bitcoin/src/dbwrapper.cpp
Maciej S. Szmigiero b73d331937 dbwrapper: Bump max file size to 32 MiB
The default max file size for LevelDB is 2 MiB, which results in the
LevelDB compaction code generating ~4 disk cache flushes per second when
syncing with the Bitcoin network.
These disk cache flushes are triggered by fdatasync() syscall issued by the
LevelDB compaction code when reaching the max file size.

If the database is on a HDD this flush rate brings the whole system to a
crawl.
It also results in very slow throughput since 2 MiB * 4 flushes per second
is about 8 MiB / second max throughput, while even an old HDD can pull
100 - 200 MiB / second streaming throughput.

Increase the max file size for LevelDB to 32 MiB instead so the flush rate
drops significantly and the system no longer gets so sluggish.

The new max file size value chosen is a compromise between the one that
works best for HDD and SSD performance, as determined by benchmarks done by
various people.
2024-11-30 20:19:08 +01:00

427 lines
15 KiB
C++

// Copyright (c) 2012-2022 The Bitcoin Core developers
// Distributed under the MIT software license, see the accompanying
// file COPYING or http://www.opensource.org/licenses/mit-license.php.
#include <dbwrapper.h>
#include <logging.h>
#include <random.h>
#include <serialize.h>
#include <span.h>
#include <streams.h>
#include <util/fs.h>
#include <util/fs_helpers.h>
#include <util/strencodings.h>
#include <algorithm>
#include <cassert>
#include <cstdarg>
#include <cstdint>
#include <cstdio>
#include <leveldb/cache.h>
#include <leveldb/db.h>
#include <leveldb/env.h>
#include <leveldb/filter_policy.h>
#include <leveldb/helpers/memenv/memenv.h>
#include <leveldb/iterator.h>
#include <leveldb/options.h>
#include <leveldb/slice.h>
#include <leveldb/status.h>
#include <leveldb/write_batch.h>
#include <memory>
#include <optional>
#include <utility>
static auto CharCast(const std::byte* data) { return reinterpret_cast<const char*>(data); }
bool DestroyDB(const std::string& path_str)
{
return leveldb::DestroyDB(path_str, {}).ok();
}
/** Handle database error by throwing dbwrapper_error exception.
*/
static void HandleError(const leveldb::Status& status)
{
if (status.ok())
return;
const std::string errmsg = "Fatal LevelDB error: " + status.ToString();
LogPrintf("%s\n", errmsg);
LogPrintf("You can use -debug=leveldb to get more complete diagnostic messages\n");
throw dbwrapper_error(errmsg);
}
class CBitcoinLevelDBLogger : public leveldb::Logger {
public:
// This code is adapted from posix_logger.h, which is why it is using vsprintf.
// Please do not do this in normal code
void Logv(const char * format, va_list ap) override {
if (!LogAcceptCategory(BCLog::LEVELDB, BCLog::Level::Debug)) {
return;
}
char buffer[500];
for (int iter = 0; iter < 2; iter++) {
char* base;
int bufsize;
if (iter == 0) {
bufsize = sizeof(buffer);
base = buffer;
}
else {
bufsize = 30000;
base = new char[bufsize];
}
char* p = base;
char* limit = base + bufsize;
// Print the message
if (p < limit) {
va_list backup_ap;
va_copy(backup_ap, ap);
// Do not use vsnprintf elsewhere in bitcoin source code, see above.
p += vsnprintf(p, limit - p, format, backup_ap);
va_end(backup_ap);
}
// Truncate to available space if necessary
if (p >= limit) {
if (iter == 0) {
continue; // Try again with larger buffer
}
else {
p = limit - 1;
}
}
// Add newline if necessary
if (p == base || p[-1] != '\n') {
*p++ = '\n';
}
assert(p <= limit);
base[std::min(bufsize - 1, (int)(p - base))] = '\0';
LogDebug(BCLog::LEVELDB, "%s\n", util::RemoveSuffixView(base, "\n"));
if (base != buffer) {
delete[] base;
}
break;
}
}
};
static void SetMaxOpenFiles(leveldb::Options *options) {
// On most platforms the default setting of max_open_files (which is 1000)
// is optimal. On Windows using a large file count is OK because the handles
// do not interfere with select() loops. On 64-bit Unix hosts this value is
// also OK, because up to that amount LevelDB will use an mmap
// implementation that does not use extra file descriptors (the fds are
// closed after being mmap'ed).
