LU-3196 tests: a few fixes for > 10 OSTs.

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

diff --git a/lustre/tests/replay-single.sh b/lustre/tests/replay-single.sh
index abc0c5e..07e7104 100755 (executable)
--- a/lustre/tests/replay-single.sh
+++ b/lustre/tests/replay-single.sh
@@ -21,16 +21,20 @@ GRANT_CHECK_LIST=${GRANT_CHECK_LIST:-""}
 require_dsh_mds || exit 0
 
 # Skip these tests
-# bug number:  17466 18857      LU1867
-ALWAYS_EXCEPT="61d   33a 33b    89      $REPLAY_SINGLE_EXCEPT"
+# bug number:  17466 18857      LU-1867 LU-1473
+ALWAYS_EXCEPT="61d   33a 33b    89      62     $REPLAY_SINGLE_EXCEPT"
 
 [ $(facet_fstype $SINGLEMDS) = "zfs" ] &&
-# bug number for skipped test:        LU-2342  LU-951
-       ALWAYS_EXCEPT="$ALWAYS_EXCEPT 20b 70a  73a"
+# bug number for skipped test:        LU-951
+       ALWAYS_EXCEPT="$ALWAYS_EXCEPT 73a"
 
 #                                                  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"
 
+[ $(facet_fstype $SINGLEMDS) = "zfs" ] &&
+# bug number for skipped test:        LU-3127
+        ALWAYS_EXCEPT="$ALWAYS_EXCEPT 73b"
+
 build_test_filter
 
 check_and_setup_lustre
@@ -438,7 +442,7 @@ test_20b() { # bug 10480
     wait_mds_ost_sync || return 3
     AFTERUSED=`df -P $DIR | tail -1 | awk '{ print $3 }'`
     log "before $BEFOREUSED, after $AFTERUSED"
-    [ $AFTERUSED -gt $((BEFOREUSED + 20)) ] && \
+    (( $AFTERUSED > $BEFOREUSED + $(fs_log_size) )) &&
         error "after $AFTERUSED > before $BEFOREUSED"
     return 0
 }
@@ -918,7 +922,8 @@ test_44a() { # was test_44
 
        for i in `seq 1 10`; do
                echo "$i of 10 ($(date +%s))"
-               do_facet $SINGLEMDS "lctl get_param -n mdt.*.mdt.timeouts | grep service"
+               do_facet $SINGLEMDS \
+                       "lctl get_param -n md[ts].*.mdt.timeouts | grep service"
 #define OBD_FAIL_TGT_CONN_RACE     0x701
                do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000701"
                 # lctl below may fail, it is valid case
@@ -940,7 +945,8 @@ test_44b() {
 
        for i in `seq 1 10`; do
                echo "$i of 10 ($(date +%s))"
-               do_facet $SINGLEMDS "lctl get_param -n mdt.*.mdt.timeouts | grep service"
+               do_facet $SINGLEMDS \
+                       "lctl get_param -n md[ts].*.mdt.timeouts | grep service"
         #define OBD_FAIL_TGT_DELAY_RECONNECT 0x704
                do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000704"
         # lctl below may fail, it is valid case
@@ -1892,6 +1898,7 @@ test_70b () {
        # set duration to 900 because it takes some time to boot node
        [ "$FAILURE_MODE" = HARD ] && duration=900
 
+       local elapsed
        local start_ts=$(date +%s)
        local cmd="rundbench 1 -t $duration"
        local pid=""
@@ -1900,15 +1907,24 @@ test_70b () {
                DBENCH_LIB=$DBENCH_LIB TESTSUITE=$TESTSUITE TESTNAME=$TESTNAME \
                MOUNT=$MOUNT DIR=$DIR/$tdir/\\\$(hostname) LCTL=$LCTL $cmd" &
        pid=$!
+
+       #LU-1897 wait for all dbench copies to start
+       while ! check_for_process $clients dbench; do
+               elapsed=$(($(date +%s) - start_ts))
+               if [ $elapsed -gt $duration]; then
+                       killall_process $clients dbench
+                       error "dbench failed to start on $clients!"
+               fi
+               sleep 1
+       done
+
        log "Started rundbench load pid=$pid ..."
 
-       # give rundbench a chance to start, bug 24118
-       sleep 12
-       local elapsed=$(($(date +%s) - start_ts))
+       elapsed=$(($(date +%s) - start_ts))
        local num_failovers=0
        while [ $elapsed -lt $duration ]; do
                if ! check_for_process $clients dbench; then
-                       error_noexit "dbench not running on some of $clients!"
+                       error_noexit "dbench stopped on some of $clients!"
                        killall_process $clients dbench
                        break
                fi
@@ -2737,7 +2753,7 @@ test_90() { # bug 19494
     # Before failing an OST, get its obd name and index
     local varsvc=${ostfail}_svc
     local obd=$(do_facet $ostfail lctl get_param -n obdfilter.${!varsvc}.uuid)
-    local index=${obd:(-6):1}
+       local index=$(($(facet_number $ostfail) - 1))
 
     echo "Fail $ostfail $obd, display the list of affected files"
     shutdown_facet $ostfail || return 2