Thanks Edward, I actually added some notes to that JIRA issue. Yes DataNucleus should work, but when debug logging is enabled I can see that Hive looks for the existence of a table called COLUMNS. By default SQL server already has a COLUMNS View and therefore the check returns true and the auto table configuration then fails on the next alter table statement. I think DataNucleus should be OK what i need to do is to tell Hive to use a different table name e.g. HIVE_COLUMNS instead of COLUMNS. Is this possible?
Cheers, Andy. On 24 March 2011 16:26, Edward Capriolo <edlinuxg...@gmail.com> wrote: > On Thu, Mar 24, 2011 at 11:36 AM, shared mailinglists > <shared.mailingli...@gmail.com> wrote: > > Thanks Bernie, hopefully they will. > > > > Were a small Java development team within a predominately MS development > > house. We’re hopefully introducing new ideas but the normal company > politics > > dictate that we should use SQL Server. That way maintenance, backup, > recover > > etc etc can be handed over to the internal MS db team while freeing us > guys > > to concentrate on better things like Hadoop & Hive :-) I assumed with the > DB > > just being a metadata store that the database wouldn’t be an issue but > were > > struggling a bit:-( > > > > On 24 March 2011 15:23, Bennie Schut <bsc...@ebuddy.com> wrote: > >> > >> Sorry to become a bit offtopic but how do you get into a situation where > >> sqlserver 2005 becomes a requirement for a hive internal meta store? > >> > >> I doubt many of the developers of hive will have access to this database > >> so I don't expect a lot of response on this. But hopefully someone can > prove > >> me wrong :) > >> > >> Bennie. > >> > >> > >> On 03/24/2011 04:01 PM, shared mailinglists wrote: > >>> > >>> Hi Hive users :-) > >>> > >>> Does anybody have experience of using Hive with MS SQL Server 2005? I’m > >>> currently stumped with the following issue > >>> https://issues.apache.org/jira/browse/HIVE-1391 where Hive (or > DataNucleus?) > >>> confuses the COLUMNS table it requires internally with that of the > default > >>> SQL Server sys.COLUMNS or information_schema.COLUMNS View and therefore > does > >>> not automatically create the required metadata table when running the > Hive > >>> CLI. > >>> > >>> > >>> Has anybody managed to get Hive to work with SQLServer 2005 or know how > I > >>> can configure Hive to use a different table name to COLUMNS ? > Unfortunately > >>> we have to use SQL Server and do not have the option to use Derby or > MySQL > >>> etc. > >>> > >>> Many thanks, > >>> > >>> > >>> Andy. > >>> > >> > > > > > > Let us not forget that M$ SQL Server is very advanced. It has for a > long time supported many types of things that mysql just plain did > not. (Did we all forget then mysql 3.X days where we had no > Transactions or Foreign keys? :) > > There was one ticket I closed on it. > https://issues.apache.org/jira/browse/HIVE-1391 > > As far as hive is concerned, m$ SQL server is JPOX/Data Nucleus > supported so it "should" work. How many deployments exist in the wild > are unknown. >
