2092 lines
77 KiB
Nim
2092 lines
77 KiB
Nim
# beacon_chain
|
|
# Copyright (c) 2018-2023 Status Research & Development GmbH
|
|
# Licensed and distributed under either of
|
|
# * MIT license (license terms in the root directory or at https://opensource.org/licenses/MIT).
|
|
# * Apache v2 license (license terms in the root directory or at https://www.apache.org/licenses/LICENSE-2.0).
|
|
# at your option. This file may not be copied, modified, or distributed except according to those terms.
|
|
|
|
{.push raises: [].}
|
|
|
|
import
|
|
std/[os, random, sequtils, terminal, times],
|
|
chronos, chronicles,
|
|
metrics, metrics/chronos_httpserver,
|
|
stew/[byteutils, io2],
|
|
eth/p2p/discoveryv5/[enr, random2],
|
|
eth/keys,
|
|
./consensus_object_pools/vanity_logs/vanity_logs,
|
|
./networking/topic_params,
|
|
./rpc/[rest_api, state_ttl_cache],
|
|
./spec/datatypes/[altair, bellatrix, phase0],
|
|
./spec/[deposit_snapshots, engine_authentication, weak_subjectivity],
|
|
./validators/[keystore_management, validator_duties],
|
|
"."/[
|
|
beacon_node, beacon_node_light_client, deposits,
|
|
nimbus_binary_common, statusbar, trusted_node_sync, wallets]
|
|
|
|
when defined(posix):
|
|
import system/ansi_c
|
|
|
|
from ./spec/datatypes/deneb import SignedBeaconBlock
|
|
|
|
from
|
|
libp2p/protocols/pubsub/gossipsub
|
|
import
|
|
TopicParams, validateParameters, init
|
|
|
|
when defined(windows):
|
|
import winlean
|
|
|
|
type
|
|
LPCSTR* = cstring
|
|
LPSTR* = cstring
|
|
|
|
SERVICE_STATUS* {.final, pure.} = object
|
|
dwServiceType*: DWORD
|
|
dwCurrentState*: DWORD
|
|
dwControlsAccepted*: DWORD
|
|
dwWin32ExitCode*: DWORD
|
|
dwServiceSpecificExitCode*: DWORD
|
|
dwCheckPoint*: DWORD
|
|
dwWaitHint*: DWORD
|
|
|
|
SERVICE_STATUS_HANDLE* = DWORD
|
|
LPSERVICE_STATUS* = ptr SERVICE_STATUS
|
|
LPSERVICE_MAIN_FUNCTION* = proc (para1: DWORD, para2: LPSTR) {.stdcall.}
|
|
|
|
SERVICE_TABLE_ENTRY* {.final, pure.} = object
|
|
lpServiceName*: LPSTR
|
|
lpServiceProc*: LPSERVICE_MAIN_FUNCTION
|
|
|
|
LPSERVICE_TABLE_ENTRY* = ptr SERVICE_TABLE_ENTRY
|
|
LPHANDLER_FUNCTION* = proc (para1: DWORD): WINBOOL{.stdcall.}
|
|
|
|
const
|
|
SERVICE_WIN32_OWN_PROCESS = 16
|
|
SERVICE_RUNNING = 4
|
|
SERVICE_STOPPED = 1
|
|
SERVICE_START_PENDING = 2
|
|
SERVICE_STOP_PENDING = 3
|
|
SERVICE_CONTROL_STOP = 1
|
|
SERVICE_CONTROL_PAUSE = 2
|
|
SERVICE_CONTROL_CONTINUE = 3
|
|
SERVICE_CONTROL_INTERROGATE = 4
|
|
SERVICE_ACCEPT_STOP = 1
|
|
NO_ERROR = 0
|
|
SERVICE_NAME = LPCSTR "NIMBUS_BEACON_NODE"
|
|
|
|
var
|
|
gSvcStatusHandle: SERVICE_STATUS_HANDLE
|
|
gSvcStatus: SERVICE_STATUS
|
|
|
|
proc reportServiceStatus*(dwCurrentState, dwWin32ExitCode, dwWaitHint: DWORD) {.gcsafe.}
|
|
|
|
proc StartServiceCtrlDispatcher*(lpServiceStartTable: LPSERVICE_TABLE_ENTRY): WINBOOL{.
|
|
stdcall, dynlib: "advapi32", importc: "StartServiceCtrlDispatcherA".}
|
|
|
|
proc SetServiceStatus*(hServiceStatus: SERVICE_STATUS_HANDLE,
|
|
lpServiceStatus: LPSERVICE_STATUS): WINBOOL{.stdcall,
|
|
dynlib: "advapi32", importc: "SetServiceStatus".}
|
|
|
|
proc RegisterServiceCtrlHandler*(lpServiceName: LPCSTR,
|
|
lpHandlerProc: LPHANDLER_FUNCTION): SERVICE_STATUS_HANDLE{.
|
|
stdcall, dynlib: "advapi32", importc: "RegisterServiceCtrlHandlerA".}
|
|
|
|
type
|
|
RpcServer = RpcHttpServer
|
|
|
|
template init(T: type RpcHttpServer, ip: ValidIpAddress, port: Port): T =
|
|
newRpcHttpServer([initTAddress(ip, port)])
|
|
|
|
# https://github.com/ethereum/eth2.0-metrics/blob/master/metrics.md#interop-metrics
|
|
declareGauge beacon_slot, "Latest slot of the beacon chain state"
|
|
declareGauge beacon_current_epoch, "Current epoch"
|
|
|
|
# Finalization tracking
|
|
declareGauge finalization_delay,
|
|
"Epoch delay between scheduled epoch and finalized epoch"
|
|
|
|
declareGauge ticks_delay,
|
|
"How long does to take to run the onSecond loop"
|
|
|
|
declareGauge next_action_wait,
|
|
"Seconds until the next attestation will be sent"
|
|
|
|
logScope: topics = "beacnde"
|
|
|
|
func getVanityLogs(stdoutKind: StdoutLogKind): VanityLogs =
|
|
case stdoutKind
|
|
of StdoutLogKind.Auto: raiseAssert "inadmissable here"
|
|
of StdoutLogKind.Colors:
|
|
VanityLogs(
|
|
onMergeTransitionBlock: color🐼,
|
|
onFinalizedMergeTransitionBlock: blink🐼,
|
|
onUpgradeToCapella: color🦉,
|
|
onKnownBlsToExecutionChange: blink🦉)
|
|
of StdoutLogKind.NoColors:
|
|
VanityLogs(
|
|
onMergeTransitionBlock: mono🐼,
|
|
onFinalizedMergeTransitionBlock: mono🐼,
|
|
onUpgradeToCapella: mono🦉,
|
|
onKnownBlsToExecutionChange: mono🦉)
|
|
of StdoutLogKind.Json, StdoutLogKind.None:
|
|
VanityLogs(
|
|
onMergeTransitionBlock:
|
|
(proc() = notice "🐼 Proof of Stake Activated 🐼"),
|
|
onFinalizedMergeTransitionBlock:
|
|
(proc() = notice "🐼 Proof of Stake Finalized 🐼"),
|
|
onUpgradeToCapella:
|
|
(proc() = notice "🦉 Withdrowls now available 🦉"),
|
|
onKnownBlsToExecutionChange:
|
|
(proc() = notice "🦉 BLS to execution changed 🦉"))
|
|
|
|
proc loadChainDag(
|
|
config: BeaconNodeConf,
|
|
cfg: RuntimeConfig,
|
|
db: BeaconChainDB,
|
|
eventBus: EventBus,
|
|
validatorMonitor: ref ValidatorMonitor,
|
|
networkGenesisValidatorsRoot: Opt[Eth2Digest]): ChainDAGRef =
|
|
info "Loading block DAG from database", path = config.databaseDir
|
|
|
|
var dag: ChainDAGRef
|
|
proc onLightClientFinalityUpdate(data: ForkedLightClientFinalityUpdate) =
|
|
if dag == nil: return
|
|
withForkyFinalityUpdate(data):
|
|
when lcDataFork > LightClientDataFork.None:
|
|
let contextFork =
|
|
dag.cfg.consensusForkAtEpoch(forkyFinalityUpdate.contextEpoch)
|
|
eventBus.finUpdateQueue.emit(
|
|
RestVersioned[ForkedLightClientFinalityUpdate](
|
|
data: data,
|
|
jsonVersion: contextFork,
|
|
sszContext: dag.forkDigests[].atConsensusFork(contextFork)))
|
|
proc onLightClientOptimisticUpdate(data: ForkedLightClientOptimisticUpdate) =
|
|
if dag == nil: return
|
|
withForkyOptimisticUpdate(data):
|
|
when lcDataFork > LightClientDataFork.None:
|
|
let contextFork =
|
|
dag.cfg.consensusForkAtEpoch(forkyOptimisticUpdate.contextEpoch)
|
|
eventBus.optUpdateQueue.emit(
|
|
RestVersioned[ForkedLightClientOptimisticUpdate](
|
|
data: data,
|
|
jsonVersion: contextFork,
|
|
sszContext: dag.forkDigests[].atConsensusFork(contextFork)))
|
|
|
|
var extraFlags = {enableTestFeatures}
|
|
if config.deploymentPhase <= DeploymentPhase.Testnet:
|
|
extraFlags.incl experimental
|
|
let
|
|
chainDagFlags =
|
|
if config.strictVerification: {strictVerification}
|
|
else: {}
|
|
onLightClientFinalityUpdateCb =
|
|
if config.lightClientDataServe: onLightClientFinalityUpdate
|
|
else: nil
|
|
onLightClientOptimisticUpdateCb =
|
|
if config.lightClientDataServe: onLightClientOptimisticUpdate
|
|
else: nil
|
|
dag = ChainDAGRef.init(
|
|
cfg, db, validatorMonitor, extraFlags + chainDagFlags, config.eraDir,
|
|
vanityLogs = getVanityLogs(detectTTY(config.logStdout)),
|
|
lcDataConfig = LightClientDataConfig(
|
|
serve: config.lightClientDataServe,
|
|
importMode: config.lightClientDataImportMode,
|
|
maxPeriods: config.lightClientDataMaxPeriods,
|
|
onLightClientFinalityUpdate: onLightClientFinalityUpdateCb,
|
|
onLightClientOptimisticUpdate: onLightClientOptimisticUpdateCb))
|
|
|
|
if networkGenesisValidatorsRoot.isSome:
|
|
let databaseGenesisValidatorsRoot =
|
|
getStateField(dag.headState, genesis_validators_root)
|
|
if networkGenesisValidatorsRoot.get != databaseGenesisValidatorsRoot:
|
|
fatal "The specified --data-dir contains data for a different network",
|
|
networkGenesisValidatorsRoot = networkGenesisValidatorsRoot.get,
|
|
databaseGenesisValidatorsRoot,
|
|
dataDir = config.dataDir
|
|
quit 1
|
|
|
|
# The first pruning after restart may take a while..
|
|
if config.historyMode == HistoryMode.Prune:
|
|
dag.pruneHistory(true)
|
|
|
|
dag
|
|
|
|
proc checkWeakSubjectivityCheckpoint(
|
|
dag: ChainDAGRef,
|
|
wsCheckpoint: Checkpoint,
|
|
beaconClock: BeaconClock) =
|
|
let
|
|
currentSlot = beaconClock.now.slotOrZero
|
|
isCheckpointStale = not is_within_weak_subjectivity_period(
|
|
dag.cfg, currentSlot, dag.headState, wsCheckpoint)
|
|
|
|
if isCheckpointStale:
|
|
error "Weak subjectivity checkpoint is stale",
|
|
currentSlot, checkpoint = wsCheckpoint,
|
|
headStateSlot = getStateField(dag.headState, slot)
|
|
quit 1
|
|
|
|
proc initFullNode(
|
|
node: BeaconNode,
|
|
rng: ref HmacDrbgContext,
|
|
dag: ChainDAGRef,
|
|
taskpool: TaskPoolPtr,
|
|
getBeaconTime: GetBeaconTimeFn) =
|
|
template config(): auto = node.config
|
|
|
|
proc onAttestationReceived(data: Attestation) =
|
|
node.eventBus.attestQueue.emit(data)
|
|
proc onSyncContribution(data: SignedContributionAndProof) =
|
|
node.eventBus.contribQueue.emit(data)
|
|
proc onVoluntaryExitAdded(data: SignedVoluntaryExit) =
|
|
node.eventBus.exitQueue.emit(data)
|
|
proc onBlockAdded(data: ForkedTrustedSignedBeaconBlock) =
|
|
let optimistic =
|
|
if node.currentSlot().epoch() >= dag.cfg.BELLATRIX_FORK_EPOCH:
|
|
some node.dag.is_optimistic(data.root)
|
|
else:
|
|
none[bool]()
|
|
node.eventBus.blocksQueue.emit(
|
|
EventBeaconBlockObject.init(data, optimistic))
|
|
proc onHeadChanged(data: HeadChangeInfoObject) =
|
|
let eventData =
|
|
if node.currentSlot().epoch() >= dag.cfg.BELLATRIX_FORK_EPOCH:
|
|
var res = data
|
|
res.optimistic = some node.dag.is_optimistic(data.block_root)
|
|
res
|
|
else:
|
|
data
|
|
node.eventBus.headQueue.emit(eventData)
|
|
proc onChainReorg(data: ReorgInfoObject) =
|
|
let eventData =
|
|
if node.currentSlot().epoch() >= dag.cfg.BELLATRIX_FORK_EPOCH:
|
|
var res = data
|
|
res.optimistic = some node.dag.is_optimistic(data.new_head_block)
|
|
res
|
|
else:
|
|
data
|
|
node.eventBus.reorgQueue.emit(eventData)
