Skip to content

nodetool gossipinfo

Displays the gossip information for all nodes in the cluster as seen by the local node.


Synopsis

nodetool [connection_options] gossipinfo

Description

nodetool gossipinfo shows detailed gossip state for every node known to the cluster, including:

  • Node status and state
  • Generation and heartbeat information
  • Schema version
  • Datacenter and rack assignment
  • Token ownership
  • Release version

This is essential for troubleshooting cluster communication and state issues.


Output Example

/192.168.1.101
  generation:1699574400
  heartbeat:12345
  STATUS:16:NORMAL,-9223372036854775808
  LOAD:12340:2.56789E11
  SCHEMA:10:a1b2c3d4-e5f6-7890-abcd-ef1234567890
  DC:8:dc1
  RACK:9:rack1
  RELEASE_VERSION:7:4.1.3
  RPC_ADDRESS:3:192.168.1.101
  NATIVE_ADDRESS_AND_PORT:12:192.168.1.101:9042
  NET_VERSION:1:12
  HOST_ID:2:a1b2c3d4-e5f6-7890-abcd-ef1234567890
  TOKENS:11:<hidden>

/192.168.1.102
  generation:1699574500
  heartbeat:12346
  STATUS:16:NORMAL,-3074457345618258602
  LOAD:12341:2.48789E11
  SCHEMA:10:a1b2c3d4-e5f6-7890-abcd-ef1234567890
  DC:8:dc1
  RACK:9:rack2
  RELEASE_VERSION:7:4.1.3
  ...

Output Fields

Node Identification

Field Description
IP Address Node's broadcast address
HOST_ID Unique node identifier
RPC_ADDRESS Address for client connections
NATIVE_ADDRESS_AND_PORT CQL native protocol endpoint

Generation and Heartbeat

Field Description
generation Epoch timestamp when node started (resets on restart)
heartbeat Counter incremented each gossip round

Status Information

Field Format Description
STATUS version:state,token Node state (NORMAL, LEAVING, etc.)
LOAD version:bytes Data size on node
SCHEMA version:uuid Schema version

Location

Field Description
DC Datacenter name
RACK Rack name

Version Information

Field Description
RELEASE_VERSION Cassandra version
NET_VERSION Protocol version

Interpreting Status Values

Node States

State Meaning
NORMAL Operating normally
LEAVING Decommission in progress
LEFT Has been decommissioned
MOVING Moving to new token
BOOT Bootstrap in progress
hibernate Node is hibernating

Generation Number

generation:1699574400

The generation is the Unix timestamp when the node started. Uses include:

  • Identifying node restarts (new generation = restart)
  • Gossip conflict resolution (higher generation wins)

Heartbeat

heartbeat:12345

Increments every gossip round (~1 second). Used to detect:

  • Node liveness
  • Gossip communication problems

When to Use

Troubleshoot Node Communication

nodetool gossipinfo | grep -E "^/|STATUS|heartbeat"

Identify nodes with stale heartbeats.

Verify Cluster Membership

nodetool gossipinfo | grep "^/"

List all nodes known to gossip.

Check Schema Agreement

nodetool gossipinfo | grep SCHEMA

All nodes should have the same SCHEMA UUID.

Investigate Topology Issues

nodetool gossipinfo | grep -E "DC|RACK"

Verify datacenter and rack assignments.

Debug Node Not Joining

# Check if new node appears in gossip
nodetool gossipinfo | grep <new-node-ip>

Examples

Basic Usage

nodetool gossipinfo

Filter for Specific Node

nodetool gossipinfo | grep -A20 "/192.168.1.102"

Check All Status Values

nodetool gossipinfo | grep STATUS

Find Nodes Not in NORMAL State

nodetool gossipinfo | grep STATUS | grep -v NORMAL

Compare Schema Across Nodes

nodetool gossipinfo | grep SCHEMA | sort | uniq -c

Check for Stale Nodes

# Nodes with old generations or low heartbeats may be stale
nodetool gossipinfo | grep -E "generation|heartbeat"

Common Issues

Node Shows OLD Generation

If a node's generation is much older than others:

/192.168.1.102
  generation:1609459200  (old)

This node may have stale gossip information. Try:

# On the problematic node
nodetool gossipinfo  # Compare with other nodes

Inconsistent Schema

nodetool gossipinfo | grep SCHEMA

Different SCHEMA UUIDs indicate schema disagreement. See describecluster for resolution.

Node Stuck in LEAVING

STATUS:16:LEAVING,-9223372036854775808

Decommission may have failed. Check logs and consider:

# If node is dead and stuck
nodetool assassinate <ip-address>

Ghost Node (LEFT but Still Appearing)

STATUS:16:LEFT,-9223372036854775808

Node was decommissioned but still in gossip. Usually clears automatically, but if persistent:

nodetool assassinate <ip-address>

Missing Node in Gossip

If a node doesn't appear:

  1. Check if node is running
  2. Check network connectivity
  3. Verify seed configuration
  4. Check firewall rules for gossip port (7000/7001)

Gossip Troubleshooting Flow

Check If No If Yes
Node appears in gossipinfo? Check network/firewall, verify seeds Continue to next check
Heartbeat increasing? Node may be frozen - check GC and resources Continue to next check
Status is NORMAL? Check for stuck operations, may need intervention Continue to next check
Schema matches other nodes? Wait for propagation or reload schema Node is healthy

Gossip vs. Status

Command Shows Use For
gossipinfo Full gossip state, all fields Debugging, detailed analysis
status Summary view, key metrics Quick health check

Use gossipinfo when status doesn't provide enough detail.


Command Relationship
status Summary cluster status
describecluster Cluster metadata and schema
ring Token distribution
enablegossip Enable gossip
disablegossip Disable gossip
statusgossip Check if gossip enabled
assassinate Force remove stuck node