Universal Agent - waste of time?

[Raymond.McCabe]

I'm hoping someone has had a look at this and has seen a way of making it work that I just can't see.

I've been reviewing the documentation for 4.3 and as far as I can tell with the new UA 'feature'

- The agent installation destination moves away from the existing structure (meaning any existing references, scripts, autodeploys need thrown away or updated dependent on how you progress UA)

- You can't attach new versions to existing instances post update unless you are using Oracle Hotspot JRE (that's about 1% of our estate)

- The rules in the rulebook for Java require an Application/Tier/Node to be defined, doesn't that mean one rule for each application node? That's 1,200 rules for us for each new agent version?? I'm sure I must have that wrong but can't see how it's meant to work.

Even if we wanted to use the above and use it for new deploys:

- we would need to maintain two deployment solutions and config

- there is more overhead creating rules than there is running a script to update a host

- we sized our AppD requirements based on original definition so now need at least double the capacity on every host

This is poor, even for AppD which is already very poor at providing means for managing the environment, roles, permissions and config.

[scott.miller]

I looked at the UA when it first came out and sort of came to the same conclusion as you. it wasn't a good fit for an environment that was already "operationalized".

But some new features have been added and using the UA could be a viable option for a new environment and they've added a lot of accommodations to convert an existing one.  The biggest being the ability to add custom variables to the Node and Tier names. While this doesn't mean you get to keep the existing familiar names that you already have, you don't have to make "1200 rules" for each one. 

You can set up a rule for an entire application or one per tier. 

One unwritten benefit is that the UA can be installed by the Unix/Linux team and you'd theoretically would never have to touch the machine. Everything can be done via the REST API Json calls. It will even add the correct params to the JVM start script. 

There is up front work to getting this up and running, but once you've created a working library of rule books, it seems like decent single point for maintaining the environment.