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
|
/*
*
* Copyright 2017 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.
*
*/
#include <memory>
#include <mutex>
#include <sstream>
#include <thread>
#include <grpc++/channel.h>
#include <grpc++/client_context.h>
#include <grpc++/create_channel.h>
#include <grpc++/server.h>
#include <grpc++/server_builder.h>
#include <grpc/grpc.h>
#include <grpc/support/alloc.h>
#include <grpc/support/log.h>
#include <grpc/support/string_util.h>
#include <grpc/support/thd.h>
#include <grpc/support/time.h>
extern "C" {
#include "src/core/ext/filters/client_channel/resolver/fake/fake_resolver.h"
#include "src/core/lib/iomgr/sockaddr.h"
}
#include "test/core/util/port.h"
#include "test/core/util/test_config.h"
#include "test/cpp/end2end/test_service_impl.h"
#include "src/proto/grpc/lb/v1/load_balancer.grpc.pb.h"
#include "src/proto/grpc/testing/echo.grpc.pb.h"
#include <gmock/gmock.h>
#include <gtest/gtest.h>
// TODO(dgq): Other scenarios in need of testing:
// - Send a serverlist with faulty ip:port addresses (port > 2^16, etc).
// - Test reception of invalid serverlist
// - Test pinging
// - Test against a non-LB server.
// - Random LB server closing the stream unexpectedly.
// - Test using DNS-resolvable names (localhost?)
// - Test handling of creation of faulty RR instance by having the LB return a
// serverlist with non-existent backends after having initially returned a
// valid one.
//
// Findings from end to end testing to be covered here:
// - Handling of LB servers restart, including reconnection after backing-off
// retries.
// - Destruction of load balanced channel (and therefore of grpclb instance)
// while:
// 1) the internal LB call is still active. This should work by virtue
// of the weak reference the LB call holds. The call should be terminated as
// part of the grpclb shutdown process.
// 2) the retry timer is active. Again, the weak reference it holds should
// prevent a premature call to \a glb_destroy.
// - Restart of backend servers with no changes to serverlist. This exercises
// the RR handover mechanism.
using std::chrono::system_clock;
using grpc::lb::v1::LoadBalanceRequest;
using grpc::lb::v1::LoadBalanceResponse;
using grpc::lb::v1::LoadBalancer;
namespace grpc {
namespace testing {
namespace {
template <typename ServiceType>
class CountedService : public ServiceType {
public:
size_t request_count() {
std::unique_lock<std::mutex> lock(mu_);
return request_count_;
}
size_t response_count() {
std::unique_lock<std::mutex> lock(mu_);
return response_count_;
}
void IncreaseResponseCount() {
std::unique_lock<std::mutex> lock(mu_);
++response_count_;
}
void IncreaseRequestCount() {
std::unique_lock<std::mutex> lock(mu_);
++request_count_;
}
void ResetCounters() {
std::unique_lock<std::mutex> lock(mu_);
request_count_ = 0;
response_count_ = 0;
}
protected:
std::mutex mu_;
private:
size_t request_count_ = 0;
size_t response_count_ = 0;
};
using BackendService = CountedService<TestServiceImpl>;
using BalancerService = CountedService<LoadBalancer::Service>;
class BackendServiceImpl : public BackendService {
public:
BackendServiceImpl() {}
Status Echo(ServerContext* context, const EchoRequest* request,
EchoResponse* response) override {
IncreaseRequestCount();
const auto status = TestServiceImpl::Echo(context, request, response);
IncreaseResponseCount();
return status;
}
// Returns true on its first invocation, false otherwise.
bool Shutdown() {
std::unique_lock<std::mutex> lock(mu_);
const bool prev = !shutdown_;
shutdown_ = true;
gpr_log(GPR_INFO, "Backend: shut down");
return prev;
}
private:
std::mutex mu_;
bool shutdown_ = false;
};
grpc::string Ip4ToPackedString(const char* ip_str) {
struct in_addr ip4;
GPR_ASSERT(inet_pton(AF_INET, ip_str, &ip4) == 1);
return grpc::string(reinterpret_cast<const char*>(&ip4), sizeof(ip4));
}
struct ClientStats {
size_t num_calls_started = 0;
size_t num_calls_finished = 0;
size_t num_calls_finished_with_client_failed_to_send = 0;
size_t num_calls_finished_known_received = 0;
std::map<grpc::string, size_t> drop_token_counts;
ClientStats& operator+=(const ClientStats& other) {
num_calls_started += other.num_calls_started;
num_calls_finished += other.num_calls_finished;
num_calls_finished_with_client_failed_to_send +=
other.num_calls_finished_with_client_failed_to_send;
num_calls_finished_known_received +=
other.num_calls_finished_known_received;
for (const auto& p : other.drop_token_counts) {
drop_token_counts[p.first] += p.second;
}
return *this;
}
};
class BalancerServiceImpl : public BalancerService {
public:
using Stream = ServerReaderWriter<LoadBalanceResponse, LoadBalanceRequest>;
using ResponseDelayPair = std::pair<LoadBalanceResponse, int>;
explicit BalancerServiceImpl(int client_load_reporting_interval_seconds)
: client_load_reporting_interval_seconds_(
client_load_reporting_interval_seconds),
shutdown_(false) {}
Status BalanceLoad(ServerContext* context, Stream* stream) override {
gpr_log(GPR_INFO, "LB[%p]: BalanceLoad", this);
LoadBalanceRequest request;
stream->Read(&request);
IncreaseRequestCount();
gpr_log(GPR_INFO, "LB[%p]: recv msg '%s'", this,
request.DebugString().c_str());
if (client_load_reporting_interval_seconds_ > 0) {
LoadBalanceResponse initial_response;
initial_response.mutable_initial_response()
->mutable_client_stats_report_interval()
->set_seconds(client_load_reporting_interval_seconds_);
stream->Write(initial_response);
}
std::vector<ResponseDelayPair> responses_and_delays;
{
std::unique_lock<std::mutex> lock(mu_);
responses_and_delays = responses_and_delays_;
}
for (const auto& response_and_delay : responses_and_delays) {
{
std::unique_lock<std::mutex> lock(mu_);
if (shutdown_) goto done;
}
SendResponse(stream, response_and_delay.first, response_and_delay.second);
}
{
std::unique_lock<std::mutex> lock(mu_);
if (shutdown_) goto done;
serverlist_cond_.wait(lock, [this] { return serverlist_ready_; });
serverlist_ready_ = false;
}
if (client_load_reporting_interval_seconds_ > 0) {
request.Clear();
stream->Read(&request);
gpr_log(GPR_INFO, "LB[%p]: recv client load report msg: '%s'", this,
request.DebugString().c_str());
GPR_ASSERT(request.has_client_stats());
// We need to acquire the lock here in order to prevent the notify_one
// below from firing before its corresponding wait is executed.
