If you can, before doing any work with the cube, create the data source on
its own pointing to the 'development' database. Then copy and paste the data
source to the same folder, renaming it to something neutral when prompted
with "already exists" (e.g. no server name in the title perhaps just the db
name) and delete the original data source.

Then when you create your cube, just use the neutral data source.

To deploy elsewhere: make sure that you have all of the locations loaded in
to the analsis manager and then it's just a case of copy and pasting the cube.

Once pasted you can then edit the datasource to point at something else
(e.g. UAT, production)

Reprocess, job done!

problems that may occur

if you didn't name your dsn with a neutral name then you can still use the
above method but your dsn will look odd as from the outside it will look like
"devServer - devDB" but inside will in fact be pointing at production.

Sometimes it gets upset about this, especially when dealing with multiple
data sources and can respond with the error: "a data source with a similar
name already exists but has a different connection string" or some such
nonsense

If this occurs, open up the offending data source and change it to point at
a different database entirely (different server)

Do the copy and paste part, then change the offending data source back to
what it was.

Reprocess, job done!

This saves messing about with the very ropey XML DSO which uses the
microsoft PissPoor(tm) technology for the documentation