[jboss-as7-dev] OSGi deployments cause missing Module for other deployment processors?

Jason T. Greene jason.greene at redhat.com
Mon Aug 22 10:23:57 EDT 2011


On 8/22/11 2:46 AM, Jaikiran Pai wrote:
> On Monday 22 August 2011 12:58 PM, Thomas Diesler wrote:
>> An OSGi bundle has a lifecycle of
>>
>> * INSTALLED
>> * RESOLVED
>> * ACTIVE
>> * RESOLVED
>> * UNINSTALLED
>>
>> The classloader is created as part of the bundle resolution phase.
>> Specifically, the OSGi resolver determines on the basis of OSGi
>> metadata a consistent wiring between the installed bundles. It is
>> therefore incorrect to create a jboss module definition for an OSGi
>> bundle at deploy time.
>>
>> The fundamental question is: Are you deploying an OSGi bundle or not?

Originally the solution we were all planning on for this problem was to 
better align the DUPs to the OSGI dep phases and use PASSIVE phase 
contexts (for everything PAST the OSGI Install [1]) to allow for the 
deployment process to complete when the bundle is finally resolved. We 
deferred this refactor though due to scheduling.

>
>    From those couple of JIRAs which have been raised, I see that the
> deployment contains a MANIFEST.MF which has some OSGi specific manifest.
>
> Is it valid case to have a deployment which has OSGi MANIFEST and at the
> same time contains EJBs? If yes, how do we handle those?

Ideally yes this should be possible, although we have had some 
disagreement on this in the past. I personally think that if an EE 
deployment uses OSGi metadata, then we should just treat that as if it 
were extra module configuration

[1] NOTE THAT OSGI "INSTALL" is not the same as DUP "INSTALL" phase. 
OSGI install is PRE module (no classloader yet).

-- 
Jason T. Greene
JBoss AS Lead / EAP Platform Architect
JBoss, a division of Red Hat


More information about the jboss-as7-dev mailing list