You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
rocksdb/db/version_set.h

573 lines
20 KiB

// 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.
//
// The representation of a DBImpl consists of a set of Versions. The
// newest version is called "current". Older versions may be kept
// around to provide a consistent view to live iterators.
//
// Each Version keeps track of a set of Table files per level. The
// entire set of versions is maintained in a VersionSet.
//
// Version,VersionSet are thread-compatible, but require external
// synchronization on all accesses.
#ifndef STORAGE_LEVELDB_DB_VERSION_SET_H_
#define STORAGE_LEVELDB_DB_VERSION_SET_H_
#include <map>
#include <memory>
#include <set>
#include <vector>
#include <deque>
#include "db/dbformat.h"
#include "db/version_edit.h"
#include "port/port.h"
#include "db/table_cache.h"
namespace leveldb {
namespace log { class Writer; }
class Compaction;
class Iterator;
class MemTable;
class TableBuilder;
class TableCache;
class Version;
class VersionSet;
class WritableFile;
// Return the smallest index i such that files[i]->largest >= key.
// Return files.size() if there is no such file.
// REQUIRES: "files" contains a sorted list of non-overlapping files.
extern int FindFile(const InternalKeyComparator& icmp,
const std::vector<FileMetaData*>& files,
const Slice& key);
// Returns true iff some file in "files" overlaps the user key range
// [*smallest,*largest].
// smallest==nullptr represents a key smaller than all keys in the DB.
// largest==nullptr represents a key largest than all keys in the DB.
// REQUIRES: If disjoint_sorted_files, files[] contains disjoint ranges
// in sorted order.
extern bool SomeFileOverlapsRange(
const InternalKeyComparator& icmp,
bool disjoint_sorted_files,
const std::vector<FileMetaData*>& files,
const Slice* smallest_user_key,
const Slice* largest_user_key);
class Version {
public:
// Append to *iters a sequence of iterators that will
// yield the contents of this Version when merged together.
// REQUIRES: This version has been saved (see VersionSet::SaveTo)
void AddIterators(const ReadOptions&, std::vector<Iterator*>* iters);
// Lookup the value for key. If found, store it in *val and
// return OK. Else return a non-OK status. Fills *stats.
// REQUIRES: lock is not held
struct GetStats {
FileMetaData* seek_file;
int seek_file_level;
};
Status Get(const ReadOptions&, const LookupKey& key, std::string* val,
GetStats* stats);
// Adds "stats" into the current state. Returns true if a new
// compaction may need to be triggered, false otherwise.
// REQUIRES: lock is held
bool UpdateStats(const GetStats& stats);
// Reference count management (so Versions do not disappear out from
// under live iterators)
void Ref();
void Unref();
void GetOverlappingInputs(
int level,
const InternalKey* begin, // nullptr means before all keys
const InternalKey* end, // nullptr means after all keys
std::vector<FileMetaData*>* inputs,
int hint_index = -1, // index of overlap file
int* file_index = nullptr); // return index of overlap file
void GetOverlappingInputsBinarySearch(
int level,
const Slice& begin, // nullptr means before all keys
const Slice& end, // nullptr means after all keys
std::vector<FileMetaData*>* inputs,
int hint_index, // index of overlap file
int* file_index); // return index of overlap file
void ExtendOverlappingInputs(
int level,
const Slice& begin, // nullptr means before all keys
const Slice& end, // nullptr means after all keys
std::vector<FileMetaData*>* inputs,
int index); // start extending from this index
// Returns true iff some file in the specified level overlaps
// some part of [*smallest_user_key,*largest_user_key].
// smallest_user_key==NULL represents a key smaller than all keys in the DB.
// largest_user_key==NULL represents a key largest than all keys in the DB.
bool OverlapInLevel(int level,
const Slice* smallest_user_key,
const Slice* largest_user_key);
// Return the level at which we should place a new memtable compaction
// result that covers the range [smallest_user_key,largest_user_key].
int PickLevelForMemTableOutput(const Slice& smallest_user_key,
const Slice& largest_user_key);
int NumFiles(int level) const { return files_[level].size(); }
// Return a human readable string that describes this version's contents.
std::string DebugString(bool hex = false) const;
// Returns the version nuber of this version
uint64_t GetVersionNumber() {
return version_number_;
}
private:
friend class Compaction;
friend class VersionSet;
class LevelFileNumIterator;
Iterator* NewConcatenatingIterator(const ReadOptions&, int level) const;
VersionSet* vset_; // VersionSet to which this Version belongs
Version* next_; // Next version in linked list
Version* prev_; // Previous version in linked list
int refs_; // Number of live refs to this version
// List of files per level, files in each level are arranged
// in increasing order of keys
std::vector<FileMetaData*>* files_;
// A list for the same set of files that are stored in files_,
// but files in each level are now sorted based on file
// size. The file with the largest size is at the front.
