Revert "cmd, core, eth, graphql, trie: no persisted clean trie cache file (#27525)"

This reverts commit 2c0c010df5.

cmd/devp2p/internal/ethtest/suite_test.go

@@ -113,6 +113,8 @@ func setupGeth(stack *node.Node) error {
 		NetworkId:               chain.genesis.Config.ChainID.Uint64(), // 19763
 		DatabaseCache:           10,
 		TrieCleanCache:          10,
+		TrieCleanCacheJournal:   "",
+		TrieCleanCacheRejournal: 60 * time.Minute,
 		TrieDirtyCache:          16,
 		TrieTimeout:             60 * time.Minute,
 		SnapshotCache:           10,

cmd/geth/config.go

@@ -276,10 +276,6 @@ func deprecated(field string) bool {
 		return true
 	case "ethconfig.Config.EWASMInterpreter":
 		return true
-	case "ethconfig.Config.TrieCleanCacheJournal":
-		return true
-	case "ethconfig.Config.TrieCleanCacheRejournal":
-		return true
 	default:
 		return false
 	}

cmd/geth/snapshot.go

@@ -50,6 +50,7 @@ var (
 				ArgsUsage: "<root>",
 				Action:    pruneState,
 				Flags: flags.Merge([]cli.Flag{
+					utils.CacheTrieJournalFlag,
 					utils.BloomFilterSizeFlag,
 				}, utils.NetworkFlags, utils.DatabasePathFlags),
 				Description: `
@@ -159,7 +160,7 @@ block is used.
 // Deprecation: this command should be deprecated once the hash-based
 // scheme is deprecated.
 func pruneState(ctx *cli.Context) error {
-	stack, _ := makeConfigNode(ctx)
+	stack, config := makeConfigNode(ctx)
 	defer stack.Close()
 
 	chaindb := utils.MakeChainDatabase(ctx, stack, false)
@@ -167,6 +168,7 @@ func pruneState(ctx *cli.Context) error {
 
 	prunerconfig := pruner.Config{
 		Datadir:   stack.ResolvePath(""),
+		Cachedir:  stack.ResolvePath(config.Eth.TrieCleanCacheJournal),
 		BloomSize: ctx.Uint64(utils.BloomFilterSizeFlag.Name),
 	}
 	pruner, err := pruner.NewPruner(chaindb, prunerconfig)

cmd/utils/flags.go

@@ -410,6 +410,18 @@ var (
 		Value:    15,
 		Category: flags.PerfCategory,
 	}
+	CacheTrieJournalFlag = &cli.StringFlag{
+		Name:     "cache.trie.journal",
+		Usage:    "Disk journal directory for trie cache to survive node restarts",
+		Value:    ethconfig.Defaults.TrieCleanCacheJournal,
+		Category: flags.PerfCategory,
+	}
+	CacheTrieRejournalFlag = &cli.DurationFlag{
+		Name:     "cache.trie.rejournal",
+		Usage:    "Time interval to regenerate the trie cache journal",
+		Value:    ethconfig.Defaults.TrieCleanCacheRejournal,
+		Category: flags.PerfCategory,
+	}
 	CacheGCFlag = &cli.IntFlag{
 		Name:     "cache.gc",
 		Usage:    "Percentage of cache memory allowance to use for trie pruning (default = 25% full mode, 0% archive mode)",
@@ -1698,6 +1710,12 @@ func SetEthConfig(ctx *cli.Context, stack *node.Node, cfg *ethconfig.Config) {
 	if ctx.IsSet(CacheFlag.Name) || ctx.IsSet(CacheTrieFlag.Name) {
 		cfg.TrieCleanCache = ctx.Int(CacheFlag.Name) * ctx.Int(CacheTrieFlag.Name) / 100
 	}
+	if ctx.IsSet(CacheTrieJournalFlag.Name) {
+		cfg.TrieCleanCacheJournal = ctx.String(CacheTrieJournalFlag.Name)
+	}
+	if ctx.IsSet(CacheTrieRejournalFlag.Name) {
+		cfg.TrieCleanCacheRejournal = ctx.Duration(CacheTrieRejournalFlag.Name)
+	}
 	if ctx.IsSet(CacheFlag.Name) || ctx.IsSet(CacheGCFlag.Name) {
 		cfg.TrieDirtyCache = ctx.Int(CacheFlag.Name) * ctx.Int(CacheGCFlag.Name) / 100
 	}

cmd/utils/flags_legacy.go

@@ -33,8 +33,6 @@ var ShowDeprecated = &cli.Command{
 
 var DeprecatedFlags = []cli.Flag{
 	NoUSBFlag,
-	CacheTrieJournalFlag,
-	CacheTrieRejournalFlag,
 }
 
 var (
@@ -44,17 +42,6 @@ var (
 		Usage:    "Disables monitoring for and managing USB hardware wallets (deprecated)",
 		Category: flags.DeprecatedCategory,
 	}
-	// (Deprecated June 2023, shown in aliased flags section)
-	CacheTrieJournalFlag = &cli.StringFlag{
-		Name:     "cache.trie.journal",
-		Usage:    "Disk journal directory for trie cache to survive node restarts",
-		Category: flags.PerfCategory,
-	}
-	CacheTrieRejournalFlag = &cli.DurationFlag{
-		Name:     "cache.trie.rejournal",
-		Usage:    "Time interval to regenerate the trie cache journal",
-		Category: flags.PerfCategory,
-	}
 )
 
 // showDeprecated displays deprecated flags that will be soon removed from the codebase.

core/blockchain.go

@@ -130,6 +130,8 @@ const (
 // that's resident in a blockchain.
 type CacheConfig struct {
 	TrieCleanLimit      int           // Memory allowance (MB) to use for caching trie nodes in memory
+	TrieCleanJournal    string        // Disk journal for saving clean cache entries.
+	TrieCleanRejournal  time.Duration // Time interval to dump clean cache to disk periodically
 	TrieCleanNoPrefetch bool          // Whether to disable heuristic state prefetching for followup blocks
 	TrieDirtyLimit      int           // Memory limit (MB) at which to start flushing dirty trie nodes to disk
 	TrieDirtyDisabled   bool          // Whether to disable trie write caching and GC altogether (archive node)
@@ -236,6 +238,7 @@ func NewBlockChain(db ethdb.Database, cacheConfig *CacheConfig, genesis *Genesis
 	// Open trie database with provided config
 	triedb := trie.NewDatabaseWithConfig(db, &trie.Config{
 		Cache:     cacheConfig.TrieCleanLimit,
+		Journal:   cacheConfig.TrieCleanJournal,
 		Preimages: cacheConfig.Preimages,
 	})
 	// Setup the genesis block, commit the provided genesis specification
@@ -408,6 +411,18 @@ func NewBlockChain(db ethdb.Database, cacheConfig *CacheConfig, genesis *Genesis
 	bc.wg.Add(1)
 	go bc.updateFutureBlocks()
 
+	// If periodic cache journal is required, spin it up.
+	if bc.cacheConfig.TrieCleanRejournal > 0 {
+		if bc.cacheConfig.TrieCleanRejournal < time.Minute {
+			log.Warn("Sanitizing invalid trie cache journal time", "provided", bc.cacheConfig.TrieCleanRejournal, "updated", time.Minute)
+			bc.cacheConfig.TrieCleanRejournal = time.Minute
+		}
+		bc.wg.Add(1)
+		go func() {
+			defer bc.wg.Done()
+			bc.triedb.SaveCachePeriodically(bc.cacheConfig.TrieCleanJournal, bc.cacheConfig.TrieCleanRejournal, bc.quit)
+		}()
+	}
 	// Rewind the chain in case of an incompatible config upgrade.
 	if compat, ok := genesisErr.(*params.ConfigCompatError); ok {
 		log.Warn("Rewinding chain to upgrade configuration", "err", compat)
@@ -972,6 +987,11 @@ func (bc *BlockChain) Stop() {
 	if err := bc.stateCache.TrieDB().Close(); err != nil {
 		log.Error("Failed to close trie db", "err", err)
 	}
+	// Ensure all live cached entries be saved into disk, so that we can skip
+	// cache warmup when node restarts.
+	if bc.cacheConfig.TrieCleanJournal != "" {
+		bc.triedb.SaveCache(bc.cacheConfig.TrieCleanJournal)
+	}
 	log.Info("Blockchain stopped")
 }
 

