On 6/22/07, Matthieu Riou <[EMAIL PROTECTED]> wrote:
I don't know what would be the best way from the Tuscany standpoint, but having some sort of configuration object (maybe a StAXArtifactProcessor) mirroring part of a global Tuscany configuration and available in the module activator would be nice. And you'll probably have people at some point who will ask you how they can configure this or that in Jetty or Tomcat. And yes, there will be only one ODE server instance on which processes will be deployed. The approach I used for now is to instantiate a server wrapper in the module activator and initialize the engine lazily in the provider, when the first implementation is started. Matthieu On 6/22/07, Venkata Krishnan <[EMAIL PROTECTED]> wrote: > > Is this something that we should consider permitting thro external config > files ? And how would it work - the contribution service picks the > resource > up and makes it available to the module activators ? > > Thanks > > - Venkat > > > On 6/22/07, Luciano Resende <[EMAIL PROTECTED]> wrote: > > > > I'm still reviewing the patch provided by Matthieu [1], but I think he > > is taking the similar approach we have for the embedded hosts. > > > > I think that, one difference, is that we don't necessary expose or let > > the user tweak server configuration when we start an embedded jetty or > > tomcat, do we ? > > > > [1] http://issues.apache.org/jira/browse/TUSCANY-1365 > > > > On 6/22/07, Venkata Krishnan <[EMAIL PROTECTED]> wrote: > > > Hi, > > > > > > I assume there is going to be just about 'one' ODE instance which > might > > be > > > running several processes that are implementations of different > > components > > > in a deployment. Is the ODE server something that we should implement > > on > > > lines with the 'hosts' that we have have ? Thoughts ? > > > > > > - Venkat > > > > > > On 6/21/07, Matthieu Riou <[EMAIL PROTECTED]> wrote: > > > > > > > > Hi guys, > > > > > > > > I'm starting on the lifecycle part of the Tuscany/ODE integration to > > get > > > > an > > > > engine started with all its necessary resources (transaction, > > persistence, > > > > ...). I can go with in-memory mode and light configuration for now > but > > > > we'll > > > > want a fully working server sooner or later. So I have a few > > questions: > > > > > > > > 1. How can I get server-wide configuration properties to configure > the > > > > behavior of the engine? Things like the size of our thread pool, if > we > > > > should use an embedded database or try to lookup a datasource, some > > memory > > > > optimization modes we can run in, ... These would be for the whole > ODE > > > > server and not on a per-process basis. > > > > > > > > 2. How can I get the a transaction manager? I haven't been able to > > find an > > > > extension point for it. If there are none available, I can still > > create my > > > > own but it'd be probably be better to reuse the Tuscany one if it's > > there. > > > > > > > > 3. In embedded persistence mode, ODE uses a Derby database. Which > > means > > > > files located somewhere. Is there a way to lookup a given directory > > under > > > > which the database could be made available? For example when we > deploy > > ODE > > > > in a webapp for example, the database is prepackage in the webapp > and > > we > > > > find it under the context root (or alternatively a path can be > > configured, > > > > see 1) > > > > > > > > I think that's all for now but I'll certainly find more in the > coming > > days > > > > :) > > > > > > > > Thanks! > > > > Matthieu > > > > > > > > > > > > > -- > > Luciano Resende > > Apache Tuscany Committer > > http://people.apache.org/~lresende > > http://lresende.blogspot.com/ > > > > --------------------------------------------------------------------- > > To unsubscribe, e-mail: [EMAIL PROTECTED] > > For additional commands, e-mail: [EMAIL PROTECTED] > > > > >
Hi Matthieu Can I just check that I understood what you have said. I.e. that "there will be only one ODE server instance on which processes will be deployed". Does this mean, 1 ODE server instance per Tuscany runtime node (a tuscany Java SCA runtime in a Java VM) - I think this is the case given what you say about module activator Or 1 ODE service instance per machine that handles any number of tuscany runtime nodes. I'm asking because one of the things I'm interested in is how we manage a domain running across multiple runtime nodes. As an aside, one of the things I have to configure for each node is the base url for each scheme (http, https etc.) that the runtime node is going to use to expose services. Currently this ability to configure runtime nodes is not included at all but the near term objective is to have this data just read into a local model from files on the classpath (the models and files exist but I haven't joined the dots between it and the runtime yet). Looking to the future I have to be able to provide such data out to nodes remotely. To allow for this each distributed runtime node starts a small SCA assembly purely for managing meta data, I.e. it defines services with interfaces for reading and writing specific meta data. In that way I can use any SCA binding to send data to a node. So a binding could still allow the data to be read from a (distributed) file system but it doesn't have to be. I'm not suggesting you do anything about this specifically yet but I would like to talk about it as the implementation progresses just in case we have to do anthing to keep ODE happy. Regards Simon
