init_test_env $@
. ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
+init_logging
TESTSUITELOG=${TESTSUITELOG:-$TMP/$(basename $0 .sh)}
DEBUGLOG=$TESTSUITELOG.debug
check_and_setup_lustre
rm -rf $DIR/[df][0-9]*
+max_recov_time=$(max_recovery_time)
+
# the test node needs to be insulated from a lustre failure as much as possible,
# so not even loading the lustre modules is ideal.
# -- umount lustre
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:
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)
-CURRENT_TS=$START_TS
-
MINSLEEP=${MINSLEEP:-120}
REQFAIL_PERCENT=${REQFAIL_PERCENT:-3} # bug17839 comment 62
REQFAIL=${REQFAIL:-$(( DURATION / SERVER_FAILOVER_PERIOD * REQFAIL_PERCENT / 100))}
reqfail=0
sleep=0
+
+START_TS=$(date +%s)
+CURRENT_TS=$START_TS
+
while [ $ELAPSED -lt $DURATION -a ! -e $END_RUN_FILE ]; do
# In order to perform the
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.
+Estimated max recovery time : $max_recov_time
+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