geronimo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Paul McMahan <paulmcma...@gmail.com>
Subject Re: web clustering componentization
Date Fri, 27 Jan 2006 16:02:50 GMT
Thanks for initiating this important conversation.  As part of the
clustering effort I would like to work on a new admin portlet that can be
used to visualize the cluster topology and do things like set properties for
the session managers, deploy/undeploy applications, etc.  I think that
making this aspect of Geronimo intuitive and easy to use will be a big
differentiator in our favor.  I would love to hear everyone's feedback about
this idea.  E.g. given our current direction does this idea make sense? and
what should it look like?

If we are generally in favor of this idea then before proceeding I would
need help understanding certain fundamental concepts of our approach, such
as:

-- Will G have explicit knowledge and control over clustering or will it
just be "embedded" within the applications' logic/configuration?
-- Will G follow a deployment manager paradigm where a certain node is in
charge of managing the entire config and pushing it to the worker nodes?  Or
will each node be responsible for its own configuration?
-- What component(s) of G will provide an interface to the clustering
metadata and config?  Would it be possible to provide a single point of
control in G that can take as input a clustering "plan" and handle the
resulting deployments and node configuration?
-- Is each member of a cluster assumed to be using an identical
configuration?
-- Will G be able to participate in heterogeneous clusters with other types
of app servers and/or versions of G?

Looking forward to your feedback and advice.

Best wishes,
Paul



On 1/26/06, David Jencks <david_jencks@yahoo.com > wrote:
>
> I have some ideas about how to set up web clustering in terms of
> geronimo modules and configurations and connections between gbeans.
> I've talked with Jules and Jeff about this and think that they agree
> that this is a good starting point so I'll try to describe it and
> possibly even help implement parts of it.
>
> There are quite a few questions I'm going to ignore, such as EJB SFSB
> clustering and the new session module James and Dain wrote and how it
> might relate to the web clustering stuff.  We have to start somewhere.
>
> For me, one of the most important factors is that the clustering
> solution be pluggable, in particular that you be able to run geronimo
> without any clustering classes.  This means that each clustering
> solution needs to be packaged in one or more Configurations or
> modules and that communication to the clustering module be done
> through interfaces defined in a more core geronimo module.  It also
> implies that the all the web containers should communicate with the
> clustering implementation through the same interface(s).
>
> I'm proposing that the interfaces be SessionManagerFactory and
> SessionManager, the factory being able to provide SessionManager
> instances.  I see the SessionManagerFactory being deployed as a GBean
> and made available to web containers through a gbean reference.
>
> My understanding is that for both jetty and tomcat, the web app
> context needs an instance of a session manager.  Therefore I propose
> giving each web app context a reference to the session manager
> factory gbean.  At an appropriate lifecycle point it can get the
> session manager it needs.  In Tomcat IIUC a host or engine can also
> be where a session manager is configured.  We don't really need to
> support that directly since we can in our deployer push a reference
> to whatever "global" session manager may be configured down into each
> web app context.
>
> I see the jetty and tomcat builders installing references to
> appropriate session manager factory gbeans.  IIUC there is a lot of
> configuration possible for at least WADI, and we should be able to
> support running multiple session manager configurations at once, as
> follows:
>
> - there can be a default session manager (say WADI) configuration,
> and the name of its session manager factory gbean can be configured
> in the web app builder(s).  This will be used for distributable web
> apps that don't specify in their plan a different session manager
> factory gbean.
>
> - a web app plan can contain an explicit reference to a specific
> session manager factory or an entire session manager factory/WADI
> configuration.
>
> - we can use a namespace driven xml-reference builder to construct a
> wadi configuration inside a SMF reference (maybe :-)
>
> There are a couple loose ends I can see and probably lots more I can't:
>
> -- I imagine we want to be able to have pluggable session managers
> for non distributable apps or in the absence of clustering.  I don't
> know if we'd want 2 "default" session manager factory gbeans
> available, one for cluster and one not, or if there is a better
> solution. Having 2 and having the builder decide which you get seems
> simplest at the moment.
>
> -- We need to be able to support the non-wadi existing tomcat-
> specific clustering, and it is unlikely this can be coerced
> effectively into the proposed geronimo SMF and SM interfaces.  We
> could support this with another gbean and interface, but at the
> moment I'm leaning towards giving the web app context either a
> separate gbean reference to the tomcat specific clustering or
> directly installing the tomcat solution into the web app context.
>
> Comments?  Flames?
>
> thanks
> david jencks
>
>

Mime
View raw message