OK, the cause is adding infinispan cluster lock jar, on DefaultCacheManager startup a cache "org.infinispan.LOCKS" with "REPL_SYNC" is requested. WHY?
Did you add the jar manually? If you want clustered locks it'd safe to assume you want to have a cluster . That's why it requests a REPL_SYNC cache.
Clustered locks use an internal cache (replicated cache by default) to be able to work. So a cluster needs to be configured. If you don't have a cluster, you don't need clustered locks.
yes i added the jar manually, but i have to explain the use case in more detail:
- we have a web application that use Infinispan as a local or replicated cache
- our custumer can decide to run the web application in standalone mode (no cluster, only one node), or in a cluster with serveral nodes (both in Infinispan embedded mode)
- on web application startup a local mode or cluster mode is activated, only depending on a infinispan.xml configuration (one with only local caches and no transport, and another one with replicated caches and an jgroups transport configuration)
- so we want to ship the web application with all the neccesarry jars (including infinispan-clustered-locks and infinispan-clustered-counter)
- with infinispan-clustered-counter we have no problem, if the application requests an clusterd-counter at some point in time (depending on the above configuration), but with infinispan-clustered-locks on classpath we have to configure allway an infinispan with transport, even if the application "does not request" an transport (only infinispan-clustered-locks creates an REPL_SYNC cache automatically, for my understanding this is a wrong assumtion in infinispan-clustered-locks)
So my wish is: let infinispan-clustered-locks work in the same way that infinispan-clustered-counter does (only create a REPL_SYN cache on demand, not in advance)
We dont want to ship our web application in two distributions (one with cluster and another on without cluster support).
With infinispan-clustered-counter it is possible, with including infinispan-clustered-locks not.
Best regards, Dario
thank you for your answer, plese take a look at my last replay to Galder.
It wold be very usefull if infinispan-clustered-locks behaves like infinispan-clustered-counter (only create a REPL_SYN cache on demand, not in advance).
Best regards, Dario