Deploy in an IBM Websphere 7 cluster with high availability nodes

Question

Wednesday:

• Java EE webApp
• JDK: 1.6,
• AS: Websphere 7 Application Server,
• OS: redhatzLinux

I am not a websphere administrator, and I am asked to develop a method or script to solve the problem below:

I have a cluster with three NodeA NodeB and NodeC nodes. My application runs on these clusters. I want to deploy my application on these nodes, so I don’t need to immediately display them all. These days, deployment is as follows: we arrive at night to stop all servers at once from the console. Then we install the application on the main node, which is located on the same computer as the deployment manager, and then we synchronize and bring all the servers back one by one.

What they ask me to do is that we update the application or install a new ear file without reducing everything, as this leads to a simple application. Is there any way to achieve this. WAS 7 is a very mature product, I am sure that there must be a way to do this.

I looked at the documentation / tutorial in which we can do something like “Update” when we select the application (from Apllications> enterprisesphere web applications) and select the update, and then select the “Replace full application” radio button and the “local file” radio button "system" and point to the new ear file. But in this case, the dock says that it will also destroy all servers during the upgrade. Its the same as before. There is no deployment on the Internet.

I am a Java programmer, so I thought about what tools I need to solve.

Tell me if this could be the problem :

1) We destroy NODEA 2) We remove NODEA from the cluster (by clicking the delete node button or using removeNode.sh) 3) Set the new ear to NODEA (can we do this in the same admin console or through the shell script or jython or maybe as a standalone server) 3) Then we start it back, and then add it to the cluster. Now we have NODEA with a new application, and node B and NODEC with old versions of applications.

Then we reduce NODEB to remove NODEB from the cluster, install the application on NODEB, run it again Add it back to the cluster

Now we have two nodes with the new application and NODEC with the old

we will try the same process for NODEC.

Will it work. Has anyone tried this. what problems can you imagine what could happen.

I will be grateful for any feedback from here. I am sure there is an experienced ppl on this forum. I don’t think this is a rare problem, I believe that this is what any organization with high availability requirements wants.

Thanks for any help in advance. Syed ...

Answer 1

This is a possible duplicate. How can I deploy zero downtime in a cluster environment? . Here is essentially my answer to this question:

After updating the application, you can use the Update Deployment feature. Instead of saving and synchronizing nodes after the upgrade, you can use this function, which automatically performs the following tasks to allow changes to propagate to all deployment goals while maintaining high availability (provided that you have a horizontal cluster so that cluster members exist on several nodes that sound like you):

• Save session changes in the main configuration
• For each node in the cluster (one for continuous availability):
  • Stop cluster members on node
  • Sync node
  • Launch the application server (which automatically launches the application)

Answer 2

Alternatively, you can perform the following procedure.

• Stop all nodes except Node A.
• Comment on or disconnect Node A from the load balancer or plugin (this way traffic will not go to node)
• Expand the app.
• Changes will only be synchronized on Node A, since its nodeagent is up.

• Uncomment / enable Node A from the plugin / load balancer.

• Comment / disconnect Node B from the plugin / load balancer to stop incoming traffic on node.

• Run nodeagent Node B to synchronize file changes on node. The ear application will stop and start after synchronization.
• Uncomment / enable Node B from the plugin / load balancer.

Repeat steps 6, 7, 8 for all other nodes.

Regards, Laika Ahmed