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
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
|
/*
*
* Copyright 2014 gRPC 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 grpc
import (
"context"
"errors"
"fmt"
"math"
"net/url"
"reflect"
"strings"
"sync"
"sync/atomic"
"time"
"google.golang.org/grpc/balancer"
"google.golang.org/grpc/balancer/base"
"google.golang.org/grpc/codes"
"google.golang.org/grpc/connectivity"
"google.golang.org/grpc/credentials"
"google.golang.org/grpc/internal/backoff"
"google.golang.org/grpc/internal/channelz"
"google.golang.org/grpc/internal/grpcsync"
iresolver "google.golang.org/grpc/internal/resolver"
"google.golang.org/grpc/internal/transport"
"google.golang.org/grpc/keepalive"
"google.golang.org/grpc/resolver"
"google.golang.org/grpc/serviceconfig"
"google.golang.org/grpc/status"
_ "google.golang.org/grpc/balancer/roundrobin" // To register roundrobin.
_ "google.golang.org/grpc/internal/resolver/dns" // To register dns resolver.
_ "google.golang.org/grpc/internal/resolver/passthrough" // To register passthrough resolver.
_ "google.golang.org/grpc/internal/resolver/unix" // To register unix resolver.
)
const (
// minimum time to give a connection to complete
minConnectTimeout = 20 * time.Second
// must match grpclbName in grpclb/grpclb.go
grpclbName = "grpclb"
)
var (
// ErrClientConnClosing indicates that the operation is illegal because
// the ClientConn is closing.
//
// Deprecated: this error should not be relied upon by users; use the status
// code of Canceled instead.
ErrClientConnClosing = status.Error(codes.Canceled, "grpc: the client connection is closing")
// errConnDrain indicates that the connection starts to be drained and does not accept any new RPCs.
errConnDrain = errors.New("grpc: the connection is drained")
// errConnClosing indicates that the connection is closing.
errConnClosing = errors.New("grpc: the connection is closing")
// invalidDefaultServiceConfigErrPrefix is used to prefix the json parsing error for the default
// service config.
invalidDefaultServiceConfigErrPrefix = "grpc: the provided default service config is invalid"
)
// The following errors are returned from Dial and DialContext
var (
// errNoTransportSecurity indicates that there is no transport security
// being set for ClientConn. Users should either set one or explicitly
// call WithInsecure DialOption to disable security.
errNoTransportSecurity = errors.New("grpc: no transport security set (use grpc.WithInsecure() explicitly or set credentials)")
// errTransportCredsAndBundle indicates that creds bundle is used together
// with other individual Transport Credentials.
errTransportCredsAndBundle = errors.New("grpc: credentials.Bundle may not be used with individual TransportCredentials")
// errNoTransportCredsInBundle indicated that the configured creds bundle
// returned a transport credentials which was nil.
errNoTransportCredsInBundle = errors.New("grpc: credentials.Bundle must return non-nil transport credentials")
// errTransportCredentialsMissing indicates that users want to transmit
// security information (e.g., OAuth2 token) which requires secure
// connection on an insecure connection.
errTransportCredentialsMissing = errors.New("grpc: the credentials require transport level security (use grpc.WithTransportCredentials() to set)")
)
const (
defaultClientMaxReceiveMessageSize = 1024 * 1024 * 4
defaultClientMaxSendMessageSize = math.MaxInt32
// http2IOBufSize specifies the buffer size for sending frames.
defaultWriteBufSize = 32 * 1024
defaultReadBufSize = 32 * 1024
)
// Dial creates a client connection to the given target.
func Dial(target string, opts ...DialOption) (*ClientConn, error) {
return DialContext(context.Background(), target, opts...)
}
type defaultConfigSelector struct {
sc *ServiceConfig
}
func (dcs *defaultConfigSelector) SelectConfig(rpcInfo iresolver.RPCInfo) (*iresolver.RPCConfig, error) {
return &iresolver.RPCConfig{
Context: rpcInfo.Context,
MethodConfig: getMethodConfig(dcs.sc, rpcInfo.Method),
}, nil
}
// DialContext creates a client connection to the given target. By default, it's
// a non-blocking dial (the function won't wait for connections to be
// established, and connecting happens in the background). To make it a blocking
// dial, use WithBlock() dial option.
//
// In the non-blocking case, the ctx does not act against the connection. It
// only controls the setup steps.
//
// In the blocking case, ctx can be used to cancel or expire the pending
// connection. Once this function returns, the cancellation and expiration of
// ctx will be noop. Users should call ClientConn.Close to terminate all the
// pending operations after this function returns.
//
// The target name syntax is defined in
// https://github.com/grpc/grpc/blob/master/doc/naming.md.
// e.g. to use dns resolver, a "dns:///" prefix should be applied to the target.
func DialContext(ctx context.Context, target string, opts ...DialOption) (conn *ClientConn, err error) {
cc := &ClientConn{
target: target,
csMgr: &connectivityStateManager{},
conns: make(map[*addrConn]struct{}),
dopts: defaultDialOptions(),
blockingpicker: newPickerWrapper(),
czData: new(channelzData),
firstResolveEvent: grpcsync.NewEvent(),
}
cc.retryThrottler.Store((*retryThrottler)(nil))
cc.safeConfigSelector.UpdateConfigSelector(&defaultConfigSelector{nil})
cc.ctx, cc.cancel = context.WithCancel(context.Background())
for _, opt := range opts {
opt.apply(&cc.dopts)
}
chainUnaryClientInterceptors(cc)
chainStreamClientInterceptors(cc)
defer func() {
if err != nil {
cc.Close()
}
}()
if channelz.IsOn() {
if cc.dopts.channelzParentID != 0 {
cc.channelzID = channelz.RegisterChannel(&channelzChannel{cc}, cc.dopts.channelzParentID, target)
channelz.AddTraceEvent(logger, cc.channelzID, 0, &channelz.TraceEventDesc{
Desc: "Channel Created",
Severity: channelz.CtInfo,
Parent: &channelz.TraceEventDesc{
Desc: fmt.Sprintf("Nested Channel(id:%d) created", cc.channelzID),
Severity: channelz.CtInfo,
},
})
} else {
cc.channelzID = channelz.RegisterChannel(&channelzChannel{cc}, 0, target)
channelz.Info(logger, cc.channelzID, "Channel Created")
}
cc.csMgr.channelzID = cc.channelzID
}
if cc.dopts.copts.TransportCredentials == nil && cc.dopts.copts.CredsBundle == nil {
return nil, errNoTransportSecurity
}
if cc.dopts.copts.TransportCredentials != nil && cc.dopts.copts.CredsBundle != nil {
return nil, errTransportCredsAndBundle
}
if cc.dopts.copts.CredsBundle != nil && cc.dopts.copts.CredsBundle.TransportCredentials() == nil {
return nil, errNoTransportCredsInBundle
}
transportCreds := cc.dopts.copts.TransportCredentials
if transportCreds == nil {
transportCreds = cc.dopts.copts.CredsBundle.TransportCredentials()
}
if transportCreds.Info().SecurityProtocol == "insecure" {
for _, cd := range cc.dopts.copts.PerRPCCredentials {
if cd.RequireTransportSecurity() {
return nil, errTransportCredentialsMissing
}
}
}
if cc.dopts.defaultServiceConfigRawJSON != nil {
scpr := parseServiceConfig(*cc.dopts.defaultServiceConfigRawJSON)
if scpr.Err != nil {
return nil, fmt.Errorf("%s: %v", invalidDefaultServiceConfigErrPrefix, scpr.Err)
}
cc.dopts.defaultServiceConfig, _ = scpr.Config.(*ServiceConfig)
}
cc.mkp = cc.dopts.copts.KeepaliveParams
if cc.dopts.copts.UserAgent != "" {
cc.dopts.copts.UserAgent += " " + grpcUA
} else {
cc.dopts.copts.UserAgent = grpcUA
}
if cc.dopts.timeout > 0 {
var cancel context.CancelFunc
ctx, cancel = context.WithTimeout(ctx, cc.dopts.timeout)
defer cancel()
}
defer func() {
select {
case <-ctx.Done():
switch {
case ctx.Err() == err:
conn = nil
case err == nil || !cc.dopts.returnLastError:
conn, err = nil, ctx.Err()
default:
conn, err = nil, fmt.Errorf("%v: %v", ctx.Err(), err)
}
default:
}
}()
scSet := false
if cc.dopts.scChan != nil {
// Try to get an initial service config.
