@peter/async padding (#219)

* support async padding
This commit is contained in:
0g-peterzhb 2024-10-08 13:13:32 +08:00 committed by GitHub
parent 395aeabde7
commit 20266e0a6c
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
15 changed files with 203 additions and 78 deletions

2
Cargo.lock generated
View File

@ -7280,7 +7280,9 @@ dependencies = [
"serde_json",
"shared_types",
"static_assertions",
"task_executor",
"tiny-keccak",
"tokio",
"tracing",
"typenum",
"zgs_seal",

View File

@ -481,7 +481,7 @@ impl LogSyncManager {
} else {
// check if current node need to save at least one segment
let store = self.store.clone();
let shard_config = store.flow().get_shard_config();
let shard_config = store.get_shard_config();
let start_segment_index = tx.start_entry_index as usize / PORA_CHUNK_SIZE;
let end_segment_index =
(tx.start_entry_index as usize + bytes_to_chunks(tx.size as usize) - 1)

View File

@ -423,7 +423,7 @@ impl Libp2pEventHandler {
let addr = self.get_listen_addr_or_add().await?;
let timestamp = timestamp_now();
let shard_config = self.store.get_store().flow().get_shard_config();
let shard_config = self.store.get_store().get_shard_config();
let msg = AnnounceFile {
tx_ids,
@ -699,7 +699,7 @@ impl Libp2pEventHandler {
}
// notify sync layer if shard config matches
let my_shard_config = self.store.get_store().flow().get_shard_config();
let my_shard_config = self.store.get_store().get_shard_config();
if my_shard_config.intersect(&announced_shard_config) {
for tx_id in msg.tx_ids.iter() {
self.send_to_sync(SyncMessage::AnnounceFileGossip {
@ -911,7 +911,9 @@ mod tests {
let (network_send, network_recv) = mpsc::unbounded_channel();
let (sync_send, sync_recv) = channel::Channel::unbounded("test");
let (chunk_pool_send, _chunk_pool_recv) = mpsc::unbounded_channel();
let store = LogManager::memorydb(LogConfig::default()).unwrap();
let executor = runtime.task_executor.clone();
let store = LogManager::memorydb(LogConfig::default(), executor).unwrap();
Self {
runtime,
network_globals: Arc::new(network_globals),

View File

@ -183,7 +183,7 @@ impl RpcServer for RpcServerImpl {
async fn get_shard_config(&self) -> RpcResult<ShardConfig> {
debug!("zgs_getShardConfig");
let shard_config = self.ctx.log_store.get_store().flow().get_shard_config();
let shard_config = self.ctx.log_store.get_store().get_shard_config();
Ok(shard_config)
}

View File

@ -89,9 +89,10 @@ impl ClientBuilder {
/// Initializes in-memory storage.
pub fn with_memory_store(mut self) -> Result<Self, String> {
let executor = require!("sync", self, runtime_context).clone().executor;
// TODO(zz): Set config.
let store = Arc::new(
LogManager::memorydb(LogConfig::default())
LogManager::memorydb(LogConfig::default(), executor)
.map_err(|e| format!("Unable to start in-memory store: {:?}", e))?,
);
@ -109,8 +110,9 @@ impl ClientBuilder {
/// Initializes RocksDB storage.
pub fn with_rocksdb_store(mut self, config: &StorageConfig) -> Result<Self, String> {
let executor = require!("sync", self, runtime_context).clone().executor;
let store = Arc::new(
LogManager::rocksdb(LogConfig::default(), &config.db_dir)
LogManager::rocksdb(LogConfig::default(), &config.db_dir, executor)
.map_err(|e| format!("Unable to start RocksDB store: {:?}", e))?,
);