//
// Increasing the value beyond the default is dangerous because LevelDB will
// fall back to a non-mmap implementation when the file count is too large.
// On 32-bit Unix host we should decrease the value because the handles use
// up real fds, and we want to avoid fd exhaustion issues.
//
// See PR #12495 for further discussion.
int default_open_files = options->max_open_files;
#ifndef WIN32
if (sizeof(void*) < 8) {
options->max_open_files = 64;
}
#endif
LogDebug(BCLog::LEVELDB, "LevelDB using max_open_files=%d (default=%d)\n",
options->max_open_files, default_open_files);
}
static leveldb::Options GetOptions(size_t nCacheSize)
{
leveldb::Options options;
options.block_cache = leveldb::NewLRUCache(nCacheSize / 2);
options.write_buffer_size = nCacheSize / 4; // up to two write buffers may be held in memory simultaneously
options.filter_policy = leveldb::NewBloomFilterPolicy(10);
options.compression = leveldb::kNoCompression;
options.info_log = new CBitcoinLevelDBLogger();
if (leveldb::kMajorVersion > 1 || (leveldb::kMajorVersion == 1 && leveldb::kMinorVersion >= 16)) {
// LevelDB versions before 1.16 consider short writes to be corruption. Only trigger error
// on corruption in later versions.
options.paranoid_checks = true;
}
options.max_file_size = std::max(options.max_file_size, DBWRAPPER_MAX_FILE_SIZE);
SetMaxOpenFiles(&options);
return options;
}
struct CDBBatch::WriteBatchImpl {
leveldb::WriteBatch batch;
};
CDBBatch::CDBBatch(const CDBWrapper& _parent)
: parent{_parent},
m_impl_batch{std::make_unique<CDBBatch::WriteBatchImpl>()} {};
CDBBatch::~CDBBatch() = default;
void CDBBatch::Clear()
{
m_impl_batch->batch.Clear();
size_estimate = 0;
}
void CDBBatch::WriteImpl(Span<const std::byte> key, DataStream& ssValue)
{
leveldb::Slice slKey(CharCast(key.data()), key.size());
ssValue.Xor(dbwrapper_private::GetObfuscateKey(parent));
leveldb::Slice slValue(CharCast(ssValue.data()), ssValue.size());
m_impl_batch->batch.Put(slKey, slValue);
// LevelDB serializes writes as:
// - byte: header
// - varint: key length (1 byte up to 127B, 2 bytes up to 16383B, ...)
// - byte[]: key
// - varint: value length
// - byte[]: value
// The formula below assumes the key and value are both less than 16k.
size_estimate += 3 + (slKey.size() > 127) + slKey.size() + (slValue.size() > 127) + slValue.size();
}
void CDBBatch::EraseImpl(Span<const std::byte> key)
{
leveldb::Slice slKey(CharCast(key.data()), key.size());
m_impl_batch->batch.Delete(slKey);
// LevelDB serializes erases as:
// - byte: header
// - varint: key length
// - byte[]: key
// The formula below assumes the key is less than 16kB.
