Split db_test.cc

Summary: Split db_test.cc into several files. Moving several helper functions into DBTestBase.

Test Plan: make check

Reviewers: sdong, yhchiang, IslamAbdelRahman

Reviewed By: IslamAbdelRahman

Subscribers: dhruba, andrewkr, kradhakrishnan, yhchiang, leveldb, sdong

Differential Revision: https://reviews.facebook.net/D56715
This commit is contained in:
Yi Wu 2016-04-18 09:42:50 -07:00
parent 40b840f294
commit 792762c42c
14 changed files with 4791 additions and 4775 deletions

View File

@ -342,6 +342,9 @@ set(TESTS
db/db_test.cc
db/db_test2.cc
db/db_block_cache_test.cc
db/db_bloom_filter_test.cc
db/db_iterator_test.cc
db/db_sst_test.cc
db/db_universal_compaction_test.cc
db/db_wal_test.cc
db/dbformat_test.cc

View File

@ -258,12 +258,15 @@ TESTS = \
db_test \
db_test2 \
db_block_cache_test \
db_bloom_filter_test \
db_iter_test \
db_log_iter_test \
db_compaction_filter_test \
db_compaction_test \
db_dynamic_level_test \
db_inplace_update_test \
db_iterator_test \
db_sst_test \
db_tailing_iter_test \
db_universal_compaction_test \
db_wal_test \
@ -872,6 +875,9 @@ db_test2: db/db_test2.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
db_block_cache_test: db/db_block_cache_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
$(AM_LINK)
db_bloom_filter_test: db/db_bloom_filter_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
$(AM_LINK)
db_log_iter_test: db/db_log_iter_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
$(AM_LINK)
@ -887,6 +893,12 @@ db_dynamic_level_test: db/db_dynamic_level_test.o db/db_test_util.o $(LIBOBJECTS
db_inplace_update_test: db/db_inplace_update_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
$(AM_LINK)
db_iterator_test: db/db_iterator_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
$(AM_LINK)
db_sst_test: db/db_sst_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
$(AM_LINK)
db_tailing_iter_test: db/db_tailing_iter_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
$(AM_LINK)