select {
case sc, ok := <-cc.dopts.scChan:
if ok {
cc.sc = &sc
cc.safeConfigSelector.UpdateConfigSelector(&defaultConfigSelector{&sc})
scSet = true
}
default:
}
}
if cc.dopts.bs == nil {
cc.dopts.bs = backoff.DefaultExponential
}
// Determine the resolver to use.
resolverBuilder, err := cc.parseTargetAndFindResolver()
if err != nil {
return nil, err
}
cc.authority, err = determineAuthority(cc.parsedTarget.Endpoint, cc.target, cc.dopts)
if err != nil {
return nil, err
}
channelz.Infof(logger, cc.channelzID, "Channel authority set to %q", cc.authority)
if cc.dopts.scChan != nil && !scSet {
// Blocking wait for the initial service config.
select {
case sc, ok := <-cc.dopts.scChan:
if ok {
cc.sc = &sc
cc.safeConfigSelector.UpdateConfigSelector(&defaultConfigSelector{&sc})
}
case <-ctx.Done():
return nil, ctx.Err()
}
}
if cc.dopts.scChan != nil {
go cc.scWatcher()
}
var credsClone credentials.TransportCredentials
if creds := cc.dopts.copts.TransportCredentials; creds != nil {
credsClone = creds.Clone()
}
cc.balancerBuildOpts = balancer.BuildOptions{
DialCreds: credsClone,
CredsBundle: cc.dopts.copts.CredsBundle,
Dialer: cc.dopts.copts.Dialer,
Authority: cc.authority,
CustomUserAgent: cc.dopts.copts.UserAgent,
ChannelzParentID: cc.channelzID,
Target: cc.parsedTarget,
}
// Build the resolver.
rWrapper, err := newCCResolverWrapper(cc, resolverBuilder)
if err != nil {
return nil, fmt.Errorf("failed to build resolver: %v", err)
}
cc.mu.Lock()
cc.resolverWrapper = rWrapper
cc.mu.Unlock()
// A blocking dial blocks until the clientConn is ready.
if cc.dopts.block {
for {
cc.Connect()
s := cc.GetState()
if s == connectivity.Ready {
break
} else if cc.dopts.copts.FailOnNonTempDialError && s == connectivity.TransientFailure {
if err = cc.connectionError(); err != nil {
terr, ok := err.(interface {
Temporary() bool
})
if ok && !terr.Temporary() {
return nil, err
}
}
}
if !cc.WaitForStateChange(ctx, s) {
// ctx got timeout or canceled.
if err = cc.connectionError(); err != nil && cc.dopts.returnLastError {
return nil, err
}
return nil, ctx.Err()
}
}
}
return cc, nil
}
// chainUnaryClientInterceptors chains all unary client interceptors into one.
func chainUnaryClientInterceptors(cc *ClientConn) {
interceptors := cc.dopts.chainUnaryInts
// Prepend dopts.unaryInt to the chaining interceptors if it exists, since unaryInt will
// be executed before any other chained interceptors.
if cc.dopts.unaryInt != nil {
interceptors = append([]UnaryClientInterceptor{cc.dopts.unaryInt}, interceptors...)
}
var chainedInt UnaryClientInterceptor
if len(interceptors) == 0 {
chainedInt = nil
} else if len(interceptors) == 1 {
chainedInt = interceptors[0]
} else {
chainedInt = func(ctx context.Context, method string, req, reply interface{}, cc *ClientConn, invoker UnaryInvoker, opts ...CallOption) error {
return interceptors[0](ctx, method, req, reply, cc, getChainUnaryInvoker(interceptors, 0, invoker), opts...)
}
}
cc.dopts.unaryInt = chainedInt
}
// getChainUnaryInvoker recursively generate the chained unary invoker.
func getChainUnaryInvoker(interceptors []UnaryClientInterceptor, curr int, finalInvoker UnaryInvoker) UnaryInvoker {
if curr == len(interceptors)-1 {
return finalInvoker
}
return func(ctx context.Context, method string, req, reply interface{}, cc *ClientConn, opts ...CallOption) error {
return interceptors[curr+1](ctx, method, req, reply, cc, getChainUnaryInvoker(interceptors, curr+1, finalInvoker), opts...)
}
}
// chainStreamClientInterceptors chains all stream client interceptors into one.
func chainStreamClientInterceptors(cc *ClientConn) {
interceptors := cc.dopts.chainStreamInts
// Prepend dopts.streamInt to the chaining interceptors if it exists, since streamInt will
// be executed before any other chained interceptors.
if cc.dopts.streamInt != nil {
interceptors = append([]StreamClientInterceptor{cc.dopts.streamInt}, interceptors...)
}
var chainedInt StreamClientInterceptor
if len(interceptors) == 0 {
chainedInt = nil
} else if len(interceptors) == 1 {
chainedInt = interceptors[0]
} else {
chainedInt = func(ctx context.Context, desc *StreamDesc, cc *ClientConn, method string, streamer Streamer, opts ...CallOption) (ClientStream, error) {
return interceptors[0](ctx, desc, cc, method, getChainStreamer(interceptors, 0, streamer), opts...)
}
}
cc.dopts.streamInt = chainedInt
}
// getChainStreamer recursively generate the chained client stream constructor.
func getChainStreamer(interceptors []StreamClientInterceptor, curr int, finalStreamer Streamer) Streamer {
if curr == len(interceptors)-1 {
return finalStreamer
}
return func(ctx context.Context, desc *StreamDesc, cc *ClientConn, method string, opts ...CallOption) (ClientStream, error) {
return interceptors[curr+1](ctx, desc, cc, method, getChainStreamer(interceptors, curr+1, finalStreamer), opts...)
}
}
// connectivityStateManager keeps the connectivity.State of ClientConn.
// This struct will eventually be exported so the balancers can access it.
type connectivityStateManager struct {
mu sync.Mutex
state connectivity.State
notifyChan chan struct{}
channelzID int64
}
// updateState updates the connectivity.State of ClientConn.