|
|
proc makeOnFinalizationCb(
|
|
# This `nimcall` functions helps for keeping track of what
|
|
# needs to be captured by the onFinalization closure.
|
|
eventBus: EventBus,
|
|
elManager: ELManager): OnFinalizedCallback {.nimcall.} =
|
|
static: doAssert (elManager is ref)
|
|
return proc(dag: ChainDAGRef, data: FinalizationInfoObject) =
|
|
if elManager != nil:
|
|
let finalizedEpochRef = dag.getFinalizedEpochRef()
|
|
discard trackFinalizedState(elManager,
|
|
finalizedEpochRef.eth1_data,
|
|
finalizedEpochRef.eth1_deposit_index)
|
|
node.updateLightClientFromDag()
|
|
let eventData =
|
|
if node.currentSlot().epoch() >= dag.cfg.BELLATRIX_FORK_EPOCH:
|
|
var res = data
|
|
res.optimistic = some node.dag.is_optimistic(data.block_root)
|
|
res
|
|
else:
|
|
data
|
|
eventBus.finalQueue.emit(eventData)
|
|
|
|
func getLocalHeadSlot(): Slot =
|
|
dag.head.slot
|
|
|
|
proc getLocalWallSlot(): Slot =
|
|
node.beaconClock.now.slotOrZero
|
|
|
|
func getFirstSlotAtFinalizedEpoch(): Slot =
|
|
dag.finalizedHead.slot
|
|
|
|
func getBackfillSlot(): Slot =
|
|
dag.backfill.slot
|
|
|
|
func getFrontfillSlot(): Slot =
|
|
max(dag.frontfill.get(BlockId()).slot, dag.horizon)
|
|
|
|
let
|
|
quarantine = newClone(
|
|
Quarantine.init())
|
|
attestationPool = newClone(
|
|
AttestationPool.init(dag, quarantine, onAttestationReceived))
|
|
syncCommitteeMsgPool = newClone(
|
|
SyncCommitteeMsgPool.init(rng, onSyncContribution))
|
|
lightClientPool = newClone(
|
|
LightClientPool())
|
|
validatorChangePool = newClone(
|
|
ValidatorChangePool.init(dag, attestationPool, onVoluntaryExitAdded))
|
|
consensusManager = ConsensusManager.new(
|
|
dag, attestationPool, quarantine, node.elManager,
|
|
ActionTracker.init(rng, config.subscribeAllSubnets),
|
|
node.dynamicFeeRecipientsStore, config.validatorsDir,
|
|
config.defaultFeeRecipient, config.suggestedGasLimit)
|
|
blockProcessor = BlockProcessor.new(
|
|
config.dumpEnabled, config.dumpDirInvalid, config.dumpDirIncoming,
|
|
rng, taskpool, consensusManager, node.validatorMonitor, getBeaconTime)
|
|
blockVerifier =
|
|
proc(signedBlock: ForkedSignedBeaconBlock, maybeFinalized: bool):
|
|
Future[Result[void, VerifierError]] =
|
|
# The design with a callback for block verification is unusual compared
|
|
# to the rest of the application, but fits with the general approach
|
|
# taken in the sync/request managers - this is an architectural compromise
|
|
# that should probably be reimagined more holistically in the future.
|
|
let resfut = newFuture[Result[void, VerifierError]]("blockVerifier")
|
|
blockProcessor[].addBlock(MsgSource.gossip, signedBlock,
|
|
BlobSidecars @[],
|
|
resfut,
|
|
maybeFinalized = maybeFinalized)
|
|
resfut
|
|
blockBlobsVerifier = proc(signedBlock: ForkedSignedBeaconBlock,
|
|
blobs: BlobSidecars,
|
|
maybeFinalized: bool):
|
|
Future[Result[void, VerifierError]] =
|
|
# The design with a callback for block verification is unusual compared
|
|
# to the rest of the application, but fits with the general approach
|
|
# taken in the sync/request managers - this is an architectural compromise
|
|
# that should probably be reimagined more holistically in the future.
|
|
let resfut = newFuture[Result[void, VerifierError]]("blockVerifier")
|
|
blockProcessor[].addBlock(MsgSource.gossip, signedBlock,
|
|
blobs, resfut, maybeFinalized = maybeFinalized)
|
|
resfut
|
|
processor = Eth2Processor.new(
|
|
config.doppelgangerDetection,
|
|
blockProcessor, node.validatorMonitor, dag, attestationPool,
|
|
validatorChangePool, node.attachedValidators, syncCommitteeMsgPool,
|
|
lightClientPool, quarantine, rng, getBeaconTime, taskpool)
|
|
syncManager = newSyncManager[Peer, PeerId](
|
|
node.network.peerPool, dag.cfg.DENEB_FORK_EPOCH, SyncQueueKind.Forward, getLocalHeadSlot,
|
|
getLocalWallSlot, getFirstSlotAtFinalizedEpoch, getBackfillSlot,
|
|
getFrontfillSlot, dag.tail.slot, blockVerifier, blockBlobsVerifier)
|
|
backfiller = newSyncManager[Peer, PeerId](
|
|
node.network.peerPool, dag.cfg.DENEB_FORK_EPOCH, SyncQueueKind.Backward, getLocalHeadSlot,
|
|
getLocalWallSlot, getFirstSlotAtFinalizedEpoch, getBackfillSlot,
|
|
getFrontfillSlot, dag.backfill.slot, blockVerifier, blockBlobsVerifier,
|
|
maxHeadAge = 0)
|
|
router = (ref MessageRouter)(
|
|
processor: processor,
|
|
network: node.network)
|
|
|
|
if node.config.lightClientDataServe:
|
|
proc scheduleSendingLightClientUpdates(slot: Slot) =
|
|
if node.lightClientPool[].broadcastGossipFut != nil:
|
|
return
|
|
if slot <= node.lightClientPool[].latestBroadcastedSlot:
|
|
return
|
|
node.lightClientPool[].latestBroadcastedSlot = slot
|
|
|
|
template fut(): auto = node.lightClientPool[].broadcastGossipFut
|
|
fut = node.handleLightClientUpdates(slot)
|
|
fut.addCallback do (p: pointer) {.gcsafe.}:
|
|
fut = nil
|
|
|
|
router.onSyncCommitteeMessage = scheduleSendingLightClientUpdates
|
|
|
|
dag.setFinalizationCb makeOnFinalizationCb(node.eventBus, node.elManager)
|
|
dag.setBlockCb(onBlockAdded)
|
|
dag.setHeadCb(onHeadChanged)
|
|
dag.setReorgCb(onChainReorg)
|
|
|
|
node.dag = dag
|
|
node.quarantine = quarantine
|
|
node.attestationPool = attestationPool
|
|
node.syncCommitteeMsgPool = syncCommitteeMsgPool
|
|
node.lightClientPool = lightClientPool
|
|
node.validatorChangePool = validatorChangePool
|
|
node.processor = processor
|
|
node.blockProcessor = blockProcessor
|
|
node.consensusManager = consensusManager
|
|
node.requestManager = RequestManager.init(node.network,
|
|
dag.cfg.DENEB_FORK_EPOCH,
|
|
getBeaconTime,
|
|
blockVerifier)
|
|
node.syncManager = syncManager
|
|
node.backfiller = backfiller
|
|
node.router = router
|
|
|
|
node.addValidators()
|
|
|
|
block:
|
|
# Add in-process validators to the list of "known" validators such that
|
|
# we start with a reasonable ENR
|
|
let wallSlot = node.beaconClock.now().slotOrZero()
|
|
for validator in node.attachedValidators[].validators.values():
|
|
if config.validatorMonitorAuto:
|
|
node.validatorMonitor[].addMonitor(validator.pubkey, validator.index)
|
|
|
|
if validator.index.isSome():
|
|
node.consensusManager[].actionTracker.knownValidators[
|
|
validator.index.get()] = wallSlot
|
|
let stabilitySubnets =
|
|
node.consensusManager[].actionTracker.stabilitySubnets(wallSlot)
|
|
# Here, we also set the correct ENR should we be in all subnets mode!
|
|
node.network.updateStabilitySubnetMetadata(stabilitySubnets)
|
|
|
|
node.network.initBeaconSync(dag, getBeaconTime)
|
|
|
|
node.updateValidatorMetrics()
|
|
|
|
const SlashingDbName = "slashing_protection"
|
|
# changing this requires physical file rename as well or history is lost.
|
|
|
|
proc init*(T: type BeaconNode,
|
|
rng: ref HmacDrbgContext,
|
|
config: BeaconNodeConf,
|
|
metadata: Eth2NetworkMetadata): BeaconNode
|
|
{.raises: [Defect, CatchableError].} =
|
|
var taskpool: TaskPoolPtr
|
|
|
|
template cfg: auto = metadata.cfg
|
|
template eth1Network: auto = metadata.eth1Network
|
|
|
|
try:
|
|
if config.numThreads < 0:
|
|
fatal "The number of threads --numThreads cannot be negative."
|
|
quit 1
|
|
elif config.numThreads == 0:
|
|
taskpool = TaskPoolPtr.new(numThreads = min(countProcessors(), 16))
|
|
else:
|
|
taskpool = TaskPoolPtr.new(numThreads = config.numThreads)
|
|
|
|
info "Threadpool started", numThreads = taskpool.numThreads
|
|
except Exception as exc:
|
|
raise newException(Defect, "Failure in taskpool initialization.")
