When you have an application using HTTP session replication setup in a HA cluster on OAS 10.1.3.2 and above you can easily ensure that when you deploy a new version of that application it will continue to run with no down time. What's important here is you select the correct deployment option.
Redeploying with ASC
If you use ASC for the deployment then your effectively redeploying to the cluster in parallel - all nodes in cluster will get notified and process the deployment in parallel. That means you will most likey have down time and get errors for those currently using your application.
Redeploy with admin_client.jar or redeploy Ant task
With these two deployment choices you can specify sequential and keepsettings options however in some circumstances the session state of an application may be lost when you redeploy an application to a cluster.
In Oracle Containers for J2EE release 10.1.3.2 or 10.1.3.1, you can use either the
waitsec parameter of the
waitsec subswitch or the
sequentialDelay property of the
redeploy Ant task to specify a number of seconds between redeployments to different OC4J instances in a cluster. This delay can provide enough time for replication of session state
For example, the following
admin_client.jar command specifies a wait time of 15 seconds between redeployments to different OC4J instances:
java -jar admin_client.jar deployer:cluster:opmn://host:port/home oc4jadminIt's documented here, make sure your using 10.1.3.2 or above for this
password -redeploy -file "myapp.ear" -deploymentName rolling -sequential 15
10.1.14 New Wait Option for Sequential Redeployment to a Cluster