Skip to end of metadata

Go to start of metadata

Versa Analytics supports high availability in a cluster for Analytics and Search personality. Following is true if the cluster is upgraded to a 22.1.4 hot fix dated Dec 2024 or higher and steps from https://docs.versa-networks.com/Solutions/System_Hardening/Perform_Manual_Hardening_for_Versa_Analytics#Change_Analytics_Database_and_Service_Passwords are performed on the cluster.

Versa Analytics platform consists of Analytics database (powered by apache cassandra), Search engine (powered by apache solr) and Zookeeper for performing resource management for solr cloud. We will discuss on the HA of each of these components below. Please note, this guidance is specifically for disaster recovery. Nodes which go down should be recovered in near future. Running a cluster with nodes down over longer period of time will lead to a degraded performance and is not at all recommended.

Guidance on size of cluster: https://docs.versa-networks.com/Getting_Started/Deployment_and_Initial_Configuration/Headend_Deployment/Headend_Basics/Hardware_and_Software_Requirements_for_Headend

Analytics (cassandra):

  • To perform HA, we recommend 2 or 4 nodes with Analytics personality. 
  • A cluster with 2 analytics db nodes can handle 1 db node going down as part of disaster recovery.
  • A cluster with 4 analytics db nodes, cluster can handle 1  db node going down as part of disaster recovery.

Zookeeper (needed for Solr):

  • Versa recommends running zookeeper on odd number of nodes, 3 or 5 such nodes.
  • With 3 nodes running zookeeper, a cluster can handle 1 zookeeper node going down as part of disaster recovery.
  • With 5 nodes running zookeeper, cluster can handle 2 zookeeper node going down as part of disaster recovery.

Search (solr):

  • To perform HA, we typically recommend 2 search nodes.
  • With 2 nodes, cluster can handle 1 node going down as part of disaster recovery.