Best-Effort durability semantics

[mnunberg]

Currently the Python SDK allows one to specify a -1 value for either persist_to or replicate_to. When this is specified, the SDK will attempt to use the maximum number of servers/replicas available for durability requirements.

In pseudo code, it might look like the following

def get_durability_requirements(self, persist_to, replicate_to):
  max_persist = self.replica_count + 1
  max_persist = min(max_persist, self.nodes_online)

  max_replica = max_persist - 1

  if persist_to < 0:
    persist_to = max_persist
  if replicate_to < 0:
    replicate_to = max_replica

  return persist_to, replicate_to

This basically allows users to worry less about the topology of their cluster, with the assumed risk that if there are no replicas online then they would be running without a replica copy.