if [[ $(uname -m) = aarch64 ]]; then
# bug number: LU-11596
ALWAYS_EXCEPT+=" $GRANT_CHECK_LIST"
- # bug number: LU-11671 LU-11667
- ALWAYS_EXCEPT+=" 45 317"
+ # bug number: LU-11671
+ ALWAYS_EXCEPT+=" 45"
# bug number: LU-14067 LU-14067
ALWAYS_EXCEPT+=" 400a 400b"
fi
ALWAYS_EXCEPT+=" 411"
fi
-# 5 12 8 12 (min)"
-[ "$SLOW" = "no" ] && EXCEPT_SLOW="27m 60i 64b 68 71 115 135 136 300o"
+# 5 12 8 12 15 (min)"
+[ "$SLOW" = "no" ] && EXCEPT_SLOW="27m 60i 64b 68 71 115 135 136 230d 300o"
if [ "$mds1_FSTYPE" = "zfs" ]; then
# bug number for skipped test:
wait_update_facet $SINGLEMDS "$get_prealloc" "" $delay
}
-__exhaust_precreations() {
- local OSTIDX=$1
- local FAILLOC=$2
- local FAILIDX=${3:-$OSTIDX}
- local ofacet=ost$((OSTIDX + 1))
-
- mkdir_on_mdt0 $DIR/$tdir
- local mdtidx=$($LFS getstripe -m $DIR/$tdir)
- local mfacet=mds$((mdtidx + 1))
- echo OSTIDX=$OSTIDX MDTIDX=$mdtidx
-
- local OST=$(ostname_from_index $OSTIDX)
-
- # on the mdt's osc
- local mdtosc_proc1=$(get_mdtosc_proc_path $mfacet $OST)
- local last_id=$(do_facet $mfacet lctl get_param -n \
- osp.$mdtosc_proc1.prealloc_last_id)
- local next_id=$(do_facet $mfacet lctl get_param -n \
- osp.$mdtosc_proc1.prealloc_next_id)
-
- local mdtosc_proc2=$(get_mdtosc_proc_path $mfacet)
- do_facet $mfacet lctl get_param osp.$mdtosc_proc2.prealloc*
-
- test_mkdir -p $DIR/$tdir/${OST}
- $LFS setstripe -i $OSTIDX -c 1 $DIR/$tdir/${OST}
-#define OBD_FAIL_OST_ENOSPC 0x215
- do_facet $ofacet lctl set_param fail_val=$FAILIDX fail_loc=0x215
- echo "Creating to objid $last_id on ost $OST..."
- createmany -o $DIR/$tdir/${OST}/f $next_id $((last_id - next_id + 2))
- do_facet $mfacet lctl get_param osp.$mdtosc_proc2.prealloc*
- do_facet $ofacet lctl set_param fail_loc=$FAILLOC
-}
-
-exhaust_precreations() {
- __exhaust_precreations $1 $2 $3
- sleep_maxage
-}
-
-exhaust_all_precreations() {
- local i
- for (( i=0; i < OSTCOUNT; i++ )) ; do
- __exhaust_precreations $i $1 -1
- done
- sleep_maxage
-}
-
test_27n() {
[[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs"
[ $PARALLEL == "yes" ] && skip "skip parallel run"
test_mkdir $DIR/$tdir
# Set default striping on directory
- $LFS setstripe -C 4 $DIR/$tdir
+ local setcount=4
+ local stripe_opt
+
+ # if we run against a 2.12 server which lacks overstring support
+ # then the connect_flag will not report overstriping, even if client
+ # is 2.14+
+ if [[ $($LCTL get_param mdc.*.connect_flags) =~ overstriping ]]; then
+ stripe_opt="-C $setcount"
+ elif (( $OSTCOUNT >= $setcount )); then
+ stripe_opt="-c $setcount"
+ else
+ skip "server does not support overstriping"
+ fi
+ $LFS setstripe $stripe_opt $DIR/$tdir
echo 1 > $DIR/$tdir/${tfile}.1
local count=$($LFS getstripe -c $DIR/$tdir/${tfile}.1)
- local setcount=4
[ $count -eq $setcount ] ||
error "(1) stripe count $count, should be $setcount"
# Clean up DOM layout
$LFS setstripe -d $DIR/$tdir
+ save_layout_restore_at_exit $MOUNT
# Now test that append striping works when layout is from root
$LFS setstripe -c 2 $MOUNT
# Make a special directory for this
mkdir $DIR/${tdir}/${tdir}.2
- stack_trap "$LFS setstripe -d $MOUNT" EXIT
# Verify for normal file
setcount=2
}
run_test 27Q "llapi_file_get_stripe() works on symlinks"
+test_27R() {
+ (( $MDS1_VERSION >= $(version_code 2.14.55) )) ||
+ skip "need MDS 2.14.55 or later"
+ (( $OSTCOUNT >= 2 )) || skip_env "needs at least 2 OSTs"
+
+ local testdir="$DIR/$tdir"
+ test_mkdir -p $testdir
+ stack_trap "rm -rf $testdir"
+ $LFS setstripe -c -1 $testdir || error "setstripe failed"
+
+ local f1="$testdir/f1"
+ touch $f1 || error "failed to touch $f1"
+ local count=$($LFS getstripe -c $f1)
+ (( $count == $OSTCOUNT )) || error "wrong stripe count"
+
+ do_facet $SINGLEMDS $LCTL set_param lod.*.max_stripecount=-1
+ (( $? == 34 )) || error "setting max_stripecount to -1 should fail and return ERANGE"
+
+ local maxcount=$(($OSTCOUNT - 1))
+ local mdts=$(comma_list $(mdts_nodes))
+ do_nodes $mdts $LCTL set_param lod.*.max_stripecount=$maxcount
+ stack_trap "do_nodes $mdts $LCTL set_param lod.*.max_stripecount=0"
+
+ local f2="$testdir/f2"
+ touch $f2 || error "failed to touch $f2"
+ local count=$($LFS getstripe -c $f2)
+ (( $count == $maxcount )) || error "wrong stripe count"
+}
+run_test 27R "test max_stripecount limitation when stripe count is set to -1"
+
# createtest also checks that device nodes are created and
# then visible correctly (#2091)
test_28() { # bug 2091
test_29() {
[ $PARALLEL == "yes" ] && skip "skip parallel run"
+ [ $MDS1_VERSION -ge $(version_code 2.14.51) ] && {
+ disable_opencache
+ stack_trap "restore_opencache"
+ }
+
sync; sleep 1; sync # flush out any dirty pages from previous tests
cancel_lru_locks
test_mkdir $DIR/d29
}
run_test 31q "create striped directory on specific MDTs"
+#LU-14949
+test_31r() {
+ touch $DIR/$tfile.target
+ touch $DIR/$tfile.source
+
+ #OBD_FAIL_LLITE_OPEN_DELAY 0x1419
+ $LCTL set_param fail_loc=0x1419 fail_val=3
+ cat $DIR/$tfile.target &
+ CATPID=$!
+
+ # Guarantee open is waiting before we get here
+ sleep 1
+ mv $DIR/$tfile.source $DIR/$tfile.target
+
+ wait $CATPID
+ RC=$?
