>Hi
>
>Our team is very new to VFP and we have a big concern in terms of data recovery especically in a multi-users environment.
>
>We did a few tests trying to understand how good VFP recovers from a power failure. For example, reboot the system or kill the application during the transaction. Unfortunately we've already got a few corruptions.
>
>-What is the best way in VFP to prevent the data corruption?

The single best method is to protect all involved systems, workstations and servers alike, from power problems. With the price of small uninterruptible power supplies adequate to power a workstation now well under the $100 mark, the cost of preventing the problem in the first place is cheaper than the solution to doing something after the power failed.

>-How can we gracefully recover from the crash?

Minimize the time that data is in flux; this means that you minimize the time that a workstation can hold data without refreshing the common data source on the server, and minimizing the time that the server delays writes to disk storage. The mechanism for the first involves acquiring data locks on an as-needed basis and wrapping changes in a transaction that must be explicitly committed to update the store (look at the buffering and transaction-related commands in VFP), the latter involves managing the server properly, and varies from environment to environment.

If there is a real need to enhance data reliability and integrity, the use of a backend such as SQL Server with transaction logging and jornalling is the real answer. Unfortunately, the cost of using a backend may not be in line with the money available to support the application - in an Enterprise environment, it's not uncommon to see server software run well over the $10K mark.

>-We know they are some tools which helps to fix the corruption problem. Which one is the best choice?

It depends on the specific nature of the data corruption, and what means you've provided to recover from errors. Without knowing what type of data corruption occurred, or what means you have for identifying the last known stable state of the database, it's impossible to answer this question adequately.