arangodb/arangod/Cache/TransactionalCache.cpp

////////////////////////////////////////////////////////////////////////////////
/// 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 Daniel H. Larkin
////////////////////////////////////////////////////////////////////////////////

#include "Cache/TransactionalCache.h"
#include "Basics/Common.h"
#include "Cache/Cache.h"
#include "Cache/CachedValue.h"
#include "Cache/Common.h"
#include "Cache/Finding.h"
#include "Cache/FrequencyBuffer.h"
#include "Cache/Metadata.h"
#include "Cache/Table.h"
#include "Cache/TransactionalBucket.h"
#include "Logger/Logger.h"

#include <stdint.h>
#include <atomic>
#include <chrono>
#include <list>

using namespace arangodb;
using namespace arangodb::cache;

Finding TransactionalCache::find(void const* key, uint32_t keySize) {
  TRI_ASSERT(key != nullptr);
  Finding result;
  uint32_t hash = hashKey(key, keySize);

  Result status;
  TransactionalBucket* bucket;
  Table* source;
  std::tie(status, bucket, source) = getBucket(hash, Cache::triesFast);
  if (status.fail()) {
    result.reportError(status);
    return result;
  }

  result.set(bucket->find(hash, key, keySize));
  if (result.found()) {
    recordStat(Stat::findHit);
  } else {
    recordStat(Stat::findMiss);
    status.reset(TRI_ERROR_ARANGO_DOCUMENT_NOT_FOUND);
    result.reportError(status);
  }
  recordStat(result.found() ? Stat::findHit : Stat::findMiss);
  bucket->unlock();

  return result;
}

Result TransactionalCache::insert(CachedValue* value) {
  TRI_ASSERT(value != nullptr);
  uint32_t hash = hashKey(value->key(), value->keySize());

  Result status;
  TransactionalBucket* bucket;
  Table* source;
  std::tie(status, bucket, source) = getBucket(hash, Cache::triesFast);
  if (status.fail()) {
    return status;
  }

  bool maybeMigrate = false;
  bool allowed = !bucket->isBlacklisted(hash);
  if (allowed) {
    int64_t change = static_cast<int64_t>(value->size());
    CachedValue* candidate = bucket->find(hash, value->key(), value->keySize());

    if (candidate == nullptr && bucket->isFull()) {
      candidate = bucket->evictionCandidate();
      if (candidate == nullptr) {
        allowed = false;
        status.reset(TRI_ERROR_ARANGO_BUSY);
      }
    }

    if (allowed) {
      if (candidate != nullptr) {
        change -= static_cast<int64_t>(candidate->size());
      }

      _metadata.readLock();
      allowed = _metadata.adjustUsageIfAllowed(change);
      _metadata.readUnlock();

      if (allowed) {
        bool eviction = false;
        if (candidate != nullptr) {
          bucket->evict(candidate, true);
          if (!candidate->sameKey(value->key(), value->keySize())) {
            eviction = true;
          }
          freeValue(candidate);
        }
        bucket->insert(hash, value);
        if (!eviction) {
          maybeMigrate = source->slotFilled();
        }
        maybeMigrate |= reportInsert(eviction);
      } else {
        requestGrow();  // let function do the hard work
        status.reset(TRI_ERROR_RESOURCE_LIMIT);
      }
    }
  } else {
    status.reset(TRI_ERROR_ARANGO_CONFLICT);
  }

  bucket->unlock();
  if (maybeMigrate) {
    requestMigrate(_table->idealSize());  // let function do the hard work
  }

  return status;
}

Result TransactionalCache::remove(void const* key, uint32_t keySize) {
  TRI_ASSERT(key != nullptr);
  uint32_t hash = hashKey(key, keySize);

  Result status;
  TransactionalBucket* bucket;
  Table* source;
  std::tie(status, bucket, source) = getBucket(hash, Cache::triesSlow);
  if (status.fail()) {
    return status;
  }

  bool maybeMigrate = false;
  CachedValue* candidate = bucket->remove(hash, key, keySize);

  if (candidate != nullptr) {
    int64_t change = -static_cast<int64_t>(candidate->size());
     _metadata.readLock();
    bool allowed = _metadata.adjustUsageIfAllowed(change);
    TRI_ASSERT(allowed);
    _metadata.readUnlock();

    freeValue(candidate);
    maybeMigrate = source->slotEmptied();
  }

  bucket->unlock();
  if (maybeMigrate) {
    requestMigrate(_table->idealSize());
  }

  return status;
}

Result TransactionalCache::blacklist(void const* key, uint32_t keySize) {
  TRI_ASSERT(key != nullptr);
  uint32_t hash = hashKey(key, keySize);

