forked from heavyai/heavydb
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathDBHandler.cpp
7331 lines (6787 loc) · 294 KB
/
DBHandler.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
/*
* Copyright 2021 OmniSci, Inc.
*
* 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.
*/
/*
* File: DBHandler.cpp
* Author: michael
*
* Created on Jan 1, 2017, 12:40 PM
*/
#include "DBHandler.h"
#include "DistributedLoader.h"
#include "TokenCompletionHints.h"
#ifdef HAVE_PROFILER
#include <gperftools/heap-profiler.h>
#endif // HAVE_PROFILER
#include "MapDRelease.h"
#include "Calcite/Calcite.h"
#include "gen-cpp/CalciteServer.h"
#include "QueryEngine/RelAlgExecutor.h"
#include "Catalog/Catalog.h"
#include "Catalog/DdlCommandExecutor.h"
#include "DataMgr/ForeignStorage/ArrowForeignStorage.h"
#include "DataMgr/ForeignStorage/DummyForeignStorage.h"
#include "DistributedHandler.h"
#include "Fragmenter/InsertOrderFragmenter.h"
#include "Geospatial/Compression.h"
#include "Geospatial/GDAL.h"
#include "Geospatial/Transforms.h"
#include "Geospatial/Types.h"
#include "ImportExport/Importer.h"
#include "LockMgr/LockMgr.h"
#include "OSDependent/omnisci_hostname.h"
#include "Parser/ParserWrapper.h"
#include "Parser/ReservedKeywords.h"
#include "Parser/parser.h"
#include "QueryEngine/ArrowResultSet.h"
#include "QueryEngine/CalciteAdapter.h"
#include "QueryEngine/Execute.h"
#include "QueryEngine/ExtensionFunctionsWhitelist.h"
#include "QueryEngine/GpuMemUtils.h"
#include "QueryEngine/JoinFilterPushDown.h"
#include "QueryEngine/JsonAccessors.h"
#include "QueryEngine/QueryDispatchQueue.h"
#include "QueryEngine/ResultSetBuilder.h"
#include "QueryEngine/TableFunctions/TableFunctionsFactory.h"
#include "QueryEngine/TableOptimizer.h"
#include "QueryEngine/ThriftSerializers.h"
#include "Shared/ArrowUtil.h"
#include "Shared/StringTransform.h"
#include "Shared/import_helpers.h"
#include "Shared/mapd_shared_mutex.h"
#include "Shared/measure.h"
#include "Shared/scope.h"
#include "UdfCompiler/UdfCompiler.h"
#ifdef HAVE_AWS_S3
#include <aws/core/auth/AWSCredentialsProviderChain.h>
#endif
#include <fcntl.h>
#include <picosha2.h>
#include <sys/types.h>
#include <algorithm>
#include <boost/algorithm/string.hpp>
#include <boost/filesystem.hpp>
#include <boost/make_shared.hpp>
#include <boost/process/search_path.hpp>
#include <boost/program_options.hpp>
#include <boost/regex.hpp>
#include <boost/tokenizer.hpp>
#include <cmath>
#include <csignal>
#include <fstream>
#include <future>
#include <map>
#include <memory>
#include <random>
#include <regex>
#include <string>
#include <thread>
#include <typeinfo>
#include <arrow/api.h>
#include <arrow/io/api.h>
#include <arrow/ipc/api.h>
#include "Shared/ArrowUtil.h"
#define ENABLE_GEO_IMPORT_COLUMN_MATCHING 0
#ifdef HAVE_AWS_S3
extern bool g_allow_s3_server_privileges;
#endif
using Catalog_Namespace::Catalog;
using Catalog_Namespace::SysCatalog;
#define INVALID_SESSION_ID ""
#define THROW_MAPD_EXCEPTION(errstr) \
{ \
TOmniSciException ex; \
ex.error_msg = errstr; \
LOG(ERROR) << ex.error_msg; \
throw ex; \
}
thread_local std::string TrackingProcessor::client_address;
thread_local ClientProtocol TrackingProcessor::client_protocol;
namespace {
SessionMap::iterator get_session_from_map(const TSessionId& session,
SessionMap& session_map) {
auto session_it = session_map.find(session);
if (session_it == session_map.end()) {
THROW_MAPD_EXCEPTION("Session not valid.");
}
return session_it;
}
struct ForceDisconnect : public std::runtime_error {
ForceDisconnect(const std::string& cause) : std::runtime_error(cause) {}
};
} // namespace
template <>
SessionMap::iterator DBHandler::get_session_it_unsafe(
const TSessionId& session,
mapd_shared_lock<mapd_shared_mutex>& read_lock) {
auto session_it = get_session_from_map(session, sessions_);
try {
check_session_exp_unsafe(session_it);
} catch (const ForceDisconnect& e) {
read_lock.unlock();
mapd_unique_lock<mapd_shared_mutex> write_lock(sessions_mutex_);
