require_dsh_mds || exit 0
-MCREATE=${MCREATE:-mcreate}
SAVED_MDSSIZE=${MDSSIZE}
SAVED_OSTSIZE=${OSTSIZE}
SAVED_OSTCOUNT=${OSTCOUNT}
ALWAYS_EXCEPT="$ALWAYS_EXCEPT 2c"
[[ $(lustre_version_code ost1) -lt $(version_code 2.5.55) ]] &&
- ALWAYS_EXCEPT="$ALWAYS_EXCEPT 11 12 13 14 15 16 17 18 19"
+ ALWAYS_EXCEPT="$ALWAYS_EXCEPT 11 12 13 14 15 16 17 18 19 20 21"
build_test_filter
local POS1=$($SHOW_NAMESPACE |
awk '/^latest_start_position/ { print $2 }' |
tr -d ',')
- [ $POS0 -lt $POS1 ] ||
+ [[ $POS0 -lt $POS1 ]] ||
error "(7) Expect larger than: $POS0, but got $POS1"
do_facet $SINGLEMDS $LCTL set_param fail_loc=0 fail_val=0
awk '/^latest_start_position/ { print $4 }')
if [ "$D_POS0" == "N/A" -o "$D_POS1" == "N/A" ]; then
- [ $O_POS0 -lt $O_POS1 ] ||
+ [[ $O_POS0 -lt $O_POS1 ]] ||
error "(7.1) $O_POS1 is not larger than $O_POS0"
else
- [ $D_POS0 -lt $D_POS1 ] ||
+ [[ $D_POS0 -lt $D_POS1 ]] ||
error "(7.2) $D_POS1 is not larger than $D_POS0"
fi
start $SINGLEMDS $MDT_DEVNAME $MOUNT_OPTS_SCRUB > /dev/null ||
error "(14) Fail to start MDS!"
+ local timeout=$(max_recovery_time)
+ local timer=0
+
+ while [ $timer -lt $timeout ]; do
+ STATUS=$(do_facet $SINGLEMDS "$LCTL get_param -n \
+ mdt.${MDT_DEV}.recovery_status |
+ awk '/^status/ { print \\\$2 }'")
+ [ "$STATUS" != "RECOVERING" ] && break;
+ sleep 1
+ timer=$((timer + 1))
+ done
+
+ [ $timer != $timeout ] ||
+ error "(14.1) recovery timeout"
+
STATUS=$($SHOW_NAMESPACE | awk '/^status/ { print $2 }')
[ "$STATUS" == "crashed" ] ||
error "(15) Expect 'crashed', but got '$STATUS'"
start $SINGLEMDS $MDT_DEVNAME $MOUNT_OPTS_SCRUB > /dev/null ||
error "(19) Fail to start MDS!"
+ timer=0
+ while [ $timer -lt $timeout ]; do
+ STATUS=$(do_facet $SINGLEMDS "$LCTL get_param -n \
+ mdt.${MDT_DEV}.recovery_status |
+ awk '/^status/ { print \\\$2 }'")
+ [ "$STATUS" != "RECOVERING" ] && break;
+ sleep 1
+ timer=$((timer + 1))
+ done
+
+ [ $timer != $timeout ] ||
+ error "(19.1) recovery timeout"
+
STATUS=$($SHOW_NAMESPACE | awk '/^status/ { print $2 }')
[ "$STATUS" == "paused" ] ||
error "(20) Expect 'paused', but got '$STATUS'"
createmany -o $DIR/$tdir/f $((count + 32))
do_facet ost1 $LCTL set_param fail_loc=0
+ start_full_debug_logging
+
# exhaust other pre-created dangling cases
count=$(precreated_ost_obj_count 0 0)
createmany -o $DIR/$tdir/a $count ||
echo "'ls' should success after layout LFSCK repairing"
ls -ail $DIR/$tdir > /dev/null || error "(9) ls should success."
+ stop_full_debug_logging
}
run_test 14 "LFSCK can repair MDT-object with dangling reference"
fi
echo "Move the files from ./lustre/lost+found/MDTxxxx to namespace"
- mv $MOUNT/.lustre/lost+found/MDT0000/R-${fid1} $DIR/$tdir/a1/f1 ||
- error "(5) Fail to move $MOUNT/.lustre/lost+found/MDT0000/R-${fid1}"
+ mv $MOUNT/.lustre/lost+found/MDT0000/${fid1}-R-0 $DIR/$tdir/a1/f1 ||
+ error "(5) Fail to move $MOUNT/.lustre/lost+found/MDT0000/${fid1}-R-0"
if [ $MDSCOUNT -ge 2 ]; then
- local name=$MOUNT/.lustre/lost+found/MDT0001/R-${fid2}
+ local name=$MOUNT/.lustre/lost+found/MDT0001/${fid2}-R-0
mv $name $DIR/$tdir/a2/f2 || error "(6) Fail to move $name"
fi
if [ $MDSCOUNT -ge 2 ]; then
$LFS mkdir -i 1 $DIR/$tdir/a2
- $LFS setstripe -c 2 -i 1 -s 1M $DIR/$tdir/a2
- do_facet ost2 $LCTL set_param fail_loc=0x1617
+ $LFS setstripe -c 1 -i 0 -s 1M $DIR/$tdir/a2
dd if=/dev/zero of=$DIR/$tdir/a2/f2 bs=1M count=2
$LFS getstripe $DIR/$tdir/a2/f2
fi
done
if [ $MDSCOUNT -ge 2 ]; then
- expected=3
+ expected=2
else
expected=1
fi
error "(5) Expect 0 fixed on mds2, but got: $repaired"
fi
- echo "There should be some stub under .lustre/lost+found/MDT0001/"
- ls -ail $MOUNT/.lustre/lost+found/MDT0001/N-* &&
+ echo "There should NOT be some stub under .lustre/lost+found/MDT0001/"
+ ls -ail $MOUNT/.lustre/lost+found/MDT0001/*-N-0 &&
error "(6) .lustre/lost+found/MDT0001/ should be empty"
echo "There should be some stub under .lustre/lost+found/MDT0000/"
- ls -ail $MOUNT/.lustre/lost+found/MDT0000/N-* ||
+ ls -ail $MOUNT/.lustre/lost+found/MDT0000/*-N-0 ||
error "(7) .lustre/lost+found/MDT0000/ should not be empty"
}
run_test 18c "Find out orphan OST-object and repair it (3)"
awk '/^status/ { print \\\$2 }'" "scanning-phase2" 6 ||
error "(3.0) MDS1 is not the expected 'scanning-phase2'"
- # LU-3469: before osp_sync() is enabled, wait for a while to guarantee
- # that former async repair operations have been executed on the OST(s).
