nimbus-eth2/ncli/ncli_db.nim

1217 lines
41 KiB
Nim
Raw Normal View History

2022-04-14 10:47:14 +00:00
# beacon_chain
# Copyright (c) 2020-2024 Status Research & Development GmbH
2022-04-14 10:47:14 +00:00
# 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, stats, tables],
snappy,
chronicles, confutils, stew/[byteutils, io2], eth/db/kvstore_sqlite3,
../beacon_chain/networking/network_metadata,
../beacon_chain/[beacon_chain_db, era_db],
../beacon_chain/consensus_object_pools/[blockchain_dag],
../beacon_chain/spec/[
beaconstate, state_transition, state_transition_epoch, validator,
ssz_codec],
disentangle eth2 types from the ssz library (#2785) * reorganize ssz dependencies This PR continues the work in https://github.com/status-im/nimbus-eth2/pull/2646, https://github.com/status-im/nimbus-eth2/pull/2779 as well as past issues with serialization and type, to disentangle SSZ from eth2 and at the same time simplify imports and exports with a structured approach. The principal idea here is that when a library wants to introduce SSZ support, they do so via 3 files: * `ssz_codecs` which imports and reexports `codecs` - this covers the basic byte conversions and ensures no overloads get lost * `xxx_merkleization` imports and exports `merkleization` to specialize and get access to `hash_tree_root` and friends * `xxx_ssz_serialization` imports and exports `ssz_serialization` to specialize ssz for a specific library Those that need to interact with SSZ always import the `xxx_` versions of the modules and never `ssz` itself so as to keep imports simple and safe. This is similar to how the REST / JSON-RPC serializers are structured in that someone wanting to serialize spec types to REST-JSON will import `eth2_rest_serialization` and nothing else. * split up ssz into a core library that is independendent of eth2 types * rename `bytes_reader` to `codec` to highlight that it contains coding and decoding of bytes and native ssz types * remove tricky List init overload that causes compile issues * get rid of top-level ssz import * reenable merkleization tests * move some "standard" json serializers to spec * remove `ValidatorIndex` serialization for now * remove test_ssz_merkleization * add tests for over/underlong byte sequences * fix broken seq[byte] test - seq[byte] is not an SSZ type There are a few things this PR doesn't solve: * like #2646 this PR is weak on how to handle root and other dontSerialize fields that "sometimes" should be computed - the same problem appears in REST / JSON-RPC etc * Fix a build problem on macOS * Another way to fix the macOS builds Co-authored-by: Zahary Karadjov <zahary@gmail.com>
2021-08-18 18:57:58 +00:00
../beacon_chain/sszdump,
../research/simutils,
./era, ./ncli_common, ./validator_db_aggregator
when defined(posix):
import system/ansi_c
type Timers = enum
tInit = "Initialize DB"
tLoadBlock = "Load block from database"
tLoadState = "Load state from database"
performance fixes (#2259) * performance fixes * don't mark tree cache as dirty on read-only List accesses * store only blob in memory for keys and signatures, parse blob lazily * compare public keys by blob instead of parsing / converting to raw * compare Eth2Digest using non-constant-time comparison * avoid some unnecessary validator copying This branch will in particular speed up deposit processing which has been slowing down block replay. Pre (mainnet, 1600 blocks): ``` All time are ms Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 3450.269, 0.000, 3450.269, 3450.269, 1, Initialize DB 0.417, 0.822, 0.036, 21.098, 1400, Load block from database 16.521, 0.000, 16.521, 16.521, 1, Load state from database 27.906, 50.846, 8.104, 1507.633, 1350, Apply block 52.617, 37.029, 20.640, 135.938, 50, Apply epoch block ``` Post: ``` 3502.715, 0.000, 3502.715, 3502.715, 1, Initialize DB 0.080, 0.560, 0.035, 21.015, 1400, Load block from database 17.595, 0.000, 17.595, 17.595, 1, Load state from database 15.706, 11.028, 8.300, 107.537, 1350, Apply block 33.217, 12.622, 17.331, 60.580, 50, Apply epoch block ``` * more perf fixes * load EpochRef cache into StateCache more aggressively * point out security concern with public key cache * reuse proposer index from state when processing block * avoid genericAssign in a few more places * don't parse key when signature is unparseable * fix `==` overload for Eth2Digest * preallocate validator list when getting active validators * speed up proposer index calculation a little bit * reuse cache when replaying blocks in ncli_db * avoid a few more copying loops ``` Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 3279.158, 0.000, 3279.158, 3279.158, 1, Initialize DB 0.072, 0.357, 0.035, 13.400, 1400, Load block from database 17.295, 0.000, 17.295, 17.295, 1, Load state from database 5.918, 9.896, 0.198, 98.028, 1350, Apply block 15.888, 10.951, 7.902, 39.535, 50, Apply epoch block 0.000, 0.000, 0.000, 0.000, 0, Database block store ``` * clear full balance cache before processing rewards and penalties ``` All time are ms Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 3947.901, 0.000, 3947.901, 3947.901, 1, Initialize DB 0.124, 0.506, 0.026, 202.370, 363345, Load block from database 97.614, 0.000, 97.614, 97.614, 1, Load state from database 0.186, 0.188, 0.012, 99.561, 357262, Advance slot, non-epoch 14.161, 5.966, 1.099, 395.511, 11524, Advance slot, epoch 1.372, 4.170, 0.017, 276.401, 363345, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database block store ```
2021-01-25 12:04:18 +00:00
tAdvanceSlot = "Advance slot, non-epoch"
tAdvanceEpoch = "Advance slot, epoch"
tApplyBlock = "Apply block, no slot processing"
tDbLoad = "Database load"
tDbStore = "Database store"
type
DbCmd* {.pure.} = enum
bench = "Run a replay benchmark for block and epoch processing"
dumpState = "Extract a state from the database as-is - only works for states that have been explicitly stored"
putState = "Store a given BeaconState in the database"
dumpBlock = "Extract a (trusted) SignedBeaconBlock from the database"
putBlock = "Store a given SignedBeaconBlock in the database, potentially updating some of the pointers"
putBlob = "Store a given BlobSidecar in the database"
rewindState = "Extract any state from the database based on a given block and slot, replaying if needed"
verifyEra = "Verify a single era file"
exportEra = "Export historical data to era store in current directory"
importEra = "Import era files to the database"
validatorPerf
validatorDb = "Create or update attestation performance database"
DbConf = object
databaseDir* {.
defaultValue: "db"
desc: "Directory where `nbc.sqlite` is stored"
name: "db".}: InputDir
eraDir* {.
defaultValue: "era"
desc: "Directory where era files are read from"
name: "era-dir".}: string
eth2Network* {.
desc: "The Eth2 network preset to use"
name: "network".}: Option[string]
case cmd* {.
command
desc: ""
.}: DbCmd
of DbCmd.bench:
benchSlot* {.
defaultValue: 0
name: "start-slot"
desc: "Starting slot, negative = backwards from head".}: int64
benchSlots* {.
defaultValue: 50000
name: "slots"
desc: "Number of slots to run benchmark for, 0 = all the way to head".}: uint64
storeBlocks* {.
defaultValue: false
desc: "Store each read block back into a separate database".}: bool
storeStates* {.
defaultValue: false
name: "store-states"
desc: "Store a state each epoch into a separate database".}: bool
performance fixes (#2259) * performance fixes * don't mark tree cache as dirty on read-only List accesses * store only blob in memory for keys and signatures, parse blob lazily * compare public keys by blob instead of parsing / converting to raw * compare Eth2Digest using non-constant-time comparison * avoid some unnecessary validator copying This branch will in particular speed up deposit processing which has been slowing down block replay. Pre (mainnet, 1600 blocks): ``` All time are ms Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 3450.269, 0.000, 3450.269, 3450.269, 1, Initialize DB 0.417, 0.822, 0.036, 21.098, 1400, Load block from database 16.521, 0.000, 16.521, 16.521, 1, Load state from database 27.906, 50.846, 8.104, 1507.633, 1350, Apply block 52.617, 37.029, 20.640, 135.938, 50, Apply epoch block ``` Post: ``` 3502.715, 0.000, 3502.715, 3502.715, 1, Initialize DB 0.080, 0.560, 0.035, 21.015, 1400, Load block from database 17.595, 0.000, 17.595, 17.595, 1, Load state from database 15.706, 11.028, 8.300, 107.537, 1350, Apply block 33.217, 12.622, 17.331, 60.580, 50, Apply epoch block ``` * more perf fixes * load EpochRef cache into StateCache more aggressively * point out security concern with public key cache * reuse proposer index from state when processing block * avoid genericAssign in a few more places * don't parse key when signature is unparseable * fix `==` overload for Eth2Digest * preallocate validator list when getting active validators * speed up proposer index calculation a little bit * reuse cache when replaying blocks in ncli_db * avoid a few more copying loops ``` Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 3279.158, 0.000, 3279.158, 3279.158, 1, Initialize DB 0.072, 0.357, 0.035, 13.400, 1400, Load block from database 17.295, 0.000, 17.295, 17.295, 1, Load state from database 5.918, 9.896, 0.198, 98.028, 1350, Apply block 15.888, 10.951, 7.902, 39.535, 50, Apply epoch block 0.000, 0.000, 0.000, 0.000, 0, Database block store ``` * clear full balance cache before processing rewards and penalties ``` All time are ms Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 3947.901, 0.000, 3947.901, 3947.901, 1, Initialize DB 0.124, 0.506, 0.026, 202.370, 363345, Load block from database 97.614, 0.000, 97.614, 97.614, 1, Load state from database 0.186, 0.188, 0.012, 99.561, 357262, Advance slot, non-epoch 14.161, 5.966, 1.099, 395.511, 11524, Advance slot, epoch 1.372, 4.170, 0.017, 276.401, 363345, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database block store ```
2021-01-25 12:04:18 +00:00
printTimes* {.
