]>
Commit | Line | Data |
---|---|---|
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 | // Copyright (c) 2011 The LevelDB Authors. All rights reserved. | |
7 | // Use of this source code is governed by a BSD-style license that can be | |
8 | // found in the LICENSE file. See the AUTHORS file for names of contributors. | |
9 | ||
10 | #pragma once | |
11 | #include <atomic> | |
12 | #include <deque> | |
13 | #include <functional> | |
14 | #include <memory> | |
15 | #include <string> | |
16 | #include <unordered_map> | |
17 | #include <vector> | |
18 | #include "db/dbformat.h" | |
19 | #include "db/range_tombstone_fragmenter.h" | |
20 | #include "db/read_callback.h" | |
21 | #include "db/version_edit.h" | |
22 | #include "memory/allocator.h" | |
23 | #include "memory/concurrent_arena.h" | |
24 | #include "monitoring/instrumented_mutex.h" | |
25 | #include "options/cf_options.h" | |
26 | #include "rocksdb/db.h" | |
27 | #include "rocksdb/env.h" | |
28 | #include "rocksdb/memtablerep.h" | |
29 | #include "table/multiget_context.h" | |
30 | #include "util/dynamic_bloom.h" | |
31 | #include "util/hash.h" | |
32 | ||
33 | namespace ROCKSDB_NAMESPACE { | |
34 | ||
35 | struct FlushJobInfo; | |
36 | class Mutex; | |
37 | class MemTableIterator; | |
38 | class MergeContext; | |
39 | ||
40 | struct ImmutableMemTableOptions { | |
41 | explicit ImmutableMemTableOptions(const ImmutableCFOptions& ioptions, | |
42 | const MutableCFOptions& mutable_cf_options); | |
43 | size_t arena_block_size; | |
44 | uint32_t memtable_prefix_bloom_bits; | |
45 | size_t memtable_huge_page_size; | |
46 | bool memtable_whole_key_filtering; | |
47 | bool inplace_update_support; | |
48 | size_t inplace_update_num_locks; | |
49 | UpdateStatus (*inplace_callback)(char* existing_value, | |
50 | uint32_t* existing_value_size, | |
51 | Slice delta_value, | |
52 | std::string* merged_value); | |
53 | size_t max_successive_merges; | |
54 | Statistics* statistics; | |
55 | MergeOperator* merge_operator; | |
56 | Logger* info_log; | |
57 | bool allow_data_in_errors; | |
58 | }; | |
59 | ||
60 | // Batched counters to updated when inserting keys in one write batch. | |
61 | // In post process of the write batch, these can be updated together. | |
62 | // Only used in concurrent memtable insert case. | |
63 | struct MemTablePostProcessInfo { | |
64 | uint64_t data_size = 0; | |
65 | uint64_t num_entries = 0; | |
66 | uint64_t num_deletes = 0; | |
67 | }; | |
68 | ||
69 | using MultiGetRange = MultiGetContext::Range; | |
70 | // Note: Many of the methods in this class have comments indicating that | |
71 | // external synchronization is required as these methods are not thread-safe. | |
72 | // It is up to higher layers of code to decide how to prevent concurrent | |
73 | // invokation of these methods. This is usually done by acquiring either | |
74 | // the db mutex or the single writer thread. | |
75 | // | |
76 | // Some of these methods are documented to only require external | |
77 | // synchronization if this memtable is immutable. Calling MarkImmutable() is | |
78 | // not sufficient to guarantee immutability. It is up to higher layers of | |
79 | // code to determine if this MemTable can still be modified by other threads. | |
80 | // Eg: The Superversion stores a pointer to the current MemTable (that can | |
81 | // be modified) and a separate list of the MemTables that can no longer be | |
82 | // written to (aka the 'immutable memtables'). | |
83 | class MemTable { | |
84 | public: | |
85 | struct KeyComparator : public MemTableRep::KeyComparator { | |
