From: G8XSU <3442979+G8XSU@users.noreply.github.com>
Date: Fri, 14 Jun 2024 23:56:36 +0000 (-0700)
Subject: Optimize ChannelMonitor persistence on block connections.
X-Git-Tag: v0.0.124-beta~71^2
X-Git-Url: http://git.bitcoin.ninja/index.cgi?a=commitdiff_plain;h=bf28957f6d488d1258a1e6f21af29344b18e2b75;p=rust-lightning

Optimize ChannelMonitor persistence on block connections.

Currently, every block connection triggers the persistence of all
ChannelMonitors with an updated best_block. This approach poses
challenges for large node operators managing thousands of channels.
Furthermore, it leads to a thundering herd problem
(https://en.wikipedia.org/wiki/Thundering_herd_problem), overwhelming
the storage with simultaneous requests.

To address this issue, we now persist ChannelMonitors at a
regular cadence, spreading their persistence across blocks to
mitigate spikes in write operations.

Outcome: After doing this, Ldk's IO footprint should be reduced
by ~50 times. The processing time required to sync each block
will be significantly reduced, particularly for nodes with 1000s
of channels, as write latency plays a significant role in this process.
As a result, the Node/ChainMonitor will be blocked for a shorter
duration, leading to further efficiency gains.
---

diff --git a/lightning/src/chain/chainmonitor.rs b/lightning/src/chain/chainmonitor.rs
index 58bf5f800..e6bb9d907 100644
--- a/lightning/src/chain/chainmonitor.rs
+++ b/lightning/src/chain/chainmonitor.rs
@@ -44,6 +44,7 @@ use crate::prelude::*;
 use crate::sync::{RwLock, RwLockReadGuard, Mutex, MutexGuard};
 use core::ops::Deref;
 use core::sync::atomic::{AtomicUsize, Ordering};
+use bitcoin::hashes::Hash;
 use bitcoin::secp256k1::PublicKey;
 
 /// `Persist` defines behavior for persisting channel monitors: this could mean
@@ -260,10 +261,11 @@ where C::Target: chain::Filter,
 	{
 		let err_str = "ChannelMonitor[Update] persistence failed unrecoverably. This indicates we cannot continue normal operation and must shut down.";
 		let funding_outpoints = hash_set_from_iter(self.monitors.read().unwrap().keys().cloned());
+		let channel_count = funding_outpoints.len();
 		for funding_outpoint in funding_outpoints.iter() {
 			let monitor_lock = self.monitors.read().unwrap();
 			if let Some(monitor_state) = monitor_lock.get(funding_outpoint) {
-				if self.update_monitor_with_chain_data(header, txdata, &process, funding_outpoint, &monitor_state).is_err() {
+				if self.update_monitor_with_chain_data(header, best_height, txdata, &process, funding_outpoint, &monitor_state, channel_count).is_err() {
 					// Take the monitors lock for writing so that we poison it and any future
 					// operations going forward fail immediately.
 					core::mem::drop(monitor_lock);
@@ -278,7 +280,7 @@ where C::Target: chain::Filter,
 		let monitor_states = self.monitors.write().unwrap();
 		for (funding_outpoint, monitor_state) in monitor_states.iter() {
 			if !funding_outpoints.contains(funding_outpoint) {
-				if self.update_monitor_with_chain_data(header, txdata, &process, funding_outpoint, &monitor_state).is_err() {
+				if self.update_monitor_with_chain_data(header, best_height, txdata, &process, funding_outpoint, &monitor_state, channel_count).is_err() {
 					log_error!(self.logger, "{}", err_str);
 					panic!("{}", err_str);
 				}
@@ -297,14 +299,29 @@ where C::Target: chain::Filter,
 	}
 