  Result status;
  TransactionalBucket* bucket;
  Table* source;
  std::tie(status, bucket, source) = getBucket(hash, Cache::triesSlow);
  if (status.fail()) {
    return status;
  }

  bool maybeMigrate = false;
  CachedValue* candidate = bucket->blacklist(hash, key, keySize);

  if (candidate != nullptr) {
    int64_t change = -static_cast<int64_t>(candidate->size());

    _metadata.readLock();
    bool allowed = _metadata.adjustUsageIfAllowed(change);
    TRI_ASSERT(allowed);
    _metadata.readUnlock();

    freeValue(candidate);
    maybeMigrate = source->slotEmptied();
  }

  bucket->unlock();
  if (maybeMigrate) {
    requestMigrate(_table->idealSize());
  }

  return status;
}

uint64_t TransactionalCache::allocationSize(bool enableWindowedStats) {
  return sizeof(TransactionalCache) +
         (enableWindowedStats ? (sizeof(StatBuffer) +
                                 StatBuffer::allocationSize(_findStatsCapacity))
                              : 0);
}

std::shared_ptr<Cache> TransactionalCache::create(Manager* manager,
                                                  uint64_t id,
                                                  Metadata&& metadata,
                                                  std::shared_ptr<Table> table,
                                                  bool enableWindowedStats) {
  return std::make_shared<TransactionalCache>(Cache::ConstructionGuard(),
                                              manager, id, std::move(metadata), table,
                                              enableWindowedStats);
}

TransactionalCache::TransactionalCache(Cache::ConstructionGuard guard,
                                       Manager* manager, uint64_t id, Metadata&& metadata,
                                       std::shared_ptr<Table> table,
                                       bool enableWindowedStats)
    : Cache(guard, manager, id, std::move(metadata), table, enableWindowedStats,
            TransactionalCache::bucketClearer, TransactionalBucket::slotsData) {
}

TransactionalCache::~TransactionalCache() {
  if (!_shutdown) {
    shutdown();
  }
}

uint64_t TransactionalCache::freeMemoryFrom(uint32_t hash) {
  uint64_t reclaimed = 0;
  Result status;
  TransactionalBucket* bucket;
  Table* source;
  std::tie(status, bucket, source) = getBucket(hash, Cache::triesFast, false);
  if (status.fail()) {
    return 0;
  }

  bool maybeMigrate = false;
  // evict LRU freeable value if exists
  CachedValue* candidate = bucket->evictionCandidate();

  if (candidate != nullptr) {
    reclaimed = candidate->size();
    bucket->evict(candidate);
    freeValue(candidate);
    maybeMigrate = source->slotEmptied();
  }

  bucket->unlock();

  uint32_t size = _table->idealSize();
  if (maybeMigrate) {
    requestMigrate(size);
  }

  return reclaimed;
}

void TransactionalCache::migrateBucket(void* sourcePtr,
                                       std::unique_ptr<Table::Subtable> targets,
                                       std::shared_ptr<Table> newTable) {
  uint64_t term = _manager->_transactions.term();

  // lock current bucket
  auto source = reinterpret_cast<TransactionalBucket*>(sourcePtr);
  source->lock(Cache::triesGuarantee);
  term = std::max(term, source->_blacklistTerm);

  // lock target bucket(s)
  targets->applyToAllBuckets([&term](void* ptr) -> bool {
    auto targetBucket = reinterpret_cast<TransactionalBucket*>(ptr);
    bool locked = targetBucket->lock(Cache::triesGuarantee);
    term = std::max(term, targetBucket->_blacklistTerm);
    return locked;
  });

