mirror of https://github.com/ethereum/go-ethereum
You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
866 lines
24 KiB
866 lines
24 KiB
// Copyright 2014 The go-ethereum Authors
|
|
// This file is part of go-ethereum.
|
|
//
|
|
// go-ethereum is free software: you can redistribute it and/or modify
|
|
// it under the terms of the GNU General Public License as published by
|
|
// the Free Software Foundation, either version 3 of the License, or
|
|
// (at your option) any later version.
|
|
//
|
|
// go-ethereum is distributed in the hope that it will be useful,
|
|
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
// GNU General Public License for more details.
|
|
//
|
|
// You should have received a copy of the GNU General Public License
|
|
// along with go-ethereum. If not, see <http://www.gnu.org/licenses/>.
|
|
|
|
// Package utils contains internal helper functions for go-ethereum commands.
|
|
package utils
|
|
|
|
import (
|
|
"bufio"
|
|
"bytes"
|
|
"compress/gzip"
|
|
"crypto/sha256"
|
|
"errors"
|
|
"fmt"
|
|
"io"
|
|
"os"
|
|
"os/signal"
|
|
"path/filepath"
|
|
"runtime"
|
|
"strings"
|
|
"syscall"
|
|
"time"
|
|
|
|
"github.com/ethereum/go-ethereum/common"
|
|
"github.com/ethereum/go-ethereum/core"
|
|
"github.com/ethereum/go-ethereum/core/rawdb"
|
|
"github.com/ethereum/go-ethereum/core/state/snapshot"
|
|
"github.com/ethereum/go-ethereum/core/types"
|
|
"github.com/ethereum/go-ethereum/crypto"
|
|
"github.com/ethereum/go-ethereum/eth/ethconfig"
|
|
"github.com/ethereum/go-ethereum/ethdb"
|
|
"github.com/ethereum/go-ethereum/internal/debug"
|
|
"github.com/ethereum/go-ethereum/internal/era"
|
|
"github.com/ethereum/go-ethereum/log"
|
|
"github.com/ethereum/go-ethereum/node"
|
|
"github.com/ethereum/go-ethereum/params"
|
|
"github.com/ethereum/go-ethereum/rlp"
|
|
"github.com/urfave/cli/v2"
|
|
)
|
|
|
|
const (
|
|
importBatchSize = 2500
|
|
)
|
|
|
|
// Fatalf formats a message to standard error and exits the program.
|
|
// The message is also printed to standard output if standard error
|
|
// is redirected to a different file.
|
|
func Fatalf(format string, args ...interface{}) {
|
|
w := io.MultiWriter(os.Stdout, os.Stderr)
|
|
if runtime.GOOS == "windows" {
|
|
// The SameFile check below doesn't work on Windows.
|
|
// stdout is unlikely to get redirected though, so just print there.
|
|
w = os.Stdout
|
|
} else {
|
|
outf, _ := os.Stdout.Stat()
|
|
errf, _ := os.Stderr.Stat()
|
|
if outf != nil && errf != nil && os.SameFile(outf, errf) {
|
|
w = os.Stderr
|
|
}
|
|
}
|
|
fmt.Fprintf(w, "Fatal: "+format+"\n", args...)
|
|
os.Exit(1)
|
|
}
|
|
|
|
func StartNode(ctx *cli.Context, stack *node.Node, isConsole bool) {
|
|
if err := stack.Start(); err != nil {
|
|
Fatalf("Error starting protocol stack: %v", err)
|
|
}
|
|
go func() {
|
|
sigc := make(chan os.Signal, 1)
|
|
signal.Notify(sigc, syscall.SIGINT, syscall.SIGTERM)
|
|
defer signal.Stop(sigc)
|
|
|
|
minFreeDiskSpace := 2 * ethconfig.Defaults.TrieDirtyCache // Default 2 * 256Mb
|
|
if ctx.IsSet(MinFreeDiskSpaceFlag.Name) {
|
|
minFreeDiskSpace = ctx.Int(MinFreeDiskSpaceFlag.Name)
|
|
} else if ctx.IsSet(CacheFlag.Name) || ctx.IsSet(CacheGCFlag.Name) {
|
|
minFreeDiskSpace = 2 * ctx.Int(CacheFlag.Name) * ctx.Int(CacheGCFlag.Name) / 100
|
|
}
|
|
if minFreeDiskSpace > 0 {
|
|
go monitorFreeDiskSpace(sigc, stack.InstanceDir(), uint64(minFreeDiskSpace)*1024*1024)
|
|
}
|
|
|
|
shutdown := func() {
|
|
log.Info("Got interrupt, shutting down...")