View File

@ -12,11 +12,6 @@
namespace rocksdb {
static uint64_t TestGetTickerCount(const Options& options,
Tickers ticker_type) {
return options.statistics->getTickerCount(ticker_type);
}
class DBBlockCacheTest : public DBTestBase {
private:
size_t miss_count_ = 0;
@ -229,6 +224,222 @@ TEST_F(DBBlockCacheTest, TestWithCompressedBlockCache) {
delete iter;
iter = nullptr;
}
// Make sure that when options.block_cache is set, after a new table is
// created its index/filter blocks are added to block cache.
TEST_F(DBBlockCacheTest, IndexAndFilterBlocksOfNewTableAddedToCache) {
Options options = CurrentOptions();
options.create_if_missing = true;
options.statistics = rocksdb::CreateDBStatistics();
BlockBasedTableOptions table_options;
table_options.cache_index_and_filter_blocks = true;
table_options.filter_policy.reset(NewBloomFilterPolicy(20));
options.table_factory.reset(new BlockBasedTableFactory(table_options));
CreateAndReopenWithCF({"pikachu"}, options);
ASSERT_OK(Put(1, "key", "val"));
// Create a new table.
ASSERT_OK(Flush(1));
// index/filter blocks added to block cache right after table creation.
ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_INDEX_MISS));
ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
ASSERT_EQ(2, /* only index/filter were added */
TestGetTickerCount(options, BLOCK_CACHE_ADD));
ASSERT_EQ(0, TestGetTickerCount(options, BLOCK_CACHE_DATA_MISS));
uint64_t int_num;
ASSERT_TRUE(
dbfull()->GetIntProperty("rocksdb.estimate-table-readers-mem", &int_num));
ASSERT_EQ(int_num, 0U);
// Make sure filter block is in cache.
std::string value;
ReadOptions ropt;
db_->KeyMayExist(ReadOptions(), handles_[1], "key", &value);
// Miss count should remain the same.
ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT));
db_->KeyMayExist(ReadOptions(), handles_[1], "key", &value);
ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
ASSERT_EQ(2, TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT));
// Make sure index block is in cache.
auto index_block_hit = TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT);
value = Get(1, "key");
ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
ASSERT_EQ(index_block_hit + 1,
TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT));
value = Get(1, "key");
ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
ASSERT_EQ(index_block_hit + 2,
TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT));
}
TEST_F(DBBlockCacheTest, ParanoidFileChecks) {
Options options = CurrentOptions();
options.create_if_missing = true;
options.statistics = rocksdb::CreateDBStatistics();
options.level0_file_num_compaction_trigger = 2;
options.paranoid_file_checks = true;
BlockBasedTableOptions table_options;
table_options.cache_index_and_filter_blocks = false;
table_options.filter_policy.reset(NewBloomFilterPolicy(20));
options.table_factory.reset(new BlockBasedTableFactory(table_options));
CreateAndReopenWithCF({"pikachu"}, options);
ASSERT_OK(Put(1, "1_key", "val"));
ASSERT_OK(Put(1, "9_key", "val"));
// Create a new table.
ASSERT_OK(Flush(1));
ASSERT_EQ(1, /* read and cache data block */
TestGetTickerCount(options, BLOCK_CACHE_ADD));
ASSERT_OK(Put(1, "1_key2", "val2"));
ASSERT_OK(Put(1, "9_key2", "val2"));
// Create a new SST file. This will further trigger a compaction
// and generate another file.
ASSERT_OK(Flush(1));
dbfull()->TEST_WaitForCompact();
ASSERT_EQ(3, /* Totally 3 files created up to now */
TestGetTickerCount(options, BLOCK_CACHE_ADD));
// After disabling options.paranoid_file_checks. NO further block
// is added after generating a new file.
ASSERT_OK(
dbfull()->SetOptions(handles_[1], {{"paranoid_file_checks", "false"}}));
ASSERT_OK(Put(1, "1_key3", "val3"));
ASSERT_OK(Put(1, "9_key3", "val3"));
ASSERT_OK(Flush(1));
ASSERT_OK(Put(1, "1_key4", "val4"));
ASSERT_OK(Put(1, "9_key4", "val4"));
ASSERT_OK(Flush(1));
dbfull()->TEST_WaitForCompact();
ASSERT_EQ(3, /* Totally 3 files created up to now */
TestGetTickerCount(options, BLOCK_CACHE_ADD));
}
TEST_F(DBBlockCacheTest, CompressedCache) {
if (!Snappy_Supported()) {
return;
}
int num_iter = 80;
// Run this test three iterations.
// Iteration 1: only a uncompressed block cache
// Iteration 2: only a compressed block cache
// Iteration 3: both block cache and compressed cache
// Iteration 4: both block cache and compressed cache, but DB is not
// compressed
for (int iter = 0; iter < 4; iter++) {
Options options = CurrentOptions();
options.write_buffer_size = 64 * 1024; // small write buffer
options.statistics = rocksdb::CreateDBStatistics();
BlockBasedTableOptions table_options;
switch (iter) {
case 0:
// only uncompressed block cache
table_options.block_cache = NewLRUCache(8 * 1024);
table_options.block_cache_compressed = nullptr;
options.table_factory.reset(NewBlockBasedTableFactory(table_options));
break;
case 1:
// no block cache, only compressed cache
table_options.no_block_cache = true;
table_options.block_cache = nullptr;
table_options.block_cache_compressed = NewLRUCache(8 * 1024);
options.table_factory.reset(NewBlockBasedTableFactory(table_options));
break;
case 2:
// both compressed and uncompressed block cache
table_options.block_cache = NewLRUCache(1024);
table_options.block_cache_compressed = NewLRUCache(8 * 1024);
options.table_factory.reset(NewBlockBasedTableFactory(table_options));
break;
case 3:
// both block cache and compressed cache, but DB is not compressed
// also, make block cache sizes bigger, to trigger block cache hits
table_options.block_cache = NewLRUCache(1024 * 1024);
table_options.block_cache_compressed = NewLRUCache(8 * 1024 * 1024);
options.table_factory.reset(NewBlockBasedTableFactory(table_options));
options.compression = kNoCompression;
break;
default:
ASSERT_TRUE(false);
}
CreateAndReopenWithCF({"pikachu"}, options);
// default column family doesn't have block cache
Options no_block_cache_opts;
no_block_cache_opts.statistics = options.statistics;
no_block_cache_opts = CurrentOptions(no_block_cache_opts);
BlockBasedTableOptions table_options_no_bc;
table_options_no_bc.no_block_cache = true;
no_block_cache_opts.table_factory.reset(
NewBlockBasedTableFactory(table_options_no_bc));
ReopenWithColumnFamilies(
{"default", "pikachu"},
std::vector<Options>({no_block_cache_opts, options}));
Random rnd(301);
// Write 8MB (80 values, each 100K)
ASSERT_EQ(NumTableFilesAtLevel(0, 1), 0);
std::vector<std::string> values;
std::string str;
for (int i = 0; i < num_iter; i++) {
if (i % 4 == 0) { // high compression ratio
str = RandomString(&rnd, 1000);
}
values.push_back(str);
ASSERT_OK(Put(1, Key(i), values[i]));
}
// flush all data from memtable so that reads are from block cache
ASSERT_OK(Flush(1));
for (int i = 0; i < num_iter; i++) {
ASSERT_EQ(Get(1, Key(i)), values[i]);
}
// check that we triggered the appropriate code paths in the cache
switch (iter) {
case 0:
// only uncompressed block cache
ASSERT_GT(TestGetTickerCount(options, BLOCK_CACHE_MISS), 0);
ASSERT_EQ(TestGetTickerCount(options, BLOCK_CACHE_COMPRESSED_MISS), 0);
break;
case 1:
// no block cache, only compressed cache
ASSERT_EQ(TestGetTickerCount(options, BLOCK_CACHE_MISS), 0);
ASSERT_GT(TestGetTickerCount(options, BLOCK_CACHE_COMPRESSED_MISS), 0);
break;
case 2:
// both compressed and uncompressed block cache
ASSERT_GT(TestGetTickerCount(options, BLOCK_CACHE_MISS), 0);
ASSERT_GT(TestGetTickerCount(options, BLOCK_CACHE_COMPRESSED_MISS), 0);
break;
case 3:
// both compressed and uncompressed block cache
ASSERT_GT(TestGetTickerCount(options, BLOCK_CACHE_MISS), 0);
ASSERT_GT(TestGetTickerCount(options, BLOCK_CACHE_HIT), 0);
ASSERT_GT(TestGetTickerCount(options, BLOCK_CACHE_COMPRESSED_MISS), 0);
// compressed doesn't have any hits since blocks are not compressed on
// storage
ASSERT_EQ(TestGetTickerCount(options, BLOCK_CACHE_COMPRESSED_HIT), 0);
break;
default:
ASSERT_TRUE(false);
}
options.create_if_missing = true;
DestroyAndReopen(options);
}
}
#endif
} // namespace rocksdb

