[
https://issues.apache.org/jira/browse/HIVE-28621?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17897143#comment-17897143
]
yongzhi.shao edited comment on HIVE-28621 at 11/11/24 10:07 AM:
----------------------------------------------------------------
[~zhangbutao] [~dkuzmenko] [~samuel281] [~seonggon]
Hello,Sir. What do you think?
Currently, our initial implementation of fs-catalog can maintain at least the
same functionality as jdbc-catalog.
It can provide safe concurrent operations beyond the scenarios faced by
HIVE-28366.
(The issues faced by HIVE-28366 are indeed quite challenging; we are
researching solutions to address this problem without using read-write locks.)
was (Author: lisoda):
[~zhangbutao] [~dkuzmenko] [~samuel281] [~seonggon]
Hello,Sir. What do you think?
Currently, our initial implementation of fs-catalog can maintain at least the
same functionality as jdbc-catalog.
It can provide safe concurrent operations beyond the scenarios faced by
HIVE-28366.
> Re implement a catalog based on file system for metadata management
> -------------------------------------------------------------------
>
> Key: HIVE-28621
> URL: https://issues.apache.org/jira/browse/HIVE-28621
> Project: Hive
> Issue Type: New Feature
> Security Level: Public(Viewable by anyone)
> Components: Iceberg integration
> Affects Versions: 4.1.0
> Reporter: yongzhi.shao
> Priority: Major
>
> Currently, the iceberg community has deprecated hadoop_catalog in version
> 1.7.0 and plans to remove it in future versions. However, the catalog
> management solution based on the file system is actually very user-friendly.
> It is lightweight, allowing users to minimize the dependencies required for
> startup. Moreover, it is also very friendly for cross-engine operations on
> Iceberg tables.
> We have implemented a version of catalog management based on the file system
> that can manage catalogs without relying on any third-party services and
> without requiring the file system to provide rename operations. We have
> verified its functionality in various file systems such as
> HDFS/S3/LOCAL-POSIX/LOCAL-LINUX/WINDOWS, and it works well.
> If we want to contribute it to the HIVE project, what should I do?
> Additionally, I would like to know what the HIVE community's opinion is on
> FS-CATALOG?
> Tks.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
