-
Notifications
You must be signed in to change notification settings - Fork 176
/
engine.go
440 lines (373 loc) · 12.5 KB
/
engine.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
// (c) 2019 Dapper Labs - ALL RIGHTS RESERVED
package synchronization
import (
"errors"
"fmt"
"math/rand"
"time"
"github.com/hashicorp/go-multierror"
"github.com/rs/zerolog"
"github.com/onflow/flow-go/engine"
clustermodel "github.com/onflow/flow-go/model/cluster"
"github.com/onflow/flow-go/model/events"
"github.com/onflow/flow-go/model/flow"
"github.com/onflow/flow-go/model/flow/filter"
"github.com/onflow/flow-go/model/messages"
"github.com/onflow/flow-go/module"
"github.com/onflow/flow-go/module/metrics"
synccore "github.com/onflow/flow-go/module/synchronization"
"github.com/onflow/flow-go/network"
"github.com/onflow/flow-go/state/cluster"
"github.com/onflow/flow-go/storage"
)
// Engine is the synchronization engine, responsible for synchronizing chain state.
type Engine struct {
unit *engine.Unit
log zerolog.Logger
metrics module.EngineMetrics
me module.Local
participants flow.IdentityList
state cluster.State
conduit network.Conduit
blocks storage.ClusterBlocks
comp network.Engine // compliance layer engine
pollInterval time.Duration
scanInterval time.Duration
core module.SyncCore
}
// New creates a new cluster synchronization engine.
func New(
log zerolog.Logger,
metrics module.EngineMetrics,
net module.Network,
me module.Local,
participants flow.IdentityList,
state cluster.State,
blocks storage.ClusterBlocks,
comp network.Engine,
core module.SyncCore,
) (*Engine, error) {
// initialize the propagation engine with its dependencies
e := &Engine{
unit: engine.NewUnit(),
log: log.With().Str("engine", "cluster_synchronization").Logger(),
metrics: metrics,
me: me,
participants: participants.Filter(filter.Not(filter.HasNodeID(me.NodeID()))),
state: state,
blocks: blocks,
comp: comp,
core: core,
pollInterval: 8 * time.Second,
scanInterval: 2 * time.Second,
}
// register the engine with the network layer and store the conduit
conduit, err := net.Register(engine.SyncCluster, e)
if err != nil {
return nil, fmt.Errorf("could not register engine: %w", err)
}
e.conduit = conduit
return e, nil
}
// Ready returns a ready channel that is closed once the engine has fully
// started. For consensus engine, this is true once the underlying consensus
// algorithm has started.
func (e *Engine) Ready() <-chan struct{} {
e.unit.Launch(e.checkLoop)
return e.unit.Ready()
}
// Done returns a done channel that is closed once the engine has fully stopped.
// For the consensus engine, we wait for hotstuff to finish.
func (e *Engine) Done() <-chan struct{} {
return e.unit.Done()
}
// SubmitLocal submits an event originating on the local node.
func (e *Engine) SubmitLocal(event interface{}) {
e.Submit(e.me.NodeID(), event)
}
// Submit submits the given event from the node with the given origin ID
// for processing in a non-blocking manner. It returns instantly and logs
// a potential processing error internally when done.
func (e *Engine) Submit(originID flow.Identifier, event interface{}) {
e.unit.Launch(func() {
err := e.Process(originID, event)
if err != nil {
engine.LogError(e.log, err)
}
})
}
// ProcessLocal processes an event originating on the local node.
func (e *Engine) ProcessLocal(event interface{}) error {
return e.Process(e.me.NodeID(), event)
}
// Process processes the given event from the node with the given origin ID in
// a blocking manner. It returns the potential processing error when done.
func (e *Engine) Process(originID flow.Identifier, event interface{}) error {
return e.unit.Do(func() error {
return e.process(originID, event)
})
}
// process processes events for the propagation engine on the consensus node.
