You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@drill.apache.org by "Aman Sinha (JIRA)" <ji...@apache.org> on 2014/11/07 16:17:33 UTC

[jira] [Updated] (DRILL-1601) Ensure a minimum allocation for variable length vectors and improve error messages

     [ https://issues.apache.org/jira/browse/DRILL-1601?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Aman Sinha updated DRILL-1601:
------------------------------
    Assignee: DrillCommitter  (was: Aman Sinha)

> Ensure a minimum allocation for variable length vectors and improve error messages
> ----------------------------------------------------------------------------------
>
>                 Key: DRILL-1601
>                 URL: https://issues.apache.org/jira/browse/DRILL-1601
>             Project: Apache Drill
>          Issue Type: Bug
>          Components: Execution - Flow
>    Affects Versions: 0.6.0
>            Reporter: Aman Sinha
>            Assignee: DrillCommitter
>             Fix For: 0.7.0
>
>         Attachments: 0001-DRILL-1601-Have-a-minimum-allocation-for-variable-le.patch
>
>
> The allocateNew() and allocateNewSafe() methods for value vectors compute the estimated allocation size based on previous requests.   There could be situations where the allocator determines a very small allocation size based on previous pattern (e.g a bunch of null values) and then encounters a large enough record size.  In such cases, we see the following types of exception: 
>   The Partitioner sometimes encounters the following exception:
> java.io.IOException: Record {} is too big to fit into the allocated memory of ValueVector.
> This issue is mainly applicable for variable length vectors since it uses bytes for the estimation whereas the fixed length vectors use the number of values for estimation. 
> A second part of this JIRA is to improve error messages in the Partitioner and UnlimitedRawBatchBuffer. 
>         



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