std::lock_guard<std::mutex> lock(mu_);
client_stats_.num_calls_started +=
request.client_stats().num_calls_started();
client_stats_.num_calls_finished +=
request.client_stats().num_calls_finished();
client_stats_.num_calls_finished_with_client_failed_to_send +=
request.client_stats()
.num_calls_finished_with_client_failed_to_send();
client_stats_.num_calls_finished_known_received +=
request.client_stats().num_calls_finished_known_received();
for (const auto& drop_token_count :
request.client_stats().calls_finished_with_drop()) {
client_stats_
.drop_token_counts[drop_token_count.load_balance_token()] +=
drop_token_count.num_calls();
}
load_report_ready_ = true;
load_report_cond_.notify_one();
}
done:
gpr_log(GPR_INFO, "LB[%p]: done", this);
return Status::OK;
}
void add_response(const LoadBalanceResponse& response, int send_after_ms) {
std::unique_lock<std::mutex> lock(mu_);
responses_and_delays_.push_back(std::make_pair(response, send_after_ms));
}
// Returns true on its first invocation, false otherwise.
bool Shutdown() {
NotifyDoneWithServerlists();
std::unique_lock<std::mutex> lock(mu_);
const bool prev = !shutdown_;
shutdown_ = true;
gpr_log(GPR_INFO, "LB[%p]: shut down", this);
return prev;
}
static LoadBalanceResponse BuildResponseForBackends(
const std::vector<int>& backend_ports,
const std::map<grpc::string, size_t>& drop_token_counts) {
LoadBalanceResponse response;
for (const auto& drop_token_count : drop_token_counts) {
for (size_t i = 0; i < drop_token_count.second; ++i) {
auto* server = response.mutable_server_list()->add_servers();
server->set_drop(true);
server->set_load_balance_token(drop_token_count.first);
}
}
for (const int& backend_port : backend_ports) {
auto* server = response.mutable_server_list()->add_servers();
server->set_ip_address(Ip4ToPackedString("127.0.0.1"));
server->set_port(backend_port);
}
return response;
}
const ClientStats& WaitForLoadReport() {
std::unique_lock<std::mutex> lock(mu_);
load_report_cond_.wait(lock, [this] { return load_report_ready_; });
load_report_ready_ = false;
return client_stats_;
}
void NotifyDoneWithServerlists() {
std::lock_guard<std::mutex> lock(mu_);
serverlist_ready_ = true;
serverlist_cond_.notify_one();
}
private:
void SendResponse(Stream* stream, const LoadBalanceResponse& response,
int delay_ms) {
gpr_log(GPR_INFO, "LB[%p]: sleeping for %d ms...", this, delay_ms);
if (delay_ms > 0) {
gpr_sleep_until(
gpr_time_add(gpr_now(GPR_CLOCK_REALTIME),
gpr_time_from_millis(delay_ms, GPR_TIMESPAN)));
}
gpr_log(GPR_INFO, "LB[%p]: Woke up! Sending response '%s'", this,
response.DebugString().c_str());
IncreaseResponseCount();
stream->Write(response);
}
const int client_load_reporting_interval_seconds_;
std::vector<ResponseDelayPair> responses_and_delays_;
std::mutex mu_;
std::condition_variable load_report_cond_;
bool load_report_ready_ = false;
std::condition_variable serverlist_cond_;
bool serverlist_ready_ = false;
ClientStats client_stats_;
bool shutdown_;
};
class GrpclbEnd2endTest : public ::testing::Test {
protected:
GrpclbEnd2endTest(int num_backends, int num_balancers,
int client_load_reporting_interval_seconds)
: server_host_("localhost"),
num_backends_(num_backends),
num_balancers_(num_balancers),
client_load_reporting_interval_seconds_(
client_load_reporting_interval_seconds),
kRequestMessage_("Live long and prosper.") {}
void SetUp() override {
response_generator_ = grpc_fake_resolver_response_generator_create();
// Start the backends.
for (size_t i = 0; i < num_backends_; ++i) {
backends_.emplace_back(new BackendServiceImpl());
backend_servers_.emplace_back(ServerThread<BackendService>(
"backend", server_host_, backends_.back().get()));
}
// Start the load balancers.
