Hello!
I've learned about cube deployment on transformer a few days ago. I know it can copy the cube to one or more locations just before building it, so the cube is not locked when building: that's great. But I haver a doubt and need to be resolved before I start using deployed copies of cubes on production.
Without knowing about deploying, I create the cube using Transformer and then, if I made changes to the structure, republished it. As far as I know, creating the cube updates the data of the Cognos package, but only republishing will update the structure.
Now my question is: if I use deployment, then in cognos connection create a new data source using the cube and then a package, what will happen when changing the cube structure? Because republishing only take effect on the package published directly from the created cube, not the package I've crated manually using the deployed cube.
Transformer does not copy the cube before building it. It actually completes the cube build, then executes a separate deployment process which can optionally keep previous versions on the filesystem. In fact if the build completes but does not deploy, within Transformer you can use the "Copy and Activate" on the cube context menu to do this same process without rerunning the build.
Using the cube deployment properties will take care of the metadata updates (structure) as well as updating the data itself. Technically, I think cubes actually have embedded metadata, but in any case the deployment will do whatever needs to be done. You do not need to interactively use the Publish menu to see the changes.
Thanks for your reply. bdbits!
In the deployment properties I've selected that after build I want the deployments locations to be copied and activated. I've also configured to keep only the last 3 versions, but there's no option indicating any action that will be made to structure. But if you say the deployment will do whatever needs to be done, then I'm ok with that.
Thanks again!
@ qvixote
There was a discussion on this here a few months ago (can't find the link). PaulM stated the following (from memory):
If you are developing your cube and making structural changes, you only need to republish the cube if you still have browser windows open.
In other words, if you close all your browser windows, there's no need to republish. Next time you open Analysis Studio or whatever, you'll see your new structure.
I tested this, and, needless to say, Paul was right.
If you are talking about batch builds and deployments, then you will have no windows open, so you do not need to republish.
Many thanks, Michael75. That resolve all my doubts.
Quote from: Michael75 on 28 Jul 2014 09:48:03 AM
@ qvixote
There was a discussion on this here a few months ago (can't find the link). PaulM stated the following (from memory):
If you are developing your cube and making structural changes, you only need to republish the cube if you still have browser windows open.
In other words, if you close all your browser windows, there's no need to republish. Next time you open Analysis Studio or whatever, you'll see your new structure.
I tested this, and, needless to say, Paul was right.
If you are talking about batch builds and deployments, then you will have no windows open, so you do not need to republish.
Yes, in recent versions of Transformer, the metadata gets updated automatically. However, the doubts people have is correct. In earlier versions, the package had to be published each time a change was made in Transformer but that has been fixed now.
From what I recall in the docs on deployment, when a new cube is in deployment, all users that have an open connect to the current (old) cube version will "see" the cube as of the time they opened it until they close and open a new session. In other words, a re-publish is still not necessarily needed, but current open connections have to get flushed out before Cognos will present the updated cube. That's why the close/re-open works. I believe this applies to both data and metadata (structure).