ALWAYS_EXCEPT="$ALWAYS_EXCEPT 11 12 13 14 15 16 17 18 19 20 21"
[[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.6.50) ]] &&
- ALWAYS_EXCEPT="$ALWAYS_EXCEPT 2d 2e 3 22 23 24"
+ ALWAYS_EXCEPT="$ALWAYS_EXCEPT 2d 2e 3 22 23 24 25 26 27 28 29 30"
build_test_filter
awk '/^updated_phase1/ { print $2 }')
[ $repaired -eq 1 ] ||
- error "(5) Fail to repair missed FID-in-LMA: $repaired"
+ error "(5) Fail to repair the missing FID-in-LMA: $repaired"
do_facet $SINGLEMDS $LCTL set_param fail_loc=0
mount_client $MOUNT || error "(6) Fail to start client!"
do_facet $SINGLEMDS $LCTL set_param fail_loc=0
}
-run_test 1b "LFSCK can find out and repair missed FID-in-LMA"
+run_test 1b "LFSCK can find out and repair the missing FID-in-LMA"
test_2a() {
lfsck_prep 1 1
[ "$dummyname" == "$DIR/$tdir/dummy" ] ||
error "(8) Fail to repair linkEA: $dummyfid $dummyname"
}
-run_test 2d "LFSCK can recover the missed linkEA entry"
+run_test 2d "LFSCK can recover the missing linkEA entry"
test_2e()
{
local MIN_SPEED=$(((BASE_SPEED1 * (RUN_TIME1 - TIME_DIFF) + \
BASE_SPEED2 * (RUN_TIME2 - TIME_DIFF)) / \
(RUN_TIME1 + RUN_TIME2) * 8 / 10))
- # Account for slow ZFS performance - LU-4934
- [ $SPEED -gt $MIN_SPEED ] || [ $(facet_fstype $SINGLEMDS) -eq zfs ] ||
- error "(5) Got speed $SPEED, expected more than $MIN_SPEED"
+ [ $SPEED -gt $MIN_SPEED ] || {
+ if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
+ error_ignore LU-5624 \
+ "(5.1) Got speed $SPEED, expected more than $MIN_SPEED"
+ else
+ error \
+ "(5.2) Got speed $SPEED, expected more than $MIN_SPEED"
+ fi
+ }
# MAX_MARGIN = 1.2 = 12 / 10
MAX_SPEED=$(((BASE_SPEED1 * (RUN_TIME1 + TIME_DIFF) + \
local MIN_SPEED=$(((BASE_SPEED1 * (RUN_TIME1 - TIME_DIFF) + \
BASE_SPEED2 * (RUN_TIME2 - TIME_DIFF)) / \
(RUN_TIME1 + RUN_TIME2) * 8 / 10))
- [ $SPEED -gt $MIN_SPEED ] ||[ $(facet_fstype $SINGLEMDS) -eq zfs ] ||
- error "(9) Got speed $SPEED, expected more than $MIN_SPEED"
+ [ $SPEED -gt $MIN_SPEED ] || {
+ if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
+ error_ignore LU-5624 \
+ "(9.1) Got speed $SPEED, expected more than $MIN_SPEED"
+ else
+ error \
+ "(9.2) Got speed $SPEED, expected more than $MIN_SPEED"
+ fi
+ }
# MAX_MARGIN = 1.2 = 12 / 10
MAX_SPEED=$(((BASE_SPEED1 * (RUN_TIME1 + TIME_DIFF) + \
test_14() {
echo "#####"
echo "The OST-object referenced by the MDT-object should be there;"
- echo "otherwise, the LFSCK should re-create the missed OST-object."
+ echo "otherwise, the LFSCK should re-create the missing OST-object."
echo "#####"
check_mount_and_prep
test_18a() {
echo "#####"
echo "The target MDT-object is there, but related stripe information"
- echo "is lost or partly lost. The LFSCK should regenerate the missed"
+ echo "is lost or partly lost. The LFSCK should regenerate the missing"
echo "layout EA entries."