auto session_it2 = get_session_from_map(session, sessions_);
disconnect_impl(session_it2, write_lock);
THROW_MAPD_EXCEPTION(e.what());
}
return session_it;
}
template <>
SessionMap::iterator DBHandler::get_session_it_unsafe(
const TSessionId& session,
mapd_unique_lock<mapd_shared_mutex>& write_lock) {
auto session_it = get_session_from_map(session, sessions_);
try {
check_session_exp_unsafe(session_it);
} catch (const ForceDisconnect& e) {
disconnect_impl(session_it, write_lock);
THROW_MAPD_EXCEPTION(e.what());
}
return session_it;
}
template <>
void DBHandler::expire_idle_sessions_unsafe(
mapd_unique_lock<mapd_shared_mutex>& write_lock) {
std::vector<std::string> expired_sessions;
for (auto session_pair : sessions_) {
auto session_it = get_session_from_map(session_pair.first, sessions_);
try {
check_session_exp_unsafe(session_it);
} catch (const ForceDisconnect& e) {
expired_sessions.emplace_back(session_it->second->get_session_id());
}
}
for (auto session_id : expired_sessions) {
if (leaf_aggregator_.leafCount() > 0) {
try {
leaf_aggregator_.disconnect(session_id);
} catch (TOmniSciException& toe) {
LOG(INFO) << " Problem disconnecting from leaves : " << toe.what();
} catch (std::exception& e) {
LOG(INFO)
<< " Problem disconnecting from leaves, check leaf logs for additonal info";
}
}
sessions_.erase(session_id);
}
if (render_handler_) {
write_lock.unlock();
for (auto session_id : expired_sessions) {
// NOTE: the render disconnect is done after the session lock is released to
// avoid a deadlock. See: https://omnisci.atlassian.net/browse/BE-3324
// This out-of-scope solution is a compromise for now until a better session
// handling/locking mechanism is developed for the renderer. Note as well that the
// session_id cannot be immediately reused. If a render request were to slip in
// after the lock is released and before the render disconnect could cause a
// problem.
render_handler_->disconnect(session_id);
}
}
}
#ifdef ENABLE_GEOS
extern std::unique_ptr<std::string> g_libgeos_so_filename;
#endif
DBHandler::DBHandler(const std::vector<LeafHostInfo>& db_leaves,
const std::vector<LeafHostInfo>& string_leaves,
const std::string& base_data_path,
const bool allow_multifrag,
const bool jit_debug,
const bool intel_jit_profile,
const bool read_only,
const bool allow_loop_joins,
const bool enable_rendering,
const bool renderer_use_vulkan_driver,
const bool enable_auto_clear_render_mem,
const int render_oom_retry_threshold,
const size_t render_mem_bytes,
const size_t max_concurrent_render_sessions,
const size_t reserved_gpu_mem,
const bool render_compositor_use_last_gpu,
const size_t num_reader_threads,
const AuthMetadata& authMetadata,
SystemParameters& system_parameters,
const bool legacy_syntax,
const int idle_session_duration,
const int max_session_duration,
const bool enable_runtime_udf_registration,
const std::string& udf_filename,
const std::string& clang_path,
const std::vector<std::string>& clang_options,
#ifdef ENABLE_GEOS
const std::string& libgeos_so_filename,
#endif
const File_Namespace::DiskCacheConfig& disk_cache_config,
const bool is_new_db)
: leaf_aggregator_(db_leaves)
, db_leaves_(db_leaves)
, string_leaves_(string_leaves)
, base_data_path_(base_data_path)
, random_gen_(std::random_device{}())
, session_id_dist_(0, INT32_MAX)
, jit_debug_(jit_debug)
, intel_jit_profile_(intel_jit_profile)
, allow_multifrag_(allow_multifrag)
, read_only_(read_only)
, allow_loop_joins_(allow_loop_joins)
, authMetadata_(authMetadata)
, system_parameters_(system_parameters)
, legacy_syntax_(legacy_syntax)
, dispatch_queue_(
std::make_unique<QueryDispatchQueue>(system_parameters.num_executors))
, super_user_rights_(false)
, idle_session_duration_(idle_session_duration * 60)
, max_session_duration_(max_session_duration * 60)
, runtime_udf_registration_enabled_(enable_runtime_udf_registration)
, enable_rendering_(enable_rendering)
, renderer_use_vulkan_driver_(renderer_use_vulkan_driver)
, enable_auto_clear_render_mem_(enable_auto_clear_render_mem)
