-
-
Notifications
You must be signed in to change notification settings - Fork 121
/
pipeline.go
647 lines (568 loc) · 18.3 KB
/
pipeline.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
// Copyright 2018-2024 Burak Sezer
//
// 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 olric
import (
"bytes"
"context"
"errors"
"runtime"
"strconv"
"sync"
"time"
"github.com/buraksezer/olric/internal/cluster/partitions"
"github.com/buraksezer/olric/internal/dmap"
"github.com/buraksezer/olric/internal/protocol"
"github.com/buraksezer/olric/internal/resp"
"github.com/redis/go-redis/v9"
"golang.org/x/sync/errgroup"
"golang.org/x/sync/semaphore"
)
var (
// ErrNotReady denotes that the Future instance you hold is not ready to read the response yet.
ErrNotReady = errors.New("not ready yet")
// ErrPipelineClosed denotes that the underlying pipeline is closed, and it's impossible to operate.
ErrPipelineClosed = errors.New("pipeline is closed")
// ErrPipelineExecuted denotes that Exec was already called on the underlying pipeline.
ErrPipelineExecuted = errors.New("pipeline already executed")
)
// DMapPipeline implements a pipeline for the following methods of the DMap API:
//
// * Put
// * Get
// * Delete
// * Incr
// * Decr
// * GetPut
// * IncrByFloat
//
// DMapPipeline enables batch operations on DMap data.
type DMapPipeline struct {
mtx sync.Mutex
dm *ClusterDMap
commands map[uint64][]redis.Cmder
result map[uint64][]redis.Cmder
ctx context.Context
cancel context.CancelFunc
closedCtx context.Context // used to detect if the pipeline is closed / discarded
closedCancel context.CancelFunc
concurrency int // defaults to runtime.NumCPU()
}
func (dp *DMapPipeline) addCommand(key string, cmd redis.Cmder) (uint64, int) {
dp.mtx.Lock()
defer dp.mtx.Unlock()
hkey := partitions.HKey(dp.dm.name, key)
partID := hkey % dp.dm.clusterClient.partitionCount
cmds, ok := dp.commands[partID]
if !ok {
// if there are no existing commands, get a new slice from the pool
cmds = getPipelineCmdsFromPool()
}
dp.commands[partID] = append(cmds, cmd)
return partID, len(dp.commands[partID]) - 1
}
// FuturePut is used to read the result of a pipelined Put command.
type FuturePut struct {
dp *DMapPipeline
partID uint64
index int
ctx context.Context
closedCtx context.Context
}
// Result returns a response for the pipelined Put command.
func (f *FuturePut) Result() error {
// this select is separate from the one below on purpose, since select is non-deterministic if multiple
// cases are available, and we need to guarantee this check first.
select {
case <-f.closedCtx.Done():
return ErrPipelineClosed
default:
}
select {
case <-f.ctx.Done():
cmd := f.dp.result[f.partID][f.index]
return processProtocolError(cmd.Err())
default:
return ErrNotReady
}
}
// Put queues a Put command. The parameters are identical to the DMap.Put,
// but it returns FuturePut to read the batched response.
func (dp *DMapPipeline) Put(ctx context.Context, key string, value interface{}, options ...PutOption) (*FuturePut, error) {
buf := bytes.NewBuffer(nil)
enc := resp.New(buf)
err := enc.Encode(value)
if err != nil {
return nil, err
}
var pc dmap.PutConfig
for _, opt := range options {
opt(&pc)
}
cmd := dp.dm.writePutCommand(&pc, key, buf.Bytes()).Command(ctx)
partID, index := dp.addCommand(key, cmd)
return &FuturePut{
dp: dp,
partID: partID,
index: index,
ctx: dp.ctx,
closedCtx: dp.closedCtx,
}, nil
}
// FutureGet is used to read result of a pipelined Get command.
type FutureGet struct {
dp *DMapPipeline
partID uint64
index int
ctx context.Context
closedCtx context.Context
}
// Result returns a response for the pipelined Get command.
func (f *FutureGet) Result() (*GetResponse, error) {
// this select is separate from the one below on purpose, since select is non-deterministic if multiple
// cases are available, and we need to guarantee this check first.
select {
case <-f.closedCtx.Done():
return nil, ErrPipelineClosed
default:
}
select {
case <-f.ctx.Done():
cmd := f.dp.result[f.partID][f.index]
if cmd.Err() != nil {
return nil, processProtocolError(cmd.Err())
}
stringCmd := redis.NewStringCmd(context.Background(), cmd.Args()...)
stringCmd.SetVal(cmd.(*redis.Cmd).Val().(string))
return f.dp.dm.makeGetResponse(stringCmd)
default:
return nil, ErrNotReady
}
}
// Get queues a Get command. The parameters are identical to the DMap.Get,
// but it returns FutureGet to read the batched response.
