beacon: raise the custody group count for attached validators (fulu validator custody)

オープン
#848 コメント 0 件 リアクション 0 件 担当者 0 名 GitHub で見る

メンテナーはふだん 6 日以内に返信

まだ誰も着手していません。

評価

難易度
4/5
見積もり時間
3〜5日
初心者へのやさしさ
35/100
issue の種類
機能追加
明瞭さ
おおむね明確
活発さ
活発
技術スタック
rust

調査の方向性

Start with crates/net/rpc/src/beacon/validator.rs to trace validator indices from beacon_committee_subscriptions and prepare_beacon_proposer. Then read bin/ethlambda/src/beacon.rs and the custody paths listed in the issue to see how the startup count is used. Done means attached validators affect the custody count and the count is advertised consistently; the issue describes additional persistence, runtime update, and backfill work.

索引モデルが issue の本文から書いたものです。

説明

Problem

ethlambda beacon does not implement fulu validator custody. It custodies and advertises whatever --custody-group-count says (default CUSTODY_REQUIREMENT = 4), regardless of how many validators are attached to it.

Fulu validator.md ("Validator custody") says a node with validators attached custodies get_validators_custody_requirement(state, validator_indices) groups. That is the attached validators' total effective balance divided by BALANCE_PER_ADDITIONAL_CUSTODY_GROUP (32 ETH), with a floor of VALIDATOR_CUSTODY_REQUIREMENT (8) and a cap of NUMBER_OF_CUSTODY_GROUPS (128). A node with 4096 ETH or more attached must therefore be a supernode.

Observed in a kurtosis test enclave: with 128 validators attached, ethlambda logged custody_group_count=4, while Lighthouse in the same network reported 128. A network laid out on the assumption that every validating node stores at least 8 groups is short on column custody wherever ethlambda runs the validators.

What the code does today (at 0ea6e214)

What the spec asks for (fulu validator.md)

  • Advertise the higher count in MetaData custody_group_count and in the ENR's cgc.
  • Adjust it dynamically, without operator input, as the attached validators' effective balances change.
  • On an increase, advertise the new count immediately. Backfill MAY follow. earliest_available_slot in Status v2 should reflect what the node can actually serve, and may move lower as backfill progresses.
  • On a decrease, keep custodying and advertising the previous (highest) count, and persist it across restarts.

Rough scope

  1. Track attached validator indices from validator client calls (beacon_committee_subscriptions, prepare_beacon_proposer), with an expiry so that a detached validator client stops counting.
  2. Once per epoch, compute get_validators_custody_requirement and take the larger of the flag, that requirement, and the persisted highest count. Store the highest count in the data directory.
  3. Make the custody set changeable at runtime. Today it is fixed in BeaconWire and in the chain actor at spawn, and these paths read it: data-column subnet subscriptions, the availability check, req/resp serving, and the engine's custody columns.
  4. When the count changes, bump the MetaData seq_number and the ENR. The ENR has the same known gap as the fork digest: ethrex's DiscoveryServer cannot replace the record it serves at runtime, so cgc would stay at its startup value until a restart.
  5. Make MetaData v3 report the real count, and add custody_group_count to /eth/v1/node/identity.
  6. After an increase, keep earliest_available_slot honest until backfill completes.

Steps 5 and 1 are small and independent, so they could land first.

主要言語
Rust
スター
86
フォーク
34
平均マージ
9日 20時間
マージ済み PR(30日)
77

環境構築

はじめの一歩

  1. issue を最後まで読み、次にプロジェクトのコントリビューションガイドを読みます。
  2. 着手することを issue にコメントします — 二人が同じ作業をするのを防げます。
  3. リポジトリをフォークし、ブランチを切って変更します。
  4. issue 番号を参照したプルリクエストを送ります。

lambdaclass/ethlambda のほかの issue

lambdaclass/ethlambda の issue をすべて見る

似ている issue

Rust の issue をもっと見る

新しい issue をメールで受け取る

初心者向けの GitHub issue を短くまとめたダイジェスト。