, render_oom_retry_threshold_(render_oom_retry_threshold)
, max_concurrent_render_sessions_(max_concurrent_render_sessions)
, reserved_gpu_mem_(reserved_gpu_mem)
, render_compositor_use_last_gpu_(render_compositor_use_last_gpu)
, render_mem_bytes_(render_mem_bytes)
, num_reader_threads_(num_reader_threads)
#ifdef ENABLE_GEOS
, libgeos_so_filename_(libgeos_so_filename)
#endif
, disk_cache_config_(disk_cache_config)
, udf_filename_(udf_filename)
, clang_path_(clang_path)
, clang_options_(clang_options)
{
LOG(INFO) << "OmniSci Server " << MAPD_RELEASE;
initialize(is_new_db);
}
void DBHandler::initialize(const bool is_new_db) {
if (!initialized_) {
initialized_ = true;
} else {
THROW_MAPD_EXCEPTION(
"Server already initialized; service restart required to activate any new "
"entitlements.");
return;
}
if (system_parameters_.cpu_only || system_parameters_.num_gpus == 0) {
executor_device_type_ = ExecutorDeviceType::CPU;
cpu_mode_only_ = true;
} else {
#ifdef HAVE_CUDA
executor_device_type_ = ExecutorDeviceType::GPU;
cpu_mode_only_ = false;
#else
executor_device_type_ = ExecutorDeviceType::CPU;
LOG(WARNING) << "This build isn't CUDA enabled, will run on CPU";
cpu_mode_only_ = true;
#endif
}
bool is_rendering_enabled = enable_rendering_;
if (system_parameters_.num_gpus == 0) {
is_rendering_enabled = false;
}
const auto data_path = boost::filesystem::path(base_data_path_) / "mapd_data";
// calculate the total amount of memory we need to reserve from each gpu that the Buffer
// manage cannot ask for
size_t total_reserved = reserved_gpu_mem_;
if (is_rendering_enabled) {
total_reserved += render_mem_bytes_;
}
std::unique_ptr<CudaMgr_Namespace::CudaMgr> cuda_mgr;
#ifdef HAVE_CUDA
if (!cpu_mode_only_ || is_rendering_enabled) {
try {
cuda_mgr = std::make_unique<CudaMgr_Namespace::CudaMgr>(
system_parameters_.num_gpus, system_parameters_.start_gpu);
} catch (const std::exception& e) {
LOG(ERROR) << "Unable to instantiate CudaMgr, falling back to CPU-only mode. "
<< e.what();
cpu_mode_only_ = true;
is_rendering_enabled = false;
}
}
#endif // HAVE_CUDA
try {
data_mgr_.reset(new Data_Namespace::DataMgr(data_path.string(),
system_parameters_,
std::move(cuda_mgr),
!cpu_mode_only_,
total_reserved,
num_reader_threads_,
disk_cache_config_));
} catch (const std::exception& e) {
LOG(FATAL) << "Failed to initialize data manager: " << e.what();
}
std::string udf_ast_filename("");
try {
if (!udf_filename_.empty()) {
const auto cuda_mgr = data_mgr_->getCudaMgr();
const CudaMgr_Namespace::NvidiaDeviceArch device_arch =
cuda_mgr ? cuda_mgr->getDeviceArch()
: CudaMgr_Namespace::NvidiaDeviceArch::Kepler;
UdfCompiler compiler(device_arch, clang_path_, clang_options_);
const auto [cpu_udf_ir_file, cuda_udf_ir_file] = compiler.compileUdf(udf_filename_);
Executor::addUdfIrToModule(cpu_udf_ir_file, /*is_cuda_ir=*/false);
if (!cuda_udf_ir_file.empty()) {
Executor::addUdfIrToModule(cuda_udf_ir_file, /*is_cuda_ir=*/true);
}
udf_ast_filename = compiler.getAstFileName(udf_filename_);
}
} catch (const std::exception& e) {
LOG(FATAL) << "Failed to initialize UDF compiler: " << e.what();
}
try {
calcite_ =
std::make_shared<Calcite>(system_parameters_, base_data_path_, udf_ast_filename);
} catch (const std::exception& e) {
LOG(FATAL) << "Failed to initialize Calcite server: " << e.what();
}
try {
ExtensionFunctionsWhitelist::add(calcite_->getExtensionFunctionWhitelist());
if (!udf_filename_.empty()) {
ExtensionFunctionsWhitelist::addUdfs(calcite_->getUserDefinedFunctionWhitelist());
}
} catch (const std::exception& e) {
LOG(FATAL) << "Failed to initialize extension functions: " << e.what();
}
try {
table_functions::TableFunctionsFactory::init();
} catch (const std::exception& e) {
LOG(FATAL) << "Failed to initialize table functions factory: " << e.what();
}
try {
auto udtfs = ThriftSerializers::to_thrift(
table_functions::TableFunctionsFactory::get_table_funcs(/*is_runtime=*/false));
std::vector<TUserDefinedFunction> udfs = {};
calcite_->setRuntimeExtensionFunctions(udfs, udtfs, /*is_runtime=*/false);
