prysm-pulse/tools/exploredb/main.go

538 lines
18 KiB
Go
Raw Normal View History

/**
* Explore DB contents
*
* Given a beacon-chain DB, This tool provides many option to
* inspect and explore it. For every non-empty bucket, print
* the number of rows, bucket size,min/average/max size of values
*/
package main
import (
"bytes"
"context"
"flag"
"os"
"path/filepath"
"sync"
"time"
"github.com/dustin/go-humanize"
"github.com/prysmaticlabs/prysm/beacon-chain/db/kv"
"github.com/prysmaticlabs/prysm/beacon-chain/state"
"github.com/prysmaticlabs/prysm/config/params"
types "github.com/prysmaticlabs/prysm/consensus-types/primitives"
"github.com/prysmaticlabs/prysm/encoding/bytesutil"
ethpb "github.com/prysmaticlabs/prysm/proto/prysm/v1alpha1"
log "github.com/sirupsen/logrus"
"github.com/status-im/keycard-go/hexutils"
bolt "go.etcd.io/bbolt"
)
const (
MaxUint64 = ^uint64(0)
maxSlotsToDisplay = 2000000
)
var (
datadir = flag.String("datadir", "", "Path to data directory.")
dbName = flag.String("dbname", "", "database name.")
command = flag.String("command", "", "command to execute.")
bucketName = flag.String("bucket-name", "", "bucket to show contents.")
rowLimit = flag.Uint64("limit", 10, "limit to rows.")
migrationName = flag.String("migration", "", "migration to cross check.")
destDatadir = flag.String("dest-datadir", "", "Path to destination data directory.")
)
// used to parallelize all the bucket stats
type bucketStat struct {
bucketName string
noOfRows uint64
totalKeySize uint64
totalValueSize uint64
minKeySize uint64
maxKeySize uint64
minValueSize uint64
maxValueSize uint64
}
// used to parallelize state bucket processing
type modifiedState struct {
state state.BeaconState
key []byte
valueSize uint64
rowCount uint64
}
// used to parallelize state summary bucket processing
type modifiedStateSummary struct {
slot types.Slot
root []byte
key []byte
valueSize uint64
rowCount uint64
}
func main() {
flag.Parse()
// Check for the mandatory flags.
if *datadir == "" {
log.Fatal("Please specify --datadir <db path> to read the database")
}
if *dbName == "" {
log.Fatal("Please specify --dbname <db file name> to specify the database file.")
}
// check if the database file is present.
dbNameWithPath := filepath.Join(*datadir, *dbName)
if _, err := os.Stat(dbNameWithPath); os.IsNotExist(err) {
log.Fatalf("could not locate database file : %s, %v", dbNameWithPath, err)
}
switch *command {
case "bucket-stats":
printBucketStats(dbNameWithPath)
case "bucket-content":
switch *bucketName {
case "state",
"state-summary":
printBucketContents(dbNameWithPath, *rowLimit, *bucketName)
default:
log.Fatal("Oops, given bucket is supported for now.")
}
case "migration-check":
destDbNameWithPath := filepath.Join(*destDatadir, *dbName)
if _, err := os.Stat(destDbNameWithPath); os.IsNotExist(err) {
log.Fatalf("could not locate destination database file : %s, %v", destDbNameWithPath, err)
}
switch *migrationName {
case "validator-entries":
checkValidatorMigration(dbNameWithPath, destDbNameWithPath)
default:
log.Fatal("Oops, given migration is not supported for now.")
