nodetool is ScyllaDB’s command-line utility for inspecting a cluster and performing operator tasks. Run it locally on a node or connect over the network to view nodes, table performance, and cluster state; it can also remove nodes. section 2.1.2
Its most common command is `nodetool status`. The output groups nodes by datacenter and shows each node’s address, load, token count, ownership, host ID, and rack. The two-letter marker combines availability and membership state: `UN` means up and normal. section 2.1.2 section 2.5
Read the status marker as two independent signals:
- `U` or `D` reports whether a node is up and taking traffic or down and unhealthy. section 2.1.2
- `N`, `J`, `L`, or `M` reports whether a node is normal, joining, leaving, or moving between datacenters or racks. section 2.1.2
Watch a node join
After starting a new node with a seed node, run `watch nodetool status --resolve-ip` from an existing live node. The new node appears as `UJ` while it joins and receives its assigned data, then becomes `UN` when it is a normal cluster member. section 2.1.3
Datacenter: datacenter1 ======================= Status=Up/Down |/ State=Normal/Leaving/Joining/Moving -- Address Load Tokens Owns Host ID Rack UN scylla-1 572 KB 256 ? 75ddfa00-9624-4137-b926-429dff20e516 rack1
Use this status output as a field guide: the header defines the status and state letters, while each row identifies one node and its placement. section 2.1.2
For a Docker-based local cluster, open a shell in the node container with `docker exec -it scylla-1 /bin/bash`, then run nodetool there. This gives a quick operational check when logs are slower to inspect or a node may be unhealthy. section 2.1.2 section 2.1.1
