-
Notifications
You must be signed in to change notification settings - Fork 1
/
db.go
677 lines (609 loc) · 18.1 KB
/
db.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
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
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
// Copyright (c) 2018-2020 Blockwatch Data Inc.
// Author: alex@blockwatch.cc
package bolt
import (
"context"
"encoding/json"
"fmt"
"io"
"os"
"path/filepath"
"strings"
"sync"
"time"
bolt "go.etcd.io/bbolt"
"blockwatch.cc/knoxdb/store"
"blockwatch.cc/knoxdb/util"
)
// db wraps a Badger instance and implements the store.DB interface.
// All database access is performed through transactions which are managed.
type db struct {
seqLock sync.RWMutex // Guard access to sequences.
closeLock sync.RWMutex // Make database close block while txns active.
txLock sync.Mutex // block creating new tx during backup/restore.
activeTx sync.WaitGroup // count active tx (needed for backup/restore).
closed bool // Is the database closed?
store *bolt.DB // the database
opts *bolt.Options
dbPath string
}
// Enforce db implements the store.DB interface.
var _ store.DB = (*db)(nil)
// Type returns the database driver type the current database instance was
// created with.
//
// This function is part of the store.DB interface implementation.
func (db *db) Type() string {
return dbType
}
func (db *db) IsReadOnly() bool {
return db.store.IsReadOnly()
}
// IsZeroCopyRead returns true if keys and values on Get and from Cursors
// are only valid within the current transaction (or iterator step).
func (db *db) IsZeroCopyRead() bool {
return true
}
// Path returns the path where the current database is stored.
//
// This function is part of the store.DB interface implementation.
func (db *db) Path() string {
return db.dbPath
}
// Manifest returns the current database manifest metadata.
//
// This function is part of the store.DB interface implementation.
func (db *db) Manifest() (store.Manifest, error) {
db.closeLock.RLock()
defer db.closeLock.RUnlock()
if db.closed {
return store.Manifest{}, makeDbErr(store.ErrDbNotOpen, errDbNotOpenStr, nil)
}
return getManifest(db.store)
}
// SetManifest overwrites the current database manifest.
//
// This function is part of the store.DB interface implementation.
func (db *db) SetManifest(manifest store.Manifest) error {
db.closeLock.RLock()
defer db.closeLock.RUnlock()
if db.closed {
return makeDbErr(store.ErrDbNotOpen, errDbNotOpenStr, nil)
}
mft, err := getManifest(db.store)
if err != nil {
return err
}
// we only allow some fields to be overwritten
mft.Name = manifest.Name
mft.Version = manifest.Version
mft.Label = manifest.Label
mft.Schema = manifest.Schema
return putManifest(db.store, mft)
}
func getManifest(bdb *bolt.DB) (store.Manifest, error) {
var mft store.Manifest
err := bdb.View(func(dbTx *bolt.Tx) error {
mftBucket := dbTx.Bucket(manifestBucketKeyName)
if mftBucket == nil {
return makeDbErr(store.ErrInvalid, "invalid database: missing manifest", nil)
}
buf := mftBucket.Get(manifestKey)
if buf != nil {
return json.Unmarshal(buf, &mft)
}
return nil
})
if err != nil {
return mft, err
}
return mft, nil
}
func putManifest(bdb *bolt.DB, manifest store.Manifest) error {
buf, err := json.Marshal(manifest)
if err != nil {
return err
}
return bdb.Update(func(dbTx *bolt.Tx) error {
return dbTx.Bucket(manifestBucketKeyName).Put(manifestKey, buf)
})
}
// Sequence creates a new managed sequence stored in the sequences bucket.
func (db *db) Sequence(key []byte, lease uint64) (store.Sequence, error) {
return &sequence{
db: db,
key: copySlice(key),
}, nil
}
// begin is the implementation function for the Begin database method. See its
// documentation for more details.
//
// This function is only separate because it returns the internal transaction
// which is used by the managed transaction code while the database method
// returns the interface.
func (db *db) begin(writable bool) (*transaction, error) {
// Whenever a new transaction is started, grab a read lock against the
// database to ensure Close will wait for the transaction to finish.
// This lock will not be released until the transaction is closed (via
// Rollback or Commit).
db.closeLock.RLock()
if db.closed {
db.closeLock.RUnlock()
return nil, makeDbErr(store.ErrDbNotOpen, errDbNotOpenStr, nil)
}
// db.txLock.Lock()
// defer db.txLock.Unlock()
dbTx, err := db.store.Begin(writable)
if err != nil {
db.closeLock.RUnlock()
return nil, convertErr("begin tx", err)
}
db.activeTx.Add(1)
tx := &transaction{
writable: writable,
db: db,
tx: dbTx,
}
return tx, nil
}
// Begin starts a transaction which is either read-only or read-write depending
// on the specified flag. Multiple read-only transactions can be started
// simultaneously while only a single read-write transaction can be started at a
// time. The call will block when starting a read-write transaction when one is
// already open.
