github.com/DxChainNetwork/dxc@v0.8.1-0.20220824085222-1162e304b6e7/ethdb/leveldb/leveldb.go (about)

     1  // Copyright 2018 The go-ethereum Authors
     2  // This file is part of the go-ethereum library.
     3  //
     4  // The go-ethereum library is free software: you can redistribute it and/or modify
     5  // it under the terms of the GNU Lesser General Public License as published by
     6  // the Free Software Foundation, either version 3 of the License, or
     7  // (at your option) any later version.
     8  //
     9  // The go-ethereum library is distributed in the hope that it will be useful,
    10  // but WITHOUT ANY WARRANTY; without even the implied warranty of
    11  // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
    12  // GNU Lesser General Public License for more details.
    13  //
    14  // You should have received a copy of the GNU Lesser General Public License
    15  // along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
    16  
    17  // +build !js
    18  
    19  // Package leveldb implements the key-value database layer based on LevelDB.
    20  package leveldb
    21  
    22  import (
    23  	"fmt"
    24  	"strconv"
    25  	"strings"
    26  	"sync"
    27  	"time"
    28  
    29  	"github.com/DxChainNetwork/dxc/common"
    30  	"github.com/DxChainNetwork/dxc/ethdb"
    31  	"github.com/DxChainNetwork/dxc/log"
    32  	"github.com/DxChainNetwork/dxc/metrics"
    33  	"github.com/syndtr/goleveldb/leveldb"
    34  	"github.com/syndtr/goleveldb/leveldb/errors"
    35  	"github.com/syndtr/goleveldb/leveldb/filter"
    36  	"github.com/syndtr/goleveldb/leveldb/opt"
    37  	"github.com/syndtr/goleveldb/leveldb/util"
    38  )
    39  
    40  const (
    41  	// degradationWarnInterval specifies how often warning should be printed if the
    42  	// leveldb database cannot keep up with requested writes.
    43  	degradationWarnInterval = time.Minute
    44  
    45  	// minCache is the minimum amount of memory in megabytes to allocate to leveldb
    46  	// read and write caching, split half and half.
    47  	minCache = 16
    48  
    49  	// minHandles is the minimum number of files handles to allocate to the open
    50  	// database files.
    51  	minHandles = 16
    52  
    53  	// metricsGatheringInterval specifies the interval to retrieve leveldb database
    54  	// compaction, io and pause stats to report to the user.
    55  	metricsGatheringInterval = 3 * time.Second
    56  )
    57  
    58  // Database is a persistent key-value store. Apart from basic data storage
    59  // functionality it also supports batch writes and iterating over the keyspace in
    60  // binary-alphabetical order.
    61  type Database struct {
    62  	fn string      // filename for reporting
    63  	db *leveldb.DB // LevelDB instance
    64  
    65  	compTimeMeter      metrics.Meter // Meter for measuring the total time spent in database compaction
    66  	compReadMeter      metrics.Meter // Meter for measuring the data read during compaction
    67  	compWriteMeter     metrics.Meter // Meter for measuring the data written during compaction
    68  	writeDelayNMeter   metrics.Meter // Meter for measuring the write delay number due to database compaction
    69  	writeDelayMeter    metrics.Meter // Meter for measuring the write delay duration due to database compaction
    70  	diskSizeGauge      metrics.Gauge // Gauge for tracking the size of all the levels in the database
    71  	diskReadMeter      metrics.Meter // Meter for measuring the effective amount of data read
    72  	diskWriteMeter     metrics.Meter // Meter for measuring the effective amount of data written
    73  	memCompGauge       metrics.Gauge // Gauge for tracking the number of memory compaction
    74  	level0CompGauge    metrics.Gauge // Gauge for tracking the number of table compaction in level0
    75  	nonlevel0CompGauge metrics.Gauge // Gauge for tracking the number of table compaction in non0 level
    76  	seekCompGauge      metrics.Gauge // Gauge for tracking the number of table compaction caused by read opt
    77  
    78  	quitLock sync.Mutex      // Mutex protecting the quit channel access
    79  	quitChan chan chan error // Quit channel to stop the metrics collection before closing the database
    80  
    81  	log log.Logger // Contextual logger tracking the database path
    82  }
    83  
    84  // New returns a wrapped LevelDB object. The namespace is the prefix that the
    85  // metrics reporting should use for surfacing internal stats.
