cerc-io/lighthouse
16
0
Fork 0
Code Issues 1 Pull Requests Packages Projects Releases Wiki Activity

Increase connection limits and allow priority connections (#2604)

Browse Source 
In previous network updates we have made our libp2p connections more lean by limiting the maximum number of connections a lighthouse node will accept before libp2p rejects new connections. 

However, we still maintain the logic that at maximum connections, we try to dial extra peers if they are needed by a validator client to publish messages on a specific subnet. The dials typically result in failures due the libp2p connection limits. 

This PR adds an extra factor, `PRIORITY_PEER_EXCESS` which sets aside a new allocation of peers we are able to dial in case we need these peers for the VC client.  This allocation sits along side the excess peer (which allows extra incoming peers on top of our target peer limit). 

The drawback here, is that libp2p now allows extra peers to connect to us (beyond the standard peer limit) which the peer manager should subsequently reject.
This commit is contained in:
Age Manning 2021-09-21 07:45:13 +00:00
parent fb1df2c926
commit 280e4fe23d
2 changed files with 35 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

40
beacon_node/eth2_libp2p/src/peer_manager/mod.rs
View File

@ -58,7 +58,12 @@ const HEARTBEAT_INTERVAL: u64 = 30;
/// PEER_EXCESS_FACTOR = 0.1 we allow 10% more nodes, i.e 55.
pub const PEER_EXCESS_FACTOR: f32 = 0.1;
/// A fraction of `PeerManager::target_peers` that need to be outbound-only connections.
pub const MIN_OUTBOUND_ONLY_FACTOR: f32 = 0.1;
pub const MIN_OUTBOUND_ONLY_FACTOR: f32 = 0.3;
/// The fraction of extra peers beyond the PEER_EXCESS_FACTOR that we allow us to dial for when
/// requiring subnet peers. More specifically, if our target peer limit is 50, and our excess peer
/// limit is 55, and we are at 55 peers, the following parameter provisions a few more slots of
/// dialing priority peers we need for validator duties.
pub const PRIORITY_PEER_EXCESS: f32 = 0.05;
/// Relative factor of peers that are allowed to have a negative gossipsub score without penalizing
/// them in lighthouse.
@ -78,8 +83,6 @@ pub struct PeerManager<TSpec: EthSpec> {
status_peers: HashSetDelay<PeerId>,
/// The target number of peers we would like to connect to.
target_peers: usize,
/// The maximum number of peers we allow (exceptions for subnet peers)
max_peers: usize,
/// A collection of sync committee subnets that we need to stay subscribed to.
/// Sync committee subnets are longer term (256 epochs). Hence, we need to re-run
/// discovery queries for subnet peers if we disconnect from existing sync
@ -137,7 +140,6 @@ impl<TSpec: EthSpec> PeerManager<TSpec> {
status_peers: HashSetDelay::new(Duration::from_secs(STATUS_INTERVAL)),
target_peers: config.target_peers,
sync_committee_subnets: Default::default(),
max_peers: (config.target_peers as f32 * (1.0 + PEER_EXCESS_FACTOR)).ceil() as usize,
heartbeat,
discovery_enabled: !config.disable_discovery,
log: log.clone(),
@ -224,9 +226,17 @@ impl<TSpec: EthSpec> PeerManager<TSpec> {
let connected_or_dialing = self.network_globals.connected_or_dialing_peers();
for (peer_id, min_ttl) in results {
// we attempt a connection if this peer is a subnet peer or if the max peer count
// is not yet filled (including dialing peers)
if (min_ttl.is_some() || connected_or_dialing + to_dial_peers.len() < self.max_peers)
// There are two conditions in deciding whether to dial this peer.
// 1. If we are less than our max connections. Discovery queries are executed to reach
// our target peers, so its fine to dial up to our max peers (which will get pruned
// in the next heartbeat down to our target).
// 2. If the peer is one our validators require for a specific subnet, then it is
// considered a priority. We have pre-allocated some extra priority slots for these
// peers as specified by PRIORITY_PEER_EXCESS. Therefore we dial these peers, even
// if we are already at our max_peer limit.
if (min_ttl.is_some()
&& connected_or_dialing + to_dial_peers.len() < self.max_priority_peers()
|| connected_or_dialing + to_dial_peers.len() < self.max_peers())
&& self.network_globals.peers.read().should_dial(&peer_id)
{
// This should be updated with the peer dialing. In fact created once the peer is
@ -300,6 +310,20 @@ impl<TSpec: EthSpec> PeerManager<TSpec> {
}
}
/// The maximum number of peers we allow to connect to us. This is `target_peers` * (1 +
/// PEER_EXCESS_FACTOR)
fn max_peers(&self) -> usize {
(self.target_peers as f32 * (1.0 + PEER_EXCESS_FACTOR)).ceil() as usize
}
/// The maximum number of peers we allow when dialing a priority peer (i.e a peer that is
/// subscribed to subnets that our validator requires. This is `target_peers` * (1 +
/// PEER_EXCESS_FACTOR + PRIORITY_PEER_EXCESS)
fn max_priority_peers(&self) -> usize {
(self.target_peers as f32 * (1.0 + PEER_EXCESS_FACTOR + PRIORITY_PEER_EXCESS)).ceil()
as usize
}
/* Notifications from the Swarm */
// A peer is being dialed.
@ -466,7 +490,7 @@ impl<TSpec: EthSpec> PeerManager<TSpec> {
/// Reports whether the peer limit is reached in which case we stop allowing new incoming
/// connections.
pub fn peer_limit_reached(&self) -> bool {
self.network_globals.connected_or_dialing_peers() >= self.max_peers
self.network_globals.connected_or_dialing_peers() >= self.max_peers()
}
/// Updates `PeerInfo` with `identify` information.

5
beacon_node/eth2_libp2p/src/service.rs
View File

@ -29,7 +29,7 @@ use std::sync::Arc;
use std::time::Duration;
use types::{ChainSpec, EnrForkId, EthSpec, ForkContext};
use crate::peer_manager::{MIN_OUTBOUND_ONLY_FACTOR, PEER_EXCESS_FACTOR};
use crate::peer_manager::{MIN_OUTBOUND_ONLY_FACTOR, PEER_EXCESS_FACTOR, PRIORITY_PEER_EXCESS};
pub const NETWORK_KEY_FILENAME: &str = "key";
/// The maximum simultaneous libp2p connections per peer.
@ -144,7 +144,8 @@ impl<TSpec: EthSpec> Service<TSpec> {
(config.target_peers as f32 * (1.0 + PEER_EXCESS_FACTOR)) as u32,
))
.with_max_established(Some(
(config.target_peers as f32 * (1.0 + PEER_EXCESS_FACTOR)) as u32,
(config.target_peers as f32 * (1.0 + PEER_EXCESS_FACTOR + PRIORITY_PEER_EXCESS))
as u32,
))
.with_max_established_per_peer(Some(MAX_CONNECTIONS_PER_PEER));