Recovery after deployment errors

If you encounter an error while deploying replication artifacts. You have to undo the actions performed within the deployment of a single business measures model to undo the changes.

All deployments are done in several stages, the following are the typical scenario:
  1. DDL deployment
    1. Deploy state.ddl.
    2. Deploy runtime.ddl.
    3. Deploy datamart.ddl.
  2. Data movement services deployment
    1. Deploy State_to_Runtime_setup_source.
    2. Deploy State_to_Runtime_setup_target.
    3. Deploy Runtime_to_historical_setup_source.
    4. Deploy Runtime_to_Historical_setup_target.
You must identify at what point the failure has occurred to determine how to take action. For example, if the state.ddl fails, then it is simply a matter of rolling back the transaction to get back to the original state. However, if the datamart.dll fails, then rolling back the datamart.ddl will only get the system back to the point after the runtime.ddl executed successfully. Failures in the middle of the data movement services deployment are the most difficult to recover from, however it is not impossible. First deployments are the easiest to recover from, deployments of new models are next, and finally deployments of change models represents the hardest recovery paths.

To recover from the replication scripts deployment errors, you go through the following stages: identifying, backing up, restoring or removing, and redeploying:

Identifying
Backing up
Restoring or removing

Redeploying

When you have removed all the artifacts supporting a business measures model, you can run the Schema Generator again with the Ignore Older Deployments option selected. If the generated schema succeeds, do not deploy the Database Definition Language (DDL) scripts; deploy the replication scripts again.


Copyright IBM Corporation 2005, 2006. All Rights Reserved.