    86  func New(file string, cache int, handles int, namespace string, readonly bool) (*Database, error) {
    87  	return NewCustom(file, namespace, func(options *opt.Options) {
    88  		// Ensure we have some minimal caching and file guarantees
    89  		if cache < minCache {
    90  			cache = minCache
    91  		}
    92  		if handles < minHandles {
    93  			handles = minHandles
    94  		}
    95  		// Set default options
    96  		options.OpenFilesCacheCapacity = handles
    97  		options.BlockCacheCapacity = cache / 2 * opt.MiB
    98  		options.WriteBuffer = cache / 4 * opt.MiB // Two of these are used internally
    99  		if readonly {
   100  			options.ReadOnly = true
   101  		}
   102  	})
   103  }
   104  
   105  // NewCustom returns a wrapped LevelDB object. The namespace is the prefix that the
   106  // metrics reporting should use for surfacing internal stats.
   107  // The customize function allows the caller to modify the leveldb options.
   108  func NewCustom(file string, namespace string, customize func(options *opt.Options)) (*Database, error) {
   109  	options := configureOptions(customize)
   110  	logger := log.New("database", file)
   111  	usedCache := options.GetBlockCacheCapacity() + options.GetWriteBuffer()*2
   112  	logCtx := []interface{}{"cache", common.StorageSize(usedCache), "handles", options.GetOpenFilesCacheCapacity()}
   113  	if options.ReadOnly {
   114  		logCtx = append(logCtx, "readonly", "true")
   115  	}
   116  	logger.Info("Allocated cache and file handles", logCtx...)
   117  
   118  	// Open the db and recover any potential corruptions
   119  	db, err := leveldb.OpenFile(file, options)
   120  	if _, corrupted := err.(*errors.ErrCorrupted); corrupted {
   121  		db, err = leveldb.RecoverFile(file, nil)
   122  	}
   123  	if err != nil {
   124  		return nil, err
   125  	}
   126  	// Assemble the wrapper with all the registered metrics
   127  	ldb := &Database{
   128  		fn:       file,
   129  		db:       db,
   130  		log:      logger,
   131  		quitChan: make(chan chan error),
   132  	}
   133  	ldb.compTimeMeter = metrics.NewRegisteredMeter(namespace+"compact/time", nil)
   134  	ldb.compReadMeter = metrics.NewRegisteredMeter(namespace+"compact/input", nil)
   135  	ldb.compWriteMeter = metrics.NewRegisteredMeter(namespace+"compact/output", nil)
   136  	ldb.diskSizeGauge = metrics.NewRegisteredGauge(namespace+"disk/size", nil)
   137  	ldb.diskReadMeter = metrics.NewRegisteredMeter(namespace+"disk/read", nil)
   138  	ldb.diskWriteMeter = metrics.NewRegisteredMeter(namespace+"disk/write", nil)
   139  	ldb.writeDelayMeter = metrics.NewRegisteredMeter(namespace+"compact/writedelay/duration", nil)
   140  	ldb.writeDelayNMeter = metrics.NewRegisteredMeter(namespace+"compact/writedelay/counter", nil)
   141  	ldb.memCompGauge = metrics.NewRegisteredGauge(namespace+"compact/memory", nil)
   142  	ldb.level0CompGauge = metrics.NewRegisteredGauge(namespace+"compact/level0", nil)
   143  	ldb.nonlevel0CompGauge = metrics.NewRegisteredGauge(namespace+"compact/nonlevel0", nil)
   144  	ldb.seekCompGauge = metrics.NewRegisteredGauge(namespace+"compact/seek", nil)
   145  
   146  	// Start up the metrics gathering and return
   147  	go ldb.meter(metricsGatheringInterval)
   148  	return ldb, nil
   149  }
   150  
   151  // configureOptions sets some default options, then runs the provided setter.
