dev@glassfish.java.net

Re: A question of deployment

From: Hong Zhang <hong.hz.zhang_at_oracle.com>
Date: Mon, 25 Feb 2013 09:07:55 -0500

On 2/24/2013 11:01 PM, Sahoo wrote:
> On Monday 25 February 2013 08:21 AM, Tang Yong wrote:
>> Hi Ancoron,
>>
>> Thanks your comments.
>>
>>> Also, GLASSFISH-16560 comes to my mind here.
>> I have not done more investigation for ear scene in depth.
>>
>> However, as to OSGi scene, just now I have done an experiment,
>>
>> Still taking the groovy-all bundle as an example, I used "asadmin osgi
>> install ..." rather than "asadmin deploy --type=osgi ..." and result is
>> that "asadmin osgi install ..." is more faster than "asadmin deploy
>> --type=osgi ..." and can be acceptable by an user although groovy-all
>> 1.8.6 bundle contains 3715 classes.
>>
>> So, I think that this issue or bottleneck should be not from OSGi comp
>> and should be from Deployment itself or Shell responsive.
>>
>> Wish that the issue can be improved as soon as possible becuase most of
>> users maybe use to deploy command.
> That's because "osgi install" does not involve annotation scanning
> unlike "deploy --type osgi." May be annotation scanning should be only
> done for Java EE application types. Hong, what do you think of this?
It's possible that we skip annotation scanning for --type osgi for
performance improvement, will this affect the deployment of hybrid OSGi
application? I think for hybrid application, you do a two-step
deployment, will the second step of the deployment calls
Deployment.deploy with a Java EE type so annotation scanning will happen
as expected? If this will not cause any issue for hybrid application,
please file a RFE for this and I will look into it.

Thanks,

- Hong