2016-05-31 01:51:59 +02:00
|
|
|
/*
|
2019-08-19 20:35:52 +02:00
|
|
|
* MinIO Cloud Storage, (C) 2016-2019 MinIO, Inc.
|
2016-05-31 01:51:59 +02:00
|
|
|
*
|
|
|
|
* Licensed under the Apache License, Version 2.0 (the "License");
|
|
|
|
* you may not use this file except in compliance with the License.
|
|
|
|
* You may obtain a copy of the License at
|
|
|
|
*
|
|
|
|
* http://www.apache.org/licenses/LICENSE-2.0
|
|
|
|
*
|
|
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
|
|
* distributed under the License is distributed on an "AS IS" BASIS,
|
|
|
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
|
|
* See the License for the specific language governing permissions and
|
|
|
|
* limitations under the License.
|
|
|
|
*/
|
|
|
|
|
2016-08-19 01:23:42 +02:00
|
|
|
package cmd
|
2016-05-26 01:42:31 +02:00
|
|
|
|
|
|
|
import (
|
2018-04-06 00:04:40 +02:00
|
|
|
"context"
|
2019-11-09 18:27:23 +01:00
|
|
|
"encoding/json"
|
2017-02-01 00:34:49 +01:00
|
|
|
"errors"
|
2016-07-08 16:33:21 +02:00
|
|
|
"hash/crc32"
|
2016-06-03 01:34:15 +02:00
|
|
|
"path"
|
2016-09-09 07:38:18 +02:00
|
|
|
|
2018-04-06 00:04:40 +02:00
|
|
|
"github.com/minio/minio/cmd/logger"
|
2019-10-14 18:44:51 +02:00
|
|
|
"github.com/minio/minio/pkg/sync/errgroup"
|
2016-05-26 01:42:31 +02:00
|
|
|
)
|
|
|
|
|
2016-07-13 20:56:25 +02:00
|
|
|
// Returns number of errors that occurred the most (incl. nil) and the
|
2017-08-15 03:09:29 +02:00
|
|
|
// corresponding error value. NB When there is more than one error value that
|
2016-07-13 20:56:25 +02:00
|
|
|
// occurs maximum number of times, the error value returned depends on how
|
|
|
|
// golang's map orders keys. This doesn't affect correctness as long as quorum
|
|
|
|
// value is greater than or equal to simple majority, since none of the equally
|
|
|
|
// maximal values would occur quorum or more number of times.
|
2016-11-21 10:47:26 +01:00
|
|
|
func reduceErrs(errs []error, ignoredErrs []error) (maxCount int, maxErr error) {
|
2016-07-13 20:56:25 +02:00
|
|
|
errorCounts := make(map[error]int)
|
2016-07-09 22:01:32 +02:00
|
|
|
for _, err := range errs {
|
2018-04-10 18:36:37 +02:00
|
|
|
if IsErrIgnored(err, ignoredErrs...) {
|
2016-07-20 04:24:32 +02:00
|
|
|
continue
|
|
|
|
}
|
2016-07-13 20:56:25 +02:00
|
|
|
errorCounts[err]++
|
2016-07-09 22:01:32 +02:00
|
|
|
}
|
2017-03-22 18:15:16 +01:00
|
|
|
|
2016-07-09 22:01:32 +02:00
|
|
|
max := 0
|
2016-07-13 20:56:25 +02:00
|
|
|
for err, count := range errorCounts {
|
2017-03-22 18:15:16 +01:00
|
|
|
switch {
|
|
|
|
case max < count:
|
2016-07-13 20:56:25 +02:00
|
|
|
max = count
|
2016-11-21 10:47:26 +01:00
|
|
|
maxErr = err
|
2017-03-22 18:15:16 +01:00
|
|
|
|
|
|
|
// Prefer `nil` over other error values with the same
|
|
|
|
// number of occurrences.
|
|
|
|
case max == count && err == nil:
|
|
|
|
maxErr = err
|
2016-07-09 22:01:32 +02:00
|
|
|
}
|
|
|
|
}
|
2016-11-21 10:47:26 +01:00
|
|
|
return max, maxErr
|
|
|
|
}
|
|
|
|
|
|
|
|
// reduceQuorumErrs behaves like reduceErrs by only for returning
|
|
|
|
// values of maximally occurring errors validated against a generic
|
2017-08-15 03:09:29 +02:00
|
|
|
// quorum number that can be read or write quorum depending on usage.
