1
0
Fork 0
arangodb/arangod/RocksDBEngine/RocksDBPrimaryIndex.cpp

645 lines
21 KiB
C++

////////////////////////////////////////////////////////////////////////////////
/// 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 Jan Steemann
////////////////////////////////////////////////////////////////////////////////
#include "RocksDBPrimaryIndex.h"
#include "Aql/AstNode.h"
#include "Basics/Exceptions.h"
#include "Basics/StaticStrings.h"
#include "Basics/VelocyPackHelper.h"
#include "Cache/CachedValue.h"
#include "Cache/TransactionalCache.h"
#include "Cluster/ServerState.h"
#include "Indexes/SimpleAttributeEqualityMatcher.h"
#include "Logger/Logger.h"
#include "RocksDBEngine/RocksDBCollection.h"
#include "RocksDBEngine/RocksDBCommon.h"
#include "RocksDBEngine/RocksDBComparator.h"
#include "RocksDBEngine/RocksDBEngine.h"
#include "RocksDBEngine/RocksDBKey.h"
#include "RocksDBEngine/RocksDBKeyBounds.h"
#include "RocksDBEngine/RocksDBMethods.h"
#include "RocksDBEngine/RocksDBToken.h"
#include "RocksDBEngine/RocksDBTransactionState.h"
#include "RocksDBEngine/RocksDBTypes.h"
#include "RocksDBEngine/RocksDBValue.h"
#include "StorageEngine/EngineSelectorFeature.h"
#include "Transaction/Context.h"
#include "Transaction/Helpers.h"
#include "Transaction/Methods.h"
#include "VocBase/LogicalCollection.h"
#include <rocksdb/iterator.h>
#include <rocksdb/utilities/transaction.h>
#include <velocypack/Builder.h>
#include <velocypack/Collection.h>
#include <velocypack/Slice.h>
#include <velocypack/velocypack-aliases.h>
using namespace arangodb;
// ================ Primary Index Iterator ================
/// @brief hard-coded vector of the index attributes
/// note that the attribute names must be hard-coded here to avoid an init-order
/// fiasco with StaticStrings::FromString etc.
static std::vector<std::vector<arangodb::basics::AttributeName>> const
IndexAttributes{{arangodb::basics::AttributeName("_id", false)},
{arangodb::basics::AttributeName("_key", false)}};
RocksDBPrimaryIndexIterator::RocksDBPrimaryIndexIterator(
LogicalCollection* collection, transaction::Methods* trx,
ManagedDocumentResult* mmdr, RocksDBPrimaryIndex* index,
std::unique_ptr<VPackBuilder>& keys)
: IndexIterator(collection, trx, mmdr, index),
_index(index),
_keys(keys.get()),
_iterator(_keys->slice()) {
keys.release(); // now we have ownership for _keys
TRI_ASSERT(_keys->slice().isArray());
}
RocksDBPrimaryIndexIterator::~RocksDBPrimaryIndexIterator() {
if (_keys != nullptr) {
// return the VPackBuilder to the transaction context
_trx->transactionContextPtr()->returnBuilder(_keys.release());
}
}
bool RocksDBPrimaryIndexIterator::next(TokenCallback const& cb, size_t limit) {
if (limit == 0 || !_iterator.valid()) {
// No limit no data, or we are actually done. The last call should have
// returned false
TRI_ASSERT(limit > 0); // Someone called with limit == 0. Api broken
return false;
}
ManagedDocumentResult result;
while (limit > 0) {
// TODO: prevent copying of the value into result, as we don't need it here!
