geronimo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dain Sundstrom <d...@coredevelopers.net>
Subject Re: DeploymentPlanner - Base class
Date Wed, 05 Nov 2003 01:10:51 GMT
On Tuesday, November 4, 2003, at 02:30 PM, gianny DAMOUR wrote:

> Dain Sundstrom wrote:
>>> 	I thought we were going to use the JMX ObjectNames to identify
>>> deployments, rather than keeping an actual repository somewhere?
>>
>> That is what I thought.  We can easily find all deployments with an 
>> ObjectName pattern query.
> Grrrrrrrr. And I really mean it: I sent a memo "JSR77 -> JSR88" on 
> Thu, 09 Oct 2003 14:02:15 +0200 proposing this idea. Excerpt of this 
> pattern:
>
> [...]
> ApplicationDeployer needs to be notified in order to track the state
> (running, stopped), the type (EJB, WAR, EAR et cetera) and various 
> other
> pieces of information.
>
> Multiple solutions could be implemented in order to "sync" the JSR77 
> and
> JSR88 models.
>
> And this is a possible one only based on naming conventions:
> ...
> This way, ApplicationDeployer "just" have to query the MBeanServer to
> retrieve the running ModuleType.WAR modules.
> [...]
> It is rather clear that I wanted to use naming conventions to identify 
> deployments and their states.
>
> Excerpt of a response to this idea from Aaron:
> [...]
> While naming conventions are convenient, I don't want to rely on that 
> for basic
> operational requirements.
> [...]
> As a conclusion, I have duplicated the deployment repository and I am 
> implementing what is required to identify the running or available 
> deployments w/o having to poll the MBenServer.

Doh.  I was out of town so I didn't see you email.  We absolutely use 
naming conventions for services.

*:role=DeploymentUnit,type=<service type>,url=<the url to the 
deployment>

If something wants to be a deployment unit, it better use a name that 
matches that pattern or we will ignore it.

>>> 	I disagree with your assertion that the server should pull code on
>>> a distribute -- among other things, that assumes that the server can
>>> freely contact that client, which I don't think is necessarily true, 
>>> or a
>>> good idea.  However I do agree that the distribute call shouldn't 
>>> block --
>>> I asked earlier if there was a way to stream data to the server over 
>>> JMX
>>> and got no answer, so I think we need to set up a standard servlet to
>>> receive the data or something.
>>
>> Agree.  We need some sort of bulk file transfer system.  I think we 
>> should really look at using WebDav for this.
> I also agree. So let me re-phrase: I will write a task to bulk 
> transfer a file from a remote host. This task will use under the cover 
> WebDAV. My point regarding the "server push" vs. "client push" idea is 
> that it is up to the server to mount a WebDAV fie system and not the 
> client to do so.

I doubt that a server will have access to mount a client's file system. 
  Clients are normally protected by firewall from a server.  Even in a 
data center you sometimes have servers on a separate physical or 
virtual lan from you admin machines.  Having the client mount the 
server webdav directory and push a deployment is way more likely to 
work.

>>>> At the end of the day, I believe that DeploymentController should 
>>>> not be the
>>>> entry point to action a start, stop, redeployment or undeployment. 
>>>> Why?
>>>> Because one knows, which planner has mounted a deployment and hence 
>>>> one can
>>>> bypass DeploymentController.
>>
>> You are assuming that only one planner was involved in the 
>> deployment.  What about ear and other meta deployments.
> My idea is that a ear containing say a rar and 2 war will be 
> represented by the following deployment-units:
> - one ear deployment unit;
> - one rar deployment unit; and
> - 2 war deployment units.
>
> When I write deployment unit, I mean a deployment meta-data 
> repository. The rar and war units are children of the ear one. When an 
> ear is deployed, a ear planner mount the ear unit. This planner then 
> calls the relevant planners for the rar and war units. These planners 
> create units, which are children of the ear unit. When a start action 
> is triggered, the ear unit requests to the ear planner to perform its 
> job. Then all the children units are started. For instance, it means 
> that a start action is triggered on the rar unit, which requests to 
> its planner to perform its job.

Agree, except you would call startRecursive.

>>>> BTW, I tried to refactor ServiceDeploymentPlanner and this is not 
>>>> so simple:
>>>> ServiceDeploymentPlanner is a kernel service. if one wants to align 
>>>> it with
>>>> the proposed approach, then one must re-package a lot of classes 
>>>> from the
>>>> core sub-project to the kernel sub-project. For instance, as the 
>>>> proposed
>>>> approach defines a class which extends AbstractManagedContainer, 
>>>> one needs
>>>> to re-package the Container, Component, AbstractManagedComponent et 
>>>> cetera
>>>> classes.
>>
>> Why is ServiceDeploymentPlanner a Container now?  I dislike the idea 
>> of pulling this stuff into kernel unless it is absolutely necessary.
> I agree. I assume that ServiceDeploymentPlanner is a container because 
> it contains the mounted services.

If we make the change I suggested at the top, it no longer tracks 
mounted services and becomes a just a planner again.

-dain


Mime
View raw message