echo "#####"
check_mount_and_prep
$LFS mkdir -i 0 $DIR/$tdir/a1
- $LFS setstripe -c 1 -i 0 -s 1M $DIR/$tdir/a1
+ $LFS setstripe -c 1 -i 0 -S 1M $DIR/$tdir/a1
dd if=/dev/zero of=$DIR/$tdir/a1/f1 bs=1M count=2
local saved_size=$(ls -il $DIR/$tdir/a1/f1 | awk '{ print $6 }')
if [ $MDSCOUNT -ge 2 ]; then
$LFS mkdir -i 1 $DIR/$tdir/a2
- $LFS setstripe -c 2 -i 1 -s 1M $DIR/$tdir/a2
+ $LFS setstripe -c 2 -i 1 -S 1M $DIR/$tdir/a2
dd if=/dev/zero of=$DIR/$tdir/a2/f2 bs=1M count=2
$LFS path2fid $DIR/$tdir/a2/f2
$LFS getstripe $DIR/$tdir/a2/f2
check_mount_and_prep
$LFS mkdir -i 0 $DIR/$tdir/a1
- $LFS setstripe -c 1 -i 0 -s 1M $DIR/$tdir/a1
+ $LFS setstripe -c 1 -i 0 -S 1M $DIR/$tdir/a1
dd if=/dev/zero of=$DIR/$tdir/a1/f1 bs=1M count=2
local saved_size=$(ls -il $DIR/$tdir/a1/f1 | awk '{ print $6 }')
local fid1=$($LFS path2fid $DIR/$tdir/a1/f1)
if [ $MDSCOUNT -ge 2 ]; then
$LFS mkdir -i 1 $DIR/$tdir/a2
- $LFS setstripe -c 2 -i 1 -s 1M $DIR/$tdir/a2
+ $LFS setstripe -c 2 -i 1 -S 1M $DIR/$tdir/a2
dd if=/dev/zero of=$DIR/$tdir/a2/f2 bs=1M count=2
fid2=$($LFS path2fid $DIR/$tdir/a2/f2)
echo ${fid2}
check_mount_and_prep
$LFS mkdir -i 0 $DIR/$tdir/a1
- $LFS setstripe -c 1 -i 0 -s 1M $DIR/$tdir/a1
+ $LFS setstripe -c 1 -i 0 -S 1M $DIR/$tdir/a1
echo "Inject failure, to simulate the case of missing parent FID"
#define OBD_FAIL_LFSCK_NOPFID 0x1617
if [ $MDSCOUNT -ge 2 ]; then
$LFS mkdir -i 1 $DIR/$tdir/a2
- $LFS setstripe -c 1 -i 0 -s 1M $DIR/$tdir/a2
+ $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
check_mount_and_prep
mkdir $DIR/$tdir/a1
- $LFS setstripe -c 1 -i 0 -s 1M $DIR/$tdir/a1
+ $LFS setstripe -c 1 -i 0 -S 1M $DIR/$tdir/a1
echo "guard" > $DIR/$tdir/a1/f1
echo "foo" > $DIR/$tdir/a1/f2
local saved_size=$(ls -il $DIR/$tdir/a1/f2 | awk '{ print $6 }')
check_mount_and_prep
mkdir $DIR/$tdir/a1
- $LFS setstripe -c 1 -i 0 -s 1M $DIR/$tdir/a1
+ $LFS setstripe -c 1 -i 0 -S 1M $DIR/$tdir/a1
echo "guard" > $DIR/$tdir/a1/f1
echo "foo" > $DIR/$tdir/a1/f2
local saved_size=$(ls -il $DIR/$tdir/a1/f2 | awk '{ print $6 }')
check_mount_and_prep
$LFS mkdir -i 0 $DIR/$tdir/a1
- $LFS setstripe -c 1 -i 0 -s 1M $DIR/$tdir/a1
+ $LFS setstripe -c 1 -i 0 -S 1M $DIR/$tdir/a1
dd if=/dev/zero of=$DIR/$tdir/a1/guard bs=1M count=2
dd if=/dev/zero of=$DIR/$tdir/a1/f1 bs=1M count=2
$LFS mkdir -i 0 $DIR/$tdir/a2
- $LFS setstripe -c 2 -i 0 -s 1M $DIR/$tdir/a2
+ $LFS setstripe -c 2 -i 0 -S 1M $DIR/$tdir/a2
dd if=/dev/zero of=$DIR/$tdir/a2/f2 bs=1M count=2
$LFS getstripe $DIR/$tdir/a1/f1
$LFS getstripe $DIR/$tdir/a2/f2
if [ $MDSCOUNT -ge 2 ]; then
$LFS mkdir -i 1 $DIR/$tdir/a3
- $LFS setstripe -c 1 -i 0 -s 1M $DIR/$tdir/a3
+ $LFS setstripe -c 1 -i 0 -S 1M $DIR/$tdir/a3
dd if=/dev/zero of=$DIR/$tdir/a3/guard bs=1M count=2
dd if=/dev/zero of=$DIR/$tdir/a3/f3 bs=1M count=2
$LFS mkdir -i 1 $DIR/$tdir/a4
- $LFS setstripe -c 2 -i 0 -s 1M $DIR/$tdir/a4
+ $LFS setstripe -c 2 -i 0 -S 1M $DIR/$tdir/a4
dd if=/dev/zero of=$DIR/$tdir/a4/f4 bs=1M count=2
$LFS getstripe $DIR/$tdir/a3/f3
$LFS getstripe $DIR/$tdir/a4/f4
check_mount_and_prep
$LFS mkdir -i 0 $DIR/$tdir/a1
if [ $OSTCOUNT -gt 2 ]; then
- $LFS setstripe -c 3 -i 0 -s 1M $DIR/$tdir/a1
+ $LFS setstripe -c 3 -i 0 -S 1M $DIR/$tdir/a1
bcount=513
else
- $LFS setstripe -c 2 -i 0 -s 1M $DIR/$tdir/a1
+ $LFS setstripe -c 2 -i 0 -S 1M $DIR/$tdir/a1
bcount=257
fi
}
run_test 24 "LFSCK can repair multiple-referenced name entry"
+test_25() {
+ [ $(facet_fstype $SINGLEMDS) != ldiskfs ] &&
+ skip "Only support to inject failure on ldiskfs" && return
+
+ echo "#####"
+ echo "The file type in the name entry does not match the file type"
+ echo "claimed by the referenced object. Then the LFSCK will update"
+ echo "the file type in the name entry."
