andreacavalli/rocksdb
0
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Merge branch 'master' into fix-range-deletion-bug

Browse Source
This commit is contained in:
奏之章 2019-12-09 22:02:10 +08:00 committed by GitHub
commit 5c02676461
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
36 changed files with 5672 additions and 4949 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

61
CMakeLists.txt
View File

@ -14,7 +14,7 @@
# cd build
# 3. Run cmake to generate project files for Windows, add more options to enable required third-party libraries.
# See thirdparty.inc for more information.
# sample command: cmake -G "Visual Studio 15 Win64" -DWITH_GFLAGS=1 -DWITH_SNAPPY=1 -DWITH_JEMALLOC=1 -DWITH_JNI=1 ..
# sample command: cmake -G "Visual Studio 15 Win64" -DCMAKE_BUILD_TYPE=Release -DWITH_GFLAGS=1 -DWITH_SNAPPY=1 -DWITH_JEMALLOC=1 -DWITH_JNI=1 ..
# 4. Then build the project in debug mode (you may want to add /m[:<N>] flag to run msbuild in <N> parallel threads
# or simply /m to use all avail cores)
# msbuild rocksdb.sln
@ -174,7 +174,7 @@ else()
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -W -Wextra -Wall")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wsign-compare -Wshadow -Wno-unused-parameter -Wno-unused-variable -Woverloaded-virtual -Wnon-virtual-dtor -Wno-missing-field-initializers -Wno-strict-aliasing")
if(MINGW)
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-format")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-format -fno-asynchronous-unwind-tables")
add_definitions(-D_POSIX_C_SOURCE=1)
endif()
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11")
@ -631,7 +631,6 @@ set(SOURCES
test_util/testutil.cc
test_util/transaction_test_util.cc
tools/block_cache_analyzer/block_cache_trace_analyzer.cc
tools/db_bench_tool.cc
tools/dump/db_dump_tool.cc
tools/ldb_cmd.cc
tools/ldb_tool.cc
@ -885,7 +884,9 @@ if(NOT WIN32 OR ROCKSDB_INSTALL_ON_WINDOWS)
endif()
option(WITH_TESTS "build with tests" ON)
if(WITH_TESTS)
# For test libraries, utilities, and exes that are build iff WITH_TESTS=ON and
# in Debug mode. Add test only code that is not #ifdefed for Release here.
if(WITH_TESTS AND CMAKE_BUILD_TYPE STREQUAL "Debug")
add_subdirectory(third-party/gtest-1.8.1/fused-src/gtest)
set(TESTS
cache/cache_test.cc
@ -1040,24 +1041,6 @@ if(WITH_TESTS)
list(APPEND TESTS third-party/folly/folly/synchronization/test/DistributedMutexTest.cpp)
endif()
set(BENCHMARKS
cache/cache_bench.cc
memtable/memtablerep_bench.cc
db/range_del_aggregator_bench.cc
tools/db_bench.cc
table/table_reader_bench.cc
util/filter_bench.cc
utilities/persistent_cache/hash_table_bench.cc)
add_library(testharness OBJECT test_util/testharness.cc)
foreach(sourcefile ${BENCHMARKS})
get_filename_component(exename ${sourcefile} NAME_WE)
add_executable(${exename}${ARTIFACT_SUFFIX} ${sourcefile}
$<TARGET_OBJECTS:testharness>)
target_link_libraries(${exename}${ARTIFACT_SUFFIX} gtest ${LIBS})
endforeach(sourcefile ${BENCHMARKS})
# For test util library that is build only in DEBUG mode
# and linked to tests. Add test only code that is not #ifdefed for Release here.
set(TESTUTIL_SOURCE
db/db_test_util.cc
monitoring/thread_status_updater_debug.cc
@ -1065,7 +1048,6 @@ if(WITH_TESTS)
test_util/fault_injection_test_env.cc
utilities/cassandra/test_utils.cc
)
# test utilities are only build in debug
enable_testing()
add_custom_target(check COMMAND ${CMAKE_CTEST_COMMAND})
set(TESTUTILLIB testutillib${ARTIFACT_SUFFIX})
@ -1080,9 +1062,7 @@ if(WITH_TESTS)
EXCLUDE_FROM_DEFAULT_BUILD_RELWITHDEBINFO 1
)
# Tests are excluded from Release builds
set(TEST_EXES ${TESTS})
foreach(sourcefile ${TEST_EXES})
get_filename_component(exename ${sourcefile} NAME_WE)
add_executable(${CMAKE_PROJECT_NAME}_${exename}${ARTIFACT_SUFFIX} ${sourcefile}
@ -1119,7 +1099,38 @@ if(WITH_TESTS)
endforeach(sourcefile ${C_TEST_EXES})
endif()
option(WITH_BENCHMARK_TOOLS "build with benchmarks" ON)
if(WITH_BENCHMARK_TOOLS)
if(NOT TARGET gtest)
add_subdirectory(third-party/gtest-1.8.1/fused-src/gtest)
endif()
set(BENCHMARKS
cache/cache_bench.cc
memtable/memtablerep_bench.cc
db/range_del_aggregator_bench.cc
table/table_reader_bench.cc
util/filter_bench.cc
utilities/persistent_cache/hash_table_bench.cc
)
add_library(testharness OBJECT test_util/testharness.cc)
foreach(sourcefile ${BENCHMARKS})
get_filename_component(exename ${sourcefile} NAME_WE)
add_executable(${exename}${ARTIFACT_SUFFIX} ${sourcefile}
$<TARGET_OBJECTS:testharness>)
target_link_libraries(${exename}${ARTIFACT_SUFFIX} gtest ${LIBS})
endforeach(sourcefile ${BENCHMARKS})
add_executable(db_bench${ARTIFACT_SUFFIX}
tools/db_bench.cc
tools/db_bench_tool.cc
$<TARGET_OBJECTS:testharness>)
target_link_libraries(db_bench${ARTIFACT_SUFFIX} gtest ${LIBS})
endif()
option(WITH_TOOLS "build with tools" ON)
if(WITH_TOOLS)
add_subdirectory(tools)
add_subdirectory(db_stress_tool)
add_custom_target(tools
DEPENDS ${tool_deps})
endif()

2
Makefile
View File

@ -1183,7 +1183,7 @@ memtablerep_bench: memtable/memtablerep_bench.o $(LIBOBJECTS) $(TESTUTIL)
filter_bench: util/filter_bench.o $(LIBOBJECTS) $(TESTUTIL)
$(AM_LINK)
db_stress: tools/db_stress.o $(STRESSTOOLOBJECTS)
db_stress: db_stress_tool/db_stress.o $(STRESSTOOLOBJECTS)
$(AM_LINK)
write_stress: tools/write_stress.o $(LIBOBJECTS) $(TESTUTIL)

13
TARGETS
View File

@ -26,6 +26,7 @@ ROCKSDB_EXTERNAL_DEPS = [
("lz4", None, "lz4"),
("zstd", None),
("tbb", None),
("liburing", None, "uring"),
("googletest", None, "gtest"),
]
@ -46,6 +47,7 @@ ROCKSDB_OS_PREPROCESSOR_FLAGS = [
"-DROCKSDB_PTHREAD_ADAPTIVE_MUTEX",
"-DROCKSDB_RANGESYNC_PRESENT",
"-DROCKSDB_SCHED_GETCPU_PRESENT",
"-DROCKSDB_IOURING_PRESENT",
"-DHAVE_SSE42",
"-DNUMA",
],
@ -70,6 +72,7 @@ ROCKSDB_PREPROCESSOR_FLAGS = [
"-DZSTD_STATIC_LINKING_ONLY",
"-DGFLAGS=gflags",
"-DTBB",
"-DLIBURING",
# Added missing flags from output of build_detect_platform
"-DROCKSDB_BACKTRACE",
@ -399,9 +402,17 @@ cpp_library(
cpp_library(
name = "rocksdb_stress_lib",
srcs = [
"db_stress_tool/batched_ops_stress.cc",
"db_stress_tool/cf_consistency_stress.cc",
"db_stress_tool/db_stress_common.cc",
"db_stress_tool/db_stress_driver.cc",
"db_stress_tool/db_stress_gflags.cc",
"db_stress_tool/db_stress_shared_state.cc",
"db_stress_tool/db_stress_test_base.cc",
"db_stress_tool/db_stress_tool.cc",
"db_stress_tool/no_batched_ops_stress.cc",
"test_util/testutil.cc",
"tools/block_cache_analyzer/block_cache_trace_analyzer.cc",
"tools/db_stress_tool.cc",
"tools/trace_analyzer_tool.cc",
],
auto_headers = AutoHeaders.RECURSIVE_GLOB,

2
appveyor.yml
View File

@ -50,7 +50,7 @@ install:
before_build:
- md %APPVEYOR_BUILD_FOLDER%\build
- cd %APPVEYOR_BUILD_FOLDER%\build
- cmake -G "Visual Studio 15 Win64" -DOPTDBG=1 -DPORTABLE=1 -DSNAPPY=1 -DLZ4=1 -DZSTD=1 -DXPRESS=1 -DJNI=1 ..
- cmake -G "Visual Studio 15 Win64" -DCMAKE_BUILD_TYPE=Debug -DOPTDBG=1 -DPORTABLE=1 -DSNAPPY=1 -DLZ4=1 -DZSTD=1 -DXPRESS=1 -DJNI=1 ..
- cd ..
build:
project: build\rocksdb.sln

3
buckifier/targets_cfg.py
View File

@ -32,6 +32,7 @@ ROCKSDB_EXTERNAL_DEPS = [
("lz4", None, "lz4"),
("zstd", None),
("tbb", None),
("liburing", None, "uring"),
("googletest", None, "gtest"),
]
@ -52,6 +53,7 @@ ROCKSDB_OS_PREPROCESSOR_FLAGS = [
"-DROCKSDB_PTHREAD_ADAPTIVE_MUTEX",
"-DROCKSDB_RANGESYNC_PRESENT",
"-DROCKSDB_SCHED_GETCPU_PRESENT",
"-DROCKSDB_IOURING_PRESENT",
"-DHAVE_SSE42",
"-DNUMA",
],
@ -76,6 +78,7 @@ ROCKSDB_PREPROCESSOR_FLAGS = [
"-DZSTD_STATIC_LINKING_ONLY",
"-DGFLAGS=gflags",
"-DTBB",
"-DLIBURING",
# Added missing flags from output of build_detect_platform
"-DROCKSDB_BACKTRACE",

13
build_tools/build_detect_platform
View File

@ -150,6 +150,19 @@ case "$TARGET_OS" in
PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -latomic"
fi
PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -lpthread -lrt"
# check for liburing
$CXX $CFLAGS -x c++ - -luring -o /dev/null 2>/dev/null <<EOF
#include <liburing.h>
int main() {
struct io_uring ring;
io_uring_queue_init(1, &ring, 0);
return 0;
}
EOF
if [ "$?" = 0 ]; then
PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -luring"
COMMON_FLAGS="$COMMON_FLAGS -DROCKSDB_IOURING_PRESENT"
fi
if test -z "$USE_FOLLY_DISTRIBUTED_MUTEX"; then
USE_FOLLY_DISTRIBUTED_MUTEX=1
fi

1
build_tools/dependencies_platform007.sh
View File

@ -13,6 +13,7 @@ JEMALLOC_BASE=/mnt/gvfs/third-party2/jemalloc/c26f08f47ac35fc31da2633b7da92d6b86
NUMA_BASE=/mnt/gvfs/third-party2/numa/3f3fb57a5ccc5fd21c66416c0b83e0aa76a05376/2.0.11/platform007/ca4da3d
LIBUNWIND_BASE=/mnt/gvfs/third-party2/libunwind/40c73d874898b386a71847f1b99115d93822d11f/1.4/platform007/6f3e0a9
TBB_BASE=/mnt/gvfs/third-party2/tbb/4ce8e8dba77cdbd81b75d6f0c32fd7a1b76a11ec/2018_U5/platform007/ca4da3d
LIBURING_BASE=/mnt/gvfs/third-party2/liburing/79427253fd0d42677255aacfe6d13bfe63f752eb/20190828/platform007/ca4da3d
KERNEL_HEADERS_BASE=/mnt/gvfs/third-party2/kernel-headers/fb251ecd2f5ae16f8671f7014c246e52a748fe0b/fb/platform007/da39a3e
BINUTILS_BASE=/mnt/gvfs/third-party2/binutils/ab9f09bba370e7066cafd4eb59752db93f2e8312/2.29.1/platform007/15a3614
VALGRIND_BASE=/mnt/gvfs/third-party2/valgrind/d42d152a15636529b0861ec493927200ebebca8e/3.15.0/platform007/ca4da3d

17
build_tools/fbcode_config_platform007.sh
View File

@ -86,6 +86,15 @@ else
fi
CFLAGS+=" -DTBB"
# location of LIBURING
LIBURING_INCLUDE=" -isystem $LIBURING_BASE/include/"
if test -z $PIC_BUILD; then
LIBURING_LIBS="$LIBURING_BASE/lib/liburing.a"
else
LIBURING_LIBS="$LIBURING_BASE/lib/liburing_pic.a"
fi
CFLAGS+=" -DLIBURING"
test "$USE_SSE" || USE_SSE=1
export USE_SSE
test "$PORTABLE" || PORTABLE=1
@ -94,7 +103,7 @@ export PORTABLE
BINUTILS="$BINUTILS_BASE/bin"
AR="$BINUTILS/ar"
DEPS_INCLUDE="$SNAPPY_INCLUDE $ZLIB_INCLUDE $BZIP_INCLUDE $LZ4_INCLUDE $ZSTD_INCLUDE $GFLAGS_INCLUDE $NUMA_INCLUDE $TBB_INCLUDE"
DEPS_INCLUDE="$SNAPPY_INCLUDE $ZLIB_INCLUDE $BZIP_INCLUDE $LZ4_INCLUDE $ZSTD_INCLUDE $GFLAGS_INCLUDE $NUMA_INCLUDE $TBB_INCLUDE $LIBURING_INCLUDE"
STDLIBS="-L $GCC_BASE/lib64"
@ -135,10 +144,10 @@ else
fi
CFLAGS+=" $DEPS_INCLUDE"
CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX -DROCKSDB_FALLOCATE_PRESENT -DROCKSDB_MALLOC_USABLE_SIZE -DROCKSDB_RANGESYNC_PRESENT -DROCKSDB_SCHED_GETCPU_PRESENT -DROCKSDB_SUPPORT_THREAD_LOCAL -DHAVE_SSE42"
CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX -DROCKSDB_FALLOCATE_PRESENT -DROCKSDB_MALLOC_USABLE_SIZE -DROCKSDB_RANGESYNC_PRESENT -DROCKSDB_SCHED_GETCPU_PRESENT -DROCKSDB_SUPPORT_THREAD_LOCAL -DHAVE_SSE42 -DROCKSDB_IOURING_PRESENT"
CXXFLAGS+=" $CFLAGS"
EXEC_LDFLAGS=" $SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $NUMA_LIB $TBB_LIBS"
EXEC_LDFLAGS=" $SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $NUMA_LIB $TBB_LIBS $LIBURING_LIBS"
EXEC_LDFLAGS+=" -B$BINUTILS/gold"
EXEC_LDFLAGS+=" -Wl,--dynamic-linker,/usr/local/fbcode/platform007/lib/ld.so"
EXEC_LDFLAGS+=" $LIBUNWIND"
@ -148,7 +157,7 @@ EXEC_LDFLAGS+=" -ldl"
PLATFORM_LDFLAGS="$LIBGCC_LIBS $GLIBC_LIBS $STDLIBS -lgcc -lstdc++"
EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $TBB_LIBS"
EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $TBB_LIBS $LIBURING_LIBS"
VALGRIND_VER="$VALGRIND_BASE/bin/"

2
build_tools/precommit_checker.py
View File

@ -1,4 +1,4 @@
#!/usr/local/fbcode/gcc-4.9-glibc-2.20-fb/bin/python2.7
#!/usr/bin/env python2.7
# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
from __future__ import absolute_import

1
build_tools/update_dependencies.sh
View File

@ -92,6 +92,7 @@ get_lib_base jemalloc LATEST platform007
get_lib_base numa LATEST platform007
get_lib_base libunwind LATEST platform007
get_lib_base tbb LATEST platform007
get_lib_base liburing LATEST platform007
get_lib_base kernel-headers fb platform007
get_lib_base binutils LATEST centos7-native

2
db/db_impl/db_secondary_test.cc
View File

@ -198,7 +198,7 @@ TEST_F(DBSecondaryTest, OpenAsSecondary) {
namespace {
class TraceFileEnv : public EnvWrapper {
public:
explicit TraceFileEnv(Env* target) : EnvWrapper(target) {}
explicit TraceFileEnv(Env* _target) : EnvWrapper(_target) {}
Status NewRandomAccessFile(const std::string& f,
std::unique_ptr<RandomAccessFile>* r,
const EnvOptions& env_options) override {

14
db_stress_tool/CMakeLists.txt Normal file
View File

@ -0,0 +1,14 @@
add_executable(db_stress${ARTIFACT_SUFFIX}
db_stress.cc
db_stress_tool.cc
batched_ops_stress.cc
cf_consistency_stress.cc
db_stress_common.cc
db_stress_driver.cc
db_stress_test_base.cc
db_stress_shared_state.cc
db_stress_gflags.cc
db_stress_tool.cc
no_batched_ops_stress.cc)
target_link_libraries(db_stress${ARTIFACT_SUFFIX} ${LIBS})
list(APPEND tool_deps db_stress)