defaultValue: true
name: "print-times"
performance fixes (#2259) * performance fixes * don't mark tree cache as dirty on read-only List accesses * store only blob in memory for keys and signatures, parse blob lazily * compare public keys by blob instead of parsing / converting to raw * compare Eth2Digest using non-constant-time comparison * avoid some unnecessary validator copying This branch will in particular speed up deposit processing which has been slowing down block replay. Pre (mainnet, 1600 blocks): ``` All time are ms Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 3450.269, 0.000, 3450.269, 3450.269, 1, Initialize DB 0.417, 0.822, 0.036, 21.098, 1400, Load block from database 16.521, 0.000, 16.521, 16.521, 1, Load state from database 27.906, 50.846, 8.104, 1507.633, 1350, Apply block 52.617, 37.029, 20.640, 135.938, 50, Apply epoch block ``` Post: ``` 3502.715, 0.000, 3502.715, 3502.715, 1, Initialize DB 0.080, 0.560, 0.035, 21.015, 1400, Load block from database 17.595, 0.000, 17.595, 17.595, 1, Load state from database 15.706, 11.028, 8.300, 107.537, 1350, Apply block 33.217, 12.622, 17.331, 60.580, 50, Apply epoch block ``` * more perf fixes * load EpochRef cache into StateCache more aggressively * point out security concern with public key cache * reuse proposer index from state when processing block * avoid genericAssign in a few more places * don't parse key when signature is unparseable * fix `==` overload for Eth2Digest * preallocate validator list when getting active validators * speed up proposer index calculation a little bit * reuse cache when replaying blocks in ncli_db * avoid a few more copying loops ``` Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 3279.158, 0.000, 3279.158, 3279.158, 1, Initialize DB 0.072, 0.357, 0.035, 13.400, 1400, Load block from database 17.295, 0.000, 17.295, 17.295, 1, Load state from database 5.918, 9.896, 0.198, 98.028, 1350, Apply block 15.888, 10.951, 7.902, 39.535, 50, Apply epoch block 0.000, 0.000, 0.000, 0.000, 0, Database block store ``` * clear full balance cache before processing rewards and penalties ``` All time are ms Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 3947.901, 0.000, 3947.901, 3947.901, 1, Initialize DB 0.124, 0.506, 0.026, 202.370, 363345, Load block from database 97.614, 0.000, 97.614, 97.614, 1, Load state from database 0.186, 0.188, 0.012, 99.561, 357262, Advance slot, non-epoch 14.161, 5.966, 1.099, 395.511, 11524, Advance slot, epoch 1.372, 4.170, 0.017, 276.401, 363345, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database block store ```
2021-01-25 12:04:18 +00:00
desc: "Print csv of block processing time".}: bool
resetCache* {.
defaultValue: false
name: "reset-cache"
performance fixes (#2259) * performance fixes * don't mark tree cache as dirty on read-only List accesses * store only blob in memory for keys and signatures, parse blob lazily * compare public keys by blob instead of parsing / converting to raw * compare Eth2Digest using non-constant-time comparison * avoid some unnecessary validator copying This branch will in particular speed up deposit processing which has been slowing down block replay. Pre (mainnet, 1600 blocks): ``` All time are ms Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 3450.269, 0.000, 3450.269, 3450.269, 1, Initialize DB 0.417, 0.822, 0.036, 21.098, 1400, Load block from database 16.521, 0.000, 16.521, 16.521, 1, Load state from database 27.906, 50.846, 8.104, 1507.633, 1350, Apply block 52.617, 37.029, 20.640, 135.938, 50, Apply epoch block ``` Post: ``` 3502.715, 0.000, 3502.715, 3502.715, 1, Initialize DB 0.080, 0.560, 0.035, 21.015, 1400, Load block from database 17.595, 0.000, 17.595, 17.595, 1, Load state from database 15.706, 11.028, 8.300, 107.537, 1350, Apply block 33.217, 12.622, 17.331, 60.580, 50, Apply epoch block ``` * more perf fixes * load EpochRef cache into StateCache more aggressively * point out security concern with public key cache * reuse proposer index from state when processing block * avoid genericAssign in a few more places * don't parse key when signature is unparseable * fix `==` overload for Eth2Digest * preallocate validator list when getting active validators * speed up proposer index calculation a little bit * reuse cache when replaying blocks in ncli_db * avoid a few more copying loops ``` Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 3279.158, 0.000, 3279.158, 3279.158, 1, Initialize DB 0.072, 0.357, 0.035, 13.400, 1400, Load block from database 17.295, 0.000, 17.295, 17.295, 1, Load state from database 5.918, 9.896, 0.198, 98.028, 1350, Apply block 15.888, 10.951, 7.902, 39.535, 50, Apply epoch block 0.000, 0.000, 0.000, 0.000, 0, Database block store ``` * clear full balance cache before processing rewards and penalties ``` All time are ms Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 3947.901, 0.000, 3947.901, 3947.901, 1, Initialize DB 0.124, 0.506, 0.026, 202.370, 363345, Load block from database 97.614, 0.000, 97.614, 97.614, 1, Load state from database 0.186, 0.188, 0.012, 99.561, 357262, Advance slot, non-epoch 14.161, 5.966, 1.099, 395.511, 11524, Advance slot, epoch 1.372, 4.170, 0.017, 276.401, 363345, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database block store ```
2021-01-25 12:04:18 +00:00
desc: "Process each block with a fresh cache".}: bool
of DbCmd.dumpState:
stateRoot* {.
argument
name: "state-root"
desc: "State root(s) to save".}: seq[string]
of DbCmd.putState:
stateFile {.
argument
name: "file"
desc: "Files to import".}: seq[string]
of DbCmd.dumpBlock:
blockRootx* {.
argument
name: "block-root"
desc: "Block root(s) to save".}: seq[string]
of DbCmd.putBlock:
blckFile {.
argument
name: "file"
desc: "Files to import".}: seq[string]
setHead {.
defaultValue: false
name: "set-head"
desc: "Update head to this block"}: bool
setTail {.
defaultValue: false
name: "set-tail"
desc: "Update tail to this block"}: bool
setGenesis {.
defaultValue: false
name: "set-genesis"
desc: "Update genesis to this block"}: bool
of DbCmd.putBlob:
blobFile {.
argument
name: "file"
desc: "Files to import".}: seq[string]
of DbCmd.rewindState:
blockRoot* {.
argument
name: "block-root"
desc: "Block root".}: string
slot* {.
argument
desc: "Slot".}: uint64
of DbCmd.verifyEra:
eraFile* {.
desc: "Era file name".}: string
of DbCmd.exportEra:
e2store: add era format (#2382) Era files contain 8192 blocks and a state corresponding to the length of the array holding block roots in the state, meaning that each block is verifiable using the pubkeys and block roots from the state. Of course, one would need to know the root of the state as well, which is available in the first block of the _next_ file - or known from outside. This PR also adds an implementation to write e2s, e2i and era files, as well as a python script to inspect them. All in all, the format is very similar to what goes on in the network requests meaning it can trivially serve as a backing format for serving said requests. Mainnet, up to the first 671k slots, take up 3.5gb - in each era file, the BeaconState contributes about 9mb at current validator set sizes, up from ~3mb in the early blocks, for a grand total of ~558mb for the 82 eras tested - this overhead could potentially be calculated but one would lose the ability to verify individual blocks (eras could still be verified using historical roots). ``` -rw-rw-r--. 1 arnetheduck arnetheduck 16 5 mar 11.47 ethereum2-mainnet-00000000-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 1,8M 5 mar 11.47 ethereum2-mainnet-00000000-00000001.e2s -rw-rw-r--. 1 arnetheduck arnetheduck 65K 5 mar 11.47 ethereum2-mainnet-00000001-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 18M 5 mar 11.47 ethereum2-mainnet-00000001-00000001.e2s ... -rw-rw-r--. 1 arnetheduck arnetheduck 65K 5 mar 11.52 ethereum2-mainnet-00000051-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 68M 5 mar 11.52 ethereum2-mainnet-00000051-00000001.e2s -rw-rw-r--. 1 arnetheduck arnetheduck 61K 5 mar 11.11 ethereum2-mainnet-00000052-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 62M 5 mar 11.11 ethereum2-mainnet-00000052-00000001.e2s ```
2021-03-15 10:31:39 +00:00
era* {.
defaultValue: 0
desc: "The era number to write".}: uint64
eraCount* {.
defaultValue: 0
name: "count"
desc: "Number of eras to write (0=all)".}: uint64
e2store: add era format (#2382) Era files contain 8192 blocks and a state corresponding to the length of the array holding block roots in the state, meaning that each block is verifiable using the pubkeys and block roots from the state. Of course, one would need to know the root of the state as well, which is available in the first block of the _next_ file - or known from outside. This PR also adds an implementation to write e2s, e2i and era files, as well as a python script to inspect them. All in all, the format is very similar to what goes on in the network requests meaning it can trivially serve as a backing format for serving said requests. Mainnet, up to the first 671k slots, take up 3.5gb - in each era file, the BeaconState contributes about 9mb at current validator set sizes, up from ~3mb in the early blocks, for a grand total of ~558mb for the 82 eras tested - this overhead could potentially be calculated but one would lose the ability to verify individual blocks (eras could still be verified using historical roots). ``` -rw-rw-r--. 1 arnetheduck arnetheduck 16 5 mar 11.47 ethereum2-mainnet-00000000-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 1,8M 5 mar 11.47 ethereum2-mainnet-00000000-00000001.e2s -rw-rw-r--. 1 arnetheduck arnetheduck 65K 5 mar 11.47 ethereum2-mainnet-00000001-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 18M 5 mar 11.47 ethereum2-mainnet-00000001-00000001.e2s ... -rw-rw-r--. 1 arnetheduck arnetheduck 65K 5 mar 11.52 ethereum2-mainnet-00000051-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 68M 5 mar 11.52 ethereum2-mainnet-00000051-00000001.e2s -rw-rw-r--. 1 arnetheduck arnetheduck 61K 5 mar 11.11 ethereum2-mainnet-00000052-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 62M 5 mar 11.11 ethereum2-mainnet-00000052-00000001.e2s ```
2021-03-15 10:31:39 +00:00
of DbCmd.importEra:
eraFiles* {.
argument
name: "file"
desc: "The name of the era file(s) to import".}: seq[string]
of DbCmd.validatorPerf:
perfSlot* {.
defaultValue: -128 * SLOTS_PER_EPOCH.int64
name: "start-slot"
desc: "Starting slot, negative = backwards from head".}: int64
perfSlots* {.
defaultValue: 0
name: "slots"
desc: "Number of slots to run benchmark for, 0 = all the way to head".}: uint64
of DbCmd.validatorDb:
outDir* {.
name: "out-dir"
abbr: "o"
desc: "Output directory".}: string
startEpoch* {.
name: "start-epoch"
abbr: "s"
desc: "Epoch from which to start recording statistics." &
"By default one past the last epoch in the output directory".}: Option[uint]
endEpoch* {.
name: "end-epoch"
abbr: "e"
desc: "The last for which to record statistics." &
"By default the last epoch in the input database".}: Option[uint]
resolution {.
defaultValue: 225,
name: "resolution"
abbr: "r"
desc: "How many epochs to be aggregated in a single compacted file" .}: uint
writeAggregatedFiles {.
name: "aggregated"
defaultValue: true
abbr: "a"
desc: "Whether to write aggregated files for a range of epochs with a given resolution" .}: bool
writeUnaggregatedFiles {.
name: "unaggregated"
defaultValue: true
abbr: "u"
desc: "Whether to write unaggregated file for each epoch" .}: bool
var shouldShutDown = false
func getSlotRange(dag: ChainDAGRef, startSlot: int64, count: uint64): (Slot, Slot) =
let
start =
if startSlot >= 0: Slot(startSlot)
elif uint64(-startSlot) >= dag.head.slot: Slot(0)
2022-02-03 19:03:01 +00:00
else: dag.head.slot - uint64(-startSlot)
ends =
if count == 0: dag.head.slot + 1
else: start + count
(start, ends)
proc cmdBench(conf: DbConf, cfg: RuntimeConfig) =
var timers: array[Timers, RunningStat]
echo "Opening database..."