+ echo "#####"
+
+ check_mount_and_prep
+
+ $LFS mkdir -i 0 $DIR/$tdir/d0 || error "(1) Fail to mkdir d0"
+
+ echo "Inject failure stub on MDT0 to simulate the case that"
+ echo "the file type stored in the name entry is wrong."
+
+ #define OBD_FAIL_LFSCK_BAD_TYPE 0x1623
+ do_facet $SINGLEMDS $LCTL set_param fail_loc=0x1623
+ touch $DIR/$tdir/d0/foo || error "(2) Fail to touch $DIR/$tdir/d0/foo"
+ do_facet $SINGLEMDS $LCTL set_param fail_loc=0
+
+ echo "Trigger namespace LFSCK to repair bad file type in the name entry"
+ $START_NAMESPACE -r || error "(3) Fail to start LFSCK for namespace"
+
+ wait_update_facet $SINGLEMDS "$LCTL get_param -n \
+ mdd.${MDT_DEV}.lfsck_namespace |
+ awk '/^status/ { print \\\$2 }'" "completed" 32 || {
+ $SHOW_NAMESPACE
+ error "(4) unexpected status"
+ }
+
+ local repaired=$($SHOW_NAMESPACE |
+ awk '/^bad_file_type_repaired/ { print $2 }')
+ [ $repaired -eq 1 ] ||
+ error "(5) Fail to repair bad file type in name entry: $repaired"
+
+ ls -ail $DIR/$tdir/d0 || error "(6) Fail to 'ls' the $DIR/$tdir/d0"
+}
+run_test 25 "LFSCK can repair bad file type in the name entry"
+
+test_26a() {
+ echo "#####"
+ echo "The local name entry back referenced by the MDT-object is lost."
+ echo "The namespace LFSCK will add the missing local name entry back"
+ echo "to the normal namespace."
+ echo "#####"
+
+ check_mount_and_prep
+
+ $LFS mkdir -i 0 $DIR/$tdir/d0 || error "(1) Fail to mkdir d0"
+ touch $DIR/$tdir/d0/foo || error "(2) Fail to create foo"
+ local foofid=$($LFS path2fid $DIR/$tdir/d0/foo)
+
+ ln $DIR/$tdir/d0/foo $DIR/$tdir/d0/dummy ||
+ error "(3) Fail to hard link to $DIR/$tdir/d0/foo"
+
+ echo "Inject failure stub on MDT0 to simulate the case that"
+ echo "foo's name entry will be removed, but the foo's object"
+ echo "and its linkEA are kept in the system."
+
+ #define OBD_FAIL_LFSCK_NO_NAMEENTRY 0x1624
+ do_facet $SINGLEMDS $LCTL set_param fail_loc=0x1624
+ rm -f $DIR/$tdir/d0/foo || error "(4) Fail to unlink $DIR/$tdir/d0/foo"
+ do_facet $SINGLEMDS $LCTL set_param fail_loc=0
+
+ ls -ail $DIR/$tdir/d0/foo > /dev/null 2>&1 && "(5) 'ls' should fail"
+
+ echo "Trigger namespace LFSCK to repair the missing remote name entry"
+ $START_NAMESPACE -r -A ||
+ error "(6) Fail to start LFSCK for namespace"
+
+ wait_update_facet $SINGLEMDS "$LCTL get_param -n \
+ mdd.${MDT_DEV}.lfsck_namespace |
+ awk '/^status/ { print \\\$2 }'" "completed" 32 || {
+ $SHOW_NAMESPACE
+ error "(7) unexpected status"
+ }
+
+ local repaired=$($SHOW_NAMESPACE |
+ awk '/^lost_dirent_repaired/ { print $2 }')
+ [ $repaired -eq 1 ] ||
+ error "(8) Fail to repair lost dirent: $repaired"
+
+ ls -ail $DIR/$tdir/d0/foo ||
+ error "(9) Fail to 'ls' $DIR/$tdir/d0/foo"
+
+ local foofid2=$($LFS path2fid $DIR/$tdir/d0/foo)
+ [ "$foofid" == "$foofid2" ] ||
+ error "(10) foo's FID changed: $foofid, $foofid2"
+}
+run_test 26a "LFSCK can add the missing local name entry back to the namespace"
+
+test_26b() {
+ [ $MDSCOUNT -lt 2 ] &&
+ skip "We need at least 2 MDSes for this test" && return
+
+ echo "#####"
+ echo "The remote name entry back referenced by the MDT-object is lost."
+ echo "The namespace LFSCK will add the missing remote name entry back"
+ echo "to the normal namespace."
+ echo "#####"
+
+ check_mount_and_prep
+
+ $LFS mkdir -i 1 $DIR/$tdir/d0 || error "(1) Fail to mkdir d0"
+ $LFS mkdir -i 0 $DIR/$tdir/d0/foo || error "(2) Fail to mkdir foo"
+ local foofid=$($LFS path2fid $DIR/$tdir/d0/foo)
+
+ echo "Inject failure stub on MDT0 to simulate the case that"
+ echo "foo's name entry will be removed, but the foo's object"
+ echo "and its linkEA are kept in the system."