// If there's a change it notifies goroutines waiting on state change to
// happen.
func (csm *connectivityStateManager) updateState(state connectivity.State) {
csm.mu.Lock()
defer csm.mu.Unlock()
if csm.state == connectivity.Shutdown {
return
}
if csm.state == state {
return
}
csm.state = state
channelz.Infof(logger, csm.channelzID, "Channel Connectivity change to %v", state)
if csm.notifyChan != nil {
// There are other goroutines waiting on this channel.
close(csm.notifyChan)
csm.notifyChan = nil
}
}
func (csm *connectivityStateManager) getState() connectivity.State {
csm.mu.Lock()
defer csm.mu.Unlock()
return csm.state
}
func (csm *connectivityStateManager) getNotifyChan() <-chan struct{} {
csm.mu.Lock()
defer csm.mu.Unlock()
if csm.notifyChan == nil {
csm.notifyChan = make(chan struct{})
}
return csm.notifyChan
}
// ClientConnInterface defines the functions clients need to perform unary and
// streaming RPCs. It is implemented by *ClientConn, and is only intended to
// be referenced by generated code.
type ClientConnInterface interface {
// Invoke performs a unary RPC and returns after the response is received
// into reply.
Invoke(ctx context.Context, method string, args interface{}, reply interface{}, opts ...CallOption) error
// NewStream begins a streaming RPC.
NewStream(ctx context.Context, desc *StreamDesc, method string, opts ...CallOption) (ClientStream, error)
}
// Assert *ClientConn implements ClientConnInterface.
var _ ClientConnInterface = (*ClientConn)(nil)
// ClientConn represents a virtual connection to a conceptual endpoint, to
// perform RPCs.
//
// A ClientConn is free to have zero or more actual connections to the endpoint
// based on configuration, load, etc. It is also free to determine which actual
// endpoints to use and may change it every RPC, permitting client-side load
// balancing.
//
// A ClientConn encapsulates a range of functionality including name
// resolution, TCP connection establishment (with retries and backoff) and TLS
// handshakes. It also handles errors on established connections by
// re-resolving the name and reconnecting.
type ClientConn struct {
ctx context.Context
cancel context.CancelFunc
target string
parsedTarget resolver.Target
authority string
dopts dialOptions
csMgr *connectivityStateManager
balancerBuildOpts balancer.BuildOptions
blockingpicker *pickerWrapper
safeConfigSelector iresolver.SafeConfigSelector
mu sync.RWMutex
resolverWrapper *ccResolverWrapper
sc *ServiceConfig
conns map[*addrConn]struct{}
// Keepalive parameter can be updated if a GoAway is received.
mkp keepalive.ClientParameters
curBalancerName string
balancerWrapper *ccBalancerWrapper
retryThrottler atomic.Value
firstResolveEvent *grpcsync.Event
channelzID int64 // channelz unique identification number
czData *channelzData
lceMu sync.Mutex // protects lastConnectionError
lastConnectionError error
}
// WaitForStateChange waits until the connectivity.State of ClientConn changes from sourceState or
// ctx expires. A true value is returned in former case and false in latter.
//
// Experimental
//
// Notice: This API is EXPERIMENTAL and may be changed or removed in a
// later release.
func (cc *ClientConn) WaitForStateChange(ctx context.Context, sourceState connectivity.State) bool {
ch := cc.csMgr.getNotifyChan()
if cc.csMgr.getState() != sourceState {
return true
}
select {
case <-ctx.Done():
return false
case <-ch:
return true
}
}
// GetState returns the connectivity.State of ClientConn.
//
// Experimental
//
// Notice: This API is EXPERIMENTAL and may be changed or removed in a later
// release.
func (cc *ClientConn) GetState() connectivity.State {
return cc.csMgr.getState()
}
// Connect causes all subchannels in the ClientConn to attempt to connect if
// the channel is idle. Does not wait for the connection attempts to begin
// before returning.
//
// Experimental
//
// Notice: This API is EXPERIMENTAL and may be changed or removed in a later
// release.
func (cc *ClientConn) Connect() {
cc.mu.Lock()
defer cc.mu.Unlock()
if cc.balancerWrapper != nil && cc.balancerWrapper.exitIdle() {
return
}
for ac := range cc.conns {
go ac.connect()
}
}
func (cc *ClientConn) scWatcher() {
for {
select {
case sc, ok := <-cc.dopts.scChan:
if !ok {
return
}
cc.mu.Lock()
// TODO: load balance policy runtime change is ignored.
// We may revisit this decision in the future.
cc.sc = &sc
cc.safeConfigSelector.UpdateConfigSelector(&defaultConfigSelector{&sc})
cc.mu.Unlock()
case <-cc.ctx.Done():
return
}
}
}
// waitForResolvedAddrs blocks until the resolver has provided addresses or the
// context expires. Returns nil unless the context expires first; otherwise
// returns a status error based on the context.
func (cc *ClientConn) waitForResolvedAddrs(ctx context.Context) error {
// This is on the RPC path, so we use a fast path to avoid the
// more-expensive "select" below after the resolver has returned once.
if cc.firstResolveEvent.HasFired() {
return nil
}
select {
case <-cc.firstResolveEvent.Done():
return nil
case <-ctx.Done():
return status.FromContextError(ctx.Err()).Err()
case <-cc.ctx.Done():
return ErrClientConnClosing
}
}
var emptyServiceConfig *ServiceConfig
func init() {
cfg := parseServiceConfig("{}")
if cfg.Err != nil {
panic(fmt.Sprintf("impossible error parsing empty service config: %v", cfg.Err))
}
emptyServiceConfig = cfg.Config.(*ServiceConfig)
}
func (cc *ClientConn) maybeApplyDefaultServiceConfig(addrs []resolver.Address) {
if cc.sc != nil {
cc.applyServiceConfigAndBalancer(cc.sc, nil, addrs)
return
}
if cc.dopts.defaultServiceConfig != nil {
cc.applyServiceConfigAndBalancer(cc.dopts.defaultServiceConfig, &defaultConfigSelector{cc.dopts.defaultServiceConfig}, addrs)
} else {
cc.applyServiceConfigAndBalancer(emptyServiceConfig, &defaultConfigSelector{emptyServiceConfig}, addrs)
}
}
func (cc *ClientConn) updateResolverState(s resolver.State, err error) error {
defer cc.firstResolveEvent.Fire()
cc.mu.Lock()
// Check if the ClientConn is already closed. Some fields (e.g.
// balancerWrapper) are set to nil when closing the ClientConn, and could
// cause nil pointer panic if we don't have this check.
if cc.conns == nil {
cc.mu.Unlock()
return nil
}
if err != nil {
// May need to apply the initial service config in case the resolver
// doesn't support service configs, or doesn't provide a service config
// with the new addresses.
cc.maybeApplyDefaultServiceConfig(nil)
if cc.balancerWrapper != nil {
cc.balancerWrapper.resolverError(err)
}
// No addresses are valid with err set; return early.
cc.mu.Unlock()
return balancer.ErrBadResolverState
}
var ret error
if cc.dopts.disableServiceConfig {
channelz.Infof(logger, cc.channelzID, "ignoring service config from resolver (%v) and applying the default because service config is disabled", s.ServiceConfig)
cc.maybeApplyDefaultServiceConfig(s.Addresses)
} else if s.ServiceConfig == nil {
cc.maybeApplyDefaultServiceConfig(s.Addresses)
// TODO: do we need to apply a failing LB policy if there is no
// default, per the error handling design?