342
db_stress_tool/batched_ops_stress.cc Normal file
View File

@ -0,0 +1,342 @@
// 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.
#ifdef GFLAGS
#include "db_stress_tool/db_stress_common.h"
namespace rocksdb {
class BatchedOpsStressTest : public StressTest {
public:
BatchedOpsStressTest() {}
virtual ~BatchedOpsStressTest() {}
// Given a key K and value V, this puts ("0"+K, "0"+V), ("1"+K, "1"+V), ...
// ("9"+K, "9"+V) in DB atomically i.e in a single batch.
// Also refer BatchedOpsStressTest::TestGet
virtual Status TestPut(ThreadState* thread, WriteOptions& write_opts,
const ReadOptions& /* read_opts */,
const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys,
char (&value)[100],
std::unique_ptr<MutexLock>& /* lock */) {
uint32_t value_base =
thread->rand.Next() % thread->shared->UNKNOWN_SENTINEL;
size_t sz = GenerateValue(value_base, value, sizeof(value));
Slice v(value, sz);
std::string keys[10] = {"9", "8", "7", "6", "5", "4", "3", "2", "1", "0"};
std::string values[10] = {"9", "8", "7", "6", "5", "4", "3", "2", "1", "0"};
Slice value_slices[10];
WriteBatch batch;
Status s;
auto cfh = column_families_[rand_column_families[0]];
std::string key_str = Key(rand_keys[0]);
for (int i = 0; i < 10; i++) {
keys[i] += key_str;
values[i] += v.ToString();
value_slices[i] = values[i];
if (FLAGS_use_merge) {
batch.Merge(cfh, keys[i], value_slices[i]);
} else {
batch.Put(cfh, keys[i], value_slices[i]);
}
}
s = db_->Write(write_opts, &batch);
if (!s.ok()) {
fprintf(stderr, "multiput error: %s\n", s.ToString().c_str());
thread->stats.AddErrors(1);
} else {
// we did 10 writes each of size sz + 1
thread->stats.AddBytesForWrites(10, (sz + 1) * 10);
}
return s;
}
// Given a key K, this deletes ("0"+K), ("1"+K),... ("9"+K)
// in DB atomically i.e in a single batch. Also refer MultiGet.
virtual Status TestDelete(ThreadState* thread, WriteOptions& writeoptions,
const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys,
std::unique_ptr<MutexLock>& /* lock */) {
std::string keys[10] = {"9", "7", "5", "3", "1", "8", "6", "4", "2", "0"};
WriteBatch batch;
Status s;
auto cfh = column_families_[rand_column_families[0]];
std::string key_str = Key(rand_keys[0]);
for (int i = 0; i < 10; i++) {
keys[i] += key_str;
batch.Delete(cfh, keys[i]);
}
s = db_->Write(writeoptions, &batch);
if (!s.ok()) {
fprintf(stderr, "multidelete error: %s\n", s.ToString().c_str());
thread->stats.AddErrors(1);
} else {
thread->stats.AddDeletes(10);
}
return s;
}
virtual Status TestDeleteRange(
ThreadState* /* thread */, WriteOptions& /* write_opts */,
const std::vector<int>& /* rand_column_families */,
const std::vector<int64_t>& /* rand_keys */,
std::unique_ptr<MutexLock>& /* lock */) {
assert(false);
return Status::NotSupported(
"BatchedOpsStressTest does not support "
"TestDeleteRange");
}
virtual void TestIngestExternalFile(
ThreadState* /* thread */,
const std::vector<int>& /* rand_column_families */,
const std::vector<int64_t>& /* rand_keys */,
std::unique_ptr<MutexLock>& /* lock */) {
assert(false);
fprintf(stderr,
"BatchedOpsStressTest does not support "
"TestIngestExternalFile\n");
std::terminate();
}
// Given a key K, this gets values for "0"+K, "1"+K,..."9"+K
// in the same snapshot, and verifies that all the values are of the form
// "0"+V, "1"+V,..."9"+V.
// ASSUMES that BatchedOpsStressTest::TestPut was used to put (K, V) into
// the DB.
virtual Status TestGet(ThreadState* thread, const ReadOptions& readoptions,
const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys) {
std::string keys[10] = {"0", "1", "2", "3", "4", "5", "6", "7", "8", "9"};
Slice key_slices[10];
std::string values[10];
ReadOptions readoptionscopy = readoptions;
readoptionscopy.snapshot = db_->GetSnapshot();
std::string key_str = Key(rand_keys[0]);
Slice key = key_str;
auto cfh = column_families_[rand_column_families[0]];
std::string from_db;
Status s;
for (int i = 0; i < 10; i++) {
keys[i] += key.ToString();
key_slices[i] = keys[i];
s = db_->Get(readoptionscopy, cfh, key_slices[i], &from_db);
if (!s.ok() && !s.IsNotFound()) {
fprintf(stderr, "get error: %s\n", s.ToString().c_str());
values[i] = "";
thread->stats.AddErrors(1);
// we continue after error rather than exiting so that we can
// find more errors if any
} else if (s.IsNotFound()) {
values[i] = "";
thread->stats.AddGets(1, 0);
} else {
values[i] = from_db;
char expected_prefix = (keys[i])[0];
char actual_prefix = (values[i])[0];
if (actual_prefix != expected_prefix) {
fprintf(stderr, "error expected prefix = %c actual = %c\n",
expected_prefix, actual_prefix);
}
(values[i])[0] = ' '; // blank out the differing character
thread->stats.AddGets(1, 1);
}
}
db_->ReleaseSnapshot(readoptionscopy.snapshot);
// Now that we retrieved all values, check that they all match
for (int i = 1; i < 10; i++) {
if (values[i] != values[0]) {
fprintf(stderr, "error : inconsistent values for key %s: %s, %s\n",
key.ToString(true).c_str(), StringToHex(values[0]).c_str(),
StringToHex(values[i]).c_str());
// we continue after error rather than exiting so that we can
// find more errors if any
}
}
return s;
}
virtual std::vector<Status> TestMultiGet(
ThreadState* thread, const ReadOptions& readoptions,
const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys) {
size_t num_keys = rand_keys.size();
std::vector<Status> ret_status(num_keys);
std::array<std::string, 10> keys = {"0", "1", "2", "3", "4",
"5", "6", "7", "8", "9"};
size_t num_prefixes = keys.size();
for (size_t rand_key = 0; rand_key < num_keys; ++rand_key) {
std::vector<Slice> key_slices;
std::vector<PinnableSlice> values(num_prefixes);
std::vector<Status> statuses(num_prefixes);
ReadOptions readoptionscopy = readoptions;
readoptionscopy.snapshot = db_->GetSnapshot();
std::vector<std::string> key_str;
key_str.reserve(num_prefixes);
key_slices.reserve(num_prefixes);
std::string from_db;
ColumnFamilyHandle* cfh = column_families_[rand_column_families[0]];
for (size_t key = 0; key < num_prefixes; ++key) {
key_str.emplace_back(keys[key] + Key(rand_keys[rand_key]));
key_slices.emplace_back(key_str.back());
}
db_->MultiGet(readoptionscopy, cfh, num_prefixes, key_slices.data(),
values.data(), statuses.data());
for (size_t i = 0; i < num_prefixes; i++) {
Status s = statuses[i];
if (!s.ok() && !s.IsNotFound()) {
fprintf(stderr, "get error: %s\n", s.ToString().c_str());
thread->stats.AddErrors(1);
ret_status[rand_key] = s;
// we continue after error rather than exiting so that we can
// find more errors if any
} else if (s.IsNotFound()) {
thread->stats.AddGets(1, 0);
ret_status[rand_key] = s;
} else {
char expected_prefix = (keys[i])[0];
char actual_prefix = (values[i])[0];
if (actual_prefix != expected_prefix) {
fprintf(stderr, "error expected prefix = %c actual = %c\n",
expected_prefix, actual_prefix);
}
std::string str;
str.assign(values[i].data(), values[i].size());
values[i].Reset();
str[0] = ' '; // blank out the differing character
values[i].PinSelf(str);
thread->stats.AddGets(1, 1);
}
}
db_->ReleaseSnapshot(readoptionscopy.snapshot);
// Now that we retrieved all values, check that they all match
for (size_t i = 1; i < num_prefixes; i++) {
if (values[i] != values[0]) {
fprintf(stderr, "error : inconsistent values for key %s: %s, %s\n",
key_str[i].c_str(), StringToHex(values[0].ToString()).c_str(),
StringToHex(values[i].ToString()).c_str());
// we continue after error rather than exiting so that we can
// find more errors if any
}
}
}
return ret_status;
}
// Given a key, this does prefix scans for "0"+P, "1"+P,..."9"+P
// in the same snapshot where P is the first FLAGS_prefix_size - 1 bytes
// of the key. Each of these 10 scans returns a series of values;
// each series should be the same length, and it is verified for each
// index i that all the i'th values are of the form "0"+V, "1"+V,..."9"+V.
// ASSUMES that MultiPut was used to put (K, V)
virtual Status TestPrefixScan(ThreadState* thread,
const ReadOptions& readoptions,
const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys) {
size_t prefix_to_use =
(FLAGS_prefix_size < 0) ? 7 : static_cast<size_t>(FLAGS_prefix_size);
std::string key_str = Key(rand_keys[0]);
Slice key = key_str;
auto cfh = column_families_[rand_column_families[0]];
std::string prefixes[10] = {"0", "1", "2", "3", "4",
"5", "6", "7", "8", "9"};
Slice prefix_slices[10];
ReadOptions readoptionscopy[10];
const Snapshot* snapshot = db_->GetSnapshot();
Iterator* iters[10];
std::string upper_bounds[10];
Slice ub_slices[10];
Status s = Status::OK();
for (int i = 0; i < 10; i++) {
prefixes[i] += key.ToString();
prefixes[i].resize(prefix_to_use);
prefix_slices[i] = Slice(prefixes[i]);
readoptionscopy[i] = readoptions;
readoptionscopy[i].snapshot = snapshot;
if (thread->rand.OneIn(2) &&
GetNextPrefix(prefix_slices[i], &(upper_bounds[i]))) {
// For half of the time, set the upper bound to the next prefix
ub_slices[i] = Slice(upper_bounds[i]);
readoptionscopy[i].iterate_upper_bound = &(ub_slices[i]);
}
iters[i] = db_->NewIterator(readoptionscopy[i], cfh);
iters[i]->Seek(prefix_slices[i]);
}
long count = 0;
while (iters[0]->Valid() && iters[0]->key().starts_with(prefix_slices[0])) {
count++;
std::string values[10];
// get list of all values for this iteration
for (int i = 0; i < 10; i++) {
// no iterator should finish before the first one
assert(iters[i]->Valid() &&
iters[i]->key().starts_with(prefix_slices[i]));
values[i] = iters[i]->value().ToString();
char expected_first = (prefixes[i])[0];
char actual_first = (values[i])[0];
if (actual_first != expected_first) {
fprintf(stderr, "error expected first = %c actual = %c\n",
expected_first, actual_first);
}
(values[i])[0] = ' '; // blank out the differing character
}
// make sure all values are equivalent
for (int i = 0; i < 10; i++) {
if (values[i] != values[0]) {
fprintf(stderr,
"error : %d, inconsistent values for prefix %s: %s, %s\n", i,
prefixes[i].c_str(), StringToHex(values[0]).c_str(),
StringToHex(values[i]).c_str());
// we continue after error rather than exiting so that we can
// find more errors if any
}
iters[i]->Next();
}
}
// cleanup iterators and snapshot
for (int i = 0; i < 10; i++) {
// if the first iterator finished, they should have all finished
assert(!iters[i]->Valid() ||
!iters[i]->key().starts_with(prefix_slices[i]));
assert(iters[i]->status().ok());
delete iters[i];
}
db_->ReleaseSnapshot(snapshot);
if (s.ok()) {
thread->stats.AddPrefixes(1, count);
} else {
thread->stats.AddErrors(1);
}
return s;
}
virtual void VerifyDb(ThreadState* /* thread */) const {}
};
StressTest* CreateBatchedOpsStressTest() { return new BatchedOpsStressTest(); }
} // namespace rocksdb
#endif // GFLAGS

510
db_stress_tool/cf_consistency_stress.cc Normal file
View File

@ -0,0 +1,510 @@
// 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.
#ifdef GFLAGS
#include "db_stress_tool/db_stress_common.h"
namespace rocksdb {
class CfConsistencyStressTest : public StressTest {
public:
CfConsistencyStressTest() : batch_id_(0) {}
virtual ~CfConsistencyStressTest() {}
virtual Status TestPut(ThreadState* thread, WriteOptions& write_opts,
const ReadOptions& /* read_opts */,
const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys,
char (&value)[100],
std::unique_ptr<MutexLock>& /* lock */) {
std::string key_str = Key(rand_keys[0]);
Slice key = key_str;
uint64_t value_base = batch_id_.fetch_add(1);
size_t sz =
GenerateValue(static_cast<uint32_t>(value_base), value, sizeof(value));
Slice v(value, sz);
WriteBatch batch;
for (auto cf : rand_column_families) {
ColumnFamilyHandle* cfh = column_families_[cf];
if (FLAGS_use_merge) {
batch.Merge(cfh, key, v);
} else { /* !FLAGS_use_merge */
batch.Put(cfh, key, v);
}
}
Status s = db_->Write(write_opts, &batch);
if (!s.ok()) {
fprintf(stderr, "multi put or merge error: %s\n", s.ToString().c_str());
thread->stats.AddErrors(1);
} else {
auto num = static_cast<long>(rand_column_families.size());
thread->stats.AddBytesForWrites(num, (sz + 1) * num);
}
return s;
}
virtual Status TestDelete(ThreadState* thread, WriteOptions& write_opts,
const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys,
std::unique_ptr<MutexLock>& /* lock */) {
std::string key_str = Key(rand_keys[0]);
Slice key = key_str;
WriteBatch batch;
for (auto cf : rand_column_families) {
ColumnFamilyHandle* cfh = column_families_[cf];
batch.Delete(cfh, key);
}
Status s = db_->Write(write_opts, &batch);
if (!s.ok()) {
fprintf(stderr, "multidel error: %s\n", s.ToString().c_str());
thread->stats.AddErrors(1);
} else {
thread->stats.AddDeletes(static_cast<long>(rand_column_families.size()));
}
return s;
}
virtual Status TestDeleteRange(ThreadState* thread, WriteOptions& write_opts,
const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys,
std::unique_ptr<MutexLock>& /* lock */) {
int64_t rand_key = rand_keys[0];
auto shared = thread->shared;
int64_t max_key = shared->GetMaxKey();
if (rand_key > max_key - FLAGS_range_deletion_width) {
rand_key =
thread->rand.Next() % (max_key - FLAGS_range_deletion_width + 1);
}
std::string key_str = Key(rand_key);
Slice key = key_str;
std::string end_key_str = Key(rand_key + FLAGS_range_deletion_width);
Slice end_key = end_key_str;
WriteBatch batch;
for (auto cf : rand_column_families) {
ColumnFamilyHandle* cfh = column_families_[rand_column_families[cf]];
batch.DeleteRange(cfh, key, end_key);
}
Status s = db_->Write(write_opts, &batch);
if (!s.ok()) {
fprintf(stderr, "multi del range error: %s\n", s.ToString().c_str());
thread->stats.AddErrors(1);
} else {
thread->stats.AddRangeDeletions(
static_cast<long>(rand_column_families.size()));
}
return s;
}
virtual void TestIngestExternalFile(
ThreadState* /* thread */,
const std::vector<int>& /* rand_column_families */,
const std::vector<int64_t>& /* rand_keys */,
std::unique_ptr<MutexLock>& /* lock */) {
assert(false);
fprintf(stderr,
"CfConsistencyStressTest does not support TestIngestExternalFile "
"because it's not possible to verify the result\n");
std::terminate();
}
virtual Status TestGet(ThreadState* thread, const ReadOptions& readoptions,
const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys) {
std::string key_str = Key(rand_keys[0]);
Slice key = key_str;
Status s;
bool is_consistent = true;
if (thread->rand.OneIn(2)) {
// 1/2 chance, does a random read from random CF
auto cfh =
column_families_[rand_column_families[thread->rand.Next() %
rand_column_families.size()]];
std::string from_db;
s = db_->Get(readoptions, cfh, key, &from_db);
} else {
// 1/2 chance, comparing one key is the same across all CFs
const Snapshot* snapshot = db_->GetSnapshot();
ReadOptions readoptionscopy = readoptions;
readoptionscopy.snapshot = snapshot;
std::string value0;
s = db_->Get(readoptionscopy, column_families_[rand_column_families[0]],
key, &value0);
if (s.ok() || s.IsNotFound()) {
bool found = s.ok();
for (size_t i = 1; i < rand_column_families.size(); i++) {
std::string value1;
s = db_->Get(readoptionscopy,
column_families_[rand_column_families[i]], key, &value1);
if (!s.ok() && !s.IsNotFound()) {
break;
}
if (!found && s.ok()) {
fprintf(stderr, "Get() return different results with key %s\n",
Slice(key_str).ToString(true).c_str());
fprintf(stderr, "CF %s is not found\n",
column_family_names_[0].c_str());
fprintf(stderr, "CF %s returns value %s\n",
column_family_names_[i].c_str(),
Slice(value1).ToString(true).c_str());
is_consistent = false;
} else if (found && s.IsNotFound()) {
fprintf(stderr, "Get() return different results with key %s\n",
Slice(key_str).ToString(true).c_str());
fprintf(stderr, "CF %s returns value %s\n",
column_family_names_[0].c_str(),
Slice(value0).ToString(true).c_str());
fprintf(stderr, "CF %s is not found\n",
column_family_names_[i].c_str());
is_consistent = false;
} else if (s.ok() && value0 != value1) {
fprintf(stderr, "Get() return different results with key %s\n",
Slice(key_str).ToString(true).c_str());
fprintf(stderr, "CF %s returns value %s\n",
column_family_names_[0].c_str(),
Slice(value0).ToString(true).c_str());
fprintf(stderr, "CF %s returns value %s\n",
column_family_names_[i].c_str(),
Slice(value1).ToString(true).c_str());
is_consistent = false;
}
if (!is_consistent) {
break;
}
}
}
db_->ReleaseSnapshot(snapshot);
}
if (!is_consistent) {
thread->stats.AddErrors(1);
// Fail fast to preserve the DB state.
thread->shared->SetVerificationFailure();
} else if (s.ok()) {
thread->stats.AddGets(1, 1);
} else if (s.IsNotFound()) {
thread->stats.AddGets(1, 0);
} else {
thread->stats.AddErrors(1);
}
return s;
}
virtual std::vector<Status> TestMultiGet(
ThreadState* thread, const ReadOptions& read_opts,
const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys) {
size_t num_keys = rand_keys.size();
std::vector<std::string> key_str;
std::vector<Slice> keys;
keys.reserve(num_keys);
key_str.reserve(num_keys);
std::vector<PinnableSlice> values(num_keys);
std::vector<Status> statuses(num_keys);
ColumnFamilyHandle* cfh = column_families_[rand_column_families[0]];
for (size_t i = 0; i < num_keys; ++i) {
key_str.emplace_back(Key(rand_keys[i]));
keys.emplace_back(key_str.back());
}
db_->MultiGet(read_opts, cfh, num_keys, keys.data(), values.data(),
statuses.data());
for (auto s : statuses) {
if (s.ok()) {
// found case
thread->stats.AddGets(1, 1);
} else if (s.IsNotFound()) {
// not found case
thread->stats.AddGets(1, 0);
} else {
// errors case
thread->stats.AddErrors(1);
}
}
return statuses;
}
virtual Status TestPrefixScan(ThreadState* thread,
const ReadOptions& readoptions,
const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys) {
size_t prefix_to_use =
(FLAGS_prefix_size < 0) ? 7 : static_cast<size_t>(FLAGS_prefix_size);
std::string key_str = Key(rand_keys[0]);
Slice key = key_str;
Slice prefix = Slice(key.data(), prefix_to_use);
std::string upper_bound;
Slice ub_slice;
ReadOptions ro_copy = readoptions;
if (thread->rand.OneIn(2) && GetNextPrefix(prefix, &upper_bound)) {
ub_slice = Slice(upper_bound);
ro_copy.iterate_upper_bound = &ub_slice;
}
auto cfh =
column_families_[rand_column_families[thread->rand.Next() %
rand_column_families.size()]];
Iterator* iter = db_->NewIterator(ro_copy, cfh);
long count = 0;
for (iter->Seek(prefix); iter->Valid() && iter->key().starts_with(prefix);
iter->Next()) {
++count;
}
assert(prefix_to_use == 0 ||
count <= (static_cast<long>(1) << ((8 - prefix_to_use) * 8)));
Status s = iter->status();
if (s.ok()) {
thread->stats.AddPrefixes(1, count);
} else {
thread->stats.AddErrors(1);
}
delete iter;
return s;
}
virtual ColumnFamilyHandle* GetControlCfh(ThreadState* thread,
int /*column_family_id*/
) {
// All column families should contain the same data. Randomly pick one.
return column_families_[thread->rand.Next() % column_families_.size()];
}
#ifdef ROCKSDB_LITE
virtual Status TestCheckpoint(
ThreadState* /* thread */,
const std::vector<int>& /* rand_column_families */,
const std::vector<int64_t>& /* rand_keys */) {
assert(false);
fprintf(stderr,
"RocksDB lite does not support "
"TestCheckpoint\n");
std::terminate();
}
#else
virtual Status TestCheckpoint(
ThreadState* thread, const std::vector<int>& /* rand_column_families */,
const std::vector<int64_t>& /* rand_keys */) {
std::string checkpoint_dir =
FLAGS_db + "/.checkpoint" + ToString(thread->tid);
DestroyDB(checkpoint_dir, options_);
Checkpoint* checkpoint = nullptr;
Status s = Checkpoint::Create(db_, &checkpoint);
if (s.ok()) {
s = checkpoint->CreateCheckpoint(checkpoint_dir);
}
std::vector<ColumnFamilyHandle*> cf_handles;
DB* checkpoint_db = nullptr;
if (s.ok()) {
delete checkpoint;
checkpoint = nullptr;
Options options(options_);
options.listeners.clear();
std::vector<ColumnFamilyDescriptor> cf_descs;
// TODO(ajkr): `column_family_names_` is not safe to access here when
// `clear_column_family_one_in != 0`. But we can't easily switch to
// `ListColumnFamilies` to get names because it won't necessarily give
// the same order as `column_family_names_`.
if (FLAGS_clear_column_family_one_in == 0) {
for (const auto& name : column_family_names_) {
cf_descs.emplace_back(name, ColumnFamilyOptions(options));
}
s = DB::OpenForReadOnly(DBOptions(options), checkpoint_dir, cf_descs,
&cf_handles, &checkpoint_db);
}
}
if (checkpoint_db != nullptr) {
for (auto cfh : cf_handles) {
delete cfh;
}
cf_handles.clear();
delete checkpoint_db;
checkpoint_db = nullptr;
}
DestroyDB(checkpoint_dir, options_);
if (!s.ok()) {
fprintf(stderr, "A checkpoint operation failed with: %s\n",
s.ToString().c_str());
}
return s;
}
#endif // !ROCKSDB_LITE
virtual void VerifyDb(ThreadState* thread) const {
ReadOptions options(FLAGS_verify_checksum, true);
// We must set total_order_seek to true because we are doing a SeekToFirst
// on a column family whose memtables may support (by default) prefix-based
// iterator. In this case, NewIterator with options.total_order_seek being
// false returns a prefix-based iterator. Calling SeekToFirst using this
// iterator causes the iterator to become invalid. That means we cannot
// iterate the memtable using this iterator any more, although the memtable
// contains the most up-to-date key-values.
options.total_order_seek = true;
assert(thread != nullptr);
auto shared = thread->shared;
std::vector<std::unique_ptr<Iterator>> iters(column_families_.size());
for (size_t i = 0; i != column_families_.size(); ++i) {
iters[i].reset(db_->NewIterator(options, column_families_[i]));
}
for (auto& iter : iters) {
iter->SeekToFirst();
}
size_t num = column_families_.size();
assert(num == iters.size());
std::vector<Status> statuses(num, Status::OK());
do {
if (shared->HasVerificationFailedYet()) {
break;
}
size_t valid_cnt = 0;
size_t idx = 0;
for (auto& iter : iters) {
if (iter->Valid()) {
++valid_cnt;
} else {
statuses[idx] = iter->status();
}
++idx;
}
if (valid_cnt == 0) {
Status status;
for (size_t i = 0; i != num; ++i) {
const auto& s = statuses[i];
if (!s.ok()) {
status = s;
fprintf(stderr, "Iterator on cf %s has error: %s\n",
column_families_[i]->GetName().c_str(),
s.ToString().c_str());
shared->SetVerificationFailure();
}
}
if (status.ok()) {
fprintf(stdout, "Finished scanning all column families.\n");
}
break;
} else if (valid_cnt != iters.size()) {
shared->SetVerificationFailure();
for (size_t i = 0; i != num; ++i) {
if (!iters[i]->Valid()) {
if (statuses[i].ok()) {
fprintf(stderr, "Finished scanning cf %s\n",
column_families_[i]->GetName().c_str());
} else {
fprintf(stderr, "Iterator on cf %s has error: %s\n",
column_families_[i]->GetName().c_str(),
statuses[i].ToString().c_str());
}
} else {
fprintf(stderr, "cf %s has remaining data to scan\n",
column_families_[i]->GetName().c_str());
}
}
break;
}
if (shared->HasVerificationFailedYet()) {
break;
}
// If the program reaches here, then all column families' iterators are
// still valid.
if (shared->PrintingVerificationResults()) {
continue;
}
Slice key;
Slice value;
int num_mismatched_cfs = 0;
for (size_t i = 0; i != num; ++i) {
if (i == 0) {
key = iters[i]->key();
value = iters[i]->value();
} else {
int cmp = key.compare(iters[i]->key());
if (cmp != 0) {
++num_mismatched_cfs;
if (1 == num_mismatched_cfs) {
fprintf(stderr, "Verification failed\n");
fprintf(stderr, "Latest Sequence Number: %" PRIu64 "\n",
db_->GetLatestSequenceNumber());
fprintf(stderr, "[%s] %s => %s\n",
column_families_[0]->GetName().c_str(),
key.ToString(true /* hex */).c_str(),
value.ToString(true /* hex */).c_str());
}
fprintf(stderr, "[%s] %s => %s\n",
column_families_[i]->GetName().c_str(),
iters[i]->key().ToString(true /* hex */).c_str(),
iters[i]->value().ToString(true /* hex */).c_str());
#ifndef ROCKSDB_LITE
Slice begin_key;
Slice end_key;
if (cmp < 0) {
begin_key = key;
end_key = iters[i]->key();
} else {
begin_key = iters[i]->key();
end_key = key;
}
std::vector<KeyVersion> versions;
const size_t kMaxNumIKeys = 8;
const auto print_key_versions = [&](ColumnFamilyHandle* cfh) {
Status s = GetAllKeyVersions(db_, cfh, begin_key, end_key,
kMaxNumIKeys, &versions);
if (!s.ok()) {
fprintf(stderr, "%s\n", s.ToString().c_str());
return;
}
assert(nullptr != cfh);
fprintf(stderr,
"Internal keys in CF '%s', [%s, %s] (max %" ROCKSDB_PRIszt
")\n",
cfh->GetName().c_str(),
begin_key.ToString(true /* hex */).c_str(),
end_key.ToString(true /* hex */).c_str(), kMaxNumIKeys);
for (const KeyVersion& kv : versions) {
fprintf(stderr, " key %s seq %" PRIu64 " type %d\n",
Slice(kv.user_key).ToString(true).c_str(), kv.sequence,
kv.type);
}
};
if (1 == num_mismatched_cfs) {
print_key_versions(column_families_[0]);
}
print_key_versions(column_families_[i]);
#endif // ROCKSDB_LITE
shared->SetVerificationFailure();
}
}
}
shared->FinishPrintingVerificationResults();
for (auto& iter : iters) {
iter->Next();
}
} while (true);
}
virtual std::vector<int> GenerateColumnFamilies(
const int /* num_column_families */, int /* rand_column_family */) const {
std::vector<int> ret;
int num = static_cast<int>(column_families_.size());
int k = 0;
std::generate_n(back_inserter(ret), num, [&k]() -> int { return k++; });
return ret;
}
private:
std::atomic<int64_t> batch_id_;
};
StressTest* CreateCfConsistencyStressTest() {
return new CfConsistencyStressTest();
}
} // namespace rocksdb
#endif // GFLAGS

