mirror of
https://github.com/facebook/rocksdb.git
synced 2024-11-29 09:36:17 +00:00
9a034801ce
Summary:
**Context/Summary:**
- Similar to https://github.com/facebook/rocksdb/pull/11288 but for user read such as `Get(), MultiGet(), DBIterator::XXX(), Verify(File)Checksum()`.
- For this, I refactored some user-facing `MultiGet` calls in `TransactionBase` and various types of `DB` so that it does not call a user-facing `Get()` but `GetImpl()` for passing the `ReadOptions::io_activity` check (see PR conversation)
- New user read stats breakdown are guarded by `kExceptDetailedTimers` since measurement shows they have 4-5% regression to the upstream/main.
- Misc
- More refactoring: with https://github.com/facebook/rocksdb/pull/11288, we complete passing `ReadOptions/IOOptions` to FS level. So we can now replace the previously [added](https://github.com/facebook/rocksdb/pull/9424) `rate_limiter_priority` parameter in `RandomAccessFileReader`'s `Read/MultiRead/Prefetch()` with `IOOptions::rate_limiter_priority`
- Also, `ReadAsync()` call time is measured in `SST_READ_MICRO` now
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11444
Test Plan:
- CI fake db crash/stress test
- Microbenchmarking
**Build** `make clean && ROCKSDB_NO_FBCODE=1 DEBUG_LEVEL=0 make -jN db_basic_bench`
- google benchmark version: 604f6fd3f4
- db_basic_bench_base: upstream
- db_basic_bench_pr: db_basic_bench_base + this PR
- asyncread_db_basic_bench_base: upstream + [db basic bench patch for IteratorNext](https://github.com/facebook/rocksdb/compare/main...hx235:rocksdb:micro_bench_async_read)
- asyncread_db_basic_bench_pr: asyncread_db_basic_bench_base + this PR
**Test**
Get
```
TEST_TMPDIR=/dev/shm ./db_basic_bench_{null_stat|base|pr} --benchmark_filter=DBGet/comp_style:0/max_data:134217728/per_key_size:256/enable_statistics:1/negative_query:0/enable_filter:0/mmap:1/threads:1 --benchmark_repetitions=1000
```
Result
```
Coming soon
```
AsyncRead
```
TEST_TMPDIR=/dev/shm ./asyncread_db_basic_bench_{base|pr} --benchmark_filter=IteratorNext/comp_style:0/max_data:134217728/per_key_size:256/enable_statistics:1/async_io:1/include_detailed_timers:0 --benchmark_repetitions=1000 > syncread_db_basic_bench_{base|pr}.out
```
Result
```
Base:
1956,1956,1968,1977,1979,1986,1988,1988,1988,1990,1991,1991,1993,1993,1993,1993,1994,1996,1997,1997,1997,1998,1999,2001,2001,2002,2004,2007,2007,2008,
PR (2.3% regression, due to measuring `SST_READ_MICRO` that wasn't measured before):
1993,2014,2016,2022,2024,2027,2027,2028,2028,2030,2031,2031,2032,2032,2038,2039,2042,2044,2044,2047,2047,2047,2048,2049,2050,2052,2052,2052,2053,2053,
```
Reviewed By: ajkr
Differential Revision: D45918925
Pulled By: hx235
fbshipit-source-id: 58a54560d9ebeb3a59b6d807639692614dad058a
280 lines
9.1 KiB
C++
280 lines
9.1 KiB
C++
// 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).
