Cassandra replication system - how it works

Question

1. Does cassandra replicates only on write procedure (with choosen consistency level)? Is there any auto-replicate option for absent nodes, if i want symetric data in every node?

2. If I plug in new node to cluster there is no auto replication - how to sync data from others nodes with new one?

3. If I want replication like multimaster (2 nodes) with slave backup (1 node) known from MySQL, what is the proper logic setup and manage that on cassandra (3 nodes)? How about two nodes?

Answer 1

Cassandra replicates on writes, yes, but it also uses Hinted Handoff, Read Repair and Anti Entropy to to reduce the inconsistency window.

If you plugin a new node with auto_bootstrap set to true that node will get the appropriate data from the other nodes in the cluster.

From cassandra.yaml (storage-conf.xml in 0.6.x) about auto_bootstrap:

Set to true to make new [non-seed] nodes automatically migrate the right data to themselves.