1047
db/db_bloom_filter_test.cc Normal file

File diff suppressed because it is too large Load Diff

View File

@ -40,34 +40,6 @@ class DBCompactionTestWithParam
};
namespace {
class OnFileDeletionListener : public EventListener {
public:
OnFileDeletionListener() :
matched_count_(0),
expected_file_name_("") {}
void SetExpectedFileName(
const std::string file_name) {
expected_file_name_ = file_name;
}
void VerifyMatchedCount(size_t expected_value) {
ASSERT_EQ(matched_count_, expected_value);
}
void OnTableFileDeleted(
const TableFileDeletionInfo& info) override {
if (expected_file_name_ != "") {
ASSERT_EQ(expected_file_name_, info.file_path);
expected_file_name_ = "";
matched_count_++;
}
}
private:
size_t matched_count_;
std::string expected_file_name_;
};
class FlushedFileCollector : public EventListener {
public:

1319
db/db_iterator_test.cc Normal file

File diff suppressed because it is too large Load Diff

1300
db/db_sst_test.cc Normal file

File diff suppressed because it is too large Load Diff

File diff suppressed because it is too large Load Diff

View File

@ -13,11 +13,6 @@
namespace rocksdb {
static uint64_t TestGetTickerCount(const Options& options,
Tickers ticker_type) {
return options.statistics->getTickerCount(ticker_type);
}
class DBTest2 : public DBTestBase {
public:
DBTest2() : DBTestBase("/db_test2") {}

View File

@ -1043,4 +1043,34 @@ std::unordered_map<std::string, uint64_t> DBTestBase::GetAllSSTFiles(
return res;
}
std::vector<std::uint64_t> DBTestBase::ListTableFiles(Env* env,
const std::string& path) {
std::vector<std::string> files;
std::vector<uint64_t> file_numbers;
env->GetChildren(path, &files);
uint64_t number;
FileType type;
for (size_t i = 0; i < files.size(); ++i) {
if (ParseFileName(files[i], &number, &type)) {
if (type == kTableFile) {
file_numbers.push_back(number);
}
}
}
return file_numbers;
}
#ifndef ROCKSDB_LITE
uint64_t DBTestBase::GetNumberOfSstFilesForColumnFamily(
DB* db, std::string column_family_name) {
std::vector<LiveFileMetaData> metadata;
db->GetLiveFilesMetaData(&metadata);
uint64_t result = 0;
for (auto& fileMetadata : metadata) {
result += (fileMetadata.column_family_name == column_family_name);
}
return result;
}
#endif // ROCKSDB_LITE
} // namespace rocksdb