|
|
|
|
#include "utilities/blob_db/blob_dump_tool.h"
|
|
|
|
#include <stdio.h>
|
|
|
|
#include <cinttypes>
|
|
#include <iostream>
|
|
#include <memory>
|
|
#include <string>
|
|
|
|
#include "file/random_access_file_reader.h"
|
|
#include "file/readahead_raf.h"
|
|
#include "port/port.h"
|
|
#include "rocksdb/convenience.h"
|
|
#include "rocksdb/file_system.h"
|
|
#include "table/format.h"
|
|
#include "util/coding.h"
|
|
#include "util/string_util.h"
|
|
|
|
namespace ROCKSDB_NAMESPACE {
|
|
namespace blob_db {
|
|
|
|
BlobDumpTool::BlobDumpTool()
|
|
: reader_(nullptr), buffer_(nullptr), buffer_size_(0) {}
|
|
|
|
Status BlobDumpTool::Run(const std::string& filename, DisplayType show_key,
|
|
DisplayType show_blob,
|
|
DisplayType show_uncompressed_blob,
|
|
bool show_summary) {
|
|
constexpr size_t kReadaheadSize = 2 * 1024 * 1024;
|
|
Status s;
|
|
const auto fs = FileSystem::Default();
|
|
IOOptions io_opts;
|
|
s = fs->FileExists(filename, io_opts, nullptr);
|
|
if (!s.ok()) {
|
|
return s;
|
|
}
|
|
uint64_t file_size = 0;
|
|
s = fs->GetFileSize(filename, io_opts, &file_size, nullptr);
|
|
if (!s.ok()) {
|
|
return s;
|
|
}
|
|
std::unique_ptr<FSRandomAccessFile> file;
|
|
s = fs->NewRandomAccessFile(filename, FileOptions(), &file, nullptr);
|
|
if (!s.ok()) {
|
|
return s;
|
|
}
|
|
file = NewReadaheadRandomAccessFile(std::move(file), kReadaheadSize);
|
|
if (file_size == 0) {
|
|
return Status::Corruption("File is empty.");
|
|
}
|
|
reader_.reset(new RandomAccessFileReader(std::move(file), filename));
|
|
uint64_t offset = 0;
|
|
uint64_t footer_offset = 0;
|
|
CompressionType compression = kNoCompression;
|
|
s = DumpBlobLogHeader(&offset, &compression);
|
|
if (!s.ok()) {
|
|
return s;
|
|
}
|
|
s = DumpBlobLogFooter(file_size, &footer_offset);
|
|
if (!s.ok()) {
|
|
return s;
|
|
}
|
|
uint64_t total_records = 0;
|
|
uint64_t total_key_size = 0;
|
|
uint64_t total_blob_size = 0;
|
|
uint64_t total_uncompressed_blob_size = 0;
|
|
if (show_key != DisplayType::kNone || show_summary) {
|
|
while (offset < footer_offset) {
|
|
s = DumpRecord(show_key, show_blob, show_uncompressed_blob, show_summary,
|
|
compression, &offset, &total_records, &total_key_size,
|
|
&total_blob_size, &total_uncompressed_blob_size);
|
|
if (!s.ok()) {
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
if (show_summary) {
|
|
fprintf(stdout, "Summary:\n");
|
|
fprintf(stdout, " total records: %" PRIu64 "\n", total_records);
|
|
fprintf(stdout, " total key size: %" PRIu64 "\n", total_key_size);
|
|
fprintf(stdout, " total blob size: %" PRIu64 "\n", total_blob_size);
|
|
if (compression != kNoCompression) {
|
|
fprintf(stdout, " total raw blob size: %" PRIu64 "\n",
|
|
total_uncompressed_blob_size);
|
|
}
|
|
}
|
|
return s;
|
|
}
|
|
|
|
Status BlobDumpTool::Read(uint64_t offset, size_t size, Slice* result) {
|
|
if (buffer_size_ < size) {
|
|
if (buffer_size_ == 0) {
|
|
buffer_size_ = 4096;
|
|
}
|
|
