Currently, isolation level is per cache instance. So you will need to have two separate cache instance to what you want.
Region based is in the roadmap though.
Doh. Perhaps I'll write my own method then...
the problem is I have an asycrounous thread that checks every n milliseconds for cache modifications (determined by a cache listener setting a boolean value). If cache has been modified I then do a recursive deep copy of the entire cache into our own tree of hashmaps and publish the "snapshot" of cache to another application using JMS.
Everything else needs to tree reads to this cache as Repeatable_Read. I only want specifically the method thats doing the deep copy to ignore lock writes. Becuase right now I get exceptions from the transaction manager if a write based cache interaction is going on while I am doing the deep copy.
I need to think on this a bit more... If I do implement my own method I'll be sure to submit a path to you guys so you can incorporate it if you want.
This may be covered by the JIRA task below. It will allow you to override selected cache properties per method call, e.g. sync-async repl, lock acquisition timeout etc: