Hawkin
/
arangodb
mirror of https://gitee.com/bigwinds/arangodb
1
0
Fork 0
Code Issues Projects Releases Wiki Activity
arangodb/arangod/RocksDBEngine/RocksDBSettingsManager.cpp

616 lines
20 KiB
C++
Raw Blame History

////////////////////////////////////////////////////////////////////////////////
/// DISCLAIMER
///
/// Copyright 2014-2017 ArangoDB GmbH, Cologne, Germany
/// Copyright 2004-2014 triAGENS GmbH, Cologne, Germany
///
/// Licensed under the Apache License, Version 2.0 (the "License");
/// you may not use this file except in compliance with the License.
/// You may obtain a copy of the License at
///
/// http://www.apache.org/licenses/LICENSE-2.0
///
/// Unless required by applicable law or agreed to in writing, software
/// distributed under the License is distributed on an "AS IS" BASIS,
/// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
/// See the License for the specific language governing permissions and
/// limitations under the License.
///
/// Copyright holder is ArangoDB GmbH, Cologne, Germany
///
/// @author Simon Grätzer
/// @author Daniel Larkin-York
////////////////////////////////////////////////////////////////////////////////
#include "RocksDBSettingsManager.h"
#include "ApplicationFeatures/ApplicationServer.h"
#include "Basics/ReadLocker.h"
#include "Basics/StringUtils.h"
#include "Basics/VelocyPackHelper.h"
#include "Basics/WriteLocker.h"
#include "Logger/Logger.h"
#include "RestServer/DatabaseFeature.h"
#include "RocksDBEngine/RocksDBCollection.h"
#include "RocksDBEngine/RocksDBColumnFamily.h"
#include "RocksDBEngine/RocksDBCommon.h"
#include "RocksDBEngine/RocksDBCuckooIndexEstimator.h"
#include "RocksDBEngine/RocksDBEdgeIndex.h"
#include "RocksDBEngine/RocksDBKey.h"
#include "RocksDBEngine/RocksDBKeyBounds.h"
#include "RocksDBEngine/RocksDBRecoveryHelper.h"
#include "RocksDBEngine/RocksDBVPackIndex.h"
#include "RocksDBEngine/RocksDBValue.h"
#include "StorageEngine/EngineSelectorFeature.h"
#include "VocBase/ticks.h"
#include <rocksdb/utilities/transaction_db.h>
#include <rocksdb/utilities/write_batch_with_index.h>
#include <rocksdb/write_batch.h>
#include <velocypack/Iterator.h>
#include <velocypack/Parser.h>
#include <velocypack/Slice.h>
#include <velocypack/velocypack-aliases.h>
namespace {
arangodb::Result writeCounterValue(
std::unordered_map<uint64_t, rocksdb::SequenceNumber>& syncedSeqNums,
rocksdb::Transaction* rtrx, VPackBuilder& b,
std::pair<uint64_t, arangodb::RocksDBSettingsManager::CMValue> const&
pair) {
using arangodb::Logger;
using arangodb::Result;
using arangodb::RocksDBColumnFamily;
using arangodb::RocksDBKey;
using arangodb::rocksutils::convertStatus;
// Skip values which we did not change
auto const& it = syncedSeqNums.find(pair.first);
if (it != syncedSeqNums.end() && it->second == pair.second._sequenceNum) {
return Result();
}
b.clear();
pair.second.serialize(b);
RocksDBKey key;
key.constructCounterValue(pair.first);
rocksdb::Slice value((char*)b.start(), b.size());
rocksdb::Status s =
rtrx->Put(RocksDBColumnFamily::definitions(), key.string(), value);
if (!s.ok()) {
rtrx->Rollback();
auto rStatus = convertStatus(s);
LOG_TOPIC(WARN, Logger::ENGINES)
<< "writing counters failed: " << rStatus.errorMessage();
return rStatus;
}
return Result();
}
} // namespace
namespace {
arangodb::Result writeSettings(rocksdb::Transaction* rtrx, VPackBuilder& b,
uint64_t seqNumber) {
using arangodb::EngineSelectorFeature;
using arangodb::Logger;
using arangodb::Result;
using arangodb::RocksDBColumnFamily;
using arangodb::RocksDBKey;
using arangodb::RocksDBSettingsType;
using arangodb::StorageEngine;
// now write global settings
StorageEngine* engine = EngineSelectorFeature::ENGINE;
b.clear();
b.openObject();
b.add("tick", VPackValue(std::to_string(TRI_CurrentTickServer())));
b.add("hlc", VPackValue(std::to_string(TRI_HybridLogicalClock())));
b.add("releasedTick", VPackValue(std::to_string(engine->releasedTick())));
