Accessing Universe/hashfiles using odbc stage in PX jobs

rkacham_DSX · Post by **rkacham_DSX** » Tue Jul 25, 2006 9:53 am

Hi ,

I am trying to access universe /hashfiles using odbc stge in parallel jobs,
i am able to connect to unverse using odbc stage from my server jobs,
when i try this in parallel jobs i'm not able to connect to universe using odbc..

i was wondering do we need to change any entries in .odbc.ini files or any other env file to access univers using odbc in parallel jobs..

thanks in advace...

kcbland · Post by **kcbland** » Tue Jul 25, 2006 10:14 am

You should NOT use ODBC to connect to Universe. If you understand Universe technology, you would know that support for multi-value attributes exists within the Universe UV/ODBC stage.

Even then, your efforts are not sufficient for connecting to Universe. You will need to setup clean dictionaries with properly defined multivalue associations and their phrases so that the multi-value attributes will be properly normalized. You also need to avoid I-descriptors, F and T correlatives that span to other files, etc.

rkacham_DSX · Post by **rkacham_DSX** » Tue Jul 25, 2006 10:46 am

In IBM Datstaage Advanced devloper calss, they said the only way to access the hash fiels in parllel jobs is using odbc stage...

is there any other way to access hashfiles in parallel jobs?

we have project in server jobs further devlopement to this project we are trying to do in parallel jobs so we need to access those hashfiles in parllel jobs

kcbland · Post by **kcbland** » Tue Jul 25, 2006 10:56 am

Universe is something else entirely. If you need to reference DataStage hashed files in a PX job, then you should consider using a container and a hashed file stage reading the hashed file and then move it into a dataset of some kind. Using the ODBC stage to read a hashed file will work if you simply need to access the physical file, but that's a messy way to do it. You need to constrain yourself to operating on the node that has the hashed file, often the Server node. It's just an awful way to do it.

Use a hashed file stage to read hashed files, makes sense don't it?

ray.wurlod · Post by **ray.wurlod** » Tue Jul 25, 2006 3:23 pm

You can also use External Source or External Target stages if the number of records is small, using the somewhat less efficient UVread and UVwrite executables. Note that these stages will need to be constrained to execute in a node pool all of whose nodes are on the same machine as the DataStage server.