local num=$(echo $facet | tr -d "mds")
local dev=$(mdsdevname $num)
echo "start mds service on `facet_active_host $facet`"
- start $facet ${dev} $MDS_MOUNT_OPTS || return 94
+ start $facet ${dev} $MDS_MOUNT_OPTS $@ || return 94
+}
+
+start_mgsmds() {
+ if ! combined_mgs_mds ; then
+ start_mgs
+ fi
+ start_mds $@
}
stop_mds() {
start_ost() {
echo "start ost1 service on `facet_active_host ost1`"
- start ost1 `ostdevname 1` $OST_MOUNT_OPTS || return 95
+ start ost1 `ostdevname 1` $OST_MOUNT_OPTS $@ || return 95
}
stop_ost() {
start_ost2() {
echo "start ost2 service on `facet_active_host ost2`"
- start ost2 `ostdevname 2` $OST_MOUNT_OPTS || return 92
+ start ost2 `ostdevname 2` $OST_MOUNT_OPTS $@ || return 92
}
stop_ost2() {
}
run_test 27b "Reacquire MGS lock after failover"
-test_28() {
+test_28a() {
setup
TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
- PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
+ PARAM="llite.$FSNAME.max_read_ahead_whole_mb"
ORIG=$($TEST)
FINAL=$(($ORIG + 1))
set_and_check client "$TEST" "$PARAM" $FINAL || return 3
set_and_check client "$TEST" "$PARAM" $ORIG || return 5
cleanup
}
-run_test 28 "permanent parameter setting"
+run_test 28a "permanent parameter setting"
+
+check_28b() {
+ local NODE=$1
+ shift
+ set_and_check $NODE "$LCTL get_param -n $1*.$2 | head -1" "$1.$2" "$3" || \
+ error "conf_param $1.$2 failed"
+}
+
+test_28b() {
+ setup > /dev/null
+ # should error
+ do_facet mgs "$LCTL conf_param foo=1 2>/dev/null" && \
+ error "Bad format should fail"
+ do_facet mgs "$LCTL conf_param osc.notanfs-OST0000.active=0 2>/dev/null" && \
+ error "Setting on unknown fs should fail"
+ do_facet mgs "$LCTL conf_param osc.$FSNAME-OST00000.active=0 2>/dev/null" && \
+ error "Bad target name should fail"
+ # should succeed
+ check_28b mds mdt.$FSNAME-MDT0000 capa_timeout 1500
+ check_28b mds mdt.$FSNAME-MDT* identity_expire 150
+ check_28b mds mdd.$FSNAME-MDT0000 atime_diff 15
+ check_28b mds mdd.$FSNAME-MDT* sync_permission 0
+ check_28b ost1 obdfilter.$FSNAME-OST0000 client_cache_seconds 15
+ check_28b ost1 obdfilter.$FSNAME-OST* client_cache_count 15
+ check_28b mds lov.$FSNAME-MDT0000 qos_maxage "15 Sec"
+ check_28b mds lov.$FSNAME-MDT0000 qos_prio_free "15%"
+ check_28b client mdc.$FSNAME-MDT0000 max_rpcs_in_flight 15
+ check_28b client osc.$FSNAME-OST0000 active 0
+ check_28b client osc.$FSNAME-OST0000 active 1
+ check_28b client osc.$FSNAME-OST0000 max_dirty_mb 15
+ check_28b client llite.$FSNAME max_read_ahead_mb 15
+ set_and_check client "$LCTL get_param -n at_max" "sys.$FSNAME.at_max" 1500 || \
+ error "conf_param sys.fsname.at_max failed"
+ cleanup > /dev/null
+}
+run_test 28b "permanent parameter setting, set_param syntax"
test_29() {
[ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
- setup > /dev/null 2>&1
+ setup > /dev/null 2>&1
start_ost2
sleep 10
- local PARAM="$FSNAME-OST0001.osc.active"
- local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
- local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
+ local PARAM="osc.$FSNAME-OST0001.active"
+ local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
+ local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
- ACTV=$(lctl get_param -n $PROC_ACT)
+ ACTV=$(lctl get_param -n $PROC_ACT)
DEAC=$((1 - $ACTV))
set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $DEAC || return 2
- # also check ost_server_uuid status
+ # also check ost_server_uuid status
RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV)
if [ -z "$RESULT" ]; then
echo "Live client not deactivated: $(lctl get_param -n $PROC_UUID)"
