On 28/11/12 17:33, Gunnar Wagenknecht
wrote:
Hi,
Am 28.11.2012 15:20, schrieb Cristiano Gavião:
I could note that gyrex product has more
than one application, being one
visible, ok.
Do you mean OSGi applications or HTTP applications?
yep. I'm referring to an OSGi app, the ones that has
org.eclipse.core.runtime.applications extension
point(org.eclipse.gyrex.cloud, org.eclipse.gyrex.http.jetty.
but what is the rule for my own apps
creation? I mean, when should I
create an application or just add features/bundles ?
I assume you are speaking about OSGi applications. So there is
some history. First we associated roles only with bundles. This
meant that everything was tight to the bundle activator
start/stop. At some point we came up with the idea of using OSGi
applications instead of the bundles. The motivation was that they
would give us more flexibility long term. Thus, we migrated to
OSGi applications. We are now no longer tight to bundles and can
refactor applications into different bundles (if necessary).
I have no sure yet, but seems that an application (the one defined
in OSGi Application Admin specification) has no lifecycle relation
with the bundles that it depends on, right? I mean, if I start/stop
one app it not necessarily will do the same with it child bundles,
right?
I could see that the
application defined in Subsystem Service Specification do that...
but only Apache Aries have implementation for
that(http://aries.apache.org/modules/applications.html)...
how could I setup bundles that needs to
auto-start, for example
gemini-jpa and eclipse-link ?
I created:
http://wiki.eclipse.org/Gyrex/Developer_Guide/Roles
Let me know if I need to add more explanations to that page.
It would be interesting to explain the existent trigger types. they
are fixed ?
when these roles are processed ? who is responsible for that?
is there an api for the roles maintenance ?
and how to setup a default p2 url ?
should I create my own product for that?
That's something we haven't had yet. p2 URLs are stored in
ZooKeeper. The p2.inf won't work here. I think one possibility
would be to write a small initialization role that checks on cloud
connect if the data has been initialized already and inserts the
necessary data into ZooKeeper if not (using
P2Activator.getInstance.getRepositoryManager).
what is the expected behavior actually defined for provisioning ?
installing a feature in the master node will trigger the same in the
other nodes?
-Gunnar
|