Orphaned FTS index generations accumulate in episode.lance/_indices/ across restarts

未关闭
#558 1 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

维护者通常 1 天内回复

还没有人认领这个 Issue。

评估

难度
4/5
预计耗时
3-5 天
新手友好度
25/100
Issue 类型
缺陷
描述清晰度
基本清楚
活跃度
活跃
技术栈
python
领域
databases

调研方向

The bug sits in how the episode.lance index directory is managed across rebuilds, so start with the everos cascade rebuild path and the startup sweep that triggers it. Check where the old index generations get dropped or replaced, and whether optimize/cleanup is called afterwards. Done means only the live generation remains in _indices/ after a rebuild, confirmed by listing that directory before and after.

由索引模型根据 Issue 内容生成。

描述

everos 1.5.1, lancedb 0.34.0 (pinned), Python 3.12 via uv tool install everos, Linux x86_64.

Observed

After a handful of server restarts plus an everos cascade rebuild, episode.lance/_indices/ accumulates multiple index-UUID directories (6 observed) while table.list_indices() reports a single episode_tokens_idx. Optimize/prune does not reclaim the stale generations, so the index directory grows with each rebuild cycle.

Repro

  1. Start/stop the server a few times (everos server start, systemd restart, etc.)
  2. Run everos cascade rebuild (or let the startup sweep rebuild)
  3. ls ~/.everos/.index/lancedb/episode.lance/_indices/ → one directory per historical generation

Expected

Dropped/replaced index generations are garbage-collected after rebuild/optimize, leaving only the live generation in _indices/.

Notes

Search correctness was never affected in our testing (BM25 via the API resolves the live index fine from fresh connections); this is disk-hygiene / index-UUID-count growth only. Possibly related to the index-merge gap tracked in lancedb/lancedb#2193, lancedb/lancedb#3177 and lance-format/lance#6711.

主要语言
Python
星标
13.3k
派生
926
平均合并
19 小时 29 分钟
30 天内合并 PR
43

环境准备

这个项目没有提供开发容器、Dockerfile 或贡献指南,环境需要你自己搭建:先看它的 README,通用步骤见我们的新手贡献指南。

从这里开始

  1. 先读完整个 Issue,再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库,在一个分支上完成修改。
  4. 提交 Pull Request,并在描述里引用这个 Issue 编号。

EverMind-AI/EverOS 的其他 Issue

查看 EverMind-AI/EverOS 的全部 Issue

相似的 Issue

更多 Python Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。