>>This error can be reported if the index is really corrupted or the part of the index that is already loaded into memory doesn't match the current index state (significant modifications have been made to the index by other clients). When VFP tries to load the rest of the index, new and old parts don't match and VFP8 reports the error. VFP7 and prior versions don't perform this kind of validation and don't report the error, however, this in-memory index corruption can have severe consequences on fast computers.
>
>>To avoid the in-memory index corruption try to do any of the following (alone or in combination):
>1) SET REFRESH to as less value as possible.
>2) Decrease buffers size using SYS(3050)
>3) Lock the file before index read operation. File lock discards all buffers for the table and its indexes the same way SET REFRESH does. Then you have a choice to unlock the table right away or after the read operation is completed.
>
>Thats good info.
>
>We are using a SET REFRESH of 2 seconds. Should I decrease it to 1 second? It may end up thrashing the server.
>
>I'll try it, but I'm wary of decreasing buffer sizes too much, as it may impact performance.
>
>Doing a file lock is not possible.
>
>The application design involved 40 clients that do a lot of reading and writing to single records (and update the keys). It always fails on a locate statement. We are experiencing about 50 failures a day. Perhaps a retry in the error handler would help.
>
>
>Thanks for the feedback.
>
>-Hans

Hi Hans,

There is another way to avoid the error - SYS(1104). It will remove all cached data from memory.

Thanks,
Aleksey.