LU-2297 test: fix replay-single test_74

unmount client before stopping OST, otherwise if client has dirty
data belongig to that OST, client umount will try to wait for
the data flushing back to the unavailable OST and waits forever
for the OST to come back.

Signed-off-by: Oleg Drokin <green@whamcloud.com>
Change-Id: I8af14134b5f417dfd96baef14e02d9f4c8912f6f
Reviewed-on: http://review.whamcloud.com/4717
Tested-by: Hudson
Reviewed-by: Nathaniel Clark <nathaniel.l.clark@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Jinshan Xiong <jinshan.xiong@whamcloud.com>

diff --git a/lustre/tests/replay-single.sh b/lustre/tests/replay-single.sh
index 94d4dfc..52a5642 100755 (executable)
--- a/lustre/tests/replay-single.sh
+++ b/lustre/tests/replay-single.sh
@@ -1961,8 +1961,8 @@ run_test 73c "open(O_CREAT), unlink, replay, reconnect at last_replay, close"
 test_74() {
     local clients=${CLIENTS:-$HOSTNAME}
 
-    stop ost1
     zconf_umount_clients $clients $MOUNT
+    stop ost1
     facet_failover $SINGLEMDS
     zconf_mount_clients $clients $MOUNT
     mount_facet ost1