b.add("lastSync", VPackValue(std::to_string(seqNumber)));
b.close();
VPackSlice slice = b.slice();
LOG_TOPIC(TRACE, Logger::ENGINES) << "writing settings: " << slice.toJson();
RocksDBKey key;
key.constructSettingsValue(RocksDBSettingsType::ServerTick);
rocksdb::Slice value(slice.startAs<char>(), slice.byteSize());
rocksdb::Status s =
rtrx->Put(RocksDBColumnFamily::definitions(), key.string(), value);
if (!s.ok()) {
auto rStatus = arangodb::rocksutils::convertStatus(s);
LOG_TOPIC(WARN, Logger::ENGINES)
<< "writing settings failed: " << rStatus.errorMessage();
rtrx->Rollback();
return rStatus;
}
return Result();
}
} // namespace
namespace {
std::pair<arangodb::Result, rocksdb::SequenceNumber>
writeIndexEstimatorsAndKeyGenerator(
rocksdb::Transaction* rtrx,
std::pair<uint64_t, arangodb::RocksDBSettingsManager::CMValue> const& pair,
rocksdb::SequenceNumber baseSeq) {
using arangodb::DatabaseFeature;
using arangodb::Logger;
using arangodb::Result;
using arangodb::RocksDBCollection;
using arangodb::application_features::ApplicationServer;
using arangodb::rocksutils::mapObjectToCollection;
auto returnSeq = baseSeq;
auto dbColPair = mapObjectToCollection(pair.first);
if (dbColPair.second == 0 && dbColPair.first == 0) {
// collection with this objectID not known.Skip.
return std::make_pair(Result(), returnSeq);
}
auto dbfeature = ApplicationServer::getFeature<DatabaseFeature>("Database");
TRI_ASSERT(dbfeature != nullptr);
auto vocbase = dbfeature->useDatabase(dbColPair.first);
if (vocbase == nullptr) {
// Bad state, we have references to a database that is not known
// anymore.
// However let's just skip in production. Not allowed to crash.
// If we cannot find this infos during recovery we can either recompute
// or start fresh.
return std::make_pair(Result(), returnSeq);
}
TRI_DEFER(vocbase->release());
auto collection = vocbase->lookupCollection(dbColPair.second);
if (collection == nullptr) {
// Bad state, we have references to a collection that is not known
// anymore.
// However let's just skip in production. Not allowed to crash.
// If we cannot find this infos during recovery we can either recompute
// or start fresh.
return std::make_pair(Result(), returnSeq);
}
auto rocksCollection =
static_cast<RocksDBCollection*>(collection->getPhysical());
TRI_ASSERT(rocksCollection != nullptr);
auto serializeResult =
rocksCollection->serializeIndexEstimates(rtrx, baseSeq);
if (!serializeResult.first.ok()) {
LOG_TOPIC(WARN, Logger::ENGINES) << "writing index estimates failed: "
<< serializeResult.first.errorMessage();
return std::make_pair(serializeResult.first, returnSeq);
}
returnSeq = std::min(returnSeq, serializeResult.second);
Result res = rocksCollection->serializeKeyGenerator(rtrx);
if (!res.ok()) {
LOG_TOPIC(WARN, Logger::ENGINES)
<< "writing key generators failed: " << res.errorMessage();
return std::make_pair(res, returnSeq);
}
return std::make_pair(Result(), returnSeq);
}
} // namespace
namespace arangodb {
RocksDBSettingsManager::CMValue::CMValue(VPackSlice const& slice)
: _sequenceNum(0), _count(0), _revisionId(0) {
if (!slice.isArray()) {
// got a somewhat invalid slice. probably old data from before the key
// structure changes
return;
}
velocypack::ArrayIterator array(slice);
if (array.valid()) {
this->_sequenceNum = (*array).getUInt();
this->_count = (*(++array)).getUInt();
this->_revisionId = (*(++array)).getUInt();
}
}
void RocksDBSettingsManager::CMValue::serialize(VPackBuilder& b) const {
b.openArray();
b.add(VPackValue(this->_sequenceNum));
b.add(VPackValue(this->_count));
b.add(VPackValue(this->_revisionId));
b.close();
}
/// Constructor needs to be called synchrunously,
/// will load counts from the db and scan the WAL
RocksDBSettingsManager::RocksDBSettingsManager(rocksdb::TransactionDB* db)
: _lastSync(0), _syncing(false), _db(db), _initialReleasedTick(0) {}
/// retrieve initial values from the database
