Repository navigation
Expand file tree
/
Copy pathdisk-cache.go
More file actions
608 lines (544 loc) · 19.1 KB
/
Copy pathdisk-cache.go
File metadata and controls
608 lines (544 loc) · 19.1 KB
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
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
/*
* MinIO Cloud Storage, (C) 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"
"fmt"
"io"
"io/ioutil"
"net/http"
"os"
"strings"
"sync"
"time"
"github.com/djherbis/atime"
"github.com/minio/minio/cmd/config/cache"
"github.com/minio/minio/cmd/logger"
"github.com/minio/minio/pkg/color"
"github.com/minio/minio/pkg/sync/errgroup"
"github.com/minio/minio/pkg/wildcard"
)
const (
cacheBlkSize = int64(1 * 1024 * 1024)
)
// CacheStorageInfo - represents total, free capacity of
// underlying cache storage.
type CacheStorageInfo struct {
Total uint64 // Total cache disk space.
Free uint64 // Free cache available space.
}
// CacheObjectLayer implements primitives for cache object API layer.
type CacheObjectLayer interface {
// Object operations.
GetObjectNInfo(ctx context.Context, bucket, object string, rs *HTTPRangeSpec, h http.Header, lockType LockType, opts ObjectOptions) (gr *GetObjectReader, err error)
GetObjectInfo(ctx context.Context, bucket, object string, opts ObjectOptions) (objInfo ObjectInfo, err error)
DeleteObject(ctx context.Context, bucket, object string) error
DeleteObjects(ctx context.Context, bucket string, objects []string) ([]error, error)
PutObject(ctx context.Context, bucket, object string, data *PutObjReader, opts ObjectOptions) (objInfo ObjectInfo, err error)
// Storage operations.
StorageInfo(ctx context.Context) CacheStorageInfo
}
// Abstracts disk caching - used by the S3 layer
type cacheObjects struct {
// slice of cache drives
cache []*diskCache
// file path patterns to exclude from cache
exclude []string
// if true migration is in progress from v1 to v2
migrating bool
// mutex to protect migration bool
migMutex sync.Mutex
// nsMutex namespace lock
nsMutex *nsLockMap
// Object functions pointing to the corresponding functions of backend implementation.
NewNSLockFn func(ctx context.Context, bucket, object string) RWLocker
GetObjectNInfoFn func(ctx context.Context, bucket, object string, rs *HTTPRangeSpec, h http.Header, lockType LockType, opts ObjectOptions) (gr *GetObjectReader, err error)
GetObjectInfoFn func(ctx context.Context, bucket, object string, opts ObjectOptions) (objInfo ObjectInfo, err error)
DeleteObjectFn func(ctx context.Context, bucket, object string) error
DeleteObjectsFn func(ctx context.Context, bucket string, objects []string) ([]error, error)
PutObjectFn func(ctx context.Context, bucket, object string, data *PutObjReader, opts ObjectOptions) (objInfo ObjectInfo, err error)
}
func (c *cacheObjects) delete(ctx context.Context, dcache *diskCache, bucket, object string) (err error) {
cLock := c.NewNSLockFn(ctx, bucket, object)
if err := cLock.GetLock(globalObjectTimeout); err != nil {
return err
}
defer cLock.Unlock()
return dcache.Delete(ctx, bucket, object)
}
func (c *cacheObjects) put(ctx context.Context, dcache *diskCache, bucket, object string, data io.Reader, size int64, opts ObjectOptions) error {
cLock := c.NewNSLockFn(ctx, bucket, object)
if err := cLock.GetLock(globalObjectTimeout); err != nil {
return err
}
defer cLock.Unlock()
return dcache.Put(ctx, bucket, object, data, size, opts)
}
func (c *cacheObjects) get(ctx context.Context, dcache *diskCache, bucket, object string, rs *HTTPRangeSpec, h http.Header, opts ObjectOptions) (gr *GetObjectReader, err error) {