for (size_t i = 0; i < num_balancers_; ++i) {
balancers_.emplace_back(
new BalancerServiceImpl(client_load_reporting_interval_seconds_));
balancer_servers_.emplace_back(ServerThread<BalancerService>(
"balancer", server_host_, balancers_.back().get()));
}
ResetStub();
std::vector<AddressData> addresses;
for (size_t i = 0; i < balancer_servers_.size(); ++i) {
addresses.emplace_back(AddressData{balancer_servers_[i].port_, true, ""});
}
SetNextResolution(addresses);
}
void TearDown() override {
for (size_t i = 0; i < backends_.size(); ++i) {
if (backends_[i]->Shutdown()) backend_servers_[i].Shutdown();
}
for (size_t i = 0; i < balancers_.size(); ++i) {
if (balancers_[i]->Shutdown()) balancer_servers_[i].Shutdown();
}
grpc_fake_resolver_response_generator_unref(response_generator_);
}
void ResetStub() {
ChannelArguments args;
args.SetPointer(GRPC_ARG_FAKE_RESOLVER_RESPONSE_GENERATOR,
response_generator_);
std::ostringstream uri;
uri << "fake:///servername_not_used";
channel_ =
CreateCustomChannel(uri.str(), InsecureChannelCredentials(), args);
stub_ = grpc::testing::EchoTestService::NewStub(channel_);
}
void ResetBackendCounters() {
for (const auto& backend : backends_) backend->ResetCounters();
}
ClientStats WaitForLoadReports() {
ClientStats client_stats;
for (const auto& balancer : balancers_) {
client_stats += balancer->WaitForLoadReport();
}
return client_stats;
}
bool SeenAllBackends() {
for (const auto& backend : backends_) {
if (backend->request_count() == 0) return false;
}
return true;
}
void SendRpcAndCount(int* num_total, int* num_ok, int* num_failure,
int* num_drops) {
const Status status = SendRpc();
if (status.ok()) {
++*num_ok;
} else {
if (status.error_message() == "Call dropped by load balancing policy") {
++*num_drops;
} else {
++*num_failure;
}
}
++*num_total;
}
std::tuple<int, int, int> WaitForAllBackends(
int num_requests_multiple_of = 1) {
int num_ok = 0;
int num_failure = 0;
int num_drops = 0;
int num_total = 0;
while (!SeenAllBackends()) {
SendRpcAndCount(&num_total, &num_ok, &num_failure, &num_drops);
}
while (num_total % num_requests_multiple_of != 0) {
SendRpcAndCount(&num_total, &num_ok, &num_failure, &num_drops);
}
ResetBackendCounters();
gpr_log(GPR_INFO,
"Performed %d warm up requests (a multiple of %d) against the "
"backends. %d succeeded, %d failed, %d dropped.",
num_total, num_requests_multiple_of, num_ok, num_failure,
num_drops);
return std::make_tuple(num_ok, num_failure, num_drops);
}
void WaitForBackend(size_t backend_idx) {
do {
CheckRpcSendOk();
} while (backends_[backend_idx]->request_count() == 0);
ResetBackendCounters();
}
struct AddressData {
int port;
bool is_balancer;
grpc::string balancer_name;
};
void SetNextResolution(const std::vector<AddressData>& address_data) {
grpc_exec_ctx exec_ctx = GRPC_EXEC_CTX_INIT;
grpc_lb_addresses* addresses =
grpc_lb_addresses_create(address_data.size(), nullptr);
for (size_t i = 0; i < address_data.size(); ++i) {
char* lb_uri_str;
gpr_asprintf(&lb_uri_str, "ipv4:127.0.0.1:%d", address_data[i].port);
grpc_uri* lb_uri = grpc_uri_parse(&exec_ctx, lb_uri_str, true);
GPR_ASSERT(lb_uri != nullptr);
grpc_lb_addresses_set_address_from_uri(
addresses, i, lb_uri, address_data[i].is_balancer,
address_data[i].balancer_name.c_str(), nullptr);
grpc_uri_destroy(lb_uri);
gpr_free(lb_uri_str);
}
grpc_arg fake_addresses = grpc_lb_addresses_create_channel_arg(addresses);
grpc_channel_args fake_result = {1, &fake_addresses};
grpc_fake_resolver_response_generator_set_response(
&exec_ctx, response_generator_, &fake_result);
grpc_lb_addresses_destroy(&exec_ctx, addresses);
grpc_exec_ctx_finish(&exec_ctx);
}
const std::vector<int> GetBackendPorts() const {
std::vector<int> backend_ports;
for (const auto& bs : backend_servers_) {
backend_ports.push_back(bs.port_);
}
return backend_ports;
}
void ScheduleResponseForBalancer(size_t i,
const LoadBalanceResponse& response,
int delay_ms) {
balancers_.at(i)->add_response(response, delay_ms);
}
Status SendRpc(EchoResponse* response = nullptr, int timeout_ms = 1000) {
const bool local_response = (response == nullptr);
if (local_response) response = new EchoResponse;
EchoRequest request;
request.set_message(kRequestMessage_);
ClientContext context;
context.set_deadline(grpc_timeout_milliseconds_to_deadline(timeout_ms));
Status status = stub_->Echo(&context, request, response);
if (local_response) delete response;
return status;
}
void CheckRpcSendOk(const size_t times = 1) {
for (size_t i = 0; i < times; ++i) {
EchoResponse response;
const Status status = SendRpc(&response);
EXPECT_TRUE(status.ok()) << "code=" << status.error_code()
<< " message=" << status.error_message();
EXPECT_EQ(response.message(), kRequestMessage_);
}
}
void CheckRpcSendFailure() {
const Status status = SendRpc();
EXPECT_FALSE(status.ok());
}
template <typename T>
struct ServerThread {
explicit ServerThread(const grpc::string& type,
const grpc::string& server_host, T* service)
: type_(type), service_(service) {
std::mutex mu;
// We need to acquire the lock here in order to prevent the notify_one
// by ServerThread::Start from firing before the wait below is hit.