ORIG=$($TEST)
LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
for i in ${LIST[@]}; do
- set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3
+ set_and_check client "$TEST" "llite.$FSNAME.max_read_ahead_whole_mb" $i || return 3
done
# make sure client restart still works
umount_client $MOUNT
pass
echo Erase parameter setting
- do_facet mgs "$LCTL conf_param -d $FSNAME.llite.max_read_ahead_whole_mb" || return 6
+ do_facet mgs "$LCTL conf_param -d llite.$FSNAME.max_read_ahead_whole_mb" || return 6
umount_client $MOUNT
mount_client $MOUNT || return 6
FINAL=$($TEST)
echo "Using fake nid $NEW"
TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
- set_and_check client "$TEST" "$FSNAME-OST0000.failover.node" $NEW || error "didn't add failover nid $NEW"
+ set_and_check client "$TEST" "osc.$FSNAME-OST0000.failover.node" $NEW || error "didn't add failover nid $NEW"
NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
echo $NIDS
NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1))
echo "should have 2 failover nids: $NIDCOUNT"
[ $NIDCOUNT -eq 2 ] || error "Failover nid not added"
- do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" || error "conf_param delete failed"
+ do_facet mgs "$LCTL conf_param -d osc.$FSNAME-OST0000.failover.node" || error "conf_param delete failed"
umount_client $MOUNT
mount_client $MOUNT || return 3
start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
start fs2ost $fs2ostdev $OST_MOUNT_OPTS
- do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" || rc=1
+ do_facet mgs "$LCTL conf_param sys.$FSNAME2.timeout=200" || rc=1
mkdir -p $MOUNT2
mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || rc=2
echo "ok."
sleep 1
cleanup
}
-run_test 34a "umount with opened file should be fail"
+run_test 34a "umount with opened file should fail"
test_34b() {
log "Set up a fake failnode for the MDS"
FAKENID="127.0.0.2"
local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
- do_facet $SINGLEMDS $LCTL conf_param ${device}.failover.node=$FAKENID || return 4
+ do_facet $SINGLEMDS $LCTL conf_param mdc.${device}.failover.node=$FAKENID || return 4
log "Wait for RECONNECT_INTERVAL seconds (10s)"
sleep 10
FAKENID="127.0.0.2"
local device=$(do_facet mds "$LCTL get_param -n devices" | \
awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
- do_facet mds "$LCTL conf_param ${device}.failover.node=$FAKENID" || \
+ do_facet mds "$LCTL conf_param mdc.${device}.failover.node=$FAKENID" || \
return 1
local at_max_saved=0
at_max_set 0 mds client
fi
- mkdir -p $MOUNT/testdir
- touch $MOUNT/testdir/test
+ mkdir -p $MOUNT/$tdir
log "Injecting EBUSY on MDS"
# Setting OBD_FAIL_MDS_RESEND=0x136
do_facet mds "$LCTL set_param fail_loc=0x80000136" || return 2
- log "Stat on a test file"
- stat $MOUNT/testdir/test
+ $LCTL set_param mdc.${FSNAME}*.stats=clear
+
+ log "Creating a test file and stat it"
+ touch $MOUNT/$tdir/$tfile
+ stat $MOUNT/$tdir/$tfile
log "Stop injecting EBUSY on MDS"
do_facet mds "$LCTL set_param fail_loc=0" || return 3
- rm -f $MOUNT/testdir/test
+ rm -f $MOUNT/$tdir/$tfile
log "done"
# restore adaptive timeout
$LCTL dk $TMP/lustre-log-$TESTNAME.log
+ CONNCNT=`$LCTL get_param mdc.${FSNAME}*.stats | awk '/mds_connect/{print $2}'`
+
# retrieve from the log if the client has ever tried to
# contact the fake server after the loss of connection
FAILCONN=`awk "BEGIN {ret = 0;}
log "ERROR: The client tried to reconnect to the failover server while the primary was busy" && \
return 5
+ # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
