Clearly with the fix it is still dangerous for them to use LOAD INTO unless 
they understand the consistency implications or have put work arounds to 
address some reader crashes. I agree though that since this is a regression, we 
should get the functionality to what it was in 0.3

Ashish 

-----Original Message-----
From: Todd Lipcon [mailto:[email protected]] 
Sent: Saturday, September 12, 2009 3:45 PM
To: [email protected]
Subject: Re: vote for a release candidate

Hi Namit,
Yes, we have customers who are using LOAD INTO without OVERWRITE. The use case 
is for collecting session data into a table partitioned by the hour of session 
start time. Since sessions are of varying lengths, incremental loads are 
necessary as sessions finish up.

There are a couple of possible workarounds, but all of them have drawbacks.

-Todd


On Thu, Sep 10, 2009 at 6:58 PM, Namit Jain <[email protected]> wrote:

> I am not sure 718 is a valid requirement. I think it got in by legacy.
>
> Should we even support LOAD INTO ?
>
> We only support INSERT OVERWRITE,  similarly, we should only support 
> LOAD OVERWRITE INTO.
>
> Is anyone using LOAD INTO without OVERWRITE ?
>
>
>
> Thanks,
> -namit
>
>
>
>
>
> -----Original Message-----
> From: Todd Lipcon [mailto:[email protected]]
> Sent: Thursday, September 10, 2009 4:28 PM
> To: [email protected]
> Subject: Re: vote for a release candidate
>
> What do you guys think the feasibility of HIVE-718 being fixed for 
> 0.4.0 is?
> I think a completely correct solution is likely to be very tough to 
> achieve, but as is it's a regression from 0.3.0 in that the 
> functionality silently fails.
>
> -Todd
>
> On Thu, Sep 10, 2009 at 3:24 PM, Namit Jain <[email protected]> wrote:
>
> > I have created a release candidate for Hive.
> >
> > https://svn.apache.org/repos/asf/hadoop/hive/tags/release-0.4.0-rc0/
> >
> >
> > Let me know if it is OK to publish this release candidate.
> >
> >
> > Thanks,
> > -namit
> >
> >
> >
> >
> >
>

Reply via email to