rocksdb/util/cache_test.cc

427 lines
11 KiB
C++
Raw Normal View History

// Copyright (c) 2013, Facebook, Inc. All rights reserved.
// This source code is licensed under the BSD-style license found in the
// LICENSE file in the root directory of this source tree. An additional grant
// of patent rights can be found in the PATENTS file in the same 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 "rocksdb/cache.h"
#include <vector>
#include <string>
#include <iostream>
#include "util/coding.h"
#include "util/testharness.h"
namespace rocksdb {
// Conversions between numeric keys/values and the types expected by Cache.
static std::string EncodeKey(int k) {
std::string result;
PutFixed32(&result, k);
return result;
}
static int DecodeKey(const Slice& k) {
assert(k.size() == 4);
return DecodeFixed32(k.data());
}
static void* EncodeValue(uintptr_t v) { return reinterpret_cast<void*>(v); }
static int DecodeValue(void* v) {
return static_cast<int>(reinterpret_cast<uintptr_t>(v));
}
rocksdb: switch to gtest Summary: Our existing test notation is very similar to what is used in gtest. It makes it easy to adopt what is different. In this diff I modify existing [[ https://code.google.com/p/googletest/wiki/Primer#Test_Fixtures:_Using_the_Same_Data_Configuration_for_Multiple_Te | test fixture ]] classes to inherit from `testing::Test`. Also for unit tests that use fixture class, `TEST` is replaced with `TEST_F` as required in gtest. There are several custom `main` functions in our existing tests. To make this transition easier, I modify all `main` functions to fallow gtest notation. But eventually we can remove them and use implementation of `main` that gtest provides. ```lang=bash % cat ~/transform #!/bin/sh files=$(git ls-files '*test\.cc') for file in $files do if grep -q "rocksdb::test::RunAllTests()" $file then if grep -Eq '^class \w+Test {' $file then perl -pi -e 's/^(class \w+Test) {/${1}: public testing::Test {/g' $file perl -pi -e 's/^(TEST)/${1}_F/g' $file fi perl -pi -e 's/(int main.*\{)/${1}::testing::InitGoogleTest(&argc, argv);/g' $file perl -pi -e 's/rocksdb::test::RunAllTests/RUN_ALL_TESTS/g' $file fi done % sh ~/transform % make format ``` Second iteration of this diff contains only scripted changes. Third iteration contains manual changes to fix last errors and make it compilable. Test Plan: Build and notice no errors. ```lang=bash % USE_CLANG=1 make check -j55 ``` Tests are still testing. Reviewers: meyering, sdong, rven, igor Reviewed By: igor Subscribers: dhruba, leveldb Differential Revision: https://reviews.facebook.net/D35157
2015-03-17 22:08:00 +01:00
class CacheTest : public testing::Test {
public:
static CacheTest* current_;
static void Deleter(const Slice& key, void* v) {
current_->deleted_keys_.push_back(DecodeKey(key));
current_->deleted_values_.push_back(DecodeValue(v));
}
static const int kCacheSize = 1000;
static const int kNumShardBits = 4;
static const int kRemoveScanCountLimit = 16;
static const int kCacheSize2 = 100;
static const int kNumShardBits2 = 2;
static const int kRemoveScanCountLimit2 = 200;
std::vector<int> deleted_keys_;
std::vector<int> deleted_values_;
shared_ptr<Cache> cache_;
shared_ptr<Cache> cache2_;
CacheTest() :
cache_(NewLRUCache(kCacheSize, kNumShardBits, kRemoveScanCountLimit)),
cache2_(NewLRUCache(kCacheSize2, kNumShardBits2,
kRemoveScanCountLimit2)) {
current_ = this;
}
~CacheTest() {
}
int Lookup(shared_ptr<Cache> cache, int key) {
Cache::Handle* handle = cache->Lookup(EncodeKey(key));
const int r = (handle == nullptr) ? -1 : DecodeValue(cache->Value(handle));
if (handle != nullptr) {
cache->Release(handle);
}
return r;
}
void Insert(shared_ptr<Cache> cache, int key, int value, int charge = 1) {
cache->Release(cache->Insert(EncodeKey(key), EncodeValue(value), charge,
&CacheTest::Deleter));
}
void Erase(shared_ptr<Cache> cache, int key) {
cache->Erase(EncodeKey(key));
}
int Lookup(int key) {
return Lookup(cache_, key);
}
void Insert(int key, int value, int charge = 1) {
Insert(cache_, key, value, charge);
}
void Erase(int key) {
Erase(cache_, key);
}
int Lookup2(int key) {
return Lookup(cache2_, key);
}
void Insert2(int key, int value, int charge = 1) {
Insert(cache2_, key, value, charge);
}
void Erase2(int key) {
Erase(cache2_, key);
}
};
CacheTest* CacheTest::current_;
namespace {
void dumbDeleter(const Slice& key, void* value) { }
} // namespace
rocksdb: switch to gtest Summary: Our existing test notation is very similar to what is used in gtest. It makes it easy to adopt what is different. In this diff I modify existing [[ https://code.google.com/p/googletest/wiki/Primer#Test_Fixtures:_Using_the_Same_Data_Configuration_for_Multiple_Te | test fixture ]] classes to inherit from `testing::Test`. Also for unit tests that use fixture class, `TEST` is replaced with `TEST_F` as required in gtest. There are several custom `main` functions in our existing tests. To make this transition easier, I modify all `main` functions to fallow gtest notation. But eventually we can remove them and use implementation of `main` that gtest provides. ```lang=bash % cat ~/transform #!/bin/sh files=$(git ls-files '*test\.cc') for file in $files do if grep -q "rocksdb::test::RunAllTests()" $file then if grep -Eq '^class \w+Test {' $file then perl -pi -e 's/^(class \w+Test) {/${1}: public testing::Test {/g' $file perl -pi -e 's/^(TEST)/${1}_F/g' $file fi perl -pi -e 's/(int main.*\{)/${1}::testing::InitGoogleTest(&argc, argv);/g' $file perl -pi -e 's/rocksdb::test::RunAllTests/RUN_ALL_TESTS/g' $file fi done % sh ~/transform % make format ``` Second iteration of this diff contains only scripted changes. Third iteration contains manual changes to fix last errors and make it compilable. Test Plan: Build and notice no errors. ```lang=bash % USE_CLANG=1 make check -j55 ``` Tests are still testing. Reviewers: meyering, sdong, rven, igor Reviewed By: igor Subscribers: dhruba, leveldb Differential Revision: https://reviews.facebook.net/D35157
2015-03-17 22:08:00 +01:00
TEST_F(CacheTest, UsageTest) {
// cache is shared_ptr and will be automatically cleaned up.