|
|
|
|
let
|
|
eventBus = EventBus(
|
|
blocksQueue: newAsyncEventQueue[EventBeaconBlockObject](),
|
|
headQueue: newAsyncEventQueue[HeadChangeInfoObject](),
|
|
reorgQueue: newAsyncEventQueue[ReorgInfoObject](),
|
|
finUpdateQueue: newAsyncEventQueue[
|
|
RestVersioned[ForkedLightClientFinalityUpdate]](),
|
|
optUpdateQueue: newAsyncEventQueue[
|
|
RestVersioned[ForkedLightClientOptimisticUpdate]](),
|
|
attestQueue: newAsyncEventQueue[Attestation](),
|
|
contribQueue: newAsyncEventQueue[SignedContributionAndProof](),
|
|
exitQueue: newAsyncEventQueue[SignedVoluntaryExit](),
|
|
finalQueue: newAsyncEventQueue[FinalizationInfoObject]()
|
|
)
|
|
db = BeaconChainDB.new(config.databaseDir, cfg, inMemory = false)
|
|
|
|
if config.finalizedCheckpointBlock.isSome:
|
|
warn "--finalized-checkpoint-block has been deprecated, ignoring"
|
|
|
|
let checkpointState = if config.finalizedCheckpointState.isSome:
|
|
let checkpointStatePath = config.finalizedCheckpointState.get.string
|
|
let tmp = try:
|
|
newClone(readSszForkedHashedBeaconState(
|
|
cfg, readAllBytes(checkpointStatePath).tryGet()))
|
|
except SszError as err:
|
|
fatal "Checkpoint state loading failed",
|
|
err = formatMsg(err, checkpointStatePath)
|
|
quit 1
|
|
except CatchableError as err:
|
|
fatal "Failed to read checkpoint state file", err = err.msg
|
|
quit 1
|
|
|
|
if not getStateField(tmp[], slot).is_epoch:
|
|
fatal "--finalized-checkpoint-state must point to a state for an epoch slot",
|
|
slot = getStateField(tmp[], slot)
|
|
quit 1
|
|
tmp
|
|
else:
|
|
nil
|
|
|
|
if config.finalizedDepositTreeSnapshot.isSome:
|
|
let
|
|
depositTreeSnapshotPath = config.finalizedDepositTreeSnapshot.get.string
|
|
depositTreeSnapshot = try:
|
|
SSZ.loadFile(depositTreeSnapshotPath, DepositTreeSnapshot)
|
|
except SszError as err:
|
|
fatal "Deposit tree snapshot loading failed",
|
|
err = formatMsg(err, depositTreeSnapshotPath)
|
|
quit 1
|
|
except CatchableError as err:
|
|
fatal "Failed to read deposit tree snapshot file", err = err.msg
|
|
quit 1
|
|
db.putDepositTreeSnapshot(depositTreeSnapshot)
|
|
|
|
let engineApiUrls = config.engineApiUrls
|
|
|
|
if engineApiUrls.len == 0:
|
|
notice "Running without execution client - validator features disabled (see https://nimbus.guide/eth1.html)"
|
|
|
|
var genesisState =
|
|
if metadata.genesisData.len > 0:
|
|
try:
|
|
newClone readSszForkedHashedBeaconState(
|
|
cfg,
|
|
metadata.genesisData.toOpenArrayByte(0, metadata.genesisData.high))
|
|
except CatchableError as err:
|
|
raiseAssert "Invalid baked-in state: " & err.msg
|
|
else:
|
|
nil
|
|
|
|
if not ChainDAGRef.isInitialized(db).isOk():
|
|
if genesisState == nil and checkpointState == nil:
|
|
fatal "No database and no genesis snapshot found. Please supply a genesis.ssz " &
|
|
"with the network configuration"
|
|
quit 1
|
|
|
|
if not genesisState.isNil and not checkpointState.isNil:
|
|
if getStateField(genesisState[], genesis_validators_root) !=
|
|
getStateField(checkpointState[], genesis_validators_root):
|
|
fatal "Checkpoint state does not match genesis - check the --network parameter",
|
|
rootFromGenesis = getStateField(
|
|
genesisState[], genesis_validators_root),
|
|
rootFromCheckpoint = getStateField(
|
|
checkpointState[], genesis_validators_root)
|
|
quit 1
|
|
|
|
try:
|
|
# Always store genesis state if we have it - this allows reindexing and
|
|
# answering genesis queries
|
|
if not genesisState.isNil:
|
|
ChainDAGRef.preInit(db, genesisState[])
|
|
|
|
if not checkpointState.isNil:
|
|
if genesisState.isNil or
|
|
getStateField(checkpointState[], slot) != GENESIS_SLOT:
|
|
ChainDAGRef.preInit(db, checkpointState[])
|
|
|
|
doAssert ChainDAGRef.isInitialized(db).isOk(), "preInit should have initialized db"
|
|
except CatchableError as exc:
|
|
error "Failed to initialize database", err = exc.msg
|
|
quit 1
|
|
else:
|
|
if not checkpointState.isNil:
|
|
fatal "A database already exists, cannot start from given checkpoint",
|
|
dataDir = config.dataDir
|
|
quit 1
|
|
|
|
# Doesn't use std/random directly, but dependencies might
|
|
randomize(rng[].rand(high(int)))
|
|
|
|
# The validatorMonitorTotals flag has been deprecated and should eventually be
|
|
# removed - until then, it's given priority if set so as not to needlessly
|
|
# break existing setups
|
|
let
|
|
validatorMonitor = newClone(ValidatorMonitor.init(
|
|
config.validatorMonitorAuto,
|
|
config.validatorMonitorTotals.get(
|
|
not config.validatorMonitorDetails)))
|
|
|
|
for key in config.validatorMonitorPubkeys:
|
|
validatorMonitor[].addMonitor(key, Opt.none(ValidatorIndex))
|
|
|
|
let
|
|
networkGenesisValidatorsRoot =
|
|
if not genesisState.isNil:
|
|
Opt.some(getStateField(genesisState[], genesis_validators_root))
|
|
else:
|
|
Opt.none(Eth2Digest)
|
|
|
|
dag = loadChainDag(
|
|
config, cfg, db, eventBus,
|
|
validatorMonitor, networkGenesisValidatorsRoot)
|
|
genesisTime = getStateField(dag.headState, genesis_time)
|
|
beaconClock = BeaconClock.init(genesisTime)
|
|
getBeaconTime = beaconClock.getBeaconTimeFn()
|
|
|
|
if config.weakSubjectivityCheckpoint.isSome:
|
|
dag.checkWeakSubjectivityCheckpoint(
|
|
config.weakSubjectivityCheckpoint.get, beaconClock)
|
|
|
|
let elManager = ELManager.new(
|
|
cfg,
|
|
metadata.depositContractBlock,
|
|
metadata.depositContractBlockHash,
|
|
db,
|
|
engineApiUrls,
|
|
eth1Network)
|
|
|
|
if config.rpcEnabled.isSome:
|
|
warn "Nimbus's JSON-RPC server has been removed. This includes the --rpc, --rpc-port, and --rpc-address configuration options. https://nimbus.guide/rest-api.html shows how to enable and configure the REST Beacon API server which replaces it."
|
|
|
|
let restServer = if config.restEnabled:
|
|
RestServerRef.init(config.restAddress, config.restPort,
|
|
config.keymanagerAllowedOrigin,
|
|
validateBeaconApiQueries,
|
|
config)
|
|
else:
|
|
nil
|
|
|
|
let
|
|
netKeys = getPersistentNetKeys(rng[], config)
|
|
nickname = if config.nodeName == "auto": shortForm(netKeys)
|
|
else: config.nodeName
|
|
network = createEth2Node(
|
|
rng, config, netKeys, cfg, dag.forkDigests, getBeaconTime,
|
|
getStateField(dag.headState, genesis_validators_root))
|
|
|
|
case config.slashingDbKind
|
|
of SlashingDbKind.v2:
|
|
discard
|
|
of SlashingDbKind.v1:
|
|
error "Slashing DB v1 is no longer supported for writing"
|
|
quit 1
|
|
of SlashingDbKind.both:
|
|
warn "Slashing DB v1 deprecated, writing only v2"
|
|
|
|
info "Loading slashing protection database (v2)",
|
|
path = config.validatorsDir()
|
|
|
|
proc getValidatorAndIdx(pubkey: ValidatorPubKey): Opt[ValidatorAndIndex] =
|
|
withState(dag.headState):
|
|
getValidator(forkyState().data.validators.asSeq(), pubkey)
|
|
|
|
let
|
|
slashingProtectionDB =
|
|
SlashingProtectionDB.init(
|
|
getStateField(dag.headState, genesis_validators_root),
|
|
config.validatorsDir(), SlashingDbName)
|
|
validatorPool = newClone(ValidatorPool.init(
|
|
slashingProtectionDB, config.doppelgangerDetection))
|
|
|
|
keymanagerInitResult = initKeymanagerServer(config, restServer)
|
|
keymanagerHost = if keymanagerInitResult.server != nil:
|
|
newClone KeymanagerHost.init(
|
|
validatorPool,
|
|
rng,
|
|
keymanagerInitResult.token,
|
|
config.validatorsDir,
|
|
config.secretsDir,
|
|
config.defaultFeeRecipient,
|
|
config.suggestedGasLimit,
|
|
getValidatorAndIdx,
|
|
getBeaconTime)
|
|
else: nil
|
|
|
|
stateTtlCache =
|
|
if config.restCacheSize > 0:
|
|
StateTtlCache.init(
|
|
cacheSize = config.restCacheSize,
|
|
cacheTtl = chronos.seconds(config.restCacheTtl))
|
|
else:
|
|
nil
|
|
|
|
let payloadBuilderRestClient =
|
|
if config.payloadBuilderEnable:
|
|
RestClientRef.new(config.payloadBuilderUrl).valueOr:
|
|
warn "Payload builder REST client setup failed",
|
|
payloadBuilderUrl = config.payloadBuilderUrl
|
|
nil
|
|
else:
|
|
nil
|
|
|
|
if config.payloadBuilderEnable and payloadBuilderRestClient != nil:
|
|
info "Using external payload builder",
|
|
payloadBuilderUrl = config.payloadBuilderUrl
|
|
|
|
let node = BeaconNode(
|
|
nickname: nickname,
|
|
graffitiBytes: if config.graffiti.isSome: config.graffiti.get
|
|
else: defaultGraffitiBytes(),
|
|
network: network,
|
|
netKeys: netKeys,
|
|
db: db,
|
|
config: config,
|
|
attachedValidators: validatorPool,
|
|
elManager: elManager,
|
|
payloadBuilderRestClient: payloadBuilderRestClient,
|
|
restServer: restServer,
|
|
keymanagerHost: keymanagerHost,
|
|
keymanagerServer: keymanagerInitResult.server,
|
|
keystoreCache: KeystoreCacheRef.init(),
|
|
eventBus: eventBus,
|
|
gossipState: {},
|
|
blocksGossipState: {},
|
|
beaconClock: beaconClock,
|
|
validatorMonitor: validatorMonitor,
|
|
stateTtlCache: stateTtlCache,
|
|
dynamicFeeRecipientsStore: newClone(DynamicFeeRecipientsStore.init()))
|
|
|
|
node.initLightClient(
|
|
rng, cfg, dag.forkDigests, getBeaconTime, dag.genesis_validators_root)
|
|
node.initFullNode(
|
|
rng, dag, taskpool, getBeaconTime)
|
|
|
|
node.updateLightClientFromDag()
|
|
|
|
node
|
|
|
|
func verifyFinalization(node: BeaconNode, slot: Slot) =
|
|
# Epoch must be >= 4 to check finalization
|
|
const SETTLING_TIME_OFFSET = 1'u64
|
|
let epoch = slot.epoch()
|
|
|
|
# Don't static-assert this -- if this isn't called, don't require it
|
|
doAssert SLOTS_PER_EPOCH > SETTLING_TIME_OFFSET
|
|
|
|
# Intentionally, loudly assert. Point is to fail visibly and unignorably
|
|
# during testing.
|
|
if epoch >= 4 and slot mod SLOTS_PER_EPOCH > SETTLING_TIME_OFFSET:
|
|
let finalizedEpoch =
|
|
node.dag.finalizedHead.slot.epoch()
|
|
# Finalization rule 234, that has the most lag slots among the cases, sets
|
|
# state.finalized_checkpoint = old_previous_justified_checkpoint.epoch + 3
|
|
# and then state.slot gets incremented, to increase the maximum offset, if
|
|
# finalization occurs every slot, to 4 slots vs scheduledSlot.
|
|
doAssert finalizedEpoch + 4 >= epoch
|
|
|
|
func subnetLog(v: BitArray): string =
|
|
$toSeq(v.oneIndices())
|
|
|
|
func forkDigests(node: BeaconNode): auto =
|
|
let forkDigestsArray: array[ConsensusFork, auto] = [
|
|
node.dag.forkDigests.phase0,
|
|
node.dag.forkDigests.altair,
|
|
node.dag.forkDigests.bellatrix,
|
|
node.dag.forkDigests.capella,
|
|
node.dag.forkDigests.deneb]
|
|
forkDigestsArray
|
|
|
|
# https://github.com/ethereum/consensus-specs/blob/v1.3.0-rc.3/specs/phase0/validator.md#phase-0-attestation-subnet-stability
|
|
proc updateAttestationSubnetHandlers(node: BeaconNode, slot: Slot) =
|
|
if node.gossipState.card == 0:
