Initial merge from 1.8.11 version

VERSION

@@ -1 +1 @@
-1.8.10
+1.8.11

cmd/geth/main.go

@@ -19,9 +19,12 @@ package main
 
 import (
 	"fmt"
+	"math"
 	"os"
 	"runtime"
+	godebug "runtime/debug"
 	"sort"
+	"strconv"
 	"strings"
 	"time"
 
@@ -35,6 +38,7 @@ import (
 	"github.com/EthereumCommonwealth/go-callisto/log"
 	"github.com/EthereumCommonwealth/go-callisto/metrics"
 	"github.com/EthereumCommonwealth/go-callisto/node"
+	"github.com/elastic/gosigar"
 	"gopkg.in/urfave/cli.v1"
 )
 
@@ -188,6 +192,22 @@ func init() {
 		if err := debug.Setup(ctx); err != nil {
 			return err
 		}
+		// Cap the cache allowance and tune the garbage colelctor
+		var mem gosigar.Mem
+		if err := mem.Get(); err == nil {
+			allowance := int(mem.Total / 1024 / 1024 / 3)
+			if cache := ctx.GlobalInt(utils.CacheFlag.Name); cache > allowance {
+				log.Warn("Sanitizing cache to Go's GC limits", "provided", cache, "updated", allowance)
+				ctx.GlobalSet(utils.CacheFlag.Name, strconv.Itoa(allowance))
+			}
+		}
+		// Ensure Go's GC ignores the database cache for trigger percentage
+		cache := ctx.GlobalInt(utils.CacheFlag.Name)
+		gogc := math.Max(20, math.Min(100, 100/(float64(cache)/1024)))
+
+		log.Debug("Sanitizing Go's GC trigger", "percent", int(gogc))
+		godebug.SetGCPercent(int(gogc))
+
 		// Start system runtime metrics collection
 		go metrics.CollectProcessMetrics(3 * time.Second)
 

consensus/ethash/algorithm.go

@@ -94,14 +94,25 @@ func calcDatasetSize(epoch int) uint64 {
 // reused between hash runs instead of requiring new ones to be created.
 type hasher func(dest []byte, data []byte)
 
-// makeHasher creates a repetitive hasher, allowing the same hash data structures
+// makeHasher creates a repetitive hasher, allowing the same hash data structures to
-// to be reused between hash runs instead of requiring new ones to be created.
+// be reused between hash runs instead of requiring new ones to be created. The returned
-// The returned function is not thread safe!
+// function is not thread safe!
 func makeHasher(h hash.Hash) hasher {
+	// sha3.state supports Read to get the sum, use it to avoid the overhead of Sum.
+	// Read alters the state but we reset the hash before every operation.
+	type readerHash interface {
+		hash.Hash
+		Read([]byte) (int, error)
+	}
+	rh, ok := h.(readerHash)
+	if !ok {
+		panic("can't find Read method on hash")
+	}
+	outputLen := rh.Size()
 	return func(dest []byte, data []byte) {
-		h.Write(data)
+		rh.Reset()
-		h.Sum(dest[:0])
+		rh.Write(data)
-		h.Reset()
+		rh.Read(dest[:outputLen])
 	}
 }
 

console/bridge.go

@@ -271,7 +271,7 @@ func (b *bridge) SleepBlocks(call otto.FunctionCall) (response otto.Value) {
 }
 
 type jsonrpcCall struct {
-	Id     int64
+	ID     int64
 	Method string
 	Params []interface{}
 }
@@ -304,7 +304,7 @@ func (b *bridge) Send(call otto.FunctionCall) (response otto.Value) {
 	resps, _ := call.Otto.Object("new Array()")
 	for _, req := range reqs {
 		resp, _ := call.Otto.Object(`({"jsonrpc":"2.0"})`)
-		resp.Set("id", req.Id)
+		resp.Set("id", req.ID)
 		var result json.RawMessage
 		err = b.client.Call(&result, req.Method, req.Params...)
 		switch err := err.(type) {

console/console.go

@@ -73,6 +73,8 @@ type Console struct {
 	printer  io.Writer    // Output writer to serialize any display strings to
 }
 
+// New initializes a JavaScript interpreted runtime environment and sets defaults
+// with the config struct.
 func New(config Config) (*Console, error) {
 	// Handle unset config values gracefully
 	if config.Prompter == nil {

core/blockchain.go

@@ -674,7 +674,7 @@ func (bc *BlockChain) Stop() {
 		for !bc.triegc.Empty() {
 			triedb.Dereference(bc.triegc.PopItem().(common.Hash), common.Hash{})
 		}
-		if size := triedb.Size(); size != 0 {
+		if size, _ := triedb.Size(); size != 0 {
 			log.Error("Dangling trie nodes after full cleanup")
 		}
 	}
@@ -916,34 +916,30 @@ func (bc *BlockChain) WriteBlockWithState(block *types.Block, receipts []*types.
 		bc.triegc.Push(root, -float32(block.NumberU64()))
 
 		if current := block.NumberU64(); current > triesInMemory {
+			// If we exceeded our memory allowance, flush matured singleton nodes to disk
+			var (
+				nodes, imgs = triedb.Size()
+				limit       = common.StorageSize(bc.cacheConfig.TrieNodeLimit) * 1024 * 1024
+			)
+			if nodes > limit || imgs > 4*1024*1024 {
+				triedb.Cap(limit - ethdb.IdealBatchSize)
+			}
 			// Find the next state trie we need to commit
 			header := bc.GetHeaderByNumber(current - triesInMemory)
 			chosen := header.Number.Uint64()
 
-			// Only write to disk if we exceeded our memory allowance *and* also have at
+			// If we exceeded out time allowance, flush an entire trie to disk
-			// least a given number of tries gapped.
+			if bc.gcproc > bc.cacheConfig.TrieTimeLimit {
-			var (
-				size  = triedb.Size()
-				limit = common.StorageSize(bc.cacheConfig.TrieNodeLimit) * 1024 * 1024
-			)
-			if size > limit || bc.gcproc > bc.cacheConfig.TrieTimeLimit {
 				// If we're exceeding limits but haven't reached a large enough memory gap,
 				// warn the user that the system is becoming unstable.
-				if chosen < lastWrite+triesInMemory {
+				if chosen < lastWrite+triesInMemory && bc.gcproc >= 2*bc.cacheConfig.TrieTimeLimit {
-					switch {
-					case size >= 2*limit:
-						log.Warn("State memory usage too high, committing", "size", size, "limit", limit, "optimum", float64(chosen-lastWrite)/triesInMemory)
-					case bc.gcproc >= 2*bc.cacheConfig.TrieTimeLimit:
 					log.Info("State in memory for too long, committing", "time", bc.gcproc, "allowance", bc.cacheConfig.TrieTimeLimit, "optimum", float64(chosen-lastWrite)/triesInMemory)
 				}
-				}
+				// Flush an entire trie and restart the counters
-				// If optimum or critical limits reached, write to disk
-				if chosen >= lastWrite+triesInMemory || size >= 2*limit || bc.gcproc >= 2*bc.cacheConfig.TrieTimeLimit {
 				triedb.Commit(header.Root, true)
 				lastWrite = chosen
 				bc.gcproc = 0
 			}
-			}
 			// Garbage collect anything below our required write retention
 			for !bc.triegc.Empty() {
 				root, number := bc.triegc.Pop()
@@ -1181,7 +1177,9 @@ func (bc *BlockChain) insertChain(chain types.Blocks) (int, []interface{}, []*ty
 		}
 		stats.processed++
 		stats.usedGas += usedGas
-		stats.report(chain, i, bc.stateCache.TrieDB().Size())
+
+		cache, _ := bc.stateCache.TrieDB().Size()
+		stats.report(chain, i, cache)
 	}
 	// Append a single chain head event if we've progressed the chain
 	if lastCanon != nil && bc.CurrentBlock().Hash() == lastCanon.Hash() {

core/state/state_object.go

@@ -219,7 +219,7 @@ func (self *stateObject) updateRoot(db Database) {
 	self.data.Root = self.trie.Hash()
 }
 
