forked from apple/foundationdb
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathFlowTransport.actor.cpp
2076 lines (1860 loc) · 77.2 KB
/
FlowTransport.actor.cpp
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
/*
* FlowTransport.actor.cpp
*
* This source file is part of the FoundationDB open source project
*
* Copyright 2013-2022 Apple Inc. and the FoundationDB project 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.
*/
#include "fdbrpc/FlowTransport.h"
#include "flow/Arena.h"
#include "flow/network.h"
#include <cstdint>
#include <unordered_map>
#if VALGRIND
#include <memcheck.h>
#endif
#include <boost/unordered_map.hpp>
#include "fdbrpc/TokenSign.h"
#include "fdbrpc/fdbrpc.h"
#include "fdbrpc/FailureMonitor.h"
#include "fdbrpc/HealthMonitor.h"
#include "fdbrpc/JsonWebKeySet.h"
#include "fdbrpc/genericactors.actor.h"
#include "fdbrpc/IPAllowList.h"
#include "fdbrpc/TokenCache.h"
#include "fdbrpc/simulator.h"
#include "flow/ActorCollection.h"
#include "flow/Error.h"
#include "flow/flow.h"
#include "flow/Net2Packet.h"
#include "flow/TDMetric.actor.h"
#include "flow/ObjectSerializer.h"
#include "flow/Platform.h"
#include "flow/ProtocolVersion.h"
#include "flow/UnitTest.h"
#include "flow/WatchFile.actor.h"
#include "flow/IConnection.h"
#define XXH_INLINE_ALL
#include "flow/xxhash.h"
#include "flow/actorcompiler.h" // This must be the last #include.
void removeCachedDNS(const std::string& host, const std::string& service) {
INetworkConnections::net()->removeCachedDNS(host, service);
}
namespace {
NetworkAddressList g_currentDeliveryPeerAddress = NetworkAddressList();
bool g_currentDeliverPeerAddressTrusted = false;
Future<Void> g_currentDeliveryPeerDisconnect;
} // namespace
constexpr int PACKET_LEN_WIDTH = sizeof(uint32_t);
const uint64_t TOKEN_STREAM_FLAG = 1;
FDB_BOOLEAN_PARAM(InReadSocket);
FDB_BOOLEAN_PARAM(IsStableConnection);
class EndpointMap : NonCopyable {
public:
// Reserve space for this many wellKnownEndpoints
explicit EndpointMap(int wellKnownEndpointCount);
void insertWellKnown(NetworkMessageReceiver* r, const Endpoint::Token& token, TaskPriority priority);
void insert(NetworkMessageReceiver* r, Endpoint::Token& token, TaskPriority priority);
const Endpoint& insert(NetworkAddressList localAddresses,
std::vector<std::pair<FlowReceiver*, TaskPriority>> const& streams);
NetworkMessageReceiver* get(Endpoint::Token const& token);
TaskPriority getPriority(Endpoint::Token const& token);
void remove(Endpoint::Token const& token, NetworkMessageReceiver* r);
private:
void realloc();
struct Entry {
union {
uint64_t
uid[2]; // priority packed into lower 32 bits; actual lower 32 bits of token are the index in data[]
uint32_t nextFree;
};
NetworkMessageReceiver* receiver = nullptr;
Endpoint::Token& token() { return *(Endpoint::Token*)uid; }
};
int wellKnownEndpointCount;
std::vector<Entry> data;
uint32_t firstFree;
};
EndpointMap::EndpointMap(int wellKnownEndpointCount)
: wellKnownEndpointCount(wellKnownEndpointCount), data(wellKnownEndpointCount), firstFree(-1) {}
void EndpointMap::realloc() {
int oldSize = data.size();
data.resize(std::max(128, oldSize * 2));
for (int i = oldSize; i < data.size(); i++) {
data[i].receiver = 0;
data[i].nextFree = i + 1;
}
data[data.size() - 1].nextFree = firstFree;
firstFree = oldSize;
}
void EndpointMap::insertWellKnown(NetworkMessageReceiver* r, const Endpoint::Token& token, TaskPriority priority) {
int index = token.second();
ASSERT(index <= wellKnownEndpointCount);
ASSERT(data[index].receiver == nullptr);
data[index].receiver = r;
data[index].token() =
Endpoint::Token(token.first(), (token.second() & 0xffffffff00000000LL) | static_cast<uint32_t>(priority));
}
void EndpointMap::insert(NetworkMessageReceiver* r, Endpoint::Token& token, TaskPriority priority) {
if (firstFree == uint32_t(-1))
realloc();
int index = firstFree;
firstFree = data[index].nextFree;
token = Endpoint::Token(token.first(), (token.second() & 0xffffffff00000000LL) | index);
data[index].token() =
Endpoint::Token(token.first(), (token.second() & 0xffffffff00000000LL) | static_cast<uint32_t>(priority));
data[index].receiver = r;
}
const Endpoint& EndpointMap::insert(NetworkAddressList localAddresses,
std::vector<std::pair<FlowReceiver*, TaskPriority>> const& streams) {
int adjacentFree = 0;
int adjacentStart = -1;
firstFree = -1;
for (int i = wellKnownEndpointCount; i < data.size(); i++) {
if (data[i].receiver) {
adjacentFree = 0;
} else {
data[i].nextFree = firstFree;
firstFree = i;
if (adjacentStart == -1 && ++adjacentFree == streams.size()) {
adjacentStart = i + 1 - adjacentFree;
firstFree = data[adjacentStart].nextFree;
}
}
}
if (adjacentStart == -1) {
data.resize(data.size() + streams.size() - adjacentFree);
adjacentStart = data.size() - streams.size();
if (adjacentFree > 0) {
firstFree = data[adjacentStart].nextFree;
}
}
UID base = deterministicRandom()->randomUniqueID();
for (uint64_t i = 0; i < streams.size(); i++) {