+ if [[ $RC -ne 0 ]]; then
+ error "open with cat failed, rc=$RC"
+ fi
+}
+run_test 31r "open-rename(replace) race"
+
cleanup_test32_mount() {
local rc=0
trap 0
test_51d() {
[ $PARALLEL == "yes" ] && skip "skip parallel run"
[[ $OSTCOUNT -lt 3 ]] && skip_env "needs >= 3 OSTs"
+ local qos_old
test_mkdir $DIR/$tdir
+ $LFS setstripe -c $OSTCOUNT $DIR/$tdir
+
+ qos_old=$(do_facet mds1 \
+ "$LCTL get_param -n lod.$FSNAME-*.qos_threshold_rr" | head -n 1)
+ do_nodes $(comma_list $(mdts_nodes)) \
+ "$LCTL set_param lod.$FSNAME-*.qos_threshold_rr=100"
+ stack_trap "do_nodes $(comma_list $(mdts_nodes)) \
+ '$LCTL set_param lod.$FSNAME-*.qos_threshold_rr=${qos_old%%%}'"
+
createmany -o $DIR/$tdir/t- 1000
$LFS getstripe $DIR/$tdir > $TMP/$tfile
- for N in $(seq 0 $((OSTCOUNT - 1))); do
- OBJS[$N]=$(awk -vobjs=0 '($1 == '$N') { objs += 1 } \
- END { printf("%0.0f", objs) }' $TMP/$tfile)
- OBJS0[$N]=$(grep -A 1 idx $TMP/$tfile | awk -vobjs=0 \
- '($1 == '$N') { objs += 1 } \
- END { printf("%0.0f", objs) }')
- log "OST$N has ${OBJS[$N]} objects, ${OBJS0[$N]} are index 0"
+ for ((n = 0; n < $OSTCOUNT; n++)); do
+ objs[$n]=$(awk -vobjs=0 '($1 == '$n') { objs += 1 } \
+ END { printf("%0.0f", objs) }' $TMP/$tfile)
+ objs0[$n]=$(grep -A 1 idx $TMP/$tfile | awk -vobjs=0 \
+ '($1 == '$n') { objs += 1 } \
+ END { printf("%0.0f", objs) }')
+ log "OST$n has ${objs[$n]} objects, ${objs0[$n]} are index 0"
done
unlinkmany $DIR/$tdir/t- 1000
- NLAST=0
- for N in $(seq 1 $((OSTCOUNT - 1))); do
- [[ ${OBJS[$N]} -lt $((${OBJS[$NLAST]} - 20)) ]] &&
- error "OST $N has less objects vs OST $NLAST" \
- " (${OBJS[$N]} < ${OBJS[$NLAST]}"
- [[ ${OBJS[$N]} -gt $((${OBJS[$NLAST]} + 20)) ]] &&
- error "OST $N has less objects vs OST $NLAST" \
- " (${OBJS[$N]} < ${OBJS[$NLAST]}"
-
- [[ ${OBJS0[$N]} -lt $((${OBJS0[$NLAST]} - 20)) ]] &&
- error "OST $N has less #0 objects vs OST $NLAST" \
- " (${OBJS0[$N]} < ${OBJS0[$NLAST]}"
- [[ ${OBJS0[$N]} -gt $((${OBJS0[$NLAST]} + 20)) ]] &&
- error "OST $N has less #0 objects vs OST $NLAST" \
- " (${OBJS0[$N]} < ${OBJS0[$NLAST]}"
- NLAST=$N
+ nlast=0
+ for ((n = 0; n < $OSTCOUNT; n++)); do
+ (( ${objs[$n]} > ${objs[$nlast]} * 4 / 5 )) ||
+ { $LFS df && $LFS df -i &&
+ error "OST $n has fewer objects vs. OST $nlast" \
+ " (${objs[$n]} < ${objs[$nlast]}"; }
+ (( ${objs[$n]} < ${objs[$nlast]} * 5 / 4 )) ||
+ { $LFS df && $LFS df -i &&
+ error "OST $n has fewer objects vs. OST $nlast" \
+ " (${objs[$n]} < ${objs[$nlast]}"; }
+
+ (( ${objs0[$n]} > ${objs0[$nlast]} * 4 / 5 )) ||
+ { $LFS df && $LFS df -i &&
+ error "OST $n has fewer #0 objects vs. OST $nlast" \
+ " (${objs0[$n]} < ${objs0[$nlast]}"; }
+ (( ${objs0[$n]} < ${objs0[$nlast]} * 5 / 4 )) ||
+ { $LFS df && $LFS df -i &&
+ error "OST $n has fewer #0 objects vs. OST $nlast" \
+ " (${objs0[$n]} < ${objs0[$nlast]}"; }
+ nlast=$n
done
- rm -f $TMP/$tfile
}
run_test 51d "check object distribution"
}
run_test 56d "'lfs df -v' prints only configured devices"
+test_56e() {
+ err_enoent=2 # No such file or directory
+ err_eopnotsupp=95 # Operation not supported
+
+ enoent_mnt=/pmt1 # Invalid dentry. Path not present
+ notsup_mnt=/tmp # Valid dentry, but Not a lustreFS
+
+ # Check for handling of path not exists
+ output=$($LFS df $enoent_mnt 2>&1)
+ ret=$?
+
+ fs=$(echo $output | awk -F: '{print $2}' | awk '{print $3}' | tr -d \')
+ [[ $fs = $enoent_mnt && $ret -eq $err_enoent ]] ||
+ error "expect failure $err_enoent, not $ret"
+
+ # Check for handling of non-Lustre FS
+ output=$($LFS df $notsup_mnt)
+ ret=$?
+
+ fs=$(echo $output | awk '{print $1}' | awk -F: '{print $2}')
+ [[ $fs = $notsup_mnt && $ret -eq $err_eopnotsupp ]] ||
+ error "expect success $err_eopnotsupp, not $ret"
+
+ # Check for multiple LustreFS argument
+ output=$($LFS df $MOUNT $MOUNT $MOUNT | grep -c "filesystem_summary:")
+ ret=$?
+
+ [[ $output -eq 3 && $ret -eq 0 ]] ||
+ error "expect success 3, not $output, rc = $ret"
+
+ # Check for correct non-Lustre FS handling among multiple
+ # LustreFS argument
+ output=$($LFS df $MOUNT $notsup_mnt $MOUNT |
+ grep -c "filesystem_summary:"; exit ${PIPESTATUS[0]})
+ ret=$?
+
+ [[ $output -eq 2 && $ret -eq $err_eopnotsupp ]] ||
+ error "expect success 2, not $output, rc = $ret"
+}
+run_test 56e "'lfs df' Handle non LustreFS & multiple LustreFS"
+
NUMFILES=3
NUMDIRS=3
setup_56() {
}
run_test 56xf "FID is not lost during migration of a composite layout file"
+check_file_ost_range() {
+ local file="$1"
+ shift
+ local range="$*"
+ local -a file_range
+ local idx
+
+ file_range=($($LFS getstripe -y "$file" |
+ awk '/l_ost_idx:/ { print $NF }'))
+
+ if [[ "${#file_range[@]}" = 0 ]]; then
+ echo "No osts found for $file"
+ return 1
+ fi
+
+ for idx in "${file_range[@]}"; do
+ [[ " $range " =~ " $idx " ]] ||
+ return 1
+ done
+
+ return 0
+}
+
+sub_test_56xg() {
+ local stripe_opt="$1"
+ local pool="$2"
+ shift 2
+ local pool_ostidx="$(seq $* | tr '\n' ' ')"
+
+ $LFS migrate $stripe_opt -p $pool $DIR/$tfile ||
+ error "Fail to migrate $tfile on $pool"
+ [[ "$($LFS getstripe -p $DIR/$tfile)" = "$pool" ]] ||
+ error "$tfile is not in pool $pool"
+ check_file_ost_range "$DIR/$tfile" $pool_ostidx ||
+ error "$tfile osts mismatch with pool $pool (osts $pool_ostidx)"
+}
+
+test_56xg() {
+ [[ $PARALLEL != "yes" ]] || skip "skip parallel run"
+ [[ $OSTCOUNT -ge 2 ]] || skip "needs >= 2 OSTs"
+ [[ $MDS1_VERSION -gt $(version_code 2.14.52) ]] ||
+ skip "Need MDS version newer than 2.14.52"
+
+ local -a pool_names=("${TESTNAME}_0" "${TESTNAME}_1" "${TESTNAME}_2")
+ local -a pool_ranges=("0 0" "1 1" "0 1")
+
+ # init pools
+ for i in "${!pool_names[@]}"; do
+ pool_add ${pool_names[$i]} ||
+ error "pool_add failed (pool: ${pool_names[$i]})"
+ pool_add_targets ${pool_names[$i]} ${pool_ranges[$i]} ||