0
tools/db_stress.cc → db_stress_tool/db_stress.cc
View File

108
db_stress_tool/db_stress_common.cc Normal file
View File

@ -0,0 +1,108 @@
// 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.
//
#ifdef GFLAGS
#include "db_stress_tool/db_stress_common.h"
rocksdb::Env* FLAGS_env = rocksdb::Env::Default();
enum rocksdb::CompressionType FLAGS_compression_type_e =
rocksdb::kSnappyCompression;
enum rocksdb::ChecksumType FLAGS_checksum_type_e = rocksdb::kCRC32c;
enum RepFactory FLAGS_rep_factory = kSkipList;
namespace rocksdb {
void PoolSizeChangeThread(void* v) {
assert(FLAGS_compaction_thread_pool_adjust_interval > 0);
ThreadState* thread = reinterpret_cast<ThreadState*>(v);
SharedState* shared = thread->shared;
while (true) {
{
MutexLock l(shared->GetMutex());
if (shared->ShoudStopBgThread()) {
shared->SetBgThreadFinish();
shared->GetCondVar()->SignalAll();
return;
}
}
auto thread_pool_size_base = FLAGS_max_background_compactions;
auto thread_pool_size_var = FLAGS_compaction_thread_pool_variations;
int new_thread_pool_size =
thread_pool_size_base - thread_pool_size_var +
thread->rand.Next() % (thread_pool_size_var * 2 + 1);
if (new_thread_pool_size < 1) {
new_thread_pool_size = 1;
}
FLAGS_env->SetBackgroundThreads(new_thread_pool_size);
// Sleep up to 3 seconds
FLAGS_env->SleepForMicroseconds(
thread->rand.Next() % FLAGS_compaction_thread_pool_adjust_interval *
1000 +
1);
}
}
void PrintKeyValue(int cf, uint64_t key, const char* value, size_t sz) {
if (!FLAGS_verbose) {
return;
}
std::string tmp;
tmp.reserve(sz * 2 + 16);
char buf[4];
for (size_t i = 0; i < sz; i++) {
snprintf(buf, 4, "%X", value[i]);
tmp.append(buf);
}
fprintf(stdout, "[CF %d] %" PRIi64 " == > (%" ROCKSDB_PRIszt ") %s\n", cf,
key, sz, tmp.c_str());
}
int64_t GenerateOneKey(ThreadState* thread, uint64_t iteration) {
const double completed_ratio =
static_cast<double>(iteration) / FLAGS_ops_per_thread;
const int64_t base_key = static_cast<int64_t>(
completed_ratio * (FLAGS_max_key - FLAGS_active_width));
return base_key + thread->rand.Next() % FLAGS_active_width;
}
std::vector<int64_t> GenerateNKeys(ThreadState* thread, int num_keys,
uint64_t iteration) {
const double completed_ratio =
static_cast<double>(iteration) / FLAGS_ops_per_thread;
const int64_t base_key = static_cast<int64_t>(
completed_ratio * (FLAGS_max_key - FLAGS_active_width));
std::vector<int64_t> keys;
keys.reserve(num_keys);
int64_t next_key = base_key + thread->rand.Next() % FLAGS_active_width;
keys.push_back(next_key);
for (int i = 1; i < num_keys; ++i) {
// This may result in some duplicate keys
next_key = next_key + thread->rand.Next() %
(FLAGS_active_width - (next_key - base_key));
keys.push_back(next_key);
}
return keys;
}
size_t GenerateValue(uint32_t rand, char* v, size_t max_sz) {
size_t value_sz =
((rand % kRandomValueMaxFactor) + 1) * FLAGS_value_size_mult;
assert(value_sz <= max_sz && value_sz >= sizeof(uint32_t));
(void)max_sz;
*((uint32_t*)v) = rand;
for (size_t i = sizeof(uint32_t); i < value_sz; i++) {
v[i] = (char)(rand ^ i);
}
v[value_sz] = '\0';
return value_sz; // the size of the value set.
}
} // namespace rocksdb
#endif // GFLAGS

361
db_stress_tool/db_stress_common.h Normal file
View File

@ -0,0 +1,361 @@
// 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.
//
// The test uses an array to compare against values written to the database.
// Keys written to the array are in 1:1 correspondence to the actual values in
// the database according to the formula in the function GenerateValue.
// Space is reserved in the array from 0 to FLAGS_max_key and values are
// randomly written/deleted/read from those positions. During verification we
// compare all the positions in the array. To shorten/elongate the running
// time, you could change the settings: FLAGS_max_key, FLAGS_ops_per_thread,
// (sometimes also FLAGS_threads).
//
// NOTE that if FLAGS_test_batches_snapshots is set, the test will have
// different behavior. See comment of the flag for details.
#ifdef GFLAGS
#pragma once
#include <fcntl.h>
#include <stdio.h>
#include <stdlib.h>
#include <sys/types.h>
#include <algorithm>
#include <array>
#include <chrono>
#include <cinttypes>
#include <exception>
#include <queue>
#include <thread>
#include "db/db_impl/db_impl.h"
#include "db/version_set.h"
#include "db_stress_tool/db_stress_listener.h"
#include "db_stress_tool/db_stress_shared_state.h"
#include "db_stress_tool/db_stress_test_base.h"
#include "hdfs/env_hdfs.h"
#include "logging/logging.h"
#include "monitoring/histogram.h"
#include "options/options_helper.h"
#include "port/port.h"
#include "rocksdb/cache.h"
#include "rocksdb/env.h"
#include "rocksdb/slice.h"
#include "rocksdb/slice_transform.h"
#include "rocksdb/statistics.h"
#include "rocksdb/utilities/backupable_db.h"
#include "rocksdb/utilities/checkpoint.h"
#include "rocksdb/utilities/db_ttl.h"
#include "rocksdb/utilities/debug.h"
#include "rocksdb/utilities/options_util.h"
#include "rocksdb/utilities/transaction.h"
#include "rocksdb/utilities/transaction_db.h"
#include "rocksdb/write_batch.h"
#include "util/coding.h"
#include "util/compression.h"
#include "util/crc32c.h"
#include "util/gflags_compat.h"
#include "util/mutexlock.h"
#include "util/random.h"
#include "util/string_util.h"
// SyncPoint is not supported in Released Windows Mode.
#if !(defined NDEBUG) || !defined(OS_WIN)
#include "test_util/sync_point.h"
#endif // !(defined NDEBUG) || !defined(OS_WIN)
#include "test_util/testutil.h"
#include "utilities/merge_operators.h"
using GFLAGS_NAMESPACE::ParseCommandLineFlags;
using GFLAGS_NAMESPACE::RegisterFlagValidator;
using GFLAGS_NAMESPACE::SetUsageMessage;
DECLARE_uint64(seed);
DECLARE_bool(read_only);
DECLARE_int64(max_key);
DECLARE_int32(column_families);
DECLARE_string(options_file);
DECLARE_int64(active_width);
DECLARE_bool(test_batches_snapshots);
DECLARE_bool(atomic_flush);
DECLARE_bool(test_cf_consistency);
DECLARE_int32(threads);
DECLARE_int32(ttl);
DECLARE_int32(value_size_mult);
DECLARE_int32(compaction_readahead_size);
DECLARE_bool(enable_pipelined_write);
DECLARE_bool(verify_before_write);
DECLARE_bool(histogram);
DECLARE_bool(destroy_db_initially);
DECLARE_bool(verbose);
DECLARE_bool(progress_reports);
DECLARE_uint64(db_write_buffer_size);
DECLARE_int32(write_buffer_size);
DECLARE_int32(max_write_buffer_number);
DECLARE_int32(min_write_buffer_number_to_merge);
DECLARE_int32(max_write_buffer_number_to_maintain);
DECLARE_int64(max_write_buffer_size_to_maintain);
DECLARE_double(memtable_prefix_bloom_size_ratio);
DECLARE_bool(memtable_whole_key_filtering);
DECLARE_int32(open_files);
DECLARE_int64(compressed_cache_size);
DECLARE_int32(compaction_style);
DECLARE_int32(level0_file_num_compaction_trigger);
DECLARE_int32(level0_slowdown_writes_trigger);
DECLARE_int32(level0_stop_writes_trigger);
DECLARE_int32(block_size);
DECLARE_int32(format_version);
DECLARE_int32(index_block_restart_interval);
DECLARE_int32(max_background_compactions);
DECLARE_int32(num_bottom_pri_threads);
DECLARE_int32(compaction_thread_pool_adjust_interval);
DECLARE_int32(compaction_thread_pool_variations);
DECLARE_int32(max_background_flushes);
DECLARE_int32(universal_size_ratio);
DECLARE_int32(universal_min_merge_width);
DECLARE_int32(universal_max_merge_width);
DECLARE_int32(universal_max_size_amplification_percent);
DECLARE_int32(clear_column_family_one_in);
DECLARE_int32(set_options_one_in);
DECLARE_int32(set_in_place_one_in);
DECLARE_int64(cache_size);
DECLARE_bool(cache_index_and_filter_blocks);
DECLARE_bool(use_clock_cache);
DECLARE_uint64(subcompactions);
DECLARE_uint64(periodic_compaction_seconds);
DECLARE_uint64(compaction_ttl);
DECLARE_bool(allow_concurrent_memtable_write);
DECLARE_bool(enable_write_thread_adaptive_yield);
DECLARE_int32(reopen);
DECLARE_int32(bloom_bits);
DECLARE_bool(use_block_based_filter);
DECLARE_bool(partition_filters);
DECLARE_int32(index_type);
DECLARE_string(db);
DECLARE_string(secondaries_base);
DECLARE_bool(enable_secondary);
DECLARE_string(expected_values_path);
DECLARE_bool(verify_checksum);
DECLARE_bool(mmap_read);
DECLARE_bool(mmap_write);
DECLARE_bool(use_direct_reads);
DECLARE_bool(use_direct_io_for_flush_and_compaction);
DECLARE_bool(statistics);
DECLARE_bool(sync);
DECLARE_bool(use_fsync);
DECLARE_int32(kill_random_test);
DECLARE_string(kill_prefix_blacklist);
DECLARE_bool(disable_wal);
DECLARE_uint64(recycle_log_file_num);
DECLARE_int64(target_file_size_base);
DECLARE_int32(target_file_size_multiplier);
DECLARE_uint64(max_bytes_for_level_base);
DECLARE_double(max_bytes_for_level_multiplier);
DECLARE_int32(range_deletion_width);
DECLARE_uint64(rate_limiter_bytes_per_sec);
DECLARE_bool(rate_limit_bg_reads);
DECLARE_bool(use_txn);
DECLARE_int32(backup_one_in);
DECLARE_int32(checkpoint_one_in);
DECLARE_int32(ingest_external_file_one_in);
DECLARE_int32(ingest_external_file_width);
DECLARE_int32(compact_files_one_in);
DECLARE_int32(compact_range_one_in);
DECLARE_int32(flush_one_in);
DECLARE_int32(compact_range_width);
DECLARE_int32(acquire_snapshot_one_in);
DECLARE_bool(compare_full_db_state_snapshot);
DECLARE_uint64(snapshot_hold_ops);
DECLARE_bool(use_multiget);
DECLARE_int32(readpercent);
DECLARE_int32(prefixpercent);
DECLARE_int32(writepercent);
DECLARE_int32(delpercent);
DECLARE_int32(delrangepercent);
DECLARE_int32(nooverwritepercent);
DECLARE_int32(iterpercent);
DECLARE_uint64(num_iterations);
DECLARE_string(compression_type);
DECLARE_int32(compression_max_dict_bytes);
DECLARE_int32(compression_zstd_max_train_bytes);
DECLARE_string(checksum_type);
DECLARE_string(hdfs);
DECLARE_string(env_uri);
DECLARE_uint64(ops_per_thread);
DECLARE_uint64(log2_keys_per_lock);
DECLARE_uint64(max_manifest_file_size);
DECLARE_bool(in_place_update);
DECLARE_int32(secondary_catch_up_one_in);
DECLARE_string(memtablerep);
DECLARE_int32(prefix_size);
DECLARE_bool(use_merge);
DECLARE_bool(use_full_merge_v1);
static const long KB = 1024;
static const int kRandomValueMaxFactor = 3;
static const int kValueMaxLen = 100;
static std::shared_ptr<rocksdb::Env> env_guard;
// posix or hdfs environment
extern rocksdb::Env* FLAGS_env;
extern enum rocksdb::CompressionType FLAGS_compression_type_e;
extern enum rocksdb::ChecksumType FLAGS_checksum_type_e;
enum RepFactory { kSkipList, kHashSkipList, kVectorRep };
inline enum RepFactory StringToRepFactory(const char* ctype) {
assert(ctype);
if (!strcasecmp(ctype, "skip_list"))
return kSkipList;
else if (!strcasecmp(ctype, "prefix_hash"))
return kHashSkipList;
else if (!strcasecmp(ctype, "vector"))
return kVectorRep;
fprintf(stdout, "Cannot parse memreptable %s\n", ctype);
return kSkipList;
}
extern enum RepFactory FLAGS_rep_factory;
namespace rocksdb {
inline enum rocksdb::CompressionType StringToCompressionType(
const char* ctype) {
assert(ctype);
if (!strcasecmp(ctype, "none"))
return rocksdb::kNoCompression;
else if (!strcasecmp(ctype, "snappy"))
return rocksdb::kSnappyCompression;
else if (!strcasecmp(ctype, "zlib"))
return rocksdb::kZlibCompression;
else if (!strcasecmp(ctype, "bzip2"))
return rocksdb::kBZip2Compression;
else if (!strcasecmp(ctype, "lz4"))
return rocksdb::kLZ4Compression;
else if (!strcasecmp(ctype, "lz4hc"))
return rocksdb::kLZ4HCCompression;
else if (!strcasecmp(ctype, "xpress"))
return rocksdb::kXpressCompression;
else if (!strcasecmp(ctype, "zstd"))
return rocksdb::kZSTD;
fprintf(stderr, "Cannot parse compression type '%s'\n", ctype);
return rocksdb::kSnappyCompression; // default value
}
inline enum rocksdb::ChecksumType StringToChecksumType(const char* ctype) {
assert(ctype);
auto iter = rocksdb::checksum_type_string_map.find(ctype);
if (iter != rocksdb::checksum_type_string_map.end()) {
return iter->second;
}
fprintf(stderr, "Cannot parse checksum type '%s'\n", ctype);
return rocksdb::kCRC32c;
}
inline std::string ChecksumTypeToString(rocksdb::ChecksumType ctype) {
auto iter = std::find_if(
rocksdb::checksum_type_string_map.begin(),
rocksdb::checksum_type_string_map.end(),
[&](const std::pair<std::string, rocksdb::ChecksumType>&
name_and_enum_val) { return name_and_enum_val.second == ctype; });
assert(iter != rocksdb::checksum_type_string_map.end());
return iter->first;
}
inline std::vector<std::string> SplitString(std::string src) {
std::vector<std::string> ret;
if (src.empty()) {
return ret;
}
size_t pos = 0;
size_t pos_comma;
while ((pos_comma = src.find(',', pos)) != std::string::npos) {
ret.push_back(src.substr(pos, pos_comma - pos));
pos = pos_comma + 1;
}
ret.push_back(src.substr(pos, src.length()));
return ret;
}
#ifdef _MSC_VER
#pragma warning(push)
// truncation of constant value on static_cast
#pragma warning(disable : 4309)
#endif
inline bool GetNextPrefix(const rocksdb::Slice& src, std::string* v) {
std::string ret = src.ToString();
for (int i = static_cast<int>(ret.size()) - 1; i >= 0; i--) {
if (ret[i] != static_cast<char>(255)) {
ret[i] = ret[i] + 1;
break;
} else if (i != 0) {
ret[i] = 0;
} else {
// all FF. No next prefix
return false;
}
}
*v = ret;
return true;
}
#ifdef _MSC_VER
#pragma warning(pop)
#endif
// convert long to a big-endian slice key
extern inline std::string Key(int64_t val) {
std::string little_endian_key;
std::string big_endian_key;
PutFixed64(&little_endian_key, val);
assert(little_endian_key.size() == sizeof(val));
big_endian_key.resize(sizeof(val));
for (size_t i = 0; i < sizeof(val); ++i) {
big_endian_key[i] = little_endian_key[sizeof(val) - 1 - i];
}
return big_endian_key;
}
extern inline bool GetIntVal(std::string big_endian_key, uint64_t* key_p) {
unsigned int size_key = sizeof(*key_p);
assert(big_endian_key.size() == size_key);
std::string little_endian_key;
little_endian_key.resize(size_key);
for (size_t i = 0; i < size_key; ++i) {
little_endian_key[i] = big_endian_key[size_key - 1 - i];
}
Slice little_endian_slice = Slice(little_endian_key);
return GetFixed64(&little_endian_slice, key_p);
}
extern inline std::string StringToHex(const std::string& str) {
std::string result = "0x";
result.append(Slice(str).ToString(true));
return result;
}
extern void PoolSizeChangeThread(void* v);
extern void PrintKeyValue(int cf, uint64_t key, const char* value, size_t sz);
extern int64_t GenerateOneKey(ThreadState* thread, uint64_t iteration);
extern std::vector<int64_t> GenerateNKeys(ThreadState* thread, int num_keys,
uint64_t iteration);
extern size_t GenerateValue(uint32_t rand, char* v, size_t max_sz);
extern StressTest* CreateCfConsistencyStressTest();
extern StressTest* CreateBatchedOpsStressTest();
extern StressTest* CreateNonBatchedOpsStressTest();
} // namespace rocksdb
#endif // GFLAGS

