forked from facebook/rocksdb
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathdb_wal_test.cc
2094 lines (1838 loc) · 73.6 KB
/
db_wal_test.cc
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 (c) 2011-present, Facebook, Inc. All rights reserved.
// This source code is licensed under both the GPLv2 (found in the
// COPYING file in the root directory) and Apache 2.0 License
// (found in the LICENSE.Apache file in the root directory).
//
// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file. See the AUTHORS file for names of contributors.
#include "db/db_test_util.h"
#include "options/options_helper.h"
#include "port/port.h"
#include "port/stack_trace.h"
#include "rocksdb/file_system.h"
#include "test_util/sync_point.h"
#include "utilities/fault_injection_env.h"
#include "utilities/fault_injection_fs.h"
namespace ROCKSDB_NAMESPACE {
class DBWALTestBase : public DBTestBase {
protected:
explicit DBWALTestBase(const std::string& dir_name)
: DBTestBase(dir_name, /*env_do_fsync=*/true) {}
#if defined(ROCKSDB_PLATFORM_POSIX)
public:
#if defined(ROCKSDB_FALLOCATE_PRESENT)
bool IsFallocateSupported() {
// Test fallocate support of running file system.
// Skip this test if fallocate is not supported.
std::string fname_test_fallocate = dbname_ + "/preallocate_testfile";
int fd = -1;
do {
fd = open(fname_test_fallocate.c_str(), O_CREAT | O_RDWR | O_TRUNC, 0644);
} while (fd < 0 && errno == EINTR);
assert(fd > 0);
int alloc_status = fallocate(fd, 0, 0, 1);
int err_number = errno;
close(fd);
assert(env_->DeleteFile(fname_test_fallocate) == Status::OK());
if (err_number == ENOSYS || err_number == EOPNOTSUPP) {
fprintf(stderr, "Skipped preallocated space check: %s\n",
errnoStr(err_number).c_str());
return false;
}
assert(alloc_status == 0);
return true;
}
#endif // ROCKSDB_FALLOCATE_PRESENT
uint64_t GetAllocatedFileSize(std::string file_name) {
struct stat sbuf;
int err = stat(file_name.c_str(), &sbuf);
assert(err == 0);
return sbuf.st_blocks * 512;
}
#endif // ROCKSDB_PLATFORM_POSIX
};
class DBWALTest : public DBWALTestBase {
public:
DBWALTest() : DBWALTestBase("/db_wal_test") {}
};
// A SpecialEnv enriched to give more insight about deleted files
class EnrichedSpecialEnv : public SpecialEnv {
public:
explicit EnrichedSpecialEnv(Env* base) : SpecialEnv(base) {}
Status NewSequentialFile(const std::string& f,
std::unique_ptr<SequentialFile>* r,
const EnvOptions& soptions) override {
InstrumentedMutexLock l(&env_mutex_);
if (f == skipped_wal) {
deleted_wal_reopened = true;
if (IsWAL(f) && largest_deleted_wal.size() != 0 &&
f.compare(largest_deleted_wal) <= 0) {
gap_in_wals = true;
}
}
return SpecialEnv::NewSequentialFile(f, r, soptions);
}
Status DeleteFile(const std::string& fname) override {
if (IsWAL(fname)) {
deleted_wal_cnt++;
InstrumentedMutexLock l(&env_mutex_);
// If this is the first WAL, remember its name and skip deleting it. We
// remember its name partly because the application might attempt to
// delete the file again.
if (skipped_wal.size() != 0 && skipped_wal != fname) {
if (largest_deleted_wal.size() == 0 ||
largest_deleted_wal.compare(fname) < 0) {
largest_deleted_wal = fname;
}
} else {
skipped_wal = fname;
return Status::OK();
}
}
return SpecialEnv::DeleteFile(fname);
}
bool IsWAL(const std::string& fname) {
// printf("iswal %s\n", fname.c_str());
return fname.compare(fname.size() - 3, 3, "log") == 0;
}
InstrumentedMutex env_mutex_;
// the wal whose actual delete was skipped by the env
std::string skipped_wal = "";
// the largest WAL that was requested to be deleted
std::string largest_deleted_wal = "";
// number of WALs that were successfully deleted
std::atomic<size_t> deleted_wal_cnt = {0};
// the WAL whose delete from fs was skipped is reopened during recovery
std::atomic<bool> deleted_wal_reopened = {false};
// whether a gap in the WALs was detected during recovery
std::atomic<bool> gap_in_wals = {false};
};
class DBWALTestWithEnrichedEnv : public DBTestBase {
public:
DBWALTestWithEnrichedEnv()
: DBTestBase("/db_wal_test", /*env_do_fsync=*/true) {
enriched_env_ = new EnrichedSpecialEnv(env_->target());
auto options = CurrentOptions();
options.env = enriched_env_;
options.allow_2pc = true;
Reopen(options);
delete env_;
// to be deleted by the parent class
env_ = enriched_env_;
}
protected:
EnrichedSpecialEnv* enriched_env_;
};
// Test that the recovery would successfully avoid the gaps between the logs.