} catch (const std::exception& e) {
LOG(FATAL) << "Failed to register compile-time table functions: " << e.what();
}
if (!data_mgr_->gpusPresent() && !cpu_mode_only_) {
executor_device_type_ = ExecutorDeviceType::CPU;
LOG(ERROR) << "No GPUs detected, falling back to CPU mode";
cpu_mode_only_ = true;
}
switch (executor_device_type_) {
case ExecutorDeviceType::GPU:
LOG(INFO) << "Started in GPU mode" << std::endl;
break;
case ExecutorDeviceType::CPU:
LOG(INFO) << "Started in CPU mode" << std::endl;
break;
}
try {
g_base_path = base_data_path_;
SysCatalog::instance().init(base_data_path_,
data_mgr_,
authMetadata_,
calcite_,
is_new_db,
!db_leaves_.empty(),
string_leaves_);
} catch (const std::exception& e) {
LOG(FATAL) << "Failed to initialize system catalog: " << e.what();
}
import_path_ = boost::filesystem::path(base_data_path_) / "mapd_import";
start_time_ = std::time(nullptr);
if (is_rendering_enabled) {
try {
render_handler_.reset(new RenderHandler(this,
render_mem_bytes_,
max_concurrent_render_sessions_,
render_compositor_use_last_gpu_,
false,
0,
false,
system_parameters_));
} catch (const std::exception& e) {
LOG(ERROR) << "Backend rendering disabled: " << e.what();
}
}
if (leaf_aggregator_.leafCount() > 0) {
try {
agg_handler_.reset(new MapDAggHandler(this));
} catch (const std::exception& e) {
LOG(ERROR) << "Distributed aggregator support disabled: " << e.what();
}
} else if (g_cluster) {
try {
leaf_handler_.reset(new MapDLeafHandler(this));
} catch (const std::exception& e) {
LOG(ERROR) << "Distributed leaf support disabled: " << e.what();
}
}
#ifdef ENABLE_GEOS
if (!libgeos_so_filename_.empty()) {
g_libgeos_so_filename.reset(new std::string(libgeos_so_filename_));
LOG(INFO) << "Overriding default geos library with '" + *g_libgeos_so_filename + "'";
}
#endif
}
DBHandler::~DBHandler() {
shutdown();
}
void DBHandler::parser_with_error_handler(
const std::string& query_str,
std::list<std::unique_ptr<Parser::Stmt>>& parse_trees) {
int num_parse_errors = 0;
std::string last_parsed;
SQLParser parser;
num_parse_errors = parser.parse(query_str, parse_trees, last_parsed);
if (num_parse_errors > 0) {
throw std::runtime_error("Syntax error at: " + last_parsed);
}
if (parse_trees.size() > 1) {
throw std::runtime_error("multiple SQL statements not allowed");
} else if (parse_trees.size() != 1) {
throw std::runtime_error("empty SQL statment not allowed");
}
}
void DBHandler::check_read_only(const std::string& str) {
if (DBHandler::read_only_) {
THROW_MAPD_EXCEPTION(str + " disabled: server running in read-only mode.");
}
}
std::string const DBHandler::createInMemoryCalciteSession(
const std::shared_ptr<Catalog_Namespace::Catalog>& catalog_ptr) {
// We would create an in memory session for calcite with super user privileges which
// would be used for getting all tables metadata when a user runs the query. The
// session would be under the name of a proxy user/password which would only persist
// till server's lifetime or execution of calcite query(in memory) whichever is the
// earliest.
mapd_lock_guard<mapd_shared_mutex> write_lock(sessions_mutex_);
std::string session_id;
do {
session_id = generate_random_string(64);
} while (sessions_.find(session_id) != sessions_.end());
Catalog_Namespace::UserMetadata user_meta(-1,
calcite_->getInternalSessionProxyUserName(),
calcite_->getInternalSessionProxyPassword(),
true,
-1,
true,
false);
const auto emplace_ret =
sessions_.emplace(session_id,
std::make_shared<Catalog_Namespace::SessionInfo>(
catalog_ptr, user_meta, executor_device_type_, session_id));
CHECK(emplace_ret.second);
return session_id;
}
bool DBHandler::isInMemoryCalciteSession(
const Catalog_Namespace::UserMetadata user_meta) {
return user_meta.userName == calcite_->getInternalSessionProxyUserName() &&
user_meta.userId == -1 && user_meta.defaultDbId == -1 &&
user_meta.isSuper.load();
}
void DBHandler::removeInMemoryCalciteSession(const std::string& session_id) {
// Remove InMemory calcite Session.