|
|
go stack.Close()
|
|
for i := 10; i > 0; i-- {
|
|
<-sigc
|
|
if i > 1 {
|
|
log.Warn("Already shutting down, interrupt more to panic.", "times", i-1)
|
|
}
|
|
}
|
|
debug.Exit() // ensure trace and CPU profile data is flushed.
|
|
debug.LoudPanic("boom")
|
|
}
|
|
|
|
if isConsole {
|
|
// In JS console mode, SIGINT is ignored because it's handled by the console.
|
|
// However, SIGTERM still shuts down the node.
|
|
for {
|
|
sig := <-sigc
|
|
if sig == syscall.SIGTERM {
|
|
shutdown()
|
|
return
|
|
}
|
|
}
|
|
} else {
|
|
<-sigc
|
|
shutdown()
|
|
}
|
|
}()
|
|
}
|
|
|
|
func monitorFreeDiskSpace(sigc chan os.Signal, path string, freeDiskSpaceCritical uint64) {
|
|
if path == "" {
|
|
return
|
|
}
|
|
for {
|
|
freeSpace, err := getFreeDiskSpace(path)
|
|
if err != nil {
|
|
log.Warn("Failed to get free disk space", "path", path, "err", err)
|
|
break
|
|
}
|
|
if freeSpace < freeDiskSpaceCritical {
|
|
log.Error("Low disk space. Gracefully shutting down Geth to prevent database corruption.", "available", common.StorageSize(freeSpace), "path", path)
|
|
sigc <- syscall.SIGTERM
|
|
break
|
|
} else if freeSpace < 2*freeDiskSpaceCritical {
|
|
log.Warn("Disk space is running low. Geth will shutdown if disk space runs below critical level.", "available", common.StorageSize(freeSpace), "critical_level", common.StorageSize(freeDiskSpaceCritical), "path", path)
|
|
}
|
|
time.Sleep(30 * time.Second)
|
|
}
|
|
}
|
|
|
|
func ImportChain(chain *core.BlockChain, fn string) error {
|
|
// Watch for Ctrl-C while the import is running.
|
|
// If a signal is received, the import will stop at the next batch.
|
|
interrupt := make(chan os.Signal, 1)
|
|
stop := make(chan struct{})
|
|
signal.Notify(interrupt, syscall.SIGINT, syscall.SIGTERM)
|
|
defer signal.Stop(interrupt)
|
|
defer close(interrupt)
|
|
go func() {
|
|
if _, ok := <-interrupt; ok {
|
|
log.Info("Interrupted during import, stopping at next batch")
|
|
}
|
|
close(stop)
|
|
}()
|
|
checkInterrupt := func() bool {
|
|
select {
|
|
case <-stop:
|
|
return true
|
|
default:
|
|
return false
|
|
}
|
|
}
|
|
|
|
log.Info("Importing blockchain", "file", fn)
|
|
|
|
// Open the file handle and potentially unwrap the gzip stream
|
|
fh, err := os.Open(fn)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
defer fh.Close()
|
|
|
|
var reader io.Reader = fh
|
|
if strings.HasSuffix(fn, ".gz") {
|
|
if reader, err = gzip.NewReader(reader); err != nil {
|
|
return err
|
|
}
|
|
}
|
|
stream := rlp.NewStream(reader, 0)
|
|
|
|
// Run actual the import.
|
|
blocks := make(types.Blocks, importBatchSize)
|
|
n := 0
|
|
for batch := 0; ; batch++ {
|
|
// Load a batch of RLP blocks.
|
|
if checkInterrupt() {
|
|
return errors.New("interrupted")
|
|
}
|
|
i := 0
|
|
for ; i < importBatchSize; i++ {
|
|
var b types.Block
|
|
if err := stream.Decode(&b); err == io.EOF {
|
|
break
|
|
} else if err != nil {
|
|
return fmt.Errorf("at block %d: %v", n, err)
|
|
}
|
|
// don't import first block
|
|
if b.NumberU64() == 0 {
|
|
i--
|
|
continue
|
|
}
|
|
blocks[i] = &b
|
|
n++
|
|
}
|
|
if i == 0 {
|
|
break
|
|
}
|
|
// Import the batch.