+
+ #define OBD_FAIL_LFSCK_NO_NAMEENTRY 0x1624
+ do_facet $SINGLEMDS $LCTL set_param fail_loc=0x1624
+ rmdir $DIR/$tdir/d0/foo || error "(3) Fail to rmdir $DIR/$tdir/d0/foo"
+ do_facet $SINGLEMDS $LCTL set_param fail_loc=0
+
+ ls -ail $DIR/$tdir/d0/foo > /dev/null 2>&1 && "(4) 'ls' should fail"
+
+ echo "Trigger namespace LFSCK to repair the missing remote name entry"
+ $START_NAMESPACE -r -A ||
+ error "(5) Fail to start LFSCK for namespace"
+
+ wait_update_facet $SINGLEMDS "$LCTL get_param -n \
+ mdd.${MDT_DEV}.lfsck_namespace |
+ awk '/^status/ { print \\\$2 }'" "completed" 32 || {
+ $SHOW_NAMESPACE
+ error "(6) unexpected status"
+ }
+
+ local repaired=$($SHOW_NAMESPACE |
+ awk '/^lost_dirent_repaired/ { print $2 }')
+ [ $repaired -eq 1 ] ||
+ error "(7) Fail to repair lost dirent: $repaired"
+
+ ls -ail $DIR/$tdir/d0/foo ||
+ error "(8) Fail to 'ls' $DIR/$tdir/d0/foo"
+
+ local foofid2=$($LFS path2fid $DIR/$tdir/d0/foo)
+ [ "$foofid" == "$foofid2" ] ||
+ error "(9) foo's FID changed: $foofid, $foofid2"
+}
+run_test 26b "LFSCK can add the missing remote name entry back to the namespace"
+
+test_27a() {
+ echo "#####"
+ echo "The local parent referenced by the MDT-object linkEA is lost."
+ echo "The namespace LFSCK will re-create the lost parent as orphan."
+ echo "#####"
+
+ check_mount_and_prep
+
+ $LFS mkdir -i 0 $DIR/$tdir/d0 || error "(1) Fail to mkdir d0"
+ touch $DIR/$tdir/d0/foo || error "(2) Fail to create foo"
+ ln $DIR/$tdir/d0/foo $DIR/$tdir/d0/dummy ||
+ error "(3) Fail to hard link to $DIR/$tdir/d0/foo"
+
+ echo "Inject failure stub on MDT0 to simulate the case that"
+ echo "foo's name entry will be removed, but the foo's object"
+ echo "and its linkEA are kept in the system. And then remove"
+ echo "another hard link and the parent directory."
+
+ #define OBD_FAIL_LFSCK_NO_NAMEENTRY 0x1624
+ do_facet $SINGLEMDS $LCTL set_param fail_loc=0x1624
+ rm -f $DIR/$tdir/d0/foo ||
+ error "(4) Fail to unlink $DIR/$tdir/d0/foo"
+ rm -f $DIR/$tdir/d0/dummy ||
+ error "(5) Fail to unlink $DIR/$tdir/d0/dummy"
+ do_facet $SINGLEMDS $LCTL set_param fail_loc=0
+
+ rm -rf $DIR/$tdir/d0 || error "(5) Fail to unlink the dir d0"
+ ls -ail $DIR/$tdir/d0 > /dev/null 2>&1 && "(6) 'ls' should fail"
+
+ echo "Trigger namespace LFSCK to repair the lost parent"
+ $START_NAMESPACE -r -A ||
+ error "(6) Fail to start LFSCK for namespace"
+
+ wait_update_facet $SINGLEMDS "$LCTL get_param -n \
+ mdd.${MDT_DEV}.lfsck_namespace |
+ awk '/^status/ { print \\\$2 }'" "completed" 32 || {
+ $SHOW_NAMESPACE
+ error "(7) unexpected status"
+ }
+
+ local repaired=$($SHOW_NAMESPACE |
+ awk '/^lost_dirent_repaired/ { print $2 }')
+ [ $repaired -eq 1 ] ||
+ error "(8) Fail to repair lost dirent: $repaired"
+
+ echo "There should be an orphan under .lustre/lost+found/MDT0000/"
+ [ -d $MOUNT/.lustre/lost+found/MDT0000 ] ||
+ error "(9) $MOUNT/.lustre/lost+found/MDT0000/ should be there"
+
+ ls -ail $MOUNT/.lustre/lost+found/MDT0000/
+
+ cname=$(find $MOUNT/.lustre/lost+found/MDT0000/ -name *-P-*)
+ [ ! -z "$cname" ] ||
+ error "(10) .lustre/lost+found/MDT0000/ should not be empty"
+}
+run_test 27a "LFSCK can recreate the lost local parent directory as orphan"
+
+test_27b() {
+ [ $MDSCOUNT -lt 2 ] &&
+ skip "We need at least 2 MDSes for this test" && return
+
+ echo "#####"
+ echo "The remote parent referenced by the MDT-object linkEA is lost."
