I'm seeing the following behaviour in Hibernate 3.3.2 which doesn't fit my mental model of how Sessions and the Query Cache work.

I'm using two sessions from the same thread:

1. Session A: do a select using a candidate key -- but defined as <properties ... unique="true">, not using <natural-id>, and the select is done via a Query, not a Criteria
2. Session B: insert a row with the candidate key we just looked for.
3. Session A: select the new row using it's unique key -- we find it.
4. Session A: select the new row using the candidate key -- we don't find it.

I expected the insert to invalidate the query cache, but it seems it doesn't. If I add a SessionFactory.evictQueries() as step 3.5 then step 4 does find the row.

The timestamp cache is correctly configured.

What's going on?

I would suggest to activate following loggings (see above), and to report then again the output.



Correctly configured means that its entries must not be evicted ever (= eternal cache).
Do you configured the update timestamp cache as eternal?