-// CommitTrie the storage trie of the object to dwb.
+// CommitTrie the storage trie of the object to db.
 // This updates the trie root.
 func (self *stateObject) CommitTrie(db Database) error {
 	self.updateTrie(db)

eth/api_tracer.go

@@ -251,7 +251,8 @@ func (api *PrivateDebugAPI) traceChain(ctx context.Context, start, end *types.Bl
 			// Print progress logs if long enough time elapsed
 			if time.Since(logged) > 8*time.Second {
 				if number > origin {
-					log.Info("Tracing chain segment", "start", origin, "end", end.NumberU64(), "current", number, "transactions", traced, "elapsed", time.Since(begin), "memory", database.TrieDB().Size())
+					nodes, imgs := database.TrieDB().Size()
+					log.Info("Tracing chain segment", "start", origin, "end", end.NumberU64(), "current", number, "transactions", traced, "elapsed", time.Since(begin), "memory", nodes+imgs)
 				} else {
 					log.Info("Preparing state for chain trace", "block", number, "start", origin, "elapsed", time.Since(begin))
 				}
@@ -298,6 +299,8 @@ func (api *PrivateDebugAPI) traceChain(ctx context.Context, start, end *types.Bl
 			// Dereference all past tries we ourselves are done working with
 			database.TrieDB().Dereference(proot, common.Hash{})
 			proot = root
+
+			// TODO(karalabe): Do we need the preimages? Won't they accumulate too much?
 		}
 	}()
 
@@ -526,7 +529,8 @@ func (api *PrivateDebugAPI) computeStateDB(block *types.Block, reexec uint64) (*
 		database.TrieDB().Dereference(proot, common.Hash{})
 		proot = root
 	}
-	log.Info("Historical state regenerated", "block", block.NumberU64(), "elapsed", time.Since(start), "size", database.TrieDB().Size())
+	nodes, imgs := database.TrieDB().Size()
+	log.Info("Historical state regenerated", "block", block.NumberU64(), "elapsed", time.Since(start), "nodes", nodes, "preimages", imgs)
 	return statedb, nil
 }
 

eth/config.go

@@ -47,8 +47,8 @@ var DefaultConfig = Config{
 	LightPeers:    100,
 	DatabaseCache: 768,
 	TrieCache:     256,
-	TrieTimeout:   5 * time.Minute,
+	TrieTimeout:   60 * time.Minute,
-	GasPrice:      big.NewInt(1800 * params.Shannon),
+	GasPrice:      big.NewInt(18 * params.Shannon),
 
 	TxPool: core.DefaultTxPoolConfig,
 	GPO: gasprice.Config{

ethstats/ethstats.go

@@ -362,7 +362,7 @@ type nodeInfo struct {
 
 // authMsg is the authentication infos needed to login to a monitoring server.
 type authMsg struct {
-	Id     string   `json:"id"`
+	ID     string   `json:"id"`
 	Info   nodeInfo `json:"info"`
 	Secret string   `json:"secret"`
 }
@@ -381,7 +381,7 @@ func (s *Service) login(conn *websocket.Conn) error {
 		protocol = fmt.Sprintf("les/%d", les.ClientProtocolVersions[0])
 	}
 	auth := &authMsg{
-		Id: s.node,
+		ID: s.node,
 		Info: nodeInfo{
 			Name:     s.node,
 			Node:     infos.Name,

metrics/exp/exp.go

@@ -134,6 +134,17 @@ func (exp *exp) publishTimer(name string, metric metrics.Timer) {
 	exp.getFloat(name + ".mean-rate").Set(t.RateMean())
 }
 
+func (exp *exp) publishResettingTimer(name string, metric metrics.ResettingTimer) {
+	t := metric.Snapshot()
+	ps := t.Percentiles([]float64{50, 75, 95, 99})
+	exp.getInt(name + ".count").Set(int64(len(t.Values())))
+	exp.getFloat(name + ".mean").Set(t.Mean())
+	exp.getInt(name + ".50-percentile").Set(ps[0])
+	exp.getInt(name + ".75-percentile").Set(ps[1])
+	exp.getInt(name + ".95-percentile").Set(ps[2])
+	exp.getInt(name + ".99-percentile").Set(ps[3])
+}
+
 func (exp *exp) syncToExpvar() {
 	exp.registry.Each(func(name string, i interface{}) {
 		switch i.(type) {
@@ -149,6 +160,8 @@ func (exp *exp) syncToExpvar() {
 			exp.publishMeter(name, i.(metrics.Meter))
 		case metrics.Timer:
 			exp.publishTimer(name, i.(metrics.Timer))
+		case metrics.ResettingTimer:
+			exp.publishResettingTimer(name, i.(metrics.ResettingTimer))
 		default:
 			panic(fmt.Sprintf("unsupported type for '%s': %T", name, i))
 		}