// One known scenario that could cause this is that the application issue the
// WAL deletion out of order. For the sake of simplicity in the test, here we
// create the gap by manipulating the env to skip deletion of the first WAL but
// not the ones after it.
TEST_F(DBWALTestWithEnrichedEnv, SkipDeletedWALs) {
auto options = last_options_;
// To cause frequent WAL deletion
options.write_buffer_size = 128;
Reopen(options);
WriteOptions writeOpt = WriteOptions();
for (int i = 0; i < 128 * 5; i++) {
ASSERT_OK(dbfull()->Put(writeOpt, "foo", "v1"));
}
FlushOptions fo;
fo.wait = true;
ASSERT_OK(db_->Flush(fo));
// some wals are deleted
ASSERT_NE(0, enriched_env_->deleted_wal_cnt);
// but not the first one
ASSERT_NE(0, enriched_env_->skipped_wal.size());
// Test that the WAL that was not deleted will be skipped during recovery
options = last_options_;
Reopen(options);
ASSERT_FALSE(enriched_env_->deleted_wal_reopened);
ASSERT_FALSE(enriched_env_->gap_in_wals);
}
TEST_F(DBWALTest, WAL) {
do {
CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
WriteOptions writeOpt = WriteOptions();
writeOpt.disableWAL = true;
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "foo", "v1"));
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "bar", "v1"));
ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
ASSERT_EQ("v1", Get(1, "foo"));
ASSERT_EQ("v1", Get(1, "bar"));
writeOpt.disableWAL = false;
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "bar", "v2"));
writeOpt.disableWAL = true;
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "foo", "v2"));
ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
// Both value's should be present.
ASSERT_EQ("v2", Get(1, "bar"));
ASSERT_EQ("v2", Get(1, "foo"));
writeOpt.disableWAL = true;
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "bar", "v3"));
writeOpt.disableWAL = false;
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "foo", "v3"));
ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
// again both values should be present.
ASSERT_EQ("v3", Get(1, "foo"));
ASSERT_EQ("v3", Get(1, "bar"));
} while (ChangeWalOptions());
}
TEST_F(DBWALTest, RollLog) {
do {
CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
ASSERT_OK(Put(1, "foo", "v1"));
ASSERT_OK(Put(1, "baz", "v5"));
ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
for (int i = 0; i < 10; i++) {
ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
}
ASSERT_OK(Put(1, "foo", "v4"));
for (int i = 0; i < 10; i++) {
ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
}
} while (ChangeWalOptions());
}
TEST_F(DBWALTest, SyncWALNotBlockWrite) {
Options options = CurrentOptions();
options.max_write_buffer_number = 4;
DestroyAndReopen(options);
ASSERT_OK(Put("foo1", "bar1"));
ASSERT_OK(Put("foo5", "bar5"));
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency({
{"WritableFileWriter::SyncWithoutFlush:1",
"DBWALTest::SyncWALNotBlockWrite:1"},
{"DBWALTest::SyncWALNotBlockWrite:2",
"WritableFileWriter::SyncWithoutFlush:2"},
});
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
ROCKSDB_NAMESPACE::port::Thread thread([&]() { ASSERT_OK(db_->SyncWAL()); });
TEST_SYNC_POINT("DBWALTest::SyncWALNotBlockWrite:1");
ASSERT_OK(Put("foo2", "bar2"));
ASSERT_OK(Put("foo3", "bar3"));
FlushOptions fo;
fo.wait = false;
ASSERT_OK(db_->Flush(fo));
ASSERT_OK(Put("foo4", "bar4"));
TEST_SYNC_POINT("DBWALTest::SyncWALNotBlockWrite:2");
thread.join();
ASSERT_EQ(Get("foo1"), "bar1");
ASSERT_EQ(Get("foo2"), "bar2");
ASSERT_EQ(Get("foo3"), "bar3");
ASSERT_EQ(Get("foo4"), "bar4");
ASSERT_EQ(Get("foo5"), "bar5");
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
}
TEST_F(DBWALTest, SyncWALNotWaitWrite) {
ASSERT_OK(Put("foo1", "bar1"));
ASSERT_OK(Put("foo3", "bar3"));
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency({
{"SpecialEnv::WalFile::Append:1", "DBWALTest::SyncWALNotWaitWrite:1"},
{"DBWALTest::SyncWALNotWaitWrite:2", "SpecialEnv::WalFile::Append:2"},
});
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
ROCKSDB_NAMESPACE::port::Thread thread(
[&]() { ASSERT_OK(Put("foo2", "bar2")); });
// Moving this to SyncWAL before the actual fsync
// TEST_SYNC_POINT("DBWALTest::SyncWALNotWaitWrite:1");
ASSERT_OK(db_->SyncWAL());
// Moving this to SyncWAL after actual fsync
// TEST_SYNC_POINT("DBWALTest::SyncWALNotWaitWrite:2");
thread.join();
ASSERT_EQ(Get("foo1"), "bar1");
ASSERT_EQ(Get("foo2"), "bar2");
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
}
TEST_F(DBWALTest, Recover) {
do {
CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
ASSERT_OK(Put(1, "foo", "v1"));
ASSERT_OK(Put(1, "baz", "v5"));
ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
ASSERT_EQ("v1", Get(1, "foo"));
ASSERT_EQ("v1", Get(1, "foo"));
ASSERT_EQ("v5", Get(1, "baz"));
ASSERT_OK(Put(1, "bar", "v2"));
ASSERT_OK(Put(1, "foo", "v3"));
ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
ASSERT_EQ("v3", Get(1, "foo"));
ASSERT_OK(Put(1, "foo", "v4"));
ASSERT_EQ("v4", Get(1, "foo"));
ASSERT_EQ("v2", Get(1, "bar"));
ASSERT_EQ("v5", Get(1, "baz"));
} while (ChangeWalOptions());
}
TEST_F(DBWALTest, RecoverWithTableHandle) {
do {
Options options = CurrentOptions();
options.create_if_missing = true;
options.disable_auto_compactions = true;
options.avoid_flush_during_recovery = false;
DestroyAndReopen(options);
CreateAndReopenWithCF({"pikachu"}, options);
ASSERT_OK(Put(1, "foo", "v1"));
ASSERT_OK(Put(1, "bar", "v2"));
ASSERT_OK(Flush(1));
ASSERT_OK(Put(1, "foo", "v3"));
ASSERT_OK(Put(1, "bar", "v4"));
ASSERT_OK(Flush(1));
ASSERT_OK(Put(1, "big", std::string(100, 'a')));
options = CurrentOptions();
const int kSmallMaxOpenFiles = 13;
if (option_config_ == kDBLogDir) {
// Use this option to check not preloading files
// Set the max open files to be small enough so no preload will
// happen.
options.max_open_files = kSmallMaxOpenFiles;
// RocksDB sanitize max open files to at least 20. Modify it back.
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
"SanitizeOptions::AfterChangeMaxOpenFiles", [&](void* arg) {
int* max_open_files = static_cast<int*>(arg);
*max_open_files = kSmallMaxOpenFiles;
});
} else if (option_config_ == kWalDirAndMmapReads) {
// Use this option to check always loading all files.
options.max_open_files = 100;
} else {
options.max_open_files = -1;
}
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
ReopenWithColumnFamilies({"default", "pikachu"}, options);
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
std::vector<std::vector<FileMetaData>> files;
dbfull()->TEST_GetFilesMetaData(handles_[1], &files);
size_t total_files = 0;
for (const auto& level : files) {
total_files += level.size();
}
ASSERT_EQ(total_files, 3);
for (const auto& level : files) {
for (const auto& file : level) {
if (options.max_open_files == kSmallMaxOpenFiles) {
ASSERT_TRUE(file.table_reader_handle == nullptr);
} else {
ASSERT_TRUE(file.table_reader_handle != nullptr);
}
}
}
} while (ChangeWalOptions());
}
TEST_F(DBWALTest, RecoverWithBlob) {
// Write a value that's below the prospective size limit for blobs and another
// one that's above. Note that blob files are not actually enabled at this
// point.
constexpr uint64_t min_blob_size = 10;
constexpr char short_value[] = "short";
static_assert(sizeof(short_value) - 1 < min_blob_size,
"short_value too long");
constexpr char long_value[] = "long_value";
static_assert(sizeof(long_value) - 1 >= min_blob_size,
"long_value too short");
ASSERT_OK(Put("key1", short_value));
ASSERT_OK(Put("key2", long_value));
// There should be no files just yet since we haven't flushed.