|
|
# When disconnected, updateGossipState is responsible for all things
|
|
# subnets - in particular, it will remove subscriptions on the edge where
|
|
# we enter the disconnected state.
|
|
return
|
|
|
|
let
|
|
aggregateSubnets =
|
|
node.consensusManager[].actionTracker.aggregateSubnets(slot)
|
|
stabilitySubnets =
|
|
node.consensusManager[].actionTracker.stabilitySubnets(slot)
|
|
subnets = aggregateSubnets + stabilitySubnets
|
|
|
|
node.network.updateStabilitySubnetMetadata(stabilitySubnets)
|
|
|
|
# Now we know what we should be subscribed to - make it so
|
|
let
|
|
prevSubnets = node.consensusManager[].actionTracker.subscribedSubnets
|
|
unsubscribeSubnets = prevSubnets - subnets
|
|
subscribeSubnets = subnets - prevSubnets
|
|
|
|
# Remember what we subscribed to, so we can unsubscribe later
|
|
node.consensusManager[].actionTracker.subscribedSubnets = subnets
|
|
|
|
let forkDigests = node.forkDigests()
|
|
|
|
for gossipFork in node.gossipState:
|
|
let forkDigest = forkDigests[gossipFork]
|
|
node.network.unsubscribeAttestationSubnets(unsubscribeSubnets, forkDigest)
|
|
node.network.subscribeAttestationSubnets(subscribeSubnets, forkDigest)
|
|
|
|
debug "Attestation subnets",
|
|
slot, epoch = slot.epoch, gossipState = node.gossipState,
|
|
stabilitySubnets = subnetLog(stabilitySubnets),
|
|
aggregateSubnets = subnetLog(aggregateSubnets),
|
|
prevSubnets = subnetLog(prevSubnets),
|
|
subscribeSubnets = subnetLog(subscribeSubnets),
|
|
unsubscribeSubnets = subnetLog(unsubscribeSubnets),
|
|
gossipState = node.gossipState
|
|
|
|
proc updateBlocksGossipStatus*(
|
|
node: BeaconNode, slot: Slot, dagIsBehind: bool) =
|
|
template cfg(): auto = node.dag.cfg
|
|
|
|
let
|
|
isBehind =
|
|
if node.shouldSyncOptimistically(slot):
|
|
# If optimistic sync is active, always subscribe to blocks gossip
|
|
false
|
|
else:
|
|
# Use DAG status to determine whether to subscribe for blocks gossip
|
|
dagIsBehind
|
|
|
|
targetGossipState = getTargetGossipState(
|
|
slot.epoch, cfg.ALTAIR_FORK_EPOCH, cfg.BELLATRIX_FORK_EPOCH,
|
|
cfg.CAPELLA_FORK_EPOCH, cfg.DENEB_FORK_EPOCH, isBehind)
|
|
|
|
template currentGossipState(): auto = node.blocksGossipState
|
|
if currentGossipState == targetGossipState:
|
|
return
|
|
|
|
if currentGossipState.card == 0 and targetGossipState.card > 0:
|
|
debug "Enabling blocks topic subscriptions",
|
|
wallSlot = slot, targetGossipState
|
|
elif currentGossipState.card > 0 and targetGossipState.card == 0:
|
|
debug "Disabling blocks topic subscriptions",
|
|
wallSlot = slot
|
|
else:
|
|
# Individual forks added / removed
|
|
discard
|
|
|
|
let
|
|
newGossipForks = targetGossipState - currentGossipState
|
|
oldGossipForks = currentGossipState - targetGossipState
|
|
|
|
for gossipFork in oldGossipForks:
|
|
let forkDigest = node.dag.forkDigests[].atConsensusFork(gossipFork)
|
|
node.network.unsubscribe(getBeaconBlocksTopic(forkDigest))
|
|
|
|
for gossipFork in newGossipForks:
|
|
let forkDigest = node.dag.forkDigests[].atConsensusFork(gossipFork)
|
|
node.network.subscribe(
|
|
getBeaconBlocksTopic(forkDigest), blocksTopicParams,
|
|
enableTopicMetrics = true)
|
|
|
|
node.blocksGossipState = targetGossipState
|
|
|
|
proc addPhase0MessageHandlers(
|
|
node: BeaconNode, forkDigest: ForkDigest, slot: Slot) =
|
|
node.network.subscribe(getAttesterSlashingsTopic(forkDigest), basicParams)
|
|
node.network.subscribe(getProposerSlashingsTopic(forkDigest), basicParams)
|
|
node.network.subscribe(getVoluntaryExitsTopic(forkDigest), basicParams)
|
|
node.network.subscribe(
|
|
getAggregateAndProofsTopic(forkDigest), aggregateTopicParams,
|
|
enableTopicMetrics = true)
|
|
|
|
# updateAttestationSubnetHandlers subscribes attestation subnets
|
|
|
|
proc removePhase0MessageHandlers(node: BeaconNode, forkDigest: ForkDigest) =
|
|
node.network.unsubscribe(getVoluntaryExitsTopic(forkDigest))
|
|
node.network.unsubscribe(getProposerSlashingsTopic(forkDigest))
|
|
node.network.unsubscribe(getAttesterSlashingsTopic(forkDigest))
|
|
node.network.unsubscribe(getAggregateAndProofsTopic(forkDigest))
|
|
|
|
for subnet_id in SubnetId:
|
|
node.network.unsubscribe(getAttestationTopic(forkDigest, subnet_id))
|
|
|
|
node.consensusManager[].actionTracker.subscribedSubnets = default(AttnetBits)
|
|
|
|
func hasSyncPubKey(node: BeaconNode, epoch: Epoch): auto =
|
|
# Only used to determine which gossip topics to which to subscribe
|
|
if node.config.subscribeAllSubnets:
|
|
(func(pubkey: ValidatorPubKey): bool {.closure.} = true)
|
|
else:
|
|
(func(pubkey: ValidatorPubKey): bool =
|
|
node.consensusManager[].actionTracker.hasSyncDuty(pubkey, epoch) or
|
|
pubkey in node.attachedValidators[].validators)
|
|
|
|
func getCurrentSyncCommiteeSubnets(node: BeaconNode, epoch: Epoch): SyncnetBits =
|
|
let syncCommittee = withState(node.dag.headState):
|
|
when consensusFork >= ConsensusFork.Altair:
|
|
forkyState.data.current_sync_committee
|
|
else:
|
|
return static(default(SyncnetBits))
|
|
|
|
getSyncSubnets(node.hasSyncPubKey(epoch), syncCommittee)
|
|
|
|
func getNextSyncCommitteeSubnets(node: BeaconNode, epoch: Epoch): SyncnetBits =
|
|
let syncCommittee = withState(node.dag.headState):
|
|
when consensusFork >= ConsensusFork.Altair:
|
|
forkyState.data.next_sync_committee
|
|
else:
|
|
return static(default(SyncnetBits))
|
|
|
|
getSyncSubnets(
|
|
node.hasSyncPubKey((epoch.sync_committee_period + 1).start_slot().epoch),
|
|
syncCommittee)
|
|
|
|
func getSyncCommitteeSubnets(node: BeaconNode, epoch: Epoch): SyncnetBits =
|
|
let
|
|
subnets = node.getCurrentSyncCommiteeSubnets(epoch)
|
|
epochsToSyncPeriod = nearSyncCommitteePeriod(epoch)
|
|
|
|
# The end-slot tracker might call this when it's theoretically applicable,
|
|
# but more than SYNC_COMMITTEE_SUBNET_COUNT epochs from when the next sync
|
|
# committee period begins, in which case `epochsToNextSyncPeriod` is none.
|
|
if epochsToSyncPeriod.isNone or
|
|
node.dag.cfg.consensusForkAtEpoch(epoch + epochsToSyncPeriod.get) <
|
|
ConsensusFork.Altair:
|
|
return subnets
|
|
|
|
subnets + node.getNextSyncCommitteeSubnets(epoch)
|
|
|
|
proc addAltairMessageHandlers(
|
|
node: BeaconNode, forkDigest: ForkDigest, slot: Slot) =
|
|
node.addPhase0MessageHandlers(forkDigest, slot)
|
|
|
|
# If this comes online near sync committee period, it'll immediately get
|
|
# replaced as usual by trackSyncCommitteeTopics, which runs at slot end.
|
|
let syncnets = node.getSyncCommitteeSubnets(slot.epoch)
|
|
|
|
for subcommitteeIdx in SyncSubcommitteeIndex:
|
|
if syncnets[subcommitteeIdx]:
|
|
node.network.subscribe(
|
|
getSyncCommitteeTopic(forkDigest, subcommitteeIdx), basicParams)
|
|
|
|
node.network.subscribe(
|
|
getSyncCommitteeContributionAndProofTopic(forkDigest), basicParams)
|
|
|
|
node.network.updateSyncnetsMetadata(syncnets)
|
|
|
|
proc addCapellaMessageHandlers(
|
|
node: BeaconNode, forkDigest: ForkDigest, slot: Slot) =
|
|
node.addAltairMessageHandlers(forkDigest, slot)
|
|
node.network.subscribe(getBlsToExecutionChangeTopic(forkDigest), basicParams)
|
|
|
|
proc removeAltairMessageHandlers(node: BeaconNode, forkDigest: ForkDigest) =
|
|
node.removePhase0MessageHandlers(forkDigest)
|
|
|
|
for subcommitteeIdx in SyncSubcommitteeIndex:
|
|
closureScope:
|
|
let idx = subcommitteeIdx
|
|
node.network.unsubscribe(getSyncCommitteeTopic(forkDigest, idx))
|
|
|
|
node.network.unsubscribe(
|
|
getSyncCommitteeContributionAndProofTopic(forkDigest))
|
|
|
|
proc removeCapellaMessageHandlers(node: BeaconNode, forkDigest: ForkDigest) =
|
|
node.removeAltairMessageHandlers(forkDigest)
|
|
node.network.unsubscribe(getBlsToExecutionChangeTopic(forkDigest))
|
|
|
|
proc updateSyncCommitteeTopics(node: BeaconNode, slot: Slot) =
|
|
template lastSyncUpdate: untyped =
|
|
node.consensusManager[].actionTracker.lastSyncUpdate
|
|
if lastSyncUpdate == Opt.some(slot.sync_committee_period()) and
|
|
nearSyncCommitteePeriod(slot.epoch).isNone():
|
|
# No need to update unless we're close to the next sync committee period or
|
|
# new validators were registered with the action tracker
|
|
# TODO we _could_ skip running this in some of the "near" slots, but..
|
|
return
|
|
|
|
lastSyncUpdate = Opt.some(slot.sync_committee_period())
|
|
|
|
let syncnets = node.getSyncCommitteeSubnets(slot.epoch)
|
|
|
|
debug "Updating sync committee subnets",
|
|
syncnets,
|
|
metadata_syncnets = node.network.metadata.syncnets,
|
|
gossipState = node.gossipState
|
|
|
|
# Assume that different gossip fork sync committee setups are in sync; this
|
|
# only remains relevant, currently, for one gossip transition epoch, so the
|
|
# consequences of this not being true aren't exceptionally dire, while this
|
|
# allows for bookkeeping simplication.
|
|
if syncnets == node.network.metadata.syncnets:
|
|
return
|
|
|
|
let
|
|
newSyncnets =
|
|
syncnets - node.network.metadata.syncnets
|
|
oldSyncnets =
|
|
node.network.metadata.syncnets - syncnets
|
|
forkDigests = node.forkDigests()
|
|
|
|
for subcommitteeIdx in SyncSubcommitteeIndex:
|
|
doAssert not (newSyncnets[subcommitteeIdx] and
|
|
oldSyncnets[subcommitteeIdx])
|
|
for gossipFork in node.gossipState:
|
|
template topic(): auto =
|
|
getSyncCommitteeTopic(forkDigests[gossipFork], subcommitteeIdx)
|
|
if oldSyncnets[subcommitteeIdx]:
|
|
node.network.unsubscribe(topic)
|
|
elif newSyncnets[subcommitteeIdx]:
|
|
node.network.subscribe(topic, basicParams)
|
|
|
|
node.network.updateSyncnetsMetadata(syncnets)
|
|
|
|
proc doppelgangerChecked(node: BeaconNode, epoch: Epoch) =
|
|
if not node.processor[].doppelgangerDetectionEnabled:
|
|
return
|
|
|
|
# broadcastStartEpoch is set to FAR_FUTURE_EPOCH when we're not monitoring
|
|
# gossip - it is only viable to assert liveness in epochs where gossip is
|
|
# active
|
|
if epoch > node.processor[].doppelgangerDetection.broadcastStartEpoch:
|
|
for validator in node.attachedValidators[]:
|
|
validator.doppelgangerChecked(epoch - 1)
