Removing 'hidden' corruption

[brian rich]

One of our hosted FMP11 files recently started misbehaving, where a script that had been running reliably for months suddenly became unreliable and did not always update records as expected. Detailed examination of the scripts and the data didn't give us any pattern for the problem so we suspected data corruption.

 

Our first recourse was to run a recover on the previous nights backup. The recovery process produced an unusually large Recovery log, nearly 4Mb in size, rather than the more usual 200Kb, but still ended with the message "Recover built a new database without detecting any problems."  When we investigated the recovery log, we saw a very large number of lines in the log file   Recovering: library '' (nnnnn), where the number went from 1 to 47302 and there is nothing between the two quote marks.

 

Slightly puzzled, we ran a recover on the backup Clone which we produce each night, and this had none of the above messages in the log file, and also ended with the "...without detecting any problems" message.

 

We then ran a consistency check on the file containing the data, this reported no problems either. More puzzled.

 

We then decided to export the data from the five tables in the file to external filemaker FP7 files (one table contains 300+ fields of which 55 are repeating fields so it had to be FP7 format) and then re-import the data into a known good backup.  After importing the data back into the backup file, it appeared sound on brief review.  However when we ran a recover on the newly updated import, the Recovery log was huge, the Recovering:  library '' (nnnn) message was back. and it still said "... without detecting any problems."  The only bright point is that there were now only 46403 lines with that mysterious message in the log, but we're still not happy with the result.

 

Can anyone suggest what might be happening here and else we might be able to do to eliminate the corruption?

 

Thanks

 

Brian