File: partitioned_index_reader.cc

package info (click to toggle)
rocksdb 9.10.0-1
  • links: PTS, VCS
  • area: main
  • in suites: forky, trixie
  • size: 46,052 kB
  • sloc: cpp: 500,768; java: 42,992; ansic: 9,789; python: 8,373; perl: 5,822; sh: 4,921; makefile: 2,386; asm: 550; xml: 342
file content (264 lines) | stat: -rw-r--r-- 10,120 bytes parent folder | download | duplicates (2)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
//  This source code is licensed under both the GPLv2 (found in the
//  COPYING file in the root directory) and Apache 2.0 License
//  (found in the LICENSE.Apache file in the root directory).
//
// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file. See the AUTHORS file for names of contributors.
#include "table/block_based/partitioned_index_reader.h"

#include "block_cache.h"
#include "file/random_access_file_reader.h"
#include "table/block_based/block_based_table_reader.h"
#include "table/block_based/partitioned_index_iterator.h"

namespace ROCKSDB_NAMESPACE {
Status PartitionIndexReader::Create(
    const BlockBasedTable* table, const ReadOptions& ro,
    FilePrefetchBuffer* prefetch_buffer, bool use_cache, bool prefetch,
    bool pin, BlockCacheLookupContext* lookup_context,
    std::unique_ptr<IndexReader>* index_reader) {
  assert(table != nullptr);
  assert(table->get_rep());
  assert(!pin || prefetch);
  assert(index_reader != nullptr);

  CachableEntry<Block> index_block;
  if (prefetch || !use_cache) {
    const Status s =
        ReadIndexBlock(table, prefetch_buffer, ro, use_cache,
                       /*get_context=*/nullptr, lookup_context, &index_block);
    if (!s.ok()) {
      return s;
    }

    if (use_cache && !pin) {
      index_block.Reset();
    }
  }

  index_reader->reset(new PartitionIndexReader(table, std::move(index_block)));

  return Status::OK();
}

InternalIteratorBase<IndexValue>* PartitionIndexReader::NewIterator(
    const ReadOptions& read_options, bool /* disable_prefix_seek */,
    IndexBlockIter* iter, GetContext* get_context,
    BlockCacheLookupContext* lookup_context) {
  CachableEntry<Block> index_block;
  const Status s = GetOrReadIndexBlock(get_context, lookup_context,
                                       &index_block, read_options);
  if (!s.ok()) {
    if (iter != nullptr) {
      iter->Invalidate(s);
      return iter;
    }

    return NewErrorInternalIterator<IndexValue>(s);
  }

  const BlockBasedTable::Rep* rep = table()->rep_;
  InternalIteratorBase<IndexValue>* it = nullptr;

  Statistics* kNullStats = nullptr;
  // Filters are already checked before seeking the index
  if (!partition_map_.empty()) {
    // We don't return pinned data from index blocks, so no need
    // to set `block_contents_pinned`.
    it = NewTwoLevelIterator(
        new BlockBasedTable::PartitionedIndexIteratorState(table(),
                                                           &partition_map_),
        index_block.GetValue()->NewIndexIterator(
            internal_comparator()->user_comparator(),
            rep->get_global_seqno(BlockType::kIndex), nullptr, kNullStats, true,
            index_has_first_key(), index_key_includes_seq(),
            index_value_is_full(), false /* block_contents_pinned */,
            user_defined_timestamps_persisted()));
  } else {
    ReadOptions ro{read_options};
    // FIXME? Possible regression seen in prefetch_test if this field is
    // propagated
    ro.readahead_size = ReadOptions{}.readahead_size;

    // We don't return pinned data from index blocks, so no need
    // to set `block_contents_pinned`.
    std::unique_ptr<InternalIteratorBase<IndexValue>> index_iter(
        index_block.GetValue()->NewIndexIterator(
            internal_comparator()->user_comparator(),
            rep->get_global_seqno(BlockType::kIndex), nullptr, kNullStats, true,
            index_has_first_key(), index_key_includes_seq(),
            index_value_is_full(), false /* block_contents_pinned */,
            user_defined_timestamps_persisted()));

    it = new PartitionedIndexIterator(
        table(), ro, *internal_comparator(), std::move(index_iter),
        lookup_context ? lookup_context->caller
                       : TableReaderCaller::kUncategorized);
  }

  assert(it != nullptr);
  index_block.TransferTo(it);

  return it;

  // TODO(myabandeh): Update TwoLevelIterator to be able to make use of
  // on-stack BlockIter while the state is on heap. Currentlly it assumes
  // the first level iter is always on heap and will attempt to delete it
  // in its destructor.
}
Status PartitionIndexReader::CacheDependencies(
    const ReadOptions& ro, bool pin, FilePrefetchBuffer* tail_prefetch_buffer) {
  if (!partition_map_.empty()) {
    // The dependencies are already cached since `partition_map_` is filled in
    // an all-or-nothing manner.
    return Status::OK();
  }
  // Before read partitions, prefetch them to avoid lots of IOs
  BlockCacheLookupContext lookup_context{TableReaderCaller::kPrefetch};
  const BlockBasedTable::Rep* rep = table()->rep_;
  IndexBlockIter biter;
  BlockHandle handle;
  Statistics* kNullStats = nullptr;

  CachableEntry<Block> index_block;
  {
    Status s = GetOrReadIndexBlock(nullptr /* get_context */, &lookup_context,
                                   &index_block, ro);
    if (!s.ok()) {
      return s;
    }
  }