while (buffer_size_ < size) {
|
|
buffer_size_ *= 2;
|
|
}
|
|
buffer_.reset(new char[buffer_size_]);
|
|
}
|
|
Status s =
|
|
reader_->Read(IOOptions(), offset, size, result, buffer_.get(), nullptr);
|
|
if (!s.ok()) {
|
|
return s;
|
|
}
|
|
if (result->size() != size) {
|
|
return Status::Corruption("Reach the end of the file unexpectedly.");
|
|
}
|
|
return s;
|
|
}
|
|
|
|
Status BlobDumpTool::DumpBlobLogHeader(uint64_t* offset,
|
|
CompressionType* compression) {
|
|
Slice slice;
|
|
Status s = Read(0, BlobLogHeader::kSize, &slice);
|
|
if (!s.ok()) {
|
|
return s;
|
|
}
|
|
BlobLogHeader header;
|
|
s = header.DecodeFrom(slice);
|
|
if (!s.ok()) {
|
|
return s;
|
|
}
|
|
fprintf(stdout, "Blob log header:\n");
|
|
fprintf(stdout, " Version : %" PRIu32 "\n", header.version);
|
|
fprintf(stdout, " Column Family ID : %" PRIu32 "\n",
|
|
header.column_family_id);
|
|
std::string compression_str;
|
|
if (!GetStringFromCompressionType(&compression_str, header.compression)
|
|
.ok()) {
|
|
compression_str = "Unrecongnized compression type (" +
|
|
std::to_string((int)header.compression) + ")";
|
|
}
|
|
fprintf(stdout, " Compression : %s\n", compression_str.c_str());
|
|
fprintf(stdout, " Expiration range : %s\n",
|
|
GetString(header.expiration_range).c_str());
|
|
*offset = BlobLogHeader::kSize;
|
|
*compression = header.compression;
|
|
return s;
|
|
}
|
|
|
|
Status BlobDumpTool::DumpBlobLogFooter(uint64_t file_size,
|
|
uint64_t* footer_offset) {
|
|
auto no_footer = [&]() {
|
|
*footer_offset = file_size;
|
|
fprintf(stdout, "No blob log footer.\n");
|
|
return Status::OK();
|
|
};
|
|
if (file_size < BlobLogHeader::kSize + BlobLogFooter::kSize) {
|
|
return no_footer();
|
|
}
|
|
Slice slice;
|
|
*footer_offset = file_size - BlobLogFooter::kSize;
|
|
Status s = Read(*footer_offset, BlobLogFooter::kSize, &slice);
|
|
if (!s.ok()) {
|
|
return s;
|
|
}
|
|
BlobLogFooter footer;
|
|
s = footer.DecodeFrom(slice);
|
|
if (!s.ok()) {
|
|
return no_footer();
|
|
}
|
|
fprintf(stdout, "Blob log footer:\n");
|
|
fprintf(stdout, " Blob count : %" PRIu64 "\n", footer.blob_count);
|
|
fprintf(stdout, " Expiration Range : %s\n",
|
|
GetString(footer.expiration_range).c_str());
|
|
return s;
|
|
}
|
|
|
|
Status BlobDumpTool::DumpRecord(DisplayType show_key, DisplayType show_blob,
|
|
DisplayType show_uncompressed_blob,
|
|
bool show_summary, CompressionType compression,
|
|
uint64_t* offset, uint64_t* total_records,
|
|
uint64_t* total_key_size,
|
|
uint64_t* total_blob_size,
|
|
uint64_t* total_uncompressed_blob_size) {
|
|
if (show_key != DisplayType::kNone) {
|
|
fprintf(stdout, "Read record with offset 0x%" PRIx64 " (%" PRIu64 "):\n",
|
|
*offset, *offset);
|
|
}
|
|
Slice slice;
|
|
Status s = Read(*offset, BlobLogRecord::kHeaderSize, &slice);
|
|
if (!s.ok()) {
|
|
return s;
|
|
}
|
|
BlobLogRecord record;
|
|
s = record.DecodeHeaderFrom(slice);
|
|
if (!s.ok()) {
|
|
return s;
|
|
}
|
|
uint64_t key_size = record.key_size;