// This vector stores the index of the file from files_.
std::vector< std::vector<int> > files_by_size_;
// An index into files_by_size_ that specifies the first
// file that is not yet compacted
std::vector<int> next_file_to_compact_by_size_;
// Only the first few entries of files_by_size_ are sorted.
// There is no need to sort all the files because it is likely
// that on a running system, we need to look at only the first
// few largest files because a new version is created every few
// seconds/minutes (because of concurrent compactions).
static const int number_of_files_to_sort_ = 50;
// Next file to compact based on seek stats.
FileMetaData* file_to_compact_;
int file_to_compact_level_;
// Level that should be compacted next and its compaction score.
// Score < 1 means compaction is not strictly needed. These fields
// are initialized by Finalize().
// The most critical level to be compacted is listed first
// These are used to pick the best compaction level
std::vector<double> compaction_score_;
std::vector<int> compaction_level_;
double max_compaction_score_; // max score in l1 to ln-1
int max_compaction_score_level_; // level on which max score occurs
// The offset in the manifest file where this version is stored.
uint64_t offset_manifest_file_;
// A version number that uniquely represents this version. This is
// used for debugging and logging purposes only.
uint64_t version_number_;
explicit Version(VersionSet* vset, uint64_t version_number = 0);
~Version();
// re-initializes the index that is used to offset into files_by_size_
// to find the next compaction candidate file.
void ResetNextCompactionIndex(int level) {
next_file_to_compact_by_size_[level] = 0;
}
// No copying allowed
Version(const Version&);
void operator=(const Version&);
};
class VersionSet {
public:
VersionSet(const std::string& dbname,
const Options* options,
TableCache* table_cache,
const InternalKeyComparator*);
~VersionSet();
// Apply *edit to the current version to form a new descriptor that
// is both saved to persistent state and installed as the new
// current version. Will release *mu while actually writing to the file.
// REQUIRES: *mu is held on entry.
// REQUIRES: no other thread concurrently calls LogAndApply()
Status LogAndApply(VersionEdit* edit, port::Mutex* mu,
bool new_descriptor_log = false);
// Recover the last saved descriptor from persistent storage.
Status Recover();
// Try to reduce the number of levels. This call is valid when
// only one level from the new max level to the old
// max level containing files.
// For example, a db currently has 7 levels [0-6], and a call to
// to reduce to 5 [0-4] can only be executed when only one level
// among [4-6] contains files.
Status ReduceNumberOfLevels(int new_levels, port::Mutex* mu);
// Return the current version.
Version* current() const { return current_; }
// Return the current manifest file number
uint64_t ManifestFileNumber() const { return manifest_file_number_; }
// Allocate and return a new file number
uint64_t NewFileNumber() { return next_file_number_++; }
// Arrange to reuse "file_number" unless a newer file number has
// already been allocated.
// REQUIRES: "file_number" was returned by a call to NewFileNumber().
void ReuseFileNumber(uint64_t file_number) {
if (next_file_number_ == file_number + 1) {
next_file_number_ = file_number;
}
}
// Return the number of Table files at the specified level.
int NumLevelFiles(int level) const;
// Return the combined file size of all files at the specified level.
int64_t NumLevelBytes(int level) const;
// Return the last sequence number.
uint64_t LastSequence() const { return last_sequence_; }
// Set the last sequence number to s.
void SetLastSequence(uint64_t s) {
assert(s >= last_sequence_);
last_sequence_ = s;
}
// Mark the specified file number as used.
void MarkFileNumberUsed(uint64_t number);
// Return the current log file number.
uint64_t LogNumber() const { return log_number_; }
// Return the log file number for the log file that is currently
// being compacted, or zero if there is no such log file.
uint64_t PrevLogNumber() const { return prev_log_number_; }
int NumberLevels() const { return num_levels_; }
// Pick level and inputs for a new compaction.
// Returns nullptr if there is no compaction to be done.
// Otherwise returns a pointer to a heap-allocated object that
// describes the compaction. Caller should delete the result.
Compaction* PickCompaction();
// Return a compaction object for compacting the range [begin,end] in
// the specified level. Returns nullptr if there is nothing in that
// level that overlaps the specified range. Caller should delete
// the result.
Compaction* CompactRange(
int level,
const InternalKey* begin,
const InternalKey* end);
// Return the maximum overlapping data (in bytes) at next level for any
// file at a level >= 1.
int64_t MaxNextLevelOverlappingBytes();
// Create an iterator that reads over the compaction inputs for "*c".
// The caller should delete the iterator when no longer needed.
Iterator* MakeInputIterator(Compaction* c);
// Returns true iff some level needs a compaction because it has
// exceeded its target size.
bool NeedsSizeCompaction() const {
for (int i = 0; i < NumberLevels()-1; i++) {
if (current_->compaction_score_[i] >= 1) {
return true;
}
}
return false;
}
// Returns true iff some level needs a compaction.
bool NeedsCompaction() const {
return ((current_->file_to_compact_ != nullptr) ||
NeedsSizeCompaction());
}
// Returns the maxmimum compaction score for levels 1 to max
double MaxCompactionScore() const {
return current_->max_compaction_score_;
}
// See field declaration
int MaxCompactionScoreLevel() const {
return current_->max_compaction_score_level_;
}
// Add all files listed in any live version to *live.
// May also mutate some internal state.
void AddLiveFiles(std::set<uint64_t>* live);
// Add all files listed in the current version to *live.
void AddLiveFilesCurrentVersion(std::set<uint64_t>* live);
// Return the approximate offset in the database of the data for
// "key" as of version "v".
uint64_t ApproximateOffsetOf(Version* v, const InternalKey& key);
// Return a human-readable short (single-line) summary of the number
// of files per level. Uses *scratch as backing store.
struct LevelSummaryStorage {
char buffer[100];
};
const char* LevelSummary(LevelSummaryStorage* scratch) const;
Utility to dump manifest contents. Summary: ./manifest_dump --file=/tmp/dbbench/MANIFEST-000002 Output looks like manifest_file_number 30 next_file_number 31 last_sequence 388082 log_number 28 prev_log_number 0 --- level 0 --- --- level 1 --- --- level 2 --- 5:3244155['0000000000000000' @ 1 : 1 .. '0000000000028220' @ 28221 : 1] 7:3244177['0000000000028221' @ 28222 : 1 .. '0000000000056441' @ 56442 : 1] 9:3244156['0000000000056442' @ 56443 : 1 .. '0000000000084662' @ 84663 : 1] 11:3244178['0000000000084663' @ 84664 : 1 .. '0000000000112883' @ 112884 : 1] 13:3244158['0000000000112884' @ 112885 : 1 .. '0000000000141104' @ 141105 : 1] 15:3244176['0000000000141105' @ 141106 : 1 .. '0000000000169325' @ 169326 : 1] 17:3244156['0000000000169326' @ 169327 : 1 .. '0000000000197546' @ 197547 : 1] 19:3244178['0000000000197547' @ 197548 : 1 .. '0000000000225767' @ 225768 : 1] 21:3244155['0000000000225768' @ 225769 : 1 .. '0000000000253988' @ 253989 : 1] 23:3244179['0000000000253989' @ 253990 : 1 .. '0000000000282209' @ 282210 : 1] 25:3244157['0000000000282210' @ 282211 : 1 .. '0000000000310430' @ 310431 : 1] 27:3244176['0000000000310431' @ 310432 : 1 .. '0000000000338651' @ 338652 : 1] 29:3244156['0000000000338652' @ 338653 : 1 .. '0000000000366872' @ 366873 : 1] --- level 3 --- --- level 4 --- --- level 5 --- --- level 6 --- Test Plan: run on test directory created by dbbench Reviewers: heyongqiang Reviewed By: heyongqiang CC: hustliubo Differential Revision: https://reviews.facebook.net/D4743
13 years ago
// printf contents (for debugging)
Status DumpManifest(Options& options, std::string& manifestFileName,
bool verbose, bool hex = false);