}
}
}
func printBucketStats(dbNameWithPath string) {
groupSize := uint64(128)
doneC := make(chan bool)
statsC := make(chan *bucketStat, groupSize)
go readBucketStat(dbNameWithPath, statsC)
go printBucketStat(statsC, doneC)
<-doneC
}
func printBucketContents(dbNameWithPath string, rowLimit uint64, bucketName string) {
// get the keys within the supplied limit for the given bucket.
bucketNameInBytes := []byte(bucketName)
keys, sizes := keysOfBucket(dbNameWithPath, bucketNameInBytes, rowLimit)
// create a new KV Store.
dbDirectory := filepath.Dir(dbNameWithPath)
db, openErr := kv.NewKVStore(context.Background(), dbDirectory, &kv.Config{})
if openErr != nil {
log.Fatalf("could not open db, %v", openErr)
}
// don't forget to close it when ejecting out of this function.
defer func() {
closeErr := db.Close()
if closeErr != nil {
log.Fatalf("could not close db, %v", closeErr)
}
}()
// retrieve every element for keys in the list and call the respective display function.
ctx := context.Background()
groupSize := uint64(128)
doneC := make(chan bool)
switch bucketName {
case "state":
stateC := make(chan *modifiedState, groupSize)
go readStates(ctx, db, stateC, keys, sizes)
go printStates(stateC, doneC)
case "state-summary":
stateSummaryC := make(chan *modifiedStateSummary, groupSize)
go readStateSummary(ctx, db, stateSummaryC, keys, sizes)
go printStateSummary(stateSummaryC, doneC)
}
<-doneC
}
func readBucketStat(dbNameWithPath string, statsC chan<- *bucketStat) {
// open the raw database file. If the file is busy, then exit.
db, openErr := bolt.Open(dbNameWithPath, 0600, &bolt.Options{Timeout: 1 * time.Second})
if openErr != nil {
log.Fatalf("could not open db to show bucket stats, %v", openErr)
}
// make sure we close the database before ejecting out of this function.
defer func() {
closeErr := db.Close()
if closeErr != nil {
log.Fatalf("could not close db after showing bucket stats, %v", closeErr)
}
}()
// get a list of all the existing buckets.
var buckets []string
if viewErr1 := db.View(func(tx *bolt.Tx) error {
return tx.ForEach(func(name []byte, buc *bolt.Bucket) error {
buckets = append(buckets, string(name))
return nil
})
}); viewErr1 != nil {
log.Fatalf("could not read buckets from db while getting list of buckets: %v", viewErr1)
}
// for every bucket, calculate the stats and send it for printing.
// calculate the state of all the buckets in parallel.
var wg sync.WaitGroup
for _, bName := range buckets {
wg.Add(1)
go func(bukName string) {
defer wg.Done()
count := uint64(0)
minValueSize := ^uint64(0)
maxValueSize := uint64(0)
totalValueSize := uint64(0)
minKeySize := ^uint64(0)
maxKeySize := uint64(0)
totalKeySize := uint64(0)
if viewErr2 := db.View(func(tx *bolt.Tx) error {
b := tx.Bucket([]byte(bukName))
if forEachErr := b.ForEach(func(k, v []byte) error {
count++
valueSize := uint64(len(v))
if valueSize < minValueSize {
minValueSize = valueSize
}
if valueSize > maxValueSize {
maxValueSize = valueSize
}
totalValueSize += valueSize
keyize := uint64(len(k))
if keyize < minKeySize {
minKeySize = keyize
}
if keyize > maxKeySize {
maxKeySize = keyize