|
|
|
|
proc updateGossipStatus(node: BeaconNode, slot: Slot) {.async.} =
|
|
## Subscribe to subnets that we are providing stability for or aggregating
|
|
## and unsubscribe from the ones that are no longer relevant.
|
|
|
|
# Let the tracker know what duties are approaching - this will tell us how
|
|
# many stability subnets we need to be subscribed to and what subnets we'll
|
|
# soon be aggregating - in addition to the in-beacon-node duties, there may
|
|
# also be duties coming from the validator client, but we don't control when
|
|
# these arrive
|
|
await node.registerDuties(slot)
|
|
|
|
# We start subscribing to gossip before we're fully synced - this allows time
|
|
# to subscribe before the sync end game
|
|
const
|
|
TOPIC_SUBSCRIBE_THRESHOLD_SLOTS = 64
|
|
HYSTERESIS_BUFFER = 16
|
|
|
|
let
|
|
head = node.dag.head
|
|
headDistance =
|
|
if slot > head.slot: (slot - head.slot).uint64
|
|
else: 0'u64
|
|
isBehind =
|
|
headDistance > TOPIC_SUBSCRIBE_THRESHOLD_SLOTS + HYSTERESIS_BUFFER
|
|
targetGossipState =
|
|
getTargetGossipState(
|
|
slot.epoch,
|
|
node.dag.cfg.ALTAIR_FORK_EPOCH,
|
|
node.dag.cfg.BELLATRIX_FORK_EPOCH,
|
|
node.dag.cfg.CAPELLA_FORK_EPOCH,
|
|
node.dag.cfg.DENEB_FORK_EPOCH,
|
|
isBehind)
|
|
|
|
doAssert targetGossipState.card <= 2
|
|
|
|
let
|
|
newGossipForks = targetGossipState - node.gossipState
|
|
oldGossipForks = node.gossipState - targetGossipState
|
|
|
|
doAssert newGossipForks.card <= 2
|
|
doAssert oldGossipForks.card <= 2
|
|
|
|
func maxGossipFork(gossipState: GossipState): int =
|
|
var res = -1
|
|
for gossipFork in gossipState:
|
|
res = max(res, gossipFork.int)
|
|
res
|
|
|
|
if maxGossipFork(targetGossipState) < maxGossipFork(node.gossipState) and
|
|
targetGossipState != {}:
|
|
warn "Unexpected clock regression during transition",
|
|
targetGossipState,
|
|
gossipState = node.gossipState
|
|
|
|
if node.gossipState.card == 0 and targetGossipState.card > 0:
|
|
# We are synced, so we will connect
|
|
debug "Enabling topic subscriptions",
|
|
wallSlot = slot,
|
|
headSlot = head.slot,
|
|
headDistance, targetGossipState
|
|
|
|
node.processor[].setupDoppelgangerDetection(slot)
|
|
|
|
# Specially when waiting for genesis, we'll already be synced on startup -
|
|
# it might also happen on a sufficiently fast restart
|
|
|
|
# We "know" the actions for the current and the next epoch
|
|
withState(node.dag.headState):
|
|
if node.consensusManager[].actionTracker.needsUpdate(
|
|
forkyState, slot.epoch):
|
|
let epochRef = node.dag.getEpochRef(head, slot.epoch, false).expect(
|
|
"Getting head EpochRef should never fail")
|
|
node.consensusManager[].actionTracker.updateActions(epochRef)
|
|
|
|
if node.consensusManager[].actionTracker.needsUpdate(
|
|
forkyState, slot.epoch + 1):
|
|
let epochRef = node.dag.getEpochRef(head, slot.epoch + 1, false).expect(
|
|
"Getting head EpochRef should never fail")
|
|
node.consensusManager[].actionTracker.updateActions(epochRef)
|
|
|
|
if node.gossipState.card > 0 and targetGossipState.card == 0:
|
|
debug "Disabling topic subscriptions",
|
|
wallSlot = slot,
|
|
headSlot = head.slot,
|
|
headDistance
|
|
|
|
node.processor[].clearDoppelgangerProtection()
|
|
|
|
let forkDigests = node.forkDigests()
|
|
|
|
const removeMessageHandlers: array[ConsensusFork, auto] = [
|
|
removePhase0MessageHandlers,
|
|
removeAltairMessageHandlers,
|
|
removeAltairMessageHandlers, # bellatrix (altair handlers, different forkDigest)
|
|
removeCapellaMessageHandlers,
|
|
removeCapellaMessageHandlers # deneb (capella handlers, different forkDigest)
|
|
]
|
|
|
|
for gossipFork in oldGossipForks:
|
|
removeMessageHandlers[gossipFork](node, forkDigests[gossipFork])
|
|
|
|
const addMessageHandlers: array[ConsensusFork, auto] = [
|
|
addPhase0MessageHandlers,
|
|
addAltairMessageHandlers,
|
|
addAltairMessageHandlers, # bellatrix (altair handlers, different forkDigest)
|
|
addCapellaMessageHandlers,
|
|
addCapellaMessageHandlers # deneb (capella handlers, different forkDigest)
|
|
]
|
|
|
|
for gossipFork in newGossipForks:
|
|
addMessageHandlers[gossipFork](node, forkDigests[gossipFork], slot)
|
|
|
|
node.gossipState = targetGossipState
|
|
node.doppelgangerChecked(slot.epoch)
|
|
node.updateAttestationSubnetHandlers(slot)
|
|
node.updateBlocksGossipStatus(slot, isBehind)
|
|
node.updateLightClientGossipStatus(slot, isBehind)
|
|
|
|
proc onSlotEnd(node: BeaconNode, slot: Slot) {.async.} =
|
|
# Things we do when slot processing has ended and we're about to wait for the
|
|
# next slot
|
|
|
|
if node.dag.needStateCachesAndForkChoicePruning():
|
|
if node.attachedValidators[].validators.len > 0:
|
|
node.attachedValidators[]
|
|
.slashingProtection
|
|
# pruning is only done if the DB is set to pruning mode.
|
|
.pruneAfterFinalization(
|
|
node.dag.finalizedHead.slot.epoch()
|
|
)
|
|
|
|
# Delay part of pruning until latency critical duties are done.
|
|
# The other part of pruning, `pruneBlocksDAG`, is done eagerly.
|
|
# ----
|
|
# This is the last pruning to do as it clears the "needPruning" condition.
|
|
node.consensusManager[].pruneStateCachesAndForkChoice()
|
|
|
|
if node.config.historyMode == HistoryMode.Prune:
|
|
node.dag.pruneHistory()
|
|
|
|
when declared(GC_fullCollect):
|
|
# The slots in the beacon node work as frames in a game: we want to make
|
|
# sure that we're ready for the next one and don't get stuck in lengthy
|
|
# garbage collection tasks when time is of essence in the middle of a slot -
|
|
# while this does not guarantee that we'll never collect during a slot, it
|
|
# makes sure that all the scratch space we used during slot tasks (logging,
|
|
# temporary buffers etc) gets recycled for the next slot that is likely to
|
|
# need similar amounts of memory.
|
|
try:
|
|
GC_fullCollect()
|
|
except Defect as exc:
|
|
raise exc # Reraise to maintain call stack
|
|
except Exception as exc:
|
|
# TODO upstream
|
|
raiseAssert "Unexpected exception during GC collection"
|
|
|
|
# Checkpoint the database to clear the WAL file and make sure changes in
|
|
# the database are synced with the filesystem.
|
|
node.db.checkpoint()
|
|
|
|
node.syncCommitteeMsgPool[].pruneData(slot)
|
|
if slot.is_epoch:
|
|
node.dynamicFeeRecipientsStore[].pruneOldMappings(slot.epoch)
|
|
|
|
# Update upcoming actions - we do this every slot in case a reorg happens
|
|
let head = node.dag.head
|
|
if node.isSynced(head) == SyncStatus.synced:
|
|
withState(node.dag.headState):
|
|
if node.consensusManager[].actionTracker.needsUpdate(
|
|
forkyState, slot.epoch + 1):
|
|
let epochRef = node.dag.getEpochRef(head, slot.epoch + 1, false).expect(
|
|
"Getting head EpochRef should never fail")
|
|
node.consensusManager[].actionTracker.updateActions(epochRef)
|
|
|
|
let
|
|
nextAttestationSlot =
|
|
node.consensusManager[].actionTracker.getNextAttestationSlot(slot)
|
|
nextProposalSlot =
|
|
node.consensusManager[].actionTracker.getNextProposalSlot(slot)
|
|
nextActionWaitTime = saturate(fromNow(
|
|
node.beaconClock, min(nextAttestationSlot, nextProposalSlot)))
|
|
|
|
# -1 is a more useful output than 18446744073709551615 as an indicator of
|
|
# no future attestation/proposal known.
|
|
template formatInt64(x: Slot): int64 =
|
|
if x == high(uint64).Slot:
|
|
-1'i64
|
|
else:
|
|
toGaugeValue(x)
|
|
|
|
template formatSyncCommitteeStatus(): string =
|
|
let slotsToNextSyncCommitteePeriod =
|
|
SLOTS_PER_SYNC_COMMITTEE_PERIOD - since_sync_committee_period_start(slot)
|
|
|
|
# int64 conversion is safe
|
|
doAssert slotsToNextSyncCommitteePeriod <= SLOTS_PER_SYNC_COMMITTEE_PERIOD
|
|
|
|
if not node.getCurrentSyncCommiteeSubnets(slot.epoch).isZeros:
|
|
"current"
|
|
elif not node.getNextSyncCommitteeSubnets(slot.epoch).isZeros:
|
|
"in " & toTimeLeftString(
|
|
SECONDS_PER_SLOT.int64.seconds * slotsToNextSyncCommitteePeriod.int64)
|
|
else:
|
|
"none"
|
|
|
|
info "Slot end",
|
|
slot = shortLog(slot),
|
|
nextActionWait =
|
|
if nextAttestationSlot == FAR_FUTURE_SLOT:
|
|
"n/a"
|
|
else:
|
|
shortLog(nextActionWaitTime),
|
|
nextAttestationSlot = formatInt64(nextAttestationSlot),
|
|
nextProposalSlot = formatInt64(nextProposalSlot),
|
|
syncCommitteeDuties = formatSyncCommitteeStatus(),
|
|
head = shortLog(head)
|
|
|
|
if nextAttestationSlot != FAR_FUTURE_SLOT:
|
|
next_action_wait.set(nextActionWaitTime.toFloatSeconds)
|
|
|
|
let epoch = slot.epoch
|
|
if epoch + 1 >= node.network.forkId.next_fork_epoch:
|
|
# Update 1 epoch early to block non-fork-ready peers
|
|
node.network.updateForkId(epoch, node.dag.genesis_validators_root)
|
|
|
|
# When we're not behind schedule, we'll speculatively update the clearance
|
|
# state in anticipation of receiving the next block - we do it after logging
|
|
# slot end since the nextActionWaitTime can be short
|
|
let
|
|
advanceCutoff = node.beaconClock.fromNow(
|
|
slot.start_beacon_time() + chronos.seconds(int(SECONDS_PER_SLOT - 1)))
|
|
if advanceCutoff.inFuture:
|
|
# We wait until there's only a second left before the next slot begins, then
|
|
# we advance the clearance state to the next slot - this gives us a high
|
|
# probability of being prepared for the block that will arrive and the
|
|
# epoch processing that follows
|
|
await sleepAsync(advanceCutoff.offset)
|
|
node.dag.advanceClearanceState()
|
|
|
|
# Prepare action tracker for the next slot
|
|
node.consensusManager[].actionTracker.updateSlot(slot + 1)
|
|
|
|
# The last thing we do is to perform the subscriptions and unsubscriptions for
|
|
# the next slot, just before that slot starts - because of the advance cuttoff
|
|
# above, this will be done just before the next slot starts
|
|
node.updateSyncCommitteeTopics(slot + 1)
|
|
|
|
await node.updateGossipStatus(slot + 1)
|
|
|
|
func syncStatus(node: BeaconNode, wallSlot: Slot): string =
|
|
let optimistic_head = node.dag.is_optimistic(node.dag.head.root)
|
|
if node.syncManager.inProgress:
|
|
let
|
|
optimisticSuffix =
|
|
if optimistic_head:
|
|
"/opt"
|
|
else:
|
|
""
|
|
lightClientSuffix =
|
|
if node.consensusManager[].shouldSyncOptimistically(wallSlot):
|
|
" - lc: " & $shortLog(node.consensusManager[].optimisticHead)
|
|
else:
|
|
""
|
|
node.syncManager.syncStatus & optimisticSuffix & lightClientSuffix
|
|
elif node.backfiller.inProgress:
|
|
"backfill: " & node.backfiller.syncStatus
|
|
elif optimistic_head:
|
|
"synced/opt"
|
|
else:
|
|
"synced"