const uint64_t kCapacity = 100000;
auto cache = NewLRUCache(kCapacity, 8, 200);
size_t usage = 0;
const char* value = "abcdef";
// make sure everything will be cached
for (int i = 1; i < 100; ++i) {
std::string key(i, 'a');
auto kv_size = key.size() + 5;
cache->Release(
cache->Insert(key, (void*)value, kv_size, dumbDeleter)
);
usage += kv_size;
ASSERT_EQ(usage, cache->GetUsage());
}
// make sure the cache will be overloaded
for (uint64_t i = 1; i < kCapacity; ++i) {
auto key = ToString(i);
cache->Release(
cache->Insert(key, (void*)value, key.size() + 5, dumbDeleter)
);
}
// the usage should be close to the capacity
ASSERT_GT(kCapacity, cache->GetUsage());
ASSERT_LT(kCapacity * 0.95, cache->GetUsage());
}
rocksdb: switch to gtest Summary: Our existing test notation is very similar to what is used in gtest. It makes it easy to adopt what is different. In this diff I modify existing [[ https://code.google.com/p/googletest/wiki/Primer#Test_Fixtures:_Using_the_Same_Data_Configuration_for_Multiple_Te | test fixture ]] classes to inherit from `testing::Test`. Also for unit tests that use fixture class, `TEST` is replaced with `TEST_F` as required in gtest. There are several custom `main` functions in our existing tests. To make this transition easier, I modify all `main` functions to fallow gtest notation. But eventually we can remove them and use implementation of `main` that gtest provides. ```lang=bash % cat ~/transform #!/bin/sh files=$(git ls-files '*test\.cc') for file in $files do if grep -q "rocksdb::test::RunAllTests()" $file then if grep -Eq '^class \w+Test {' $file then perl -pi -e 's/^(class \w+Test) {/${1}: public testing::Test {/g' $file perl -pi -e 's/^(TEST)/${1}_F/g' $file fi perl -pi -e 's/(int main.*\{)/${1}::testing::InitGoogleTest(&argc, argv);/g' $file perl -pi -e 's/rocksdb::test::RunAllTests/RUN_ALL_TESTS/g' $file fi done % sh ~/transform % make format ``` Second iteration of this diff contains only scripted changes. Third iteration contains manual changes to fix last errors and make it compilable. Test Plan: Build and notice no errors. ```lang=bash % USE_CLANG=1 make check -j55 ``` Tests are still testing. Reviewers: meyering, sdong, rven, igor Reviewed By: igor Subscribers: dhruba, leveldb Differential Revision: https://reviews.facebook.net/D35157
2015-03-17 22:08:00 +01:00
TEST_F(CacheTest, HitAndMiss) {
ASSERT_EQ(-1, Lookup(100));
Insert(100, 101);
ASSERT_EQ(101, Lookup(100));
ASSERT_EQ(-1, Lookup(200));
ASSERT_EQ(-1, Lookup(300));
Insert(200, 201);
ASSERT_EQ(101, Lookup(100));
ASSERT_EQ(201, Lookup(200));
ASSERT_EQ(-1, Lookup(300));
Insert(100, 102);
ASSERT_EQ(102, Lookup(100));
ASSERT_EQ(201, Lookup(200));
ASSERT_EQ(-1, Lookup(300));
ASSERT_EQ(1U, deleted_keys_.size());
ASSERT_EQ(100, deleted_keys_[0]);
ASSERT_EQ(101, deleted_values_[0]);
}
rocksdb: switch to gtest Summary: Our existing test notation is very similar to what is used in gtest. It makes it easy to adopt what is different. In this diff I modify existing [[ https://code.google.com/p/googletest/wiki/Primer#Test_Fixtures:_Using_the_Same_Data_Configuration_for_Multiple_Te | test fixture ]] classes to inherit from `testing::Test`. Also for unit tests that use fixture class, `TEST` is replaced with `TEST_F` as required in gtest. There are several custom `main` functions in our existing tests. To make this transition easier, I modify all `main` functions to fallow gtest notation. But eventually we can remove them and use implementation of `main` that gtest provides. ```lang=bash % cat ~/transform #!/bin/sh files=$(git ls-files '*test\.cc') for file in $files do if grep -q "rocksdb::test::RunAllTests()" $file then if grep -Eq '^class \w+Test {' $file then perl -pi -e 's/^(class \w+Test) {/${1}: public testing::Test {/g' $file perl -pi -e 's/^(TEST)/${1}_F/g' $file fi perl -pi -e 's/(int main.*\{)/${1}::testing::InitGoogleTest(&argc, argv);/g' $file perl -pi -e 's/rocksdb::test::RunAllTests/RUN_ALL_TESTS/g' $file fi done % sh ~/transform % make format ``` Second iteration of this diff contains only scripted changes. Third iteration contains manual changes to fix last errors and make it compilable. Test Plan: Build and notice no errors. ```lang=bash % USE_CLANG=1 make check -j55 ``` Tests are still testing. Reviewers: meyering, sdong, rven, igor Reviewed By: igor Subscribers: dhruba, leveldb Differential Revision: https://reviews.facebook.net/D35157
2015-03-17 22:08:00 +01:00
TEST_F(CacheTest, Erase) {
Erase(200);
ASSERT_EQ(0U, deleted_keys_.size());
Insert(100, 101);
Insert(200, 201);
Erase(100);
ASSERT_EQ(-1, Lookup(100));
ASSERT_EQ(201, Lookup(200));
ASSERT_EQ(1U, deleted_keys_.size());
ASSERT_EQ(100, deleted_keys_[0]);
ASSERT_EQ(101, deleted_values_[0]);
Erase(100);
ASSERT_EQ(-1, Lookup(100));
ASSERT_EQ(201, Lookup(200));
ASSERT_EQ(1U, deleted_keys_.size());
}
rocksdb: switch to gtest Summary: Our existing test notation is very similar to what is used in gtest. It makes it easy to adopt what is different. In this diff I modify existing [[ https://code.google.com/p/googletest/wiki/Primer#Test_Fixtures:_Using_the_Same_Data_Configuration_for_Multiple_Te | test fixture ]] classes to inherit from `testing::Test`. Also for unit tests that use fixture class, `TEST` is replaced with `TEST_F` as required in gtest. There are several custom `main` functions in our existing tests. To make this transition easier, I modify all `main` functions to fallow gtest notation. But eventually we can remove them and use implementation of `main` that gtest provides. ```lang=bash % cat ~/transform #!/bin/sh files=$(git ls-files '*test\.cc') for file in $files do if grep -q "rocksdb::test::RunAllTests()" $file then if grep -Eq '^class \w+Test {' $file then perl -pi -e 's/^(class \w+Test) {/${1}: public testing::Test {/g' $file perl -pi -e 's/^(TEST)/${1}_F/g' $file fi perl -pi -e 's/(int main.*\{)/${1}::testing::InitGoogleTest(&argc, argv);/g' $file perl -pi -e 's/rocksdb::test::RunAllTests/RUN_ALL_TESTS/g' $file fi done % sh ~/transform % make format ``` Second iteration of this diff contains only scripted changes. Third iteration contains manual changes to fix last errors and make it compilable. Test Plan: Build and notice no errors. ```lang=bash % USE_CLANG=1 make check -j55 ``` Tests are still testing. Reviewers: meyering, sdong, rven, igor Reviewed By: igor Subscribers: dhruba, leveldb Differential Revision: https://reviews.facebook.net/D35157