cLock := c.NewNSLockFn(ctx, bucket, object)
if err := cLock.GetRLock(globalObjectTimeout); err != nil {
return nil, err
}
defer cLock.RUnlock()
return dcache.Get(ctx, bucket, object, rs, h, opts)
}
func (c *cacheObjects) stat(ctx context.Context, dcache *diskCache, bucket, object string) (oi ObjectInfo, err error) {
cLock := c.NewNSLockFn(ctx, bucket, object)
if err := cLock.GetRLock(globalObjectTimeout); err != nil {
return oi, err
}
defer cLock.RUnlock()
return dcache.Stat(ctx, bucket, object)
}
// DeleteObject clears cache entry if backend delete operation succeeds
func (c *cacheObjects) DeleteObject(ctx context.Context, bucket, object string) (err error) {
if err = c.DeleteObjectFn(ctx, bucket, object); err != nil {
return
}
if c.isCacheExclude(bucket, object) || c.skipCache() {
return
}
dcache, cerr := c.getCacheLoc(bucket, object)
if cerr != nil {
return
}
if dcache.Exists(ctx, bucket, object) {
c.delete(ctx, dcache, bucket, object)
}
return
}
// DeleteObjects batch deletes objects in slice, and clears any cached entries
func (c *cacheObjects) DeleteObjects(ctx context.Context, bucket string, objects []string) ([]error, error) {
errs := make([]error, len(objects))
for idx, object := range objects {
errs[idx] = c.DeleteObject(ctx, bucket, object)
}
return errs, nil
}
// construct a metadata k-v map
func getMetadata(objInfo ObjectInfo) map[string]string {
metadata := make(map[string]string)
metadata["etag"] = objInfo.ETag
metadata["content-type"] = objInfo.ContentType
if objInfo.ContentEncoding != "" {
metadata["content-encoding"] = objInfo.ContentEncoding
}
if objInfo.Expires != timeSentinel {
metadata["expires"] = objInfo.Expires.Format(http.TimeFormat)
}
for k, v := range objInfo.UserDefined {
metadata[k] = v
}
return metadata
}
func (c *cacheObjects) GetObjectNInfo(ctx context.Context, bucket, object string, rs *HTTPRangeSpec, h http.Header, lockType LockType, opts ObjectOptions) (gr *GetObjectReader, err error) {
if c.isCacheExclude(bucket, object) || c.skipCache() {
return c.GetObjectNInfoFn(ctx, bucket, object, rs, h, lockType, opts)
}
var cc cacheControl
// fetch diskCache if object is currently cached or nearest available cache drive
dcache, err := c.getCacheToLoc(ctx, bucket, object)
if err != nil {
return c.GetObjectNInfoFn(ctx, bucket, object, rs, h, lockType, opts)
}
cacheReader, cacheErr := c.get(ctx, dcache, bucket, object, rs, h, opts)
if cacheErr == nil {
cc = cacheControlOpts(cacheReader.ObjInfo)
if !cc.isEmpty() && !cc.isStale(cacheReader.ObjInfo.ModTime) {
return cacheReader, nil
}
}
objInfo, err := c.GetObjectInfoFn(ctx, bucket, object, opts)
if backendDownError(err) && cacheErr == nil {
return cacheReader, nil
} else if err != nil {
if _, ok := err.(ObjectNotFound); ok {
if cacheErr == nil {
cacheReader.Close()
// Delete cached entry if backend object
// was deleted.
dcache.Delete(ctx, bucket, object)
}
}
return nil, err
}
if !objInfo.IsCacheable() {
return c.GetObjectNInfoFn(ctx, bucket, object, rs, h, lockType, opts)
}
if cacheErr == nil {
// if ETag matches for stale cache entry, serve from cache
if cacheReader.ObjInfo.ETag == objInfo.ETag {
// Update metadata in case server-side copy might have changed object metadata
dcache.updateMetadataIfChanged(ctx, bucket, object, objInfo, cacheReader.ObjInfo)
return cacheReader, nil
}
cacheReader.Close()
// Object is stale, so delete from cache
c.delete(ctx, dcache, bucket, object)
}
// Since we got here, we are serving the request from backend,
// and also adding the object to the cache.
