Full Node Crashes with Different Proposer Address on Contract Deployment

[khalifaT]

Description

When deploying a contract to a full node synced with our sequencer, the node crashes with a panic. While simple transfer transactions work fine, contract deployment fails with an error indicating the validator does not exist for a particular consensus address.
Error Message

goroutine 361 [running]: github.com/rollkit/rollkit/block.(*Manager).trySyncNextBlock(0xc001367800, {0x935a710, 0xc0028a0500}, 0x51c691)
/home/root/go/pkg/mod/github.com/rollkit/rollkit@v0.13.10/block/manager.go:548 +0x81c
ERR Transaction verification failed error="failed to get evm config from context: failed to obtain coinbase address: failed to retrieve validator from block proposer address artvalcons1cmgs93cgwmhpzwdyvw80zm4av20jdrvgh9lz8f: validator does not exist [artela-rollkit/x/evm/keeper/proposer.go:16]" module=server

Investigation
I've added logging to print the proposerAddress inside the EVM config.
The issue appears in:

• artela-rollkit/app/ante/evm/aspect_context.go

  Line 56 in a2e9af5

  return ctx, fmt.Errorf("failed to get evm config from context: %w", err)

• artela-rollkit/x/evm/keeper/config.go

  Line 29 in a2e9af5

  return nil, errorsmod.Wrap(err, "failed to obtain coinbase address")

Key Finding
Most log messages show the correct sequencer address, matching the sequencer logs. However, the last message shows a different validator address, which causes the error.
Hypothesis
The GetProposerAddress function in the code takes a parameter (proposerAddress) but should potentially be retrieving the consensus address . For some reason, during contract deployment, the function is being called with an incorrect validator address that doesn't exist in the registry.

Reproduction Steps

• Deploy a full node with the sequencer
• Sync the node (transfer transactions work fine)
• Attempt to deploy a contract
• Node crashes with the validator error

[lukeweb3]

​I attempted to deploy a contract via a full node and did not encounter any node crashes. Please let me know if there are any differences in our reproduction steps.

Here are the steps I followed to reproduce the process:​

1. Start the sequencer using the attached Docker Compose configuration.
2. Launch a full node on the local network without the --rollkit.aggregator flag, and set the seed to the sequencer node.​
3. Connect MetaMask wallet to the full node's JSON-RPC port (8545).​
4. Deploy the 1_Storage.sol contract to the full node via Remix Ethereum using the "Injected Provider - MetaMask" option.​

Review the deployment logs as shown below; everything appears to be normal:

Docker Compose configuration:

# Docker Compose is used for spinning up the local-da docker container along with the GM container.

version: "3.8"
services:
  artroll:
    container_name: artroll
    # The build config can be used for debugging and testing changes to the Dockerfile
    # build: .

    # The image config can be used for running against published images
    image: artelanetwork/rollkit:latest

    # Used for targetting localhost as if you were outside the container
    ports:
      - "26656:26656"

    # Always pull the latest image
    pull_policy: always

    # The command config is used for launching the GM rollup once the Docker container is running
    command: rollkit start --rollkit.aggregator --rollkit.da_address http://local-da:7980
    depends_on:
      - local-da
    networks:
      - art_net

  local-da:
    image: ghcr.io/rollkit/local-da:v0.2.1
    container_name: local-da
    ports:
      - "7980:7980"
    networks:
      - art_net

networks:
  art_net:
    driver: bridge