deal with rpc blobs in groups per block in the da checker. don't cache missing blob ids in the da checker.

This commit is contained in:
realbigsean
2023-04-04 10:29:07 -04:00
parent 22694d1c89
commit 6f12df37cf
8 changed files with 247 additions and 79 deletions

View File

@@ -625,15 +625,17 @@ impl<T: BeaconChainTypes> WorkEvent<T> {
}
}
pub fn rpc_blob(
blob: Arc<BlobSidecar<T::EthSpec>>,
pub fn rpc_blobs(
block_root: Hash256,
blobs: Vec<Arc<BlobSidecar<T::EthSpec>>>,
seen_timestamp: Duration,
process_type: BlockProcessType,
) -> Self {
Self {
drop_during_sync: false,
work: Work::RpcBlob {
block: blob,
work: Work::RpcBlobs {
block_root,
blobs,
seen_timestamp,
process_type,
},
@@ -936,8 +938,9 @@ pub enum Work<T: BeaconChainTypes> {
process_type: BlockProcessType,
should_process: bool,
},
RpcBlob {
block: Arc<BlobSidecar<T::EthSpec>>,
RpcBlobs {
block_root: Hash256,
blobs: Vec<Arc<BlobSidecar<T::EthSpec>>>,
seen_timestamp: Duration,
process_type: BlockProcessType,
},
@@ -1000,7 +1003,7 @@ impl<T: BeaconChainTypes> Work<T> {
Work::GossipLightClientFinalityUpdate { .. } => GOSSIP_LIGHT_CLIENT_FINALITY_UPDATE,
Work::GossipLightClientOptimisticUpdate { .. } => GOSSIP_LIGHT_CLIENT_OPTIMISTIC_UPDATE,
Work::RpcBlock { .. } => RPC_BLOCK,
Work::RpcBlob { .. } => RPC_BLOB,
Work::RpcBlobs { .. } => RPC_BLOB,
Work::ChainSegment { .. } => CHAIN_SEGMENT,
Work::Status { .. } => STATUS_PROCESSING,
Work::BlocksByRangeRequest { .. } => BLOCKS_BY_RANGE_REQUEST,
@@ -1513,7 +1516,7 @@ impl<T: BeaconChainTypes> BeaconProcessor<T> {
optimistic_update_queue.push(work, work_id, &self.log)
}
Work::RpcBlock { .. } => rpc_block_queue.push(work, work_id, &self.log),
Work::RpcBlob { .. } => rpc_blob_queue.push(work, work_id, &self.log),
Work::RpcBlobs { .. } => rpc_blob_queue.push(work, work_id, &self.log),
Work::ChainSegment { ref process_id, .. } => match process_id {
ChainSegmentProcessId::RangeBatchId { .. }
| ChainSegmentProcessId::ParentLookup { .. } => {
@@ -1936,12 +1939,14 @@ impl<T: BeaconChainTypes> BeaconProcessor<T> {
duplicate_cache,
should_process,
)),
Work::RpcBlob {
block,
Work::RpcBlobs {
block_root,
blobs,
seen_timestamp,
process_type,
} => task_spawner.spawn_async(worker.process_rpc_blob(
block,
} => task_spawner.spawn_async(worker.process_rpc_blobs(
block_root,
blobs,
seen_timestamp,
process_type,
)),

View File

@@ -136,16 +136,21 @@ impl<T: BeaconChainTypes> Worker<T> {
drop(handle);
}
pub async fn process_rpc_blob(
pub async fn process_rpc_blobs(
self,
blob: Arc<BlobSidecar<T::EthSpec>>,
block_root: Hash256,
blobs: Vec<Arc<BlobSidecar<T::EthSpec>>>,
seen_timestamp: Duration,
process_type: BlockProcessType,
) {
let result = self
.chain
.check_availability_and_maybe_import(
|chain| chain.data_availability_checker.put_rpc_blob(blob),
|chain| {
chain
.data_availability_checker
.put_rpc_blobs(block_root, blobs)
},
CountUnrealized::True,
)
.await;

