nimbus-eth2/beacon_chain/trusted_node_sync.nim

342 lines
12 KiB
Nim
Raw Normal View History

# Copyright (c) 2018-2022 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.
2022-07-29 10:53:42 +00:00
when (NimMajor, NimMinor) < (1, 4):
{.push raises: [Defect].}
else:
{.push raises: [].}
import
stew/[base10, results],
chronicles, chronos, eth/async_utils,
./sync/sync_manager,
./consensus_object_pools/[block_clearance, blockchain_dag],
./spec/eth2_apis/rest_beacon_client,
./spec/[beaconstate, eth2_merkleization, forks, presets,
state_transition, deposit_snapshots],
"."/[beacon_clock, beacon_chain_db, era_db]
from presto import RestDecodingError
const
largeRequestsTimeout = 60.seconds # Downloading large items such as states.
smallRequestsTimeout = 30.seconds # Downloading smaller items such as blocks and deposit snapshots.
proc fetchDepositSnapshot(client: RestClientRef):
Future[Result[DepositTreeSnapshot, string]] {.async.} =
let resp = try:
awaitWithTimeout(client.getDepositSnapshot(), smallRequestsTimeout):
return err "Fetching /eth/v1/beacon/deposit_snapshot timed out"
except CatchableError as e:
return err("The trusted node likely does not support the /eth/v1/beacon/deposit_snapshot end-point:" & e.msg)
let data = resp.data.data
let snapshot = DepositTreeSnapshot(
eth1Block: data.execution_block_hash,
depositContractState: DepositContractState(
branch: data.finalized,
deposit_count: depositCountBytes(data.deposit_count)),
blockHeight: data.execution_block_height)
if not snapshot.isValid(data.deposit_root):
return err "The obtained deposit snapshot contains self-contradictory data"
return ok snapshot
from ./spec/datatypes/eip4844 import asSigVerified, shortLog
proc doTrustedNodeSync*(
cfg: RuntimeConfig,
databaseDir: string,
eraDir: string,
restUrl: string,
stateId: string,
backfill: bool,
reindex: bool,
downloadDepositSnapshot: bool,
genesisState: ref ForkedHashedBeaconState = nil) {.async.} =
State-only checkpoint state startup (#4251) Currently, we require genesis and a checkpoint block and state to start from an arbitrary slot - this PR relaxes this requirement so that we can start with a state alone. The current trusted-node-sync algorithm works by first downloading blocks until we find an epoch aligned non-empty slot, then downloads the state via slot. However, current [proposals](https://github.com/ethereum/beacon-APIs/pull/226) for checkpointing prefer finalized state as the main reference - this allows more simple access control and caching on the server side - in particular, this should help checkpoint-syncing from sources that have a fast `finalized` state download (like infura and teku) but are slow when accessing state via slot. Earlier versions of Nimbus will not be able to read databases created without a checkpoint block and genesis. In most cases, backfilling makes the database compatible except where genesis is also missing (custom networks). * backfill checkpoint block from libp2p instead of checkpoint source, when doing trusted node sync * allow starting the client without genesis / checkpoint block * perform epoch start slot lookahead when loading tail state, so as to deal with the case where the epoch start slot does not have a block * replace `--blockId` with `--state-id` in TNS command line * when replaying, also look at the parent of the last-known-block (even if we don't have the parent block data, we can still replay from a "parent" state) - in particular, this clears the way for implementing state pruning * deprecate `--finalized-checkpoint-block` option (no longer needed)
2022-11-02 10:02:38 +00:00
logScope:
restUrl
stateId
notice "Starting trusted node sync",
State-only checkpoint state startup (#4251) Currently, we require genesis and a checkpoint block and state to start from an arbitrary slot - this PR relaxes this requirement so that we can start with a state alone. The current trusted-node-sync algorithm works by first downloading blocks until we find an epoch aligned non-empty slot, then downloads the state via slot. However, current [proposals](https://github.com/ethereum/beacon-APIs/pull/226) for checkpointing prefer finalized state as the main reference - this allows more simple access control and caching on the server side - in particular, this should help checkpoint-syncing from sources that have a fast `finalized` state download (like infura and teku) but are slow when accessing state via slot. Earlier versions of Nimbus will not be able to read databases created without a checkpoint block and genesis. In most cases, backfilling makes the database compatible except where genesis is also missing (custom networks). * backfill checkpoint block from libp2p instead of checkpoint source, when doing trusted node sync * allow starting the client without genesis / checkpoint block * perform epoch start slot lookahead when loading tail state, so as to deal with the case where the epoch start slot does not have a block * replace `--blockId` with `--state-id` in TNS command line * when replaying, also look at the parent of the last-known-block (even if we don't have the parent block data, we can still replay from a "parent" state) - in particular, this clears the way for implementing state pruning * deprecate `--finalized-checkpoint-block` option (no longer needed)
2022-11-02 10:02:38 +00:00
databaseDir, backfill, reindex