|
|
if checkInterrupt() {
|
|
return errors.New("interrupted")
|
|
}
|
|
missing := missingBlocks(chain, blocks[:i])
|
|
if len(missing) == 0 {
|
|
log.Info("Skipping batch as all blocks present", "batch", batch, "first", blocks[0].Hash(), "last", blocks[i-1].Hash())
|
|
continue
|
|
}
|
|
if failindex, err := chain.InsertChain(missing); err != nil {
|
|
var failnumber uint64
|
|
if failindex > 0 && failindex < len(missing) {
|
|
failnumber = missing[failindex].NumberU64()
|
|
} else {
|
|
failnumber = missing[0].NumberU64()
|
|
}
|
|
return fmt.Errorf("invalid block %d: %v", failnumber, err)
|
|
}
|
|
}
|
|
return nil
|
|
}
|
|
|
|
func readList(filename string) ([]string, error) {
|
|
b, err := os.ReadFile(filename)
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
return strings.Split(string(b), "\n"), nil
|
|
}
|
|
|
|
// ImportHistory imports Era1 files containing historical block information,
|
|
// starting from genesis.
|
|
func ImportHistory(chain *core.BlockChain, db ethdb.Database, dir string, network string) error {
|
|
if chain.CurrentSnapBlock().Number.BitLen() != 0 {
|
|
return errors.New("history import only supported when starting from genesis")
|
|
}
|
|
entries, err := era.ReadDir(dir, network)
|
|
if err != nil {
|
|
return fmt.Errorf("error reading %s: %w", dir, err)
|
|
}
|
|
checksums, err := readList(filepath.Join(dir, "checksums.txt"))
|
|
if err != nil {
|
|
return fmt.Errorf("unable to read checksums.txt: %w", err)
|
|
}
|
|
if len(checksums) != len(entries) {
|
|
return fmt.Errorf("expected equal number of checksums and entries, have: %d checksums, %d entries", len(checksums), len(entries))
|
|
}
|
|
var (
|
|
start = time.Now()
|
|
reported = time.Now()
|
|
imported = 0
|
|
forker = core.NewForkChoice(chain, nil)
|
|
h = sha256.New()
|
|
buf = bytes.NewBuffer(nil)
|
|
)
|
|
for i, filename := range entries {
|
|
err := func() error {
|
|
f, err := os.Open(filepath.Join(dir, filename))
|
|
if err != nil {
|
|
return fmt.Errorf("unable to open era: %w", err)
|
|
}
|
|
defer f.Close()
|
|
|
|
// Validate checksum.
|
|
if _, err := io.Copy(h, f); err != nil {
|
|
return fmt.Errorf("unable to recalculate checksum: %w", err)
|
|
}
|
|
if have, want := common.BytesToHash(h.Sum(buf.Bytes()[:])).Hex(), checksums[i]; have != want {
|
|
return fmt.Errorf("checksum mismatch: have %s, want %s", have, want)
|
|
}
|
|
h.Reset()
|
|
buf.Reset()
|
|
|
|
// Import all block data from Era1.
|
|
e, err := era.From(f)
|
|
if err != nil {
|
|
return fmt.Errorf("error opening era: %w", err)
|
|
}
|
|
it, err := era.NewIterator(e)
|
|
if err != nil {
|
|
return fmt.Errorf("error making era reader: %w", err)
|
|
}
|
|
for it.Next() {
|
|
block, err := it.Block()
|
|
if err != nil {
|
|
return fmt.Errorf("error reading block %d: %w", it.Number(), err)
|
|
}
|
|
if block.Number().BitLen() == 0 {
|
|
continue // skip genesis
|
|
}
|
|
receipts, err := it.Receipts()
|
|
if err != nil {
|
|
return fmt.Errorf("error reading receipts %d: %w", it.Number(), err)
|
|
}
|
|
if status, err := chain.HeaderChain().InsertHeaderChain([]*types.Header{block.Header()}, start, forker); err != nil {
|
|
return fmt.Errorf("error inserting header %d: %w", it.Number(), err)
|
|
} else if status != core.CanonStatTy {
|
|
return fmt.Errorf("error inserting header %d, not canon: %v", it.Number(), status)
|
|
}
|
|
if _, err := chain.InsertReceiptChain([]*types.Block{block}, []types.Receipts{receipts}, 2^64-1); err != nil {
|
|
return fmt.Errorf("error inserting body %d: %w", it.Number(), err)
|
|
}
|
|
imported += 1
|
|
|
|
// Give the user some feedback that something is happening.