View File

@ -95,23 +95,22 @@ impl Store {
&self,
seal_index_max: usize,
) -> anyhow::Result<Option<Vec<SealTask>>> {
self.spawn(move |store| store.flow().pull_seal_chunk(seal_index_max))
self.spawn(move |store| store.pull_seal_chunk(seal_index_max))
.await
}
pub async fn submit_seal_result(&self, answers: Vec<SealAnswer>) -> anyhow::Result<()> {
self.spawn(move |store| store.flow().submit_seal_result(answers))
self.spawn(move |store| store.submit_seal_result(answers))
.await
}
pub async fn load_sealed_data(&self, chunk_index: u64) -> Result<Option<MineLoadChunk>> {
self.spawn(move |store| store.flow().load_sealed_data(chunk_index))
self.spawn(move |store| store.load_sealed_data(chunk_index))
.await
}
pub async fn get_num_entries(&self) -> Result<u64> {
self.spawn(move |store| store.flow().get_num_entries())
.await
self.spawn(move |store| store.get_num_entries()).await
}
pub async fn remove_chunks_batch(&self, batch_list: &[u64]) -> Result<()> {
@ -122,7 +121,7 @@ impl Store {
pub async fn update_shard_config(&self, shard_config: ShardConfig) {
self.spawn(move |store| {
store.flow().update_shard_config(shard_config);
store.update_shard_config(shard_config);
Ok(())
})
.await

View File

@ -29,6 +29,8 @@ itertools = "0.13.0"
serde = { version = "1.0.197", features = ["derive"] }
parking_lot = "0.12.3"
serde_json = "1.0.127"
tokio = { version = "1.10.0", features = ["sync"] }
task_executor = { path = "../../common/task_executor" }
[dev-dependencies]
rand = "0.8.5"

View File

@ -14,14 +14,18 @@ use storage::{
},
LogManager,
};
use task_executor::test_utils::TestRuntime;
fn write_performance(c: &mut Criterion) {
if Path::new("db_write").exists() {
fs::remove_dir_all("db_write").unwrap();
}
let runtime = TestRuntime::default();
let executor = runtime.task_executor.clone();
let store: Arc<RwLock<dyn Store>> = Arc::new(RwLock::new(
LogManager::rocksdb(LogConfig::default(), "db_write")
LogManager::rocksdb(LogConfig::default(), "db_write", executor)
.map_err(|e| format!("Unable to start RocksDB store: {:?}", e))
.unwrap(),
));
@ -105,8 +109,12 @@ fn read_performance(c: &mut Criterion) {
fs::remove_dir_all("db_read").unwrap();
}
let runtime = TestRuntime::default();
let executor = runtime.task_executor.clone();
let store: Arc<RwLock<dyn Store>> = Arc::new(RwLock::new(
LogManager::rocksdb(LogConfig::default(), "db_read")
LogManager::rocksdb(LogConfig::default(), "db_read", executor)
.map_err(|e| format!("Unable to start RocksDB store: {:?}", e))
.unwrap(),
));

