feat(cli): debug replay engine (#5972)

This commit is contained in:
Roman Krasiuk
2024-01-09 13:06:12 +01:00
committed by GitHub
parent 8078c515c7
commit ac104d09eb
5 changed files with 353 additions and 3 deletions

View File

@ -2,6 +2,7 @@
use clap::Args;
use reth_primitives::{TxHash, B256};
use std::path::PathBuf;
/// Parameters for debugging purposes
#[derive(Debug, Args, PartialEq, Default)]
@ -57,6 +58,12 @@ pub struct DebugArgs {
conflicts_with = "hook_transaction"
)]
pub hook_all: bool,
/// The path to store engine API messages at.
/// If specified, all of the intercepted engine API messages
/// will be written to specified location.
#[arg(long = "debug.engine-api-store", help_heading = "Debug", value_name = "PATH")]
pub engine_api_store: Option<PathBuf>,
}
#[cfg(test)]

View File

@ -11,7 +11,10 @@ use crate::{
db_type::{DatabaseBuilder, DatabaseInstance},
ext::{RethCliExt, RethNodeCommandConfig},
},
commands::node::{cl_events::ConsensusLayerHealthEvents, events},
commands::{
debug_cmd::EngineApiStore,
node::{cl_events::ConsensusLayerHealthEvents, events},
},
dirs::{ChainPath, DataDirPath, MaybePlatformPath},
init::init_genesis,
prometheus_exporter,
@ -1109,7 +1112,16 @@ impl<DB: Database + DatabaseMetrics + DatabaseMetadata + 'static> NodeBuilderWit
let payload_builder =
ext.spawn_payload_builder_service(&self.config.builder, &components)?;
let (consensus_engine_tx, consensus_engine_rx) = unbounded_channel();
let (consensus_engine_tx, mut consensus_engine_rx) = unbounded_channel();
if let Some(store_path) = self.config.debug.engine_api_store.clone() {
let (engine_intercept_tx, engine_intercept_rx) = unbounded_channel();
let engine_api_store = EngineApiStore::new(store_path);
executor.spawn_critical(
"engine api interceptor",
engine_api_store.intercept(consensus_engine_rx, engine_intercept_tx),
);
consensus_engine_rx = engine_intercept_rx;
};
let max_block = self.config.max_block(&network_client, provider_factory.clone()).await?;
// Configure the pipeline

View File

@ -9,6 +9,9 @@ mod execution;
mod in_memory_merkle;
mod merkle;
mod replay_engine;
pub(crate) use replay_engine::EngineApiStore;
/// `reth debug` command
#[derive(Debug, Parser)]
pub struct Command {
@ -27,6 +30,8 @@ pub enum Subcommands {
InMemoryMerkle(in_memory_merkle::Command),
/// Debug block building.
BuildBlock(build_block::Command),
/// Debug engine API by replaying stored messages.
ReplayEngine(replay_engine::Command),
}
impl Command {
@ -37,6 +42,7 @@ impl Command {
Subcommands::Merkle(command) => command.execute(ctx).await,
Subcommands::InMemoryMerkle(command) => command.execute(ctx).await,
Subcommands::BuildBlock(command) => command.execute(ctx).await,
Subcommands::ReplayEngine(command) => command.execute(ctx).await,
}
}
}