Utility to dump manifest contents. Summary: ./manifest_dump --file=/tmp/dbbench/MANIFEST-000002 Output looks like manifest_file_number 30 next_file_number 31 last_sequence 388082 log_number 28 prev_log_number 0 --- level 0 --- --- level 1 --- --- level 2 --- 5:3244155['0000000000000000' @ 1 : 1 .. '0000000000028220' @ 28221 : 1] 7:3244177['0000000000028221' @ 28222 : 1 .. '0000000000056441' @ 56442 : 1] 9:3244156['0000000000056442' @ 56443 : 1 .. '0000000000084662' @ 84663 : 1] 11:3244178['0000000000084663' @ 84664 : 1 .. '0000000000112883' @ 112884 : 1] 13:3244158['0000000000112884' @ 112885 : 1 .. '0000000000141104' @ 141105 : 1] 15:3244176['0000000000141105' @ 141106 : 1 .. '0000000000169325' @ 169326 : 1] 17:3244156['0000000000169326' @ 169327 : 1 .. '0000000000197546' @ 197547 : 1] 19:3244178['0000000000197547' @ 197548 : 1 .. '0000000000225767' @ 225768 : 1] 21:3244155['0000000000225768' @ 225769 : 1 .. '0000000000253988' @ 253989 : 1] 23:3244179['0000000000253989' @ 253990 : 1 .. '0000000000282209' @ 282210 : 1] 25:3244157['0000000000282210' @ 282211 : 1 .. '0000000000310430' @ 310431 : 1] 27:3244176['0000000000310431' @ 310432 : 1 .. '0000000000338651' @ 338652 : 1] 29:3244156['0000000000338652' @ 338653 : 1 .. '0000000000366872' @ 366873 : 1] --- level 3 --- --- level 4 --- --- level 5 --- --- level 6 --- Test Plan: run on test directory created by dbbench Reviewers: heyongqiang Reviewed By: heyongqiang CC: hustliubo Differential Revision: https://reviews.facebook.net/D4743
13 years ago
// Return a human-readable short (single-line) summary of the data size
// of files per level. Uses *scratch as backing store.
const char* LevelDataSizeSummary(LevelSummaryStorage* scratch) const;
// Return the size of the current manifest file
const uint64_t ManifestFileSize() { return current_->offset_manifest_file_; }
// For the specfied level, pick a compaction.
// Returns nullptr if there is no compaction to be done.
// If level is 0 and there is already a compaction on that level, this
// function will return nullptr.
Compaction* PickCompactionBySize(int level, double score);
// Free up the files that were participated in a compaction
void ReleaseCompactionFiles(Compaction* c, Status status);
// verify that the files that we started with for a compaction
// still exist in the current version and in the same original level.
// This ensures that a concurrent compaction did not erroneously
// pick the same files to compact.
bool VerifyCompactionFileConsistency(Compaction* c);
// used to sort files by size
typedef struct fsize {
int index;
FileMetaData* file;
} Fsize;
// Sort all files for this version based on their file size and
// record results in files_by_size_. The largest files are listed first.
void UpdateFilesBySize(Version *v);
// Get the max file size in a given level.
uint64_t MaxFileSizeForLevel(int level);
private:
class Builder;
struct ManifestWriter;
friend class Compaction;
friend class Version;
void Init(int num_levels);
void Finalize(Version* v);
void GetRange(const std::vector<FileMetaData*>& inputs,
InternalKey* smallest,
InternalKey* largest);
void GetRange2(const std::vector<FileMetaData*>& inputs1,
const std::vector<FileMetaData*>& inputs2,
InternalKey* smallest,
InternalKey* largest);
void SetupOtherInputs(Compaction* c);
// Save current contents to *log
Status WriteSnapshot(log::Writer* log);
void AppendVersion(Version* v);
bool ManifestContains(const std::string& record) const;
double MaxBytesForLevel(int level);
int64_t ExpandedCompactionByteSizeLimit(int level);
int64_t MaxGrandParentOverlapBytes(int level);
Env* const env_;
const std::string dbname_;
const Options* const options_;
TableCache* const table_cache_;
const InternalKeyComparator icmp_;
uint64_t next_file_number_;
uint64_t manifest_file_number_;
uint64_t last_sequence_;
uint64_t log_number_;
uint64_t prev_log_number_; // 0 or backing store for memtable being compacted
int num_levels_;
// Opened lazily
unique_ptr<log::Writer> descriptor_log_;
Version dummy_versions_; // Head of circular doubly-linked list of versions.
Version* current_; // == dummy_versions_.prev_
// Per-level key at which the next compaction at that level should start.
