You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
rocksdb/table/format.cc

465 lines
17 KiB

// Copyright (c) 2011-present, Facebook, Inc. All rights reserved.
// This source code is licensed under both the GPLv2 (found in the
// COPYING file in the root directory) and Apache 2.0 License
// (found in the LICENSE.Apache file in the root directory).
//
// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file. See the AUTHORS file for names of contributors.
#include "table/format.h"
#include <cinttypes>
#include <string>
#include "block_fetcher.h"
#include "logging/logging.h"
#include "memory/memory_allocator.h"
#include "monitoring/perf_context_imp.h"
#include "monitoring/statistics.h"
#include "rocksdb/env.h"
#include "table/block_based/block.h"
#include "table/block_based/block_based_table_reader.h"
#include "table/persistent_cache_helper.h"
#include "util/coding.h"
#include "util/compression.h"
#include "util/crc32c.h"
#include "util/file_reader_writer.h"
#include "util/stop_watch.h"
#include "util/string_util.h"
#include "util/xxhash.h"
namespace rocksdb {
extern const uint64_t kLegacyBlockBasedTableMagicNumber;
extern const uint64_t kBlockBasedTableMagicNumber;
11 years ago
#ifndef ROCKSDB_LITE
extern const uint64_t kLegacyPlainTableMagicNumber;
extern const uint64_t kPlainTableMagicNumber;
11 years ago
#else
// ROCKSDB_LITE doesn't have plain table
const uint64_t kLegacyPlainTableMagicNumber = 0;
const uint64_t kPlainTableMagicNumber = 0;
#endif
bool ShouldReportDetailedTime(Env* env, Statistics* stats) {
return env != nullptr && stats != nullptr &&
stats->get_stats_level() > kExceptDetailedTimers;
}
void BlockHandle::EncodeTo(std::string* dst) const {
// Sanity check that all fields have been set
assert(offset_ != ~static_cast<uint64_t>(0));
assert(size_ != ~static_cast<uint64_t>(0));
Miscellaneous performance improvements Summary: I was investigating performance issues in the SstFileWriter and found all of the following: - The SstFileWriter::Add() function created a local InternalKey every time it was called generating a allocation and free each time. Changed to have an InternalKey member variable that can be reset with the new InternalKey::Set() function. - In SstFileWriter::Add() the smallest_key and largest_key values were assigned the result of a ToString() call, but it is simpler to just assign them directly from the user's key. - The Slice class had no move constructor so each time one was returned from a function a new one had to be allocated, the old data copied to the new, and the old one was freed. I added the move constructor which also required a copy constructor and assignment operator. - The BlockBuilder::CurrentSizeEstimate() function calculates the current estimate size, but was being called 2 or 3 times for each key added. I changed the class to maintain a running estimate (equal to the original calculation) so that the function can return an already calculated value. - The code in BlockBuilder::Add() that calculated the shared bytes between the last key and the new key duplicated what Slice::difference_offset does, so I replaced it with the standard function. - BlockBuilder::Add() had code to copy just the changed portion into the last key value (and asserted that it now matched the new key). It is more efficient just to copy the whole new key over. - Moved this same code up into the 'if (use_delta_encoding_)' since the last key value is only needed when delta encoding is on. - FlushBlockBySizePolicy::BlockAlmostFull calculated a standard deviation value each time it was called, but this information would only change if block_size of block_size_deviation changed, so I created a member variable to hold the value to avoid the calculation each time. - Each PutVarint??() function has a buffer and calls std::string::append(). Two or three calls in a row could share a buffer and a single call to std::string::append(). Some of these will be helpful outside of the SstFileWriter. I'm not 100% the addition of the move constructor is appropriate as I wonder why this wasn't done before - maybe because of compiler compatibility? I tried it on gcc 4.8 and 4.9. Test Plan: The changes should not affect the results so the existing tests should all still work and no new tests were added. The value of the changes was seen by manually testing the SstFileWriter class through MyRocks and adding timing code to identify problem areas. Reviewers: sdong, IslamAbdelRahman Reviewed By: IslamAbdelRahman Subscribers: andrewkr, dhruba Differential Revision: https://reviews.facebook.net/D59607
9 years ago
PutVarint64Varint64(dst, offset_, size_);
}
Status BlockHandle::DecodeFrom(Slice* input) {
if (GetVarint64(input, &offset_) && GetVarint64(input, &size_)) {
return Status::OK();
} else {
// reset in case failure after partially decoding
offset_ = 0;
size_ = 0;
return Status::Corruption("bad block handle");
}
}
Status BlockHandle::DecodeSizeFrom(uint64_t _offset, Slice* input) {
if (GetVarint64(input, &size_)) {
offset_ = _offset;
return Status::OK();
} else {
// reset in case failure after partially decoding
offset_ = 0;
size_ = 0;
return Status::Corruption("bad block handle");
}
}
// Return a string that contains the copy of handle.