mapd_lock_guard<mapd_shared_mutex> write_lock(sessions_mutex_);
const auto it = sessions_.find(session_id);
CHECK(it != sessions_.end());
sessions_.erase(it);
}
// internal connection for connections with no password
void DBHandler::internal_connect(TSessionId& session,
const std::string& username,
const std::string& dbname) {
auto stdlog = STDLOG(); // session_info set by connect_impl()
std::string username2 = username; // login() may reset username given as argument
std::string dbname2 = dbname; // login() may reset dbname given as argument
Catalog_Namespace::UserMetadata user_meta;
std::shared_ptr<Catalog> cat = nullptr;
try {
cat =
SysCatalog::instance().login(dbname2, username2, std::string(), user_meta, false);
} catch (std::exception& e) {
THROW_MAPD_EXCEPTION(e.what());
}
DBObject dbObject(dbname2, DatabaseDBObjectType);
dbObject.loadKey(*cat);
dbObject.setPrivileges(AccessPrivileges::ACCESS);
std::vector<DBObject> dbObjects;
dbObjects.push_back(dbObject);
if (!SysCatalog::instance().checkPrivileges(user_meta, dbObjects)) {
THROW_MAPD_EXCEPTION("Unauthorized Access: user " + user_meta.userLoggable() +
" is not allowed to access database " + dbname2 + ".");
}
connect_impl(session, std::string(), dbname2, user_meta, cat, stdlog);
}
bool DBHandler::isAggregator() const {
return leaf_aggregator_.leafCount() > 0;
}
void DBHandler::krb5_connect(TKrb5Session& session,
const std::string& inputToken,
const std::string& dbname) {
THROW_MAPD_EXCEPTION("Unauthrorized Access. Kerberos login not supported");
}
void DBHandler::connect(TSessionId& session,
const std::string& username,
const std::string& passwd,
const std::string& dbname) {
auto stdlog = STDLOG(); // session_info set by connect_impl()
stdlog.appendNameValuePairs("client", getConnectionInfo().toString());
std::string username2 = username; // login() may reset username given as argument
std::string dbname2 = dbname; // login() may reset dbname given as argument
Catalog_Namespace::UserMetadata user_meta;
std::shared_ptr<Catalog> cat = nullptr;
try {
cat = SysCatalog::instance().login(
dbname2, username2, passwd, user_meta, !super_user_rights_);
} catch (std::exception& e) {
stdlog.appendNameValuePairs("user", username, "db", dbname, "exception", e.what());
THROW_MAPD_EXCEPTION(e.what());
}
DBObject dbObject(dbname2, DatabaseDBObjectType);
dbObject.loadKey(*cat);
dbObject.setPrivileges(AccessPrivileges::ACCESS);
std::vector<DBObject> dbObjects;
dbObjects.push_back(dbObject);
if (!SysCatalog::instance().checkPrivileges(user_meta, dbObjects)) {
stdlog.appendNameValuePairs(
"user", username, "db", dbname, "exception", "Missing Privileges");
THROW_MAPD_EXCEPTION("Unauthorized Access: user " + user_meta.userLoggable() +
" is not allowed to access database " + dbname2 + ".");
}
connect_impl(session, passwd, dbname2, user_meta, cat, stdlog);
// Restriction is returned as part of the users metadata on login but
// is per session so transfering it over here
// Currently only SAML can even set a Restriction
auto restriction = std::make_shared<Restriction>(user_meta.restriction);
auto login_session = get_session_ptr(session);
login_session->set_restriction(restriction);
// if pki auth session will come back encrypted with user pubkey
SysCatalog::instance().check_for_session_encryption(passwd, session);
}
std::shared_ptr<Catalog_Namespace::SessionInfo> DBHandler::create_new_session(
TSessionId& session,
const std::string& dbname,
const Catalog_Namespace::UserMetadata& user_meta,
std::shared_ptr<Catalog> cat) {
do {
session = generate_random_string(32);
} while (sessions_.find(session) != sessions_.end());
std::pair<SessionMap::iterator, bool> emplace_retval =
sessions_.emplace(session,
std::make_shared<Catalog_Namespace::SessionInfo>(
cat, user_meta, executor_device_type_, session));
CHECK(emplace_retval.second);
auto& session_ptr = emplace_retval.first->second;
LOG(INFO) << "User " << user_meta.userLoggable() << " connected to database " << dbname;
return session_ptr;
}
void DBHandler::connect_impl(TSessionId& session,
const std::string& passwd,
const std::string& dbname,
const Catalog_Namespace::UserMetadata& user_meta,
std::shared_ptr<Catalog> cat,
query_state::StdLog& stdlog) {
// TODO(sy): Is there any reason to have dbname as a parameter
// here when the cat parameter already provides cat->name()?
