RFE: choose which sections of the model to discover [Discover Tool]

[rssh22]

In some scenarios only a bunch of information about the domain model is required. For example, topology section provides enough information for operating over a domain (start, stop, health-checks). Getting all the model is a long process that can take minutes, depending on the domain. However, getting just some sections would be faster.

So, for example, a simple approximation could be done by adding a "-section

" argument to tell the tool which section to discover:

• topology
• resources
• appDeployments
• domaInfo

and advanced solution could allow to filter more specifically which components to discover. For example, which weblogic clusters, or servers or kind of resources, etc.

[ddsharpe]

We will consider this enhancement.

[ddsharpe]

This enhancement is easy enough to do with the current code. My biggest concern with implementing this is the confusion that it would cause. Models generated with this -section option would be incomplete an unusable by themselves (mostly). For example, an application would be targeted at a server that did not exist in the model. As long as that partial model was used against a similar domain, the user is fine. But for every other use case, that generated model requires editing. My concern is that this would generate a significant number of customer issues outweighing the benefits of the limited use cases where this is valuable.

[rssh22]

I don't understand the technical complexity of the user's confusion. From my point of view, considering I'm not understanding the technical implications in deep, I think users will still be able to use the tool as before. And this new feature could be helpful in some other scenarios. For example:

Imagine that you have a big WLS Domain with hundreds of applications, shared libraries and resources deployed . What If I just want to replicate the topology or edit some parameters? With the current discovery tool I'll have to manage a very big model file, which could be very inconvenient and prone to errors.

As I said, I may be wrong in the conception of the idea. But in my mind it sounds right.

[robertpatrick]

@rssh22 The problem we see is that a partial discovery will typically result in "dangling references". For example, if I only discover the topology section of a domain that has the DataSourceForAutomaticMigration configured, the discovered model will reference a DataSource name that is not present in the discovered model. Likewise, discovering resources without the topology will result in those resources referencing Target values that do not exist in the model.

We struggle with enabling this not due to technical complexity from an implementation point of view but more from a support point of view where users that do not fully understand the implications of using this feature will struggle to use the resulting model that is not referentially complete.