86 | const InternalKeyComparator comparator; | |
87 | explicit KeyComparator(const InternalKeyComparator& c) : comparator(c) { } | |
88 | virtual int operator()(const char* prefix_len_key1, | |
89 | const char* prefix_len_key2) const override; | |
90 | virtual int operator()(const char* prefix_len_key, | |
91 | const DecodedType& key) const override; | |
92 | }; | |
93 | ||
94 | // MemTables are reference counted. The initial reference count | |
95 | // is zero and the caller must call Ref() at least once. | |
96 | // | |
97 | // earliest_seq should be the current SequenceNumber in the db such that any | |
98 | // key inserted into this memtable will have an equal or larger seq number. | |
99 | // (When a db is first created, the earliest sequence number will be 0). | |
100 | // If the earliest sequence number is not known, kMaxSequenceNumber may be | |
101 | // used, but this may prevent some transactions from succeeding until the | |
102 | // first key is inserted into the memtable. | |
103 | explicit MemTable(const InternalKeyComparator& comparator, | |
104 | const ImmutableCFOptions& ioptions, | |
105 | const MutableCFOptions& mutable_cf_options, | |
106 | WriteBufferManager* write_buffer_manager, | |
107 | SequenceNumber earliest_seq, uint32_t column_family_id); | |
108 | // No copying allowed | |
109 | MemTable(const MemTable&) = delete; | |
110 | MemTable& operator=(const MemTable&) = delete; | |
111 | ||
112 | // Do not delete this MemTable unless Unref() indicates it not in use. | |
113 | ~MemTable(); | |
114 | ||
115 | // Increase reference count. | |
116 | // REQUIRES: external synchronization to prevent simultaneous | |
117 | // operations on the same MemTable. | |
118 | void Ref() { ++refs_; } | |
119 | ||
120 | // Drop reference count. | |
121 | // If the refcount goes to zero return this memtable, otherwise return null. | |
122 | // REQUIRES: external synchronization to prevent simultaneous | |
123 | // operations on the same MemTable. | |
124 | MemTable* Unref() { | |
125 | --refs_; | |
126 | assert(refs_ >= 0); | |
127 | if (refs_ <= 0) { | |
128 | return this; | |
129 | } | |
130 | return nullptr; | |
131 | } | |
132 | ||
133 | // Returns an estimate of the number of bytes of data in use by this | |
134 | // data structure. | |
135 | // | |
136 | // REQUIRES: external synchronization to prevent simultaneous | |
137 | // operations on the same MemTable (unless this Memtable is immutable). | |
138 | size_t ApproximateMemoryUsage(); | |
139 | ||
140 | // As a cheap version of `ApproximateMemoryUsage()`, this function doens't | |
141 | // require external synchronization. The value may be less accurate though | |
142 | size_t ApproximateMemoryUsageFast() const { | |
143 | return approximate_memory_usage_.load(std::memory_order_relaxed); | |
144 | } | |
145 | ||
146 | // This method heuristically determines if the memtable should continue to | |
147 | // host more data. | |
148 | bool ShouldScheduleFlush() const { | |
149 | return flush_state_.load(std::memory_order_relaxed) == FLUSH_REQUESTED; | |
150 | } | |
151 | ||
152 | // Returns true if a flush should be scheduled and the caller should | |
153 | // be the one to schedule it | |
154 | bool MarkFlushScheduled() { | |
155 | auto before = FLUSH_REQUESTED; | |
156 | return flush_state_.compare_exchange_strong(before, FLUSH_SCHEDULED, | |
157 | std::memory_order_relaxed, | |
158 | std::memory_order_relaxed); | |
159 | } | |
160 | ||
161 | // Return an iterator that yields the contents of the memtable. | |
162 | // | |
163 | // The caller must ensure that the underlying MemTable remains live | |