2015-03-17 22:08:00 +01:00
TEST_F(CacheTest, EntriesArePinned) {
Insert(100, 101);
Cache::Handle* h1 = cache_->Lookup(EncodeKey(100));
ASSERT_EQ(101, DecodeValue(cache_->Value(h1)));
ASSERT_EQ(1U, cache_->GetUsage());
Insert(100, 102);
Cache::Handle* h2 = cache_->Lookup(EncodeKey(100));
ASSERT_EQ(102, DecodeValue(cache_->Value(h2)));
ASSERT_EQ(0U, deleted_keys_.size());
ASSERT_EQ(2U, cache_->GetUsage());
cache_->Release(h1);
ASSERT_EQ(1U, deleted_keys_.size());
ASSERT_EQ(100, deleted_keys_[0]);
ASSERT_EQ(101, deleted_values_[0]);
ASSERT_EQ(1U, cache_->GetUsage());
Erase(100);
ASSERT_EQ(-1, Lookup(100));
ASSERT_EQ(1U, deleted_keys_.size());
ASSERT_EQ(1U, cache_->GetUsage());
cache_->Release(h2);
ASSERT_EQ(2U, deleted_keys_.size());
ASSERT_EQ(100, deleted_keys_[1]);
ASSERT_EQ(102, deleted_values_[1]);
ASSERT_EQ(0U, cache_->GetUsage());
}
rocksdb: switch to gtest Summary: Our existing test notation is very similar to what is used in gtest. It makes it easy to adopt what is different. In this diff I modify existing [[ https://code.google.com/p/googletest/wiki/Primer#Test_Fixtures:_Using_the_Same_Data_Configuration_for_Multiple_Te | test fixture ]] classes to inherit from `testing::Test`. Also for unit tests that use fixture class, `TEST` is replaced with `TEST_F` as required in gtest. There are several custom `main` functions in our existing tests. To make this transition easier, I modify all `main` functions to fallow gtest notation. But eventually we can remove them and use implementation of `main` that gtest provides. ```lang=bash % cat ~/transform #!/bin/sh files=$(git ls-files '*test\.cc') for file in $files do if grep -q "rocksdb::test::RunAllTests()" $file then if grep -Eq '^class \w+Test {' $file then perl -pi -e 's/^(class \w+Test) {/${1}: public testing::Test {/g' $file perl -pi -e 's/^(TEST)/${1}_F/g' $file fi perl -pi -e 's/(int main.*\{)/${1}::testing::InitGoogleTest(&argc, argv);/g' $file perl -pi -e 's/rocksdb::test::RunAllTests/RUN_ALL_TESTS/g' $file fi done % sh ~/transform % make format ``` Second iteration of this diff contains only scripted changes. Third iteration contains manual changes to fix last errors and make it compilable. Test Plan: Build and notice no errors. ```lang=bash % USE_CLANG=1 make check -j55 ``` Tests are still testing. Reviewers: meyering, sdong, rven, igor Reviewed By: igor Subscribers: dhruba, leveldb Differential Revision: https://reviews.facebook.net/D35157
2015-03-17 22:08:00 +01:00
TEST_F(CacheTest, EvictionPolicy) {
Insert(100, 101);
Insert(200, 201);
// Frequently used entry must be kept around
for (int i = 0; i < kCacheSize + 100; i++) {
Insert(1000+i, 2000+i);
ASSERT_EQ(2000+i, Lookup(1000+i));
ASSERT_EQ(101, Lookup(100));
}
ASSERT_EQ(101, Lookup(100));
ASSERT_EQ(-1, Lookup(200));
}
rocksdb: switch to gtest Summary: Our existing test notation is very similar to what is used in gtest. It makes it easy to adopt what is different. In this diff I modify existing [[ https://code.google.com/p/googletest/wiki/Primer#Test_Fixtures:_Using_the_Same_Data_Configuration_for_Multiple_Te | test fixture ]] classes to inherit from `testing::Test`. Also for unit tests that use fixture class, `TEST` is replaced with `TEST_F` as required in gtest. There are several custom `main` functions in our existing tests. To make this transition easier, I modify all `main` functions to fallow gtest notation. But eventually we can remove them and use implementation of `main` that gtest provides. ```lang=bash % cat ~/transform #!/bin/sh files=$(git ls-files '*test\.cc') for file in $files do if grep -q "rocksdb::test::RunAllTests()" $file then if grep -Eq '^class \w+Test {' $file then perl -pi -e 's/^(class \w+Test) {/${1}: public testing::Test {/g' $file perl -pi -e 's/^(TEST)/${1}_F/g' $file fi perl -pi -e 's/(int main.*\{)/${1}::testing::InitGoogleTest(&argc, argv);/g' $file perl -pi -e 's/rocksdb::test::RunAllTests/RUN_ALL_TESTS/g' $file fi done % sh ~/transform % make format ``` Second iteration of this diff contains only scripted changes. Third iteration contains manual changes to fix last errors and make it compilable. Test Plan: Build and notice no errors. ```lang=bash % USE_CLANG=1 make check -j55 ``` Tests are still testing. Reviewers: meyering, sdong, rven, igor Reviewed By: igor Subscribers: dhruba, leveldb Differential Revision: https://reviews.facebook.net/D35157
2015-03-17 22:08:00 +01:00
TEST_F(CacheTest, EvictionPolicyRef) {
Insert(100, 101);
Insert(101, 102);
Insert(102, 103);
Insert(103, 104);
Insert(200, 101);
Insert(201, 102);
Insert(202, 103);
Insert(203, 104);
Cache::Handle* h201 = cache_->Lookup(EncodeKey(200));
Cache::Handle* h202 = cache_->Lookup(EncodeKey(201));
Cache::Handle* h203 = cache_->Lookup(EncodeKey(202));
Cache::Handle* h204 = cache_->Lookup(EncodeKey(203));
Insert(300, 101);
Insert(301, 102);
Insert(302, 103);
Insert(303, 104);
// Insert entries much more than Cache capacity
for (int i = 0; i < kCacheSize + 100; i++) {
Insert(1000 + i, 2000 + i);
}