+ error "pool_add_targets failed (pool: ${pool_names[$i]})"
+ done
+
+ # init the file to migrate
+ $LFS setstripe -c1 -i1 $DIR/$tfile ||
+ error "Unable to create $tfile on OST1"
+ dd if=/dev/urandom of=$DIR/$tfile bs=1M count=4 status=none ||
+ error "Unable to write on $tfile"
+
+ echo "1. migrate $tfile on pool ${pool_names[0]}"
+ sub_test_56xg "-c-1" "${pool_names[0]}" ${pool_ranges[0]}
+
+ echo "2. migrate $tfile on pool ${pool_names[2]}"
+ sub_test_56xg "-c-1 -S2M" "${pool_names[2]}" ${pool_ranges[2]}
+
+ echo "3. migrate $tfile on pool ${pool_names[1]}"
+ sub_test_56xg "-n -c-1" "${pool_names[1]}" ${pool_ranges[1]}
+
+ echo "4. migrate $tfile on pool ${pool_names[2]} with default stripe parameters"
+ sub_test_56xg "" "${pool_names[2]}" ${pool_ranges[2]}
+ echo
+
+ # Clean pools
+ destroy_test_pools ||
+ error "pool_destroy failed"
+}
+run_test 56xg "lfs migrate pool support"
+
test_56y() {
[ $MDS1_VERSION -lt $(version_code 2.4.53) ] &&
skip "No HSM $(lustre_build_version $SINGLEMDS) MDS < 2.4.53"
}
run_test 64f "check grant consumption (with grant allocation)"
+test_64g() {
+ #[ $MDS1_VERSION -lt $(version_code 2.14.54) ] &&
+ # skip "Need MDS version at least 2.14.54"
+
+ local mdts=$(comma_list $(mdts_nodes))
+
+ local old=$($LCTL get_param mdc.$FSNAME-*.grant_shrink_interval |
+ tr '\n' ' ')
+ stack_trap "$LCTL set_param $old"
+
+ # generate dirty pages and increase dirty granted on MDT
+ stack_trap "rm -f $DIR/$tfile-*"
+ for (( i = 0; i < 10; i++)); do
+ $LFS setstripe -E 1M -L mdt $DIR/$tfile-$i ||
+ error "can't set stripe"
+ dd if=/dev/zero of=$DIR/$tfile-$i bs=128k count=1 ||
+ error "can't dd"
+ $LFS getstripe $DIR/$tfile-$i | grep -q pattern.*mdt || {
+ $LFS getstripe $DIR/$tfile-$i
+ error "not DoM file"
+ }
+ done
+
+ # flush dirty pages
+ sync
+
+ # wait until grant shrink reset grant dirty on MDTs
+ for ((i = 0; i < 120; i++)); do
+ grant_dirty=$(do_nodes $mdts $LCTL get_param -n mdt.*.tot_dirty |
+ awk '{sum=sum+$1} END {print sum}')
+ vm_dirty=$(awk '/Dirty:/{print $2}' /proc/meminfo)
+ echo "$grant_dirty grants, $vm_dirty pages"
+ (( grant_dirty + vm_dirty == 0 )) && break
+ (( i == 3 )) && sync &&
+ $LCTL set_param mdc.$FSNAME-*.grant_shrink_interval=5
+ sleep 1
+ done
+
+ grant_dirty=$(do_nodes $mdts $LCTL get_param -n mdt.*.tot_dirty |
+ awk '{sum=sum+$1} END {print sum}')
+ (( grant_dirty == 0 )) || error "$grant_dirty on MDT"
+}
+run_test 64g "grant shrink on MDT"
+
+test_64h() {
+ local instance=$($LFS getname -i $DIR)
+ local osc_tgt="$FSNAME-OST0000-osc-$instance"
+ local num_exps=$(do_facet ost1 \
+ $LCTL get_param -n obdfilter.*OST0000*.num_exports)
+ local max_brw_size=$(import_param $osc_tgt max_brw_size)
+ local avail=$($LCTL get_param -n osc.*OST0000-osc-$instance.kbytesavail)
+ local p="$TMP/$TESTSUITE-$TESTNAME.parameters"
+
+ # 10MiB is for file to be written, max_brw_size * 16 *
+ # num_exps is space reserve so that tgt_grant_shrink() decided
+ # to not shrink
+ local expect=$((max_brw_size * 16 * num_exps + 10 * 1048576))
+ (( avail * 1024 < expect )) &&
+ skip "need $expect bytes on ost1, have $(( avail * 1024 )) only"
+
+ save_lustre_params client "osc.*OST0000*.grant_shrink" > $p
+ save_lustre_params client "osc.*OST0000*.grant_shrink_interval" >> $p
+ stack_trap "restore_lustre_params < $p; rm -f $save" EXIT
+ $LCTL set_param osc.*OST0000*.grant_shrink=1
+ $LCTL set_param osc.*OST0000*.grant_shrink_interval=10
+
+ $LFS setstripe -c 1 -i 0 $DIR/$tfile
+ dd if=/dev/zero of=$DIR/$tfile bs=1M count=10 oflag=sync
+
+ # drop cache so that coming read would do rpc
+ cancel_lru_locks osc
+
+ # shrink interval is set to 10, pause for 7 seconds so that
+ # grant thread did not wake up yet but coming read entered
+ # shrink mode for rpc (osc_should_shrink_grant())
+ sleep 7
+
+ declare -a cur_grant_bytes
+ declare -a tot_granted
+ cur_grant_bytes[0]=$($LCTL get_param -n osc.*OST0000*.cur_grant_bytes)
+ tot_granted[0]=$(do_facet ost1 \
+ $LCTL get_param -n obdfilter.*OST0000*.tot_granted)
+
+ dd if=$DIR/$tfile bs=4K count=1 of=/dev/null
+
+ cur_grant_bytes[1]=$($LCTL get_param -n osc.*OST0000*.cur_grant_bytes)
+ tot_granted[1]=$(do_facet ost1 \
+ $LCTL get_param -n obdfilter.*OST0000*.tot_granted)
+
+ # grant change should be equal on both sides
+ (( cur_grant_bytes[0] - cur_grant_bytes[1] ==
+ tot_granted[0] - tot_granted[1])) ||
+ error "grant change mismatch, " \
+ "server: ${tot_granted[0]} to ${tot_granted[1]}, " \
+ "client: ${cur_grant_bytes[0]} to ${cur_grant_bytes[1]}"
+}
+run_test 64h "grant shrink on read"
+
+test_64i() {
+ (( $OST1_VERSION >= $(version_code 2.14.55) )) ||
+ skip "need OST at least 2.14.55 to avoid grant shrink on replay"
+
+ [ $PARALLEL == "yes" ] && skip "skip parallel run"
+ remote_ost_nodsh && skip "remote OSTs with nodsh"
+
+ $LFS setstripe -c 1 -i 0 $DIR/$tfile
+
+ dd if=/dev/zero of=$DIR/$tfile bs=1M count=64
+
+ # lustre-ffff9fc75e850800 /mnt/lustre -> ffff9fc75e850800
+ local instance=$($LFS getname -i $DIR)
+
+ local osc_tgt="$FSNAME-OST0000-osc-$instance"
+ local cgb=$($LCTL get_param -n osc.$osc_tgt.cur_grant_bytes)
+
+ # shrink grants and simulate rpc loss
+ #define OBD_FAIL_PTLRPC_DROP_REQ_OPC 0x513
+ do_facet ost1 "$LCTL set_param fail_loc=0x80000513 fail_val=17"
+ $LCTL set_param osc.$osc_tgt.cur_grant_bytes=$((cgb/2))B
+
+ fail ost1
+
+ dd if=/dev/zero of=$DIR/$tfile oflag=append bs=1M count=8 conv=notrunc
+
+ local testid=$(echo $TESTNAME | tr '_' ' ')
+
+ do_facet ost1 dmesg | tac | sed "/$testid/,$ d" |
+ grep "GRANT, real grant" &&
+ error "client has more grants then it owns" || true
+}
+run_test 64i "shrink on reconnect"
+
# bug 1414 - set/get directories' stripe info
test_65a() {
[ $PARALLEL == "yes" ] && skip "skip parallel run"
}
run_test 77m "Verify checksum_speed is correctly read"