View File

@ -3,7 +3,8 @@ use super::{MineLoadChunk, SealAnswer, SealTask};
use crate::config::ShardConfig;
use crate::error::Error;
use crate::log_store::log_manager::{
bytes_to_entries, COL_ENTRY_BATCH, COL_ENTRY_BATCH_ROOT, COL_FLOW_MPT_NODES, PORA_CHUNK_SIZE,
bytes_to_entries, data_to_merkle_leaves, COL_ENTRY_BATCH, COL_ENTRY_BATCH_ROOT,
COL_FLOW_MPT_NODES, ENTRY_SIZE, PORA_CHUNK_SIZE,
};
use crate::log_store::{FlowRead, FlowSeal, FlowWrite};
use crate::{try_option, ZgsKeyValueDB};
@ -11,7 +12,7 @@ use anyhow::{anyhow, bail, Result};
use append_merkle::{MerkleTreeInitialData, MerkleTreeRead};
use itertools::Itertools;
use parking_lot::RwLock;
use shared_types::{ChunkArray, DataRoot, FlowProof};
use shared_types::{ChunkArray, DataRoot, FlowProof, Merkle};
use ssz::{Decode, Encode};
use ssz_derive::{Decode as DeriveDecode, Encode as DeriveEncode};
use std::cmp::Ordering;
@ -441,6 +442,10 @@ impl FlowDBStore {
// and they will be updated in the merkle tree with `fill_leaf` by the caller.
let mut leaf_list = Vec::new();
let mut expected_index = 0;
let empty_data = vec![0; PORA_CHUNK_SIZE * ENTRY_SIZE];
let empty_root = *Merkle::new(data_to_merkle_leaves(&empty_data)?, 0, None).root();
for r in self.kvdb.iter(COL_ENTRY_BATCH_ROOT) {
let (index_bytes, root_bytes) = r?;
let (batch_index, subtree_depth) = decode_batch_root_key(index_bytes.as_ref())?;
@ -475,25 +480,26 @@ impl FlowDBStore {
expected_index += 1;
}
Ordering::Greater => {
bail!(
"unexpected chunk leaf in range, expected={}, get={}, range={:?}",
expected_index,
batch_index,
range_root,
);
while batch_index > expected_index {
// Fill the gap with empty leaves.
root_list.push((1, empty_root));
expected_index += 1;
}
range_root = None;
root_list.push((1, root));
expected_index += 1;
}
}
}
} else if expected_index == batch_index {
} else {
while batch_index > expected_index {
// Fill the gap with empty leaves.
root_list.push((1, empty_root));
expected_index += 1;
}
range_root = Some(BatchRoot::Multiple((subtree_depth, root)));
root_list.push((subtree_depth, root));
expected_index += 1 << (subtree_depth - 1);
} else {
bail!(
"unexpected range root: expected={} get={}",
expected_index,
batch_index
);
}
}
let extra_node_list = self.get_mpt_node_list()?;

