Electric resume check puts a lagging tab in error when a peer deleted the row it updates

Aberta
#2,267 0 comentários 0 reações 0 responsáveis Ver no GitHub

Mantenedores costumam responder em até 1 dia

Ninguém assumiu esta issue ainda.

Avaliação

Dificuldade
4/5
Tempo estimado
3-5 dias
Facilidade para iniciantes
48/100
Tipo de issue
Bug
Clareza
Razoavelmente clara
Status de atividade
Ativa
Stack de tecnologia
sqlite, typescript
Domínio
database, testing

Direção de pesquisa

Start with the pinned replay in packages/browser-db-sqlite-persistence/tests/multi-tab-electric-oracle.test.ts and run the Vitest command from the issue to confirm the failure. Then trace the resume check and shared durable row handling; the issue points to peer commits mixing with this tab's stream state. Done means the replay passes, the resumed tab avoids error, and tabs show the server rows after catching up.

Escrita pelo modelo de indexação a partir do texto da issue.

Descrição

Summary

Electric's resume check can put a persisted Collection in error when several tabs share one persisted database. A resumed tab that lags behind a peer receives an update for a row that the peer's commit already deleted from the shared durable rows. Electric treats the update as a row it never sent and fails with Electric resume state referenced an unseen row; a full snapshot is required.

The resume check assumes that the local rows came from this tab's own stream. With shared persistence, the local rows also reflect peers' commits, which can be ahead of this tab's stream.

main has the same behavior (checked at b8d8ee07c).

Reproduction

The generated multi-tab oracle in #2265 found this history and shrank it:

  1. Two eager tabs share one database. Both deliver the initial snapshot (rows 1 and 2).
  2. Tab B reloads. Its new stream resumes from the persisted position.
  3. The server updates row 2, deletes row 1, deletes row 2, and inserts row 1 again.
  4. Tab A's stream delivers all four changes and up-to-date. A commits them, so the durable rows hold only row 1.
  5. Tab B applies A's commit, so it no longer holds row 2.
  6. Tab B's stream delivers the same changes. The first message updates row 2. B fails with the error above.

The history is pinned as an it.fails replay in packages/browser-db-sqlite-persistence/tests/multi-tab-electric-oracle.test.ts:

cd packages/browser-db-sqlite-persistence
npx vitest run tests/multi-tab-electric-oracle.test.ts -t "keeps a resumed lagging tab out of error"

Expected

A tab whose own stream is legal does not enter error. After every stream catches up, every tab shows the server's rows.

Notes

  • The generated campaigns leave this history out until a fix lands. The coverage map records the exclusion.
  • A fix probably needs the resume check to use evidence from this stream only, not rows that peers' commits supplied. #2209 (ordering source writes by server position) is related, because both come from tabs sharing durable rows at different stream positions.

This issue was written by Isaac.

Linguagem predominante
TypeScript
Estrelas
4k
Forks
286
Merge médio
1d 8h
PRs com merge (30d)
105

Preparar o ambiente

Primeiros passos

  1. Leia a issue inteira e depois o guia de contribuição do projeto.
  2. Comente na issue dizendo que vai assumir — evita que duas pessoas façam o mesmo trabalho.
  3. Faça um fork do repositório e trabalhe em uma branch.
  4. Abra um pull request que referencie o número da issue.

Mais de TanStack/db

Todas as issues de TanStack/db

Issues semelhantes

Mais issues de TypeScript

Receba novas issues na sua caixa de entrada

Um resumo curto de issues do GitHub para quem está começando.