[Hawkular-dev] [GSOC 2016], "Hawkular-agent for vert.x"

[Thomas Segismont]

Hi Anton,

You will find my answers inline.

Le 12/03/2016 23:52, Anton Okolnychyi a écrit :
> Hi, all!
>
> My name is Anton Okolnychyi. I am keen on participating in GSOC 2016.
> That's why I am; trying to gather relevant information to write a nice
> proposal for a "Hawkular-agent for vert.x" project on GSoC 2016.
>
> Thanks to Thomas Segismont for his initial advice. I have analyzed the
> existing functionality of the vertx-hawkular-metrics module, its current
> features, and architecture. In addition, I have investigated Hawkular's
> REST APIs of Metrics and Inventory services. Currently, the
> vertx-hawkular-metrics module reports only to the Metrics service. The
> idea that is available in the scope of GSOC 2016 is to enhance
> vertx-hawkular-metrics' capabilities. In particular, enable reporting to
> the Inventory service of Hawkular.

Exactly.

>
> As a result of my research, I have some questions. It would be great if
> someone can answer them.
>
> 1. Since the main goal of the project is to actually enable reporting to
> the Inventory service, one should come up with an appropriate graph
> architecture of Inventory which should contain resources(vertices),
> relationships between them(edges) and optionally assigned metrics to
> each resource (I am aware of many to many relationships here). The
> vertx-hawkular-metrics module currently has metrics that cover the
> following: datagram/udp, event bus, tcp, http client, http server.
> That's why one can define the corresponding resources for these metrics
> in the Inventory (e.g. Event Bus, Net Client, HTTP Client, Datagram
> socket, HTTP Server, Net Server etc.). Will that make sense? Am I going
> in the right direction? Alternatively, we can define a resource that
> will cover several metrics at once. If you have anything that is
> important to include as a resource in the Inventory, please, share that
> with me.

I believe your model is correct:
Event bus resource has event bus metrics
NetClient resource has NetClient metrics
... etc

We might want to have event bus handlers as separate resources with 
their own metrics.

>
> 2. The vertx-hawkular-metrics module collects only counters and gauges
> and reports them via the POST request to /metrics/data (correct me if I
> am wrong). Would it make sense/be interesting to also handle some
> availability metrics? The Hawkular API allows to do that and the current
> architecture of the vertx-hawkular-metrics is flexible enough to
> integrate a new metric type quite easily. Probably, this question should
> be asked on the vertx-dev list. But if someone knows the answer or has
> any possible applications/suggestions in mind I would be glad to hear.

I can definitely answer that as I wrote the vertx-hawkular-metrics module :)

It would make sense to add availability metrics as well, yes. I didn't 
include them in the current implementation because to me, they make 
sense when you report to inventory as well. Otherwise it is not clear 
what a resource is (and why it would be up/down/etc...).

>
> Thanks in advance! Any help is highly appreciated!

Thank you for your interest!

>
> Sincerely,
> Anton Okolnychyi
>
>
> _______________________________________________
> hawkular-dev mailing list
> hawkular-dev at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/hawkular-dev
>