// Check whether the entries inserted in the beginning
// are evicted. Ones without extra ref are evicted and
// those with are not.
ASSERT_EQ(-1, Lookup(100));
ASSERT_EQ(-1, Lookup(101));
ASSERT_EQ(-1, Lookup(102));
ASSERT_EQ(-1, Lookup(103));
ASSERT_EQ(-1, Lookup(300));
ASSERT_EQ(-1, Lookup(301));
ASSERT_EQ(-1, Lookup(302));
ASSERT_EQ(-1, Lookup(303));
ASSERT_EQ(101, Lookup(200));
ASSERT_EQ(102, Lookup(201));
ASSERT_EQ(103, Lookup(202));
ASSERT_EQ(104, Lookup(203));
// Cleaning up all the handles
cache_->Release(h201);
cache_->Release(h202);
cache_->Release(h203);
cache_->Release(h204);
}
rocksdb: switch to gtest Summary: Our existing test notation is very similar to what is used in gtest. It makes it easy to adopt what is different. In this diff I modify existing [[ https://code.google.com/p/googletest/wiki/Primer#Test_Fixtures:_Using_the_Same_Data_Configuration_for_Multiple_Te | test fixture ]] classes to inherit from `testing::Test`. Also for unit tests that use fixture class, `TEST` is replaced with `TEST_F` as required in gtest. There are several custom `main` functions in our existing tests. To make this transition easier, I modify all `main` functions to fallow gtest notation. But eventually we can remove them and use implementation of `main` that gtest provides. ```lang=bash % cat ~/transform #!/bin/sh files=$(git ls-files '*test\.cc') for file in $files do if grep -q "rocksdb::test::RunAllTests()" $file then if grep -Eq '^class \w+Test {' $file then perl -pi -e 's/^(class \w+Test) {/${1}: public testing::Test {/g' $file perl -pi -e 's/^(TEST)/${1}_F/g' $file fi perl -pi -e 's/(int main.*\{)/${1}::testing::InitGoogleTest(&argc, argv);/g' $file perl -pi -e 's/rocksdb::test::RunAllTests/RUN_ALL_TESTS/g' $file fi done % sh ~/transform % make format ``` Second iteration of this diff contains only scripted changes. Third iteration contains manual changes to fix last errors and make it compilable. Test Plan: Build and notice no errors. ```lang=bash % USE_CLANG=1 make check -j55 ``` Tests are still testing. Reviewers: meyering, sdong, rven, igor Reviewed By: igor Subscribers: dhruba, leveldb Differential Revision: https://reviews.facebook.net/D35157
2015-03-17 22:08:00 +01:00
TEST_F(CacheTest, ErasedHandleState) {
Modifed the LRU cache eviction code so that it doesn't evict blocks which have exteranl references Summary: Currently, blocks which have more than one reference (ie referenced by something other than cache itself) are evicted from cache. This doesn't make much sense: - blocks are still in RAM, so the RAM usage reported by the cache is incorrect - if the same block is needed by another iterator, it will be loaded and decompressed again This diff changes the reference counting scheme a bit. Previously, if the cache contained the block, this was accounted for in its refcount. After this change, the refcount is only used to track external references. There is a boolean flag which indicates whether or not the block is contained in the cache. This diff also changes how LRU list is used. Previously, both hashtable and the LRU list contained all blocks. After this change, the LRU list contains blocks with the refcount==0, ie those which can be evicted from the cache. Note that this change still allows for cache to grow beyond its capacity. This happens when all blocks are pinned (ie refcount>0). This is consistent with the current behavior. The cache's insert function never fails. I spent lots of time trying to make table_reader and other places work with the insert which might failed. It turned out to be pretty hard. It might really destabilize some customers, so finally, I decided against doing this. table_cache_remove_scan_count_limit option will be unneeded after this change, but I will remove it in the following diff, if this one gets approved Test Plan: Ran tests, made sure they pass Reviewers: sdong, ljin Differential Revision: https://reviews.facebook.net/D25503
2014-10-21 20:49:13 +02:00
// insert a key and get two handles
Insert(100, 1000);
Cache::Handle* h1 = cache_->Lookup(EncodeKey(100));
Cache::Handle* h2 = cache_->Lookup(EncodeKey(100));
ASSERT_EQ(h1, h2);
ASSERT_EQ(DecodeValue(cache_->Value(h1)), 1000);
ASSERT_EQ(DecodeValue(cache_->Value(h2)), 1000);
Modifed the LRU cache eviction code so that it doesn't evict blocks which have exteranl references Summary: Currently, blocks which have more than one reference (ie referenced by something other than cache itself) are evicted from cache. This doesn't make much sense: - blocks are still in RAM, so the RAM usage reported by the cache is incorrect - if the same block is needed by another iterator, it will be loaded and decompressed again This diff changes the reference counting scheme a bit. Previously, if the cache contained the block, this was accounted for in its refcount. After this change, the refcount is only used to track external references. There is a boolean flag which indicates whether or not the block is contained in the cache. This diff also changes how LRU list is used. Previously, both hashtable and the LRU list contained all blocks. After this change, the LRU list contains blocks with the refcount==0, ie those which can be evicted from the cache. Note that this change still allows for cache to grow beyond its capacity. This happens when all blocks are pinned (ie refcount>0). This is consistent with the current behavior. The cache's insert function never fails. I spent lots of time trying to make table_reader and other places work with the insert which might failed. It turned out to be pretty hard. It might really destabilize some customers, so finally, I decided against doing this. table_cache_remove_scan_count_limit option will be unneeded after this change, but I will remove it in the following diff, if this one gets approved Test Plan: Ran tests, made sure they pass Reviewers: sdong, ljin Differential Revision: https://reviews.facebook.net/D25503
