Create a Single-token DSE Cluster

Overview

Carefully generating tokens is key to maintaining overall cluster balance, and not a trivial task. Before the advent of Mission Control, using previous tools or manually allocating a token were the choices for setting a token and ensuring the best cluster balance. Even so, optimal balance is not guaranteed. Mission Control eases the task of creating single-token clusters.

Single-token Cassandra clusters

each physical node owns only one token range in the token ring.

Virtual nodes (vnodes)

each node in the cluster owns two or more token ranges. Vnodes mitigate the challenge of managing unbalanced clusters, where data is not distributed evenly amongst the nodes.

Historically, and by default, tokens are assigned through a Cassandra (C*) allocation algorithm or DSE 6.8’s token-generator tool. In Mission Control, ConfigMaps aid in implementing token-assignments.

Recommended token ranges
Version # token ranges

Cassandra v3.x

256 tokens per node

Cassandra 4.x

256 tokens per node

DSE 6.8

16 tokens per node

Mission Control

Create a single-token cluster

This details how to set the num_tokens: option to 1.

Create a single-token cluster with Mission Control. This example works with a single datacenter that has a single-rack cluster with six nodes. The replication factor (RF) is set to Mission Control’s default of 3. Modify the MissionControlCluster manifest to explicitly override the default of 16 in the config:cassandraYaml:num_tokens section. From this definition, Mission Control automatically generates initial tokens.

  1. Required: Modify the MissionControlCluster manifest, explicitly specifying config.cassandraYaml.num_tokens: 1, as follows:

    apiVersion: missioncontrol.datastax.com/v1beta1
    kind: MissionControlCluster
    metadata:
      name: demo
    spec:
      k8ssandra:
        cassandra:
          serverVersion: 6.8.26
          serverType: dse
          config:
            cassandraYaml:
              num_tokens: 1
          datacenters:
            - metadata:
                name: dc1
              size: 6

    This change necessarily overrides the default number of tokens (16) that DSE sets per node.

    Do not adjust the initial_token specification in the MissionControlCluster manifest. From this manifest you cannot manually vary the configuration option based on the node to which it is applied. Let Mission Control generate the initial tokens per node for you. See Configure a DSE node.

  2. Issue the following command from a pod running DSE in the cluster and review the resulting cluster:

    kubectl exec demo-dc1-rack1-sts-0 -c cassandra -- nodetool -u demo-superuser -pw <omitted> ring
    Sample results
    Datacenter: dc1
    ==========
    Address      Rack    Status State   Load        Owns    Token
                                                             6148914691236517204
    10.100.3.9   default Up     Normal  144.48 KiB  50.00%  -9223372036854775808
    10.100.6.10  default Up     Normal  209.15 KiB  50.00%  -6148914691236517206
    10.100.1.10  default Up     Normal  118.17 KiB  50.00%  -3074457345618258603
    10.100.0.11  default Up     Normal  209.87 KiB  50.00%  -1
    10.100.2.9   default Up     Normal  170.66 KiB  50.00%  3074457345618258602
    10.100.5.12  default Up     Normal  200.67 KiB  50.00%  6148914691236517204

    The result shows a perfectly balanced ring with 6 token ranges of equal size. The Owns column reveals that each node holds 50% of the data. This is the expected data ownership because every node is replicating 3 token ranges of ⅙ each of the whole ring.

Allowing Mission Control to assign tokens requires that you do not delete or modify the ConfigMaps that Mission Control uses to create the required per-node configuration.

If you intend to use custom per-node configurations, then you must manually compute tokens per node. See Configure a DSE node.

What’s next

Rebalance an existing cluster. Assign tokens to new nodes that are joining an existing cluster.

Was this helpful?

Give Feedback

How can we improve the documentation?

© 2024 DataStax | Privacy policy | Terms of use

Apache, Apache Cassandra, Cassandra, Apache Tomcat, Tomcat, Apache Lucene, Apache Solr, Apache Hadoop, Hadoop, Apache Pulsar, Pulsar, Apache Spark, Spark, Apache TinkerPop, TinkerPop, Apache Kafka and Kafka are either registered trademarks or trademarks of the Apache Software Foundation or its subsidiaries in Canada, the United States and/or other countries. Kubernetes is the registered trademark of the Linux Foundation.

General Inquiries: +1 (650) 389-6000, info@datastax.com