forked from polarsignals/frostdb
-
Notifications
You must be signed in to change notification settings - Fork 0
/
levels.go
333 lines (278 loc) · 9.09 KB
/
levels.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
package index
import (
"bytes"
"encoding/binary"
"fmt"
"io"
"os"
"path/filepath"
"strconv"
"sync"
"github.com/go-kit/log"
"github.com/go-kit/log/level"
"github.com/parquet-go/parquet-go"
"github.com/dreamsxin/frostdb/dynparquet"
"github.com/dreamsxin/frostdb/parts"
)
const (
IndexFileExtension = ".idx"
ParquetCompactionTXKey = "compaction_tx"
dirPerms = os.FileMode(0o755)
filePerms = os.FileMode(0o640)
)
type Compaction func(w io.Writer, compact []parts.Part, options ...parquet.WriterOption) (int64, error)
type FileCompaction struct {
// settings
dir string
compact Compaction
maxSize int64
// internal data
indexFiles []*os.File
offset int64 // Writing offsets into the file
parts sync.WaitGroup // Wait group for parts that are currently reference in this level.
// Options
logger log.Logger
}
func NewFileCompaction(dir string, maxSize int64, compact Compaction, logger log.Logger) (*FileCompaction, error) {
f := &FileCompaction{
dir: dir,
compact: compact,
maxSize: maxSize,
logger: logger,
}
if err := os.MkdirAll(dir, dirPerms); err != nil {
return nil, err
}
return f, nil
}
func (f *FileCompaction) MaxSize() int64 { return f.maxSize }
// Snapshot takes a snapshot of the current level. It ignores the parts and just hard links the files into the snapshot directory.
// It will rotate the active file if it has data in it rendering all snapshotted files as immutable.
func (f *FileCompaction) Snapshot(_ []parts.Part, _ func(parts.Part) error, dir string) error {
if err := os.MkdirAll(dir, dirPerms); err != nil {
return err
}
for i, file := range f.indexFiles {
if i == len(f.indexFiles)-1 {
// Sync the last file if it has data in it.
if f.offset > 0 {
if err := f.Sync(); err != nil {
return err
}
} else {
return nil // Skip empty file.
}
}
// Hard link the file into the snapshot directory.
if err := os.Link(file.Name(), filepath.Join(dir, filepath.Base(file.Name()))); err != nil {
return err
}
}
// Rotate the active file if it has data in it.
_, err := f.createIndexFile(len(f.indexFiles))
return err
}
func (f *FileCompaction) createIndexFile(id int) (*os.File, error) {
file, err := os.OpenFile(filepath.Join(f.dir, fmt.Sprintf("%020d%s", id, IndexFileExtension)), os.O_CREATE|os.O_RDWR, filePerms)
if err != nil {
return nil, err
}
f.offset = 0
f.indexFiles = append(f.indexFiles, file)
return file, nil
}
func (f *FileCompaction) openIndexFile(path string) (*os.File, error) {
file, err := os.Open(path)
if err != nil {
return nil, err
}
f.indexFiles = append(f.indexFiles, file)
return file, nil
}
// file returns the currently active index file.
func (f *FileCompaction) file() *os.File {
return f.indexFiles[len(f.indexFiles)-1]
}
// accountingWriter is a writer that accounts for the number of bytes written.
type accountingWriter struct {
w io.Writer
n int64
}
func (a *accountingWriter) Write(p []byte) (int, error) {
n, err := a.w.Write(p)
a.n += int64(n)
return n, err
}
// Compact will compact the given parts into a Parquet file written to the next level file.
func (f *FileCompaction) Compact(compact []parts.Part, options ...parts.Option) ([]parts.Part, int64, int64, error) {
if len(compact) == 0 {
return nil, 0, 0, fmt.Errorf("no parts to compact")
}
accountant := &accountingWriter{w: f.file()}
preCompactionSize, err := f.compact(accountant, compact,
parquet.KeyValueMetadata(
ParquetCompactionTXKey, // Compacting up through this transaction.
fmt.Sprintf("%v", compact[0].TX()),
),
) // compact into the next level
if err != nil {
return nil, 0, 0, err
}
// Record the writing offset into the file.
prevOffset := f.offset
// Record the file size for recovery.
size := make([]byte, 8)
binary.LittleEndian.PutUint64(size, uint64(accountant.n))
if n, err := f.file().Write(size); n != 8 {
return nil, 0, 0, fmt.Errorf("failed to write size to file: %v", err)
}
f.offset += accountant.n + 8
// Sync file after writing.
if err := f.Sync(); err != nil {
return nil, 0, 0, fmt.Errorf("failed to sync file: %v", err)
}
pf, err := parquet.OpenFile(io.NewSectionReader(f.file(), prevOffset, accountant.n), accountant.n)
if err != nil {
return nil, 0, 0, fmt.Errorf("failed to open file after compaction: %w", err)
}
buf, err := dynparquet.NewSerializedBuffer(pf)
if err != nil {
return nil, 0, 0, err
}
f.parts.Add(1)
return []parts.Part{parts.NewParquetPart(compact[0].TX(), buf, append(options, parts.WithRelease(f.parts.Done))...)}, preCompactionSize, accountant.n, nil
}
// Reset is called when the level no longer has active parts in it at the end of a compaction.
