github.com/dominant-strategies/go-quai@v0.28.2/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  //go:build !js
    18  // +build !js
    19  
    20  // Package leveldb implements the key-value database layer based on LevelDB.
    21  package leveldb
    22  
    23  import (
    24  	"fmt"
    25  	"strconv"
    26  	"strings"
    27  	"sync"
    28  	"time"
    29  
    30  	"github.com/dominant-strategies/go-quai/common"
    31  	"github.com/dominant-strategies/go-quai/ethdb"
    32  	"github.com/dominant-strategies/go-quai/log"
    33  	"github.com/dominant-strategies/go-quai/metrics"
    34  	"github.com/syndtr/goleveldb/leveldb"
    35  	"github.com/syndtr/goleveldb/leveldb/errors"
    36  	"github.com/syndtr/goleveldb/leveldb/filter"
    37  	"github.com/syndtr/goleveldb/leveldb/opt"
    38  	"github.com/syndtr/goleveldb/leveldb/util"
    39  )
    40  
    41  const (
    42  	// degradationWarnInterval specifies how often warning should be printed if the
    43  	// leveldb database cannot keep up with requested writes.
    44  	degradationWarnInterval = time.Minute
    45  
    46  	// minCache is the minimum amount of memory in megabytes to allocate to leveldb
    47  	// read and write caching, split half and half.
    48  	minCache = 16
    49  
    50  	// minHandles is the minimum number of files handles to allocate to the open
    51  	// database files.
    52  	minHandles = 16
    53  
    54  	// metricsGatheringInterval specifies the interval to retrieve leveldb database
    55  	// compaction, io and pause stats to report to the user.
    56  	metricsGatheringInterval = 3 * time.Second
    57  )
    58  
    59  // Database is a persistent key-value store. Apart from basic data storage
    60  // functionality it also supports batch writes and iterating over the keyspace in
    61  // binary-alphabetical order.
    62  type Database struct {
    63  	fn string      // filename for reporting
    64  	db *leveldb.DB // LevelDB instance
    65  
    66  	compTimeMeter       metrics.Meter // Meter for measuring the total time spent in database compaction
    67  	compReadMeter       metrics.Meter // Meter for measuring the data read during compaction
    68  	compWriteMeter      metrics.Meter // Meter for measuring the data written during compaction
    69  	writeDelayNMeter    metrics.Meter // Meter for measuring the write delay number due to database compaction
    70  	writeDelayMeter     metrics.Meter // Meter for measuring the write delay duration due to database compaction
    71  	diskSizeGauge       metrics.Gauge // Gauge for tracking the size of all the levels in the database
    72  	diskReadMeter       metrics.Meter // Meter for measuring the effective amount of data read
    73  	diskWriteMeter      metrics.Meter // Meter for measuring the effective amount of data written
    74  	memCompGauge        metrics.Gauge // Gauge for tracking the number of memory compaction
    75  	level0CompGauge     metrics.Gauge // Gauge for tracking the number of table compaction in level0
    76  	nonlevel0CompGauge  metrics.Gauge // Gauge for tracking the number of table compaction in non0 level
    77  	seekCompGauge       metrics.Gauge // Gauge for tracking the number of table compaction caused by read opt
    78  	manualMemAllocGauge metrics.Gauge // Gauge to track the amount of memory that has been manually allocated (not a part of runtime/GC)
    79  
    80  	quitLock sync.Mutex      // Mutex protecting the quit channel access
    81  	quitChan chan chan error // Quit channel to stop the metrics collection before closing the database
    82  
    83  	log log.Logger // Contextual logger tracking the database path
    84  }
    85  
    86  // New returns a wrapped LevelDB object. The namespace is the prefix that the
    87  // metrics reporting should use for surfacing internal stats.
    88  func New(file string, cache int, handles int, namespace string, readonly bool) (*Database, error) {
    89  	return NewCustom(file, namespace, func(options *opt.Options) {
    90  		// Ensure we have some minimal caching and file guarantees
    91  		if cache < minCache {
    92  			cache = minCache
    93  		}
    94  		if handles < minHandles {
    95  			handles = minHandles
    96  		}
    97  		// Set default options
    98  		options.OpenFilesCacheCapacity = handles
    99  		options.BlockCacheCapacity = cache / 2 * opt.MiB
   100  		options.WriteBuffer = cache / 4 * opt.MiB // Two of these are used internally
   101  		if readonly {
   102  			options.ReadOnly = true
   103  		}
   104  	})
   105  }
   106  
   107  // NewCustom returns a wrapped LevelDB object. The namespace is the prefix that the
   108  // metrics reporting should use for surfacing internal stats.