RocksDBToken token = _index->lookupKey(_trx, *_iterator, result);
cb(token);
--limit;
_iterator.next();
if (!_iterator.valid()) {
return false;
}
}
return true;
}
void RocksDBPrimaryIndexIterator::reset() { _iterator.reset(); }
// ================ All Iterator ==================
RocksDBAllIndexIterator::RocksDBAllIndexIterator(
LogicalCollection* collection, transaction::Methods* trx,
ManagedDocumentResult* mmdr, RocksDBPrimaryIndex const* index, bool reverse)
: IndexIterator(collection, trx, mmdr, index),
_reverse(reverse),
_iterator(rocksutils::toRocksMethods(trx)->NewIterator()),
_bounds(RocksDBKeyBounds::PrimaryIndex(index->objectId()))
{
// acquire rocksdb transaction
RocksDBTransactionState* state = rocksutils::toRocksTransactionState(trx);
TRI_ASSERT(state != nullptr);
RocksDBMethods* mthds = rocksutils::toRocksMethods(trx);
auto const& options = mthds->readOptions();
TRI_ASSERT(options.snapshot != nullptr);
TRI_ASSERT(options.prefix_same_as_start);
if (reverse) {
_iterator->SeekForPrev(_bounds.end());
} else {
_iterator->Seek(_bounds.start());
}
}
bool RocksDBAllIndexIterator::next(TokenCallback const& cb, size_t limit) {
TRI_ASSERT(_trx->state()->isRunning());
if (limit == 0 || !_iterator->Valid()) {
// No limit no data, or we are actually done. The last call should have
// returned false
TRI_ASSERT(limit > 0); // Someone called with limit == 0. Api broken
return false;
}
while (limit > 0) {
RocksDBToken token(RocksDBValue::revisionId(_iterator->value()));
cb(token);
--limit;
if (_reverse) {
_iterator->Prev();
} else {
_iterator->Next();
}
if (!_iterator->Valid()) {
return false;
}
}
return true;
}
/// special method to expose the document key for incremental replication
bool RocksDBAllIndexIterator::nextWithKey(TokenKeyCallback const& cb,
size_t limit) {
TRI_ASSERT(_trx->state()->isRunning());
if (limit == 0 || !_iterator->Valid()) {
// No limit no data, or we are actually done. The last call should have
// returned false
TRI_ASSERT(limit > 0); // Someone called with limit == 0. Api broken
return false;
}
while (limit > 0) {
RocksDBToken token(RocksDBValue::revisionId(_iterator->value()));
StringRef key = RocksDBKey::primaryKey(_iterator->key());
cb(token, key);
--limit;
if (_reverse) {
_iterator->Prev();
} else {
_iterator->Next();
}
if (!_iterator->Valid()) {
return false;
}
}
return true;
}
void RocksDBAllIndexIterator::seek(StringRef const& key) {
TRI_ASSERT(_trx->state()->isRunning());
// don't want to get the index pointer just for this
uint64_t objectId = _bounds.objectId();
RocksDBKey val = RocksDBKey::PrimaryIndexValue(objectId, key);
if (_reverse) {
_iterator->SeekForPrev(val.string());
} else {
_iterator->Seek(val.string());
}
}
void RocksDBAllIndexIterator::reset() {
TRI_ASSERT(_trx->state()->isRunning());
if (_reverse) {
_iterator->SeekForPrev(_bounds.end());
} else {
_iterator->Seek(_bounds.start());
}
}
// ================ Any Iterator ================
RocksDBAnyIndexIterator::RocksDBAnyIndexIterator(
LogicalCollection* collection, transaction::Methods* trx,
ManagedDocumentResult* mmdr, RocksDBPrimaryIndex const* index)
: IndexIterator(collection, trx, mmdr, index),
_cmp(index->comparator()),
_iterator(rocksutils::toRocksMethods(trx)->NewIterator()),
_bounds(RocksDBKeyBounds::PrimaryIndex(index->objectId())),
_total(0),
_returned(0) {
// acquire rocksdb transaction
RocksDBTransactionState* state = rocksutils::toRocksTransactionState(trx);
TRI_ASSERT(state != nullptr);
auto const& options = rocksutils::toRocksMethods(trx)->readOptions();
TRI_ASSERT(options.snapshot != nullptr);
_total = collection->numberDocuments(trx);
uint64_t off = RandomGenerator::interval(_total - 1);
// uint64_t goal = off;
if (_total > 0) {
if (off <= _total / 2) {
_iterator->Seek(_bounds.start());
while (_iterator->Valid() && off-- > 0) {
_iterator->Next();
}
} else {
off = _total - (off + 1);
_iterator->SeekForPrev(_bounds.end());
while (_iterator->Valid() && off-- > 0) {
_iterator->Prev();
}
}
if (!_iterator->Valid()) {
_iterator->Seek(_bounds.start());
}
}
}
bool RocksDBAnyIndexIterator::next(TokenCallback const& cb, size_t limit) {
TRI_ASSERT(_trx->state()->isRunning());
if (limit == 0 || !_iterator->Valid()) {
// No limit no data, or we are actually done. The last call should have
// returned false
TRI_ASSERT(limit > 0); // Someone called with limit == 0. Api broken
return false;
}
while (limit > 0) {