+check_filefrag_77n() {
+ local nr_ext=0
+ local starts=()
+ local ends=()
+
+ while read extidx a b start end rest; do
+ if [[ "${extidx}" =~ ^[0-9]+: ]]; then
+ nr_ext=$(( $nr_ext + 1 ))
+ starts+=( ${start%..} )
+ ends+=( ${end%:} )
+ fi
+ done < <( filefrag -sv $1 )
+
+ [[ $nr_ext -eq 2 ]] && [[ "${starts[-1]}" == $(( ${ends[0]} + 1 )) ]] && return 0
+ return 1
+}
+
+test_77n() {
+ [[ "$CKSUM_TYPES" =~ t10 ]] || skip "no T10 checksum support on osc"
+
+ touch $DIR/$tfile
+ $TRUNCATE $DIR/$tfile 0
+ dd if=/dev/urandom of=$DIR/$tfile bs=4k conv=notrunc count=1 seek=0
+ dd if=/dev/urandom of=$DIR/$tfile bs=4k conv=notrunc count=1 seek=2
+ check_filefrag_77n $DIR/$tfile ||
+ skip "$tfile blocks not contiguous around hole"
+
+ set_checksums 1
+ stack_trap "set_checksums $ORIG_CSUM" EXIT
+ stack_trap "set_checksum_type $ORIG_CSUM_TYPE" EXIT
+ stack_trap "rm -f $DIR/$tfile"
+
+ for algo in $CKSUM_TYPES; do
+ if [[ "$algo" =~ ^t10 ]]; then
+ set_checksum_type $algo ||
+ error "fail to set checksum type $algo"
+ dd if=$DIR/$tfile of=/dev/null bs=12k count=1 iflag=direct ||
+ error "fail to read $tfile with $algo"
+ fi
+ done
+ rm -f $DIR/$tfile
+ return 0
+}
+run_test 77n "Verify read from a hole inside contiguous blocks with T10PI"
+
+test_77o() {
+ (( $CLIENT_VERSION >= $(version_code 2.14.54) )) ||
+ skip "Need at least version 2.14.54"
+ local ofd=obdfilter
+ local mdt=mdt
+
+ # print OST checksum_type
+ echo "$ofd.$FSNAME-*.checksum_type:"
+ do_nodes $(comma_list $(osts_nodes)) \
+ $LCTL get_param -n $ofd.$FSNAME-*.checksum_type
+
+ # print MDT checksum_type
+ echo "$mdt.$FSNAME-*.checksum_type:"
+ do_nodes $(comma_list $(mdts_nodes)) \
+ $LCTL get_param -n $mdt.$FSNAME-*.checksum_type
+
+ local o_count=$(do_nodes $(comma_list $(osts_nodes)) \
+ $LCTL get_param -n $ofd.$FSNAME-*.checksum_type | wc -l)
+
+ (( $o_count == $OSTCOUNT )) ||
+ error "found $o_count checksums, not \$MDSCOUNT=$OSTCOUNT"
+
+ local m_count=$(do_nodes $(comma_list $(mdts_nodes)) \
+ $LCTL get_param -n $mdt.$FSNAME-*.checksum_type | wc -l)
+
+ (( $m_count == $MDSCOUNT )) ||
+ error "found $m_count checksums, not \$MDSCOUNT=$MDSCOUNT"
+}
+run_test 77o "Verify checksum_type for server (mdt and ofd(obdfilter))"
+
cleanup_test_78() {
trap 0
rm -f $DIR/$tfile
test_150c() {
check_set_fallocate_or_skip
+ local striping="-c2"
stack_trap "rm -f $DIR/$tfile; wait_delete_completed"
$LFS setstripe -c $OSTCOUNT -S1M $DIR/$tfile || error "setstripe failed"
fallocate -l ${OSTCOUNT}m $DIR/$tfile || error "fallocate failed"
- sync; sync_all_data
- cancel_lru_locks $OSC
- sleep 5
- bytes=$(($(stat -c '%b * %B' $DIR/$tfile)))
- want=$((OSTCOUNT * 1048576))
+ local bytes=$(($(stat -c '%b * %B' $DIR/$tfile)))
+ local want=$((OSTCOUNT * 1048576))
# Must allocate all requested space, not more than 5% extra
(( $bytes >= $want && $bytes < $want * 105 / 100 )) ||
error "bytes $bytes is not $want"
rm -f $DIR/$tfile
- # verify fallocate on PFL file
- $LFS setstripe -E1M -c1 -E16M -c3 -Eeof -c 4 $DIR/$tfile ||
+
+ echo "verify fallocate on PFL file"
+
+ [[ "x$DOM" == "xyes" ]] && striping="-L mdt"
+
+ $LFS setstripe -E1M $striping -E16M -c3 -Eeof -c 4 $DIR/$tfile ||
error "Create $DIR/$tfile failed"
- fallocate -l $((1048576 * 1024)) $DIR/$tfile ||
+ fallocate -l $((1048576 * 512)) $DIR/$tfile ||
error "fallocate failed"
- sync; sync_all_data
- cancel_lru_locks $OSC
- sleep 5
- local bytes=$(($(stat -c '%b * %B' $DIR/$tfile)))
- local want=$((1024 * 1048576))
+ bytes=$(($(stat -c '%b * %B' $DIR/$tfile)))
+ want=$((512 * 1048576))
# Must allocate all requested space, not more than 5% extra
(( $bytes >= $want && $bytes < $want * 105 / 100 )) ||
test_150d() {
check_set_fallocate_or_skip
+ local striping="-c2"
+
+ [[ "x$DOM" == "xyes" ]] && striping="-L mdt"
stack_trap "rm -f $DIR/$tfile; wait_delete_completed"
- $LFS setstripe -c $OSTCOUNT -S1M $DIR/$tdir || error "setstripe failed"
+ $LFS setstripe -E1M $striping -E eof -c $OSTCOUNT -S1M $DIR/$tdir ||
+ error "setstripe failed"
fallocate -o 1G -l ${OSTCOUNT}m $DIR/$tdir || error "fallocate failed"
- sync; sync_all_data
- cancel_lru_locks $OSC
- sleep 5
local bytes=$(($(stat -c '%b * %B' $DIR/$tdir)))
local want=$((OSTCOUNT * 1048576))
check_set_fallocate_or_skip
stack_trap "rm -f $DIR/$tfile; wait_delete_completed"
+ [[ "x$DOM" == "xyes" ]] &&
+ $LFS setstripe -E1M -L mdt -E eof $DIR/$tfile
+
echo "Verify fallocate punch: Range within the file range"
yes 'A' | dd of=$DIR/$tfile bs=4096 count=5 ||
error "dd failed for bs 4096 and count 5"
check_set_fallocate_or_skip
stack_trap "rm -f $DIR/$tfile; wait_delete_completed"
- $LFS setstripe -c${OSTCOUNT} $DIR/$tfile ||
- error "$LFS setstripe -c${OSTCOUNT} $DIR/$tfile failed"
+ if [[ "x$DOM" == "xyes" ]]; then
+ $LFS setstripe -E2M -L mdt -E eof -c${OSTCOUNT} $DIR/$tfile ||
+ error "$LFS setstripe DoM + ${OSTCOUNT} OST failed"
+ else
+ $LFS setstripe -c${OSTCOUNT} $DIR/$tfile ||
+ error "$LFS setstripe -c${OSTCOUNT} $DIR/$tfile failed"
+ fi
# Get 100MB per OST of the available space to reduce run time
# else 60% of the available space if we are running SLOW tests
# Create a user
changelog_register || error "changelog_register failed"
- # Delete a future user (expect fail)
local MDT0=$(facet_svc $SINGLEMDS)
- do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_deregister "cl77"
- local rc=$?
+ local rc
+
+ # No user (expect fail)
+ do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_deregister
+ rc=$?
+ if [ $rc -eq 0 ]; then
+ error "Should fail without user"
+ elif [ $rc -ne 4 ]; then
+ error "changelog_deregister failed with $rc, expect 4(CMD_HELP)"
+ fi
+ # Delete a future user (expect fail)
+ do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_deregister "cl77"
+ rc=$?