+ # Reconnects are supposed to be rate limited to one every 5s
+ [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] && \
+ log "ERROR: Too many reconnects $CONNCNT" && \
+ return 6
+
cleanup
# remove nid settings
writeconf
test_42() { #bug 14693
setup
check_mount || return 2
- do_facet mgs $LCTL conf_param lustre.llite.some_wrong_param=10
+ do_facet mgs $LCTL conf_param llite.$FSNAME.some_wrong_param=10
umount_client $MOUNT
mount_client $MOUNT || return 1
cleanup
chmod ugo+x $DIR || error "chmod 0 failed"
set_and_check mds \
"lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
- "$FSNAME.mdt.root_squash" \
+ "mdt.$FSNAME-MDT*.root_squash" \
"0:0"
set_and_check mds \
"lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
- "$FSNAME.mdt.nosquash_nids" \
+ "mdt.$FSNAME-MDT*.nosquash_nids" \
"NONE"
#
#
set_and_check mds \
"lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
- "$FSNAME.mdt.root_squash" \
+ "mdt.$FSNAME-MDT*.root_squash" \
"$RUNAS_ID:$RUNAS_ID"
ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
set_and_check mds \
"lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
- "$FSNAME-MDTall.mdt.nosquash_nids" \
+ "mdt.$FSNAME-MDT*.nosquash_nids" \
"$NIDLIST"
ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
# We need to expand $parampat, but it may match multiple parameters, so
# we'll pick the first one
- if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
+ paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1)
+ if [ -z "$paramp" ]; then
error "Couldn't expand ${parampat}.threads_min parameter name"
return 22
fi
}
run_test 53b "check MDT thread count params"
-if ! combined_mgs_mds ; then
- stop mgs
-fi
-
run_llverfs()
{
local dir=$1
}
run_test 56 "check big indexes"
+test_57() { # bug 22656
+ local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
+ writeconf
+ do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" || error "tunefs failed"
+ start_mgsmds
+ start_ost && error "OST registration from failnode should fail"
+ stop_mds
+ reformat
+}
+run_test 57 "initial registration from failnode should fail (should return errs)"
+
+count_osts() {
+ do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
+}
+
+test_59() {
+ start_mgsmds >> /dev/null
+ local C1=$(count_osts)
+ if [ $C1 -eq 0 ]; then
+ start_ost >> /dev/null
+ C1=$(count_osts)
+ fi
+ stopall
+ echo "original ost count: $C1 (expect > 0)"
+ [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
+ start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
+ local C2=$(count_osts)
+ echo "after mdt writeconf count: $C2 (expect 0)"
+ [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
+ echo "OST start without writeconf should fail:"
+ start_ost >> /dev/null && error "OST start without writeconf didn't fail"
+ echo "OST start with writeconf should succeed:"
+ start_ost -o writeconf >> /dev/null || error "OST1 start failed"
+ local C3=$(count_osts)
+ echo "after ost writeconf count: $C3 (expect 1)"
+ [ $C3 -eq 1 ] || error "new OST writeconf should add:"
+ start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
+ local C4=$(count_osts)
+ echo "after ost2 writeconf count: $C4 (expect 2)"
+ [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
+ stop_ost2 >> /dev/null
+ cleanup_nocli >> /dev/null
+}
+run_test 59 "writeconf mount option"
+
+
+if ! combined_mgs_mds ; then
+ stop mgs
+fi
cleanup_gss
equals_msg `basename $0`: test complete
[ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG && grep -q FAIL $TESTSUITELOG && exit 1 || true
git://git.whamcloud.com / fs / lustre-release.git / blobdiff