//
// NOTE: The transaction must be closed by calling Rollback or Commit on it when
// it is no longer needed. Failure to do so will result in unclaimed memory.
//
// This function is part of the store.DB interface implementation.
func (db *db) Begin(writable bool) (store.Tx, error) {
return db.begin(writable)
}
// rollbackOnPanic rolls the passed transaction back if the code in the calling
// function panics. This is needed since the mutex on a transaction must be
// released and a panic in called code would prevent that from happening.
//
// NOTE: This can only be handled manually for managed transactions since they
// control the life-cycle of the transaction. As the documentation on Begin
// calls out, callers opting to use manual transactions will have to ensure the
// transaction is rolled back on panic if it desires that functionality as well
// or the database will fail to close since the read-lock will never be
// released.
// func rollbackOnPanic(tx *transaction) {
// if err := recover(); err != nil {
// tx.managed = false
// _ = tx.Rollback()
// panic(err)
// }
// }
// View invokes the passed function in the context of a managed read-only
// transaction with the root bucket for the namespace. Any errors returned from
// the user-supplied function are returned from this function.
//
// This function is part of the store.DB interface implementation.
func (db *db) View(fn func(store.Tx) error) error {
// check for close and hold close lock
db.closeLock.RLock()
defer db.closeLock.RUnlock()
if db.closed {
return makeDbErr(store.ErrDbNotOpen, errDbNotOpenStr, nil)
}
// count active tx
db.activeTx.Add(1)
defer db.activeTx.Done()
var err error
dberr := db.store.View(func(tx *bolt.Tx) error {
dbtx := &transaction{
writable: false,
db: db,
tx: tx,
}
err = fn(dbtx)
return nil
})
if dberr != nil {
return convertErr("view tx", dberr)
}
return err
}
// Update invokes the passed function in the context of a managed read-write
// transaction with the root bucket for the namespace. Any errors returned from
// the user-supplied function will cause the transaction to be rolled back and
// are returned from this function. Otherwise, the transaction is committed
// when the user-supplied function returns a nil error.
//
// This function is part of the store.DB interface implementation.
func (db *db) Update(fn func(store.Tx) error) error {
// check for close and hold close lock
db.closeLock.RLock()
defer db.closeLock.RUnlock()
if db.closed {
return makeDbErr(store.ErrDbNotOpen, errDbNotOpenStr, nil)
}
// count active tx
db.activeTx.Add(1)
defer db.activeTx.Done()
var err error
dberr := db.store.Update(func(tx *bolt.Tx) error {
dbtx := &transaction{
writable: true,
db: db,
tx: tx,
}
err = fn(dbtx)
return nil
})
if dberr != nil {
return convertErr("update tx", dberr)
}
return err
}
// Close cleanly shuts down the database and syncs all data. It will block
// until all database transactions have been finalized (rolled back or
// committed).
//
// This function is part of the store.DB interface implementation.
func (db *db) Close() error {
// Since all transactions have a read lock on this mutex, this will
// cause Close to wait for all readers to complete.
db.closeLock.Lock()
defer db.closeLock.Unlock()
if db.closed {
return makeDbErr(store.ErrDbNotOpen, errDbNotOpenStr, nil)
}
return db.close()
}
func (db *db) close() error {
mft, err := getManifest(db.store)
if err != nil {
return err
}
if !db.store.IsReadOnly() {
// write manifest
mft.IsLocked = false
if err := putManifest(db.store, mft); err != nil {
return err
}
}
// NOTE: Since the above lock waits for all transactions to finish and
// prevents any new ones from being started, it is safe to close the
// underlying db here.
if err := db.store.Close(); err != nil {
return convertErr("close", err)
}
if mft.Name != "" {
log.Debugf("%s database closed successfully.", strings.Title(mft.Name))
} else {
log.Debugf("Database closed successfully.")