157
db_stress_tool/db_stress_driver.cc Normal file
View File

@ -0,0 +1,157 @@
// 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.
//
#ifdef GFLAGS
#include "db_stress_tool/db_stress_common.h"
namespace rocksdb {
void ThreadBody(void* v) {
ThreadState* thread = reinterpret_cast<ThreadState*>(v);
SharedState* shared = thread->shared;
if (shared->ShouldVerifyAtBeginning()) {
thread->shared->GetStressTest()->VerifyDb(thread);
}
{
MutexLock l(shared->GetMutex());
shared->IncInitialized();
if (shared->AllInitialized()) {
shared->GetCondVar()->SignalAll();
}
while (!shared->Started()) {
shared->GetCondVar()->Wait();
}
}
thread->shared->GetStressTest()->OperateDb(thread);
{
MutexLock l(shared->GetMutex());
shared->IncOperated();
if (shared->AllOperated()) {
shared->GetCondVar()->SignalAll();
}
while (!shared->VerifyStarted()) {
shared->GetCondVar()->Wait();
}
}
thread->shared->GetStressTest()->VerifyDb(thread);
{
MutexLock l(shared->GetMutex());
shared->IncDone();
if (shared->AllDone()) {
shared->GetCondVar()->SignalAll();
}
}
}
bool RunStressTest(StressTest* stress) {
stress->InitDb();
SharedState shared(FLAGS_env, stress);
if (FLAGS_read_only) {
stress->InitReadonlyDb(&shared);
}
uint32_t n = shared.GetNumThreads();
uint64_t now = FLAGS_env->NowMicros();
fprintf(stdout, "%s Initializing worker threads\n",
FLAGS_env->TimeToString(now / 1000000).c_str());
std::vector<ThreadState*> threads(n);
for (uint32_t i = 0; i < n; i++) {
threads[i] = new ThreadState(i, &shared);
FLAGS_env->StartThread(ThreadBody, threads[i]);
}
ThreadState bg_thread(0, &shared);
if (FLAGS_compaction_thread_pool_adjust_interval > 0) {
FLAGS_env->StartThread(PoolSizeChangeThread, &bg_thread);
}
// Each thread goes through the following states:
// initializing -> wait for others to init -> read/populate/depopulate
// wait for others to operate -> verify -> done
{
MutexLock l(shared.GetMutex());
while (!shared.AllInitialized()) {
shared.GetCondVar()->Wait();
}
if (shared.ShouldVerifyAtBeginning()) {
if (shared.HasVerificationFailedYet()) {
printf("Crash-recovery verification failed :(\n");
} else {
printf("Crash-recovery verification passed :)\n");
}
}
now = FLAGS_env->NowMicros();
fprintf(stdout, "%s Starting database operations\n",
FLAGS_env->TimeToString(now / 1000000).c_str());
shared.SetStart();
shared.GetCondVar()->SignalAll();
while (!shared.AllOperated()) {
shared.GetCondVar()->Wait();
}
now = FLAGS_env->NowMicros();
if (FLAGS_test_batches_snapshots) {
fprintf(stdout, "%s Limited verification already done during gets\n",
FLAGS_env->TimeToString((uint64_t)now / 1000000).c_str());
} else {
fprintf(stdout, "%s Starting verification\n",
FLAGS_env->TimeToString((uint64_t)now / 1000000).c_str());
}
shared.SetStartVerify();
shared.GetCondVar()->SignalAll();
while (!shared.AllDone()) {
shared.GetCondVar()->Wait();
}
}
for (unsigned int i = 1; i < n; i++) {
threads[0]->stats.Merge(threads[i]->stats);
}
threads[0]->stats.Report("Stress Test");
for (unsigned int i = 0; i < n; i++) {
delete threads[i];
threads[i] = nullptr;
}
now = FLAGS_env->NowMicros();
if (!FLAGS_test_batches_snapshots && !shared.HasVerificationFailedYet()) {
fprintf(stdout, "%s Verification successful\n",
FLAGS_env->TimeToString(now / 1000000).c_str());
}
stress->PrintStatistics();
if (FLAGS_compaction_thread_pool_adjust_interval > 0) {
MutexLock l(shared.GetMutex());
shared.SetShouldStopBgThread();
while (!shared.BgThreadFinished()) {
shared.GetCondVar()->Wait();
}
}
if (!stress->VerifySecondaries()) {
return false;
}
if (shared.HasVerificationFailedYet()) {
printf("Verification failed :(\n");
return false;
}
return true;
}
} // namespace rocksdb
#endif // GFLAGS

17
db_stress_tool/db_stress_driver.h Normal file
View File

@ -0,0 +1,17 @@
// 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.
#ifdef GFLAGS
#pragma once
#include "db_stress_tool/db_stress_test_base.h"
namespace rocksdb {
extern void ThreadBody(void* /*thread_state*/);
extern bool RunStressTest(StressTest*);
} // namespace rocksdb
#endif // GFLAGS

