Merge pull request #306 from OffchainLabs/pebble-extra-options

expose pebble open options, add more compact metrics

Author: PlasmaPower

core/rawdb/database.go

@@ -357,8 +357,8 @@ func NewLevelDBDatabase(file string, cache int, handles int, namespace string, r
 
 // NewPebbleDBDatabase creates a persistent key-value database without a freezer
 // moving immutable chain segments into cold storage.
-func NewPebbleDBDatabase(file string, cache int, handles int, namespace string, readonly, ephemeral bool) (ethdb.Database, error) {
-	db, err := pebble.New(file, cache, handles, namespace, readonly, ephemeral)
+func NewPebbleDBDatabase(file string, cache int, handles int, namespace string, readonly, ephemeral bool, extraOptions *pebble.ExtraOptions) (ethdb.Database, error) {
+	db, err := pebble.New(file, cache, handles, namespace, readonly, ephemeral, extraOptions)
 	if err != nil {
 		return nil, err
 	}
@@ -399,6 +399,8 @@ type OpenOptions struct {
 	// Ephemeral means that filesystem sync operations should be avoided: data integrity in the face of
 	// a crash is not important. This option should typically be used in tests.
 	Ephemeral bool
+
+	PebbleExtraOptions *pebble.ExtraOptions
 }
 
 // openKeyValueDatabase opens a disk-based key-value database, e.g. leveldb or pebble.
@@ -420,15 +422,15 @@ func openKeyValueDatabase(o OpenOptions) (ethdb.Database, error) {
 	}
 	if o.Type == dbPebble || existingDb == dbPebble {
 		log.Info("Using pebble as the backing database")
-		return NewPebbleDBDatabase(o.Directory, o.Cache, o.Handles, o.Namespace, o.ReadOnly, o.Ephemeral)
+		return NewPebbleDBDatabase(o.Directory, o.Cache, o.Handles, o.Namespace, o.ReadOnly, o.Ephemeral, o.PebbleExtraOptions)
 	}
 	if o.Type == dbLeveldb || existingDb == dbLeveldb {
 		log.Info("Using leveldb as the backing database")
 		return NewLevelDBDatabase(o.Directory, o.Cache, o.Handles, o.Namespace, o.ReadOnly)
 	}
 	// No pre-existing database, no user-requested one either. Default to Pebble.
 	log.Info("Defaulting to pebble as the backing database")
-	return NewPebbleDBDatabase(o.Directory, o.Cache, o.Handles, o.Namespace, o.ReadOnly, o.Ephemeral)
+	return NewPebbleDBDatabase(o.Directory, o.Cache, o.Handles, o.Namespace, o.ReadOnly, o.Ephemeral, o.PebbleExtraOptions)
 }
 
 // Open opens both a disk-based key-value database such as leveldb or pebble, but also

ethdb/pebble/extraoptions.go

@@ -0,0 +1,35 @@
+package pebble
+
+import "time"
+
+type ExtraOptions struct {
+	BytesPerSync                int
+	L0CompactionFileThreshold   int
+	L0CompactionThreshold       int
+	L0StopWritesThreshold       int
+	LBaseMaxBytes               int64
+	MemTableStopWritesThreshold int
+	MaxConcurrentCompactions    func() int
+	DisableAutomaticCompactions bool
+	WALBytesPerSync             int
+	WALDir                      string
+	WALMinSyncInterval          func() time.Duration
+	TargetByteDeletionRate      int
+	Experimental                ExtraOptionsExperimental
+	Levels                      []ExtraLevelOptions
+}
+
+type ExtraOptionsExperimental struct {
+	L0CompactionConcurrency   int
+	CompactionDebtConcurrency uint64
+	ReadCompactionRate        int64
+	ReadSamplingMultiplier    int64
+	MaxWriterConcurrency      int
+	ForceWriterParallelism    bool
+}
+
+type ExtraLevelOptions struct {
+	BlockSize      int
+	IndexBlockSize int
+	TargetFileSize int64
+}

ethdb/pebble/pebble.go

@@ -70,6 +70,25 @@ type Database struct {
 	seekCompGauge       metrics.Gauge // Gauge for tracking the number of table compaction caused by read opt
 	manualMemAllocGauge metrics.Gauge // Gauge for tracking amount of non-managed memory currently allocated
 