func (dp *DMapPipeline) Get(ctx context.Context, key string) *FutureGet {
cmd := protocol.NewGet(dp.dm.name, key).SetRaw().Command(ctx)
partID, index := dp.addCommand(key, cmd)
return &FutureGet{
dp: dp,
partID: partID,
index: index,
ctx: dp.ctx,
closedCtx: dp.closedCtx,
}
}
// FutureDelete is used to read the result of a pipelined Delete command.
type FutureDelete struct {
dp *DMapPipeline
partID uint64
index int
ctx context.Context
closedCtx context.Context
}
// Result returns a response for the pipelined Delete command.
func (f *FutureDelete) Result() (int, error) {
// this select is separate from the one below on purpose, since select is non-deterministic if multiple
// cases are available, and we need to guarantee this check first.
select {
case <-f.closedCtx.Done():
return 0, ErrPipelineClosed
default:
}
select {
case <-f.ctx.Done():
cmd := f.dp.result[f.partID][f.index]
if cmd.Err() != nil {
return 0, processProtocolError(cmd.Err())
}
return int(cmd.(*redis.Cmd).Val().(int64)), nil
default:
return 0, ErrNotReady
}
}
// Delete queues a Delete command. The parameters are identical to the DMap.Delete,
// but it returns FutureDelete to read the batched response.
func (dp *DMapPipeline) Delete(ctx context.Context, key string) *FutureDelete {
cmd := protocol.NewDel(dp.dm.name, []string{key}...).Command(ctx)
partID, index := dp.addCommand(key, cmd)
return &FutureDelete{
dp: dp,
partID: partID,
index: index,
ctx: dp.ctx,
closedCtx: dp.closedCtx,
}
}
// FutureExpire is used to read the result of a pipelined Expire command.
type FutureExpire struct {
dp *DMapPipeline
partID uint64
index int
ctx context.Context
closedCtx context.Context
}
// Result returns a response for the pipelined Expire command.
func (f *FutureExpire) Result() error {
// this select is separate from the one below on purpose, since select is non-deterministic if multiple
// cases are available, and we need to guarantee this check first.
select {
case <-f.closedCtx.Done():
return ErrPipelineClosed
default:
}
select {
case <-f.ctx.Done():
cmd := f.dp.result[f.partID][f.index]
return processProtocolError(cmd.Err())
default:
return ErrNotReady
}
}
// Expire queues an Expire command. The parameters are identical to the DMap.Expire,
// but it returns FutureExpire to read the batched response.
func (dp *DMapPipeline) Expire(ctx context.Context, key string, timeout time.Duration) (*FutureExpire, error) {
cmd := protocol.NewExpire(dp.dm.name, key, timeout).Command(ctx)
partID, index := dp.addCommand(key, cmd)
return &FutureExpire{
dp: dp,
partID: partID,
index: index,
ctx: dp.ctx,
closedCtx: dp.closedCtx,
}, nil
}
// FutureIncr is used to read the result of a pipelined Incr command.
type FutureIncr struct {
dp *DMapPipeline
partID uint64
index int
ctx context.Context
closedCtx context.Context
}
// Result returns a response for the pipelined Incr command.
func (f *FutureIncr) Result() (int, error) {
// this select is separate from the one below on purpose, since select is non-deterministic if multiple
// cases are available, and we need to guarantee this check first.
select {
case <-f.closedCtx.Done():
return 0, ErrPipelineClosed
default:
}
select {
case <-f.ctx.Done():
cmd := f.dp.result[f.partID][f.index]
if cmd.Err() != nil {
return 0, processProtocolError(cmd.Err())
}
return int(cmd.(*redis.Cmd).Val().(int64)), nil
default:
return 0, ErrNotReady
}
}
// Incr queues an Incr command. The parameters are identical to the DMap.Incr,
// but it returns FutureIncr to read the batched response.
func (dp *DMapPipeline) Incr(ctx context.Context, key string, delta int) (*FutureIncr, error) {
cmd := protocol.NewIncr(dp.dm.name, key, delta).Command(ctx)
partID, index := dp.addCommand(key, cmd)
return &FutureIncr{
dp: dp,
partID: partID,
index: index,
ctx: dp.ctx,
closedCtx: dp.closedCtx,
}, nil
}
// FutureDecr is used to read the result of a pipelined Decr command.
type FutureDecr struct {
dp *DMapPipeline
partID uint64
index int
ctx context.Context
closedCtx context.Context
}
// Result returns a response for the pipelined Decr command.
func (f *FutureDecr) Result() (int, error) {
// this select is separate from the one below on purpose, since select is non-deterministic if multiple
// cases are available, and we need to guarantee this check first.