} else {
if sc, ok := s.ServiceConfig.Config.(*ServiceConfig); s.ServiceConfig.Err == nil && ok {
configSelector := iresolver.GetConfigSelector(s)
if configSelector != nil {
if len(s.ServiceConfig.Config.(*ServiceConfig).Methods) != 0 {
channelz.Infof(logger, cc.channelzID, "method configs in service config will be ignored due to presence of config selector")
}
} else {
configSelector = &defaultConfigSelector{sc}
}
cc.applyServiceConfigAndBalancer(sc, configSelector, s.Addresses)
} else {
ret = balancer.ErrBadResolverState
if cc.balancerWrapper == nil {
var err error
if s.ServiceConfig.Err != nil {
err = status.Errorf(codes.Unavailable, "error parsing service config: %v", s.ServiceConfig.Err)
} else {
err = status.Errorf(codes.Unavailable, "illegal service config type: %T", s.ServiceConfig.Config)
}
cc.safeConfigSelector.UpdateConfigSelector(&defaultConfigSelector{cc.sc})
cc.blockingpicker.updatePicker(base.NewErrPicker(err))
cc.csMgr.updateState(connectivity.TransientFailure)
cc.mu.Unlock()
return ret
}
}
}
var balCfg serviceconfig.LoadBalancingConfig
if cc.dopts.balancerBuilder == nil && cc.sc != nil && cc.sc.lbConfig != nil {
balCfg = cc.sc.lbConfig.cfg
}
cbn := cc.curBalancerName
bw := cc.balancerWrapper
cc.mu.Unlock()
if cbn != grpclbName {
// Filter any grpclb addresses since we don't have the grpclb balancer.
for i := 0; i < len(s.Addresses); {
if s.Addresses[i].Type == resolver.GRPCLB {
copy(s.Addresses[i:], s.Addresses[i+1:])
s.Addresses = s.Addresses[:len(s.Addresses)-1]
continue
}
i++
}
}
uccsErr := bw.updateClientConnState(&balancer.ClientConnState{ResolverState: s, BalancerConfig: balCfg})
if ret == nil {
ret = uccsErr // prefer ErrBadResolver state since any other error is
// currently meaningless to the caller.
}
return ret
}
// switchBalancer starts the switching from current balancer to the balancer
// with the given name.
//
// It will NOT send the current address list to the new balancer. If needed,
// caller of this function should send address list to the new balancer after
// this function returns.
//
// Caller must hold cc.mu.
func (cc *ClientConn) switchBalancer(name string) {
if strings.EqualFold(cc.curBalancerName, name) {
return
}
channelz.Infof(logger, cc.channelzID, "ClientConn switching balancer to %q", name)
if cc.dopts.balancerBuilder != nil {
channelz.Info(logger, cc.channelzID, "ignoring balancer switching: Balancer DialOption used instead")
return
}
if cc.balancerWrapper != nil {
// Don't hold cc.mu while closing the balancers. The balancers may call
// methods that require cc.mu (e.g. cc.NewSubConn()). Holding the mutex
// would cause a deadlock in that case.
cc.mu.Unlock()
cc.balancerWrapper.close()
cc.mu.Lock()
}
builder := balancer.Get(name)
if builder == nil {
channelz.Warningf(logger, cc.channelzID, "Channel switches to new LB policy %q due to fallback from invalid balancer name", PickFirstBalancerName)
channelz.Infof(logger, cc.channelzID, "failed to get balancer builder for: %v, using pick_first instead", name)
builder = newPickfirstBuilder()
} else {
channelz.Infof(logger, cc.channelzID, "Channel switches to new LB policy %q", name)
}
cc.curBalancerName = builder.Name()
cc.balancerWrapper = newCCBalancerWrapper(cc, builder, cc.balancerBuildOpts)
}
func (cc *ClientConn) handleSubConnStateChange(sc balancer.SubConn, s connectivity.State, err error) {
cc.mu.Lock()
if cc.conns == nil {
cc.mu.Unlock()
return
}
// TODO(bar switching) send updates to all balancer wrappers when balancer
// gracefully switching is supported.
cc.balancerWrapper.handleSubConnStateChange(sc, s, err)
cc.mu.Unlock()
}
// newAddrConn creates an addrConn for addrs and adds it to cc.conns.
//
// Caller needs to make sure len(addrs) > 0.
func (cc *ClientConn) newAddrConn(addrs []resolver.Address, opts balancer.NewSubConnOptions) (*addrConn, error) {
ac := &addrConn{
state: connectivity.Idle,
cc: cc,
addrs: addrs,
scopts: opts,
dopts: cc.dopts,
czData: new(channelzData),
resetBackoff: make(chan struct{}),
}
ac.ctx, ac.cancel = context.WithCancel(cc.ctx)
// Track ac in cc. This needs to be done before any getTransport(...) is called.
cc.mu.Lock()
if cc.conns == nil {
cc.mu.Unlock()
return nil, ErrClientConnClosing
}
if channelz.IsOn() {
ac.channelzID = channelz.RegisterSubChannel(ac, cc.channelzID, "")
channelz.AddTraceEvent(logger, ac.channelzID, 0, &channelz.TraceEventDesc{
Desc: "Subchannel Created",
Severity: channelz.CtInfo,
Parent: &channelz.TraceEventDesc{
Desc: fmt.Sprintf("Subchannel(id:%d) created", ac.channelzID),
Severity: channelz.CtInfo,
},
})
}
cc.conns[ac] = struct{}{}
cc.mu.Unlock()
return ac, nil
}
// removeAddrConn removes the addrConn in the subConn from clientConn.
// It also tears down the ac with the given error.
func (cc *ClientConn) removeAddrConn(ac *addrConn, err error) {
cc.mu.Lock()
if cc.conns == nil {
cc.mu.Unlock()
return
}
delete(cc.conns, ac)
cc.mu.Unlock()
ac.tearDown(err)
}
func (cc *ClientConn) channelzMetric() *channelz.ChannelInternalMetric {
return &channelz.ChannelInternalMetric{
State: cc.GetState(),
Target: cc.target,
CallsStarted: atomic.LoadInt64(&cc.czData.callsStarted),
CallsSucceeded: atomic.LoadInt64(&cc.czData.callsSucceeded),
CallsFailed: atomic.LoadInt64(&cc.czData.callsFailed),
LastCallStartedTimestamp: time.Unix(0, atomic.LoadInt64(&cc.czData.lastCallStartedTime)),
}
}
// Target returns the target string of the ClientConn.
//
// Experimental
//
// Notice: This API is EXPERIMENTAL and may be changed or removed in a
// later release.
func (cc *ClientConn) Target() string {
return cc.target
}
func (cc *ClientConn) incrCallsStarted() {
atomic.AddInt64(&cc.czData.callsStarted, 1)
atomic.StoreInt64(&cc.czData.lastCallStartedTime, time.Now().UnixNano())
}
func (cc *ClientConn) incrCallsSucceeded() {
atomic.AddInt64(&cc.czData.callsSucceeded, 1)
}
func (cc *ClientConn) incrCallsFailed() {
atomic.AddInt64(&cc.czData.callsFailed, 1)
}
// connect starts creating a transport.
// It does nothing if the ac is not IDLE.