{
VersionSet* const versions = dbfull()->TEST_GetVersionSet();
ASSERT_NE(versions, nullptr);
ColumnFamilyData* const cfd = versions->GetColumnFamilySet()->GetDefault();
ASSERT_NE(cfd, nullptr);
Version* const current = cfd->current();
ASSERT_NE(current, nullptr);
const VersionStorageInfo* const storage_info = current->storage_info();
ASSERT_NE(storage_info, nullptr);
ASSERT_EQ(storage_info->num_non_empty_levels(), 0);
ASSERT_TRUE(storage_info->GetBlobFiles().empty());
}
// Reopen the database with blob files enabled. A new table file/blob file
// pair should be written during recovery.
Options options;
options.enable_blob_files = true;
options.min_blob_size = min_blob_size;
options.avoid_flush_during_recovery = false;
options.disable_auto_compactions = true;
options.env = env_;
Reopen(options);
ASSERT_EQ(Get("key1"), short_value);
ASSERT_EQ(Get("key2"), long_value);
VersionSet* const versions = dbfull()->TEST_GetVersionSet();
ASSERT_NE(versions, nullptr);
ColumnFamilyData* const cfd = versions->GetColumnFamilySet()->GetDefault();
ASSERT_NE(cfd, nullptr);
Version* const current = cfd->current();
ASSERT_NE(current, nullptr);
const VersionStorageInfo* const storage_info = current->storage_info();
ASSERT_NE(storage_info, nullptr);
const auto& l0_files = storage_info->LevelFiles(0);
ASSERT_EQ(l0_files.size(), 1);
const FileMetaData* const table_file = l0_files[0];
ASSERT_NE(table_file, nullptr);
const auto& blob_files = storage_info->GetBlobFiles();
ASSERT_EQ(blob_files.size(), 1);
const auto& blob_file = blob_files.begin()->second;
ASSERT_NE(blob_file, nullptr);
ASSERT_EQ(table_file->smallest.user_key(), "key1");
ASSERT_EQ(table_file->largest.user_key(), "key2");
ASSERT_EQ(table_file->fd.smallest_seqno, 1);
ASSERT_EQ(table_file->fd.largest_seqno, 2);
ASSERT_EQ(table_file->oldest_blob_file_number,
blob_file->GetBlobFileNumber());
ASSERT_EQ(blob_file->GetTotalBlobCount(), 1);
#ifndef ROCKSDB_LITE
const InternalStats* const internal_stats = cfd->internal_stats();
ASSERT_NE(internal_stats, nullptr);
const auto& compaction_stats = internal_stats->TEST_GetCompactionStats();
ASSERT_FALSE(compaction_stats.empty());
ASSERT_EQ(compaction_stats[0].bytes_written, table_file->fd.GetFileSize());
ASSERT_EQ(compaction_stats[0].bytes_written_blob,
blob_file->GetTotalBlobBytes());
ASSERT_EQ(compaction_stats[0].num_output_files, 1);
ASSERT_EQ(compaction_stats[0].num_output_files_blob, 1);
const uint64_t* const cf_stats_value = internal_stats->TEST_GetCFStatsValue();
ASSERT_EQ(cf_stats_value[InternalStats::BYTES_FLUSHED],
compaction_stats[0].bytes_written +
compaction_stats[0].bytes_written_blob);
#endif // ROCKSDB_LITE
}
TEST_F(DBWALTest, RecoverWithBlobMultiSST) {
// Write several large (4 KB) values without flushing. Note that blob files
// are not actually enabled at this point.
std::string large_value(1 << 12, 'a');
constexpr int num_keys = 64;
for (int i = 0; i < num_keys; ++i) {
ASSERT_OK(Put(Key(i), large_value));
}
// There should be no files just yet since we haven't flushed.
{
VersionSet* const versions = dbfull()->TEST_GetVersionSet();
ASSERT_NE(versions, nullptr);
ColumnFamilyData* const cfd = versions->GetColumnFamilySet()->GetDefault();
ASSERT_NE(cfd, nullptr);
Version* const current = cfd->current();
ASSERT_NE(current, nullptr);
const VersionStorageInfo* const storage_info = current->storage_info();
ASSERT_NE(storage_info, nullptr);
ASSERT_EQ(storage_info->num_non_empty_levels(), 0);
ASSERT_TRUE(storage_info->GetBlobFiles().empty());
}
// Reopen the database with blob files enabled and write buffer size set to a
// smaller value. Multiple table files+blob files should be written and added
// to the Version during recovery.