// Should dbname and cat->name() ever differ?
{
mapd_unique_lock<mapd_shared_mutex> write_lock(sessions_mutex_);
expire_idle_sessions_unsafe(write_lock);
if (system_parameters_.num_sessions > 0 &&
sessions_.size() + 1 > static_cast<size_t>(system_parameters_.num_sessions)) {
THROW_MAPD_EXCEPTION("Too many active sessions");
}
}
{
mapd_lock_guard<mapd_shared_mutex> write_lock(sessions_mutex_);
auto session_ptr = create_new_session(session, dbname, user_meta, cat);
stdlog.setSessionInfo(session_ptr);
session_ptr->set_connection_info(getConnectionInfo().toString());
if (!super_user_rights_) { // no need to connect to leaf_aggregator_ at this time
// while doing warmup
if (leaf_aggregator_.leafCount() > 0) {
leaf_aggregator_.connect(*session_ptr, user_meta.userName, passwd, dbname);
return;
}
}
}
auto const roles =
stdlog.getConstSessionInfo()->get_currentUser().isSuper
? std::vector<std::string>{{"super"}}
: SysCatalog::instance().getRoles(
false, false, stdlog.getConstSessionInfo()->get_currentUser().userName);
stdlog.appendNameValuePairs("roles", boost::algorithm::join(roles, ","));
}
void DBHandler::disconnect(const TSessionId& session) {
auto stdlog = STDLOG();
stdlog.appendNameValuePairs("client", getConnectionInfo().toString());
mapd_unique_lock<mapd_shared_mutex> write_lock(sessions_mutex_);
auto session_it = get_session_it_unsafe(session, write_lock);
stdlog.setSessionInfo(session_it->second);
const auto dbname = session_it->second->getCatalog().getCurrentDB().dbName;
LOG(INFO) << "User " << session_it->second->get_currentUser().userLoggable()
<< " disconnected from database " << dbname
<< " with public_session_id: " << session_it->second->get_public_session_id();
disconnect_impl(session_it, write_lock);
}
void DBHandler::disconnect_impl(const SessionMap::iterator& session_it,
mapd_unique_lock<mapd_shared_mutex>& write_lock) {
// session_it existence should already have been checked (i.e. called via
// get_session_it_unsafe(...))
const auto session_id = session_it->second->get_session_id();
std::exception_ptr leaf_exception = nullptr;
try {
if (leaf_aggregator_.leafCount() > 0) {
leaf_aggregator_.disconnect(session_id);
}
} catch (...) {
leaf_exception = std::current_exception();
}
{
std::lock_guard<std::mutex> lock(render_group_assignment_mutex_);
render_group_assignment_map_.erase(session_id);
}
sessions_.erase(session_it);
write_lock.unlock();
if (render_handler_) {
render_handler_->disconnect(session_id);
}
}
void DBHandler::switch_database(const TSessionId& session, const std::string& dbname) {
auto stdlog = STDLOG(get_session_ptr(session));
stdlog.appendNameValuePairs("client", getConnectionInfo().toString());
mapd_unique_lock<mapd_shared_mutex> write_lock(sessions_mutex_);
auto session_it = get_session_it_unsafe(session, write_lock);
std::string dbname2 = dbname; // switchDatabase() may reset dbname given as argument
try {
std::shared_ptr<Catalog> cat = SysCatalog::instance().switchDatabase(
dbname2, session_it->second->get_currentUser().userName);
session_it->second->set_catalog_ptr(cat);
if (leaf_aggregator_.leafCount() > 0) {
leaf_aggregator_.switch_database(session, dbname);
return;
}
} catch (std::exception& e) {
THROW_MAPD_EXCEPTION(e.what());
}
}
void DBHandler::clone_session(TSessionId& session2, const TSessionId& session1) {
auto stdlog = STDLOG(get_session_ptr(session1));
stdlog.appendNameValuePairs("client", getConnectionInfo().toString());
mapd_unique_lock<mapd_shared_mutex> write_lock(sessions_mutex_);
auto session_it = get_session_it_unsafe(session1, write_lock);
try {
const Catalog_Namespace::UserMetadata& user_meta =
session_it->second->get_currentUser();
std::shared_ptr<Catalog> cat = session_it->second->get_catalog_ptr();
auto session2_ptr = create_new_session(session2, cat->name(), user_meta, cat);
if (leaf_aggregator_.leafCount() > 0) {
leaf_aggregator_.clone_session(session1, session2);
return;
}
} catch (std::exception& e) {
THROW_MAPD_EXCEPTION(e.what());
}
}
void DBHandler::interrupt(const TSessionId& query_session,
const TSessionId& interrupt_session) {
