[jira] [Commented] (SENTRY-1101) When edit log for HDFS sync in Sentry Server is full, the next Path update is not correctly setup

["Sravya Tirukkovalur (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/SENTRY-1101?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15225049#comment-15225049 ] 

Sravya Tirukkovalur commented on SENTRY-1101:
---------------------------------------------

public void handleUpdateNotification(final K update) throws SentryPluginException {
   ...
            updateable.updatePartial(Collections.singletonList(update), lock);
        appendToUpdateLog(toUpdate);
}

If you see the above code snippet, we are updating the updatable, which in turn adds the path change to HMSPaths, before adding it to the updatelog. So, when full image is built inside appendToUpdateLog, full image is basically a dump of HMSPaths, so I think we should be good. Unless there is another bug:-). Might be good to write a deterministic test to reproduce this for the edit_log_size + 1th edit by loweing the size to something low as 5?

> When edit log for HDFS sync in Sentry Server is full, the next Path update is not correctly setup
> -------------------------------------------------------------------------------------------------
>
>                 Key: SENTRY-1101
>                 URL: https://issues.apache.org/jira/browse/SENTRY-1101
>             Project: Sentry
>          Issue Type: Bug
>          Components: Sentry
>            Reporter: Hao Hao
>            Assignee: Hao Hao
>             Fix For: 1.7.0
>
>         Attachments: SENTRY-1101.0.patch
>
>
> The edit log for caching the path updates is not updated properly once the cache is full in Sentry Server side. In that case, even though the path updates will be send to NN plugin, the path updates itself is not containing the correct paths. It can cause unexpected ACL behavior issues during table creation. 



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)