void RocksDBSettingsManager::retrieveInitialValues() {
loadSettings();
loadIndexEstimates();
loadCounterValues();
loadKeyGenerators();
EngineSelectorFeature::ENGINE->releaseTick(_initialReleasedTick);
}
RocksDBSettingsManager::CounterAdjustment RocksDBSettingsManager::loadCounter(
uint64_t objectId) const {
TRI_ASSERT(objectId != 0); // TODO fix this
READ_LOCKER(guard, _rwLock);
auto const& it = _counters.find(objectId);
if (it != _counters.end()) {
return CounterAdjustment(it->second._sequenceNum, it->second._count, 0,
it->second._revisionId);
}
return CounterAdjustment(); // do not create
}
/// collections / views / indexes can call this method to update
/// their total counts. Thread-Safe needs the snapshot so we know
/// the sequence number used
void RocksDBSettingsManager::updateCounter(uint64_t objectId,
CounterAdjustment const& update) {
bool needsSync = false;
{
WRITE_LOCKER(guard, _rwLock);
auto it = _counters.find(objectId);
if (it != _counters.end()) {
it->second._count += update.added();
it->second._count -= update.removed();
// just use the latest trx info
if (update.sequenceNumber() > it->second._sequenceNum) {
it->second._sequenceNum = update.sequenceNumber();
it->second._revisionId = update.revisionId();
}
} else {
// insert new counter
_counters.emplace(std::make_pair(
objectId,
CMValue(update.sequenceNumber(), update.added() - update.removed(),
update.revisionId())));
needsSync = true; // only count values from WAL if they are in the DB
}
}
if (needsSync) {
sync(true);
}
}
arangodb::Result RocksDBSettingsManager::setAbsoluteCounter(uint64_t objectId,
uint64_t value) {
arangodb::Result res;
WRITE_LOCKER(guard, _rwLock);
auto it = _counters.find(objectId);
if (it != _counters.end()) {
rocksdb::TransactionDB* db = rocksutils::globalRocksDB();
it->second._sequenceNum = db->GetLatestSequenceNumber();
it->second._count = value;
} else {
// nothing to do as the counter has never been written it can not be set to
// a value that would require correction. but we use the return value to
// signal that no sync is rquired
res.reset(TRI_ERROR_INTERNAL, "counter value not found - no sync required");
}
return res;
}
void RocksDBSettingsManager::removeCounter(uint64_t objectId) {
WRITE_LOCKER(guard, _rwLock);
auto const& it = _counters.find(objectId);
if (it != _counters.end()) {
RocksDBKey key;
key.constructCounterValue(it->first);
rocksdb::WriteOptions options;
rocksdb::Status s =
_db->Delete(options, RocksDBColumnFamily::definitions(), key.string());
if (!s.ok()) {
LOG_TOPIC(ERR, Logger::ENGINES) << "deleting counter failed";
}
_counters.erase(it);
}
}
std::unordered_map<uint64_t, rocksdb::SequenceNumber>
RocksDBSettingsManager::counterSeqs() {
std::unordered_map<uint64_t, rocksdb::SequenceNumber> seqs;
{ // block all updates while we copy
READ_LOCKER(guard, _rwLock);
for (auto it : _counters) {
seqs.emplace(it.first, it.second._sequenceNum);
}
}
return seqs;
}
bool RocksDBSettingsManager::lockForSync(bool force) {
if (force) {
while (true) {
bool expected = false;
bool res = _syncing.compare_exchange_strong(
expected, true, std::memory_order_acquire, std::memory_order_relaxed);
if (res) {
break;
}
std::this_thread::sleep_for(std::chrono::milliseconds(10));
}
} else {
bool expected = false;
if (!_syncing.compare_exchange_strong(expected, true,
std::memory_order_acquire,
std::memory_order_relaxed)) {
return false;
}
}
return true;
}
/// Thread-Safe force sync
Result RocksDBSettingsManager::sync(bool force) {
TRI_IF_FAILURE("RocksDBSettingsManagerSync") { return Result(); }
if (!lockForSync(force)) {
return Result();
}
TRI_DEFER(_syncing = false);
std::unordered_map<uint64_t, CMValue> copy;
{ // block all updates
WRITE_LOCKER(guard, _rwLock);
copy = _counters;
}
rocksdb::WriteOptions writeOptions;
// fetch the seq number prior to any writes; this guarantees that we save
// any subsequent updates in the WAL to replay if we crash in the middle