size_estimate += 2 + (slKey.size() > 127) + slKey.size();
}
struct LevelDBContext {
//! custom environment this database is using (may be nullptr in case of default environment)
leveldb::Env* penv;
//! database options used
leveldb::Options options;
//! options used when reading from the database
leveldb::ReadOptions readoptions;
//! options used when iterating over values of the database
leveldb::ReadOptions iteroptions;
//! options used when writing to the database
leveldb::WriteOptions writeoptions;
//! options used when sync writing to the database
leveldb::WriteOptions syncoptions;
//! the database itself
leveldb::DB* pdb;
};
CDBWrapper::CDBWrapper(const DBParams& params)
: m_db_context{std::make_unique<LevelDBContext>()}, m_name{fs::PathToString(params.path.stem())}, m_path{params.path}, m_is_memory{params.memory_only}
{
DBContext().penv = nullptr;
DBContext().readoptions.verify_checksums = true;
DBContext().iteroptions.verify_checksums = true;
DBContext().iteroptions.fill_cache = false;
DBContext().syncoptions.sync = true;
DBContext().options = GetOptions(params.cache_bytes);
DBContext().options.create_if_missing = true;
if (params.memory_only) {
DBContext().penv = leveldb::NewMemEnv(leveldb::Env::Default());
DBContext().options.env = DBContext().penv;
} else {
if (params.wipe_data) {
LogPrintf("Wiping LevelDB in %s\n", fs::PathToString(params.path));
leveldb::Status result = leveldb::DestroyDB(fs::PathToString(params.path), DBContext().options);
HandleError(result);
}
TryCreateDirectories(params.path);
LogPrintf("Opening LevelDB in %s\n", fs::PathToString(params.path));
}
// PathToString() return value is safe to pass to leveldb open function,
// because on POSIX leveldb passes the byte string directly to ::open(), and
// on Windows it converts from UTF-8 to UTF-16 before calling ::CreateFileW
// (see env_posix.cc and env_windows.cc).
leveldb::Status status = leveldb::DB::Open(DBContext().options, fs::PathToString(params.path), &DBContext().pdb);
HandleError(status);
LogPrintf("Opened LevelDB successfully\n");
if (params.options.force_compact) {
LogPrintf("Starting database compaction of %s\n", fs::PathToString(params.path));
DBContext().pdb->CompactRange(nullptr, nullptr);
LogPrintf("Finished database compaction of %s\n", fs::PathToString(params.path));
}
// The base-case obfuscation key, which is a noop.
obfuscate_key = std::vector<unsigned char>(OBFUSCATE_KEY_NUM_BYTES, '\000');
bool key_exists = Read(OBFUSCATE_KEY_KEY, obfuscate_key);
if (!key_exists && params.obfuscate && IsEmpty()) {
// Initialize non-degenerate obfuscation if it won't upset
// existing, non-obfuscated data.
std::vector<unsigned char> new_key = CreateObfuscateKey();
// Write `new_key` so we don't obfuscate the key with itself
Write(OBFUSCATE_KEY_KEY, new_key);
obfuscate_key = new_key;
LogPrintf("Wrote new obfuscate key for %s: %s\n", fs::PathToString(params.path), HexStr(obfuscate_key));
}
LogPrintf("Using obfuscation key for %s: %s\n", fs::PathToString(params.path), HexStr(obfuscate_key));
}
CDBWrapper::~CDBWrapper()
{
delete DBContext().pdb;
DBContext().pdb = nullptr;
delete DBContext().options.filter_policy;
DBContext().options.filter_policy = nullptr;
delete DBContext().options.info_log;
DBContext().options.info_log = nullptr;
delete DBContext().options.block_cache;
DBContext().options.block_cache = nullptr;
delete DBContext().penv;
DBContext().options.env = nullptr;
}
bool CDBWrapper::WriteBatch(CDBBatch& batch, bool fSync)
{
const bool log_memory = LogAcceptCategory(BCLog::LEVELDB, BCLog::Level::Debug);
double mem_before = 0;
if (log_memory) {
mem_before = DynamicMemoryUsage() / 1024.0 / 1024;
}
leveldb::Status status = DBContext().pdb->Write(fSync ? DBContext().syncoptions : DBContext().writeoptions, &batch.m_impl_batch->batch);
HandleError(status);
if (log_memory) {
double mem_after = DynamicMemoryUsage() / 1024.0 / 1024;
LogDebug(BCLog::LEVELDB, "WriteBatch memory usage: db=%s, before=%.1fMiB, after=%.1fMiB\n",
m_name, mem_before, mem_after);
}
return true;
}
size_t CDBWrapper::DynamicMemoryUsage() const
{
std::string memory;
std::optional<size_t> parsed;
if (!DBContext().pdb->GetProperty("leveldb.approximate-memory-usage", &memory) || !(parsed = ToIntegral<size_t>(memory))) {
LogDebug(BCLog::LEVELDB, "Failed to get approximate-memory-usage property\n");
return 0;
}
return parsed.value();
}
// Prefixed with null character to avoid collisions with other keys
//
// We must use a string constructor which specifies length so that we copy
// past the null-terminator.