View File

@ -1,3 +1,4 @@
use crate::config::ShardConfig;
use crate::log_store::flow_store::{batch_iter_sharded, FlowConfig, FlowStore};
use crate::log_store::tx_store::TransactionStore;
use crate::log_store::{
@ -20,11 +21,12 @@ use shared_types::{
use std::cmp::Ordering;
use std::collections::BTreeMap;
use std::path::Path;
use std::sync::mpsc;
use std::sync::Arc;
use tracing::{debug, error, info, instrument, trace, warn};
use super::tx_store::BlockHashAndSubmissionIndex;
use super::LogStoreInner;
use super::{FlowSeal, MineLoadChunk, SealAnswer, SealTask};
/// 256 Bytes
pub const ENTRY_SIZE: usize = 256;
@ -45,11 +47,18 @@ pub const COL_NUM: u32 = 9;
// Process at most 1M entries (256MB) pad data at a time.
const PAD_MAX_SIZE: usize = 1 << 20;
pub struct UpdateFlowMessage {
pub root_map: BTreeMap<usize, (H256, usize)>,
pub pad_data: usize,
pub tx_start_flow_index: u64,
}
pub struct LogManager {
pub(crate) db: Arc<dyn ZgsKeyValueDB>,
tx_store: TransactionStore,
flow_store: FlowStore,
flow_store: Arc<FlowStore>,
merkle: RwLock<MerkleManager>,
sender: mpsc::Sender<UpdateFlowMessage>,
}
struct MerkleManager {
@ -139,16 +148,6 @@ pub struct LogConfig {
pub flow: FlowConfig,
}
impl LogStoreInner for LogManager {
fn flow(&self) -> &dyn super::Flow {
&self.flow_store
}
fn flow_mut(&mut self) -> &mut dyn super::Flow {
&mut self.flow_store
}
}
impl LogStoreChunkWrite for LogManager {
fn put_chunks(&self, tx_seq: u64, chunks: ChunkArray) -> Result<()> {
let mut merkle = self.merkle.write();
@ -388,6 +387,14 @@ impl LogStoreWrite for LogManager {
fn delete_block_hash_by_number(&self, block_number: u64) -> Result<()> {
self.tx_store.delete_block_hash_by_number(block_number)
}
fn update_shard_config(&self, shard_config: ShardConfig) {
self.flow_store.update_shard_config(shard_config)
}
fn submit_seal_result(&self, answers: Vec<SealAnswer>) -> Result<()> {
self.flow_store.submit_seal_result(answers)
}
}
impl LogStoreChunkRead for LogManager {
@ -578,24 +585,48 @@ impl LogStoreRead for LogManager {
fn check_tx_pruned(&self, tx_seq: u64) -> crate::error::Result<bool> {
self.tx_store.check_tx_pruned(tx_seq)
}
fn pull_seal_chunk(&self, seal_index_max: usize) -> Result<Option<Vec<SealTask>>> {
self.flow_store.pull_seal_chunk(seal_index_max)
}
fn get_num_entries(&self) -> Result<u64> {
self.flow_store.get_num_entries()
}
fn load_sealed_data(&self, chunk_index: u64) -> Result<Option<MineLoadChunk>> {
self.flow_store.load_sealed_data(chunk_index)
}
fn get_shard_config(&self) -> ShardConfig {
self.flow_store.get_shard_config()
}
}
impl LogManager {
pub fn rocksdb(config: LogConfig, path: impl AsRef<Path>) -> Result<Self> {
pub fn rocksdb(
config: LogConfig,
path: impl AsRef<Path>,
executor: task_executor::TaskExecutor,
) -> Result<Self> {
let mut db_config = DatabaseConfig::with_columns(COL_NUM);
db_config.enable_statistics = true;
let db = Arc::new(Database::open(&db_config, path)?);
Self::new(db, config)
Self::new(db, config, executor)
}
pub fn memorydb(config: LogConfig) -> Result<Self> {
pub fn memorydb(config: LogConfig, executor: task_executor::TaskExecutor) -> Result<Self> {
let db = Arc::new(kvdb_memorydb::create(COL_NUM));
Self::new(db, config)
Self::new(db, config, executor)
}
fn new(db: Arc<dyn ZgsKeyValueDB>, config: LogConfig) -> Result<Self> {
fn new(
db: Arc<dyn ZgsKeyValueDB>,
config: LogConfig,
executor: task_executor::TaskExecutor,
) -> Result<Self> {
let tx_store = TransactionStore::new(db.clone())?;
let flow_store = FlowStore::new(db.clone(), config.flow);
let flow_store = Arc::new(FlowStore::new(db.clone(), config.flow));
let mut initial_data = flow_store.get_chunk_root_list()?;
// If the last tx `put_tx` does not complete, we will revert it in `initial_data.subtree_list`
// first and call `put_tx` later. The known leaves in its data will be saved in `extra_leaves`
@ -699,13 +730,19 @@ impl LogManager {
pora_chunks_merkle,
last_chunk_merkle,
});
let log_manager = Self {
let (sender, receiver) = mpsc::channel();
let mut log_manager = Self {
db,
tx_store,
flow_store,
merkle,
sender,
};
log_manager.start_receiver(receiver, executor);
if let Some(tx) = last_tx_to_insert {
log_manager.revert_to(tx.seq - 1)?;
log_manager.put_tx(tx)?;
@ -727,6 +764,41 @@ impl LogManager {
Ok(log_manager)
}
fn start_receiver(
&mut self,
rx: mpsc::Receiver<UpdateFlowMessage>,
executor: task_executor::TaskExecutor,
) {
let flow_store = self.flow_store.clone();
executor.spawn(
async move {
loop {
match rx.recv() {
std::result::Result::Ok(data) => {
// Update the root index.
flow_store.put_batch_root_list(data.root_map).unwrap();
// Update the flow database.
// This should be called before `complete_last_chunk_merkle` so that we do not save
// subtrees with data known.
flow_store
.append_entries(ChunkArray {
data: vec![0; data.pad_data],
start_index: data.tx_start_flow_index,
})
.unwrap();
}
std::result::Result::Err(_) => {
error!("Receiver error");
}
};
}
},
"pad_tx",
);
// Wait for the spawned thread to finish
// let _ = handle.join().expect("Thread panicked");
}
fn gen_proof(&self, flow_index: u64, maybe_root: Option<DataRoot>) -> Result<FlowProof> {
match maybe_root {
None => self.gen_proof_at_version(flow_index, None),
@ -863,6 +935,7 @@ impl LogManager {
);
if extra != 0 {
for pad_data in Self::padding((first_subtree_size - extra) as usize) {
let mut is_full_empty = true;
let mut root_map = BTreeMap::new();
// Update the in-memory merkle tree.
@ -874,6 +947,7 @@ impl LogManager {
let mut completed_chunk_index = None;
if pad_data.len() < last_chunk_pad {
is_full_empty = false;
merkle
.last_chunk_merkle
.append_list(data_to_merkle_leaves(&pad_data)?);
@ -882,6 +956,7 @@ impl LogManager {
.update_last(*merkle.last_chunk_merkle.root());
} else {
if last_chunk_pad != 0 {
is_full_empty = false;
// Pad the last chunk.
merkle
.last_chunk_merkle
@ -910,16 +985,26 @@ impl LogManager {
assert_eq!(pad_data.len(), start_index * ENTRY_SIZE);
}
// Update the root index.
self.flow_store.put_batch_root_list(root_map)?;
// Update the flow database.
// This should be called before `complete_last_chunk_merkle` so that we do not save
// subtrees with data known.
let data_size = pad_data.len() / ENTRY_SIZE;
self.flow_store.append_entries(ChunkArray {
data: pad_data,
start_index: tx_start_flow_index,
})?;
if is_full_empty {
self.sender.send(UpdateFlowMessage {
root_map,
pad_data: pad_data.len(),
tx_start_flow_index,
})?;
} else {
self.flow_store.put_batch_root_list(root_map).unwrap();
// Update the flow database.
// This should be called before `complete_last_chunk_merkle` so that we do not save
// subtrees with data known.
self.flow_store
.append_entries(ChunkArray {
data: pad_data.to_vec(),
start_index: tx_start_flow_index,
})
.unwrap();
}
tx_start_flow_index += data_size as u64;
if let Some(index) = completed_chunk_index {
self.complete_last_chunk_merkle(index, &mut *merkle)?;