   152  func configureOptions(customizeFn func(*opt.Options)) *opt.Options {
   153  	// Set default options
   154  	options := &opt.Options{
   155  		Filter:                 filter.NewBloomFilter(10),
   156  		DisableSeeksCompaction: true,
   157  	}
   158  	// Allow caller to make custom modifications to the options
   159  	if customizeFn != nil {
   160  		customizeFn(options)
   161  	}
   162  	return options
   163  }
   164  
   165  // Close stops the metrics collection, flushes any pending data to disk and closes
   166  // all io accesses to the underlying key-value store.
   167  func (db *Database) Close() error {
   168  	db.quitLock.Lock()
   169  	defer db.quitLock.Unlock()
   170  
   171  	if db.quitChan != nil {
   172  		errc := make(chan error)
   173  		db.quitChan <- errc
   174  		if err := <-errc; err != nil {
   175  			db.log.Error("Metrics collection failed", "err", err)
   176  		}
   177  		db.quitChan = nil
   178  	}
   179  	return db.db.Close()
   180  }
   181  
   182  // Has retrieves if a key is present in the key-value store.
   183  func (db *Database) Has(key []byte) (bool, error) {
   184  	return db.db.Has(key, nil)
   185  }
   186  
   187  // Get retrieves the given key if it's present in the key-value store.
   188  func (db *Database) Get(key []byte) ([]byte, error) {
   189  	dat, err := db.db.Get(key, nil)
   190  	if err != nil {
   191  		return nil, err
   192  	}
   193  	return dat, nil
   194  }
   195  
   196  // Put inserts the given value into the key-value store.
   197  func (db *Database) Put(key []byte, value []byte) error {
   198  	return db.db.Put(key, value, nil)
   199  }
   200  
   201  // Delete removes the key from the key-value store.
   202  func (db *Database) Delete(key []byte) error {
   203  	return db.db.Delete(key, nil)
   204  }
   205  
   206  // NewBatch creates a write-only key-value store that buffers changes to its host
   207  // database until a final write is called.
   208  func (db *Database) NewBatch() ethdb.Batch {
   209  	return &batch{
   210  		db: db.db,
   211  		b:  new(leveldb.Batch),
   212  	}
   213  }
   214  
   215  // NewIterator creates a binary-alphabetical iterator over a subset
   216  // of database content with a particular key prefix, starting at a particular
   217  // initial key (or after, if it does not exist).
   218  func (db *Database) NewIterator(prefix []byte, start []byte) ethdb.Iterator {
   219  	return db.db.NewIterator(bytesPrefixRange(prefix, start), nil)
   220  }
   221  
   222  // Stat returns a particular internal stat of the database.
   223  func (db *Database) Stat(property string) (string, error) {
   224  	return db.db.GetProperty(property)
   225  }
   226  
   227  // Compact flattens the underlying data store for the given key range. In essence,
   228  // deleted and overwritten versions are discarded, and the data is rearranged to
   229  // reduce the cost of operations needed to access them.
   230  //
   231  // A nil start is treated as a key before all keys in the data store; a nil limit
   232  // is treated as a key after all keys in the data store. If both is nil then it
   233  // will compact entire data store.
   234  func (db *Database) Compact(start []byte, limit []byte) error {
   235  	return db.db.CompactRange(util.Range{Start: start, Limit: limit})
   236  }
   237  
   238  // Path returns the path to the database directory.
   239  func (db *Database) Path() string {
   240  	return db.fn
   241  }
   242  
   243  // meter periodically retrieves internal leveldb counters and reports them to
   244  // the metrics subsystem.