+	compDebtGauge       metrics.Gauge
+	compInProgressGauge metrics.Gauge
+
+	commitCountMeter               metrics.Meter
+	commitTotalDurationMeter       metrics.Meter
+	commitSemaphoreWaitMeter       metrics.Meter
+	commitMemTableWriteStallMeter  metrics.Meter
+	commitL0ReadAmpWriteStallMeter metrics.Meter
+	commitWALRotationMeter         metrics.Meter
+	commitWaitMeter                metrics.Meter
+
+	commitCount               atomic.Int64
+	commitTotalDuration       atomic.Int64
+	commitSemaphoreWait       atomic.Int64
+	commitMemTableWriteStall  atomic.Int64
+	commitL0ReadAmpWriteStall atomic.Int64
+	commitWALRotation         atomic.Int64
+	commitWait                atomic.Int64
+
 	levelsGauge []metrics.Gauge // Gauge for tracking the number of tables in levels
 
 	quitLock sync.RWMutex    // Mutex protecting the quit channel and the closed flag
@@ -137,7 +156,38 @@ func (l panicLogger) Fatalf(format string, args ...interface{}) {
 
 // New returns a wrapped pebble DB object. The namespace is the prefix that the
 // metrics reporting should use for surfacing internal stats.
-func New(file string, cache int, handles int, namespace string, readonly bool, ephemeral bool) (*Database, error) {
+func New(file string, cache int, handles int, namespace string, readonly bool, ephemeral bool, extraOptions *ExtraOptions) (*Database, error) {
+	if extraOptions == nil {
+		extraOptions = &ExtraOptions{}
+	}
+	if extraOptions.MemTableStopWritesThreshold <= 0 {
+		extraOptions.MemTableStopWritesThreshold = 2
+	}
+	if extraOptions.MaxConcurrentCompactions == nil {
+		extraOptions.MaxConcurrentCompactions = func() int { return runtime.NumCPU() }
+	}
+	var levels []pebble.LevelOptions
+	if len(extraOptions.Levels) == 0 {
+		levels = []pebble.LevelOptions{
+			{TargetFileSize: 2 * 1024 * 1024, FilterPolicy: bloom.FilterPolicy(10)},
+			{TargetFileSize: 2 * 1024 * 1024, FilterPolicy: bloom.FilterPolicy(10)},
+			{TargetFileSize: 2 * 1024 * 1024, FilterPolicy: bloom.FilterPolicy(10)},
+			{TargetFileSize: 2 * 1024 * 1024, FilterPolicy: bloom.FilterPolicy(10)},
+			{TargetFileSize: 2 * 1024 * 1024, FilterPolicy: bloom.FilterPolicy(10)},
+			{TargetFileSize: 2 * 1024 * 1024, FilterPolicy: bloom.FilterPolicy(10)},
+			{TargetFileSize: 2 * 1024 * 1024, FilterPolicy: bloom.FilterPolicy(10)},
+		}
+	} else {
+		for _, level := range extraOptions.Levels {
+			levels = append(levels, pebble.LevelOptions{
+				BlockSize:      level.BlockSize,
+				IndexBlockSize: level.IndexBlockSize,
+				TargetFileSize: level.TargetFileSize,
+				FilterPolicy:   bloom.FilterPolicy(10),
+			})
+		}
+	}
+
 	// Ensure we have some minimal caching and file guarantees
 	if cache < minCache {
 		cache = minCache
@@ -162,7 +212,7 @@ func New(file string, cache int, handles int, namespace string, readonly bool, e
 
 	// Two memory tables is configured which is identical to leveldb,
 	// including a frozen memory table and another live one.
-	memTableLimit := 2
+	memTableLimit := extraOptions.MemTableStopWritesThreshold
 	memTableSize := cache * 1024 * 1024 / 2 / memTableLimit
 
 	// The memory table size is currently capped at maxMemTableSize-1 due to a
@@ -200,19 +250,11 @@ func New(file string, cache int, handles int, namespace string, readonly bool, e
 
 		// The default compaction concurrency(1 thread),
 		// Here use all available CPUs for faster compaction.
-		MaxConcurrentCompactions: func() int { return runtime.NumCPU() },
+		MaxConcurrentCompactions: extraOptions.MaxConcurrentCompactions,
 
