LU-5810 tests: add client hostname to lctl mark

[fs/lustre-release.git] / lustre / tests / replay-single.sh

diff --git a/lustre/tests/replay-single.sh b/lustre/tests/replay-single.sh
index feef23b..0df8804 100755 (executable)
--- a/lustre/tests/replay-single.sh
+++ b/lustre/tests/replay-single.sh
@@ -22,7 +22,7 @@ require_dsh_mds || exit 0
 
 # Skip these tests
 # bug number for skipped tests:
-#                                    b=17466/LU-472
+#                                    LU-472
 ALWAYS_EXCEPT="$REPLAY_SINGLE_EXCEPT 61d"
 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
 
@@ -32,8 +32,8 @@ case "$(lsb_release -sr)" in  # only disable tests for el7
        ;;
 esac
 
-#                                                  63 min  7 min  AT AT AT AT"
-[ "$SLOW" = "no" ] && EXCEPT_SLOW="1 2 3 4 6 12 16 44a      44b    65 66 67 68"
+#                                  7.5  (min)"
+[ "$SLOW" = "no" ] && EXCEPT_SLOW="44b"
 
 [ $(facet_fstype $SINGLEMDS) = "zfs" ] &&
 # bug number for skipped test:        LU-1867  LU-3127
@@ -873,7 +873,7 @@ test_40(){
        lctl get_param mdc.*.connect_flags | grep -q layout_lock &&
                skip "layout_lock needs MDS connection for IO" && return 0
 
-       $LCTL mark multiop $MOUNT/$tfile OS_c
+       $LCTL mark "$HOSTNAME multiop $MOUNT/$tfile OS_c"
        multiop $MOUNT/$tfile OS_c  &
        PID=$!
        writeme -s $MOUNT/${tfile}-2 &
@@ -2036,6 +2036,10 @@ check_for_process () {
 
 test_70b () {
        local clients=${CLIENTS:-$HOSTNAME}
+       local mdscount=$MDSCOUNT
+
+       # until LU-6844 is fixed, run on one MDT instead of disabling test
+       mdscount=1
 
        zconf_mount_clients $clients $MOUNT
 
@@ -2048,9 +2052,9 @@ test_70b () {
        local start_ts=$(date +%s)
        local cmd="rundbench 1 -t $duration"
        local pid=""
-       if [ $MDSCOUNT -ge 2 ]; then
-               test_mkdir -p -c$MDSCOUNT $DIR/$tdir
-               $LFS setdirstripe -D -c$MDSCOUNT $DIR/$tdir
+       if [ $mdscount -ge 2 ]; then
+               test_mkdir -p -c$mdscount $DIR/$tdir
+               $LFS setdirstripe -D -c$mdscount $DIR/$tdir
        fi
        do_nodesv $clients "set -x; MISSING_DBENCH_OK=$MISSING_DBENCH_OK \
                PATH=\$PATH:$LUSTRE/utils:$LUSTRE/tests/:$DBENCH_LIB \
@@ -2087,7 +2091,7 @@ test_70b () {
                log "$TESTNAME fail mds$fail_index $num_failovers times"
                fail mds$fail_index
                elapsed=$(($(date +%s) - start_ts))
-               if [ $fail_index -ge $MDSCOUNT ]; then
+               if [ $fail_index -ge $mdscount ]; then
                        fail_index=1
                else
                        fail_index=$((fail_index+1))
@@ -2126,8 +2130,10 @@ random_fail_mdt() {
 
 cleanup_70c() {
        trap 0
-       kill -9 $tar_70c_pid
+       rm -f $DIR/replay-single.70c.lck
+       rm -rf /$DIR/$tdir
 }
+
 test_70c () {
        local clients=${CLIENTS:-$HOSTNAME}
        local rc=0
@@ -2144,16 +2150,14 @@ test_70c () {
 
        trap cleanup_70c EXIT
        (
-               while true; do
+               while [ ! -e $DIR/replay-single.70c.lck ]; do
                        test_mkdir -p -c$MDSCOUNT $DIR/$tdir || break
                        if [ $MDSCOUNT -ge 2 ]; then
                                $LFS setdirstripe -D -c$MDSCOUNT $DIR/$tdir ||
                                error "set default dirstripe failed"
                        fi
                        cd $DIR/$tdir || break
-                       tar cf - /etc | tar xf - || error "tar failed"
-                       cd $DIR || break
-                       rm -rf $DIR/$tdir || break
+                       tar cf - /etc | tar xf - || error "tar failed in loop"
                done
        )&
        tar_70c_pid=$!
@@ -2162,6 +2166,9 @@ test_70c () {
        random_fail_mdt $MDSCOUNT $duration $tar_70c_pid
        kill -0 $tar_70c_pid || error "tar $tar_70c_pid stopped"
 
+       touch $DIR/replay-single.70c.lck
+       wait $tar_70c_pid || error "$?: tar failed"
+
        cleanup_70c
        true
 }