func (e *Engine) process(originID flow.Identifier, event interface{}) error {
switch ev := event.(type) {
case *messages.SyncRequest:
e.before(metrics.MessageSyncRequest)
defer e.after(metrics.MessageSyncRequest)
return e.onSyncRequest(originID, ev)
case *messages.SyncResponse:
e.before(metrics.MessageSyncResponse)
defer e.after(metrics.MessageSyncResponse)
return e.onSyncResponse(originID, ev)
case *messages.RangeRequest:
e.before(metrics.MessageRangeRequest)
defer e.after(metrics.MessageRangeRequest)
return e.onRangeRequest(originID, ev)
case *messages.BatchRequest:
e.before(metrics.MessageBatchRequest)
defer e.after(metrics.MessageBatchRequest)
return e.onBatchRequest(originID, ev)
case *messages.ClusterBlockResponse:
e.before(metrics.MessageClusterBlockResponse)
defer e.after(metrics.MessageClusterBlockResponse)
return e.onBlockResponse(originID, ev)
default:
return fmt.Errorf("invalid event type (%T)", event)
}
}
func (e *Engine) before(msg string) {
e.metrics.MessageReceived(metrics.EngineClusterSynchronization, msg)
e.unit.Lock()
}
func (e *Engine) after(msg string) {
e.unit.Unlock()
e.metrics.MessageHandled(metrics.EngineClusterSynchronization, msg)
}
// onSyncRequest processes an outgoing handshake; if we have a higher height, we
// inform the other node of it, so they can organize their block downloads. If
// we have a lower height, we add the difference to our own download queue.
func (e *Engine) onSyncRequest(originID flow.Identifier, req *messages.SyncRequest) error {
final, err := e.state.Final().Head()
if err != nil {
return fmt.Errorf("could not get finalized height: %w", err)
}
// queue any missing heights as needed
e.core.HandleHeight(final, req.Height)
// don't bother sending a response if we're within tolerance or if we're
// behind the requester
if e.core.WithinTolerance(final, req.Height) || req.Height > final.Height {
return nil
}
// if we're sufficiently ahead of the requester, send a response
res := &messages.SyncResponse{
Height: final.Height,
Nonce: req.Nonce,
}
err = e.conduit.Unicast(res, originID)
if err != nil {
return fmt.Errorf("could not send sync response: %w", err)
}
e.metrics.MessageSent(metrics.EngineClusterSynchronization, metrics.MessageSyncResponse)
return nil
}
// onSyncResponse processes a synchronization response.
func (e *Engine) onSyncResponse(originID flow.Identifier, res *messages.SyncResponse) error {
final, err := e.state.Final().Head()
if err != nil {
return fmt.Errorf("could not get finalized height: %w", err)
}
e.core.HandleHeight(final, res.Height)
return nil
}
// onRangeRequest processes a request for a range of blocks by height.
func (e *Engine) onRangeRequest(originID flow.Identifier, req *messages.RangeRequest) error {
// get the latest final state to know if we can fulfill the request
head, err := e.state.Final().Head()
if err != nil {
return fmt.Errorf("could not get finalized head: %w", err)
}
// if we don't have anything to send, we can bail right away
if head.Height < req.FromHeight || req.FromHeight > req.ToHeight {
return nil
}
// get all of the blocks, one by one
blocks := make([]*clustermodel.Block, 0, req.ToHeight-req.FromHeight+1)
for height := req.FromHeight; height <= req.ToHeight; height++ {
block, err := e.blocks.ByHeight(height)
if errors.Is(err, storage.ErrNotFound) {
e.log.Error().Uint64("height", height).Msg("skipping unknown heights")
break
}
if err != nil {
return fmt.Errorf("could not get block for height (%d): %w", height, err)
}
blocks = append(blocks, block)
}
// if there are no blocks to send, skip network message
if len(blocks) == 0 {
e.log.Debug().Msg("skipping empty range response")
return nil
}
// send the response
res := &messages.ClusterBlockResponse{
Nonce: req.Nonce,
Blocks: blocks,
}
err = e.conduit.Unicast(res, originID)
if err != nil {
return fmt.Errorf("could not send range response: %w", err)
}
e.metrics.MessageSent(metrics.EngineClusterSynchronization, metrics.MessageBlockResponse)
return nil
}
// onBatchRequest processes a request for a specific block by block ID.
