Cube Swap - Error Handling

[kado]

Transformer Gurus,

We have a batch script that is called after our ETL loads finish which triggers our cube builds and then publishes them (pretty standard I think). Unfortunately though, if the cube build fails for any reason or is corrupt, it still gets published.

We'd like to embed some error handling such that if the build chokes for any reason that the new cube is not published. IBM recommends using their approach (http://publib.boulder.ibm.com/infocenter/cbi/v10r1m0/index.jsp?topic=%2Fcom.ibm.swg.im.cognos.ug_cogtr.10.1.0.doc%2Fug_cogtr_id10824CopyAndActivatePowerCube.html) but I'm pretty certain it's doing the same thing we are with our batch scripts (even more generic though).

Does anyone know if this approach will keep corrupt cubes from being published? Or another approach for Error Handling?

Thanks,
Kado

[dcsoke]

This won't keep corrupt cubes from being published, however, by keeping atleast one older version you can always revert back.

If you have the audit db running then there are options. An error check that you may want to try is to setup something similar to what I've done below.
1. Create a simple report that accesses each cube and shows the update date/time on the report.
2. Have this report run incrementally throughout the time range of when cubes publish.
3. Setup a second report that uses the audit db and checks for any report failures. If failures exist then email the error details to you so that you'll know which cube has the issue. If failures don't exist then do nothing.

[kado]

That's a good idea, Thanks!