std::string BlockHandle::ToString(bool hex) const {
std::string handle_str;
EncodeTo(&handle_str);
if (hex) {
return Slice(handle_str).ToString(true);
} else {
return handle_str;
}
}
const BlockHandle BlockHandle::kNullBlockHandle(0, 0);
Add an option to put first key of each sst block in the index (#5289) Summary: The first key is used to defer reading the data block until this file gets to the top of merging iterator's heap. For short range scans, most files never make it to the top of the heap, so this change can reduce read amplification by a lot sometimes. Consider the following workload. There are a few data streams (we'll be calling them "logs"), each stream consisting of a sequence of blobs (we'll be calling them "records"). Each record is identified by log ID and a sequence number within the log. RocksDB key is concatenation of log ID and sequence number (big endian). Reads are mostly relatively short range scans, each within a single log. Writes are mostly sequential for each log, but writes to different logs are randomly interleaved. Compactions are disabled; instead, when we accumulate a few tens of sst files, we create a new column family and start writing to it. So, a typical sst file consists of a few ranges of blocks, each range corresponding to one log ID (we use FlushBlockPolicy to cut blocks at log boundaries). A typical read would go like this. First, iterator Seek() reads one block from each sst file. Then a series of Next()s move through one sst file (since writes to each log are mostly sequential) until the subiterator reaches the end of this log in this sst file; then Next() switches to the next sst file and reads sequentially from that, and so on. Often a range scan will only return records from a small number of blocks in small number of sst files; in this case, the cost of initial Seek() reading one block from each file may be bigger than the cost of reading the actually useful blocks. Neither iterate_upper_bound nor bloom filters can prevent reading one block from each file in Seek(). But this PR can: if the index contains first key from each block, we don't have to read the block until this block actually makes it to the top of merging iterator's heap, so for short range scans we won't read any blocks from most of the sst files. This PR does the deferred block loading inside value() call. This is not ideal: there's no good way to report an IO error from inside value(). As discussed with siying offline, it would probably be better to change InternalIterator's interface to explicitly fetch deferred value and get status. I'll do it in a separate PR. Pull Request resolved: https://github.com/facebook/rocksdb/pull/5289 Differential Revision: D15256423 Pulled By: al13n321 fbshipit-source-id: 750e4c39ce88e8d41662f701cf6275d9388ba46a
6 years ago
void IndexValue::EncodeTo(std::string* dst, bool have_first_key,
const BlockHandle* previous_handle) const {
if (previous_handle) {
assert(handle.offset() == previous_handle->offset() +
previous_handle->size() + kBlockTrailerSize);
PutVarsignedint64(dst, handle.size() - previous_handle->size());
} else {
handle.EncodeTo(dst);
}
assert(dst->size() != 0);
if (have_first_key) {
PutLengthPrefixedSlice(dst, first_internal_key);
}
}
Status IndexValue::DecodeFrom(Slice* input, bool have_first_key,
const BlockHandle* previous_handle) {
if (previous_handle) {
int64_t delta;
if (!GetVarsignedint64(input, &delta)) {
return Status::Corruption("bad delta-encoded index value");
}
handle = BlockHandle(
previous_handle->offset() + previous_handle->size() + kBlockTrailerSize,
previous_handle->size() + delta);
} else {
Status s = handle.DecodeFrom(input);
if (!s.ok()) {
return s;
}
}
if (!have_first_key) {
