forked from influxdata/influxdb
/
index_files.go
457 lines (388 loc) · 10.4 KB
/
index_files.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
package tsi1
import (
"bufio"
"io"
"os"
"sort"
"time"
"github.com/influxdata/influxdb/pkg/bytesutil"
"github.com/influxdata/influxdb/pkg/estimator/hll"
"github.com/influxdata/influxdb/tsdb"
)
// IndexFiles represents a layered set of index files.
type IndexFiles []*IndexFile
// IDs returns the ids for all index files.
func (p IndexFiles) IDs() []int {
a := make([]int, len(p))
for i, f := range p {
a[i] = f.ID()
}
return a
}
// Retain adds a reference count to all files.
func (p IndexFiles) Retain() {
for _, f := range p {
f.Retain()
}
}
// Release removes a reference count from all files.
func (p IndexFiles) Release() {
for _, f := range p {
f.Release()
}
}
// Files returns p as a list of File objects.
func (p IndexFiles) Files() []File {
other := make([]File, len(p))
for i, f := range p {
other[i] = f
}
return other
}
func (p IndexFiles) buildSeriesIDSets() (seriesIDSet, tombstoneSeriesIDSet *tsdb.SeriesIDSet, err error) {
if len(p) == 0 {
return tsdb.NewSeriesIDSet(), tsdb.NewSeriesIDSet(), nil
}
// Start with sets from last file.
if seriesIDSet, err = p[len(p)-1].SeriesIDSet(); err != nil {
return nil, nil, err
} else if tombstoneSeriesIDSet, err = p[len(p)-1].TombstoneSeriesIDSet(); err != nil {
return nil, nil, err
}
// Build sets in reverse order.
// This assumes that bits in both sets are mutually exclusive.
for i := len(p) - 2; i >= 0; i-- {
ss, err := p[i].SeriesIDSet()
if err != nil {
return nil, nil, err
}
ts, err := p[i].TombstoneSeriesIDSet()
if err != nil {
return nil, nil, err
}
// Add tombstones and remove from old series existence set.
seriesIDSet.Diff(ts)
tombstoneSeriesIDSet.Merge(ts)
// Add new series and remove from old series tombstone set.
tombstoneSeriesIDSet.Diff(ss)
seriesIDSet.Merge(ss)
}
return seriesIDSet, tombstoneSeriesIDSet, nil
}
// MeasurementNames returns a sorted list of all measurement names for all files.
func (p *IndexFiles) MeasurementNames() [][]byte {
itr := p.MeasurementIterator()
if itr == nil {
return nil
}
var names [][]byte
for e := itr.Next(); e != nil; e = itr.Next() {
names = append(names, bytesutil.Clone(e.Name()))
}
sort.Sort(byteSlices(names))
return names
}
// MeasurementIterator returns an iterator that merges measurements across all files.
func (p IndexFiles) MeasurementIterator() MeasurementIterator {
a := make([]MeasurementIterator, 0, len(p))
for i := range p {
itr := p[i].MeasurementIterator()
if itr == nil {
continue
}
a = append(a, itr)
}
return MergeMeasurementIterators(a...)
}
// TagKeyIterator returns an iterator that merges tag keys across all files.
func (p *IndexFiles) TagKeyIterator(name []byte) (TagKeyIterator, error) {
a := make([]TagKeyIterator, 0, len(*p))
for _, f := range *p {
itr := f.TagKeyIterator(name)
if itr == nil {
continue
}
a = append(a, itr)
}
return MergeTagKeyIterators(a...), nil
}
// MeasurementSeriesIDIterator returns an iterator that merges series across all files.
func (p IndexFiles) MeasurementSeriesIDIterator(name []byte) tsdb.SeriesIDIterator {
a := make([]tsdb.SeriesIDIterator, 0, len(p))
for _, f := range p {
itr := f.MeasurementSeriesIDIterator(name)
if itr == nil {
continue
}
a = append(a, itr)
}
return tsdb.MergeSeriesIDIterators(a...)