RocksDBToken token(RocksDBValue::revisionId(_iterator->value()));
cb(token);
--limit;
_returned++;
_iterator->Next();
if (!_iterator->Valid()) {
if (_returned < _total) {
_iterator->Seek(_bounds.start());
continue;
}
return false;
}
}
return true;
}
void RocksDBAnyIndexIterator::reset() { _iterator->Seek(_bounds.start()); }
// ================ PrimaryIndex ================
RocksDBPrimaryIndex::RocksDBPrimaryIndex(
arangodb::LogicalCollection* collection, VPackSlice const& info)
: RocksDBIndex(0, collection,
std::vector<std::vector<arangodb::basics::AttributeName>>(
{{arangodb::basics::AttributeName(
StaticStrings::KeyString, false)}}),
true, false, RocksDBColumnFamily::none(),
basics::VelocyPackHelper::stringUInt64(info, "objectId"),
false) {
// !ServerState::instance()->isCoordinator() /*useCache*/) {
TRI_ASSERT(_objectId != 0);
}
RocksDBPrimaryIndex::~RocksDBPrimaryIndex() {}
/// @brief return the number of documents from the index
size_t RocksDBPrimaryIndex::size() const {
// TODO
return 0;
}
/// @brief return the memory usage of the index
size_t RocksDBPrimaryIndex::memory() const {
rocksdb::TransactionDB* db = rocksutils::globalRocksDB();
RocksDBKeyBounds bounds = RocksDBKeyBounds::PrimaryIndex(_objectId);
rocksdb::Range r(bounds.start(), bounds.end());
uint64_t out;
db->GetApproximateSizes(&r, 1, &out, true);
return (size_t)out;
}
int RocksDBPrimaryIndex::cleanup() {
rocksdb::TransactionDB* db = rocksutils::globalRocksDB();
rocksdb::CompactRangeOptions opts;
RocksDBKeyBounds bounds = RocksDBKeyBounds::PrimaryIndex(_objectId);
rocksdb::Slice b = bounds.start(), e = bounds.end();
db->CompactRange(opts, &b, &e);
return TRI_ERROR_NO_ERROR;
}
/// @brief return a VelocyPack representation of the index
void RocksDBPrimaryIndex::toVelocyPack(VPackBuilder& builder, bool withFigures,
bool forPersistence) const {
builder.openObject();
RocksDBIndex::toVelocyPack(builder, withFigures, forPersistence);
// hard-coded
builder.add("unique", VPackValue(true));
builder.add("sparse", VPackValue(false));
builder.close();
}
RocksDBToken RocksDBPrimaryIndex::lookupKey(transaction::Methods* trx,
arangodb::StringRef keyRef) const {
auto key = RocksDBKey::PrimaryIndexValue(_objectId, keyRef);
auto value = RocksDBValue::Empty(RocksDBEntryType::PrimaryIndexValue);
if (useCache()) {
TRI_ASSERT(_cache != nullptr);
// check cache first for fast path
auto f = _cache->find(key.string().data(),
static_cast<uint32_t>(key.string().size()));
if (f.found()) {
value.buffer()->append(reinterpret_cast<char const*>(f.value()->value()),
static_cast<size_t>(f.value()->valueSize));
return RocksDBToken(RocksDBValue::revisionId(value));
}
}
// acquire rocksdb transaction
RocksDBMethods* mthds = rocksutils::toRocksMethods(trx);
auto& options = mthds->readOptions();
TRI_ASSERT(options.snapshot != nullptr);
arangodb::Result r = mthds->Get(key, value.buffer());
if (!r.ok()) {
return RocksDBToken();
}
if (useCache()) {
TRI_ASSERT(_cache != nullptr);
// write entry back to cache
auto entry = cache::CachedValue::construct(
key.string().data(), static_cast<uint32_t>(key.string().size()),
value.buffer()->data(), static_cast<uint64_t>(value.buffer()->size()));
bool cached = _cache->insert(entry);
if (!cached) {
delete entry;
}
}
return RocksDBToken(RocksDBValue::revisionId(value));
}
// TODO: remove this method?
RocksDBToken RocksDBPrimaryIndex::lookupKey(
transaction::Methods* trx, VPackSlice slice,
ManagedDocumentResult& result) const {
return lookupKey(trx, StringRef(slice));
}
int RocksDBPrimaryIndex::insert(transaction::Methods* trx,
TRI_voc_rid_t revisionId,
VPackSlice const& slice, bool) {
auto key = RocksDBKey::PrimaryIndexValue(
_objectId, StringRef(slice.get(StaticStrings::KeyString)));
auto value = RocksDBValue::PrimaryIndexValue(revisionId);
// acquire rocksdb transaction
RocksDBMethods* mthd = rocksutils::toRocksMethods(trx);
if (mthd->Exists(key)) {
return TRI_ERROR_ARANGO_UNIQUE_CONSTRAINT_VIOLATED;
}
blackListKey(key.string().data(), static_cast<uint32_t>(key.string().size()));
Result status = mthd->Put(key, value.string(), rocksutils::index);
return status.errorNumber();
}
int RocksDBPrimaryIndex::insertRaw(RocksDBMethods*, TRI_voc_rid_t,
VPackSlice const&) {
THROW_ARANGO_EXCEPTION(TRI_ERROR_NOT_IMPLEMENTED);
}
int RocksDBPrimaryIndex::remove(transaction::Methods* trx,
TRI_voc_rid_t revisionId,
VPackSlice const& slice, bool) {
// TODO: deal with matching revisions?