   245  //
   246  // This is how a LevelDB stats table looks like (currently):
   247  //   Compactions
   248  //    Level |   Tables   |    Size(MB)   |    Time(sec)  |    Read(MB)   |   Write(MB)
   249  //   -------+------------+---------------+---------------+---------------+---------------
   250  //      0   |          0 |       0.00000 |       1.27969 |       0.00000 |      12.31098
   251  //      1   |         85 |     109.27913 |      28.09293 |     213.92493 |     214.26294
   252  //      2   |        523 |    1000.37159 |       7.26059 |      66.86342 |      66.77884
   253  //      3   |        570 |    1113.18458 |       0.00000 |       0.00000 |       0.00000
   254  //
   255  // This is how the write delay look like (currently):
   256  // DelayN:5 Delay:406.604657ms Paused: false
   257  //
   258  // This is how the iostats look like (currently):
   259  // Read(MB):3895.04860 Write(MB):3654.64712
   260  func (db *Database) meter(refresh time.Duration) {
   261  	// Create the counters to store current and previous compaction values
   262  	compactions := make([][]float64, 2)
   263  	for i := 0; i < 2; i++ {
   264  		compactions[i] = make([]float64, 4)
   265  	}
   266  	// Create storage for iostats.
   267  	var iostats [2]float64
   268  
   269  	// Create storage and warning log tracer for write delay.
   270  	var (
   271  		delaystats      [2]int64
   272  		lastWritePaused time.Time
   273  	)
   274  
   275  	var (
   276  		errc chan error
   277  		merr error
   278  	)
   279  
   280  	timer := time.NewTimer(refresh)
   281  	defer timer.Stop()
   282  
   283  	// Iterate ad infinitum and collect the stats
   284  	for i := 1; errc == nil && merr == nil; i++ {
   285  		// Retrieve the database stats
   286  		stats, err := db.db.GetProperty("leveldb.stats")
   287  		if err != nil {
   288  			db.log.Error("Failed to read database stats", "err", err)
   289  			merr = err
   290  			continue
   291  		}
   292  		// Find the compaction table, skip the header
   293  		lines := strings.Split(stats, "\n")
   294  		for len(lines) > 0 && strings.TrimSpace(lines[0]) != "Compactions" {
   295  			lines = lines[1:]
   296  		}
   297  		if len(lines) <= 3 {
   298  			db.log.Error("Compaction leveldbTable not found")
   299  			merr = errors.New("compaction leveldbTable not found")
   300  			continue
   301  		}
   302  		lines = lines[3:]
   303  
   304  		// Iterate over all the leveldbTable rows, and accumulate the entries
   305  		for j := 0; j < len(compactions[i%2]); j++ {
   306  			compactions[i%2][j] = 0
   307  		}
   308  		for _, line := range lines {
   309  			parts := strings.Split(line, "|")
   310  			if len(parts) != 6 {
   311  				break
   312  			}
   313  			for idx, counter := range parts[2:] {
   314  				value, err := strconv.ParseFloat(strings.TrimSpace(counter), 64)
   315  				if err != nil {
   316  					db.log.Error("Compaction entry parsing failed", "err", err)
   317  					merr = err
   318  					continue
   319  				}
   320  				compactions[i%2][idx] += value
   321  			}
   322  		}
   323  		// Update all the requested meters
   324  		if db.diskSizeGauge != nil {
   325  			db.diskSizeGauge.Update(int64(compactions[i%2][0] * 1024 * 1024))
   326  		}
   327  		if db.compTimeMeter != nil {
   328  			db.compTimeMeter.Mark(int64((compactions[i%2][1] - compactions[(i-1)%2][1]) * 1000 * 1000 * 1000))
   329  		}
   330  		if db.compReadMeter != nil {