2014-10-21 20:49:13 +02:00
// delete the key from the cache
Erase(100);
// can no longer find in the cache
ASSERT_EQ(-1, Lookup(100));
Modifed the LRU cache eviction code so that it doesn't evict blocks which have exteranl references Summary: Currently, blocks which have more than one reference (ie referenced by something other than cache itself) are evicted from cache. This doesn't make much sense: - blocks are still in RAM, so the RAM usage reported by the cache is incorrect - if the same block is needed by another iterator, it will be loaded and decompressed again This diff changes the reference counting scheme a bit. Previously, if the cache contained the block, this was accounted for in its refcount. After this change, the refcount is only used to track external references. There is a boolean flag which indicates whether or not the block is contained in the cache. This diff also changes how LRU list is used. Previously, both hashtable and the LRU list contained all blocks. After this change, the LRU list contains blocks with the refcount==0, ie those which can be evicted from the cache. Note that this change still allows for cache to grow beyond its capacity. This happens when all blocks are pinned (ie refcount>0). This is consistent with the current behavior. The cache's insert function never fails. I spent lots of time trying to make table_reader and other places work with the insert which might failed. It turned out to be pretty hard. It might really destabilize some customers, so finally, I decided against doing this. table_cache_remove_scan_count_limit option will be unneeded after this change, but I will remove it in the following diff, if this one gets approved Test Plan: Ran tests, made sure they pass Reviewers: sdong, ljin Differential Revision: https://reviews.facebook.net/D25503
2014-10-21 20:49:13 +02:00
// release one handle
cache_->Release(h1);
// still can't find in cache
ASSERT_EQ(-1, Lookup(100));
Modifed the LRU cache eviction code so that it doesn't evict blocks which have exteranl references Summary: Currently, blocks which have more than one reference (ie referenced by something other than cache itself) are evicted from cache. This doesn't make much sense: - blocks are still in RAM, so the RAM usage reported by the cache is incorrect - if the same block is needed by another iterator, it will be loaded and decompressed again This diff changes the reference counting scheme a bit. Previously, if the cache contained the block, this was accounted for in its refcount. After this change, the refcount is only used to track external references. There is a boolean flag which indicates whether or not the block is contained in the cache. This diff also changes how LRU list is used. Previously, both hashtable and the LRU list contained all blocks. After this change, the LRU list contains blocks with the refcount==0, ie those which can be evicted from the cache. Note that this change still allows for cache to grow beyond its capacity. This happens when all blocks are pinned (ie refcount>0). This is consistent with the current behavior. The cache's insert function never fails. I spent lots of time trying to make table_reader and other places work with the insert which might failed. It turned out to be pretty hard. It might really destabilize some customers, so finally, I decided against doing this. table_cache_remove_scan_count_limit option will be unneeded after this change, but I will remove it in the following diff, if this one gets approved Test Plan: Ran tests, made sure they pass Reviewers: sdong, ljin Differential Revision: https://reviews.facebook.net/D25503
2014-10-21 20:49:13 +02:00
cache_->Release(h2);
}
rocksdb: switch to gtest Summary: Our existing test notation is very similar to what is used in gtest. It makes it easy to adopt what is different. In this diff I modify existing [[ https://code.google.com/p/googletest/wiki/Primer#Test_Fixtures:_Using_the_Same_Data_Configuration_for_Multiple_Te | test fixture ]] classes to inherit from `testing::Test`. Also for unit tests that use fixture class, `TEST` is replaced with `TEST_F` as required in gtest. There are several custom `main` functions in our existing tests. To make this transition easier, I modify all `main` functions to fallow gtest notation. But eventually we can remove them and use implementation of `main` that gtest provides. ```lang=bash % cat ~/transform #!/bin/sh files=$(git ls-files '*test\.cc') for file in $files do if grep -q "rocksdb::test::RunAllTests()" $file then if grep -Eq '^class \w+Test {' $file then perl -pi -e 's/^(class \w+Test) {/${1}: public testing::Test {/g' $file perl -pi -e 's/^(TEST)/${1}_F/g' $file fi perl -pi -e 's/(int main.*\{)/${1}::testing::InitGoogleTest(&argc, argv);/g' $file perl -pi -e 's/rocksdb::test::RunAllTests/RUN_ALL_TESTS/g' $file fi done % sh ~/transform % make format ``` Second iteration of this diff contains only scripted changes. Third iteration contains manual changes to fix last errors and make it compilable. Test Plan: Build and notice no errors. ```lang=bash % USE_CLANG=1 make check -j55 ``` Tests are still testing. Reviewers: meyering, sdong, rven, igor Reviewed By: igor Subscribers: dhruba, leveldb Differential Revision: https://reviews.facebook.net/D35157
2015-03-17 22:08:00 +01:00
TEST_F(CacheTest, HeavyEntries) {
// Add a bunch of light and heavy entries and then count the combined
// size of items still in the cache, which must be approximately the
// same as the total capacity.