let
db = BeaconChainDB.new(conf.databaseDir.string, readOnly = true)
dbBenchmark = BeaconChainDB.new("benchmark")
defer:
db.close()
dbBenchmark.close()
if (let v = ChainDAGRef.isInitialized(db); v.isErr()):
echo "Database not initialized: ", v.error()
quit 1
echo "Initializing block pool..."
let
validatorMonitor = newClone(ValidatorMonitor.init())
dag = withTimerRet(timers[tInit]):
ChainDAGRef.init(cfg, db, validatorMonitor, {}, conf.eraDir)
var
(start, ends) = dag.getSlotRange(conf.benchSlot, conf.benchSlots)
blockRefs = dag.getBlockRange(max(start, Slot 1), ends)
blocks: (
seq[phase0.TrustedSignedBeaconBlock],
seq[altair.TrustedSignedBeaconBlock],
seq[bellatrix.TrustedSignedBeaconBlock],
seq[capella.TrustedSignedBeaconBlock],
seq[deneb.TrustedSignedBeaconBlock],
seq[electra.TrustedSignedBeaconBlock])
echo "Loaded head slot ", dag.head.slot,
" selected ", blockRefs.len, " blocks"
doAssert blockRefs.len() > 0, "Must select at least one block"
for b in 0 ..< blockRefs.len:
let blck = blockRefs[blockRefs.len - b - 1]
withTimer(timers[tLoadBlock]):
case cfg.consensusForkAtEpoch(blck.slot.epoch)
of ConsensusFork.Phase0:
blocks[0].add dag.db.getBlock(
blck.root, phase0.TrustedSignedBeaconBlock).get()
of ConsensusFork.Altair:
blocks[1].add dag.db.getBlock(
blck.root, altair.TrustedSignedBeaconBlock).get()
of ConsensusFork.Bellatrix:
blocks[2].add dag.db.getBlock(
blck.root, bellatrix.TrustedSignedBeaconBlock).get()
of ConsensusFork.Capella:
blocks[3].add dag.db.getBlock(
blck.root, capella.TrustedSignedBeaconBlock).get()
of ConsensusFork.Deneb:
blocks[4].add dag.db.getBlock(
blck.root, deneb.TrustedSignedBeaconBlock).get()
of ConsensusFork.Electra:
blocks[5].add dag.db.getBlock(
blck.root, electra.TrustedSignedBeaconBlock).get()
let stateData = newClone(dag.headState)
var
cache = StateCache()
info = ForkedEpochInfo()
loadedState = (
(ref phase0.HashedBeaconState)(),
(ref altair.HashedBeaconState)(),
(ref bellatrix.HashedBeaconState)(),
2023-03-28 13:44:38 +00:00
(ref capella.HashedBeaconState)(),
(ref deneb.HashedBeaconState)(),
(ref electra.HashedBeaconState)())
performance fixes (#2259) * performance fixes * don't mark tree cache as dirty on read-only List accesses * store only blob in memory for keys and signatures, parse blob lazily * compare public keys by blob instead of parsing / converting to raw * compare Eth2Digest using non-constant-time comparison * avoid some unnecessary validator copying This branch will in particular speed up deposit processing which has been slowing down block replay. Pre (mainnet, 1600 blocks): ``` All time are ms Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 3450.269, 0.000, 3450.269, 3450.269, 1, Initialize DB 0.417, 0.822, 0.036, 21.098, 1400, Load block from database 16.521, 0.000, 16.521, 16.521, 1, Load state from database 27.906, 50.846, 8.104, 1507.633, 1350, Apply block 52.617, 37.029, 20.640, 135.938, 50, Apply epoch block ``` Post: ``` 3502.715, 0.000, 3502.715, 3502.715, 1, Initialize DB 0.080, 0.560, 0.035, 21.015, 1400, Load block from database 17.595, 0.000, 17.595, 17.595, 1, Load state from database 15.706, 11.028, 8.300, 107.537, 1350, Apply block 33.217, 12.622, 17.331, 60.580, 50, Apply epoch block ``` * more perf fixes * load EpochRef cache into StateCache more aggressively * point out security concern with public key cache * reuse proposer index from state when processing block * avoid genericAssign in a few more places * don't parse key when signature is unparseable * fix `==` overload for Eth2Digest * preallocate validator list when getting active validators * speed up proposer index calculation a little bit * reuse cache when replaying blocks in ncli_db * avoid a few more copying loops ``` Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 3279.158, 0.000, 3279.158, 3279.158, 1, Initialize DB 0.072, 0.357, 0.035, 13.400, 1400, Load block from database 17.295, 0.000, 17.295, 17.295, 1, Load state from database 5.918, 9.896, 0.198, 98.028, 1350, Apply block 15.888, 10.951, 7.902, 39.535, 50, Apply epoch block 0.000, 0.000, 0.000, 0.000, 0, Database block store ``` * clear full balance cache before processing rewards and penalties ``` All time are ms Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 3947.901, 0.000, 3947.901, 3947.901, 1, Initialize DB 0.124, 0.506, 0.026, 202.370, 363345, Load block from database 97.614, 0.000, 97.614, 97.614, 1, Load state from database 0.186, 0.188, 0.012, 99.561, 357262, Advance slot, non-epoch 14.161, 5.966, 1.099, 395.511, 11524, Advance slot, epoch 1.372, 4.170, 0.017, 276.401, 363345, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database block store ```
2021-01-25 12:04:18 +00:00
withTimer(timers[tLoadState]):
doAssert dag.updateState(
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
stateData[],
dag.atSlot(blockRefs[^1], blockRefs[^1].slot - 1).expect("not nil"),
false, cache)
template processBlocks(blocks: auto) =
for b in blocks.mitems():
if shouldShutDown: quit QuitSuccess
while getStateField(stateData[], slot) < b.message.slot:
let isEpoch = (getStateField(stateData[], slot) + 1).is_epoch()
withTimer(timers[if isEpoch: tAdvanceEpoch else: tAdvanceSlot]):
process_slots(
dag.cfg, stateData[], getStateField(stateData[], slot) + 1, cache,
info, {}).expect("Slot processing can't fail with correct inputs")
var start = Moment.now()
withTimer(timers[tApplyBlock]):
if conf.resetCache:
cache = StateCache()
let res = state_transition_block(
dag.cfg, stateData[], b, cache, {}, noRollback)
if res.isErr():
dump("./", b)
echo "State transition failed (!) ", res.error()
quit 1
if conf.printTimes:
echo b.message.slot, ",", toHex(b.root.data), ",", nanoseconds(Moment.now() - start)
if conf.storeBlocks:
immutable validator database factoring (#2297) * initial immutable validator database factoring * remove changes from chain_dag: this abstraction properly belongs in beacon_chain_db * add merging mutable/immutable validator portions; individually test database roundtripping of immutable validators and states-sans-immutable-validators * update test summaries * use stew/assign2 instead of Nim assignment * add reading/writing of immutable validators in chaindag * remove unused import * replace chunked k/v store of immutable validators with per-row SQL table storage * use List instead of HashList * un-stub some ncli_db code so that it uses * switch HashArray to array; move BeaconStateNoImmutableValidators from datatypes to beacon_chain_db * begin only-mutable-part state storage * uncomment some assigns * work around https://github.com/nim-lang/Nim/issues/17253 * fix most of the issues/oversights; local sim runs again * fix test suite by adding missing beaconstate field to copy function * have ncli bench also store immutable validators * extract some immutable-validator-specific code from the beacon chain db module * add more rigorous database state roundtripping, with changing validator sets * adjust ncli_db to use new schema * simplify putState/getState by moving all immutable validator accounting into beacon state DB * remove redundant test case and move code to immutable-beacon-chain module * more efficient, but still brute-force, mutable+immutable validator merging * reuse BeaconState in getState * ensure HashList/HashArray caches are cleared when reusing getState buffers; add ncli_db and a unit test to verify this * HashList.clear() -> HashList.clearCache() * only copy incrementally necessary immutable validators * increase strictness of test cases and fix/work around resulting HashList cache invalidation issues * remove explanatory scaffolding * allow for storage of full (with all validators) states for backwards/forwards-compatibility * adjust DbSeq type usage * store full, with-validators, state every 64 epochs to enable reverting versions * reduce memory allocation and intermediate objects in state storage codepath * eliminate allocation/copying through intermediate BeaconStateNoImmutableValidators objects * skip benchmarking initial genesis-validator-heavy state store * always store new-style state and sometimes old-style state * document intent behind BeaconState/Validator type-punnery * more accurate failure message on SQLite in-memory database initialization failure
2021-03-15 14:11:51 +00:00
withTimer(timers[tDbStore]):
dbBenchmark.putBlock(b)
withState(stateData[]):
if forkyState.data.slot.is_epoch and conf.storeStates:
if forkyState.data.slot.epoch < 2:
dbBenchmark.putState(forkyState.root, forkyState.data)
dbBenchmark.checkpoint()
else:
withTimer(timers[tDbStore]):
dbBenchmark.putState(forkyState.root, forkyState.data)
dbBenchmark.checkpoint()
withTimer(timers[tDbLoad]):
case consensusFork
of ConsensusFork.Phase0:
doAssert dbBenchmark.getState(
forkyState.root, loadedState[0][].data, noRollback)
of ConsensusFork.Altair:
doAssert dbBenchmark.getState(
forkyState.root, loadedState[1][].data, noRollback)
of ConsensusFork.Bellatrix:
doAssert dbBenchmark.getState(
forkyState.root, loadedState[2][].data, noRollback)
of ConsensusFork.Capella:
doAssert dbBenchmark.getState(
forkyState.root, loadedState[3][].data, noRollback)
of ConsensusFork.Deneb:
2023-03-28 13:44:38 +00:00
doAssert dbBenchmark.getState(
forkyState.root, loadedState[4][].data, noRollback)
of ConsensusFork.Electra:
doAssert dbBenchmark.getState(
forkyState.root, loadedState[5][].data, noRollback)
if forkyState.data.slot.epoch mod 16 == 0:
let loadedRoot = case consensusFork
of ConsensusFork.Phase0: hash_tree_root(loadedState[0][].data)
of ConsensusFork.Altair: hash_tree_root(loadedState[1][].data)
of ConsensusFork.Bellatrix: hash_tree_root(loadedState[2][].data)
of ConsensusFork.Capella: hash_tree_root(loadedState[3][].data)
2023-03-28 13:44:38 +00:00
of ConsensusFork.Deneb: hash_tree_root(loadedState[4][].data)
of ConsensusFork.Electra: hash_tree_root(loadedState[5][].data)
doAssert hash_tree_root(forkyState.data) == loadedRoot
processBlocks(blocks[0])
processBlocks(blocks[1])
processBlocks(blocks[2])
processBlocks(blocks[3])
2023-08-23 16:37:25 +00:00
processBlocks(blocks[4])
processBlocks(blocks[5])
printTimers(false, timers)
proc cmdDumpState(conf: DbConf) =
let db = BeaconChainDB.new(conf.databaseDir.string, readOnly = true)
defer: db.close()
let
phase0State = (ref phase0.HashedBeaconState)()
altairState = (ref altair.HashedBeaconState)()
bellatrixState = (ref bellatrix.HashedBeaconState)()
capellaState = (ref capella.HashedBeaconState)()
2023-08-23 16:37:25 +00:00
denebState = (ref deneb.HashedBeaconState)()
electraState = (ref electra.HashedBeaconState)()
for stateRoot in conf.stateRoot:
if shouldShutDown: quit QuitSuccess
template doit(state: untyped) =
try:
state.root = Eth2Digest.fromHex(stateRoot)
if db.getState(state.root, state.data, noRollback):
dump("./", state)
continue
except CatchableError as e:
echo "Couldn't load ", state.root, ": ", e.msg
doit(phase0State[])
doit(altairState[])
doit(bellatrixState[])
doit(capellaState[])
2023-08-23 16:37:25 +00:00
doit(denebState[])