if [ $rc -eq 0 ]; then
error "Deleted non-existant user cl77"
elif [ $rc -ne 2 ]; then
test_160g() {
remote_mds_nodsh && skip "remote MDS with nodsh"
- [[ $MDS1_VERSION -ge $(version_code 2.10.56) ]] ||
- skip "Need MDS version at least 2.10.56"
+ [[ $MDS1_VERSION -ge $(version_code 2.14.55) ]] ||
+ skip "Need MDS version at least 2.14.55"
local mdts=$(comma_list $(mdts_nodes))
- #define OBD_FAIL_TIME_IN_CHLOG_USER 0x1314
- do_nodes $mdts $LCTL set_param fail_loc=0x1314
-
# Create a user
changelog_register || error "first changelog_register failed"
changelog_register || error "second changelog_register failed"
(( $nbcl > 0 )) || error "no changelogs found"
# reduce the max_idle_indexes value to make sure we exceed it
- for param in "changelog_max_idle_indexes=1" \
+ for param in "changelog_max_idle_indexes=2" \
"changelog_gc=1" \
- "changelog_min_gc_interval=2" \
- "changelog_min_free_cat_entries=3"; do
+ "changelog_min_gc_interval=2"; do
local MDT0=$(facet_svc $SINGLEMDS)
local var="${param%=*}"
local old=$(do_facet mds1 "$LCTL get_param -n mdd.$MDT0.$var")
error "unable to set mdd.*.$param"
done
- # simulate changelog catalog almost full
- #define OBD_FAIL_CAT_FREE_RECORDS 0x1313
- do_nodes $mdts "$LCTL set_param fail_loc=0x1313 fail_val=3"
-
local start=$SECONDS
for i in $(seq $MDSCOUNT); do
cl_users=(${CL_USERS[mds$i]})
cl_user2[mds$i]="${cl_users[1]}"
[ -n "${cl_user1[mds$i]}" ] ||
- error "mds$i: no user registered"
+ error "mds$i: user1 is not registered"
[ -n "${cl_user2[mds$i]}" ] ||
error "mds$i: only ${cl_user1[mds$i]} is registered"
user_rec1=$(changelog_user_rec mds$i ${cl_user1[mds$i]})
[ -n "$user_rec1" ] ||
- error "mds$i: User ${cl_user1[mds$i]} not registered"
+ error "mds$i: user1 ${cl_user1[mds$i]} not found"
__changelog_clear mds$i ${cl_user1[mds$i]} +2
user_rec2=$(changelog_user_rec mds$i ${cl_user1[mds$i]})
[ -n "$user_rec2" ] ||
- error "mds$i: User ${cl_user1[mds$i]} not registered"
- echo "mds$i: verifying user ${cl_user1[mds$i]} clear: " \
+ error "mds$i: user1 ${cl_user1[mds$i]} not found (2)"
+ echo "mds$i: verifying user1 ${cl_user1[mds$i]} clear: " \
"$user_rec1 + 2 == $user_rec2"
[ $((user_rec1 + 2)) == $user_rec2 ] ||
- error "mds$i: user ${cl_user1[mds$i]} index expected " \
- "$user_rec1 + 2, but is $user_rec2"
+ error "mds$i: user1 ${cl_user1[mds$i]} index " \
+ "expected $user_rec1 + 2, but is $user_rec2"
user_rec2=$(changelog_user_rec mds$i ${cl_user2[mds$i]})
[ -n "$user_rec2" ] ||
- error "mds$i: User ${cl_user2[mds$i]} not registered"
+ error "mds$i: user2 ${cl_user2[mds$i]} not found"
[ $user_rec1 == $user_rec2 ] ||
- error "mds$i: user ${cl_user2[mds$i]} index expected " \
- "$user_rec1, but is $user_rec2"
+ error "mds$i: user2 ${cl_user2[mds$i]} index " \
+ "expected $user_rec1, but is $user_rec2"
done
# ensure we are past the previous changelog_min_gc_interval set above
local sleep2=$((start + 2 - SECONDS))
(( sleep2 > 0 )) && echo "sleep $sleep2 for interval" && sleep $sleep2
-
# Generate one more changelog to trigger GC at fail_loc for cl_user2.
# cl_user1 should be OK because it recently processed records.
for ((i = 0; i < MDSCOUNT; i++)); do
- $LFS mkdir -i $i $DIR/$tdir/d$i.3 $DIR/$tdir/d$i.4 ||
+ $LFS mkdir -i $i $DIR/$tdir/d$i.3 ||
error "create $DIR/$tdir/d$i.3 failed"
done
for (( i = 1; i <= MDSCOUNT; i++ )); do
# check cl_user1 still registered
changelog_users mds$i | grep -q "${cl_user1[mds$i]}" ||
- error "mds$i: User ${cl_user1[mds$i]} not registered"
+ error "mds$i: user1 ${cl_user1[mds$i]} not found (3)"
# check cl_user2 unregistered
changelog_users mds$i | grep -q "${cl_user2[mds$i]}" &&
- error "mds$i: User ${cl_user2[mds$i]} still registered"
+ error "mds$i: user2 ${cl_user2[mds$i]} is registered"
# check changelogs are present and starting at $user_rec1 + 1
user_rec1=$(changelog_user_rec mds$i ${cl_user1[mds$i]})
[ -n "$user_rec1" ] ||
- error "mds$i: User ${cl_user1[mds$i]} not registered"
+ error "mds$i: user1 ${cl_user1[mds$i]} not found (4)"
first_rec=$($LFS changelog $(facet_svc mds$i) |
awk '{ print $1; exit; }')
error "mds$i: rec $first_rec != $user_rec1 + 1"
done
}
-run_test 160g "changelog garbage collect (old users)"
+run_test 160g "changelog garbage collect on idle records"
test_160h() {
remote_mds_nodsh && skip "remote MDS with nodsh" && return
changelog_register --user test_160o -m unlnk+close+open ||
error "changelog_register failed"
- # drop server mask so it doesn't interfere
+
do_facet $SINGLEMDS $LCTL --device $mdt \
changelog_register -u "Tt3_-#" &&
error "bad symbols in name should fail"
}
run_test 160p "Changelog orphan cleanup with no users"
+test_160q() {
+ local mdt="$(facet_svc $SINGLEMDS)"
+ local clu
+
+ [[ $PARALLEL != "yes" ]] || skip "skip parallel run"
+ remote_mds_nodsh && skip "remote MDS with nodsh"
+ [ $MDS1_VERSION -ge $(version_code 2.14.54) ] ||
+ skip "Need MDS version at least 2.14.54"
+
+ # set server mask to minimal value like server init does
+ changelog_chmask "MARK"
+ clu=$(do_facet $SINGLEMDS $LCTL --device $mdt changelog_register -n) ||
+ error "changelog_register failed"
+ # check effective mask again, should be treated as DEFMASK now
+ mask=$(do_facet $SINGLEMDS $LCTL get_param \
+ mdd.$mdt.changelog_current_mask -n)
+ do_facet $SINGLEMDS $LCTL --device $mdt changelog_deregister $clu ||
+ error "changelog_deregister failed"
+ [[ $mask == *"HLINK"* ]] || error "mask is not DEFMASK as expected"
+}
+run_test 160q "changelog effective mask is DEFMASK if not set"
+
+test_160s() {
+ remote_mds_nodsh && skip "remote MDS with nodsh"
+ (( $MDS1_VERSION >= $(version_code 2.14.55) )) ||
+ skip "Need MDS version at least 2.14.55"
+
+ local mdts=$(comma_list $(mdts_nodes))
+
+ #define OBD_FAIL_TIME_IN_CHLOG_USER 0x1314
+ do_nodes $mdts $LCTL set_param fail_loc=0x1314 \
+ fail_val=$((24 * 3600 * 10))
+
+ # Create a user which is 10 days old
+ changelog_register || error "first changelog_register failed"
+ local cl_users
+ declare -A cl_user1
+ local i
+
+ # generate some changelog records to accumulate on each MDT
+ # use all_char because created files should be evenly distributed
+ test_mkdir -c $MDSCOUNT -H all_char $DIR/$tdir ||
+ error "test_mkdir $tdir failed"
+ for ((i = 0; i < MDSCOUNT; i++)); do
+ $LFS mkdir -i $i $DIR/$tdir/d$i.1 $DIR/$tdir/d$i.2 ||
+ error "create $DIR/$tdir/d$i.1 failed"
+ done
+
+ # check changelogs have been generated
+ local nbcl=$(changelog_dump | wc -l)
+ (( nbcl > 0 )) || error "no changelogs found"
+
+ # reduce the max_idle_indexes value to make sure we exceed it
+ for param in "changelog_max_idle_indexes=2097446912" \
+ "changelog_max_idle_time=2592000" \
+ "changelog_gc=1" \
+ "changelog_min_gc_interval=2"; do
+ local MDT0=$(facet_svc $SINGLEMDS)
+ local var="${param%=*}"
+ local old=$(do_facet mds1 "$LCTL get_param -n mdd.$MDT0.$var")
+
+ stack_trap "do_nodes $mdts $LCTL set_param mdd.*.$var=$old" EXIT
+ do_nodes $mdts $LCTL set_param mdd.*.$param ||
+ error "unable to set mdd.*.$param"
+ done
+
+ local start=$SECONDS
+ for i in $(seq $MDSCOUNT); do
+ cl_users=(${CL_USERS[mds$i]})
+ cl_user1[mds$i]="${cl_users[0]}"
+
+ [[ -n "${cl_user1[mds$i]}" ]] ||
+ error "mds$i: no user registered"
+ done
+
+ #define OBD_FAIL_MDS_CHANGELOG_IDX_PUMP 0x16d
+ do_nodes $mdts $LCTL set_param fail_loc=0x16d fail_val=500000000
+
+ # ensure we are past the previous changelog_min_gc_interval set above
+ local sleep2=$((start + 2 - SECONDS))
+ (( sleep2 > 0 )) && echo "sleep $sleep2 for interval" && sleep $sleep2
+
+ # Generate one more changelog to trigger GC
+ for ((i = 0; i < MDSCOUNT; i++)); do
+ $LFS mkdir -i $i $DIR/$tdir/d$i.3 $DIR/$tdir/d$i.4 ||
+ error "create $DIR/$tdir/d$i.3 failed"
+ done
+
+ # ensure gc thread is done
+ for node in $(mdts_nodes); do
+ wait_update $node "pgrep chlg_gc_thread" "" 20 ||
+ error "$node: GC-thread not done"
+ done
+
+ do_nodes $mdts $LCTL set_param fail_loc=0
+
+ for (( i = 1; i <= MDSCOUNT; i++ )); do
+ # check cl_user1 is purged
+ changelog_users mds$i | grep -q "${cl_user1[mds$i]}" &&
+ error "mds$i: User ${cl_user1[mds$i]} is registered"
+ done
+ return 0
+}
+run_test 160s "changelog garbage collect on idle records * time"
+
test_161a() {
[ $PARALLEL == "yes" ] && skip "skip parallel run"
$MULTIOP $DIR/$tfile oO_CREAT:O_RDWR:eRE+eU || error "get lease error"
echo "==== test 2: verify lease can be broken by upcoming open"
- $MULTIOP $DIR/$tfile oO_RDONLY:eR_E-eUc &
+ $MULTIOP $DIR/$tfile oO_RDWR:eR_E-eUc &
local PID=$!