532
db_stress_tool/db_stress_gflags.cc Normal file
View File

@ -0,0 +1,532 @@
// 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.
#ifdef GFLAGS
#include "db_stress_tool/db_stress_common.h"
static bool ValidateUint32Range(const char* flagname, uint64_t value) {
if (value > std::numeric_limits<uint32_t>::max()) {
fprintf(stderr, "Invalid value for --%s: %lu, overflow\n", flagname,
(unsigned long)value);
return false;
}
return true;
}
DEFINE_uint64(seed, 2341234, "Seed for PRNG");
static const bool FLAGS_seed_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_seed, &ValidateUint32Range);
DEFINE_bool(read_only, false, "True if open DB in read-only mode during tests");
DEFINE_int64(max_key, 1 * KB * KB,
"Max number of key/values to place in database");
DEFINE_int32(column_families, 10, "Number of column families");
DEFINE_string(
options_file, "",
"The path to a RocksDB options file. If specified, then db_stress will "
"run with the RocksDB options in the default column family of the "
"specified options file. Note that, when an options file is provided, "
"db_stress will ignore the flag values for all options that may be passed "
"via options file.");
DEFINE_int64(
active_width, 0,
"Number of keys in active span of the key-range at any given time. The "
"span begins with its left endpoint at key 0, gradually moves rightwards, "
"and ends with its right endpoint at max_key. If set to 0, active_width "
"will be sanitized to be equal to max_key.");
// TODO(noetzli) Add support for single deletes
DEFINE_bool(test_batches_snapshots, false,
"If set, the test uses MultiGet(), MultiPut() and MultiDelete()"
" which read/write/delete multiple keys in a batch. In this mode,"
" we do not verify db content by comparing the content with the "
"pre-allocated array. Instead, we do partial verification inside"
" MultiGet() by checking various values in a batch. Benefit of"
" this mode:\n"
"\t(a) No need to acquire mutexes during writes (less cache "
"flushes in multi-core leading to speed up)\n"
"\t(b) No long validation at the end (more speed up)\n"
"\t(c) Test snapshot and atomicity of batch writes");
DEFINE_bool(atomic_flush, false,
"If set, enables atomic flush in the options.\n");
DEFINE_bool(test_cf_consistency, false,
"If set, runs the stress test dedicated to verifying writes to "
"multiple column families are consistent. Setting this implies "
"`atomic_flush=true` is set true if `disable_wal=false`.\n");
DEFINE_int32(threads, 32, "Number of concurrent threads to run.");
DEFINE_int32(ttl, -1,
"Opens the db with this ttl value if this is not -1. "
"Carefully specify a large value such that verifications on "
"deleted values don't fail");
DEFINE_int32(value_size_mult, 8,
"Size of value will be this number times rand_int(1,3) bytes");
DEFINE_int32(compaction_readahead_size, 0, "Compaction readahead size");
DEFINE_bool(enable_pipelined_write, false, "Pipeline WAL/memtable writes");
DEFINE_bool(verify_before_write, false, "Verify before write");
DEFINE_bool(histogram, false, "Print histogram of operation timings");
DEFINE_bool(destroy_db_initially, true,
"Destroys the database dir before start if this is true");
DEFINE_bool(verbose, false, "Verbose");
DEFINE_bool(progress_reports, true,
"If true, db_stress will report number of finished operations");
DEFINE_uint64(db_write_buffer_size, rocksdb::Options().db_write_buffer_size,
"Number of bytes to buffer in all memtables before compacting");
DEFINE_int32(write_buffer_size,
static_cast<int32_t>(rocksdb::Options().write_buffer_size),
"Number of bytes to buffer in memtable before compacting");
DEFINE_int32(max_write_buffer_number,
rocksdb::Options().max_write_buffer_number,
"The number of in-memory memtables. "
"Each memtable is of size FLAGS_write_buffer_size.");
DEFINE_int32(min_write_buffer_number_to_merge,
rocksdb::Options().min_write_buffer_number_to_merge,
"The minimum number of write buffers that will be merged together "
"before writing to storage. This is cheap because it is an "
"in-memory merge. If this feature is not enabled, then all these "
"write buffers are flushed to L0 as separate files and this "
"increases read amplification because a get request has to check "
"in all of these files. Also, an in-memory merge may result in "
"writing less data to storage if there are duplicate records in"
" each of these individual write buffers.");
DEFINE_int32(max_write_buffer_number_to_maintain,
rocksdb::Options().max_write_buffer_number_to_maintain,
"The total maximum number of write buffers to maintain in memory "
"including copies of buffers that have already been flushed. "
"Unlike max_write_buffer_number, this parameter does not affect "
"flushing. This controls the minimum amount of write history "
"that will be available in memory for conflict checking when "
"Transactions are used. If this value is too low, some "
"transactions may fail at commit time due to not being able to "
"determine whether there were any write conflicts. Setting this "
"value to 0 will cause write buffers to be freed immediately "
"after they are flushed. If this value is set to -1, "
"'max_write_buffer_number' will be used.");
DEFINE_int64(max_write_buffer_size_to_maintain,
rocksdb::Options().max_write_buffer_size_to_maintain,
"The total maximum size of write buffers to maintain in memory "
"including copies of buffers that have already been flushed. "
"Unlike max_write_buffer_number, this parameter does not affect "
"flushing. This controls the minimum amount of write history "
"that will be available in memory for conflict checking when "
"Transactions are used. If this value is too low, some "
"transactions may fail at commit time due to not being able to "
"determine whether there were any write conflicts. Setting this "
"value to 0 will cause write buffers to be freed immediately "
"after they are flushed. If this value is set to -1, "
"'max_write_buffer_number' will be used.");
DEFINE_double(memtable_prefix_bloom_size_ratio,
rocksdb::Options().memtable_prefix_bloom_size_ratio,
"creates prefix blooms for memtables, each with size "
"`write_buffer_size * memtable_prefix_bloom_size_ratio`.");
DEFINE_bool(memtable_whole_key_filtering,
rocksdb::Options().memtable_whole_key_filtering,
"Enable whole key filtering in memtables.");
DEFINE_int32(open_files, rocksdb::Options().max_open_files,
"Maximum number of files to keep open at the same time "
"(use default if == 0)");
DEFINE_int64(compressed_cache_size, -1,
"Number of bytes to use as a cache of compressed data."
" Negative means use default settings.");
DEFINE_int32(compaction_style, rocksdb::Options().compaction_style, "");
DEFINE_int32(level0_file_num_compaction_trigger,
rocksdb::Options().level0_file_num_compaction_trigger,
"Level0 compaction start trigger");
DEFINE_int32(level0_slowdown_writes_trigger,
rocksdb::Options().level0_slowdown_writes_trigger,
"Number of files in level-0 that will slow down writes");
DEFINE_int32(level0_stop_writes_trigger,
rocksdb::Options().level0_stop_writes_trigger,
"Number of files in level-0 that will trigger put stop.");
DEFINE_int32(block_size,
static_cast<int32_t>(rocksdb::BlockBasedTableOptions().block_size),
"Number of bytes in a block.");
DEFINE_int32(
format_version,
static_cast<int32_t>(rocksdb::BlockBasedTableOptions().format_version),
"Format version of SST files.");
DEFINE_int32(index_block_restart_interval,
rocksdb::BlockBasedTableOptions().index_block_restart_interval,
"Number of keys between restart points "
"for delta encoding of keys in index block.");
DEFINE_int32(max_background_compactions,
rocksdb::Options().max_background_compactions,
"The maximum number of concurrent background compactions "
"that can occur in parallel.");
DEFINE_int32(num_bottom_pri_threads, 0,
"The number of threads in the bottom-priority thread pool (used "
"by universal compaction only).");
DEFINE_int32(compaction_thread_pool_adjust_interval, 0,
"The interval (in milliseconds) to adjust compaction thread pool "
"size. Don't change it periodically if the value is 0.");
DEFINE_int32(compaction_thread_pool_variations, 2,
"Range of background thread pool size variations when adjusted "
"periodically.");
DEFINE_int32(max_background_flushes, rocksdb::Options().max_background_flushes,
"The maximum number of concurrent background flushes "
"that can occur in parallel.");
DEFINE_int32(universal_size_ratio, 0,
"The ratio of file sizes that trigger"
" compaction in universal style");
DEFINE_int32(universal_min_merge_width, 0,
"The minimum number of files to "
"compact in universal style compaction");
DEFINE_int32(universal_max_merge_width, 0,
"The max number of files to compact"
" in universal style compaction");
DEFINE_int32(universal_max_size_amplification_percent, 0,
"The max size amplification for universal style compaction");
DEFINE_int32(clear_column_family_one_in, 1000000,
"With a chance of 1/N, delete a column family and then recreate "
"it again. If N == 0, never drop/create column families. "
"When test_batches_snapshots is true, this flag has no effect");
DEFINE_int32(set_options_one_in, 0,
"With a chance of 1/N, change some random options");
DEFINE_int32(set_in_place_one_in, 0,
"With a chance of 1/N, toggle in place support option");
DEFINE_int64(cache_size, 2LL * KB * KB * KB,
"Number of bytes to use as a cache of uncompressed data.");
DEFINE_bool(cache_index_and_filter_blocks, false,
"True if indexes/filters should be cached in block cache.");
DEFINE_bool(use_clock_cache, false,
"Replace default LRU block cache with clock cache.");
DEFINE_uint64(subcompactions, 1,
"Maximum number of subcompactions to divide L0-L1 compactions "
"into.");
DEFINE_uint64(periodic_compaction_seconds, 1000,
"Files older than this value will be picked up for compaction.");
DEFINE_uint64(compaction_ttl, 1000,
"Files older than TTL will be compacted to the next level.");
DEFINE_bool(allow_concurrent_memtable_write, false,
"Allow multi-writers to update mem tables in parallel.");
DEFINE_bool(enable_write_thread_adaptive_yield, true,
"Use a yielding spin loop for brief writer thread waits.");
static const bool FLAGS_subcompactions_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_subcompactions, &ValidateUint32Range);
static bool ValidateInt32Positive(const char* flagname, int32_t value) {
if (value < 0) {
fprintf(stderr, "Invalid value for --%s: %d, must be >=0\n", flagname,
value);
return false;
}
return true;
}
DEFINE_int32(reopen, 10, "Number of times database reopens");
static const bool FLAGS_reopen_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_reopen, &ValidateInt32Positive);
DEFINE_int32(bloom_bits, 10,
"Bloom filter bits per key. "
"Negative means use default settings.");
DEFINE_bool(use_block_based_filter, false,
"use block based filter"
"instead of full filter for block based table");
DEFINE_bool(partition_filters, false,
"use partitioned filters "
"for block-based table");
DEFINE_int32(
index_type,
static_cast<int32_t>(rocksdb::BlockBasedTableOptions::kBinarySearch),
"Type of block-based table index (see `enum IndexType` in table.h)");
DEFINE_string(db, "", "Use the db with the following name.");
DEFINE_string(secondaries_base, "",
"Use this path as the base path for secondary instances.");
DEFINE_bool(enable_secondary, false, "Enable secondary instance.");
DEFINE_string(
expected_values_path, "",
"File where the array of expected uint32_t values will be stored. If "
"provided and non-empty, the DB state will be verified against these "
"values after recovery. --max_key and --column_family must be kept the "
"same across invocations of this program that use the same "
"--expected_values_path.");
DEFINE_bool(verify_checksum, false,
"Verify checksum for every block read from storage");
DEFINE_bool(mmap_read, rocksdb::Options().allow_mmap_reads,
"Allow reads to occur via mmap-ing files");
DEFINE_bool(mmap_write, rocksdb::Options().allow_mmap_writes,
"Allow writes to occur via mmap-ing files");
DEFINE_bool(use_direct_reads, rocksdb::Options().use_direct_reads,
"Use O_DIRECT for reading data");
DEFINE_bool(use_direct_io_for_flush_and_compaction,
rocksdb::Options().use_direct_io_for_flush_and_compaction,
"Use O_DIRECT for writing data");
DEFINE_bool(statistics, false, "Create database statistics");
DEFINE_bool(sync, false, "Sync all writes to disk");
DEFINE_bool(use_fsync, false, "If true, issue fsync instead of fdatasync");
DEFINE_int32(kill_random_test, 0,
"If non-zero, kill at various points in source code with "
"probability 1/this");
static const bool FLAGS_kill_random_test_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_kill_random_test, &ValidateInt32Positive);
extern int rocksdb_kill_odds;
DEFINE_string(kill_prefix_blacklist, "",
"If non-empty, kill points with prefix in the list given will be"
" skipped. Items are comma-separated.");
extern std::vector<std::string> rocksdb_kill_prefix_blacklist;
DEFINE_bool(disable_wal, false, "If true, do not write WAL for write.");
DEFINE_uint64(recycle_log_file_num, rocksdb::Options().recycle_log_file_num,
"Number of old WAL files to keep around for later recycling");
DEFINE_int64(target_file_size_base, rocksdb::Options().target_file_size_base,
"Target level-1 file size for compaction");
DEFINE_int32(target_file_size_multiplier, 1,
"A multiplier to compute target level-N file size (N >= 2)");
DEFINE_uint64(max_bytes_for_level_base,
rocksdb::Options().max_bytes_for_level_base,
"Max bytes for level-1");
DEFINE_double(max_bytes_for_level_multiplier, 2,
"A multiplier to compute max bytes for level-N (N >= 2)");
DEFINE_int32(range_deletion_width, 10,
"The width of the range deletion intervals.");
DEFINE_uint64(rate_limiter_bytes_per_sec, 0, "Set options.rate_limiter value.");
DEFINE_bool(rate_limit_bg_reads, false,
"Use options.rate_limiter on compaction reads");
DEFINE_bool(use_txn, false,
"Use TransactionDB. Currently the default write policy is "
"TxnDBWritePolicy::WRITE_PREPARED");
DEFINE_int32(backup_one_in, 0,
"If non-zero, then CreateNewBackup() will be called once for "
"every N operations on average. 0 indicates CreateNewBackup() "
"is disabled.");
DEFINE_int32(checkpoint_one_in, 0,
"If non-zero, then CreateCheckpoint() will be called once for "
"every N operations on average. 0 indicates CreateCheckpoint() "
"is disabled.");
DEFINE_int32(ingest_external_file_one_in, 0,
"If non-zero, then IngestExternalFile() will be called once for "
"every N operations on average. 0 indicates IngestExternalFile() "
"is disabled.");
DEFINE_int32(ingest_external_file_width, 1000,
"The width of the ingested external files.");
DEFINE_int32(compact_files_one_in, 0,
"If non-zero, then CompactFiles() will be called once for every N "
"operations on average. 0 indicates CompactFiles() is disabled.");
DEFINE_int32(compact_range_one_in, 0,
"If non-zero, then CompactRange() will be called once for every N "
"operations on average. 0 indicates CompactRange() is disabled.");
DEFINE_int32(flush_one_in, 0,
"If non-zero, then Flush() will be called once for every N ops "
"on average. 0 indicates calls to Flush() are disabled.");
DEFINE_int32(compact_range_width, 10000,
"The width of the ranges passed to CompactRange().");
DEFINE_int32(acquire_snapshot_one_in, 0,
"If non-zero, then acquires a snapshot once every N operations on "
"average.");
DEFINE_bool(compare_full_db_state_snapshot, false,
"If set we compare state of entire db (in one of the threads) with"
"each snapshot.");
DEFINE_uint64(snapshot_hold_ops, 0,
"If non-zero, then releases snapshots N operations after they're "
"acquired.");
DEFINE_bool(use_multiget, false,
"If set, use the batched MultiGet API for reads");
static bool ValidateInt32Percent(const char* flagname, int32_t value) {
if (value < 0 || value > 100) {
fprintf(stderr, "Invalid value for --%s: %d, 0<= pct <=100 \n", flagname,
value);
return false;
}
return true;
}
DEFINE_int32(readpercent, 10,
"Ratio of reads to total workload (expressed as a percentage)");
static const bool FLAGS_readpercent_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_readpercent, &ValidateInt32Percent);
DEFINE_int32(prefixpercent, 20,
"Ratio of prefix iterators to total workload (expressed as a"
" percentage)");
static const bool FLAGS_prefixpercent_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_prefixpercent, &ValidateInt32Percent);
DEFINE_int32(writepercent, 45,
"Ratio of writes to total workload (expressed as a percentage)");
static const bool FLAGS_writepercent_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_writepercent, &ValidateInt32Percent);
DEFINE_int32(delpercent, 15,
"Ratio of deletes to total workload (expressed as a percentage)");
static const bool FLAGS_delpercent_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_delpercent, &ValidateInt32Percent);
DEFINE_int32(delrangepercent, 0,
"Ratio of range deletions to total workload (expressed as a "
"percentage). Cannot be used with test_batches_snapshots");
static const bool FLAGS_delrangepercent_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_delrangepercent, &ValidateInt32Percent);
DEFINE_int32(nooverwritepercent, 60,
"Ratio of keys without overwrite to total workload (expressed as "
" a percentage)");
static const bool FLAGS_nooverwritepercent_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_nooverwritepercent, &ValidateInt32Percent);
DEFINE_int32(iterpercent, 10,
"Ratio of iterations to total workload"
" (expressed as a percentage)");
static const bool FLAGS_iterpercent_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_iterpercent, &ValidateInt32Percent);
DEFINE_uint64(num_iterations, 10, "Number of iterations per MultiIterate run");
static const bool FLAGS_num_iterations_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_num_iterations, &ValidateUint32Range);
DEFINE_string(compression_type, "snappy",
"Algorithm to use to compress the database");
DEFINE_int32(compression_max_dict_bytes, 0,
"Maximum size of dictionary used to prime the compression "
"library.");
DEFINE_int32(compression_zstd_max_train_bytes, 0,
"Maximum size of training data passed to zstd's dictionary "
"trainer.");
DEFINE_string(checksum_type, "kCRC32c", "Algorithm to use to checksum blocks");
DEFINE_string(hdfs, "", "Name of hdfs environment");
DEFINE_string(env_uri, "",
"URI for env lookup. Mutually exclusive with --hdfs");
DEFINE_uint64(ops_per_thread, 1200000, "Number of operations per thread.");
static const bool FLAGS_ops_per_thread_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_ops_per_thread, &ValidateUint32Range);
DEFINE_uint64(log2_keys_per_lock, 2, "Log2 of number of keys per lock");
static const bool FLAGS_log2_keys_per_lock_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_log2_keys_per_lock, &ValidateUint32Range);
DEFINE_uint64(max_manifest_file_size, 16384, "Maximum size of a MANIFEST file");
DEFINE_bool(in_place_update, false, "On true, does inplace update in memtable");
DEFINE_int32(secondary_catch_up_one_in, 0,
"If non-zero, the secondaries attemp to catch up with the primary "
"once for every N operations on average. 0 indicates the "
"secondaries do not try to catch up after open.");
DEFINE_string(memtablerep, "skip_list", "");
inline static bool ValidatePrefixSize(const char* flagname, int32_t value) {
if (value < -1 || value > 8) {
fprintf(stderr, "Invalid value for --%s: %d. -1 <= PrefixSize <= 8\n",
flagname, value);
return false;
}
return true;
}
DEFINE_int32(prefix_size, 7,
"Control the prefix size for HashSkipListRep. "
"-1 is disabled.");
static const bool FLAGS_prefix_size_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_prefix_size, &ValidatePrefixSize);
DEFINE_bool(use_merge, false,
"On true, replaces all writes with a Merge "
"that behaves like a Put");
DEFINE_bool(use_full_merge_v1, false,
"On true, use a merge operator that implement the deprecated "
"version of FullMerge");
#endif // GFLAGS

142
db_stress_tool/db_stress_listener.h Normal file
View File

@ -0,0 +1,142 @@
// 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).
#ifdef GFLAGS
#pragma once
#include "rocksdb/listener.h"
#include "util/gflags_compat.h"
DECLARE_int32(compact_files_one_in);
namespace rocksdb {
class DbStressListener : public EventListener {
public:
DbStressListener(const std::string& db_name,
const std::vector<DbPath>& db_paths,
const std::vector<ColumnFamilyDescriptor>& column_families)
: db_name_(db_name),
db_paths_(db_paths),
column_families_(column_families),
num_pending_file_creations_(0) {}
virtual ~DbStressListener() { assert(num_pending_file_creations_ == 0); }
#ifndef ROCKSDB_LITE
virtual void OnFlushCompleted(DB* /*db*/, const FlushJobInfo& info) override {
assert(IsValidColumnFamilyName(info.cf_name));
VerifyFilePath(info.file_path);
// pretending doing some work here
std::this_thread::sleep_for(
std::chrono::microseconds(Random::GetTLSInstance()->Uniform(5000)));
}
virtual void OnCompactionCompleted(DB* /*db*/,
const CompactionJobInfo& ci) override {
assert(IsValidColumnFamilyName(ci.cf_name));
assert(ci.input_files.size() + ci.output_files.size() > 0U);
for (const auto& file_path : ci.input_files) {
VerifyFilePath(file_path);
}
for (const auto& file_path : ci.output_files) {
VerifyFilePath(file_path);
}
// pretending doing some work here
std::this_thread::sleep_for(
std::chrono::microseconds(Random::GetTLSInstance()->Uniform(5000)));
}
virtual void OnTableFileCreationStarted(
const TableFileCreationBriefInfo& /*info*/) override {
++num_pending_file_creations_;
}
virtual void OnTableFileCreated(const TableFileCreationInfo& info) override {
assert(info.db_name == db_name_);
assert(IsValidColumnFamilyName(info.cf_name));
if (info.file_size) {
VerifyFilePath(info.file_path);
}
assert(info.job_id > 0 || FLAGS_compact_files_one_in > 0);
if (info.status.ok() && info.file_size > 0) {
assert(info.table_properties.data_size > 0 ||
info.table_properties.num_range_deletions > 0);
assert(info.table_properties.raw_key_size > 0);
assert(info.table_properties.num_entries > 0);
}
--num_pending_file_creations_;
}
protected:
bool IsValidColumnFamilyName(const std::string& cf_name) const {
if (cf_name == kDefaultColumnFamilyName) {
return true;
}
// The column family names in the stress tests are numbers.
for (size_t i = 0; i < cf_name.size(); ++i) {
if (cf_name[i] < '0' || cf_name[i] > '9') {
return false;
}
}
return true;
}
void VerifyFileDir(const std::string& file_dir) {
#ifndef NDEBUG
if (db_name_ == file_dir) {
return;
}
for (const auto& db_path : db_paths_) {
if (db_path.path == file_dir) {
return;
}
}
for (auto& cf : column_families_) {
for (const auto& cf_path : cf.options.cf_paths) {
if (cf_path.path == file_dir) {
return;
}
}
}
assert(false);
#else
(void)file_dir;
#endif // !NDEBUG
}
void VerifyFileName(const std::string& file_name) {
#ifndef NDEBUG
uint64_t file_number;
FileType file_type;
bool result = ParseFileName(file_name, &file_number, &file_type);
assert(result);
assert(file_type == kTableFile);
#else
(void)file_name;
#endif // !NDEBUG
}
void VerifyFilePath(const std::string& file_path) {
#ifndef NDEBUG
size_t pos = file_path.find_last_of("/");
if (pos == std::string::npos) {
VerifyFileName(file_path);
} else {
if (pos > 0) {
VerifyFileDir(file_path.substr(0, pos));
}
VerifyFileName(file_path.substr(pos));
}
#else
(void)file_path;
#endif // !NDEBUG
}
#endif // !ROCKSDB_LITE
private:
std::string db_name_;
std::vector<DbPath> db_paths_;
std::vector<ColumnFamilyDescriptor> column_families_;
std::atomic<int> num_pending_file_creations_;
};
} // namespace rocksdb
#endif // GFLAGS

19
db_stress_tool/db_stress_shared_state.cc Normal file
View File

@ -0,0 +1,19 @@
// 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.
//
#ifdef GFLAGS
#include "db_stress_tool/db_stress_shared_state.h"
namespace rocksdb {
const uint32_t SharedState::UNKNOWN_SENTINEL = 0xfffffffe;
// indicates a key should definitely be deleted
const uint32_t SharedState::DELETION_SENTINEL = 0xffffffff;
} // namespace rocksdb
#endif // GFLAGS