const int kLight = 1;
const int kHeavy = 10;
int added = 0;
int index = 0;
while (added < 2*kCacheSize) {
const int weight = (index & 1) ? kLight : kHeavy;
Insert(index, 1000+index, weight);
added += weight;
index++;
}
int cached_weight = 0;
for (int i = 0; i < index; i++) {
const int weight = (i & 1 ? kLight : kHeavy);
int r = Lookup(i);
if (r >= 0) {
cached_weight += weight;
ASSERT_EQ(1000+i, r);
}
}
ASSERT_LE(cached_weight, kCacheSize + kCacheSize/10);
}
rocksdb: switch to gtest Summary: Our existing test notation is very similar to what is used in gtest. It makes it easy to adopt what is different. In this diff I modify existing [[ https://code.google.com/p/googletest/wiki/Primer#Test_Fixtures:_Using_the_Same_Data_Configuration_for_Multiple_Te | test fixture ]] classes to inherit from `testing::Test`. Also for unit tests that use fixture class, `TEST` is replaced with `TEST_F` as required in gtest. There are several custom `main` functions in our existing tests. To make this transition easier, I modify all `main` functions to fallow gtest notation. But eventually we can remove them and use implementation of `main` that gtest provides. ```lang=bash % cat ~/transform #!/bin/sh files=$(git ls-files '*test\.cc') for file in $files do if grep -q "rocksdb::test::RunAllTests()" $file then if grep -Eq '^class \w+Test {' $file then perl -pi -e 's/^(class \w+Test) {/${1}: public testing::Test {/g' $file perl -pi -e 's/^(TEST)/${1}_F/g' $file fi perl -pi -e 's/(int main.*\{)/${1}::testing::InitGoogleTest(&argc, argv);/g' $file perl -pi -e 's/rocksdb::test::RunAllTests/RUN_ALL_TESTS/g' $file fi done % sh ~/transform % make format ``` Second iteration of this diff contains only scripted changes. Third iteration contains manual changes to fix last errors and make it compilable. Test Plan: Build and notice no errors. ```lang=bash % USE_CLANG=1 make check -j55 ``` Tests are still testing. Reviewers: meyering, sdong, rven, igor Reviewed By: igor Subscribers: dhruba, leveldb Differential Revision: https://reviews.facebook.net/D35157
2015-03-17 22:08:00 +01:00
TEST_F(CacheTest, NewId) {
uint64_t a = cache_->NewId();
uint64_t b = cache_->NewId();
ASSERT_NE(a, b);
}
class Value {
private:
size_t v_;
public:
explicit Value(size_t v) : v_(v) { }
~Value() { std::cout << v_ << " is destructed\n"; }
};
namespace {
void deleter(const Slice& key, void* value) {
delete static_cast<Value *>(value);
}
} // namespace
rocksdb: switch to gtest Summary: Our existing test notation is very similar to what is used in gtest. It makes it easy to adopt what is different. In this diff I modify existing [[ https://code.google.com/p/googletest/wiki/Primer#Test_Fixtures:_Using_the_Same_Data_Configuration_for_Multiple_Te | test fixture ]] classes to inherit from `testing::Test`. Also for unit tests that use fixture class, `TEST` is replaced with `TEST_F` as required in gtest. There are several custom `main` functions in our existing tests. To make this transition easier, I modify all `main` functions to fallow gtest notation. But eventually we can remove them and use implementation of `main` that gtest provides. ```lang=bash % cat ~/transform #!/bin/sh files=$(git ls-files '*test\.cc') for file in $files do if grep -q "rocksdb::test::RunAllTests()" $file then if grep -Eq '^class \w+Test {' $file then perl -pi -e 's/^(class \w+Test) {/${1}: public testing::Test {/g' $file perl -pi -e 's/^(TEST)/${1}_F/g' $file fi perl -pi -e 's/(int main.*\{)/${1}::testing::InitGoogleTest(&argc, argv);/g' $file perl -pi -e 's/rocksdb::test::RunAllTests/RUN_ALL_TESTS/g' $file fi done % sh ~/transform % make format ``` Second iteration of this diff contains only scripted changes. Third iteration contains manual changes to fix last errors and make it compilable. Test Plan: Build and notice no errors. ```lang=bash % USE_CLANG=1 make check -j55 ``` Tests are still testing. Reviewers: meyering, sdong, rven, igor Reviewed By: igor Subscribers: dhruba, leveldb Differential Revision: https://reviews.facebook.net/D35157
2015-03-17 22:08:00 +01:00
TEST_F(CacheTest, OverCapacity) {
size_t n = 10;
// a LRUCache with n entries and one shard only
std::shared_ptr<Cache> cache = NewLRUCache(n, 0);
std::vector<Cache::Handle*> handles(n+1);
// Insert n+1 entries, but not releasing.
for (size_t i = 0; i < n + 1; i++) {
std::string key = ToString(i+1);
handles[i] = cache->Insert(key, new Value(i+1), 1, &deleter);
}
// Guess what's in the cache now?
