Weblogic 7.0 JDBC Culster

[Santhosh Kumar]

I'm trying to configure the Weblogic JDBC in cluster. I'm getting confused at the conceptual level.
Let's say I have following two Two m/cs
1. MC1 (Hosting Adming server AS1 and managed server MS2
2. MC2 (Hosting managed server MS2)
MS1 and MS2 are part of the cluster CLUSTER1.
I have created a JDBC pool and corresponding datasource in managed server and given "CLUSTER1" as the target. As MS1 and MS2 are part of the cluster, pool should be automatically be available in the both managed server. My qustions are,
1. When we look up to get the Datasource context, which server should I lookup to? MS1 or MS2?
2. If we lookup to MS1 or MS2, its bound to one sever. How it will work in cluster?
3. Is it possible to look up to Admin server?
Your help is greatly appreciated.
Santhosh.

[Chris Mathews]

You is taken of when you get your initial context.
You have two choices:
1) Give a comma separated list of nodes in the cluster as the PROVIDER_URL.
2) Give a cluster DNS name the contains the full list of all nodes in the cluster.

Hashtable ht = new Hashtable();
ht.put(Context.INITIAL_CONTEXT_FACTORY,"weblogic.jndi.WLInitialContextFactory");
ht.put(Context.PROVIDER_URL, "t3://node1,node2,node3:7001");

or

Hashtable ht = new Hashtable();
ht.put(Context.INITIAL_CONTEXT_FACTORY,"weblogic.jndi.WLInitialContextFactory");
ht.put(Context.PROVIDER_URL, "t3://clusterDNS:7001");

Of the two, the second choice is best. It is more maintainable and BEA admits to a bug related to the resolution of comma separated cluster lists.
Regardless of how you do the lookup, the naming service is then chosen by DNS Round-Robin.
[ September 17, 2002: Message edited by: Chris Mathews ]

[Santhosh Kumar]

Cris, I understood this concept now. I'll configure the same and test it. If I have any more questions, I'll continue back here.
Thank you so much,
Santhosh.

[arun sahni]

Chris
Just thinking WLS 7.x allows to run multiple instances of MS on single IP address. If I run 2 instances on same m/c then how can I figure out using DNS or configuring network file for name resoloution?
Sinc most of us for testing do not have many m/c to test true clustering what is the best soln. using 1 or 2 m/c's?

[Chris Mathews]

In 7.0 you can specify comma separated name :p ort pairs in your initial context lookup.

Hashtable ht = new Hashtable();
ht.put(Context.INITIAL_CONTEXT_FACTORY,"weblogic.jndi.WLInitialContextFactory");
ht.put(Context.PROVIDER_URL, "t3://node1:7001,node1:7002,node1:7003");

Definitely for development clustering on a single machine is the way to go. I don't know many places that are willing to give developers multiple machines. In fact, some places that I have worked wouldn't have given us a single machine if they could get around it .
[ September 19, 2002: Message edited by: Chris Mathews ]

[Rahul Mahindrakar]

Chris,
Can you point me to the weblogic 7.0 documentation for this.

[Chris Mathews]

Rahul,
Here is your link.

[Rahul Mahindrakar]

Thanks
It looks like I never read this document and focused on the other areas like ejb etc . Strange I still got 100% in my Weblogic Exam in the clustering section.

[Chris Mathews]

We can't read everything... try as we might.
I still have a wish list miles long with books that I want to read. Sometimes I think software development was invented by publishers to sell more books