select {
case <-f.closedCtx.Done():
return 0, ErrPipelineClosed
default:
}
select {
case <-f.ctx.Done():
cmd := f.dp.result[f.partID][f.index]
if cmd.Err() != nil {
return 0, processProtocolError(cmd.Err())
}
return int(cmd.(*redis.Cmd).Val().(int64)), nil
default:
return 0, ErrNotReady
}
}
// Decr queues a Decr command. The parameters are identical to the DMap.Decr,
// but it returns FutureDecr to read the batched response.
func (dp *DMapPipeline) Decr(ctx context.Context, key string, delta int) (*FutureDecr, error) {
cmd := protocol.NewDecr(dp.dm.name, key, delta).Command(ctx)
partID, index := dp.addCommand(key, cmd)
return &FutureDecr{
dp: dp,
partID: partID,
index: index,
ctx: dp.ctx,
closedCtx: dp.closedCtx,
}, nil
}
// FutureGetPut is used to read the result of a pipelined GetPut command.
type FutureGetPut struct {
dp *DMapPipeline
partID uint64
index int
ctx context.Context
closedCtx context.Context
}
// Result returns a response for the pipelined GetPut command.
func (f *FutureGetPut) Result() (*GetResponse, error) {
// this select is separate from the one below on purpose, since select is non-deterministic if multiple
// cases are available, and we need to guarantee this check first.
select {
case <-f.closedCtx.Done():
return nil, ErrPipelineClosed
default:
}
select {
case <-f.ctx.Done():
cmd := f.dp.result[f.partID][f.index]
if cmd.Err() == redis.Nil {
// This should be the first run.
return nil, nil
}
if cmd.Err() != nil {
return nil, processProtocolError(cmd.Err())
}
stringCmd := redis.NewStringCmd(context.Background(), cmd.Args()...)
stringCmd.SetVal(cmd.(*redis.Cmd).Val().(string))
return f.dp.dm.makeGetResponse(stringCmd)
default:
return nil, ErrNotReady
}
}
// GetPut queues a GetPut command. The parameters are identical to the DMap.GetPut,
// but it returns FutureGetPut to read the batched response.
func (dp *DMapPipeline) GetPut(ctx context.Context, key string, value interface{}) (*FutureGetPut, error) {
buf := bytes.NewBuffer(nil)
enc := resp.New(buf)
err := enc.Encode(value)
if err != nil {
return nil, err
}
cmd := protocol.NewGetPut(dp.dm.name, key, buf.Bytes()).SetRaw().Command(ctx)
partID, index := dp.addCommand(key, cmd)
return &FutureGetPut{
dp: dp,
partID: partID,
index: index,
ctx: dp.ctx,
closedCtx: dp.closedCtx,
}, nil
}
// FutureIncrByFloat is used to read the result of a pipelined IncrByFloat command.
type FutureIncrByFloat struct {
dp *DMapPipeline
partID uint64
index int
ctx context.Context
closedCtx context.Context
}
// Result returns a response for the pipelined IncrByFloat command.
func (f *FutureIncrByFloat) Result() (float64, error) {
// this select is separate from the one below on purpose, since select is non-deterministic if multiple
// cases are available, and we need to guarantee this check first.
select {
case <-f.closedCtx.Done():
return 0, ErrPipelineClosed
default:
}
select {
case <-f.ctx.Done():
cmd := f.dp.result[f.partID][f.index]
if cmd.Err() != nil {
return 0, processProtocolError(cmd.Err())
}
stringRes := cmd.(*redis.Cmd).Val().(string)
return strconv.ParseFloat(stringRes, 64)
default:
return 0, ErrNotReady
}
}
// IncrByFloat queues an IncrByFloat command. The parameters are identical to the DMap.IncrByFloat,
// but it returns FutureIncrByFloat to read the batched response.
func (dp *DMapPipeline) IncrByFloat(ctx context.Context, key string, delta float64) (*FutureIncrByFloat, error) {
cmd := protocol.NewIncrByFloat(dp.dm.name, key, delta).Command(ctx)
partID, index := dp.addCommand(key, cmd)
return &FutureIncrByFloat{
dp: dp,
partID: partID,
index: index,
ctx: dp.ctx,
closedCtx: dp.closedCtx,
}, nil
}
func (dp *DMapPipeline) execOnPartition(ctx context.Context, partID uint64) error {
rc, err := dp.dm.clusterClient.clientByPartID(partID)
if err != nil {
return err
}
// There is no need to protect dp.commands map and its content.
// It's already filled before running Exec, and it's now a read-only
// data structure
commands := dp.commands[partID]
pipe := rc.Pipeline()
for _, cmd := range commands {
pipe.Do(ctx, cmd.Args()...)
}
// Exec executes all previously queued commands using one
// client-server roundtrip.