View File

@ -76,6 +76,14 @@ pub trait LogStoreRead: LogStoreChunkRead {
/// Return flow root and length.
fn get_context(&self) -> Result<(DataRoot, u64)>;
fn pull_seal_chunk(&self, seal_index_max: usize) -> Result<Option<Vec<SealTask>>>;
fn get_num_entries(&self) -> Result<u64>;
fn load_sealed_data(&self, chunk_index: u64) -> Result<Option<MineLoadChunk>>;
fn get_shard_config(&self) -> ShardConfig;
}
pub trait LogStoreChunkRead {
@ -145,6 +153,10 @@ pub trait LogStoreWrite: LogStoreChunkWrite {
) -> Result<bool>;
fn delete_block_hash_by_number(&self, block_number: u64) -> Result<()>;
fn update_shard_config(&self, shard_config: ShardConfig);
fn submit_seal_result(&self, answers: Vec<SealAnswer>) -> Result<()>;
}
pub trait LogStoreChunkWrite {
@ -168,19 +180,10 @@ pub trait LogChunkStore: LogStoreChunkRead + LogStoreChunkWrite + Send + Sync +
impl<T: LogStoreChunkRead + LogStoreChunkWrite + Send + Sync + 'static> LogChunkStore for T {}
pub trait Store:
LogStoreRead + LogStoreWrite + LogStoreInner + config::Configurable + Send + Sync + 'static
LogStoreRead + LogStoreWrite + config::Configurable + Send + Sync + 'static
{
}
impl<
T: LogStoreRead + LogStoreWrite + LogStoreInner + config::Configurable + Send + Sync + 'static,
> Store for T
{
}
pub trait LogStoreInner {
fn flow(&self) -> &dyn Flow;
fn flow_mut(&mut self) -> &mut dyn Flow;
}
impl<T: LogStoreRead + LogStoreWrite + config::Configurable + Send + Sync + 'static> Store for T {}
pub struct MineLoadChunk {
// Use `Vec` instead of array to avoid thread stack overflow.