var
client = RestClientRef.new(restUrl).valueOr:
error "Cannot connect to server", error = error
quit 1
let
db = BeaconChainDB.new(databaseDir, cfg, inMemory = false)
defer:
db.close()
State-only checkpoint state startup (#4251) Currently, we require genesis and a checkpoint block and state to start from an arbitrary slot - this PR relaxes this requirement so that we can start with a state alone. The current trusted-node-sync algorithm works by first downloading blocks until we find an epoch aligned non-empty slot, then downloads the state via slot. However, current [proposals](https://github.com/ethereum/beacon-APIs/pull/226) for checkpointing prefer finalized state as the main reference - this allows more simple access control and caching on the server side - in particular, this should help checkpoint-syncing from sources that have a fast `finalized` state download (like infura and teku) but are slow when accessing state via slot. Earlier versions of Nimbus will not be able to read databases created without a checkpoint block and genesis. In most cases, backfilling makes the database compatible except where genesis is also missing (custom networks). * backfill checkpoint block from libp2p instead of checkpoint source, when doing trusted node sync * allow starting the client without genesis / checkpoint block * perform epoch start slot lookahead when loading tail state, so as to deal with the case where the epoch start slot does not have a block * replace `--blockId` with `--state-id` in TNS command line * when replaying, also look at the parent of the last-known-block (even if we don't have the parent block data, we can still replay from a "parent" state) - in particular, this clears the way for implementing state pruning * deprecate `--finalized-checkpoint-block` option (no longer needed)
2022-11-02 10:02:38 +00:00
# If possible, we'll store the genesis state in the database - this is not
# strictly necessary but renders the resulting database compatible with
# versions prior to 22.11 and makes reindexing possible
let genesisState =
if (let genesisRoot = db.getGenesisBlock(); genesisRoot.isSome()):
let
genesisBlock = db.getForkedBlock(genesisRoot.get()).valueOr:
error "Cannot load genesis block from database",
genesisRoot = genesisRoot.get()
quit 1
State-only checkpoint state startup (#4251) Currently, we require genesis and a checkpoint block and state to start from an arbitrary slot - this PR relaxes this requirement so that we can start with a state alone. The current trusted-node-sync algorithm works by first downloading blocks until we find an epoch aligned non-empty slot, then downloads the state via slot. However, current [proposals](https://github.com/ethereum/beacon-APIs/pull/226) for checkpointing prefer finalized state as the main reference - this allows more simple access control and caching on the server side - in particular, this should help checkpoint-syncing from sources that have a fast `finalized` state download (like infura and teku) but are slow when accessing state via slot. Earlier versions of Nimbus will not be able to read databases created without a checkpoint block and genesis. In most cases, backfilling makes the database compatible except where genesis is also missing (custom networks). * backfill checkpoint block from libp2p instead of checkpoint source, when doing trusted node sync * allow starting the client without genesis / checkpoint block * perform epoch start slot lookahead when loading tail state, so as to deal with the case where the epoch start slot does not have a block * replace `--blockId` with `--state-id` in TNS command line * when replaying, also look at the parent of the last-known-block (even if we don't have the parent block data, we can still replay from a "parent" state) - in particular, this clears the way for implementing state pruning * deprecate `--finalized-checkpoint-block` option (no longer needed)
2022-11-02 10:02:38 +00:00
genesisStateRoot = getForkedBlockField(genesisBlock, state_root)
stateFork = cfg.stateForkAtEpoch(GENESIS_EPOCH)
State-only checkpoint state startup (#4251) Currently, we require genesis and a checkpoint block and state to start from an arbitrary slot - this PR relaxes this requirement so that we can start with a state alone. The current trusted-node-sync algorithm works by first downloading blocks until we find an epoch aligned non-empty slot, then downloads the state via slot. However, current [proposals](https://github.com/ethereum/beacon-APIs/pull/226) for checkpointing prefer finalized state as the main reference - this allows more simple access control and caching on the server side - in particular, this should help checkpoint-syncing from sources that have a fast `finalized` state download (like infura and teku) but are slow when accessing state via slot. Earlier versions of Nimbus will not be able to read databases created without a checkpoint block and genesis. In most cases, backfilling makes the database compatible except where genesis is also missing (custom networks). * backfill checkpoint block from libp2p instead of checkpoint source, when doing trusted node sync * allow starting the client without genesis / checkpoint block * perform epoch start slot lookahead when loading tail state, so as to deal with the case where the epoch start slot does not have a block * replace `--blockId` with `--state-id` in TNS command line * when replaying, also look at the parent of the last-known-block (even if we don't have the parent block data, we can still replay from a "parent" state) - in particular, this clears the way for implementing state pruning * deprecate `--finalized-checkpoint-block` option (no longer needed)
