Skip to content

nodetool setdefaultrf

Sets the default replication factor used by NetworkTopologyStrategy auto-expansion.

Synopsis

nodetool [connection_options] setdefaultrf <replication_factor>

Description

nodetool setdefaultrf configures the cluster-wide default replication factor that is applied when creating or altering keyspaces using NetworkTopologyStrategy with the replication_factor shorthand instead of explicit datacenter mappings.

How Auto-Expansion Works

When a keyspace is created using NetworkTopologyStrategy with the generic replication_factor parameter, Cassandra automatically expands this to all known datacenters:

-- Using replication_factor shorthand
CREATE KEYSPACE my_keyspace WITH replication = {
    'class': 'NetworkTopologyStrategy',
    'replication_factor': 3
};

-- Cassandra expands this to explicit datacenter mappings:
-- 'dc1': 3, 'dc2': 3, 'dc3': 3 (for all datacenters in the cluster)

The value set by setdefaultrf becomes the default when replication_factor is not explicitly specified in the CQL statement.

Arguments

Argument Description
replication_factor The default replication factor to apply. Must be a positive integer.

When to Use

Standardizing Replication Across Teams

In multi-tenant or multi-team environments, setting a cluster-wide default ensures new keyspaces follow organizational standards without requiring explicit specification:

# Set production standard of RF=3
nodetool setdefaultrf 3

This prevents accidental creation of under-replicated keyspaces when developers omit explicit replication settings.

Preparing for Cluster Expansion

Before adding new datacenters, set the default RF to ensure auto-expansion applies the correct replication factor:

# Before adding dc3 to the cluster
nodetool setdefaultrf 3

# When dc3 joins, keyspaces using auto-expansion will include dc3 with RF=3

Automation and Infrastructure-as-Code

When using automation tools that create keyspaces, setdefaultrf provides a safety net:

# Ensure minimum replication regardless of script defaults
nodetool setdefaultrf 3

Behavior Details

Node-Level Setting

This setting is per-node and is not persisted across restarts. Each node in the cluster maintains its own default RF value. For consistent behavior:

  1. Run the command on all nodes
  2. Or configure default_keyspace_rf in cassandra.yaml for persistence

Does Not Affect Existing Keyspaces

Changing the default RF does not alter existing keyspaces. It only affects:

  • New keyspaces created with replication_factor shorthand
  • ALTER KEYSPACE commands using replication_factor shorthand

Interaction with Explicit Settings

Explicit datacenter settings always override the default:

-- DC1 gets RF=5, DC2 uses the default from setdefaultrf (e.g., 3)
CREATE KEYSPACE mixed_rf WITH replication = {
    'class': 'NetworkTopologyStrategy',
    'replication_factor': 3,
    'dc1': 5
};

Examples

Set Default RF to Production Standard

nodetool setdefaultrf 3

Verify the Setting

nodetool getdefaultrf

Output:

Current default replication factor: 3

Applying Across All Nodes

# Run on each node or use parallel execution
for host in node1 node2 node3; do
    ssh "$host" "nodetool setdefaultrf 3"
done

Best Practices

Recommendations

  1. Set on all nodes - The setting is per-node; ensure consistency across the cluster
  2. Use with cassandra.yaml - For persistence, also configure default_keyspace_rf in cassandra.yaml
  3. Match node count - The default RF should not exceed the number of nodes in the smallest datacenter
  4. Production standard: RF=3 - Provides fault tolerance for one node failure while maintaining QUORUM availability

Limitations

  • Does not affect keyspaces created with explicit datacenter mappings
  • Does not retroactively change existing keyspaces
  • Resets to default (1) on node restart unless configured in cassandra.yaml

The persistent equivalent in cassandra.yaml:

# Default replication factor for auto-expanding keyspaces
default_keyspace_rf: 3
Command Relationship
getdefaultrf View current default RF
describecluster View cluster topology and datacenter information
status Check node count per datacenter