auto key = RocksDBKey::PrimaryIndexValue(
_objectId, StringRef(slice.get(StaticStrings::KeyString)));
blackListKey(key.string().data(), static_cast<uint32_t>(key.string().size()));
// acquire rocksdb transaction
RocksDBMethods* mthds = rocksutils::toRocksMethods(trx);
Result r = mthds->Delete(key);
//rocksutils::convertStatus(status, rocksutils::StatusHint::index);
return r.errorNumber();
}
/// optimization for truncateNoTrx, never called in fillIndex
int RocksDBPrimaryIndex::removeRaw(RocksDBMethods*, TRI_voc_rid_t,
VPackSlice const&) {
THROW_ARANGO_EXCEPTION(TRI_ERROR_NOT_IMPLEMENTED);
}
/// @brief called when the index is dropped
int RocksDBPrimaryIndex::drop() {
// First drop the cache all indexes can work without it.
RocksDBIndex::drop();
return rocksutils::removeLargeRange(rocksutils::globalRocksDB(),
RocksDBKeyBounds::PrimaryIndex(_objectId))
.errorNumber();
}
/// @brief checks whether the index supports the condition
bool RocksDBPrimaryIndex::supportsFilterCondition(
arangodb::aql::AstNode const* node,
arangodb::aql::Variable const* reference, size_t itemsInIndex,
size_t& estimatedItems, double& estimatedCost) const {
SimpleAttributeEqualityMatcher matcher(IndexAttributes);
return matcher.matchOne(this, node, reference, itemsInIndex, estimatedItems,
estimatedCost);
}
/// @brief creates an IndexIterator for the given Condition
IndexIterator* RocksDBPrimaryIndex::iteratorForCondition(
transaction::Methods* trx, ManagedDocumentResult* mmdr,
arangodb::aql::AstNode const* node,
arangodb::aql::Variable const* reference, bool reverse) {
TRI_ASSERT(node->type == aql::NODE_TYPE_OPERATOR_NARY_AND);
TRI_ASSERT(node->numMembers() == 1);
auto comp = node->getMember(0);
// assume a.b == value
auto attrNode = comp->getMember(0);
auto valNode = comp->getMember(1);
if (attrNode->type != aql::NODE_TYPE_ATTRIBUTE_ACCESS) {
// value == a.b -> flip the two sides
attrNode = comp->getMember(1);
valNode = comp->getMember(0);
}
TRI_ASSERT(attrNode->type == aql::NODE_TYPE_ATTRIBUTE_ACCESS);
if (comp->type == aql::NODE_TYPE_OPERATOR_BINARY_EQ) {
// a.b == value
return createEqIterator(trx, mmdr, attrNode, valNode);
} else if (comp->type == aql::NODE_TYPE_OPERATOR_BINARY_IN) {
// a.b IN values
if (!valNode->isArray()) {
// a.b IN non-array
return new EmptyIndexIterator(_collection, trx, mmdr, this);
}
return createInIterator(trx, mmdr, attrNode, valNode);
}
// operator type unsupported
return new EmptyIndexIterator(_collection, trx, mmdr, this);
}
/// @brief specializes the condition for use with the index
arangodb::aql::AstNode* RocksDBPrimaryIndex::specializeCondition(
arangodb::aql::AstNode* node,
arangodb::aql::Variable const* reference) const {
SimpleAttributeEqualityMatcher matcher(IndexAttributes);
return matcher.specializeOne(this, node, reference);
}
/// @brief request an iterator over all elements in the index in
/// a sequential order.
IndexIterator* RocksDBPrimaryIndex::allIterator(transaction::Methods* trx,
ManagedDocumentResult* mmdr,
bool reverse) const {
return new RocksDBAllIndexIterator(_collection, trx, mmdr, this, reverse);
}
/// @brief request an iterator over all elements in the index in
/// a sequential order.