+ echo "The namespace LFSCK will re-create the lost parent as orphan."
+ echo "#####"
+
+ check_mount_and_prep
+
+ $LFS mkdir -i 1 $DIR/$tdir/d0 || error "(1) Fail to mkdir d0"
+ $LFS mkdir -i 0 $DIR/$tdir/d0/foo || error "(2) Fail to mkdir foo"
+
+ $LFS path2fid $DIR/$tdir/d0
+
+ echo "Inject failure stub on MDT0 to simulate the case that"
+ echo "foo's name entry will be removed, but the foo's object"
+ echo "and its linkEA are kept in the system. And then remove"
+ echo "the parent directory."
+
+ #define OBD_FAIL_LFSCK_NO_NAMEENTRY 0x1624
+ do_facet $SINGLEMDS $LCTL set_param fail_loc=0x1624
+ rmdir $DIR/$tdir/d0/foo || error "(3) Fail to rmdir $DIR/$tdir/d0/foo"
+ do_facet $SINGLEMDS $LCTL set_param fail_loc=0
+
+ rmdir $DIR/$tdir/d0 || error "(4) Fail to unlink the dir d0"
+ ls -ail $DIR/$tdir/d0 > /dev/null 2>&1 && "(5) 'ls' should fail"
+
+ echo "Trigger namespace LFSCK to repair the missing remote name entry"
+ $START_NAMESPACE -r -A ||
+ error "(6) Fail to start LFSCK for namespace"
+
+ wait_update_facet $SINGLEMDS "$LCTL get_param -n \
+ mdd.${MDT_DEV}.lfsck_namespace |
+ awk '/^status/ { print \\\$2 }'" "completed" 32 || {
+ $SHOW_NAMESPACE
+ error "(7) unexpected status"
+ }
+
+ local repaired=$($SHOW_NAMESPACE |
+ awk '/^lost_dirent_repaired/ { print $2 }')
+ [ $repaired -eq 1 ] ||
+ error "(8) Fail to repair lost dirent: $repaired"
+
+ ls -ail $MOUNT/.lustre/lost+found/
+
+ echo "There should be an orphan under .lustre/lost+found/MDT0001/"
+ [ -d $MOUNT/.lustre/lost+found/MDT0001 ] ||
+ error "(9) $MOUNT/.lustre/lost+found/MDT0001/ should be there"
+
+ ls -ail $MOUNT/.lustre/lost+found/MDT0001/
+
+ cname=$(find $MOUNT/.lustre/lost+found/MDT0001/ -name *-P-*)
+ [ ! -z "$cname" ] ||
+ error "(10) .lustre/lost+found/MDT0001/ should not be empty"
+}
+run_test 27b "LFSCK can recreate the lost remote parent directory as orphan"
+
+test_28() {
+ [ $MDSCOUNT -lt 2 ] &&
+ skip "The test needs at least 2 MDTs" && return
+
+ echo "#####"
+ echo "The target name entry is lost. The LFSCK should insert the"
+ echo "orphan MDT-object under .lustre/lost+found/MDTxxxx. But if"
+ echo "the MDT (on which the orphan MDT-object resides) has ever"
+ echo "failed to respond some name entry verification during the"
+ echo "first stage-scanning, then the LFSCK should skip to handle"
+ echo "orphan MDT-object on this MDT. But other MDTs should not"
+ echo "be affected."
+ echo "#####"
+
+ check_mount_and_prep
+ $LFS mkdir -i 0 $DIR/$tdir/d1
+ $LFS mkdir -i 1 $DIR/$tdir/d1/a1
+ $LFS mkdir -i 1 $DIR/$tdir/d1/a2
+
+ $LFS mkdir -i 1 $DIR/$tdir/d2
+ $LFS mkdir -i 0 $DIR/$tdir/d2/a1
+ $LFS mkdir -i 0 $DIR/$tdir/d2/a2
+
+ echo "Inject failure stub on MDT0 to simulate the case that"
+ echo "d1/a1's name entry will be removed, but the d1/a1's object"
+ echo "and its linkEA are kept in the system. And the case that"
+ echo "d2/a2's name entry will be removed, but the d2/a2's object"
+ echo "and its linkEA are kept in the system."
+
+ #define OBD_FAIL_LFSCK_NO_NAMEENTRY 0x1624
+ do_facet mds1 $LCTL set_param fail_loc=0x1624
+ do_facet mds2 $LCTL set_param fail_loc=0x1624
+ rmdir $DIR/$tdir/d1/a1 || error "(1) Fail to rmdir $DIR/$tdir/d1/a1"
+ rmdir $DIR/$tdir/d2/a2 || error "(2) Fail to rmdir $DIR/$tdir/d2/a2"
+ do_facet mds1 $LCTL set_param fail_loc=0
+ do_facet mds2 $LCTL set_param fail_loc=0
+
+ cancel_lru_locks mdc
+ cancel_lru_locks osc
+
+ echo "Inject failure, to simulate the MDT0 fail to handle"
+ echo "MDT1 LFSCK request during the first-stage scanning."