int index = adjacentStart + i;
uint64_t first = (base.first() + (i << 32)) | TOKEN_STREAM_FLAG;
streams[i].first->setEndpoint(
Endpoint(localAddresses, UID(first, (base.second() & 0xffffffff00000000LL) | index)));
data[index].token() =
Endpoint::Token(first, (base.second() & 0xffffffff00000000LL) | static_cast<uint32_t>(streams[i].second));
data[index].receiver = (NetworkMessageReceiver*)streams[i].first;
}
return streams[0].first->getEndpoint(TaskPriority::DefaultEndpoint);
}
NetworkMessageReceiver* EndpointMap::get(Endpoint::Token const& token) {
uint32_t index = token.second();
if (index < wellKnownEndpointCount && data[index].receiver == nullptr) {
TraceEvent(SevWarnAlways, "WellKnownEndpointNotAdded")
.detail("Token", token)
.detail("Index", index)
.backtrace();
}
if (index < data.size() && data[index].token().first() == token.first() &&
((data[index].token().second() & 0xffffffff00000000LL) | index) == token.second())
return data[index].receiver;
return 0;
}
TaskPriority EndpointMap::getPriority(Endpoint::Token const& token) {
uint32_t index = token.second();
if (index < data.size() && data[index].token().first() == token.first() &&
((data[index].token().second() & 0xffffffff00000000LL) | index) == token.second()) {
auto res = static_cast<TaskPriority>(data[index].token().second());
// we don't allow this priority to be "misused" for other stuff as we won't even
// attempt to find an endpoint if UnknownEndpoint is returned here
ASSERT(res != TaskPriority::UnknownEndpoint);
return res;
}
return TaskPriority::UnknownEndpoint;
}
void EndpointMap::remove(Endpoint::Token const& token, NetworkMessageReceiver* r) {
uint32_t index = token.second();
if (index < wellKnownEndpointCount) {
data[index].receiver = nullptr;
} else if (index < data.size() && data[index].token().first() == token.first() &&
((data[index].token().second() & 0xffffffff00000000LL) | index) == token.second() &&
data[index].receiver == r) {
data[index].receiver = 0;
data[index].nextFree = firstFree;
firstFree = index;
}
}
struct EndpointNotFoundReceiver final : NetworkMessageReceiver {
EndpointNotFoundReceiver(EndpointMap& endpoints) {
endpoints.insertWellKnown(
this, Endpoint::wellKnownToken(WLTOKEN_ENDPOINT_NOT_FOUND), TaskPriority::DefaultEndpoint);
}
void receive(ArenaObjectReader& reader) override {
// Remote machine tells us it doesn't have endpoint e
UID token;
reader.deserialize(token);
Endpoint e = FlowTransport::transport().loadedEndpoint(token);
IFailureMonitor::failureMonitor().endpointNotFound(e);
}
bool isPublic() const override { return true; }
};
struct PingRequest {
constexpr static FileIdentifier file_identifier = 4707015;
ReplyPromise<Void> reply{ PeerCompatibilityPolicy{ RequirePeer::AtLeast,
ProtocolVersion::withStableInterfaces() } };
template <class Ar>
void serialize(Ar& ar) {
serializer(ar, reply);
}
};
struct PingReceiver final : NetworkMessageReceiver {
PingReceiver(EndpointMap& endpoints) {
endpoints.insertWellKnown(this, Endpoint::wellKnownToken(WLTOKEN_PING_PACKET), TaskPriority::ReadSocket);
}
void receive(ArenaObjectReader& reader) override {
PingRequest req;
reader.deserialize(req);
req.reply.send(Void());
}
PeerCompatibilityPolicy peerCompatibilityPolicy() const override {
return PeerCompatibilityPolicy{ RequirePeer::AtLeast, ProtocolVersion::withStableInterfaces() };
}
bool isPublic() const override { return true; }
};
struct UnauthorizedEndpointReceiver final : NetworkMessageReceiver {
UnauthorizedEndpointReceiver(EndpointMap& endpoints) {
endpoints.insertWellKnown(
this, Endpoint::wellKnownToken(WLTOKEN_UNAUTHORIZED_ENDPOINT), TaskPriority::ReadSocket);
}
void receive(ArenaObjectReader& reader) override {
UID token;
reader.deserialize(token);
Endpoint e = FlowTransport::transport().loadedEndpoint(token);
IFailureMonitor::failureMonitor().unauthorizedEndpoint(e);
}
bool isPublic() const override { return true; }
};
// NetworkAddressCachedString retains a cached Standalone<StringRef> of
// a NetworkAddressList.address.toString() value. This cached value is useful
// for features in the hot path (i.e. Tracing), which need the String formatted value
// frequently and do not wish to pay the formatting cost. If the underlying NetworkAddressList
// needs to change, do not attempt to update it directly, use the setNetworkAddress API as it
// will ensure the new toString() cached value is updated.
class NetworkAddressCachedString {
public:
NetworkAddressCachedString() { setAddressList(NetworkAddressList()); }
NetworkAddressCachedString(NetworkAddressList const& list) { setAddressList(list); }
NetworkAddressList const& getAddressList() const { return addressList; }
void setAddressList(NetworkAddressList const& list) {
cachedStr = Standalone<StringRef>(StringRef(list.address.toString()));
addressList = list;
}
void setNetworkAddress(NetworkAddress const& addr) {
addressList.address = addr;
setAddressList(addressList); // force the recaching of the string.
