|
There is a small potential for that. IBatis allows for you to configure your cache to flush when a statement is called. So, in a case where someone was performing an update that flushed a cache and then called a select that repopulated that same cache in the same transaction which subsequently failed, then that there would be invalid cached objects. I suppose we could fix that by retaining a session scoped cache for that particular transaction/thread and committing it to the global cache upon completion of the transaction.
Brandon
>>> [EMAIL PROTECTED] 11/30/2004 4:36:24 PM >>> I was wondering if caches get cleared once a transaction has failed or
has been rolled back? If not, couldn't they contain invalid entries? As caches seem to be global, what about isolation of different concurrent transactions from each other? Oliver |
- Caching and transactions Oliver Zeigermann
- Brandon Goodin
