mirror of
https://github.com/matrix-construct/construct
synced 2024-11-15 22:41:12 +01:00
1439 lines
33 KiB
C++
1439 lines
33 KiB
C++
/*
|
|
* Copyright (C) 2016 Charybdis Development Team
|
|
* Copyright (C) 2016 Jason Volk <jason@zemos.net>
|
|
*
|
|
* Permission to use, copy, modify, and/or distribute this software for any
|
|
* purpose with or without fee is hereby granted, provided that the above
|
|
* copyright notice and this permission notice is present in all copies.
|
|
*
|
|
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
|
|
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
|
|
* WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
|
|
* DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
|
|
* INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
|
|
* (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
|
|
* SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
|
|
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
|
|
* STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
|
|
* IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
|
|
* POSSIBILITY OF SUCH DAMAGE.
|
|
*
|
|
*/
|
|
|
|
#include <rocksdb/db.h>
|
|
#include <rocksdb/cache.h>
|
|
#include <rocksdb/merge_operator.h>
|
|
#include <rocksdb/perf_level.h>
|
|
#include <rocksdb/listener.h>
|
|
#include <rocksdb/statistics.h>
|
|
#include <rocksdb/convenience.h>
|
|
#include <rocksdb/env.h>
|
|
|
|
namespace ircd {
|
|
namespace db {
|
|
|
|
struct log::log log
|
|
{
|
|
"db", 'D' // Database subsystem takes SNOMASK +D
|
|
};
|
|
|
|
const std::string &reflect(const rocksdb::Tickers &);
|
|
const std::string &reflect(const rocksdb::Histograms &);
|
|
|
|
void throw_on_error(const rocksdb::Status &);
|
|
bool valid(const rocksdb::Iterator &);
|
|
void valid_or_throw(const rocksdb::Iterator &);
|
|
void valid_equal_or_throw(const rocksdb::Iterator &, const rocksdb::Slice &);
|
|
|
|
const auto BLOCKING = rocksdb::ReadTier::kReadAllTier;
|
|
const auto NON_BLOCKING = rocksdb::ReadTier::kBlockCacheTier;
|
|
enum class pos
|
|
{
|
|
FRONT = -2, // .front() | first element
|
|
PREV = -1, // std::prev() | previous element
|
|
END = 0, // break; | exit iteration (or past the end)
|
|
NEXT = 1, // continue; | next element
|
|
BACK = 2, // .back() | last element
|
|
};
|
|
void seek(rocksdb::Iterator &, const pos &);
|
|
void seek(rocksdb::Iterator &, const rocksdb::Slice &);
|
|
template<class pos> void seek(rocksdb::DB &, rocksdb::ColumnFamilyHandle *const &, const pos &, rocksdb::ReadOptions &, std::unique_ptr<rocksdb::Iterator> &);
|
|
std::unique_ptr<rocksdb::Iterator> seek(rocksdb::DB &, rocksdb::ColumnFamilyHandle *const &, const rocksdb::Slice &, rocksdb::ReadOptions &);
|
|
|
|
// This is important to prevent thrashing the iterators which have to reset on iops
|
|
const auto DEFAULT_READAHEAD = 4_MiB;
|
|
|
|
rocksdb::WriteOptions make_opts(const sopts &);
|
|
rocksdb::ReadOptions make_opts(const gopts &, const bool &iterator = false);
|
|
rocksdb::Options make_opts(const opts &);
|
|
|
|
std::vector<std::string> column_names(const std::string &path, const rocksdb::DBOptions &);
|
|
std::vector<std::string> column_names(const std::string &path, const opts &);
|
|
|
|
struct meta
|
|
:std::enable_shared_from_this<meta>
|
|
{
|
|
static std::map<std::string, meta *> dbs; // open databases
|
|
|
|
struct logs;
|
|
struct stats;
|
|
struct events;
|
|
struct mergeop;
|
|
|
|
std::string name;
|
|
std::string path;
|
|
rocksdb::Options opts;
|
|
std::shared_ptr<struct logs> logs;
|
|
std::shared_ptr<struct stats> stats;
|
|
std::shared_ptr<struct events> events;
|
|
std::shared_ptr<struct mergeop> mergeop;
|
|
std::shared_ptr<rocksdb::Cache> cache;
|
|
std::vector<rocksdb::ColumnFamilyDescriptor> columns;
|
|
std::vector<rocksdb::ColumnFamilyHandle *> handles;
|
|
custom_ptr<rocksdb::DB> d;
|
|
|
|
meta(const std::string &name,
|
|
const db::opts &opts,
|
|
merge_function mf);
|
|
|
|
~meta() noexcept;
|
|
};
|
|
|
|
struct meta::logs
|
|
:std::enable_shared_from_this<struct meta::logs>
|
|
,rocksdb::Logger
|
|
{
|
|
struct meta *meta;
|
|
|
|
// Logger
|
|
void Logv(const rocksdb::InfoLogLevel level, const char *fmt, va_list ap) override;
|
|
void Logv(const char *fmt, va_list ap) override;
|
|
void LogHeader(const char *fmt, va_list ap) override;
|
|
|
|
logs(struct meta *const &meta)
|
|
:meta{meta}
|
|
{}
|
|
};
|
|
|
|
struct meta::stats
|
|
:std::enable_shared_from_this<struct meta::stats>
|
|
,rocksdb::Statistics
|
|
{
|
|
struct meta *meta;
|
|
std::array<uint64_t, rocksdb::TICKER_ENUM_MAX> ticker {{0}};
|
|
std::array<rocksdb::HistogramData, rocksdb::HISTOGRAM_ENUM_MAX> histogram;
|
|
|
|
uint64_t getTickerCount(const uint32_t tickerType) const override;
|
|
void recordTick(const uint32_t tickerType, const uint64_t count) override;
|
|
void setTickerCount(const uint32_t tickerType, const uint64_t count) override;
|
|
void histogramData(const uint32_t type, rocksdb::HistogramData *) const override;
|
|
void measureTime(const uint32_t histogramType, const uint64_t time) override;
|
|
bool HistEnabledForType(const uint32_t type) const override;
|
|
|
|
stats(struct meta *const &meta)
|
|
:meta{meta}
|
|
{}
|
|
};
|
|
|
|
struct meta::events
|
|
:std::enable_shared_from_this<struct meta::events>
|
|
,rocksdb::EventListener
|
|
{
|
|
struct meta *meta;
|
|
|
|
