-
Notifications
You must be signed in to change notification settings - Fork 146
/
runtime.go
469 lines (383 loc) · 10.6 KB
/
runtime.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
/*
* Copyright 2018 The CovenantSQL Authors.
*
* 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 kayak
import (
"context"
"math"
"sync"
"sync/atomic"
"time"
"github.com/pkg/errors"
kt "github.com/CovenantSQL/CovenantSQL/kayak/types"
"github.com/CovenantSQL/CovenantSQL/proto"
"github.com/CovenantSQL/CovenantSQL/utils/log"
"github.com/CovenantSQL/CovenantSQL/utils/timer"
"github.com/CovenantSQL/CovenantSQL/utils/trace"
)
const (
// commit channel window size
commitWindow = 0
)
// Runtime defines the main kayak Runtime.
type Runtime struct {
/// Indexes
// index for next log.
nextIndexLock sync.Mutex
nextIndex uint64
// lastCommit, last commit log index
lastCommit uint64
// pendingPrepares, prepares needs to be committed/rollback
pendingPrepares map[uint64]bool
pendingPreparesLock sync.RWMutex
/// Runtime entities
// current node id.
nodeID proto.NodeID
// instance identifies kayak in multi-instance environment
// e.g. use database id for SQLChain scenario.
instanceID string
// wal defines the wal for kayak.
wal kt.Wal
// underlying handler
sh kt.Handler
/// Peers info
// peers defines the server peers.
peers *proto.Peers
// cached role of current node in peers, calculated from peers info.
role proto.ServerRole
// cached followers in peers, calculated from peers info.
followers []proto.NodeID
// peers lock for peers update logic.
peersLock sync.RWMutex
// calculated min follower nodes for prepare.
minPreparedFollowers int
// calculated min follower nodes for commit.
minCommitFollowers int
/// RPC related
// new caller functions: wrap for mocking testable purpose.
TrackerNewCallerFunc NewCallerFunc
WaiterNewCallerFunc NewCallerFunc
// service name for mux service.
serviceName string
// rpc method for apply requests.
applyRPCMethod string
// rpc method for startFetch requests.
fetchRPCMethod string
//// Parameters
// prepare threshold defines the minimum node count requirement for prepare operation.
prepareThreshold float64
// commit threshold defines the minimum node count requirement for commit operation.
commitThreshold float64
// prepare timeout defines the max allowed time for prepare operation.
prepareTimeout time.Duration
// commit timeout defines the max allowed time for commit operation.
commitTimeout time.Duration
// log wait timeout to startFetch missing logs.
logWaitTimeout time.Duration
// channel for awaiting commits.
commitCh chan *commitReq
waitLogMap sync.Map // map[uint64]*waitItem
/// Sub-routines management.
started uint32
stopCh chan struct{}
wg sync.WaitGroup
}
// commitReq defines the commit operation input.
type commitReq struct {
ctx context.Context
data interface{}
index uint64
lastCommit uint64
log *kt.Log
result *commitFuture
tm *timer.Timer
}
// commitResult defines the commit operation result.
type commitResult struct {
index uint64
result interface{}
err error
storageErr error
rpc *rpcTracker
}
type commitFuture struct {
ch chan *commitResult
}
func newCommitFuture() *commitFuture {
return &commitFuture{
ch: make(chan *commitResult, 1),
}
}
func (f *commitFuture) Get(ctx context.Context) (cr *commitResult, err error) {
if f == nil || f.ch == nil {
err = errors.Wrap(ctx.Err(), "enqueue commit timeout")
return
}
select {
case <-ctx.Done():
err = errors.Wrap(ctx.Err(), "get commit result timeout")
return
case cr = <-f.ch:
return
}
}
func (f *commitFuture) Set(cr *commitResult) {
select {
case f.ch <- cr:
default:
}
}
// NewRuntime creates new kayak Runtime.