sleep 1
- $MULTIOP $DIR/$tfile oO_RDONLY:c
+ $MULTIOP $DIR/$tfile oO_RDWR:c
kill -USR1 $PID && wait $PID || error "break lease error"
echo "==== test 3: verify lease can't be granted if an open already exists"
- $MULTIOP $DIR/$tfile oO_RDONLY:_c &
+ $MULTIOP $DIR/$tfile oO_RDWR:_c &
local PID=$!
sleep 1
- $MULTIOP $DIR/$tfile oO_RDONLY:eReUc && error "apply lease should fail"
+ $MULTIOP $DIR/$tfile oO_RDWR:eReUc && error "apply lease should fail"
kill -USR1 $PID && wait $PID || error "open file error"
echo "==== test 4: lease can sustain over recovery"
- $MULTIOP $DIR/$tfile oO_RDONLY:eR_E+eUc &
+ $MULTIOP $DIR/$tfile oO_RDWR:eR_E+eUc &
PID=$!
sleep 1
kill -USR1 $PID && wait $PID || error "lease broken over recovery"
echo "==== test 5: lease broken can't be regained by replay"
- $MULTIOP $DIR/$tfile oO_RDONLY:eR_E-eUc &
+ $MULTIOP $DIR/$tfile oO_RDWR:eR_E-eUc &
PID=$!
sleep 1
test_224a() { # LU-1039, MRP-303
[ $PARALLEL == "yes" ] && skip "skip parallel run"
-
#define OBD_FAIL_PTLRPC_CLIENT_BULK_CB 0x508
$LCTL set_param fail_loc=0x508
- dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 conv=fsync
+ dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 conv=fsync
$LCTL set_param fail_loc=0
df $DIR
}
run_test 224a "Don't panic on bulk IO failure"
-test_224b() { # LU-1039, MRP-303
+test_224bd_sub() { # LU-1039, MRP-303
[ $PARALLEL == "yes" ] && skip "skip parallel run"
+ local timeout=$1
- dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1
+ shift
+ dd if=/dev/urandom of=$TMP/$tfile bs=1M count=1
+
+ $LFS setstripe -c 1 -i 0 $DIR/$tfile
+
+ dd if=$TMP/$tfile of=$DIR/$tfile bs=1M count=1
cancel_lru_locks osc
+ set_checksums 0
+ stack_trap "set_checksums $ORIG_CSUM" EXIT
+ local at_max_saved=0
+
+ # adaptive timeouts may prevent seeing the issue
+ if at_is_enabled; then
+ at_max_saved=$(at_max_get mds)
+ at_max_set 0 mds client
+ stack_trap "at_max_set $at_max_saved mds client" EXIT
+ fi
+
#define OBD_FAIL_PTLRPC_CLIENT_BULK_CB2 0x515
- $LCTL set_param fail_loc=0x515
- dd of=/dev/null if=$DIR/$tfile bs=4096 count=1
- $LCTL set_param fail_loc=0
+ do_facet ost1 $LCTL set_param fail_val=$timeout fail_loc=0x80000515
+ dd of=$TMP/$tfile.new if=$DIR/$tfile bs=1M count=1 || "$@"
+
+ do_facet ost1 $LCTL set_param fail_loc=0
+ cmp $TMP/$tfile $TMP/$tfile.new || error "file contents wrong"
df $DIR
}
+
+test_224b() {
+ test_224bd_sub 3 error "dd failed"
+}
run_test 224b "Don't panic on bulk IO failure"
test_224c() { # LU-6441
}
run_test 224c "Don't hang if one of md lost during large bulk RPC"
+test_224d() { # LU-11169
+ test_224bd_sub $((TIMEOUT + 2)) error "dd failed"
+}
+run_test 224d "Don't corrupt data on bulk IO timeout"
+
MDSSURVEY=${MDSSURVEY:-$(which mds-survey 2>/dev/null || true)}
test_225a () {
[ $PARALLEL == "yes" ] && skip "skip parallel run"
error "migrate remote dir error"
echo "Finish migration, then checking.."
- for file in $(find $migrate_dir); do
+ for file in $(find $migrate_dir -maxdepth 1); do
mdt_index=$($LFS getstripe -m $file)
if [ $mdt_index -lt $new_index ] ||
[ $mdt_index -gt $((new_index + new_count - 1)) ]; then
}
run_test 230t "migrate directory with project ID set"
+test_230u()
+{
+ (( MDSCOUNT > 3 )) || skip_env "needs >= 4 MDTs"
+ (( MDS1_VERSION >= $(version_code 2.14.53) )) ||
+ skip "Need MDS version at least 2.14.53"
+
+ local count
+
+ mkdir_on_mdt0 $DIR/$tdir || error "mkdir $tdir failed"
+ mkdir $DIR/$tdir/sub{0..99} || error "mkdir sub failed"
+ $LFS migrate -m -1 $DIR/$tdir/sub{0..99} || error "migrate sub failed"
+ for i in $(seq 0 $((MDSCOUNT - 1))); do
+ count=$($LFS getstripe -m $DIR/$tdir/sub* | grep -c ^$i)
+ echo "$count dirs migrated to MDT$i"
+ done
+ count=$($LFS getstripe -m $DIR/$tdir/sub* | sort -u | wc -l)
+ (( count >= MDSCOUNT - 1 )) || error "dirs migrated to $count MDTs"
+}
+run_test 230u "migrate directory by QOS"
+
+test_230v()
+{
+ (( MDSCOUNT > 3 )) || skip_env "needs >= 4 MDTs"
+ (( MDS1_VERSION >= $(version_code 2.14.53) )) ||
+ skip "Need MDS version at least 2.14.53"
+
+ local count
+
+ mkdir $DIR/$tdir || error "mkdir $tdir failed"
+ mkdir $DIR/$tdir/sub{0..99} || error "mkdir sub failed"
+ $LFS migrate -m 0,2,1 $DIR/$tdir || error "migrate $tdir failed"
+ for i in $(seq 0 $((MDSCOUNT - 1))); do
+ count=$($LFS getstripe -m $DIR/$tdir/sub* | grep -c ^$i)
+ echo "$count subdirs migrated to MDT$i"
+ (( i == 3 )) && (( count > 0 )) &&
+ error "subdir shouldn't be migrated to MDT3"
+ done
+ count=$($LFS getstripe -m $DIR/$tdir/sub* | sort -u | wc -l)
+ (( count == 3 )) || error "dirs migrated to $count MDTs"
+}
+run_test 230v "subdir migrated to the MDT where its parent is located"
+
+test_230w() {
+ (( MDSCOUNT > 1 )) || skip "needs >= 2 MDTs"
+ (( MDS1_VERSION >= $(version_code 2.14.53) )) ||
+ skip "Need MDS version at least 2.14.53"
+
+ mkdir -p $DIR/$tdir/sub || error "mkdir failed"
+
+ $LFS migrate -m 1 -c $MDSCOUNT -d $DIR/$tdir ||
+ error "migrate failed"
+
+ (( $($LFS getdirstripe -c $DIR/$tdir) == MDSCOUNT )) ||
+ error "$tdir stripe count mismatch"
+
+ (( $($LFS getdirstripe -c $DIR/$tdir/sub) == 0 )) ||
+ error "$tdir/sub is striped"
+}
+run_test 230w "non-recursive mode dir migration"
+
test_231a()
{
# For simplicity this test assumes that max_pages_per_rpc
test_254() {
[ $PARALLEL == "yes" ] && skip "skip parallel run"
remote_mds_nodsh && skip "remote MDS with nodsh"
- do_facet $SINGLEMDS $LCTL get_param -n mdd.$MDT0.changelog_size ||
+
+ local mdt=$(facet_svc $SINGLEMDS)
+
+ do_facet $SINGLEMDS $LCTL get_param -n mdd.$mdt.changelog_size ||
skip "MDS does not support changelog_size"
local cl_user
- local MDT0=$(facet_svc $SINGLEMDS)
changelog_register || error "changelog_register failed"
changelog_clear 0 || error "changelog_clear failed"
local size1=$(do_facet $SINGLEMDS \
- $LCTL get_param -n mdd.$MDT0.changelog_size)
+ $LCTL get_param -n mdd.$mdt.changelog_size)
echo "Changelog size $size1"
rm -rf $DIR/$tdir
rm $DIR/$tdir/pics/desktop.jpg
local size2=$(do_facet $SINGLEMDS \
- $LCTL get_param -n mdd.$MDT0.changelog_size)
+ $LCTL get_param -n mdd.$mdt.changelog_size)
echo "Changelog size after work $size2"
(( $size2 > $size1 )) ||
}
run_test 270h "DoM: DoM stripe removal when disabled on server"
+test_270i() {
+ (( $MDS1_VERSION >= $(version_code 2.14.54) )) ||