|
|
uint64_t value_size = record.value_size;
|
|
if (show_key != DisplayType::kNone) {
|
|
fprintf(stdout, " key size : %" PRIu64 "\n", key_size);
|
|
fprintf(stdout, " value size : %" PRIu64 "\n", value_size);
|
|
fprintf(stdout, " expiration : %" PRIu64 "\n", record.expiration);
|
|
}
|
|
*offset += BlobLogRecord::kHeaderSize;
|
|
s = Read(*offset, static_cast<size_t>(key_size + value_size), &slice);
|
|
if (!s.ok()) {
|
|
return s;
|
|
}
|
|
// Decompress value
|
|
std::string uncompressed_value;
|
|
if (compression != kNoCompression &&
|
|
(show_uncompressed_blob != DisplayType::kNone || show_summary)) {
|
|
BlockContents contents;
|
|
UncompressionContext context(compression);
|
|
UncompressionInfo info(context, UncompressionDict::GetEmptyDict(),
|
|
compression);
|
|
s = UncompressBlockData(
|
|
info, slice.data() + key_size, static_cast<size_t>(value_size),
|
|
&contents, 2 /*compress_format_version*/, ImmutableOptions(Options()));
|
|
if (!s.ok()) {
|
|
return s;
|
|
}
|
|
uncompressed_value = contents.data.ToString();
|
|
}
|
|
if (show_key != DisplayType::kNone) {
|
|
fprintf(stdout, " key : ");
|
|
DumpSlice(Slice(slice.data(), static_cast<size_t>(key_size)), show_key);
|
|
if (show_blob != DisplayType::kNone) {
|
|
fprintf(stdout, " blob : ");
|
|
DumpSlice(Slice(slice.data() + static_cast<size_t>(key_size),
|
|
static_cast<size_t>(value_size)),
|
|
show_blob);
|
|
}
|
|
if (show_uncompressed_blob != DisplayType::kNone) {
|
|
fprintf(stdout, " raw blob : ");
|
|
DumpSlice(Slice(uncompressed_value), show_uncompressed_blob);
|
|
}
|
|
}
|
|
*offset += key_size + value_size;
|
|
*total_records += 1;
|
|
*total_key_size += key_size;
|
|
*total_blob_size += value_size;
|
|
*total_uncompressed_blob_size += uncompressed_value.size();
|
|
return s;
|
|
}
|
|
|
|
void BlobDumpTool::DumpSlice(const Slice s, DisplayType type) {
|
|
if (type == DisplayType::kRaw) {
|
|
fprintf(stdout, "%s\n", s.ToString().c_str());
|
|
} else if (type == DisplayType::kHex) {
|
|
fprintf(stdout, "%s\n", s.ToString(true /*hex*/).c_str());
|
|
} else if (type == DisplayType::kDetail) {
|
|
char buf[100];
|
|
for (size_t i = 0; i < s.size(); i += 16) {
|
|
memset(buf, 0, sizeof(buf));
|
|
for (size_t j = 0; j < 16 && i + j < s.size(); j++) {
|
|
unsigned char c = s[i + j];
|
|
snprintf(buf + j * 3 + 15, 2, "%x", c >> 4);
|
|
snprintf(buf + j * 3 + 16, 2, "%x", c & 0xf);
|
|
snprintf(buf + j + 65, 2, "%c", (0x20 <= c && c <= 0x7e) ? c : '.');
|
|
}
|
|
for (size_t p = 0; p + 1 < sizeof(buf); p++) {
|
|
if (buf[p] == 0) {
|
|
buf[p] = ' ';
|
|
}
|
|
}
|
|
fprintf(stdout, "%s\n", i == 0 ? buf + 15 : buf);
|
|
}
|
|
}
|
|
}
|
|
|
|
template <class T>
|
|
std::string BlobDumpTool::GetString(std::pair<T, T> p) {
|
|
if (p.first == 0 && p.second == 0) {
|
|
return "nil";
|
|
}
|
|
return "(" + std::to_string(p.first) + ", " + std::to_string(p.second) + ")";
|
|
}
|
|
|
|
} // namespace blob_db
|
|
} // namespace ROCKSDB_NAMESPACE
|