doit(electraState[])
echo "Couldn't load ", stateRoot
proc cmdPutState(conf: DbConf, cfg: RuntimeConfig) =
let db = BeaconChainDB.new(conf.databaseDir.string)
defer: db.close()
for file in conf.stateFile:
if shouldShutDown: quit QuitSuccess
let state =
try:
newClone(readSszForkedHashedBeaconState(
cfg, readAllBytes(file).tryGet()))
except ResultError[IoErrorCode] as e:
echo "Couldn't load ", file, ": ", e.msg
continue
except SerializationError as e:
echo "Malformed ", file, ": ", e.msg
continue
withState(state[]):
db.putState(forkyState)
proc cmdDumpBlock(conf: DbConf) =
let db = BeaconChainDB.new(conf.databaseDir.string, readOnly = true)
defer: db.close()
for blockRoot in conf.blockRootx:
if shouldShutDown: quit QuitSuccess
try:
let root = Eth2Digest.fromHex(blockRoot)
if (let blck = db.getBlock(
root, phase0.TrustedSignedBeaconBlock); blck.isSome):
dump("./", blck.get())
elif (let blck = db.getBlock(
root, altair.TrustedSignedBeaconBlock); blck.isSome):
dump("./", blck.get())
elif (let blck = db.getBlock(root, bellatrix.TrustedSignedBeaconBlock); blck.isSome):
dump("./", blck.get())
elif (let blck = db.getBlock(root, capella.TrustedSignedBeaconBlock); blck.isSome):
dump("./", blck.get())
2023-08-23 16:37:25 +00:00
elif (let blck = db.getBlock(root, deneb.TrustedSignedBeaconBlock); blck.isSome):
dump("./", blck.get())
else:
echo "Couldn't load ", blockRoot
except CatchableError as e:
echo "Couldn't load ", blockRoot, ": ", e.msg
proc cmdPutBlock(conf: DbConf, cfg: RuntimeConfig) =
let db = BeaconChainDB.new(conf.databaseDir.string)
defer: db.close()
for file in conf.blckFile:
if shouldShutDown: quit QuitSuccess
let blck =
try:
readSszForkedSignedBeaconBlock(
cfg, readAllBytes(file).tryGet())
except ResultError[IoErrorCode] as e:
echo "Couldn't load ", file, ": ", e.msg
continue
except SerializationError as e:
echo "Malformed ", file, ": ", e.msg
continue
withBlck(blck.asTrusted()):
db.putBlock(forkyBlck)
if conf.setHead:
db.putHeadBlock(forkyBlck.root)
if conf.setTail:
db.putTailBlock(forkyBlck.root)
if conf.setGenesis:
db.putGenesisBlock(forkyBlck.root)
proc cmdPutBlob(conf: DbConf, cfg: RuntimeConfig) =
let db = BeaconChainDB.new(conf.databaseDir.string, cfg)
defer: db.close()
for file in conf.blobFile:
if shouldShutDown: quit QuitSuccess
let
blob =
try:
SSZ.decode(readAllBytes(file).tryGet(), BlobSidecar)
except ResultError[IoErrorCode] as e:
echo "Couldn't load ", file, ": ", e.msg
continue
except SerializationError as e:
echo "Malformed ", file, ": ", e.msg
continue
res = blob.verify_blob_sidecar_inclusion_proof()
if res.isErr:
echo "Invalid ", file, ": ", res.error
continue
db.putBlobSidecar(blob)
proc cmdRewindState(conf: DbConf, cfg: RuntimeConfig) =
echo "Opening database..."
let db = BeaconChainDB.new(conf.databaseDir.string, readOnly = true)
defer: db.close()
if (let v = ChainDAGRef.isInitialized(db); v.isErr()):
echo "Database not initialized: ", v.error()
quit 1
echo "Initializing block pool..."
let
validatorMonitor = newClone(ValidatorMonitor.init())
dag = ChainDAGRef.init(cfg, db, validatorMonitor, {}, conf.eraDir)
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
let bid = dag.getBlockId(fromHex(Eth2Digest, conf.blockRoot)).valueOr:
echo "Block not found in database"
return
let tmpState = assignClone(dag.headState)
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
dag.withUpdatedState(
tmpState[], dag.atSlot(bid, Slot(conf.slot)).expect("block found")) do:
echo "Writing state..."
withState(updatedState):
dump("./", forkyState)
do: raiseAssert "withUpdatedState failed"
proc cmdVerifyEra(conf: DbConf, cfg: RuntimeConfig) =
let
f = EraFile.open(conf.eraFile).valueOr:
echo error
quit 1
root = f.verify(cfg).valueOr:
echo error
quit 1
echo root
proc cmdExportEra(conf: DbConf, cfg: RuntimeConfig) =
let db = BeaconChainDB.new(conf.databaseDir.string, readOnly = true)
e2store: add era format (#2382) Era files contain 8192 blocks and a state corresponding to the length of the array holding block roots in the state, meaning that each block is verifiable using the pubkeys and block roots from the state. Of course, one would need to know the root of the state as well, which is available in the first block of the _next_ file - or known from outside. This PR also adds an implementation to write e2s, e2i and era files, as well as a python script to inspect them. All in all, the format is very similar to what goes on in the network requests meaning it can trivially serve as a backing format for serving said requests. Mainnet, up to the first 671k slots, take up 3.5gb - in each era file, the BeaconState contributes about 9mb at current validator set sizes, up from ~3mb in the early blocks, for a grand total of ~558mb for the 82 eras tested - this overhead could potentially be calculated but one would lose the ability to verify individual blocks (eras could still be verified using historical roots). ``` -rw-rw-r--. 1 arnetheduck arnetheduck 16 5 mar 11.47 ethereum2-mainnet-00000000-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 1,8M 5 mar 11.47 ethereum2-mainnet-00000000-00000001.e2s -rw-rw-r--. 1 arnetheduck arnetheduck 65K 5 mar 11.47 ethereum2-mainnet-00000001-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 18M 5 mar 11.47 ethereum2-mainnet-00000001-00000001.e2s ... -rw-rw-r--. 1 arnetheduck arnetheduck 65K 5 mar 11.52 ethereum2-mainnet-00000051-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 68M 5 mar 11.52 ethereum2-mainnet-00000051-00000001.e2s -rw-rw-r--. 1 arnetheduck arnetheduck 61K 5 mar 11.11 ethereum2-mainnet-00000052-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 62M 5 mar 11.11 ethereum2-mainnet-00000052-00000001.e2s ```
2021-03-15 10:31:39 +00:00
defer: db.close()
if (let v = ChainDAGRef.isInitialized(db); v.isErr()):
fatal "Database not initialized", error = v.error()
e2store: add era format (#2382) Era files contain 8192 blocks and a state corresponding to the length of the array holding block roots in the state, meaning that each block is verifiable using the pubkeys and block roots from the state. Of course, one would need to know the root of the state as well, which is available in the first block of the _next_ file - or known from outside. This PR also adds an implementation to write e2s, e2i and era files, as well as a python script to inspect them. All in all, the format is very similar to what goes on in the network requests meaning it can trivially serve as a backing format for serving said requests. Mainnet, up to the first 671k slots, take up 3.5gb - in each era file, the BeaconState contributes about 9mb at current validator set sizes, up from ~3mb in the early blocks, for a grand total of ~558mb for the 82 eras tested - this overhead could potentially be calculated but one would lose the ability to verify individual blocks (eras could still be verified using historical roots). ``` -rw-rw-r--. 1 arnetheduck arnetheduck 16 5 mar 11.47 ethereum2-mainnet-00000000-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 1,8M 5 mar 11.47 ethereum2-mainnet-00000000-00000001.e2s -rw-rw-r--. 1 arnetheduck arnetheduck 65K 5 mar 11.47 ethereum2-mainnet-00000001-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 18M 5 mar 11.47 ethereum2-mainnet-00000001-00000001.e2s ... -rw-rw-r--. 1 arnetheduck arnetheduck 65K 5 mar 11.52 ethereum2-mainnet-00000051-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 68M 5 mar 11.52 ethereum2-mainnet-00000051-00000001.e2s -rw-rw-r--. 1 arnetheduck arnetheduck 61K 5 mar 11.11 ethereum2-mainnet-00000052-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 62M 5 mar 11.11 ethereum2-mainnet-00000052-00000001.e2s ```
2021-03-15 10:31:39 +00:00
quit 1
type Timers = enum
tState
tBlocks
e2store: add era format (#2382) Era files contain 8192 blocks and a state corresponding to the length of the array holding block roots in the state, meaning that each block is verifiable using the pubkeys and block roots from the state. Of course, one would need to know the root of the state as well, which is available in the first block of the _next_ file - or known from outside. This PR also adds an implementation to write e2s, e2i and era files, as well as a python script to inspect them. All in all, the format is very similar to what goes on in the network requests meaning it can trivially serve as a backing format for serving said requests. Mainnet, up to the first 671k slots, take up 3.5gb - in each era file, the BeaconState contributes about 9mb at current validator set sizes, up from ~3mb in the early blocks, for a grand total of ~558mb for the 82 eras tested - this overhead could potentially be calculated but one would lose the ability to verify individual blocks (eras could still be verified using historical roots). ``` -rw-rw-r--. 1 arnetheduck arnetheduck 16 5 mar 11.47 ethereum2-mainnet-00000000-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 1,8M 5 mar 11.47 ethereum2-mainnet-00000000-00000001.e2s -rw-rw-r--. 1 arnetheduck arnetheduck 65K 5 mar 11.47 ethereum2-mainnet-00000001-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 18M 5 mar 11.47 ethereum2-mainnet-00000001-00000001.e2s ... -rw-rw-r--. 1 arnetheduck arnetheduck 65K 5 mar 11.52 ethereum2-mainnet-00000051-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 68M 5 mar 11.52 ethereum2-mainnet-00000051-00000001.e2s -rw-rw-r--. 1 arnetheduck arnetheduck 61K 5 mar 11.11 ethereum2-mainnet-00000052-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 62M 5 mar 11.11 ethereum2-mainnet-00000052-00000001.e2s ```
2021-03-15 10:31:39 +00:00
let
validatorMonitor = newClone(ValidatorMonitor.init())
dag = ChainDAGRef.init(cfg, db, validatorMonitor, {}, conf.eraDir)
e2store: add era format (#2382) Era files contain 8192 blocks and a state corresponding to the length of the array holding block roots in the state, meaning that each block is verifiable using the pubkeys and block roots from the state. Of course, one would need to know the root of the state as well, which is available in the first block of the _next_ file - or known from outside. This PR also adds an implementation to write e2s, e2i and era files, as well as a python script to inspect them. All in all, the format is very similar to what goes on in the network requests meaning it can trivially serve as a backing format for serving said requests. Mainnet, up to the first 671k slots, take up 3.5gb - in each era file, the BeaconState contributes about 9mb at current validator set sizes, up from ~3mb in the early blocks, for a grand total of ~558mb for the 82 eras tested - this overhead could potentially be calculated but one would lose the ability to verify individual blocks (eras could still be verified using historical roots). ``` -rw-rw-r--. 1 arnetheduck arnetheduck 16 5 mar 11.47 ethereum2-mainnet-00000000-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 1,8M 5 mar 11.47 ethereum2-mainnet-00000000-00000001.e2s -rw-rw-r--. 1 arnetheduck arnetheduck 65K 5 mar 11.47 ethereum2-mainnet-00000001-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 18M 5 mar 11.47 ethereum2-mainnet-00000001-00000001.e2s ... -rw-rw-r--. 1 arnetheduck arnetheduck 65K 5 mar 11.52 ethereum2-mainnet-00000051-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 68M 5 mar 11.52 ethereum2-mainnet-00000051-00000001.e2s -rw-rw-r--. 1 arnetheduck arnetheduck 61K 5 mar 11.11 ethereum2-mainnet-00000052-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 62M 5 mar 11.11 ethereum2-mainnet-00000052-00000001.e2s ```