   331  			db.compReadMeter.Mark(int64((compactions[i%2][2] - compactions[(i-1)%2][2]) * 1024 * 1024))
   332  		}
   333  		if db.compWriteMeter != nil {
   334  			db.compWriteMeter.Mark(int64((compactions[i%2][3] - compactions[(i-1)%2][3]) * 1024 * 1024))
   335  		}
   336  		// Retrieve the write delay statistic
   337  		writedelay, err := db.db.GetProperty("leveldb.writedelay")
   338  		if err != nil {
   339  			db.log.Error("Failed to read database write delay statistic", "err", err)
   340  			merr = err
   341  			continue
   342  		}
   343  		var (
   344  			delayN        int64
   345  			delayDuration string
   346  			duration      time.Duration
   347  			paused        bool
   348  		)
   349  		if n, err := fmt.Sscanf(writedelay, "DelayN:%d Delay:%s Paused:%t", &delayN, &delayDuration, &paused); n != 3 || err != nil {
   350  			db.log.Error("Write delay statistic not found")
   351  			merr = err
   352  			continue
   353  		}
   354  		duration, err = time.ParseDuration(delayDuration)
   355  		if err != nil {
   356  			db.log.Error("Failed to parse delay duration", "err", err)
   357  			merr = err
   358  			continue
   359  		}
   360  		if db.writeDelayNMeter != nil {
   361  			db.writeDelayNMeter.Mark(delayN - delaystats[0])
   362  		}
   363  		if db.writeDelayMeter != nil {
   364  			db.writeDelayMeter.Mark(duration.Nanoseconds() - delaystats[1])
   365  		}
   366  		// If a warning that db is performing compaction has been displayed, any subsequent
   367  		// warnings will be withheld for one minute not to overwhelm the user.
   368  		if paused && delayN-delaystats[0] == 0 && duration.Nanoseconds()-delaystats[1] == 0 &&
   369  			time.Now().After(lastWritePaused.Add(degradationWarnInterval)) {
   370  			db.log.Warn("Database compacting, degraded performance")
   371  			lastWritePaused = time.Now()
   372  		}
   373  		delaystats[0], delaystats[1] = delayN, duration.Nanoseconds()
   374  
   375  		// Retrieve the database iostats.
   376  		ioStats, err := db.db.GetProperty("leveldb.iostats")
   377  		if err != nil {
   378  			db.log.Error("Failed to read database iostats", "err", err)
   379  			merr = err
   380  			continue
   381  		}
   382  		var nRead, nWrite float64
   383  		parts := strings.Split(ioStats, " ")
   384  		if len(parts) < 2 {
   385  			db.log.Error("Bad syntax of ioStats", "ioStats", ioStats)
   386  			merr = fmt.Errorf("bad syntax of ioStats %s", ioStats)
   387  			continue
   388  		}
   389  		if n, err := fmt.Sscanf(parts[0], "Read(MB):%f", &nRead); n != 1 || err != nil {
   390  			db.log.Error("Bad syntax of read entry", "entry", parts[0])
   391  			merr = err
   392  			continue
   393  		}
   394  		if n, err := fmt.Sscanf(parts[1], "Write(MB):%f", &nWrite); n != 1 || err != nil {
   395  			db.log.Error("Bad syntax of write entry", "entry", parts[1])
   396  			merr = err
   397  			continue
   398  		}
   399  		if db.diskReadMeter != nil {
   400  			db.diskReadMeter.Mark(int64((nRead - iostats[0]) * 1024 * 1024))
   401  		}
   402  		if db.diskWriteMeter != nil {
   403  			db.diskWriteMeter.Mark(int64((nWrite - iostats[1]) * 1024 * 1024))
   404  		}
   405  		iostats[0], iostats[1] = nRead, nWrite
   406  
   407  		compCount, err := db.db.GetProperty("leveldb.compcount")
   408  		if err != nil {
   409  			db.log.Error("Failed to read database iostats", "err", err)
   410  			merr = err
   411  			continue
   412  		}
   413  
   414  		var (
   415  			memComp       uint32
