Hacktoberfest 2026：メンテナが10月に向けて印を付けた、オープンで初心者向けの issue。 Hacktoberfest の issue を見る

Cassandra: four catalog and queue metadata reads on the data path run at LOCAL_ONE and feed a write or a cleanup

オープン
  • 何か問題がありますか？
#592 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 2 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
54/100
issue の種類
バグ
明瞭さ
おおむね明確
活発さ
活発
技術スタック
cassandra, rust
領域
backend, databases

調査の方向性

Start with the four reads in crates/storage-cassandra/src/data/index.rs and data/ttl.rs, then trace ttl_config_for_table in metadata_engine.rs and the catalog writes in create_table.rs and update_table.rs. Check how Cassandra consistency levels are set and which writes each read must intersect; the issue identifies the affected operations and failure cases. Done means those reads cannot miss the catalog or queue state they use for writes and cleanup.

索引モデルが issue の本文から書いたものです。

説明

backend:cassandra

Summary

Four reads of catalog and queue metadata on the Cassandra data path run at LOCAL_ONE and feed a write or a cleanup. Each can miss a row written moments earlier at LOCAL_QUORUM on a keyspace with a replication factor above 1. None changes what a base-table GetItem, Query or Scan with ConsistentRead: true returns, so they were left out of #560 (fixed in #591).

The four reads

Read File Feeds A missed row means
fetch_indexes_for_table on the put, update and delete paths crates/storage-cassandra/src/data/index.rs (fetch_indexes_for_table_at(.., false)) the set of GSI rows a write maintains a write after an UpdateTable that added a GSI writes no index row for the item; if the backfill already passed its range, the GSI never holds the item
the gsi_pending last_ready_at read data/index.rs the clamp ready_at_ms = max(jittered, last_ready_at + 1) that orders async GSI mutations a later mutation of the same worker partition can take an earlier ready_at, so the GSI can end on the older image
delete_indexes_for_table data/index.rs the index ids whose data tables DeleteTable drops an index's data table is left undropped after its catalog rows are deleted
the TTL bucket read in clear_ttl_entries data/ttl.rs the queue partitions deleted when TTL is disabled queue rows left behind after the registration is removed

The write paths also read the TTL configuration through ttl_config_for_table (metadata_engine.rs), which runs through cassandra_util::query_rows at the driver default outside data/; a write that misses a TTL just enabled inserts no expiration entry.

Expected

The gsi_pending read decides a write and the rows it reads are written at LOCAL_QUORUM inside data/, so passing true there closes it in one argument. The other three read catalog rows that CreateTable and UpdateTable write at the driver default outside data/, so a quorum read alone does not intersect those writes; the catalog writes need a stated level first.

Where

  • crates/storage-cassandra/src/data/index.rs
  • crates/storage-cassandra/src/data/ttl.rs
  • crates/storage-cassandra/src/metadata_engine.rs (ttl_config_for_table)
  • crates/storage-cassandra/src/create_table.rs, update_table.rs (the catalog writes)
主要言語
Rust
スター
517
フォーク
60
平均マージ
2日 22時間
マージ済み PR（30日）
25

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。

ExtendDB/extenddb のほかの issue

ExtendDB/extenddb 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。