View File

@ -0,0 +1,325 @@
use crate::{
args::{
get_secret_key,
utils::{chain_help, genesis_value_parser, SUPPORTED_CHAINS},
DatabaseArgs, NetworkArgs,
},
dirs::{DataDirPath, MaybePlatformPath},
runner::CliContext,
};
use clap::Parser;
use eyre::Context;
use reth_basic_payload_builder::{BasicPayloadJobGenerator, BasicPayloadJobGeneratorConfig};
use reth_beacon_consensus::{
hooks::EngineHooks, BeaconConsensus, BeaconConsensusEngine, BeaconEngineMessage,
};
use reth_blockchain_tree::{
BlockchainTree, BlockchainTreeConfig, ShareableBlockchainTree, TreeExternals,
};
use reth_config::Config;
use reth_db::{init_db, DatabaseEnv};
use reth_interfaces::consensus::Consensus;
use reth_network::NetworkHandle;
use reth_network_api::NetworkInfo;
use reth_payload_builder::PayloadBuilderService;
use reth_primitives::{
fs::{self},
ChainSpec,
};
use reth_provider::{providers::BlockchainProvider, ProviderFactory};
use reth_revm::EvmProcessorFactory;
use reth_rpc_types::{
engine::{CancunPayloadFields, ForkchoiceState, PayloadAttributes},
ExecutionPayload,
};
use reth_stages::Pipeline;
use reth_tasks::TaskExecutor;
use reth_transaction_pool::noop::NoopTransactionPool;
use serde::{Deserialize, Serialize};
use std::{
collections::BTreeMap,
net::{SocketAddr, SocketAddrV4},
path::PathBuf,
sync::Arc,
time::{Duration, SystemTime},
};
use tokio::sync::{
mpsc::{self, UnboundedReceiver, UnboundedSender},
oneshot,
};
use tracing::*;
/// `reth debug replay-engine` command
/// This script will read stored engine API messages and replay them by the timestamp.
/// It does not require
#[derive(Debug, Parser)]
pub struct Command {
/// The path to the data dir for all reth files and subdirectories.
///
/// Defaults to the OS-specific data directory:
///
/// - Linux: `$XDG_DATA_HOME/reth/` or `$HOME/.local/share/reth/`
/// - Windows: `{FOLDERID_RoamingAppData}/reth/`
/// - macOS: `$HOME/Library/Application Support/reth/`
#[arg(long, value_name = "DATA_DIR", verbatim_doc_comment, default_value_t)]
datadir: MaybePlatformPath<DataDirPath>,
/// The chain this node is running.
///
/// Possible values are either a built-in chain or the path to a chain specification file.
#[arg(
long,
value_name = "CHAIN_OR_PATH",
long_help = chain_help(),
default_value = SUPPORTED_CHAINS[0],
value_parser = genesis_value_parser
)]
chain: Arc<ChainSpec>,
#[clap(flatten)]
db: DatabaseArgs,
#[clap(flatten)]
network: NetworkArgs,
/// The path to read engine API messages from.
#[arg(long = "engine-api-store", value_name = "PATH")]
engine_api_store: PathBuf,
/// The number of milliseconds between Engine API messages.
#[arg(long = "interval", default_value_t = 1_000)]
interval: u64,
}
impl Command {
async fn build_network(
&self,
config: &Config,
task_executor: TaskExecutor,
db: Arc<DatabaseEnv>,
network_secret_path: PathBuf,
default_peers_path: PathBuf,
) -> eyre::Result<NetworkHandle> {
let secret_key = get_secret_key(&network_secret_path)?;
let network = self
.network
.network_config(config, self.chain.clone(), secret_key, default_peers_path)
.with_task_executor(Box::new(task_executor))
.listener_addr(SocketAddr::V4(SocketAddrV4::new(self.network.addr, self.network.port)))
.discovery_addr(SocketAddr::V4(SocketAddrV4::new(
self.network.discovery.addr,
self.network.discovery.port,
)))
.build(ProviderFactory::new(db, self.chain.clone()))
.start_network()
.await?;
info!(target: "reth::cli", peer_id = %network.peer_id(), local_addr = %network.local_addr(), "Connected to P2P network");
debug!(target: "reth::cli", peer_id = ?network.peer_id(), "Full peer ID");
Ok(network)
}
/// Execute `debug replay-engine` command