|
|
|
|
proc onSlotStart(node: BeaconNode, wallTime: BeaconTime,
|
|
lastSlot: Slot): Future[bool] {.async.} =
|
|
## Called at the beginning of a slot - usually every slot, but sometimes might
|
|
## skip a few in case we're running late.
|
|
## wallTime: current system time - we will strive to perform all duties up
|
|
## to this point in time
|
|
## lastSlot: the last slot that we successfully processed, so we know where to
|
|
## start work from - there might be jumps if processing is delayed
|
|
let
|
|
# The slot we should be at, according to the clock
|
|
wallSlot = wallTime.slotOrZero
|
|
# If everything was working perfectly, the slot that we should be processing
|
|
expectedSlot = lastSlot + 1
|
|
finalizedEpoch = node.dag.finalizedHead.blck.slot.epoch()
|
|
delay = wallTime - expectedSlot.start_beacon_time()
|
|
|
|
info "Slot start",
|
|
slot = shortLog(wallSlot),
|
|
epoch = shortLog(wallSlot.epoch),
|
|
sync = node.syncStatus(wallSlot),
|
|
peers = len(node.network.peerPool),
|
|
head = shortLog(node.dag.head),
|
|
finalized = shortLog(getStateField(
|
|
node.dag.headState, finalized_checkpoint)),
|
|
delay = shortLog(delay)
|
|
|
|
# Check before any re-scheduling of onSlotStart()
|
|
if checkIfShouldStopAtEpoch(wallSlot, node.config.stopAtEpoch):
|
|
quit(0)
|
|
|
|
when defined(windows):
|
|
if node.config.runAsService:
|
|
reportServiceStatus(SERVICE_RUNNING, NO_ERROR, 0)
|
|
|
|
beacon_slot.set wallSlot.toGaugeValue
|
|
beacon_current_epoch.set wallSlot.epoch.toGaugeValue
|
|
|
|
# both non-negative, so difference can't overflow or underflow int64
|
|
finalization_delay.set(
|
|
wallSlot.epoch.toGaugeValue - finalizedEpoch.toGaugeValue)
|
|
|
|
if node.config.strictVerification:
|
|
verifyFinalization(node, wallSlot)
|
|
|
|
node.consensusManager[].updateHead(wallSlot)
|
|
|
|
await node.handleValidatorDuties(lastSlot, wallSlot)
|
|
|
|
await onSlotEnd(node, wallSlot)
|
|
|
|
# https://github.com/ethereum/builder-specs/blob/v0.3.0/specs/bellatrix/validator.md#registration-dissemination
|
|
# This specification suggests validators re-submit to builder software every
|
|
# `EPOCHS_PER_VALIDATOR_REGISTRATION_SUBMISSION` epochs.
|
|
if wallSlot.is_epoch and
|
|
wallSlot.epoch mod EPOCHS_PER_VALIDATOR_REGISTRATION_SUBMISSION == 0:
|
|
asyncSpawn node.registerValidators(wallSlot.epoch)
|
|
|
|
return false
|
|
|
|
proc handleMissingBlocks(node: BeaconNode) =
|
|
let missingBlocks = node.quarantine[].checkMissing()
|
|
if missingBlocks.len > 0:
|
|
debug "Requesting detected missing blocks", blocks = shortLog(missingBlocks)
|
|
node.requestManager.fetchAncestorBlocks(missingBlocks)
|
|
|
|
proc onSecond(node: BeaconNode, time: Moment) =
|
|
## This procedure will be called once per second.
|
|
if not(node.syncManager.inProgress):
|
|
node.handleMissingBlocks()
|
|
|
|
# Nim GC metrics (for the main thread)
|
|
updateThreadMetrics()
|
|
|
|
if node.config.stopAtSyncedEpoch != 0 and
|
|
node.dag.head.slot.epoch >= node.config.stopAtSyncedEpoch:
|
|
notice "Shutting down after having reached the target synced epoch"
|
|
bnStatus = BeaconNodeStatus.Stopping
|
|
|
|
proc runOnSecondLoop(node: BeaconNode) {.async.} =
|
|
const
|
|
sleepTime = chronos.seconds(1)
|
|
nanosecondsIn1s = float(sleepTime.nanoseconds)
|
|
while true:
|
|
let start = chronos.now(chronos.Moment)
|
|
await chronos.sleepAsync(sleepTime)
|
|
let afterSleep = chronos.now(chronos.Moment)
|
|
let sleepTime = afterSleep - start
|
|
node.onSecond(start)
|
|
let finished = chronos.now(chronos.Moment)
|
|
let processingTime = finished - afterSleep
|
|
ticks_delay.set(sleepTime.nanoseconds.float / nanosecondsIn1s)
|
|
trace "onSecond task completed", sleepTime, processingTime
|
|
|
|
func connectedPeersCount(node: BeaconNode): int =
|
|
len(node.network.peerPool)
|
|
|
|
proc installRestHandlers(restServer: RestServerRef, node: BeaconNode) =
|
|
restServer.router.installBeaconApiHandlers(node)
|
|
restServer.router.installConfigApiHandlers(node)
|
|
restServer.router.installDebugApiHandlers(node)
|
|
restServer.router.installEventApiHandlers(node)
|
|
restServer.router.installNimbusApiHandlers(node)
|
|
restServer.router.installNodeApiHandlers(node)
|
|
restServer.router.installValidatorApiHandlers(node)
|
|
if node.dag.lcDataStore.serve:
|
|
restServer.router.installLightClientApiHandlers(node)
|
|
|
|
from ./spec/datatypes/capella import SignedBeaconBlock
|
|
|
|
proc installMessageValidators(node: BeaconNode) =
|
|
# https://github.com/ethereum/consensus-specs/blob/v1.3.0-rc.3/specs/phase0/p2p-interface.md#attestations-and-aggregation
|
|
# These validators stay around the whole time, regardless of which specific
|
|
# subnets are subscribed to during any given epoch.
|
|
let forkDigests = node.dag.forkDigests
|
|
|
|
node.network.addValidator(
|
|
getBeaconBlocksTopic(forkDigests.phase0),
|
|
proc (signedBlock: phase0.SignedBeaconBlock): ValidationResult =
|
|
if node.shouldSyncOptimistically(node.currentSlot):
|
|
toValidationResult(
|
|
node.optimisticProcessor.processSignedBeaconBlock(signedBlock))
|
|
else:
|
|
toValidationResult(node.processor[].processSignedBeaconBlock(
|
|
MsgSource.gossip, signedBlock)))
|
|
|
|
template installPhase0Validators(digest: auto) =
|
|
for it in SubnetId:
|
|
closureScope:
|
|
let subnet_id = it
|
|
node.network.addAsyncValidator(
|
|
getAttestationTopic(digest, subnet_id),
|
|
# This proc needs to be within closureScope; don't lift out of loop.
|
|
proc(attestation: Attestation): Future[ValidationResult] {.async.} =
|
|
return toValidationResult(
|
|
await node.processor.processAttestation(
|
|
MsgSource.gossip, attestation, subnet_id)))
|
|
|
|
node.network.addAsyncValidator(
|
|
getAggregateAndProofsTopic(digest),
|
|
proc(signedAggregateAndProof: SignedAggregateAndProof):
|
|
Future[ValidationResult] {.async.} =
|
|
return toValidationResult(
|
|
await node.processor.processSignedAggregateAndProof(
|
|
MsgSource.gossip, signedAggregateAndProof, false)))
|
|
|
|
node.network.addValidator(
|
|
getAttesterSlashingsTopic(digest),
|
|
proc (attesterSlashing: AttesterSlashing): ValidationResult =
|
|
toValidationResult(
|
|
node.processor[].processAttesterSlashing(
|
|
MsgSource.gossip, attesterSlashing)))
|
|
|
|
node.network.addValidator(
|
|
getProposerSlashingsTopic(digest),
|
|
proc (proposerSlashing: ProposerSlashing): ValidationResult =
|
|
toValidationResult(
|
|
node.processor[].processProposerSlashing(
|
|
MsgSource.gossip, proposerSlashing)))
|
|
|
|
node.network.addValidator(
|
|
getVoluntaryExitsTopic(digest),
|
|
proc (signedVoluntaryExit: SignedVoluntaryExit): ValidationResult =
|
|
toValidationResult(
|
|
node.processor[].processSignedVoluntaryExit(
|
|
MsgSource.gossip, signedVoluntaryExit)))
|
|
|
|
installPhase0Validators(forkDigests.phase0)
|
|
|
|
# Validators introduced in phase0 are also used in Altair and Bellatrix, but
|
|
# with different fork digests
|
|
installPhase0Validators(forkDigests.altair)
|
|
installPhase0Validators(forkDigests.bellatrix)
|
|
installPhase0Validators(forkDigests.capella)
|
|
if node.dag.cfg.DENEB_FORK_EPOCH != FAR_FUTURE_EPOCH:
|
|
installPhase0Validators(forkDigests.deneb)
|
|
|
|
node.network.addValidator(
|
|
getBeaconBlocksTopic(forkDigests.altair),
|
|
proc (signedBlock: altair.SignedBeaconBlock): ValidationResult =
|
|
if node.shouldSyncOptimistically(node.currentSlot):
|
|
toValidationResult(
|
|
node.optimisticProcessor.processSignedBeaconBlock(signedBlock))
|
|
else:
|
|
toValidationResult(node.processor[].processSignedBeaconBlock(
|
|
MsgSource.gossip, signedBlock)))
|
|
|
|
node.network.addValidator(
|
|
getBeaconBlocksTopic(forkDigests.bellatrix),
|
|
proc (signedBlock: bellatrix.SignedBeaconBlock): ValidationResult =
|
|
if node.shouldSyncOptimistically(node.currentSlot):
|
|
toValidationResult(
|
|
node.optimisticProcessor.processSignedBeaconBlock(signedBlock))
|
|
else:
|
|
toValidationResult(node.processor[].processSignedBeaconBlock(
|
|
MsgSource.gossip, signedBlock)))
|
|
|
|
node.network.addValidator(
|
|
getBeaconBlocksTopic(forkDigests.capella),
|
|
proc (signedBlock: capella.SignedBeaconBlock): ValidationResult =
|
|
if node.shouldSyncOptimistically(node.currentSlot):
|
|
toValidationResult(
|
|
node.optimisticProcessor.processSignedBeaconBlock(signedBlock))
|
|
else:
|
|
toValidationResult(node.processor[].processSignedBeaconBlock(
|
|
MsgSource.gossip, signedBlock)))
|
|
|
|
node.network.addValidator(
|
|
getBeaconBlocksTopic(forkDigests.deneb),
|
|
proc (signedBlock: deneb.SignedBeaconBlock): ValidationResult =
|
|
if node.shouldSyncOptimistically(node.currentSlot):
|
|
toValidationResult(
|
|
node.optimisticProcessor.processSignedBeaconBlock(signedBlock))
|
|
else:
|
|
toValidationResult(node.processor[].processSignedBeaconBlock(
|
|
MsgSource.gossip, signedBlock)))
|
|
|
|
template installSyncCommitteeeValidators(digest: auto) =
|
|
for subcommitteeIdx in SyncSubcommitteeIndex:
|
|
closureScope:
