User scenario: add resources to the inventory and start monitoring

[theute]

For RHQ.next, I am looking for the best user experience.

Putting aside the implementation details, I wanted to discuss the flow for a user to start monitoring "something" on bare-metal setup (/non-OpenShift)

I imagine the following steps (not necessarily the only way, but something that needs to be supported):

1. User has a list of platforms to monitor and the option to add a new one
   
2. When the user wants to monitor a new platform, the server propose an agent to download and basic instructions to start the agent (Ideally it would provision this)
   
3. Once the agent is running, the server automatically adds the platform to the inventory (I think we can/should skip the discovery queue)

Now the questions are:

• Do we also add to the inventory, everything that has been discovered (disks, CPU, servers, servlets, JMS queues...) ? (I would say yes, with an option to hide from the UI certain resources and/or types of resources)
• Do we also start monitoring (collect metrics, setup alerts) everything that was discovered and added to the inventory ? (I would say ideally yes, but no since we need to be cautious on disk space and false alerts)
• If we do not start monitoring everything, do we automatically start monitoring "some" resources ? (Such as high level EAP servers, could be advised by the agent plugin and overridden by server preferences

Answers, comments ?

[jayshaughnessy]

I'm not sure a discovery queue should go away completely, although it could get absorbed into a Platforms/Machines/Feeds view that showed you the discovered servers for each, and which had been imported and which had not. For example.  Maybe have certain server types set for auto import.  Once imported I think monitoring would start.

I like the overall flow you show.

[rutlucas]

Hi,

A comment from past experiences with other products and previous RHQ/JON experiences: normally, an agent is installed configured as part of the OS layer, via an .rpm or similar of other platforms.

So, from flow perspective, I would expect first a "ping"/"active" machine you want to add, if not responding/present, then offer him to download agent.

Is there a possibility that an agent is installed and "hibernated" until a server makes that ping with initial order of monitoring ?