/
erasure-metadata-utils.go
286 lines (261 loc) · 9.06 KB
/
erasure-metadata-utils.go
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
/*
* MinIO Cloud Storage, (C) 2016-2019 MinIO, Inc.
*
* 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.
*/
package cmd
import (
"context"
"errors"
"hash/crc32"
"github.com/minio/minio/cmd/logger"
"github.com/minio/minio/pkg/sync/errgroup"
)
// Returns number of errors that occurred the most (incl. nil) and the
// corresponding error value. NB When there is more than one error value that
// 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.
func reduceErrs(errs []error, ignoredErrs []error) (maxCount int, maxErr error) {
errorCounts := make(map[error]int)
for _, err := range errs {
if IsErrIgnored(err, ignoredErrs...) {
continue
}
errorCounts[err]++
}
max := 0
for err, count := range errorCounts {
switch {
case max < count:
max = count
maxErr = err
// Prefer `nil` over other error values with the same
// number of occurrences.
case max == count && err == nil:
maxErr = err
}
}
return max, maxErr
}
// reduceQuorumErrs behaves like reduceErrs by only for returning
// values of maximally occurring errors validated against a generic
// quorum number that can be read or write quorum depending on usage.
func reduceQuorumErrs(ctx context.Context, errs []error, ignoredErrs []error, quorum int, quorumErr error) error {
maxCount, maxErr := reduceErrs(errs, ignoredErrs)
if maxCount >= quorum {
return maxErr
}
return quorumErr
}
// reduceReadQuorumErrs behaves like reduceErrs but only for returning
// values of maximally occurring errors validated against readQuorum.
func reduceReadQuorumErrs(ctx context.Context, errs []error, ignoredErrs []error, readQuorum int) (maxErr error) {
return reduceQuorumErrs(ctx, errs, ignoredErrs, readQuorum, errErasureReadQuorum)
}
// reduceWriteQuorumErrs behaves like reduceErrs but only for returning
// values of maximally occurring errors validated against writeQuorum.
func reduceWriteQuorumErrs(ctx context.Context, errs []error, ignoredErrs []error, writeQuorum int) (maxErr error) {
return reduceQuorumErrs(ctx, errs, ignoredErrs, writeQuorum, errErasureWriteQuorum)
}
// Similar to 'len(slice)' but returns the actual elements count
// skipping the unallocated elements.
func diskCount(disks []StorageAPI) int {
diskCount := 0
for _, disk := range disks {
if disk == nil {
continue
}
diskCount++
}
return diskCount
}
// hashOrder - hashes input key to return consistent
// 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 {
if cardinality <= 0 {
// Returns an empty int slice for cardinality < 0.
return nil
}
nums := make([]int, cardinality)
keyCrc := crc32.Checksum([]byte(key), crc32.IEEETable)
start := int(keyCrc % uint32(cardinality))
for i := 1; i <= cardinality; i++ {
nums[i-1] = 1 + ((start + i) % cardinality)
}
return nums
}
// Reads all `xl.meta` metadata as a FileInfo slice.
// Returns error slice indicating the failed metadata reads.
func readAllFileInfo(ctx context.Context, disks []StorageAPI, bucket, object, versionID string, readData bool) ([]FileInfo, []error) {
metadataArray := make([]FileInfo, len(disks))
g := errgroup.WithNErrs(len(disks))
// Read `xl.meta` in parallel across disks.
for index := range disks {
index := index
g.Go(func() (err error) {
if disks[index] == nil {
return errDiskNotFound
}
metadataArray[index], err = disks[index].ReadVersion(ctx, bucket, object, versionID, readData)
if err != nil {
if !IsErr(err, []error{
errFileNotFound,
errVolumeNotFound,
errFileVersionNotFound,
errDiskNotFound,
}...) {
logger.LogOnceIf(ctx, err, disks[index].String())
}
}
return err
}, index)
}
// Return all the metadata.