  // We don't return pinned data from index blocks, so no need
  // to set `block_contents_pinned`.
  index_block.GetValue()->NewIndexIterator(
      internal_comparator()->user_comparator(),
      rep->get_global_seqno(BlockType::kIndex), &biter, kNullStats, true,
      index_has_first_key(), index_key_includes_seq(), index_value_is_full(),
      false /* block_contents_pinned */, user_defined_timestamps_persisted());
  // Index partitions are assumed to be consecuitive. Prefetch them all.
  // Read the first block offset
  biter.SeekToFirst();
  if (!biter.Valid()) {
    // Empty index.
    return biter.status();
  }
  handle = biter.value().handle;
  uint64_t prefetch_off = handle.offset();

  // Read the last block's offset
  biter.SeekToLast();
  if (!biter.Valid()) {
    // Empty index.
    return biter.status();
  }
  handle = biter.value().handle;
  uint64_t last_off =
      handle.offset() + BlockBasedTable::BlockSizeWithTrailer(handle);
  uint64_t prefetch_len = last_off - prefetch_off;
  std::unique_ptr<FilePrefetchBuffer> prefetch_buffer;
  if (tail_prefetch_buffer == nullptr || !tail_prefetch_buffer->Enabled() ||
      tail_prefetch_buffer->GetPrefetchOffset() > prefetch_off) {
    rep->CreateFilePrefetchBuffer(ReadaheadParams(), &prefetch_buffer,
                                  /*readaheadsize_cb*/ nullptr,
                                  /*usage=*/FilePrefetchBufferUsage::kUnknown);
    IOOptions opts;
    {
      Status s = rep->file->PrepareIOOptions(ro, opts);
      if (s.ok()) {
        s = prefetch_buffer->Prefetch(opts, rep->file.get(), prefetch_off,
                                      static_cast<size_t>(prefetch_len));
      }
      if (!s.ok()) {
        return s;
      }
    }
  }
  // For saving "all or nothing" to partition_map_
  UnorderedMap<uint64_t, CachableEntry<Block>> map_in_progress;

  // After prefetch, read the partitions one by one
  biter.SeekToFirst();
  size_t partition_count = 0;
  for (; biter.Valid(); biter.Next()) {
    handle = biter.value().handle;
    CachableEntry<Block> block;
    ++partition_count;
    // TODO: Support counter batch update for partitioned index and
    // filter blocks
    Status s = table()->MaybeReadBlockAndLoadToCache(
        prefetch_buffer ? prefetch_buffer.get() : tail_prefetch_buffer, ro,
        handle, UncompressionDict::GetEmptyDict(),
        /*for_compaction=*/false, &block.As<Block_kIndex>(),
        /*get_context=*/nullptr, &lookup_context, /*contents=*/nullptr,
        /*async_read=*/false, /*use_block_cache_for_lookup=*/true);

    if (!s.ok()) {
      return s;
    }
    if (block.GetValue() != nullptr) {
      // Might need to "pin" some mmap-read blocks (GetOwnValue) if some
      // partitions are successfully compressed (cached) and some are not
      // compressed (mmap eligible)
      if (block.IsCached() || block.GetOwnValue()) {
        if (pin) {
          map_in_progress[handle.offset()] = std::move(block);
        }
      }
    }
  }
  Status s = biter.status();
  // Save (pin) them only if everything checks out
  if (map_in_progress.size() == partition_count && s.ok()) {
    std::swap(partition_map_, map_in_progress);
  }
  return s;
}

void PartitionIndexReader::EraseFromCacheBeforeDestruction(
    uint32_t uncache_aggressiveness) {
  // NOTE: essentially a copy of
  // PartitionedFilterBlockReader::EraseFromCacheBeforeDestruction
  if (uncache_aggressiveness > 0) {
    CachableEntry<Block> top_level_block;

    ReadOptions ro_no_io;
    ro_no_io.read_tier = ReadTier::kBlockCacheTier;
    GetOrReadIndexBlock(/*get_context=*/nullptr,
                        /*lookup_context=*/nullptr, &top_level_block, ro_no_io)
        .PermitUncheckedError();

    if (!partition_map_.empty()) {
      // All partitions present if any
      for (auto& e : partition_map_) {
        e.second.ResetEraseIfLastRef();
      }
    } else if (!top_level_block.IsEmpty()) {
      IndexBlockIter biter;
      const InternalKeyComparator* const comparator = internal_comparator();
      Statistics* kNullStats = nullptr;
      top_level_block.GetValue()->NewIndexIterator(
          comparator->user_comparator(),
          table()->get_rep()->get_global_seqno(BlockType::kIndex), &biter,
          kNullStats, true /* total_order_seek */, index_has_first_key(),
          index_key_includes_seq(), index_value_is_full(),
          false /* block_contents_pinned */,
          user_defined_timestamps_persisted());

      UncacheAggressivenessAdvisor advisor(uncache_aggressiveness);
      for (biter.SeekToFirst(); biter.Valid() && advisor.ShouldContinue();
           biter.Next()) {
        bool erased = table()->EraseFromCache(biter.value().handle);
        advisor.Report(erased);
      }
      biter.status().PermitUncheckedError();
    }
    top_level_block.ResetEraseIfLastRef();
  }
  // Might be needed to un-cache a pinned top-level block
  BlockBasedTable::IndexReaderCommon::EraseFromCacheBeforeDestruction(
      uncache_aggressiveness);
}
}  // namespace ROCKSDB_NAMESPACE