View File

@ -39,6 +39,7 @@
#include "rocksdb/filter_policy.h"
#include "rocksdb/options.h"
#include "rocksdb/slice.h"
#include "rocksdb/statistics.h"
#include "rocksdb/table.h"
#include "rocksdb/utilities/checkpoint.h"
#include "table/block_based_table_factory.h"
@ -491,6 +492,31 @@ class SpecialEnv : public EnvWrapper {
std::atomic<bool> is_wal_sync_thread_safe_{true};
};
class OnFileDeletionListener : public EventListener {
public:
OnFileDeletionListener() : matched_count_(0), expected_file_name_("") {}
void SetExpectedFileName(const std::string file_name) {
expected_file_name_ = file_name;
}
void VerifyMatchedCount(size_t expected_value) {
ASSERT_EQ(matched_count_, expected_value);
}
void OnTableFileDeleted(const TableFileDeletionInfo& info) override {
if (expected_file_name_ != "") {
ASSERT_EQ(expected_file_name_, info.file_path);
expected_file_name_ = "";
matched_count_++;
}
}
private:
size_t matched_count_;
std::string expected_file_name_;
};
class DBTestBase : public testing::Test {
protected:
// Sequence of option configurations to try
@ -756,6 +782,17 @@ class DBTestBase : public testing::Test {
std::unordered_map<std::string, uint64_t> GetAllSSTFiles(
uint64_t* total_size = nullptr);
std::vector<std::uint64_t> ListTableFiles(Env* env, const std::string& path);
#ifndef ROCKSDB_LITE
uint64_t GetNumberOfSstFilesForColumnFamily(DB* db,
std::string column_family_name);
#endif // ROCKSDB_LITE
uint64_t TestGetTickerCount(const Options& options, Tickers ticker_type) {
return options.statistics->getTickerCount(ticker_type);
}
};
} // namespace rocksdb

View File

@ -14,11 +14,6 @@
namespace rocksdb {
static uint64_t TestGetTickerCount(const Options& options,
Tickers ticker_type) {
return options.statistics->getTickerCount(ticker_type);
}
static std::string CompressibleString(Random* rnd, int len) {
std::string r;
test::CompressibleString(rnd, 0.8, len, &r);