first_internal_key = Slice();
} else if (!GetLengthPrefixedSlice(input, &first_internal_key)) {
return Status::Corruption("bad first key in block info");
}
return Status::OK();
}
std::string IndexValue::ToString(bool hex, bool have_first_key) const {
std::string s;
EncodeTo(&s, have_first_key, nullptr);
if (hex) {
return Slice(s).ToString(true);
} else {
return s;
}
}
namespace {
inline bool IsLegacyFooterFormat(uint64_t magic_number) {
return magic_number == kLegacyBlockBasedTableMagicNumber ||
magic_number == kLegacyPlainTableMagicNumber;
}
inline uint64_t UpconvertLegacyFooterFormat(uint64_t magic_number) {
if (magic_number == kLegacyBlockBasedTableMagicNumber) {
return kBlockBasedTableMagicNumber;
}
if (magic_number == kLegacyPlainTableMagicNumber) {
return kPlainTableMagicNumber;
}
assert(false);
return 0;
}
} // namespace
// legacy footer format:
// metaindex handle (varint64 offset, varint64 size)
// index handle (varint64 offset, varint64 size)
// <padding> to make the total size 2 * BlockHandle::kMaxEncodedLength
// table_magic_number (8 bytes)
// new footer format:
// checksum type (char, 1 byte)
// metaindex handle (varint64 offset, varint64 size)
// index handle (varint64 offset, varint64 size)
// <padding> to make the total size 2 * BlockHandle::kMaxEncodedLength + 1
// footer version (4 bytes)
// table_magic_number (8 bytes)
void Footer::EncodeTo(std::string* dst) const {
assert(HasInitializedTableMagicNumber());
if (IsLegacyFooterFormat(table_magic_number())) {
// has to be default checksum with legacy footer
assert(checksum_ == kCRC32c);
const size_t original_size = dst->size();
metaindex_handle_.EncodeTo(dst);
index_handle_.EncodeTo(dst);
dst->resize(original_size + 2 * BlockHandle::kMaxEncodedLength); // Padding
PutFixed32(dst, static_cast<uint32_t>(table_magic_number() & 0xffffffffu));
PutFixed32(dst, static_cast<uint32_t>(table_magic_number() >> 32));
assert(dst->size() == original_size + kVersion0EncodedLength);
} else {
const size_t original_size = dst->size();
dst->push_back(static_cast<char>(checksum_));
metaindex_handle_.EncodeTo(dst);
index_handle_.EncodeTo(dst);
dst->resize(original_size + kNewVersionsEncodedLength - 12); // Padding
PutFixed32(dst, version());
PutFixed32(dst, static_cast<uint32_t>(table_magic_number() & 0xffffffffu));
PutFixed32(dst, static_cast<uint32_t>(table_magic_number() >> 32));
assert(dst->size() == original_size + kNewVersionsEncodedLength);
}
}
Footer::Footer(uint64_t _table_magic_number, uint32_t _version)
: version_(_version),
checksum_(kCRC32c),
table_magic_number_(_table_magic_number) {
// This should be guaranteed by constructor callers
assert(!IsLegacyFooterFormat(_table_magic_number) || version_ == 0);
}
Status Footer::DecodeFrom(Slice* input) {
assert(!HasInitializedTableMagicNumber());
assert(input != nullptr);
assert(input->size() >= kMinEncodedLength);
const char* magic_ptr =
input->data() + input->size() - kMagicNumberLengthByte;
const uint32_t magic_lo = DecodeFixed32(magic_ptr);
const uint32_t magic_hi = DecodeFixed32(magic_ptr + 4);
uint64_t magic = ((static_cast<uint64_t>(magic_hi) << 32) |
(static_cast<uint64_t>(magic_lo)));
// We check for legacy formats here and silently upconvert them
bool legacy = IsLegacyFooterFormat(magic);
if (legacy) {
magic = UpconvertLegacyFooterFormat(magic);
}
set_table_magic_number(magic);
if (legacy) {
// The size is already asserted to be at least kMinEncodedLength
// at the beginning of the function
input->remove_prefix(input->size() - kVersion0EncodedLength);
version_ = 0 /* legacy */;
checksum_ = kCRC32c;
} else {
version_ = DecodeFixed32(magic_ptr - 4);
// Footer version 1 and higher will always occupy exactly this many bytes.