}
totalKeySize += uint64(len(k))
return nil
}); forEachErr != nil {
log.WithError(forEachErr).Errorf("could not process row %d for bucket: %s", count, bukName)
return forEachErr
}
return nil
}); viewErr2 != nil {
log.WithError(viewErr2).Errorf("could not get stats for bucket: %s", bukName)
return
}
stat := &bucketStat{
bucketName: bukName,
noOfRows: count,
totalKeySize: totalKeySize,
totalValueSize: totalValueSize,
minKeySize: minKeySize,
maxKeySize: maxKeySize,
minValueSize: minValueSize,
maxValueSize: maxValueSize,
}
statsC <- stat
}(bName)
}
wg.Wait()
close(statsC)
}
func readStates(ctx context.Context, db *kv.Store, stateC chan<- *modifiedState, keys [][]byte, sizes []uint64) {
stateMap := make(map[uint64]*modifiedState)
for rowCount, key := range keys {
st, stateErr := db.State(ctx, bytesutil.ToBytes32(key))
if stateErr != nil {
log.WithError(stateErr).Errorf("could not get state for key : %s", hexutils.BytesToHex(key))
continue
}
mst := &modifiedState{
state: st,
key: key,
valueSize: sizes[rowCount],
rowCount: uint64(rowCount),
}
stateMap[uint64(st.Slot())] = mst
}
for i := uint64(0); i < maxSlotsToDisplay; i++ {
if _, ok := stateMap[i]; ok {
stateC <- stateMap[i]
}
}
close(stateC)
}
func readStateSummary(ctx context.Context, db *kv.Store, stateSummaryC chan<- *modifiedStateSummary, keys [][]byte, sizes []uint64) {
for rowCount, key := range keys {
ss, ssErr := db.StateSummary(ctx, bytesutil.ToBytes32(key))
if ssErr != nil {
log.WithError(ssErr).Errorf("could not get state summary for key : %s", hexutils.BytesToHex(key))
continue
}
mst := &modifiedStateSummary{
slot: ss.Slot,
root: ss.Root,
key: key,
valueSize: sizes[rowCount],
rowCount: uint64(rowCount),
}
stateSummaryC <- mst
}
close(stateSummaryC)
}
func printBucketStat(statsC <-chan *bucketStat, doneC chan<- bool) {
for stat := range statsC {
if stat.noOfRows != 0 {
averageValueSize := stat.totalValueSize / stat.noOfRows
averageKeySize := stat.totalKeySize / stat.noOfRows
log.Infof("------ %s ---------", stat.bucketName)
log.Infof("NumberOfRows = %d", stat.noOfRows)
log.Infof("TotalBucketSize = %s", humanize.Bytes(stat.totalValueSize+stat.totalKeySize))
log.Infof("KeySize = %s, (min = %s, avg = %s, max = %s)",
humanize.Bytes(stat.totalKeySize),
humanize.Bytes(stat.minKeySize),
humanize.Bytes(averageKeySize),
humanize.Bytes(stat.maxKeySize))
log.Infof("ValueSize = %s, (min = %s, avg = %s, max = %s)",
humanize.Bytes(stat.totalValueSize),
humanize.Bytes(stat.minValueSize),
humanize.Bytes(averageValueSize),
humanize.Bytes(stat.maxValueSize))
}
}
doneC <- true
}
func printStates(stateC <-chan *modifiedState, doneC chan<- bool) {
for mst := range stateC {
st := mst.state
log.Infof("---- row = %04d, slot = %8d, epoch = %8d, key = %s ----", mst.rowCount, st.Slot(), st.Slot()/params.BeaconConfig().SlotsPerEpoch, hexutils.BytesToHex(mst.key))
log.Infof("key : %s", hexutils.BytesToHex(mst.key))
log.Infof("value : compressed size = %s", humanize.Bytes(mst.valueSize))
t := time.Unix(int64(st.GenesisTime()), 0) // lint:ignore uintcast -- Genesis time will not exceed int64 in your lifetime.