|
2018-04-06 00:04:40 +02:00
|
|
|
func reduceQuorumErrs(ctx context.Context, errs []error, ignoredErrs []error, quorum int, quorumErr error) error {
|
|
|
|
maxCount, maxErr := reduceErrs(errs, ignoredErrs)
|
|
|
|
if maxCount >= quorum {
|
|
|
|
return maxErr
|
2016-11-21 10:47:26 +01:00
|
|
|
}
|
2018-04-06 00:04:40 +02:00
|
|
|
return quorumErr
|
2016-11-21 10:47:26 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
// reduceReadQuorumErrs behaves like reduceErrs but only for returning
|
|
|
|
// values of maximally occurring errors validated against readQuorum.
|
2018-04-06 00:04:40 +02:00
|
|
|
func reduceReadQuorumErrs(ctx context.Context, errs []error, ignoredErrs []error, readQuorum int) (maxErr error) {
|
|
|
|
return reduceQuorumErrs(ctx, errs, ignoredErrs, readQuorum, errXLReadQuorum)
|
2016-11-21 10:47:26 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
// reduceWriteQuorumErrs behaves like reduceErrs but only for returning
|
|
|
|
// values of maximally occurring errors validated against writeQuorum.
|
2018-04-06 00:04:40 +02:00
|
|
|
func reduceWriteQuorumErrs(ctx context.Context, errs []error, ignoredErrs []error, writeQuorum int) (maxErr error) {
|
|
|
|
return reduceQuorumErrs(ctx, errs, ignoredErrs, writeQuorum, errXLWriteQuorum)
|
2016-07-09 22:01:32 +02:00
|
|
|
}
|
|
|
|
|
2017-08-15 03:09:29 +02:00
|
|
|
// Similar to 'len(slice)' but returns the actual elements count
|
2016-06-17 20:57:51 +02:00
|
|
|
// skipping the unallocated elements.
|
|
|
|
func diskCount(disks []StorageAPI) int {
|
|
|
|
diskCount := 0
|
|
|
|
for _, disk := range disks {
|
|
|
|
if disk == nil {
|
|
|
|
continue
|
|
|
|
}
|
|
|
|
diskCount++
|
|
|
|
}
|
|
|
|
return diskCount
|
|
|
|
}
|
|
|
|
|
2017-08-15 03:09:29 +02:00
|
|
|
// hashOrder - hashes input key to return consistent
|
2016-07-22 04:07:00 +02:00
|
|
|
// hashed integer slice. Returned integer order is salted
|
|
|
|
// with an input key. This results in consistent order.
|
|
|
|
// NOTE: collisions are fine, we are not looking for uniqueness
|
|
|
|
// in the slices returned.
|
|
|
|
func hashOrder(key string, cardinality int) []int {
|
2018-02-16 02:45:57 +01:00
|
|
|
if cardinality <= 0 {
|
|
|
|
// Returns an empty int slice for cardinality < 0.
|
2016-07-22 04:07:00 +02:00
|
|
|
return nil
|
2016-05-26 01:42:31 +02:00
|
|
|
}
|
2018-02-16 02:45:57 +01:00
|
|
|
|
2016-07-22 04:07:00 +02:00
|
|
|
nums := make([]int, cardinality)
|
|
|
|
keyCrc := crc32.Checksum([]byte(key), crc32.IEEETable)
|
2016-07-08 16:33:21 +02:00
|
|
|
|
2017-11-30 21:57:03 +01:00
|
|
|
start := int(keyCrc % uint32(cardinality))
|
2016-07-22 04:07:00 +02:00
|
|
|
for i := 1; i <= cardinality; i++ {
|
|
|
|
nums[i-1] = 1 + ((start + i) % cardinality)
|
2016-05-26 01:42:31 +02:00
|
|
|
}
|
2016-07-08 16:33:21 +02:00
|
|
|
return nums
|
2016-05-26 01:42:31 +02:00
|
|
|
}
|
|
|
|
|
2019-09-06 00:51:27 +02:00
|
|
|
// Constructs xlMetaV1 using `jsoniter` lib.
