Do any of the forms of clustering perform re-balancing of redundant data if a node goes down? Take for example the distributed cache: https://docs.jboss.org/author/display/WFLY8/Infinispan+Subsystem#InfinispanSubsystem-%3Cdistributedcache%3Eextends%3Cabstractsharedcache%3E Let's say I define each key needs at least 2 owners. Node A goes down. Keys K, L, M, and N were stored on it. Will the other nodes that hold the remaining copies of those keys then make new copies onto the remaining nodes? This way, even if Node A is down for a long time, the cluster is considered to have "healed" after some time, since every key now has a copy on at least two remaining nodes.
Re-balancing also goes in the other direction. If the cluster was fine, but now I add nodes X, Y, and Z. Will only new keys get copies to these new nodes? Or will an algorithm start redistributing the key copies on the fly, to make the cluster as a whole probabilistically more resilient?