p2p: allowed_peers list propagated network-wide instead of being local to the node

まず、P2P peer のアドバタイズおよび接続処理のパスを通じて --allowed_peers CLI 設定を追跡します。Issue に記載されている sequencer/fullnode トポロジーを再現し、その後、制限がそれを設定した node によってのみ適用され、制限のない fullnodes が新しい peers を受け入れることを確認します。

Sequencer version : v1.1.0-rc.1

Description

When running a sequencer node with --allowed_peers=fullnode-1,fullnode-2, the allowed peers list appears to be propagated to the rest of the network. As a result, other fullnodes that try to connect via P2P to fullnode-1 or fullnode-2 are rejected, even though those fullnodes never set allowed_peers themselves.

Expected Behavior

The allowed_peers configuration should be local to the node that runs the argument. It should only restrict which peers that specific node accepts connections from. It must not be advertised or enforced on other nodes in the network.

Actual Behavior

The allowed_peers list set on the sequencer is propagated across the network. Nodes that did not set allowed_peers still refuse connections from nodes not on the sequencer's list.

Network Topology

The following diagram illustrates the affected network and the expected isolation of the allowed_peers constraint:

                        Eden Testnet Network
  ┌─────────────────────────────────────────────────────────────────────┐
  │                                                                     │
  │   ┌───────────────────────────────┐                                 │
  │   │  Sequencer                    │                                 │
  │   │  allowed_peers=fullnode-1,    │                                 │
  │   │             fullnode-2        │                                 │
  │   └──────────────┬────────────────┘                                │
  │                  │ P2P                                              │
  │         ┌────────┴────────┐                                        │
  │         │                 │                                        │
  │   ┌─────▼──────┐   ┌──────▼─────┐                                 │
  │   │ fullnode-1 │   │ fullnode-2 │                                  │
  │   └─────┬──────┘   └──────┬─────┘                                 │
  │         │                 │                                        │
  │         │ P2P  (EXPECTED) │ P2P  (EXPECTED)                       │
  │         │                 │                                        │
  │   ┌─────▼─────────────────▼─────┐                                 │
  │   │     new-fullnode-N ...       │                                 │
  │   │  (unable to connect — BUG)  │                                 │
  │   └─────────────────────────────┘                                 │
  │                                                                     │
  └─────────────────────────────────────────────────────────────────────┘

  Legend:
    ──►  P2P connection (established / expected)
    BUG: new-fullnode-N cannot connect to fullnode-1 or fullnode-2
         because the sequencer's allowed_peers list is enforced there

Intended behavior (allowed_peers should be local)

  Sequencer  ←──(only accepts fullnode-1, fullnode-2)──►  fullnode-1
                                                       ──►  fullnode-2

  fullnode-1  ←──(no restriction; accepts any peer)──►  new-fullnode-N
  fullnode-2  ←──(no restriction; accepts any peer)──►  new-fullnode-N

Steps to Reproduce

1. Start a network with: sequencer, fullnode-1, fullnode-2.
2. Run the sequencer with --allowed_peers=fullnode-1,fullnode-2.
3. Attempt to join a new fullnode that connects via P2P to fullnode-1 or fullnode-2.
4. Observe: the new fullnode is rejected by fullnode-1/fullnode-2 despite them not having allowed_peers set.

Environment

• Network: eden-testnet
• Observed: week of 2026-04-14

Impact

Any new node attempting to sync via fullnode-1 or fullnode-2 is silently blocked, making fullnode only syncing through DA unless modifying the sequencer's allowed_peers list.