metrics/resetting_timer.go

@@ -210,7 +210,7 @@ func (t *ResettingTimerSnapshot) calc(percentiles []float64) {
 				// poor man's math.Round(x):
 				// math.Floor(x + 0.5)
 				indexOfPerc := int(math.Floor(((abs / 100.0) * float64(count)) + 0.5))
-				if pct >= 0 {
+				if pct >= 0 && indexOfPerc > 0 {
 					indexOfPerc -= 1 // index offset=0
 				}
 				thresholdBoundary = t.values[indexOfPerc]

metrics/resetting_timer_test.go

@@ -104,3 +104,113 @@ func TestResettingTimer(t *testing.T) {
 		}
 	}
 }
+
+func TestResettingTimerWithFivePercentiles(t *testing.T) {
+	tests := []struct {
+		values   []int64
+		start    int
+		end      int
+		wantP05  int64
+		wantP20  int64
+		wantP50  int64
+		wantP95  int64
+		wantP99  int64
+		wantMean float64
+		wantMin  int64
+		wantMax  int64
+	}{
+		{
+			values:  []int64{},
+			start:   1,
+			end:     11,
+			wantP05: 1, wantP20: 2, wantP50: 5, wantP95: 10, wantP99: 10,
+			wantMin: 1, wantMax: 10, wantMean: 5.5,
+		},
+		{
+			values:  []int64{},
+			start:   1,
+			end:     101,
+			wantP05: 5, wantP20: 20, wantP50: 50, wantP95: 95, wantP99: 99,
+			wantMin: 1, wantMax: 100, wantMean: 50.5,
+		},
+		{
+			values:  []int64{1},
+			start:   0,
+			end:     0,
+			wantP05: 1, wantP20: 1, wantP50: 1, wantP95: 1, wantP99: 1,
+			wantMin: 1, wantMax: 1, wantMean: 1,
+		},
+		{
+			values:  []int64{0},
+			start:   0,
+			end:     0,
+			wantP05: 0, wantP20: 0, wantP50: 0, wantP95: 0, wantP99: 0,
+			wantMin: 0, wantMax: 0, wantMean: 0,
+		},
+		{
+			values:  []int64{},
+			start:   0,
+			end:     0,
+			wantP05: 0, wantP20: 0, wantP50: 0, wantP95: 0, wantP99: 0,
+			wantMin: 0, wantMax: 0, wantMean: 0,
+		},
+		{
+			values:  []int64{1, 10},
+			start:   0,
+			end:     0,
+			wantP05: 1, wantP20: 1, wantP50: 1, wantP95: 10, wantP99: 10,
+			wantMin: 1, wantMax: 10, wantMean: 5.5,
+		},
+	}
+	for ind, tt := range tests {
+		timer := NewResettingTimer()
+
+		for i := tt.start; i < tt.end; i++ {
+			tt.values = append(tt.values, int64(i))
+		}
+
+		for _, v := range tt.values {
+			timer.Update(time.Duration(v))
+		}
+
+		snap := timer.Snapshot()
+
+		ps := snap.Percentiles([]float64{5, 20, 50, 95, 99})
+
+		val := snap.Values()
+
+		if len(val) > 0 {
+			if tt.wantMin != val[0] {
+				t.Fatalf("%d: min: got %d, want %d", ind, val[0], tt.wantMin)
+			}
+
+			if tt.wantMax != val[len(val)-1] {
+				t.Fatalf("%d: max: got %d, want %d", ind, val[len(val)-1], tt.wantMax)
+			}
+		}
+
+		if tt.wantMean != snap.Mean() {
+			t.Fatalf("%d: mean: got %.2f, want %.2f", ind, snap.Mean(), tt.wantMean)
+		}
+
+		if tt.wantP05 != ps[0] {
+			t.Fatalf("%d: p05: got %d, want %d", ind, ps[0], tt.wantP05)
+		}
+
+		if tt.wantP20 != ps[1] {
+			t.Fatalf("%d: p20: got %d, want %d", ind, ps[1], tt.wantP20)
+		}
+
+		if tt.wantP50 != ps[2] {
+			t.Fatalf("%d: p50: got %d, want %d", ind, ps[2], tt.wantP50)
+		}
+
+		if tt.wantP95 != ps[3] {
+			t.Fatalf("%d: p95: got %d, want %d", ind, ps[3], tt.wantP95)
+		}
+
+		if tt.wantP99 != ps[4] {
+			t.Fatalf("%d: p99: got %d, want %d", ind, ps[4], tt.wantP99)
+		}
+	}
+}

node/api.go

@@ -338,6 +338,21 @@ func (api *PublicDebugAPI) Metrics(raw bool) (map[string]interface{}, error) {
 					},
 				}
 