|
2019-08-19 20:35:52 +02:00
|
|
|
func xlMetaV1UnmarshalJSON(ctx context.Context, xlMetaBuf []byte) (xlMeta xlMetaV1, err error) {
|
2019-09-06 00:51:27 +02:00
|
|
|
err = json.Unmarshal(xlMetaBuf, &xlMeta)
|
|
|
|
return xlMeta, err
|
2016-09-09 07:38:18 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
// read xl.json from the given disk, parse and return xlV1MetaV1.Parts.
|
2019-01-05 23:16:43 +01:00
|
|
|
func readXLMetaParts(ctx context.Context, disk StorageAPI, bucket string, object string) ([]ObjectPartInfo, map[string]string, error) {
|
2016-09-09 07:38:18 +02:00
|
|
|
// Reads entire `xl.json`.
|
|
|
|
xlMetaBuf, err := disk.ReadAll(bucket, path.Join(object, xlMetaJSONFile))
|
|
|
|
if err != nil {
|
2018-04-06 00:04:40 +02:00
|
|
|
logger.LogIf(ctx, err)
|
|
|
|
return nil, nil, err
|
2016-09-09 07:38:18 +02:00
|
|
|
}
|
2018-03-01 20:37:57 +01:00
|
|
|
|
2019-08-19 20:35:52 +02:00
|
|
|
var xlMeta xlMetaV1
|
|
|
|
xlMeta, err = xlMetaV1UnmarshalJSON(ctx, xlMetaBuf)
|
|
|
|
if err != nil {
|
|
|
|
return nil, nil, err
|
|
|
|
}
|
2016-09-09 07:38:18 +02:00
|
|
|
|
2019-08-19 20:35:52 +02:00
|
|
|
return xlMeta.Parts, xlMeta.Meta, nil
|
2016-09-09 07:38:18 +02:00
|
|
|
}
|
|
|
|
|
2019-09-06 00:51:27 +02:00
|
|
|
// read xl.json from the given disk and parse xlV1Meta.Stat and xlV1Meta.Meta using jsoniter.
|
2019-08-19 20:35:52 +02:00
|
|
|
func readXLMetaStat(ctx context.Context, disk StorageAPI, bucket string, object string) (si statInfo,
|
|
|
|
mp map[string]string, e error) {
|
2016-09-09 07:38:18 +02:00
|
|
|
// Reads entire `xl.json`.
|
|
|
|
xlMetaBuf, err := disk.ReadAll(bucket, path.Join(object, xlMetaJSONFile))
|
|
|
|
if err != nil {
|
2018-04-06 00:04:40 +02:00
|
|
|
logger.LogIf(ctx, err)
|
|
|
|
return si, nil, err
|
2016-09-09 07:38:18 +02:00
|
|
|
}
|
2017-05-14 21:05:51 +02:00
|
|
|
|
2019-08-19 20:35:52 +02:00
|
|
|
var xlMeta xlMetaV1
|
|
|
|
xlMeta, err = xlMetaV1UnmarshalJSON(ctx, xlMetaBuf)
|
2016-09-09 07:38:18 +02:00
|
|
|
if err != nil {
|
2019-08-19 20:35:52 +02:00
|
|
|
return si, mp, err
|
2016-09-09 07:38:18 +02:00
|
|
|
}
|
2017-05-14 21:05:51 +02:00
|
|
|
|
2016-09-09 07:38:18 +02:00
|
|
|
// Return structured `xl.json`.
|
2019-08-19 20:35:52 +02:00
|
|
|
return xlMeta.Stat, xlMeta.Meta, nil
|
2016-09-09 07:38:18 +02:00
|
|
|
}
|
|
|
|
|
2016-06-25 23:51:06 +02:00
|
|
|
// readXLMeta reads `xl.json` and returns back XL metadata structure.
|
2018-04-06 00:04:40 +02:00
|
|
|
func readXLMeta(ctx context.Context, disk StorageAPI, bucket string, object string) (xlMeta xlMetaV1, err error) {
|
2016-06-24 11:06:23 +02:00
|
|
|
// Reads entire `xl.json`.
