> Clearly it's trying to use an OID it calculated for one of these
> tables after the table has been dropped, and I suspect that the lock
> is released between gathering the data and sorting it. I don't have
> any 8.2 databases around to try this on, but perhaps you would avoid
> it with a slight rearrangement of your monitoring query:
Thanks for both making this clear and providing an rearranged/modified
version of my original query. I'll try out this option also.
> If that doesn't do it I might try adding zero to numbers and
> concatenating empty strings to try to prevent late use of the OID.
> (Essentially as a form of optimization barrier.)
I couldn't understand this approach clearly. Can you help explain me with
some example?