X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Ftests%2Fsanity-scrub.sh;h=29ba554db4757ae634ee941630add20fa264bd8c;hb=3b3d8a35b7eb6de8c637e154f9509bb78695fece;hp=2000ed85006b7a7b00eb07a3b11a9adc344e1ee2;hpb=0f37c051158a399f7b00536eeec27f5dbdd54168;p=fs%2Flustre-release.git

diff --git a/lustre/tests/sanity-scrub.sh b/lustre/tests/sanity-scrub.sh
index 2000ed8..29ba554 100644
--- a/lustre/tests/sanity-scrub.sh
+++ b/lustre/tests/sanity-scrub.sh
@@ -128,13 +128,8 @@ scrub_prep() {
 	echo "preparing... $(date)"
 	for n in $(seq $MDSCOUNT); do
 		echo "creating $nfiles files on mds$n"
-		if [ $n -eq 1 ]; then
-			mkdir $DIR/$tdir/mds$n ||
-				error "Failed to create directory mds$n"
-		else
-			$LFS mkdir -i $((n - 1)) $DIR/$tdir/mds$n ||
-				error "Failed to create remote directory mds$n"
-		fi
+		test_mkdir -i $((n - 1)) $DIR/$tdir/mds$n ||
+			error "Failed to create directory mds$n"
 		cp $LUSTRE/tests/*.sh $DIR/$tdir/mds$n ||
 			error "Failed to copy files to mds$n"
 		mkdir -p $DIR/$tdir/mds$n/d_$tfile ||
@@ -148,6 +143,18 @@ scrub_prep() {
 	done
 	echo "prepared $(date)."
 	cleanup_mount $MOUNT > /dev/null || error "Fail to stop client!"
+
+	# sync local transactions on every MDT
+	do_nodes $(comma_list $(mdts_nodes)) \
+		"$LCTL set_param -n osd*.*MDT*.force_sync=1"
+
+	# wait for a while to cancel update logs after transactions committed.
+	sleep 3
+
+	# sync again to guarantee all things done.
+	do_nodes $(comma_list $(mdts_nodes)) \
+		"$LCTL set_param -n osd*.*MDT*.force_sync=1"
+
 	for n in $(seq $MDSCOUNT); do
 		echo "stop mds$n"
 		stop mds$n > /dev/null || error "Fail to stop MDS$n!"
@@ -287,7 +294,7 @@ scrub_backup_restore() {
 
 	for n in $(seq $MDSCOUNT); do
 		mds_backup_restore mds$n $igif ||
-			error "(error_id) Backup/restore on mds$n failed"
+			error "($error_id) Backup/restore on mds$n failed"
 	done
 }
 
@@ -466,6 +473,10 @@ test_4b() {
 	for n in $(seq $MDSCOUNT); do
 		updated0[$n]=$(scrub_status $n |
 			       awk '/^prior_updated/ { print $2 }')
+
+		echo "OI scrub on MDS$n status for the 1st time:"
+		do_facet mds$n $LCTL get_param -n \
+			osd-ldiskfs.$(facet_svc mds$n).oi_scrub
 	done
 
 	scrub_check_data2 sanity-scrub.sh 9
@@ -478,6 +489,11 @@ test_4b() {
 	for n in $(seq $MDSCOUNT); do
 		updated1[$n]=$(scrub_status $n |
 			       awk '/^prior_updated/ { print $2 }')
+
+		echo "OI scrub on MDS$n status for the 2nd time:"
+		do_facet mds$n $LCTL get_param -n \
+			osd-ldiskfs.$(facet_svc mds$n).oi_scrub
+
 		[ ${updated0[$n]} -lt ${updated1[$n]} ] ||
 			error "(12) Auto trigger full scrub unexpectedly"
 	done
@@ -494,6 +510,11 @@ test_4b() {
 	for n in $(seq $MDSCOUNT); do
 		updated0[$n]=$(scrub_status $n |
 			       awk '/^prior_updated/ { print $2 }')
+
+		echo "OI scrub on MDS$n status for the 3rd time:"
+		do_facet mds$n $LCTL get_param -n \
+			osd-ldiskfs.$(facet_svc mds$n).oi_scrub
+
 		[ ${updated0[$n]} -gt ${updated1[$n]} ] ||
 			error "(16) Auto trigger full scrub unexpectedly"
 	done
@@ -506,8 +527,13 @@ test_4b() {
 	for n in $(seq $MDSCOUNT); do
 		updated1[$n]=$(scrub_status $n |
 			       awk '/^prior_updated/ { print $2 }')
-		[ ${updated0[$n]} -eq ${updated1[$n]} ] ||
+		[ ${updated0[$n]} -eq ${updated1[$n]} ] || {
+			echo "OI scrub on MDS$n status for the 4th time:"
+			do_facet mds$n $LCTL get_param -n \
+				osd-ldiskfs.$(facet_svc mds$n).oi_scrub
+
 			error "(18) NOT auto trigger full scrub as expected"
+		}
 	done
 }
 run_test 4b "Auto trigger OI scrub if bad OI mapping was found (2)"
@@ -532,6 +558,10 @@ test_4c() {
 	for n in $(seq $MDSCOUNT); do
 		updated0[$n]=$(scrub_status $n |
 			       awk '/^prior_updated/ { print $2 }')
+
+		echo "OI scrub on MDS$n status for the 1st time:"
+		do_facet mds$n $LCTL get_param -n \
+			osd-ldiskfs.$(facet_svc mds$n).oi_scrub
 	done
 
