0 Replies Latest reply on May 11, 2011 1:44 PM by Adam Fitzgerald

    Synchronous cache & consistency

    Adam Fitzgerald Newbie

      Hi,

       

      I have some questions about the behavior of a clustered (sync replicated), configured with a sync singleton store CacheLoader.  The cache would be used transactionally and we want to ensure as high a degree of consistency as possible, both between cache instances within the cluster and the persistent store.

       

      The modifications to the cache are replicated upon commit, are changes replicated to the other instances in the cluster sequentially or concurrently? Is the commit failed if the replication fails to some instances?  Are the changes to the successfully replicated instances rolled back?

       

      With the CacheLoader configured as a singleton store with sync writes (and sync replication), if the cache instance elected to do the writing fails what happens to transactions that try to commit during the failover/election period (do they fail)? After the election, with pushStateWhenCoordinator=true, are transactions blocked during the writing of in memory state to the persistent store?  If not why is there a timeout?

       

      Thanks in advance!

       

      Adam

      Hi,

       

      I have some questions about the behavior of a clustered (sync replicated), configured with a sync singleton store

       

      CacheLoader.  The cache would be used transactionally and we want to ensure as high a degree of consistency as possible, both

       

      between cache instances within the cluster and the persistent store.

       

      The modifications to the cache are replicated upon commit, are changes replicated to the other instances in the cluster

       

      sequentially or concurrently? Is the commit failed if the replication fails to some instances?  Are the changes to the

       

      successfully replicated instances rolled back?

       

      With the CacheLoader configured as a singleton store with sync writes (and sync replication), if the cache instance elected

       

      to do the writing fails what happens to transactions that try to commit during the failover/election period (do they fail)?

       

      After the election, with pushStateWhenCoordinator=true, are transactions blocked during the writing of in memory state to the

       

      persistent store?  If not why is there a timeout?

       

      Thanks in advance!

       

      Adam