Hey Andrus-- Thanks for the response and sorry for the delay in getting back to you.
Good point about the local context, that makes sense and I should have seen that. So my source studying took me to EventManager and DataChannelListener. It *looks* like the GraphEvents should be communicated over the JMS EventBridge but I wanted to confirm that with you. Assuming so, I think I can see how I could respond to those events similarly to how I was trying to respond to lifecycle events. Thanks again for your help-- Dave On Thu, May 26, 2011 at 12:26 PM, Andrus Adamchik <and...@objectstyle.org>wrote: > Good question... There are 2 things that are easily clustered out of the > box: > > 1. Snapshot events (events used internally by Cayenne to synchronize object > state). Those are triggered by ObjectContext commit operations. > > 2. Cache group events used to invalidate query cache groups across the > cluster. Those are triggered manually by the application code by calling > QueryCache.removeGroup(..) if the underlying QueryCache supports clustering. > In 3.1 we provided some utilities on automating of these events: > http://svn.apache.org/repos/asf/cayenne/main/trunk/framework/cayenne-lifecycle/src/main/java/org/apache/cayenne/lifecycle/cache/CacheInvalidationFilter.java > > Lifecycle events are not clustered. Object callbacks are local to > ObjectContext (so peer objects in other contexts in the same app do not > receive them), so clustering them makes no sense. Lifecycle *listeners* > (that are not persistent objects) are only notified of events on local VM > objects. But the listeners can generate their own clustered events of course > (CacheInvalidationFilter mentioned above is one example of this). > > Andrus > > > On May 17, 2011, at 4:15 PM, Dave Lamy wrote: > > Hey guys-- > > > > In a clustered cayenne environment (I'm using the JMS event bridge), do > all > > nodes receive all cayenne events (talking preInsert, postUpdate, etc > etc). > > Only some of them? None of them? I just realized I may be making some > > massive assumptions here. > > > > Thanks, > > Dave > >