|
|
let idx = subcommitteeIdx
|
|
node.network.addAsyncValidator(
|
|
getSyncCommitteeTopic(digest, idx),
|
|
# This proc needs to be within closureScope; don't lift out of loop.
|
|
proc(msg: SyncCommitteeMessage): Future[ValidationResult] {.async.} =
|
|
return toValidationResult(
|
|
await node.processor.processSyncCommitteeMessage(
|
|
MsgSource.gossip, msg, idx)))
|
|
|
|
node.network.addAsyncValidator(
|
|
getSyncCommitteeContributionAndProofTopic(digest),
|
|
proc(msg: SignedContributionAndProof): Future[ValidationResult] {.async.} =
|
|
return toValidationResult(
|
|
await node.processor.processSignedContributionAndProof(
|
|
MsgSource.gossip, msg)))
|
|
|
|
installSyncCommitteeeValidators(forkDigests.altair)
|
|
installSyncCommitteeeValidators(forkDigests.bellatrix)
|
|
installSyncCommitteeeValidators(forkDigests.capella)
|
|
if node.dag.cfg.DENEB_FORK_EPOCH != FAR_FUTURE_EPOCH:
|
|
installSyncCommitteeeValidators(forkDigests.deneb)
|
|
|
|
template installBlsToExecutionChangeValidators(digest: auto) =
|
|
node.network.addAsyncValidator(
|
|
getBlsToExecutionChangeTopic(digest),
|
|
proc(msg: SignedBLSToExecutionChange):
|
|
Future[ValidationResult] {.async.} =
|
|
return toValidationResult(
|
|
await node.processor.processBlsToExecutionChange(
|
|
MsgSource.gossip, msg)))
|
|
|
|
installBlsToExecutionChangeValidators(forkDigests.capella)
|
|
if node.dag.cfg.DENEB_FORK_EPOCH != FAR_FUTURE_EPOCH:
|
|
installBlsToExecutionChangeValidators(forkDigests.deneb)
|
|
|
|
node.installLightClientMessageValidators()
|
|
|
|
proc stop(node: BeaconNode) =
|
|
bnStatus = BeaconNodeStatus.Stopping
|
|
notice "Graceful shutdown"
|
|
if not node.config.inProcessValidators:
|
|
try:
|
|
node.vcProcess.close()
|
|
except Exception as exc:
|
|
warn "Couldn't close vc process", msg = exc.msg
|
|
try:
|
|
waitFor node.network.stop()
|
|
except CatchableError as exc:
|
|
warn "Couldn't stop network", msg = exc.msg
|
|
|
|
node.attachedValidators[].slashingProtection.close()
|
|
node.attachedValidators[].close()
|
|
node.db.close()
|
|
notice "Databases closed"
|
|
|
|
proc startBackfillTask(node: BeaconNode) {.async.} =
|
|
while node.dag.needsBackfill:
|
|
if not node.syncManager.inProgress:
|
|
# Only start the backfiller if it's needed _and_ head sync has completed -
|
|
# if we lose sync after having synced head, we could stop the backfilller,
|
|
# but this should be a fringe case - might as well keep the logic simple for
|
|
# now
|
|
node.backfiller.start()
|
|
return
|
|
|
|
await sleepAsync(chronos.seconds(2))
|
|
|
|
proc run(node: BeaconNode) {.raises: [Defect, CatchableError].} =
|
|
bnStatus = BeaconNodeStatus.Running
|
|
|
|
if not isNil(node.restServer):
|
|
node.restServer.installRestHandlers(node)
|
|
node.restServer.start()
|
|
|
|
if not isNil(node.keymanagerServer):
|
|
doAssert not isNil(node.keymanagerHost)
|
|
node.keymanagerServer.router.installKeymanagerHandlers(node.keymanagerHost[])
|
|
if node.keymanagerServer != node.restServer:
|
|
node.keymanagerServer.start()
|
|
|
|
let
|
|
wallTime = node.beaconClock.now()
|
|
wallSlot = wallTime.slotOrZero()
|
|
|
|
node.startLightClient()
|
|
node.requestManager.start()
|
|
node.syncManager.start()
|
|
|
|
if node.dag.needsBackfill(): asyncSpawn node.startBackfillTask()
|
|
|
|
waitFor node.updateGossipStatus(wallSlot)
|
|
|
|
asyncSpawn runSlotLoop(node, wallTime, onSlotStart)
|
|
asyncSpawn runOnSecondLoop(node)
|
|
asyncSpawn runQueueProcessingLoop(node.blockProcessor)
|
|
asyncSpawn runKeystoreCachePruningLoop(node.keystoreCache)
|
|
|
|
## Ctrl+C handling
|
|
proc controlCHandler() {.noconv.} =
|
|
when defined(windows):
|
|
# workaround for https://github.com/nim-lang/Nim/issues/4057
|
|
try:
|
|
setupForeignThreadGc()
|
|
except Exception as exc: raiseAssert exc.msg # shouldn't happen
|
|
notice "Shutting down after having received SIGINT"
|
|
bnStatus = BeaconNodeStatus.Stopping
|
|
try:
|
|
setControlCHook(controlCHandler)
|
|
except Exception as exc: # TODO Exception
|
|
warn "Cannot set ctrl-c handler", msg = exc.msg
|
|
|
|
# equivalent SIGTERM handler
|
|
when defined(posix):
|
|
proc SIGTERMHandler(signal: cint) {.noconv.} =
|
|
notice "Shutting down after having received SIGTERM"
|
|
bnStatus = BeaconNodeStatus.Stopping
|
|
c_signal(ansi_c.SIGTERM, SIGTERMHandler)
|
|
|
|
# main event loop
|
|
while bnStatus == BeaconNodeStatus.Running:
|
|
poll() # if poll fails, the network is broken
|
|
|
|
# time to say goodbye
|
|
node.stop()
|
|
|
|
var gPidFile: string
|
|
proc createPidFile(filename: string) {.raises: [Defect, IOError].} =
|
|
writeFile filename, $os.getCurrentProcessId()
|
|
gPidFile = filename
|
|
addQuitProc proc {.noconv.} = discard io2.removeFile(gPidFile)
|
|
|
|
proc initializeNetworking(node: BeaconNode) {.async.} =
|
|
node.installMessageValidators()
|
|
|
|
info "Listening to incoming network requests"
|
|
await node.network.startListening()
|
|
|
|
let addressFile = node.config.dataDir / "beacon_node.enr"
|
|
writeFile(addressFile, node.network.announcedENR.toURI)
|
|
|
|
await node.network.start()
|
|
|
|
proc start*(node: BeaconNode) {.raises: [Defect, CatchableError].} =
|
|
let
|
|
head = node.dag.head
|
|
finalizedHead = node.dag.finalizedHead
|
|
genesisTime = node.beaconClock.fromNow(start_beacon_time(Slot 0))
|
|
|
|
notice "Starting beacon node",
|
|
version = fullVersionStr,
|
|
nimVersion = NimVersion,
|
|
enr = node.network.announcedENR.toURI,
|
|
peerId = $node.network.switch.peerInfo.peerId,
|
|
timeSinceFinalization =
|
|
node.beaconClock.now() - finalizedHead.slot.start_beacon_time(),
|
|
head = shortLog(head),
|
|
justified = shortLog(getStateField(
|
|
node.dag.headState, current_justified_checkpoint)),
|
|
finalized = shortLog(getStateField(
|
|
node.dag.headState, finalized_checkpoint)),
|
|
finalizedHead = shortLog(finalizedHead),
|
|
SLOTS_PER_EPOCH,
|
|
SECONDS_PER_SLOT,
|
|
SPEC_VERSION,
|
|
dataDir = node.config.dataDir.string,
|
|
validators = node.attachedValidators[].count
|
|
|
|
if genesisTime.inFuture:
|
|
notice "Waiting for genesis", genesisIn = genesisTime.offset
|
|
|
|
waitFor node.initializeNetworking()
|
|
|
|
node.elManager.start()
|
|
node.run()
|
|
|
|
func formatGwei(amount: uint64): string =
|
|
# TODO This is implemented in a quite a silly way.
|
|
# Better routines for formatting decimal numbers
|
|
# should exists somewhere else.
|
|
let
|
|
eth = amount div 1000000000
|
|
remainder = amount mod 1000000000
|
|
|
|
result = $eth
|
|
if remainder != 0:
|
|
result.add '.'
|
|
let remainderStr = $remainder
|
|
for i in remainderStr.len ..< 9:
|
|
result.add '0'
|
|
result.add remainderStr
|
|
while result[^1] == '0':
|
|
result.setLen(result.len - 1)
|
|
|
|
when not defined(windows):
|
|
proc initStatusBar(node: BeaconNode) {.raises: [Defect, ValueError].} =
|
|
if not isatty(stdout): return
|
|
if not node.config.statusBarEnabled: return
|
|
|
|
try:
|
|
enableTrueColors()
|
|
except Exception as exc: # TODO Exception
|
|
error "Couldn't enable colors", err = exc.msg
|
|
|
|
proc dataResolver(expr: string): string {.raises: [Defect].} =
|
|
template justified: untyped = node.dag.head.atEpochStart(
|
|
getStateField(
|
|
node.dag.headState, current_justified_checkpoint).epoch)
|
|
# TODO:
|
|
# We should introduce a general API for resolving dot expressions
|
|
# such as `db.latest_block.slot` or `metrics.connected_peers`.
|
|
# Such an API can be shared between the RPC back-end, CLI tools
|
|
# such as ncli, a potential GraphQL back-end and so on.
|
|
# The status bar feature would allow the user to specify an
|
|
# arbitrary expression that is resolvable through this API.
|
|
case expr.toLowerAscii
|
|
of "version":
|
|
versionAsStr
|
|
|
|
of "full_version":
|
|
fullVersionStr
|
|
|
|
of "connected_peers":
|
|
$(node.connectedPeersCount)
|
|
|
|
of "head_root":
|
|
shortLog(node.dag.head.root)
|
|
of "head_epoch":
|
|
$(node.dag.head.slot.epoch)
|
|
of "head_epoch_slot":
|
|
$(node.dag.head.slot.since_epoch_start)
|
|
of "head_slot":
|
|
$(node.dag.head.slot)
|
|
|
|
of "justifed_root":
|
|
shortLog(justified.blck.root)
|
|
of "justifed_epoch":
|
|
$(justified.slot.epoch)
|
|
of "justifed_epoch_slot":
|
|
$(justified.slot.since_epoch_start)
|
|
of "justifed_slot":
|
|
$(justified.slot)
|
|
|
|
of "finalized_root":
|
|
shortLog(node.dag.finalizedHead.blck.root)
|
|
of "finalized_epoch":
|
|
$(node.dag.finalizedHead.slot.epoch)
|
|
of "finalized_epoch_slot":
|
|
$(node.dag.finalizedHead.slot.since_epoch_start)
|
|
of "finalized_slot":
|
|
$(node.dag.finalizedHead.slot)
|
|
|
|
of "epoch":
|
|
$node.currentSlot.epoch
|
|
|
|
of "epoch_slot":
|
|
$(node.currentSlot.since_epoch_start)
|
|
|
|
of "slot":
|
|
$node.currentSlot
|
|
|
|
of "slots_per_epoch":
|
|
$SLOTS_PER_EPOCH
|
|
|
|
of "slot_trailing_digits":
|
|
var slotStr = $node.currentSlot
|
|
if slotStr.len > 3: slotStr = slotStr[^3..^1]
|
|
slotStr
|
|
|
|
of "attached_validators_balance":
|
|
formatGwei(node.attachedValidatorBalanceTotal)
|
|
|
|
of "sync_status":
|
|
node.syncStatus(node.currentSlot)
|
|
else:
|
|
# We ignore typos for now and just render the expression
|
|
# as it was written. TODO: come up with a good way to show
|
|
# an error message to the user.
|
|
"$" & expr
|
|
|
|
var statusBar = StatusBarView.init(
|
|
node.config.statusBarContents,
|
|
dataResolver)
|
|
|
|
when compiles(defaultChroniclesStream.outputs[0].writer):
|
|
let tmp = defaultChroniclesStream.outputs[0].writer
|
|
|
|
defaultChroniclesStream.outputs[0].writer =
|
|
proc (logLevel: LogLevel, msg: LogOutputStr) {.raises: [Defect].} =
|
|
try:
|
|
# p.hidePrompt
|
|
erase statusBar
|
|
# p.writeLine msg
|
|
tmp(logLevel, msg)
|
|
render statusBar
|
|
# p.showPrompt
|
|
except Exception as e: # render raises Exception
|
|
logLoggingFailure(cstring(msg), e)
|
|
|
|
proc statusBarUpdatesPollingLoop() {.async.} =
|
|
try:
|
|
while true:
|
|
update statusBar
|
|
erase statusBar
|
|
render statusBar
|
|
await sleepAsync(chronos.seconds(1))
|
|
except CatchableError as exc:
|
|
warn "Failed to update status bar, no further updates", err = exc.msg
|
|
|
|
asyncSpawn statusBarUpdatesPollingLoop()
|
|
|
|
proc doRunBeaconNode(config: var BeaconNodeConf, rng: ref HmacDrbgContext) {.raises: [Defect, CatchableError].} =
|
|
info "Launching beacon node",
|
|
version = fullVersionStr,
|
|
bls_backend = $BLS_BACKEND,
|
|
cmdParams = commandLineParams(),
|
|
config
|
|
|
|
template ignoreDeprecatedOption(option: untyped): untyped =
|
|
if config.option.isSome:
|
|
warn "Config option is deprecated",
|
|
option = config.option.get
|
|
ignoreDeprecatedOption requireEngineAPI
|
|
ignoreDeprecatedOption safeSlotsToImportOptimistically
|
|
ignoreDeprecatedOption terminalTotalDifficultyOverride
|
|
ignoreDeprecatedOption optimistic
|
|
ignoreDeprecatedOption validatorMonitorTotals
|
|
|
|
createPidFile(config.dataDir.string / "beacon_node.pid")
|
|
|
|
config.createDumpDirs()
|
|
|
|
if config.metricsEnabled:
|
|
let metricsAddress = config.metricsAddress
|
|
notice "Starting metrics HTTP server",
|
|
url = "http://" & $metricsAddress & ":" & $config.metricsPort & "/metrics"
|
|
try:
|
|
startMetricsHttpServer($metricsAddress, config.metricsPort)
|
|
except CatchableError as exc:
|
|
raise exc
|
|
except Exception as exc:
|
|
raiseAssert exc.msg # TODO fix metrics
|
|
|
|
# Nim GC metrics (for the main thread) will be collected in onSecond(), but
|
|
# we disable piggy-backing on other metrics here.
|
|
setSystemMetricsAutomaticUpdate(false)
|
|
|
|
# There are no managed event loops in here, to do a graceful shutdown, but
|
|
# letting the default Ctrl+C handler exit is safe, since we only read from
|
|
# the db.
|
|
let metadata = config.loadEth2Network()
|
|
|
|
# Updating the config based on the metadata certainly is not beautiful but it
|
|
# works
|
|
for node in metadata.bootstrapNodes:
|
|
config.bootstrapNodes.add node
|
|
if not (metadata.cfg.CAPELLA_FORK_EPOCH == FAR_FUTURE_EPOCH or
|
|
config.deploymentPhase == DeploymentPhase.None):
|
|
config.deploymentPhase = DeploymentPhase.CapellaReady
|
|
|
|
let node = BeaconNode.init(rng, config, metadata)
|
|
|
|
if bnStatus == BeaconNodeStatus.Stopping:
|
|
return
|
|
|
|
when not defined(windows):
|
|
# This status bar can lock a Windows terminal emulator, blocking the whole
|
|
# event loop (seen on Windows 10, with a default MSYS2 terminal).
|
|
initStatusBar(node)
|
|
|
|
if node.nickname != "":
|
|
dynamicLogScope(node = node.nickname): node.start()
|
|
else:
|
|
node.start()
|
|
|
|
proc doRecord(config: BeaconNodeConf, rng: var HmacDrbgContext) {.
