Hi,
I have some concerns here as it starts to mix a few conceptually independent
perspectives for a SCA domain. Simon has already pointed out some of the
players. Let me start with a few simple questions:
1) What's the result of the discovery? A list of SCA nodes (for example,
node.composite that describes the node configuration)?
2) Do we have a "manager" that initiate the discovery and maintain the
membership dynamically?
3) Where are the SCA contributions from? Is this part of the discovery?
4) Which part is responsible to manage the SCA contributions and resolve
composites at the domain level? The input to a SCA node is a deployable
composite and a list of SCA contributions (contribution uri + location).
5) I cannot find samples/helloworld-distributed. Where is it?
6) Can you describe the steps in sequence for this approach? I would like to
see how the players interact in this dynamic fashion and what information is
produced from each interaction.
Thanks,
Raymond
From: Simon Laws
Sent: Wednesday, September 10, 2008 5:36 AM
To: [email protected] ; [EMAIL PROTECTED]
Subject: Re: A new JMS impl of binding.sca and "dynamic" SCA domains
On Wed, Sep 10, 2008 at 8:32 AM, ant elder <[EMAIL PROTECTED]> wrote:
On Thu, Sep 4, 2008 at 7:53 AM, ant elder <[EMAIL PROTECTED]> wrote:
While implementing part of the webapp support in the JEE spec i've made a
JMS version of binding.sca, its still a bit rough around the edges but its
got to the stage where it does work.
One feature i think is cool is that by default it uses UDP multicast to
discover nodes, so you can just start up nodes and they automatically find
each other and become part of the domain.
There is a sample showing using it in samples/helloworld-distributed, the
README explains how to run it. Try it out, I'd be interested in any
feedback and if anyone is interested in helping there is still lots to do.
One thing that would be useful is to configure it from a binding.jms in the
definitions file so you could then use custom domain names by including a
definitions.xml when starting a node.
...ant
Ping?
I've been playing with this for a while now and i really like how easy it
makes using distributed SCA domains so i'd like to make this a first class
part of Tuscany. The type of things it needs are adding methods to the Node
APIs to start up a node using a list of contributions and a domain name -
programmatically and from the cmd line, the same for webapps so something
like a new ServletListener and a way to configure the domain name used. Have
a way to configure the network of brokers if auto discovery is wanted or
static node IP addresses and if security is needed etc. And have more
samples and demos showing how to use it. I'll start on this over the next
days if anyone would like to help that would be great, this is interesting
and has a lot of potential IMHO so i encourage anyone to get involved.
...ant
Hi
I'm interested in the prospect of dynamic discovery of some form. There are
two natural parts to the domain.
1) The domain manager. Currently this takes contributions and munges all the
information together to deliver configured composites to the node runtimes.
There is a user configuration step where node information is added to the
domain.
2) The runtime (nodes). These read the preconfigured composites and run them
up.
If the nodes were able to find each other and the domain manager this would
be an interesting alternative to the manual configuration of nodes in the
domain manager.
A simpler first experiement would be to take the nodes as we have them,
install your binding.sca.jms in place of binding.sca.ws, set the default URI
for binding.sca to something like "DYNAMIC" and then put a tweak in the wire
builder to spot this special case and leave the reference unwired.
Simon