auto seqNumber = _db->GetLatestSequenceNumber();
std::unique_ptr<rocksdb::Transaction> rtrx(
_db->BeginTransaction(writeOptions));
VPackBuilder b;
for (std::pair<uint64_t, CMValue> const& pair : copy) {
Result res = writeCounterValue(_syncedSeqNums, rtrx.get(), b, pair);
if (res.fail()) {
return res;
}
auto writeResult =
writeIndexEstimatorsAndKeyGenerator(rtrx.get(), pair, seqNumber);
if (writeResult.first.fail()) {
return writeResult.first;
}
seqNumber = std::min(seqNumber, writeResult.second);
}
Result res = writeSettings(rtrx.get(), b, seqNumber);
if (res.fail()) {
return res;
}
// we have to commit all counters in one batch
auto s = rtrx->Commit();
if (s.ok()) {
{
WRITE_LOCKER(guard, _rwLock);
_lastSync = seqNumber;
}
for (std::pair<uint64_t, CMValue> const& pair : copy) {
_syncedSeqNums[pair.first] = pair.second._sequenceNum;
}
}
return rocksutils::convertStatus(s);
}
void RocksDBSettingsManager::loadSettings() {
WRITE_LOCKER(guard, _rwLock);
RocksDBKey key;
key.constructSettingsValue(RocksDBSettingsType::ServerTick);
rocksdb::PinnableSlice result;
rocksdb::Status status =
_db->Get(rocksdb::ReadOptions(), RocksDBColumnFamily::definitions(),
key.string(), &result);
if (status.ok()) {
// key may not be there, so don't fail when not found
VPackSlice slice = VPackSlice(result.data());
TRI_ASSERT(slice.isObject());
LOG_TOPIC(TRACE, Logger::ENGINES)
<< "read initial settings: " << slice.toJson();
if (!result.empty()) {
try {
if (slice.hasKey("tick")) {
uint64_t lastTick =
basics::VelocyPackHelper::stringUInt64(slice.get("tick"));
LOG_TOPIC(TRACE, Logger::ENGINES) << "using last tick: " << lastTick;
TRI_UpdateTickServer(lastTick);
}
if (slice.hasKey("hlc")) {
uint64_t lastHlc =
basics::VelocyPackHelper::stringUInt64(slice.get("hlc"));
LOG_TOPIC(TRACE, Logger::ENGINES) << "using last hlc: " << lastHlc;
TRI_HybridLogicalClock(lastHlc);
}
if (slice.hasKey("releasedTick")) {
_initialReleasedTick =
basics::VelocyPackHelper::stringUInt64(slice.get("releasedTick"));
LOG_TOPIC(TRACE, Logger::ENGINES)
<< "using released tick: " << _initialReleasedTick;
EngineSelectorFeature::ENGINE->releaseTick(_initialReleasedTick);
}
if (slice.hasKey("lastSync")) {
_lastSync =
basics::VelocyPackHelper::stringUInt64(slice.get("lastSync"));
LOG_TOPIC(TRACE, Logger::ENGINES)
<< "last background settings sync: " << _lastSync;
}
} catch (...) {
LOG_TOPIC(WARN, Logger::ENGINES)
<< "unable to read initial settings: invalid data";
}
}
}
}
void RocksDBSettingsManager::loadIndexEstimates() {
WRITE_LOCKER(guard, _rwLock);
RocksDBKeyBounds bounds = RocksDBKeyBounds::IndexEstimateValues();
auto cf = RocksDBColumnFamily::definitions();
rocksdb::Comparator const* cmp = cf->GetComparator();
rocksdb::ReadOptions readOptions;
std::unique_ptr<rocksdb::Iterator> iter(_db->NewIterator(readOptions, cf));
iter->Seek(bounds.start());
for (; iter->Valid() && cmp->Compare(iter->key(), bounds.end()) < 0;
iter->Next()) {
uint64_t objectId = RocksDBKey::definitionsObjectId(iter->key());
uint64_t lastSeqNumber =
rocksutils::uint64FromPersistent(iter->value().data());
StringRef estimateSerialisation(iter->value().data() + sizeof(uint64_t),
iter->value().size() - sizeof(uint64_t));
// If this hits we have two estimates for the same index
TRI_ASSERT(_estimators.find(objectId) == _estimators.end());
try {
if (RocksDBCuckooIndexEstimator<uint64_t>::isFormatSupported(
estimateSerialisation)) {
auto it = _estimators.emplace(
objectId, std::make_unique<RocksDBCuckooIndexEstimator<uint64_t>>(
lastSeqNumber, estimateSerialisation));
if (it.second) {
auto estimator = it.first->second.get();
LOG_TOPIC(TRACE, Logger::ENGINES)
<< "found index estimator for objectId '" << objectId
<< "' last synced at " << lastSeqNumber << " with estimate "
<< estimator->computeEstimate();
}
}
} catch (...) {
// Nothing to do, if the estimator fails to create we let it be recreated.