   109  // The customize function allows the caller to modify the leveldb options.
   110  func NewCustom(file string, namespace string, customize func(options *opt.Options)) (*Database, error) {
   111  	options := configureOptions(customize)
   112  	usedCache := options.GetBlockCacheCapacity() + options.GetWriteBuffer()*2
   113  	logCtx := []interface{}{"cache", common.StorageSize(usedCache), "handles", options.GetOpenFilesCacheCapacity()}
   114  	if options.ReadOnly {
   115  		logCtx = append(logCtx, "readonly", "true")
   116  	}
   117  	log.Info("Allocated cache and file handles", logCtx...)
   118  
   119  	// Open the db and recover any potential corruptions
   120  	db, err := leveldb.OpenFile(file, options)
   121  	if _, corrupted := err.(*errors.ErrCorrupted); corrupted {
   122  		db, err = leveldb.RecoverFile(file, nil)
   123  	}
   124  	if err != nil {
   125  		return nil, err
   126  	}
   127  	// Assemble the wrapper with all the registered metrics
   128  	ldb := &Database{
   129  		fn:       file,
   130  		db:       db,
   131  		log:      log.Log,
   132  		quitChan: make(chan chan error),
   133  	}
   134  	ldb.compTimeMeter = metrics.NewRegisteredMeter(namespace+"compact/time", nil)
   135  	ldb.compReadMeter = metrics.NewRegisteredMeter(namespace+"compact/input", nil)
   136  	ldb.compWriteMeter = metrics.NewRegisteredMeter(namespace+"compact/output", nil)
   137  	ldb.diskSizeGauge = metrics.NewRegisteredGauge(namespace+"disk/size", nil)
   138  	ldb.diskReadMeter = metrics.NewRegisteredMeter(namespace+"disk/read", nil)
   139  	ldb.diskWriteMeter = metrics.NewRegisteredMeter(namespace+"disk/write", nil)
   140  	ldb.writeDelayMeter = metrics.NewRegisteredMeter(namespace+"compact/writedelay/duration", nil)
   141  	ldb.writeDelayNMeter = metrics.NewRegisteredMeter(namespace+"compact/writedelay/counter", nil)
   142  	ldb.memCompGauge = metrics.NewRegisteredGauge(namespace+"compact/memory", nil)
   143  	ldb.level0CompGauge = metrics.NewRegisteredGauge(namespace+"compact/level0", nil)
   144  	ldb.nonlevel0CompGauge = metrics.NewRegisteredGauge(namespace+"compact/nonlevel0", nil)
   145  	ldb.seekCompGauge = metrics.NewRegisteredGauge(namespace+"compact/seek", nil)
   146  	ldb.manualMemAllocGauge = metrics.NewRegisteredGauge(namespace+"memory/manualalloc", nil)
   147  
   148  	// Start up the metrics gathering and return
   149  	go ldb.meter(metricsGatheringInterval)
   150  	return ldb, nil
   151  }
   152  
   153  // configureOptions sets some default options, then runs the provided setter.
   154  func configureOptions(customizeFn func(*opt.Options)) *opt.Options {
   155  	// Set default options
   156  	options := &opt.Options{
   157  		Filter:                 filter.NewBloomFilter(10),
   158  		DisableSeeksCompaction: true,
   159  	}
   160  	// Allow caller to make custom modifications to the options
   161  	if customizeFn != nil {
   162  		customizeFn(options)
   163  	}
   164  	return options
   165  }
   166  
   167  // Close stops the metrics collection, flushes any pending data to disk and closes
   168  // all io accesses to the underlying key-value store.
   169  func (db *Database) Close() error {
   170  	db.quitLock.Lock()
   171  	defer db.quitLock.Unlock()
   172  
   173  	if db.quitChan != nil {
   174  		errc := make(chan error)
   175  		db.quitChan <- errc
   176  		if err := <-errc; err != nil {
   177  			db.log.Error("Metrics collection failed", "err", err)
   178  		}
   179  		db.quitChan = nil
   180  	}
   181  	return db.db.Close()
   182  }
   183  
   184  // Has retrieves if a key is present in the key-value store.
   185  func (db *Database) Has(key []byte) (bool, error) {
   186  	return db.db.Has(key, nil)
   187  }
   188  
   189  // Get retrieves the given key if it's present in the key-value store.
   190  func (db *Database) Get(key []byte) ([]byte, error) {
   191  	dat, err := db.db.Get(key, nil)
   192  	if err != nil {
   193  		return nil, err
   194  	}
   195  	return dat, nil
   196  }
   197  
   198  // Put inserts the given value into the key-value store.
