b=19919 Supply a absolute path.

[fs/lustre-release.git] / lustre / tests / recovery-mds-scale.sh

diff --git a/lustre/tests/recovery-mds-scale.sh b/lustre/tests/recovery-mds-scale.sh
index d4867f5..4ad886f 100644 (file)
--- a/lustre/tests/recovery-mds-scale.sh
+++ b/lustre/tests/recovery-mds-scale.sh
@@ -14,6 +14,7 @@ CLEANUP=${CLEANUP:-""}
 init_test_env $@
 
 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
+init_logging
 
 TESTSUITELOG=${TESTSUITELOG:-$TMP/$(basename $0 .sh)}
 DEBUGLOG=$TESTSUITELOG.debug
@@ -124,7 +125,7 @@ summary_and_cleanup () {
     local result=PASS
     [ $rc -eq 0 ] || result=FAIL
 
-    log "Duraion:                $DURATION
+    log "Duration:                $DURATION
 Server failover period: $SERVER_FAILOVER_PERIOD seconds
 Exited after:           $ELAPSED seconds
 Number of failovers before exit:
@@ -181,12 +182,8 @@ fi
 start_client_loads $NODES_TO_USE
 
 echo clients load pids:
-if ! do_nodes $NODES_TO_USE "set -x; echo \$(hostname): && cat $LOAD_PID_FILE"; then
-    if [ -e $DEBUGLOG ]; then
-        exec 2<&-
-        cat $DEBUGLOG
+if ! do_nodesv $NODES_TO_USE "cat $LOAD_PID_FILE"; then
         exit 3
-    fi
 fi
 
 START_TS=$(date +%s)
@@ -261,7 +258,10 @@ while [ $ELAPSED -lt $DURATION -a ! -e $END_RUN_FILE ]; do
     if [ $sleep -lt $MINSLEEP ]; then
         reqfail=$((reqfail +1))
         log "WARNING: failover and two check_client_loads time exceeded SERVER_FAILOVER_PERIOD - MINSLEEP !
-Failed to meet interval $reqfail times ( REQFAIL=$REQFAIL ); have sleep=$sleep"
+Failed to load the filesystem with I/O for a minimum period of $MINSLEEP $reqfail times ( REQFAIL=$REQFAIL ).
+This iteration, the load was only applied for sleep=$sleep seconds.
+Probably the hardware is taking excessively long to boot.
+Try to increase SERVER_FAILOVER_PERIOD (current is $SERVER_FAILOVER_PERIOD), bug 20918"
         [ $reqfail -gt $REQFAIL ] && exit 6 
     fi