core/state/pruner/pruner.go

@@ -57,6 +57,7 @@ const (
 // Config includes all the configurations for pruning.
 type Config struct {
 	Datadir   string // The directory of the state database
+	Cachedir  string // The directory of state clean cache
 	BloomSize uint64 // The Megabytes of memory allocated to bloom-filter
 }
 
@@ -240,7 +241,7 @@ func (p *Pruner) Prune(root common.Hash) error {
 		return err
 	}
 	if stateBloomRoot != (common.Hash{}) {
-		return RecoverPruning(p.config.Datadir, p.db)
+		return RecoverPruning(p.config.Datadir, p.db, p.config.Cachedir)
 	}
 	// If the target state root is not specified, use the HEAD-127 as the
 	// target. The reason for picking it is:
@@ -298,6 +299,12 @@ func (p *Pruner) Prune(root common.Hash) error {
 			log.Info("Selecting user-specified state as the pruning target", "root", root)
 		}
 	}
+	// Before start the pruning, delete the clean trie cache first.
+	// It's necessary otherwise in the next restart we will hit the
+	// deleted state root in the "clean cache" so that the incomplete
+	// state is picked for usage.
+	deleteCleanTrieCache(p.config.Cachedir)
+
 	// All the state roots of the middle layer should be forcibly pruned,
 	// otherwise the dangling state will be left.
 	middleRoots := make(map[common.Hash]struct{})
@@ -335,7 +342,7 @@ func (p *Pruner) Prune(root common.Hash) error {
 // pruning can be resumed. What's more if the bloom filter is constructed, the
 // pruning **has to be resumed**. Otherwise a lot of dangling nodes may be left
 // in the disk.
-func RecoverPruning(datadir string, db ethdb.Database) error {
+func RecoverPruning(datadir string, db ethdb.Database, trieCachePath string) error {
 	stateBloomPath, stateBloomRoot, err := findBloomFilter(datadir)
 	if err != nil {
 		return err
@@ -371,6 +378,12 @@ func RecoverPruning(datadir string, db ethdb.Database) error {
 	}
 	log.Info("Loaded state bloom filter", "path", stateBloomPath)
 
+	// Before start the pruning, delete the clean trie cache first.
+	// It's necessary otherwise in the next restart we will hit the
+	// deleted state root in the "clean cache" so that the incomplete
+	// state is picked for usage.
+	deleteCleanTrieCache(trieCachePath)
+
 	// All the state roots of the middle layers should be forcibly pruned,
 	// otherwise the dangling state will be left.
 	var (
@@ -484,3 +497,23 @@ func findBloomFilter(datadir string) (string, common.Hash, error) {
 	}
 	return stateBloomPath, stateBloomRoot, nil
 }
+
+const warningLog = `
+
+WARNING!
+
+The clean trie cache is not found. Please delete it by yourself after the 
+pruning. Remember don't start the Geth without deleting the clean trie cache
+otherwise the entire database may be damaged!
+
+Check the command description "geth snapshot prune-state --help" for more details.
+`
+
+func deleteCleanTrieCache(path string) {
+	if !common.FileExist(path) {
+		log.Warn(warningLog)
+		return
+	}
+	os.RemoveAll(path)
+	log.Info("Deleted trie clean cache", "path", path)
+}