}
db.closed = true
db.store = nil
db.opts = nil
return nil
}
// filesExists reports whether the named file or directory exists.
func fileExists(name string) bool {
if _, err := os.Stat(name); err != nil {
if os.IsNotExist(err) {
return false
}
}
return true
}
// initDB creates the initial buckets and values used by the package. This is
// mainly in a separate function for testing purposes.
func initDB(db *bolt.DB) error {
// init manifest
now := time.Now().UTC()
mft := store.Manifest{
CreatedAt: now,
IsLocked: true,
}
buf, err := json.Marshal(mft)
if err != nil {
return err
}
// init sequences bucket
err = db.Update(func(dbTx *bolt.Tx) error {
mftBucket, err := dbTx.CreateBucketIfNotExists(manifestBucketKeyName)
if err != nil {
return err
}
return mftBucket.Put(manifestKey, buf)
})
if err != nil {
return convertErr("init db", err)
}
return nil
}
// openDB opens the database at the provided path. store.ErrDbDoesNotExist
// is returned if the database doesn't exist and the create flag is not set.
func openDB(dbPath string, opts *bolt.Options, create bool) (store.DB, error) {
dbExists := fileExists(dbPath)
if !create && !dbExists {
str := fmt.Sprintf("database file %q does not exist", dbPath)
return nil, makeDbErr(store.ErrDbDoesNotExist, str, nil)
}
if create && dbExists {
str := fmt.Sprintf("database file %q exists", dbPath)
return nil, makeDbErr(store.ErrDbExists, str, nil)
}
// Ensure the full path to the database exists.
if !dbExists {
// The error can be ignored here since the call to
// badger.Open will fail if the directory couldn't be
// created.
_ = os.MkdirAll(filepath.Dir(dbPath), 0700)
}
// bolt will create any non-existing database file automatically
bdb, err := bolt.Open(dbPath, 0600, opts)
if err != nil {
return nil, convertErr("open db", err)
}
if create && !dbExists {
log.Debug("Initializing database...")
if err := initDB(bdb); err != nil {
bdb.Close()
return nil, convertErr("init db", err)
}
} else {
// update manifest
mft, err := getManifest(bdb)
if err != nil {
bdb.Close()
return nil, err
}
if !bdb.IsReadOnly() {
mft.IsLocked = true
if err := putManifest(bdb, mft); err != nil {
bdb.Close()
return nil, err
}
}
if mft.Name != "" {
log.Debugf("%s database opened successfully.", strings.Title(mft.Name))
} else {
log.Debug("Database opened successfully.")
}
}
return &db{store: bdb, dbPath: dbPath, opts: opts}, nil
}
// Database maintenance functions
// Export all database contents as protobuf data.
func (db *db) Dump(w io.Writer) error {
db.closeLock.RLock()
defer db.closeLock.RUnlock()
if db.closed {
return makeDbErr(store.ErrDbNotOpen, errDbNotOpenStr, nil)
}
// backup may run in parallel to any tx and will be using a snapshot copy
err := db.store.View(func(dbTx *bolt.Tx) error {
log.Debugf("Exporting database of size %s (this may take a while)...",
util.ByteSize(dbTx.Size()).String())
n, err := dbTx.WriteTo(w)
if err != nil {
return err
}
log.Debugf("Successfully wrote %s of data.", util.ByteSize(n).String())
return nil
})
if err != nil {
return convertErr("dump db", err)
}
return nil
}
// Should be called on a database that is not running any other
// concurrent transactions while it is running.
func (db *db) Restore(r io.Reader) error {
// not implemented; to do so implement
// - close bolt db (waiting for any open tx)
// - restore/overwrite file with reader contents
// - open bolt db from restored file
return nil
}
// Should be called on a database that is not running any concurrent tx.
//
// Garbage collect database. This will create a new file, stream all keys into
// that file, replace the existing DB with the new file and reopen the DB.
func (db *db) GC(ctx context.Context, ratio float64) error {
// hold close lock
db.closeLock.RLock()
defer db.closeLock.RUnlock()
if db.closed {
return makeDbErr(store.ErrDbNotOpen, errDbNotOpenStr, nil)
}
// make wait interruptable
run := make(chan struct{})
go func() {
// wait for tx to finish
db.activeTx.Wait()
run <- struct{}{}
}()
select {
case <-ctx.Done():
return ctx.Err()
case <-run:
}
// prevent parallel GC
db.activeTx.Add(1)
defer db.activeTx.Done()
// init
start := time.Now()
srcPath := db.dbPath
dstPath := db.dbPath + ".temp"
fi, err := os.Stat(srcPath)
if err != nil {
return convertErr("cannot stat source db", err)
}
initialSize := fi.Size()
// Open destination database.
dstOpts := *db.opts
dstOpts.ReadOnly = false
dst, err := bolt.Open(dstPath, fi.Mode(), &dstOpts)
if err != nil {
return convertErr("cannot open compaction db", err)
}
defer func(dst *bolt.DB, dstPath string) {
if err != nil {
dst.Close()
os.Remove(dstPath)
}
}(dst, dstPath)
// Run compaction.
log.Infof("[GC] Compacting database %s (%s).", db.dbPath, util.ByteSize(initialSize))
if err = compact(ctx, dst, db.store, compactTxSize, ratio); err != nil {
return convertErr("compact db", err)
}
// Report stats on new size.