+			case metrics.ResettingTimer:
+				t := metric.Snapshot()
+				ps := t.Percentiles([]float64{5, 20, 50, 80, 95})
+				root[name] = map[string]interface{}{
+					"Measurements": len(t.Values()),
+					"Mean":         time.Duration(t.Mean()).String(),
+					"Percentiles": map[string]interface{}{
+						"5":  time.Duration(ps[0]).String(),
+						"20": time.Duration(ps[1]).String(),
+						"50": time.Duration(ps[2]).String(),
+						"80": time.Duration(ps[3]).String(),
+						"95": time.Duration(ps[4]).String(),
+					},
+				}
+
 			default:
 				root[name] = "Unknown metric type"
 			}
@@ -373,6 +388,21 @@ func (api *PublicDebugAPI) Metrics(raw bool) (map[string]interface{}, error) {
 					},
 				}
 
+			case metrics.ResettingTimer:
+				t := metric.Snapshot()
+				ps := t.Percentiles([]float64{5, 20, 50, 80, 95})
+				root[name] = map[string]interface{}{
+					"Measurements": len(t.Values()),
+					"Mean":         time.Duration(t.Mean()).String(),
+					"Percentiles": map[string]interface{}{
+						"5":  time.Duration(ps[0]).String(),
+						"20": time.Duration(ps[1]).String(),
+						"50": time.Duration(ps[2]).String(),
+						"80": time.Duration(ps[3]).String(),
+						"95": time.Duration(ps[4]).String(),
+					},
+				}
+
 			default:
 				root[name] = "Unknown metric type"
 			}

params/version.go

@@ -23,7 +23,7 @@ import (
 const (
 	VersionMajor = 1            // Major version component of the current release
 	VersionMinor = 8            // Minor version component of the current release
-	VersionPatch = 10           // Patch version component of the current release
+	VersionPatch = 11           // Patch version component of the current release
 	VersionMeta  = "CLO stable" // Version metadata to append to the version string
 )
 

rpc/http.go

@@ -165,7 +165,12 @@ func NewHTTPServer(cors []string, vhosts []string, srv *Server) *http.Server {
 	// Wrap the CORS-handler within a host-handler
 	handler := newCorsHandler(srv, cors)
 	handler = newVHostHandler(vhosts, handler)
-	return &http.Server{Handler: handler}
+	return &http.Server{
+		Handler:      handler,
+		ReadTimeout:  5 * time.Second,
+		WriteTimeout: 10 * time.Second,
+		IdleTimeout:  120 * time.Second,
+	}
 }
 
 // ServeHTTP serves JSON-RPC requests over HTTP.
@@ -181,7 +186,7 @@ func (srv *Server) ServeHTTP(w http.ResponseWriter, r *http.Request) {
 	// All checks passed, create a codec that reads direct from the request body
 	// untilEOF and writes the response to w and order the server to process a
 	// single request.
-	ctx := context.Background()
+	ctx := r.Context()
 	ctx = context.WithValue(ctx, "remote", r.RemoteAddr)
 	ctx = context.WithValue(ctx, "scheme", r.Proto)
 	ctx = context.WithValue(ctx, "local", r.Host)

trie/database.go

@@ -23,6 +23,21 @@ import (
 	"github.com/EthereumCommonwealth/go-callisto/common"
 	"github.com/EthereumCommonwealth/go-callisto/ethdb"
 	"github.com/EthereumCommonwealth/go-callisto/log"
+	"github.com/EthereumCommonwealth/go-callisto/metrics"
+)
+
+var (
+	memcacheFlushTimeTimer  = metrics.NewRegisteredResettingTimer("trie/memcache/flush/time", nil)
+	memcacheFlushNodesMeter = metrics.NewRegisteredMeter("trie/memcache/flush/nodes", nil)
+	memcacheFlushSizeMeter  = metrics.NewRegisteredMeter("trie/memcache/flush/size", nil)
+
+	memcacheGCTimeTimer  = metrics.NewRegisteredResettingTimer("trie/memcache/gc/time", nil)
+	memcacheGCNodesMeter = metrics.NewRegisteredMeter("trie/memcache/gc/nodes", nil)
+	memcacheGCSizeMeter  = metrics.NewRegisteredMeter("trie/memcache/gc/size", nil)
+
+	memcacheCommitTimeTimer  = metrics.NewRegisteredResettingTimer("trie/memcache/commit/time", nil)
+	memcacheCommitNodesMeter = metrics.NewRegisteredMeter("trie/memcache/commit/nodes", nil)
+	memcacheCommitSizeMeter  = metrics.NewRegisteredMeter("trie/memcache/commit/size", nil)
 )
 
