Archive of non-existent streams
维护者通常 7 天内回复
还没有人认领这个 Issue。
评估
调研方向
The issue describes archive and hot store behavior in Eventuous, with EventStoreDB (ESDB) and Elastic as the stores; it names no files or tests. Start by reading the archive loading and appending paths, then trace how unarchiving would interact with subscriptions and the archive connector. Done means choosing and implementing a strategy that preserves optimistic concurrency without duplicate archive reads or unwanted subscription effects.
由索引模型根据 Issue 内容生成。
描述
Aiming to solve the following scenario:
- We are importing data from a legacy system directly to the events archive (Elastic)
- There's one event per aggregate in the archive, its version is zero
- When executing a command on an aggregate, events can only be found in the archive
- When appending the event to the hot store (ESDB), we provide the expected version = 1
- The append fails as ESDB doesn't have the stream
It doesn't work as the archive use case expected all the events to be first appended to the hot store (ESDB), and then archived rea time. The hot store stream gets truncated by time or size, so the stream version is kept although events are gone.
One thing we tried is to put the expected version to the meta. However, it effectively disables the native optimistic concurrency checks provided by ESDB or any other store that keeps the event number per stream as a unique constraint. In that case, this complexity moves to Eventuous code, which is not desirable.
Proposed solutions:
- Append dummy events to the imported streams when importing to the archive, then delete those streams. Pro: everything works as it should, no code changes are required in Eventuous. Cons: ESDB data set will include metadata for all the deleted streams. We are talking about hundreds of millions of streams, and most of them will never get new events, so having all those streams in ESDB is inefficient.
- When loading the aggregate, read events from the archive (as it works today) and append them to ESDB right after. Basically, it's an "unarchive on load" feature. Pros: little code change needed, everything will work as it should. Cons: there's no guarantee that any append will happen after load. It would work quite ok if the aggregate state is only loaded for executing commands, and not for queries.
- Unarchive the stream when appending new events. Pros: events will be appended to the hot store if the command execution produced new events. Cons: more code changes are required, as the aggregate needs to keep the collection of original events for unarchiving purposes.
Both unarchiving strategies have two potential issues:
- Unarchived events will get to all the subscriptions, potentially resulting in undesired side effects
- The connector will try to archive those events too, which will cause data duplication, and the next read will get the archived events twice (unless it fails with a duplicate message-id during the archive)
Potential solutions:
- Force delete stream after unarchive
- Annotate specific (import) events so they aren't archived. With this, we'll have a more generic function to prevent some events from being archived at all, which might be useful (but dangerous). It will also spread out to the connector.
- 主要语言
- C#
- 星标
- 512
- 派生
- 99
- 平均合并
- 5 天 6 小时
- 30 天内合并 PR
- 5
环境准备
- 提供 Dockerfile 或 Docker Compose 文件
- 没有 Pull Request 模板
- 阅读贡献指南
从这里开始
- 先读完整个 Issue，再读项目的贡献指南。
- 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
- Fork 仓库，在一个分支上完成修改。
- 提交 Pull Request，并在描述里引用这个 Issue 编号。
Eventuous/eventuous 的其他 Issue
-
MongoDB integration tests fail on newer GitHub Actions runner kernels可能已有人在做 @nmummau 于 1 天前认领。 未关闭
难度 2/5 1-3 小时 新手友好度 74/100
Eventuous/eventuous#626 · 2 条评论 ·
维护者通常 7 天内回复
-
`dotnet format style --verify-no-changes` fails with naming violations and unnecessary imports可能已有人在做 @nmummau 于 1 天前认领。 未关闭
难度 2/5 1-3 小时 新手友好度 80/100
Eventuous/eventuous#624 · 1 条评论 ·
维护者通常 7 天内回复
-
难度 5/5 一周以上 新手友好度 35/100
Eventuous/eventuous#627 · 1 条评论 ·
维护者通常 7 天内回复
-
SQL Server: `append_events` does not need the `@inserted` table variable可能已有人在做 @nmummau 于 7 天前认领。 未关闭
难度 3/5 1-2 天 新手友好度 20/100
Eventuous/eventuous#619 · 1 条评论 ·
维护者通常 7 天内回复
-
SQL Server: stream reads do one key lookup per event, and `IDX_EventsStream` duplicates `UQ_StreamIdAndStreamPosition`可能已有人在做 @nmummau 于 7 天前认领。 未关闭
难度 3/5 1-2 天 新手友好度 18/100
Eventuous/eventuous#618 · 1 条评论 ·
维护者通常 7 天内回复
查看 Eventuous/eventuous 的全部 Issue
相似的 Issue
-
/api/oauth/introspect emits clientId and tokenType instead of RFC 7662's client_id and token_type未关闭bug P2
难度 2/5 1-3 小时 新手友好度 78/100
nightscout/nocturne#2317 ·
维护者通常 1 天内回复
-
[BUG]: Position intent set on a simple order is dropped when it becomes a bracket, OCO or OTO order可能已有人在做 @RaphaelFakhri 今天认领。 未关闭
难度 2/5 1-3 小时 新手友好度 78/100
alpacahq/alpaca-trade-api-csharp#903 ·
维护者通常 2 天内回复
-
难度 2/5 1-3 小时 新手友好度 64/100
joinrpg/joinrpg-net#5752 ·
维护者通常 1 天内回复
-
bug ux
难度 2/5 1-3 小时 新手友好度 78/100
CodeForFire/lagebuch#714 ·
维护者通常 1 天内回复
-
难度 2/5 1-3 小时 新手友好度 82/100
dimitar-radenkov/Pointframe#294 ·
维护者通常 1 天内回复