//
// Exec always returns list of commands and error of the first failed
// command if any.
result, _ := pipe.Exec(ctx)
dp.mtx.Lock()
dp.result[partID] = result
dp.mtx.Unlock()
return nil
}
// Exec executes all queued commands using one client-server roundtrip per partition.
func (dp *DMapPipeline) Exec(ctx context.Context) error {
// this select is separate from the one below on purpose, since select is non-deterministic if multiple
// cases are available, and we need to guarantee this check first.
select {
case <-dp.closedCtx.Done():
return ErrPipelineClosed
default:
}
// this checks to see if Exec has already run. While Exec should only be called once, it is possible that
// the user could call Exec multiple times. If we stored the result of errGr.Wait on the pipeline, we could
// return that error and make Exec idempotent.
select {
case <-dp.ctx.Done():
return ErrPipelineExecuted
default:
}
defer dp.cancel()
var errGr errgroup.Group
sem := semaphore.NewWeighted(int64(dp.concurrency))
for i := uint64(0); i < dp.dm.clusterClient.partitionCount; i++ {
err := sem.Acquire(ctx, 1)
if err != nil {
return err
}
partID := i
errGr.Go(func() error {
defer sem.Release(1)
// If execOnPartition returns an error, it will eventually stop
// all flush operation.
return dp.execOnPartition(ctx, partID)
})
}
return errGr.Wait()
}
// Discard discards the pipelined commands and resets all internal states.
// A pipeline can be reused after calling Discard.
func (dp *DMapPipeline) Discard() error {
select {
case <-dp.closedCtx.Done():
return ErrPipelineClosed
default:
}
dp.closedCancel()
dp.mtx.Lock()
defer dp.mtx.Unlock()
// return all command slices to the pool
for _, v := range dp.commands {
putPipelineCmdsIntoPool(v)
}
for _, v := range dp.result {
putPipelineCmdsIntoPool(v)
}
// the deletes below are purposefully not combined with the loops above, as these are recognized and optimized
// by the compiler. https://go-review.googlesource.com/c/go/+/110055
for k := range dp.commands {
delete(dp.commands, k)
}
for k := range dp.result {
delete(dp.result, k)
}
dp.initContexts()
return nil
}
// Close closes the pipeline and frees the allocated resources. You shouldn't try to
// reuse a closed pipeline.
func (dp *DMapPipeline) Close() {
dp.closedCancel()
}
// Pipeline is a mechanism to realise Redis Pipeline technique.
//
// Pipelining is a technique to extremely speed up processing by packing
// operations to batches, send them at once to Redis and read a replies in a
// singe step.
// See https://redis.io/topics/pipelining
//
// Pay attention, that Pipeline is not a transaction, so you can get unexpected
// results in case of big pipelines and small read/write timeouts.
// Redis client has retransmission logic in case of timeouts, pipeline
// can be retransmitted and commands can be executed more than once.
func (dm *ClusterDMap) Pipeline(opts ...PipelineOption) (*DMapPipeline, error) {
dp := &DMapPipeline{
dm: dm,
commands: make(map[uint64][]redis.Cmder),
result: make(map[uint64][]redis.Cmder),
concurrency: runtime.NumCPU(),
}
for _, opt := range opts {
opt(dp)
}
dp.initContexts()
return dp, nil
}
// initContexts sets up chained contexts for the pipeline. The base is closedCtx, which is closed either in
// Close or Discard. ctx is a child of closedCtx, as we want to cancel the pipeline if it is closed. It is
// canceled in Exec, and used to block FutureXXX.Result() calls until Exec has completed.
func (dp *DMapPipeline) initContexts() {
dp.closedCtx, dp.closedCancel = context.WithCancel(context.Background())
dp.ctx, dp.cancel = context.WithCancel(dp.closedCtx)
}
// This stores a slice of commands for each partition. There is a possibility that a single
// large slice could be allocated with an unusually large number of commands in a single pipeline that
// are very unbalanced across partitions, but that is unlikely to be a problem in practice.
//
// It does not store a pointer to the slice as recommended by staticcheck because that is harder to reason
// about, and a single allocation is not a big deal compared to the slices we're able to reuse.
// https://staticcheck.io/docs/checks#SA6002
// https://github.com/dominikh/go-tools/issues/1336#issuecomment-1331206290
var pipelineCmdPool = sync.Pool{
New: func() interface{} {
return make([]redis.Cmder, 0)
},
}
func getPipelineCmdsFromPool() []redis.Cmder {
return pipelineCmdPool.Get().([]redis.Cmder)
}
func putPipelineCmdsIntoPool(cmds []redis.Cmder) {
// remove references to underlying commands so they can be GCed
for i := range cmds {
cmds[i] = nil
}
cmds = cmds[:0]
pipelineCmdPool.Put(cmds)
}