std::unique_lock<std::mutex> lock(mu);
port_ = grpc_pick_unused_port_or_die();
gpr_log(GPR_INFO, "starting %s server on port %d", type_.c_str(), port_);
std::condition_variable cond;
thread_.reset(new std::thread(
std::bind(&ServerThread::Start, this, server_host, &mu, &cond)));
cond.wait(lock);
gpr_log(GPR_INFO, "%s server startup complete", type_.c_str());
}
void Start(const grpc::string& server_host, std::mutex* mu,
std::condition_variable* cond) {
// We need to acquire the lock here in order to prevent the notify_one
// below from firing before its corresponding wait is executed.
std::lock_guard<std::mutex> lock(*mu);
std::ostringstream server_address;
server_address << server_host << ":" << port_;
ServerBuilder builder;
builder.AddListeningPort(server_address.str(),
InsecureServerCredentials());
builder.RegisterService(service_);
server_ = builder.BuildAndStart();
cond->notify_one();
}
void Shutdown() {
gpr_log(GPR_INFO, "%s about to shutdown", type_.c_str());
server_->Shutdown(grpc_timeout_milliseconds_to_deadline(0));
thread_->join();
gpr_log(GPR_INFO, "%s shutdown completed", type_.c_str());
}
int port_;
grpc::string type_;
std::unique_ptr<Server> server_;
T* service_;
std::unique_ptr<std::thread> thread_;
};
const grpc::string kMessage_ = "Live long and prosper.";
const grpc::string server_host_;
const size_t num_backends_;
const size_t num_balancers_;
const int client_load_reporting_interval_seconds_;
std::shared_ptr<Channel> channel_;
std::unique_ptr<grpc::testing::EchoTestService::Stub> stub_;
std::vector<std::unique_ptr<BackendServiceImpl>> backends_;
std::vector<std::unique_ptr<BalancerServiceImpl>> balancers_;
std::vector<ServerThread<BackendService>> backend_servers_;
std::vector<ServerThread<BalancerService>> balancer_servers_;
grpc_fake_resolver_response_generator* response_generator_;
const grpc::string kRequestMessage_;
};
class SingleBalancerTest : public GrpclbEnd2endTest {
public:
SingleBalancerTest() : GrpclbEnd2endTest(4, 1, 0) {}
};
TEST_F(SingleBalancerTest, Vanilla) {
const size_t kNumRpcsPerAddress = 100;
ScheduleResponseForBalancer(
0, BalancerServiceImpl::BuildResponseForBackends(GetBackendPorts(), {}),
0);
// Make sure that trying to connect works without a call.
channel_->GetState(true /* try_to_connect */);
// We need to wait for all backends to come online.
WaitForAllBackends();
// Send kNumRpcsPerAddress RPCs per server.
CheckRpcSendOk(kNumRpcsPerAddress * num_backends_);
// Each backend should have gotten 100 requests.
for (size_t i = 0; i < backends_.size(); ++i) {
EXPECT_EQ(kNumRpcsPerAddress,
backend_servers_[i].service_->request_count());
}
balancers_[0]->NotifyDoneWithServerlists();
// The balancer got a single request.
EXPECT_EQ(1U, balancer_servers_[0].service_->request_count());
// and sent a single response.
EXPECT_EQ(1U, balancer_servers_[0].service_->response_count());
// Check LB policy name for the channel.
EXPECT_EQ("grpclb", channel_->GetLoadBalancingPolicyName());
}
TEST_F(SingleBalancerTest, InitiallyEmptyServerlist) {
const int kServerlistDelayMs = 500 * grpc_test_slowdown_factor();
const int kCallDeadlineMs = 1000 * grpc_test_slowdown_factor();
// First response is an empty serverlist, sent right away.
ScheduleResponseForBalancer(0, LoadBalanceResponse(), 0);
// Send non-empty serverlist only after kServerlistDelayMs
ScheduleResponseForBalancer(
0, BalancerServiceImpl::BuildResponseForBackends(GetBackendPorts(), {}),
kServerlistDelayMs);
const auto t0 = system_clock::now();
// Client will block: LB will initially send empty serverlist.
CheckRpcSendOk(num_backends_);
const auto ellapsed_ms =
std::chrono::duration_cast<std::chrono::milliseconds>(
system_clock::now() - t0);
// but eventually, the LB sends a serverlist update that allows the call to
// proceed. The call delay must be larger than the delay in sending the
// populated serverlist but under the call's deadline.
EXPECT_GT(ellapsed_ms.count(), kServerlistDelayMs);
EXPECT_LT(ellapsed_ms.count(), kCallDeadlineMs);
// Each backend should have gotten 1 request.
for (size_t i = 0; i < backends_.size(); ++i) {
EXPECT_EQ(1U, backend_servers_[i].service_->request_count());
}
balancers_[0]->NotifyDoneWithServerlists();
// The balancer got a single request.
EXPECT_EQ(1U, balancer_servers_[0].service_->request_count());
// and sent two responses.
EXPECT_EQ(2U, balancer_servers_[0].service_->response_count());
// Check LB policy name for the channel.