-		// Per-level options. Options for at least one level must be specified. The
-		// options for the last level are used for all subsequent levels.
-		Levels: []pebble.LevelOptions{
-			{TargetFileSize: 2 * 1024 * 1024, FilterPolicy: bloom.FilterPolicy(10)},
-			{TargetFileSize: 2 * 1024 * 1024, FilterPolicy: bloom.FilterPolicy(10)},
-			{TargetFileSize: 2 * 1024 * 1024, FilterPolicy: bloom.FilterPolicy(10)},
-			{TargetFileSize: 2 * 1024 * 1024, FilterPolicy: bloom.FilterPolicy(10)},
-			{TargetFileSize: 2 * 1024 * 1024, FilterPolicy: bloom.FilterPolicy(10)},
-			{TargetFileSize: 2 * 1024 * 1024, FilterPolicy: bloom.FilterPolicy(10)},
-			{TargetFileSize: 2 * 1024 * 1024, FilterPolicy: bloom.FilterPolicy(10)},
-		},
+		// Per-level extraOptions. Options for at least one level must be specified. The
+		// extraOptions for the last level are used for all subsequent levels.
+		Levels:   levels,
 		ReadOnly: readonly,
 		EventListener: &pebble.EventListener{
 			CompactionBegin: db.onCompactionBegin,
@@ -221,11 +263,31 @@ func New(file string, cache int, handles int, namespace string, readonly bool, e
 			WriteStallEnd:   db.onWriteStallEnd,
 		},
 		Logger: panicLogger{}, // TODO(karalabe): Delete when this is upstreamed in Pebble
+
+		BytesPerSync:                extraOptions.BytesPerSync,
+		L0CompactionFileThreshold:   extraOptions.L0CompactionFileThreshold,
+		L0CompactionThreshold:       extraOptions.L0CompactionThreshold,
+		L0StopWritesThreshold:       extraOptions.L0StopWritesThreshold,
+		LBaseMaxBytes:               extraOptions.LBaseMaxBytes,
+		DisableAutomaticCompactions: extraOptions.DisableAutomaticCompactions,
+		WALBytesPerSync:             extraOptions.WALBytesPerSync,
+		WALDir:                      extraOptions.WALDir,
+		WALMinSyncInterval:          extraOptions.WALMinSyncInterval,
+		TargetByteDeletionRate:      extraOptions.TargetByteDeletionRate,
 	}
 	// Disable seek compaction explicitly. Check https://github.com/ethereum/go-ethereum/pull/20130
 	// for more details.
 	opt.Experimental.ReadSamplingMultiplier = -1
 
+	if opt.Experimental.ReadSamplingMultiplier != 0 {
+		opt.Experimental.ReadSamplingMultiplier = extraOptions.Experimental.ReadSamplingMultiplier
+	}
+	opt.Experimental.L0CompactionConcurrency = extraOptions.Experimental.L0CompactionConcurrency
+	opt.Experimental.CompactionDebtConcurrency = extraOptions.Experimental.CompactionDebtConcurrency
+	opt.Experimental.ReadCompactionRate = extraOptions.Experimental.ReadCompactionRate
+	opt.Experimental.MaxWriterConcurrency = extraOptions.Experimental.MaxWriterConcurrency
+	opt.Experimental.ForceWriterParallelism = extraOptions.Experimental.ForceWriterParallelism
+
 	// Open the db and recover any potential corruptions
 	innerDB, err := pebble.Open(file, opt)
 	if err != nil {
@@ -247,6 +309,17 @@ func New(file string, cache int, handles int, namespace string, readonly bool, e
 	db.seekCompGauge = metrics.GetOrRegisterGauge(namespace+"compact/seek", nil)
 	db.manualMemAllocGauge = metrics.GetOrRegisterGauge(namespace+"memory/manualalloc", nil)
 
+	db.compDebtGauge = metrics.GetOrRegisterGauge(namespace+"compact/debt", nil)
+	db.compInProgressGauge = metrics.GetOrRegisterGauge(namespace+"compact/inprogress", nil)
+
+	db.commitCountMeter = metrics.GetOrRegisterMeter(namespace+"commit/counter", nil)
+	db.commitTotalDurationMeter = metrics.GetOrRegisterMeter(namespace+"commit/duration/total", nil)
+	db.commitSemaphoreWaitMeter = metrics.GetOrRegisterMeter(namespace+"commit/duration/semaphorewait", nil)
+	db.commitMemTableWriteStallMeter = metrics.GetOrRegisterMeter(namespace+"commit/duration/memtablewritestall", nil)
+	db.commitL0ReadAmpWriteStallMeter = metrics.GetOrRegisterMeter(namespace+"commit/duration/l0readampwritestall", nil)
+	db.commitWALRotationMeter = metrics.GetOrRegisterMeter(namespace+"commit/duration/walrotation", nil)
+	db.commitWaitMeter = metrics.GetOrRegisterMeter(namespace+"commit/duration/commitwait", nil)
+
 	// Start up the metrics gathering and return
 	go db.meter(metricsGatheringInterval, namespace)
 	return db, nil
@@ -459,6 +532,14 @@ func (d *Database) meter(refresh time.Duration, namespace string) {
 		compReads        [2]int64
 
 		nWrites [2]int64
+
+		commitCounts               [2]int64
+		commitTotalDurations       [2]int64
+		commitSemaphoreWaits       [2]int64
+		commitMemTableWriteStalls  [2]int64
+		commitL0ReadAmpWriteStalls [2]int64
+		commitWALRotations         [2]int64
+		commitWaits                [2]int64
 	)
 
