1// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
2// Use of this source code is governed by a BSD-style license that can be
3// found in the LICENSE file. See the AUTHORS file for names of contributors.
4
5#ifndef STORAGE_LEVELDB_DB_DB_IMPL_H_
6#define STORAGE_LEVELDB_DB_DB_IMPL_H_
7
8#include <atomic>
9#include <deque>
10#include <set>
11#include <string>
12
13#include "db/dbformat.h"
14#include "db/log_writer.h"
15#include "db/snapshot.h"
16#include "leveldb/db.h"
17#include "leveldb/env.h"
18#include "port/port.h"
19#include "port/thread_annotations.h"
20
21namespace leveldb {
22
23class MemTable;
24class TableCache;
25class Version;
26class VersionEdit;
27class VersionSet;
28
29class DBImpl : public DB {
30 public:
31 DBImpl(const Options& options, const std::string& dbname);
32
33 DBImpl(const DBImpl&) = delete;
34 DBImpl& operator=(const DBImpl&) = delete;
35
36 ~DBImpl() override;
37
38 // Implementations of the DB interface
39 Status Put(const WriteOptions&, const Slice& key,
40 const Slice& value) override;
41 Status Delete(const WriteOptions&, const Slice& key) override;
42 Status Write(const WriteOptions& options, WriteBatch* updates) override;
43 Status Get(const ReadOptions& options, const Slice& key,
44 std::string* value) override;
45 Iterator* NewIterator(const ReadOptions&) override;
46 const Snapshot* GetSnapshot() override;
47 void ReleaseSnapshot(const Snapshot* snapshot) override;
48 bool GetProperty(const Slice& property, std::string* value) override;
49 void GetApproximateSizes(const Range* range, int n, uint64_t* sizes) override;
50 void CompactRange(const Slice* begin, const Slice* end) override;
51
52 // Extra methods (for testing) that are not in the public DB interface
53
54 // Compact any files in the named level that overlap [*begin,*end]
55 void TEST_CompactRange(int level, const Slice* begin, const Slice* end);
56
57 // Force current memtable contents to be compacted.
58 Status TEST_CompactMemTable();
59
60 // Return an internal iterator over the current state of the database.
61 // The keys of this iterator are internal keys (see format.h).
62 // The returned iterator should be deleted when no longer needed.
63 Iterator* TEST_NewInternalIterator();
64
65 // Return the maximum overlapping data (in bytes) at next level for any
66 // file at a level >= 1.
67 int64_t TEST_MaxNextLevelOverlappingBytes();
68
69 // Record a sample of bytes read at the specified internal key.
70 // Samples are taken approximately once every config::kReadBytesPeriod
71 // bytes.
72 void RecordReadSample(Slice key);
73
74 private:
75 friend class DB;
76 struct CompactionState;
77 struct Writer;
78
79 // Information for a manual compaction
80 struct ManualCompaction {
81 int level;
82 bool done;
83 const InternalKey* begin; // null means beginning of key range
84 const InternalKey* end; // null means end of key range
85 InternalKey tmp_storage; // Used to keep track of compaction progress
86 };
87
88 // Per level compaction stats. stats_[level] stores the stats for
89 // compactions that produced data for the specified "level".
90 struct CompactionStats {
91 CompactionStats() : micros(0), bytes_read(0), bytes_written(0) {}
92
93 void Add(const CompactionStats& c) {
94 this->micros += c.micros;
95 this->bytes_read += c.bytes_read;
96 this->bytes_written += c.bytes_written;
97 }
98
99 int64_t micros;
100 int64_t bytes_read;
101 int64_t bytes_written;
102 };
103
104 Iterator* NewInternalIterator(const ReadOptions&,
105 SequenceNumber* latest_snapshot,
106 uint32_t* seed);
107
108 Status NewDB();
109
110 // Recover the descriptor from persistent storage. May do a significant
111 // amount of work to recover recently logged updates. Any changes to
112 // be made to the descriptor are added to *edit.
113 Status Recover(VersionEdit* edit, bool* save_manifest)
114 EXCLUSIVE_LOCKS_REQUIRED(mutex_);
115
116 void MaybeIgnoreError(Status* s) const;
117
118 // Delete any unneeded files and stale in-memory entries.
119 void RemoveObsoleteFiles() EXCLUSIVE_LOCKS_REQUIRED(mutex_);
120
121 // Compact the in-memory write buffer to disk. Switches to a new
122 // log-file/memtable and writes a new descriptor iff successful.
