Connecting to Different Physical Node on Partitioned Database on DB2 8.1.6

[stabbert]

We are running DB2 UDB 8.1.6 in a partitioned environment where we have
8 physical nodes. We have a process that remotely connects to each
individual node and loads some data. By setting the DB2NODE variable
to a different value then the coordinator node is the only way I am
aware for a remote application to do this so I am making the assumption
that is how this application connects. What I am more interested in is
that after logging into any other node besides the coordinator node and
running NETSTAT I don't see a physical TCP connection to the remote box
that the application is running on. I only see a physical TCP
connection to the coordinator node. Why is this? It would appear that
even though we are connecting to a different node it is still making
some sort of a connection through the coordinator node and then the
coordinator routes the traffic to the appropriate node. Can anyone
confirm this or lend more information?
Spencer

[Ronen]

DB2NODE creates a gateway to the remote node through the node your
connected to instead of directly connecting to the node specified in
DB2NODE. If you want to connect directly to the remote node i would
catalog each remote node locally, and connect directly.

currently with that way you have it setup you have

client ---> node0 (gateway connection) ----> <node specified in
DB2NODE>

How to connect directly to a node:

cataloging each node:

db2 catalog tcpip node node0 remote <server0name> server <servicename>
db2 catalog tcpip node node1 remote <server1name> server <servicename>
....

catalog database on each node:

db2 catalog db <dbname> as <dbname0> at node node0
db2 catalog db <dbname> as <dbname1> at node node1
...

then instead of using DB2NODE=X connect directly dbnameX

Ronen

stabbert wrote:

We are running DB2 UDB 8.1.6 in a partitioned environment where we have 8 physical nodes. We have a process that remotely connects to each
individual node and loads some data. By setting the DB2NODE variable
to a different value then the coordinator node is the only way I am
aware for a remote application to do this so I am making the assumption that is how this application connects. What I am more interested in is that after logging into any other node besides the coordinator node and running NETSTAT I don't see a physical TCP connection to the remote box that the application is running on. I only see a physical TCP
connection to the coordinator node. Why is this? It would appear that even though we are connecting to a different node it is still making
some sort of a connection through the coordinator node and then the
coordinator routes the traffic to the appropriate node. Can anyone
confirm this or lend more information?
Spencer

[stabbert]

Great information! Thank you for the confirmation. We are having some
routing issues occur when the node0 is acting as a gateway connection
and we have some traffic going over a network that it shouldn't be.
This information will be of great help. I am assuming also that when I
specify something other then the controlling node the controlling node
then uses the db2nodes.cfg file to find the hostname that partition
resides on and that is the host the controlling node makes a connection
too. Is this correct or is the partition number to hostname resolved
someway else?
Spencer

[Thiru]

Yes Stabbert. You are perfectly right.

Cheers,
thiru
WantedToBeDBA.