I'm using the EHCache as a second level cache, which is working fine when I store an object, re-loading it several times in different sessions (i.e. multiple requests) uses the cache and avoids the database hit. Which is great.

However, if I update the object, it forces the cache to be invalidated and reloads the complete object graph from the database. I wouldn't expect this behaviour as hibernate should know what has changed, and so should need to invalidate the second level cache object.

Am I missing something, is is possible to ensure that objects in second level are not evicted on update (i.e. I want to stop the object being re-loaded from the database as hibernate will know what has changed and could just update the second level cahce with the changes, rather than just invalidating.

Many thanks, Pete.

evicting an object from the cache when it's updated is the way a cache should work

Seems strange to me.... I'm sure it should evict - but should it not update the object in the cache, rather than force it to re-read from the database. I've noticed that it behaves differently depending on whether you use "read-write" - I was using "nonstrict-read-write".

My object are being found in the cache (so some progress) although it hasn't stopped it also hitting the database. If anyone can help I'd be grateful for some advice, otherwise I'll carryon stepping through hibernate code.

Pete.

i think you're forgetting concurrence hell....

Yes, anthony is correct.

In certain circumstances (when it is safe, depending upon concurrency strategy, and the type of "thing" being cached), Hibernate does directly update the cache.

But remember that Hibernate is MUCH more fussy about transaction isolation than the typical cache you would build in your application!