   199  func (db *Database) Put(key []byte, value []byte) error {
   200  	return db.db.Put(key, value, nil)
   201  }
   202  
   203  // Delete removes the key from the key-value store.
   204  func (db *Database) Delete(key []byte) error {
   205  	return db.db.Delete(key, nil)
   206  }
   207  
   208  // NewBatch creates a write-only key-value store that buffers changes to its host
   209  // database until a final write is called.
   210  func (db *Database) NewBatch() ethdb.Batch {
   211  	return &batch{
   212  		db: db.db,
   213  		b:  new(leveldb.Batch),
   214  	}
   215  }
   216  
   217  // NewIterator creates a binary-alphabetical iterator over a subset
   218  // of database content with a particular key prefix, starting at a particular
   219  // initial key (or after, if it does not exist).
   220  func (db *Database) NewIterator(prefix []byte, start []byte) ethdb.Iterator {
   221  	return db.db.NewIterator(bytesPrefixRange(prefix, start), nil)
   222  }
   223  
   224  // Stat returns a particular internal stat of the database.
   225  func (db *Database) Stat(property string) (string, error) {
   226  	return db.db.GetProperty(property)
   227  }
   228  
   229  // Compact flattens the underlying data store for the given key range. In essence,
   230  // deleted and overwritten versions are discarded, and the data is rearranged to
   231  // reduce the cost of operations needed to access them.
   232  //
   233  // A nil start is treated as a key before all keys in the data store; a nil limit
   234  // is treated as a key after all keys in the data store. If both is nil then it
   235  // will compact entire data store.
   236  func (db *Database) Compact(start []byte, limit []byte) error {
   237  	return db.db.CompactRange(util.Range{Start: start, Limit: limit})
   238  }
   239  
   240  // Path returns the path to the database directory.
   241  func (db *Database) Path() string {
   242  	return db.fn
   243  }
   244  
   245  // meter periodically retrieves internal leveldb counters and reports them to
   246  // the metrics subsystem.
   247  //
   248  // This is how a LevelDB stats table looks like (currently):
   249  //
   250  //	Compactions
   251  //	 Level |   Tables   |    Size(MB)   |    Time(sec)  |    Read(MB)   |   Write(MB)
   252  //	-------+------------+---------------+---------------+---------------+---------------
   253  //	   0   |          0 |       0.00000 |       1.27969 |       0.00000 |      12.31098
   254  //	   1   |         85 |     109.27913 |      28.09293 |     213.92493 |     214.26294
   255  //	   2   |        523 |    1000.37159 |       7.26059 |      66.86342 |      66.77884
   256  //	   3   |        570 |    1113.18458 |       0.00000 |       0.00000 |       0.00000
   257  //
   258  // This is how the write delay look like (currently):
   259  // DelayN:5 Delay:406.604657ms Paused: false
   260  //
   261  // This is how the iostats look like (currently):
   262  // Read(MB):3895.04860 Write(MB):3654.64712
   263  func (db *Database) meter(refresh time.Duration) {
   264  	// Create the counters to store current and previous compaction values
   265  	compactions := make([][]float64, 2)
   266  	for i := 0; i < 2; i++ {
   267  		compactions[i] = make([]float64, 4)
   268  	}
   269  	// Create storage for iostats.
   270  	var iostats [2]float64
   271  
   272  	// Create storage and warning log tracer for write delay.
