[Hawkular-dev] need design decisions to be made for server-side config of agent

[John Mazzitelli]

> It might also be good to reserve a time to have a call to discuss this over
> bluejeans.
> 
> As far as I can tell, the plan isn't to have server side configurations
> (like you would with a pull model) but instead to continue to have client
> side configurations but have the server be able to push out updates to the
> client.

The idea isn't a "push from server to agent", rather its the agent pulling its config from the server. 

> Will this require a new component to expose REST endpoints at the server
> level? or are we planning on reusing an existing component?

We already have something in place today:

https://github.com/hawkular/hawkular-commons/blob/mwm-wildfly/hawkular-inventory-parent/hawkular-inventory-service/src/main/java/org/hawkular/inventory/handlers/InventoryHandlers.java#L78-L94

> Are we going to be able to have individual configurations per server, or
> are we lumping things into server types (eg all EAP7 instances have the
> same configuration)

Preferably we are going to be grouping config based on server type. I mention an example in the JIRA comment:

These server-side configuration files are centrally located and will define types for all servers to be managed. There is not going to be one uber file. There will be two files per server kind (two files because one is for jmx exporter, the other is for our agent).

e.g.
EAP-7.0.0-Final-jmx-exporter.yaml
EAP-7.0.0-Final-inventory.yaml
Fuse-1.1.0-RC2-jmx-exporter.yaml
Fuse-1.1.0-RC2-inventory.yaml

So when EAP 7.1 is released, we'll just add two new files to the central location on the server and the agent can start downloading those to managed EAP 7.1 servers.