ceph/src/rocksdb/utilities/transactions/transaction_util.cc

   1 //  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
   2 //  This source code is licensed under both the GPLv2 (found in the
   3 //  COPYING file in the root directory) and Apache 2.0 License
   4 //  (found in the LICENSE.Apache file in the root directory).
   5
   6 #ifndef ROCKSDB_LITE
   7
   8 #include "utilities/transactions/transaction_util.h"
   9
  10 #include <cinttypes>
  11 #include <string>
  12 #include <vector>
  13
  14 #include "db/db_impl/db_impl.h"
  15 #include "rocksdb/status.h"
  16 #include "rocksdb/utilities/write_batch_with_index.h"
  17 #include "util/cast_util.h"
  18 #include "util/string_util.h"
  19
  20 namespace ROCKSDB_NAMESPACE {
  21
  22 Status TransactionUtil::CheckKeyForConflicts(
  23     DBImpl* db_impl, ColumnFamilyHandle* column_family, const std::string& key,
  24     SequenceNumber snap_seq, bool cache_only, ReadCallback* snap_checker,
  25     SequenceNumber min_uncommitted) {
  26   Status result;
  27
  28   auto cfh = static_cast_with_check<ColumnFamilyHandleImpl>(column_family);
  29   auto cfd = cfh->cfd();
  30   SuperVersion* sv = db_impl->GetAndRefSuperVersion(cfd);
  31
  32   if (sv == nullptr) {
  33     result = Status::InvalidArgument("Could not access column family " +
  34                                      cfh->GetName());
  35   }
  36
  37   if (result.ok()) {
  38     SequenceNumber earliest_seq =
  39         db_impl->GetEarliestMemTableSequenceNumber(sv, true);
  40
  41     result = CheckKey(db_impl, sv, earliest_seq, snap_seq, key, cache_only,
  42                       snap_checker, min_uncommitted);
  43
  44     db_impl->ReturnAndCleanupSuperVersion(cfd, sv);
  45   }
  46
  47   return result;
  48 }
  49
  50 Status TransactionUtil::CheckKey(DBImpl* db_impl, SuperVersion* sv,
  51                                  SequenceNumber earliest_seq,
  52                                  SequenceNumber snap_seq,
  53                                  const std::string& key, bool cache_only,
  54                                  ReadCallback* snap_checker,
  55                                  SequenceNumber min_uncommitted) {
  56   // When `min_uncommitted` is provided, keys are not always committed
  57   // in sequence number order, and `snap_checker` is used to check whether
  58   // specific sequence number is in the database is visible to the transaction.
  59   // So `snap_checker` must be provided.
  60   assert(min_uncommitted == kMaxSequenceNumber || snap_checker != nullptr);
  61
  62   Status result;
  63   bool need_to_read_sst = false;
  64
  65   // Since it would be too slow to check the SST files, we will only use
  66   // the memtables to check whether there have been any recent writes
  67   // to this key after it was accessed in this transaction.  But if the
  68   // Memtables do not contain a long enough history, we must fail the
  69   // transaction.
  70   if (earliest_seq == kMaxSequenceNumber) {
  71     // The age of this memtable is unknown.  Cannot rely on it to check
  72     // for recent writes.  This error shouldn't happen often in practice as
  73     // the Memtable should have a valid earliest sequence number except in some
  74     // corner cases (such as error cases during recovery).
  75     need_to_read_sst = true;
  76
  77     if (cache_only) {
  78       result = Status::TryAgain(
  79           "Transaction could not check for conflicts as the MemTable does not "
  80           "contain a long enough history to check write at SequenceNumber: ",
  81           ToString(snap_seq));
  82     }
  83   } else if (snap_seq < earliest_seq || min_uncommitted <= earliest_seq) {
  84     // Use <= for min_uncommitted since earliest_seq is actually the largest sec
  85     // before this memtable was created
  86     need_to_read_sst = true;
  87
  88     if (cache_only) {
  89       // The age of this memtable is too new to use to check for recent
  90       // writes.