 	// Iterate ad infinitum and collect the stats
@@ -474,6 +555,14 @@ func (d *Database) meter(refresh time.Duration, namespace string) {
 			writeDelayTime     = d.writeDelayTime.Load()
 			nonLevel0CompCount = int64(d.nonLevel0Comp.Load())
 			level0CompCount    = int64(d.level0Comp.Load())
+
+			commitCount               = d.commitCount.Load()
+			commitTotalDuration       = d.commitTotalDuration.Load()
+			commitSemaphoreWait       = d.commitSemaphoreWait.Load()
+			commitMemTableWriteStall  = d.commitMemTableWriteStall.Load()
+			commitL0ReadAmpWriteStall = d.commitL0ReadAmpWriteStall.Load()
+			commitWALRotation         = d.commitWALRotation.Load()
+			commitWait                = d.commitWait.Load()
 		)
 		writeDelayTimes[i%2] = writeDelayTime
 		writeDelayCounts[i%2] = writeDelayCount
@@ -524,6 +613,25 @@ func (d *Database) meter(refresh time.Duration, namespace string) {
 		d.level0CompGauge.Update(level0CompCount)
 		d.seekCompGauge.Update(stats.Compact.ReadCount)
 
+		commitCounts[i%2] = commitCount
+		commitTotalDurations[i%2] = commitTotalDuration
+		commitSemaphoreWaits[i%2] = commitSemaphoreWait
+		commitMemTableWriteStalls[i%2] = commitMemTableWriteStall
+		commitL0ReadAmpWriteStalls[i%2] = commitL0ReadAmpWriteStall
+		commitWALRotations[i%2] = commitWALRotation
+		commitWaits[i%2] = commitWait
+
+		d.commitCountMeter.Mark(commitCounts[i%2] - commitCounts[(i-1)%2])
+		d.commitTotalDurationMeter.Mark(commitTotalDurations[i%2] - commitTotalDurations[(i-1)%2])
+		d.commitSemaphoreWaitMeter.Mark(commitSemaphoreWaits[i%2] - commitSemaphoreWaits[(i-1)%2])
+		d.commitMemTableWriteStallMeter.Mark(commitMemTableWriteStalls[i%2] - commitMemTableWriteStalls[(i-1)%2])
+		d.commitL0ReadAmpWriteStallMeter.Mark(commitL0ReadAmpWriteStalls[i%2] - commitL0ReadAmpWriteStalls[(i-1)%2])
+		d.commitWALRotationMeter.Mark(commitWALRotations[i%2] - commitWALRotations[(i-1)%2])
+		d.commitWaitMeter.Mark(commitWaits[i%2] - commitWaits[(i-1)%2])
+
+		d.compDebtGauge.Update(int64(stats.Compact.EstimatedDebt))
+		d.compInProgressGauge.Update(stats.Compact.NumInProgress)
+
 		for i, level := range stats.Levels {
 			// Append metrics for additional layers
 			if i >= len(d.levelsGauge) {
@@ -578,7 +686,20 @@ func (b *batch) Write() error {
 	if b.db.closed {
 		return pebble.ErrClosed
 	}
-	return b.b.Commit(b.db.writeOptions)
+	err := b.b.Commit(b.db.writeOptions)
+	if err != nil {
+		return err
+	}
+	stats := b.b.CommitStats()
+	b.db.commitCount.Add(1)
+	b.db.commitTotalDuration.Add(int64(stats.TotalDuration))
+	b.db.commitSemaphoreWait.Add(int64(stats.SemaphoreWaitDuration))
+	b.db.commitMemTableWriteStall.Add(int64(stats.MemTableWriteStallDuration))
+	b.db.commitL0ReadAmpWriteStall.Add(int64(stats.L0ReadAmpWriteStallDuration))
+	b.db.commitWALRotation.Add(int64(stats.WALRotationDuration))
+	b.db.commitWait.Add(int64(stats.CommitWaitDuration))
+	// TODO add metric for stats.WALQueueWaitDuration when it will be used by pebble (currently it is always 0)
+	return nil
 }
 
