[jira] [Resolved] (OAK-5863) SegmentNodeBuilder getNodeState can trigger eager flush of child node builder

["Alex Parvulescu (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/OAK-5863?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Alex Parvulescu resolved OAK-5863.
----------------------------------
       Resolution: Fixed
         Assignee: Alex Parvulescu
    Fix Version/s: 1.7.0

impact seems very low, the condition to reproduce here is to have 10k changes on that specific nodebuilder which is quite a lot.

fixed with http://svn.apache.org/viewvc?rev=1784953&view=rev

> SegmentNodeBuilder getNodeState can trigger eager flush of child node builder
> -----------------------------------------------------------------------------
>
>                 Key: OAK-5863
>                 URL: https://issues.apache.org/jira/browse/OAK-5863
>             Project: Jackrabbit Oak
>          Issue Type: Bug
>          Components: segment-tar
>            Reporter: Alex Parvulescu
>            Assignee: Alex Parvulescu
>             Fix For: 1.7.0, 1.8
>
>
> It seems that calling {{getNodeState}} on a {{SegmentNodeBuilder}} that is not root can reset the {{updateCount}} even if the builder is not root, and will start eagerly flushing the state to disk once the update limit is reached. [0]
> The fix is relatively simple, don't reset the {{updateCount}} to {{0}} if the builder is not a root. I'm now trying to determine the impact on write throughput and on the size of the repository, as this is not an easy situation to reproduce. It needs to have a very large transaction happening, and a call to {{getNodeState}} down the tree.
> [0] https://github.com/apache/jackrabbit-oak/blob/trunk/oak-segment-tar/src/main/java/org/apache/jackrabbit/oak/segment/SegmentNodeBuilder.java#L114



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)