2022-11-02 10:02:38 +00:00
tmp = (ref ForkedHashedBeaconState)(kind: stateFork)
if not db.getState(stateFork, genesisStateRoot, tmp[], noRollback):
error "Cannot load genesis state from database",
genesisStateRoot
quit 1
if (genesisState != nil) and
(getStateRoot(tmp[]) != getStateRoot(genesisState[])):
error "Unexpected genesis state in database, is this the same network?",
databaseRoot = getStateRoot(tmp[]),
genesisRoot = getStateRoot(genesisState[])
quit 1
tmp
else:
let tmp = if genesisState != nil:
genesisState
else:
notice "Downloading genesis state", restUrl
try:
awaitWithTimeout(
client.getStateV2(StateIdent.init(StateIdentType.Genesis), cfg),
largeRequestsTimeout):
info "Attempt to download genesis state timed out"
nil
except CatchableError as exc:
State-only checkpoint state startup (#4251) Currently, we require genesis and a checkpoint block and state to start from an arbitrary slot - this PR relaxes this requirement so that we can start with a state alone. The current trusted-node-sync algorithm works by first downloading blocks until we find an epoch aligned non-empty slot, then downloads the state via slot. However, current [proposals](https://github.com/ethereum/beacon-APIs/pull/226) for checkpointing prefer finalized state as the main reference - this allows more simple access control and caching on the server side - in particular, this should help checkpoint-syncing from sources that have a fast `finalized` state download (like infura and teku) but are slow when accessing state via slot. Earlier versions of Nimbus will not be able to read databases created without a checkpoint block and genesis. In most cases, backfilling makes the database compatible except where genesis is also missing (custom networks). * backfill checkpoint block from libp2p instead of checkpoint source, when doing trusted node sync * allow starting the client without genesis / checkpoint block * perform epoch start slot lookahead when loading tail state, so as to deal with the case where the epoch start slot does not have a block * replace `--blockId` with `--state-id` in TNS command line * when replaying, also look at the parent of the last-known-block (even if we don't have the parent block data, we can still replay from a "parent" state) - in particular, this clears the way for implementing state pruning * deprecate `--finalized-checkpoint-block` option (no longer needed)
2022-11-02 10:02:38 +00:00
info "Unable to download genesis state",
error = exc.msg, restUrl
State-only checkpoint state startup (#4251) Currently, we require genesis and a checkpoint block and state to start from an arbitrary slot - this PR relaxes this requirement so that we can start with a state alone. The current trusted-node-sync algorithm works by first downloading blocks until we find an epoch aligned non-empty slot, then downloads the state via slot. However, current [proposals](https://github.com/ethereum/beacon-APIs/pull/226) for checkpointing prefer finalized state as the main reference - this allows more simple access control and caching on the server side - in particular, this should help checkpoint-syncing from sources that have a fast `finalized` state download (like infura and teku) but are slow when accessing state via slot. Earlier versions of Nimbus will not be able to read databases created without a checkpoint block and genesis. In most cases, backfilling makes the database compatible except where genesis is also missing (custom networks). * backfill checkpoint block from libp2p instead of checkpoint source, when doing trusted node sync * allow starting the client without genesis / checkpoint block * perform epoch start slot lookahead when loading tail state, so as to deal with the case where the epoch start slot does not have a block * replace `--blockId` with `--state-id` in TNS command line * when replaying, also look at the parent of the last-known-block (even if we don't have the parent block data, we can still replay from a "parent" state) - in particular, this clears the way for implementing state pruning * deprecate `--finalized-checkpoint-block` option (no longer needed)
2022-11-02 10:02:38 +00:00
nil
State-only checkpoint state startup (#4251) Currently, we require genesis and a checkpoint block and state to start from an arbitrary slot - this PR relaxes this requirement so that we can start with a state alone. The current trusted-node-sync algorithm works by first downloading blocks until we find an epoch aligned non-empty slot, then downloads the state via slot. However, current [proposals](https://github.com/ethereum/beacon-APIs/pull/226) for checkpointing prefer finalized state as the main reference - this allows more simple access control and caching on the server side - in particular, this should help checkpoint-syncing from sources that have a fast `finalized` state download (like infura and teku) but are slow when accessing state via slot. Earlier versions of Nimbus will not be able to read databases