You are viewing a plain text version of this content. The canonical link for it is here.

Posted to user@hive.apache.org by james warren <ja...@rockyou.com> on 2009/09/09 07:44:35 UTC

possible INSERT OVERWRITE TABLE bug

I encountered this behavior twice today, but before filing a JIRA I thought
I'd seek confirmation this is indeed a bug and not a potential configuration
issue.
While running queries of the form

INSERT OVERWRITE TABLE 'hdfs://server:port/foo/bar/out" SELECT...


where the "foo" and "bar" directories don't exist on HDFS, the job runs
without error but the "/foo/bar/" path is not created (and hence the output
is not written to HDFS).  After manually creating the path, the output is
indeed written correctly.

Regarding our configuration, we're running the Cloudera 0.18.3 installation.

cheers,
-James Warren

Re: possible INSERT OVERWRITE TABLE bug

Posted by Prasad Chakka <pc...@facebook.com>.

The latest version does create all the parents if they do not exist. Can you try with the latest version (or 0.4)?

Prasad

________________________________
From: james warren <ja...@rockyou.com>
Reply-To: <hi...@hadoop.apache.org>
Date: Tue, 8 Sep 2009 22:44:35 -0700
To: <hi...@hadoop.apache.org>
Subject: possible INSERT OVERWRITE TABLE bug

I encountered this behavior twice today, but before filing a JIRA I thought I'd seek confirmation this is indeed a bug and not a potential configuration issue.

While running queries of the form

INSERT OVERWRITE TABLE 'hdfs://server:port/foo/bar/out" SELECT...

where the "foo" and "bar" directories don't exist on HDFS, the job runs without error but the "/foo/bar/" path is not created (and hence the output is not written to HDFS).  After manually creating the path, the output is indeed written correctly.

Regarding our configuration, we're running the Cloudera 0.18.3 installation.

cheers,
-James Warren