if !dcache.diskUsageLow() {
select {
case dcache.purgeChan <- struct{}{}:
default:
}
}
if !dcache.diskAvailable(objInfo.Size) {
return c.GetObjectNInfoFn(ctx, bucket, object, rs, h, lockType, opts)
}
if rs != nil {
go func() {
// fill cache in the background for range GET requests
bReader, bErr := c.GetObjectNInfoFn(ctx, bucket, object, nil, h, lockType, opts)
if bErr != nil {
return
}
defer bReader.Close()
oi, err := c.stat(ctx, dcache, bucket, object)
// avoid cache overwrite if another background routine filled cache
if err != nil || oi.ETag != bReader.ObjInfo.ETag {
c.put(ctx, dcache, bucket, object, bReader, bReader.ObjInfo.Size, ObjectOptions{UserDefined: getMetadata(bReader.ObjInfo)})
}
}()
return c.GetObjectN
8000
InfoFn(ctx, bucket, object, rs, h, lockType, opts)
}
bkReader, bkErr := c.GetObjectNInfoFn(ctx, bucket, object, rs, h, lockType, opts)
if bkErr != nil {
return nil, bkErr
}
// Initialize pipe.
pipeReader, pipeWriter := io.Pipe()
teeReader := io.TeeReader(bkReader, pipeWriter)
go func() {
putErr := c.put(ctx, dcache, bucket, object, io.LimitReader(pipeReader, bkReader.ObjInfo.Size), bkReader.ObjInfo.Size, ObjectOptions{UserDefined: getMetadata(bkReader.ObjInfo)})
// close the write end of the pipe, so the error gets
// propagated to getObjReader
pipeWriter.CloseWithError(putErr)
}()
cleanupBackend := func() { bkReader.Close() }
cleanupPipe := func() { pipeWriter.Close() }
return NewGetObjectReaderFromReader(teeReader, bkReader.ObjInfo, opts.CheckCopyPrecondFn, cleanupBackend, cleanupPipe)
}
// Returns ObjectInfo from cache if available.
func (c *cacheObjects) GetObjectInfo(ctx context.Context, bucket, object string, opts ObjectOptions) (ObjectInfo, error) {
getObjectInfoFn := c.GetObjectInfoFn
if c.isCacheExclude(bucket, object) || c.skipCache() {
return getObjectInfoFn(ctx, bucket, object, opts)
}
// fetch diskCache if object is currently cached or nearest available cache drive
dcache, err := c.getCacheToLoc(ctx, bucket, object)
if err != nil {
return getObjectInfoFn(ctx, bucket, object, opts)
}
var cc cacheControl
// if cache control setting is valid, avoid HEAD operation to backend
cachedObjInfo, cerr := c.stat(ctx, dcache, bucket, object)
if cerr == nil {
cc = cacheControlOpts(cachedObjInfo)
if !cc.isEmpty() && !cc.isStale(cachedObjInfo.ModTime) {
return cachedObjInfo, nil
}
}
objInfo, err := getObjectInfoFn(ctx, bucket, object, opts)
if err != nil {
if _, ok := err.(ObjectNotFound); ok {
// Delete the cached entry if backend object was deleted.
c.delete(ctx, dcache, bucket, object)
return ObjectInfo{}, err
}
if !backendDownError(err) {
return ObjectInfo{}, err
}
if cerr == nil {
return cachedObjInfo, nil
}
return ObjectInfo{}, BackendDown{}
}
// when backend is up, do a sanity check on cached object
if cerr != nil {
return objInfo, nil
}
if cachedObjInfo.ETag != objInfo.ETag {
// Delete the cached entry if the backend object was replaced.
c.delete(ctx, dcache, bucket, object)
}
return objInfo, nil
}
// StorageInfo - returns underlying storage statistics.
func (c *cacheObjects) StorageInfo(ctx context.Context) (cInfo CacheStorageInfo) {
var total, free uint64
for _, cache := range c.cache {
if cache == nil {
continue
}
info, err := getDiskInfo(cache.dir)
logger.GetReqInfo(ctx).AppendTags("cachePath", cache.dir)
logger.LogIf(ctx, err)
total += info.Total
free += info.Free
}
return CacheStorageInfo{
Total: total,
Free: free,
}
}
// skipCache() returns true if cache migration is in progress
func (c *cacheObjects) skipCache() bool {
c.migMutex.Lock()
defer c.migMutex.Unlock()
return c.migrating
}
// Returns true if object should be excluded from cache
func (c *cacheObjects) isCacheExclude(bucket, object string) bool {
// exclude directories from cache
if strings.HasSuffix(object, SlashSeparator) {
return true
}
for _, pattern := range c.exclude {
matchStr := fmt.Sprintf("%s/%s", bucket, object)
if ok := wildcard.MatchSimple(pattern, matchStr); ok {
return true
}
}
return false
}
// choose a cache deterministically based on hash of bucket,object. The hash index is treated as
// a hint. In the event that the cache drive at hash index is offline, treat the list of cache drives
// as a circular buffer and walk through them starting at hash index until an online drive is found.
