7 @@ test_51() {
 	local LOCAL_TIMEOUT=20
 
 	reformat
-	start_mds
-	start_ost
-	mount_client $MOUNT
+	setup_noconfig
 	check_mount || return 1
 
 	mkdir $MOUNT/d1
@@ -2076,6 +2356,8 @@ test_51() {
 	wait $pid
 	stop_ost2 || return 3
 	cleanup
+	#writeconf to remove all ost2 traces for subsequent tests
+	writeconf
 }
 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
 
@@ -2136,7 +2418,7 @@ test_52() {
 	[ $? -eq 0 ] || { error "Unable to mount client"; return 3; }
 
 	local nrfiles=8
-	local ost1mnt=${MOUNT%/*}/ost1
+	local ost1mnt=$(facet_mntpt ost1)
 	local ost1node=$(facet_active_host ost1)
 	local ost1tmp=$TMP/conf52
 
@@ -2144,7 +2426,9 @@ test_52() {
 	[ $? -eq 0 ] || { error "Unable to create tdir"; return 4; }
 	touch $TMP/modified_first
 	[ $? -eq 0 ] || { error "Unable to create temporary file"; return 5; }
-	do_node $ost1node "mkdir -p $ost1tmp && touch $ost1tmp/modified_first"
+	local mtime=$(stat -c %Y $TMP/modified_first)
+	do_node $ost1node "mkdir -p $ost1tmp && touch -m -d @$mtime $ost1tmp/modified_first"
+
 	[ $? -eq 0 ] || { error "Unable to create temporary file"; return 6; }
 	sleep 1
 
@@ -2170,6 +2454,8 @@ test_52() {
 	[ $? -eq 0 ] || { error "Unable to stop ost1"; return 11; }
 
 	echo mount ost1 as ldiskfs
+	do_node $ost1node mkdir -p $ost1mnt
+	[ $? -eq 0 ] || { error "Unable to create $ost1mnt"; return 23; }
 	do_node $ost1node mount -t $FSTYPE $ost1_dev $ost1mnt $OST_MOUNT_OPTS
 	[ $? -eq 0 ] || { error "Unable to mount ost1 as ldiskfs"; return 12; }
 
@@ -2192,7 +2478,7 @@ test_52() {
 	diff_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
 	[ $? -eq 0 ] || { error "Unable to diff objects"; return 16; }
 
-	do_node $ost1node "umount $ost1_dev"
+	do_node $ost1node "umount $ost1mnt"
 	[ $? -eq 0 ] || { error "Unable to umount ost1 as ldiskfs"; return 17; }
 
 	start_ost
@@ -2212,6 +2498,279 @@ test_52() {
 }
 run_test 52 "check recovering objects from lost+found"
 