 	fn update_monitor_with_chain_data<FN>(
-		&self, header: &Header, txdata: &TransactionData, process: FN, funding_outpoint: &OutPoint,
-		monitor_state: &MonitorHolder<ChannelSigner>
+		&self, header: &Header, best_height: Option<u32>, txdata: &TransactionData, process: FN, funding_outpoint: &OutPoint,
+		monitor_state: &MonitorHolder<ChannelSigner>, channel_count: usize,
 	) -> Result<(), ()> where FN: Fn(&ChannelMonitor<ChannelSigner>, &TransactionData) -> Vec<TransactionOutputs> {
 		let monitor = &monitor_state.monitor;
 		let logger = WithChannelMonitor::from(&self.logger, &monitor, None);
-		let mut txn_outputs;
-		{
-			txn_outputs = process(monitor, txdata);
+
+		let mut txn_outputs = process(monitor, txdata);
+
+		let get_partition_key = |funding_outpoint: &OutPoint| {
+			let funding_txid_hash = funding_outpoint.txid.to_raw_hash();
+			let funding_txid_hash_bytes = funding_txid_hash.as_byte_array();
+			let funding_txid_u32 = u32::from_be_bytes([funding_txid_hash_bytes[0], funding_txid_hash_bytes[1], funding_txid_hash_bytes[2], funding_txid_hash_bytes[3]]);
+			funding_txid_u32.wrapping_add(best_height.unwrap_or_default())
+		};
+
+		let partition_factor = if channel_count < 15 {
+			5
+		} else {
+			50 // ~ 8hours
+		};
+
+		let has_pending_claims = monitor_state.monitor.has_pending_claims();
+		if has_pending_claims || get_partition_key(funding_outpoint) % partition_factor == 0 {
 			log_trace!(logger, "Syncing Channel Monitor for channel {}", log_funding_info!(monitor));
 			match self.persister.update_persisted_channel(*funding_outpoint, None, monitor) {
 				ChannelMonitorUpdateStatus::Completed =>
@@ -313,10 +330,10 @@ where C::Target: chain::Filter,
 					),
 				ChannelMonitorUpdateStatus::InProgress => {
 					log_trace!(logger, "Channel Monitor sync for channel {} in progress.", log_funding_info!(monitor));
-				},
+				}
 				ChannelMonitorUpdateStatus::UnrecoverableError => {
 					return Err(());
-				},
+				}
 			}
 		}
 
@@ -870,14 +887,17 @@ impl<ChannelSigner: EcdsaChannelSigner, C: Deref, T: Deref, F: Deref, L: Deref,
 
 #[cfg(test)]
 mod tests {
-	use crate::check_added_monitors;
+	use crate::{check_added_monitors, check_closed_event};
 	use crate::{expect_payment_path_successful, get_event_msg};
 	use crate::{get_htlc_update_msgs, get_revoke_commit_msgs};
 	use crate::chain::{ChannelMonitorUpdateStatus, Watch};
-	use crate::events::{Event, MessageSendEvent, MessageSendEventsProvider};
+	use crate::chain::channelmonitor::ANTI_REORG_DELAY;
+	use crate::events::{ClosureReason, Event, MessageSendEvent, MessageSendEventsProvider};
 	use crate::ln::functional_test_utils::*;
 	use crate::ln::msgs::ChannelMessageHandler;
 
+	const CHAINSYNC_MONITOR_PARTITION_FACTOR: u32 = 5;
+
 	#[test]
 	fn test_async_ooo_offchain_updates() {
 		// Test that if we have multiple offchain updates being persisted and they complete
@@ -983,6 +1003,79 @@ mod tests {
 		check_added_monitors!(nodes[0], 1);
 	}
 