void OnFlushCompleted(rocksdb::DB *, const rocksdb::FlushJobInfo &) override;
|
|
void OnCompactionCompleted(rocksdb::DB *, const rocksdb::CompactionJobInfo &) override;
|
|
void OnTableFileDeleted(const rocksdb::TableFileDeletionInfo &) override;
|
|
void OnTableFileCreated(const rocksdb::TableFileCreationInfo &) override;
|
|
void OnTableFileCreationStarted(const rocksdb::TableFileCreationBriefInfo &) override;
|
|
void OnMemTableSealed(const rocksdb::MemTableInfo &) override;
|
|
void OnColumnFamilyHandleDeletionStarted(rocksdb::ColumnFamilyHandle *) override;
|
|
|
|
events(struct meta *const &meta)
|
|
:meta{meta}
|
|
{}
|
|
};
|
|
|
|
struct meta::mergeop
|
|
:std::enable_shared_from_this<struct meta::mergeop>
|
|
,rocksdb::AssociativeMergeOperator
|
|
{
|
|
struct meta *meta;
|
|
merge_function merger;
|
|
|
|
bool Merge(const rocksdb::Slice &, const rocksdb::Slice *, const rocksdb::Slice &, std::string *, rocksdb::Logger *) const override;
|
|
const char *Name() const override;
|
|
|
|
mergeop(struct meta *const &meta, merge_function merger = nullptr)
|
|
:meta{meta}
|
|
,merger{std::move(merger)}
|
|
{}
|
|
};
|
|
|
|
std::map<std::string, meta *>
|
|
meta::dbs
|
|
{};
|
|
|
|
} // namespace db
|
|
} // namespace ircd
|
|
|
|
using namespace ircd;
|
|
|
|
db::init::init()
|
|
{
|
|
}
|
|
|
|
db::init::~init()
|
|
noexcept
|
|
{
|
|
}
|
|
|
|
///////////////////////////////////////////////////////////////////////////////
|
|
//
|
|
// Meta
|
|
//
|
|
|
|
db::meta::meta(const std::string &name,
|
|
const db::opts &opts,
|
|
merge_function merger)
|
|
try
|
|
:name{name}
|
|
,path{db::path(name)}
|
|
,opts{make_opts(opts)}
|
|
,logs{std::make_shared<struct logs>(this)}
|
|
,stats{std::make_shared<struct stats>(this)}
|
|
,events{std::make_shared<struct events>(this)}
|
|
,mergeop{std::make_shared<struct mergeop>(this, std::move(merger))}
|
|
,cache{[this, &opts]() -> std::shared_ptr<rocksdb::Cache>
|
|
{
|
|
const auto lru_cache_size
|
|
{
|
|
opt_val(opts, opt::LRU_CACHE)
|
|
};
|
|
|
|
if(lru_cache_size > 0)
|
|
{
|
|
const auto ret(rocksdb::NewLRUCache(lru_cache_size));
|
|
this->opts.row_cache = ret;
|
|
return ret;
|
|
}
|
|
else return {};
|
|
}()}
|
|
,d{[this, &opts, &name]() -> custom_ptr<rocksdb::DB>
|
|
{
|
|
// Setup logging
|
|
logs->SetInfoLogLevel(ircd::debugmode? rocksdb::DEBUG_LEVEL : rocksdb::WARN_LEVEL);
|
|
this->opts.info_log_level = logs->GetInfoLogLevel();
|
|
this->opts.info_log = logs;
|
|
|
|
// Setup event and statistics callbacks
|
|
this->opts.listeners.emplace_back(this->events);
|
|
//this->opts.statistics = this->stats; // broken?
|
|
|
|
// Setup performance metric options
|
|
//rocksdb::SetPerfLevel(rocksdb::PerfLevel::kDisable);
|
|
|
|
// Setup journal recovery options
|
|
this->opts.wal_recovery_mode = rocksdb::WALRecoveryMode::kAbsoluteConsistency;
|
|
//this->opts.wal_recovery_mode = rocksdb::WALRecoveryMode::kPointInTimeRecovery;
|
|
|
|
// Setup user operators
|
|
if(mergeop->merger)
|
|
this->opts.merge_operator = this->mergeop;
|
|
|
|
// Setup column families
|
|
const auto column_names{db::column_names(path, this->opts)};
|
|
this->opts.create_missing_column_families = true;
|
|
for(const auto &name : column_names)
|
|
columns.emplace_back(name, this->opts);
|
|
|
|
// Setup the database closer.
|
|
// We need customization here because of the column family thing.
|
|
const auto deleter([this](rocksdb::DB *const d)
|
|
noexcept
|
|
{
|
|
for(auto it(std::rbegin(this->handles)); it != std::rend(this->handles); ++it)
|
|
d->DestroyColumnFamilyHandle(*it);
|
|
|
|
throw_on_error(d->SyncWAL()); // blocking
|
|
rocksdb::CancelAllBackgroundWork(d, true); // true = blocking
|
|
const auto seq(d->GetLatestSequenceNumber());
|
|
delete d;
|
|
|
|
log.info("'%s': closed database @ `%s' seq[%zu]",
|
|
this->name,
|
|
this->path,
|
|
seq);
|
|
});
|
|
|
|
// Announce attempt before usual point where exceptions are thrown
|
|
log.debug("Opening database \"%s\" @ `%s' columns: %zu",
|
|
name,
|
|
path,
|
|
columns.size());
|
|
|
|
// Open DB into ptr
|
|
rocksdb::DB *ptr;
|
|
if(has_opt(opts, opt::READ_ONLY))
|
|
throw_on_error(rocksdb::DB::OpenForReadOnly(this->opts, path, columns, &handles, &ptr));
|
|
else
|
|
throw_on_error(rocksdb::DB::Open(this->opts, path, columns, &handles, &ptr));
|
|
|
|
// re-establish RAII here
|
|
return { ptr, deleter };
|
|
}()}
|
|
{
|
|
log.info("'%s': Opened database @ `%s' (handle: %p) columns: %zu",
|
|
name,
|
|
path,
|
|
(const void *)this,
|
|
handles.size());
|
|
|
|
dbs.emplace(name, this);
|
|
}
|
|
catch(const invalid_argument &e)
|
|
{
|
|
const bool no_create(has_opt(opts, opt::NO_CREATE));
|
|
const bool no_existing(has_opt(opts, opt::NO_EXISTING));
|
|
const auto helpstr
|
|
{
|
|
no_create? " (The database is missing and will not be created)":
|
|
no_existing? " (The database already exists but must be fresh)":
|
|
""
|
|
};
|
|
|
|
throw error("Failed to open db '%s': %s%s", name, e, helpstr);
|
|
}
|
|
catch(const std::exception &e)
|
|
{
|
|
throw error("Failed to open db '%s': %s", name, e);
|
|
}
|
|
|
|
ircd::db::meta::~meta()
|
|
noexcept
|
|
{
|
|
log.debug("'%s': closing database @ `%s'", name, path);
|
|
dbs.erase(name);
|
|
}
|
|
|
|
static
|
|
ircd::log::facility
|
|
translate(const rocksdb::InfoLogLevel &level)
|
|
{
|
|
switch(level)
|
|
{
|
|
// Treat all infomational messages from rocksdb as debug here for now.
