<smarlow> pferraro:  I was able to recreate the reported issue with deploying two 2lc apps and undeploying on and redeploying.
<smarlow> pferraro:  I was able to recreate the reported issue with deploying two 2lc apps and undeploying one and redeploying.
<smarlow> when the app redeploys, it seems to find the global component registry from before which doesn't have the infinispan custom commands anymore
<smarlow> pferraro:  so it could have something to do with how we stop the 2lc (currently ignore the stop I think)
<smarlow> pferraro:  smells like an infinispan bug to me or maybe a bug in org.jboss.as.jpa.hibernate4.infinispan.SharedInfinispanRegionFactory.stop() which currently doesn't do anything.  I think we need to do the mirror of start()
<pferraro> doesn't do anything?
<smarlow> pferraro:  it has a comment "// Do not attempt to stop our cache manager because it wasn't created by this region factory."
<pferraro> right - the cache manager in this case is shared across multiple apps
<pferraro> so - we wouldn't want to stop it
<smarlow> oh, right, I forgot that
<pferraro> the cache instances should get stopped though
<pferraro> so long as they are restarted on redeploy
<pferraro> which the explicit Cache.start() is meant to do
<smarlow> hmm, I guess the mystery is still finding out why the hibernate-infinispan commands disappear from the global registry
<pferraro> hmm - maybe nothing is stopping the cache instances on undeploy?
<smarlow> I don't think anything is doing that
<pferraro> actuall, that's not it - Region.destroy() stops the cache instance
<smarlow> I think this was on the timestamp regiion
<pferraro> I wonder why the custom commands would be gone
<smarlow> where is the region.destroy logic?
<pferraro> ah
<pferraro> In BaseRegion
<pferraro> I see the problem
<pferraro> the CacheCommandFactory is only designed to be used with a single RegionFactory
<pferraro> ugh
<pferraro> yet it's part of the GlobalComponentRegistry
<pferraro> which is a property of the cache manager