// TODO(bar) Move this to the addrConn section.
func (ac *addrConn) connect() error {
ac.mu.Lock()
if ac.state == connectivity.Shutdown {
ac.mu.Unlock()
return errConnClosing
}
if ac.state != connectivity.Idle {
ac.mu.Unlock()
return nil
}
// Update connectivity state within the lock to prevent subsequent or
// concurrent calls from resetting the transport more than once.
ac.updateConnectivityState(connectivity.Connecting, nil)
ac.mu.Unlock()
ac.resetTransport()
return nil
}
// tryUpdateAddrs tries to update ac.addrs with the new addresses list.
//
// If ac is Connecting, it returns false. The caller should tear down the ac and
// create a new one. Note that the backoff will be reset when this happens.
//
// If ac is TransientFailure, it updates ac.addrs and returns true. The updated
// addresses will be picked up by retry in the next iteration after backoff.
//
// If ac is Shutdown or Idle, it updates ac.addrs and returns true.
//
// If ac is Ready, it checks whether current connected address of ac is in the
// new addrs list.
// - If true, it updates ac.addrs and returns true. The ac will keep using
// the existing connection.
// - If false, it does nothing and returns false.
func (ac *addrConn) tryUpdateAddrs(addrs []resolver.Address) bool {
ac.mu.Lock()
defer ac.mu.Unlock()
channelz.Infof(logger, ac.channelzID, "addrConn: tryUpdateAddrs curAddr: %v, addrs: %v", ac.curAddr, addrs)
if ac.state == connectivity.Shutdown ||
ac.state == connectivity.TransientFailure ||
ac.state == connectivity.Idle {
ac.addrs = addrs
return true
}
if ac.state == connectivity.Connecting {
return false
}
// ac.state is Ready, try to find the connected address.
var curAddrFound bool
for _, a := range addrs {
a.ServerName = ac.cc.getServerName(a)
if reflect.DeepEqual(ac.curAddr, a) {
curAddrFound = true
break
}
}
channelz.Infof(logger, ac.channelzID, "addrConn: tryUpdateAddrs curAddrFound: %v", curAddrFound)
if curAddrFound {
ac.addrs = addrs
}
return curAddrFound
}
// getServerName determines the serverName to be used in the connection
// handshake. The default value for the serverName is the authority on the
// ClientConn, which either comes from the user's dial target or through an
// authority override specified using the WithAuthority dial option. Name
// resolvers can specify a per-address override for the serverName through the
// resolver.Address.ServerName field which is used only if the WithAuthority
// dial option was not used. The rationale is that per-address authority
// overrides specified by the name resolver can represent a security risk, while
// an override specified by the user is more dependable since they probably know
// what they are doing.
func (cc *ClientConn) getServerName(addr resolver.Address) string {
if cc.dopts.authority != "" {
return cc.dopts.authority
}
if addr.ServerName != "" {
return addr.ServerName
}
return cc.authority
}
func getMethodConfig(sc *ServiceConfig, method string) MethodConfig {
if sc == nil {
return MethodConfig{}
}
if m, ok := sc.Methods[method]; ok {
return m
}
i := strings.LastIndex(method, "/")
if m, ok := sc.Methods[method[:i+1]]; ok {
return m
}
return sc.Methods[""]
}
// GetMethodConfig gets the method config of the input method.
// If there's an exact match for input method (i.e. /service/method), we return
// the corresponding MethodConfig.
// If there isn't an exact match for the input method, we look for the service's default
// config under the service (i.e /service/) and then for the default for all services (empty string).
//
// If there is a default MethodConfig for the service, we return it.
// Otherwise, we return an empty MethodConfig.
func (cc *ClientConn) GetMethodConfig(method string) MethodConfig {
// TODO: Avoid the locking here.
cc.mu.RLock()
defer cc.mu.RUnlock()
return getMethodConfig(cc.sc, method)
}
func (cc *ClientConn) healthCheckConfig() *healthCheckConfig {
cc.mu.RLock()
defer cc.mu.RUnlock()
if cc.sc == nil {
return nil
}
return cc.sc.healthCheckConfig
}
func (cc *ClientConn) getTransport(ctx context.Context, failfast bool, method string) (transport.ClientTransport, func(balancer.DoneInfo), error) {
t, done, err := cc.blockingpicker.pick(ctx, failfast, balancer.PickInfo{
Ctx: ctx,
FullMethodName: method,
})
if err != nil {
return nil, nil, toRPCErr(err)
}
return t, done, nil
}
func (cc *ClientConn) applyServiceConfigAndBalancer(sc *ServiceConfig, configSelector iresolver.ConfigSelector, addrs []resolver.Address) {
if sc == nil {
// should never reach here.
return
}
cc.sc = sc
if configSelector != nil {
cc.safeConfigSelector.UpdateConfigSelector(configSelector)
}
if cc.sc.retryThrottling != nil {
newThrottler := &retryThrottler{
tokens: cc.sc.retryThrottling.MaxTokens,
max: cc.sc.retryThrottling.MaxTokens,
thresh: cc.sc.retryThrottling.MaxTokens / 2,
ratio: cc.sc.retryThrottling.TokenRatio,
}
cc.retryThrottler.Store(newThrottler)
} else {
cc.retryThrottler.Store((*retryThrottler)(nil))
}
if cc.dopts.balancerBuilder == nil {
// Only look at balancer types and switch balancer if balancer dial
// option is not set.
var newBalancerName string
if cc.sc != nil && cc.sc.lbConfig != nil {
newBalancerName = cc.sc.lbConfig.name
} else {
var isGRPCLB bool
for _, a := range addrs {
if a.Type == resolver.GRPCLB {
isGRPCLB = true
break
}
}
if isGRPCLB {
newBalancerName = grpclbName
} else if cc.sc != nil && cc.sc.LB != nil {
newBalancerName = *cc.sc.LB
} else {
newBalancerName = PickFirstBalancerName
}
}
cc.switchBalancer(newBalancerName)
} else if cc.balancerWrapper == nil {
// Balancer dial option was set, and this is the first time handling
// resolved addresses. Build a balancer with dopts.balancerBuilder.
cc.curBalancerName = cc.dopts.balancerBuilder.Name()
cc.balancerWrapper = newCCBalancerWrapper(cc, cc.dopts.balancerBuilder, cc.balancerBuildOpts)
}
}
func (cc *ClientConn) resolveNow(o resolver.ResolveNowOptions) {
cc.mu.RLock()
r := cc.resolverWrapper
cc.mu.RUnlock()
if r == nil {
return
}
go r.resolveNow(o)
}
// ResetConnectBackoff wakes up all subchannels in transient failure and causes
// them to attempt another connection immediately. It also resets the backoff
// times used for subsequent attempts regardless of the current state.
//
// In general, this function should not be used. Typical service or network
// outages result in a reasonable client reconnection strategy by default.
// However, if a previously unavailable network becomes available, this may be
// used to trigger an immediate reconnect.