}
Standalone<StringRef> getLocalAddressAsString() const { return cachedStr; }
operator NetworkAddressList const&() { return addressList; }
private:
NetworkAddressList addressList;
Standalone<StringRef> cachedStr;
};
class TransportData {
public:
TransportData(uint64_t transportId, int maxWellKnownEndpoints, IPAllowList const* allowList);
~TransportData();
void initMetrics() {
bytesSent.init("Net2.BytesSent"_sr);
countPacketsReceived.init("Net2.CountPacketsReceived"_sr);
countPacketsGenerated.init("Net2.CountPacketsGenerated"_sr);
countConnEstablished.init("Net2.CountConnEstablished"_sr);
countConnClosedWithError.init("Net2.CountConnClosedWithError"_sr);
countConnClosedWithoutError.init("Net2.CountConnClosedWithoutError"_sr);
}
Reference<struct Peer> getPeer(NetworkAddress const& address);
Reference<struct Peer> getOrOpenPeer(NetworkAddress const& address, bool startConnectionKeeper = true);
// Returns true if given network address 'address' is one of the address we are listening on.
bool isLocalAddress(const NetworkAddress& address) const;
void applyPublicKeySet(StringRef jwkSetString);
NetworkAddressCachedString localAddresses;
std::vector<Future<Void>> listeners;
std::unordered_map<NetworkAddress, Reference<struct Peer>> peers;
std::unordered_map<NetworkAddress, std::pair<double, double>> closedPeers;
HealthMonitor healthMonitor;
std::set<NetworkAddress> orderedAddresses;
Reference<AsyncVar<bool>> degraded;
EndpointMap endpoints;
EndpointNotFoundReceiver endpointNotFoundReceiver{ endpoints };
PingReceiver pingReceiver{ endpoints };
UnauthorizedEndpointReceiver unauthorizedEndpointReceiver{ endpoints };
Int64MetricHandle bytesSent;
Int64MetricHandle countPacketsReceived;
Int64MetricHandle countPacketsGenerated;
Int64MetricHandle countConnEstablished;
Int64MetricHandle countConnClosedWithError;
Int64MetricHandle countConnClosedWithoutError;
std::map<NetworkAddress, std::pair<uint64_t, double>> incompatiblePeers;
AsyncTrigger incompatiblePeersChanged;
uint32_t numIncompatibleConnections;
std::map<uint64_t, double> multiVersionConnections;
double lastIncompatibleMessage;
uint64_t transportId;
IPAllowList allowList;
Future<Void> multiVersionCleanup;
Future<Void> pingLogger;
Future<Void> publicKeyFileWatch;
std::unordered_map<Standalone<StringRef>, PublicKey> publicKeys;
};
ACTOR Future<Void> pingLatencyLogger(TransportData* self) {
state NetworkAddress lastAddress = NetworkAddress();
loop {
if (self->orderedAddresses.size()) {
auto it = self->orderedAddresses.upper_bound(lastAddress);
if (it == self->orderedAddresses.end()) {
it = self->orderedAddresses.begin();
}
lastAddress = *it;
auto peer = self->getPeer(lastAddress);
if (!peer) {
TraceEvent(SevWarnAlways, "MissingNetworkAddress").suppressFor(10.0).detail("PeerAddr", lastAddress);
}
if (peer->lastLoggedTime <= 0.0) {
peer->lastLoggedTime = peer->lastConnectTime;
}
if (peer && (peer->pingLatencies.getPopulationSize() >= 10 || peer->connectFailedCount > 0 ||
peer->timeoutCount > 0)) {
TraceEvent("PingLatency")
.detail("Elapsed", now() - peer->lastLoggedTime)
.detail("PeerAddr", lastAddress)
.detail("MinLatency", peer->pingLatencies.min())
.detail("MaxLatency", peer->pingLatencies.max())
.detail("MeanLatency", peer->pingLatencies.mean())
.detail("MedianLatency", peer->pingLatencies.median())
.detail("P90Latency", peer->pingLatencies.percentile(0.90))
.detail("Count", peer->pingLatencies.getPopulationSize())
.detail("BytesReceived", peer->bytesReceived - peer->lastLoggedBytesReceived)
.detail("BytesSent", peer->bytesSent - peer->lastLoggedBytesSent)
.detail("TimeoutCount", peer->timeoutCount)
.detail("ConnectOutgoingCount", peer->connectOutgoingCount)
.detail("ConnectIncomingCount", peer->connectIncomingCount)
.detail("ConnectFailedCount", peer->connectFailedCount)
.detail("ConnectMinLatency", peer->connectLatencies.min())
.detail("ConnectMaxLatency", peer->connectLatencies.max())
.detail("ConnectMeanLatency", peer->connectLatencies.mean())
.detail("ConnectMedianLatency", peer->connectLatencies.median())