func (c *cacheObjects) getCacheLoc(bucket, object string) (*diskCache, error) {
index := c.hashIndex(bucket, object)
numDisks := len(c.cache)
for k := 0; k < numDisks; k++ {
i := (index + k) % numDisks
if c.cache[i] == nil {
continue
}
if c.cache[i].IsOnline() {
return c.cache[i], nil
}
}
return nil, errDiskNotFound
}
// get cache disk where object is currently cached for a GET op
8000
eration. If object does not exist at that location,
// treat the list of cache drives as a circular buffer and walk through them starting at hash index
// until an online drive is found.If object is not found, fall back to the first online cache drive
// closest to the hash index, so that object can be re-cached.
func (c *cacheObjects) getCacheToLoc(ctx context.Context, bucket, object string) (*diskCache, error) {
index := c.hashIndex(bucket, object)
numDisks := len(c.cache)
// save first online cache disk closest to the hint index
var firstOnlineDisk *diskCache
for k := 0; k < numDisks; k++ {
i := (index + k) % numDisks
if c.cache[i] == nil {
continue
}
if c.cache[i].IsOnline() {
if firstOnlineDisk == nil {
firstOnlineDisk = c.cache[i]
}
if c.cache[i].Exists(ctx, bucket, object) {
return c.cache[i], nil
}
}
}
if firstOnlineDisk != nil {
return firstOnlineDisk, nil
}
return nil, errDiskNotFound
}
// Compute a unique hash sum for bucket and object
func (c *cacheObjects) hashIndex(bucket, object string) int {
return crcHashMod(pathJoin(bucket, object), len(c.cache))
}
// newCache initializes the cacheFSObjects for the "drives" specified in config.json
// or the global env overrides.
func newCache(config cache.Config) ([]*diskCache, bool, error) {
var caches []*diskCache
ctx := logger.SetReqInfo(context.Background(), &logger.ReqInfo{})
formats, migrating, err := loadAndValidateCacheFormat(ctx, config.Drives)
if err != nil {
return nil, false, err
}
for i, dir := range config.Drives {
// skip diskCache creation for cache drives missing a format.json
if formats[i] == nil {
caches = append(caches, nil)
continue
}
if err := checkAtimeSupport(dir); err != nil {
return nil, false, errors.New("Atime support required for disk caching")
}
quota := config.MaxUse
if quota == 0 {
quota = config.Quota
}
cache, err := newDiskCache(dir, config.Expiry, quota)
if err != nil {
return nil, false, err
}
// Start the purging go-routine for entries that have expired if no migration in progress
if !migrating {
go cache.purge()
}
caches = append(caches, cache)
}
return caches, migrating, nil
}
// Return error if Atime is disabled on the O/S
func checkAtimeSupport(dir string) (err error) {
file, err := ioutil.TempFile(dir, "prefix")
if err != nil {
return
}
defer os.Remove(file.Name())
finfo1, err := os.Stat(file.Name())
if err != nil {
return
}
// add a sleep to ensure atime change is detected
time.Sleep(10 * time.Millisecond)
if _, err = io.Copy(ioutil.Discard, file); err != nil {
return
}
finfo2, err := os.Stat(file.Name())
if atime.Get(finfo2).Equal(atime.Get(finfo1)) {
return errors.New("Atime not supported")
}
return
}
func (c *cacheObjects) migrateCacheFromV1toV2(ctx context.Context) {
logStartupMessage(color.Blue("Cache migration initiated ...."))
g := errgroup.WithNErrs(len(c.cache))
for index, dc := range c.cache {
if dc == nil {
continue
}
index := index
g.Go(func() error {
// start migration from V1 to V2
return migrateOldCache(ctx, c.cache[index])
}, index)
}
errCnt := 0
for index, err := range g.Wait() {
if err != nil {
errCnt++
logger.LogIf(ctx, err)
continue
}
go c.cache[index].purge()
}
if errCnt > 0 {
return
}
// update migration status
c.migMutex.Lock()
defer c.migMutex.Unlock()
c.migrating = false
logStartupMessage(color.Blue("Cache migration completed successfully."))
