Hi all

In the new code base I'm preparing for the Karaf 3.x line I'll provide only one assembly. The other can be added later if needed.

We should decide what to do with the full assembly in the current servicemix5 codebase. The full assembly makes actually problems with upgrade to newer Karaf version. Another problem - as soon as we upgrade to Karaf version including Hibernate in enterprise features (e.g. 2.3.4 or 2.4.0) the hibernate bundles will be included in the full distribution, but it is not allowed (due to the LGPL license)

Regards
Krzysztof

On 19.02.2014 13:18, Filippo Balicchia wrote:
Hi,
i'm  full assembly user, but the solution proposed from Raul  imho
is interesting.
Say that keep the default assembly should be OK.

Regards

--Filippo


2014-02-19 12:51 GMT+01:00 Raul Kripalani <[email protected]>:

I agree with Freeman. Even if you're an avid user, you probably won't use
*all* the dependencies that we bundle in the full distribution. Conclusion:
for most users, with the full distro we're delivering a large footprint
overhead. Think of it: just by embedding all of Camel features,
transitively we include hundreds of dependencies. Even if most
organizations using Camel typically end up using around 20 components, not
more...

On the other hand, the goal of the full distro was legit: make sure that
SMX doesn't ever need access to the Internet. But the current approach is
like "killing flies with a bazooka".

What I propose is to create a Karaf command
"servicemix:bundles-persist-in-local-repo" which uses the bundle locations
of all installed bundles to copy them to the SMX system/ directory.

So basically, to create a distribution that is both (a) suitable for
offline usage and (b) customized for your organization, you would follow
this workflow:

   1. Download the default SMX distro. Start it.
   2. Install all features you will need for your organization (camel-ftp,
camel-mongodb, hibernate, activiti, etc.). This step installs transitive
dependencies too.
   3. Call the servicemix:bundles-persist-in-local-repo command, which will
persist all bundles in the system/ directory.
   4. Stop SMX.
   5. Delete the data/ directory.
   6. Voilà, you have a SMX distro adapted to your organization which is
suitable for offline usage.

We could probably group step 4 and step 5 into a single command
servicemix:shutdown-and-clean, or enhance the original Karaf shutdown
command to include a --clean switch, which deletes the data directory just
before exiting the Java process.

WYDT?

Regards,

*Raúl Kripalani*
Apache Camel PMC Member & Committer | Enterprise Architect, Open Source
Integration specialist
http://about.me/raulkripalani | http://www.linkedin.com/in/raulkripalani
http://blog.raulkr.net | twitter: @raulvk

On Wed, Feb 19, 2014 at 11:36 AM, Freeman Fang <[email protected]
wrote:
I'd say keep the default assembly should be OK, only including bundles
used by default boot features should be good.
-------------
Freeman(Yue) Fang

Red Hat, Inc.
FuseSource is now part of Red Hat



On 2014-2-19, at 下午6:46, Gert Vanthienen wrote:

L.S.,


Just noticed Krzysztof's comment about the how and why of the
apache-servicemix-full assembly in JIRA and think it's worth having a
quick discussion about which of these assemblies we want to keep
around going forward.

Right now, we have 3 variations of our assembly in ServiceMix 5:
- The default assembly has a setup that includes the base CXF, Camel
and ActiveMQ bits and preinstalls those as boot features
- The minimal assembly actually has none of these things preinstalled,
it's just a plain Karaf with the feature URLs (among a few other
things) preconfigured - not sure how much this one is used though
- The full assembly is identical to default assembly, but the /system
folder contains all bundles for all optional features.  This makes it
easy for people to use ServiceMix on servers that have no direct
connection to the internet.  This one is probably used a bit more (I
usually install this one myself), but we are struggling to keep it
below the maximum 350000000 bytes distribution size (which is already
an exception to the rule at the ASF) and as issue
https://issues.apache.org/jira/browse/SM-2241 is showing, it is prone
to be impacted by any kind of error in any of our dependencies'
features files.

For ServiceMix 5 and other upcoming versions, which of these
assemblies do we want to keep around?


Regards,

Gert Vanthienen



--
Krzysztof Sobkowiak

JEE & OSS Architect | Technical Architect @ Capgemini
Capgemini <http://www.pl.capgemini.com/> | Software Solutions Center <http://www.pl.capgemini-sdm.com/> | Wroclaw e-mail: [email protected] <mailto:[email protected]> | Twitter: @KSobkowiak

Reply via email to