func (e *Engine) onBatchRequest(originID flow.Identifier, req *messages.BatchRequest) error {
// we should bail and send nothing on empty request
if len(req.BlockIDs) == 0 {
return nil
}
// deduplicate the block IDs in the batch request
blockIDs := make(map[flow.Identifier]struct{})
for _, blockID := range req.BlockIDs {
blockIDs[blockID] = struct{}{}
}
// try to get all the blocks by ID
blocks := make([]*clustermodel.Block, 0, len(blockIDs))
for blockID := range blockIDs {
block, err := e.blocks.ByID(blockID)
if errors.Is(err, storage.ErrNotFound) {
e.log.Debug().Hex("block_id", blockID[:]).Msg("skipping unknown block")
continue
}
if err != nil {
return fmt.Errorf("could not get block by ID (%s): %w", blockID, err)
}
blocks = append(blocks, block)
}
// if there are no blocks to send, skip network message
if len(blocks) == 0 {
e.log.Debug().Msg("skipping empty batch response")
return nil
}
// send the response
res := &messages.ClusterBlockResponse{
Nonce: req.Nonce,
Blocks: blocks,
}
err := e.conduit.Unicast(res, originID)
if err != nil {
return fmt.Errorf("could not send batch response: %w", err)
}
e.metrics.MessageSent(metrics.EngineClusterSynchronization, metrics.MessageBlockResponse)
return nil
}
// onBlockResponse processes a response containing a specifically requested block.
func (e *Engine) onBlockResponse(originID flow.Identifier, res *messages.ClusterBlockResponse) error {
// process the blocks one by one
for _, block := range res.Blocks {
e.processIncomingBlock(originID, block)
}
return nil
}
// processIncoming processes an incoming block, so we can take into account the
// overlap between block IDs and heights.
func (e *Engine) processIncomingBlock(originID flow.Identifier, block *clustermodel.Block) {
shouldProcess := e.core.HandleBlock(block.Header)
if !shouldProcess {
return
}
synced := &events.SyncedClusterBlock{
OriginID: originID,
Block: block,
}
e.comp.SubmitLocal(synced)
}
// checkLoop will regularly scan for items that need requesting.
func (e *Engine) checkLoop() {
poll := time.NewTicker(e.pollInterval)
scan := time.NewTicker(e.scanInterval)
CheckLoop:
for {
// give the quit channel a priority to be selected
select {
case <-e.unit.Quit():
break CheckLoop
default:
}
select {
case <-e.unit.Quit():
break CheckLoop
case <-poll.C:
err := e.pollHeight()
if err != nil {
e.log.Error().Err(err).Msg("could not poll heights")
}
case <-scan.C:
final, err := e.state.Final().Head()
if err != nil {
e.log.Error().Err(err).Msg("could not get final height")
continue
}
e.unit.Lock()
ranges, batches := e.core.ScanPending(final)
err = e.sendRequests(ranges, batches)
if err != nil {
e.log.Error().Err(err).Msg("could not send requests")
}
e.unit.Unlock()
}
}
// some minor cleanup
scan.Stop()
poll.Stop()
}
// pollHeight will send a synchronization request to three random nodes.
func (e *Engine) pollHeight() error {
// get the last finalized header
final, err := e.state.Final().Head()
if err != nil {
return fmt.Errorf("could not get last finalized header: %w", err)
}
// send the request for synchronization
req := &messages.SyncRequest{
Nonce: rand.Uint64(),
Height: final.Height,
}
err = e.conduit.Multicast(req, synccore.DefaultPollNodes, e.participants.NodeIDs()...)
if err != nil {
return fmt.Errorf("could not send sync request: %w", err)
}
e.metrics.MessageSent(metrics.EngineClusterSynchronization, metrics.MessageSyncRequest)
return nil
}
// sendRequests sends a request for each range and batch.
func (e *Engine) sendRequests(ranges []flow.Range, batches []flow.Batch) error {
var errs error
for _, ran := range ranges {
req := &messages.RangeRequest{
Nonce: rand.Uint64(),
FromHeight: ran.From,
ToHeight: ran.To,
}
err := e.conduit.Multicast(req, synccore.DefaultBlockRequestNodes, e.participants.NodeIDs()...)
if err != nil {
errs = multierror.Append(errs, fmt.Errorf("could not submit range request (from=%d, to=%d): %w", ran.From, ran.To, err))
continue
}
e.core.RangeRequested(ran)
e.metrics.MessageSent(metrics.EngineClusterSynchronization, metrics.MessageRangeRequest)
}
for _, batch := range batches {
req := &messages.BatchRequest{
Nonce: rand.Uint64(),
BlockIDs: batch.BlockIDs,
}
err := e.conduit.Multicast(req, synccore.DefaultBlockRequestNodes, e.participants.NodeIDs()...)
if err != nil {
errs = multierror.Append(errs, fmt.Errorf("could not submit batch request (size=%d): %w", len(batch.BlockIDs), err))
continue
}
e.core.BatchRequested(batch)
e.metrics.MessageSent(metrics.EngineClusterSynchronization, metrics.MessageBatchRequest)
}
return errs
}