+ skip "Need MDS version at least 2.14.54"
+
+ mkdir $DIR/$tdir
+ $LFS setstripe -L mdt -S 128k -c -1 $DIR/$tdir &&
+ error "setstripe should fail" || true
+}
+run_test 270i "DoM: setting invalid DoM striping should fail"
+
test_271a() {
[ $MDS1_VERSION -lt $(version_code 2.10.55) ] &&
skip "Need MDS version at least 2.10.55"
#
# sparse file test
- # Create file with a hole and write actual two blocks. Block count
- # must be 16.
+ # Create file with a hole and write actual 65536 bytes which aligned
+ # with 4K and 64K PAGE_SIZE. Block count must be 128.
#
- dd if=/dev/zero of=$DIR/$tfile bs=$grant_blk_size count=2 seek=5 \
- conv=fsync || error "Create file : $DIR/$tfile"
-
- # Calculate the final truncate size.
- trunc_sz=$(($(stat --format=%s $DIR/$tfile) - (grant_blk_size + 1)))
+ local bs=65536
+ dd if=/dev/zero of=$DIR/$tfile bs=$bs count=1 seek=5 conv=fsync ||
+ error "Create file : $DIR/$tfile"
#
- # truncate to size $trunc_sz bytes. Strip the last block
- # The block count must drop to 8
+ # Truncate to size $trunc_sz bytes. Strip tail blocks and leave only 8
+ # blocks. The block count must drop to 8.
#
+ trunc_sz=$(($(stat --format=%s $DIR/$tfile) - \
+ ((bs - grant_blk_size) + 1)))
$TRUNCATE $DIR/$tfile $trunc_sz ||
error "truncate $tfile to $trunc_sz failed"
generate_uneven_mdts() {
local threshold=$1
+ local lmv_qos_maxage
+ local lod_qos_maxage
local ffree
local bavail
local max
local tmp
local i
+ lmv_qos_maxage=$($LCTL get_param -n lmv.*.qos_maxage)
+ $LCTL set_param lmv.*.qos_maxage=1
+ stack_trap "$LCTL set_param \
+ lmv.*.qos_maxage=$lmv_qos_maxage > /dev/null" RETURN
+ lod_qos_maxage=$(do_facet mds1 $LCTL get_param -n \
+ lod.$FSNAME-MDT0000-mdtlov.qos_maxage | awk '{ print $1 }')
+ do_nodes $(comma_list $(mdts_nodes)) $LCTL set_param \
+ lod.*.mdt_qos_maxage=1
+ stack_trap "do_nodes $(comma_list $(mdts_nodes)) $LCTL set_param \
+ lod.*.mdt_qos_maxage=$lod_qos_maxage > /dev/null" RETURN
+
echo
echo "Check for uneven MDTs: "
fi
done
+ (( ${ffree[min_index]} > 0 )) ||
+ skip "no free files in MDT$min_index"
+ (( ${ffree[min_index]} < 10000000 )) ||
+ skip "too many free files in MDT$min_index"
+
# Check if we need to generate uneven MDTs
local diff=$(((max - min) * 100 / min))
local testdir=$DIR/$tdir-fillmdt
+ local start
mkdir -p $testdir
while (( diff < threshold )); do
# generate uneven MDTs, create till $threshold% diff
echo -n "weight diff=$diff% must be > $threshold% ..."
- echo "Fill MDT$min_index with 100 files: loop $i"
+ echo "Fill MDT$min_index with 1000 files: loop $i"
testdir=$DIR/$tdir-fillmdt/$i
[ -d $testdir ] || $LFS mkdir -i $min_index $testdir ||
error "mkdir $testdir failed"
$LFS setstripe -E 1M -L mdt $testdir ||
error "setstripe $testdir failed"
- for F in f.{0..99}; do
- dd if=/dev/zero of=$testdir/$F bs=1M count=1 > \
+ start=$SECONDS
+ for F in f.{0..999}; do
+ dd if=/dev/zero of=$testdir/$F bs=64K count=1 > \
/dev/null 2>&1 || error "dd $F failed"
done
+ # wait for QOS to update
+ (( SECONDS < start + 1 )) && sleep $((start + 1 - SECONDS))
+
ffree=($(lctl get_param -n mdc.*[mM][dD][cC]-*.filesfree))
bavail=($(lctl get_param -n mdc.*[mM][dD][cC]-*.kbytesavail))
max=$(((${ffree[max_index]} >> 8) * \
local stripe_index=$($LFS getstripe -m $testdir)
local test_mkdir_rr=true
- echo "dirstripe: '$($LFS getdirstripe $testdir)'"
getfattr -d -m dmv -e hex $testdir | grep dmv
if (( $? == 0 && $MDS1_VERSION >= $(version_code 2.14.51) )); then
echo "defstripe: '$($LFS getdirstripe -D $testdir)'"
(( ${ffree[min_index]} > 0 )) ||
skip "no free files in MDT$min_index"
- (( ${ffree[min_index]} < 100000000 )) ||
+ (( ${ffree[min_index]} < 10000000 )) ||
skip "too many free files in MDT$min_index"
echo "MDT filesfree available: ${ffree[@]}"
error "$mkdir_cmd subdir$i failed"
done
+ max=0
for (( i = 0; i < $MDSCOUNT; i++ )); do
count=$($LFS getdirstripe -i $testdir/* | grep -c "^$i$")
+ (( count > max )) && max=$count
echo "$count directories created on MDT$i"
-
- if [ $stripe_count -gt 1 ]; then
- count=$($LFS getdirstripe $testdir/* |
- grep -c -P "^\s+$i\t")
- echo "$count stripes created on MDT$i"
- fi
done
- max=$($LFS getdirstripe -i $testdir/* | grep -c "^$max_index$")
min=$($LFS getdirstripe -i $testdir/* | grep -c "^$min_index$")
# D-value should > 10% of averge
- (( max - min >= num / 10 )) ||
+ (( max - min > num / 10 )) ||
error "subdirs shouldn't be evenly distributed: $max - $min < $((num / 10))"
- # 5% for stripes
+ # ditto for stripes
if (( stripe_count > 1 )); then
- max=$($LFS getdirstripe $testdir/* |
- grep -c -P "^\s+$max_index\t")
+ max=0
+ for (( i = 0; i < $MDSCOUNT; i++ )); do
+ count=$($LFS getdirstripe $testdir/* |
+ grep -c -P "^\s+$i\t")
+ (( count > max )) && max=$count
+ echo "$count stripes created on MDT$i"
+ done
+
min=$($LFS getdirstripe $testdir/* |
grep -c -P "^\s+$min_index\t")
- (( max - min >= num * stripe_count / 20 )) ||
- error "stripes shouldn't be evenly distributed: $max - $min < $((num / 20)) * $stripe_count"
+ (( max - min > num * stripe_count / 10 )) ||
+ error "stripes shouldn't be evenly distributed: $max - $min < $((num / 10)) * $stripe_count"
fi
}
}
run_test 413c "mkdir with default LMV max inherit rr"
+test_413d() {
+ (( MDSCOUNT >= 2 )) ||
+ skip "We need at least 2 MDTs for this test"
+
+ (( MDS1_VERSION >= $(version_code 2.14.51) )) ||
+ skip "Need server version at least 2.14.51"
+
+ local lmv_qos_threshold_rr
+
+ lmv_qos_threshold_rr=$($LCTL get_param -n lmv.*.qos_threshold_rr |
+ head -n1)
+ stack_trap "$LCTL set_param \
+ lmv.*.qos_threshold_rr=$lmv_qos_threshold_rr > /dev/null" EXIT
+
+ $LCTL set_param lmv.*.qos_threshold_rr=100 > /dev/null
+ mkdir -p $DIR/$tdir || error "mkdir $tdir failed"
+ getfattr -d -m dmv -e hex $DIR/$tdir | grep dmv &&
+ error "$tdir shouldn't have default LMV"
+ createmany -d $DIR/$tdir/sub $((100 * MDSCOUNT)) ||
+ error "mkdir sub failed"
+
+ local count=$($LFS getstripe -m $DIR/$tdir/* | grep -c ^0)
+
+ (( count == 100 )) || error "$count subdirs on MDT0"
+}
+run_test 413d "inherit ROOT default LMV"
+
test_413z() {
local pids=""
local subdir
local pid
for subdir in $(\ls -1 -d $DIR/d413*-fillmdt/*); do
- unlinkmany $subdir/f. 100 &
+ unlinkmany $subdir/f. 1000 &
pids="$pids $!"