}
// TagValueSeriesIDSet returns an iterator that merges series across all files.
func (p IndexFiles) TagValueSeriesIDSet(name, key, value []byte) (*tsdb.SeriesIDSet, error) {
ss := tsdb.NewSeriesIDSet()
for i := range p {
if fss, err := p[i].TagValueSeriesIDSet(name, key, value); err != nil {
return nil, err
} else if fss != nil {
ss.Merge(fss)
}
}
return ss, nil
}
// CompactTo merges all index files and writes them to w.
func (p IndexFiles) CompactTo(w io.Writer, sfile *tsdb.SeriesFile, m, k uint64, cancel <-chan struct{}) (n int64, err error) {
var t IndexFileTrailer
// Check for cancellation.
select {
case <-cancel:
return n, ErrCompactionInterrupted
default:
}
// Wrap writer in buffered I/O.
bw := bufio.NewWriter(w)
// Setup context object to track shared data for this compaction.
var info indexCompactInfo
info.cancel = cancel
info.tagSets = make(map[string]indexTagSetPos)
// Write magic number.
if err := writeTo(bw, []byte(FileSignature), &n); err != nil {
return n, err
}
// Flush buffer before re-mapping.
if err := bw.Flush(); err != nil {
return n, err
}
// Write tagset blocks in measurement order.
if err := p.writeTagsetsTo(bw, &info, &n); err != nil {
return n, err
}
// Ensure block is word aligned.
// if offset := n % 8; offset != 0 {
// if err := writeTo(bw, make([]byte, 8-offset), &n); err != nil {
// return n, err
// }
// }
// Write measurement block.
t.MeasurementBlock.Offset = n
if err := p.writeMeasurementBlockTo(bw, &info, &n); err != nil {
return n, err
}
t.MeasurementBlock.Size = n - t.MeasurementBlock.Offset
// Build series sets.
seriesIDSet, tombstoneSeriesIDSet, err := p.buildSeriesIDSets()
if err != nil {
return n, err
}
// Generate sketches from series sets.
sketch := hll.NewDefaultPlus()
seriesIDSet.ForEach(func(id uint64) {
if key := sfile.SeriesKey(id); key != nil {
sketch.Add(key)
}
})
tSketch := hll.NewDefaultPlus()
tombstoneSeriesIDSet.ForEach(func(id uint64) {
if key := sfile.SeriesKey(id); key != nil {
tSketch.Add(key)
}
})
// Write series set.
t.SeriesIDSet.Offset = n
nn, err := seriesIDSet.WriteTo(bw)
if n += nn; err != nil {
return n, err
}
t.SeriesIDSet.Size = n - t.SeriesIDSet.Offset
// Write tombstone series set.
t.TombstoneSeriesIDSet.Offset = n
nn, err = tombstoneSeriesIDSet.WriteTo(bw)
if n += nn; err != nil {
return n, err
}
t.TombstoneSeriesIDSet.Size = n - t.TombstoneSeriesIDSet.Offset
// Write series sketches. TODO(edd): Implement WriterTo on HLL++.
t.SeriesSketch.Offset = n
data, err := sketch.MarshalBinary()
if err != nil {
return n, err
} else if _, err := bw.Write(data); err != nil {
return n, err
}
t.SeriesSketch.Size = int64(len(data))
n += t.SeriesSketch.Size
t.TombstoneSeriesSketch.Offset = n
if data, err = tSketch.MarshalBinary(); err != nil {
return n, err
} else if _, err := bw.Write(data); err != nil {
return n, err
}
t.TombstoneSeriesSketch.Size = int64(len(data))
n += t.TombstoneSeriesSketch.Size
// Write trailer.
nn, err = t.WriteTo(bw)
n += nn
if err != nil {
return n, err
}
// Flush file.
if err := bw.Flush(); err != nil {
return n, err
}
return n, nil
}
func (p IndexFiles) writeTagsetsTo(w io.Writer, info *indexCompactInfo, n *int64) error {
mitr := p.MeasurementIterator()
if mitr == nil {
return nil
}
for m := mitr.Next(); m != nil; m = mitr.Next() {
if err := p.writeTagsetTo(w, m.Name(), info, n); err != nil {
return err
}
}
return nil
}
// writeTagsetTo writes a single tagset to w and saves the tagset offset.
func (p IndexFiles) writeTagsetTo(w io.Writer, name []byte, info *indexCompactInfo, n *int64) error {
var seriesIDs []uint64
// Check for cancellation.
select {
case <-info.cancel:
return ErrCompactionInterrupted
default:
}
// Ensure block is word aligned.