View File

@@ -1,5 +1,5 @@
use std::collections::hash_map::Entry;
use std::collections::HashMap;
use std::collections::{HashMap, HashSet};
use std::sync::Arc;
use std::thread::sleep;
use std::time::Duration;
@@ -19,6 +19,7 @@ use types::{BlobSidecar, SignedBeaconBlock};
use crate::beacon_processor::{ChainSegmentProcessId, WorkEvent};
use crate::metrics;
use crate::sync::block_lookups::single_block_lookup::SingleBlobRequest;
use self::parent_lookup::PARENT_FAIL_TOLERANCE;
use self::{
@@ -59,6 +60,8 @@ pub(crate) struct BlockLookups<T: BeaconChainTypes> {
/// The flag allows us to determine if the peer returned data or sent us nothing.
single_block_lookups: FnvHashMap<Id, SingleBlockRequest<SINGLE_BLOCK_LOOKUP_MAX_ATTEMPTS>>,
single_blob_lookups: FnvHashMap<Id, SingleBlobRequest<SINGLE_BLOCK_LOOKUP_MAX_ATTEMPTS>>,
/// The logger for the import manager.
log: Logger,
}
@@ -72,6 +75,7 @@ impl<T: BeaconChainTypes> BlockLookups<T> {
FAILED_CHAINS_CACHE_EXPIRY_SECONDS,
)),
single_block_lookups: Default::default(),
single_blob_lookups: Default::default(),
log,
}
}
@@ -137,6 +141,56 @@ impl<T: BeaconChainTypes> BlockLookups<T> {
cx: &mut SyncNetworkContext<T>,
) {
todo!()
//
// let hash = Hash256::zero();
//
// // Do not re-request a blo that is already being requested
// if self
// .single_blob_lookups
// .values_mut()
// .any(|single_block_request| single_block_request.add_peer(&hash, &peer_id))
// {
// return;
// }
//
// if self.parent_lookups.iter_mut().any(|parent_req| {
// parent_req.add_peer(&hash, &peer_id) || parent_req.contains_block(&hash)
// }) {
// // If the block was already downloaded, or is being downloaded in this moment, do not
// // request it.
// return;
// }
//
// if self
// .processing_parent_lookups
// .values()
// .any(|(hashes, _last_parent_request)| hashes.contains(&hash))
// {
// // we are already processing this block, ignore it.
// return;
// }
//
// debug!(
// self.log,
// "Searching for block";
// "peer_id" => %peer_id,
// "block" => %hash
// );
//
// let mut single_block_request = SingleBlobRequest::new(hash, peer_id);
//
// let (peer_id, request) = single_block_request
// .request_block()
// .expect("none of the possible failure cases apply for a newly created block lookup");
// if let Ok(request_id) = cx.single_block_lookup_request(peer_id, request) {
// self.single_blob_lookups
// .insert(request_id, single_block_request);
//
// metrics::set_gauge(
// &metrics::SYNC_SINGLE_BLOB_LOOKUPS,
// self.single_blob_lookups.len() as i64,
// );
}
pub fn search_block_delayed(
@@ -171,6 +225,13 @@ impl<T: BeaconChainTypes> BlockLookups<T> {
peer_id: PeerId,
cx: &mut SyncNetworkContext<T>,
) {
//
// let missing_ids = cx.chain.data_availability_checker.get_missing_blob_ids(block, Some(root));
// // TODO(sean) how do we handle this erroring?
// if let Ok(missing_ids) = missing_ids {
// self.search_blobs(missing_ids, peer_id, cx);
// }
let parent_root = block.parent_root();
// If this block or it's parent is part of a known failed chain, ignore it.
if self.failed_chains.contains(&parent_root) || self.failed_chains.contains(&block_root) {
@@ -497,19 +558,17 @@ impl<T: BeaconChainTypes> BlockLookups<T> {
};
match result {
BlockProcessResult::Ok(status) => {
match status {
AvailabilityProcessingStatus::Imported(hash) => {
trace!(self.log, "Single block processing succeeded"; "block" => %root);
}
AvailabilityProcessingStatus::PendingBlobs(blobs) => {
// trigger?
}
AvailabilityProcessingStatus::PendingBlock(hash) => {
// logic error
}
BlockProcessResult::Ok(status) => match status {
AvailabilityProcessingStatus::Imported(hash) => {
trace!(self.log, "Single block processing succeeded"; "block" => %root);
}
}
AvailabilityProcessingStatus::PendingBlobs(blobs_ids) => {
self.search_blobs(blobs_ids, peer_id, cx);
}
AvailabilityProcessingStatus::PendingBlock(hash) => {
warn!(self.log, "Block processed but returned PendingBlock"; "block" => %hash);
}
},
BlockProcessResult::Ignored => {
// Beacon processor signalled to ignore the block processing result.
// This implies that the cpu is overloaded. Drop the request.
@@ -620,13 +679,18 @@ impl<T: BeaconChainTypes> BlockLookups<T> {
BlockProcessResult::Ok(AvailabilityProcessingStatus::PendingBlock(_)) => {
// doesn't make sense
}
BlockProcessResult::Ok(AvailabilityProcessingStatus::PendingBlobs(blobs)) => {
// trigger
BlockProcessResult::Ok(AvailabilityProcessingStatus::PendingBlobs(blobs_ids)) => {
self.search_blobs(blobs_ids, peer_id, cx);
}
BlockProcessResult::Err(BlockError::ParentUnknown(block)) => {
// need to keep looking for parents
// add the block back to the queue and continue the search
// TODO(sean) how do we handle this erroring?
let missing_ids = cx
.chain
.data_availability_checker
.get_missing_blob_ids(block.clone(), None)
.unwrap_or_default();
parent_lookup.add_block(block);
self.search_blobs(missing_ids, peer_id, cx);
self.request_parent(parent_lookup, cx);
}
BlockProcessResult::Ok(AvailabilityProcessingStatus::Imported(_))

