Olivier Corbun wrote:
> Hi,
>
> I was wondering what are the plan for the node agents in V3.
> Are they planned to be "OSGified"? And will it be possible to run server
> instances without them in V3?
Node Agent can be designed to be extensible, but it generally doesn't need
to be. It's supposed to be a low memory footprint process with defined
functionality for life cycle management of server instances, which are
of course extensible.
In V2 as well, it is not a must that you start the node agent. If you want
to manually start the server instances you can use the startserv script
but that's a generally discouraged option.
At least for now, we plan to have node agents on all non-DAS machines.
There may be a possibility that on DAS-machines, you don't need node
agents. This is yet to be seen however, since if node agents are
supposed to be low overhead.
We definitely will beef up the synchronization process and the distribution
of modules/bundles to the instance machines, when one has distributed
the modules/bundles to DAS machine.
What advantages do you see from providing OSGi support in the Node Agents?
None of the user code runs in the Node Agents.
>
> Regarding clustering, is there any concept of having OSGi controlling
> bundles over a JavaEE cluster?
I am not sure what you mean by OSGi controlling the bundles over EE clusters.
If a Java EE application deployed to a v3 cluster depends on services
in OSGi bundles, we need to have a cluster-wide way of making those bundles
available to those clusters, which run the application code. That's the work
we need to undertake during the synchronization of bits from DAS to clusters.
At the moment, we are figuring out how to expose OSGi to applications, I guess.
> How does this new Felix based architecture relates to GMS/Shoal?
> I guess the way OSGi is used is limited to a single JVM without
> interaction in between. Shoal covers this inter JVM relations. Will
> Shoal members in one JVM be seen as OSGi bundles?
Don't know yet.