for (size_t i = 0; i < n + 1; i++) {
std::string key = ToString(i+1);
auto h = cache->Lookup(key);
std::cout << key << (h?" found\n":" not found\n");
Modifed the LRU cache eviction code so that it doesn't evict blocks which have exteranl references Summary: Currently, blocks which have more than one reference (ie referenced by something other than cache itself) are evicted from cache. This doesn't make much sense: - blocks are still in RAM, so the RAM usage reported by the cache is incorrect - if the same block is needed by another iterator, it will be loaded and decompressed again This diff changes the reference counting scheme a bit. Previously, if the cache contained the block, this was accounted for in its refcount. After this change, the refcount is only used to track external references. There is a boolean flag which indicates whether or not the block is contained in the cache. This diff also changes how LRU list is used. Previously, both hashtable and the LRU list contained all blocks. After this change, the LRU list contains blocks with the refcount==0, ie those which can be evicted from the cache. Note that this change still allows for cache to grow beyond its capacity. This happens when all blocks are pinned (ie refcount>0). This is consistent with the current behavior. The cache's insert function never fails. I spent lots of time trying to make table_reader and other places work with the insert which might failed. It turned out to be pretty hard. It might really destabilize some customers, so finally, I decided against doing this. table_cache_remove_scan_count_limit option will be unneeded after this change, but I will remove it in the following diff, if this one gets approved Test Plan: Ran tests, made sure they pass Reviewers: sdong, ljin Differential Revision: https://reviews.facebook.net/D25503
2014-10-21 20:49:13 +02:00
ASSERT_TRUE(h != nullptr);
if (h) cache->Release(h);
}
Modifed the LRU cache eviction code so that it doesn't evict blocks which have exteranl references Summary: Currently, blocks which have more than one reference (ie referenced by something other than cache itself) are evicted from cache. This doesn't make much sense: - blocks are still in RAM, so the RAM usage reported by the cache is incorrect - if the same block is needed by another iterator, it will be loaded and decompressed again This diff changes the reference counting scheme a bit. Previously, if the cache contained the block, this was accounted for in its refcount. After this change, the refcount is only used to track external references. There is a boolean flag which indicates whether or not the block is contained in the cache. This diff also changes how LRU list is used. Previously, both hashtable and the LRU list contained all blocks. After this change, the LRU list contains blocks with the refcount==0, ie those which can be evicted from the cache. Note that this change still allows for cache to grow beyond its capacity. This happens when all blocks are pinned (ie refcount>0). This is consistent with the current behavior. The cache's insert function never fails. I spent lots of time trying to make table_reader and other places work with the insert which might failed. It turned out to be pretty hard. It might really destabilize some customers, so finally, I decided against doing this. table_cache_remove_scan_count_limit option will be unneeded after this change, but I will remove it in the following diff, if this one gets approved Test Plan: Ran tests, made sure they pass Reviewers: sdong, ljin Differential Revision: https://reviews.facebook.net/D25503
2014-10-21 20:49:13 +02:00
// the cache is over capacity since nothing could be evicted
ASSERT_EQ(n + 1U, cache->GetUsage());
for (size_t i = 0; i < n + 1; i++) {
cache->Release(handles[i]);
}
Modifed the LRU cache eviction code so that it doesn't evict blocks which have exteranl references Summary: Currently, blocks which have more than one reference (ie referenced by something other than cache itself) are evicted from cache. This doesn't make much sense: - blocks are still in RAM, so the RAM usage reported by the cache is incorrect - if the same block is needed by another iterator, it will be loaded and decompressed again This diff changes the reference counting scheme a bit. Previously, if the cache contained the block, this was accounted for in its refcount. After this change, the refcount is only used to track external references. There is a boolean flag which indicates whether or not the block is contained in the cache. This diff also changes how LRU list is used. Previously, both hashtable and the LRU list contained all blocks. After this change, the LRU list contains blocks with the refcount==0, ie those which can be evicted from the cache. Note that this change still allows for cache to grow beyond its capacity. This happens when all blocks are pinned (ie refcount>0). This is consistent with the current behavior. The cache's insert function never fails. I spent lots of time trying to make table_reader and other places work with the insert which might failed. It turned out to be pretty hard. It might really destabilize some customers, so finally, I decided against doing this. table_cache_remove_scan_count_limit option will be unneeded after this change, but I will remove it in the following diff, if this one gets approved Test Plan: Ran tests, made sure they pass Reviewers: sdong, ljin Differential Revision: https://reviews.facebook.net/D25503
2014-10-21 20:49:13 +02:00
// cache is under capacity now since elements were released
ASSERT_EQ(n, cache->GetUsage());
// element 0 is evicted and the rest is there
// This is consistent with the LRU policy since the element 0
// was released first
for (size_t i = 0; i < n + 1; i++) {
Modifed the LRU cache eviction code so that it doesn't evict blocks which have exteranl references Summary: Currently, blocks which have more than one reference (ie referenced by something other than cache itself) are evicted from cache. This doesn't make much sense: - blocks are still in RAM, so the RAM usage reported by the cache is incorrect - if the same block is needed by another iterator, it will be loaded and decompressed again This diff changes the reference counting scheme a bit. Previously, if the cache contained the block, this was accounted for in its refcount. After this change, the refcount is only used to track external references. There is a boolean flag which indicates whether or not the block is contained in the cache. This diff also changes how LRU list is used. Previously, both hashtable and the LRU list contained all blocks. After this change, the LRU list contains blocks with the refcount==0, ie those which can be evicted from the cache. Note that this change still allows for cache to grow beyond its capacity. This happens when all blocks are pinned (ie refcount>0). This is consistent with the current behavior. The cache's insert function never fails. I spent lots of time trying to make table_reader and other places work with the insert which might failed. It turned out to be pretty hard. It might really destabilize some customers, so finally, I decided against doing this. table_cache_remove_scan_count_limit option will be unneeded after this change, but I will remove it in the following diff, if this one gets approved Test Plan: Ran tests, made sure they pass Reviewers: sdong, ljin Differential Revision: https://reviews.facebook.net/D25503
2014-10-21 20:49:13 +02:00