View File

@@ -29,7 +29,23 @@ pub struct SingleBlockRequest<const MAX_ATTEMPTS: u8> {
failed_processing: u8,
/// How many times have we attempted to download this block.
failed_downloading: u8,
missing_blobs: Vec<BlobIdentifier>,
}
#[derive(PartialEq, Eq)]
pub struct SingleBlobRequest<const MAX_ATTEMPTS: u8> {
/// The hash of the requested block.
pub hash: Hash256,
pub blob_ids: Vec<BlobIdentifier>,
/// State of this request.
pub state: State,
/// Peers that should have this block.
pub available_peers: HashSet<PeerId>,
/// Peers from which we have requested this block.
pub used_peers: HashSet<PeerId>,
/// How many times have we attempted to process this block.
failed_processing: u8,
/// How many times have we attempted to download this block.
failed_downloading: u8,
}
#[derive(Debug, PartialEq, Eq)]
@@ -65,7 +81,6 @@ impl<const MAX_ATTEMPTS: u8> SingleBlockRequest<MAX_ATTEMPTS> {
used_peers: HashSet::default(),
failed_processing: 0,
failed_downloading: 0,
missing_blobs: vec![],
}
}

View File

@@ -64,7 +64,7 @@ pub struct SyncNetworkContext<T: BeaconChainTypes> {
/// Channel to send work to the beacon processor.
beacon_processor_send: mpsc::Sender<WorkEvent<T>>,
chain: Arc<BeaconChain<T>>,
pub chain: Arc<BeaconChain<T>>,
/// Logger for the `SyncNetworkContext`.
log: slog::Logger,
@@ -411,6 +411,8 @@ impl<T: BeaconChainTypes> SyncNetworkContext<T> {
}
}
// TODO(sean) add single blob lookup + parent lookup request methods
/// Sends a blocks by root request for a single block lookup.
pub fn single_block_lookup_request(
&mut self,