func (f *FileCompaction) Reset() {
f.parts.Wait() // Wait for all parts to be released.
for _, file := range f.indexFiles {
if err := file.Close(); err != nil {
level.Error(f.logger).Log("msg", "failed to close level file", "err", err)
}
}
// Delete all the files in the directory level. And open a new file.
if err := os.RemoveAll(f.dir); err != nil {
level.Error(f.logger).Log("msg", "failed to remove level directory", "err", err)
}
if err := os.MkdirAll(f.dir, dirPerms); err != nil {
level.Error(f.logger).Log("msg", "failed to create level directory", "err", err)
}
f.indexFiles = nil
_, err := f.createIndexFile(len(f.indexFiles))
if err != nil {
level.Error(f.logger).Log("msg", "failed to create new level file", "err", err)
}
}
// recovery the level from the given directory.
func (f *FileCompaction) recover(options ...parts.Option) ([]parts.Part, error) {
defer func() {
_, err := f.createIndexFile(len(f.indexFiles))
if err != nil {
level.Error(f.logger).Log("msg", "failed to create new level file", "err", err)
}
}()
recovered := []parts.Part{}
err := filepath.WalkDir(f.dir, func(path string, d os.DirEntry, err error) error {
if err != nil {
return err
}
if filepath.Ext(path) != IndexFileExtension {
return nil
}
info, err := d.Info()
if err != nil {
return fmt.Errorf("failed to get file info: %v", err)
}
if info.Size() == 0 { // file empty, nothing to recover.
return nil
}
file, err := f.openIndexFile(path)
if err != nil {
return fmt.Errorf("failed to open file: %v", err)
}
// Recover all parts from file.
fileParts := []parts.Part{}
if err := func() error {
for offset := info.Size(); offset > 0; {
offset -= 8
size := make([]byte, 8)
if n, err := file.ReadAt(size, offset); n != 8 {
return fmt.Errorf("failed to read size from file: %v", err)
}
parquetSize := int64(binary.LittleEndian.Uint64(size))
offset -= parquetSize
pf, err := parquet.OpenFile(io.NewSectionReader(file, offset, parquetSize), parquetSize)
if err != nil {
return err
}
buf, err := dynparquet.NewSerializedBuffer(pf)
if err != nil {
return err
}
var tx int
txstr, ok := buf.ParquetFile().Lookup(ParquetCompactionTXKey)
if !ok {
level.Warn(f.logger).Log("msg", "failed to find compaction_tx metadata", "file", file.Name())
tx = 0 // Downgrade the compaction tx so that all future reads will be able to read this part.
} else {
tx, err = strconv.Atoi(txstr)
if err != nil {
level.Warn(f.logger).Log("msg", "failed to parse compaction_tx metadata", "file", file.Name(), "err", err)
tx = 0 // Downgrade the compaction tx so that all future reads will be able to read this part.
}
}
f.parts.Add(1)
fileParts = append(fileParts, parts.NewParquetPart(uint64(tx), buf, append(options, parts.WithRelease(f.parts.Done))...))
}
return nil
}(); err != nil {
for _, part := range fileParts {
part.Release()
}
// If we failed to recover the file, remove it.
if err := f.file().Close(); err != nil {
level.Error(f.logger).Log("msg", "failed to close level file after failed recovery", "err", err)
}
f.indexFiles = f.indexFiles[:len(f.indexFiles)-1] // Remove the file from the list of files.
return err
}
recovered = append(recovered, fileParts...)
return nil
})
if err != nil {
return nil, err
}
return recovered, nil
}
// Sync calls Sync on the underlying file.
func (f *FileCompaction) Sync() error { return f.file().Sync() }
type inMemoryLevel struct {
compact Compaction
maxSize int64
}
func (l *inMemoryLevel) MaxSize() int64 { return l.maxSize }
func (l *inMemoryLevel) Snapshot(snapshot []parts.Part, writer func(parts.Part) error, _ string) error {
for _, part := range snapshot {
if err := writer(part); err != nil {
return err
}
}
return nil
}
func (l *inMemoryLevel) Reset() {}
func (l *inMemoryLevel) Compact(toCompact []parts.Part, options ...parts.Option) ([]parts.Part, int64, int64, error) {
if len(toCompact) == 0 {
return nil, 0, 0, fmt.Errorf("no parts to compact")
}
var b bytes.Buffer
preCompactionSize, err := l.compact(&b, toCompact)
if err != nil {
return nil, 0, 0, err
}
buf, err := dynparquet.ReaderFromBytes(b.Bytes())
if err != nil {
return nil, 0, 0, err
}
postCompactionSize := int64(b.Len())
return []parts.Part{parts.NewParquetPart(toCompact[0].TX(), buf, options...)}, preCompactionSize, postCompactionSize, nil
}