You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "Jark Wu (Jira)" <ji...@apache.org> on 2020/05/25 05:53:00 UTC

[jira] [Commented] (FLINK-17909) Make the GenericInMemoryCatalog to hold the serialized meta data to uncover more potential bugs

    [ https://issues.apache.org/jira/browse/FLINK-17909?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17115719#comment-17115719 ] 

Jark Wu commented on FLINK-17909:
---------------------------------

What do you think about this [~twalthr], [~phoenixjiangnan]? 

> Make the GenericInMemoryCatalog to hold the serialized meta data to uncover more potential bugs
> -----------------------------------------------------------------------------------------------
>
>                 Key: FLINK-17909
>                 URL: https://issues.apache.org/jira/browse/FLINK-17909
>             Project: Flink
>          Issue Type: Improvement
>          Components: Table SQL / API, Table SQL / Planner
>            Reporter: Jark Wu
>            Priority: Major
>             Fix For: 1.11.0
>
>
> Currently, the builtin {{GenericInMemoryCatalog}} hold the meta objects in HashMap. However, this lead to many bugs when users switch to some persisted catalogs, e.g. Hive Metastore. For example, FLINK-17189, FLINK-17868, FLINK-16021. 
> That is because the builtin {{GenericInMemoryCatalog}} doesn't cover the important path of serialization and deserialization of meta data. We missed some important meta information (PK, time attributes) when serialization and deserialization which lead to bugs. 
> So I propose to hold the serialized meta data in {{GenericInMemoryCatalog}} to cover the serialization and deserializtion path. 
> We may lose some performance here, but {{GenericInMemoryCatalog}} is mostly used in demo/experiment/testing, so I think it's fine. 



--
This message was sent by Atlassian Jira
(v8.3.4#803005)