+	#[test]
+	fn test_chainsync_triggers_distributed_monitor_persistence() {
+		let chanmon_cfgs = create_chanmon_cfgs(3);
+		let node_cfgs = create_node_cfgs(3, &chanmon_cfgs);
+		let node_chanmgrs = create_node_chanmgrs(3, &node_cfgs, &[None, None, None]);
+		let nodes = create_network(3, &node_cfgs, &node_chanmgrs);
+
+		// Use FullBlockViaListen to avoid duplicate calls to process_chain_data and skips_blocks() in
+		// case of other connect_styles.
+		*nodes[0].connect_style.borrow_mut() = ConnectStyle::FullBlockViaListen;
+		*nodes[1].connect_style.borrow_mut() = ConnectStyle::FullBlockViaListen;
+		*nodes[2].connect_style.borrow_mut() = ConnectStyle::FullBlockViaListen;
+
+		let _channel_1 = create_announced_chan_between_nodes(&nodes, 0, 1).2;
+		let channel_2 = create_announced_chan_between_nodes_with_value(&nodes, 0, 2, 1_000_000, 0).2;
+
+		chanmon_cfgs[0].persister.chain_sync_monitor_persistences.lock().unwrap().clear();
+		chanmon_cfgs[1].persister.chain_sync_monitor_persistences.lock().unwrap().clear();
+		chanmon_cfgs[2].persister.chain_sync_monitor_persistences.lock().unwrap().clear();
+
+		connect_blocks(&nodes[0], CHAINSYNC_MONITOR_PARTITION_FACTOR * 2);
+		connect_blocks(&nodes[1], CHAINSYNC_MONITOR_PARTITION_FACTOR * 2);
+		connect_blocks(&nodes[2], CHAINSYNC_MONITOR_PARTITION_FACTOR * 2);
+
+		// Connecting [`DEFAULT_CHAINSYNC_PARTITION_FACTOR`] * 2 blocks should trigger only 2 writes
+		// per monitor/channel.
+		assert_eq!(2 * 2, chanmon_cfgs[0].persister.chain_sync_monitor_persistences.lock().unwrap().len());
+		assert_eq!(2, chanmon_cfgs[1].persister.chain_sync_monitor_persistences.lock().unwrap().len());
+		assert_eq!(2, chanmon_cfgs[2].persister.chain_sync_monitor_persistences.lock().unwrap().len());
+
+		// Test that monitors with pending_claims are persisted on every block.
+		// Now, close channel_2 i.e. b/w node-0 and node-2 to create pending_claim in node[0].
+		nodes[0].node.force_close_broadcasting_latest_txn(&channel_2, &nodes[2].node.get_our_node_id(), "Channel force-closed".to_string()).unwrap();
+		check_closed_event!(&nodes[0], 1, ClosureReason::HolderForceClosed { broadcasted_latest_txn: Some(true) }, false,
+			[nodes[2].node.get_our_node_id()], 1000000);
+		check_closed_broadcast(&nodes[0], 1, true);
+		let close_tx = nodes[0].tx_broadcaster.txn_broadcasted.lock().unwrap().split_off(0);
+		assert_eq!(close_tx.len(), 1);
+
+		mine_transaction(&nodes[2], &close_tx[0]);
+		check_added_monitors(&nodes[2], 1);
+		check_closed_broadcast(&nodes[2], 1, true);
+		check_closed_event!(&nodes[2], 1, ClosureReason::CommitmentTxConfirmed, false,
+			[nodes[0].node.get_our_node_id()], 1000000);
+
+		chanmon_cfgs[0].persister.chain_sync_monitor_persistences.lock().unwrap().clear();
+		chanmon_cfgs[2].persister.chain_sync_monitor_persistences.lock().unwrap().clear();
+
+		// For channel_2, there should be a monitor write for every block connection.
+		// We connect [`DEFAULT_CHAINSYNC_MONITOR_PARTITION_FACTOR`] blocks since we don't know when
+		// channel_1 monitor persistence will occur, with [`DEFAULT_CHAINSYNC_MONITOR_PARTITION_FACTOR`]
+		// it will be persisted exactly once.
+		connect_blocks(&nodes[0], CHAINSYNC_MONITOR_PARTITION_FACTOR);
+		connect_blocks(&nodes[2], CHAINSYNC_MONITOR_PARTITION_FACTOR);
+
+		// DEFAULT_CHAINSYNC_MONITOR_PARTITION_FACTOR writes for channel_2 due to pending_claim, 1 for
+		// channel_1
+		assert_eq!((CHAINSYNC_MONITOR_PARTITION_FACTOR + 1) as usize, chanmon_cfgs[0].persister.chain_sync_monitor_persistences.lock().unwrap().len());
+		// For node[2], there is no pending_claim
+		assert_eq!(1, chanmon_cfgs[2].persister.chain_sync_monitor_persistences.lock().unwrap().len());
+
+		// Confirm claim for node[0] with ANTI_REORG_DELAY and reset monitor write counter.
+		mine_transaction(&nodes[0], &close_tx[0]);
+		connect_blocks(&nodes[0], ANTI_REORG_DELAY - 1);
+		check_added_monitors(&nodes[0], 1);
+		chanmon_cfgs[0].persister.chain_sync_monitor_persistences.lock().unwrap().clear();
+
+		// Again connect 1 full cycle of DEFAULT_CHAINSYNC_MONITOR_PARTITION_FACTOR blocks, it should only
+		// result in 1 write per monitor/channel.
+		connect_blocks(&nodes[0], CHAINSYNC_MONITOR_PARTITION_FACTOR);
+		assert_eq!(2, chanmon_cfgs[0].persister.chain_sync_monitor_persistences.lock().unwrap().len());
+	}
+
 	#[test]
 	#[cfg(feature = "std")]
 	fn update_during_chainsync_poisons_channel() {
@@ -991,12 +1084,15 @@ mod tests {
 		let node_chanmgrs = create_node_chanmgrs(2, &node_cfgs, &[None, None]);
 		let nodes = create_network(2, &node_cfgs, &node_chanmgrs);
 		create_announced_chan_between_nodes(&nodes, 0, 1);
+		*nodes[0].connect_style.borrow_mut() = ConnectStyle::FullBlockViaListen;
 
