The cache's transaction boundary for db cacheStore

Question

Hi !
I implement a CacheStore interface to cache db data,when I use cache.put() , the value can be persist into db,it's a statement level transaction, how can i control a code fragment's transaction boundary
such as:
cache.put(key1,value1);
cache.put(key2,value2);
when the last operation fail the first operation can be rollback by theory, the conclusion the other question is the cache.storeAll is handled by multi-thread , the some values has persited into db ,but one fail ,it's violates business's atomicity ,how can handle this situation?

Answer 1

Hi Stephen,
You can do the cache operations as a part of a single transaction by using a transactional cache. Something like ...
Connection conn =  new DefaultConnectionFactory().createConnection();
OptimisticNamedCache cache1 = conn.getNamedCache("tx-cache1");
OptimisticNamedCache cache2 = conn.getNamedCache("tx-cache2");
OptimisticNamedCache cache3 = conn.getNamedCache("tx-cache3");
conn.setAutoCommit(false);
cache1.insert(key1, value1);
cache2.insert(key2, value2);
cache3.remove(key3);
conn.commit();You are correct that you cannot currently use write-behind with a transactional cache. I'm not sure of your requirements but if you don't need the database updates to be part of the transaction then you could possibly do something by registering a MapListener with the transactional caches. Events occur on the transactional cache at commit time.
Another thing to keep in mind is that there are some other limitations with transactional caches and there is a performance cost. See the documentation for more details.
Thanks,
Tom