+ #define OBD_FAIL_LFSCK_BAD_NETWORK 0x161c
+ do_facet mds2 $LCTL set_param fail_loc=0x161c fail_val=0
+
+ echo "Trigger namespace LFSCK on all devices to find out orphan object"
+ $START_NAMESPACE -r -A ||
+ error "(3) Fail to start LFSCK for namespace"
+
+ wait_update_facet mds1 "$LCTL get_param -n \
+ mdd.$(facet_svc mds1).lfsck_namespace |
+ awk '/^status/ { print \\\$2 }'" "partial" 32 || {
+ error "(4) mds1 is not the expected 'partial'"
+ }
+
+ wait_update_facet mds2 "$LCTL get_param -n \
+ mdd.$(facet_svc mds2).lfsck_namespace |
+ awk '/^status/ { print \\\$2 }'" "completed" 32 || {
+ error "(5) mds2 is not the expected 'completed'"
+ }
+
+ do_facet mds2 $LCTL set_param fail_loc=0 fail_val=0
+
+ local repaired=$(do_facet mds1 $LCTL get_param -n \
+ mdd.$(facet_svc mds1).lfsck_namespace |
+ awk '/^lost_dirent_repaired/ { print $2 }')
+ [ $repaired -eq 0 ] ||
+ error "(6) Expect 0 fixed on mds1, but got: $repaired"
+
+ repaired=$(do_facet mds2 $LCTL get_param -n \
+ mdd.$(facet_svc mds2).lfsck_namespace |
+ awk '/^lost_dirent_repaired/ { print $2 }')
+ [ $repaired -eq 1 ] ||
+ error "(7) Expect 1 fixed on mds2, but got: $repaired"
+
+ echo "Trigger namespace LFSCK on all devices again to cleanup"
+ $START_NAMESPACE -r -A ||
+ error "(8) Fail to start LFSCK for namespace"
+
+ for k in $(seq $MDSCOUNT); do
+ # The LFSCK status query internal is 30 seconds. For the case
+ # of some LFSCK_NOTIFY RPCs failure/lost, we will wait enough
+ # time to guarantee the status sync up.
+ wait_update_facet mds${k} "$LCTL get_param -n \
+ mdd.$(facet_svc mds${k}).lfsck_namespace |
+ awk '/^status/ { print \\\$2 }'" "completed" 32 ||
+ error "(9) MDS${k} is not the expected 'completed'"
+ done
+
+ local repaired=$(do_facet mds1 $LCTL get_param -n \
+ mdd.$(facet_svc mds1).lfsck_namespace |
+ awk '/^lost_dirent_repaired/ { print $2 }')
+ [ $repaired -eq 1 ] ||
+ error "(10) Expect 1 fixed on mds1, but got: $repaired"
+
+ repaired=$(do_facet mds2 $LCTL get_param -n \
+ mdd.$(facet_svc mds2).lfsck_namespace |
+ awk '/^lost_dirent_repaired/ { print $2 }')
+ [ $repaired -eq 0 ] ||
+ error "(11) Expect 0 fixed on mds2, but got: $repaired"
+}
+run_test 28 "Skip the failed MDT(s) when handle orphan MDT-objects"
+
+test_29a() {
+ echo "#####"
+ echo "The object's nlink attribute is larger than the object's known"
+ echo "name entries count. The LFSCK will repair the object's nlink"
+ echo "attribute to match the known name entries count"
+ echo "#####"
+
+ check_mount_and_prep
+
+ $LFS mkdir -i 0 $DIR/$tdir/d0 || error "(1) Fail to mkdir d0"
+ touch $DIR/$tdir/d0/foo || error "(2) Fail to create foo"
+
+ echo "Inject failure stub on MDT0 to simulate the case that foo's"
+ echo "nlink attribute is larger than its name entries count."
+
+ #define OBD_FAIL_LFSCK_MORE_NLINK 0x1625
+ do_facet $SINGLEMDS $LCTL set_param fail_loc=0x1625
+ ln $DIR/$tdir/d0/foo $DIR/$tdir/d0/h1 ||
+ error "(3) Fail to hard link to $DIR/$tdir/d0/foo"
+ do_facet $SINGLEMDS $LCTL set_param fail_loc=0
+
+ cancel_lru_locks mdc
+ local count=$(stat --format=%h $DIR/$tdir/d0/foo)
+ [ $count -eq 3 ] || error "(4) Cannot inject error: $count"
+
+ echo "Trigger namespace LFSCK to repair the nlink count"
+ $START_NAMESPACE -r -A ||
+ error "(5) Fail to start LFSCK for namespace"
+
+ wait_update_facet $SINGLEMDS "$LCTL get_param -n \
+ mdd.${MDT_DEV}.lfsck_namespace |
+ awk '/^status/ { print \\\$2 }'" "completed" 32 || {
+ $SHOW_NAMESPACE
+ error "(6) unexpected status"
+ }
+
+ local repaired=$($SHOW_NAMESPACE |
+ awk '/^nlinks_repaired/ { print $2 }')
+ [ $repaired -eq 1 ] ||
+ error "(7) Fail to repair nlink count: $repaired"
+
+ cancel_lru_locks mdc
+ count=$(stat --format=%h $DIR/$tdir/d0/foo)
+ [ $count -eq 2 ] || error "(8) Fail to repair nlink count: $count"
+}
+run_test 29a "LFSCK can repair bad nlink count (1)"
+
+test_29b() {
+ echo "#####"
+ echo "The object's nlink attribute is smaller than the object's known"
+ echo "name entries count. The LFSCK will repair the object's nlink"
+ echo "attribute to match the known name entries count"
+ echo "#####"
+
+ check_mount_and_prep
+
+ $LFS mkdir -i 0 $DIR/$tdir/d0 || error "(1) Fail to mkdir d0"
+ touch $DIR/$tdir/d0/foo || error "(2) Fail to create foo"
+
+ echo "Inject failure stub on MDT0 to simulate the case that foo's"
+ echo "nlink attribute is smaller than its name entries count."
