[Hawkular-dev] wildfly agent news

[John Mazzitelli]

Just wanted to give a brief summary of some changes that went into the agent recently.

1) We now produce our own WildFly feature pack and can run in Swarm

Thanks to Bob McWhirter, the agent build now produces a feature pack that can be used by WildFly for installing the agent so it comes with Wildfly out of box. I don't know much more than that - but it sounds cool :)

Bob also got the agent running as a swarm app, so by running an uber-jar (via a command like "java -jar ...") you can run the agent. See https://wildfly-swarm.gitbooks.io/wildfly-swarm-users-guide/content/hawkular.html

2) We now monitor native platform resources.

I think I already announced this but I'll repeat. The agent can now monitor basic native platform resources like CPUs, system memory, and file systems - supported on Linux, Windows, and MacOS. When you run the wildfly agent, by default you will not only get WildFly servers in inventory, but you'll also get platform resources including a root "Operating System" resource along with its children that include Processor resources, File Store resources, and Memory resources. Collected are metrics such as CPU usage, total and free system memory, total and free disk space.

3) We now utilize the inventory batch API

The agent uses the new inventory batch API to store a resource, its type, its resource config, and its relationships to its parent and metrics in one REST API call. This speeds things up somewhat, but hopefully in the future we can optimize this even further to bulk insert multiple resources at a time. As part of this work, you will now see inventory storage diagnostic data logged in the server log. This can help us determine the performance of our rest calls into inventory. For example, in the diagnostic log message, you'll now see something like this: "feedid.diagnostics.inventory.storage-request-timer: type=[timer], count=[115], min=[283.090987], max=[6260.832863], mean=[3732.066908]" meaning we made 115 inventory requests (which means one inventory batch REST call ) that took an average of 3.7s.

All of this is in master and will go out in the next agent release (I actually think the platform resource stuff is already released and in kettle).