|
|
// We can clean them up and make better reports for our users eventually.
|
|
default:
|
|
case rocksdb::InfoLogLevel::DEBUG_LEVEL: return ircd::log::facility::DEBUG;
|
|
case rocksdb::InfoLogLevel::INFO_LEVEL: return ircd::log::facility::DEBUG;
|
|
|
|
case rocksdb::InfoLogLevel::WARN_LEVEL: return ircd::log::facility::WARNING;
|
|
case rocksdb::InfoLogLevel::ERROR_LEVEL: return ircd::log::facility::ERROR;
|
|
case rocksdb::InfoLogLevel::FATAL_LEVEL: return ircd::log::facility::CRITICAL;
|
|
case rocksdb::InfoLogLevel::HEADER_LEVEL: return ircd::log::facility::NOTICE;
|
|
}
|
|
}
|
|
|
|
void
|
|
ircd::db::meta::logs::Logv(const char *const fmt,
|
|
va_list ap)
|
|
{
|
|
Logv(rocksdb::InfoLogLevel::DEBUG_LEVEL, fmt, ap);
|
|
}
|
|
|
|
void
|
|
ircd::db::meta::logs::LogHeader(const char *const fmt,
|
|
va_list ap)
|
|
{
|
|
Logv(rocksdb::InfoLogLevel::DEBUG_LEVEL, fmt, ap);
|
|
}
|
|
|
|
void
|
|
ircd::db::meta::logs::Logv(const rocksdb::InfoLogLevel level,
|
|
const char *const fmt,
|
|
va_list ap)
|
|
{
|
|
if(level < GetInfoLogLevel())
|
|
return;
|
|
|
|
char buf[1024]; const auto len
|
|
{
|
|
std::vsnprintf(buf, sizeof(buf), fmt, ap)
|
|
};
|
|
|
|
const auto str
|
|
{
|
|
// RocksDB adds annoying leading whitespace to attempt to right-justify things and idc
|
|
lstrip(buf, ' ')
|
|
};
|
|
|
|
log(translate(level), "'%s': (rdb) %s", meta->name, str);
|
|
}
|
|
|
|
const char *
|
|
ircd::db::meta::mergeop::Name()
|
|
const
|
|
{
|
|
return "<unnamed>";
|
|
}
|
|
|
|
bool
|
|
ircd::db::meta::mergeop::Merge(const rocksdb::Slice &_key,
|
|
const rocksdb::Slice *const _exist,
|
|
const rocksdb::Slice &_update,
|
|
std::string *const newval,
|
|
rocksdb::Logger *const)
|
|
const try
|
|
{
|
|
const string_view key
|
|
{
|
|
_key.data(), _key.size()
|
|
};
|
|
|
|
const string_view exist
|
|
{
|
|
_exist? string_view { _exist->data(), _exist->size() } : string_view{}
|
|
};
|
|
|
|
const string_view update
|
|
{
|
|
_update.data(), _update.size()
|
|
};
|
|
|
|
if(exist.empty())
|
|
{
|
|
*newval = std::string(update);
|
|
return true;
|
|
}
|
|
|
|
//XXX caching opportunity?