+
+ #define OBD_FAIL_LFSCK_LESS_NLINK 0x1626
+ do_facet $SINGLEMDS $LCTL set_param fail_loc=0x1626
+ ln $DIR/$tdir/d0/foo $DIR/$tdir/d0/h1 ||
+ error "(3) Fail to hard link to $DIR/$tdir/d0/foo"
+ do_facet $SINGLEMDS $LCTL set_param fail_loc=0
+
+ cancel_lru_locks mdc
+ local count=$(stat --format=%h $DIR/$tdir/d0/foo)
+ [ $count -eq 1 ] || error "(4) Cannot inject error: $count"
+
+ echo "Trigger namespace LFSCK to repair the nlink count"
+ $START_NAMESPACE -r -A ||
+ error "(5) Fail to start LFSCK for namespace"
+
+ wait_update_facet $SINGLEMDS "$LCTL get_param -n \
+ mdd.${MDT_DEV}.lfsck_namespace |
+ awk '/^status/ { print \\\$2 }'" "completed" 32 || {
+ $SHOW_NAMESPACE
+ error "(6) unexpected status"
+ }
+
+ local repaired=$($SHOW_NAMESPACE |
+ awk '/^nlinks_repaired/ { print $2 }')
+ [ $repaired -eq 1 ] ||
+ error "(7) Fail to repair nlink count: $repaired"
+
+ cancel_lru_locks mdc
+ count=$(stat --format=%h $DIR/$tdir/d0/foo)
+ [ $count -eq 2 ] || error "(8) Fail to repair nlink count: $count"
+}
+run_test 29b "LFSCK can repair bad nlink count (2)"
+
+test_29c() {
+ echo "#####"
+ echo "There are too much hard links to the object, and exceeds the
+ echo object's linkEA limitation, as to NOT all the known name entries"
+ echo "will be recorded in the linkEA. Under such case, LFSCK should"
+ echo "skip the nlink verification for this object."
+ echo "#####"
+
+ check_mount_and_prep
+
+ $LFS mkdir -i 0 $DIR/$tdir/d0 || error "(1) Fail to mkdir d0"
+ touch $DIR/$tdir/d0/foo || error "(2) Fail to create foo"
+ ln $DIR/$tdir/d0/foo $DIR/$tdir/d0/h1 ||
+ error "(3) Fail to hard link to $DIR/$tdir/d0/foo"
+
+ echo "Inject failure stub on MDT0 to simulate the case that"
+ echo "foo's hard links exceed the object's linkEA limitation."
+
+ #define OBD_FAIL_LFSCK_LINKEA_OVERFLOW 0x1627
+ do_facet $SINGLEMDS $LCTL set_param fail_loc=0x1627
+ ln $DIR/$tdir/d0/foo $DIR/$tdir/d0/h2 ||
+ error "(4) Fail to hard link to $DIR/$tdir/d0/foo"
+
+ cancel_lru_locks mdc
+
+ local count1=$(stat --format=%h $DIR/$tdir/d0/foo)
+ [ $count1 -eq 3 ] || error "(5) Stat failure: $count1"
+
+ local foofid=$($LFS path2fid $DIR/$tdir/d0/foo)
+ $LFS fid2path $DIR $foofid
+ local count2=$($LFS fid2path $DIR $foofid | wc -l)
+ [ $count2 -eq 2 ] || "(6) Fail to inject error: $count2"
+
+ echo "Trigger namespace LFSCK to repair the nlink count"
+ $START_NAMESPACE -r -A ||
+ error "(7) Fail to start LFSCK for namespace"
+
+ wait_update_facet $SINGLEMDS "$LCTL get_param -n \
+ mdd.${MDT_DEV}.lfsck_namespace |
+ awk '/^status/ { print \\\$2 }'" "completed" 32 || {
+ $SHOW_NAMESPACE
+ error "(8) unexpected status"
+ }
+
+ do_facet $SINGLEMDS $LCTL set_param fail_loc=0
+ local repaired=$($SHOW_NAMESPACE |
+ awk '/^nlinks_repaired/ { print $2 }')
+ [ $repaired -eq 0 ] ||
+ error "(9) Repair nlink count unexpcetedly: $repaired"
+
+ cancel_lru_locks mdc
+
+ count1=$(stat --format=%h $DIR/$tdir/d0/foo)
+ [ $count1 -eq 3 ] || error "(10) Stat failure: $count1"
+
+ count2=$($LFS fid2path $DIR $foofid | wc -l)
+ [ $count2 -eq 2 ] ||
+ error "(11) Repaired something unexpectedly: $count2"
+}
+run_test 29c "Not verify nlink attr if hark links exceed linkEA limitation"
+
+test_30() {
+ [ $(facet_fstype $SINGLEMDS) != ldiskfs ] &&
+ skip "Only support backend /lost+found for ldiskfs" && return
+
+ echo "#####"
+ echo "The namespace LFSCK will move the orphans from backend"
+ echo "/lost+found directory to normal client visible namespace"
+ echo "or to global visible ./lustre/lost+found/MDTxxxx/ directory"
+ echo "#####"
+
+ check_mount_and_prep
+
+ $LFS mkdir -i 0 $DIR/$tdir/foo || error "(1) Fail to mkdir foo"
+ touch $DIR/$tdir/foo/f0 || error "(2) Fail to touch f1"
+
+ echo "Inject failure stub on MDT0 to simulate the case that"
+ echo "directory d0 has no linkEA entry, then the LFSCK will"
+ echo "move it into .lustre/lost+found/MDTxxxx/ later."