View File

@ -8,11 +8,15 @@ use ethereum_types::H256;
use rand::random;
use shared_types::{compute_padded_chunk_size, ChunkArray, Transaction, CHUNK_SIZE};
use std::cmp;
use task_executor::test_utils::TestRuntime;
#[test]
fn test_put_get() {
let config = LogConfig::default();
let store = LogManager::memorydb(config.clone()).unwrap();
let runtime = TestRuntime::default();
let executor = runtime.task_executor.clone();
let store = LogManager::memorydb(config.clone(), executor).unwrap();
let chunk_count = config.flow.batch_size + config.flow.batch_size / 2 - 1;
// Aligned with size.
let start_offset = 1024;
@ -169,8 +173,10 @@ fn test_put_tx() {
fn create_store() -> LogManager {
let config = LogConfig::default();
let runtime = TestRuntime::default();
let executor = runtime.task_executor.clone();
LogManager::memorydb(config).unwrap()
LogManager::memorydb(config, executor).unwrap()
}
fn put_tx(store: &mut LogManager, chunk_count: usize, seq: u64) {

View File

@ -492,7 +492,7 @@ impl SerialSyncController {
metrics::SERIAL_SYNC_SEGMENT_LATENCY.update_since(since.0);
let shard_config = self.store.get_store().flow().get_shard_config();
let shard_config = self.store.get_store().get_shard_config();
let next_chunk = segment_to_sector(shard_config.next_segment_index(
sector_to_segment(from_chunk),
sector_to_segment(self.tx_start_chunk_in_flow),
@ -1622,7 +1622,7 @@ mod tests {
let num_chunks = 123;
let config = LogConfig::default();
let store = Arc::new(LogManager::memorydb(config).unwrap());
let store = Arc::new(LogManager::memorydb(config, task_executor.clone()).unwrap());
create_controller(task_executor, peer_id, store, tx_id, num_chunks)
}

View File

@ -807,7 +807,7 @@ impl SyncService {
}
async fn tx_sync_start_index(store: &Store, tx: &Transaction) -> Result<Option<u64>> {
let shard_config = store.get_store().flow().get_shard_config();
let shard_config = store.get_store().get_shard_config();
let start_segment = sector_to_segment(tx.start_entry_index());
let end =
bytes_to_chunks(usize::try_from(tx.size).map_err(|e| anyhow!("tx size e={}", e))?);
@ -1294,7 +1294,9 @@ mod tests {
let config = LogConfig::default();
let store = Arc::new(LogManager::memorydb(config.clone()).unwrap());
let executor = runtime.task_executor.clone();
let store = Arc::new(LogManager::memorydb(config.clone(), executor).unwrap());
let init_peer_id = identity::Keypair::generate_ed25519().public().to_peer_id();
let file_location_cache: Arc<FileLocationCache> =

View File

@ -9,6 +9,8 @@ use storage::{
LogManager,
};
use task_executor::test_utils::TestRuntime;
/// Creates stores for local node and peers with initialized transaction of specified chunk count.
/// The first store is for local node, and data not stored. The second store is for peers, and all
/// transactions are finalized for file sync.
@ -22,8 +24,11 @@ pub fn create_2_store(
Vec<Vec<u8>>,
) {
let config = LogConfig::default();
let mut store = LogManager::memorydb(config.clone()).unwrap();
let mut peer_store = LogManager::memorydb(config).unwrap();
let runtime = TestRuntime::default();
let executor = runtime.task_executor.clone();
let mut store = LogManager::memorydb(config.clone(), executor.clone()).unwrap();
let mut peer_store = LogManager::memorydb(config, executor).unwrap();
let mut offset = 1;
let mut txs = vec![];
@ -115,7 +120,10 @@ pub mod tests {
impl TestStoreRuntime {
pub fn new_store() -> impl LogStore {
LogManager::memorydb(LogConfig::default()).unwrap()
let runtime = TestRuntime::default();
let executor = runtime.task_executor.clone();
LogManager::memorydb(LogConfig::default(), executor).unwrap()
}
pub fn new(store: Arc<dyn LogStore>) -> TestStoreRuntime {