   273  	var (
   274  		delaystats      [2]int64
   275  		lastWritePaused time.Time
   276  	)
   277  
   278  	var (
   279  		errc chan error
   280  		merr error
   281  	)
   282  
   283  	timer := time.NewTimer(refresh)
   284  	defer timer.Stop()
   285  
   286  	// Iterate ad infinitum and collect the stats
   287  	for i := 1; errc == nil && merr == nil; i++ {
   288  		// Retrieve the database stats
   289  		stats, err := db.db.GetProperty("leveldb.stats")
   290  		if err != nil {
   291  			db.log.Error("Failed to read database stats", "err", err)
   292  			merr = err
   293  			continue
   294  		}
   295  		// Find the compaction table, skip the header
   296  		lines := strings.Split(stats, "\n")
   297  		for len(lines) > 0 && strings.TrimSpace(lines[0]) != "Compactions" {
   298  			lines = lines[1:]
   299  		}
   300  		if len(lines) <= 3 {
   301  			db.log.Error("Compaction leveldbTable not found")
   302  			merr = errors.New("compaction leveldbTable not found")
   303  			continue
   304  		}
   305  		lines = lines[3:]
   306  
   307  		// Iterate over all the leveldbTable rows, and accumulate the entries
   308  		for j := 0; j < len(compactions[i%2]); j++ {
   309  			compactions[i%2][j] = 0
   310  		}
   311  		for _, line := range lines {
   312  			parts := strings.Split(line, "|")
   313  			if len(parts) != 6 {
   314  				break
   315  			}
   316  			for idx, counter := range parts[2:] {
   317  				value, err := strconv.ParseFloat(strings.TrimSpace(counter), 64)
   318  				if err != nil {
   319  					db.log.Error("Compaction entry parsing failed", "err", err)
   320  					merr = err
   321  					continue
   322  				}
   323  				compactions[i%2][idx] += value
   324  			}
   325  		}
   326  		// Update all the requested meters
   327  		if db.diskSizeGauge != nil {
   328  			db.diskSizeGauge.Update(int64(compactions[i%2][0] * 1024 * 1024))
   329  		}
   330  		if db.compTimeMeter != nil {
   331  			db.compTimeMeter.Mark(int64((compactions[i%2][1] - compactions[(i-1)%2][1]) * 1000 * 1000 * 1000))
   332  		}
   333  		if db.compReadMeter != nil {
   334  			db.compReadMeter.Mark(int64((compactions[i%2][2] - compactions[(i-1)%2][2]) * 1024 * 1024))
   335  		}
   336  		if db.compWriteMeter != nil {
   337  			db.compWriteMeter.Mark(int64((compactions[i%2][3] - compactions[(i-1)%2][3]) * 1024 * 1024))
   338  		}
   339  		// Retrieve the write delay statistic
   340  		writedelay, err := db.db.GetProperty("leveldb.writedelay")
   341  		if err != nil {
   342  			db.log.Error("Failed to read database write delay statistic", "err", err)
   343  			merr = err
   344  			continue
   345  		}
   346  		var (
   347  			delayN        int64
   348  			delayDuration string
   349  			duration      time.Duration
   350  			paused        bool
   351  		)
   352  		if n, err := fmt.Sscanf(writedelay, "DelayN:%d Delay:%s Paused:%t", &delayN, &delayDuration, &paused); n != 3 || err != nil {
   353  			db.log.Error("Write delay statistic not found")
   354  			merr = err
   355  			continue
   356  		}
   357  		duration, err = time.ParseDuration(delayDuration)
   358  		if err != nil {
   359  			db.log.Error("Failed to parse delay duration", "err", err)
   360  			merr = err
   361  			continue
   362  		}
   363  		if db.writeDelayNMeter != nil {
   364  			db.writeDelayNMeter.Mark(delayN - delaystats[0])
   365  		}
   366  		if db.writeDelayMeter != nil {
   367  			db.writeDelayMeter.Mark(duration.Nanoseconds() - delaystats[1])
   368  		}
   369  		// If a warning that db is performing compaction has been displayed, any subsequent
   370  		// warnings will be withheld for one minute not to overwhelm the user.
   371  		if paused && delayN-delaystats[0] == 0 && duration.Nanoseconds()-delaystats[1] == 0 &&
   372  			time.Now().After(lastWritePaused.Add(degradationWarnInterval)) {
   373  			db.log.Warn("Database compacting, degraded performance")
   374  			lastWritePaused = time.Now()
   375  		}
   376  		delaystats[0], delaystats[1] = delayN, duration.Nanoseconds()
   377  
   378  		// Retrieve the database iostats.