|
|
if time.Since(reported) >= 8*time.Second {
|
|
log.Info("Importing Era files", "head", it.Number(), "imported", imported, "elapsed", common.PrettyDuration(time.Since(start)))
|
|
imported = 0
|
|
reported = time.Now()
|
|
}
|
|
}
|
|
return nil
|
|
}()
|
|
if err != nil {
|
|
return err
|
|
}
|
|
}
|
|
|
|
return nil
|
|
}
|
|
|
|
func missingBlocks(chain *core.BlockChain, blocks []*types.Block) []*types.Block {
|
|
head := chain.CurrentBlock()
|
|
for i, block := range blocks {
|
|
// If we're behind the chain head, only check block, state is available at head
|
|
if head.Number.Uint64() > block.NumberU64() {
|
|
if !chain.HasBlock(block.Hash(), block.NumberU64()) {
|
|
return blocks[i:]
|
|
}
|
|
continue
|
|
}
|
|
// If we're above the chain head, state availability is a must
|
|
if !chain.HasBlockAndState(block.Hash(), block.NumberU64()) {
|
|
return blocks[i:]
|
|
}
|
|
}
|
|
return nil
|
|
}
|
|
|
|
// ExportChain exports a blockchain into the specified file, truncating any data
|
|
// already present in the file.
|
|
func ExportChain(blockchain *core.BlockChain, fn string) error {
|
|
log.Info("Exporting blockchain", "file", fn)
|
|
|
|
// Open the file handle and potentially wrap with a gzip stream
|
|
fh, err := os.OpenFile(fn, os.O_CREATE|os.O_WRONLY|os.O_TRUNC, os.ModePerm)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
defer fh.Close()
|
|
|
|
var writer io.Writer = fh
|
|
if strings.HasSuffix(fn, ".gz") {
|
|
writer = gzip.NewWriter(writer)
|
|
defer writer.(*gzip.Writer).Close()
|
|
}
|
|
// Iterate over the blocks and export them
|
|
if err := blockchain.Export(writer); err != nil {
|
|
return err
|
|
}
|
|
log.Info("Exported blockchain", "file", fn)
|
|
|
|
return nil
|
|
}
|
|
|
|
// ExportAppendChain exports a blockchain into the specified file, appending to
|
|
// the file if data already exists in it.
|
|
func ExportAppendChain(blockchain *core.BlockChain, fn string, first uint64, last uint64) error {
|
|
log.Info("Exporting blockchain", "file", fn)
|
|
|
|
// Open the file handle and potentially wrap with a gzip stream
|
|
fh, err := os.OpenFile(fn, os.O_CREATE|os.O_APPEND|os.O_WRONLY, os.ModePerm)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
defer fh.Close()
|
|
|
|
var writer io.Writer = fh
|
|
if strings.HasSuffix(fn, ".gz") {
|
|
writer = gzip.NewWriter(writer)
|
|
defer writer.(*gzip.Writer).Close()
|
|
}
|
|
// Iterate over the blocks and export them
|
|
if err := blockchain.ExportN(writer, first, last); err != nil {
|
|
return err
|
|
}
|
|
log.Info("Exported blockchain to", "file", fn)
|
|
return nil
|
|
}
|
|
|
|
// ExportHistory exports blockchain history into the specified directory,
|
|
// following the Era format.