 	scrub_check_data2 sanity-scrub.sh 9
@@ -544,6 +574,11 @@ test_4c() {
 	for n in $(seq $MDSCOUNT); do
 		updated1[$n]=$(scrub_status $n |
 			       awk '/^prior_updated/ { print $2 }')
+
+		echo "OI scrub on MDS$n status for the 2nd time:"
+		do_facet mds$n $LCTL get_param -n \
+			osd-ldiskfs.$(facet_svc mds$n).oi_scrub
+
 		[ ${updated0[$n]} -lt ${updated1[$n]} ] ||
 			error "(12) Auto trigger full scrub unexpectedly"
 	done
@@ -560,6 +595,11 @@ test_4c() {
 	for n in $(seq $MDSCOUNT); do
 		updated0[$n]=$(scrub_status $n |
 			       awk '/^prior_updated/ { print $2 }')
+
+		echo "OI scrub on MDS$n status for the 3rd time:"
+		do_facet mds$n $LCTL get_param -n \
+			osd-ldiskfs.$(facet_svc mds$n).oi_scrub
+
 		[ ${updated0[$n]} -gt ${updated1[$n]} ] ||
 			error "(16) Auto trigger full scrub unexpectedly"
 	done
@@ -572,8 +612,13 @@ test_4c() {
 	for n in $(seq $MDSCOUNT); do
 		updated1[$n]=$(scrub_status $n |
 			       awk '/^prior_updated/ { print $2 }')
-		[ ${updated0[$n]} -eq ${updated1[$n]} ] ||
+		[ ${updated0[$n]} -eq ${updated1[$n]} ] || {
+			echo "OI scrub on MDS$n status for the 4th time:"
+			do_facet mds$n $LCTL get_param -n \
+				osd-ldiskfs.$(facet_svc mds$n).oi_scrub
+
 			error "(18) NOT auto trigger full scrub as expected"
+		}
 	done
 }
 run_test 4c "Auto trigger OI scrub if bad OI mapping was found (3)"
@@ -590,12 +635,12 @@ test_5() {
 	scrub_check_flags 4 recreated,inconsistent
 	mount_client $MOUNT || error "(5) Fail to start client!"
 	scrub_enable_auto
+	full_scrub_ratio 0
 
 	#define OBD_FAIL_OSD_SCRUB_DELAY	 0x190
 	do_nodes $(comma_list $(mdts_nodes)) \
 		$LCTL set_param fail_val=3 fail_loc=0x190
 
-	full_scrub_ratio 0
 	scrub_check_data 6
 	umount_client $MOUNT || error "(7) Fail to stop client!"
 	scrub_check_status 8 scanning
@@ -634,16 +679,24 @@ test_5() {
 		$LCTL set_param fail_val=3 fail_loc=0x190
 
 	local n
+	declare -a pids
+
 	for n in $(seq $MDSCOUNT); do
-		stat $DIR/$tdir/mds$n/${tfile}800 ||
-			error "(17) Failed to stat mds$n/${tfile}800"
+		stat $DIR/$tdir/mds$n/${tfile}800 &
+		pids[$n]=$!
 	done
 
-	scrub_check_status 18 scanning
+	sleep 3
+
+	scrub_check_status 17 scanning
 
 	do_nodes $(comma_list $(mdts_nodes)) \
 		$LCTL set_param fail_loc=0 fail_val=0
 
+	for n in $(seq $MDSCOUNT); do
+		wait ${pids[$n]} || error "(18) Fail to stat mds$n/${tfile}800"
+	done
+
 	scrub_check_status 19 completed
 	scrub_check_flags 20 ""
 }
@@ -657,12 +710,12 @@ test_6() {
 	scrub_check_flags 4 recreated,inconsistent
 	mount_client $MOUNT || error "(5) Fail to start client!"
 	scrub_enable_auto
+	full_scrub_ratio 0
 
 	#define OBD_FAIL_OSD_SCRUB_DELAY	 0x190
 	do_nodes $(comma_list $(mdts_nodes)) \
 		$LCTL set_param fail_val=2 fail_loc=0x190
 
-	full_scrub_ratio 0
 	scrub_check_data 6
 
 	# Sleep 5 sec to guarantee at least one object processed by OI scrub
@@ -735,12 +788,12 @@ test_7() {
 	scrub_check_flags 4 recreated,inconsistent
 	mount_client $MOUNT || error "(5) Fail to start client!"
 	scrub_enable_auto
+	full_scrub_ratio 0
 
 	#define OBD_FAIL_OSD_SCRUB_DELAY	 0x190
 	do_nodes $(comma_list $(mdts_nodes)) \
 		$LCTL set_param fail_val=3 fail_loc=0x190
 
-	full_scrub_ratio 0
 	scrub_check_data 6
 
 	local n
@@ -874,12 +927,12 @@ test_10a() {
 	scrub_check_flags 4 recreated,inconsistent
 	mount_client $MOUNT || error "(5) Fail to start client!"
 	scrub_enable_auto
+	full_scrub_ratio 0
 
 	#define OBD_FAIL_OSD_SCRUB_DELAY	 0x190
 	do_nodes $(comma_list $(mdts_nodes)) \
 		$LCTL set_param fail_val=1 fail_loc=0x190
 
-	full_scrub_ratio 0
 	scrub_check_data 6
 	scrub_check_status 7 scanning
 	umount_client $MOUNT || error "(8) Fail to stop client!"
@@ -938,7 +991,7 @@ test_11() {
 	check_mount_and_prep
 
 	for n in $(seq $MDSCOUNT); do
-		$LFS mkdir -i $((n - 1)) $DIR/$tdir/mds$n ||
+		test_mkdir -i $((n - 1)) $DIR/$tdir/mds$n ||
 			error "(1) Fail to mkdir $DIR/$tdir/mds$n"
 
 		createmany -o $DIR/$tdir/mds$n/f $CREATED ||