View File

@ -127,6 +127,634 @@ TEST_F(DBWALTest, SyncWALNotWaitWrite) {
ASSERT_EQ(Get("foo2"), "bar2");
rocksdb::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 (ChangeOptions());
}
TEST_F(DBWALTest, RecoverWithTableHandle) {
do {
Options options = CurrentOptions();
options.create_if_missing = true;
options.disable_auto_compactions = true;
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')));
ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
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 (kInfiniteMaxOpenFiles == option_config_) {
ASSERT_TRUE(file.table_reader_handle != nullptr);
} else {
ASSERT_TRUE(file.table_reader_handle == nullptr);
}
}
}
} while (ChangeOptions());
}
TEST_F(DBWALTest, IgnoreRecoveredLog) {
std::string backup_logs = dbname_ + "/backup_logs";
// delete old files in backup_logs directory
env_->CreateDirIfMissing(backup_logs);
std::vector<std::string> old_files;
env_->GetChildren(backup_logs, &old_files);
for (auto& file : old_files) {
if (file != "." && file != "..") {
env_->DeleteFile(backup_logs + "/" + file);
}
}
do {
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;
env_->GetChildren(options.wal_dir, &logs);
for (auto& log : logs) {
if (log != ".." && log != ".") {
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) {
if (log != ".." && log != ".") {
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
env_->CreateDirIfMissing(options.wal_dir);
for (auto& log : logs) {
if (log != ".." && log != ".") {
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
env_->CreateDirIfMissing(options.wal_dir);
for (auto& log : logs) {
if (log != ".." && log != ".") {
CopyFile(backup_logs + "/" + log, options.wal_dir + "/" + log);
// we won't be needing this file no more
env_->DeleteFile(backup_logs + "/" + log);
}
}
Status s = TryReopen(options);
ASSERT_TRUE(!s.ok());
} while (ChangeOptions(kSkipHashCuckoo));
}
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 (ChangeOptions());
}
#ifndef ROCKSDB_LITE
TEST_F(DBWALTest, RecoverWithLargeLog) {
do {
{
Options options = CurrentOptions();
CreateAndReopenWithCF({"pikachu"}, options);
ASSERT_OK(Put(1, "big1", std::string(200000, '1')));
ASSERT_OK(Put(1, "big2", std::string(200000, '2')));
ASSERT_OK(Put(1, "small3", std::string(10, '3')));
ASSERT_OK(Put(1, "small4", std::string(10, '4')));
ASSERT_EQ(NumTableFilesAtLevel(0, 1), 0);
}
// Make sure that if we re-open with a small write buffer size that
// we flush table files in the middle of a large log file.
Options options;
options.write_buffer_size = 100000;
options = CurrentOptions(options);
ReopenWithColumnFamilies({"default", "pikachu"}, options);
ASSERT_EQ(NumTableFilesAtLevel(0, 1), 3);
ASSERT_EQ(std::string(200000, '1'), Get(1, "big1"));
ASSERT_EQ(std::string(200000, '2'), Get(1, "big2"));
ASSERT_EQ(std::string(10, '3'), Get(1, "small3"));
ASSERT_EQ(std::string(10, '4'), Get(1, "small4"));
ASSERT_GT(NumTableFilesAtLevel(0, 1), 1);
} while (ChangeCompactOptions());
}
// In https://reviews.facebook.net/D20661 we change
// recovery behavior: previously for each log file each column family
// memtable was flushed, even it was empty. Now it's changed:
// we try to create the smallest number of table files by merging
// updates from multiple logs
TEST_F(DBWALTest, RecoverCheckFileAmountWithSmallWriteBuffer) {
Options options = CurrentOptions();
options.write_buffer_size = 5000000;
CreateAndReopenWithCF({"pikachu", "dobrynia", "nikitich"}, options);
// Since we will reopen DB with smaller write_buffer_size,
// each key will go to new SST file
ASSERT_OK(Put(1, Key(10), DummyString(1000000)));
ASSERT_OK(Put(1, Key(10), DummyString(1000000)));
ASSERT_OK(Put(1, Key(10), DummyString(1000000)));
ASSERT_OK(Put(1, Key(10), DummyString(1000000)));
ASSERT_OK(Put(3, Key(10), DummyString(1)));
// Make 'dobrynia' to be flushed and new WAL file to be created
ASSERT_OK(Put(2, Key(10), DummyString(7500000)));
ASSERT_OK(Put(2, Key(1), DummyString(1)));
dbfull()->TEST_WaitForFlushMemTable(handles_[2]);
{
auto tables = ListTableFiles(env_, dbname_);
ASSERT_EQ(tables.size(), static_cast<size_t>(1));
// Make sure 'dobrynia' was flushed: check sst files amount
ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "dobrynia"),
static_cast<uint64_t>(1));
}
// New WAL file
ASSERT_OK(Put(1, Key(1), DummyString(1)));
ASSERT_OK(Put(1, Key(1), DummyString(1)));
ASSERT_OK(Put(3, Key(10), DummyString(1)));
ASSERT_OK(Put(3, Key(10), DummyString(1)));