164 | // while the returned iterator is live. The keys returned by this | |
165 | // iterator are internal keys encoded by AppendInternalKey in the | |
166 | // db/dbformat.{h,cc} module. | |
167 | // | |
168 | // By default, it returns an iterator for prefix seek if prefix_extractor | |
169 | // is configured in Options. | |
170 | // arena: If not null, the arena needs to be used to allocate the Iterator. | |
171 | // Calling ~Iterator of the iterator will destroy all the states but | |
172 | // those allocated in arena. | |
173 | InternalIterator* NewIterator(const ReadOptions& read_options, Arena* arena); | |
174 | ||
175 | FragmentedRangeTombstoneIterator* NewRangeTombstoneIterator( | |
176 | const ReadOptions& read_options, SequenceNumber read_seq); | |
177 | ||
178 | // Add an entry into memtable that maps key to value at the | |
179 | // specified sequence number and with the specified type. | |
180 | // Typically value will be empty if type==kTypeDeletion. | |
181 | // | |
182 | // REQUIRES: if allow_concurrent = false, external synchronization to prevent | |
183 | // simultaneous operations on the same MemTable. | |
184 | // | |
185 | // Returns false if MemTableRepFactory::CanHandleDuplicatedKey() is true and | |
186 | // the <key, seq> already exists. | |
187 | bool Add(SequenceNumber seq, ValueType type, const Slice& key, | |
188 | const Slice& value, bool allow_concurrent = false, | |
189 | MemTablePostProcessInfo* post_process_info = nullptr, | |
190 | void** hint = nullptr); | |
191 | ||
192 | // Used to Get value associated with key or Get Merge Operands associated | |
193 | // with key. | |
194 | // If do_merge = true the default behavior which is Get value for key is | |
195 | // executed. Expected behavior is described right below. | |
196 | // If memtable contains a value for key, store it in *value and return true. | |
197 | // If memtable contains a deletion for key, store a NotFound() error | |
198 | // in *status and return true. | |
199 | // If memtable contains Merge operation as the most recent entry for a key, | |
200 | // and the merge process does not stop (not reaching a value or delete), | |
201 | // prepend the current merge operand to *operands. | |
202 | // store MergeInProgress in s, and return false. | |
203 | // Else, return false. | |
204 | // If any operation was found, its most recent sequence number | |
205 | // will be stored in *seq on success (regardless of whether true/false is | |
206 | // returned). Otherwise, *seq will be set to kMaxSequenceNumber. | |
207 | // On success, *s may be set to OK, NotFound, or MergeInProgress. Any other | |
208 | // status returned indicates a corruption or other unexpected error. | |
209 | // If do_merge = false then any Merge Operands encountered for key are simply | |
210 | // stored in merge_context.operands_list and never actually merged to get a | |
211 | // final value. The raw Merge Operands are eventually returned to the user. | |
212 | bool Get(const LookupKey& key, std::string* value, Status* s, | |
213 | MergeContext* merge_context, | |
214 | SequenceNumber* max_covering_tombstone_seq, SequenceNumber* seq, | |
215 | const ReadOptions& read_opts, ReadCallback* callback = nullptr, | |
216 | bool* is_blob_index = nullptr, bool do_merge = true) { | |
217 | return Get(key, value, /*timestamp=*/nullptr, s, merge_context, | |
218 | max_covering_tombstone_seq, seq, read_opts, callback, | |
219 | is_blob_index, do_merge); | |
220 | } | |
221 | ||
222 | bool Get(const LookupKey& key, std::string* value, std::string* timestamp, | |