Options options;
options.write_buffer_size = 1 << 16; // 64 KB
options.enable_blob_files = true;
options.avoid_flush_during_recovery = false;
options.disable_auto_compactions = true;
options.env = env_;
Reopen(options);
for (int i = 0; i < num_keys; ++i) {
ASSERT_EQ(Get(Key(i)), large_value);
}
VersionSet* const versions = dbfull()->TEST_GetVersionSet();
ASSERT_NE(versions, nullptr);
ColumnFamilyData* const cfd = versions->GetColumnFamilySet()->GetDefault();
ASSERT_NE(cfd, nullptr);
Version* const current = cfd->current();
ASSERT_NE(current, nullptr);
const VersionStorageInfo* const storage_info = current->storage_info();
ASSERT_NE(storage_info, nullptr);
const auto& l0_files = storage_info->LevelFiles(0);
ASSERT_GT(l0_files.size(), 1);
const auto& blob_files = storage_info->GetBlobFiles();
ASSERT_GT(blob_files.size(), 1);
ASSERT_EQ(l0_files.size(), blob_files.size());
}
TEST_F(DBWALTest, WALWithChecksumHandoff) {
#ifndef ROCKSDB_ASSERT_STATUS_CHECKED
if (mem_env_ || encrypted_env_) {
ROCKSDB_GTEST_SKIP("Test requires non-mem or non-encrypted environment");
return;
}
std::shared_ptr<FaultInjectionTestFS> fault_fs(
new FaultInjectionTestFS(FileSystem::Default()));
std::unique_ptr<Env> fault_fs_env(NewCompositeEnv(fault_fs));
do {
Options options = CurrentOptions();
options.checksum_handoff_file_types.Add(FileType::kWalFile);
options.env = fault_fs_env.get();
fault_fs->SetChecksumHandoffFuncType(ChecksumType::kCRC32c);
CreateAndReopenWithCF({"pikachu"}, options);
WriteOptions writeOpt = WriteOptions();
writeOpt.disableWAL = true;
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "foo", "v1"));
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "bar", "v1"));
ReopenWithColumnFamilies({"default", "pikachu"}, options);
ASSERT_EQ("v1", Get(1, "foo"));
ASSERT_EQ("v1", Get(1, "bar"));
writeOpt.disableWAL = false;
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "bar", "v2"));
writeOpt.disableWAL = true;
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "foo", "v2"));
ReopenWithColumnFamilies({"default", "pikachu"}, options);
// Both value's should be present.
ASSERT_EQ("v2", Get(1, "bar"));
ASSERT_EQ("v2", Get(1, "foo"));
writeOpt.disableWAL = true;
// This put, data is persisted by Flush
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "bar", "v3"));
ReopenWithColumnFamilies({"default", "pikachu"}, options);
writeOpt.disableWAL = false;
// Data is persisted in the WAL
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "zoo", "v3"));
// The hash does not match, write fails
fault_fs->SetChecksumHandoffFuncType(ChecksumType::kxxHash);
writeOpt.disableWAL = false;
ASSERT_NOK(dbfull()->Put(writeOpt, handles_[1], "foo", "v3"));
ReopenWithColumnFamilies({"default", "pikachu"}, options);
// Due to the write failure, Get should not find
ASSERT_NE("v3", Get(1, "foo"));
ASSERT_EQ("v3", Get(1, "zoo"));
ASSERT_EQ("v3", Get(1, "bar"));
fault_fs->SetChecksumHandoffFuncType(ChecksumType::kCRC32c);
// Each write will be similated as corrupted.
fault_fs->IngestDataCorruptionBeforeWrite();
writeOpt.disableWAL = true;
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "bar", "v4"));
writeOpt.disableWAL = false;
ASSERT_NOK(dbfull()->Put(writeOpt, handles_[1], "foo", "v4"));
ReopenWithColumnFamilies({"default", "pikachu"}, options);
ASSERT_NE("v4", Get(1, "foo"));
ASSERT_NE("v4", Get(1, "bar"));
fault_fs->NoDataCorruptionBeforeWrite();
fault_fs->SetChecksumHandoffFuncType(ChecksumType::kNoChecksum);
// The file system does not provide checksum method and verification.
writeOpt.disableWAL = true;
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "bar", "v5"));
writeOpt.disableWAL = false;
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "foo", "v5"));
ReopenWithColumnFamilies({"default", "pikachu"}, options);
ASSERT_EQ("v5", Get(1, "foo"));
ASSERT_EQ("v5", Get(1, "bar"));
Destroy(options);
} while (ChangeWalOptions());
#endif // ROCKSDB_ASSERT_STATUS_CHECKED
}
class DBRecoveryTestBlobError
: public DBWALTest,
public testing::WithParamInterface<std::string> {
public:
DBRecoveryTestBlobError() : sync_point_(GetParam()) {}
std::string sync_point_;
};
INSTANTIATE_TEST_CASE_P(DBRecoveryTestBlobError, DBRecoveryTestBlobError,
::testing::ValuesIn(std::vector<std::string>{
"BlobFileBuilder::WriteBlobToFile:AddRecord",
"BlobFileBuilder::WriteBlobToFile:AppendFooter"}));
TEST_P(DBRecoveryTestBlobError, RecoverWithBlobError) {
// Write a value. Note that blob files are not actually enabled at this point.