2021-03-15 10:31:39 +00:00
let tmpState = assignClone(dag.headState)
var
tmp: seq[byte]
timers: array[Timers, RunningStat]
var
era = Era(conf.era)
missingHistory = false
era: load blocks and states (#3394) * era: load blocks and states Era files contain finalized history and can be thought of as an alternative source for block and state data that allows clients to avoid syncing this information from the P2P network - the P2P network is then used to "top up" the client with the most recent data. They can be freely shared in the community via whatever means (http, torrent, etc) and serve as a permanent cold store of consensus data (and, after the merge, execution data) for history buffs and bean counters alike. This PR gently introduces support for loading blocks and states in two cases: block requests from rest/p2p and frontfilling when doing checkpoint sync. The era files are used as a secondary source if the information is not found in the database - compared to the database, there are a few key differences: * the database stores the block indexed by block root while the era file indexes by slot - the former is used only in rest, while the latter is used both by p2p and rest. * when loading blocks from era files, the root is no longer trivially available - if it is needed, it must either be computed (slow) or cached (messy) - the good news is that for p2p requests, it is not needed * in era files, "framed" snappy encoding is used while in the database we store unframed snappy - for p2p2 requests, the latter requires recompression while the former could avoid it * front-filling is the process of using era files to replace backfilling - in theory this front-filling could happen from any block and front-fills with gaps could also be entertained, but our backfilling algorithm cannot take advantage of this because there's no (simple) way to tell it to "skip" a range. * front-filling, as implemented, is a bit slow (10s to load mainnet): we load the full BeaconState for every era to grab the roots of the blocks - it would be better to partially load the state - as such, it would also be good to be able to partially decompress snappy blobs * lookups from REST via root are served by first looking up a block summary in the database, then using the slot to load the block data from the era file - however, there needs to be an option to create the summary table from era files to fully support historical queries To test this, `ncli_db` has an era file exporter: the files it creates should be placed in an `era` folder next to `db` in the data directory. What's interesting in particular about this setup is that `db` remains as the source of truth for security purposes - it stores the latest synced head root which in turn determines where a node "starts" its consensus participation - the era directory however can be freely shared between nodes / people without any (significant) security implications, assuming the era files are consistent / not broken. There's lots of future improvements to be had: * we can drop the in-memory `BlockRef` index almost entirely - at this point, resident memory usage of Nimbus should drop to a cool 500-600 mb * we could serve era files via REST trivially: this would drop backfill times to whatever time it takes to download the files - unlike the current implementation that downloads block by block, downloading an era at a time almost entirely cuts out request overhead * we can "reasonably" recreate detailed state history from almost any point in time, turning an O(slot) process into O(1) effectively - we'll still need caches and indices to do this with sufficient efficiency for the rest api, but at least it cuts the whole process down to minutes instead of hours, for arbitrary points in time * CI: ignore failures with Nim-1.6 (temporary) * test fixes Co-authored-by: Ștefan Talpalaru <stefantalpalaru@yahoo.com>
2022-03-23 08:58:17 +00:00
while conf.eraCount == 0 or era < Era(conf.era) + conf.eraCount:
defer: era += 1
if shouldShutDown:
break
era: load blocks and states (#3394) * era: load blocks and states Era files contain finalized history and can be thought of as an alternative source for block and state data that allows clients to avoid syncing this information from the P2P network - the P2P network is then used to "top up" the client with the most recent data. They can be freely shared in the community via whatever means (http, torrent, etc) and serve as a permanent cold store of consensus data (and, after the merge, execution data) for history buffs and bean counters alike. This PR gently introduces support for loading blocks and states in two cases: block requests from rest/p2p and frontfilling when doing checkpoint sync. The era files are used as a secondary source if the information is not found in the database - compared to the database, there are a few key differences: * the database stores the block indexed by block root while the era file indexes by slot - the former is used only in rest, while the latter is used both by p2p and rest. * when loading blocks from era files, the root is no longer trivially available - if it is needed, it must either be computed (slow) or cached (messy) - the good news is that for p2p requests, it is not needed * in era files, "framed" snappy encoding is used while in the database we store unframed snappy - for p2p2 requests, the latter requires recompression while the former could avoid it * front-filling is the process of using era files to replace backfilling - in theory this front-filling could happen from any block and front-fills with gaps could also be entertained, but our backfilling algorithm cannot take advantage of this because there's no (simple) way to tell it to "skip" a range. * front-filling, as implemented, is a bit slow (10s to load mainnet): we load the full BeaconState for every era to grab the roots of the blocks - it would be better to partially load the state - as such, it would also be good to be able to partially decompress snappy blobs * lookups from REST via root are served by first looking up a block summary in the database, then using the slot to load the block data from the era file - however, there needs to be an option to create the summary table from era files to fully support historical queries To test this, `ncli_db` has an era file exporter: the files it creates should be placed in an `era` folder next to `db` in the data directory. What's interesting in particular about this setup is that `db` remains as the source of truth for security purposes - it stores the latest synced head root which in turn determines where a node "starts" its consensus participation - the era directory however can be freely shared between nodes / people without any (significant) security implications, assuming the era files are consistent / not broken. There's lots of future improvements to be had: * we can drop the in-memory `BlockRef` index almost entirely - at this point, resident memory usage of Nimbus should drop to a cool 500-600 mb * we could serve era files via REST trivially: this would drop backfill times to whatever time it takes to download the files - unlike the current implementation that downloads block by block, downloading an era at a time almost entirely cuts out request overhead * we can "reasonably" recreate detailed state history from almost any point in time, turning an O(slot) process into O(1) effectively - we'll still need caches and indices to do this with sufficient efficiency for the rest api, but at least it cuts the whole process down to minutes instead of hours, for arbitrary points in time * CI: ignore failures with Nim-1.6 (temporary) * test fixes Co-authored-by: Ștefan Talpalaru <stefantalpalaru@yahoo.com>
2022-03-23 08:58:17 +00:00
# Era files hold the blocks for the "previous" era, and the first state in
# the era itself
e2store: add era format (#2382) Era files contain 8192 blocks and a state corresponding to the length of the array holding block roots in the state, meaning that each block is verifiable using the pubkeys and block roots from the state. Of course, one would need to know the root of the state as well, which is available in the first block of the _next_ file - or known from outside. This PR also adds an implementation to write e2s, e2i and era files, as well as a python script to inspect them. All in all, the format is very similar to what goes on in the network requests meaning it can trivially serve as a backing format for serving said requests. Mainnet, up to the first 671k slots, take up 3.5gb - in each era file, the BeaconState contributes about 9mb at current validator set sizes, up from ~3mb in the early blocks, for a grand total of ~558mb for the 82 eras tested - this overhead could potentially be calculated but one would lose the ability to verify individual blocks (eras could still be verified using historical roots). ``` -rw-rw-r--. 1 arnetheduck arnetheduck 16 5 mar 11.47 ethereum2-mainnet-00000000-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 1,8M 5 mar 11.47 ethereum2-mainnet-00000000-00000001.e2s -rw-rw-r--. 1 arnetheduck arnetheduck 65K 5 mar 11.47 ethereum2-mainnet-00000001-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 18M 5 mar 11.47 ethereum2-mainnet-00000001-00000001.e2s ... -rw-rw-r--. 1 arnetheduck arnetheduck 65K 5 mar 11.52 ethereum2-mainnet-00000051-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 68M 5 mar 11.52 ethereum2-mainnet-00000051-00000001.e2s -rw-rw-r--. 1 arnetheduck arnetheduck 61K 5 mar 11.11 ethereum2-mainnet-00000052-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 62M 5 mar 11.11 ethereum2-mainnet-00000052-00000001.e2s ```
2021-03-15 10:31:39 +00:00
let
firstSlot =
if era == 0: none(Slot)