|
|
raises: [Defect, CatchableError].} =
|
|
case config.recordCmd:
|
|
of RecordCmd.create:
|
|
let netKeys = getPersistentNetKeys(rng, config)
|
|
|
|
var fieldPairs: seq[FieldPair]
|
|
for field in config.fields:
|
|
let fieldPair = field.split(":")
|
|
if fieldPair.len > 1:
|
|
fieldPairs.add(toFieldPair(fieldPair[0], hexToSeqByte(fieldPair[1])))
|
|
else:
|
|
fatal "Invalid field pair"
|
|
quit QuitFailure
|
|
|
|
let record = enr.Record.init(
|
|
config.seqNumber,
|
|
netKeys.seckey.asEthKey,
|
|
some(config.ipExt),
|
|
some(config.tcpPortExt),
|
|
some(config.udpPortExt),
|
|
fieldPairs).expect("Record within size limits")
|
|
|
|
echo record.toURI()
|
|
|
|
of RecordCmd.print:
|
|
echo $config.recordPrint
|
|
|
|
proc doWeb3Cmd(config: BeaconNodeConf, rng: var HmacDrbgContext)
|
|
{.raises: [Defect, CatchableError].} =
|
|
case config.web3Cmd:
|
|
of Web3Cmd.test:
|
|
let metadata = config.loadEth2Network()
|
|
|
|
waitFor testWeb3Provider(config.web3TestUrl,
|
|
metadata.cfg.DEPOSIT_CONTRACT_ADDRESS,
|
|
rng.loadJwtSecret(config, allowCreate = true))
|
|
|
|
proc doSlashingExport(conf: BeaconNodeConf) {.raises: [IOError, Defect].}=
|
|
let
|
|
dir = conf.validatorsDir()
|
|
filetrunc = SlashingDbName
|
|
# TODO: Make it read-only https://github.com/status-im/nim-eth/issues/312
|
|
let db = SlashingProtectionDB.loadUnchecked(dir, filetrunc, readOnly = false)
|
|
|
|
let interchange = conf.exportedInterchangeFile.string
|
|
db.exportSlashingInterchange(interchange, conf.exportedValidators)
|
|
echo "Export finished: '", dir/filetrunc & ".sqlite3" , "' into '", interchange, "'"
|
|
|
|
proc doSlashingImport(conf: BeaconNodeConf) {.raises: [SerializationError, IOError, Defect].} =
|
|
let
|
|
dir = conf.validatorsDir()
|
|
filetrunc = SlashingDbName
|
|
# TODO: Make it read-only https://github.com/status-im/nim-eth/issues/312
|
|
|
|
let interchange = conf.importedInterchangeFile.string
|
|
|
|
var spdir: SPDIR
|
|
try:
|
|
spdir = Json.loadFile(interchange, SPDIR,
|
|
requireAllFields = true)
|
|
except SerializationError as err:
|
|
writeStackTrace()
|
|
stderr.write $Json & " load issue for file \"", interchange, "\"\n"
|
|
stderr.write err.formatMsg(interchange), "\n"
|
|
quit 1
|
|
|
|
# Open DB and handle migration from v1 to v2 if needed
|
|
let db = SlashingProtectionDB.init(
|
|
genesis_validators_root = Eth2Digest spdir.metadata.genesis_validators_root,
|
|
basePath = dir,
|
|
dbname = filetrunc,
|
|
modes = {kCompleteArchive}
|
|
)
|
|
|
|
# Now import the slashing interchange file
|
|
# Failures mode:
|
|
# - siError can only happen with invalid genesis_validators_root which would be caught above
|
|
# - siPartial can happen for invalid public keys, slashable blocks, slashable votes
|
|
let status = db.inclSPDIR(spdir)
|
|
doAssert status in {siSuccess, siPartial}
|
|
|
|
echo "Import finished: '", interchange, "' into '", dir/filetrunc & ".sqlite3", "'"
|
|
|
|
proc doSlashingInterchange(conf: BeaconNodeConf) {.raises: [Defect, CatchableError].} =
|
|
case conf.slashingdbCmd
|
|
of SlashProtCmd.`export`:
|
|
conf.doSlashingExport()
|
|
of SlashProtCmd.`import`:
|
|
conf.doSlashingImport()
|
|
|
|
proc handleStartUpCmd(config: var BeaconNodeConf) {.raises: [Defect, CatchableError].} =
|
|
# Single RNG instance for the application - will be seeded on construction
|
|
# and avoid using system resources (such as urandom) after that
|
|
let rng = keys.newRng()
|
|
|
|
case config.cmd
|
|
of BNStartUpCmd.noCommand: doRunBeaconNode(config, rng)
|
|
of BNStartUpCmd.deposits: doDeposits(config, rng[])
|
|
of BNStartUpCmd.wallets: doWallets(config, rng[])
|
|
of BNStartUpCmd.record: doRecord(config, rng[])
|
|
of BNStartUpCmd.web3: doWeb3Cmd(config, rng[])
|
|
of BNStartUpCmd.slashingdb: doSlashingInterchange(config)
|
|
of BNStartUpCmd.trustedNodeSync:
|
|
let
|
|
network = loadEth2Network(config)
|
|
cfg = network.cfg
|
|
genesis =
|
|
if network.genesisData.len > 0:
|
|
newClone(readSszForkedHashedBeaconState(
|
|
cfg,
|
|
network.genesisData.toOpenArrayByte(0, network.genesisData.high())))
|
|
else: nil
|
|
|
|
if config.blockId.isSome():
|
|
error "--blockId option has been removed - use --state-id instead!"
|
|
quit 1
|
|
|
|
waitFor doTrustedNodeSync(
|
|
cfg,
|
|
config.databaseDir,
|
|
config.eraDir,
|
|
config.trustedNodeUrl,
|
|
config.stateId,
|
|
config.backfillBlocks,
|
|
config.reindex,
|
|
config.downloadDepositSnapshot,
|
|
genesis)
|
|
|
|
{.pop.} # TODO moduletests exceptions
|
|
|
|
when defined(windows):
|
|
proc reportServiceStatus*(dwCurrentState, dwWin32ExitCode, dwWaitHint: DWORD) {.gcsafe.} =
|
|
gSvcStatus.dwCurrentState = dwCurrentState
|
|
gSvcStatus.dwWin32ExitCode = dwWin32ExitCode
|
|
gSvcStatus.dwWaitHint = dwWaitHint
|
|
if dwCurrentState == SERVICE_START_PENDING:
|
|
gSvcStatus.dwControlsAccepted = 0
|
|
else:
|
|
gSvcStatus.dwControlsAccepted = SERVICE_ACCEPT_STOP
|
|
|
|
# TODO
|
|
# We can use non-zero values for the `dwCheckPoint` parameter to report
|
|
# progress during lengthy operations such as start-up and shut down.
|
|
gSvcStatus.dwCheckPoint = 0
|
|
|
|
# Report the status of the service to the SCM.
|
|
let status = SetServiceStatus(gSvcStatusHandle, addr gSvcStatus)
|
|
debug "Service status updated", status
|
|
|
|
proc serviceControlHandler(dwCtrl: DWORD): WINBOOL {.stdcall.} =
|
|
case dwCtrl
|
|
of SERVICE_CONTROL_STOP:
|
|
# We re reporting that we plan stop the service in 10 seconds
|
|
reportServiceStatus(SERVICE_STOP_PENDING, NO_ERROR, 10_000)
|
|
bnStatus = BeaconNodeStatus.Stopping
|
|
of SERVICE_CONTROL_PAUSE, SERVICE_CONTROL_CONTINUE:
|
|
warn "The Nimbus service cannot be paused and resimed"
|
|
of SERVICE_CONTROL_INTERROGATE:
|
|
# The default behavior is correct.
|
|
# The service control manager will report our last status.
|
|
discard
|
|
else:
|
|
debug "Service received an unexpected user-defined control message",
|
|
msg = dwCtrl
|
|
|
|
proc serviceMainFunction(dwArgc: DWORD, lpszArgv: LPSTR) {.stdcall.} =
|
|
# The service is launched in a fresh thread created by Windows, so
|
|
# we must initialize the Nim GC here
|
|
setupForeignThreadGc()
|
|
|
|
gSvcStatusHandle = RegisterServiceCtrlHandler(
|
|
SERVICE_NAME,
|
|
serviceControlHandler)
|
|
|
|
gSvcStatus.dwServiceType = SERVICE_WIN32_OWN_PROCESS
|
|
gSvcStatus.dwServiceSpecificExitCode = 0
|
|
reportServiceStatus(SERVICE_RUNNING, NO_ERROR, 0)
|
|
|
|
info "Service thread started"
|
|
|
|
var config = makeBannerAndConfig(clientId, BeaconNodeConf)
|
|
handleStartUpCmd(config)
|
|
|
|
info "Service thread stopped"
|
|
reportServiceStatus(SERVICE_STOPPED, NO_ERROR, 0) # we have to report back when we stopped!
|
|
|
|
programMain:
|
|
var config = makeBannerAndConfig(clientId, BeaconNodeConf)
|
|
|
|
if not(checkAndCreateDataDir(string(config.dataDir))):
|
|
# We are unable to access/create data folder or data folder's
|
|
# permissions are insecure.
|
|
quit QuitFailure
|
|
|
|
setupLogging(config.logLevel, config.logStdout, config.logFile)
|
|
|
|
## This Ctrl+C handler exits the program in non-graceful way.
|
|
## It's responsible for handling Ctrl+C in sub-commands such
|
|
## as `wallets *` and `deposits *`. In a regular beacon node
|
|
## run, it will be overwritten later with a different handler
|
|
## performing a graceful exit.
|
|
proc exitImmediatelyOnCtrlC() {.noconv.} =
|
|
when defined(windows):
|
|
# workaround for https://github.com/nim-lang/Nim/issues/4057
|
|
setupForeignThreadGc()
|
|
# in case a password prompt disabled echoing
|
|
resetStdin()
|
|
echo "" # If we interrupt during an interactive prompt, this
|
|
# will move the cursor to the next line
|
|
notice "Shutting down after having received SIGINT"
|
|
quit 0
|
|
setControlCHook(exitImmediatelyOnCtrlC)
|
|
# equivalent SIGTERM handler
|
|
when defined(posix):
|
|
proc exitImmediatelyOnSIGTERM(signal: cint) {.noconv.} =
|
|
notice "Shutting down after having received SIGTERM"
|
|
quit 0
|
|
c_signal(ansi_c.SIGTERM, exitImmediatelyOnSIGTERM)
|
|
|
|
when defined(windows):
|
|
if config.runAsService:
|
|
var dispatchTable = [
|
|
SERVICE_TABLE_ENTRY(lpServiceName: SERVICE_NAME, lpServiceProc: serviceMainFunction),
|
|
SERVICE_TABLE_ENTRY(lpServiceName: nil, lpServiceProc: nil) # last entry must be nil
|
|
]
|
|
|
|
let status = StartServiceCtrlDispatcher(LPSERVICE_TABLE_ENTRY(addr dispatchTable[0]))
|
|
if status == 0:
|
|
fatal "Failed to start Windows service", errorCode = getLastError()
|
|
quit 1
|
|
else:
|
|
handleStartUpCmd(config)
|
|
else:
|
|
handleStartUpCmd(config)
|