hive-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Sushanth Sowmyan (JIRA)" <>
Subject [jira] [Updated] (HIVE-7072) HCatLoader only loads first region of hbase table
Date Mon, 09 Jun 2014 22:03:03 GMT


Sushanth Sowmyan updated HIVE-7072:

    Attachment: HIVE-7072.3.patch

Updated patch.

> HCatLoader only loads first region of hbase table
> -------------------------------------------------
>                 Key: HIVE-7072
>                 URL:
>             Project: Hive
>          Issue Type: Bug
>    Affects Versions: 0.14.0
>            Reporter: Sushanth Sowmyan
>            Assignee: Sushanth Sowmyan
>         Attachments: HIVE-7072.2.patch, HIVE-7072.3.patch
> Pig needs a config parameter 'pig.noSplitCombination' set to 'true' for it to be able
to read HBaseStorageHandler-based tables.
> This is done in the HBaseLoader at getSplits time, but HCatLoader does not do so, which
results in only a partial data load.
> Thus, we need one more special case definition in HCat, that sets this parameter in the
job properties if we detect that we're loading a HBaseStorageHandler based table. (Note, also,
that we should not depend directly on the HBaseStorageHandler class, and instead depend on
the name of the class, since we do not want a mvn dependency on hive-hbase-handler to be able
to compile HCatalog core, since it's conceivable that at some time, there might be a reverse
dependency.) The primary issue is one of where this code should go, since it doesn't belong
in pig (pig does not know what loader behaviour should be, and this parameter is its interface
to a loader), and doesn't belong in the HBaseStorageHandler either, since that's implementing
a HiveStorageHandler and is connecting up the two. Thus, this should belong to HCatLoader.
Setting this parameter across the board results in poor performance for HCatLoader, so it
must only be set when using with HBase.
> Thus, it belongs in the SpecialCases definition as that was created specifically for
these kinds of odd cases, and can be called from within HCatLoader.

This message was sent by Atlassian JIRA

View raw message