Archives of the TeradataForum

Message Posted: Fri, 16 Apr 2004 @ 20:36:57 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

Kevin,

We have set up DNS entries for both ODBC connections and a CLI connections.

For ODBC
We have entries in our DNS servers for our Teradata production and development servers; however, you could just enter the IP addresses if you wanted.

When using DNS, if you select the "do not resolve the alias name to IP address" checkbox, you are telling ODBC that you are checking and resolving this 'name to IP address', right now, at the time you are configuring the ODBC connection. If this checkbox is not selected (unchecked) the 'name to IP address' resolution will occur each time you connect to your server (or node). If your DNS name\IP config does not change much you may want this checked; but, it is safer to leave it unchecked so that ODBC checks the 'name to IP' resolution each time you connect.

     DNS NAME        IP ADDRESS
     prodnode1       xx.xx.xx.xx
     prodnode2       xx.xx.xx.xx
     ....and so on

For CLI
We have been experimenting with a DNS setup for CLI - eliminating the need for the hosts file. We have are using DNS similar to the way we use it for ODBC with a slight difference; we have entries in DNS for our development and production "Database Names" the same as we did for ODBC but we asked our DNS administrator to use an Alias name - the same alias name - for each separate node. This DNS alias name matches the name chosen as the "Database Name". You can use whatever name you want for the "DNS Alias and Database Name" as long as it is the same name; and, the same name as the front of the COPx portion of the entire DNS name.

See below:

     NAME            IP ADDR ALIAS
     devsysCOP1      xx.xx.xx.xx     devsys (DNS alias and your Database Name)
     devsysCOP2      xx.xx.xx.xx     devsys (DNS alias and your Database Name)
     ...and so on.

     prdsysCOP1      xx.xx.xx.xx prdsys (DNS alias and your Database Name)
     prdsysCOP2      xx.xx.xx.xx prdsys (DNS alias and your Database Name)
     ...and so on.

So when connecting through any CLI application I would use prdsys or devsys as the Database Name for these applications.

I never really understood the COPx portion of this name but I have heard that it is a legacy type naming system that identifies, or identified, the system as one single system - perhaps someone else could contribute to this. This is probably why I am concerned about rolling out this"alias" type setup to our masses; we are concerned about balanced distribution of sessions. On the mainframe there is the TDP which handles the even distribution of these sessions. On a LAN based client the TDP is replaced with the MTDP and we are not sure whether the distribution of sessions will be balanced when integrated with DNS in this manner - i.e. does DNS do a round-robin distribution to the different IP addresses for your aliased system name; and then, if it does, do you assume that Teradata then handles the even distribution of sessions - does the COPx name even come into play? We need to do some further testing.

Hope this helps or at least furthers the discussion around the name to IP address resolution.

Steve Rockarts
Database Analyst
Information Resource Management (IRM)
Hudson's Bay Company

	<Prev	Next>		<<First	<Prev	Next>	Last>>