// Just validate that no corrupted memory was produced.
TRI_ASSERT(_estimators.find(objectId) == _estimators.end());
}
}
}
void RocksDBSettingsManager::loadKeyGenerators() {
WRITE_LOCKER(guard, _rwLock);
RocksDBKeyBounds bounds = RocksDBKeyBounds::KeyGenerators();
auto cf = RocksDBColumnFamily::definitions();
rocksdb::Comparator const* cmp = cf->GetComparator();
rocksdb::ReadOptions readOptions;
std::unique_ptr<rocksdb::Iterator> iter(_db->NewIterator(readOptions, cf));
iter->Seek(bounds.start());
for (; iter->Valid() && cmp->Compare(iter->key(), bounds.end()) < 0;
iter->Next()) {
uint64_t objectId = RocksDBKey::definitionsObjectId(iter->key());
auto properties = RocksDBValue::data(iter->value());
VPackSlice s = properties.get(StaticStrings::LastValue);
if (!s.isNone()) {
uint64_t lastValue = properties.get(StaticStrings::LastValue).getUInt();
// If this hits we have two generators for the same collection
TRI_ASSERT(_generators.find(objectId) == _generators.end());
try {
_generators.emplace(objectId, lastValue);
} catch (...) {
// Nothing to do, just validate that no corrupted memory was produced.
TRI_ASSERT(_generators.find(objectId) == _generators.end());
}
}
}
}
std::unique_ptr<RocksDBCuckooIndexEstimator<uint64_t>>
RocksDBSettingsManager::stealIndexEstimator(uint64_t objectId) {
std::unique_ptr<RocksDBCuckooIndexEstimator<uint64_t>> res;
WRITE_LOCKER(guard, _rwLock);
auto it = _estimators.find(objectId);
if (it != _estimators.end()) {
// We swap out the stored estimate in order to move it to the caller
res.swap(it->second);
// Drop the now empty estimator
_estimators.erase(objectId);
}
return res;
}
uint64_t RocksDBSettingsManager::stealKeyGenerator(uint64_t objectId) {
WRITE_LOCKER(guard, _rwLock);
uint64_t res = 0;
auto it = _generators.find(objectId);
if (it != _generators.end()) {
// We swap out the stored estimate in order to move it to the caller
res = it->second;
// Drop the now empty estimator
_generators.erase(objectId);
}
return res;
}
void RocksDBSettingsManager::clearIndexEstimators() {
// We call this to remove all index estimators that have been stored but are
// no longer read
// by recovery.
// TODO REMOVE RocksDB Keys of all not stolen values?
WRITE_LOCKER(guard, _rwLock);
_estimators.clear();
}
void RocksDBSettingsManager::clearKeyGenerators() {
WRITE_LOCKER(guard, _rwLock);
_generators.clear();
}
/// Parse counter values from rocksdb
void RocksDBSettingsManager::loadCounterValues() {
WRITE_LOCKER(guard, _rwLock);
RocksDBKeyBounds bounds = RocksDBKeyBounds::CounterValues();
auto cf = RocksDBColumnFamily::definitions();
rocksdb::Comparator const* cmp = cf->GetComparator();
rocksdb::ReadOptions readOptions;
std::unique_ptr<rocksdb::Iterator> iter(_db->NewIterator(readOptions, cf));
iter->Seek(bounds.start());
while (iter->Valid() && cmp->Compare(iter->key(), bounds.end()) < 0) {
uint64_t objectId = RocksDBKey::definitionsObjectId(iter->key());
auto const& it =
_counters.emplace(objectId, CMValue(VPackSlice(iter->value().data())));
_syncedSeqNums[objectId] = it.first->second._sequenceNum;
LOG_TOPIC(TRACE, Logger::ENGINES)
<< "found count marker for objectId '" << objectId
<< "' last synced at " << it.first->first << " with count "
<< it.first->second._count;
iter->Next();
}
}
/// earliest safe sequence number to throw away from wal
rocksdb::SequenceNumber RocksDBSettingsManager::earliestSeqNeeded() const {
READ_LOCKER(guard, _rwLock);
return _lastSync;
}
} // namespace arangodb
Reference in New Issue View Git Blame Copy Permalink