Meeraj,
I got the JXTA working for sending messages. However, what I did was
just finding the error and patching it, so I just modified a line of
code and got to know that the problem was not on the dissemination of
messages by JXTA but on the peer name being used to dispatch the name
to.
So, I saw that the problem was that the sender of the message was trying
to send a msg to a peer called "slave" (as seeing in the following
exception that was being displayed on the browser after pressing the
"Contribute SCDL" button):
org.apache.tuscany.spi.services.discovery.DiscoveryException:
Unrecognized runtime slave
What I did was just commenting a line of code out and hardcoding the
name of the runtime in the JxtaDiscoveryService class (when the runtime
is registering itself):
/**
* Configures the platform.
*
*/
private void configure() throws DiscoveryException {
try {
//String runtimeId = getRuntimeInfo().getRuntimeId();
String runtimeId = "slave";
configurator.setName(runtimeId);
configurator.setHome(new File(runtimeId));
if (configurator.exists()) {
File pc = new File(configurator.getHome(),
"PlatformConfig");
configurator.load(pc.toURI());
configurator.save();
} else {
configurator.save();
}
} catch (IOException ex) {
throw new DiscoveryException(ex);
} catch (CertificateException ex) {
throw new DiscoveryException(ex);
}
}
After doing that, the SCDL was successfully processed.
So, as you can see, the problem was not completely solved (the runtimeId
is hardcoded). But, at least I found why the messages were not being
delivered.
Now, I was trying to understand where the target name comes wrong from
and I think the problem could be that the AssemblyServiceImpl class is
setting the wrong id in the include method:
.....
// create physical wire definitions
for (ComponentDefinition<?> child :
type.getDeclaredComponents().values()) {
URI id = child.getRuntimeId()
.....
Since, it finally invokes the marshallAndSend(id, context), which in
turn invokes the
discoveryService.sendMessage(id.toASCIIString(), pcsReader) method,
which ends up in an invocation to JxtaDiscoveryService.sendMessage(...)
with the wrong runtimeId (i.e.; slave)
So, as you can see, it seems that the problem comes from some place
outside of the scope of JXTA and I am not experienced enough to deal
with such issue. Do you have any idea where the "slave" id is being
wrongly set?
Best regards,
Mario
-----Original Message-----
From: Meeraj Kunnumpurath [mailto:[EMAIL PROTECTED]
Sent: Tuesday, March 20, 2007 12:39 PM
To: tuscany-dev@ws.apache.org
Subject: Re: Discovery update
Mario,
I will try to be as detailed as I can, if you need further info, pls
ask.
Tuscany code structure is roughly organized into kernel, runtime,
services
and extensions. There are other modules like plugins, console etc, which
are
not relavant in the context of this discussion. There is also a contrib
module, where we keep artifacts that are not ready to go, it is
important in
the context of this discussion because the JXTA impl is currently in
contrib, because of some issues we are investigating with some patented
code
with BouncyCastle.
tuscany-spi in kernel, contains the key model classes and service
provider
interfaces for Tuscany. Some of these are implemented by core in kernel
and
some of them are implemented outside kernel. DiscoveryService is an SPI
defined in tuscany-spi that is used by the runtime fabric for enabling
communication between multiple profiles participating in a logical SCA
domain. A profile is basically a group of services, both system and
user,
that are managed together. Multiple profiles make up a logical SCA
domain.
You can run profiles in the same VM or different VMs.
Discovery service provides basically one method,
1. Targeted delivery of a message to a given profile.
However, the JXTA impl, also provides a listener for receiving those
messages and so does the JMS impl. JMS is not a real option for us, as
long
term we want to enable multi-fabric profiles in the same domain, Java
and
C++ for example. This is where JXTA fits in nicely.
Profile names are important, as the JXTA implementation of the discovery
service uses the profile names as logical endpoints and use them to map
to
JXTA peer ids. The JXTA implementation uses a peer group specific to the
domain in which the profile is participating. The domain name is defined
in
/etc/runtime.properties of the profile. We do that to restrict
communication
between the profiles only in the same domain. We use PDP (Peer Discovery
Protocol), for maintaining a view of all profiles available in the
domain
and PRP (Peer Rsolver Protocol) for sending directed messages. PDP seems
to
work fine, however, PRP is not delivering the messages. I have posted
this
on the JXTA list and I can forward you the emails if you want (I will
forward it to this list)
Here are the key areas you can look at,
/java/sca/kernel/tuscany-spi: For the discovery service abstractions.
/java/sca/kernel/core: Usage patterns for discovery service
/java/sca/contrib/discovery/jxta: The JXTA impl of discovery
/java/sca/runtime/standalone/server.start: The basic shell for starting
a
tuscany server standlone
/java/sca/console: For a browser based console (it is pretty minimal
now)
/java/distribution/sca/demo: A maven assembly for creating an
installation
image for three feedrated profiles master, slave1 and slave2. In the
src/profile directory you will find the teh different profiles and their
system SCDLs. Currentlly, they use JMS, however, I have commented
component
definitions for JXTA.
you can start all three profiles in one vm using
java -jar server.start.jar master slave1 slave2
or
java -Dtuscany.adminPort-2000 -jar server.start.jar master
java -Dtuscany.adminPort-3000 -jar server.start.jar slave1 etc.
You can access the console using http://<server>:7000/scdlForm. As I
said it
is pretty basic, I am sure it will evolve :)
Once again, great to have you. You can send the patches to the list, I
will
test it and apply it.
Ta
Meeraj
>From: Jeremy Boynes <[EMAIL PROTECTED]>
>Reply-To: tuscany-dev@ws.apache.org
>To: tuscany-dev@ws.apache.org
>Subject: Re: Discovery update
>Date: Tue, 20 Mar 2007 07:39:59 -0700
>
>On Mar 20, 2007, at 7:26 AM, Antollini, Mario wrote:
>
>>Meeraj,
>>
>>I am willing to help you. However, keep in mind that I am neither a
>>Tuscany developer nor a committer. Therefore you must give me a task I
>>can actually work on.
>>
>>In case you do write to me, please be very specific since I do not
have
>>much experience with Tuscany's code.
>>
>>Looking forward to your reply.
>
>I'll leave technical details to Meeraj as he's the one fighting the
>transport issue, but any contribution is welcome. For code changes,
the
>best way is to send a patch generated with "svn diff" once you have
the
>change working - either sent as a text attachment to this list, or for
>larger changes attached to a JIRA.
>
>Welcome aboard!
>Jeremy
>
>
>---------------------------------------------------------------------
>To unsubscribe, e-mail: [EMAIL PROTECTED]
>For additional commands, e-mail: [EMAIL PROTECTED]
>
_________________________________________________________________
Solve the Conspiracy and win fantastic prizes.
http://www.theconspiracygame.co.uk/
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