pub async fn execute(self, ctx: CliContext) -> eyre::Result<()> {
let config = Config::default();
// Add network name to data dir
let data_dir = self.datadir.unwrap_or_chain_default(self.chain.chain);
let db_path = data_dir.db_path();
fs::create_dir_all(&db_path)?;
// Initialize the database
let db = Arc::new(init_db(db_path, self.db.log_level)?);
let provider_factory = ProviderFactory::new(db.clone(), self.chain.clone());
let consensus: Arc<dyn Consensus> = Arc::new(BeaconConsensus::new(Arc::clone(&self.chain)));
// Configure blockchain tree
let tree_externals = TreeExternals::new(
provider_factory.clone(),
Arc::clone(&consensus),
EvmProcessorFactory::new(self.chain.clone()),
);
let tree = BlockchainTree::new(tree_externals, BlockchainTreeConfig::default(), None)?;
let blockchain_tree = ShareableBlockchainTree::new(tree);
// Set up the blockchain provider
let blockchain_db =
BlockchainProvider::new(provider_factory.clone(), blockchain_tree.clone())?;
// Set up network
let network_secret_path =
self.network.p2p_secret_key.clone().unwrap_or_else(|| data_dir.p2p_secret_path());
let network = self
.build_network(
&config,
ctx.task_executor.clone(),
db.clone(),
network_secret_path,
data_dir.known_peers_path(),
)
.await?;
// Set up payload builder
#[cfg(not(feature = "optimism"))]
let payload_builder = reth_ethereum_payload_builder::EthereumPayloadBuilder::default();
// Optimism's payload builder is implemented on the OptimismPayloadBuilder type.
#[cfg(feature = "optimism")]
let payload_builder = reth_optimism_payload_builder::OptimismPayloadBuilder::default();
let payload_generator = BasicPayloadJobGenerator::with_builder(
blockchain_db.clone(),
NoopTransactionPool::default(),
ctx.task_executor.clone(),
BasicPayloadJobGeneratorConfig::default(),
self.chain.clone(),
payload_builder,
);
let (payload_service, payload_builder) = PayloadBuilderService::new(payload_generator);
ctx.task_executor.spawn_critical("payload builder service", Box::pin(payload_service));
// Configure the consensus engine
let network_client = network.fetch_client().await?;
let (consensus_engine_tx, consensus_engine_rx) = mpsc::unbounded_channel();
let (beacon_consensus_engine, beacon_engine_handle) = BeaconConsensusEngine::with_channel(
network_client,
Pipeline::builder().build(provider_factory),
blockchain_db.clone(),
Box::new(ctx.task_executor.clone()),
Box::new(network),
None,
false,
payload_builder,
None,
u64::MAX,
consensus_engine_tx,
consensus_engine_rx,
EngineHooks::new(),
)?;
info!(target: "reth::cli", "Consensus engine initialized");
// Run consensus engine to completion
let (tx, rx) = oneshot::channel();
info!(target: "reth::cli", "Starting consensus engine");
ctx.task_executor.spawn_critical_blocking("consensus engine", async move {
let res = beacon_consensus_engine.await;
let _ = tx.send(res);
});
let engine_api_store = EngineApiStore::new(self.engine_api_store.clone());
for filepath in engine_api_store.engine_messages_iter()? {
let contents =
fs::read(&filepath).wrap_err(format!("failed to read: {}", filepath.display()))?;
let message = serde_json::from_slice(&contents)
.wrap_err(format!("failed to parse: {}", filepath.display()))?;
debug!(target: "reth::cli", filepath = %filepath.display(), ?message, "Forwarding Engine API message");
match message {
StoredEngineApiMessage::ForkchoiceUpdated { state, payload_attrs } => {
let response =
beacon_engine_handle.fork_choice_updated(state, payload_attrs).await?;
debug!(target: "reth::cli", ?response, "Received for forkchoice updated");
}
StoredEngineApiMessage::NewPayload { payload, cancun_fields } => {