|
2016-09-09 07:38:18 +02:00
|
|
|
xlMetaBuf, err := disk.ReadAll(bucket, path.Join(object, xlMetaJSONFile))
|
2016-06-25 23:51:06 +02:00
|
|
|
if err != nil {
|
2018-10-31 00:07:57 +01:00
|
|
|
if err != errFileNotFound && err != errVolumeNotFound {
|
2018-09-14 06:42:50 +02:00
|
|
|
logger.GetReqInfo(ctx).AppendTags("disk", disk.String())
|
2018-04-25 20:46:49 +02:00
|
|
|
logger.LogIf(ctx, err)
|
|
|
|
}
|
2018-04-06 00:04:40 +02:00
|
|
|
return xlMetaV1{}, err
|
2016-06-03 01:34:15 +02:00
|
|
|
}
|
2018-11-13 20:57:03 +01:00
|
|
|
if len(xlMetaBuf) == 0 {
|
|
|
|
return xlMetaV1{}, errFileNotFound
|
|
|
|
}
|
2019-09-11 19:22:12 +02:00
|
|
|
return xlMetaV1UnmarshalJSON(ctx, xlMetaBuf)
|
2016-05-26 01:42:31 +02:00
|
|
|
}
|
2016-07-13 03:23:40 +02:00
|
|
|
|
2016-07-26 20:34:48 +02:00
|
|
|
// Reads all `xl.json` metadata as a xlMetaV1 slice.
|
|
|
|
// Returns error slice indicating the failed metadata reads.
|
2018-04-06 00:04:40 +02:00
|
|
|
func readAllXLMetadata(ctx context.Context, disks []StorageAPI, bucket, object string) ([]xlMetaV1, []error) {
|
2016-07-26 20:34:48 +02:00
|
|
|
metadataArray := make([]xlMetaV1, len(disks))
|
2019-10-14 18:44:51 +02:00
|
|
|
|
|
|
|
g := errgroup.WithNErrs(len(disks))
|
2016-07-26 20:34:48 +02:00
|
|
|
// Read `xl.json` parallelly across disks.
|
2019-10-14 18:44:51 +02:00
|
|
|
for index := range disks {
|
|
|
|
index := index
|
|
|
|
g.Go(func() (err error) {
|
|
|
|
if disks[index] == nil {
|
|
|
|
return errDiskNotFound
|
|
|
|
}
|
|
|
|
metadataArray[index], err = readXLMeta(ctx, disks[index], bucket, object)
|
|
|
|
return err
|
|
|
|
}, index)
|
2016-07-26 20:34:48 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
// Return all the metadata.
|
2019-10-14 18:44:51 +02:00
|
|
|
return metadataArray, g.Wait()
|
2016-07-26 20:34:48 +02:00
|
|
|
}
|
|
|
|
|
2017-02-24 18:20:40 +01:00
|
|
|
// Return shuffled partsMetadata depending on distribution.
|
|
|
|
func shufflePartsMetadata(partsMetadata []xlMetaV1, distribution []int) (shuffledPartsMetadata []xlMetaV1) {
|
|
|
|
if distribution == nil {
|
|
|
|
return partsMetadata
|
|
|
|
}
|
|
|
|
shuffledPartsMetadata = make([]xlMetaV1, len(partsMetadata))
|
|
|
|
// Shuffle slice xl metadata for expected distribution.
|
2016-07-16 17:35:30 +02:00
|
|
|
for index := range partsMetadata {
|
|
|
|
blockIndex := distribution[index]
|
2017-02-24 18:20:40 +01:00
|
|
|
shuffledPartsMetadata[blockIndex-1] = partsMetadata[index]
|
2016-07-16 17:35:30 +02:00
|
|
|
}
|
2017-02-24 18:20:40 +01:00
|
|
|
return shuffledPartsMetadata
|
2016-07-16 17:35:30 +02:00
|
|
|
}
|
|
|
|
|
2017-02-24 18:20:40 +01:00
|
|
|
// shuffleDisks - shuffle input disks slice depending on the
|
2017-08-15 03:09:29 +02:00
|
|
|
// erasure distribution. Return shuffled slice of disks with
|
2017-02-24 18:20:40 +01:00
|
|
|
// their expected distribution.
|
|
|
|
func shuffleDisks(disks []StorageAPI, distribution []int) (shuffledDisks []StorageAPI) {
|
|
|
|
if distribution == nil {
|
|
|
|
return disks
|
|
|
|
}
|
|
|
|
shuffledDisks = make([]StorageAPI, len(disks))
|
|
|
|
// Shuffle disks for expected distribution.