|
|
func ExportHistory(bc *core.BlockChain, dir string, first, last, step uint64) error {
|
|
log.Info("Exporting blockchain history", "dir", dir)
|
|
if head := bc.CurrentBlock().Number.Uint64(); head < last {
|
|
log.Warn("Last block beyond head, setting last = head", "head", head, "last", last)
|
|
last = head
|
|
}
|
|
network := "unknown"
|
|
if name, ok := params.NetworkNames[bc.Config().ChainID.String()]; ok {
|
|
network = name
|
|
}
|
|
if err := os.MkdirAll(dir, os.ModePerm); err != nil {
|
|
return fmt.Errorf("error creating output directory: %w", err)
|
|
}
|
|
var (
|
|
start = time.Now()
|
|
reported = time.Now()
|
|
h = sha256.New()
|
|
buf = bytes.NewBuffer(nil)
|
|
checksums []string
|
|
)
|
|
for i := first; i <= last; i += step {
|
|
err := func() error {
|
|
filename := filepath.Join(dir, era.Filename(network, int(i/step), common.Hash{}))
|
|
f, err := os.Create(filename)
|
|
if err != nil {
|
|
return fmt.Errorf("could not create era file: %w", err)
|
|
}
|
|
defer f.Close()
|
|
|
|
w := era.NewBuilder(f)
|
|
for j := uint64(0); j < step && j <= last-i; j++ {
|
|
var (
|
|
n = i + j
|
|
block = bc.GetBlockByNumber(n)
|
|
)
|
|
if block == nil {
|
|
return fmt.Errorf("export failed on #%d: not found", n)
|
|
}
|
|
receipts := bc.GetReceiptsByHash(block.Hash())
|
|
if receipts == nil {
|
|
return fmt.Errorf("export failed on #%d: receipts not found", n)
|
|
}
|
|
td := bc.GetTd(block.Hash(), block.NumberU64())
|
|
if td == nil {
|
|
return fmt.Errorf("export failed on #%d: total difficulty not found", n)
|
|
}
|
|
if err := w.Add(block, receipts, td); err != nil {
|
|
return err
|
|
}
|
|
}
|
|
root, err := w.Finalize()
|
|
if err != nil {
|
|
return fmt.Errorf("export failed to finalize %d: %w", step/i, err)
|
|
}
|
|
// Set correct filename with root.
|
|
os.Rename(filename, filepath.Join(dir, era.Filename(network, int(i/step), root)))
|
|
|
|
// Compute checksum of entire Era1.
|
|
if _, err := f.Seek(0, io.SeekStart); err != nil {
|
|
return err
|
|
}
|
|
if _, err := io.Copy(h, f); err != nil {
|
|
return fmt.Errorf("unable to calculate checksum: %w", err)
|
|
}
|
|
checksums = append(checksums, common.BytesToHash(h.Sum(buf.Bytes()[:])).Hex())
|
|
h.Reset()
|
|
buf.Reset()
|
|
return nil
|
|
}()
|
|
if err != nil {
|
|
return err
|
|
}
|
|
if time.Since(reported) >= 8*time.Second {
|
|
log.Info("Exporting blocks", "exported", i, "elapsed", common.PrettyDuration(time.Since(start)))
|
|
reported = time.Now()
|
|
}
|
|
}
|
|
|
|
os.WriteFile(filepath.Join(dir, "checksums.txt"), []byte(strings.Join(checksums, "\n")), os.ModePerm)
|
|
|
|
log.Info("Exported blockchain to", "dir", dir)
|
|
|
|
return nil
|
|
}
|
|
|
|
// ImportPreimages imports a batch of exported hash preimages into the database.
|
|
// It's a part of the deprecated functionality, should be removed in the future.
|
|
func ImportPreimages(db ethdb.Database, fn string) error {
|
|
log.Info("Importing preimages", "file", fn)
|
|
|
|
// Open the file handle and potentially unwrap the gzip stream
|
|
fh, err := os.Open(fn)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
defer fh.Close()
|
|
|
|
var reader io.Reader = bufio.NewReader(fh)
|
|
if strings.HasSuffix(fn, ".gz") {
|
|
if reader, err = gzip.NewReader(reader); err != nil {
|
|
return err
|
|
}
|
|
}
|
|
stream := rlp.NewStream(reader, 0)
|
|
|
|
// Import the preimages in batches to prevent disk thrashing
|
|
preimages := make(map[common.Hash][]byte)
|
|
|
|
for {
|
|
// Read the next entry and ensure it's not junk
|
|
var blob []byte
|
|
|
|
if err := stream.Decode(&blob); err != nil {
|
|
if err == io.EOF {
|
|
break
|
|
}
|
|
return err
|
|
}
|
|
// Accumulate the preimages and flush when enough ws gathered
|
|
preimages[crypto.Keccak256Hash(blob)] = common.CopyBytes(blob)
|
|
if len(preimages) > 1024 {
|
|
rawdb.WritePreimages(db, preimages)
|
|
preimages = make(map[common.Hash][]byte)
|
|
}
|
|
}
|
|
// Flush the last batch preimage data
|
|
if len(preimages) > 0 {
|
|
rawdb.WritePreimages(db, preimages)
|
|
}
|
|
return nil
|
|
}
|
|
|
|
// ExportPreimages exports all known hash preimages into the specified file,
|
|
// truncating any data already present in the file.