223 | Status* s, MergeContext* merge_context, | |
224 | SequenceNumber* max_covering_tombstone_seq, SequenceNumber* seq, | |
225 | const ReadOptions& read_opts, ReadCallback* callback = nullptr, | |
226 | bool* is_blob_index = nullptr, bool do_merge = true); | |
227 | ||
228 | bool Get(const LookupKey& key, std::string* value, std::string* timestamp, | |
229 | Status* s, MergeContext* merge_context, | |
230 | SequenceNumber* max_covering_tombstone_seq, | |
231 | const ReadOptions& read_opts, ReadCallback* callback = nullptr, | |
232 | bool* is_blob_index = nullptr, bool do_merge = true) { | |
233 | SequenceNumber seq; | |
234 | return Get(key, value, timestamp, s, merge_context, | |
235 | max_covering_tombstone_seq, &seq, read_opts, callback, | |
236 | is_blob_index, do_merge); | |
237 | } | |
238 | ||
239 | void MultiGet(const ReadOptions& read_options, MultiGetRange* range, | |
240 | ReadCallback* callback, bool* is_blob); | |
241 | ||
242 | // Attempts to update the new_value inplace, else does normal Add | |
243 | // Pseudocode | |
244 | // if key exists in current memtable && prev_value is of type kTypeValue | |
245 | // if new sizeof(new_value) <= sizeof(prev_value) | |
246 | // update inplace | |
247 | // else add(key, new_value) | |
248 | // else add(key, new_value) | |
249 | // | |
250 | // REQUIRES: external synchronization to prevent simultaneous | |
251 | // operations on the same MemTable. | |
252 | void Update(SequenceNumber seq, | |
253 | const Slice& key, | |
254 | const Slice& value); | |
255 | ||
256 | // If prev_value for key exists, attempts to update it inplace. | |
257 | // else returns false | |
258 | // Pseudocode | |
259 | // if key exists in current memtable && prev_value is of type kTypeValue | |
260 | // new_value = delta(prev_value) | |
261 | // if sizeof(new_value) <= sizeof(prev_value) | |
262 | // update inplace | |
263 | // else add(key, new_value) | |
264 | // else return false | |
265 | // | |
266 | // REQUIRES: external synchronization to prevent simultaneous | |
267 | // operations on the same MemTable. | |
268 | bool UpdateCallback(SequenceNumber seq, | |
269 | const Slice& key, | |
270 | const Slice& delta); | |
271 | ||
272 | // Returns the number of successive merge entries starting from the newest | |
273 | // entry for the key up to the last non-merge entry or last entry for the | |
274 | // key in the memtable. | |
275 | size_t CountSuccessiveMergeEntries(const LookupKey& key); | |
276 | ||
277 | // Update counters and flush status after inserting a whole write batch | |
278 | // Used in concurrent memtable inserts. | |
279 | void BatchPostProcess(const MemTablePostProcessInfo& update_counters) { | |
280 | num_entries_.fetch_add(update_counters.num_entries, | |
281 | std::memory_order_relaxed); | |
282 | data_size_.fetch_add(update_counters.data_size, std::memory_order_relaxed); | |
283 | if (update_counters.num_deletes != 0) { | |
284 | num_deletes_.fetch_add(update_counters.num_deletes, | |
285 | std::memory_order_relaxed); | |
286 | } | |
287 | UpdateFlushState(); | |
288 | } | |
289 | ||
290 | // Get total number of entries in the mem table. | |
291 | // REQUIRES: external synchronization to prevent simultaneous | |
292 | // operations on the same MemTable (unless this Memtable is immutable). | |
293 | uint64_t num_entries() const { | |
294 | return num_entries_.load(std::memory_order_relaxed); | |
295 | } | |
296 | ||
297 | // Get total number of deletes in the mem table. | |
298 | // REQUIRES: external synchronization to prevent simultaneous | |
