public class LockingInterceptor
With regular NonTransactionalLockingInterceptor,
async replication does not work in combination with synchronous replication: sync replication
relies on locking to order writes on backup while async replication relies on FIFO-ordering
from primary to backup. If these two combine, there's a possibility that on backup two modifications
modifications will proceed concurrently.
Similar issue threatens consistency when the command has Flag.CACHE_MODE_LOCAL
- these commands don't acquire locks either.
Therefore, this interceptor locks the entry all the time. UnorderedDistributionInterceptor does not forward
the message from non-origin to any other node, and the distribution interceptor won't block on RPC but will return
CompletableFuture and we'll wait for it here.
Fields inherited from class org.infinispan.interceptors.locking.AbstractLockingInterceptor
cdl, dataContainer, lockManager
Fields inherited from class org.infinispan.interceptors.base.CommandInterceptor