//
// Experimental
//
// Notice: This API is EXPERIMENTAL and may be changed or removed in a
// later release.
func (cc *ClientConn) ResetConnectBackoff() {
cc.mu.Lock()
conns := cc.conns
cc.mu.Unlock()
for ac := range conns {
ac.resetConnectBackoff()
}
}
// Close tears down the ClientConn and all underlying connections.
func (cc *ClientConn) Close() error {
defer cc.cancel()
cc.mu.Lock()
if cc.conns == nil {
cc.mu.Unlock()
return ErrClientConnClosing
}
conns := cc.conns
cc.conns = nil
cc.csMgr.updateState(connectivity.Shutdown)
rWrapper := cc.resolverWrapper
cc.resolverWrapper = nil
bWrapper := cc.balancerWrapper
cc.balancerWrapper = nil
cc.mu.Unlock()
cc.blockingpicker.close()
if bWrapper != nil {
bWrapper.close()
}
if rWrapper != nil {
rWrapper.close()
}
for ac := range conns {
ac.tearDown(ErrClientConnClosing)
}
if channelz.IsOn() {
ted := &channelz.TraceEventDesc{
Desc: "Channel Deleted",
Severity: channelz.CtInfo,
}
if cc.dopts.channelzParentID != 0 {
ted.Parent = &channelz.TraceEventDesc{
Desc: fmt.Sprintf("Nested channel(id:%d) deleted", cc.channelzID),
Severity: channelz.CtInfo,
}
}
channelz.AddTraceEvent(logger, cc.channelzID, 0, ted)
// TraceEvent needs to be called before RemoveEntry, as TraceEvent may add trace reference to
// the entity being deleted, and thus prevent it from being deleted right away.
channelz.RemoveEntry(cc.channelzID)
}
return nil
}
// addrConn is a network connection to a given address.
type addrConn struct {
ctx context.Context
cancel context.CancelFunc
cc *ClientConn
dopts dialOptions
acbw balancer.SubConn
scopts balancer.NewSubConnOptions
// transport is set when there's a viable transport (note: ac state may not be READY as LB channel
// health checking may require server to report healthy to set ac to READY), and is reset
// to nil when the current transport should no longer be used to create a stream (e.g. after GoAway
// is received, transport is closed, ac has been torn down).
transport transport.ClientTransport // The current transport.
mu sync.Mutex
curAddr resolver.Address // The current address.
addrs []resolver.Address // All addresses that the resolver resolved to.
// Use updateConnectivityState for updating addrConn's connectivity state.
state connectivity.State
backoffIdx int // Needs to be stateful for resetConnectBackoff.
resetBackoff chan struct{}
channelzID int64 // channelz unique identification number.
czData *channelzData
}
// Note: this requires a lock on ac.mu.
func (ac *addrConn) updateConnectivityState(s connectivity.State, lastErr error) {
if ac.state == s {
return
}
ac.state = s
channelz.Infof(logger, ac.channelzID, "Subchannel Connectivity change to %v", s)
ac.cc.handleSubConnStateChange(ac.acbw, s, lastErr)
}
// adjustParams updates parameters used to create transports upon
// receiving a GoAway.
func (ac *addrConn) adjustParams(r transport.GoAwayReason) {
switch r {
case transport.GoAwayTooManyPings:
v := 2 * ac.dopts.copts.KeepaliveParams.Time
ac.cc.mu.Lock()
if v > ac.cc.mkp.Time {
ac.cc.mkp.Time = v
}
ac.cc.mu.Unlock()
}
}
func (ac *addrConn) resetTransport() {
ac.mu.Lock()
if ac.state == connectivity.Shutdown {
ac.mu.Unlock()
return
}
addrs := ac.addrs
backoffFor := ac.dopts.bs.Backoff(ac.backoffIdx)
// This will be the duration that dial gets to finish.
dialDuration := minConnectTimeout
if ac.dopts.minConnectTimeout != nil {
dialDuration = ac.dopts.minConnectTimeout()
}
if dialDuration < backoffFor {
// Give dial more time as we keep failing to connect.
dialDuration = backoffFor
}
// We can potentially spend all the time trying the first address, and
// if the server accepts the connection and then hangs, the following
// addresses will never be tried.
//
// The spec doesn't mention what should be done for multiple addresses.
// https://github.com/grpc/grpc/blob/master/doc/connection-backoff.md#proposed-backoff-algorithm
connectDeadline := time.Now().Add(dialDuration)
ac.updateConnectivityState(connectivity.Connecting, nil)
ac.mu.Unlock()
if err := ac.tryAllAddrs(addrs, connectDeadline); err != nil {
ac.cc.resolveNow(resolver.ResolveNowOptions{})
// After exhausting all addresses, the addrConn enters
// TRANSIENT_FAILURE.
ac.mu.Lock()
if ac.state == connectivity.Shutdown {
ac.mu.Unlock()
return
}
ac.updateConnectivityState(connectivity.TransientFailure, err)
// Backoff.
b := ac.resetBackoff
ac.mu.Unlock()
timer := time.NewTimer(backoffFor)
select {
case <-timer.C:
ac.mu.Lock()
ac.backoffIdx++
ac.mu.Unlock()
case <-b:
timer.Stop()
case <-ac.ctx.Done():
timer.Stop()
return
}
ac.mu.Lock()
if ac.state != connectivity.Shutdown {
ac.updateConnectivityState(connectivity.Idle, err)
}
ac.mu.Unlock()
return
}
// Success; reset backoff.
ac.mu.Lock()
ac.backoffIdx = 0
ac.mu.Unlock()
}
// tryAllAddrs tries to creates a connection to the addresses, and stop when at
// the first successful one. It returns an error if no address was successfully
// connected, or updates ac appropriately with the new transport.
func (ac *addrConn) tryAllAddrs(addrs []resolver.Address, connectDeadline time.Time) error {
var firstConnErr error
for _, addr := range addrs {
ac.mu.Lock()
if ac.state == connectivity.Shutdown {
ac.mu.Unlock()
return errConnClosing
}
ac.cc.mu.RLock()
ac.dopts.copts.KeepaliveParams = ac.cc.mkp
ac.cc.mu.RUnlock()
copts := ac.dopts.copts
if ac.scopts.CredsBundle != nil {
copts.CredsBundle = ac.scopts.CredsBundle
}
ac.mu.Unlock()
channelz.Infof(logger, ac.channelzID, "Subchannel picks a new address %q to connect", addr.Addr)
err := ac.createTransport(addr, copts, connectDeadline)
if err == nil {
return nil
}
if firstConnErr == nil {
firstConnErr = err
}
ac.cc.updateConnectionError(err)
}
// Couldn't connect to any address.
return firstConnErr
}
// createTransport creates a connection to addr. It returns an error if the
// address was not successfully connected, or updates ac appropriately with the
// new transport.
func (ac *addrConn) createTransport(addr resolver.Address, copts transport.ConnectOptions, connectDeadline time.Time) error {
// TODO: Delete prefaceReceived and move the logic to wait for it into the
// transport.
prefaceReceived := grpcsync.NewEvent()
connClosed := grpcsync.NewEvent()
addr.ServerName = ac.cc.getServerName(addr)
hctx, hcancel := context.WithCancel(ac.ctx)
hcStarted := false // protected by ac.mu
onClose := func() {
ac.mu.Lock()
defer ac.mu.Unlock()
defer connClosed.Fire()
if !hcStarted || hctx.Err() != nil {
// We didn't start the health check or set the state to READY, so
// no need to do anything else here.
