Sync fixes for `fusaka-devnet-4` #7876

jimmygchen · 2025-08-14T01:17:29Z

Issue Addressed

Remove penalties on failed block sidecar coupling as this is problematic in forky situations. Add them to fail peer only and deprioritise them instead.
- May no longer be required with the new range sync approach, as we now query columns by root.
Add logic to maintain custody column peers and prevent pruning
- Maintain peers across all data column subnets #7915
Prioritise status v2
- Prioritise StatusV2 over StatusV1 RPC protocol #7912
Penalise peers for sending blocks containing invalid execution payload.
Only send lookup request to synced / advanced peers
- Only send lookup requests to peers that are synced or advacned #7913

This reverts commit 1f776a4

…shold.

beacon_node/src/config.rs

beacon_node/lighthouse_network/src/peer_manager/mod.rs

dapplion · 2025-08-14T20:48:32Z

beacon_node/lighthouse_network/src/peer_manager/mod.rs

+/// For non sampling columns, requirements is lower, but need to ensure there is peer for
+/// publishing.
+pub const MIN_SAMPLING_COLUMN_SUBNET_PEERS: u64 = TARGET_SUBNET_PEERS as u64;
+pub const MIN_NON_SAMPLING_COLUMN_SUBNET_PEERS: u64 = 1;


Why do we need 1 peer and not 0 in this subnets?

I thought it would be useful to have at least one peer per column in the case where we're publishing data columns.

beacon_node/network/src/sync/range_sync/chain.rs

dapplion · 2025-08-14T20:53:19Z

Did a first review pass and approach looks solid, just some minor comments

This reverts commit 5f38f31.

jimmygchen · 2025-08-15T06:07:07Z

beacon_node/network/src/network_beacon_processor/sync_methods.rs

+                            err
+                        ),
+                        peer_action: Some(PeerAction::LowToleranceError),
+                    })


@pawanjay176 do we need this change?

This may be due to optimistic EL right?

lighthouse/beacon_node/beacon_chain/src/block_verification.rs

Lines 430 to 432 in 522bd9e

// An honest optimistic node may propagate blocks which are rejected by an EE, do not

// penalize them.

ExecutionPayloadError::RejectedByExecutionEngine { .. } => false,

jimmygchen · 2025-08-15T06:15:08Z

beacon_node/lighthouse_network/src/peer_manager/peerdb.rs

@@ -300,6 +300,7 @@ impl<E: EthSpec> PeerDB<E> {
            .filter(move |(_, info)| {
                // We check both the metadata and gossipsub data as we only want to count long-lived subscribed peers
                info.is_connected()
+                    && info.is_synced_or_advanced()


Need to think again about this change - this is used to determine if we have enough good peers when starting discovery query:

lighthouse/beacon_node/lighthouse_network/src/service/mod.rs

Lines 1093 to 1103 in c8bbe47

// Already have target number of peers, no need for subnet discovery

let peers_on_subnet = self

.network_globals

.peers

.read()

.good_peers_on_subnet(s.subnet)

.count();

if peers_on_subnet >= TARGET_SUBNET_PEERS {

trace!(

subnet = ?s.subnet,

reason = "Already connected to desired peers",

So the condition to push a PeerManagerEvent::DiscoverSubnetPeers below must also satisfy this

lighthouse/beacon_node/lighthouse_network/src/peer_manager/mod.rs

Lines 969 to 992 in a250daa

fn maintain_custody_peers(&mut self) {

let subnets_to_discover: Vec<SubnetDiscovery> = self

.network_globals

.sampling_subnets()

.iter()

.filter_map(|custody_subnet| {

if self

.network_globals

.peers

.read()

.has_good_peers_in_custody_subnet(

custody_subnet,

MIN_SAMPLING_COLUMN_SUBNET_PEERS as usize,

)

{

None

} else {

Some(SubnetDiscovery {

subnet: Subnet::DataColumn(*custody_subnet),

min_ttl: None,

})

}

})

.collect();

Otherwise it might result in pushing a DiscoverSubnetPeers event to network but then it gets ignored by network.

mergify · 2025-08-15T06:26:12Z

Some required checks have failed. Could you please take a look @jimmygchen? 🙏

jimmygchen · 2025-08-15T06:27:22Z

@pawanjay176 I've done a bit of clean up, but there's still a bit of mess around maintaining peers & discovery:

#7876 (comment)

I'll revisit next week.