|
|
// It's a part of the deprecated functionality, should be removed in the future.
|
|
func ExportPreimages(db ethdb.Database, fn string) error {
|
|
log.Info("Exporting preimages", "file", fn)
|
|
|
|
// Open the file handle and potentially wrap with a gzip stream
|
|
fh, err := os.OpenFile(fn, os.O_CREATE|os.O_WRONLY|os.O_TRUNC, os.ModePerm)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
defer fh.Close()
|
|
|
|
var writer io.Writer = fh
|
|
if strings.HasSuffix(fn, ".gz") {
|
|
writer = gzip.NewWriter(writer)
|
|
defer writer.(*gzip.Writer).Close()
|
|
}
|
|
// Iterate over the preimages and export them
|
|
it := db.NewIterator([]byte("secure-key-"), nil)
|
|
defer it.Release()
|
|
|
|
for it.Next() {
|
|
if err := rlp.Encode(writer, it.Value()); err != nil {
|
|
return err
|
|
}
|
|
}
|
|
log.Info("Exported preimages", "file", fn)
|
|
return nil
|
|
}
|
|
|
|
// ExportSnapshotPreimages exports the preimages corresponding to the enumeration of
|
|
// the snapshot for a given root.
|
|
func ExportSnapshotPreimages(chaindb ethdb.Database, snaptree *snapshot.Tree, fn string, root common.Hash) error {
|
|
log.Info("Exporting preimages", "file", fn)
|
|
|
|
fh, err := os.OpenFile(fn, os.O_CREATE|os.O_WRONLY|os.O_TRUNC, os.ModePerm)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
defer fh.Close()
|
|
|
|
// Enable gzip compressing if file name has gz suffix.
|
|
var writer io.Writer = fh
|
|
if strings.HasSuffix(fn, ".gz") {
|
|
gz := gzip.NewWriter(writer)
|
|
defer gz.Close()
|
|
writer = gz
|
|
}
|
|
buf := bufio.NewWriter(writer)
|
|
defer buf.Flush()
|
|
writer = buf
|
|
|
|
type hashAndPreimageSize struct {
|
|
Hash common.Hash
|
|
Size int
|
|
}
|
|
hashCh := make(chan hashAndPreimageSize)
|
|
|
|
var (
|
|
start = time.Now()
|
|
logged = time.Now()
|
|
preimages int
|
|
)
|
|
go func() {
|
|
defer close(hashCh)
|
|
accIt, err := snaptree.AccountIterator(root, common.Hash{})
|
|
if err != nil {
|
|
log.Error("Failed to create account iterator", "error", err)
|
|
return
|
|
}
|
|
defer accIt.Release()
|
|
|
|
for accIt.Next() {
|
|
acc, err := types.FullAccount(accIt.Account())
|
|
if err != nil {
|
|
log.Error("Failed to get full account", "error", err)
|
|
return
|
|
}
|
|
preimages += 1
|
|
hashCh <- hashAndPreimageSize{Hash: accIt.Hash(), Size: common.AddressLength}
|
|
|
|
if acc.Root != (common.Hash{}) && acc.Root != types.EmptyRootHash {
|
|
stIt, err := snaptree.StorageIterator(root, accIt.Hash(), common.Hash{})
|
|
if err != nil {
|
|
log.Error("Failed to create storage iterator", "error", err)
|
|
return
|
|
}
|
|
for stIt.Next() {
|
|
preimages += 1
|
|
hashCh <- hashAndPreimageSize{Hash: stIt.Hash(), Size: common.HashLength}
|
|
|
|
if time.Since(logged) > time.Second*8 {
|
|
logged = time.Now()
|
|
log.Info("Exporting preimages", "count", preimages, "elapsed", common.PrettyDuration(time.Since(start)))
|
|
}
|
|
}
|
|
stIt.Release()
|
|
}
|
|
if time.Since(logged) > time.Second*8 {
|
|
logged = time.Now()
|
|
log.Info("Exporting preimages", "count", preimages, "elapsed", common.PrettyDuration(time.Since(start)))
|
|
}
|
|
}
|
|
}()
|
|
|
|
for item := range hashCh {
|
|
preimage := rawdb.ReadPreimage(chaindb, item.Hash)
|
|
if len(preimage) == 0 {
|
|
return fmt.Errorf("missing preimage for %v", item.Hash)
|
|
}
|
|
if len(preimage) != item.Size {
|
|
return fmt.Errorf("invalid preimage size, have %d", len(preimage))
|
|
}
|
|
rlpenc, err := rlp.EncodeToBytes(preimage)
|
|
if err != nil {
|
|
return fmt.Errorf("error encoding preimage: %w", err)
|
|
}
|
|
if _, err := writer.Write(rlpenc); err != nil {
|
|
return fmt.Errorf("failed to write preimage: %w", err)
|
|
}
|
|
}
|
|
log.Info("Exported preimages", "count", preimages, "elapsed", common.PrettyDuration(time.Since(start)), "file", fn)
|
|
return nil
|
|
}
|
|
|
|
// exportHeader is used in the export/import flow. When we do an export,
|
|
// the first element we output is the exportHeader.