 // secureKeyPrefix is the database key prefix used to store trie node preimages.
@@ -47,6 +62,9 @@ type Database struct {
 	diskdb ethdb.Database // Persistent storage for matured trie nodes
 
 	nodes  map[common.Hash]*cachedNode // Data and references relationships of a node
+	oldest common.Hash                 // Oldest tracked node, flush-list head
+	newest common.Hash                 // Newest tracked node, flush-list tail
+
 	preimages map[common.Hash][]byte // Preimages of nodes from the secure trie
 	seckeybuf [secureKeyLength]byte  // Ephemeral buffer for calculating preimage keys
 
@@ -54,7 +72,11 @@ type Database struct {
 	gcnodes uint64             // Nodes garbage collected since last commit
 	gcsize  common.StorageSize // Data storage garbage collected since last commit
 
-	nodesSize     common.StorageSize // Storage size of the nodes cache
+	flushtime  time.Duration      // Time spent on data flushing since last commit
+	flushnodes uint64             // Nodes flushed since last commit
+	flushsize  common.StorageSize // Data storage flushed since last commit
+
+	nodesSize     common.StorageSize // Storage size of the nodes cache (exc. flushlist)
 	preimagesSize common.StorageSize // Storage size of the preimages cache
 
 	lock sync.RWMutex
@@ -66,6 +88,9 @@ type cachedNode struct {
 	blob     []byte              // Cached data block of the trie node
 	parents  int                 // Number of live nodes referencing this one
 	children map[common.Hash]int // Children referenced by this nodes
+
+	flushPrev common.Hash // Previous node in the flush-list
+	flushNext common.Hash // Next node in the flush-list
 }
 
 // NewDatabase creates a new trie database to store ephemeral trie content before
@@ -96,12 +121,20 @@ func (db *Database) Insert(hash common.Hash, blob []byte) {
 
 // insert is the private locked version of Insert.
 func (db *Database) insert(hash common.Hash, blob []byte) {
+	// If the node's already cached, skip
 	if _, ok := db.nodes[hash]; ok {
 		return
 	}
 	db.nodes[hash] = &cachedNode{
 		blob:      common.CopyBytes(blob),
 		children:  make(map[common.Hash]int),
+		flushPrev: db.newest,
+	}
+	// Update the flush-list endpoints
+	if db.oldest == (common.Hash{}) {
+		db.oldest, db.newest = hash, hash
+	} else {
+		db.nodes[db.newest].flushNext, db.newest = hash, hash
 	}
 	db.nodesSize += common.StorageSize(common.HashLength + len(blob))
 }
@@ -208,6 +241,10 @@ func (db *Database) Dereference(child common.Hash, parent common.Hash) {
 	db.gcsize += storage - db.nodesSize
 	db.gctime += time.Since(start)
 
+	memcacheGCTimeTimer.Update(time.Since(start))
+	memcacheGCSizeMeter.Mark(int64(storage - db.nodesSize))
+	memcacheGCNodesMeter.Mark(int64(nodes - len(db.nodes)))
+
 	log.Debug("Dereferenced trie from memory database", "nodes", nodes-len(db.nodes), "size", storage-db.nodesSize, "time", time.Since(start),
 		"gcnodes", db.gcnodes, "gcsize", db.gcsize, "gctime", db.gctime, "livenodes", len(db.nodes), "livesize", db.nodesSize)
 }
@@ -221,7 +258,7 @@ func (db *Database) dereference(child common.Hash, parent common.Hash) {
 	if node.children[child] == 0 {
 		delete(node.children, child)
 	}
-	// If the node does not exist, it's a previously committed node.
+	// If the child does not exist, it's a previously committed node.
 	node, ok := db.nodes[child]
 	if !ok {
 		return
@@ -229,6 +266,14 @@ func (db *Database) dereference(child common.Hash, parent common.Hash) {
 	// If there are no more references to the child, delete it and cascade
 	node.parents--
 	if node.parents == 0 {
+		// Remove the node from the flush-list
+		if child == db.oldest {
+			db.oldest = node.flushNext
+		} else {
+			db.nodes[node.flushPrev].flushNext = node.flushNext
+			db.nodes[node.flushNext].flushPrev = node.flushPrev
+		}
+		// Dereference all children and delete the node
 		for hash := range node.children {
 			db.dereference(hash, child)
 		}
@@ -237,6 +282,107 @@ func (db *Database) dereference(child common.Hash, parent common.Hash) {
 	}
 }
 