   379  		ioStats, err := db.db.GetProperty("leveldb.iostats")
   380  		if err != nil {
   381  			db.log.Error("Failed to read database iostats", "err", err)
   382  			merr = err
   383  			continue
   384  		}
   385  		var nRead, nWrite float64
   386  		parts := strings.Split(ioStats, " ")
   387  		if len(parts) < 2 {
   388  			db.log.Error("Bad syntax of ioStats", "ioStats", ioStats)
   389  			merr = fmt.Errorf("bad syntax of ioStats %s", ioStats)
   390  			continue
   391  		}
   392  		if n, err := fmt.Sscanf(parts[0], "Read(MB):%f", &nRead); n != 1 || err != nil {
   393  			db.log.Error("Bad syntax of read entry", "entry", parts[0])
   394  			merr = err
   395  			continue
   396  		}
   397  		if n, err := fmt.Sscanf(parts[1], "Write(MB):%f", &nWrite); n != 1 || err != nil {
   398  			db.log.Error("Bad syntax of write entry", "entry", parts[1])
   399  			merr = err
   400  			continue
   401  		}
   402  		if db.diskReadMeter != nil {
   403  			db.diskReadMeter.Mark(int64((nRead - iostats[0]) * 1024 * 1024))
   404  		}
   405  		if db.diskWriteMeter != nil {
   406  			db.diskWriteMeter.Mark(int64((nWrite - iostats[1]) * 1024 * 1024))
   407  		}
   408  		iostats[0], iostats[1] = nRead, nWrite
   409  
   410  		compCount, err := db.db.GetProperty("leveldb.compcount")
   411  		if err != nil {
   412  			db.log.Error("Failed to read database iostats", "err", err)
   413  			merr = err
   414  			continue
   415  		}
   416  
   417  		var (
   418  			memComp       uint32
   419  			level0Comp    uint32
   420  			nonLevel0Comp uint32
   421  			seekComp      uint32
   422  		)
   423  		if n, err := fmt.Sscanf(compCount, "MemComp:%d Level0Comp:%d NonLevel0Comp:%d SeekComp:%d", &memComp, &level0Comp, &nonLevel0Comp, &seekComp); n != 4 || err != nil {
   424  			db.log.Error("Compaction count statistic not found")
   425  			merr = err
   426  			continue
   427  		}
   428  		db.memCompGauge.Update(int64(memComp))
   429  		db.level0CompGauge.Update(int64(level0Comp))
   430  		db.nonlevel0CompGauge.Update(int64(nonLevel0Comp))
   431  		db.seekCompGauge.Update(int64(seekComp))
   432  
   433  		// Sleep a bit, then repeat the stats collection
   434  		select {
   435  		case errc = <-db.quitChan:
   436  			// Quit requesting, stop hammering the database
   437  		case <-timer.C:
   438  			timer.Reset(refresh)
   439  			// Timeout, gather a new set of stats
   440  		}
   441  	}
   442  
   443  	if errc == nil {
   444  		errc = <-db.quitChan
   445  	}
   446  	errc <- merr
   447  }
   448  
   449  // batch is a write-only leveldb batch that commits changes to its host database
   450  // when Write is called. A batch cannot be used concurrently.
   451  type batch struct {
   452  	db   *leveldb.DB
   453  	b    *leveldb.Batch
   454  	size int
   455  }
   456  
   457  // Put inserts the given value into the batch for later committing.
   458  func (b *batch) Put(key, value []byte) error {
   459  	b.b.Put(key, value)
   460  	b.size += len(value)
   461  	return nil
   462  }
   463  
   464  // Delete inserts the a key removal into the batch for later committing.
   465  func (b *batch) Delete(key []byte) error {
   466  	b.b.Delete(key)
   467  	b.size += len(key)
   468  	return nil
   469  }
   470  
   471  // ValueSize retrieves the amount of data queued up for writing.
   472  func (b *batch) ValueSize() int {
   473  	return b.size
   474  }
   475  
   476  // Write flushes any accumulated data to disk.
   477  func (b *batch) Write() error {
   478  	return b.db.Write(b.b, nil)
   479  }
   480  
   481  // Reset resets the batch for reuse.
   482  func (b *batch) Reset() {
   483  	b.b.Reset()
   484  	b.size = 0
   485  }
   486  
   487  // Replay replays the batch contents.
   488  func (b *batch) Replay(w ethdb.KeyValueWriter) error {
   489  	return b.b.Replay(&replayer{writer: w})
   490  }
   491  
   492  // replayer is a small wrapper to implement the correct replay methods.
   493  type replayer struct {
   494  	writer  ethdb.KeyValueWriter
   495  	failure error
   496  }
   497  
   498  // Put inserts the given value into the key-value data store.
   499  func (r *replayer) Put(key, value []byte) {
   500  	// If the replay already failed, stop executing ops
   501  	if r.failure != nil {
   502  		return
   503  	}
   504  	r.failure = r.writer.Put(key, value)
   505  }
   506  
   507  // Delete removes the key from the key-value data store.
   508  func (r *replayer) Delete(key []byte) {
   509  	// If the replay already failed, stop executing ops
   510  	if r.failure != nil {
   511  		return
   512  	}
   513  	r.failure = r.writer.Delete(key)
   514  }
   515  
   516  // bytesPrefixRange returns key range that satisfy
   517  // - the given prefix, and
   518  // - the given seek position
   519  func bytesPrefixRange(prefix, start []byte) *util.Range {
   520  	r := util.BytesPrefix(prefix)
   521  	r.Start = append(r.Start, start...)
   522  	return r
   523  }