std::string key = ToString(i+1);
auto h = cache->Lookup(key);
if (h) {
ASSERT_NE(i, 0U);
Modifed the LRU cache eviction code so that it doesn't evict blocks which have exteranl references Summary: Currently, blocks which have more than one reference (ie referenced by something other than cache itself) are evicted from cache. This doesn't make much sense: - blocks are still in RAM, so the RAM usage reported by the cache is incorrect - if the same block is needed by another iterator, it will be loaded and decompressed again This diff changes the reference counting scheme a bit. Previously, if the cache contained the block, this was accounted for in its refcount. After this change, the refcount is only used to track external references. There is a boolean flag which indicates whether or not the block is contained in the cache. This diff also changes how LRU list is used. Previously, both hashtable and the LRU list contained all blocks. After this change, the LRU list contains blocks with the refcount==0, ie those which can be evicted from the cache. Note that this change still allows for cache to grow beyond its capacity. This happens when all blocks are pinned (ie refcount>0). This is consistent with the current behavior. The cache's insert function never fails. I spent lots of time trying to make table_reader and other places work with the insert which might failed. It turned out to be pretty hard. It might really destabilize some customers, so finally, I decided against doing this. table_cache_remove_scan_count_limit option will be unneeded after this change, but I will remove it in the following diff, if this one gets approved Test Plan: Ran tests, made sure they pass Reviewers: sdong, ljin Differential Revision: https://reviews.facebook.net/D25503
2014-10-21 20:49:13 +02:00
cache->Release(h);
} else {
ASSERT_EQ(i, 0U);
Modifed the LRU cache eviction code so that it doesn't evict blocks which have exteranl references Summary: Currently, blocks which have more than one reference (ie referenced by something other than cache itself) are evicted from cache. This doesn't make much sense: - blocks are still in RAM, so the RAM usage reported by the cache is incorrect - if the same block is needed by another iterator, it will be loaded and decompressed again This diff changes the reference counting scheme a bit. Previously, if the cache contained the block, this was accounted for in its refcount. After this change, the refcount is only used to track external references. There is a boolean flag which indicates whether or not the block is contained in the cache. This diff also changes how LRU list is used. Previously, both hashtable and the LRU list contained all blocks. After this change, the LRU list contains blocks with the refcount==0, ie those which can be evicted from the cache. Note that this change still allows for cache to grow beyond its capacity. This happens when all blocks are pinned (ie refcount>0). This is consistent with the current behavior. The cache's insert function never fails. I spent lots of time trying to make table_reader and other places work with the insert which might failed. It turned out to be pretty hard. It might really destabilize some customers, so finally, I decided against doing this. table_cache_remove_scan_count_limit option will be unneeded after this change, but I will remove it in the following diff, if this one gets approved Test Plan: Ran tests, made sure they pass Reviewers: sdong, ljin Differential Revision: https://reviews.facebook.net/D25503
2014-10-21 20:49:13 +02:00
}
}
}
namespace {
std::vector<std::pair<int, int>> callback_state;
void callback(void* entry, size_t charge) {
callback_state.push_back({DecodeValue(entry), static_cast<int>(charge)});
}
};
rocksdb: switch to gtest Summary: Our existing test notation is very similar to what is used in gtest. It makes it easy to adopt what is different. In this diff I modify existing [[ https://code.google.com/p/googletest/wiki/Primer#Test_Fixtures:_Using_the_Same_Data_Configuration_for_Multiple_Te | test fixture ]] classes to inherit from `testing::Test`. Also for unit tests that use fixture class, `TEST` is replaced with `TEST_F` as required in gtest. There are several custom `main` functions in our existing tests. To make this transition easier, I modify all `main` functions to fallow gtest notation. But eventually we can remove them and use implementation of `main` that gtest provides. ```lang=bash % cat ~/transform #!/bin/sh files=$(git ls-files '*test\.cc') for file in $files do if grep -q "rocksdb::test::RunAllTests()" $file then if grep -Eq '^class \w+Test {' $file then perl -pi -e 's/^(class \w+Test) {/${1}: public testing::Test {/g' $file perl -pi -e 's/^(TEST)/${1}_F/g' $file fi perl -pi -e 's/(int main.*\{)/${1}::testing::InitGoogleTest(&argc, argv);/g' $file perl -pi -e 's/rocksdb::test::RunAllTests/RUN_ALL_TESTS/g' $file fi done % sh ~/transform % make format ``` Second iteration of this diff contains only scripted changes. Third iteration contains manual changes to fix last errors and make it compilable. Test Plan: Build and notice no errors. ```lang=bash % USE_CLANG=1 make check -j55 ``` Tests are still testing. Reviewers: meyering, sdong, rven, igor Reviewed By: igor Subscribers: dhruba, leveldb Differential Revision: https://reviews.facebook.net/D35157
2015-03-17 22:08:00 +01:00
TEST_F(CacheTest, ApplyToAllCacheEntiresTest) {
std::vector<std::pair<int, int>> inserted;
callback_state.clear();
for (int i = 0; i < 10; ++i) {
Insert(i, i * 2, i + 1);
inserted.push_back({i * 2, i + 1});
}
cache_->ApplyToAllCacheEntries(callback, true);
sort(inserted.begin(), inserted.end());
sort(callback_state.begin(), callback_state.end());
ASSERT_TRUE(inserted == callback_state);
}
} // namespace rocksdb
int main(int argc, char** argv) {
rocksdb: switch to gtest Summary: Our existing test notation is very similar to what is used in gtest. It makes it easy to adopt what is different. In this diff I modify existing [[ https://code.google.com/p/googletest/wiki/Primer#Test_Fixtures:_Using_the_Same_Data_Configuration_for_Multiple_Te | test fixture ]] classes to inherit from `testing::Test`. Also for unit tests that use fixture class, `TEST` is replaced with `TEST_F` as required in gtest. There are several custom `main` functions in our existing tests. To make this transition easier, I modify all `main` functions to fallow gtest notation. But eventually we can remove them and use implementation of `main` that gtest provides. ```lang=bash % cat ~/transform #!/bin/sh files=$(git ls-files '*test\.cc') for file in $files do if grep -q "rocksdb::test::RunAllTests()" $file then if grep -Eq '^class \w+Test {' $file then perl -pi -e 's/^(class \w+Test) {/${1}: public testing::Test {/g' $file perl -pi -e 's/^(TEST)/${1}_F/g' $file fi perl -pi -e 's/(int main.*\{)/${1}::testing::InitGoogleTest(&argc, argv);/g' $file perl -pi -e 's/rocksdb::test::RunAllTests/RUN_ALL_TESTS/g' $file fi done % sh ~/transform % make format ``` Second iteration of this diff contains only scripted changes. Third iteration contains manual changes to fix last errors and make it compilable. Test Plan: Build and notice no errors. ```lang=bash % USE_CLANG=1 make check -j55 ``` Tests are still testing. Reviewers: meyering, sdong, rven, igor Reviewed By: igor Subscribers: dhruba, leveldb Differential Revision: https://reviews.facebook.net/D35157
2015-03-17 22:08:00 +01:00
::testing::InitGoogleTest(&argc, argv);
return RUN_ALL_TESTS();
}