Event bus: build high-rate topics only for clients that subscribed to them

Aperta
#845 0 commenti 0 reazioni 0 assegnatari Vedi su GitHub

I maintainer di solito rispondono entro 6 giorni

Nessuno ha ancora preso questa issue.

Valutazione

Difficoltà
4/5
Tempo stimato
3-5 giorni
Idoneità per principianti
42/100
Tipo di issue
Funzionalità
Chiarezza
Abbastanza chiara
Stato di attività
Attiva
Stack tecnologico
rust

Direzione di ricerca

Start with crates/blockchain/src/events.rs and crates/net/rpc/src/events.rs, then trace the producer gates and sink trait in ethlambda-network-api; the issue describes two possible designs. Run or add a test where a client subscribes only to head. Done when that subscription does not build single_attestation, and the relevant high-rate topics are handled by the chosen design.

Scritto dal modello di indicizzazione a partire dal testo della issue.

Descrizione

Problem. EventBus::has_subscribers is receiver_count() > 0 (crates/blockchain/src/events.rs:644). Every producer gates on it, so a single SSE client turns on every topic's payload construction, whatever ?topics= it asked for: a validator client that only wants head, for example. The filtering happens afterwards, in the stream (crates/net/rpc/src/events.rs, !topics.contains(&ev.topic())).

For the two high-rate beacon topics that means:

  • single_attestation (lambdaclass/ethlambda#757) clones every subnet attestation the node accepts: every vote on its backbone subnets, plus the duty subnets it joins.
  • attestation clones every accepted aggregate.

Both go into the one ring all topics share (CHAIN_EVENT_CHANNEL_CAPACITY, 8192, events.rs:600). A client whose connection stalls lags after capacity / total rate, and loses the low-rate events it did ask for, which it learns only from the dropped N messages comment frame. The doc on CHAIN_EVENT_CHANNEL_CAPACITY already names this and calls a per-topic split "the escape hatch". The rates on a mainnet node are not measured; lean_-prefixed counters per topic emitted would measure them.

Why not in lambdaclass/ethlambda#757. 757 adds one producer and follows the bus's existing gate, as attestation already does. Per-topic interest changes the bus, the SSE handler and every producer's gate.

Fix. Either of two:

  • Count subscriptions per Topic. The SSE handler registers its topic set with a guard that decrements on drop, and EventBus::wants(Topic) replaces has_subscribers at each producer; the sink trait in ethlambda-network-api grows the same method.
  • Keep has_subscribers, but give single_attestation and attestation a ring of their own, so their rate cannot push low-rate events out of a slow client's window.

A test: a client subscribed to head alone causes no single_attestation to be built.

Also high-rate (from lambdaclass/ethlambda#748): the gloas payload_attestation_message and execution_payload_bid events are built per message too, so the same per-topic interest check should cover them.

Lingua principale
Rust
Stelle
86
Fork
34
Merge medio
9g 20h
PR unite (30g)
77

Preparare l'ambiente

Come iniziare

  1. Leggi tutta la issue e poi la guida ai contributi del progetto.
  2. Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
  3. Fai un fork del repository e lavora su un branch.
  4. Apri una pull request che faccia riferimento al numero della issue.

Altre issue di lambdaclass/ethlambda

Tutte le issue di lambdaclass/ethlambda

Issue simili

Altre issue su Rust

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.