}
// PutObject - caches the uploaded object for single Put operations
func (c *cacheObjects) PutObject(ctx context.Context, bucket, object string, r *PutObjReader, opts ObjectOptions) (objInfo ObjectInfo, err error) {
putObjectFn := c.PutObjectFn
dcache, err := c.getCacheToLoc(ctx, bucket, object)
if err != nil {
// disk cache could not be located,execute backend call.
return putObjectFn(ctx, bucket, object, r, opts)
}
size := r.Size()
if c.skipCache() {
return putObjectFn(ctx, bucket, object, r, opts)
}
// fetch from backend if there is no space on cache drive
if !dcache.diskAvailable(size) {
return putObjectFn(ctx, bucket, object, r, opts)
}
if opts.ServerSideEncryption != nil {
return putObjectFn(ctx, bucket, object, r, opts)
}
// skip cache for objects with locks
objRetention := getObjectRetentionMeta(opts.UserDefined)
if objRetention.Mode == Governance || objRetention.Mode == Compliance {
return putObjectFn(ctx, bucket, object, r, opts)
}
// fetch from backend if cache exclude pattern or cache-control
// directive set to exclude
if c.isCacheExclude(bucket, object) {
dcache.Delete(ctx, bucket, object)
return putObjectFn(ctx, bucket, object, r, opts)
}
objInfo, err = putObjectFn(ctx, bucket, object, r, opts)
if err == nil {
go func() {
// fill cache in the background
bReader, bErr := c.GetObjectNInfoFn(ctx, bucket, object, nil, http.Header{}, readLock, ObjectOptions{})
if bErr != nil {
return
}
defer bReader.Close()
oi, err := c.stat(ctx, dcache, bucket, object)
// avoid cache overwrite if another background routine filled cache
if err != nil || oi.ETag != bReader.ObjInfo.ETag {
c.put(ctx, dcache, bucket, object, bReader, bReader.ObjInfo.Size, ObjectOptions{UserDefined: getMetadata(bReader.ObjInfo)})
}
}()
}
return objInfo, err
}
// Returns cacheObjects for use by Server.
func newServerCacheObjects(ctx context.Context, config cache.Config) (CacheObjectLayer, error) {
// list of disk caches for cache "drives" specified in config.json or MINIO_CACHE_DRIVES env var.
cache, migrateSw, err := newCache(config)
if err != nil {
return nil, err
}
c := &cacheObjects{
cache: cache,
exclude: config.Exclude,
migrating: migrateSw,
migMutex: sync.Mutex{},
nsMutex: newNSLock(false),
GetObjectInfoFn: func(ctx context.Context, bucket, object string, opts ObjectOptions) (ObjectInfo, error) {
return newObjectLayerFn().GetObjectInfo(ctx, bucket, object, opts)
},
GetObjectNInfoFn: func(ctx context.Context, bucket, object string, rs *HTTPRangeSpec, h http.Header, lockType LockType, opts ObjectOptions) (gr *GetObjectReader, err error) {
return newObjectLayerFn().GetObjectNInfo(ctx, bucket, object, rs, h, lockType, opts)
},
DeleteObjectFn: func(ctx context.Context, bucket, object string) error {
return newObjectLayerFn().DeleteObject(ctx, bucket, object)
},
DeleteObjectsFn: func(ctx context.Context, bucket string, objects []string) ([]error, error) {
errs := make([]error, len(objects))
for idx, object := range objects {
errs[idx] = newObjectLayerFn().DeleteObject(ctx, bucket, object)
}
return errs, nil
},
PutObjectFn: func(ctx context.Context, bucket, object string, data *PutObjReader, opts ObjectOptions) (objInfo ObjectInfo, err error) {
return newObjectLayerFn().PutObject(ctx, bucket, object, data, opts)
},
}
c.NewNSLockFn = func(ctx context.Context, bucket, object string) RWLocker {
return c.nsMutex.NewNSLock(ctx, nil, bucket, object)
}
if migrateSw {
go c.migrateCacheFromV1toV2(ctx)
}
return c, nil
}