eth/backend.go

@@ -132,7 +132,7 @@ func New(stack *node.Node, config *ethconfig.Config) (*Ethereum, error) {
 	if err != nil {
 		return nil, err
 	}
-	if err := pruner.RecoverPruning(stack.ResolvePath(""), chainDb); err != nil {
+	if err := pruner.RecoverPruning(stack.ResolvePath(""), chainDb, stack.ResolvePath(config.TrieCleanCacheJournal)); err != nil {
 		log.Error("Failed to recover state", "error", err)
 	}
 	// Transfer mining-related config to the ethash config.
@@ -184,6 +184,8 @@ func New(stack *node.Node, config *ethconfig.Config) (*Ethereum, error) {
 		}
 		cacheConfig = &core.CacheConfig{
 			TrieCleanLimit:      config.TrieCleanCache,
+			TrieCleanJournal:    stack.ResolvePath(config.TrieCleanCacheJournal),
+			TrieCleanRejournal:  config.TrieCleanCacheRejournal,
 			TrieCleanNoPrefetch: config.NoPrefetch,
 			TrieDirtyLimit:      config.TrieDirtyCache,
 			TrieDirtyDisabled:   config.NoPruning,

eth/ethconfig/config.go

@@ -64,6 +64,8 @@ var Defaults = Config{
 	UltraLightFraction:      75,
 	DatabaseCache:           512,
 	TrieCleanCache:          154,
+	TrieCleanCacheJournal:   "triecache",
+	TrieCleanCacheRejournal: 60 * time.Minute,
 	TrieDirtyCache:          256,
 	TrieTimeout:             60 * time.Minute,
 	SnapshotCache:           102,
@@ -123,6 +125,8 @@ type Config struct {
 	DatabaseFreezer    string
 
 	TrieCleanCache          int
+	TrieCleanCacheJournal   string        `toml:",omitempty"` // Disk journal directory for trie cache to survive node restarts
+	TrieCleanCacheRejournal time.Duration `toml:",omitempty"` // Time interval to regenerate the journal for clean cache
 	TrieDirtyCache          int
 	TrieTimeout             time.Duration
 	SnapshotCache           int

eth/ethconfig/gen_config.go

@@ -39,6 +39,8 @@ func (c Config) MarshalTOML() (interface{}, error) {
 		DatabaseCache           int
 		DatabaseFreezer         string
 		TrieCleanCache          int
+		TrieCleanCacheJournal   string        `toml:",omitempty"`
+		TrieCleanCacheRejournal time.Duration `toml:",omitempty"`
 		TrieDirtyCache          int
 		TrieTimeout             time.Duration
 		SnapshotCache           int
@@ -79,6 +81,8 @@ func (c Config) MarshalTOML() (interface{}, error) {
 	enc.DatabaseCache = c.DatabaseCache
 	enc.DatabaseFreezer = c.DatabaseFreezer
 	enc.TrieCleanCache = c.TrieCleanCache
+	enc.TrieCleanCacheJournal = c.TrieCleanCacheJournal
+	enc.TrieCleanCacheRejournal = c.TrieCleanCacheRejournal
 	enc.TrieDirtyCache = c.TrieDirtyCache
 	enc.TrieTimeout = c.TrieTimeout
 	enc.SnapshotCache = c.SnapshotCache
@@ -123,6 +127,8 @@ func (c *Config) UnmarshalTOML(unmarshal func(interface{}) error) error {
 		DatabaseCache           *int
 		DatabaseFreezer         *string
 		TrieCleanCache          *int
+		TrieCleanCacheJournal   *string        `toml:",omitempty"`
+		TrieCleanCacheRejournal *time.Duration `toml:",omitempty"`
 		TrieDirtyCache          *int
 		TrieTimeout             *time.Duration
 		SnapshotCache           *int
@@ -212,6 +218,12 @@ func (c *Config) UnmarshalTOML(unmarshal func(interface{}) error) error {
 	if dec.TrieCleanCache != nil {
 		c.TrieCleanCache = *dec.TrieCleanCache
 	}
+	if dec.TrieCleanCacheJournal != nil {
+		c.TrieCleanCacheJournal = *dec.TrieCleanCacheJournal
+	}
+	if dec.TrieCleanCacheRejournal != nil {
+		c.TrieCleanCacheRejournal = *dec.TrieCleanCacheRejournal
+	}
 	if dec.TrieDirtyCache != nil {
 		c.TrieDirtyCache = *dec.TrieDirtyCache
 	}