|
|
*newval = merger(key, {exist, update}); // call the user
|
|
return true;
|
|
}
|
|
catch(const std::bad_function_call &e)
|
|
{
|
|
log.critical("merge: missing merge operator (%s)", e);
|
|
return false;
|
|
}
|
|
catch(const std::exception &e)
|
|
{
|
|
log.error("merge: %s", e);
|
|
return false;
|
|
}
|
|
|
|
bool
|
|
ircd::db::meta::stats::HistEnabledForType(const uint32_t type)
|
|
const
|
|
{
|
|
return type < histogram.size();
|
|
}
|
|
|
|
void
|
|
ircd::db::meta::stats::measureTime(const uint32_t type,
|
|
const uint64_t time)
|
|
{
|
|
}
|
|
|
|
void
|
|
ircd::db::meta::stats::histogramData(const uint32_t type,
|
|
rocksdb::HistogramData *const data)
|
|
const
|
|
{
|
|
assert(data);
|
|
|
|
const auto &median(data->median);
|
|
const auto &percentile95(data->percentile95);
|
|
const auto &percentile88(data->percentile99);
|
|
const auto &average(data->average);
|
|
const auto &standard_deviation(data->standard_deviation);
|
|
}
|
|
|
|
void
|
|
ircd::db::meta::stats::recordTick(const uint32_t type,
|
|
const uint64_t count)
|
|
{
|
|
ticker.at(type) += count;
|
|
}
|
|
|
|
void
|
|
ircd::db::meta::stats::setTickerCount(const uint32_t type,
|
|
const uint64_t count)
|
|
{
|
|
ticker.at(type) = count;
|
|
}
|
|
|
|
uint64_t
|
|
ircd::db::meta::stats::getTickerCount(const uint32_t type)
|
|
const
|
|
{
|
|
return ticker.at(type);
|
|
}
|
|
|
|
void
|
|
ircd::db::meta::events::OnFlushCompleted(rocksdb::DB *const db,
|
|
const rocksdb::FlushJobInfo &info)
|
|
{
|
|
log.debug("'%s' @%p: flushed: column[%s] path[%s] tid[%lu] job[%d] writes[slow:%d stop:%d]",
|
|
meta->name,
|
|
db,
|
|
info.cf_name,
|
|
info.file_path,
|
|
info.thread_id,
|
|
info.job_id,
|
|
info.triggered_writes_slowdown,
|
|
info.triggered_writes_stop);
|
|
}
|
|
|
|
void
|
|
ircd::db::meta::events::OnCompactionCompleted(rocksdb::DB *const db,
|
|
const rocksdb::CompactionJobInfo &info)
|
|
{
|
|
log.debug("'%s' @%p: compacted: column[%s] status[%d] tid[%lu] job[%d]",
|
|
meta->name,
|
|
db,
|
|
info.cf_name,
|
|
int(info.status.code()),
|
|
info.thread_id,
|
|
info.job_id);
|
|
}
|
|
|
|
void
|
|
ircd::db::meta::events::OnTableFileDeleted(const rocksdb::TableFileDeletionInfo &info)
|
|
{
|
|
log.debug("'%s': table file deleted: db[%s] path[%s] status[%d] job[%d]",
|
|
meta->name,
|
|
info.db_name,
|
|
info.file_path,
|
|
int(info.status.code()),
|
|
info.job_id);
|
|
}
|
|
|
|
void
|
|
ircd::db::meta::events::OnTableFileCreated(const rocksdb::TableFileCreationInfo &info)
|
|
{
|
|
log.debug("'%s': table file created: db[%s] path[%s] status[%d] job[%d]",
|
|
meta->name,
|
|
info.db_name,
|
|
info.file_path,
|
|
int(info.status.code()),
|
|
info.job_id);
|
|
}
|
|
|
|
void
|
|
ircd::db::meta::events::OnTableFileCreationStarted(const rocksdb::TableFileCreationBriefInfo &info)
|
|
{
|
|
log.debug("'%s': table file creating: db[%s] column[%s] path[%s] job[%d]",
|
|
meta->name,
|
|
info.db_name,
|
|
info.cf_name,
|
|
info.file_path,
|
|
info.job_id);
|
|
}
|
|
|
|
void
|
|
ircd::db::meta::events::OnMemTableSealed(const rocksdb::MemTableInfo &info)
|
|
{
|
|
log.debug("'%s': memory table sealed: column[%s] entries[%lu] deletes[%lu]",
|
|
meta->name,
|
|
info.cf_name,
|
|
info.num_entries,
|
|
info.num_deletes);
|
|
}
|
|
|
|
void
|
|
ircd::db::meta::events::OnColumnFamilyHandleDeletionStarted(rocksdb::ColumnFamilyHandle *const h)
|
|
{
|
|
log.debug("'%s': column family handle deletion started: %p",
|
|
meta->name,
|
|
h);
|
|
}
|
|
|
|
///////////////////////////////////////////////////////////////////////////////
|
|
//
|
|
// Handle
|
|
//
|
|
|
|
db::handle::handle()
|
|
:meta{}
|
|
,h{nullptr}
|
|
{
|
|
}
|
|
|
|
db::handle::handle(const std::string &name,
|
|
const std::string &column,
|
|
const opts &opts,
|
|
merge_function mf)
|
|
try
|
|
:meta{[&name, &column, &opts, &mf]
|
|
{
|
|
const auto it(meta::dbs.find(name));
|
|
if(it != std::end(meta::dbs))
|
|
{
|
|
const auto &meta(it->second);
|
|
return meta->shared_from_this();
|
|
}
|
|
|
|
return std::make_shared<struct meta>(name, opts, std::move(mf));
|
|
}()}
|
|
,h{[this, &name, &column]
|
|
{
|
|
auto &handles(meta->handles);
|
|
const auto it(std::find_if(std::begin(handles), std::end(handles), [&column]
|
|
(const auto &cf)
|
|
{
|
|
return cf->GetName() == column;
|
|
}));
|
|
|
|
if(it != std::end(handles))
|
|
return *it;
|
|
|
|
log.debug("'%s': Creating new column '%s'", name, column);
|
|
|
|
rocksdb::ColumnFamilyHandle *ret;
|
|
throw_on_error(meta->d->CreateColumnFamily(meta->opts, column, &ret));
|
|
meta->handles.emplace_back(ret);
|
|
return ret;
|
|
}()}
|
|
{
|
|
}
|
|
catch(const std::exception &e)
|
|
{