era: load blocks and states (#3394) * era: load blocks and states Era files contain finalized history and can be thought of as an alternative source for block and state data that allows clients to avoid syncing this information from the P2P network - the P2P network is then used to "top up" the client with the most recent data. They can be freely shared in the community via whatever means (http, torrent, etc) and serve as a permanent cold store of consensus data (and, after the merge, execution data) for history buffs and bean counters alike. This PR gently introduces support for loading blocks and states in two cases: block requests from rest/p2p and frontfilling when doing checkpoint sync. The era files are used as a secondary source if the information is not found in the database - compared to the database, there are a few key differences: * the database stores the block indexed by block root while the era file indexes by slot - the former is used only in rest, while the latter is used both by p2p and rest. * when loading blocks from era files, the root is no longer trivially available - if it is needed, it must either be computed (slow) or cached (messy) - the good news is that for p2p requests, it is not needed * in era files, "framed" snappy encoding is used while in the database we store unframed snappy - for p2p2 requests, the latter requires recompression while the former could avoid it * front-filling is the process of using era files to replace backfilling - in theory this front-filling could happen from any block and front-fills with gaps could also be entertained, but our backfilling algorithm cannot take advantage of this because there's no (simple) way to tell it to "skip" a range. * front-filling, as implemented, is a bit slow (10s to load mainnet): we load the full BeaconState for every era to grab the roots of the blocks - it would be better to partially load the state - as such, it would also be good to be able to partially decompress snappy blobs * lookups from REST via root are served by first looking up a block summary in the database, then using the slot to load the block data from the era file - however, there needs to be an option to create the summary table from era files to fully support historical queries To test this, `ncli_db` has an era file exporter: the files it creates should be placed in an `era` folder next to `db` in the data directory. What's interesting in particular about this setup is that `db` remains as the source of truth for security purposes - it stores the latest synced head root which in turn determines where a node "starts" its consensus participation - the era directory however can be freely shared between nodes / people without any (significant) security implications, assuming the era files are consistent / not broken. There's lots of future improvements to be had: * we can drop the in-memory `BlockRef` index almost entirely - at this point, resident memory usage of Nimbus should drop to a cool 500-600 mb * we could serve era files via REST trivially: this would drop backfill times to whatever time it takes to download the files - unlike the current implementation that downloads block by block, downloading an era at a time almost entirely cuts out request overhead * we can "reasonably" recreate detailed state history from almost any point in time, turning an O(slot) process into O(1) effectively - we'll still need caches and indices to do this with sufficient efficiency for the rest api, but at least it cuts the whole process down to minutes instead of hours, for arbitrary points in time * CI: ignore failures with Nim-1.6 (temporary) * test fixes Co-authored-by: Ștefan Talpalaru <stefantalpalaru@yahoo.com>
2022-03-23 08:58:17 +00:00
else: some((era - 1).start_slot)
endSlot = era.start_slot
e2store: add era format (#2382) Era files contain 8192 blocks and a state corresponding to the length of the array holding block roots in the state, meaning that each block is verifiable using the pubkeys and block roots from the state. Of course, one would need to know the root of the state as well, which is available in the first block of the _next_ file - or known from outside. This PR also adds an implementation to write e2s, e2i and era files, as well as a python script to inspect them. All in all, the format is very similar to what goes on in the network requests meaning it can trivially serve as a backing format for serving said requests. Mainnet, up to the first 671k slots, take up 3.5gb - in each era file, the BeaconState contributes about 9mb at current validator set sizes, up from ~3mb in the early blocks, for a grand total of ~558mb for the 82 eras tested - this overhead could potentially be calculated but one would lose the ability to verify individual blocks (eras could still be verified using historical roots). ``` -rw-rw-r--. 1 arnetheduck arnetheduck 16 5 mar 11.47 ethereum2-mainnet-00000000-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 1,8M 5 mar 11.47 ethereum2-mainnet-00000000-00000001.e2s -rw-rw-r--. 1 arnetheduck arnetheduck 65K 5 mar 11.47 ethereum2-mainnet-00000001-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 18M 5 mar 11.47 ethereum2-mainnet-00000001-00000001.e2s ... -rw-rw-r--. 1 arnetheduck arnetheduck 65K 5 mar 11.52 ethereum2-mainnet-00000051-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 68M 5 mar 11.52 ethereum2-mainnet-00000051-00000001.e2s -rw-rw-r--. 1 arnetheduck arnetheduck 61K 5 mar 11.11 ethereum2-mainnet-00000052-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 62M 5 mar 11.11 ethereum2-mainnet-00000052-00000001.e2s ```
2021-03-15 10:31:39 +00:00
if endSlot > dag.head.slot:
notice "Written all complete eras", era, endSlot, head = dag.head
e2store: add era format (#2382) Era files contain 8192 blocks and a state corresponding to the length of the array holding block roots in the state, meaning that each block is verifiable using the pubkeys and block roots from the state. Of course, one would need to know the root of the state as well, which is available in the first block of the _next_ file - or known from outside. This PR also adds an implementation to write e2s, e2i and era files, as well as a python script to inspect them. All in all, the format is very similar to what goes on in the network requests meaning it can trivially serve as a backing format for serving said requests. Mainnet, up to the first 671k slots, take up 3.5gb - in each era file, the BeaconState contributes about 9mb at current validator set sizes, up from ~3mb in the early blocks, for a grand total of ~558mb for the 82 eras tested - this overhead could potentially be calculated but one would lose the ability to verify individual blocks (eras could still be verified using historical roots). ``` -rw-rw-r--. 1 arnetheduck arnetheduck 16 5 mar 11.47 ethereum2-mainnet-00000000-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 1,8M 5 mar 11.47 ethereum2-mainnet-00000000-00000001.e2s -rw-rw-r--. 1 arnetheduck arnetheduck 65K 5 mar 11.47 ethereum2-mainnet-00000001-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 18M 5 mar 11.47 ethereum2-mainnet-00000001-00000001.e2s ... -rw-rw-r--. 1 arnetheduck arnetheduck 65K 5 mar 11.52 ethereum2-mainnet-00000051-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 68M 5 mar 11.52 ethereum2-mainnet-00000051-00000001.e2s -rw-rw-r--. 1 arnetheduck arnetheduck 61K 5 mar 11.11 ethereum2-mainnet-00000052-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 62M 5 mar 11.11 ethereum2-mainnet-00000052-00000001.e2s ```
2021-03-15 10:31:39 +00:00
break
let
eraRoot = withState(dag.headState):
eraRoot(
forkyState.data.genesis_validators_root,
forkyState.data.historical_roots.asSeq,
dag.headState.historical_summaries().asSeq,
era).expect("have era root since we checked slot")
name = eraFileName(cfg, era, eraRoot)
era: load blocks and states (#3394) * era: load blocks and states Era files contain finalized history and can be thought of as an alternative source for block and state data that allows clients to avoid syncing this information from the P2P network - the P2P network is then used to "top up" the client with the most recent data. They can be freely shared in the community via whatever means (http, torrent, etc) and serve as a permanent cold store of consensus data (and, after the merge, execution data) for history buffs and bean counters alike. This PR gently introduces support for loading blocks and states in two cases: block requests from rest/p2p and frontfilling when doing checkpoint sync. The era files are used as a secondary source if the information is not found in the database - compared to the database, there are a few key differences: * the database stores the block indexed by block root while the era file indexes by slot - the former is used only in rest, while the latter is used both by p2p and rest. * when loading blocks from era files, the root is no longer trivially available - if it is needed, it must either be computed (slow) or cached (messy) - the good news is that for p2p requests, it is not needed * in era files, "framed" snappy encoding is used while in the database we store unframed snappy - for p2p2 requests, the latter requires recompression while the former could avoid it * front-filling is the process of using era files to replace backfilling - in theory this front-filling could happen from any block and front-fills with gaps could also be entertained, but our backfilling algorithm cannot take advantage of this because there's no (simple) way to tell it to "skip" a range. * front-filling, as implemented, is a bit slow (10s to load mainnet): we load the full BeaconState for every era to grab the roots of the blocks - it would be better to partially load the state - as such, it would also be good to be able to partially decompress snappy blobs * lookups from REST via root are served by first looking up a block summary in the database, then using the slot to load the block data from the era file - however, there needs to be an option to create the summary table from era files to fully support historical queries To test this, `ncli_db` has an era file exporter: the files it creates should be placed in an `era` folder next to `db` in the data directory. What's interesting in particular about this setup is that `db` remains as the source of truth for security purposes - it stores the latest synced head root which in turn determines where a node "starts" its consensus participation - the era directory however can be freely shared between nodes / people without any (significant) security implications, assuming the era files are consistent / not broken. There's lots of future improvements to be had: * we can drop the in-memory `BlockRef` index almost entirely - at this point, resident memory usage of Nimbus should drop to a cool 500-600 mb * we could serve era files via REST trivially: this would drop backfill times to whatever time it takes to download the files - unlike the current implementation that downloads block by block, downloading an era at a time almost entirely cuts out request overhead * we can "reasonably" recreate detailed state history from almost any point in time, turning an O(slot) process into O(1) effectively - we'll still need caches and indices to do this with sufficient efficiency for the rest api, but at least it cuts the whole process down to minutes instead of hours, for arbitrary points in time * CI: ignore failures with Nim-1.6 (temporary) * test fixes Co-authored-by: Ștefan Talpalaru <stefantalpalaru@yahoo.com>
2022-03-23 08:58:17 +00:00
if isFile(name):