|
2016-07-16 17:35:30 +02:00
|
|
|
for index := range disks {
|
|
|
|
blockIndex := distribution[index]
|
2017-02-24 18:20:40 +01:00
|
|
|
shuffledDisks[blockIndex-1] = disks[index]
|
2016-07-16 17:35:30 +02:00
|
|
|
}
|
2017-02-24 18:20:40 +01:00
|
|
|
return shuffledDisks
|
2016-07-16 17:35:30 +02:00
|
|
|
}
|
2017-01-31 00:44:42 +01:00
|
|
|
|
2017-06-15 02:14:27 +02:00
|
|
|
// evalDisks - returns a new slice of disks where nil is set if
|
2017-08-15 03:09:29 +02:00
|
|
|
// the corresponding error in errs slice is not nil
|
2017-06-15 02:14:27 +02:00
|
|
|
func evalDisks(disks []StorageAPI, errs []error) []StorageAPI {
|
|
|
|
if len(errs) != len(disks) {
|
2018-04-06 00:04:40 +02:00
|
|
|
logger.LogIf(context.Background(), errors.New("unexpected disks/errors slice length"))
|
2017-06-15 02:14:27 +02:00
|
|
|
return nil
|
|
|
|
}
|
|
|
|
newDisks := make([]StorageAPI, len(disks))
|
|
|
|
for index := range errs {
|
|
|
|
if errs[index] == nil {
|
|
|
|
newDisks[index] = disks[index]
|
|
|
|
} else {
|
|
|
|
newDisks[index] = nil
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return newDisks
|
|
|
|
}
|
|
|
|
|
2017-10-06 18:38:01 +02:00
|
|
|
// Errors specifically generated by calculatePartSizeFromIdx function.
|
2017-02-01 00:34:49 +01:00
|
|
|
var (
|
|
|
|
errPartSizeZero = errors.New("Part size cannot be zero")
|
|
|
|
errPartSizeIndex = errors.New("Part index cannot be smaller than 1")
|
|
|
|
)
|
|
|
|
|
2017-10-06 18:38:01 +02:00
|
|
|
// calculatePartSizeFromIdx calculates the part size according to input index.
|
|
|
|
// returns error if totalSize is -1, partSize is 0, partIndex is 0.
|
2018-04-06 00:04:40 +02:00
|
|
|
func calculatePartSizeFromIdx(ctx context.Context, totalSize int64, partSize int64, partIndex int) (currPartSize int64, err error) {
|
2018-09-28 05:36:17 +02:00
|
|
|
if totalSize < -1 {
|
2018-04-06 00:04:40 +02:00
|
|
|
logger.LogIf(ctx, errInvalidArgument)
|
|
|
|
return 0, errInvalidArgument
|
2017-10-06 18:38:01 +02:00
|
|
|
}
|
2017-01-31 00:44:42 +01:00
|
|
|
if partSize == 0 {
|
2018-04-06 00:04:40 +02:00
|
|
|
logger.LogIf(ctx, errPartSizeZero)
|
|
|
|
return 0, errPartSizeZero
|
2017-01-31 00:44:42 +01:00
|
|
|
}
|
|
|
|
if partIndex < 1 {
|
2018-04-06 00:04:40 +02:00
|
|
|
logger.LogIf(ctx, errPartSizeIndex)
|
|
|
|
return 0, errPartSizeIndex
|
2017-01-31 00:44:42 +01:00
|
|
|
}
|
2019-05-01 01:27:31 +02:00
|
|
|
if totalSize == -1 {
|
|
|
|
return -1, nil
|
|
|
|
}
|
2017-10-06 18:38:01 +02:00
|
|
|
if totalSize > 0 {
|
|
|
|
// Compute the total count of parts
|
|
|
|
partsCount := totalSize/partSize + 1
|
|
|
|
// Return the part's size
|
|
|
|
switch {
|
|
|
|
case int64(partIndex) < partsCount:
|
|
|
|
currPartSize = partSize
|
|
|
|
case int64(partIndex) == partsCount:
|
|
|
|
// Size of last part
|
|
|
|
currPartSize = totalSize % partSize
|
|
|
|
default:
|
|
|
|
currPartSize = 0
|
|
|
|
}
|
2017-01-31 00:44:42 +01:00
|
|
|
}
|
2017-10-06 18:38:01 +02:00
|
|
|
return currPartSize, nil
|
2017-01-31 00:44:42 +01:00
|
|
|
}
|