// if this is for distributed setting, query_session becomes a parent session (agg)
// and the interrupt session is one of existing session in the leaf node (leaf)
// so we can think there exists a logical mapping
// between query_session (agg) and interrupt_session (leaf)
auto stdlog = STDLOG(get_session_ptr(interrupt_session));
stdlog.appendNameValuePairs("client", getConnectionInfo().toString());
const auto allow_query_interrupt =
g_enable_runtime_query_interrupt || g_enable_non_kernel_time_query_interrupt;
if (g_enable_dynamic_watchdog || allow_query_interrupt) {
// Shared lock to allow simultaneous interrupts of multiple sessions
mapd_shared_lock<mapd_shared_mutex> read_lock(sessions_mutex_);
auto session_it = get_session_it_unsafe(interrupt_session, read_lock);
auto& cat = session_it->second.get()->getCatalog();
const auto dbname = cat.getCurrentDB().dbName;
auto executor = Executor::getExecutor(Executor::UNITARY_EXECUTOR_ID,
jit_debug_ ? "/tmp" : "",
jit_debug_ ? "mapdquery" : "",
system_parameters_);
CHECK(executor);
VLOG(1) << "Received interrupt: "
<< "Session " << *session_it->second << ", Executor " << executor
<< ", leafCount " << leaf_aggregator_.leafCount() << ", User "
<< session_it->second->get_currentUser().userLoggable() << ", Database "
<< dbname << std::endl;
if (leaf_aggregator_.leafCount() > 0) {
leaf_aggregator_.interrupt(query_session, interrupt_session);
}
executor->interrupt(query_session, interrupt_session);
LOG(INFO) << "User " << session_it->second->get_currentUser().userName
<< " interrupted session with database " << dbname << std::endl;
}
}
TRole::type DBHandler::getServerRole() const {
if (g_cluster) {
if (leaf_aggregator_.leafCount() > 0) {
return TRole::type::AGGREGATOR;
}
return TRole::type::LEAF;
}
return TRole::type::SERVER;
}
void DBHandler::get_server_status(TServerStatus& _return, const TSessionId& session) {
auto stdlog = STDLOG(get_session_ptr(session));
stdlog.appendNameValuePairs("client", getConnectionInfo().toString());
const auto rendering_enabled = bool(render_handler_);
_return.read_only = read_only_;
_return.version = MAPD_RELEASE;
_return.rendering_enabled = rendering_enabled;
_return.start_time = start_time_;
_return.edition = MAPD_EDITION;
_return.host_name = omnisci::get_hostname();
_return.poly_rendering_enabled = rendering_enabled;
_return.role = getServerRole();
_return.renderer_status_json =
render_handler_ ? render_handler_->get_renderer_status_json() : "";
}
void DBHandler::get_status(std::vector<TServerStatus>& _return,
const TSessionId& session) {
//
// get_status() is now called locally at startup on the aggregator
// in order to validate that all nodes of a cluster are running the
// same software version and the same renderer status
//
// In that context, it is called with the InvalidSessionID, and
// with the local super-user flag set.
//
// Hence, we allow this session-less mode only in distributed mode, and
// then on a leaf (always), or on the aggregator (only in super-user mode)
//
auto const allow_invalid_session = g_cluster && (!isAggregator() || super_user_rights_);
if (!allow_invalid_session || session != getInvalidSessionId()) {
auto stdlog = STDLOG(get_session_ptr(session));
stdlog.appendNameValuePairs("client", getConnectionInfo().toString());
} else {
LOG(INFO) << "get_status() called in session-less mode";
}
const auto rendering_enabled = bool(render_handler_);
TServerStatus ret;
ret.read_only = read_only_;
ret.version = MAPD_RELEASE;
ret.rendering_enabled = rendering_enabled;
ret.start_time = start_time_;
ret.edition = MAPD_EDITION;
ret.host_name = omnisci::get_hostname();
ret.poly_rendering_enabled = rendering_enabled;
ret.role = getServerRole();
ret.renderer_status_json =
render_handler_ ? render_handler_->get_renderer_status_json() : "";
_return.push_back(ret);
if (leaf_aggregator_.leafCount() > 0) {
std::vector<TServerStatus> leaf_status = leaf_aggregator_.getLeafStatus(session);
_return.insert(_return.end(), leaf_status.begin(), leaf_status.end());
}
}
void DBHandler::get_hardware_info(TClusterHardwareInfo& _return,
const TSessionId& session) {
auto stdlog = STDLOG(get_session_ptr(session));