299 | // operations on the same MemTable (unless this Memtable is immutable). | |
300 | uint64_t num_deletes() const { | |
301 | return num_deletes_.load(std::memory_order_relaxed); | |
302 | } | |
303 | ||
304 | uint64_t get_data_size() const { | |
305 | return data_size_.load(std::memory_order_relaxed); | |
306 | } | |
307 | ||
308 | // Dynamically change the memtable's capacity. If set below the current usage, | |
309 | // the next key added will trigger a flush. Can only increase size when | |
310 | // memtable prefix bloom is disabled, since we can't easily allocate more | |
311 | // space. | |
312 | void UpdateWriteBufferSize(size_t new_write_buffer_size) { | |
313 | if (bloom_filter_ == nullptr || | |
314 | new_write_buffer_size < write_buffer_size_) { | |
315 | write_buffer_size_.store(new_write_buffer_size, | |
316 | std::memory_order_relaxed); | |
317 | } | |
318 | } | |
319 | ||
320 | // Returns the edits area that is needed for flushing the memtable | |
321 | VersionEdit* GetEdits() { return &edit_; } | |
322 | ||
323 | // Returns if there is no entry inserted to the mem table. | |
324 | // REQUIRES: external synchronization to prevent simultaneous | |
325 | // operations on the same MemTable (unless this Memtable is immutable). | |
326 | bool IsEmpty() const { return first_seqno_ == 0; } | |
327 | ||
328 | // Returns the sequence number of the first element that was inserted | |
329 | // into the memtable. | |
330 | // REQUIRES: external synchronization to prevent simultaneous | |
331 | // operations on the same MemTable (unless this Memtable is immutable). | |
332 | SequenceNumber GetFirstSequenceNumber() { | |
333 | return first_seqno_.load(std::memory_order_relaxed); | |
334 | } | |
335 | ||
336 | // Returns the sequence number that is guaranteed to be smaller than or equal | |
337 | // to the sequence number of any key that could be inserted into this | |
338 | // memtable. It can then be assumed that any write with a larger(or equal) | |
339 | // sequence number will be present in this memtable or a later memtable. | |
340 | // | |
341 | // If the earliest sequence number could not be determined, | |
342 | // kMaxSequenceNumber will be returned. | |
343 | SequenceNumber GetEarliestSequenceNumber() { | |
344 | return earliest_seqno_.load(std::memory_order_relaxed); | |
345 | } | |
346 | ||
347 | // DB's latest sequence ID when the memtable is created. This number | |
348 | // may be updated to a more recent one before any key is inserted. | |
349 | SequenceNumber GetCreationSeq() const { return creation_seq_; } | |
350 | ||
351 | void SetCreationSeq(SequenceNumber sn) { creation_seq_ = sn; } | |
352 | ||
353 | // Returns the next active logfile number when this memtable is about to | |
354 | // be flushed to storage | |
355 | // REQUIRES: external synchronization to prevent simultaneous | |
356 | // operations on the same MemTable. | |
357 | uint64_t GetNextLogNumber() { return mem_next_logfile_number_; } | |
358 | ||
359 | // Sets the next active logfile number when this memtable is about to | |
360 | // be flushed to storage | |
361 | // REQUIRES: external synchronization to prevent simultaneous | |
362 | // operations on the same MemTable. | |
363 | void SetNextLogNumber(uint64_t num) { mem_next_logfile_number_ = num; } | |
364 | ||
365 | // if this memtable contains data from a committed | |
366 | // two phase transaction we must take note of the | |
367 | // log which contains that data so we can know | |
368 | // when to relese that log | |
369 | void RefLogContainingPrepSection(uint64_t log); | |