// Either an empty string, or a valid InternalKey.
std::string* compact_pointer_;
// Per-level target file size.
uint64_t* max_file_size_;
// Per-level max bytes
uint64_t* level_max_bytes_;
// record all the ongoing compactions for all levels
std::vector<std::set<Compaction*> > compactions_in_progress_;
// generates a increasing version number for every new version
uint64_t current_version_number_;
// Queue of writers to the manifest file
std::deque<ManifestWriter*> manifest_writers_;
// Store the manifest file size when it is checked.
// Save us the cost of checking file size twice in LogAndApply
uint64_t last_observed_manifest_size_;
// No copying allowed
VersionSet(const VersionSet&);
void operator=(const VersionSet&);
// Return the total amount of data that is undergoing
// compactions at this level
uint64_t SizeBeingCompacted(int level);
// Returns true if any one of the parent files are being compacted
bool ParentRangeInCompaction(const InternalKey* smallest,
const InternalKey* largest, int level, int* index);
// Returns true if any one of the specified files are being compacted
bool FilesInCompaction(std::vector<FileMetaData*>& files);
void LogAndApplyHelper(Builder*b, Version* v,
VersionEdit* edit, port::Mutex* mu);
};
// A Compaction encapsulates information about a compaction.
class Compaction {
public:
~Compaction();
// Return the level that is being compacted. Inputs from "level"
// and "level+1" will be merged to produce a set of "level+1" files.
int level() const { return level_; }
// Return the object that holds the edits to the descriptor done
// by this compaction.
VersionEdit* edit() { return edit_; }
// "which" must be either 0 or 1
int num_input_files(int which) const { return inputs_[which].size(); }
// Return the ith input file at "level()+which" ("which" must be 0 or 1).
FileMetaData* input(int which, int i) const { return inputs_[which][i]; }
// Maximum size of files to build during this compaction.
uint64_t MaxOutputFileSize() const { return max_output_file_size_; }
// Is this a trivial compaction that can be implemented by just
// moving a single input file to the next level (no merging or splitting)
bool IsTrivialMove() const;
// Add all inputs to this compaction as delete operations to *edit.
void AddInputDeletions(VersionEdit* edit);
// Returns true if the information we have available guarantees that
// the compaction is producing data in "level+1" for which no data exists
// in levels greater than "level+1".
bool IsBaseLevelForKey(const Slice& user_key);
// Returns true iff we should stop building the current output
// before processing "internal_key".
bool ShouldStopBefore(const Slice& internal_key);
// Release the input version for the compaction, once the compaction
// is successful.
void ReleaseInputs();
void Summary(char* output, int len);
// Return the score that was used to pick this compaction run.
double score() const { return score_; }
private:
friend class Version;
friend class VersionSet;
explicit Compaction(int level, uint64_t target_file_size,
uint64_t max_grandparent_overlap_bytes, int number_levels,
bool seek_compaction = false);
int level_;
uint64_t max_output_file_size_;
int64_t maxGrandParentOverlapBytes_;
Version* input_version_;
VersionEdit* edit_;
int number_levels_;
bool seek_compaction_;
// Each compaction reads inputs from "level_" and "level_+1"
std::vector<FileMetaData*> inputs_[2]; // The two sets of inputs
// State used to check for number of of overlapping grandparent files
// (parent == level_ + 1, grandparent == level_ + 2)
std::vector<FileMetaData*> grandparents_;
size_t grandparent_index_; // Index in grandparent_starts_
bool seen_key_; // Some output key has been seen
int64_t overlapped_bytes_; // Bytes of overlap between current output
// and grandparent files
int base_index_; // index of the file in files_[level_]
int parent_index_; // index of some file with same range in files_[level_+1]
double score_; // score that was used to pick this compaction.
// State for implementing IsBaseLevelForKey
// level_ptrs_ holds indices into input_version_->levels_: our state
// is that we are positioned at one of the file ranges for each
// higher level than the ones involved in this compaction (i.e. for
// all L >= level_ + 2).
size_t* level_ptrs_;
// mark (or clear) all files that are being compacted
void MarkFilesBeingCompacted(bool);
// In case of compaction error, reset the nextIndex that is used
// to pick up the next file to be compacted from files_by_size_
void ResetNextCompactionIndex();
};
} // namespace leveldb
#endif // STORAGE_LEVELDB_DB_VERSION_SET_H_