Squashed commit of the following: commit d8b6a99 Author: Jimmy Chen <[email protected]> Date: Mon Aug 18 23:48:28 2025 +1000 Remove unnecessary test commit 832d862 Author: Jimmy Chen <[email protected]> Date: Fri Aug 15 16:22:48 2025 +1000 Clippy commit 0437243 Merge: a250daa 317dc0f Author: Jimmy Chen <[email protected]> Date: Fri Aug 15 16:22:30 2025 +1000 Merge branch 'unstable' into fusaka-devnet-4-revive commit a250daa Author: Jimmy Chen <[email protected]> Date: Fri Aug 15 16:02:05 2025 +1000 Clean up and refactor custody peer checks commit ec7956b Author: Jimmy Chen <[email protected]> Date: Fri Aug 15 14:40:06 2025 +1000 Revert "Increase target peer count to 200." This reverts commit 5f38f31. commit cdc417f Author: Jimmy Chen <[email protected]> Date: Thu Aug 14 23:39:10 2025 +1000 Prevent sampling subnets from getting pruned if below min target threshold. commit 9ce9318 Author: Jimmy Chen <[email protected]> Date: Thu Aug 14 17:39:57 2025 +1000 Check for sync status for custody subnet discovery commit 1b8f0e4 Author: Jimmy Chen <[email protected]> Date: Thu Aug 14 17:30:30 2025 +1000 Fix tests. commit 1465d72 Author: Jimmy Chen <[email protected]> Date: Thu Aug 14 11:47:33 2025 +1000 Only send lookup requests to peers that are synced or advacned. commit 5f38f31 Author: Jimmy Chen <[email protected]> Date: Thu Aug 14 11:30:04 2025 +1000 Increase target peer count to 200. commit 56ca5bf Author: Jimmy Chen <[email protected]> Date: Thu Aug 14 11:14:20 2025 +1000 Remove penalties for block sidecar coupling. commit 8e95a54 Author: Jimmy Chen <[email protected]> Date: Thu Aug 14 11:07:50 2025 +1000 Revert "Reduce number of head chains we sync to" This reverts commit 1f776a4 commit 0144205 Author: Pawan Dhananjay <[email protected]> Date: Wed Aug 13 17:33:47 2025 -0700 Request columns from global peer pool commit a8f6801 Author: Pawan Dhananjay <[email protected]> Date: Wed Aug 13 16:37:12 2025 -0700 Priorotize status v2 commit c8bbe47 Author: Pawan Dhananjay <[email protected]> Date: Wed Aug 13 16:36:56 2025 -0700 Penalize if invalid EL block commit 1f776a4 Author: Pawan Dhananjay <[email protected]> Date: Wed Aug 13 16:36:13 2025 -0700 Reduce number of head chains we sync to

Prioritise `StatusV2` over `StatusV1` RPC protocol. A bug discovered during devnet-4 testing and extracted from the sync fixes PR #7876.

pawanjay176 and others added 6 commits August 13, 2025 16:36

Reduce number of head chains we sync to

1f776a4

Penalize if invalid EL block

c8bbe47

Priorotize status v2

a8f6801

Request columns from global peer pool

0144205

Revert "Reduce number of head chains we sync to"

8e95a54

This reverts commit 1f776a4

Remove penalties for block sidecar coupling.

56ca5bf

jimmygchen requested a review from jxs as a code owner August 14, 2025 01:17

jimmygchen added 5 commits August 14, 2025 11:30

Increase target peer count to 200.

5f38f31

Only send lookup requests to peers that are synced or advacned.

1465d72

Fix tests.

1b8f0e4

Check for sync status for custody subnet discovery

9ce9318

Prevent sampling subnets from getting pruned if below min target thre…

cdc417f

…shold.

dapplion reviewed Aug 14, 2025

View reviewed changes

beacon_node/src/config.rs Outdated Show resolved Hide resolved

dapplion reviewed Aug 14, 2025

View reviewed changes

jimmygchen added the do-not-merge label Aug 14, 2025

jimmygchen added 2 commits August 15, 2025 14:40

Revert "Increase target peer count to 200."

ec7956b

This reverts commit 5f38f31.