EXPECT_EQ("grpclb", channel_->GetLoadBalancingPolicyName());
}
TEST_F(SingleBalancerTest, BackendsRestart) {
const size_t kNumRpcsPerAddress = 100;
ScheduleResponseForBalancer(
0, BalancerServiceImpl::BuildResponseForBackends(GetBackendPorts(), {}),
0);
// Make sure that trying to connect works without a call.
channel_->GetState(true /* try_to_connect */);
// Send kNumRpcsPerAddress RPCs per server.
CheckRpcSendOk(kNumRpcsPerAddress * num_backends_);
balancers_[0]->NotifyDoneWithServerlists();
// The balancer got a single request.
EXPECT_EQ(1U, balancer_servers_[0].service_->request_count());
// and sent a single response.
EXPECT_EQ(1U, balancer_servers_[0].service_->response_count());
for (size_t i = 0; i < backends_.size(); ++i) {
if (backends_[i]->Shutdown()) backend_servers_[i].Shutdown();
}
CheckRpcSendFailure();
for (size_t i = 0; i < num_backends_; ++i) {
backends_.emplace_back(new BackendServiceImpl());
backend_servers_.emplace_back(ServerThread<BackendService>(
"backend", server_host_, backends_.back().get()));
}
// The following RPC will fail due to the backend ports having changed. It
// will nonetheless exercise the grpclb-roundrobin handling of the RR policy
// having gone into shutdown.
// TODO(dgq): implement the "backend restart" component as well. We need extra
// machinery to either update the LB responses "on the fly" or instruct
// backends which ports to restart on.
CheckRpcSendFailure();
// Check LB policy name for the channel.
EXPECT_EQ("grpclb", channel_->GetLoadBalancingPolicyName());
}
class UpdatesTest : public GrpclbEnd2endTest {
public:
UpdatesTest() : GrpclbEnd2endTest(4, 3, 0) {}
};
TEST_F(UpdatesTest, UpdateBalancers) {
const std::vector<int> first_backend{GetBackendPorts()[0]};
const std::vector<int> second_backend{GetBackendPorts()[1]};
ScheduleResponseForBalancer(
0, BalancerServiceImpl::BuildResponseForBackends(first_backend, {}), 0);
ScheduleResponseForBalancer(
1, BalancerServiceImpl::BuildResponseForBackends(second_backend, {}), 0);
// Start servers and send 10 RPCs per server.
gpr_log(GPR_INFO, "========= BEFORE FIRST BATCH ==========");
CheckRpcSendOk(10);
gpr_log(GPR_INFO, "========= DONE WITH FIRST BATCH ==========");
// All 10 requests should have gone to the first backend.
EXPECT_EQ(10U, backend_servers_[0].service_->request_count());
balancers_[0]->NotifyDoneWithServerlists();
balancers_[1]->NotifyDoneWithServerlists();
balancers_[2]->NotifyDoneWithServerlists();
// Balancer 0 got a single request.
EXPECT_EQ(1U, balancer_servers_[0].service_->request_count());
// and sent a single response.
EXPECT_EQ(1U, balancer_servers_[0].service_->response_count());
EXPECT_EQ(0U, balancer_servers_[1].service_->request_count());
EXPECT_EQ(0U, balancer_servers_[1].service_->response_count());
EXPECT_EQ(0U, balancer_servers_[2].service_->request_count());
EXPECT_EQ(0U, balancer_servers_[2].service_->response_count());
std::vector<AddressData> addresses;
addresses.emplace_back(AddressData{balancer_servers_[1].port_, true, ""});
gpr_log(GPR_INFO, "========= ABOUT TO UPDATE 1 ==========");
SetNextResolution(addresses);
gpr_log(GPR_INFO, "========= UPDATE 1 DONE ==========");
// Wait until update has been processed, as signaled by the second backend
// receiving a request.
EXPECT_EQ(0U, backend_servers_[1].service_->request_count());
WaitForBackend(1);
backend_servers_[1].service_->ResetCounters();
gpr_log(GPR_INFO, "========= BEFORE SECOND BATCH ==========");
CheckRpcSendOk(10);
gpr_log(GPR_INFO, "========= DONE WITH SECOND BATCH ==========");
// All 10 requests should have gone to the second backend.
EXPECT_EQ(10U, backend_servers_[1].service_->request_count());
balancers_[0]->NotifyDoneWithServerlists();
balancers_[1]->NotifyDoneWithServerlists();
balancers_[2]->NotifyDoneWithServerlists();
EXPECT_EQ(1U, balancer_servers_[0].service_->request_count());
EXPECT_EQ(1U, balancer_servers_[0].service_->response_count());
EXPECT_EQ(1U, balancer_servers_[1].service_->request_count());
EXPECT_EQ(1U, balancer_servers_[1].service_->response_count());
EXPECT_EQ(0U, balancer_servers_[2].service_->request_count());
EXPECT_EQ(0U, balancer_servers_[2].service_->response_count());
// Check LB policy name for the channel.
EXPECT_EQ("grpclb", channel_->GetLoadBalancingPolicyName());
}
// Send an update with the same set of LBs as the one in SetUp() in order to
// verify that the LB channel inside grpclb keeps the initial connection (which
// by definition is also present in the update).