func NewRuntime(cfg *kt.RuntimeConfig) (rt *Runtime, err error) {
if cfg == nil {
err = errors.Wrap(kt.ErrInvalidConfig, "nil config")
return
}
peers := cfg.Peers
if peers == nil {
err = errors.Wrap(kt.ErrInvalidConfig, "nil peers")
return
}
// verify peers
if err = peers.Verify(); err != nil {
err = errors.Wrap(err, "verify peers during kayak init failed")
return
}
followers := make([]proto.NodeID, 0, len(peers.Servers))
exists := false
var role proto.ServerRole
for _, v := range peers.Servers {
if !v.IsEqual(&peers.Leader) {
followers = append(followers, v)
}
if v.IsEqual(&cfg.NodeID) {
exists = true
if v.IsEqual(&peers.Leader) {
role = proto.Leader
} else {
role = proto.Follower
}
}
}
if !exists {
err = errors.Wrapf(kt.ErrNotInPeer, "node %v not in peers %v", cfg.NodeID, peers)
return
}
// calculate fan-out count according to threshold and peers info
minPreparedFollowers := int(math.Max(math.Ceil(cfg.PrepareThreshold*float64(len(peers.Servers))), 1) - 1)
minCommitFollowers := int(math.Max(math.Ceil(cfg.CommitThreshold*float64(len(peers.Servers))), 1) - 1)
rt = &Runtime{
// indexes
pendingPrepares: make(map[uint64]bool, commitWindow*2),
// handler and logs
sh: cfg.Handler,
wal: cfg.Wal,
instanceID: cfg.InstanceID,
// peers
peers: cfg.Peers,
nodeID: cfg.NodeID,
followers: followers,
role: role,
minPreparedFollowers: minPreparedFollowers,
minCommitFollowers: minCommitFollowers,
// rpc related
TrackerNewCallerFunc: defaultNewCallerFunc,
WaiterNewCallerFunc: defaultNewCallerFunc,
serviceName: cfg.ServiceName,
applyRPCMethod: cfg.ServiceName + "." + cfg.ApplyMethodName,
fetchRPCMethod: cfg.ServiceName + "." + cfg.FetchMethodName,
// commits related
prepareThreshold: cfg.PrepareThreshold,
prepareTimeout: cfg.PrepareTimeout,
commitThreshold: cfg.CommitThreshold,
commitTimeout: cfg.CommitTimeout,
logWaitTimeout: cfg.LogWaitTimeout,
commitCh: make(chan *commitReq, commitWindow),
// stop coordinator
stopCh: make(chan struct{}),
}
// read from pool to rebuild uncommitted log map
if err = rt.readLogs(); err != nil {
return
}
return
}
// Start starts the Runtime.
func (r *Runtime) Start() (err error) {
if !atomic.CompareAndSwapUint32(&r.started, 0, 1) {
return
}
// start commit cycle
r.goFunc(r.commitCycle)
return
}
// Shutdown waits for the Runtime to stop.
func (r *Runtime) Shutdown() (err error) {
if !atomic.CompareAndSwapUint32(&r.started, 1, 2) {
return
}
select {
case <-r.stopCh:
default:
close(r.stopCh)
}
r.wg.Wait()
return
}
// Apply defines entry for Leader node.
func (r *Runtime) Apply(ctx context.Context, req interface{}) (result interface{}, logIndex uint64, err error) {
if atomic.LoadUint32(&r.started) != 1 {
err = kt.ErrStopped
return
}
ctx, task := trace.NewTask(ctx, "Kayak.Apply")
defer task.End()
tm := timer.NewTimer()
defer func() {
log.WithField("r", logIndex).
WithFields(tm.ToLogFields()).
WithError(err).