|
|
throw error("Opening handle to '%s' column '%s': %s", name, column, e);
|
|
}
|
|
|
|
db::handle::handle(handle &&other)
|
|
noexcept
|
|
:meta{std::move(other.meta)}
|
|
,h{std::move(other.h)}
|
|
{
|
|
}
|
|
|
|
db::handle &
|
|
db::handle::operator=(handle &&other)
|
|
noexcept
|
|
{
|
|
meta = std::move(other.meta);
|
|
h = std::move(other.h);
|
|
return *this;
|
|
}
|
|
|
|
db::handle::~handle()
|
|
noexcept
|
|
{
|
|
}
|
|
|
|
void
|
|
db::handle::sync()
|
|
{
|
|
throw_on_error(meta->d->SyncWAL());
|
|
}
|
|
|
|
void
|
|
db::handle::flush(const bool &blocking)
|
|
{
|
|
rocksdb::FlushOptions opts;
|
|
opts.wait = blocking;
|
|
throw_on_error(meta->d->Flush(opts, h));
|
|
}
|
|
|
|
void
|
|
db::handle::del(const string_view &key,
|
|
const sopts &sopts)
|
|
{
|
|
using rocksdb::Slice;
|
|
|
|
auto opts(make_opts(sopts));
|
|
const Slice k(key.data(), key.size());
|
|
throw_on_error(meta->d->Delete(opts, h, k));
|
|
}
|
|
|
|
void
|
|
db::handle::set(const string_view &key,
|
|
const uint8_t *const &buf,
|
|
const size_t &size,
|
|
const sopts &sopts)
|
|
{
|
|
const string_view val{reinterpret_cast<const char *>(buf), size};
|
|
set(key, key, sopts);
|
|
}
|
|
|
|
void
|
|
db::handle::set(const string_view &key,
|
|
const string_view &val,
|
|
const sopts &sopts)
|
|
{
|
|
using rocksdb::Slice;
|
|
|
|
auto opts(make_opts(sopts));
|
|
const Slice k(key.data(), key.size());
|
|
const Slice v(val.data(), val.size());
|
|
throw_on_error(meta->d->Put(opts, h, k, v));
|
|
}
|
|
|
|
std::string
|
|
db::handle::get(const string_view &key,
|
|
const gopts &gopts)
|
|
{
|
|
std::string ret;
|
|
const auto copy([&ret]
|
|
(const string_view &src)
|
|
{
|
|
ret.assign(begin(src), end(src));
|
|
});
|
|
|
|
operator()(key, copy, gopts);
|
|
return ret;
|
|
}
|
|
|
|
size_t
|
|
db::handle::get(const string_view &key,
|
|
uint8_t *const &buf,
|
|
const size_t &max,
|
|
const gopts &gopts)
|
|
{
|
|
size_t ret(0);
|
|
const auto copy([&ret, &buf, &max]
|
|
(const string_view &src)
|
|
{
|
|
ret = std::min(src.size(), max);
|
|
memcpy(buf, src.data(), ret);
|
|
});
|
|
|
|
operator()(key, copy, gopts);
|
|
return ret;
|
|
}
|
|
|
|
size_t
|
|
db::handle::get(const string_view &key,
|
|
char *const &buf,
|
|
const size_t &max,
|
|
const gopts &gopts)
|
|
{
|
|
size_t ret(0);
|
|
const auto copy([&ret, &buf, &max]
|
|
(const string_view &src)
|
|
{
|
|
ret = strlcpy(buf, src.data(), std::min(src.size(), max));
|
|
});
|
|
|
|
operator()(key, copy, gopts);
|
|
return ret;
|
|
}
|
|
|
|
namespace ircd {
|
|
namespace db {
|
|
|
|
static
|
|
void append(rocksdb::WriteBatch &batch,
|
|
rocksdb::ColumnFamilyHandle *const &h,
|
|
const delta &delta)
|
|
{
|
|
const rocksdb::Slice k
|
|
{
|
|
std::get<1>(delta).data(), std::get<1>(delta).size()
|
|
};
|
|
|
|
const rocksdb::Slice v
|
|
{
|
|
std::get<2>(delta).data(), std::get<2>(delta).size()
|
|
};
|
|
|
|
switch(std::get<0>(delta))
|
|
{
|
|
case op::GET: assert(0); break;
|
|
case op::SET: batch.Put(h, k, v); break;
|
|
case op::MERGE: batch.Merge(h, k, v); break;
|
|
case op::DELETE: batch.Delete(h, k); break;
|
|
case op::DELETE_RANGE: batch.DeleteRange(h, k, v); break;
|
|
case op::SINGLE_DELETE: batch.SingleDelete(h, k); break;
|
|
}
|
|
}
|
|
|
|
} // namespace db
|
|
} // namespace ircd
|
|
|
|
void
|
|
db::handle::operator()(const op &op,
|
|
const string_view &key,
|
|
const string_view &val,
|
|
const sopts &sopts)
|
|
{
|
|
operator()(delta{op, key, val}, sopts);
|
|
}
|
|
|
|
void
|
|
db::handle::operator()(const std::initializer_list<delta> &deltas,
|
|
const sopts &sopts)
|
|
{
|
|
rocksdb::WriteBatch batch;
|
|
for(const auto &delta : deltas)
|
|
append(batch, h, delta);
|
|
|
|
auto opts(make_opts(sopts));
|
|
throw_on_error(meta->d->Write(opts, &batch));
|
|
}
|
|
|
|
void
|
|
db::handle::operator()(const delta &delta,
|
|
const sopts &sopts)
|
|
{
|
|
rocksdb::WriteBatch batch;
|
|
append(batch, h, delta);
|
|
|
|
auto opts(make_opts(sopts));
|
|
throw_on_error(meta->d->Write(opts, &batch));
|
|
}
|
|
|
|
void
|
|
db::handle::operator()(const string_view &key,
|
|
const gopts &gopts,
|
|
const closure &func)
|
|
{
|
|
return operator()(key, func, gopts);
|
|
}
|
|
|
|
void
|
|
db::handle::operator()(const string_view &key,
|
|
const closure &func,
|
|
const gopts &gopts)
|
|
{
|
|
using rocksdb::Slice;
|
|
using rocksdb::Iterator;
|
|
|
|
auto opts(make_opts(gopts));
|
|
const Slice sk(key.data(), key.size());
|
|
const auto it(seek(*meta->d, h, sk, opts));
|
|
valid_equal_or_throw(*it, sk);
|
|
const auto &v(it->value());
|
|
func(string_view{v.data(), v.size()});
|
|
}
|
|
|
|
bool
|
|