 // Reset resets the batch for reuse.

ethdb/pebble/pebble_non64bit.go

@@ -8,6 +8,6 @@ import (
 	"github.com/ethereum/go-ethereum/ethdb"
 )
 
-func New(file string, cache int, handles int, namespace string, readonly bool, ephemeral bool) (ethdb.Database, error) {
+func New(file string, cache int, handles int, namespace string, readonly bool, ephemeral bool, extraOptions *ExtraOptions) (ethdb.Database, error) {
 	return nil, errors.New("pebble is not supported on this platform")
 }

node/node.go

@@ -34,6 +34,7 @@ import (
 	"github.com/ethereum/go-ethereum/common/hexutil"
 	"github.com/ethereum/go-ethereum/core/rawdb"
 	"github.com/ethereum/go-ethereum/ethdb"
+	"github.com/ethereum/go-ethereum/ethdb/pebble"
 	"github.com/ethereum/go-ethereum/event"
 	"github.com/ethereum/go-ethereum/log"
 	"github.com/ethereum/go-ethereum/p2p"
@@ -746,6 +747,10 @@ func (n *Node) EventMux() *event.TypeMux {
 // previous can be found) from within the node's instance directory. If the node is
 // ephemeral, a memory database is returned.
 func (n *Node) OpenDatabase(name string, cache, handles int, namespace string, readonly bool) (ethdb.Database, error) {
+	return n.OpenDatabaseWithExtraOptions(name, cache, handles, namespace, readonly, nil)
+}
+
+func (n *Node) OpenDatabaseWithExtraOptions(name string, cache, handles int, namespace string, readonly bool, pebbleExtraOptions *pebble.ExtraOptions) (ethdb.Database, error) {
 	n.lock.Lock()
 	defer n.lock.Unlock()
 	if n.state == closedState {
@@ -758,12 +763,13 @@ func (n *Node) OpenDatabase(name string, cache, handles int, namespace string, r
 		db = rawdb.NewMemoryDatabase()
 	} else {
 		db, err = rawdb.Open(rawdb.OpenOptions{
-			Type:      n.config.DBEngine,
-			Directory: n.ResolvePath(name),
-			Namespace: namespace,
-			Cache:     cache,
-			Handles:   handles,
-			ReadOnly:  readonly,
+			Type:               n.config.DBEngine,
+			Directory:          n.ResolvePath(name),
+			Namespace:          namespace,
+			Cache:              cache,
+			Handles:            handles,
+			ReadOnly:           readonly,
+			PebbleExtraOptions: pebbleExtraOptions,
 		})
 	}
 
@@ -779,6 +785,10 @@ func (n *Node) OpenDatabase(name string, cache, handles int, namespace string, r
 // database to immutable append-only files. If the node is an ephemeral one, a
 // memory database is returned.
 func (n *Node) OpenDatabaseWithFreezer(name string, cache, handles int, ancient string, namespace string, readonly bool) (ethdb.Database, error) {
+	return n.OpenDatabaseWithFreezerWithExtraOptions(name, cache, handles, ancient, namespace, readonly, nil)
+}
+
+func (n *Node) OpenDatabaseWithFreezerWithExtraOptions(name string, cache, handles int, ancient string, namespace string, readonly bool, pebbleExtraOptions *pebble.ExtraOptions) (ethdb.Database, error) {
 	n.lock.Lock()
 	defer n.lock.Unlock()
 	if n.state == closedState {
@@ -790,13 +800,14 @@ func (n *Node) OpenDatabaseWithFreezer(name string, cache, handles int, ancient
 		db = rawdb.NewMemoryDatabase()
 	} else {
 		db, err = rawdb.Open(rawdb.OpenOptions{
-			Type:              n.config.DBEngine,
-			Directory:         n.ResolvePath(name),
-			AncientsDirectory: n.ResolveAncient(name, ancient),
-			Namespace:         namespace,
-			Cache:             cache,
-			Handles:           handles,
-			ReadOnly:          readonly,
+			Type:               n.config.DBEngine,
+			Directory:          n.ResolvePath(name),
+			AncientsDirectory:  n.ResolveAncient(name, ancient),
+			Namespace:          namespace,
+			Cache:              cache,
+			Handles:            handles,
+			ReadOnly:           readonly,
+			PebbleExtraOptions: pebbleExtraOptions,
 		})
 	}
 	if err == nil {