let response = beacon_engine_handle.new_payload(payload, cancun_fields).await?;
debug!(target: "reth::cli", ?response, "Received for new payload");
}
};
// Pause before next message
tokio::time::sleep(Duration::from_millis(self.interval)).await;
}
info!(target: "reth::cli", "Finished replaying engine API messages");
match rx.await? {
Ok(()) => info!("Beacon consensus engine exited successfully"),
Err(error) => {
error!(target: "reth::cli", %error, "Beacon consensus engine exited with an error")
}
};
Ok(())
}
}
#[derive(Debug, Serialize, Deserialize)]
#[serde(rename_all = "camelCase")]
enum StoredEngineApiMessage {
ForkchoiceUpdated { state: ForkchoiceState, payload_attrs: Option<PayloadAttributes> },
NewPayload { payload: ExecutionPayload, cancun_fields: Option<CancunPayloadFields> },
}
#[derive(Debug)]
pub(crate) struct EngineApiStore {
path: PathBuf,
}
impl EngineApiStore {
pub(crate) fn new(path: PathBuf) -> Self {
Self { path }
}
fn on_message(&self, msg: &BeaconEngineMessage, received_at: SystemTime) -> eyre::Result<()> {
fs::create_dir_all(&self.path)?; // ensure that store path had been created
let timestamp = received_at.duration_since(SystemTime::UNIX_EPOCH).unwrap().as_millis();
match msg {
BeaconEngineMessage::ForkchoiceUpdated { state, payload_attrs, tx: _tx } => {
let filename = format!("{}-fcu-{}.json", timestamp, state.head_block_hash);
fs::write(
self.path.join(filename),
serde_json::to_vec(&StoredEngineApiMessage::ForkchoiceUpdated {
state: *state,
payload_attrs: payload_attrs.clone(),
})?,
)?;
}
BeaconEngineMessage::NewPayload { payload, cancun_fields, tx: _tx } => {
let filename = format!("{}-new_payload-{}.json", timestamp, payload.block_hash());
fs::write(
self.path.join(filename),
serde_json::to_vec(&StoredEngineApiMessage::NewPayload {
payload: payload.clone(),
cancun_fields: cancun_fields.clone(),
})?,
)?;
}
// noop
BeaconEngineMessage::TransitionConfigurationExchanged |
BeaconEngineMessage::EventListener(_) => (),
};
Ok(())
}
pub(crate) fn engine_messages_iter(&self) -> eyre::Result<impl Iterator<Item = PathBuf>> {
let mut filenames_by_ts = BTreeMap::<u64, Vec<PathBuf>>::default();
for entry in fs::read_dir(&self.path)? {
let entry = entry?;
let filename = entry.file_name();
if let Some(filename) = filename.to_str().filter(|n| n.ends_with(".json")) {
if let Some(Ok(timestamp)) = filename.split('-').next().map(|n| n.parse::<u64>()) {
filenames_by_ts.entry(timestamp).or_default().push(entry.path());
tracing::debug!(target: "engine::store", timestamp, filename, "Queued engine API message");
} else {
tracing::warn!(target: "engine::store", %filename, "Could not parse timestamp from filename")
}
} else {
tracing::warn!(target: "engine::store", ?filename, "Skipping non json file");
}
}
Ok(filenames_by_ts.into_iter().flat_map(|(_, paths)| paths))
}
pub(crate) async fn intercept(
self,
mut rx: UnboundedReceiver<BeaconEngineMessage>,
to_engine: UnboundedSender<BeaconEngineMessage>,
) {
loop {
let Some(msg) = rx.recv().await else { break };
if let Err(error) = self.on_message(&msg, SystemTime::now()) {
error!(target: "engine::intercept", ?msg, %error, "Error handling Engine API message");
}
let _ = to_engine.send(msg);
}
}
}

View File

@ -8,7 +8,7 @@ use alloy_primitives::B256;
///
/// See also:
/// <https://github.com/ethereum/execution-apis/blob/fe8e13c288c592ec154ce25c534e26cb7ce0530d/src/engine/cancun.md#request>
#[derive(Debug, Clone, PartialEq, Eq, Hash, Default)]
#[derive(Debug, Clone, PartialEq, Eq, Hash, Default, serde::Serialize, serde::Deserialize)]
pub struct CancunPayloadFields {
/// The parent beacon block root.
pub parent_beacon_block_root: B256,