Measure the payload copy in the engine request, and share the block if it shows

未關閉
#819 0 則留言 0 個 reaction 已指派 0 人 在 GitHub 檢視

維護者通常 6 天內回覆

還沒有人認領這個 Issue。

評估

難度
4/5
預估耗時
3-5 天
新手友好度
48/100
Issue 類型
功能
描述清晰度
基本清楚
活躍度
活躍
技術堆疊
rust
領域
performance

研究方向

Read crates/blockchain/src/beacon_engine.rs and the changes in lambdaclass/ethlambda#755 to understand NewPayloadRequest::into_owned and the engine import phase metric. Measure request construction on a live node with an execution client for typical and transaction-heavy blocks, comparing it with the roughly 157 ms transition time cited in the issue. If the copy is material, apply the outlined Arc ownership approach; done means the measurement is reported and any resulting change avoids the actor-thread copy while preserving the block for on_block_speculative.

由索引模型根據 Issue 內容生成。

描述

Since lambdaclass/ethlambda#755, a pre-gloas beacon import with an execution client sends engine_newPayloadV4 from a spawned task while the chain actor computes the block's state transition. The task cannot borrow the block, so NewPayloadRequest::into_owned (crates/blockchain/src/beacon_engine.rs) clones the block's ExecutionPayload, transactions included, on the actor thread before the request is spawned. Three automated reviewers raised the copy across two review rounds of that PR.

Its cost has not been measured, neither typically nor for a transaction-heavy worst case, so it is not known how it compares with the transition it overlaps (~157 ms at the docs' p50). Since #755 the engine import phase (lean_block_import_phase_seconds{phase="engine"}) includes building the request, so the copy is visible there on a live node: the phase's floor, for blocks whose answer arrived before the transition finished, is roughly the request construction.

Measuring it was out of scope for #755, since it needs a live node with an execution client.

What a fix would look like, if the floor is material:

  • Move the block into an Arc for the duration of the question.
  • Have the task build the request from its clone, so nothing is copied on the actor.
  • Take the block back with Arc::unwrap_or_clone before on_block_speculative. The task has finished by then, so this does not clone in practice.

OwnedNewPayloadRequest and into_owned would then go away.

主要語言
Rust
星號
86
分支
34
平均合併
9 天 20 小時
30 天內合併 PR
77

環境準備

從這裡開始

  1. 先讀完整個 Issue,再讀專案的貢獻指南。
  2. 在 Issue 下留言說明你要接手 —— 這能避免兩個人做同樣的事。
  3. Fork 儲存庫,在一個分支上完成修改。
  4. 送出 Pull Request,並在描述裡引用這個 Issue 編號。

lambdaclass/ethlambda 的其他 Issue

查看 lambdaclass/ethlambda 的全部 Issue

相似的 Issue

更多 Rust Issue

把新 issue 寄到你的電子郵件信箱

精選適合新手參與的 GitHub issue 摘要。