Branch b1_4

Use a timeout instead of killing writemany - the OST failover may not have
happened yet.
r=nathan

diff --git a/lustre/tests/recovery-small.sh b/lustre/tests/recovery-small.sh
index 4497fb0..5f5bb60 100755 (executable)
--- a/lustre/tests/recovery-small.sh
+++ b/lustre/tests/recovery-small.sh
@@ -512,15 +512,13 @@ test_51() {
 run_test 51 "failover MDS during recovery"
 
 test_52_guts() {
-       do_facet client "writemany -q -a $DIR/$tdir/$tfile 0 5" &
+       do_facet client "writemany -q -a $DIR/$tdir/$tfile 300 5" &
        CLIENT_PID=$!
        echo writemany pid $CLIENT_PID
        sleep 10
        FAILURE_MODE="SOFT"
        fail ost
        rc=0
-       sleep 5
-       kill -USR1 $CLIENT_PID
        wait $CLIENT_PID || rc=$?
        # active client process should see an EIO for down OST
        [ $rc -eq 5 ] && { echo "writemany correctly failed $rc" && return 0; }