created without a checkpoint block and genesis. In most cases, backfilling makes the database compatible except where genesis is also missing (custom networks). * backfill checkpoint block from libp2p instead of checkpoint source, when doing trusted node sync * allow starting the client without genesis / checkpoint block * perform epoch start slot lookahead when loading tail state, so as to deal with the case where the epoch start slot does not have a block * replace `--blockId` with `--state-id` in TNS command line * when replaying, also look at the parent of the last-known-block (even if we don't have the parent block data, we can still replay from a "parent" state) - in particular, this clears the way for implementing state pruning * deprecate `--finalized-checkpoint-block` option (no longer needed)
2022-11-02 10:02:38 +00:00
if isNil(tmp):
notice "Server is missing genesis state, node will not be able to reindex history",
restUrl
tmp
let
dbHead = db.getHeadBlock()
head = if dbHead.isSome():
let
bid = db.getBlockId(dbHead.get()).valueOr:
error "Database missing head block summary - database too old or corrupt",
headRoot = dbHead.get()
quit 1
Opt.some bid
else:
# When we don't have a head, we'll use the given checkpoint as head
Opt.none(BlockId)
if head.isNone:
State-only checkpoint state startup (#4251) Currently, we require genesis and a checkpoint block and state to start from an arbitrary slot - this PR relaxes this requirement so that we can start with a state alone. The current trusted-node-sync algorithm works by first downloading blocks until we find an epoch aligned non-empty slot, then downloads the state via slot. However, current [proposals](https://github.com/ethereum/beacon-APIs/pull/226) for checkpointing prefer finalized state as the main reference - this allows more simple access control and caching on the server side - in particular, this should help checkpoint-syncing from sources that have a fast `finalized` state download (like infura and teku) but are slow when accessing state via slot. Earlier versions of Nimbus will not be able to read databases created without a checkpoint block and genesis. In most cases, backfilling makes the database compatible except where genesis is also missing (custom networks). * backfill checkpoint block from libp2p instead of checkpoint source, when doing trusted node sync * allow starting the client without genesis / checkpoint block * perform epoch start slot lookahead when loading tail state, so as to deal with the case where the epoch start slot does not have a block * replace `--blockId` with `--state-id` in TNS command line * when replaying, also look at the parent of the last-known-block (even if we don't have the parent block data, we can still replay from a "parent" state) - in particular, this clears the way for implementing state pruning * deprecate `--finalized-checkpoint-block` option (no longer needed)
2022-11-02 10:02:38 +00:00
notice "Downloading checkpoint state"
State-only checkpoint state startup (#4251) Currently, we require genesis and a checkpoint block and state to start from an arbitrary slot - this PR relaxes this requirement so that we can start with a state alone. The current trusted-node-sync algorithm works by first downloading blocks until we find an epoch aligned non-empty slot, then downloads the state via slot. However, current [proposals](https://github.com/ethereum/beacon-APIs/pull/226) for checkpointing prefer finalized state as the main reference - this allows more simple access control and caching on the server side - in particular, this should help checkpoint-syncing from sources that have a fast `finalized` state download (like infura and teku) but are slow when accessing state via slot. Earlier versions of Nimbus will not be able to read databases created without a checkpoint block and genesis. In most cases, backfilling makes the database compatible except where genesis is also missing (custom networks). * backfill checkpoint block from libp2p instead of checkpoint source, when doing trusted node sync * allow starting the client without genesis / checkpoint block * perform epoch start slot lookahead when loading tail state, so as to deal with the case where the epoch start slot does not have a block * replace `--blockId` with `--state-id` in TNS command line * when replaying, also look at the parent of the last-known-block (even if we don't have the parent block data, we can still replay from a "parent" state) - in particular, this clears the way for implementing state pruning * deprecate `--finalized-checkpoint-block` option (no longer needed)
2022-11-02 10:02:38 +00:00
let
state = try:
let id = block:
let tmp = StateIdent.decodeString(stateId).valueOr:
error "Cannot decode checkpoint state id, must be a slot, hash, 'finalized' or 'head'"
quit 1
if tmp.kind == StateQueryKind.Slot and not tmp.slot.is_epoch():
notice "Rounding given slot to epoch"
StateIdent.init(tmp.slot.epoch().start_slot)
else:
tmp
awaitWithTimeout(client.getStateV2(id, cfg), largeRequestsTimeout):
error "Attempt to download checkpoint state timed out"
quit 1
