Cassandra Scaling

Google Cloud Apigee
1

Hi All,

We are have Apigee Setup on OPDK 4.19.01 version. I want to add additional cassandra nodes to my existing planet. I have 3 DCs which is having 3 cassandra nodes each. That mean 3 cassandra nodes per DC. Now I want to add 3 new cassandra nodes to each DC. which are of total of 9 new nodes for all the 3 DCs( Total are18 nodes with 9 old + 9 new nodes). Right now i have few queries below regarding scaling of cassandra.

  1. Can I scale only Cassandra or do I need to scale Zookeeper along with the Cassandra for the new nodes what I am planning for?

  2. Right now I have DC1 with 3 followers and the other 2 DCs with 2 followers and 1 observer each. What will be the type of nodes do i need to define for the new cassandra nodes? Is there any ratio to be followed for the voters VS observers?

Thanks,

2

  1. Scaling Cassandra: In order to add new Cassandra nodes, you only need to scale Cassandra itself. ZooKeeper does not need to be scaled alongside Cassandra. ZooKeeper is responsible for coordinating the distributed nature of Cassandra, and its scaling is typically handled separately if required.

  2. Node Types and Ratios: The node types for the new Cassandra nodes depend on your specific requirements and the desired configuration. The current setup you mentioned has DC1 with 3 followers and the other 2 DCs with 2 followers and 1 observer each. When adding new nodes, you can follow a similar pattern based on your needs.

    • Followers: Followers are fully participating nodes in the Cassandra cluster. They handle read and write operations and participate in the consensus process for data replication. You can consider adding new Cassandra nodes as followers to maintain a balanced configuration across the data centers.

    • Observers: Observers are lightweight nodes that do not participate in the write consensus process. They are helpful for offloading read traffic and maintaining a higher read throughput. If you require increased read performance, you can consider adding new Cassandra nodes as observers. However, it’s important to note that observers do not contribute to the write capacity of the cluster.

    The ratio of followers to observers depends on your specific workload and performance requirements. There is no fixed ratio that applies universally. You may need to evaluate the read and write patterns, data consistency needs, and overall cluster performance to determine the optimal ratio for your setup