370
db_stress_tool/db_stress_shared_state.h Normal file
View File

@ -0,0 +1,370 @@
// 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
#ifdef GFLAGS
#pragma once
#include "db_stress_tool/db_stress_stat.h"
#include "util/gflags_compat.h"
DECLARE_uint64(seed);
DECLARE_int64(max_key);
DECLARE_uint64(log2_keys_per_lock);
DECLARE_int32(threads);
DECLARE_int32(column_families);
DECLARE_int32(nooverwritepercent);
DECLARE_string(expected_values_path);
DECLARE_int32(clear_column_family_one_in);
DECLARE_bool(test_batches_snapshots);
namespace rocksdb {
class StressTest;
// State shared by all concurrent executions of the same benchmark.
class SharedState {
public:
// indicates a key may have any value (or not be present) as an operation on
// it is incomplete.
static const uint32_t UNKNOWN_SENTINEL;
// indicates a key should definitely be deleted
static const uint32_t DELETION_SENTINEL;
SharedState(Env* env, StressTest* stress_test)
: cv_(&mu_),
seed_(static_cast<uint32_t>(FLAGS_seed)),
max_key_(FLAGS_max_key),
log2_keys_per_lock_(static_cast<uint32_t>(FLAGS_log2_keys_per_lock)),
num_threads_(FLAGS_threads),
num_initialized_(0),
num_populated_(0),
vote_reopen_(0),
num_done_(0),
start_(false),
start_verify_(false),
should_stop_bg_thread_(false),
bg_thread_finished_(false),
stress_test_(stress_test),
verification_failure_(false),
no_overwrite_ids_(FLAGS_column_families),
values_(nullptr),
printing_verification_results_(false) {
// Pick random keys in each column family that will not experience
// overwrite
printf("Choosing random keys with no overwrite\n");
Random64 rnd(seed_);
// Start with the identity permutation. Subsequent iterations of
// for loop below will start with perm of previous for loop
int64_t* permutation = new int64_t[max_key_];
for (int64_t i = 0; i < max_key_; i++) {
permutation[i] = i;
}
// Now do the Knuth shuffle
int64_t num_no_overwrite_keys = (max_key_ * FLAGS_nooverwritepercent) / 100;
// Only need to figure out first num_no_overwrite_keys of permutation
no_overwrite_ids_.reserve(num_no_overwrite_keys);
for (int64_t i = 0; i < num_no_overwrite_keys; i++) {
int64_t rand_index = i + rnd.Next() % (max_key_ - i);
// Swap i and rand_index;
int64_t temp = permutation[i];
permutation[i] = permutation[rand_index];
permutation[rand_index] = temp;
// Fill no_overwrite_ids_ with the first num_no_overwrite_keys of
// permutation
no_overwrite_ids_.insert(permutation[i]);
}
delete[] permutation;
size_t expected_values_size =
sizeof(std::atomic<uint32_t>) * FLAGS_column_families * max_key_;
bool values_init_needed = false;
Status status;
if (!FLAGS_expected_values_path.empty()) {
if (!std::atomic<uint32_t>{}.is_lock_free()) {
status = Status::InvalidArgument(
"Cannot use --expected_values_path on platforms without lock-free "
"std::atomic<uint32_t>");
}
if (status.ok() && FLAGS_clear_column_family_one_in > 0) {
status = Status::InvalidArgument(
"Cannot use --expected_values_path on when "
"--clear_column_family_one_in is greater than zero.");
}
uint64_t size = 0;
if (status.ok()) {
status = env->GetFileSize(FLAGS_expected_values_path, &size);
}
std::unique_ptr<WritableFile> wfile;
if (status.ok() && size == 0) {
const EnvOptions soptions;
status =
env->NewWritableFile(FLAGS_expected_values_path, &wfile, soptions);
}
if (status.ok() && size == 0) {
std::string buf(expected_values_size, '\0');
status = wfile->Append(buf);
values_init_needed = true;
}
if (status.ok()) {
status = env->NewMemoryMappedFileBuffer(FLAGS_expected_values_path,
&expected_mmap_buffer_);
}
if (status.ok()) {
assert(expected_mmap_buffer_->GetLen() == expected_values_size);
values_ = static_cast<std::atomic<uint32_t>*>(
expected_mmap_buffer_->GetBase());
assert(values_ != nullptr);
} else {
fprintf(stderr, "Failed opening shared file '%s' with error: %s\n",
FLAGS_expected_values_path.c_str(), status.ToString().c_str());
assert(values_ == nullptr);
}
}
if (values_ == nullptr) {
values_allocation_.reset(
new std::atomic<uint32_t>[FLAGS_column_families * max_key_]);
values_ = &values_allocation_[0];
values_init_needed = true;
}
assert(values_ != nullptr);
if (values_init_needed) {
for (int i = 0; i < FLAGS_column_families; ++i) {
for (int j = 0; j < max_key_; ++j) {
Delete(i, j, false /* pending */);
}
}
}
if (FLAGS_test_batches_snapshots) {
fprintf(stdout, "No lock creation because test_batches_snapshots set\n");
return;
}
long num_locks = static_cast<long>(max_key_ >> log2_keys_per_lock_);
if (max_key_ & ((1 << log2_keys_per_lock_) - 1)) {
num_locks++;
}
fprintf(stdout, "Creating %ld locks\n", num_locks * FLAGS_column_families);
key_locks_.resize(FLAGS_column_families);
for (int i = 0; i < FLAGS_column_families; ++i) {
key_locks_[i].resize(num_locks);
for (auto& ptr : key_locks_[i]) {
ptr.reset(new port::Mutex);
}
}
}
~SharedState() {}
port::Mutex* GetMutex() { return &mu_; }
port::CondVar* GetCondVar() { return &cv_; }
StressTest* GetStressTest() const { return stress_test_; }
int64_t GetMaxKey() const { return max_key_; }
uint32_t GetNumThreads() const { return num_threads_; }
void IncInitialized() { num_initialized_++; }
void IncOperated() { num_populated_++; }
void IncDone() { num_done_++; }
void IncVotedReopen() { vote_reopen_ = (vote_reopen_ + 1) % num_threads_; }
bool AllInitialized() const { return num_initialized_ >= num_threads_; }
bool AllOperated() const { return num_populated_ >= num_threads_; }
bool AllDone() const { return num_done_ >= num_threads_; }
bool AllVotedReopen() { return (vote_reopen_ == 0); }
void SetStart() { start_ = true; }
void SetStartVerify() { start_verify_ = true; }
bool Started() const { return start_; }
bool VerifyStarted() const { return start_verify_; }
void SetVerificationFailure() { verification_failure_.store(true); }
bool HasVerificationFailedYet() { return verification_failure_.load(); }
port::Mutex* GetMutexForKey(int cf, int64_t key) {
return key_locks_[cf][key >> log2_keys_per_lock_].get();
}
void LockColumnFamily(int cf) {
for (auto& mutex : key_locks_[cf]) {
mutex->Lock();
}
}
void UnlockColumnFamily(int cf) {
for (auto& mutex : key_locks_[cf]) {
mutex->Unlock();
}
}
std::atomic<uint32_t>& Value(int cf, int64_t key) const {
return values_[cf * max_key_ + key];
}
void ClearColumnFamily(int cf) {
std::fill(&Value(cf, 0 /* key */), &Value(cf + 1, 0 /* key */),
DELETION_SENTINEL);
}
// @param pending True if the update may have started but is not yet
// guaranteed finished. This is useful for crash-recovery testing when the
// process may crash before updating the expected values array.
void Put(int cf, int64_t key, uint32_t value_base, bool pending) {
if (!pending) {
// prevent expected-value update from reordering before Write
std::atomic_thread_fence(std::memory_order_release);
}
Value(cf, key).store(pending ? UNKNOWN_SENTINEL : value_base,
std::memory_order_relaxed);
if (pending) {
// prevent Write from reordering before expected-value update
std::atomic_thread_fence(std::memory_order_release);
}
}
uint32_t Get(int cf, int64_t key) const { return Value(cf, key); }
// @param pending See comment above Put()
// Returns true if the key was not yet deleted.
bool Delete(int cf, int64_t key, bool pending) {
if (Value(cf, key) == DELETION_SENTINEL) {
return false;
}
Put(cf, key, DELETION_SENTINEL, pending);
return true;
}
// @param pending See comment above Put()
// Returns true if the key was not yet deleted.
bool SingleDelete(int cf, int64_t key, bool pending) {
return Delete(cf, key, pending);
}
// @param pending See comment above Put()
// Returns number of keys deleted by the call.
int DeleteRange(int cf, int64_t begin_key, int64_t end_key, bool pending) {
int covered = 0;
for (int64_t key = begin_key; key < end_key; ++key) {
if (Delete(cf, key, pending)) {
++covered;
}
}
return covered;
}
bool AllowsOverwrite(int64_t key) {
return no_overwrite_ids_.find(key) == no_overwrite_ids_.end();
}
bool Exists(int cf, int64_t key) {
// UNKNOWN_SENTINEL counts as exists. That assures a key for which overwrite
// is disallowed can't be accidentally added a second time, in which case
// SingleDelete wouldn't be able to properly delete the key. It does allow
// the case where a SingleDelete might be added which covers nothing, but
// that's not a correctness issue.
uint32_t expected_value = Value(cf, key).load();
return expected_value != DELETION_SENTINEL;
}
uint32_t GetSeed() const { return seed_; }
void SetShouldStopBgThread() { should_stop_bg_thread_ = true; }
bool ShoudStopBgThread() { return should_stop_bg_thread_; }
void SetBgThreadFinish() { bg_thread_finished_ = true; }
bool BgThreadFinished() const { return bg_thread_finished_; }
bool ShouldVerifyAtBeginning() const {
return expected_mmap_buffer_.get() != nullptr;
}
bool PrintingVerificationResults() {
bool tmp = false;
return !printing_verification_results_.compare_exchange_strong(
tmp, true, std::memory_order_relaxed);
}
void FinishPrintingVerificationResults() {
printing_verification_results_.store(false, std::memory_order_relaxed);
}
private:
port::Mutex mu_;
port::CondVar cv_;
const uint32_t seed_;
const int64_t max_key_;
const uint32_t log2_keys_per_lock_;
const int num_threads_;
long num_initialized_;
long num_populated_;
long vote_reopen_;
long num_done_;
bool start_;
bool start_verify_;
bool should_stop_bg_thread_;
bool bg_thread_finished_;
StressTest* stress_test_;
std::atomic<bool> verification_failure_;
// Keys that should not be overwritten
std::unordered_set<size_t> no_overwrite_ids_;
std::atomic<uint32_t>* values_;
std::unique_ptr<std::atomic<uint32_t>[]> values_allocation_;
// Has to make it owned by a smart ptr as port::Mutex is not copyable
// and storing it in the container may require copying depending on the impl.
std::vector<std::vector<std::unique_ptr<port::Mutex>>> key_locks_;
std::unique_ptr<MemoryMappedFileBuffer> expected_mmap_buffer_;
std::atomic<bool> printing_verification_results_;
};
// Per-thread state for concurrent executions of the same benchmark.
struct ThreadState {
uint32_t tid; // 0..n-1
Random rand; // Has different seeds for different threads
SharedState* shared;
Stats stats;
struct SnapshotState {
const Snapshot* snapshot;
// The cf from which we did a Get at this snapshot
int cf_at;
// The name of the cf at the time that we did a read
std::string cf_at_name;
// The key with which we did a Get at this snapshot
std::string key;
// The status of the Get
Status status;
// The value of the Get
std::string value;
// optional state of all keys in the db
std::vector<bool>* key_vec;
};
std::queue<std::pair<uint64_t, SnapshotState>> snapshot_queue;
ThreadState(uint32_t index, SharedState* _shared)
: tid(index), rand(1000 + index + _shared->GetSeed()), shared(_shared) {}
};
} // namespace rocksdb
#endif // GFLAGS

213
db_stress_tool/db_stress_stat.h Normal file
View File

@ -0,0 +1,213 @@
// 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).
#pragma once
#include <cinttypes>
#include <memory>
#include <queue>
#include <unordered_set>
#include "monitoring/histogram.h"
#include "port/port.h"
#include "rocksdb/env.h"
#include "rocksdb/snapshot.h"
#include "rocksdb/statistics.h"
#include "util/gflags_compat.h"
#include "util/random.h"
DECLARE_bool(histogram);
DECLARE_bool(progress_reports);
namespace rocksdb {
// Database statistics
static std::shared_ptr<rocksdb::Statistics> dbstats;
static std::shared_ptr<rocksdb::Statistics> dbstats_secondaries;
class Stats {
private:
uint64_t start_;
uint64_t finish_;
double seconds_;
long done_;
long gets_;
long prefixes_;
long writes_;
long deletes_;
size_t single_deletes_;
long iterator_size_sums_;
long founds_;
long iterations_;
long range_deletions_;
long covered_by_range_deletions_;
long errors_;
long num_compact_files_succeed_;
long num_compact_files_failed_;
int next_report_;
size_t bytes_;
uint64_t last_op_finish_;
HistogramImpl hist_;
public:
Stats() {}
void Start() {
next_report_ = 100;
hist_.Clear();
done_ = 0;
gets_ = 0;
prefixes_ = 0;
writes_ = 0;
deletes_ = 0;
single_deletes_ = 0;
iterator_size_sums_ = 0;
founds_ = 0;
iterations_ = 0;
range_deletions_ = 0;
covered_by_range_deletions_ = 0;
errors_ = 0;
bytes_ = 0;
seconds_ = 0;
num_compact_files_succeed_ = 0;
num_compact_files_failed_ = 0;
start_ = Env::Default()->NowMicros();
last_op_finish_ = start_;
finish_ = start_;
}
void Merge(const Stats& other) {
hist_.Merge(other.hist_);
done_ += other.done_;
gets_ += other.gets_;
prefixes_ += other.prefixes_;
writes_ += other.writes_;
deletes_ += other.deletes_;
single_deletes_ += other.single_deletes_;
iterator_size_sums_ += other.iterator_size_sums_;
founds_ += other.founds_;
iterations_ += other.iterations_;
range_deletions_ += other.range_deletions_;
covered_by_range_deletions_ = other.covered_by_range_deletions_;
errors_ += other.errors_;
bytes_ += other.bytes_;
seconds_ += other.seconds_;
num_compact_files_succeed_ += other.num_compact_files_succeed_;
num_compact_files_failed_ += other.num_compact_files_failed_;
if (other.start_ < start_) start_ = other.start_;
if (other.finish_ > finish_) finish_ = other.finish_;
}
void Stop() {
finish_ = Env::Default()->NowMicros();
seconds_ = (finish_ - start_) * 1e-6;
}
void FinishedSingleOp() {
if (FLAGS_histogram) {
auto now = Env::Default()->NowMicros();
auto micros = now - last_op_finish_;
hist_.Add(micros);
if (micros > 20000) {
fprintf(stdout, "long op: %" PRIu64 " micros%30s\r", micros, "");
}
last_op_finish_ = now;
}
done_++;
if (FLAGS_progress_reports) {
if (done_ >= next_report_) {
if (next_report_ < 1000)
next_report_ += 100;
else if (next_report_ < 5000)
next_report_ += 500;
else if (next_report_ < 10000)
next_report_ += 1000;
else if (next_report_ < 50000)
next_report_ += 5000;
else if (next_report_ < 100000)
next_report_ += 10000;
else if (next_report_ < 500000)
next_report_ += 50000;
else
next_report_ += 100000;
fprintf(stdout, "... finished %ld ops%30s\r", done_, "");
}
}
}
void AddBytesForWrites(long nwrites, size_t nbytes) {
writes_ += nwrites;
bytes_ += nbytes;
}
void AddGets(long ngets, long nfounds) {
founds_ += nfounds;
gets_ += ngets;
}
void AddPrefixes(long nprefixes, long count) {
prefixes_ += nprefixes;
iterator_size_sums_ += count;
}
void AddIterations(long n) { iterations_ += n; }
void AddDeletes(long n) { deletes_ += n; }
void AddSingleDeletes(size_t n) { single_deletes_ += n; }
void AddRangeDeletions(long n) { range_deletions_ += n; }
void AddCoveredByRangeDeletions(long n) { covered_by_range_deletions_ += n; }
void AddErrors(long n) { errors_ += n; }
void AddNumCompactFilesSucceed(long n) { num_compact_files_succeed_ += n; }
void AddNumCompactFilesFailed(long n) { num_compact_files_failed_ += n; }
void Report(const char* name) {
std::string extra;
if (bytes_ < 1 || done_ < 1) {
fprintf(stderr, "No writes or ops?\n");
return;
}
double elapsed = (finish_ - start_) * 1e-6;
double bytes_mb = bytes_ / 1048576.0;
double rate = bytes_mb / elapsed;
double throughput = (double)done_ / elapsed;
fprintf(stdout, "%-12s: ", name);
fprintf(stdout, "%.3f micros/op %ld ops/sec\n", seconds_ * 1e6 / done_,
(long)throughput);
fprintf(stdout, "%-12s: Wrote %.2f MB (%.2f MB/sec) (%ld%% of %ld ops)\n",
"", bytes_mb, rate, (100 * writes_) / done_, done_);
fprintf(stdout, "%-12s: Wrote %ld times\n", "", writes_);
fprintf(stdout, "%-12s: Deleted %ld times\n", "", deletes_);
fprintf(stdout, "%-12s: Single deleted %" ROCKSDB_PRIszt " times\n", "",
single_deletes_);
fprintf(stdout, "%-12s: %ld read and %ld found the key\n", "", gets_,
founds_);
fprintf(stdout, "%-12s: Prefix scanned %ld times\n", "", prefixes_);
fprintf(stdout, "%-12s: Iterator size sum is %ld\n", "",
iterator_size_sums_);
fprintf(stdout, "%-12s: Iterated %ld times\n", "", iterations_);
fprintf(stdout, "%-12s: Deleted %ld key-ranges\n", "", range_deletions_);
fprintf(stdout, "%-12s: Range deletions covered %ld keys\n", "",
covered_by_range_deletions_);
fprintf(stdout, "%-12s: Got errors %ld times\n", "", errors_);
fprintf(stdout, "%-12s: %ld CompactFiles() succeed\n", "",
num_compact_files_succeed_);
fprintf(stdout, "%-12s: %ld CompactFiles() did not succeed\n", "",
num_compact_files_failed_);
if (FLAGS_histogram) {
fprintf(stdout, "Microseconds per op:\n%s\n", hist_.ToString().c_str());
}
fflush(stdout);
}
};
} // namespace rocksdb

1681
db_stress_tool/db_stress_test_base.cc Normal file
View File

File diff suppressed because it is too large Load Diff

186
db_stress_tool/db_stress_test_base.h Normal file
View File

@ -0,0 +1,186 @@
// 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.
#ifdef GFLAGS
#pragma once
#include "db_stress_tool/db_stress_common.h"
#include "db_stress_tool/db_stress_shared_state.h"
namespace rocksdb {
class Transaction;
class TransactionDB;
class StressTest {
public:
StressTest();
virtual ~StressTest();
std::shared_ptr<Cache> NewCache(size_t capacity);
bool BuildOptionsTable();
void InitDb();
void InitReadonlyDb(SharedState*);
// Return false if verification fails.
bool VerifySecondaries();
void OperateDb(ThreadState* thread);
virtual void VerifyDb(ThreadState* thread) const = 0;
void PrintStatistics();
protected:
Status AssertSame(DB* db, ColumnFamilyHandle* cf,
ThreadState::SnapshotState& snap_state);
// Currently PreloadDb has to be single-threaded.
void PreloadDbAndReopenAsReadOnly(int64_t number_of_keys,
SharedState* shared);
Status SetOptions(ThreadState* thread);
#ifndef ROCKSDB_LITE
Status NewTxn(WriteOptions& write_opts, Transaction** txn);
Status CommitTxn(Transaction* txn);
#endif
virtual void MaybeClearOneColumnFamily(ThreadState* /* thread */) {}
virtual bool ShouldAcquireMutexOnKey() const { return false; }
virtual std::vector<int> GenerateColumnFamilies(
const int /* num_column_families */, int rand_column_family) const {
return {rand_column_family};
}
virtual std::vector<int64_t> GenerateKeys(int64_t rand_key) const {
return {rand_key};
}
virtual Status TestGet(ThreadState* thread, const ReadOptions& read_opts,
const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys) = 0;
virtual std::vector<Status> TestMultiGet(
ThreadState* thread, const ReadOptions& read_opts,
const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys) = 0;
virtual Status TestPrefixScan(ThreadState* thread,
const ReadOptions& read_opts,
const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys) = 0;
virtual Status TestPut(ThreadState* thread, WriteOptions& write_opts,
const ReadOptions& read_opts,
const std::vector<int>& cf_ids,
const std::vector<int64_t>& keys, char (&value)[100],
std::unique_ptr<MutexLock>& lock) = 0;
virtual Status TestDelete(ThreadState* thread, WriteOptions& write_opts,
const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys,
std::unique_ptr<MutexLock>& lock) = 0;
virtual Status TestDeleteRange(ThreadState* thread, WriteOptions& write_opts,
const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys,
std::unique_ptr<MutexLock>& lock) = 0;
virtual void TestIngestExternalFile(
ThreadState* thread, const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys,
std::unique_ptr<MutexLock>& lock) = 0;
// Return a column family handle that mirrors what is pointed by
// `column_family_id`, which will be used to validate data to be correct.
// By default, the column family itself will be returned.
virtual ColumnFamilyHandle* GetControlCfh(ThreadState* /* thread*/,
int column_family_id) {
return column_families_[column_family_id];
}
#ifndef ROCKSDB_LITE
// Generated a list of keys that close to boundaries of SST keys.
// If there isn't any SST file in the DB, return empty list.
std::vector<std::string> GetWhiteBoxKeys(ThreadState* thread, DB* db,
ColumnFamilyHandle* cfh,
size_t num_keys);
#else // !ROCKSDB_LITE
std::vector<std::string> GetWhiteBoxKeys(ThreadState*, DB*,
ColumnFamilyHandle*, size_t) {
// Not supported in LITE mode.
return {};
}
#endif // !ROCKSDB_LITE
// Given a key K, this creates an iterator which scans to K and then
// does a random sequence of Next/Prev operations.
virtual Status TestIterate(ThreadState* thread, const ReadOptions& read_opts,
const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys);
// Enum used by VerifyIterator() to identify the mode to validate.
enum LastIterateOp { kLastOpSeek, kLastOpSeekForPrev, kLastOpNextOrPrev };
// Compare the two iterator, iter and cmp_iter are in the same position,
// unless iter might be made invalidate or undefined because of
// upper or lower bounds, or prefix extractor.
// Will flag failure if the verification fails.
// diverged = true if the two iterator is already diverged.
// True if verification passed, false if not.
void VerifyIterator(ThreadState* thread, ColumnFamilyHandle* cmp_cfh,
const ReadOptions& ro, Iterator* iter, Iterator* cmp_iter,
LastIterateOp op, const Slice& seek_key, bool* diverged);
virtual Status TestBackupRestore(ThreadState* thread,
const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys);
virtual Status TestCheckpoint(ThreadState* thread,
const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys);
void VerificationAbort(SharedState* shared, std::string msg, Status s) const;
void VerificationAbort(SharedState* shared, std::string msg, int cf,
int64_t key) const;
void PrintEnv() const;
void Open();
void Reopen();
std::shared_ptr<Cache> cache_;
std::shared_ptr<Cache> compressed_cache_;
std::shared_ptr<const FilterPolicy> filter_policy_;
DB* db_;
#ifndef ROCKSDB_LITE
TransactionDB* txn_db_;
#endif
Options options_;
std::vector<ColumnFamilyHandle*> column_families_;
std::vector<std::string> column_family_names_;
std::atomic<int> new_column_family_name_;
int num_times_reopened_;
std::unordered_map<std::string, std::vector<std::string>> options_table_;
std::vector<std::string> options_index_;
std::atomic<bool> db_preload_finished_;
// Fields used for stress-testing secondary instance in the same process
std::vector<DB*> secondaries_;
std::vector<std::vector<ColumnFamilyHandle*>> secondary_cfh_lists_;
};
} // namespace rocksdb
#endif // GFLAGS