db::handle::has(const string_view &key,
|
|
const gopts &gopts)
|
|
{
|
|
using rocksdb::Slice;
|
|
using rocksdb::Status;
|
|
|
|
const Slice k(key.data(), key.size());
|
|
auto opts(make_opts(gopts));
|
|
|
|
// Perform queries which are stymied from any sysentry
|
|
opts.read_tier = NON_BLOCKING;
|
|
|
|
// Perform a co-RP query to the filtration
|
|
if(!meta->d->KeyMayExist(opts, h, k, nullptr, nullptr))
|
|
return false;
|
|
|
|
// Perform a query to the cache
|
|
auto status(meta->d->Get(opts, h, k, nullptr));
|
|
if(status.IsIncomplete())
|
|
{
|
|
// DB cache miss; next query requires I/O, offload it
|
|
opts.read_tier = BLOCKING;
|
|
ctx::offload([this, &opts, &k, &status]
|
|
{
|
|
status = meta->d->Get(opts, h, k, nullptr);
|
|
});
|
|
}
|
|
|
|
// Finally the result
|
|
switch(status.code())
|
|
{
|
|
case Status::kOk: return true;
|
|
case Status::kNotFound: return false;
|
|
default:
|
|
throw_on_error(status);
|
|
__builtin_unreachable();
|
|
}
|
|
}
|
|
|
|
///////////////////////////////////////////////////////////////////////////////
|
|
//
|
|
// handle::const_iterator
|
|
//
|
|
|
|
namespace ircd {
|
|
namespace db {
|
|
|
|
struct handle::const_iterator::state
|
|
{
|
|
struct handle *handle;
|
|
rocksdb::ReadOptions ropts;
|
|
std::shared_ptr<const rocksdb::Snapshot> snap;
|
|
std::unique_ptr<rocksdb::Iterator> it;
|
|
|
|
state(struct handle *const & = nullptr, const gopts & = {});
|
|
};
|
|
|
|
} // namespace db
|
|
} // namespace ircd
|
|
|
|
db::handle::const_iterator
|
|
db::handle::cend(const gopts &gopts)
|
|
{
|
|
return {};
|
|
}
|
|
|
|
db::handle::const_iterator
|
|
db::handle::cbegin(const gopts &gopts)
|
|
{
|
|
const_iterator ret
|
|
{
|
|
std::make_unique<struct const_iterator::state>(this, gopts)
|
|
};
|
|
|
|
auto &state(*ret.state);
|
|
if(!has_opt(gopts, db::get::READAHEAD))
|
|
state.ropts.readahead_size = DEFAULT_READAHEAD;
|
|
|
|
seek(*state.handle->meta->d, state.handle->h, pos::FRONT, state.ropts, state.it);
|
|
return std::move(ret);
|
|
}
|
|
|
|
db::handle::const_iterator
|
|
db::handle::upper_bound(const string_view &key,
|
|
const gopts &gopts)
|
|
{
|
|
using rocksdb::Slice;
|
|
|
|
auto it(lower_bound(key, gopts));
|
|
const Slice sk(key.data(), key.size());
|
|
if(it && it.state->it->key().compare(sk) == 0)
|
|
++it;
|
|
|
|
return std::move(it);
|
|
}
|
|
|
|
db::handle::const_iterator
|
|
db::handle::lower_bound(const string_view &key,
|
|
const gopts &gopts)
|
|
{
|
|
using rocksdb::Slice;
|
|
|
|
const_iterator ret
|
|
{
|
|
std::make_unique<struct const_iterator::state>(this, gopts)
|
|
};
|
|
|
|
auto &state(*ret.state);
|
|
if(!has_opt(gopts, db::get::READAHEAD))
|
|
state.ropts.readahead_size = DEFAULT_READAHEAD;
|
|
|
|
const Slice sk(key.data(), key.size());
|
|
seek(*state.handle->meta->d, state.handle->h, sk, state.ropts, state.it);
|
|
return std::move(ret);
|
|
}
|
|
|
|
db::handle::const_iterator::state::state(db::handle *const &handle,
|
|
const gopts &gopts)
|
|
:handle{handle}
|
|
,ropts{make_opts(gopts, true)}
|
|
,snap
|
|
{
|
|
[this, &handle, &gopts]() -> const rocksdb::Snapshot *
|
|
{
|
|
if(handle && !has_opt(gopts, db::get::NO_SNAPSHOT))
|
|
ropts.snapshot = handle->meta->d->GetSnapshot();
|
|
|
|
return ropts.snapshot;
|
|
}()
|
|
,[this](const auto *const &snap)
|
|
{
|
|
if(this->handle && this->handle->meta && this->handle->meta->d)
|
|
this->handle->meta->d->ReleaseSnapshot(snap);
|
|
}
|
|
}
|
|
{
|
|
}
|
|
|
|
db::handle::const_iterator::const_iterator(std::unique_ptr<struct state> &&state)
|
|
:state{std::move(state)}
|
|
{
|
|
}
|
|
|
|
db::handle::const_iterator::const_iterator(const_iterator &&o)
|
|
noexcept
|
|
:state{std::move(o.state)}
|
|
{
|
|
}
|
|
|
|
db::handle::const_iterator::~const_iterator()
|
|
noexcept
|
|
{
|
|
}
|
|
|
|
db::handle::const_iterator &
|
|
db::handle::const_iterator::operator--()
|
|
{
|
|
seek(*state->handle->meta->d, state->handle->h, pos::PREV, state->ropts, state->it);
|
|
return *this;
|
|
}
|
|
|
|
db::handle::const_iterator &
|
|
db::handle::const_iterator::operator++()
|
|
{
|
|
seek(*state->handle->meta->d, state->handle->h, pos::NEXT, state->ropts, state->it);
|
|
return *this;
|
|
}
|
|
|
|
const db::handle::const_iterator::value_type &
|
|
db::handle::const_iterator::operator*()
|
|
const
|
|
{
|
|
const auto &k(state->it->key());
|
|
const auto &v(state->it->value());
|
|
|
|
val.first = { k.data(), k.size() };
|
|
val.second = { v.data(), v.size() };
|
|
|
|
return val;
|
|
}
|
|
|
|
const db::handle::const_iterator::value_type *
|
|
db::handle::const_iterator::operator->()
|
|
const
|
|
{
|
|
return &operator*();
|
|
}
|
|
|
|
bool
|
|