+// Cap iteratively flushes old but still referenced trie nodes until the total
+// memory usage goes below the given threshold.
+func (db *Database) Cap(limit common.StorageSize) error {
+	// Create a database batch to flush persistent data out. It is important that
+	// outside code doesn't see an inconsistent state (referenced data removed from
+	// memory cache during commit but not yet in persistent storage). This is ensured
+	// by only uncaching existing data when the database write finalizes.
+	db.lock.RLock()
+
+	nodes, storage, start := len(db.nodes), db.nodesSize, time.Now()
+	batch := db.diskdb.NewBatch()
+
+	// db.nodesSize only contains the useful data in the cache, but when reporting
+	// the total memory consumption, the maintenance metadata is also needed to be
+	// counted. For every useful node, we track 2 extra hashes as the flushlist.
+	size := db.nodesSize + common.StorageSize(len(db.nodes)*2*common.HashLength)
+
+	// If the preimage cache got large enough, push to disk. If it's still small
+	// leave for later to deduplicate writes.
+	flushPreimages := db.preimagesSize > 4*1024*1024
+	if flushPreimages {
+		for hash, preimage := range db.preimages {
+			if err := batch.Put(db.secureKey(hash[:]), preimage); err != nil {
+				log.Error("Failed to commit preimage from trie database", "err", err)
+				db.lock.RUnlock()
+				return err
+			}
+			if batch.ValueSize() > ethdb.IdealBatchSize {
+				if err := batch.Write(); err != nil {
+					db.lock.RUnlock()
+					return err
+				}
+				batch.Reset()
+			}
+		}
+	}
+	// Keep committing nodes from the flush-list until we're below allowance
+	oldest := db.oldest
+	for size > limit && oldest != (common.Hash{}) {
+		// Fetch the oldest referenced node and push into the batch
+		node := db.nodes[oldest]
+		if err := batch.Put(oldest[:], node.blob); err != nil {
+			db.lock.RUnlock()
+			return err
+		}
+		// If we exceeded the ideal batch size, commit and reset
+		if batch.ValueSize() >= ethdb.IdealBatchSize {
+			if err := batch.Write(); err != nil {
+				log.Error("Failed to write flush list to disk", "err", err)
+				db.lock.RUnlock()
+				return err
+			}
+			batch.Reset()
+		}
+		// Iterate to the next flush item, or abort if the size cap was achieved. Size
+		// is the total size, including both the useful cached data (hash -> blob), as
+		// well as the flushlist metadata (2*hash). When flushing items from the cache,
+		// we need to reduce both.
+		size -= common.StorageSize(3*common.HashLength + len(node.blob))
+		oldest = node.flushNext
+	}
+	// Flush out any remainder data from the last batch
+	if err := batch.Write(); err != nil {
+		log.Error("Failed to write flush list to disk", "err", err)
+		db.lock.RUnlock()
+		return err
+	}
+	db.lock.RUnlock()
+
+	// Write successful, clear out the flushed data
+	db.lock.Lock()
+	defer db.lock.Unlock()
+
+	if flushPreimages {
+		db.preimages = make(map[common.Hash][]byte)
+		db.preimagesSize = 0
+	}
+	for db.oldest != oldest {
+		node := db.nodes[db.oldest]
+		delete(db.nodes, db.oldest)
+		db.oldest = node.flushNext
+
+		db.nodesSize -= common.StorageSize(common.HashLength + len(node.blob))
+	}
+	if db.oldest != (common.Hash{}) {
+		db.nodes[db.oldest].flushPrev = common.Hash{}
+	}
+	db.flushnodes += uint64(nodes - len(db.nodes))
+	db.flushsize += storage - db.nodesSize
+	db.flushtime += time.Since(start)
+
+	memcacheFlushTimeTimer.Update(time.Since(start))
+	memcacheFlushSizeMeter.Mark(int64(storage - db.nodesSize))
+	memcacheFlushNodesMeter.Mark(int64(nodes - len(db.nodes)))
+
+	log.Debug("Persisted nodes from memory database", "nodes", nodes-len(db.nodes), "size", storage-db.nodesSize, "time", time.Since(start),
+		"flushnodes", db.flushnodes, "flushsize", db.flushsize, "flushtime", db.flushtime, "livenodes", len(db.nodes), "livesize", db.nodesSize)
+
+	return nil
+}
+
 // Commit iterates over all the children of a particular node, writes them out
 // to disk, forcefully tearing down all references in both directions.
 //
@@ -266,7 +412,7 @@ func (db *Database) Commit(node common.Hash, report bool) error {
 		}
 	}
 	// Move the trie itself into the batch, flushing if enough data is accumulated
-	nodes, storage := len(db.nodes), db.nodesSize+db.preimagesSize
+	nodes, storage := len(db.nodes), db.nodesSize
 	if err := db.commit(node, batch); err != nil {
 		log.Error("Failed to commit trie from trie database", "err", err)
 		db.lock.RUnlock()
@@ -289,15 +435,20 @@ func (db *Database) Commit(node common.Hash, report bool) error {
 