+# Checks threads_min/max/started for some service
+#
+# Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
+# parameter pattern prefix like 'ost.*.ost'.
+thread_sanity() {
+        local modname=$1
+        local facet=$2
+        local parampat=$3
+        local opts=$4
+        local tmin
+        local tmin2
+        local tmax
+        local tmax2
+        local tstarted
+        local paramp
+        local msg="Insane $modname thread counts"
+        shift 4
+
+        setup
+        check_mount || return 41
+
+        # We need to expand $parampat, but it may match multiple parameters, so
+        # we'll pick the first one
+        if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
+                error "Couldn't expand ${parampat}.threads_min parameter name"
+                return 22
+        fi
+
+        # Remove the .threads_min part
+        paramp=${paramp%.threads_min}
+
+        # Check for sanity in defaults
+        tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
+        tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
+        tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started" || echo 0)
+        lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' || return $?
+        lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' || return $?
+
+        # Check that we can change min/max
+        do_facet $facet "lctl set_param ${paramp}.threads_min=$((tmin + 1))"
+        do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmax - 1))"
+        tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
+        tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
+        lassert 25 "$msg" '(($tmin2 == ($tmin + 1) && $tmax2 == ($tmax -1)))' || return $?
+
+        # Check that we can set min/max to the same value
+        tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
+        do_facet $facet "lctl set_param ${paramp}.threads_max=$tmin"
+        tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
+        tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
+        lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
+
+        # Check that we can't set max < min
+        do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmin - 1))"
+        tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
+        tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
+        lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
+
+        # We need to ensure that we get the module options desired; to do this
+        # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
+        LOAD_MODULES_REMOTE=true
+        cleanup
+        local oldvalue
+        setmodopts -a $modname "$opts" oldvalue
+
+        load_modules
+        setup
+        check_mount || return 41
+
+        # Restore previous setting of MODOPTS_*
+        setmodopts $modname "$oldvalue"
+
+        # Check that $opts took
+        tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min")
+        tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max")
+        tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started")
+        lassert 28 "$msg" '(($tstarted == $tmin && $tstarted == $tmax ))' || return $?
+        cleanup
+
+        # Workaround a YALA bug where YALA expects that modules will remain
+        # loaded on the servers
+        LOAD_MODULES_REMOTE=false
+        load_modules
+        setup
+        cleanup
+}
+
+test_53a() {
+        thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads=64'
+}
+run_test 53a "check OSS thread count params"
+
+test_53b() {
+        thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads=64'
+}
+run_test 53b "check MDT thread count params"
+
+test_54a() {
+    do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p
+    [ $? -eq 0 ] || error "llverdev failed!"
+    reformat_and_config
+}
+run_test 54a "test llverdev and partial verify of device"
+
+test_54b() {
+    setup
+    run_llverfs $MOUNT -p
+    [ $? -eq 0 ] || error "llverfs failed!"
+    cleanup
+}
+run_test 54b "test llverfs and partial verify of filesystem"
+
+lov_objid_size()
+{
+	local max_ost_index=$1
+	echo -n $(((max_ost_index + 1) * 8))
+}
+
+test_55() {
+	local mdsdev=$(mdsdevname 1)
+	local ostdev=$(ostdevname 1)
+	local saved_opts=$OST_MKFS_OPTS
+
+	for i in 1023 2048
+	do
+		OST_MKFS_OPTS="$saved_opts --index $i"
+		reformat
+
+		setup_noconfig
+		stopall
+
+		setup
+		sync
+		echo checking size of lov_objid for ost index $i
+		LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
+		if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
+			error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
+		else
+			echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
+		fi
+		stopall
+	done
+
+	OST_MKFS_OPTS=$saved_opts
+	reformat
+}
+run_test 55 "check lov_objid size"
+
+test_56() {
+	add mds1 $MDS_MKFS_OPTS --mkfsoptions='\"-J size=16\"' --reformat $(mdsdevname 1)
+	add ost1 $OST_MKFS_OPTS --index=1000 --reformat $(ostdevname 1)
+	add ost2 $OST_MKFS_OPTS --index=10000 --reformat $(ostdevname 2)
+
+	start_mds
+	start_ost
+	start_ost2 || error "Unable to start second ost"
+	mount_client $MOUNT || error "Unable to mount client"
+	echo ok
+	$LFS osts
+	[ -n "$ENABLE_QUOTA" ] && { $LFS quotacheck -ug $MOUNT || error "quotacheck has failed" ; }
+	stopall
+	reformat
+}
+run_test 56 "check big indexes"
+
+test_57a() { # bug 22656
+	local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
+	writeconf
+	do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" || error "tunefs failed"
+	start_mgsmds
+	start_ost && error "OST registration from failnode should fail"
+	reformat
+}
+run_test 57a "initial registration from failnode should fail (should return errs)"
+
+test_57b() {
+	local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
+	writeconf
+	do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" || error "tunefs failed"
+	start_mgsmds
+	start_ost || error "OST registration from servicenode should not fail"
+	reformat
+}
+run_test 57b "initial registration from servicenode should not fail"
+
+count_osts() {
+        do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
+}
+
+test_58() { # bug 22658
+        [ "$FSTYPE" != "ldiskfs" ] && skip "not supported for $FSTYPE" && return
+	setup
+	mkdir -p $DIR/$tdir
+	createmany -o $DIR/$tdir/$tfile-%d 100
+	# make sure that OSTs do not cancel llog cookies before we unmount the MDS
+#define OBD_FAIL_OBD_LOG_CANCEL_NET      0x601
+	do_facet mds "lctl set_param fail_loc=0x601"
+	unlinkmany $DIR/$tdir/$tfile-%d 100
+	stop mds
+	local MNTDIR=$(facet_mntpt mds)
+	# remove all files from the OBJECTS dir
+	do_facet mds "mount -t ldiskfs $MDSDEV $MNTDIR"
+	do_facet mds "find $MNTDIR/OBJECTS -type f -delete"
+	do_facet mds "umount $MNTDIR"
+	# restart MDS with missing llog files
+	start_mds
+	do_facet mds "lctl set_param fail_loc=0"
+	reformat
+}
+run_test 58 "missing llog files must not prevent MDT from mounting"
+
+test_59() {
+	start_mgsmds >> /dev/null
+	local C1=$(count_osts)
+	if [ $C1 -eq 0 ]; then
+		start_ost >> /dev/null
+		C1=$(count_osts)
+	fi
+	stopall
+	echo "original ost count: $C1 (expect > 0)"
+	[ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
+	start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
+	local C2=$(count_osts)
+	echo "after mdt writeconf count: $C2 (expect 0)"
+	[ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
+	echo "OST start without writeconf should fail:"
+	start_ost >> /dev/null && error "OST start without writeconf didn't fail"
+	echo "OST start with writeconf should succeed:"
+	start_ost -o writeconf >> /dev/null || error "OST1 start failed"
+	local C3=$(count_osts)
+	echo "after ost writeconf count: $C3 (expect 1)"
+	[ $C3 -eq 1 ] || error "new OST writeconf should add:"
+	start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
+	local C4=$(count_osts)
+	echo "after ost2 writeconf count: $C4 (expect 2)"
+	[ $C4 -eq 2 ] || error "OST2 writeconf should add log"
+	stop_ost2 >> /dev/null
+	cleanup_nocli >> /dev/null
+	#writeconf to remove all ost2 traces for subsequent tests
+	writeconf
+}
+run_test 59 "writeconf mount option"
+
+test_60() { # LU-471
+	add mds1 $MDS_MKFS_OPTS --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' --reformat $(mdsdevname 1)
+
+	dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
+	rc=${PIPESTATUS[0]}
+	[ $rc -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
+
+	# MDT default has dirdata feature
+	echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
+	# we disable uninit_bg feature
+	echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
+	# we set stride extended options
+	echo $dump | grep stride > /dev/null || error "stride is not set"
+	reformat
+}
+run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
+
+if ! combined_mgs_mds ; then
+	stop mgs
+fi
+
 cleanup_gss
-equals_msg `basename $0`: test complete
-[ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG && grep -q FAIL $TESTSUITELOG && exit 1 || true
+
+# restore the ${facet}_MKFS_OPTS variables
+for facet in MGS MDS OST; do
+    opts=SAVED_${facet}_MKFS_OPTS
+    if [[ -n ${!opts} ]]; then
+        eval ${facet}_MKFS_OPTS=\"${!opts}\"
+    fi
+done
+
+complete $(basename $0) $SECONDS
+exit_status