ASSERT_OK(Put("key", "blob"));
// Reopen with blob files enabled but make blob file writing fail during
// recovery.
SyncPoint::GetInstance()->SetCallBack(sync_point_, [this](void* arg) {
Status* const s = static_cast<Status*>(arg);
assert(s);
(*s) = Status::IOError(sync_point_);
});
SyncPoint::GetInstance()->EnableProcessing();
Options options;
options.enable_blob_files = true;
options.avoid_flush_during_recovery = false;
options.disable_auto_compactions = true;
options.env = env_;
ASSERT_NOK(TryReopen(options));
SyncPoint::GetInstance()->DisableProcessing();
SyncPoint::GetInstance()->ClearAllCallBacks();
// Make sure the files generated by the failed recovery have been deleted.
std::vector<std::string> files;
ASSERT_OK(env_->GetChildren(dbname_, &files));
for (const auto& file : files) {
uint64_t number = 0;
FileType type = kTableFile;
if (!ParseFileName(file, &number, &type)) {
continue;
}
ASSERT_NE(type, kTableFile);
ASSERT_NE(type, kBlobFile);
}
}
TEST_F(DBWALTest, IgnoreRecoveredLog) {
std::string backup_logs = dbname_ + "/backup_logs";
do {
// delete old files in backup_logs directory
ASSERT_OK(env_->CreateDirIfMissing(backup_logs));
std::vector<std::string> old_files;
ASSERT_OK(env_->GetChildren(backup_logs, &old_files));
for (auto& file : old_files) {
ASSERT_OK(env_->DeleteFile(backup_logs + "/" + file));
}
Options options = CurrentOptions();
options.create_if_missing = true;
options.merge_operator = MergeOperators::CreateUInt64AddOperator();
options.wal_dir = dbname_ + "/logs";
DestroyAndReopen(options);
// fill up the DB
std::string one, two;
PutFixed64(&one, 1);
PutFixed64(&two, 2);
ASSERT_OK(db_->Merge(WriteOptions(), Slice("foo"), Slice(one)));
ASSERT_OK(db_->Merge(WriteOptions(), Slice("foo"), Slice(one)));
ASSERT_OK(db_->Merge(WriteOptions(), Slice("bar"), Slice(one)));
// copy the logs to backup
std::vector<std::string> logs;
ASSERT_OK(env_->GetChildren(options.wal_dir, &logs));
for (auto& log : logs) {
CopyFile(options.wal_dir + "/" + log, backup_logs + "/" + log);
}
// recover the DB
Reopen(options);
ASSERT_EQ(two, Get("foo"));
ASSERT_EQ(one, Get("bar"));
Close();
// copy the logs from backup back to wal dir
for (auto& log : logs) {
CopyFile(backup_logs + "/" + log, options.wal_dir + "/" + log);
}
// this should ignore the log files, recovery should not happen again
// if the recovery happens, the same merge operator would be called twice,
// leading to incorrect results
Reopen(options);
ASSERT_EQ(two, Get("foo"));
ASSERT_EQ(one, Get("bar"));
Close();
Destroy(options);
Reopen(options);
Close();
// copy the logs from backup back to wal dir
ASSERT_OK(env_->CreateDirIfMissing(options.wal_dir));
for (auto& log : logs) {
CopyFile(backup_logs + "/" + log, options.wal_dir + "/" + log);
}
// assert that we successfully recovered only from logs, even though we
// destroyed the DB
Reopen(options);
ASSERT_EQ(two, Get("foo"));
ASSERT_EQ(one, Get("bar"));
// Recovery will fail if DB directory doesn't exist.