 		chanmon_cfgs[0].persister.set_update_ret(ChannelMonitorUpdateStatus::UnrecoverableError);
 
 		assert!(std::panic::catch_unwind(|| {
 			// Returning an UnrecoverableError should always panic immediately
-			connect_blocks(&nodes[0], 1);
+			// Connecting [`DEFAULT_CHAINSYNC_PARTITION_FACTOR`] blocks so that we trigger some persistence
+			// after accounting for block-height based partitioning/distribution.
+			connect_blocks(&nodes[0], CHAINSYNC_MONITOR_PARTITION_FACTOR);
 		}).is_err());
 		assert!(std::panic::catch_unwind(|| {
 			// ...and also poison our locks causing later use to panic as well
diff --git a/lightning/src/chain/channelmonitor.rs b/lightning/src/chain/channelmonitor.rs
index cd6061a20..c6d1bcdd2 100644
--- a/lightning/src/chain/channelmonitor.rs
+++ b/lightning/src/chain/channelmonitor.rs
@@ -1812,6 +1812,12 @@ impl<Signer: EcdsaChannelSigner> ChannelMonitor<Signer> {
 		);
 	}
 
+	/// Returns true if the monitor has pending claim requests that are not fully confirmed yet.
+	pub fn has_pending_claims(&self) -> bool
+	{
+		self.inner.lock().unwrap().onchain_tx_handler.has_pending_claims()
+	}
+
 	/// Triggers rebroadcasts of pending claims from a force-closed channel after a transaction
 	/// signature generation failure.
 	pub fn signer_unblocked<B: Deref, F: Deref, L: Deref>(
diff --git a/lightning/src/chain/onchaintx.rs b/lightning/src/chain/onchaintx.rs
index e5757cfd1..07a75361d 100644
--- a/lightning/src/chain/onchaintx.rs
+++ b/lightning/src/chain/onchaintx.rs
@@ -535,6 +535,13 @@ impl<ChannelSigner: EcdsaChannelSigner> OnchainTxHandler<ChannelSigner> {
 		}
 	}
 
+	/// Returns true if we are currently tracking any pending claim requests that are not fully 
+	/// confirmed yet.
+	pub(super) fn has_pending_claims(&self) -> bool
+	{
+		self.pending_claim_requests.len() != 0
+	}
+
 	/// Lightning security model (i.e being able to redeem/timeout HTLC or penalize counterparty
 	/// onchain) lays on the assumption of claim transactions getting confirmed before timelock
 	/// expiration (CSV or CLTV following cases). In case of high-fee spikes, claim tx may get stuck