Adding 'cluster' property to resource 'wls_server'.

[diegoliber]

On issue #317 , I mentioned an inability on configuring a Weblogic Cluster without knowing a priori all servers and machines on the infrastructure.

With this update, "cluster" property to the "wls_server" resource that allows the server to provide which cluster it will belong to. This partially solves the problem, since a puppet script can declare which clusters are available in the admin server and its main properties, and on each managed server, which cluster it will belong.

There's a remaining issue on updating the wls_server cluster and machine information, since it requires a managed server restart, before activation. This contribution still don't solve these edge cases.

[coveralls]

Coverage remained the same at 38.714% when pulling 1cb5166797483b1882676da18cbf979eeb1801bc on diegoliber:master into 59e04a80aa78fa72b162ca4948f66d53feeb3a03 on biemond:master.

[biemond]

Hi,

Can you check this #200, it should does the same only delays it to the cluster creation. In this you can create clusters first , create servers and run it again or first servers and then clusters. It works in every situation.

you can do this on a wls_server

server_parameters: 'WebCluster' and on wls_cluster use this.

servers: 'inherited'

[diegoliber]

Hi,

I understand the features provided by #200. I just believe it's a fragile implementation, since it depends on an specific format on the Notes attribute on the managed server (which is just a String of comma-separated values), and require a puppet agent execution on the managed server, followed by a puppet execution on the admin server, in this order. I still have to check if this solution solves the problem of requiring a managed server restart.

[biemond]

Ok,

but you can add the cluster config also to all nodes, no need to run it again on the adminserver

so basically you want to do 1 time provisioning on all nodes. So on the admin ,just a domain with some empty clusters . After that the server nodes which will create a machine, managed server add itself to the cluster. This can also work for persistence and a jms server.

How do you want to handle datasources and jms modules. You also need to have something to add the server to a datasource target or a jms server to a subdeployment . If you don't like this and not using notes it will remove the other targets.

also I have managed server control which can subscribe to managed server changes and do an auto restart when the config changes. This can solve your restart problem.

Maybe we should make a shared vagrant box where we can make this use case.

Thanks