 	db.uncache(node)
 
+	memcacheCommitTimeTimer.Update(time.Since(start))
+	memcacheCommitSizeMeter.Mark(int64(storage - db.nodesSize))
+	memcacheCommitNodesMeter.Mark(int64(nodes - len(db.nodes)))
+
 	logger := log.Info
 	if !report {
 		logger = log.Debug
 	}
-	logger("Persisted trie from memory database", "nodes", nodes-len(db.nodes), "size", storage-db.nodesSize, "time", time.Since(start),
+	logger("Persisted trie from memory database", "nodes", nodes-len(db.nodes)+int(db.flushnodes), "size", storage-db.nodesSize+db.flushsize, "time", time.Since(start)+db.flushtime,
 		"gcnodes", db.gcnodes, "gcsize", db.gcsize, "gctime", db.gctime, "livenodes", len(db.nodes), "livesize", db.nodesSize)
 
 	// Reset the garbage collection statistics
 	db.gcnodes, db.gcsize, db.gctime = 0, 0, 0
+	db.flushnodes, db.flushsize, db.flushtime = 0, 0, 0
 
 	return nil
 }
@@ -317,7 +468,7 @@ func (db *Database) commit(hash common.Hash, batch ethdb.Batch) error {
 	if err := batch.Put(hash[:], node.blob); err != nil {
 		return err
 	}
-	// If we've reached an optimal match size, commit and start over
+	// If we've reached an optimal batch size, commit and start over
 	if batch.ValueSize() >= ethdb.IdealBatchSize {
 		if err := batch.Write(); err != nil {
 			return err
@@ -337,7 +488,14 @@ func (db *Database) uncache(hash common.Hash) {
 	if !ok {
 		return
 	}
-	// Otherwise uncache the node's subtries and remove the node itself too
+	// Node still exists, remove it from the flush-list
+	if hash == db.oldest {
+		db.oldest = node.flushNext
+	} else {
+		db.nodes[node.flushPrev].flushNext = node.flushNext
+		db.nodes[node.flushNext].flushPrev = node.flushPrev
+	}
+	// Uncache the node's subtries and remove the node itself too
 	for child := range node.children {
 		db.uncache(child)
 	}
@@ -347,9 +505,13 @@ func (db *Database) uncache(hash common.Hash) {
 
 // Size returns the current storage size of the memory cache in front of the
 // persistent database layer.
-func (db *Database) Size() common.StorageSize {
+func (db *Database) Size() (common.StorageSize, common.StorageSize) {
 	db.lock.RLock()
 	defer db.lock.RUnlock()
 
-	return db.nodesSize + db.preimagesSize
+	// db.nodesSize only contains the useful data in the cache, but when reporting
+	// the total memory consumption, the maintenance metadata is also needed to be
+	// counted. For every useful node, we track 2 extra hashes as the flushlist.
+	var flushlistSize = common.StorageSize(len(db.nodes) * 2 * common.HashLength)
+	return db.nodesSize + flushlistSize, db.preimagesSize
 }