.detail("ConnectP90Latency", peer->connectLatencies.percentile(0.90));
peer->lastLoggedTime = now();
peer->connectOutgoingCount = 0;
peer->connectIncomingCount = 0;
peer->connectFailedCount = 0;
peer->pingLatencies.clear();
peer->connectLatencies.clear();
peer->lastLoggedBytesReceived = peer->bytesReceived;
peer->lastLoggedBytesSent = peer->bytesSent;
peer->timeoutCount = 0;
wait(delay(FLOW_KNOBS->PING_LOGGING_INTERVAL));
} else if (it == self->orderedAddresses.begin()) {
wait(delay(FLOW_KNOBS->PING_LOGGING_INTERVAL));
}
} else {
wait(delay(FLOW_KNOBS->PING_LOGGING_INTERVAL));
}
}
}
TransportData::TransportData(uint64_t transportId, int maxWellKnownEndpoints, IPAllowList const* allowList)
: endpoints(maxWellKnownEndpoints), endpointNotFoundReceiver(endpoints), pingReceiver(endpoints),
numIncompatibleConnections(0), lastIncompatibleMessage(0), transportId(transportId),
allowList(allowList == nullptr ? IPAllowList() : *allowList) {
degraded = makeReference<AsyncVar<bool>>(false);
pingLogger = pingLatencyLogger(this);
}
#define CONNECT_PACKET_V0 0x0FDB00A444020001LL
#define CONNECT_PACKET_V0_SIZE 14
#pragma pack(push, 1)
struct ConnectPacket {
// The value does not include the size of `connectPacketLength` itself,
// but only the other fields of this structure.
uint32_t connectPacketLength = 0;
ProtocolVersion protocolVersion; // Expect currentProtocolVersion
uint16_t canonicalRemotePort = 0; // Port number to reconnect to the originating process
uint64_t connectionId = 0; // Multi-version clients will use the same Id for both connections, other connections
// will set this to zero. Added at protocol Version 0x0FDB00A444020001.
// IP Address to reconnect to the originating process. Only one of these must be populated.
uint32_t canonicalRemoteIp4 = 0;
enum ConnectPacketFlags { FLAG_IPV6 = 1 };
uint16_t flags = 0;
uint8_t canonicalRemoteIp6[16] = { 0 };
ConnectPacket() = default;
IPAddress canonicalRemoteIp() const {
if (isIPv6()) {
IPAddress::IPAddressStore store;
memcpy(store.data(), canonicalRemoteIp6, sizeof(canonicalRemoteIp6));
return IPAddress(store);
} else {
return IPAddress(canonicalRemoteIp4);
}
}
void setCanonicalRemoteIp(const IPAddress& ip) {
if (ip.isV6()) {
flags = flags | FLAG_IPV6;
memcpy(&canonicalRemoteIp6, ip.toV6().data(), 16);
} else {
flags = flags & ~FLAG_IPV6;
canonicalRemoteIp4 = ip.toV4();
}
}
bool isIPv6() const { return flags & FLAG_IPV6; }
uint32_t totalPacketSize() const { return connectPacketLength + sizeof(connectPacketLength); }
template <class Ar>
void serialize(Ar& ar) {
serializer(ar, connectPacketLength);
if (connectPacketLength > sizeof(ConnectPacket) - sizeof(connectPacketLength)) {
ASSERT(!g_network->isSimulated());
TraceEvent("SerializationFailed").backtrace();
throw serialization_failed();
}
serializer(ar, protocolVersion, canonicalRemotePort, connectionId, canonicalRemoteIp4);
if (ar.isDeserializing && !ar.protocolVersion().hasIPv6()) {
flags = 0;
} else {
// We can send everything in serialized packet, since the current version of ConnectPacket
// is backward compatible with CONNECT_PACKET_V0.
serializer(ar, flags);
ar.serializeBytes(&canonicalRemoteIp6, sizeof(canonicalRemoteIp6));
}
}
};
#pragma pack(pop)
ACTOR static Future<Void> connectionReader(TransportData* transport,
Reference<IConnection> conn,
Reference<struct Peer> peer,
Promise<Reference<struct Peer>> onConnected);
static void sendLocal(TransportData* self, ISerializeSource const& what, const Endpoint& destination);
static ReliablePacket* sendPacket(TransportData* self,
Reference<Peer> peer,
ISerializeSource const& what,
const Endpoint& destination,
bool reliable);
ACTOR Future<Void> connectionMonitor(Reference<Peer> peer) {
state Endpoint remotePingEndpoint({ peer->destination }, Endpoint::wellKnownToken(WLTOKEN_PING_PACKET));
loop {
if (!FlowTransport::isClient() && !peer->destination.isPublic() && peer->compatible) {
// Don't send ping messages to clients unless necessary. Instead monitor incoming client pings.
