6 Replies Latest reply on Sep 12, 2007 3:50 PM by starksm64

DeployerConfig vs. WebMetaData

brian.stansberry Sep 12, 2007 2:57 PM

Wanting to clean up some stuff related to clustered webapp configuration, but first want to be sure I understand how things are meant to work.

Looking at DeployerConfig and WebMetaData, it seems like there's some overlap in their usage. Both contain some metadata about a web app. AFAICT, DeployerConfig should contain information that solely tells the deployer how to do the deployment, while WebMetaData would be the place to put things that the webapp itself would need to understand. Metadata that the deployer doesn't care about but the webapp does should be stored in WebMetaData.

DeployerConfig has a number of clustering-related properties that based on the above should be stored in WebMetaData rather than in DeployerConfig:

cacheName
useJK
useLocalCache
snapshotMode
snapshotInterval

This will eliminate a bunch of code where TomcatDeployment is passing these values into the clustered session manager one at a time. Instead they all would get passed with WebMetaData.

Am I wrong-headed here?

1. Re: DeployerConfig vs. WebMetaData

starksm64 Sep 12, 2007 3:02 PM (in response to brian.stansberry)

That is right. Really, the DeployerConfig is obsolete as the notion of multiple web deployers from a common core is not a reality we have supported for years now. The other point of the current separation was better class loader isolation, but with the new class loading model that will also be obsolete.
Actions
2. Re: DeployerConfig vs. WebMetaData

anil.saldhana Sep 12, 2007 3:05 PM (in response to brian.stansberry)

So, either the metadata or the deploymentunit sitting in the WebApplication are the current bridge from the Deployer to a Deployment in the web space?
Actions
3. Re: DeployerConfig vs. WebMetaData

starksm64 Sep 12, 2007 3:12 PM (in response to brian.stansberry)

The DeploymentUnit should just the way for accessing deployment specific metadata, the WebMetaData being the typical piece. At some point we should just completely pojoize the web deployer layer and throw out the legacy stuff.
Actions
4. Re: DeployerConfig vs. WebMetaData

brian.stansberry Sep 12, 2007 3:19 PM (in response to brian.stansberry)

For the clustering bit, http://jira.jboss.com/jira/browse/JBAS-4708 .
Actions
5. Re: DeployerConfig vs. WebMetaData

anil.saldhana Sep 12, 2007 3:25 PM (in response to brian.stansberry)

Is it right, if I say that the WebMetaData is the glorified view of settings emanating from web.xml, jboss-web.xml and deployer custom settings?

I am still uncomfortable with the deployer custom settings leaking into the WebMetaData. DeployerConfig was for that particular reason.
Actions
6. Re: DeployerConfig vs. WebMetaData

starksm64 Sep 12, 2007 3:50 PM (in response to brian.stansberry)

WebMetaData should be the application specific metadata. According to Brian the DeployerConfig elements he is talking about are really web app specific. Technically there should not be anything other than config like what the descriptor names are, packages to exclude from the web app class loader config, etc. that should be at the deployer config level that would be concerns with regard to leaking to the WebMetaData. If they did it would just be noise.

Its reasonable to have WebMetaData defaults at the deployer level though. This is no different than the shared web.xml/context.xml descriptors.
Actions

Go to original post