I am looking for useful information on the subject of recovery testing. Generic test cases, ideas for fault injection to create disasters, general properties of a system that should be able to perform recovery etc.
The question is intentionally posed for a generic system since I am interested in the general knowledge, however, if there are system type (e.g., client server) specific answers or ideas, these are interesting too.
Re: Recovery Testing
Wipe out the entire database. Verify that a backup can be installed and used properly.
Drop a connection between your server and a third party server. Verify that when you attempt to restablish the connection, everything works properly.
Crash the server (or just unplug it) during a transaction. Reboot and verify that nothing got severly jacked up.
If it can rollover to a redundant server then you want cause the conditions that warrant a rollover, make sure the rollover occurs and verify that the correct data is on the rollover server (that depends on how often the data is replicated.)
Tim Van Tongeren