+
+ #define OBD_FAIL_LFSCK_NO_LINKEA 0x161d
+ do_facet $SINGLEMDS $LCTL set_param fail_loc=0x161d
+ mkdir $DIR/$tdir/foo/d0 || error "(3) Fail to mkdir d0"
+ do_facet $SINGLEMDS $LCTL set_param fail_loc=0
+
+ touch $DIR/$tdir/foo/d0/f1 || error "(4) Fail to touch f1"
+ mkdir $DIR/$tdir/foo/d0/d1 || error "(5) Fail to mkdir d1"
+
+ echo "Inject failure stub on MDT0 to simulate the case that the"
+ echo "object's name entry will be removed, but not destroy the"
+ echo "object. Then backend e2fsck will handle it as orphan and"
+ echo "add them into the backend /lost+found directory."
+
+ #define OBD_FAIL_LFSCK_NO_NAMEENTRY 0x1624
+ do_facet $SINGLEMDS $LCTL set_param fail_loc=0x1624
+ rmdir $DIR/$tdir/foo/d0/d1 || error "(6) Fail to rmdir d1"
+ rm -f $DIR/$tdir/foo/d0/f1 || error "(7) Fail to unlink f1"
+ rmdir $DIR/$tdir/foo/d0 || error "(8) Fail to rmdir d0"
+ rm -f $DIR/$tdir/foo/f0 || error "(9) Fail to unlink f0"
+ do_facet $SINGLEMDS $LCTL set_param fail_loc=0
+
+ umount_client $MOUNT || error "(10) Fail to stop client!"
+
+ stop $SINGLEMDS || error "(11) Fail to stop MDT0"
+
+ echo "run e2fsck"
+ run_e2fsck $(facet_host $SINGLEMDS) $MDT_DEVNAME "-y" ||
+ error "(12) Fail to run e2fsck"
+
+ start $SINGLEMDS $MDT_DEVNAME $MOUNT_OPTS_NOSCRUB > /dev/null ||
+ error "(13) Fail to start MDT0"
+
+ echo "Trigger namespace LFSCK to recover backend orphans"
+ $START_NAMESPACE -r -A ||
+ error "(14) Fail to start LFSCK for namespace"
+
+ wait_update_facet $SINGLEMDS "$LCTL get_param -n \
+ mdd.${MDT_DEV}.lfsck_namespace |
+ awk '/^status/ { print \\\$2 }'" "completed" 32 || {
+ $SHOW_NAMESPACE
+ error "(15) unexpected status"
+ }
+
+ local repaired=$($SHOW_NAMESPACE |
+ awk '/^local_lost_found_moved/ { print $2 }')
+ [ $repaired -ge 4 ] ||
+ error "(16) Fail to recover backend orphans: $repaired"
+
+ mount_client $MOUNT || error "(17) Fail to start client!"
+
+ stat $DIR/$tdir/foo/f0 || "(18) f0 is not recovered"
+
+ ls -ail $MOUNT/.lustre/lost+found/
+
+ echo "d0 should become orphan under .lustre/lost+found/MDT0000/"
+ [ -d $MOUNT/.lustre/lost+found/MDT0000 ] ||
+ error "(19) $MOUNT/.lustre/lost+found/MDT0000/ should be there"
+
+ ls -ail $MOUNT/.lustre/lost+found/MDT0000/
+
+ cname=$(find $MOUNT/.lustre/lost+found/MDT0000/ -name *-*-D-*)
+ [ ! -z "$cname" ] || error "(20) d0 is not recovered"
+
+ stat ${cname}/d1 || error "(21) d0 is not recovered"
+ stat ${cname}/f1 || error "(22) f1 is not recovered"
+}
+run_test 30 "LFSCK can recover the orphans from backend /lost+found"
+
$LCTL set_param debug=-lfsck > /dev/null || true
# restore MDS/OST size
git://git.whamcloud.com / fs / lustre-release.git / blobdiff