State-only checkpoint state startup (#4251) Currently, we require genesis and a checkpoint block and state to start from an arbitrary slot - this PR relaxes this requirement so that we can start with a state alone. The current trusted-node-sync algorithm works by first downloading blocks until we find an epoch aligned non-empty slot, then downloads the state via slot. However, current [proposals](https://github.com/ethereum/beacon-APIs/pull/226) for checkpointing prefer finalized state as the main reference - this allows more simple access control and caching on the server side - in particular, this should help checkpoint-syncing from sources that have a fast `finalized` state download (like infura and teku) but are slow when accessing state via slot. Earlier versions of Nimbus will not be able to read databases created without a checkpoint block and genesis. In most cases, backfilling makes the database compatible except where genesis is also missing (custom networks). * backfill checkpoint block from libp2p instead of checkpoint source, when doing trusted node sync * allow starting the client without genesis / checkpoint block * perform epoch start slot lookahead when loading tail state, so as to deal with the case where the epoch start slot does not have a block * replace `--blockId` with `--state-id` in TNS command line * when replaying, also look at the parent of the last-known-block (even if we don't have the parent block data, we can still replay from a "parent" state) - in particular, this clears the way for implementing state pruning * deprecate `--finalized-checkpoint-block` option (no longer needed)
2022-11-02 10:02:38 +00:00
except CatchableError as exc:
error "Unable to download checkpoint state",
error = exc.msg
quit 1
State-only checkpoint state startup (#4251) Currently, we require genesis and a checkpoint block and state to start from an arbitrary slot - this PR relaxes this requirement so that we can start with a state alone. The current trusted-node-sync algorithm works by first downloading blocks until we find an epoch aligned non-empty slot, then downloads the state via slot. However, current [proposals](https://github.com/ethereum/beacon-APIs/pull/226) for checkpointing prefer finalized state as the main reference - this allows more simple access control and caching on the server side - in particular, this should help checkpoint-syncing from sources that have a fast `finalized` state download (like infura and teku) but are slow when accessing state via slot. Earlier versions of Nimbus will not be able to read databases created without a checkpoint block and genesis. In most cases, backfilling makes the database compatible except where genesis is also missing (custom networks). * backfill checkpoint block from libp2p instead of checkpoint source, when doing trusted node sync * allow starting the client without genesis / checkpoint block * perform epoch start slot lookahead when loading tail state, so as to deal with the case where the epoch start slot does not have a block * replace `--blockId` with `--state-id` in TNS command line * when replaying, also look at the parent of the last-known-block (even if we don't have the parent block data, we can still replay from a "parent" state) - in particular, this clears the way for implementing state pruning * deprecate `--finalized-checkpoint-block` option (no longer needed)
2022-11-02 10:02:38 +00:00
if state == nil:
error "No state found a given checkpoint",
stateId
quit 1
State-only checkpoint state startup (#4251) Currently, we require genesis and a checkpoint block and state to start from an arbitrary slot - this PR relaxes this requirement so that we can start with a state alone. The current trusted-node-sync algorithm works by first downloading blocks until we find an epoch aligned non-empty slot, then downloads the state via slot. However, current [proposals](https://github.com/ethereum/beacon-APIs/pull/226) for checkpointing prefer finalized state as the main reference - this allows more simple access control and caching on the server side - in particular, this should help checkpoint-syncing from sources that have a fast `finalized` state download (like infura and teku) but are slow when accessing state via slot. Earlier versions of Nimbus will not be able to read databases created without a checkpoint block and genesis. In most cases, backfilling makes the database compatible except where genesis is also missing (custom networks). * backfill checkpoint block from libp2p instead of checkpoint source, when doing trusted node sync * allow starting the client without genesis / checkpoint block * perform epoch start slot lookahead when loading tail state, so as to deal with the case where the epoch start slot does not have a block * replace `--blockId` with `--state-id` in TNS command line * when replaying, also look at the parent of the last-known-block (even if we don't have the parent block data, we can still replay from a "parent" state) - in particular, this clears the way for implementing state pruning * deprecate `--finalized-checkpoint-block` option (no longer needed)
2022-11-02 10:02:38 +00:00
if not getStateField(state[], slot).is_epoch():
error "State slot must fall on an epoch boundary",
slot = getStateField(state[], slot),
offset = getStateField(state[], slot) -
getStateField(state[], slot).epoch.start_slot
quit 1