//
// OR, we have already cancelled the health check context, meaning
// we have already called onClose once for this transport. In this
// case it would be dangerous to clear the transport and update the
// state, since there may be a new transport in this addrConn.
return
}
hcancel()
ac.transport = nil
// Refresh the name resolver
ac.cc.resolveNow(resolver.ResolveNowOptions{})
if ac.state != connectivity.Shutdown {
ac.updateConnectivityState(connectivity.Idle, nil)
}
}
onGoAway := func(r transport.GoAwayReason) {
ac.mu.Lock()
ac.adjustParams(r)
ac.mu.Unlock()
onClose()
}
connectCtx, cancel := context.WithDeadline(ac.ctx, connectDeadline)
defer cancel()
if channelz.IsOn() {
copts.ChannelzParentID = ac.channelzID
}
newTr, err := transport.NewClientTransport(connectCtx, ac.cc.ctx, addr, copts, func() { prefaceReceived.Fire() }, onGoAway, onClose)
if err != nil {
// newTr is either nil, or closed.
channelz.Warningf(logger, ac.channelzID, "grpc: addrConn.createTransport failed to connect to %v. Err: %v", addr, err)
return err
}
select {
case <-connectCtx.Done():
// We didn't get the preface in time.
// The error we pass to Close() is immaterial since there are no open
// streams at this point, so no trailers with error details will be sent
// out. We just need to pass a non-nil error.
newTr.Close(transport.ErrConnClosing)
if connectCtx.Err() == context.DeadlineExceeded {
err := errors.New("failed to receive server preface within timeout")
channelz.Warningf(logger, ac.channelzID, "grpc: addrConn.createTransport failed to connect to %v: %v", addr, err)
return err
}
return nil
case <-prefaceReceived.Done():
// We got the preface - huzzah! things are good.
ac.mu.Lock()
defer ac.mu.Unlock()
if connClosed.HasFired() {
// onClose called first; go idle but do nothing else.
if ac.state != connectivity.Shutdown {
ac.updateConnectivityState(connectivity.Idle, nil)
}
return nil
}
if ac.state == connectivity.Shutdown {
// This can happen if the subConn was removed while in `Connecting`
// state. tearDown() would have set the state to `Shutdown`, but
// would not have closed the transport since ac.transport would not
// been set at that point.
//
// We run this in a goroutine because newTr.Close() calls onClose()
// inline, which requires locking ac.mu.
//
// The error we pass to Close() is immaterial since there are no open
// streams at this point, so no trailers with error details will be sent
// out. We just need to pass a non-nil error.
go newTr.Close(transport.ErrConnClosing)
return nil
}
ac.curAddr = addr
ac.transport = newTr
hcStarted = true
ac.startHealthCheck(hctx) // Will set state to READY if appropriate.
return nil
case <-connClosed.Done():
// The transport has already closed. If we received the preface, too,
// this is not an error.
select {
case <-prefaceReceived.Done():
return nil
default:
return errors.New("connection closed before server preface received")
}
}
}
// startHealthCheck starts the health checking stream (RPC) to watch the health
// stats of this connection if health checking is requested and configured.
//
// LB channel health checking is enabled when all requirements below are met:
// 1. it is not disabled by the user with the WithDisableHealthCheck DialOption
// 2. internal.HealthCheckFunc is set by importing the grpc/health package
// 3. a service config with non-empty healthCheckConfig field is provided
// 4. the load balancer requests it
//
// It sets addrConn to READY if the health checking stream is not started.
//
// Caller must hold ac.mu.
func (ac *addrConn) startHealthCheck(ctx context.Context) {
var healthcheckManagingState bool
defer func() {
if !healthcheckManagingState {
ac.updateConnectivityState(connectivity.Ready, nil)
}
}()
if ac.cc.dopts.disableHealthCheck {
return
}
healthCheckConfig := ac.cc.healthCheckConfig()
if healthCheckConfig == nil {
return
}
if !ac.scopts.HealthCheckEnabled {
return
}
healthCheckFunc := ac.cc.dopts.healthCheckFunc
if healthCheckFunc == nil {
// The health package is not imported to set health check function.
//
// TODO: add a link to the health check doc in the error message.
channelz.Error(logger, ac.channelzID, "Health check is requested but health check function is not set.")
return
}
healthcheckManagingState = true
// Set up the health check helper functions.
currentTr := ac.transport
newStream := func(method string) (interface{}, error) {
ac.mu.Lock()
if ac.transport != currentTr {
ac.mu.Unlock()
return nil, status.Error(codes.Canceled, "the provided transport is no longer valid to use")
}
ac.mu.Unlock()
return newNonRetryClientStream(ctx, &StreamDesc{ServerStreams: true}, method, currentTr, ac)
}
setConnectivityState := func(s connectivity.State, lastErr error) {
ac.mu.Lock()
defer ac.mu.Unlock()
if ac.transport != currentTr {
return
}
ac.updateConnectivityState(s, lastErr)
}
// Start the health checking stream.
go func() {
err := ac.cc.dopts.healthCheckFunc(ctx, newStream, setConnectivityState, healthCheckConfig.ServiceName)
if err != nil {
if status.Code(err) == codes.Unimplemented {
channelz.Error(logger, ac.channelzID, "Subchannel health check is unimplemented at server side, thus health check is disabled")
} else {
channelz.Errorf(logger, ac.channelzID, "HealthCheckFunc exits with unexpected error %v", err)
}
}
}()
}
func (ac *addrConn) resetConnectBackoff() {
ac.mu.Lock()
close(ac.resetBackoff)
ac.backoffIdx = 0
ac.resetBackoff = make(chan struct{})
ac.mu.Unlock()
}
// getReadyTransport returns the transport if ac's state is READY or nil if not.
func (ac *addrConn) getReadyTransport() transport.ClientTransport {
ac.mu.Lock()
defer ac.mu.Unlock()
if ac.state == connectivity.Ready {
return ac.transport
}
return nil
}
// tearDown starts to tear down the addrConn.
//
// Note that tearDown doesn't remove ac from ac.cc.conns, so the addrConn struct
// will leak. In most cases, call cc.removeAddrConn() instead.
func (ac *addrConn) tearDown(err error) {
ac.mu.Lock()
if ac.state == connectivity.Shutdown {
ac.mu.Unlock()
return
}
curTr := ac.transport
ac.transport = nil
// We have to set the state to Shutdown before anything else to prevent races
// between setting the state and logic that waits on context cancellation / etc.
ac.updateConnectivityState(connectivity.Shutdown, nil)
ac.cancel()
ac.curAddr = resolver.Address{}
if err == errConnDrain && curTr != nil {
// GracefulClose(...) may be executed multiple times when
// i) receiving multiple GoAway frames from the server; or
// ii) there are concurrent name resolver/Balancer triggered
// address removal and GoAway.
// We have to unlock and re-lock here because GracefulClose => Close => onClose, which requires locking ac.mu.
ac.mu.Unlock()
curTr.GracefulClose()
ac.mu.Lock()
}
if channelz.IsOn() {
channelz.AddTraceEvent(logger, ac.channelzID, 0, &channelz.TraceEventDesc{
Desc: "Subchannel Deleted",
Severity: channelz.CtInfo,
Parent: &channelz.TraceEventDesc{
Desc: fmt.Sprintf("Subchanel(id:%d) deleted", ac.channelzID),
Severity: channelz.CtInfo,
},
})
// TraceEvent needs to be called before RemoveEntry, as TraceEvent may add trace reference to
// the entity being deleted, and thus prevent it from being deleted right away.
