Database connections in datastage

[skathaitrooney]

Hello Experts,

Just had a query.

Suppose my jobs runs on a head node and a compute node.

Now if i am using a database in the job, will the network sockets for the database be opened from the head node or the compute node?

Currently i am seeing sockets opened for both head node and compute node.

i.e. from where exactly are the database connections made?

[electajay]

If i understand correctly. you are trying to connect to database ex: oracle from Datastage server A (head Node) server B (Compute node). yes you need to open the firewall and ports from both the Servers A and B and update the tnsnames.ora file on both the Servers A and B.

I think Database connections are made from Conductor node (Head Node)
lets see what others say.

[AshishDevassy]

from what i remember , connections are made thru the head nodes.

[skathaitrooney]

Ashsish/electajay,

Let me explain further, suppose i am using a DB2 DB in my job. I am connecting to db server using port say 65432. Now when i do a

Code: Select all

netstat -a|grep 65432

on both head nodes and compute node, i can see connections established.

So i was wondering where exactly are the connections made from? Head node or compute node.
Also the section leaders for the connector stage runs on compute nodes, so shouldnt the DB connections be made from the compute nodes?

Currently i can see DB connections on both head and compute node.

[PaulVL]

"Validate Column type and size mismatches" happen off the head node I believe.

Some connector stages have that as options.

[AshishDevassy]

here is something you can try.
I am sure that the DB2 Database must be cataloged on both the head and the compute node. UnCatalog the database in question from all the compute nodes and run the job.
And when you run send the job to these compute nodes.

[PaulVL]

umm that's a little harsh and unwise since that could mess up your developers.

I vote nay on that last recommendation.

You could do it in an Admin Only environment, but not a live DEV/TEST/PROD where users will chew you out.

[skathaitrooney]

Hey guys,

I just asked my DB2 DBA for a application snap, in that i can see the connections being made from the compute nodes only(that are in Executing state).

Also, on "VIEW Data", connections are made from the head node

[qt_ky]

"View Data" at design time uses a different connection method than connections made at run time, so that would make sense.

The DB2 Connector also has a setting (not sure its value, as I'm not on a grid): In the Alternate conductor settings field, specify whether you want to provide alternate connection settings to use on the conductor node. The default value is No.

Also, here is some interesting reading on using direct connections:
http://www-01.ibm.com/support/knowledge ... tions.html