I think additional table to keep hash values ect is unnecessary overhead.
May be we can have an additional directory and move the file system
deployed or uploaded bpmn artifact with a generated time stamp id to that
directory in case the developer want to refer to that archive later.

Regards
Nandika


On Wed, Feb 26, 2014 at 10:10 AM, Chathura Ekanayake <[email protected]>wrote:

>
> Yes, file system based undeployment cannot be supported with this model.
> Undeployment has to be done either using the WS API or from the UI.
>
> If we want to support file system based undeployment, we have to maintain
> an additional table containing the mapping (deployment ID -> hash).
> Furthermore, then we have to handle additional things like versioning of
> archives, synchronization, etc. So I think it is better to keep this as
> close as possible to the Activiti deployment model.
>
> Regards,
> Chathura
>
>
>
> On Wed, Feb 26, 2014 at 12:16 AM, Nandika Jayawardana <[email protected]>wrote:
>
>> This approach allows us to be aligned with activiti deployment model.
>> Also it should work without any issue in the cluster also and would not
>> require deployment synchronizer as well. Only draw back is that we do not
>> have a file system based un deployment method. But i think we should be
>> able to live with that.
>>
>> Regards
>> Nandika
>>
>>
>> On Tue, Feb 25, 2014 at 2:56 PM, Chathura Ekanayake <[email protected]>wrote:
>>
>>>
>>> Hi All,
>>>
>>> Activiti stores deployment artifacts in its database without keeping
>>> extracted versions of BPMN archives. I think it is better to follow a
>>> similar approach when implementing BPMN support in BPS. Similar to
>>> deploying BPEL packages, we can support deployments through file system and
>>> file upload UI.
>>>
>>> File system based deployment works as follows:
>>> - A BPMN deployer is started on server start up to listen for bar files
>>> in the bpmn folder
>>> - BPMN archives (i.e. bar files) are copied to the bpmn folder
>>> - Artifacts in the copied archive are persisted in the database and the
>>> archive is deleted from the file system
>>> - In a clustering environment, manager node sends cluster messages to
>>> all members to load necessary data from the database.
>>>
>>> This approach does not  cause redeployments upon server restarts as the
>>> archives are deleted as soon as the deployment is complete. Activiti
>>> assigns a unique ID for each deployment (even for deployments with the same
>>> name), so that versioning can be handled with a minor effort.
>>>
>>> We have to maintain tenant details (i.e. tenant ID -> deployment ID
>>> mappings) in BPS. When a BPMN archive is deployed, we have to update these
>>> mappings in the database. Furthermore, in a clustered setting, manager node
>>> has to inform all members about new deployments, so that members can update
>>> their tenant mappings. By following this approach, we can simplify the
>>> deployment and clustering, while aligning the deployment model of our BPMN
>>> implementation with that of Activiti.
>>>
>>> Thoughts...
>>>
>>> Regards,
>>> Chathura
>>>
>>>
>>> _______________________________________________
>>> Architecture mailing list
>>> [email protected]
>>> https://mail.wso2.org/cgi-bin/mailman/listinfo/architecture
>>>
>>>
>>
>>
>> --
>> Nandika Jayawardana
>> Senior Technical Lead
>> WSO2 Inc ; http://wso2.com
>> lean.enterprise.middleware
>>
>> _______________________________________________
>> Architecture mailing list
>> [email protected]
>> https://mail.wso2.org/cgi-bin/mailman/listinfo/architecture
>>
>>
>
> _______________________________________________
> Architecture mailing list
> [email protected]
> https://mail.wso2.org/cgi-bin/mailman/listinfo/architecture
>
>


-- 
Nandika Jayawardana
Senior Technical Lead
WSO2 Inc ; http://wso2.com
lean.enterprise.middleware
_______________________________________________
Architecture mailing list
[email protected]
https://mail.wso2.org/cgi-bin/mailman/listinfo/architecture

Reply via email to