return metadataArray, g.Wait()
}
func shuffleDisksAndPartsMetadataByIndex(disks []StorageAPI, metaArr []FileInfo, fi FileInfo) (shuffledDisks []StorageAPI, shuffledPartsMetadata []FileInfo) {
shuffledDisks = make([]StorageAPI, len(disks))
shuffledPartsMetadata = make([]FileInfo, len(disks))
var inconsistent int
distribution := fi.Erasure.Distribution
for i, meta := range metaArr {
if disks[i] == nil {
// Assuming offline drives as inconsistent,
// to be safe and fallback to original
// distribution order.
inconsistent++
continue
}
// check if erasure distribution order matches the index
// position if this is not correct we discard the disk
// and move to collect others
if distribution[i] != meta.Erasure.Index {
inconsistent++ // keep track of inconsistent entries
continue
}
shuffledDisks[meta.Erasure.Index-1] = disks[i]
shuffledPartsMetadata[meta.Erasure.Index-1] = metaArr[i]
}
// Inconsistent meta info is with in the limit of
// expected quorum, proceed with EcIndex based
// disk order.
if inconsistent < fi.Erasure.ParityBlocks {
return shuffledDisks, shuffledPartsMetadata
}
// fall back to original distribution based order.
return shuffleDisksAndPartsMetadata(disks, metaArr, distribution)
}
// Return shuffled partsMetadata depending on distribution.
func shuffleDisksAndPartsMetadata(disks []StorageAPI, partsMetadata []FileInfo, distribution []int) (shuffledDisks []StorageAPI, shuffledPartsMetadata []FileInfo) {
if distribution == nil {
return disks, partsMetadata
}
shuffledDisks = make([]StorageAPI, len(disks))
shuffledPartsMetadata = make([]FileInfo, len(partsMetadata))
// Shuffle slice xl metadata for expected distribution.
for index := range partsMetadata {
blockIndex := distribution[index]
shuffledPartsMetadata[blockIndex-1] = partsMetadata[index]
shuffledDisks[blockIndex-1] = disks[index]
}
return shuffledDisks, shuffledPartsMetadata
}
// Return shuffled partsMetadata depending on distribution.
func shufflePartsMetadata(partsMetadata []FileInfo, distribution []int) (shuffledPartsMetadata []FileInfo) {
if distribution == nil {
return partsMetadata
}
shuffledPartsMetadata = make([]FileInfo, len(partsMetadata))
// Shuffle slice xl metadata for expected distribution.
for index := range partsMetadata {
blockIndex := distribution[index]
shuffledPartsMetadata[blockIndex-1] = partsMetadata[index]
}
return shuffledPartsMetadata
}
// shuffleDisks - shuffle input disks slice depending on the
// erasure distribution. Return shuffled slice of disks with
// 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.
for index := range disks {
blockIndex := distribution[index]
shuffledDisks[blockIndex-1] = disks[index]
}
return shuffledDisks
}
// evalDisks - returns a new slice of disks where nil is set if
// the corresponding error in errs slice is not nil
func evalDisks(disks []StorageAPI, errs []error) []StorageAPI {
if len(errs) != len(disks) {
logger.LogIf(GlobalContext, errors.New("unexpected disks/errors slice length"))
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
}
// Errors specifically generated by calculatePartSizeFromIdx function.
var (
errPartSizeZero = errors.New("Part size cannot be zero")
errPartSizeIndex = errors.New("Part index cannot be smaller than 1")
)
// calculatePartSizeFromIdx calculates the part size according to input index.
// returns error if totalSize is -1, partSize is 0, partIndex is 0.
func calculatePartSizeFromIdx(ctx context.Context, totalSize int64, partSize int64, partIndex int) (currPartSize int64, err error) {
if totalSize < -1 {
logger.LogIf(ctx, errInvalidArgument)
return 0, errInvalidArgument
}
if partSize == 0 {
logger.LogIf(ctx, errPartSizeZero)
return 0, errPartSizeZero
}
if partIndex < 1 {
logger.LogIf(ctx, errPartSizeIndex)
return 0, errPartSizeIndex
}
if totalSize == -1 {
return -1, nil
}
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
}
}
return currPartSize, nil
}