370 | uint64_t GetMinLogContainingPrepSection(); | |
371 | ||
372 | // Notify the underlying storage that no more items will be added. | |
373 | // REQUIRES: external synchronization to prevent simultaneous | |
374 | // operations on the same MemTable. | |
375 | // After MarkImmutable() is called, you should not attempt to | |
376 | // write anything to this MemTable(). (Ie. do not call Add() or Update()). | |
377 | void MarkImmutable() { | |
378 | table_->MarkReadOnly(); | |
379 | mem_tracker_.DoneAllocating(); | |
380 | } | |
381 | ||
382 | // Notify the underlying storage that all data it contained has been | |
383 | // persisted. | |
384 | // REQUIRES: external synchronization to prevent simultaneous | |
385 | // operations on the same MemTable. | |
386 | void MarkFlushed() { | |
387 | table_->MarkFlushed(); | |
388 | } | |
389 | ||
390 | // return true if the current MemTableRep supports merge operator. | |
391 | bool IsMergeOperatorSupported() const { | |
392 | return table_->IsMergeOperatorSupported(); | |
393 | } | |
394 | ||
395 | // return true if the current MemTableRep supports snapshots. | |
396 | // inplace update prevents snapshots, | |
397 | bool IsSnapshotSupported() const { | |
398 | return table_->IsSnapshotSupported() && !moptions_.inplace_update_support; | |
399 | } | |
400 | ||
401 | struct MemTableStats { | |
402 | uint64_t size; | |
403 | uint64_t count; | |
404 | }; | |
405 | ||
406 | MemTableStats ApproximateStats(const Slice& start_ikey, | |
407 | const Slice& end_ikey); | |
408 | ||
409 | // Get the lock associated for the key | |
410 | port::RWMutex* GetLock(const Slice& key); | |
411 | ||
412 | const InternalKeyComparator& GetInternalKeyComparator() const { | |
413 | return comparator_.comparator; | |
414 | } | |
415 | ||
416 | const ImmutableMemTableOptions* GetImmutableMemTableOptions() const { | |
417 | return &moptions_; | |
418 | } | |
419 | ||
420 | uint64_t ApproximateOldestKeyTime() const { | |
421 | return oldest_key_time_.load(std::memory_order_relaxed); | |
422 | } | |
423 | ||
424 | // REQUIRES: db_mutex held. | |
425 | void SetID(uint64_t id) { id_ = id; } | |
426 | ||
427 | uint64_t GetID() const { return id_; } | |
428 | ||
429 | void SetFlushCompleted(bool completed) { flush_completed_ = completed; } | |
430 | ||
431 | uint64_t GetFileNumber() const { return file_number_; } | |
432 | ||
433 | void SetFileNumber(uint64_t file_num) { file_number_ = file_num; } | |
434 | ||
435 | void SetFlushInProgress(bool in_progress) { | |
436 | flush_in_progress_ = in_progress; | |
437 | } | |
438 | ||
439 | #ifndef ROCKSDB_LITE | |
440 | void SetFlushJobInfo(std::unique_ptr<FlushJobInfo>&& info) { | |
441 | flush_job_info_ = std::move(info); | |
442 | } | |
443 | ||
444 | std::unique_ptr<FlushJobInfo> ReleaseFlushJobInfo() { | |
445 | return std::move(flush_job_info_); | |
446 | } | |
447 | #endif // !ROCKSDB_LITE | |
448 | ||
449 | private: | |
450 | enum FlushStateEnum { FLUSH_NOT_REQUESTED, FLUSH_REQUESTED, FLUSH_SCHEDULED }; | |
451 | ||
452 | friend class MemTableIterator; | |
453 | friend class MemTableBackwardIterator; | |
454 | friend class MemTableList; | |
455 | ||
456 | KeyComparator comparator_; | |
457 | const ImmutableMemTableOptions moptions_; | |
458 | int refs_; | |
459 | const size_t kArenaBlockSize; | |
460 | AllocTracker mem_tracker_; | |
461 | ConcurrentArena arena_; | |
462 | std::unique_ptr<MemTableRep> table_; | |
463 | std::unique_ptr<MemTableRep> range_del_table_; | |
464 | std::atomic_bool is_range_del_table_empty_; | |
465 | ||
466 | // Total data size of all data inserted | |
