Hello again, I've started writing a RoutingSqlMapClient that implements the SqlMapInterface and delegates all calls to Map of SqlMapClients based on a ThreadLocal. Each target SqlMapClient will have its own specific datasource.
My concern with this approach is the Spring declarative transaction handling, but I think it will work. Transactions are handled on a per-thread basis, so as along as the key to the routing sqlmapclient says the same on a per-thread call it should all work. I would really appreciate any thoughts on this. Thanks, Chris On Fri, 2007-04-06 at 10:03 -0600, Chris Lamey wrote: > Hello all, > > I'm looking at using Spring's new-ish AbstractRoutingDataSource to hit > multiple DataSources with iBATIS. I think it'll work fine in terms of > routing to the right DataSource via a ThreadLocal, but I'm concerned > about the iBATIS caching mechanism. > > Specifically, it seems to me that if I have a single > SqlMapClientFactoryBean with RoutingDataSource as the DataSource, the > cache in the SqlMapClient won't know which 'real' DataSource the results > came from and mix them all together. So if the databases in question > have the same schemas and some overlapping OIDs, the cache would be > messed up. > > Does that sound about right? Is there a way to avoid that other than > not using cache? Is there a better way to use Spring and iBATIS with > multiple DataSources? > > Thanks, > Chris
