Re: How this is possible?

I can think of two reasons why the export count would exceed the results of a select count(*) on the same table.

First, a delete operation was committed between the two events.

Second, an index scan was used to get the select count(*) results and the index is out of sync (corrupted) but the rdbms has not yet accessed a non-indexed row as part of an index update operation to find this out.

Run an explan plan on the select count(*) and if an index is used run an analyze index validate structure command on it.

But first you should probably run a select /*+ Full g */ count(*) from gravity g; and then rerun the export. If the counts now match there almost had to have been unexpected activity pending. Perhaps an implicit commit cause by exiting SQLPlus after a delete process had run but had not been committed so that a point in time view (export) still sees the "to be deleted" rows.

If the counts still are different then the analyze index might provide the answer.

HTH -- Mark D Powell -- Received on Wed Mar 02 2005 - 07:37:45 CST