  // update all buckets to maximum term found (guaranteed at most the current)
  source->updateBlacklistTerm(term);
  targets->applyToAllBuckets([&term](void* ptr) -> bool {
    auto targetBucket = reinterpret_cast<TransactionalBucket*>(ptr);
    targetBucket->updateBlacklistTerm(term);
    return true;
  });
  // now actually migrate any relevant blacklist terms
  if (source->isFullyBlacklisted()) {
    targets->applyToAllBuckets([](void* ptr) -> bool {
      auto targetBucket = reinterpret_cast<TransactionalBucket*>(ptr);
      if (!targetBucket->isFullyBlacklisted()) {
        targetBucket->_state.toggleFlag(BucketState::Flag::blacklisted);
      }
      return true;
    });
  } else {
    for (size_t j = 0; j < TransactionalBucket::slotsBlacklist; j++) {
      uint32_t hash = source->_blacklistHashes[j];
      if (hash != 0) {
        auto targetBucket =
            reinterpret_cast<TransactionalBucket*>(targets->fetchBucket(hash));
        CachedValue* candidate = targetBucket->blacklist(hash, nullptr, 0);
        if (candidate != nullptr) {
          uint64_t size = candidate->size();
          freeValue(candidate);
          reclaimMemory(size);
          newTable->slotEmptied();
        }
        source->_blacklistHashes[j] = 0;
      }
    }
  }

  // migrate actual values
  for (size_t j = 0; j < TransactionalBucket::slotsData; j++) {
    size_t k = TransactionalBucket::slotsData - (j + 1);
    if (source->_cachedData[k] != nullptr) {
      uint32_t hash = source->_cachedHashes[k];
      CachedValue* value = source->_cachedData[k];

      auto targetBucket =
          reinterpret_cast<TransactionalBucket*>(targets->fetchBucket(hash));
      if (targetBucket->isBlacklisted(hash)) {
        uint64_t size = value->size();
        freeValue(value);
        reclaimMemory(size);
      } else {
        bool haveSpace = true;
        if (targetBucket->isFull()) {
          CachedValue* candidate = targetBucket->evictionCandidate();
          if (candidate != nullptr) {
            targetBucket->evict(candidate, true);
            uint64_t size = candidate->size();
            freeValue(candidate);
            reclaimMemory(size);
            newTable->slotEmptied();
          } else {
            haveSpace = false;
          }
        }
        if (haveSpace) {
          targetBucket->insert(hash, value);
          newTable->slotFilled();
        } else {
          uint64_t size = value->size();
          freeValue(value);
          reclaimMemory(size);
        }
      }

      source->_cachedHashes[k] = 0;
      source->_cachedData[k] = nullptr;
    }
  }

  // unlock targets
  targets->applyToAllBuckets([](void* ptr) -> bool {
    auto bucket = reinterpret_cast<TransactionalBucket*>(ptr);
    bucket->unlock();
    return true;
  });

  // finish up this bucket's migration
  source->_state.toggleFlag(BucketState::Flag::migrated);
  source->unlock();
}

std::tuple<Result, TransactionalBucket*, Table*>
TransactionalCache::getBucket(uint32_t hash, uint64_t maxTries,
                              bool singleOperation) {
  Result status;
  TransactionalBucket* bucket = nullptr;
  Table* source = nullptr;

  Table* table = _table;
  if (isShutdown() || table == nullptr) {
    status.reset(TRI_ERROR_SHUTTING_DOWN);
    return std::make_tuple(status, bucket, source);
  }

  if (singleOperation) {
    _manager->reportAccess(_id);
  }

  uint64_t term = _manager->_transactions.term();
  auto pair = table->fetchAndLockBucket(hash, maxTries);
  bucket = reinterpret_cast<TransactionalBucket*>(pair.first);
  source = pair.second;
  bool ok = (bucket != nullptr);
  if (ok) {
    bucket->updateBlacklistTerm(term);
  } else {
    status.reset(TRI_ERROR_LOCK_TIMEOUT);
  }

  return std::make_tuple(status, bucket, source);
}

Table::BucketClearer TransactionalCache::bucketClearer(Metadata* metadata) {
  return [metadata](void* ptr) -> void {
    auto bucket = reinterpret_cast<TransactionalBucket*>(ptr);
    bucket->lock(Cache::triesGuarantee);
    for (size_t j = 0; j < TransactionalBucket::slotsData; j++) {
      if (bucket->_cachedData[j] != nullptr) {
        uint64_t size = bucket->_cachedData[j]->size();
        freeValue(bucket->_cachedData[j]);
        metadata->readLock();
        metadata->adjustUsageIfAllowed(-static_cast<int64_t>(size));
        metadata->readUnlock();
      }
    }
    bucket->clear();
  };
}