State-only checkpoint state startup (#4251) Currently, we require genesis and a checkpoint block and state to start from an arbitrary slot - this PR relaxes this requirement so that we can start with a state alone. The current trusted-node-sync algorithm works by first downloading blocks until we find an epoch aligned non-empty slot, then downloads the state via slot. However, current [proposals](https://github.com/ethereum/beacon-APIs/pull/226) for checkpointing prefer finalized state as the main reference - this allows more simple access control and caching on the server side - in particular, this should help checkpoint-syncing from sources that have a fast `finalized` state download (like infura and teku) but are slow when accessing state via slot. Earlier versions of Nimbus will not be able to read databases created without a checkpoint block and genesis. In most cases, backfilling makes the database compatible except where genesis is also missing (custom networks). * backfill checkpoint block from libp2p instead of checkpoint source, when doing trusted node sync * allow starting the client without genesis / checkpoint block * perform epoch start slot lookahead when loading tail state, so as to deal with the case where the epoch start slot does not have a block * replace `--blockId` with `--state-id` in TNS command line * when replaying, also look at the parent of the last-known-block (even if we don't have the parent block data, we can still replay from a "parent" state) - in particular, this clears the way for implementing state pruning * deprecate `--finalized-checkpoint-block` option (no longer needed)
2022-11-02 10:02:38 +00:00
if genesisState != nil:
if getStateField(state[], genesis_validators_root) !=
getStateField(genesisState[], genesis_validators_root):
error "Checkpoint state does not match genesis",
rootInCheckpoint = getStateField(state[], genesis_validators_root),
rootInGenesis = getStateField(genesisState[], genesis_validators_root)
quit 1
State-only checkpoint state startup (#4251) Currently, we require genesis and a checkpoint block and state to start from an arbitrary slot - this PR relaxes this requirement so that we can start with a state alone. The current trusted-node-sync algorithm works by first downloading blocks until we find an epoch aligned non-empty slot, then downloads the state via slot. However, current [proposals](https://github.com/ethereum/beacon-APIs/pull/226) for checkpointing prefer finalized state as the main reference - this allows more simple access control and caching on the server side - in particular, this should help checkpoint-syncing from sources that have a fast `finalized` state download (like infura and teku) but are slow when accessing state via slot. Earlier versions of Nimbus will not be able to read databases created without a checkpoint block and genesis. In most cases, backfilling makes the database compatible except where genesis is also missing (custom networks). * backfill checkpoint block from libp2p instead of checkpoint source, when doing trusted node sync * allow starting the client without genesis / checkpoint block * perform epoch start slot lookahead when loading tail state, so as to deal with the case where the epoch start slot does not have a block * replace `--blockId` with `--state-id` in TNS command line * when replaying, also look at the parent of the last-known-block (even if we don't have the parent block data, we can still replay from a "parent" state) - in particular, this clears the way for implementing state pruning * deprecate `--finalized-checkpoint-block` option (no longer needed)
2022-11-02 10:02:38 +00:00
ChainDAGRef.preInit(db, genesisState[])
State-only checkpoint state startup (#4251) Currently, we require genesis and a checkpoint block and state to start from an arbitrary slot - this PR relaxes this requirement so that we can start with a state alone. The current trusted-node-sync algorithm works by first downloading blocks until we find an epoch aligned non-empty slot, then downloads the state via slot. However, current [proposals](https://github.com/ethereum/beacon-APIs/pull/226) for checkpointing prefer finalized state as the main reference - this allows more simple access control and caching on the server side - in particular, this should help checkpoint-syncing from sources that have a fast `finalized` state download (like infura and teku) but are slow when accessing state via slot. Earlier versions of Nimbus will not be able to read databases created without a checkpoint block and genesis. In most cases, backfilling makes the database compatible except where genesis is also missing (custom networks). * backfill checkpoint block from libp2p instead of checkpoint source, when doing trusted node sync * allow starting the client without genesis / checkpoint block * perform epoch start slot lookahead when loading tail state, so as to deal with the case where the epoch start slot does not have a block * replace `--blockId` with `--state-id` in TNS command line * when replaying, also look at the parent of the last-known-block (even if we don't have the parent block data, we can still replay from a "parent" state) - in particular, this clears the way for implementing state pruning * deprecate `--finalized-checkpoint-block` option (no longer needed)
2022-11-02 10:02:38 +00:00
if getStateField(genesisState[], slot) != getStateField(state[], slot):
ChainDAGRef.preInit(db, state[])
else:
ChainDAGRef.preInit(db, state[])
if downloadDepositSnapshot:
# Fetch deposit snapshot. This API endpoint is still optional.
let depositSnapshot = await fetchDepositSnapshot(client)
if depositSnapshot.isOk:
if depositSnapshot.get.matches(getStateField(state[], eth1_data)):
info "Writing deposit contracts snapshot",
depositRoot = depositSnapshot.get.getDepositRoot(),
depositCount = depositSnapshot.get.getDepositCountU64
db.putDepositTreeSnapshot(depositSnapshot.get)
else:
warn "The downloaded deposit snapshot does not agree with the downloaded state"
else:
warn "Deposit tree snapshot was not imported", reason = depositSnapshot.error
else:
notice "Skipping checkpoint download, database already exists (remove db directory to get a fresh snapshot)",
databaseDir, head = shortLog(head.get())
# Coming this far, we've done what ChainDAGRef.preInit would normally do -
# we can now load a ChainDAG to start backfilling it
State-only checkpoint state startup (#4251) Currently, we require genesis and a checkpoint block and state to start from an arbitrary slot - this PR relaxes this requirement so that we can start with a state alone. The current trusted-node-sync algorithm works by first downloading blocks until we find an epoch aligned non-empty slot, then downloads the state via slot. However, current [proposals](https://github.com/ethereum/beacon-APIs/pull/226) for checkpointing prefer finalized state as the main reference - this allows more simple access control and caching on the server side - in particular, this should help checkpoint-syncing from sources that have a fast `finalized` state download (like infura and teku) but are slow when accessing state via slot. Earlier versions of Nimbus will not be able to read databases created without a checkpoint block and genesis. In most cases, backfilling makes the database compatible except where genesis is also missing (custom networks). * backfill checkpoint block from libp2p instead of checkpoint source, when doing trusted node sync * allow starting the client without genesis / checkpoint block * perform epoch start slot lookahead when loading tail state, so as to deal with the case where the epoch start slot does not have a block * replace `--blockId` with `--state-id` in TNS command line * when replaying, also look at the parent of the last-known-block (even if we don't have the parent block data, we can still replay from a "parent" state) - in particular, this clears the way for implementing state pruning * deprecate `--finalized-checkpoint-block` option (no longer needed)
2022-11-02 10:02:38 +00:00
let
validatorMonitor = newClone(ValidatorMonitor.init(false, false))
dag = ChainDAGRef.init(cfg, db, validatorMonitor, {}, eraPath = eraDir)
backfillSlot = dag.backfill.slot
horizon = max(dag.horizon, dag.frontfill.valueOr(BlockId()).slot)
let canReindex = if backfillSlot <= horizon:
info "Database backfilled", backfill = dag.backfill, horizon
true
elif backfill:
# +1 because we need to download the frontfill slot for the frontfill match
# detection to kick in, in addBackfillBlock
let missingSlots = dag.backfill.slot - horizon + 1
notice "Downloading historical blocks - you can interrupt this process at any time and it automatically be completed when you start the beacon node",
backfillSlot, horizon, missingSlots
var # Same averaging as SyncManager
syncCount = 0
processed = 0'u64
avgSyncSpeed = 0.0
stamp = SyncMoment.now(0)
proc downloadBlock(slot: Slot):
Future[Option[ref ForkedSignedBeaconBlock]] {.async.} =
# Download block at given slot, retrying a few times,
var lastError: ref CatchableError
for i in 0..<3:
try:
return awaitWithTimeout(client.getBlockV2(BlockIdent.init(slot), cfg),
smallRequestsTimeout):
raise newException(CatchableError, "Request timed out")
except RestResponseError as exc:
lastError = exc
notice "Server does not support block downloads / backfilling - blocks will be downloaded later",
msg = exc.msg
break
except CatchableError as exc:
# We'll assume this may be a connectivity error or something similar
lastError = exc
warn "Retrying download of block", slot, err = exc.msg
client = RestClientRef.new(restUrl).valueOr:
error "Cannot connect to server", url = restUrl, error = error
quit 1
raise lastError
# Download several blocks in parallel but process them serially
proc processBlock(blck: Option[ref ForkedSignedBeaconBlock]) =
let newStamp = SyncMoment.now(processed)
if newStamp.stamp - stamp.stamp > 12.seconds:
syncCount += 1
let
remaining = dag.backfill.slot - horizon
slotsPerSec = speed(stamp, newStamp)
avgSyncSpeed = avgSyncSpeed + (slotsPerSec - avgSyncSpeed) / float(syncCount)
info "Backfilling",
timeleft = toTimeLeftString(
if avgSyncSpeed >= 0.001:
Duration.fromFloatSeconds(remaining.float / avgSyncSpeed)
else: InfiniteDuration),
slotsPerSecond = avgSyncSpeed,
remainingSlots = remaining
stamp = newStamp
processed += 1
if blck.isSome():
let
data = blck.get()
withBlck(data[]):
if (let res = dag.addBackfillBlock(blck.asSigVerified()); res.isErr()):
case res.error()
of VerifierError.Invalid,
VerifierError.MissingParent,
VerifierError.UnviableFork:
error "Got invalid block from trusted node - is it on the right network?",
blck = shortLog(blck), err = res.error()
quit 1
of VerifierError.Duplicate:
discard