debug "Era file already exists", era, name
continue
# Check if we reasonably could write the era file given what's in the
# database - we perform this check after checking for existing era files
# since the database might have been pruned up to the "existing" era files!
if endSlot < dag.tail.slot and era != 0:
notice "Skipping era, state history not available",
era, tail = shortLog(dag.tail)
missingHistory = true
continue
let
eraBid = dag.atSlot(dag.head.bid, endSlot).valueOr:
notice "Skipping era, blocks not available", era, name
missingHistory = true
continue
withTimer(timers[tState]):
var cache: StateCache
if not updateState(dag, tmpState[], eraBid, false, cache):
notice "Skipping era, state history not available", era, name
missingHistory = true
continue
info "Writing ", name
let tmpName = name & ".tmp"
var completed = false
block writeFileBlock:
let e2 = openFile(tmpName, {OpenFlags.Write, OpenFlags.Create, OpenFlags.Truncate}).get()
defer: discard closeFile(e2)
var group = EraGroup.init(e2, firstSlot).get()
if firstSlot.isSome():
withTimer(timers[tBlocks]):
var blocks: array[SLOTS_PER_HISTORICAL_ROOT.int, BlockId]
for i in dag.getBlockRange(firstSlot.get(), 1, blocks)..<blocks.len:
if not dag.getBlockSZ(blocks[i], tmp):
break writeFileBlock
group.update(e2, blocks[i].slot, tmp).get()
withState(tmpState[]):
group.finish(e2, forkyState.data).get()
completed = true
if completed:
try:
moveFile(tmpName, name)
except IOError as e:
warn "Failed to rename era file to its final name (IOError)",
name, tmpName, error = e.msg
except OSError as e:
warn "Failed to rename era file to its final name (OSError)",
name, tmpName, error = e.msg
except Exception as e:
warn "Failed to rename era file to its final name (Exception)",
name, tmpName, error = e.msg
else:
if (let e = io2.removeFile(name); e.isErr):
warn "Failed to clean up incomplete era file", tmpName, error = e.error
if missingHistory:
notice "Some era files were not written due to missing state history - see https://nimbus.guide/trusted-node-sync.html#recreate-historical-state-access-indices for more information"
printTimers(true, timers)
e2store: add era format (#2382) Era files contain 8192 blocks and a state corresponding to the length of the array holding block roots in the state, meaning that each block is verifiable using the pubkeys and block roots from the state. Of course, one would need to know the root of the state as well, which is available in the first block of the _next_ file - or known from outside. This PR also adds an implementation to write e2s, e2i and era files, as well as a python script to inspect them. All in all, the format is very similar to what goes on in the network requests meaning it can trivially serve as a backing format for serving said requests. Mainnet, up to the first 671k slots, take up 3.5gb - in each era file, the BeaconState contributes about 9mb at current validator set sizes, up from ~3mb in the early blocks, for a grand total of ~558mb for the 82 eras tested - this overhead could potentially be calculated but one would lose the ability to verify individual blocks (eras could still be verified using historical roots). ``` -rw-rw-r--. 1 arnetheduck arnetheduck 16 5 mar 11.47 ethereum2-mainnet-00000000-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 1,8M 5 mar 11.47 ethereum2-mainnet-00000000-00000001.e2s -rw-rw-r--. 1 arnetheduck arnetheduck 65K 5 mar 11.47 ethereum2-mainnet-00000001-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 18M 5 mar 11.47 ethereum2-mainnet-00000001-00000001.e2s ... -rw-rw-r--. 1 arnetheduck arnetheduck 65K 5 mar 11.52 ethereum2-mainnet-00000051-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 68M 5 mar 11.52 ethereum2-mainnet-00000051-00000001.e2s -rw-rw-r--. 1 arnetheduck arnetheduck 61K 5 mar 11.11 ethereum2-mainnet-00000052-00000001.e2i -rw-rw-r--. 1 arnetheduck arnetheduck 62M 5 mar 11.11 ethereum2-mainnet-00000052-00000001.e2s ```
2021-03-15 10:31:39 +00:00
proc cmdImportEra(conf: DbConf, cfg: RuntimeConfig) =
let db = BeaconChainDB.new(conf.databaseDir.string)
defer: db.close()
type Timers = enum
tBlock
tState
var
blocks = 0
states = 0
others = 0
timers: array[Timers, RunningStat]
var data: seq[byte]
for file in conf.eraFiles:
if shouldShutDown: quit QuitSuccess
let f = openFile(file, {OpenFlags.Read}).valueOr:
warn "Can't open ", file
continue
defer: discard closeFile(f)
while true:
let header = readRecord(f, data).valueOr:
break
if header.typ == SnappyBeaconBlock:
withTimer(timers[tBlock]):
let uncompressed = decodeFramed(data, checkIntegrity = false)
let blck = try: readSszForkedSignedBeaconBlock(cfg, uncompressed)
except CatchableError as exc:
error "Invalid snappy block", msg = exc.msg, file
continue
withBlck(blck.asTrusted()):
db.putBlock(forkyBlck)
blocks += 1
elif header.typ == SnappyBeaconState:
info "Skipping beacon state (use reindexing to recreate state snapshots)"
states += 1
else:
info "Skipping record", typ = toHex(header.typ)
others += 1
notice "Done", blocks, states, others
printTimers(true, timers)
type
# Validator performance metrics tool based on
# https://github.com/paulhauner/lighthouse/blob/etl/lcli/src/etl/validator_performance.rs
# Credits to Paul Hauner
ValidatorPerformance = object
attestation_hits: uint64
attestation_misses: uint64
head_attestation_hits: uint64
head_attestation_misses: uint64
target_attestation_hits: uint64
target_attestation_misses: uint64
first_slot_head_attester_when_first_slot_empty: uint64
first_slot_head_attester_when_first_slot_not_empty: uint64
delays: Table[uint64, uint64]
proc cmdValidatorPerf(conf: DbConf, cfg: RuntimeConfig) =
echo "Opening database..."
let
db = BeaconChainDB.new(conf.databaseDir.string, readOnly = true)
defer:
db.close()
if (let v = ChainDAGRef.isInitialized(db); v.isErr()):
echo "Database not initialized: ", v.error()
quit 1
echo "# Initializing block pool..."
let
validatorMonitor = newClone(ValidatorMonitor.init())
dag = ChainDAGRef.init(cfg, db, validatorMonitor, {}, conf.eraDir)
var
(start, ends) = dag.getSlotRange(conf.perfSlot, conf.perfSlots)
blockRefs = dag.getBlockRange(start, ends)
perfs = newSeq[ValidatorPerformance](
getStateField(dag.headState, validators).len())
cache = StateCache()
info = ForkedEpochInfo()
blck: phase0.TrustedSignedBeaconBlock
doAssert blockRefs.len() > 0, "Must select at least one block"
echo "# Analyzing performance for epochs ",
blockRefs[^1].slot.epoch, " - ", blockRefs[0].slot.epoch
let state = newClone(dag.headState)
doAssert dag.updateState(
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
state[],
dag.atSlot(blockRefs[^1], blockRefs[^1].slot - 1).expect("block found"),
false, cache)
proc processEpoch() =
let
prev_epoch_target_slot =
state[].get_previous_epoch().start_slot()
penultimate_epoch_end_slot =
if prev_epoch_target_slot == 0: Slot(0)
else: prev_epoch_target_slot - 1
first_slot_empty =
state[].get_block_root_at_slot(prev_epoch_target_slot) ==
state[].get_block_root_at_slot(penultimate_epoch_end_slot)
let first_slot_attesters = block:
let committees_per_slot = state[].get_committee_count_per_slot(
prev_epoch_target_slot.epoch, cache)
var indices = HashSet[ValidatorIndex]()
for committee_index in get_committee_indices(committees_per_slot):
for validator_index in state[].get_beacon_committee(
prev_epoch_target_slot, committee_index, cache):
indices.incl(validator_index)
indices
case info.kind
of EpochInfoFork.Phase0:
template info: untyped = info.phase0Data
for i, s in info.validators:
let perf = addr perfs[i]
if RewardFlags.isActiveInPreviousEpoch in s.flags:
if s.is_previous_epoch_attester.isSome():
perf.attestation_hits += 1;
if RewardFlags.isPreviousEpochHeadAttester in s.flags:
perf.head_attestation_hits += 1
else:
perf.head_attestation_misses += 1
if RewardFlags.isPreviousEpochTargetAttester in s.flags:
perf.target_attestation_hits += 1
else:
perf.target_attestation_misses += 1
if i.ValidatorIndex in first_slot_attesters:
if first_slot_empty:
perf.first_slot_head_attester_when_first_slot_empty += 1
else:
perf.first_slot_head_attester_when_first_slot_not_empty += 1
if s.is_previous_epoch_attester.isSome():
perf.delays.mgetOrPut(
s.is_previous_epoch_attester.get().delay, 0'u64) += 1
else:
perf.attestation_misses += 1;
of EpochInfoFork.Altair:
echo "TODO altair"
if shouldShutDown: quit QuitSuccess
for bi in 0 ..< blockRefs.len:
blck = db.getBlock(
blockRefs[blockRefs.len - bi - 1].root,
phase0.TrustedSignedBeaconBlock).get()
while getStateField(state[], slot) < blck.message.slot:
let
nextSlot = getStateField(state[], slot) + 1
flags =
if nextSlot == blck.message.slot: {skipLastStateRootCalculation}
else: {}
process_slots(
dag.cfg, state[], nextSlot, cache, info, flags).expect(
"Slot processing can't fail with correct inputs")
if getStateField(state[], slot).is_epoch():
processEpoch()
let res = state_transition_block(
dag.cfg, state[], blck, cache, {}, noRollback)
if res.isErr:
echo "State transition failed (!) ", res.error()
quit 1
# Capture rewards of empty slots as well
while getStateField(state[], slot) < ends:
process_slots(
dag.cfg, state[], getStateField(state[], slot) + 1, cache,
info, {}).expect("Slot processing can't fail with correct inputs")
if getStateField(state[], slot).is_epoch():
processEpoch()
echo "validator_index,attestation_hits,attestation_misses,head_attestation_hits,head_attestation_misses,target_attestation_hits,target_attestation_misses,delay_avg,first_slot_head_attester_when_first_slot_empty,first_slot_head_attester_when_first_slot_not_empty"
for i, perf in perfs:
var
count = 0'u64
sum = 0'u64
for delay, n in perf.delays:
count += n
sum += delay * n
echo i,",",
perf.attestation_hits,",",
perf.attestation_misses,",",
perf.head_attestation_hits,",",
perf.head_attestation_misses,",",
perf.target_attestation_hits,",",
perf.target_attestation_misses,",",
if count == 0: 0.0
else: sum.float / count.float,",",
perf.first_slot_head_attester_when_first_slot_empty,",",
perf.first_slot_head_attester_when_first_slot_not_empty
proc createValidatorsRawTable(db: SqStoreRef) =
db.exec("""
CREATE TABLE IF NOT EXISTS validators_raw(
validator_index INTEGER PRIMARY KEY,
pubkey BLOB NOT NULL UNIQUE
);
""").expect("DB")
proc createValidatorsView(db: SqStoreRef) =
db.exec("""
CREATE VIEW IF NOT EXISTS validators AS
SELECT
validator_index,
'0x' || lower(hex(pubkey)) as pubkey
FROM validators_raw;
""").expect("DB")
proc createInsertValidatorProc(db: SqStoreRef): auto =
db.prepareStmt("""
INSERT OR IGNORE INTO validators_raw(
validator_index,
pubkey)
VALUES(?, ?);""",
(int64, array[48, byte]), void).expect("DB")
proc collectBalances(balances: var seq[uint64], forkedState: ForkedHashedBeaconState) =
withState(forkedState):
balances = seq[uint64](forkyState.data.balances.data)
proc calculateDelta(info: RewardsAndPenalties): int64 =
info.source_outcome +
info.target_outcome +
info.head_outcome +
info.inclusion_delay_outcome +
info.sync_committee_outcome +
info.proposer_outcome +
info.slashing_outcome -
info.inactivity_penalty.int64 +
info.deposits.int64
proc printComponents(info: RewardsAndPenalties) =
echo "Components:"
echo "Source outcome: ", info.source_outcome
echo "Target outcome: ", info.target_outcome
echo "Head outcome: ", info.head_outcome
echo "Inclusion delay outcome: ", info.inclusion_delay_outcome
echo "Sync committee outcome: ", info.sync_committee_outcome
echo "Proposer outcome: ", info.proposer_outcome
echo "Slashing outcome: ", info.slashing_outcome
echo "Inactivity penalty: ", info.inactivity_penalty
echo "Deposits: ", info.deposits
proc checkBalance(validatorIndex: int64,
validator: RewardStatus | ParticipationInfo,
currentEpochBalance, previousEpochBalance: int64,
validatorInfo: RewardsAndPenalties) =
let delta = validatorInfo.calculateDelta
if currentEpochBalance == previousEpochBalance + delta:
return
echo "Validator: ", validatorIndex
echo "Is eligible: ", is_eligible_validator(validator)
echo "Current epoch balance: ", currentEpochBalance
echo "Previous epoch balance: ", previousEpochBalance
echo "State delta: ", currentEpochBalance - previousEpochBalance
echo "Computed delta: ", delta
printComponents(validatorInfo)
raiseAssert("Validator's previous epoch balance plus computed validator's " &
"delta is not equal to the validator's current epoch balance.")