Clean up and refactor custody peer checks

a250daa

jimmygchen commented Aug 15, 2025

View reviewed changes

jimmygchen and others added 2 commits August 15, 2025 16:22

Merge branch 'unstable' into fusaka-devnet-4-revive

0437243

Clippy

832d862

jimmygchen added the ready-for-review The code is ready for review label Aug 15, 2025

mergify bot added waiting-on-author The reviewer has suggested changes and awaits thier implementation. and removed ready-for-review The code is ready for review labels Aug 15, 2025

jimmygchen self-assigned this Aug 18, 2025

Remove unnecessary test

d8b6a99

This was referenced Aug 18, 2025

fusaka-devnet-3 test branch #7893

Closed

Prioritise StatusV2 over StatusV1 RPC protocol #7912

Merged

Maintain peers across all data column subnets #7915

Open

mergify bot pushed a commit that referenced this pull request Aug 21, 2025

Prioritise StatusV2 over StatusV1 RPC protocol (#7912)

d24a6d2

Prioritise `StatusV2` over `StatusV1` RPC protocol. A bug discovered during devnet-4 testing and extracted from the sync fixes PR #7876.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Sync fixes for `fusaka-devnet-4` #7876

Sync fixes for `fusaka-devnet-4` #7876

Uh oh!

jimmygchen commented Aug 14, 2025 •

edited

Loading

Uh oh!

Uh oh!

Uh oh!

dapplion Aug 14, 2025

Uh oh!

jimmygchen Aug 14, 2025

Uh oh!

Uh oh!

dapplion commented Aug 14, 2025

Uh oh!

jimmygchen Aug 15, 2025

Uh oh!

jimmygchen Aug 15, 2025 •

edited

Loading

Uh oh!

mergify bot commented Aug 15, 2025

Uh oh!

jimmygchen commented Aug 15, 2025

Uh oh!

Uh oh!

	// An honest optimistic node may propagate blocks which are rejected by an EE, do not
	// penalize them.
	ExecutionPayloadError::RejectedByExecutionEngine { .. } => false,

	// Already have target number of peers, no need for subnet discovery
	let peers_on_subnet = self
	.network_globals
	.peers
	.read()
	.good_peers_on_subnet(s.subnet)
	.count();
	if peers_on_subnet >= TARGET_SUBNET_PEERS {
	trace!(
	subnet = ?s.subnet,
	reason = "Already connected to desired peers",

	fn maintain_custody_peers(&mut self) {
	let subnets_to_discover: Vec<SubnetDiscovery> = self
	.network_globals
	.sampling_subnets()
	.iter()
	.filter_map(\|custody_subnet\| {
	if self
	.network_globals
	.peers
	.read()
	.has_good_peers_in_custody_subnet(
	custody_subnet,
	MIN_SAMPLING_COLUMN_SUBNET_PEERS as usize,
	)
	{
	None
	} else {
	Some(SubnetDiscovery {
	subnet: Subnet::DataColumn(*custody_subnet),
	min_ttl: None,
	})
	}
	})
	.collect();

Sync fixes for fusaka-devnet-4 #7876

Are you sure you want to change the base?

Sync fixes for fusaka-devnet-4 #7876

Uh oh!

Conversation

jimmygchen commented Aug 14, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Issue Addressed

Uh oh!

Uh oh!

Uh oh!

dapplion Aug 14, 2025

Choose a reason for hiding this comment

Uh oh!

jimmygchen Aug 14, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

dapplion commented Aug 14, 2025

Uh oh!

jimmygchen Aug 15, 2025

Choose a reason for hiding this comment

Uh oh!

jimmygchen Aug 15, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mergify bot commented Aug 15, 2025

Uh oh!

jimmygchen commented Aug 15, 2025

Uh oh!

Uh oh!

Sync fixes for `fusaka-devnet-4` #7876

Sync fixes for `fusaka-devnet-4` #7876

jimmygchen commented Aug 14, 2025 •

edited

Loading

jimmygchen Aug 15, 2025 •

edited

Loading