graphql/graphql_test.go

@@ -441,6 +441,8 @@ func newGQLService(t *testing.T, stack *node.Node, shanghai bool, gspec *core.Ge
 		Genesis:                 gspec,
 		NetworkId:               1337,
 		TrieCleanCache:          5,
+		TrieCleanCacheJournal:   "triecache",
+		TrieCleanCacheRejournal: 60 * time.Minute,
 		TrieDirtyCache:          5,
 		TrieTimeout:             60 * time.Minute,
 		SnapshotCache:           5,

tests/fuzzers/snap/fuzz_handler.go

@@ -71,6 +71,7 @@ func getChain() *core.BlockChain {
 		TrieDirtyLimit:      0,
 		TrieTimeLimit:       5 * time.Minute,
 		TrieCleanNoPrefetch: true,
+		TrieCleanRejournal:  0,
 		SnapshotLimit:       100,
 		SnapshotWait:        true,
 	}

trie/database_wrap.go

@@ -18,10 +18,13 @@ package trie
 
 import (
 	"errors"
+	"runtime"
+	"time"
 
 	"github.com/VictoriaMetrics/fastcache"
 	"github.com/ethereum/go-ethereum/common"
 	"github.com/ethereum/go-ethereum/ethdb"
+	"github.com/ethereum/go-ethereum/log"
 	"github.com/ethereum/go-ethereum/trie/triedb/hashdb"
 	"github.com/ethereum/go-ethereum/trie/trienode"
 )
@@ -29,6 +32,7 @@ import (
 // Config defines all necessary options for database.
 type Config struct {
 	Cache     int    // Memory allowance (MB) to use for caching trie nodes in memory
+	Journal   string // Journal of clean cache to survive node restarts
 	Preimages bool   // Flag whether the preimage of trie key is recorded
 }
 
@@ -75,7 +79,11 @@ type Database struct {
 func prepare(diskdb ethdb.Database, config *Config) *Database {
 	var cleans *fastcache.Cache
 	if config != nil && config.Cache > 0 {
+		if config.Journal == "" {
 			cleans = fastcache.New(config.Cache * 1024 * 1024)
+		} else {
+			cleans = fastcache.LoadFromFileOrNew(config.Journal, config.Cache*1024*1024)
+		}
 	}
 	var preimages *preimageStore
 	if config != nil && config.Preimages {
@@ -171,6 +179,46 @@ func (db *Database) WritePreimages() {
 	}
 }
 
+// saveCache saves clean state cache to given directory path
+// using specified CPU cores.
+func (db *Database) saveCache(dir string, threads int) error {
+	if db.cleans == nil {
+		return nil
+	}
+	log.Info("Writing clean trie cache to disk", "path", dir, "threads", threads)
+
+	start := time.Now()
+	err := db.cleans.SaveToFileConcurrent(dir, threads)
+	if err != nil {
+		log.Error("Failed to persist clean trie cache", "error", err)
+		return err
+	}
+	log.Info("Persisted the clean trie cache", "path", dir, "elapsed", common.PrettyDuration(time.Since(start)))
+	return nil
+}
+
+// SaveCache atomically saves fast cache data to the given dir using all
+// available CPU cores.
+func (db *Database) SaveCache(dir string) error {
+	return db.saveCache(dir, runtime.GOMAXPROCS(0))
+}
+
+// SaveCachePeriodically atomically saves fast cache data to the given dir with
+// the specified interval. All dump operation will only use a single CPU core.
+func (db *Database) SaveCachePeriodically(dir string, interval time.Duration, stopCh <-chan struct{}) {
+	ticker := time.NewTicker(interval)
+	defer ticker.Stop()
+
+	for {
+		select {
+		case <-ticker.C:
+			db.saveCache(dir, 1)
+		case <-stopCh:
+			return
+		}
+	}
+}
+
 // Cap iteratively flushes old but still referenced trie nodes until the total
 // memory usage goes below the given threshold. The held pre-images accumulated
 // up to this point will be flushed in case the size exceeds the threshold.