TEST_F(UpdatesTest, UpdateBalancersRepeated) {
const std::vector<int> first_backend{GetBackendPorts()[0]};
const std::vector<int> second_backend{GetBackendPorts()[0]};
ScheduleResponseForBalancer(
0, BalancerServiceImpl::BuildResponseForBackends(first_backend, {}), 0);
ScheduleResponseForBalancer(
1, BalancerServiceImpl::BuildResponseForBackends(second_backend, {}), 0);
// Start servers and send 10 RPCs per server.
gpr_log(GPR_INFO, "========= BEFORE FIRST BATCH ==========");
CheckRpcSendOk(10);
gpr_log(GPR_INFO, "========= DONE WITH FIRST BATCH ==========");
// All 10 requests should have gone to the first backend.
EXPECT_EQ(10U, backend_servers_[0].service_->request_count());
balancers_[0]->NotifyDoneWithServerlists();
// Balancer 0 got a single request.
EXPECT_EQ(1U, balancer_servers_[0].service_->request_count());
// and sent a single response.
EXPECT_EQ(1U, balancer_servers_[0].service_->response_count());
EXPECT_EQ(0U, balancer_servers_[1].service_->request_count());
EXPECT_EQ(0U, balancer_servers_[1].service_->response_count());
EXPECT_EQ(0U, balancer_servers_[2].service_->request_count());
EXPECT_EQ(0U, balancer_servers_[2].service_->response_count());
std::vector<AddressData> addresses;
addresses.emplace_back(AddressData{balancer_servers_[0].port_, true, ""});
addresses.emplace_back(AddressData{balancer_servers_[1].port_, true, ""});
addresses.emplace_back(AddressData{balancer_servers_[2].port_, true, ""});
gpr_log(GPR_INFO, "========= ABOUT TO UPDATE 1 ==========");
SetNextResolution(addresses);
gpr_log(GPR_INFO, "========= UPDATE 1 DONE ==========");
EXPECT_EQ(0U, backend_servers_[1].service_->request_count());
gpr_timespec deadline = gpr_time_add(
gpr_now(GPR_CLOCK_REALTIME), gpr_time_from_millis(10000, GPR_TIMESPAN));
// Send 10 seconds worth of RPCs
do {
CheckRpcSendOk();
} while (gpr_time_cmp(gpr_now(GPR_CLOCK_REALTIME), deadline) < 0);
// grpclb continued using the original LB call to the first balancer, which
// doesn't assign the second backend.
EXPECT_EQ(0U, backend_servers_[1].service_->request_count());
balancers_[0]->NotifyDoneWithServerlists();
addresses.clear();
addresses.emplace_back(AddressData{balancer_servers_[0].port_, true, ""});
addresses.emplace_back(AddressData{balancer_servers_[1].port_, true, ""});
gpr_log(GPR_INFO, "========= ABOUT TO UPDATE 2 ==========");
SetNextResolution(addresses);
gpr_log(GPR_INFO, "========= UPDATE 2 DONE ==========");
EXPECT_EQ(0U, backend_servers_[1].service_->request_count());
deadline = gpr_time_add(gpr_now(GPR_CLOCK_REALTIME),
gpr_time_from_millis(10000, GPR_TIMESPAN));
// Send 10 seconds worth of RPCs
do {
CheckRpcSendOk();
} while (gpr_time_cmp(gpr_now(GPR_CLOCK_REALTIME), deadline) < 0);
// grpclb continued using the original LB call to the first balancer, which
// doesn't assign the second backend.
EXPECT_EQ(0U, backend_servers_[1].service_->request_count());
balancers_[0]->NotifyDoneWithServerlists();
// Check LB policy name for the channel.
EXPECT_EQ("grpclb", channel_->GetLoadBalancingPolicyName());
}
TEST_F(UpdatesTest, UpdateBalancersDeadUpdate) {
const std::vector<int> first_backend{GetBackendPorts()[0]};
const std::vector<int> second_backend{GetBackendPorts()[1]};
ScheduleResponseForBalancer(
0, BalancerServiceImpl::BuildResponseForBackends(first_backend, {}), 0);
ScheduleResponseForBalancer(
1, BalancerServiceImpl::BuildResponseForBackends(second_backend, {}), 0);
// Start servers and send 10 RPCs per server.
gpr_log(GPR_INFO, "========= BEFORE FIRST BATCH ==========");
CheckRpcSendOk(10);
gpr_log(GPR_INFO, "========= DONE WITH FIRST BATCH ==========");
// All 10 requests should have gone to the first backend.
EXPECT_EQ(10U, backend_servers_[0].service_->request_count());
// Kill balancer 0
gpr_log(GPR_INFO, "********** ABOUT TO KILL BALANCER 0 *************");
balancers_[0]->NotifyDoneWithServerlists();
if (balancers_[0]->Shutdown()) balancer_servers_[0].Shutdown();
gpr_log(GPR_INFO, "********** KILLED BALANCER 0 *************");
// This is serviced by the existing RR policy
gpr_log(GPR_INFO, "========= BEFORE SECOND BATCH ==========");
CheckRpcSendOk(10);
gpr_log(GPR_INFO, "========= DONE WITH SECOND BATCH ==========");
// All 10 requests should again have gone to the first backend.
EXPECT_EQ(20U, backend_servers_[0].service_->request_count());
EXPECT_EQ(0U, backend_servers_[1].service_->request_count());
balancers_[0]->NotifyDoneWithServerlists();
balancers_[1]->NotifyDoneWithServerlists();
balancers_[2]->NotifyDoneWithServerlists();
// Balancer 0 got a single request.
EXPECT_EQ(1U, balancer_servers_[0].service_->request_count());
// and sent a single response.
