>Hi All,
>
>Our Application Support came to me to ask what could be wrong. We have a big FP 2.5 DOS app which works for about 8-9 years. They decided to test it on the new server. It is Pentium 3 dual processor 1 Ghz machine with 512 mb RAM running Windows 2000 Server, and the data are located on Compaq Storage Area Network.
>

Is the app being run on the server itself, or from a workstation that uses data residing on the new server? The problem tends to suggest widely diffeent issues for each situation - if running from a workstation working using the new server for data access, I'd generally look at sharing behaviors - things like optimistic locking, release of cache data when each access handle is closed (rather than when the connection for the workstation is redefined) and full flushing of the cache and share tables when each resource handle is closed, rather than when the system is idle or requires the cache to be released before it gets reused for something else. Running on the server, issues of the relative priority of foreground and server processes, the tendency for DOS apps to use far more cycles while forced to wait for external events like reads, writes and events, issues with local reference using DOS file services rather than Windows services and such.

>What happens is that sometimes (often enough) they get Error 1705 - File Access is Denied (all tables, naturally, are shared). Restarting the workstation does not release the file, only rebooting the server helps.
>

Have you examined the server's lock table to ensure that the connection is cleanly released? It's pretty safe to bet you are relying on drive mapping, probably with persistent mappings. This may allow the server to persist cached data and file handles until the workstation explicitly breaks the mapping. If you do operate in this way, try explicitly mapping and unmapping the network drive letters using the NET USE command in a batch file, to request a connection on program startup, and releasing the connection at completion.

>All application settings (config.fp and .nt, etc.) are identical to those on the running production server, which is Pentium single processor 333 MHz, Windows NT 4.0 Service Pack 3. All production data are located on local drives on this server.
>
>Anybody have any ideas? Could it be related to dual processor configuration?

It's unlikely that the dual processor configuration is at fault; if you suspect this, you might try to force Win2K to restrict processor affinity for a task to a single processor.