log.Infof("genesis_time : %s", t.Format(time.UnixDate))
log.Infof("genesis_validators_root : %s", hexutils.BytesToHex(st.GenesisValidatorsRoot()))
log.Infof("slot : %d", st.Slot())
log.Infof("fork : previous_version = %b, current_version = %b", st.Fork().PreviousVersion, st.Fork().CurrentVersion)
log.Infof("latest_block_header : sizeSSZ = %s", humanize.Bytes(uint64(st.LatestBlockHeader().SizeSSZ())))
size, count := sizeAndCountOfByteList(st.BlockRoots())
log.Infof("block_roots : size = %s, count = %d", humanize.Bytes(size), count)
size, count = sizeAndCountOfByteList(st.StateRoots())
log.Infof("state_roots : size = %s, count = %d", humanize.Bytes(size), count)
size, count = sizeAndCountOfByteList(st.HistoricalRoots())
log.Infof("historical_roots : size = %s, count = %d", humanize.Bytes(size), count)
log.Infof("eth1_data : sizeSSZ = %s", humanize.Bytes(uint64(st.Eth1Data().SizeSSZ())))
size, count = sizeAndCountGeneric(st.Eth1DataVotes(), nil)
log.Infof("eth1_data_votes : sizeSSZ = %s, count = %d", humanize.Bytes(size), count)
log.Infof("eth1_deposit_index : %d", st.Eth1DepositIndex())
size, count = sizeAndCountGeneric(st.Validators(), nil)
log.Infof("validators : sizeSSZ = %s, count = %d", humanize.Bytes(size), count)
size, count = sizeAndCountOfUin64List(st.Balances())
log.Infof("balances : size = %s, count = %d", humanize.Bytes(size), count)
size, count = sizeAndCountOfByteList(st.RandaoMixes())
log.Infof("randao_mixes : size = %s, count = %d", humanize.Bytes(size), count)
size, count = sizeAndCountOfUin64List(st.Slashings())
log.Infof("slashings : size = %s, count = %d", humanize.Bytes(size), count)
size, count = sizeAndCountGeneric(st.PreviousEpochAttestations())
log.Infof("previous_epoch_attestations : sizeSSZ = %s, count = %d", humanize.Bytes(size), count)
size, count = sizeAndCountGeneric(st.CurrentEpochAttestations())
log.Infof("current_epoch_attestations : sizeSSZ = %s, count = %d", humanize.Bytes(size), count)
Deduplicate native state (a.k.a. One State to rule them all) (#10483) * v0 * getters/setters * init and copy * hasher * all the nice stuff * make bazel happy * remove tests for smaller PR * remove old states * move files * import fixes * custom MarshalSSZ * fixed deadlock * copy version when copying state * correct issues in state_trie * fix Copy() * better e2e comment * add code to minimal state * spectest test * Revert "Auxiliary commit to revert individual files from 84154423464e8372f7e0a03367403656ac5cd78e" This reverts commit 9602599d183081291dfa0ba4f1036430f63a7822. * native state assert * always error * always log * more native state usage * cleanup * remove empty line * Revert "spectests" This reverts commit 1c49bed5d1cf6224afaf21e18562bf72fae5d2b6. # Conflicts: # beacon-chain/powchain/service.go # beacon-chain/state/v1/state_trie.go # beacon-chain/state/v2/state_trie.go # beacon-chain/state/v3/state_trie.go # testing/spectest/shared/phase0/finality/BUILD.bazel # testing/spectest/shared/phase0/finality/runner.go * dedup field trie * fix test issues * cleanup * use correct field num in FinalizedRootProof * use existing version constant * halfway there * "working" version * some fixes * fix field nums in tests * rename v0types to nativetypes * Revert "Auxiliary commit to revert individual files from dc549b1cf8e724bd08cee1ecc760ff3771d5592d" This reverts commit 7254d3070d8693b283fc686a2e01a822ecbac1b3. * uncomment code * remove map size * Revert "Revert "spectests"" This reverts commit 