|
|
// Whenever a backwards-incompatible change is made, the Version header
|
|
// should be bumped.
|
|
// If the importer sees a higher version, it should reject the import.
|
|
type exportHeader struct {
|
|
Magic string // Always set to 'gethdbdump' for disambiguation
|
|
Version uint64
|
|
Kind string
|
|
UnixTime uint64
|
|
}
|
|
|
|
const exportMagic = "gethdbdump"
|
|
const (
|
|
OpBatchAdd = 0
|
|
OpBatchDel = 1
|
|
)
|
|
|
|
// ImportLDBData imports a batch of snapshot data into the database
|
|
func ImportLDBData(db ethdb.Database, f string, startIndex int64, interrupt chan struct{}) error {
|
|
log.Info("Importing leveldb data", "file", f)
|
|
|
|
// Open the file handle and potentially unwrap the gzip stream
|
|
fh, err := os.Open(f)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
defer fh.Close()
|
|
|
|
var reader io.Reader = bufio.NewReader(fh)
|
|
if strings.HasSuffix(f, ".gz") {
|
|
if reader, err = gzip.NewReader(reader); err != nil {
|
|
return err
|
|
}
|
|
}
|
|
stream := rlp.NewStream(reader, 0)
|
|
|
|
// Read the header
|
|
var header exportHeader
|
|
if err := stream.Decode(&header); err != nil {
|
|
return fmt.Errorf("could not decode header: %v", err)
|
|
}
|
|
if header.Magic != exportMagic {
|
|
return errors.New("incompatible data, wrong magic")
|
|
}
|
|
if header.Version != 0 {
|
|
return fmt.Errorf("incompatible version %d, (support only 0)", header.Version)
|
|
}
|
|
log.Info("Importing data", "file", f, "type", header.Kind, "data age",
|
|
common.PrettyDuration(time.Since(time.Unix(int64(header.UnixTime), 0))))
|
|
|
|
// Import the snapshot in batches to prevent disk thrashing
|
|
var (
|
|
count int64
|
|
start = time.Now()
|
|
logged = time.Now()
|
|
batch = db.NewBatch()
|
|
)
|
|
for {
|
|
// Read the next entry
|
|
var (
|
|
op byte
|
|
key, val []byte
|
|
)
|
|
if err := stream.Decode(&op); err != nil {
|
|
if err == io.EOF {
|
|
break
|
|
}
|
|
return err
|
|
}
|
|
if err := stream.Decode(&key); err != nil {
|
|
return err
|
|
}
|
|
if err := stream.Decode(&val); err != nil {
|
|
return err
|
|
}
|
|
if count < startIndex {
|
|
count++
|
|
continue
|
|
}
|
|
switch op {
|
|
case OpBatchDel:
|
|
batch.Delete(key)
|
|
case OpBatchAdd:
|
|
batch.Put(key, val)
|
|
default:
|
|
return fmt.Errorf("unknown op %d", op)
|
|
}
|
|
if batch.ValueSize() > ethdb.IdealBatchSize {
|
|
if err := batch.Write(); err != nil {
|
|
return err
|
|
}
|
|
batch.Reset()
|
|
}
|
|
// Check interruption emitted by ctrl+c
|
|
if count%1000 == 0 {
|
|
select {
|
|
case <-interrupt:
|
|
if err := batch.Write(); err != nil {
|
|
return err
|
|
}
|
|
log.Info("External data import interrupted", "file", f, "count", count, "elapsed", common.PrettyDuration(time.Since(start)))
|
|
return nil
|
|
default:
|
|
}
|
|
}
|
|
if count%1000 == 0 && time.Since(logged) > 8*time.Second {
|
|
log.Info("Importing external data", "file", f, "count", count, "elapsed", common.PrettyDuration(time.Since(start)))
|
|
logged = time.Now()
|
|
}
|
|
count += 1
|
|
}
|
|
// Flush the last batch snapshot data
|
|
if batch.ValueSize() > 0 {
|
|
if err := batch.Write(); err != nil {
|
|
return err
|
|
}
|
|
}
|
|
log.Info("Imported chain data", "file", f, "count", count,
|
|
"elapsed", common.PrettyDuration(time.Since(start)))
|
|
return nil
|
|
}
|
|
|
|
// ChainDataIterator is an interface wraps all necessary functions to iterate
|
|
// the exporting chain data.