193
db_stress_tool/db_stress_tool.cc Normal file
View File

@ -0,0 +1,193 @@
// 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.
//
// The test uses an array to compare against values written to the database.
// Keys written to the array are in 1:1 correspondence to the actual values in
// the database according to the formula in the function GenerateValue.
// Space is reserved in the array from 0 to FLAGS_max_key and values are
// randomly written/deleted/read from those positions. During verification we
// compare all the positions in the array. To shorten/elongate the running
// time, you could change the settings: FLAGS_max_key, FLAGS_ops_per_thread,
// (sometimes also FLAGS_threads).
//
// NOTE that if FLAGS_test_batches_snapshots is set, the test will have
// different behavior. See comment of the flag for details.
#ifdef GFLAGS
#include "db_stress_tool/db_stress_common.h"
#include "db_stress_tool/db_stress_driver.h"
namespace rocksdb {
int db_stress_tool(int argc, char** argv) {
SetUsageMessage(std::string("\nUSAGE:\n") + std::string(argv[0]) +
" [OPTIONS]...");
ParseCommandLineFlags(&argc, &argv, true);
if (FLAGS_statistics) {
dbstats = rocksdb::CreateDBStatistics();
if (FLAGS_enable_secondary) {
dbstats_secondaries = rocksdb::CreateDBStatistics();
}
}
FLAGS_compression_type_e =
StringToCompressionType(FLAGS_compression_type.c_str());
FLAGS_checksum_type_e = StringToChecksumType(FLAGS_checksum_type.c_str());
if (!FLAGS_hdfs.empty()) {
if (!FLAGS_env_uri.empty()) {
fprintf(stderr, "Cannot specify both --hdfs and --env_uri.\n");
exit(1);
}
FLAGS_env = new rocksdb::HdfsEnv(FLAGS_hdfs);
} else if (!FLAGS_env_uri.empty()) {
Status s = Env::LoadEnv(FLAGS_env_uri, &FLAGS_env, &env_guard);
if (FLAGS_env == nullptr) {
fprintf(stderr, "No Env registered for URI: %s\n", FLAGS_env_uri.c_str());
exit(1);
}
}
FLAGS_rep_factory = StringToRepFactory(FLAGS_memtablerep.c_str());
// The number of background threads should be at least as much the
// max number of concurrent compactions.
FLAGS_env->SetBackgroundThreads(FLAGS_max_background_compactions);
FLAGS_env->SetBackgroundThreads(FLAGS_num_bottom_pri_threads,
rocksdb::Env::Priority::BOTTOM);
if (FLAGS_prefixpercent > 0 && FLAGS_prefix_size < 0) {
fprintf(stderr,
"Error: prefixpercent is non-zero while prefix_size is "
"not positive!\n");
exit(1);
}
if (FLAGS_test_batches_snapshots && FLAGS_prefix_size <= 0) {
fprintf(stderr,
"Error: please specify prefix_size for "
"test_batches_snapshots test!\n");
exit(1);
}
if (FLAGS_memtable_prefix_bloom_size_ratio > 0.0 && FLAGS_prefix_size < 0) {
fprintf(stderr,
"Error: please specify positive prefix_size in order to use "
"memtable_prefix_bloom_size_ratio\n");
exit(1);
}
if ((FLAGS_readpercent + FLAGS_prefixpercent + FLAGS_writepercent +
FLAGS_delpercent + FLAGS_delrangepercent + FLAGS_iterpercent) != 100) {
fprintf(stderr,
"Error: Read+Prefix+Write+Delete+DeleteRange+Iterate percents != "
"100!\n");
exit(1);
}
if (FLAGS_disable_wal == 1 && FLAGS_reopen > 0) {
fprintf(stderr, "Error: Db cannot reopen safely with disable_wal set!\n");
exit(1);
}
if ((unsigned)FLAGS_reopen >= FLAGS_ops_per_thread) {
fprintf(stderr,
"Error: #DB-reopens should be < ops_per_thread\n"
"Provided reopens = %d and ops_per_thread = %lu\n",
FLAGS_reopen, (unsigned long)FLAGS_ops_per_thread);
exit(1);
}
if (FLAGS_test_batches_snapshots && FLAGS_delrangepercent > 0) {
fprintf(stderr,
"Error: nonzero delrangepercent unsupported in "
"test_batches_snapshots mode\n");
exit(1);
}
if (FLAGS_active_width > FLAGS_max_key) {
fprintf(stderr, "Error: active_width can be at most max_key\n");
exit(1);
} else if (FLAGS_active_width == 0) {
FLAGS_active_width = FLAGS_max_key;
}
if (FLAGS_value_size_mult * kRandomValueMaxFactor > kValueMaxLen) {
fprintf(stderr, "Error: value_size_mult can be at most %d\n",
kValueMaxLen / kRandomValueMaxFactor);
exit(1);
}
if (FLAGS_use_merge && FLAGS_nooverwritepercent == 100) {
fprintf(
stderr,
"Error: nooverwritepercent must not be 100 when using merge operands");
exit(1);
}
if (FLAGS_ingest_external_file_one_in > 0 && FLAGS_nooverwritepercent > 0) {
fprintf(stderr,
"Error: nooverwritepercent must be 0 when using file ingestion\n");
exit(1);
}
if (FLAGS_clear_column_family_one_in > 0 && FLAGS_backup_one_in > 0) {
fprintf(stderr,
"Error: clear_column_family_one_in must be 0 when using backup\n");
exit(1);
}
if (FLAGS_test_cf_consistency && FLAGS_disable_wal) {
FLAGS_atomic_flush = true;
}
if (FLAGS_read_only) {
if (FLAGS_writepercent != 0 || FLAGS_delpercent != 0 ||
FLAGS_delrangepercent != 0) {
fprintf(stderr, "Error: updates are not supported in read only mode\n");
exit(1);
} else if (FLAGS_checkpoint_one_in > 0 &&
FLAGS_clear_column_family_one_in > 0) {
fprintf(stdout,
"Warn: checkpoint won't be validated since column families may "
"be dropped.\n");
}
}
// Choose a location for the test database if none given with --db=<path>
if (FLAGS_db.empty()) {
std::string default_db_path;
FLAGS_env->GetTestDirectory(&default_db_path);
default_db_path += "/dbstress";
FLAGS_db = default_db_path;
}
if (FLAGS_enable_secondary && FLAGS_secondaries_base.empty()) {
std::string default_secondaries_path;
FLAGS_env->GetTestDirectory(&default_secondaries_path);
default_secondaries_path += "/dbstress_secondaries";
rocksdb::Status s = FLAGS_env->CreateDirIfMissing(default_secondaries_path);
if (!s.ok()) {
fprintf(stderr, "Failed to create directory %s: %s\n",
default_secondaries_path.c_str(), s.ToString().c_str());
exit(1);
}
FLAGS_secondaries_base = default_secondaries_path;
}
if (!FLAGS_enable_secondary && FLAGS_secondary_catch_up_one_in > 0) {
fprintf(stderr, "Secondary instance is disabled.\n");
exit(1);
}
rocksdb_kill_odds = FLAGS_kill_random_test;
rocksdb_kill_prefix_blacklist = SplitString(FLAGS_kill_prefix_blacklist);
std::unique_ptr<rocksdb::StressTest> stress;
if (FLAGS_test_cf_consistency) {
stress.reset(CreateCfConsistencyStressTest());
} else if (FLAGS_test_batches_snapshots) {
stress.reset(CreateBatchedOpsStressTest());
} else {
stress.reset(CreateNonBatchedOpsStressTest());
}
if (RunStressTest(stress.get())) {
return 0;
} else {
return 1;
}
}
} // namespace rocksdb
#endif // GFLAGS

553
db_stress_tool/no_batched_ops_stress.cc Normal file
View File

@ -0,0 +1,553 @@
// 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.
#ifdef GFLAGS
#include "db_stress_tool/db_stress_common.h"
namespace rocksdb {
class NonBatchedOpsStressTest : public StressTest {
public:
NonBatchedOpsStressTest() {}
virtual ~NonBatchedOpsStressTest() {}
virtual void VerifyDb(ThreadState* thread) const {
ReadOptions options(FLAGS_verify_checksum, true);
auto shared = thread->shared;
const int64_t max_key = shared->GetMaxKey();
const int64_t keys_per_thread = max_key / shared->GetNumThreads();
int64_t start = keys_per_thread * thread->tid;
int64_t end = start + keys_per_thread;
uint64_t prefix_to_use =
(FLAGS_prefix_size < 0) ? 1 : static_cast<size_t>(FLAGS_prefix_size);
if (thread->tid == shared->GetNumThreads() - 1) {
end = max_key;
}
for (size_t cf = 0; cf < column_families_.size(); ++cf) {
if (thread->shared->HasVerificationFailedYet()) {
break;
}
if (!thread->rand.OneIn(2)) {
// Use iterator to verify this range
std::unique_ptr<Iterator> iter(
db_->NewIterator(options, column_families_[cf]));
iter->Seek(Key(start));
for (auto i = start; i < end; i++) {
if (thread->shared->HasVerificationFailedYet()) {
break;
}
// TODO(ljin): update "long" to uint64_t
// Reseek when the prefix changes
if (prefix_to_use > 0 &&
i % (static_cast<int64_t>(1) << 8 * (8 - prefix_to_use)) == 0) {
iter->Seek(Key(i));
}
std::string from_db;
std::string keystr = Key(i);
Slice k = keystr;
Status s = iter->status();
if (iter->Valid()) {
if (iter->key().compare(k) > 0) {
s = Status::NotFound(Slice());
} else if (iter->key().compare(k) == 0) {
from_db = iter->value().ToString();
iter->Next();
} else if (iter->key().compare(k) < 0) {
VerificationAbort(shared, "An out of range key was found",
static_cast<int>(cf), i);
}
} else {
// The iterator found no value for the key in question, so do not
// move to the next item in the iterator
s = Status::NotFound(Slice());
}
VerifyValue(static_cast<int>(cf), i, options, shared, from_db, s,
true);
if (from_db.length()) {
PrintKeyValue(static_cast<int>(cf), static_cast<uint32_t>(i),
from_db.data(), from_db.length());
}
}
} else {
// Use Get to verify this range
for (auto i = start; i < end; i++) {
if (thread->shared->HasVerificationFailedYet()) {
break;
}
std::string from_db;
std::string keystr = Key(i);
Slice k = keystr;
Status s = db_->Get(options, column_families_[cf], k, &from_db);
VerifyValue(static_cast<int>(cf), i, options, shared, from_db, s,
true);
if (from_db.length()) {
PrintKeyValue(static_cast<int>(cf), static_cast<uint32_t>(i),
from_db.data(), from_db.length());
}
}
}
}
}
virtual void MaybeClearOneColumnFamily(ThreadState* thread) {
if (FLAGS_clear_column_family_one_in != 0 && FLAGS_column_families > 1) {
if (thread->rand.OneIn(FLAGS_clear_column_family_one_in)) {
// drop column family and then create it again (can't drop default)
int cf = thread->rand.Next() % (FLAGS_column_families - 1) + 1;
std::string new_name = ToString(new_column_family_name_.fetch_add(1));
{
MutexLock l(thread->shared->GetMutex());
fprintf(
stdout,
"[CF %d] Dropping and recreating column family. new name: %s\n",
cf, new_name.c_str());
}
thread->shared->LockColumnFamily(cf);
Status s = db_->DropColumnFamily(column_families_[cf]);
delete column_families_[cf];
if (!s.ok()) {
fprintf(stderr, "dropping column family error: %s\n",
s.ToString().c_str());
std::terminate();
}
s = db_->CreateColumnFamily(ColumnFamilyOptions(options_), new_name,
&column_families_[cf]);
column_family_names_[cf] = new_name;
thread->shared->ClearColumnFamily(cf);
if (!s.ok()) {
fprintf(stderr, "creating column family error: %s\n",
s.ToString().c_str());
std::terminate();
}
thread->shared->UnlockColumnFamily(cf);
}
}
}
virtual bool ShouldAcquireMutexOnKey() const { return true; }
virtual Status TestGet(ThreadState* thread, const ReadOptions& read_opts,
const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys) {
auto cfh = column_families_[rand_column_families[0]];
std::string key_str = Key(rand_keys[0]);
Slice key = key_str;
std::string from_db;
Status s = db_->Get(read_opts, cfh, key, &from_db);
if (s.ok()) {
// found case
thread->stats.AddGets(1, 1);
} else if (s.IsNotFound()) {
// not found case
thread->stats.AddGets(1, 0);
} else {
// errors case
thread->stats.AddErrors(1);
}
return s;
}
virtual std::vector<Status> TestMultiGet(
ThreadState* thread, const ReadOptions& read_opts,
const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys) {
size_t num_keys = rand_keys.size();
std::vector<std::string> key_str;
std::vector<Slice> keys;
key_str.reserve(num_keys);
keys.reserve(num_keys);
std::vector<PinnableSlice> values(num_keys);
std::vector<Status> statuses(num_keys);
ColumnFamilyHandle* cfh = column_families_[rand_column_families[0]];
for (size_t i = 0; i < num_keys; ++i) {
key_str.emplace_back(Key(rand_keys[i]));
keys.emplace_back(key_str.back());
}
db_->MultiGet(read_opts, cfh, num_keys, keys.data(), values.data(),
statuses.data());
for (const auto& s : statuses) {
if (s.ok()) {
// found case
thread->stats.AddGets(1, 1);
} else if (s.IsNotFound()) {
// not found case
thread->stats.AddGets(1, 0);
} else {
// errors case
thread->stats.AddErrors(1);
}
}
return statuses;
}
virtual Status TestPrefixScan(ThreadState* thread,
const ReadOptions& read_opts,
const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys) {
auto cfh = column_families_[rand_column_families[0]];
std::string key_str = Key(rand_keys[0]);
Slice key = key_str;
Slice prefix = Slice(key.data(), FLAGS_prefix_size);
std::string upper_bound;
Slice ub_slice;
ReadOptions ro_copy = read_opts;
if (thread->rand.OneIn(2) && GetNextPrefix(prefix, &upper_bound)) {
// For half of the time, set the upper bound to the next prefix
ub_slice = Slice(upper_bound);
ro_copy.iterate_upper_bound = &ub_slice;
}
Iterator* iter = db_->NewIterator(ro_copy, cfh);
long count = 0;
for (iter->Seek(prefix); iter->Valid() && iter->key().starts_with(prefix);
iter->Next()) {
++count;
}
assert(count <= (static_cast<long>(1) << ((8 - FLAGS_prefix_size) * 8)));
Status s = iter->status();
if (iter->status().ok()) {
thread->stats.AddPrefixes(1, count);
} else {
thread->stats.AddErrors(1);
}
delete iter;
return s;
}
virtual Status TestPut(ThreadState* thread, WriteOptions& write_opts,
const ReadOptions& read_opts,
const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys,
char (&value)[100], std::unique_ptr<MutexLock>& lock) {
auto shared = thread->shared;
int64_t max_key = shared->GetMaxKey();
int64_t rand_key = rand_keys[0];
int rand_column_family = rand_column_families[0];
while (!shared->AllowsOverwrite(rand_key) &&
(FLAGS_use_merge || shared->Exists(rand_column_family, rand_key))) {
lock.reset();
rand_key = thread->rand.Next() % max_key;
rand_column_family = thread->rand.Next() % FLAGS_column_families;
lock.reset(
new MutexLock(shared->GetMutexForKey(rand_column_family, rand_key)));
}
std::string key_str = Key(rand_key);
Slice key = key_str;
ColumnFamilyHandle* cfh = column_families_[rand_column_family];
if (FLAGS_verify_before_write) {
std::string key_str2 = Key(rand_key);
Slice k = key_str2;
std::string from_db;
Status s = db_->Get(read_opts, cfh, k, &from_db);
if (!VerifyValue(rand_column_family, rand_key, read_opts, shared, from_db,
s, true)) {
return s;
}
}
uint32_t value_base = thread->rand.Next() % shared->UNKNOWN_SENTINEL;
size_t sz = GenerateValue(value_base, value, sizeof(value));
Slice v(value, sz);
shared->Put(rand_column_family, rand_key, value_base, true /* pending */);
Status s;
if (FLAGS_use_merge) {
if (!FLAGS_use_txn) {
s = db_->Merge(write_opts, cfh, key, v);
} else {
#ifndef ROCKSDB_LITE
Transaction* txn;
s = NewTxn(write_opts, &txn);
if (s.ok()) {
s = txn->Merge(cfh, key, v);
if (s.ok()) {
s = CommitTxn(txn);
}
}
#endif
}
} else {
if (!FLAGS_use_txn) {
s = db_->Put(write_opts, cfh, key, v);
} else {
#ifndef ROCKSDB_LITE
Transaction* txn;
s = NewTxn(write_opts, &txn);
if (s.ok()) {
s = txn->Put(cfh, key, v);
if (s.ok()) {
s = CommitTxn(txn);
}
}
#endif
}
}
shared->Put(rand_column_family, rand_key, value_base, false /* pending */);
if (!s.ok()) {
fprintf(stderr, "put or merge error: %s\n", s.ToString().c_str());
std::terminate();
}
thread->stats.AddBytesForWrites(1, sz);
PrintKeyValue(rand_column_family, static_cast<uint32_t>(rand_key), value,
sz);
return s;
}
virtual Status TestDelete(ThreadState* thread, WriteOptions& write_opts,
const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys,
std::unique_ptr<MutexLock>& lock) {
int64_t rand_key = rand_keys[0];
int rand_column_family = rand_column_families[0];
auto shared = thread->shared;
int64_t max_key = shared->GetMaxKey();
// OPERATION delete
// If the chosen key does not allow overwrite and it does not exist,
// choose another key.
while (!shared->AllowsOverwrite(rand_key) &&
!shared->Exists(rand_column_family, rand_key)) {
lock.reset();
rand_key = thread->rand.Next() % max_key;
rand_column_family = thread->rand.Next() % FLAGS_column_families;
lock.reset(
new MutexLock(shared->GetMutexForKey(rand_column_family, rand_key)));
}
std::string key_str = Key(rand_key);
Slice key = key_str;
auto cfh = column_families_[rand_column_family];
// Use delete if the key may be overwritten and a single deletion
// otherwise.
Status s;
if (shared->AllowsOverwrite(rand_key)) {
shared->Delete(rand_column_family, rand_key, true /* pending */);
if (!FLAGS_use_txn) {
s = db_->Delete(write_opts, cfh, key);
} else {
#ifndef ROCKSDB_LITE
Transaction* txn;
s = NewTxn(write_opts, &txn);
if (s.ok()) {
s = txn->Delete(cfh, key);
if (s.ok()) {
s = CommitTxn(txn);
}
}
#endif
}
shared->Delete(rand_column_family, rand_key, false /* pending */);
thread->stats.AddDeletes(1);
if (!s.ok()) {
fprintf(stderr, "delete error: %s\n", s.ToString().c_str());
std::terminate();
}
} else {
shared->SingleDelete(rand_column_family, rand_key, true /* pending */);
if (!FLAGS_use_txn) {
s = db_->SingleDelete(write_opts, cfh, key);
} else {
#ifndef ROCKSDB_LITE
Transaction* txn;
s = NewTxn(write_opts, &txn);
if (s.ok()) {
s = txn->SingleDelete(cfh, key);
if (s.ok()) {
s = CommitTxn(txn);
}
}
#endif
}
shared->SingleDelete(rand_column_family, rand_key, false /* pending */);
thread->stats.AddSingleDeletes(1);
if (!s.ok()) {
fprintf(stderr, "single delete error: %s\n", s.ToString().c_str());
std::terminate();
}
}
return s;
}
virtual Status TestDeleteRange(ThreadState* thread, WriteOptions& write_opts,
const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys,
std::unique_ptr<MutexLock>& lock) {
// OPERATION delete range
std::vector<std::unique_ptr<MutexLock>> range_locks;
// delete range does not respect disallowed overwrites. the keys for
// which overwrites are disallowed are randomly distributed so it
// could be expensive to find a range where each key allows
// overwrites.
int64_t rand_key = rand_keys[0];
int rand_column_family = rand_column_families[0];
auto shared = thread->shared;
int64_t max_key = shared->GetMaxKey();
if (rand_key > max_key - FLAGS_range_deletion_width) {
lock.reset();
rand_key =
thread->rand.Next() % (max_key - FLAGS_range_deletion_width + 1);
range_locks.emplace_back(
new MutexLock(shared->GetMutexForKey(rand_column_family, rand_key)));
} else {
range_locks.emplace_back(std::move(lock));
}
for (int j = 1; j < FLAGS_range_deletion_width; ++j) {
if (((rand_key + j) & ((1 << FLAGS_log2_keys_per_lock) - 1)) == 0) {
range_locks.emplace_back(new MutexLock(
shared->GetMutexForKey(rand_column_family, rand_key + j)));
}
}
shared->DeleteRange(rand_column_family, rand_key,
rand_key + FLAGS_range_deletion_width,
true /* pending */);
std::string keystr = Key(rand_key);
Slice key = keystr;
auto cfh = column_families_[rand_column_family];
std::string end_keystr = Key(rand_key + FLAGS_range_deletion_width);
Slice end_key = end_keystr;
Status s = db_->DeleteRange(write_opts, cfh, key, end_key);
if (!s.ok()) {
fprintf(stderr, "delete range error: %s\n", s.ToString().c_str());
std::terminate();
}
int covered = shared->DeleteRange(rand_column_family, rand_key,
rand_key + FLAGS_range_deletion_width,
false /* pending */);
thread->stats.AddRangeDeletions(1);
thread->stats.AddCoveredByRangeDeletions(covered);
return s;
}
#ifdef ROCKSDB_LITE
virtual void TestIngestExternalFile(
ThreadState* /* thread */,
const std::vector<int>& /* rand_column_families */,
const std::vector<int64_t>& /* rand_keys */,
std::unique_ptr<MutexLock>& /* lock */) {
assert(false);
fprintf(stderr,
"RocksDB lite does not support "
"TestIngestExternalFile\n");
std::terminate();
}
#else
virtual void TestIngestExternalFile(
ThreadState* thread, const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys, std::unique_ptr<MutexLock>& lock) {
const std::string sst_filename =
FLAGS_db + "/." + ToString(thread->tid) + ".sst";
Status s;
if (FLAGS_env->FileExists(sst_filename).ok()) {
// Maybe we terminated abnormally before, so cleanup to give this file
// ingestion a clean slate
s = FLAGS_env->DeleteFile(sst_filename);
}
SstFileWriter sst_file_writer(EnvOptions(options_), options_);
if (s.ok()) {
s = sst_file_writer.Open(sst_filename);
}
int64_t key_base = rand_keys[0];
int column_family = rand_column_families[0];
std::vector<std::unique_ptr<MutexLock>> range_locks;
std::vector<uint32_t> values;
SharedState* shared = thread->shared;
// Grab locks, set pending state on expected values, and add keys
for (int64_t key = key_base;
s.ok() && key < std::min(key_base + FLAGS_ingest_external_file_width,
shared->GetMaxKey());
++key) {
if (key == key_base) {
range_locks.emplace_back(std::move(lock));
} else if ((key & ((1 << FLAGS_log2_keys_per_lock) - 1)) == 0) {
range_locks.emplace_back(
new MutexLock(shared->GetMutexForKey(column_family, key)));
}
uint32_t value_base = thread->rand.Next() % shared->UNKNOWN_SENTINEL;
values.push_back(value_base);
shared->Put(column_family, key, value_base, true /* pending */);
char value[100];
size_t value_len = GenerateValue(value_base, value, sizeof(value));
auto key_str = Key(key);
s = sst_file_writer.Put(Slice(key_str), Slice(value, value_len));
}
if (s.ok()) {
s = sst_file_writer.Finish();
}
if (s.ok()) {
s = db_->IngestExternalFile(column_families_[column_family],
{sst_filename}, IngestExternalFileOptions());
}
if (!s.ok()) {
fprintf(stderr, "file ingestion error: %s\n", s.ToString().c_str());
std::terminate();
}
int64_t key = key_base;
for (int32_t value : values) {
shared->Put(column_family, key, value, false /* pending */);
++key;
}
}
#endif // ROCKSDB_LITE
bool VerifyValue(int cf, int64_t key, const ReadOptions& /*opts*/,
SharedState* shared, const std::string& value_from_db,
Status s, bool strict = false) const {
if (shared->HasVerificationFailedYet()) {
return false;
}
// compare value_from_db with the value in the shared state
char value[kValueMaxLen];
uint32_t value_base = shared->Get(cf, key);
if (value_base == SharedState::UNKNOWN_SENTINEL) {
return true;
}
if (value_base == SharedState::DELETION_SENTINEL && !strict) {
return true;
}
if (s.ok()) {
if (value_base == SharedState::DELETION_SENTINEL) {
VerificationAbort(shared, "Unexpected value found", cf, key);
return false;
}
size_t sz = GenerateValue(value_base, value, sizeof(value));
if (value_from_db.length() != sz) {
VerificationAbort(shared, "Length of value read is not equal", cf, key);
return false;
}
if (memcmp(value_from_db.data(), value, sz) != 0) {
VerificationAbort(shared, "Contents of value read don't match", cf,
key);
return false;
}
} else {
if (value_base != SharedState::DELETION_SENTINEL) {
VerificationAbort(shared, "Value not found: " + s.ToString(), cf, key);
return false;
}
}
return true;
}
};
StressTest* CreateNonBatchedOpsStressTest() {
return new NonBatchedOpsStressTest();
}
} // namespace rocksdb
#endif // GFLAGS

