[jira] [Commented] (OAK-9358) DocumentNodeStore may accumulate split candidates

["Marcel Reutegger (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/OAK-9358?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17285228#comment-17285228 ] 

Marcel Reutegger commented on OAK-9358:
---------------------------------------

Added ignored tests to trunk: http://svn.apache.org/r1886595

> DocumentNodeStore may accumulate split candidates
> -------------------------------------------------
>
>                 Key: OAK-9358
>                 URL: https://issues.apache.org/jira/browse/OAK-9358
>             Project: Jackrabbit Oak
>          Issue Type: Bug
>          Components: documentmk
>    Affects Versions: 1.38.0
>            Reporter: Marcel Reutegger
>            Assignee: Marcel Reutegger
>            Priority: Critical
>             Fix For: 1.40.0
>
>
> This is a regression introduced by OAK-9149. The split candidates in DocumentNodeStore may contain document IDs that actually don't need to be split. {{Commit#checkSplitCandidate()}} does a very quick check based on the size of the document and then leaves it to the background update to perform more thorough check and perform the split if necessary. As of OAK-9149, an ID for a document that does not need to be split in the background update is not removed from the split candidates and the background update checks it in every cycle. It is also possible that distinct split candidates accumulate over time and slow down the background operation significantly.



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