const std::string CDBWrapper::OBFUSCATE_KEY_KEY("\000obfuscate_key", 14);
const unsigned int CDBWrapper::OBFUSCATE_KEY_NUM_BYTES = 8;
/**
* Returns a string (consisting of 8 random bytes) suitable for use as an
* obfuscating XOR key.
*/
std::vector<unsigned char> CDBWrapper::CreateObfuscateKey() const
{
std::vector<uint8_t> ret(OBFUSCATE_KEY_NUM_BYTES);
GetRandBytes(ret);
return ret;
}
std::optional<std::string> CDBWrapper::ReadImpl(Span<const std::byte> key) const
{
leveldb::Slice slKey(CharCast(key.data()), key.size());
std::string strValue;
leveldb::Status status = DBContext().pdb->Get(DBContext().readoptions, slKey, &strValue);
if (!status.ok()) {
if (status.IsNotFound())
return std::nullopt;
LogPrintf("LevelDB read failure: %s\n", status.ToString());
HandleError(status);
}
return strValue;
}
bool CDBWrapper::ExistsImpl(Span<const std::byte> key) const
{
leveldb::Slice slKey(CharCast(key.data()), key.size());
std::string strValue;
leveldb::Status status = DBContext().pdb->Get(DBContext().readoptions, slKey, &strValue);
if (!status.ok()) {
if (status.IsNotFound())
return false;
LogPrintf("LevelDB read failure: %s\n", status.ToString());
HandleError(status);
}
return true;
}
size_t CDBWrapper::EstimateSizeImpl(Span<const std::byte> key1, Span<const std::byte> key2) const
{
leveldb::Slice slKey1(CharCast(key1.data()), key1.size());
leveldb::Slice slKey2(CharCast(key2.data()), key2.size());
uint64_t size = 0;
leveldb::Range range(slKey1, slKey2);
DBContext().pdb->GetApproximateSizes(&range, 1, &size);
return size;
}
bool CDBWrapper::IsEmpty()
{
std::unique_ptr<CDBIterator> it(NewIterator());
it->SeekToFirst();
return !(it->Valid());
}
struct CDBIterator::IteratorImpl {
const std::unique_ptr<leveldb::Iterator> iter;
explicit IteratorImpl(leveldb::Iterator* _iter) : iter{_iter} {}
};
CDBIterator::CDBIterator(const CDBWrapper& _parent, std::unique_ptr<IteratorImpl> _piter) : parent(_parent),
m_impl_iter(std::move(_piter)) {}
CDBIterator* CDBWrapper::NewIterator()
{
return new CDBIterator{*this, std::make_unique<CDBIterator::IteratorImpl>(DBContext().pdb->NewIterator(DBContext().iteroptions))};
}
void CDBIterator::SeekImpl(Span<const std::byte> key)
{
leveldb::Slice slKey(CharCast(key.data()), key.size());
m_impl_iter->iter->Seek(slKey);
}
Span<const std::byte> CDBIterator::GetKeyImpl() const
{
return MakeByteSpan(m_impl_iter->iter->key());
}
Span<const std::byte> CDBIterator::GetValueImpl() const
{
return MakeByteSpan(m_impl_iter->iter->value());
}
CDBIterator::~CDBIterator() = default;
bool CDBIterator::Valid() const { return m_impl_iter->iter->Valid(); }
void CDBIterator::SeekToFirst() { m_impl_iter->iter->SeekToFirst(); }
void CDBIterator::Next() { m_impl_iter->iter->Next(); }
namespace dbwrapper_private {
const std::vector<unsigned char>& GetObfuscateKey(const CDBWrapper &w)
{
return w.obfuscate_key;
}
} // namespace dbwrapper_private