// We ignore this block for incompatible clients because pings from server would trigger the
// peer->resetPing and prevent 'connection_failed' due to ping timeout.
state double lastRefreshed = now();
state int64_t lastBytesReceived = peer->bytesReceived;
loop {
wait(delay(FLOW_KNOBS->CONNECTION_MONITOR_LOOP_TIME, TaskPriority::ReadSocket));
if (lastBytesReceived < peer->bytesReceived) {
lastRefreshed = now();
lastBytesReceived = peer->bytesReceived;
} else if (lastRefreshed < now() - FLOW_KNOBS->CONNECTION_MONITOR_IDLE_TIMEOUT *
FLOW_KNOBS->CONNECTION_MONITOR_INCOMING_IDLE_MULTIPLIER) {
// If we have not received anything in this period, client must have closed
// connection by now. Break loop to check if it is still alive by sending a ping.
break;
}
}
}
// We cannot let an error be thrown from connectionMonitor while still on the stack from scanPackets in
// connectionReader because then it would not call the destructor of connectionReader when connectionReader is
// cancelled.
wait(delay(0, TaskPriority::ReadSocket));
if (peer->reliable.empty() && peer->unsent.empty() && peer->outstandingReplies == 0) {
if (peer->peerReferences == 0 &&
(peer->lastDataPacketSentTime < now() - FLOW_KNOBS->CONNECTION_MONITOR_UNREFERENCED_CLOSE_DELAY)) {
// TODO: What about when peerReference == -1?
throw connection_unreferenced();
} else if (FlowTransport::isClient() && peer->compatible && peer->destination.isPublic() &&
(peer->lastConnectTime < now() - FLOW_KNOBS->CONNECTION_MONITOR_IDLE_TIMEOUT) &&
(peer->lastDataPacketSentTime < now() - FLOW_KNOBS->CONNECTION_MONITOR_IDLE_TIMEOUT)) {
// First condition is necessary because we may get here if we are server.
throw connection_idle();
}
}
wait(delayJittered(FLOW_KNOBS->CONNECTION_MONITOR_LOOP_TIME, TaskPriority::ReadSocket));
// TODO: Stop monitoring and close the connection with no onDisconnect requests outstanding
state PingRequest pingRequest;
FlowTransport::transport().sendUnreliable(SerializeSource<PingRequest>(pingRequest), remotePingEndpoint, true);
state int64_t startingBytes = peer->bytesReceived;
state int timeouts = 0;
state double startTime = now();
loop {
choose {
when(wait(delay(FLOW_KNOBS->CONNECTION_MONITOR_TIMEOUT))) {
peer->timeoutCount++;
if (startingBytes == peer->bytesReceived) {
if (peer->destination.isPublic()) {
peer->pingLatencies.addSample(now() - startTime);
}
TraceEvent("ConnectionTimeout").suppressFor(1.0).detail("WithAddr", peer->destination);
throw connection_failed();
}
if (timeouts > 1) {
TraceEvent(SevWarnAlways, "ConnectionSlowPing")
.suppressFor(1.0)
.detail("WithAddr", peer->destination)
.detail("Timeouts", timeouts);
}
startingBytes = peer->bytesReceived;
timeouts++;
}
when(wait(pingRequest.reply.getFuture())) {
if (peer->destination.isPublic()) {
peer->pingLatencies.addSample(now() - startTime);
}
break;
}
when(wait(peer->resetPing.onTrigger())) {
break;
}
}
}
}
}
ACTOR Future<Void> connectionWriter(Reference<Peer> self, Reference<IConnection> conn) {
state double lastWriteTime = now();
loop {
// wait( delay(0, TaskPriority::WriteSocket) );
wait(delayJittered(
std::max<double>(FLOW_KNOBS->MIN_COALESCE_DELAY, FLOW_KNOBS->MAX_COALESCE_DELAY - (now() - lastWriteTime)),
TaskPriority::WriteSocket));
// wait( delay(500e-6, TaskPriority::WriteSocket) );
// wait( yield(TaskPriority::WriteSocket) );
// Send until there is nothing left to send
loop {
lastWriteTime = now();
int sent = conn->write(self->unsent.getUnsent(), /* limit= */ FLOW_KNOBS->MAX_PACKET_SEND_BYTES);
if (sent) {
self->bytesSent += sent;
self->transport->bytesSent += sent;
self->unsent.sent(sent);
}
if (self->unsent.empty()) {
break;
}
CODE_PROBE(
true, "We didn't write everything, so apparently the write buffer is full. Wait for it to be nonfull");
wait(conn->onWritable());
wait(yield(TaskPriority::WriteSocket));
}
// Wait until there is something to send
while (self->unsent.empty())
wait(self->dataToSend.onTrigger());
}
}
ACTOR Future<Void> delayedHealthUpdate(NetworkAddress address, bool* tooManyConnectionsClosed) {
state double start = now();
loop {
if (FLOW_KNOBS->HEALTH_MONITOR_MARK_FAILED_UNSTABLE_CONNECTIONS &&
FlowTransport::transport().healthMonitor()->tooManyConnectionsClosed(address) && address.isPublic()) {
wait(delayJittered(FLOW_KNOBS->MAX_RECONNECTION_TIME * 2.0));
} else {
if (*tooManyConnectionsClosed) {
TraceEvent("TooManyConnectionsClosedMarkAvailable")
.detail("Dest", address)
.detail("StartTime", start)
.detail("TimeElapsed", now() - start)
.detail("ClosedCount", FlowTransport::transport().healthMonitor()->closedConnectionsCount(address));
*tooManyConnectionsClosed = false;
}
IFailureMonitor::failureMonitor().setStatus(address, FailureStatus(false));
break;
}
}
return Void();
}
ACTOR Future<Void> connectionKeeper(Reference<Peer> self,
Reference<IConnection> conn = Reference<IConnection>(),
Future<Void> reader = Void()) {
TraceEvent(SevDebug, "ConnectionKeeper", conn ? conn->getDebugID() : UID())
.detail("PeerAddr", self->destination)
.detail("ConnSet", (bool)conn);
ASSERT_WE_THINK(FlowTransport::transport().getLocalAddress() != self->destination);
state Future<Void> delayedHealthUpdateF;
state Optional<double> firstConnFailedTime = Optional<double>();
state int retryConnect = false;
state bool tooManyConnectionsClosed = false;
loop {
try {
delayedHealthUpdateF = Future<Void>();
if (!conn) { // Always, except for the first loop with an incoming connection
self->outgoingConnectionIdle = true;
// Wait until there is something to send.