ASSERT_OK(Put(3, Key(10), DummyString(1)));
options.write_buffer_size = 4096;
options.arena_block_size = 4096;
ReopenWithColumnFamilies({"default", "pikachu", "dobrynia", "nikitich"},
options);
{
// No inserts => default is empty
ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "default"),
static_cast<uint64_t>(0));
// First 4 keys goes to separate SSTs + 1 more SST for 2 smaller keys
ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "pikachu"),
static_cast<uint64_t>(5));
// 1 SST for big key + 1 SST for small one
ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "dobrynia"),
static_cast<uint64_t>(2));
// 1 SST for all keys
ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "nikitich"),
static_cast<uint64_t>(1));
}
}
// In https://reviews.facebook.net/D20661 we change
// recovery behavior: previously for each log file each column family
// memtable was flushed, even it wasn't empty. Now it's changed:
// we try to create the smallest number of table files by merging
// updates from multiple logs
TEST_F(DBWALTest, RecoverCheckFileAmount) {
Options options = CurrentOptions();
options.write_buffer_size = 100000;
options.arena_block_size = 4 * 1024;
CreateAndReopenWithCF({"pikachu", "dobrynia", "nikitich"}, options);
ASSERT_OK(Put(0, Key(1), DummyString(1)));
ASSERT_OK(Put(1, Key(1), DummyString(1)));
ASSERT_OK(Put(2, Key(1), DummyString(1)));
// Make 'nikitich' memtable to be flushed
ASSERT_OK(Put(3, Key(10), DummyString(1002400)));
ASSERT_OK(Put(3, Key(1), DummyString(1)));
dbfull()->TEST_WaitForFlushMemTable(handles_[3]);
// 4 memtable are not flushed, 1 sst file
{
auto tables = ListTableFiles(env_, dbname_);
ASSERT_EQ(tables.size(), static_cast<size_t>(1));
ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "nikitich"),
static_cast<uint64_t>(1));
}
// Memtable for 'nikitich' has flushed, new WAL file has opened
// 4 memtable still not flushed
// Write to new WAL file
ASSERT_OK(Put(0, Key(1), DummyString(1)));
ASSERT_OK(Put(1, Key(1), DummyString(1)));
ASSERT_OK(Put(2, Key(1), DummyString(1)));
// Fill up 'nikitich' one more time
ASSERT_OK(Put(3, Key(10), DummyString(1002400)));
// make it flush
ASSERT_OK(Put(3, Key(1), DummyString(1)));
dbfull()->TEST_WaitForFlushMemTable(handles_[3]);
// There are still 4 memtable not flushed, and 2 sst tables
ASSERT_OK(Put(0, Key(1), DummyString(1)));
ASSERT_OK(Put(1, Key(1), DummyString(1)));
ASSERT_OK(Put(2, Key(1), DummyString(1)));
{
auto tables = ListTableFiles(env_, dbname_);
ASSERT_EQ(tables.size(), static_cast<size_t>(2));
ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "nikitich"),
static_cast<uint64_t>(2));
}
ReopenWithColumnFamilies({"default", "pikachu", "dobrynia", "nikitich"},
options);
{
std::vector<uint64_t> table_files = ListTableFiles(env_, dbname_);
// Check, that records for 'default', 'dobrynia' and 'pikachu' from
// first, second and third WALs went to the same SST.
// So, there is 6 SSTs: three for 'nikitich', one for 'default', one for
// 'dobrynia', one for 'pikachu'
ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "default"),
static_cast<uint64_t>(1));
ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "nikitich"),
static_cast<uint64_t>(3));
ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "dobrynia"),
static_cast<uint64_t>(1));
ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "pikachu"),
static_cast<uint64_t>(1));
}
}
TEST_F(DBWALTest, SyncMultipleLogs) {
const uint64_t kNumBatches = 2;
const int kBatchSize = 1000;
Options options = CurrentOptions();
options.create_if_missing = true;
options.write_buffer_size = 4096;
Reopen(options);
WriteBatch batch;
WriteOptions wo;
wo.sync = true;
for (uint64_t b = 0; b < kNumBatches; b++) {
batch.Clear();
for (int i = 0; i < kBatchSize; i++) {
batch.Put(Key(i), DummyString(128));
}
dbfull()->Write(wo, &batch);
}
ASSERT_OK(dbfull()->SyncWAL());
}
//
// Test WAL recovery for the various modes available
//
class RecoveryTestHelper {
public:
// Number of WAL files to generate
static const int kWALFilesCount = 10;
// Starting number for the WAL file name like 00010.log
static const int kWALFileOffset = 10;
// Keys to be written per WAL file
static const int kKeysPerWALFile = 1024;
// Size of the value
static const int kValueSize = 10;
// Create WAL files with values filled in
static void FillData(DBWALTest* test, const Options& options,
const size_t wal_count, size_t* count) {
const DBOptions& db_options = options;
*count = 0;
shared_ptr<Cache> table_cache = NewLRUCache(50000, 16);