123 // Errors are recorded in bg_error_.
124 void CompactMemTable() EXCLUSIVE_LOCKS_REQUIRED(mutex_);
125
126 Status RecoverLogFile(uint64_t log_number, bool last_log, bool* save_manifest,
127 VersionEdit* edit, SequenceNumber* max_sequence)
128 EXCLUSIVE_LOCKS_REQUIRED(mutex_);
129
130 Status WriteLevel0Table(MemTable* mem, VersionEdit* edit, Version* base)
131 EXCLUSIVE_LOCKS_REQUIRED(mutex_);
132
133 Status MakeRoomForWrite(bool force /* compact even if there is room? */)
134 EXCLUSIVE_LOCKS_REQUIRED(mutex_);
135 WriteBatch* BuildBatchGroup(Writer** last_writer)
136 EXCLUSIVE_LOCKS_REQUIRED(mutex_);
137
138 void RecordBackgroundError(const Status& s);
139
140 void MaybeScheduleCompaction() EXCLUSIVE_LOCKS_REQUIRED(mutex_);
141 static void BGWork(void* db);
142 void BackgroundCall();
143 void BackgroundCompaction() EXCLUSIVE_LOCKS_REQUIRED(mutex_);
144 void CleanupCompaction(CompactionState* compact)
145 EXCLUSIVE_LOCKS_REQUIRED(mutex_);
146 Status DoCompactionWork(CompactionState* compact)
147 EXCLUSIVE_LOCKS_REQUIRED(mutex_);
148
149 Status OpenCompactionOutputFile(CompactionState* compact);
150 Status FinishCompactionOutputFile(CompactionState* compact, Iterator* input);
151 Status InstallCompactionResults(CompactionState* compact)
152 EXCLUSIVE_LOCKS_REQUIRED(mutex_);
153
154 const Comparator* user_comparator() const {
155 return internal_comparator_.user_comparator();
156 }
157
158 // Constant after construction
159 Env* const env_;
160 const InternalKeyComparator internal_comparator_;
161 const InternalFilterPolicy internal_filter_policy_;
162 const Options options_; // options_.comparator == &internal_comparator_
163 const bool owns_info_log_;
164 const bool owns_cache_;
165 const std::string dbname_;
166
167 // table_cache_ provides its own synchronization
168 TableCache* const table_cache_;
169
170 // Lock over the persistent DB state. Non-null iff successfully acquired.
171 FileLock* db_lock_;
172
173 // State below is protected by mutex_
174 port::Mutex mutex_;
175 std::atomic<bool> shutting_down_;
176 port::CondVar background_work_finished_signal_ GUARDED_BY(mutex_);
177 MemTable* mem_;
178 MemTable* imm_ GUARDED_BY(mutex_); // Memtable being compacted
179 std::atomic<bool> has_imm_; // So bg thread can detect non-null imm_
180 WritableFile* logfile_;
181 uint64_t logfile_number_ GUARDED_BY(mutex_);
182 log::Writer* log_;
183 uint32_t seed_ GUARDED_BY(mutex_); // For sampling.
184
185 // Queue of writers.
186 std::deque<Writer*> writers_ GUARDED_BY(mutex_);
187 WriteBatch* tmp_batch_ GUARDED_BY(mutex_);
188
189 SnapshotList snapshots_ GUARDED_BY(mutex_);
190
191 // Set of table files to protect from deletion because they are
192 // part of ongoing compactions.
193 std::set<uint64_t> pending_outputs_ GUARDED_BY(mutex_);
194
195 // Has a background compaction been scheduled or is running?
196 bool background_compaction_scheduled_ GUARDED_BY(mutex_);
197
198 ManualCompaction* manual_compaction_ GUARDED_BY(mutex_);
199
200 VersionSet* const versions_ GUARDED_BY(mutex_);
201
202 // Have we encountered a background error in paranoid mode?
203 Status bg_error_ GUARDED_BY(mutex_);
204
205 CompactionStats stats_[config::kNumLevels] GUARDED_BY(mutex_);
206};
207
208// Sanitize db options. The caller should delete result.info_log if
209// it is not equal to src.info_log.
210Options SanitizeOptions(const std::string& db,
211 const InternalKeyComparator* icmp,
212 const InternalFilterPolicy* ipolicy,
213 const Options& src);
214
215} // namespace leveldb
216
217#endif // STORAGE_LEVELDB_DB_DB_IMPL_H_
218