39c271ae6b57d6a3737e2c202cd8407857475e56. * use reverse map * Revert "Revert "Revert "spectests""" This reverts commit 19ba8cf95cdca689357c8234a262e08cccbafef4. * finally found the bug (cherry picked from commit a5414c4be1bdb61a50b391ea5301895e772cc5e9) * simplify populateFieldIndexes * fix copy (cherry picked from commit 7da4fb8cf51557ef931bb781872ea52fc6731af5) * remove native state from e2e * remove index map * unsupported functions * Use ProtobufBeaconState() from native state * tests * typo * reduce complexity of `SaveStatesEfficient` * remove unused receiver name * update doc.go * fix test assertion * fix test assertion 2 * Phase0 justification bits * bring back state tests * rename fieldIndexRev * versioning of ToProto * remove version check from unexported function * hasher tests * don't return error from JustificationBits * extract fieldConvertersNative * helper error function * use fieldConvertersNative * Introduce RealPosition method on FieldIndex * use RealPosition in hasher * remove unused fields * remove TestAppendBeyondIndicesLimit (cherry picked from commit 3017e700282969c30006b64c95c21ffe6b166f8b) * simplify RealPosition * rename field interface * use helper in proofs.go * Update beacon-chain/core/altair/upgrade.go Co-authored-by: Nishant Das <nishdas93@gmail.com> Co-authored-by: Raul Jordan <raul@prysmaticlabs.com>
2022-05-09 13:02:34 +00:00
justificationBits := st.JustificationBits()
log.Infof("justification_bits : size = %s, count = %d", humanize.Bytes(justificationBits.Len()), justificationBits.Count())
log.Infof("previous_justified_checkpoint : sizeSSZ = %s", humanize.Bytes(uint64(st.PreviousJustifiedCheckpoint().SizeSSZ())))
log.Infof("current_justified_checkpoint : sizeSSZ = %s", humanize.Bytes(uint64(st.CurrentJustifiedCheckpoint().SizeSSZ())))
log.Infof("finalized_checkpoint : sizeSSZ = %s", humanize.Bytes(uint64(st.FinalizedCheckpoint().SizeSSZ())))
}
doneC <- true
}
func printStateSummary(stateSummaryC <-chan *modifiedStateSummary, doneC chan<- bool) {
for msts := range stateSummaryC {
log.Infof("row : %04d, slot : %d, root = %s", msts.rowCount, msts.slot, hexutils.BytesToHex(msts.root))
}
doneC <- true
}
func checkValidatorMigration(dbNameWithPath, destDbNameWithPath string) {
// get the keys within the supplied limit for the given bucket.
sourceStateKeys, _ := keysOfBucket(dbNameWithPath, []byte("state"), MaxUint64)
destStateKeys, _ := keysOfBucket(destDbNameWithPath, []byte("state"), MaxUint64)
if len(destStateKeys) < len(sourceStateKeys) {
log.Fatalf("destination keys are lesser then source keys (%d/%d)", len(sourceStateKeys), len(destStateKeys))
}
// create the source and destination KV stores.
sourceDbDirectory := filepath.Dir(dbNameWithPath)
sourceDB, openErr := kv.NewKVStore(context.Background(), sourceDbDirectory, &kv.Config{})
if openErr != nil {
log.Fatalf("could not open sourceDB: %v", openErr)
}
destinationDbDirectory := filepath.Dir(destDbNameWithPath)
destDB, openErr := kv.NewKVStore(context.Background(), destinationDbDirectory, &kv.Config{})
if openErr != nil {
// dirty hack alert: Ignore this prometheus error as we are opening two DB with same metric name
// if you want to avoid this then we should pass the metric name when opening the DB which touches
// too many places.
if openErr.Error() != "duplicate metrics collector registration attempted" {
log.Fatalf("could not open sourceDB, %v", openErr)
}
}
// don't forget to close it when ejecting out of this function.