Destroy(options);
// copy the logs from backup back to wal dir
ASSERT_OK(env_->CreateDirIfMissing(options.wal_dir));
for (auto& log : logs) {
CopyFile(backup_logs + "/" + log, options.wal_dir + "/" + log);
// we won't be needing this file no more
ASSERT_OK(env_->DeleteFile(backup_logs + "/" + log));
}
Status s = TryReopen(options);
ASSERT_NOK(s);
Destroy(options);
} while (ChangeWalOptions());
}
TEST_F(DBWALTest, RecoveryWithEmptyLog) {
do {
CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
ASSERT_OK(Put(1, "foo", "v1"));
ASSERT_OK(Put(1, "foo", "v2"));
ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
ASSERT_OK(Put(1, "foo", "v3"));
ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
ASSERT_EQ("v3", Get(1, "foo"));
} while (ChangeWalOptions());
}
#if !(defined NDEBUG) || !defined(OS_WIN)
TEST_F(DBWALTest, PreallocateBlock) {
Options options = CurrentOptions();
options.write_buffer_size = 10 * 1000 * 1000;
options.max_total_wal_size = 0;
size_t expected_preallocation_size = static_cast<size_t>(
options.write_buffer_size + options.write_buffer_size / 10);
DestroyAndReopen(options);
std::atomic<int> called(0);
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
"DBTestWalFile.GetPreallocationStatus", [&](void* arg) {
ASSERT_TRUE(arg != nullptr);
size_t preallocation_size = *(static_cast<size_t*>(arg));
ASSERT_EQ(expected_preallocation_size, preallocation_size);
called.fetch_add(1);
});
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
ASSERT_OK(Put("", ""));
ASSERT_OK(Flush());
ASSERT_OK(Put("", ""));
Close();
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
ASSERT_EQ(2, called.load());
options.max_total_wal_size = 1000 * 1000;
expected_preallocation_size = static_cast<size_t>(options.max_total_wal_size);
Reopen(options);
called.store(0);
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
"DBTestWalFile.GetPreallocationStatus", [&](void* arg) {
ASSERT_TRUE(arg != nullptr);
size_t preallocation_size = *(static_cast<size_t*>(arg));
ASSERT_EQ(expected_preallocation_size, preallocation_size);
called.fetch_add(1);
});
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
ASSERT_OK(Put("", ""));
ASSERT_OK(Flush());
ASSERT_OK(Put("", ""));
Close();
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
ASSERT_EQ(2, called.load());
options.db_write_buffer_size = 800 * 1000;
expected_preallocation_size =
static_cast<size_t>(options.db_write_buffer_size);
Reopen(options);
called.store(0);
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
"DBTestWalFile.GetPreallocationStatus", [&](void* arg) {
ASSERT_TRUE(arg != nullptr);
size_t preallocation_size = *(static_cast<size_t*>(arg));
ASSERT_EQ(expected_preallocation_size, preallocation_size);
called.fetch_add(1);
});
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
ASSERT_OK(Put("", ""));
ASSERT_OK(Flush());
ASSERT_OK(Put("", ""));
Close();
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
ASSERT_EQ(2, called.load());
expected_preallocation_size = 700 * 1000;
std::shared_ptr<WriteBufferManager> write_buffer_manager =
std::make_shared<WriteBufferManager>(static_cast<uint64_t>(700 * 1000));
options.write_buffer_manager = write_buffer_manager;
Reopen(options);
called.store(0);
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
"DBTestWalFile.GetPreallocationStatus", [&](void* arg) {
ASSERT_TRUE(arg != nullptr);
size_t preallocation_size = *(static_cast<size_t*>(arg));
ASSERT_EQ(expected_preallocation_size, preallocation_size);
called.fetch_add(1);
});
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
ASSERT_OK(Put("", ""));
ASSERT_OK(Flush());
ASSERT_OK(Put("", ""));
Close();
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
ASSERT_EQ(2, called.load());
}
#endif // !(defined NDEBUG) || !defined(OS_WIN)
#ifndef ROCKSDB_LITE
TEST_F(DBWALTest, DISABLED_FullPurgePreservesRecycledLog) {
// TODO(ajkr): Disabled until WAL recycling is fixed for
// `kPointInTimeRecovery`.
// For github issue #1303
for (int i = 0; i < 2; ++i) {
Options options = CurrentOptions();
options.create_if_missing = true;
options.recycle_log_file_num = 2;
if (i != 0) {
options.wal_dir = alternative_wal_dir_;
}
DestroyAndReopen(options);
ASSERT_OK(Put("foo", "v1"));
VectorLogPtr log_files;
ASSERT_OK(dbfull()->GetSortedWalFiles(log_files));
ASSERT_GT(log_files.size(), 0);
ASSERT_OK(Flush());
// Now the original WAL is in log_files[0] and should be marked for
// recycling.
// Verify full purge cannot remove this file.
