Background
A Documentum installation consists of (minimally) the content server software (docbroker, content server), database, and the content. I know there is other supporting software but for the purposes of this question this is all we need to consider.
I would like to run a Virtual Machine for each of the three:
1 VM Content Server
1 VM Database
SAN managed storage mounted to Content Server VM as filestore_01
A complete lifecycle contains three of these groupings 1 for each of development, staging and production.
So we have:
Development:
1 VM Content Server
1 VM Database
SAN managed storage mounted to Content Server VM as filestore_01 (~2TB)
Staging:
1 VM Content Server
1 VM Database
SAN managed storage mounted to Content Server VM as filestore_01 (~200GB)
Production:
1 VM Content Server
1 VM Database
SAN managed storage mounted to Content Server VM as filestore_01 (~200GB)
Currently when an upgrade is performed it is repeated for each in turn (devlopment, then staging, then production). This also go for changes to the object model, etc.
Instead, and in keeping with the benefits of virtualization, we'd like to manage the VMs as a lifecyle and promote our development server to staging (and create a new development from an image), and our staging to production, repeating this lifecycle as necessary.
Problem
Content server database and content are tightly coupled, meaning that we need to keep the content and database tied to the docbase for which they were created. If we promote staging to production we would be replacing our production database and content with staging data and content. How can both promote our virtual machines through a lifecycle but also ensure that the production server always has the correct content and metadata?
So far the only solutions that look like they could work are based on content migration after the fact, which is not good enough in my opinion. Our VMware is managed by another group in the data center and they maintain a strict hands-off approach. We have little to know access to VMware systems but if we had a strong argument for how it could be done we may be able to make it happen.