EXPECT_EQ(1U, balancer_servers_[0].service_->response_count());
EXPECT_EQ(0U, balancer_servers_[1].service_->request_count());
EXPECT_EQ(0U, balancer_servers_[1].service_->response_count());
EXPECT_EQ(0U, balancer_servers_[2].service_->request_count());
EXPECT_EQ(0U, balancer_servers_[2].service_->response_count());
std::vector<AddressData> addresses;
addresses.emplace_back(AddressData{balancer_servers_[1].port_, true, ""});
gpr_log(GPR_INFO, "========= ABOUT TO UPDATE 1 ==========");
SetNextResolution(addresses);
gpr_log(GPR_INFO, "========= UPDATE 1 DONE ==========");
// Wait until update has been processed, as signaled by the second backend
// receiving a request. In the meantime, the client continues to be serviced
// (by the first backend) without interruption.
EXPECT_EQ(0U, backend_servers_[1].service_->request_count());
WaitForBackend(1);
// This is serviced by the existing RR policy
backend_servers_[1].service_->ResetCounters();
gpr_log(GPR_INFO, "========= BEFORE THIRD BATCH ==========");
CheckRpcSendOk(10);
gpr_log(GPR_INFO, "========= DONE WITH THIRD BATCH ==========");
// All 10 requests should have gone to the second backend.
EXPECT_EQ(10U, backend_servers_[1].service_->request_count());
balancers_[0]->NotifyDoneWithServerlists();
balancers_[1]->NotifyDoneWithServerlists();
balancers_[2]->NotifyDoneWithServerlists();
EXPECT_EQ(1U, balancer_servers_[0].service_->request_count());
EXPECT_EQ(1U, balancer_servers_[0].service_->response_count());
EXPECT_EQ(1U, balancer_servers_[1].service_->request_count());
EXPECT_EQ(1U, balancer_servers_[1].service_->response_count());
EXPECT_EQ(0U, balancer_servers_[2].service_->request_count());
EXPECT_EQ(0U, balancer_servers_[2].service_->response_count());
// Check LB policy name for the channel.
EXPECT_EQ("grpclb", channel_->GetLoadBalancingPolicyName());
}
TEST_F(SingleBalancerTest, Drop) {
const size_t kNumRpcsPerAddress = 100;
const int num_of_drop_by_rate_limiting_addresses = 1;
const int num_of_drop_by_load_balancing_addresses = 2;
const int num_of_drop_addresses = num_of_drop_by_rate_limiting_addresses +
num_of_drop_by_load_balancing_addresses;
const int num_total_addresses = num_backends_ + num_of_drop_addresses;
ScheduleResponseForBalancer(
0, BalancerServiceImpl::BuildResponseForBackends(
GetBackendPorts(),
{{"rate_limiting", num_of_drop_by_rate_limiting_addresses},
{"load_balancing", num_of_drop_by_load_balancing_addresses}}),
0);
// Wait until all backends are ready.
WaitForAllBackends();
// Send kNumRpcsPerAddress RPCs for each server and drop address.
size_t num_drops = 0;
for (size_t i = 0; i < kNumRpcsPerAddress * num_total_addresses; ++i) {
EchoResponse response;
const Status status = SendRpc(&response);
if (!status.ok() &&
status.error_message() == "Call dropped by load balancing policy") {
++num_drops;
} else {
EXPECT_TRUE(status.ok()) << "code=" << status.error_code()
<< " message=" << status.error_message();
EXPECT_EQ(response.message(), kMessage_);
}
}
EXPECT_EQ(kNumRpcsPerAddress * num_of_drop_addresses, num_drops);
// Each backend should have gotten 100 requests.
for (size_t i = 0; i < backends_.size(); ++i) {
EXPECT_EQ(kNumRpcsPerAddress,
backend_servers_[i].service_->request_count());
}
// The balancer got a single request.
EXPECT_EQ(1U, balancer_servers_[0].service_->request_count());
// and sent a single response.
EXPECT_EQ(1U, balancer_servers_[0].service_->response_count());
}
TEST_F(SingleBalancerTest, DropAllFirst) {
// All registered addresses are marked as "drop".
const int num_of_drop_by_rate_limiting_addresses = 1;
const int num_of_drop_by_load_balancing_addresses = 1;
ScheduleResponseForBalancer(
0, BalancerServiceImpl::BuildResponseForBackends(
{}, {{"rate_limiting", num_of_drop_by_rate_limiting_addresses},
{"load_balancing", num_of_drop_by_load_balancing_addresses}}),
0);
const Status status = SendRpc();
EXPECT_FALSE(status.ok());
EXPECT_EQ(status.error_message(), "Call dropped by load balancing policy");
}
TEST_F(SingleBalancerTest, DropAll) {
ScheduleResponseForBalancer(
0, BalancerServiceImpl::BuildResponseForBackends(GetBackendPorts(), {}),
0);
const int num_of_drop_by_rate_limiting_addresses = 1;
const int num_of_drop_by_load_balancing_addresses = 1;
ScheduleResponseForBalancer(
0, BalancerServiceImpl::BuildResponseForBackends(
{}, {{"rate_limiting", num_of_drop_by_rate_limiting_addresses},
{"load_balancing", num_of_drop_by_load_balancing_addresses}}),
1000);
// First call succeeds.
CheckRpcSendOk();
// But eventually, the update with only dropped servers is processed and calls
// fail.