fi, err = os.Stat(dstPath)
if err != nil {
return convertErr("cannot stat destination db", err)
} else if fi.Size() == 0 {
err = fmt.Errorf("zero size after compaction")
return convertErr("compact db", err)
}
log.Infof("[GC] Database %s successfully compacted %s -> %s (gain=%.2fx) in %s.",
db.dbPath,
util.ByteSize(initialSize),
util.ByteSize(fi.Size()),
float64(initialSize)/float64(fi.Size()),
time.Since(start))
// replace db - point of no return
// also, don't overwrite err to avoid triggering defer
if err := dst.Close(); err != nil {
return convertErr("close after compact", err)
}
if err := db.store.Close(); err != nil {
return convertErr("close after compact", err)
}
db.closed = true
if err := os.Rename(srcPath, srcPath+".backup"); err != nil {
return convertErr("rename source db", err)
}
if err := os.Rename(dstPath, srcPath); err != nil {
return convertErr("rename compacted db", err)
}
db.store, err = bolt.Open(srcPath, 0600, db.opts)
if err != nil {
return convertErr("open compacted db", err)
}
// update manifest
if !db.store.IsReadOnly() {
mft, err := getManifest(db.store)
if err != nil {
db.store.Close()
return convertErr("get manifest after compact", err)
}
mft.IsLocked = true
if err := putManifest(db.store, mft); err != nil {
db.store.Close()
return convertErr("store manifest after compact", err)
}
}
log.Debugf("[GC] Database %s reopened successfully.", db.dbPath)
db.closed = false
// when all is good, remove the old database, ignoring errors
os.Remove(srcPath + ".backup")
log.Info("[GC] Using compacted database from now.")
return nil
}
func compact(ctx context.Context, dst, src *bolt.DB, txMaxSize int64, fillPercent float64) error {
// commit regularly, or we'll run out of memory for large datasets if using one transaction.
var size int64
tx, err := dst.Begin(true)
if err != nil {
return err
}
defer tx.Rollback()
if err := walk(src, func(keys [][]byte, k, v []byte, seq uint64) error {
// On each key/value, check if we have exceeded tx size.
sz := int64(len(k) + len(v))
if size+sz > txMaxSize && txMaxSize != 0 {
// Commit previous transaction.
if err := tx.Commit(); err != nil {
return err
}
// check context
select {
case <-ctx.Done():
return ctx.Err()
default:
}
// Start new transaction.
tx, err = dst.Begin(true)
if err != nil {
return err
}
size = 0
}
size += sz
// Create bucket on the root transaction if this is the first level.
nk := len(keys)
if nk == 0 {
bkt, err := tx.CreateBucket(k)
if err != nil {
return err
}
if err := bkt.SetSequence(seq); err != nil {
return err
}
return nil
}
// Create buckets on subsequent levels, if necessary.
b := tx.Bucket(keys[0])
if nk > 1 {
for _, k := range keys[1:] {
b = b.Bucket(k)
}
}
// Fill the entire page for best compaction.
b.FillPercent = fillPercent
// If there is no value then this is a bucket call.
if v == nil {
bkt, err := b.CreateBucket(k)
if err != nil {
return err
}
if err := bkt.SetSequence(seq); err != nil {
return err
}
return nil
}
// Otherwise treat it as a key/value pair.
return b.Put(k, v)
}); err != nil {
return err
}
return tx.Commit()
}
// walkFunc is the type of the function called for keys (buckets and "normal"
// values) discovered by Walk. keys is the list of keys to descend to the bucket
// owning the discovered key/value pair k/v.
type walkFunc func(keys [][]byte, k, v []byte, seq uint64) error
// walk walks recursively the bolt database db, calling walkFn for each key it finds.
func walk(db *bolt.DB, walkFn walkFunc) error {
return db.View(func(tx *bolt.Tx) error {
return tx.ForEach(func(name []byte, b *bolt.Bucket) error {
return walkBucket(b, nil, name, nil, b.Sequence(), walkFn)
})
})
}
func walkBucket(b *bolt.Bucket, keypath [][]byte, k, v []byte, seq uint64, fn walkFunc) error {
// Execute callback.
if err := fn(keypath, k, v, seq); err != nil {
return err
}
// If this is not a bucket then stop.
if v != nil {
return nil
}
// Iterate over each child key/value.
keypath = append(keypath, k)
return b.ForEach(func(k, v []byte) error {
if v == nil {
bkt := b.Bucket(k)
return walkBucket(bkt, keypath, k, nil, bkt.Sequence(), fn)
}
return walkBucket(b, keypath, k, v, b.Sequence(), fn)
})
}