// It consists of the checksum type, two block handles, padding,
// a version number, and a magic number
if (input->size() < kNewVersionsEncodedLength) {
return Status::Corruption("input is too short to be an sstable");
} else {
input->remove_prefix(input->size() - kNewVersionsEncodedLength);
}
uint32_t chksum;
if (!GetVarint32(input, &chksum)) {
return Status::Corruption("bad checksum type");
}
checksum_ = static_cast<ChecksumType>(chksum);
}
Status result = metaindex_handle_.DecodeFrom(input);
if (result.ok()) {
result = index_handle_.DecodeFrom(input);
}
if (result.ok()) {
// We skip over any leftover data (just padding for now) in "input"
const char* end = magic_ptr + kMagicNumberLengthByte;
*input = Slice(end, input->data() + input->size() - end);
}
return result;
}
std::string Footer::ToString() const {
std::string result;
result.reserve(1024);
bool legacy = IsLegacyFooterFormat(table_magic_number_);
if (legacy) {
result.append("metaindex handle: " + metaindex_handle_.ToString() + "\n ");
result.append("index handle: " + index_handle_.ToString() + "\n ");
result.append("table_magic_number: " +
rocksdb::ToString(table_magic_number_) + "\n ");
} else {
result.append("checksum: " + rocksdb::ToString(checksum_) + "\n ");
result.append("metaindex handle: " + metaindex_handle_.ToString() + "\n ");
result.append("index handle: " + index_handle_.ToString() + "\n ");
result.append("footer version: " + rocksdb::ToString(version_) + "\n ");
result.append("table_magic_number: " +
rocksdb::ToString(table_magic_number_) + "\n ");
}
return result;
}
Status ReadFooterFromFile(RandomAccessFileReader* file,
FilePrefetchBuffer* prefetch_buffer,
uint64_t file_size, Footer* footer,
uint64_t enforce_table_magic_number) {
if (file_size < Footer::kMinEncodedLength) {
return Status::Corruption("file is too short (" + ToString(file_size) +
" bytes) to be an "
"sstable: " +
file->file_name());
}
char footer_space[Footer::kMaxEncodedLength];
Slice footer_input;
size_t read_offset =
(file_size > Footer::kMaxEncodedLength)
? static_cast<size_t>(file_size - Footer::kMaxEncodedLength)
: 0;
Status s;
if (prefetch_buffer == nullptr ||
!prefetch_buffer->TryReadFromCache(read_offset, Footer::kMaxEncodedLength,
&footer_input)) {
s = file->Read(read_offset, Footer::kMaxEncodedLength, &footer_input,
footer_space);
if (!s.ok()) return s;
}
// Check that we actually read the whole footer from the file. It may be
// that size isn't correct.
if (footer_input.size() < Footer::kMinEncodedLength) {
return Status::Corruption("file is too short (" + ToString(file_size) +
" bytes) to be an "
"sstable" +
file->file_name());
}
s = footer->DecodeFrom(&footer_input);
if (!s.ok()) {
return s;
}
if (enforce_table_magic_number != 0 &&
enforce_table_magic_number != footer->table_magic_number()) {
return Status::Corruption(
"Bad table magic number: expected " +
ToString(enforce_table_magic_number) + ", found " +
ToString(footer->table_magic_number()) + " in " + file->file_name());
}
return Status::OK();
}
Status UncompressBlockContentsForCompressionType(
const UncompressionInfo& uncompression_info, const char* data, size_t n,
BlockContents* contents, uint32_t format_version,
const ImmutableCFOptions& ioptions, MemoryAllocator* allocator) {
CacheAllocationPtr ubuf;
assert(uncompression_info.type() != kNoCompression &&
"Invalid compression type");
StopWatchNano timer(ioptions.env, ShouldReportDetailedTime(
ioptions.env, ioptions.statistics));
int decompress_size = 0;
switch (uncompression_info.type()) {
case kSnappyCompression: {
size_t ulength = 0;
static char snappy_corrupt_msg[] =
"Snappy not supported or corrupted Snappy compressed block contents";
if (!Snappy_GetUncompressedLength(data, n, &ulength)) {