stdlog.appendNameValuePairs("client", getConnectionInfo().toString());
THardwareInfo ret;
const auto cuda_mgr = data_mgr_->getCudaMgr();
if (cuda_mgr) {
ret.num_gpu_hw = cuda_mgr->getDeviceCount();
ret.start_gpu = cuda_mgr->getStartGpu();
if (ret.start_gpu >= 0) {
ret.num_gpu_allocated = cuda_mgr->getDeviceCount() - cuda_mgr->getStartGpu();
// ^ This will break as soon as we allow non contiguous GPU allocations to MapD
}
for (int16_t device_id = 0; device_id < ret.num_gpu_hw; device_id++) {
TGpuSpecification gpu_spec;
auto deviceProperties = cuda_mgr->getDeviceProperties(device_id);
gpu_spec.num_sm = deviceProperties->numMPs;
gpu_spec.clock_frequency_kHz = deviceProperties->clockKhz;
gpu_spec.memory = deviceProperties->globalMem;
gpu_spec.compute_capability_major = deviceProperties->computeMajor;
gpu_spec.compute_capability_minor = deviceProperties->computeMinor;
ret.gpu_info.push_back(gpu_spec);
}
}
// start hardware/OS dependent code
ret.num_cpu_hw = std::thread::hardware_concurrency();
// ^ This might return diffrent results in case of hyper threading
// end hardware/OS dependent code
_return.hardware_info.push_back(ret);
if (leaf_aggregator_.leafCount() > 0) {
ret.host_name = "aggregator";
TClusterHardwareInfo leaf_hardware = leaf_aggregator_.getHardwareInfo(session);
_return.hardware_info.insert(_return.hardware_info.end(),
leaf_hardware.hardware_info.begin(),
leaf_hardware.hardware_info.end());
}
}
void DBHandler::get_session_info(TSessionInfo& _return, const TSessionId& session) {
auto session_ptr = get_session_ptr(session);
auto stdlog = STDLOG(session_ptr);
stdlog.appendNameValuePairs("client", getConnectionInfo().toString());
auto user_metadata = session_ptr->get_currentUser();
_return.user = user_metadata.userName;
_return.database = session_ptr->getCatalog().getCurrentDB().dbName;
_return.start_time = session_ptr->get_start_time();
_return.is_super = user_metadata.isSuper;
}
void DBHandler::value_to_thrift_column(const TargetValue& tv,
const SQLTypeInfo& ti,
TColumn& column) {
if (ti.is_array()) {
TColumn tColumn;
const auto array_tv = boost::get<ArrayTargetValue>(&tv);
CHECK(array_tv);
bool is_null = !array_tv->is_initialized();
if (!is_null) {
const auto& vec = array_tv->get();
for (const auto& elem_tv : vec) {
value_to_thrift_column(elem_tv, ti.get_elem_type(), tColumn);
}
}
column.data.arr_col.push_back(tColumn);
column.nulls.push_back(is_null && !ti.get_notnull());
} else if (ti.is_geometry()) {
const auto scalar_tv = boost::get<ScalarTargetValue>(&tv);
if (scalar_tv) {
auto s_n = boost::get<NullableString>(scalar_tv);
auto s = boost::get<std::string>(s_n);
if (s) {
column.data.str_col.push_back(*s);
} else {
column.data.str_col.emplace_back(""); // null string
auto null_p = boost::get<void*>(s_n);
CHECK(null_p && !*null_p);
}
column.nulls.push_back(!s && !ti.get_notnull());
} else {
const auto array_tv = boost::get<ArrayTargetValue>(&tv);
CHECK(array_tv);
bool is_null = !array_tv->is_initialized();
if (!is_null) {
auto elem_type = SQLTypeInfo(kDOUBLE, false);
TColumn tColumn;
const auto& vec = array_tv->get();
for (const auto& elem_tv : vec) {
value_to_thrift_column(elem_tv, elem_type, tColumn);
}
column.data.arr_col.push_back(tColumn);
column.nulls.push_back(false);
} else {
TColumn tColumn;
column.data.arr_col.push_back(tColumn);
column.nulls.push_back(is_null && !ti.get_notnull());
}
}
} else {
CHECK(!ti.is_column());
const auto scalar_tv = boost::get<ScalarTargetValue>(&tv);
CHECK(scalar_tv);
if (boost::get<int64_t>(scalar_tv)) {
int64_t data = *(boost::get<int64_t>(scalar_tv));
if (ti.is_decimal()) {
double val = static_cast<double>(data);
if (ti.get_scale() > 0) {
val /= pow(10.0, std::abs(ti.get_scale()));
}
column.data.real_col.push_back(val);
} else {
column.data.int_col.push_back(data);
}
switch (ti.get_type()) {
case kBOOLEAN:
column.nulls.push_back(data == NULL_BOOLEAN && !ti.get_notnull());
break;
case kTINYINT:
column.nulls.push_back(data == NULL_TINYINT && !ti.get_notnull());
break;
case kSMALLINT:
column.nulls.push_back(data == NULL_SMALLINT && !ti.get_notnull());
break;