Load balancing and cluster

[Former Member]

Please help on this,

Steps I have created

1. Install Portal on node A

2. Create j2ee cluster instance on node B pointing to node A

3. Software load balance both the nodes

Even after load balancing the two portal nodes,if I shutdown Portal node A dispatcher,then the portal node B dispatcher will automatically shutdown, since the portal node B is a j2ee cluster for Portal node A.In this case how the URL will work from node B ? Please explain

[Former Member · Accepted solution]

From your description and the documents I understood, that state controller is running only on server A, and that no backup state controller is installed on server B, so the cluster nodes on server B will shut down, when we shutdown state-controller on server A.

Since I have installed on J2ee cluster node on server B is it possible to configure the server as backup state controller, if so, please let me know the steps or refer some documents to remove additional node of J2ee cluster from server B.

I will provide you star as soon as possible.

[MarcelRabe · Accepted solution]

You're right. Once the state controller fails, the cluster information is gone. The backup state controller is optional but recommended, so yes you can configure your server B as the backup state controller.

If you do not see the state controller in the configtool, then to be honest I don't know if you can manually add it. I would run a cleanapp.bat or just delete the entire usr directory and sapinst directory on node B and start again. If you keep hostnames and ip's the same, my experience is that it works fine (even the clusterid's remain the same).

but there's a part in the pdf about removing a cluster node. I don;t have it by hand here.

[Former Member · Accepted solution]

Marcel,

Thank you very much for the information.I uninstalled the cluster node and installed again with the option

of "backup state controller". Now it is working fine. When i shutdown the dispatcher of Server A, now the Server B is not going down. And I am able to see iViews and users on Server B.

I have one more question,that do I need to copy any files physically from Server A(main server)to Server B(Additional node with backup state controller) manually.

Note : I have given you star.

[MarcelRabe · Accepted solution]

Hi,

thanx for the star. Good to hear it works now.

You do not need to copy any files manually. In principal all components are replicated when needed. But the truth is that this doesn't work properly sometimes.

In the case when a iview works on node a but not on node b, find out which par file the iview uses and use the Console Administrator to install the missing component on the other node.

rdgs Marcel

[Former Member]

I did portal install on server A. I started dispatcher and I am logging with the URL http://ServerA:52100/irj on server A. And I did install of J2EE cluster on another server called ServerB. I started Server B's dispatcher. When I am starting Server B's dispatcher in Server A I am able to see another time Element joined.The URL for the serverB is http://ServerB:52100/irj. When I created users on Server A, I am able to see the users on Server B.When server B goes down or if dispatcher of B stopped even then Server A is working because it is independent install. But when the Server A goes down which is the main server then the additional J2EE cluster node that is Server B is going down automatically because it is dependent on Server A. But I want to make it as complete high availability of failover,means if Server A goes down it should fail over to Server B and if Server B goes down it should fail over to Server A. I understood we can do software load balancing from windows level or hardware load balancing to get the high availability. But now with my existing installs, I want to make failover to work, please suggest me which is the best way to do without changing any of the existing installs.

My environment is :

Operating system: Windows 2003 Enterprise Edition

Data base: SQL 2000(Already configured as cluster)

Portal: EP6.0 SP2

[MarcelRabe]

Make sure you use J2EE PL21 or higher (25+ is my recommendation) that uses the State Controller principal. If you are using this, then you've appearently setup the cluster wrong. The dispactcher should only serve the local server node and the state controller controls the cluster.

the 2nd node serverB can be setup as backup state controller. In both state controllers point them to one another (and not in the server and dispatcher node).

I suggest you read the cluster config guide in the service market place (or in here).

For URL load balancing you need something like MSCS, or a hardware load balancer. The problem with MSCS is that is does not do high availability (it does not check to see if the portal server A or B is working).

If you get it to work, please press the star button.

[Former Member]

I would like to re phrase my question in detail,

I have SQL cluster installed and configured,after that I have installed Enterprise Portal 6.0 on node A,pointing to the cluster database.Next step I have installed EP60 on Node B as j2ee cluster for Portal node A.Please explain me in detail how to achive failover between in these nodes.

Note : Installed version Enterpirse Portal 6.0

[MarcelRabe]

Just a question. Did you read the documentation on service.sap.com/epinstall? The way to go depends on the exact version of J2EE, portal (SP2 or other) and what you are trying to achieve: fail over or true cluster with load balancing?

Even though you have two portal nodes, you will need some kind of load balancing for the portal URL (hardware or software).

[MarcelRabe]

What version of the portal?

[Former Member]

To clarify my understanding:

On node A you have the SCS and JC instance running?

On node B you have a JC instance (dispatcher + server processes)?

I'm assuming you're using the SAP Web dispatcher for load balancing.

What command do you use to shutdown the instances on Node A? You need to make sure you do NOT shutdown the SCS instance -- the JC instance on node B will still keep running if the SCS instance is still available on node A.

This is the recommended HA setup from SAP.

[Former Member]

Somckit,

Have you tried this build (node A: with SCS and JC instance), and Node B: with JC instance (dispatcher + server processes) with a single/or clustered SQL Server DB+CI on NW04 SR1? Need some assistance in the process flow of sapinst.exe in building a distributed model. Thanks for your time.

Regards,

James