> -----Original Message-----
> From: Ann W. Harrison [mailto:aharrison@...]
> Sent: Friday, December 15, 2006 12:48 PM
>
> But, for the particular bug to happen, there have to be at
> least two connections active - one removing old versions and
> one inserting new versions. The first could be a report or
> other read-only app that happens to be dragooned into garbage
> collection.

That is definitely a possibility. A user could be running a report
while the delete\load transaction is in progress. In fact, that's
exactly why we don't commit after the delete, as it takes a few
minutes to load the data (there is a bunch of lookups and ancillary
inserts that could happen during the insert).

Would doing a count(*) from the table after commit help prevent this
error from occurring by incurring the garbage collection tax itself?


Bob M..