done
[ "$1" == "blocks" ] && inodes= || inodes="-i"
for count in {1..100}; do
- cancel_lru_locks
+ do_nodes "$CLIENTS" \
+ $LCTL set_param ldlm.namespaces.*.lru_size=clear
sync; sleep 0.2
# read the lines of interest
# "<NID>:/<fsname>" for df, "filesystem_summary:" for lfs df
# compare the two outputs
passed=true
- for i in {1..5}; do
+ # skip "available" on MDT until LU-13997 is fixed.
+ #for i in {1..5}; do
+ for i in 1 2 4 5; do
[ "${df_out[i]}" != "${lfs_df_out[i]}" ] && passed=false
done
$passed && break
run_test 817 "nfsd won't cache write lock for exec file"
test_818() {
- mkdir $DIR/$tdir
- $LFS setstripe -c1 -i0 $DIR/$tfile
- $LFS setstripe -c1 -i1 $DIR/$tfile
+ test_mkdir -i0 -c1 $DIR/$tdir
+ $LFS setstripe -c1 -i0 $DIR/$tdir/$tfile
+ $LFS setstripe -c1 -i1 $DIR/$tdir/$tfile
stop $SINGLEMDS
+
+ # restore osp-syn threads
+ stack_trap "fail $SINGLEMDS"
+
#define OBD_FAIL_OSP_CANT_PROCESS_LLOG 0x2105
do_facet $SINGLEMDS lctl set_param fail_loc=0x80002105
start $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) $MDS_MOUNT_OPTS ||
error "start $SINGLEMDS failed"
rm -rf $DIR/$tdir
+
+ local testid=$(echo $TESTNAME | tr '_' ' ')
+
+ do_facet mds1 dmesg | tac | sed "/$testid/,$ d" |
+ grep "run LFSCK" || error "run LFSCK is not suggested"
}
run_test 818 "unlink with failed llog"
}
run_test 822 "test precreate failure"
+test_823() {
+ local p="$TMP/$TESTSUITE-$TESTNAME.parameters"
+ local OST_MAX_PRECREATE=20000
+
+ save_lustre_params mds1 \
+ "osp.$FSNAME-OST*-osc-MDT0000.max_create_count" > $p
+ do_facet $SINGLEMDS "$LCTL set_param -n \
+ osp.$FSNAME-OST*MDT0000.max_create_count=0"
+ do_facet $SINGLEMDS "$LCTL set_param -n \
+ osp.$FSNAME-OST0000*MDT0000.max_create_count=$OST_MAX_PRECREATE"
+
+ stack_trap "restore_lustre_params < $p; rm $p"
+
+ do_facet $SINGLEMDS "$LCTL set_param -n \
+ osp.$FSNAME-OST*-osc-MDT*.create_count=100200"
+
+ local count=$(do_facet $SINGLEMDS "$LCTL get_param -n \
+ osp.$FSNAME-OST0000*MDT0000.create_count")
+ local max=$(do_facet $SINGLEMDS "$LCTL get_param -n \
+ osp.$FSNAME-OST0000*MDT0000.max_create_count")
+ local expect_count=$(((($max/2)/256) * 256))
+
+ log "setting create_count to 100200:"
+ log " -result- count: $count with max: $max, expecting: $expect_count"
+
+ [[ $count -eq expect_count ]] ||
+ error "Create count not set to max precreate."
+}
+run_test 823 "Setting create_count > OST_MAX_PRECREATE is lowered to maximum"
+
+test_831() {
+ local sync_changes=$(do_facet $SINGLEMDS \
+ $LCTL get_param -n osp.$FSNAME-OST0000-osc-MDT0000.sync_changes)
+
+ [ "$sync_changes" -gt 100 ] &&
+ skip "Sync changes $sync_changes > 100 already"
+
+ local p="$TMP/$TESTSUITE-$TESTNAME.parameters"
+
+ $LFS mkdir -i 0 $DIR/$tdir
+ $LFS setstripe -c 1 -i 0 $DIR/$tdir
+
+ save_lustre_params mds1 \
+ "osp.$FSNAME-OST*-osc-MDT0000.max_sync_changes" > $p
+ save_lustre_params mds1 \
+ "osp.$FSNAME-OST*-osc-MDT0000.max_rpcs_in_progress" >> $p
+
+ do_facet mds1 "$LCTL set_param -n \
+ osp.$FSNAME-OST*-osc-MDT0000.max_sync_changes=100 \
+ osp.$FSNAME-OST*-osc-MDT0000.max_rpcs_in_progress=128"
+ stack_trap "restore_lustre_params < $p" EXIT
+
+ createmany -o $DIR/$tdir/f- 1000
+ unlinkmany $DIR/$tdir/f- 1000 &
+ local UNLINK_PID=$!
+
+ while sleep 1; do
+ sync_changes=$(do_facet mds1 \
+ $LCTL get_param -n osp.$FSNAME-OST0000-osc-MDT0000.sync_changes)
+ # the check in the code is racy, fail the test
+ # if the value above the limit by 10.
+ [ $sync_changes -gt 110 ] && {
+ kill -2 $UNLINK_PID
+ wait
+ error "osp changes throttling failed, $sync_changes>110"
+ }
+ kill -0 $UNLINK_PID 2> /dev/null || break
+ done
+ wait
+}
+run_test 831 "throttling unlink/setattr queuing on OSP"
+
#
# tests that do cleanup/setup should be run at the end
#
}
run_test 902 "test short write doesn't hang lustre"
+# LU-14711
+test_903() {
+ $LFS setstripe -i 0 -c 1 $DIR/$tfile $DIR/${tfile}-2
+ echo "blah" > $DIR/${tfile}-2
+ dd if=/dev/zero of=$DIR/$tfile bs=1M count=6 conv=fsync
+ #define OBD_FAIL_OSC_SLOW_PAGE_EVICT 0x417
+ $LCTL set_param fail_loc=0x417 fail_val=20
+
+ mv $DIR/${tfile}-2 $DIR/$tfile # Destroys the big object
+ sleep 1 # To start the destroy
+ wait_destroy_complete 150 || error "Destroy taking too long"
+ cat $DIR/$tfile > /dev/null || error "Evicted"
+}
+run_test 903 "Test long page discard does not cause evictions"
+
complete $SECONDS
[ -f $EXT2_DEV ] && rm $EXT2_DEV || true
check_and_cleanup_lustre