29
env/env_posix.cc vendored
View File

@ -16,6 +16,9 @@
#if defined(OS_LINUX)
#include <linux/fs.h>
#endif
#if defined(ROCKSDB_IOURING_PRESENT)
#include <liburing.h>
#endif
#include <pthread.h>
#include <signal.h>
#include <stdio.h>
@ -32,6 +35,9 @@
#include <sys/statvfs.h>
#include <sys/time.h>
#include <sys/types.h>
#if defined(ROCKSDB_IOURING_PRESENT)
#include <sys/uio.h>
#endif
#include <time.h>
#include <algorithm>
// Get nano time includes
@ -286,7 +292,12 @@ class PosixEnv : public Env {
}
#endif
}
result->reset(new PosixRandomAccessFile(fname, fd, options));
result->reset(new PosixRandomAccessFile(fname, fd, options
#if defined(ROCKSDB_IOURING_PRESENT)
,
thread_local_io_urings_.get()
#endif
));
}
return s;
}
@ -1105,6 +1116,11 @@ class PosixEnv : public Env {
#endif
}
#if defined(ROCKSDB_IOURING_PRESENT)
// io_uring instance
std::unique_ptr<ThreadLocalPtr> thread_local_io_urings_;
#endif
size_t page_size_;
std::vector<ThreadPoolImpl> thread_pools_;
@ -1129,6 +1145,17 @@ PosixEnv::PosixEnv()
thread_pools_[pool_id].SetHostEnv(this);
}
thread_status_updater_ = CreateThreadStatusUpdater();
#if defined(ROCKSDB_IOURING_PRESENT)
// Test whether IOUring is supported, and if it does, create a managing
// object for thread local point so that in the future thread-local
// io_uring can be created.
struct io_uring* new_io_uring = CreateIOUring();
if (new_io_uring != nullptr) {
thread_local_io_urings_.reset(new ThreadLocalPtr(DeleteIOUring));
delete new_io_uring;
}
#endif
}
void PosixEnv::Schedule(void (*function)(void* arg1), void* arg, Priority pri,

107
env/io_posix.cc vendored
View File

@ -34,6 +34,7 @@
#include "port/port.h"
#include "rocksdb/slice.h"
#include "test_util/sync_point.h"
#include "util/autovector.h"
#include "util/coding.h"
#include "util/string_util.h"
@ -409,12 +410,22 @@ size_t PosixHelper::GetUniqueIdFromFile(int fd, char* id, size_t max_size) {
*
* pread() based random-access
*/
PosixRandomAccessFile::PosixRandomAccessFile(const std::string& fname, int fd,
const EnvOptions& options)
PosixRandomAccessFile::PosixRandomAccessFile(
const std::string& fname, int fd, const EnvOptions& options
#if defined(ROCKSDB_IOURING_PRESENT)
,
ThreadLocalPtr* thread_local_io_urings
#endif
)
: filename_(fname),
fd_(fd),
use_direct_io_(options.use_direct_reads),
logical_sector_size_(GetLogicalBufferSize(fd_)) {
logical_sector_size_(GetLogicalBufferSize(fd_))
#if defined(ROCKSDB_IOURING_PRESENT)
,
thread_local_io_urings_(thread_local_io_urings)
#endif
{
assert(!options.use_direct_reads || !options.use_mmap_reads);
assert(!options.use_mmap_reads || sizeof(void*) < 8);
}
@ -460,6 +471,96 @@ Status PosixRandomAccessFile::Read(uint64_t offset, size_t n, Slice* result,
return s;
}
Status PosixRandomAccessFile::MultiRead(ReadRequest* reqs, size_t num_reqs) {
#if defined(ROCKSDB_IOURING_PRESENT)
size_t reqs_off;
ssize_t ret __attribute__((__unused__));
struct io_uring* iu = nullptr;
if (thread_local_io_urings_) {
iu = static_cast<struct io_uring*>(thread_local_io_urings_->Get());
if (iu == nullptr) {
iu = CreateIOUring();
if (iu != nullptr) {
thread_local_io_urings_->Reset(iu);
}
}
}
// Init failed, platform doesn't support io_uring. Fall back to
// serialized reads
if (iu == nullptr) {
return RandomAccessFile::MultiRead(reqs, num_reqs);
}
struct WrappedReadRequest {
ReadRequest* req;
struct iovec iov;
explicit WrappedReadRequest(ReadRequest* r) : req(r) {}
};
autovector<WrappedReadRequest, 32> req_wraps;
for (size_t i = 0; i < num_reqs; i++) {
req_wraps.emplace_back(&reqs[i]);
}
reqs_off = 0;
while (num_reqs) {
size_t this_reqs = num_reqs;
// If requests exceed depth, split it into batches
if (this_reqs > kIoUringDepth) this_reqs = kIoUringDepth;
for (size_t i = 0; i < this_reqs; i++) {
size_t index = i + reqs_off;
struct io_uring_sqe* sqe;
sqe = io_uring_get_sqe(iu);
req_wraps[index].iov.iov_base = reqs[index].scratch;
req_wraps[index].iov.iov_len = reqs[index].len;
reqs[index].result = reqs[index].scratch;
io_uring_prep_readv(sqe, fd_, &req_wraps[index].iov, 1,
reqs[index].offset);
io_uring_sqe_set_data(sqe, &req_wraps[index]);
}
ret = io_uring_submit_and_wait(iu, static_cast<unsigned int>(this_reqs));
if (static_cast<size_t>(ret) != this_reqs) {
fprintf(stderr, "ret = %ld this_reqs: %ld\n", (long)ret, (long)this_reqs);
}
assert(static_cast<size_t>(ret) == this_reqs);
for (size_t i = 0; i < this_reqs; i++) {
struct io_uring_cqe* cqe;
WrappedReadRequest* req_wrap;
// We could use the peek variant here, but this seems safer in terms
// of our initial wait not reaping all completions
ret = io_uring_wait_cqe(iu, &cqe);
assert(!ret);
req_wrap = static_cast<WrappedReadRequest*>(io_uring_cqe_get_data(cqe));
ReadRequest* req = req_wrap->req;
if (static_cast<size_t>(cqe->res) == req_wrap->iov.iov_len) {
req->result = Slice(req->scratch, cqe->res);
req->status = Status::OK();
} else if (cqe->res >= 0) {
req->result = Slice(req->scratch, req_wrap->iov.iov_len - cqe->res);
} else {
req->result = Slice(req->scratch, 0);
req->status = IOError("Req failed", filename_, cqe->res);
}
io_uring_cqe_seen(iu, cqe);
}
num_reqs -= this_reqs;
reqs_off += this_reqs;
}
return Status::OK();
#else
return RandomAccessFile::MultiRead(reqs, num_reqs);
#endif
}
Status PosixRandomAccessFile::Prefetch(uint64_t offset, size_t n) {
Status s;
if (!use_direct_io()) {

37
env/io_posix.h vendored
View File

@ -8,10 +8,15 @@
// found in the LICENSE file. See the AUTHORS file for names of contributors.
#pragma once
#include <errno.h>
#if defined(ROCKSDB_IOURING_PRESENT)
#include <liburing.h>
#include <sys/uio.h>
#endif
#include <unistd.h>
#include <atomic>
#include <string>
#include "rocksdb/env.h"
#include "util/thread_local.h"
// For non linux platform, the following macros are used only as place
// holder.
@ -79,21 +84,51 @@ class PosixSequentialFile : public SequentialFile {
}
};
#if defined(ROCKSDB_IOURING_PRESENT)
// io_uring instance queue depth
const unsigned int kIoUringDepth = 256;
inline void DeleteIOUring(void* p) {
struct io_uring* iu = static_cast<struct io_uring*>(p);
delete iu;
}
inline struct io_uring* CreateIOUring() {
struct io_uring* new_io_uring = new struct io_uring;
int ret = io_uring_queue_init(kIoUringDepth, new_io_uring, 0);
if (ret) {
delete new_io_uring;
new_io_uring = nullptr;
}
return new_io_uring;
}
#endif // defined(ROCKSDB_IOURING_PRESENT)
class PosixRandomAccessFile : public RandomAccessFile {
protected:
std::string filename_;
int fd_;
bool use_direct_io_;
size_t logical_sector_size_;
#if defined(ROCKSDB_IOURING_PRESENT)
ThreadLocalPtr* thread_local_io_urings_;
#endif
public:
PosixRandomAccessFile(const std::string& fname, int fd,
const EnvOptions& options);
const EnvOptions& options
#if defined(ROCKSDB_IOURING_PRESENT)
,
ThreadLocalPtr* thread_local_io_urings
#endif
);
virtual ~PosixRandomAccessFile();
virtual Status Read(uint64_t offset, size_t n, Slice* result,
char* scratch) const override;
virtual Status MultiRead(ReadRequest* reqs, size_t num_reqs) override;
virtual Status Prefetch(uint64_t offset, size_t n) override;
#if defined(OS_LINUX) || defined(OS_MACOSX) || defined(OS_AIX)

12
src.mk
View File

@ -265,7 +265,15 @@ BENCH_LIB_SOURCES = \
tools/db_bench_tool.cc \
STRESS_LIB_SOURCES = \
tools/db_stress_tool.cc \
db_stress_tool/batched_ops_stress.cc \
db_stress_tool/cf_consistency_stress.cc \
db_stress_tool/db_stress_common.cc \
db_stress_tool/db_stress_driver.cc \
db_stress_tool/db_stress_test_base.cc \
db_stress_tool/db_stress_gflags.cc \
db_stress_tool/db_stress_shared_state.cc \
db_stress_tool/db_stress_tool.cc \
db_stress_tool/no_batched_ops_stress.cc \
TEST_LIB_SOURCES = \
db/db_test_util.cc \
@ -283,6 +291,7 @@ FOLLY_SOURCES = \
MAIN_SOURCES = \
cache/cache_bench.cc \
cache/cache_test.cc \
db_stress_tool/db_stress.cc \
db/column_family_test.cc \
db/compact_files_test.cc \
db/compaction/compaction_iterator_test.cc \
@ -398,7 +407,6 @@ MAIN_SOURCES = \
tools/db_bench.cc \
tools/db_bench_tool_test.cc \
tools/db_sanity_test.cc \
tools/db_stress.cc \
tools/ldb_cmd_test.cc \
tools/reduce_levels_test.cc \
tools/sst_dump_test.cc \

3
table/block_based/filter_policy.cc
View File

@ -443,7 +443,8 @@ void BloomFilterPolicy::CreateFilter(const Slice* keys, int n,
std::string* dst) const {
// We should ideally only be using this deprecated interface for
// appropriately constructed BloomFilterPolicy
assert(mode_ == kDeprecatedBlock);
// FIXME disabled because of bug in C interface; see issue #6129
//assert(mode_ == kDeprecatedBlock);
// Compute bloom filter size (in both bits and bytes)
uint32_t bits = static_cast<uint32_t>(n * whole_bits_per_key_);

10
third-party/folly/folly/synchronization/test/DistributedMutexTest.cpp vendored
View File

@ -13,7 +13,7 @@
#include <gtest/gtest.h>
#endif
#ifndef ROCKSDB_LITE
#if !defined(ROCKSDB_LITE) && !defined(__ARM_ARCH)
#include <chrono>
#include <cmath>
@ -1128,9 +1128,15 @@ TEST(DistributedMutex, StressBigValueReturnSixtyFourThreads) {
}
} // namespace folly
#endif // ROCKSDB_LITE
int main(int argc, char** argv) {
::testing::InitGoogleTest(&argc, argv);
return RUN_ALL_TESTS();
}
#else
int main(int /*argc*/, char** /*argv*/) {
printf("DistributedMutex is not supported in ROCKSDB_LITE or on ARM\n");
return 0;
}
#endif // !ROCKSDB_LITE && !__ARM_ARCH

6
tools/CMakeLists.txt
View File

@ -1,7 +1,6 @@
set(TOOLS
sst_dump.cc
db_sanity_test.cc
db_stress.cc
write_stress.cc
ldb.cc
db_repl_stress.cc
@ -14,8 +13,9 @@ foreach(src ${TOOLS})
target_link_libraries(${exename}${ARTIFACT_SUFFIX} ${LIBS})
list(APPEND tool_deps ${exename})
endforeach()
add_custom_target(tools
DEPENDS ${tool_deps})
list(APPEND tool_deps)
add_custom_target(ldb_tests
COMMAND python ${CMAKE_CURRENT_SOURCE_DIR}/ldb_tests.py
DEPENDS ldb)

4902
tools/db_stress_tool.cc
View File

File diff suppressed because it is too large Load Diff