  91       char msg[300];
  92       snprintf(msg, sizeof(msg),
  93                "Transaction could not check for conflicts for operation at "
  94                "SequenceNumber %" PRIu64
  95                " as the MemTable only contains changes newer than "
  96                "SequenceNumber %" PRIu64
  97                ".  Increasing the value of the "
  98                "max_write_buffer_size_to_maintain option could reduce the "
  99                "frequency "
 100                "of this error.",
 101                snap_seq, earliest_seq);
 102       result = Status::TryAgain(msg);
 103     }
 104   }
 105
 106   if (result.ok()) {
 107     SequenceNumber seq = kMaxSequenceNumber;
 108     bool found_record_for_key = false;
 109
 110     // When min_uncommitted == kMaxSequenceNumber, writes are committed in
 111     // sequence number order, so only keys larger than `snap_seq` can cause
 112     // conflict.
 113     // When min_uncommitted != kMaxSequenceNumber, keys lower than
 114     // min_uncommitted will not triggered conflicts, while keys larger than
 115     // min_uncommitted might create conflicts, so we need  to read them out
 116     // from the DB, and call callback to snap_checker to determine. So only
 117     // keys lower than min_uncommitted can be skipped.
 118     SequenceNumber lower_bound_seq =
 119         (min_uncommitted == kMaxSequenceNumber) ? snap_seq : min_uncommitted;
 120     Status s = db_impl->GetLatestSequenceForKey(sv, key, !need_to_read_sst,
 121                                                 lower_bound_seq, &seq,
 122                                                 &found_record_for_key);
 123
 124     if (!(s.ok() || s.IsNotFound() || s.IsMergeInProgress())) {
 125       result = s;
 126     } else if (found_record_for_key) {
 127       bool write_conflict = snap_checker == nullptr
 128                                 ? snap_seq < seq
 129                                 : !snap_checker->IsVisible(seq);
 130       if (write_conflict) {
 131         result = Status::Busy();
 132       }
 133     }
 134   }
 135
 136   return result;
 137 }
 138
 139 Status TransactionUtil::CheckKeysForConflicts(DBImpl* db_impl,
 140                                               const LockTracker& tracker,
 141                                               bool cache_only) {
 142   Status result;
 143
 144   std::unique_ptr<LockTracker::ColumnFamilyIterator> cf_it(
 145       tracker.GetColumnFamilyIterator());
 146   assert(cf_it != nullptr);
 147   while (cf_it->HasNext()) {
 148     ColumnFamilyId cf = cf_it->Next();
 149
 150     SuperVersion* sv = db_impl->GetAndRefSuperVersion(cf);
 151     if (sv == nullptr) {
 152       result = Status::InvalidArgument("Could not access column family " +
 153                                        ToString(cf));
 154       break;
 155     }
 156
 157     SequenceNumber earliest_seq =
 158         db_impl->GetEarliestMemTableSequenceNumber(sv, true);
 159
 160     // For each of the keys in this transaction, check to see if someone has
 161     // written to this key since the start of the transaction.
 162     std::unique_ptr<LockTracker::KeyIterator> key_it(
 163         tracker.GetKeyIterator(cf));
 164     assert(key_it != nullptr);
 165     while (key_it->HasNext()) {
 166       const std::string& key = key_it->Next();
 167       PointLockStatus status = tracker.GetPointLockStatus(cf, key);
 168       const SequenceNumber key_seq = status.seq;
 169
 170       result = CheckKey(db_impl, sv, earliest_seq, key_seq, key, cache_only);
 171       if (!result.ok()) {
 172         break;
 173       }
 174     }
 175
 176     db_impl->ReturnAndCleanupSuperVersion(cf, sv);
 177
 178     if (!result.ok()) {
 179       break;
 180     }
 181   }
 182
 183   return result;
 184 }
 185
 186 }  // namespace ROCKSDB_NAMESPACE
 187
 188 #endif  // ROCKSDB_LITE