IndexIterator* RocksDBPrimaryIndex::anyIterator(
transaction::Methods* trx, ManagedDocumentResult* mmdr) const {
return new RocksDBAnyIndexIterator(_collection, trx, mmdr, this);
}
void RocksDBPrimaryIndex::invokeOnAllElements(
transaction::Methods* trx,
std::function<bool(DocumentIdentifierToken const&)> callback) const {
ManagedDocumentResult mmdr;
std::unique_ptr<IndexIterator> cursor(allIterator(trx, &mmdr, false));
bool cnt = true;
auto cb = [&](DocumentIdentifierToken token) {
if (cnt) {
cnt = callback(token);
}
};
while (cursor->next(cb, 1000) && cnt) {
}
}
Result RocksDBPrimaryIndex::postprocessRemove(transaction::Methods* trx,
rocksdb::Slice const& key,
rocksdb::Slice const& value) {
blackListKey(key.data(), key.size());
return {TRI_ERROR_NO_ERROR};
}
/// @brief create the iterator, for a single attribute, IN operator
IndexIterator* RocksDBPrimaryIndex::createInIterator(
transaction::Methods* trx, ManagedDocumentResult* mmdr,
arangodb::aql::AstNode const* attrNode,
arangodb::aql::AstNode const* valNode) {
// _key or _id?
bool const isId = (attrNode->stringEquals(StaticStrings::IdString));
TRI_ASSERT(valNode->isArray());
// lease builder, but immediately pass it to the unique_ptr so we don't leak
transaction::BuilderLeaser builder(trx);
std::unique_ptr<VPackBuilder> keys(builder.steal());
keys->openArray();
size_t const n = valNode->numMembers();
// only leave the valid elements
for (size_t i = 0; i < n; ++i) {
handleValNode(trx, keys.get(), valNode->getMemberUnchecked(i), isId);
TRI_IF_FAILURE("PrimaryIndex::iteratorValNodes") {
THROW_ARANGO_EXCEPTION(TRI_ERROR_DEBUG);
}
}
TRI_IF_FAILURE("PrimaryIndex::noIterator") {
THROW_ARANGO_EXCEPTION(TRI_ERROR_DEBUG);
}
keys->close();
return new RocksDBPrimaryIndexIterator(_collection, trx, mmdr, this, keys);
}
/// @brief create the iterator, for a single attribute, EQ operator
IndexIterator* RocksDBPrimaryIndex::createEqIterator(
transaction::Methods* trx, ManagedDocumentResult* mmdr,
arangodb::aql::AstNode const* attrNode,
arangodb::aql::AstNode const* valNode) {
// _key or _id?
bool const isId = (attrNode->stringEquals(StaticStrings::IdString));
// lease builder, but immediately pass it to the unique_ptr so we don't leak
transaction::BuilderLeaser builder(trx);
std::unique_ptr<VPackBuilder> keys(builder.steal());
keys->openArray();
// handle the sole element
handleValNode(trx, keys.get(), valNode, isId);
TRI_IF_FAILURE("PrimaryIndex::noIterator") {
THROW_ARANGO_EXCEPTION(TRI_ERROR_DEBUG);
}
keys->close();
return new RocksDBPrimaryIndexIterator(_collection, trx, mmdr, this, keys);
}
/// @brief add a single value node to the iterator's keys
void RocksDBPrimaryIndex::handleValNode(transaction::Methods* trx,
VPackBuilder* keys,
arangodb::aql::AstNode const* valNode,
bool isId) const {
if (!valNode->isStringValue() || valNode->getStringLength() == 0) {
return;
}
if (isId) {
// lookup by _id. now validate if the lookup is performed for the
// correct collection (i.e. _collection)
TRI_voc_cid_t cid;
char const* key;
size_t outLength;
Result res =
trx->resolveId(valNode->getStringValue(),
valNode->getStringLength(), cid, key, outLength);
if (!res.ok()) {
return;
}
TRI_ASSERT(cid != 0);
TRI_ASSERT(key != nullptr);
if (!trx->state()->isRunningInCluster() && cid != _collection->cid()) {
// only continue lookup if the id value is syntactically correct and
// refers to "our" collection, using local collection id
return;
}
if (trx->state()->isRunningInCluster() && cid != _collection->planId()) {
// only continue lookup if the id value is syntactically correct and
// refers to "our" collection, using cluster collection id
return;
}
// use _key value from _id
keys->add(VPackValuePair(key, outLength, VPackValueType::String));
} else {
keys->add(VPackValuePair(valNode->getStringValue(),
valNode->getStringLength(),
VPackValueType::String));
}
}