db::handle::const_iterator::operator>=(const const_iterator &o)
|
|
const
|
|
{
|
|
return (*this > o) || (*this == o);
|
|
}
|
|
|
|
bool
|
|
db::handle::const_iterator::operator<=(const const_iterator &o)
|
|
const
|
|
{
|
|
return (*this < o) || (*this == o);
|
|
}
|
|
|
|
bool
|
|
db::handle::const_iterator::operator!=(const const_iterator &o)
|
|
const
|
|
{
|
|
return !(*this == o);
|
|
}
|
|
|
|
bool
|
|
db::handle::const_iterator::operator==(const const_iterator &o)
|
|
const
|
|
{
|
|
if(*this && o)
|
|
{
|
|
const auto &a(state->it->key());
|
|
const auto &b(o.state->it->key());
|
|
return a.compare(b) == 0;
|
|
}
|
|
|
|
if(!*this && !o)
|
|
return true;
|
|
|
|
return false;
|
|
}
|
|
|
|
bool
|
|
db::handle::const_iterator::operator>(const const_iterator &o)
|
|
const
|
|
{
|
|
if(*this && o)
|
|
{
|
|
const auto &a(state->it->key());
|
|
const auto &b(o.state->it->key());
|
|
return a.compare(b) == 1;
|
|
}
|
|
|
|
if(!*this && o)
|
|
return true;
|
|
|
|
if(!*this && !o)
|
|
return false;
|
|
|
|
assert(!*this && o);
|
|
return false;
|
|
}
|
|
|
|
bool
|
|
db::handle::const_iterator::operator<(const const_iterator &o)
|
|
const
|
|
{
|
|
if(*this && o)
|
|
{
|
|
const auto &a(state->it->key());
|
|
const auto &b(o.state->it->key());
|
|
return a.compare(b) == -1;
|
|
}
|
|
|
|
if(!*this && o)
|
|
return false;
|
|
|
|
if(!*this && !o)
|
|
return false;
|
|
|
|
assert(*this && !o);
|
|
return true;
|
|
}
|
|
|
|
bool
|
|
db::handle::const_iterator::operator!()
|
|
const
|
|
{
|
|
if(!state)
|
|
return true;
|
|
|
|
if(!state->it)
|
|
return true;
|
|
|
|
if(!state->it->Valid())
|
|
return true;
|
|
|
|
return false;
|
|
}
|
|
|
|
db::handle::const_iterator::operator bool()
|
|
const
|
|
{
|
|
return !!*this;
|
|
}
|
|
|
|
///////////////////////////////////////////////////////////////////////////////
|
|
//
|
|
// Misc
|
|
//
|
|
|
|
std::vector<std::string>
|
|
ircd::db::column_names(const std::string &path,
|
|
const opts &opts)
|
|
{
|
|
return column_names(path, make_opts(opts));
|
|
}
|
|
|
|
std::vector<std::string>
|
|
ircd::db::column_names(const std::string &path,
|
|
const rocksdb::DBOptions &opts)
|
|
try
|
|
{
|
|
std::vector<std::string> ret;
|
|
throw_on_error(rocksdb::DB::ListColumnFamilies(opts, path, &ret));
|
|
return ret;
|
|
}
|
|
catch(const io_error &e)
|
|
{
|
|
return // No database found at path. Assume fresh.
|
|
{
|
|
{ rocksdb::kDefaultColumnFamilyName }
|
|
};
|
|
}
|
|
|
|
rocksdb::Options
|
|
db::make_opts(const opts &opts)
|
|
{
|
|
rocksdb::Options ret;
|
|
ret.create_if_missing = true; // They default this to false, but we invert the option
|
|
|
|
for(const auto &o : opts) switch(o.first)
|
|
{
|
|
case opt::NO_CREATE:
|
|
ret.create_if_missing = false;
|
|
continue;
|
|
|
|
case opt::NO_EXISTING:
|
|
ret.error_if_exists = true;
|
|
continue;
|
|
|
|
case opt::NO_CHECKSUM:
|
|
ret.paranoid_checks = false;
|
|
continue;
|
|
|
|
case opt::NO_MADV_DONTNEED:
|
|
ret.allow_os_buffer = false;
|
|
continue;
|
|
|
|
case opt::NO_MADV_RANDOM:
|
|
ret.advise_random_on_open = false;
|
|
continue;
|
|
|
|
case opt::FALLOCATE:
|
|
ret.allow_fallocate = true;
|
|
continue;
|
|
|
|
case opt::NO_FALLOCATE:
|
|
ret.allow_fallocate = false;
|
|
continue;
|
|
|
|
case opt::NO_FDATASYNC:
|
|
ret.disableDataSync = true;
|
|
continue;
|
|
|
|
case opt::FSYNC:
|
|
ret.use_fsync = true;
|
|
continue;
|
|
|
|
case opt::MMAP_READS:
|
|
ret.allow_mmap_reads = true;
|
|
continue;
|
|
|
|
case opt::MMAP_WRITES:
|
|
ret.allow_mmap_writes = true;
|
|
continue;
|
|
|
|
case opt::STATS_THREAD:
|
|
ret.enable_thread_tracking = true;
|
|
continue;
|
|
|
|
case opt::STATS_MALLOC:
|
|
ret.dump_malloc_stats = true;
|
|
continue;
|
|
|
|
case opt::OPEN_FAST:
|
|
ret.skip_stats_update_on_db_open = true;
|
|
continue;
|
|
|
|
case opt::OPEN_BULKLOAD:
|
|
ret.PrepareForBulkLoad();
|
|
continue;
|
|
|
|
case opt::OPEN_SMALL:
|
|
ret.OptimizeForSmallDb();
|
|
continue;
|
|
|
|
default:
|
|
continue;
|
|
}
|
|
|
|
return ret;
|
|
}
|
|
|
|
rocksdb::ReadOptions
|
|
db::make_opts(const gopts &opts,
|
|
const bool &iterator)
|
|
{
|
|
rocksdb::ReadOptions ret;
|
|
|
|
if(iterator)
|
|
ret.fill_cache = false;
|
|
|
|
for(const auto &opt : opts) switch(opt.first)
|
|
{
|
|
case get::PIN:
|
|
ret.pin_data = true;
|
|
continue;
|
|
|
|
case get::CACHE:
|
|
ret.fill_cache = true;
|
|
continue;
|
|
|
|
case get::NO_CACHE:
|
|
ret.fill_cache = false;
|
|
continue;
|
|
|
|
case get::NO_CHECKSUM:
|
|
ret.verify_checksums = false;
|
|
continue;
|
|
|
|
case get::READAHEAD:
|
|
ret.readahead_size = opt.second;
|
|
continue;
|
|
|
|
default:
|
|
continue;
|
|
}
|
|
|
|