while (self->unsent.empty()) {
// Override waiting, if we are in failed state to update failure monitoring status.
Future<Void> retryConnectF = Never();
if (retryConnect) {
retryConnectF = IFailureMonitor::failureMonitor().getState(self->destination).isAvailable()
? delay(FLOW_KNOBS->FAILURE_DETECTION_DELAY)
: delay(FLOW_KNOBS->SERVER_REQUEST_INTERVAL);
}
choose {
when(wait(self->dataToSend.onTrigger())) {}
when(wait(retryConnectF)) {
break;
}
}
}
ASSERT(self->destination.isPublic());
self->outgoingConnectionIdle = false;
wait(delayJittered(std::max(0.0,
self->lastConnectTime + self->reconnectionDelay -
now()))); // Don't connect() to the same peer more than once per 2 sec
self->lastConnectTime = now();
TraceEvent("ConnectingTo", conn ? conn->getDebugID() : UID())
.suppressFor(1.0)
.detail("PeerAddr", self->destination)
.detail("PeerReferences", self->peerReferences)
.detail("FailureStatus",
IFailureMonitor::failureMonitor().getState(self->destination).isAvailable() ? "OK"
: "FAILED");
++self->connectOutgoingCount;
try {
choose {
when(Reference<IConnection> _conn =
wait(INetworkConnections::net()->connect(self->destination))) {
conn = _conn;
wait(conn->connectHandshake());
self->connectLatencies.addSample(now() - self->lastConnectTime);
if (FlowTransport::isClient()) {
IFailureMonitor::failureMonitor().setStatus(self->destination, FailureStatus(false));
}
if (self->unsent.empty()) {
delayedHealthUpdateF =
delayedHealthUpdate(self->destination, &tooManyConnectionsClosed);
choose {
when(wait(delayedHealthUpdateF)) {
conn->close();
conn = Reference<IConnection>();
retryConnect = false;
continue;
}
when(wait(self->dataToSend.onTrigger())) {}
}
}
TraceEvent("ConnectionExchangingConnectPacket", conn->getDebugID())
.suppressFor(1.0)
.detail("PeerAddr", self->destination);
self->prependConnectPacket();
reader = connectionReader(self->transport, conn, self, Promise<Reference<Peer>>());
}
when(wait(delay(FLOW_KNOBS->CONNECTION_MONITOR_TIMEOUT))) {
throw connection_failed();
}
}
} catch (Error& e) {
++self->connectFailedCount;
if (e.code() != error_code_connection_failed) {
throw;
}
TraceEvent("ConnectionTimedOut", conn ? conn->getDebugID() : UID())
.suppressFor(1.0)
.detail("PeerAddr", self->destination);
throw;
}
} else {
self->outgoingConnectionIdle = false;
self->lastConnectTime = now();
}
firstConnFailedTime.reset();
try {
self->transport->countConnEstablished++;
if (!delayedHealthUpdateF.isValid())
delayedHealthUpdateF = delayedHealthUpdate(self->destination, &tooManyConnectionsClosed);
self->connected = true;
wait(connectionWriter(self, conn) || reader || connectionMonitor(self) ||
self->resetConnection.onTrigger());
TraceEvent("ConnectionReset", conn ? conn->getDebugID() : UID())
.suppressFor(1.0)
.detail("PeerAddr", self->destination);
throw connection_failed();
} catch (Error& e) {
if (e.code() == error_code_connection_failed || e.code() == error_code_actor_cancelled ||
e.code() == error_code_connection_unreferenced ||
(g_network->isSimulated() && e.code() == error_code_checksum_failed))
self->transport->countConnClosedWithoutError++;
else
self->transport->countConnClosedWithError++;
throw e;
}
} catch (Error& e) {
self->connected = false;
delayedHealthUpdateF.cancel();
if (now() - self->lastConnectTime > FLOW_KNOBS->RECONNECTION_RESET_TIME) {
self->reconnectionDelay = FLOW_KNOBS->INITIAL_RECONNECTION_TIME;
} else {
self->reconnectionDelay = std::min(FLOW_KNOBS->MAX_RECONNECTION_TIME,
self->reconnectionDelay * FLOW_KNOBS->RECONNECTION_TIME_GROWTH_RATE);
}
if (firstConnFailedTime.present()) {
if (now() - firstConnFailedTime.get() > FLOW_KNOBS->PEER_UNAVAILABLE_FOR_LONG_TIME_TIMEOUT) {
TraceEvent(SevWarnAlways, "PeerUnavailableForLongTime", conn ? conn->getDebugID() : UID())
.suppressFor(1.0)
.detail("PeerAddr", self->destination);
firstConnFailedTime = now() - FLOW_KNOBS->PEER_UNAVAILABLE_FOR_LONG_TIME_TIMEOUT / 2.0;
}
} else {
firstConnFailedTime = now();
}
// Don't immediately mark connection as failed. To stay closed to earlier behaviour of centralized
// failure monitoring, wait until connection stays failed for FLOW_KNOBS->FAILURE_DETECTION_DELAY timeout.