Status status;
do {
status = SendRpc();
} while (status.ok());
EXPECT_FALSE(status.ok());
EXPECT_EQ(status.error_message(), "Call dropped by load balancing policy");
}
class SingleBalancerWithClientLoadReportingTest : public GrpclbEnd2endTest {
public:
SingleBalancerWithClientLoadReportingTest() : GrpclbEnd2endTest(4, 1, 2) {}
};
TEST_F(SingleBalancerWithClientLoadReportingTest, Vanilla) {
const size_t kNumRpcsPerAddress = 100;
ScheduleResponseForBalancer(
0, BalancerServiceImpl::BuildResponseForBackends(GetBackendPorts(), {}),
0);
// Wait until all backends are ready.
int num_ok = 0;
int num_failure = 0;
int num_drops = 0;
std::tie(num_ok, num_failure, num_drops) = WaitForAllBackends();
// Send kNumRpcsPerAddress RPCs per server.
CheckRpcSendOk(kNumRpcsPerAddress * num_backends_);
// Each backend should have gotten 100 requests.
for (size_t i = 0; i < backends_.size(); ++i) {
EXPECT_EQ(kNumRpcsPerAddress,
backend_servers_[i].service_->request_count());
}
balancers_[0]->NotifyDoneWithServerlists();
// The balancer got a single request.
EXPECT_EQ(1U, balancer_servers_[0].service_->request_count());
// and sent a single response.
EXPECT_EQ(1U, balancer_servers_[0].service_->response_count());
const ClientStats client_stats = WaitForLoadReports();
EXPECT_EQ(kNumRpcsPerAddress * num_backends_ + num_ok,
client_stats.num_calls_started);
EXPECT_EQ(kNumRpcsPerAddress * num_backends_ + num_ok,
client_stats.num_calls_finished);
EXPECT_EQ(0U, client_stats.num_calls_finished_with_client_failed_to_send);
EXPECT_EQ(kNumRpcsPerAddress * num_backends_ + (num_ok + num_drops),
client_stats.num_calls_finished_known_received);
EXPECT_THAT(client_stats.drop_token_counts, ::testing::ElementsAre());
}
TEST_F(SingleBalancerWithClientLoadReportingTest, Drop) {
const size_t kNumRpcsPerAddress = 3;
const int num_of_drop_by_rate_limiting_addresses = 2;
const int num_of_drop_by_load_balancing_addresses = 1;
const int num_of_drop_addresses = num_of_drop_by_rate_limiting_addresses +
num_of_drop_by_load_balancing_addresses;
const int num_total_addresses = num_backends_ + num_of_drop_addresses;
ScheduleResponseForBalancer(
0, BalancerServiceImpl::BuildResponseForBackends(
GetBackendPorts(),
{{"rate_limiting", num_of_drop_by_rate_limiting_addresses},
{"load_balancing", num_of_drop_by_load_balancing_addresses}}),
0);
// Wait until all backends are ready.
int num_warmup_ok = 0;
int num_warmup_failure = 0;
int num_warmup_drops = 0;
std::tie(num_warmup_ok, num_warmup_failure, num_warmup_drops) =
WaitForAllBackends(num_total_addresses /* num_requests_multiple_of */);
const int num_total_warmup_requests =
num_warmup_ok + num_warmup_failure + num_warmup_drops;
size_t num_drops = 0;
for (size_t i = 0; i < kNumRpcsPerAddress * num_total_addresses; ++i) {
EchoResponse response;
const Status status = SendRpc(&response);
if (!status.ok() &&
status.error_message() == "Call dropped by load balancing policy") {
++num_drops;
} else {
EXPECT_TRUE(status.ok()) << "code=" << status.error_code()
<< " message=" << status.error_message();
EXPECT_EQ(response.message(), kMessage_);
}
}
EXPECT_EQ(kNumRpcsPerAddress * num_of_drop_addresses, num_drops);
// Each backend should have gotten 100 requests.
for (size_t i = 0; i < backends_.size(); ++i) {
EXPECT_EQ(kNumRpcsPerAddress,
backend_servers_[i].service_->request_count());
}
balancers_[0]->NotifyDoneWithServerlists();
// The balancer got a single request.
EXPECT_EQ(1U, balancer_servers_[0].service_->request_count());
// and sent a single response.
EXPECT_EQ(1U, balancer_servers_[0].service_->response_count());
const ClientStats client_stats = WaitForLoadReports();
EXPECT_EQ(
kNumRpcsPerAddress * num_total_addresses + num_total_warmup_requests,
client_stats.num_calls_started);
EXPECT_EQ(
kNumRpcsPerAddress * num_total_addresses + num_total_warmup_requests,
client_stats.num_calls_finished);
EXPECT_EQ(0U, client_stats.num_calls_finished_with_client_failed_to_send);
EXPECT_EQ(kNumRpcsPerAddress * num_backends_ + num_warmup_ok,
client_stats.num_calls_finished_known_received);
// The number of warmup request is a multiple of the number of addresses.
// Therefore, all addresses in the scheduled balancer response are hit the
// same number of times.
const int num_times_drop_addresses_hit =
num_warmup_drops / num_of_drop_addresses;
EXPECT_THAT(
client_stats.drop_token_counts,
::testing::ElementsAre(
::testing::Pair("load_balancing",
(kNumRpcsPerAddress + num_times_drop_addresses_hit)),
::testing::Pair(
"rate_limiting",
(kNumRpcsPerAddress + num_times_drop_addresses_hit) * 2)));
}
} // namespace
} // namespace testing
} // namespace grpc
int main(int argc, char** argv) {
grpc_init();
grpc_test_init(argc, argv);
::testing::InitGoogleTest(&argc, argv);
const auto result = RUN_ALL_TESTS();
grpc_shutdown();
return result;
}
|