467 | std::atomic<uint64_t> data_size_; | |
468 | std::atomic<uint64_t> num_entries_; | |
469 | std::atomic<uint64_t> num_deletes_; | |
470 | ||
471 | // Dynamically changeable memtable option | |
472 | std::atomic<size_t> write_buffer_size_; | |
473 | ||
474 | // These are used to manage memtable flushes to storage | |
475 | bool flush_in_progress_; // started the flush | |
476 | bool flush_completed_; // finished the flush | |
477 | uint64_t file_number_; // filled up after flush is complete | |
478 | ||
479 | // The updates to be applied to the transaction log when this | |
480 | // memtable is flushed to storage. | |
481 | VersionEdit edit_; | |
482 | ||
483 | // The sequence number of the kv that was inserted first | |
484 | std::atomic<SequenceNumber> first_seqno_; | |
485 | ||
486 | // The db sequence number at the time of creation or kMaxSequenceNumber | |
487 | // if not set. | |
488 | std::atomic<SequenceNumber> earliest_seqno_; | |
489 | ||
490 | SequenceNumber creation_seq_; | |
491 | ||
492 | // The log files earlier than this number can be deleted. | |
493 | uint64_t mem_next_logfile_number_; | |
494 | ||
495 | // the earliest log containing a prepared section | |
496 | // which has been inserted into this memtable. | |
497 | std::atomic<uint64_t> min_prep_log_referenced_; | |
498 | ||
499 | // rw locks for inplace updates | |
500 | std::vector<port::RWMutex> locks_; | |
501 | ||
502 | const SliceTransform* const prefix_extractor_; | |
503 | std::unique_ptr<DynamicBloom> bloom_filter_; | |
504 | ||
505 | std::atomic<FlushStateEnum> flush_state_; | |
506 | ||
507 | Env* env_; | |
508 | ||
509 | // Extract sequential insert prefixes. | |
510 | const SliceTransform* insert_with_hint_prefix_extractor_; | |
511 | ||
512 | // Insert hints for each prefix. | |
513 | std::unordered_map<Slice, void*, SliceHasher> insert_hints_; | |
514 | ||
515 | // Timestamp of oldest key | |
516 | std::atomic<uint64_t> oldest_key_time_; | |
517 | ||
518 | // Memtable id to track flush. | |
519 | uint64_t id_ = 0; | |
520 | ||
521 | // Sequence number of the atomic flush that is responsible for this memtable. | |
522 | // The sequence number of atomic flush is a seq, such that no writes with | |
523 | // sequence numbers greater than or equal to seq are flushed, while all | |
524 | // writes with sequence number smaller than seq are flushed. | |
525 | SequenceNumber atomic_flush_seqno_; | |
526 | ||
527 | // keep track of memory usage in table_, arena_, and range_del_table_. | |
528 | // Gets refrshed inside `ApproximateMemoryUsage()` or `ShouldFlushNow` | |
529 | std::atomic<uint64_t> approximate_memory_usage_; | |
530 | ||
531 | #ifndef ROCKSDB_LITE | |
532 | // Flush job info of the current memtable. | |
533 | std::unique_ptr<FlushJobInfo> flush_job_info_; | |
534 | #endif // !ROCKSDB_LITE | |
535 | ||
536 | // Returns a heuristic flush decision | |
537 | bool ShouldFlushNow(); | |
538 | ||
539 | // Updates flush_state_ using ShouldFlushNow() | |
540 | void UpdateFlushState(); | |
541 | ||
542 | void UpdateOldestKeyTime(); | |
543 | ||
544 | void GetFromTable(const LookupKey& key, | |
545 | SequenceNumber max_covering_tombstone_seq, bool do_merge, | |
546 | ReadCallback* callback, bool* is_blob_index, | |
547 | std::string* value, std::string* timestamp, Status* s, | |
548 | MergeContext* merge_context, SequenceNumber* seq, | |
549 | bool* found_final_value, bool* merge_in_progress); | |
550 | }; | |
551 | ||
552 | extern const char* EncodeKey(std::string* scratch, const Slice& target); | |
553 | ||
554 | } // namespace ROCKSDB_NAMESPACE |