retryConnect = true;
if (e.code() == error_code_connection_failed) {
if (!self->destination.isPublic()) {
// Can't connect back to non-public addresses.
IFailureMonitor::failureMonitor().setStatus(self->destination, FailureStatus(true));
} else if (now() - firstConnFailedTime.get() > FLOW_KNOBS->FAILURE_DETECTION_DELAY) {
IFailureMonitor::failureMonitor().setStatus(self->destination, FailureStatus(true));
}
}
self->discardUnreliablePackets();
reader = Future<Void>();
bool ok = e.code() == error_code_connection_failed || e.code() == error_code_actor_cancelled ||
e.code() == error_code_connection_unreferenced || e.code() == error_code_connection_idle ||
(g_network->isSimulated() && e.code() == error_code_checksum_failed);
if (self->compatible) {
TraceEvent(ok ? SevInfo : SevWarnAlways, "ConnectionClosed", conn ? conn->getDebugID() : UID())
.errorUnsuppressed(e)
.suppressFor(1.0)
.detail("PeerAddr", self->destination);
} else {
TraceEvent(
ok ? SevInfo : SevWarnAlways, "IncompatibleConnectionClosed", conn ? conn->getDebugID() : UID())
.errorUnsuppressed(e)
.suppressFor(1.0)
.detail("PeerAddr", self->destination);
// Since the connection has closed, we need to check the protocol version the next time we connect
self->compatible = true;
}
if (self->destination.isPublic() &&
IFailureMonitor::failureMonitor().getState(self->destination).isAvailable() &&
!FlowTransport::isClient()) {
auto& it = self->transport->closedPeers[self->destination];
if (now() - it.second > FLOW_KNOBS->TOO_MANY_CONNECTIONS_CLOSED_RESET_DELAY) {
it.first = now();
} else if (now() - it.first > FLOW_KNOBS->TOO_MANY_CONNECTIONS_CLOSED_TIMEOUT) {
TraceEvent(SevWarnAlways, "TooManyConnectionsClosed", conn ? conn->getDebugID() : UID())
.suppressFor(5.0)
.detail("PeerAddr", self->destination);
self->transport->degraded->set(true);
}
it.second = now();
}
if (conn) {
if (self->destination.isPublic() && e.code() == error_code_connection_failed) {
FlowTransport::transport().healthMonitor()->reportPeerClosed(self->destination);
if (FLOW_KNOBS->HEALTH_MONITOR_MARK_FAILED_UNSTABLE_CONNECTIONS &&
FlowTransport::transport().healthMonitor()->tooManyConnectionsClosed(self->destination) &&
self->destination.isPublic()) {
TraceEvent("TooManyConnectionsClosedMarkFailed")
.detail("Dest", self->destination)
.detail(
"ClosedCount",
FlowTransport::transport().healthMonitor()->closedConnectionsCount(self->destination));
tooManyConnectionsClosed = true;
IFailureMonitor::failureMonitor().setStatus(self->destination, FailureStatus(true));
}
}
conn->close();
conn = Reference<IConnection>();
// Old versions will throw this error, and we don't want to forget their protocol versions.
// This means we can't tell the difference between an old protocol version and one we
// can no longer connect to.
