You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@phoenix.apache.org by "James Taylor (JIRA)" <ji...@apache.org> on 2017/10/20 05:27:00 UTC

[jira] [Commented] (PHOENIX-4100) Simplify mutable secondary index implementation

    [ https://issues.apache.org/jira/browse/PHOENIX-4100?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16212186#comment-16212186 ] 

James Taylor commented on PHOENIX-4100:
---------------------------------------

One thought I had which would be pretty straightforward would be to use the PhoenixTransactionalIndexer for global, mutable non transactional index maintenance too. Everything would be the same, we'd just have to allow for the lack of the transaction state to be passed over.

> Simplify mutable secondary index implementation
> -----------------------------------------------
>
>                 Key: PHOENIX-4100
>                 URL: https://issues.apache.org/jira/browse/PHOENIX-4100
>             Project: Phoenix
>          Issue Type: Improvement
>            Reporter: James Taylor
>              Labels: HBase-2.0
>
> There's a lot of code for mutable secondary indexes, a lot of which is commented out (because it's not working under load). With all the latest patches, in particular PHOENIX-4089, we don't need most of it. Instead, we can do a simple scan to find the current value of the data row. We won't have mutations at the same time due to our locking and due to us timestamping the rows.
> This will get rid of many private HBase classes used by Phoenix, such as IndexMemStore, KeyValueStore, etc. and thus is likely a prerequisite to run on top of HBase 2.0.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)