   416  			level0Comp    uint32
   417  			nonLevel0Comp uint32
   418  			seekComp      uint32
   419  		)
   420  		if n, err := fmt.Sscanf(compCount, "MemComp:%d Level0Comp:%d NonLevel0Comp:%d SeekComp:%d", &memComp, &level0Comp, &nonLevel0Comp, &seekComp); n != 4 || err != nil {
   421  			db.log.Error("Compaction count statistic not found")
   422  			merr = err
   423  			continue
   424  		}
   425  		db.memCompGauge.Update(int64(memComp))
   426  		db.level0CompGauge.Update(int64(level0Comp))
   427  		db.nonlevel0CompGauge.Update(int64(nonLevel0Comp))
   428  		db.seekCompGauge.Update(int64(seekComp))
   429  
   430  		// Sleep a bit, then repeat the stats collection
   431  		select {
   432  		case errc = <-db.quitChan:
   433  			// Quit requesting, stop hammering the database
   434  		case <-timer.C:
   435  			timer.Reset(refresh)
   436  			// Timeout, gather a new set of stats
   437  		}
   438  	}
   439  
   440  	if errc == nil {
   441  		errc = <-db.quitChan
   442  	}
   443  	errc <- merr
   444  }
   445  
   446  // batch is a write-only leveldb batch that commits changes to its host database
   447  // when Write is called. A batch cannot be used concurrently.
   448  type batch struct {
   449  	db   *leveldb.DB
   450  	b    *leveldb.Batch
   451  	size int
   452  }
   453  
   454  // Put inserts the given value into the batch for later committing.
   455  func (b *batch) Put(key, value []byte) error {
   456  	b.b.Put(key, value)
   457  	b.size += len(value)
   458  	return nil
   459  }
   460  
   461  // Delete inserts the a key removal into the batch for later committing.
   462  func (b *batch) Delete(key []byte) error {
   463  	b.b.Delete(key)
   464  	b.size += len(key)
   465  	return nil
   466  }
   467  
   468  // ValueSize retrieves the amount of data queued up for writing.
   469  func (b *batch) ValueSize() int {
   470  	return b.size
   471  }
   472  
   473  // Write flushes any accumulated data to disk.
   474  func (b *batch) Write() error {
   475  	return b.db.Write(b.b, nil)
   476  }
   477  
   478  // Reset resets the batch for reuse.
   479  func (b *batch) Reset() {
   480  	b.b.Reset()
   481  	b.size = 0
   482  }
   483  
   484  // Replay replays the batch contents.
   485  func (b *batch) Replay(w ethdb.KeyValueWriter) error {
   486  	return b.b.Replay(&replayer{writer: w})
   487  }
   488  
   489  // replayer is a small wrapper to implement the correct replay methods.
   490  type replayer struct {
   491  	writer  ethdb.KeyValueWriter
   492  	failure error
   493  }
   494  
   495  // Put inserts the given value into the key-value data store.
   496  func (r *replayer) Put(key, value []byte) {
   497  	// If the replay already failed, stop executing ops
   498  	if r.failure != nil {
   499  		return
   500  	}
   501  	r.failure = r.writer.Put(key, value)
   502  }
   503  
   504  // Delete removes the key from the key-value data store.
   505  func (r *replayer) Delete(key []byte) {
   506  	// If the replay already failed, stop executing ops
   507  	if r.failure != nil {
   508  		return
   509  	}
   510  	r.failure = r.writer.Delete(key)
   511  }
   512  
   513  // bytesPrefixRange returns key range that satisfy
   514  // - the given prefix, and
   515  // - the given seek position
   516  func bytesPrefixRange(prefix, start []byte) *util.Range {
   517  	r := util.BytesPrefix(prefix)
   518  	r.Start = append(r.Start, start...)
   519  	return r
   520  }