return Status::Corruption(snappy_corrupt_msg);
}
ubuf = AllocateBlock(ulength, allocator);
if (!Snappy_Uncompress(data, n, ubuf.get())) {
return Status::Corruption(snappy_corrupt_msg);
}
*contents = BlockContents(std::move(ubuf), ulength);
break;
}
case kZlibCompression:
ubuf = Zlib_Uncompress(
uncompression_info, data, n, &decompress_size,
GetCompressFormatForVersion(kZlibCompression, format_version),
allocator);
if (!ubuf) {
static char zlib_corrupt_msg[] =
"Zlib not supported or corrupted Zlib compressed block contents";
return Status::Corruption(zlib_corrupt_msg);
}
*contents = BlockContents(std::move(ubuf), decompress_size);
break;
case kBZip2Compression:
ubuf = BZip2_Uncompress(
data, n, &decompress_size,
GetCompressFormatForVersion(kBZip2Compression, format_version),
allocator);
if (!ubuf) {
static char bzip2_corrupt_msg[] =
"Bzip2 not supported or corrupted Bzip2 compressed block contents";
return Status::Corruption(bzip2_corrupt_msg);
}
*contents = BlockContents(std::move(ubuf), decompress_size);
break;
case kLZ4Compression:
ubuf = LZ4_Uncompress(
uncompression_info, data, n, &decompress_size,
GetCompressFormatForVersion(kLZ4Compression, format_version),
allocator);
if (!ubuf) {
static char lz4_corrupt_msg[] =
"LZ4 not supported or corrupted LZ4 compressed block contents";
return Status::Corruption(lz4_corrupt_msg);
}
*contents = BlockContents(std::move(ubuf), decompress_size);
break;
case kLZ4HCCompression:
ubuf = LZ4_Uncompress(
uncompression_info, data, n, &decompress_size,
GetCompressFormatForVersion(kLZ4HCCompression, format_version),
allocator);
if (!ubuf) {
static char lz4hc_corrupt_msg[] =
"LZ4HC not supported or corrupted LZ4HC compressed block contents";
return Status::Corruption(lz4hc_corrupt_msg);
}
*contents = BlockContents(std::move(ubuf), decompress_size);
break;
case kXpressCompression:
// XPRESS allocates memory internally, thus no support for custom
// allocator.
ubuf.reset(XPRESS_Uncompress(data, n, &decompress_size));
if (!ubuf) {
static char xpress_corrupt_msg[] =
"XPRESS not supported or corrupted XPRESS compressed block "
"contents";
return Status::Corruption(xpress_corrupt_msg);
}
*contents = BlockContents(std::move(ubuf), decompress_size);
break;
case kZSTD:
case kZSTDNotFinalCompression:
ubuf = ZSTD_Uncompress(uncompression_info, data, n, &decompress_size,
allocator);
if (!ubuf) {
static char zstd_corrupt_msg[] =
"ZSTD not supported or corrupted ZSTD compressed block contents";
return Status::Corruption(zstd_corrupt_msg);
}
*contents = BlockContents(std::move(ubuf), decompress_size);
break;
default:
return Status::Corruption("bad block type");
}
if (ShouldReportDetailedTime(ioptions.env, ioptions.statistics)) {
RecordTimeToHistogram(ioptions.statistics, DECOMPRESSION_TIMES_NANOS,
timer.ElapsedNanos());
}
RecordTimeToHistogram(ioptions.statistics, BYTES_DECOMPRESSED,
contents->data.size());
RecordTick(ioptions.statistics, NUMBER_BLOCK_DECOMPRESSED);
return Status::OK();
}
//
// The 'data' points to the raw block contents that was read in from file.
// This method allocates a new heap buffer and the raw block
// contents are uncompresed into this buffer. This
// buffer is returned via 'result' and it is upto the caller to
// free this buffer.
// format_version is the block format as defined in include/rocksdb/table.h
Status UncompressBlockContents(const UncompressionInfo& uncompression_info,
const char* data, size_t n,
BlockContents* contents, uint32_t format_version,
const ImmutableCFOptions& ioptions,
MemoryAllocator* allocator) {
assert(data[n] != kNoCompression);
assert(data[n] == uncompression_info.type());
return UncompressBlockContentsForCompressionType(uncompression_info, data, n,
contents, format_version,
ioptions, allocator);
}
} // namespace rocksdb