Run a cluster

For fault tolerance and higher throughput, run multiple nodes. The c8ctl plugin handles the wiring — c8ctl nano start 3 --rf 3 brings up a 3-node, Raft-replicated cluster and configures every node for you. Use c8ctl nano status, logs, pause, and resume to operate it.

A few rules of thumb if you configure topology yourself:

• Partitions ≥ nodes, always. Each partition has exactly one writer node; the clean choice is one partition led per node. The partition count can't be changed in place — start fresh to change it.
• Replication factor = copies per partition. RF=1 (default) is no replication; RF=3 survives one node loss. Keep RF ≤ node count and use an odd voter count.
• Throughput scales with led partitions, not nodes alone — each partition is a single writer, so aggregate write throughput tracks the number of led partitions.

Durability and replication tiers are chosen at startup from the on-disk log; switching tiers on an existing data directory is unsupported. Start each reconfigured cluster from a fresh data directory.