if (e.code() != error_code_incompatible_protocol_version) {
self->protocolVersion->set(Optional<ProtocolVersion>());
}
}
// Clients might send more packets in response, which needs to go out on the next connection
IFailureMonitor::failureMonitor().notifyDisconnect(self->destination);
Promise<Void> disconnect = self->disconnect;
self->disconnect = Promise<Void>();
disconnect.send(Void());
if (e.code() == error_code_actor_cancelled)
throw;
// Try to recover, even from serious errors, by retrying
if (self->peerReferences <= 0 && self->reliable.empty() && self->unsent.empty() &&
self->outstandingReplies == 0) {
TraceEvent("PeerDestroy").errorUnsuppressed(e).suppressFor(1.0).detail("PeerAddr", self->destination);
self->connect.cancel();
self->transport->peers.erase(self->destination);
self->transport->orderedAddresses.erase(self->destination);
return Void();
}
}
}
}
Peer::Peer(TransportData* transport, NetworkAddress const& destination)
: transport(transport), destination(destination), compatible(true), connected(false), outgoingConnectionIdle(true),
lastConnectTime(0.0), reconnectionDelay(FLOW_KNOBS->INITIAL_RECONNECTION_TIME), peerReferences(-1),
bytesReceived(0), bytesSent(0), lastDataPacketSentTime(now()), outstandingReplies(0),
pingLatencies(destination.isPublic() ? FLOW_KNOBS->PING_SKETCH_ACCURACY : 0.1), lastLoggedTime(0.0),
lastLoggedBytesReceived(0), lastLoggedBytesSent(0), timeoutCount(0),
protocolVersion(Reference<AsyncVar<Optional<ProtocolVersion>>>(new AsyncVar<Optional<ProtocolVersion>>())),
connectOutgoingCount(0), connectIncomingCount(0), connectFailedCount(0),
connectLatencies(destination.isPublic() ? FLOW_KNOBS->PING_SKETCH_ACCURACY : 0.1) {
IFailureMonitor::failureMonitor().setStatus(destination, FailureStatus(false));
}
void Peer::send(PacketBuffer* pb, ReliablePacket* rp, bool firstUnsent) {
unsent.setWriteBuffer(pb);
if (rp)
reliable.insert(rp);
if (firstUnsent)
dataToSend.trigger();
}
void Peer::prependConnectPacket() {
// Send the ConnectPacket expected at the beginning of a new connection
ConnectPacket pkt;
if (transport->localAddresses.getAddressList().address.isTLS() == destination.isTLS()) {
pkt.canonicalRemotePort = transport->localAddresses.getAddressList().address.port;
pkt.setCanonicalRemoteIp(transport->localAddresses.getAddressList().address.ip);
} else if (transport->localAddresses.getAddressList().secondaryAddress.present()) {
pkt.canonicalRemotePort = transport->localAddresses.getAddressList().secondaryAddress.get().port;
pkt.setCanonicalRemoteIp(transport->localAddresses.getAddressList().secondaryAddress.get().ip);
} else {
// a "mixed" TLS/non-TLS connection is like a client/server connection - there's no way to reverse it
pkt.canonicalRemotePort = 0;
pkt.setCanonicalRemoteIp(IPAddress(0));
}
pkt.connectPacketLength = sizeof(pkt) - sizeof(pkt.connectPacketLength);
pkt.protocolVersion = g_network->protocolVersion();
pkt.protocolVersion.addObjectSerializerFlag();
pkt.connectionId = transport->transportId;
PacketBuffer *pb_first = PacketBuffer::create(), *pb_end = nullptr;
PacketWriter wr(pb_first, nullptr, Unversioned());
pkt.serialize(wr);
pb_end = wr.finish();
#if VALGRIND
SendBuffer* checkbuf = pb_first;
while (checkbuf) {
int size = checkbuf->bytes_written;
const uint8_t* data = checkbuf->data();
VALGRIND_CHECK_MEM_IS_DEFINED(data, size);
checkbuf = checkbuf->next;
}
#endif
unsent.prependWriteBuffer(pb_first, pb_end);
}
void Peer::discardUnreliablePackets() {
// Throw away the current unsent list, dropping the reference count on each PacketBuffer that accounts for presence
// in the unsent list
unsent.discardAll();
// If there are reliable packets, compact reliable packets into a new unsent range
if (!reliable.empty()) {
PacketBuffer* pb = unsent.getWriteBuffer();
pb = reliable.compact(pb, nullptr);
unsent.setWriteBuffer(pb);
}
}
void Peer::onIncomingConnection(Reference<Peer> self, Reference<IConnection> conn, Future<Void> reader) {
// In case two processes are trying to connect to each other simultaneously, the process with the larger canonical
// NetworkAddress gets to keep its outgoing connection.
++self->connectIncomingCount;
if (!destination.isPublic() && !outgoingConnectionIdle)
throw address_in_use();
NetworkAddress compatibleAddr = transport->localAddresses.getAddressList().address;
if (transport->localAddresses.getAddressList().secondaryAddress.present() &&
transport->localAddresses.getAddressList().secondaryAddress.get().isTLS() == destination.isTLS()) {
compatibleAddr = transport->localAddresses.getAddressList().secondaryAddress.get();
}
if (!destination.isPublic() || outgoingConnectionIdle || destination > compatibleAddr ||
(lastConnectTime > 1.0 && now() - lastConnectTime > FLOW_KNOBS->ALWAYS_ACCEPT_DELAY)) {
// Keep the new connection
TraceEvent("IncomingConnection"_audit, conn->getDebugID())
.suppressFor(1.0)
.detail("FromAddr", conn->getPeerAddress())
.detail("CanonicalAddr", destination)
.detail("IsPublic", destination.isPublic())
.detail("Trusted", self->transport->allowList(conn->getPeerAddress().ip) && conn->hasTrustedPeer());
connect.cancel();
prependConnectPacket();
connect = connectionKeeper(self, conn, reader);
} else {
TraceEvent("RedundantConnection", conn->getDebugID())
.suppressFor(1.0)
.detail("FromAddr", conn->getPeerAddress().toString())
.detail("CanonicalAddr", destination)
.detail("LocalAddr", compatibleAddr);
// Keep our prior connection
reader.cancel();
conn->close();
// Send an (ignored) packet to make sure that, if our outgoing connection died before the peer made this
// connection attempt, we eventually find out that our connection is dead, close it, and then respond to the
// next connection reattempt from peer.
}
}