# Download blocks backwards from the backfill slot, ie the first slot for
# which we don't have a block, when walking backwards from the head
try:
var
gets: array[16, Future[Option[ref ForkedSignedBeaconBlock]]]
for i in 0.uint64..missingSlots + gets.lenu64:
if i >= gets.lenu64():
let
fut = gets[int(i mod gets.lenu64)]
processBlock(await fut)
if i <= backfillSlot:
let slot = backfillSlot - i
gets[int(i mod gets.lenu64)] = downloadBlock(slot)
if i mod 1024 == 0:
db.checkpoint() # Transfer stuff from wal periodically
true
except CatchableError as exc: # Block download failed
notice "Backfilling incomplete - blocks will be downloaded when starting the node", msg = exc.msg
false
else:
let missingSlots = dag.backfill.slot - horizon
notice "Database initialized, historical blocks will be backfilled when starting the node",
missingSlots, backfill = dag.backfill, horizon
false
if reindex and canReindex:
notice "Reindexing historical state lookup tables (you can interrupt this process at any time)"
# Build a DAG
dag.rebuildIndex()
notice "Done, your beacon node is ready to serve you! Don't forget to check that you're on the canonical chain by comparing the checkpoint root with other online sources. See https://nimbus.guide/trusted-node-sync.html for more information.",
checkpoint = dag.head
when isMainModule:
import
std/[os],
networking/network_metadata
limit by-root requests to non-finalized blocks (#3293) * limit by-root requests to non-finalized blocks Presently, we keep a mapping from block root to `BlockRef` in memory - this has simplified reasoning about the dag, but is not sustainable with the chain growing. We can distinguish between two cases where by-root access is useful: * unfinalized blocks - this is where the beacon chain is operating generally, by validating incoming data as interesting for future fork choice decisions - bounded by the length of the unfinalized period * finalized blocks - historical access in the REST API etc - no bounds, really In this PR, we limit the by-root block index to the first use case: finalized chain data can more efficiently be addressed by slot number. Future work includes: * limiting the `BlockRef` horizon in general - each instance is 40 bytes+overhead which adds up - this needs further refactoring to deal with the tail vs state problem * persisting the finalized slot-to-hash index - this one also keeps growing unbounded (albeit slowly) Anyway, this PR easily shaves ~128mb of memory usage at the time of writing. * No longer honor `BeaconBlocksByRoot` requests outside of the non-finalized period - previously, Nimbus would generously return any block through this libp2p request - per the spec, finalized blocks should be fetched via `BeaconBlocksByRange` instead. * return `Opt[BlockRef]` instead of `nil` when blocks can't be found - this becomes a lot more common now and thus deserves more attention * `dag.blocks` -> `dag.forkBlocks` - this index only carries unfinalized blocks from now - `finalizedBlocks` covers the other `BlockRef` instances * in backfill, verify that the last backfilled block leads back to genesis, or panic * add backfill timings to log * fix missing check that `BlockRef` block can be fetched with `getForkedBlock` reliably * shortcut doppelganger check when feature is not enabled * in REST/JSON-RPC, fetch blocks without involving `BlockRef` * fix dag.blocks ref
2022-01-21 11:33:16 +00:00
let backfill = os.paramCount() > 5 and os.paramStr(6) == "true"
waitFor doTrustedNodeSync(
State-only checkpoint state startup (#4251) Currently, we require genesis and a checkpoint block and state to start from an arbitrary slot - this PR relaxes this requirement so that we can start with a state alone. The current trusted-node-sync algorithm works by first downloading blocks until we find an epoch aligned non-empty slot, then downloads the state via slot. However, current [proposals](https://github.com/ethereum/beacon-APIs/pull/226) for checkpointing prefer finalized state as the main reference - this allows more simple access control and caching on the server side - in particular, this should help checkpoint-syncing from sources that have a fast `finalized` state download (like infura and teku) but are slow when accessing state via slot. Earlier versions of Nimbus will not be able to read databases created without a checkpoint block and genesis. In most cases, backfilling makes the database compatible except where genesis is also missing (custom networks). * backfill checkpoint block from libp2p instead of checkpoint source, when doing trusted node sync * allow starting the client without genesis / checkpoint block * perform epoch start slot lookahead when loading tail state, so as to deal with the case where the epoch start slot does not have a block * replace `--blockId` with `--state-id` in TNS command line * when replaying, also look at the parent of the last-known-block (even if we don't have the parent block data, we can still replay from a "parent" state) - in particular, this clears the way for implementing state pruning * deprecate `--finalized-checkpoint-block` option (no longer needed)
2022-11-02 10:02:38 +00:00
getRuntimeConfig(some os.paramStr(1)), os.paramStr(2), os.paramStr(3),
os.paramStr(4), os.paramStr(5), backfill, false, true)