Le 13/02/2015 19:01, Randall Hauch a écrit :
Oh, one more thing:
>> On Feb 13, 2015, at 10:49 AM, Thomas Segismont <tsegismo(a)redhat.com
>> <mailto:tsegismo@redhat.com>> wrote:
>> ...
>> Hawkular metrics comes in two forms:
>> * a Java library (metrics-core)
>> * a Java EE web application (built on top of the library)
>>
>> metrics-core can be embedded in any sort of JVM application but it
>> expects to find a Cassandra cluster somewhere.
I understand that the collectors will be libraries; you’ll probably
eventually want these in multiple languages. But it sure seems strange
to have the Hawkular engine (for lack of a better term) be a library.
Not only is that more complicated to set up, but if you design it to be
embeddable then it will be harder to also make it scale.
Most consumers of the metrics tool will use the web application. The
metrics-core library might be used in projects in advanced integration
scenarios.
It sure seems easier to focus on only installing Hawkular as a
separate
(potentially distributed) system that is set up and used. Sure, config
it to point to the Cassandra cluster. But that can be easily wrapped up
in a Docker image that people download and simply start, pointing via
ENV variables to a Cassandra instance/cluster (perhaps also started from
a Docker image). Better still is that these Docker images can be wrapped
in a Kubernetes pod (that uses a Cassandra pod) and deployed on
OpenShift. Now you have Hawkular as a service, but still easily
installed on a customer’s network.
This is already in our plans.