proc getDbValidatorsCount(db: SqStoreRef): int64 =
var res: int64
discard db.exec("SELECT count(*) FROM validators", ()) do (r: int64):
res = r
return res
template inTransaction(db: SqStoreRef, dbName: string, body: untyped) =
try:
db.exec("BEGIN TRANSACTION;").expect(dbName)
body
finally:
db.exec("END TRANSACTION;").expect(dbName)
proc insertValidators(db: SqStoreRef, state: ForkedHashedBeaconState,
startIndex, endIndex: int64) =
var insertValidator {.global.}: SqliteStmt[
(int64, array[48, byte]), void]
once: insertValidator = db.createInsertValidatorProc
withState(state):
db.inTransaction("DB"):
for i in startIndex ..< endIndex:
insertValidator.exec(
(i, forkyState.data.validators[i].pubkey.toRaw)).expect("DB")
proc cmdValidatorDb(conf: DbConf, cfg: RuntimeConfig) =
# Create a database with performance information for every epoch
info "Opening database..."
let db = BeaconChainDB.new(conf.databaseDir.string, readOnly = true)
defer: db.close()
if (let v = ChainDAGRef.isInitialized(db); v.isErr()):
echo "Database not initialized"
quit 1
echo "Initializing block pool..."
let
validatorMonitor = newClone(ValidatorMonitor.init())
dag = ChainDAGRef.init(cfg, db, validatorMonitor, {}, conf.eraDir)
let outDb = SqStoreRef.init(conf.outDir, "validatorDb").expect("DB")
defer: outDb.close()
outDb.createValidatorsRawTable
outDb.createValidatorsView
let
unaggregatedFilesOutputDir = conf.outDir / "unaggregated"
aggregatedFilesOutputDir = conf.outDir / "aggregated"
startEpoch =
if conf.startEpoch.isSome:
Epoch(conf.startEpoch.get)
else:
try:
let unaggregatedFilesNextEpoch = getUnaggregatedFilesLastEpoch(
unaggregatedFilesOutputDir) + 1
let aggregatedFilesNextEpoch = getAggregatedFilesLastEpoch(
aggregatedFilesOutputDir) + 1
if conf.writeUnaggregatedFiles and conf.writeAggregatedFiles:
min(unaggregatedFilesNextEpoch, aggregatedFilesNextEpoch)
elif conf.writeUnaggregatedFiles:
unaggregatedFilesNextEpoch
elif conf.writeAggregatedFiles:
aggregatedFilesNextEpoch
else:
min(unaggregatedFilesNextEpoch, aggregatedFilesNextEpoch)
except OSError as e:
fatal "Failed to iterate epoch files", e = e.msg
quit QuitFailure
except ValueError as e:
fatal "Failed to parse epoch file name", e = e.msg
quit QuitFailure
endEpoch =
if conf.endEpoch.isSome:
Epoch(conf.endEpoch.get)
else:
dag.finalizedHead.slot.epoch # Avoid dealing with changes
if startEpoch > endEpoch:
fatal "Start epoch cannot be bigger than end epoch.",
startEpoch = startEpoch, endEpoch = endEpoch
quit QuitFailure
info "Analyzing performance for epochs.",
startEpoch = startEpoch, endEpoch = endEpoch
let
startEpochSlot = startEpoch.start_slot
endSlot = endEpoch.start_slot + SLOTS_PER_EPOCH
blockRefs = dag.getBlockRange(startEpochSlot, endSlot)
if not unaggregatedFilesOutputDir.dirExists:
try:
unaggregatedFilesOutputDir.createDir()
except IOError as e:
fatal "Failed to create unaggregated files directory (IOError)",
dirName = unaggregatedFilesOutputDir, e = e.msg
quit QuitFailure
except OSError as e:
fatal "Failed to create unaggregated files directory (OSError)",
dirName = unaggregatedFilesOutputDir, e = e.msg
quit QuitFailure
if not aggregatedFilesOutputDir.dirExists:
try:
aggregatedFilesOutputDir.createDir()
except IOError as e:
fatal "Failed to create aggregated files directory (IOError)",
dirName = aggregatedFilesOutputDir, e = e.msg
quit QuitFailure
except OSError as e:
fatal "Failed to create aggregated files directory (OSError)",
dirName = aggregatedFilesOutputDir, e = e.msg
quit QuitFailure
let tmpState = newClone(dag.headState)
var cache = StateCache()
let slot = if startEpochSlot > 0: startEpochSlot - 1 else: 0.Slot
if blockRefs.len > 0:
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
discard dag.updateState(
tmpState[], dag.atSlot(blockRefs[^1], slot).expect("block"), false, cache)
else:
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
discard dag.updateState(
tmpState[], dag.getBlockIdAtSlot(slot).expect("block"), false, cache)
let savedValidatorsCount = outDb.getDbValidatorsCount
var validatorsCount = getStateField(tmpState[], validators).len
outDb.insertValidators(tmpState[], savedValidatorsCount, validatorsCount)
var previousEpochBalances: seq[uint64]
collectBalances(previousEpochBalances, tmpState[])
var forkedInfo = ForkedEpochInfo()
var rewardsAndPenalties: seq[RewardsAndPenalties]
rewardsAndPenalties.setLen(validatorsCount)
var auxiliaryState: AuxiliaryState
auxiliaryState.copyParticipationFlags(tmpState[])
var aggregator = ValidatorDbAggregator.init(
aggregatedFilesOutputDir, conf.resolution, endEpoch)
proc processEpoch() =
let epoch = getStateField(tmpState[], slot).epoch
info "Processing epoch ...", epoch = epoch
var csvLines = newStringOfCap(1000000)
withState(tmpState[]):
withEpochInfo(forkedInfo):
doAssert forkyState.data.balances.len == info.validators.len
doAssert forkyState.data.balances.len == previousEpochBalances.len
doAssert forkyState.data.balances.len == rewardsAndPenalties.len
for index, validator in info.validators:
template rp: untyped = rewardsAndPenalties[index]
checkBalance(
index, validator, forkyState.data.balances.item(index).int64,
previousEpochBalances[index].int64, rp)
when infoFork == EpochInfoFork.Phase0:
rp.inclusion_delay = block:
let notSlashed = (RewardFlags.isSlashed notin validator.flags)
if notSlashed and validator.is_previous_epoch_attester.isSome():
2024-09-23 08:02:09 +00:00
Opt.some(validator.is_previous_epoch_attester.get().delay.uint64)
else:
2024-09-23 08:02:09 +00:00
Opt.none(uint64)
if conf.writeUnaggregatedFiles:
csvLines.add rp.serializeToCsv
if conf.writeAggregatedFiles:
aggregator.addValidatorData(index, rp)
if conf.writeUnaggregatedFiles:
let fileName = getFilePathForEpoch(epoch, unaggregatedFilesOutputDir)
var res = io2.removeFile(fileName)
doAssert res.isOk
res = io2.writeFile(fileName, snappy.encode(csvLines.toBytes))
doAssert res.isOk
if conf.writeAggregatedFiles:
aggregator.advanceEpochs(epoch, shouldShutDown)
if shouldShutDown: quit QuitSuccess
collectBalances(previousEpochBalances, tmpState[])
proc processSlots(ends: Slot, endsFlags: UpdateFlags) =
var currentSlot = getStateField(tmpState[], slot)
while currentSlot < ends:
let nextSlot = currentSlot + 1
let flags = if nextSlot == ends: endsFlags else: {}
2022-04-14 10:47:14 +00:00
if nextSlot.is_epoch:
withState(tmpState[]):
var stateData = newClone(forkyState.data)
rewardsAndPenalties.collectEpochRewardsAndPenalties(
stateData[], cache, cfg, flags)
let res = process_slots(cfg, tmpState[], nextSlot, cache, forkedInfo, flags)
doAssert res.isOk, "Slot processing can't fail with correct inputs"
currentSlot = nextSlot
2022-04-14 10:47:14 +00:00
if currentSlot.is_epoch:
processEpoch()
rewardsAndPenalties.setLen(0)
rewardsAndPenalties.setLen(validatorsCount)
auxiliaryState.copyParticipationFlags(tmpState[])
clear cache
for bi in 0 ..< blockRefs.len:
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
let forkedBlock = dag.getForkedBlock(blockRefs[blockRefs.len - bi - 1]).get()
withBlck(forkedBlock):
processSlots(forkyBlck.message.slot, {skipLastStateRootCalculation})
rewardsAndPenalties.collectBlockRewardsAndPenalties(
tmpState[], forkedBlock, auxiliaryState, cache, cfg)
let res = state_transition_block(
cfg, tmpState[], forkyBlck, cache, {}, noRollback)
if res.isErr:
fatal "State transition failed (!)"
quit QuitFailure
let newValidatorsCount = getStateField(tmpState[], validators).len
if newValidatorsCount > validatorsCount:
# Resize the structures in case a new validator has appeared after
# the state_transition_block procedure call ...
rewardsAndPenalties.setLen(newValidatorsCount)
previousEpochBalances.setLen(newValidatorsCount)
# ... and add the new validators to the database.
outDb.insertValidators(
tmpState[], validatorsCount, newValidatorsCount)
validatorsCount = newValidatorsCount
# Capture rewards of empty slots as well, including the epoch that got
# finalized
processSlots(endSlot, {})
proc controlCHook {.noconv.} =
notice "Shutting down after having received SIGINT."
shouldShutDown = true
proc exitOnSigterm(signal: cint) {.noconv.} =
notice "Shutting down after having received SIGTERM."
shouldShutDown = true
when isMainModule:
setControlCHook(controlCHook)
when defined(posix):
c_signal(SIGTERM, exitOnSigterm)
var
conf = DbConf.load()
cfg = getRuntimeConfig(conf.eth2Network)
case conf.cmd
of DbCmd.bench:
cmdBench(conf, cfg)
of DbCmd.dumpState:
cmdDumpState(conf)
of DbCmd.putState:
cmdPutState(conf, cfg)
of DbCmd.dumpBlock:
cmdDumpBlock(conf)
of DbCmd.putBlock:
cmdPutBlock(conf, cfg)
of DbCmd.putBlob:
cmdPutBlob(conf, cfg)
of DbCmd.rewindState:
cmdRewindState(conf, cfg)
of DbCmd.verifyEra:
cmdVerifyEra(conf, cfg)
of DbCmd.exportEra:
cmdExportEra(conf, cfg)
of DbCmd.importEra:
cmdImportEra(conf, cfg)
of DbCmd.validatorPerf:
cmdValidatorPerf(conf, cfg)
of DbCmd.validatorDb:
cmdValidatorDb(conf, cfg)