// if offset := (*n) % 8; offset != 0 {
// if err := writeTo(w, make([]byte, 8-offset), n); err != nil {
// return err
// }
// }
kitr, err := p.TagKeyIterator(name)
if err != nil {
return err
}
enc := NewTagBlockEncoder(w)
for ke := kitr.Next(); ke != nil; ke = kitr.Next() {
// Encode key.
if err := enc.EncodeKey(ke.Key(), ke.Deleted()); err != nil {
return err
}
// Iterate over tag values.
vitr := ke.TagValueIterator()
for ve := vitr.Next(); ve != nil; ve = vitr.Next() {
seriesIDs = seriesIDs[:0]
// Merge all series together.
if err := func() error {
ss, err := p.TagValueSeriesIDSet(name, ke.Key(), ve.Value())
if err != nil {
return err
}
return enc.EncodeValue(ve.Value(), ve.Deleted(), ss)
}(); err != nil {
return nil
}
}
}
// Save tagset offset to measurement.
pos := info.tagSets[string(name)]
pos.offset = *n
// Flush data to writer.
err = enc.Close()
*n += enc.N()
if err != nil {
return err
}
// Save tagset size to measurement.
pos.size = *n - pos.offset
info.tagSets[string(name)] = pos
return nil
}
func (p IndexFiles) writeMeasurementBlockTo(w io.Writer, info *indexCompactInfo, n *int64) error {
mw := NewMeasurementBlockWriter()
// Check for cancellation.
select {
case <-info.cancel:
return ErrCompactionInterrupted
default:
}
// Add measurement data & compute sketches.
mitr := p.MeasurementIterator()
if mitr != nil {
var seriesN int
for m := mitr.Next(); m != nil; m = mitr.Next() {
name := m.Name()
// Look-up series ids.
if err := func() error {
itr := p.MeasurementSeriesIDIterator(name)
defer itr.Close()
var seriesIDs []uint64
for {
e, err := itr.Next()
if err != nil {
return err
} else if e.SeriesID == 0 {
break
}
seriesIDs = append(seriesIDs, e.SeriesID)
// Check for cancellation periodically.
if seriesN++; seriesN%1000 == 0 {
select {
case <-info.cancel:
return ErrCompactionInterrupted
default:
}
}
}
sort.Sort(uint64Slice(seriesIDs))
// Add measurement to writer.
pos := info.tagSets[string(name)]
mw.Add(name, m.Deleted(), pos.offset, pos.size, seriesIDs)
return nil
}(); err != nil {
return err
}
}
}
// Flush data to writer.
nn, err := mw.WriteTo(w)
*n += nn
return err
}
// Stat returns the max index file size and the total file size for all index files.
func (p IndexFiles) Stat() (*IndexFilesInfo, error) {
var info IndexFilesInfo
for _, f := range p {
fi, err := os.Stat(f.Path())
if os.IsNotExist(err) {
continue
} else if err != nil {
return nil, err
}
if fi.Size() > info.MaxSize {
info.MaxSize = fi.Size()
}
if fi.ModTime().After(info.ModTime) {
info.ModTime = fi.ModTime()
}
info.Size += fi.Size()
}
return &info, nil
}
type IndexFilesInfo struct {
MaxSize int64 // largest file size
Size int64 // total file size
ModTime time.Time // last modified
}
// indexCompactInfo is a context object used for tracking position information
// during the compaction of index files.
type indexCompactInfo struct {
cancel <-chan struct{}
// Tracks offset/size for each measurement's tagset.
tagSets map[string]indexTagSetPos
}
// indexTagSetPos stores the offset/size of tagsets.
type indexTagSetPos struct {
offset int64
size int64
}