Debug("kayak leader apply")
}()
r.peersLock.RLock()
defer r.peersLock.RUnlock()
tm.Add("peers_lock")
if r.role != proto.Leader {
// not leader
err = kt.ErrNotLeader
return
}
// prepare
prepareLog, err := r.doLeaderPrepare(ctx, tm, req)
if prepareLog != nil {
defer r.markPrepareFinished(ctx, prepareLog.Index)
}
if err == nil {
// commit
return r.doLeaderCommit(ctx, tm, prepareLog, req)
}
// rollback
if prepareLog != nil {
r.doLeaderRollback(ctx, tm, prepareLog)
}
return
}
// Fetch defines entry for missing log startFetch.
func (r *Runtime) Fetch(ctx context.Context, index uint64) (l *kt.Log, err error) {
if atomic.LoadUint32(&r.started) != 1 {
err = kt.ErrStopped
return
}
tm := timer.NewTimer()
defer func() {
log.WithField("l", index).
WithFields(tm.ToLogFields()).
WithError(err).
Debug("kayak log startFetch")
}()
r.peersLock.RLock()
defer r.peersLock.RUnlock()
tm.Add("peers_lock")
if r.role != proto.Leader {
// not leader
err = kt.ErrNotLeader
return
}
// wal get
return r.wal.Get(index)
}
// FollowerApply defines entry for follower node.
func (r *Runtime) FollowerApply(l *kt.Log) (err error) {
return r.followerApply(l, true)
}
// UpdatePeers defines entry for peers update logic.
func (r *Runtime) UpdatePeers(peers *proto.Peers) (err error) {
r.peersLock.Lock()
defer r.peersLock.Unlock()
return
}
func (r *Runtime) updateNextIndex(ctx context.Context, l *kt.Log) {
defer trace.StartRegion(ctx, "updateNextIndex").End()
r.nextIndexLock.Lock()
defer r.nextIndexLock.Unlock()
if r.nextIndex < l.Index+1 {
r.nextIndex = l.Index + 1
}
}
func (r *Runtime) checkIfPrepareFinished(ctx context.Context, index uint64) (finished bool) {
defer trace.StartRegion(ctx, "checkIfPrepareFinished").End()
r.pendingPreparesLock.RLock()
defer r.pendingPreparesLock.RUnlock()
return !r.pendingPrepares[index]
}
func (r *Runtime) markPendingPrepare(ctx context.Context, index uint64) {
defer trace.StartRegion(ctx, "markPendingPrepare").End()
r.pendingPreparesLock.Lock()
defer r.pendingPreparesLock.Unlock()
r.pendingPrepares[index] = true
}
func (r *Runtime) markPrepareFinished(ctx context.Context, index uint64) {
defer trace.StartRegion(ctx, "markPrepareFinished").End()
r.pendingPreparesLock.Lock()
defer r.pendingPreparesLock.Unlock()
delete(r.pendingPrepares, index)
}
func (r *Runtime) followerApply(l *kt.Log, checkPrepare bool) (err error) {
if l == nil {
err = errors.Wrap(kt.ErrInvalidLog, "log is nil")
return
}
if atomic.LoadUint32(&r.started) != 1 {
err = kt.ErrStopped
return
}
ctx, task := trace.NewTask(context.Background(), "Kayak.FollowerApply."+l.Type.String())
defer task.End()
tm := timer.NewTimer()
var storageErr error
defer func() {
log.
WithFields(log.Fields{
"t": l.Type.String(),
"i": l.Index,
"se": storageErr,
}).
WithFields(tm.ToLogFields()).
WithError(err).
Debug("kayak follower apply")
}()
r.peersLock.RLock()
defer r.peersLock.RUnlock()
tm.Add("peers_lock")
if r.role == proto.Leader {
// not follower
err = kt.ErrNotFollower
return
}
// verify log structure
switch l.Type {
case kt.LogPrepare:
err = r.followerPrepare(ctx, tm, l, checkPrepare)
case kt.LogRollback:
err = r.followerRollback(ctx, tm, l)
case kt.LogCommit:
storageErr, err = r.followerCommit(ctx, tm, l)
}
if err == nil {
r.updateNextIndex(ctx, l)
r.triggerLogAwaits(l)
}
return
}