return ret;
|
|
}
|
|
|
|
rocksdb::WriteOptions
|
|
db::make_opts(const sopts &opts)
|
|
{
|
|
rocksdb::WriteOptions ret;
|
|
for(const auto &opt : opts) switch(opt.first)
|
|
{
|
|
case set::FSYNC:
|
|
ret.sync = true;
|
|
continue;
|
|
|
|
case set::NO_JOURNAL:
|
|
ret.disableWAL = true;
|
|
continue;
|
|
|
|
case set::MISSING_COLUMNS:
|
|
ret.ignore_missing_column_families = true;
|
|
continue;
|
|
|
|
default:
|
|
continue;
|
|
}
|
|
|
|
return ret;
|
|
}
|
|
|
|
std::unique_ptr<rocksdb::Iterator>
|
|
db::seek(rocksdb::DB &db,
|
|
rocksdb::ColumnFamilyHandle *const &h,
|
|
const rocksdb::Slice &key,
|
|
rocksdb::ReadOptions &opts)
|
|
{
|
|
using rocksdb::Iterator;
|
|
|
|
// Perform a query which won't be allowed to do kernel IO
|
|
opts.read_tier = NON_BLOCKING;
|
|
|
|
std::unique_ptr<Iterator> it(db.NewIterator(opts, h));
|
|
seek(*it, key);
|
|
|
|
if(it->status().IsIncomplete())
|
|
{
|
|
// DB cache miss: reset the iterator to blocking mode and offload it
|
|
opts.read_tier = BLOCKING;
|
|
it.reset(db.NewIterator(opts, h));
|
|
ctx::offload([&] { seek(*it, key); });
|
|
}
|
|
// else DB cache hit; no context switch; no thread switch; no kernel I/O; gg
|
|
|
|
return std::move(it);
|
|
}
|
|
|
|
template<class pos>
|
|
void
|
|
db::seek(rocksdb::DB &db,
|
|
rocksdb::ColumnFamilyHandle *const &h,
|
|
const pos &p,
|
|
rocksdb::ReadOptions &opts,
|
|
std::unique_ptr<rocksdb::Iterator> &it)
|
|
{
|
|
// Start with a non-blocking query
|
|
if(!it || opts.read_tier == BLOCKING)
|
|
{
|
|
opts.read_tier = NON_BLOCKING;
|
|
it.reset(db.NewIterator(opts, h));
|
|
}
|
|
|
|
seek(*it, p);
|
|
if(it->status().IsIncomplete())
|
|
{
|
|
// DB cache miss: reset the iterator to blocking mode and offload it
|
|
opts.read_tier = BLOCKING;
|
|
it.reset(db.NewIterator(opts, h));
|
|
ctx::offload([&] { seek(*it, p); });
|
|
}
|
|
}
|
|
|
|
void
|
|
db::seek(rocksdb::Iterator &it,
|
|
const rocksdb::Slice &sk)
|
|
{
|
|
it.Seek(sk);
|
|
}
|
|
|
|
void
|
|
db::seek(rocksdb::Iterator &it,
|
|
const pos &p)
|
|
{
|
|
switch(p)
|
|
{
|
|
case pos::NEXT: it.Next(); break;
|
|
case pos::PREV: it.Prev(); break;
|
|
case pos::FRONT: it.SeekToFirst(); break;
|
|
case pos::BACK: it.SeekToLast(); break;
|
|
default:
|
|
case pos::END:
|
|
{
|
|
it.SeekToLast();
|
|
if(it.Valid())
|
|
it.Next();
|
|
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
|
|
void
|
|
db::valid_equal_or_throw(const rocksdb::Iterator &it,
|
|
const rocksdb::Slice &sk)
|
|
{
|
|
valid_or_throw(it);
|
|
if(it.key().compare(sk) != 0)
|
|
throw not_found();
|
|
}
|
|
|
|
void
|
|
db::valid_or_throw(const rocksdb::Iterator &it)
|
|
{
|
|
if(!valid(it))
|
|
{
|
|
throw_on_error(it.status());
|
|
throw not_found();
|
|
//assert(0); // status == ok + !Valid() == ???
|
|
}
|
|
}
|
|
|
|
bool
|
|
db::valid(const rocksdb::Iterator &it)
|
|
{
|
|
return it.Valid();
|
|
}
|
|
|
|
void
|
|
db::throw_on_error(const rocksdb::Status &s)
|
|
{
|
|
using rocksdb::Status;
|
|
|
|
switch(s.code())
|
|
{
|
|
case Status::kOk: return;
|
|
case Status::kNotFound: throw not_found();
|
|
case Status::kCorruption: throw corruption();
|
|
case Status::kNotSupported: throw not_supported();
|
|
case Status::kInvalidArgument: throw invalid_argument();
|
|
case Status::kIOError: throw io_error();
|
|
case Status::kMergeInProgress: throw merge_in_progress();
|
|
case Status::kIncomplete: throw incomplete();
|
|
case Status::kShutdownInProgress: throw shutdown_in_progress();
|
|
case Status::kTimedOut: throw timed_out();
|
|
case Status::kAborted: throw aborted();
|
|
case Status::kBusy: throw busy();
|
|
case Status::kExpired: throw expired();
|
|
case Status::kTryAgain: throw try_again();
|
|
default:
|
|
throw error("Unknown error");
|
|
}
|
|
}
|
|
|
|
std::string
|
|
db::path(const std::string &name)
|
|
{
|
|
const auto prefix(path::get(path::DB));
|
|
return path::build({prefix, name});
|
|
}
|
|
|
|
const std::string &
|
|
ircd::db::reflect(const rocksdb::Tickers &type)
|
|
{
|
|
const auto &names(rocksdb::TickersNameMap);
|
|
const auto it(std::find_if(begin(names), end(names), [&type]
|
|
(const auto &pair)
|
|
{
|
|
return pair.first == type;
|
|
}));
|
|
|
|
static const auto empty{"<ticker>?????"s};
|
|
return it != end(names)? it->second : empty;
|
|
}
|
|
|
|
const std::string &
|
|
ircd::db::reflect(const rocksdb::Histograms &type)
|
|
{
|
|
const auto &names(rocksdb::HistogramsNameMap);
|
|
const auto it(std::find_if(begin(names), end(names), [&type]
|
|
(const auto &pair)
|
|
{
|
|
return pair.first == type;
|
|
}));
|
|
|
|
static const auto empty{"<histogram>?????"s};
|
|
return it != end(names)? it->second : empty;
|
|
}
|