EnvOptions env_options;
WriteBuffer write_buffer(db_options.db_write_buffer_size);
unique_ptr<VersionSet> versions;
unique_ptr<WalManager> wal_manager;
WriteController write_controller;
versions.reset(new VersionSet(test->dbname_, &db_options, env_options,
table_cache.get(), &write_buffer,
&write_controller));
wal_manager.reset(new WalManager(db_options, env_options));
std::unique_ptr<log::Writer> current_log_writer;
for (size_t j = kWALFileOffset; j < wal_count + kWALFileOffset; j++) {
uint64_t current_log_number = j;
std::string fname = LogFileName(test->dbname_, current_log_number);
unique_ptr<WritableFile> file;
ASSERT_OK(db_options.env->NewWritableFile(fname, &file, env_options));
unique_ptr<WritableFileWriter> file_writer(
new WritableFileWriter(std::move(file), env_options));
current_log_writer.reset(
new log::Writer(std::move(file_writer), current_log_number,
db_options.recycle_log_file_num > 0));
for (int i = 0; i < kKeysPerWALFile; i++) {
std::string key = "key" + ToString((*count)++);
std::string value = test->DummyString(kValueSize);
assert(current_log_writer.get() != nullptr);
uint64_t seq = versions->LastSequence() + 1;
WriteBatch batch;
batch.Put(key, value);
WriteBatchInternal::SetSequence(&batch, seq);
current_log_writer->AddRecord(WriteBatchInternal::Contents(&batch));
versions->SetLastSequence(seq);
}
}
}
// Recreate and fill the store with some data
static size_t FillData(DBWALTest* test, Options* options) {
options->create_if_missing = true;
test->DestroyAndReopen(*options);
test->Close();
size_t count = 0;
FillData(test, *options, kWALFilesCount, &count);
return count;
}
// Read back all the keys we wrote and return the number of keys found
static size_t GetData(DBWALTest* test) {
size_t count = 0;
for (size_t i = 0; i < kWALFilesCount * kKeysPerWALFile; i++) {
if (test->Get("key" + ToString(i)) != "NOT_FOUND") {
++count;
}
}
return count;
}
// Manuall corrupt the specified WAL
static void CorruptWAL(DBWALTest* test, const Options& options,
const double off, const double len,
const int wal_file_id, const bool trunc = false) {
Env* env = options.env;
std::string fname = LogFileName(test->dbname_, wal_file_id);
uint64_t size;
ASSERT_OK(env->GetFileSize(fname, &size));
ASSERT_GT(size, 0);
#ifdef OS_WIN
// Windows disk cache behaves differently. When we truncate
// the original content is still in the cache due to the original
// handle is still open. Generally, in Windows, one prohibits
// shared access to files and it is not needed for WAL but we allow
// it to induce corruption at various tests.
test->Close();
#endif
if (trunc) {
ASSERT_EQ(0, truncate(fname.c_str(), static_cast<int64_t>(size * off)));
} else {
InduceCorruption(fname, static_cast<size_t>(size * off),
static_cast<size_t>(size * len));
}
}
// Overwrite data with 'a' from offset for length len
static void InduceCorruption(const std::string& filename, size_t offset,
size_t len) {
ASSERT_GT(len, 0U);
int fd = open(filename.c_str(), O_RDWR);
ASSERT_GT(fd, 0);
ASSERT_EQ(offset, lseek(fd, offset, SEEK_SET));
void* buf = alloca(len);
memset(buf, 'a', len);
ASSERT_EQ(len, write(fd, buf, static_cast<unsigned int>(len)));
close(fd);
}
};
// Test scope:
// - We expect to open the data store when there is incomplete trailing writes
// at the end of any of the logs
// - We do not expect to open the data store for corruption
TEST_F(DBWALTest, kTolerateCorruptedTailRecords) {
const int jstart = RecoveryTestHelper::kWALFileOffset;
const int jend = jstart + RecoveryTestHelper::kWALFilesCount;
for (auto trunc : {true, false}) { /* Corruption style */
for (int i = 0; i < 4; i++) { /* Corruption offset position */
for (int j = jstart; j < jend; j++) { /* WAL file */
// Fill data for testing
Options options = CurrentOptions();
const size_t row_count = RecoveryTestHelper::FillData(this, &options);
// test checksum failure or parsing
RecoveryTestHelper::CorruptWAL(this, options, /*off=*/i * .3,
/*len%=*/.1, /*wal=*/j, trunc);
if (trunc) {
options.wal_recovery_mode =
WALRecoveryMode::kTolerateCorruptedTailRecords;
options.create_if_missing = false;
ASSERT_OK(TryReopen(options));
const size_t recovered_row_count = RecoveryTestHelper::GetData(this);
ASSERT_TRUE(i == 0 || recovered_row_count > 0);
ASSERT_LT(recovered_row_count, row_count);
} else {
options.wal_recovery_mode =
WALRecoveryMode::kTolerateCorruptedTailRecords;
ASSERT_NOK(TryReopen(options));
}
}
}
}
}
// Test scope:
// We don't expect the data store to be opened if there is any corruption
// (leading, middle or trailing -- incomplete writes or corruption)
TEST_F(DBWALTest, kAbsoluteConsistency) {
const int jstart = RecoveryTestHelper::kWALFileOffset;
const int jend = jstart + RecoveryTestHelper::kWALFilesCount;
// Verify clean slate behavior
Options options = CurrentOptions();
const size_t row_count = RecoveryTestHelper::FillData(this, &options);
options.wal_recovery_mode = WALRecoveryMode::kAbsoluteConsistency;
options.create_if_missing = false;
ASSERT_OK(TryReopen(options));
ASSERT_EQ(RecoveryTestHelper::GetData(this), row_count);
for (auto trunc : {true, false}) { /* Corruption style */
for (int i = 0; i < 4; i++) { /* Corruption offset position */
if (trunc && i == 0) {
continue;
}
for (int j = jstart; j < jend; j++) { /* wal files */
// fill with new date
RecoveryTestHelper::FillData(this, &options);
// corrupt the wal
RecoveryTestHelper::CorruptWAL(this, options, /*off=*/i * .3,
/*len%=*/.1, j, trunc);
// verify
options.wal_recovery_mode = WALRecoveryMode::kAbsoluteConsistency;
options.create_if_missing = false;
ASSERT_NOK(TryReopen(options));
}
}
}
}
// Test scope:
// - We expect to open data store under all circumstances
// - We expect only data upto the point where the first error was encountered
TEST_F(DBWALTest, kPointInTimeRecovery) {
const int jstart = RecoveryTestHelper::kWALFileOffset;
const int jend = jstart + RecoveryTestHelper::kWALFilesCount;
const int maxkeys =
RecoveryTestHelper::kWALFilesCount * RecoveryTestHelper::kKeysPerWALFile;
for (auto trunc : {true, false}) { /* Corruption style */
for (int i = 0; i < 4; i++) { /* Offset of corruption */
for (int j = jstart; j < jend; j++) { /* WAL file */
// Fill data for testing
Options options = CurrentOptions();
const size_t row_count = RecoveryTestHelper::FillData(this, &options);
// Corrupt the wal
RecoveryTestHelper::CorruptWAL(this, options, /*off=*/i * .3,
/*len%=*/.1, j, trunc);
// Verify
options.wal_recovery_mode = WALRecoveryMode::kPointInTimeRecovery;
options.create_if_missing = false;
ASSERT_OK(TryReopen(options));
// Probe data for invariants
size_t recovered_row_count = RecoveryTestHelper::GetData(this);
ASSERT_LT(recovered_row_count, row_count);
bool expect_data = true;
for (size_t k = 0; k < maxkeys; ++k) {
bool found = Get("key" + ToString(i)) != "NOT_FOUND";
if (expect_data && !found) {
expect_data = false;
}
ASSERT_EQ(found, expect_data);
}
const size_t min = RecoveryTestHelper::kKeysPerWALFile *
(j - RecoveryTestHelper::kWALFileOffset);
ASSERT_GE(recovered_row_count, min);
if (!trunc && i != 0) {
const size_t max = RecoveryTestHelper::kKeysPerWALFile *
(j - RecoveryTestHelper::kWALFileOffset + 1);
ASSERT_LE(recovered_row_count, max);
}
}
}
}
}
// Test scope:
// - We expect to open the data store under all scenarios
// - We expect to have recovered records past the corruption zone
TEST_F(DBWALTest, kSkipAnyCorruptedRecords) {
const int jstart = RecoveryTestHelper::kWALFileOffset;
const int jend = jstart + RecoveryTestHelper::kWALFilesCount;
for (auto trunc : {true, false}) { /* Corruption style */
for (int i = 0; i < 4; i++) { /* Corruption offset */
for (int j = jstart; j < jend; j++) { /* wal files */
// Fill data for testing
Options options = CurrentOptions();
const size_t row_count = RecoveryTestHelper::FillData(this, &options);
// Corrupt the WAL
RecoveryTestHelper::CorruptWAL(this, options, /*off=*/i * .3,
/*len%=*/.1, j, trunc);
// Verify behavior
options.wal_recovery_mode = WALRecoveryMode::kSkipAnyCorruptedRecords;
options.create_if_missing = false;
ASSERT_OK(TryReopen(options));
// Probe data for invariants
size_t recovered_row_count = RecoveryTestHelper::GetData(this);
ASSERT_LT(recovered_row_count, row_count);
if (!trunc) {
ASSERT_TRUE(i != 0 || recovered_row_count > 0);
}
}
}
}
}
#endif // ROCKSDB_LITE
} // namespace rocksdb
int main(int argc, char** argv) {

6
src.mk
View File

@ -198,13 +198,17 @@ TEST_BENCH_SOURCES = \
db/dbformat_test.cc \
db/db_iter_test.cc \
db/db_test.cc \
db/db_block_cache_test.cc \
db/db_bloom_filter_test.cc \
db/db_compaction_filter_test.cc \
db/db_compaction_test.cc \
db/db_dynamic_level_test.cc \
db/db_inplace_update_test.cc \
db/db_iterator_test.cc \
db/db_log_iter_test.cc \
db/db_universal_compaction_test.cc \
db/db_sst_test.cc \
db/db_tailing_iter_test.cc \
db/db_universal_compaction_test.cc \
db/db_wal_test.cc \
db/db_table_properties_test.cc \
db/deletefile_test.cc \