defer func() {
closeErr := sourceDB.Close()
if closeErr != nil {
log.Fatalf("could not close sourceDB: %v", closeErr)
}
}()
defer func() {
closeErr := destDB.Close()
if closeErr != nil {
log.Fatalf("could not close sourceDB: %v", closeErr)
}
}()
ctx := context.Background()
failCount := 0
for rowCount, key := range sourceStateKeys[910:] {
sourceState, stateErr := sourceDB.State(ctx, bytesutil.ToBytes32(key))
if stateErr != nil {
log.Fatalf("could not get from source db, the state for key : %s, %v", hexutils.BytesToHex(key), stateErr)
}
destinationState, stateErr := destDB.State(ctx, bytesutil.ToBytes32(key))
if stateErr != nil {
log.Fatalf("could not get destination db, the state for key : %s, %v", hexutils.BytesToHex(key), stateErr)
}
if destinationState == nil {
log.Infof("could not find state in migrated DB: index = %d, slot = %d, epoch = %d, numOfValidators = %d, key = %s",
rowCount, sourceState.Slot(), sourceState.Slot()/params.BeaconConfig().SlotsPerEpoch, sourceState.NumValidators(), hexutils.BytesToHex(key))
failCount++
continue
}
if len(sourceState.Validators()) != len(destinationState.Validators()) {
log.Fatalf("validator mismatch : source = %d, dest = %d", len(sourceState.Validators()), len(destinationState.Validators()))
}
sourceStateHash, err := sourceState.HashTreeRoot(ctx)
if err != nil {
log.Fatalf("could not find hash of source state: %v", err)
}
destinationSatteHash, err := destinationState.HashTreeRoot(ctx)
if err != nil {
log.Fatalf("could not find hash of destination state: %v", err)
}
if !bytes.Equal(sourceStateHash[:], destinationSatteHash[:]) {
log.Fatalf("state mismatch : key = %s", hexutils.BytesToHex(key))
}
}
log.Infof("number of state that did not match: %d", failCount)
}
func keysOfBucket(dbNameWithPath string, bucketName []byte, rowLimit uint64) ([][]byte, []uint64) {
// open the raw database file. If the file is busy, then exit.
db, openErr := bolt.Open(dbNameWithPath, 0600, &bolt.Options{Timeout: 1 * time.Second})
if openErr != nil {
log.Fatalf("could not open db while getting keys of a bucket, %v", openErr)
}
// make sure we close the database before ejecting out of this function.
defer func() {
closeErr := db.Close()
if closeErr != nil {
log.Fatalf("could not close db while getting keys of a bucket, %v", closeErr)
}
}()
// get all the keys of the given bucket.
var keys [][]byte
var sizes []uint64
if viewErr := db.View(func(tx *bolt.Tx) error {
b := tx.Bucket(bucketName)
c := b.Cursor()
count := uint64(0)
for k, v := c.First(); k != nil; k, v = c.Next() {
if count >= rowLimit {
return nil
}
actualKey := make([]byte, len(k))
actualSizes := make([]byte, len(v))
copy(actualKey, k)
copy(actualSizes, v)
keys = append(keys, actualKey)
sizes = append(sizes, uint64(len(v)))
count++
}
return nil
}); viewErr != nil {
log.Fatalf("could not read keys of bucket from db: %v", viewErr)
}
return keys, sizes
}
func sizeAndCountOfByteList(list [][]byte) (uint64, uint64) {
size := uint64(0)
count := uint64(0)
for _, root := range list {
size += uint64(len(root))
count += 1
}
return size, count
}
func sizeAndCountOfUin64List(list []uint64) (uint64, uint64) {
size := uint64(0)
count := uint64(0)
for i := 0; i < len(list); i++ {
size += uint64(8)
count += 1
}
return size, count
}
func sizeAndCountGeneric(genericItems interface{}, err error) (uint64, uint64) {
size := uint64(0)
count := uint64(0)
if err != nil {
return size, count
}
switch items := genericItems.(type) {
case []*ethpb.Eth1Data:
for _, item := range items {
size += uint64(item.SizeSSZ())
}
count = uint64(len(items))
case []*ethpb.Validator:
for _, item := range items {
size += uint64(item.SizeSSZ())
}
count = uint64(len(items))
case []*ethpb.PendingAttestation:
for _, item := range items {
size += uint64(item.SizeSSZ())
}
count = uint64(len(items))
default:
return 0, 0
}
return size, count
}