channelz.RemoveEntry(ac.channelzID)
}
ac.mu.Unlock()
}
func (ac *addrConn) getState() connectivity.State {
ac.mu.Lock()
defer ac.mu.Unlock()
return ac.state
}
func (ac *addrConn) ChannelzMetric() *channelz.ChannelInternalMetric {
ac.mu.Lock()
addr := ac.curAddr.Addr
ac.mu.Unlock()
return &channelz.ChannelInternalMetric{
State: ac.getState(),
Target: addr,
CallsStarted: atomic.LoadInt64(&ac.czData.callsStarted),
CallsSucceeded: atomic.LoadInt64(&ac.czData.callsSucceeded),
CallsFailed: atomic.LoadInt64(&ac.czData.callsFailed),
LastCallStartedTimestamp: time.Unix(0, atomic.LoadInt64(&ac.czData.lastCallStartedTime)),
}
}
func (ac *addrConn) incrCallsStarted() {
atomic.AddInt64(&ac.czData.callsStarted, 1)
atomic.StoreInt64(&ac.czData.lastCallStartedTime, time.Now().UnixNano())
}
func (ac *addrConn) incrCallsSucceeded() {
atomic.AddInt64(&ac.czData.callsSucceeded, 1)
}
func (ac *addrConn) incrCallsFailed() {
atomic.AddInt64(&ac.czData.callsFailed, 1)
}
type retryThrottler struct {
max float64
thresh float64
ratio float64
mu sync.Mutex
tokens float64 // TODO(dfawley): replace with atomic and remove lock.
}
// throttle subtracts a retry token from the pool and returns whether a retry
// should be throttled (disallowed) based upon the retry throttling policy in
// the service config.
func (rt *retryThrottler) throttle() bool {
if rt == nil {
return false
}
rt.mu.Lock()
defer rt.mu.Unlock()
rt.tokens--
if rt.tokens < 0 {
rt.tokens = 0
}
return rt.tokens <= rt.thresh
}
func (rt *retryThrottler) successfulRPC() {
if rt == nil {
return
}
rt.mu.Lock()
defer rt.mu.Unlock()
rt.tokens += rt.ratio
if rt.tokens > rt.max {
rt.tokens = rt.max
}
}
type channelzChannel struct {
cc *ClientConn
}
func (c *channelzChannel) ChannelzMetric() *channelz.ChannelInternalMetric {
return c.cc.channelzMetric()
}
// ErrClientConnTimeout indicates that the ClientConn cannot establish the
// underlying connections within the specified timeout.
//
// Deprecated: This error is never returned by grpc and should not be
// referenced by users.
var ErrClientConnTimeout = errors.New("grpc: timed out when dialing")
func (cc *ClientConn) getResolver(scheme string) resolver.Builder {
for _, rb := range cc.dopts.resolvers {
if scheme == rb.Scheme() {
return rb
}
}
return resolver.Get(scheme)
}
func (cc *ClientConn) updateConnectionError(err error) {
cc.lceMu.Lock()
cc.lastConnectionError = err
cc.lceMu.Unlock()
}
func (cc *ClientConn) connectionError() error {
cc.lceMu.Lock()
defer cc.lceMu.Unlock()
return cc.lastConnectionError
}
func (cc *ClientConn) parseTargetAndFindResolver() (resolver.Builder, error) {
channelz.Infof(logger, cc.channelzID, "original dial target is: %q", cc.target)
var rb resolver.Builder
parsedTarget, err := parseTarget(cc.target)
if err != nil {
channelz.Infof(logger, cc.channelzID, "dial target %q parse failed: %v", cc.target, err)
} else {
channelz.Infof(logger, cc.channelzID, "parsed dial target is: %+v", parsedTarget)
rb = cc.getResolver(parsedTarget.Scheme)
if rb != nil {
cc.parsedTarget = parsedTarget
return rb, nil
}
}
// We are here because the user's dial target did not contain a scheme or
// specified an unregistered scheme. We should fallback to the default
// scheme, except when a custom dialer is specified in which case, we should
// always use passthrough scheme.
defScheme := resolver.GetDefaultScheme()
channelz.Infof(logger, cc.channelzID, "fallback to scheme %q", defScheme)
canonicalTarget := defScheme + ":///" + cc.target
parsedTarget, err = parseTarget(canonicalTarget)
if err != nil {
channelz.Infof(logger, cc.channelzID, "dial target %q parse failed: %v", canonicalTarget, err)
return nil, err
}
channelz.Infof(logger, cc.channelzID, "parsed dial target is: %+v", parsedTarget)
rb = cc.getResolver(parsedTarget.Scheme)
if rb == nil {
return nil, fmt.Errorf("could not get resolver for default scheme: %q", parsedTarget.Scheme)
}
cc.parsedTarget = parsedTarget
return rb, nil
}
// parseTarget uses RFC 3986 semantics to parse the given target into a
// resolver.Target struct containing scheme, authority and endpoint. Query
// params are stripped from the endpoint.
func parseTarget(target string) (resolver.Target, error) {
u, err := url.Parse(target)
if err != nil {
return resolver.Target{}, err
}
// For targets of the form "[scheme]://[authority]/endpoint, the endpoint
// value returned from url.Parse() contains a leading "/". Although this is
// in accordance with RFC 3986, we do not want to break existing resolver
// implementations which expect the endpoint without the leading "/". So, we
// end up stripping the leading "/" here. But this will result in an
// incorrect parsing for something like "unix:///path/to/socket". Since we
// own the "unix" resolver, we can workaround in the unix resolver by using
// the `URL` field instead of the `Endpoint` field.
endpoint := u.Path
if endpoint == "" {
endpoint = u.Opaque
}
endpoint = strings.TrimPrefix(endpoint, "/")
return resolver.Target{
Scheme: u.Scheme,
Authority: u.Host,
Endpoint: endpoint,
URL: *u,
}, nil
}
// Determine channel authority. The order of precedence is as follows:
// - user specified authority override using `WithAuthority` dial option
// - creds' notion of server name for the authentication handshake
// - endpoint from dial target of the form "scheme://[authority]/endpoint"
func determineAuthority(endpoint, target string, dopts dialOptions) (string, error) {
// Historically, we had two options for users to specify the serverName or
// authority for a channel. One was through the transport credentials
// (either in its constructor, or through the OverrideServerName() method).
// The other option (for cases where WithInsecure() dial option was used)
// was to use the WithAuthority() dial option.
//
// A few things have changed since:
// - `insecure` package with an implementation of the `TransportCredentials`
// interface for the insecure case
// - WithAuthority() dial option support for secure credentials
authorityFromCreds := ""
if creds := dopts.copts.TransportCredentials; creds != nil && creds.Info().ServerName != "" {
authorityFromCreds = creds.Info().ServerName
}
authorityFromDialOption := dopts.authority
if (authorityFromCreds != "" && authorityFromDialOption != "") && authorityFromCreds != authorityFromDialOption {
return "", fmt.Errorf("ClientConn's authority from transport creds %q and dial option %q don't match", authorityFromCreds, authorityFromDialOption)
}
switch {
case authorityFromDialOption != "":
return authorityFromDialOption, nil
case authorityFromCreds != "":
return authorityFromCreds, nil
case strings.HasPrefix(target, "unix:") || strings.HasPrefix(target, "unix-abstract:"):
// TODO: remove when the unix resolver implements optional interface to
// return channel authority.
return "localhost", nil
case strings.HasPrefix(endpoint, ":"):
return "localhost" + endpoint, nil
default:
// TODO: Define an optional interface on the resolver builder to return
// the channel authority given the user's dial target. For resolvers
// which don't implement this interface, we will use the endpoint from
// "scheme://authority/endpoint" as the default authority.
return endpoint, nil
}
}
|