JobContext job_context(0);
dbfull()->TEST_LockMutex();
dbfull()->FindObsoleteFiles(&job_context, true /* force */);
dbfull()->TEST_UnlockMutex();
dbfull()->PurgeObsoleteFiles(job_context);
if (i == 0) {
ASSERT_OK(
env_->FileExists(LogFileName(dbname_, log_files[0]->LogNumber())));
} else {
ASSERT_OK(env_->FileExists(
LogFileName(alternative_wal_dir_, log_files[0]->LogNumber())));
}
}
}
TEST_F(DBWALTest, DISABLED_FullPurgePreservesLogPendingReuse) {
// TODO(ajkr): Disabled until WAL recycling is fixed for
// `kPointInTimeRecovery`.
// Ensures full purge cannot delete a WAL while it's in the process of being
// recycled. In particular, we force the full purge after a file has been
// chosen for reuse, but before it has been renamed.
for (int i = 0; i < 2; ++i) {
Options options = CurrentOptions();
options.recycle_log_file_num = 1;
if (i != 0) {
options.wal_dir = alternative_wal_dir_;
}
DestroyAndReopen(options);
// The first flush creates a second log so writes can continue before the
// flush finishes.
ASSERT_OK(Put("foo", "bar"));
ASSERT_OK(Flush());
// The second flush can recycle the first log. Sync points enforce the
// full purge happens after choosing the log to recycle and before it is
// renamed.
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency({
{"DBImpl::CreateWAL:BeforeReuseWritableFile1",
"DBWALTest::FullPurgePreservesLogPendingReuse:PreFullPurge"},
{"DBWALTest::FullPurgePreservesLogPendingReuse:PostFullPurge",
"DBImpl::CreateWAL:BeforeReuseWritableFile2"},
});
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
ROCKSDB_NAMESPACE::port::Thread thread([&]() {
TEST_SYNC_POINT(
"DBWALTest::FullPurgePreservesLogPendingReuse:PreFullPurge");
ASSERT_OK(db_->EnableFileDeletions(true));
TEST_SYNC_POINT(
"DBWALTest::FullPurgePreservesLogPendingReuse:PostFullPurge");
});
ASSERT_OK(Put("foo", "bar"));
ASSERT_OK(Flush());
thread.join();
}
}
TEST_F(DBWALTest, GetSortedWalFiles) {
do {
CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
VectorLogPtr log_files;
ASSERT_OK(dbfull()->GetSortedWalFiles(log_files));
ASSERT_EQ(0, log_files.size());
ASSERT_OK(Put(1, "foo", "v1"));
ASSERT_OK(dbfull()->GetSortedWalFiles(log_files));
ASSERT_EQ(1, log_files.size());
} while (ChangeWalOptions());
}
TEST_F(DBWALTest, GetCurrentWalFile) {
do {
CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
std::unique_ptr<LogFile>* bad_log_file = nullptr;
ASSERT_NOK(dbfull()->GetCurrentWalFile(bad_log_file));
std::unique_ptr<LogFile> log_file;
ASSERT_OK(dbfull()->GetCurrentWalFile(&log_file));
// nothing has been written to the log yet
ASSERT_EQ(log_file->StartSequence(), 0);
ASSERT_EQ(log_file->SizeFileBytes(), 0);
ASSERT_EQ(log_file->Type(), kAliveLogFile);
ASSERT_GT(log_file->LogNumber(), 0);
// add some data and verify that the file size actually moves foward
ASSERT_OK(Put(0, "foo", "v1"));
ASSERT_OK(Put(0, "foo2", "v2"));
ASSERT_OK(Put(0, "foo3", "v3"));
ASSERT_OK(dbfull()->GetCurrentWalFile(&log_file));
ASSERT_EQ(log_file->StartSequence(), 0);
ASSERT_GT(log_file->SizeFileBytes(), 0);
ASSERT_EQ(log_file->Type(), kAliveLogFile);
ASSERT_GT(log_file->LogNumber(), 0);
// force log files to cycle and add some more data, then check if
// log number moves forward
ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
for (int i = 0; i < 10; i++) {
ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
}
ASSERT_OK(Put(0, "foo4", "v4"));
ASSERT_OK(Put(0, "foo5", "v5"));
ASSERT_OK(Put(0, "foo6", "v6"));
ASSERT_OK(dbfull()->GetCurrentWalFile(&log_file));
ASSERT_EQ(log_file->StartSequence(), 0);
ASSERT_GT(log_file->SizeFileBytes(), 0);
ASSERT_EQ(log_file->Type(), kAliveLogFile);
ASSERT_GT(log_file->LogNumber(), 0);
} while (ChangeWalOptions());
}
TEST_F(DBWALTest, RecoveryWithLogDataForSomeCFs) {
// Test for regression of WAL cleanup missing files that don't contain data
// for every column family.
do {
CreateAndReopenWithCF({"pikachu"}, CurrentOptions());