You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@ignite.apache.org by "Andrey N. Gura (Jira)" <ji...@apache.org> on 2022/04/28 14:03:00 UTC

[jira] [Updated] (IGNITE-16595) Change planning approach to optimistic (in sense of data colocation)

     [ https://issues.apache.org/jira/browse/IGNITE-16595?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Andrey N. Gura updated IGNITE-16595:
------------------------------------
    Epic Link:   (was: IGNITE-16603)

> Change planning approach to optimistic (in sense of data colocation)
> --------------------------------------------------------------------
>
>                 Key: IGNITE-16595
>                 URL: https://issues.apache.org/jira/browse/IGNITE-16595
>             Project: Ignite
>          Issue Type: Improvement
>            Reporter: Yury Gerzhedovich
>            Priority: Major
>              Labels: ignite-3
>         Attachments: bugreport-blackjack-QODS30.163-7-27-2022-03-13-13-48-43-dumpstate_log-17151.txt
>
>
> Because of the nature of the Raft, we can't guarantee the leaders won't be changed in between the mapping of the query to particular nodes and its actual execution. So the pessimistic approach was chosen where table has no determined distribution.
> However, the change of leader is actually a deterministic process, not a random one. With that said, let's switch to the optimistic approach and start to respect the actual distribution of the tables. Also it would be great to log every false-positive colocation



--
This message was sent by Atlassian Jira
(v8.20.7#820007)