Hello, The environment we work in is Windows (no luck getting a *ux), and our applications run in four different areas: development, testing, quality assurance and production. All these environments are alike, except for some parameters which are defined at build time.
I was disappointed by the fact that it is not possible to share jobs and transforms between repositories. The job I created is only dependent upon one Oracle connection and one SQL Server connection. What I really wanted to do was to set up 4 repositories, each which their own connections (same name, different contents) and then use shared jobs and transforms so that a single code base could be used for all four environments.
Alas, no such luck. I suppose that when working on Linux/Unix it is possible to do this, by defining jobs and transforms in one directory, and then adding a symbolic link to this in the different repositories. Now I am really into cut-and-paste land: copy the complete environment in a repository, update the connections, and when an error is discovered somewhere it must be manually applied across all environments.
I think it would already help if it was possible to the name of a connection in a variable and have the possibility to select a variable in a connection field. It would then at least be possible to define a completely parametrised ETL job.
So, any practical advice on this? We do have the Pentaho Kettle Solutions book by Casters/Bouman/van Dongen. While a good book, it does not really go in deep in such issues.
Regards, Jurgen
(And what's up with the weird formatting in vBulletin?)