search

paradigmxyz / reth

Modular, contributor-friendly and blazing-fast implementation of the Ethereum protocol, in Rust
https://reth.rs/
Apache License 2.0
3.59k stars 936 forks source link

`ethereum/sync` hive failure: reth does not set FCU blocks in RPC after successful pipeline run #8579

Open Rjected opened 1 month ago

Rjected commented 1 month ago

After applying this fix in hive: https://github.com/ethereum/hive/pull/1112

We still fail the ethereum/sync test because we do not set the latest block after a successful pipeline run.

For example, the test sends a newPayload, then FCU which triggers sync using the pipeline:

2024-06-03T22:16:33.074258Z TRACE rpc::engine: Serving engine_newPayloadV2
2024-06-03T22:16:33.074924Z TRACE rpc::engine: Serving engine_forkchoiceUpdatedV2
2024-06-03T22:16:33.075108Z  INFO reth_node_events::node: Live sync in progress, downloading blocks remaining_blocks=1 target_block_hash=0x3303c6e3577579843c427f795d7485abdddc55694b5a37af0b32b48259ebe103
2024-06-03T22:16:33.075141Z  INFO reth_node_events::node: Received forkchoice updated message when syncing head_block_hash=0x07f7d3c5744ffddc5441cf31074c91a934afbf6ddb7e8234a081ee1f21140455 safe_block_hash=0x07f7d3c5744ffddc5441cf31074c91a934afbf6ddb7e8234a081ee1f21140455 finalized_block_hash=0x07f7d3c5744ffddc5441cf31074c91a934afbf6ddb7e8234a081ee1f21140455

It then checks if the node has finished syncing by calling eth_getHeaderByNumber and the latest tag:

2024-06-03T22:16:48.100023Z  INFO sync::stages::index_history: Writing indices progress=99.71%
2024-06-03T22:16:48.100112Z  INFO reth_node_events::node: Finished stage pipeline_stages=11/12 stage=IndexAccountHistory checkpoint=2000 target=2000
2024-06-03T22:16:48.101517Z  INFO reth_node_events::node: Preparing stage pipeline_stages=12/12 stage=Finish checkpoint=0 target=2000
2024-06-03T22:16:48.101532Z  INFO reth_node_events::node: Executing stage pipeline_stages=12/12 stage=Finish checkpoint=0 target=2000
2024-06-03T22:16:48.101589Z  INFO reth_node_events::node: Finished stage pipeline_stages=12/12 stage=Finish checkpoint=2000 target=2000
2024-06-03T22:16:48.147681Z TRACE rpc::eth: Serving eth_getBlockByNumber number=Latest full=false
2024-06-03T22:16:49.153164Z TRACE rpc::eth: Serving eth_getBlockByNumber number=Latest full=false
2024-06-03T22:16:50.159387Z TRACE rpc::eth: Serving eth_getBlockByNumber number=Latest full=false
2024-06-03T22:16:51.166554Z TRACE rpc::eth: Serving eth_getBlockByNumber number=Latest full=false

We return zero because we still have unset ChainInfo after the pipeline run completes.

Rjected commented 1 month ago

thinking this could be solved by fcu buffering, for example, while the fcu does kick off a pipeline run, we do "skip" the head / safe parts of the FCU in this case. So we should buffer every unique forkchoice state which we respond SYNCING to. This would then be solved because the FCU would be buffered, applied later, and the latest block would be properly set.

github-actions[bot] commented 3 weeks ago

This issue is stale because it has been open for 21 days with no activity.