|
|
type ChainDataIterator interface {
|
|
// Next returns the key-value pair for next exporting entry in the iterator.
|
|
// When the end is reached, it will return (0, nil, nil, false).
|
|
Next() (byte, []byte, []byte, bool)
|
|
|
|
// Release releases associated resources. Release should always succeed and can
|
|
// be called multiple times without causing error.
|
|
Release()
|
|
}
|
|
|
|
// ExportChaindata exports the given data type (truncating any data already present)
|
|
// in the file. If the suffix is 'gz', gzip compression is used.
|
|
func ExportChaindata(fn string, kind string, iter ChainDataIterator, interrupt chan struct{}) error {
|
|
log.Info("Exporting chain data", "file", fn, "kind", kind)
|
|
defer iter.Release()
|
|
|
|
// Open the file handle and potentially wrap with a gzip stream
|
|
fh, err := os.OpenFile(fn, os.O_CREATE|os.O_WRONLY|os.O_TRUNC, os.ModePerm)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
defer fh.Close()
|
|
|
|
var writer io.Writer = fh
|
|
if strings.HasSuffix(fn, ".gz") {
|
|
writer = gzip.NewWriter(writer)
|
|
defer writer.(*gzip.Writer).Close()
|
|
}
|
|
// Write the header
|
|
if err := rlp.Encode(writer, &exportHeader{
|
|
Magic: exportMagic,
|
|
Version: 0,
|
|
Kind: kind,
|
|
UnixTime: uint64(time.Now().Unix()),
|
|
}); err != nil {
|
|
return err
|
|
}
|
|
// Extract data from source iterator and dump them out to file
|
|
var (
|
|
count int64
|
|
start = time.Now()
|
|
logged = time.Now()
|
|
)
|
|
for {
|
|
op, key, val, ok := iter.Next()
|
|
if !ok {
|
|
break
|
|
}
|
|
if err := rlp.Encode(writer, op); err != nil {
|
|
return err
|
|
}
|
|
if err := rlp.Encode(writer, key); err != nil {
|
|
return err
|
|
}
|
|
if err := rlp.Encode(writer, val); err != nil {
|
|
return err
|
|
}
|
|
if count%1000 == 0 {
|
|
// Check interruption emitted by ctrl+c
|
|
select {
|
|
case <-interrupt:
|
|
log.Info("Chain data exporting interrupted", "file", fn,
|
|
"kind", kind, "count", count, "elapsed", common.PrettyDuration(time.Since(start)))
|
|
return nil
|
|
default:
|
|
}
|
|
if time.Since(logged) > 8*time.Second {
|
|
log.Info("Exporting chain data", "file", fn, "kind", kind,
|
|
"count", count, "elapsed", common.PrettyDuration(time.Since(start)))
|
|
logged = time.Now()
|
|
}
|
|
}
|
|
count++
|
|
}
|
|
log.Info("Exported chain data", "file", fn, "kind", kind, "count", count,
|
|
"elapsed", common.PrettyDuration(time.Since(start)))
|
|
return nil
|
|
}
|
|
|