- sync
- sleep 2
-
do_facet $SINGLEMDS $LCTL set_param fail_val=0 fail_loc=0
for k in $(seq $MDSCOUNT); do
error "(6) Expect 1 orphan has been fixed, but got: $repaired"
echo "There should be stub file under .lustre/lost+found/MDT0000/"
- local cname=$(ls $MOUNT/.lustre/lost+found/MDT0000/C-*)
+ local cname=$(ls $MOUNT/.lustre/lost+found/MDT0000/*-C-0)
[ ! -z $name ] ||
error "(7) .lustre/lost+found/MDT0000/ should not be empty"
LOV_PATTERN_F_HOLE=0x40000000
#
- # R-${fid0} is the old f0
+ # ${fid0}-R-0 is the old f0
#
- local name="$MOUNT/.lustre/lost+found/MDT0000/R-${fid0}"
+ local name="$MOUNT/.lustre/lost+found/MDT0000/${fid0}-R-0"
echo "Check $name, which is the old f0"
$LFS getstripe -v $name || error "(5.1) cannot getstripe on $name"
rm -f $name || error "(5.9) cannot unlink $name"
#
- # R-${fid1} contains the old f1's stripe1 (and stripe2 if OSTs > 2)
+ # ${fid1}-R-0 contains the old f1's stripe1 (and stripe2 if OSTs > 2)
#
- name="$MOUNT/.lustre/lost+found/MDT0000/R-${fid1}"
+ name="$MOUNT/.lustre/lost+found/MDT0000/${fid1}-R-0"
if [ $OSTCOUNT -gt 2 ]; then
echo "Check $name, it contains the old f1's stripe1 and stripe2"
else
rm -f $name || error "(6.13) cannot unlink $name"
#
- # R-${fid2} it contains the old f2's stripe0 (and stripe2 if OSTs > 2)
+ # ${fid2}-R-0 it contains the old f2's stripe0 (and stripe2 if OSTs > 2)
#
- name="$MOUNT/.lustre/lost+found/MDT0000/R-${fid2}"
+ name="$MOUNT/.lustre/lost+found/MDT0000/${fid2}-R-0"
if [ $OSTCOUNT -gt 2 ]; then
echo "Check $name, it contains the old f2's stripe0 and stripe2"
else
[ $OSTCOUNT -le 2 ] && return
#
- # R-${fid3} should contains the old f3's stripe0 and stripe1
+ # ${fid3}-R-0 should contains the old f3's stripe0 and stripe1
#
- name="$MOUNT/.lustre/lost+found/MDT0000/R-${fid3}"
+ name="$MOUNT/.lustre/lost+found/MDT0000/${fid3}-R-0"
echo "Check $name, which contains the old f3's stripe0 and stripe1"
$LFS getstripe -v $name || error "(8.1) cannot getstripe on $name"
}
run_test 20 "Handle the orphan with dummy LOV EA slot properly"
+test_21() {
+ [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.5.59) ]] &&
+ skip "ignore the test if MDS is older than 2.5.59" && exit 0
+
+ check_mount_and_prep
+ createmany -o $DIR/$tdir/f 100 || error "(0) Fail to create 100 files"
+
+ echo "Start all LFSCK components by default (-s 1)"
+ do_facet mds1 $LCTL lfsck_start -M ${FSNAME}-MDT0000 -s 1 -r ||
+ error "Fail to start LFSCK"
+
+ echo "namespace LFSCK should be in 'scanning-phase1' status"
+ local STATUS=$($SHOW_NAMESPACE | awk '/^status/ { print $2 }')
+ [ "$STATUS" == "scanning-phase1" ] ||
+ error "Expect namespace 'scanning-phase1', but got '$STATUS'"
+
+ echo "layout LFSCK should be in 'scanning-phase1' status"
+ STATUS=$($SHOW_LAYOUT | awk '/^status/ { print $2 }')
+ [ "$STATUS" == "scanning-phase1" ] ||
+ error "Expect layout 'scanning-phase1', but got '$STATUS'"
+
+ echo "Stop all LFSCK components by default"
+ do_facet mds1 $LCTL lfsck_stop -M ${FSNAME}-MDT0000 ||
+ error "Fail to stop LFSCK"
+}
+run_test 21 "run all LFSCK components by default"
+
$LCTL set_param debug=-lfsck > /dev/null || true
# restore MDS/OST size
git://git.whamcloud.com / fs / lustre-release.git / blobdiff