High Availability

The Varada Data Platform supports high availability (HA) as follows:

  • If the coordinator falls, all running queries are terminated, and the cluster automatically switches over to a new coordinator to accept new queries. You will need to explicitly retry any failed queries. If required, a query retry mechanism should be implemented to handle queries that fail as a result of a high availability event.
  • If any of the worker nodes fall, all running queries are terminated, and the index and cache on that node are lost. The cluster immediately begins accepting and executing new queries, and spins up a new worker node to auto-scale to the required cluster size.

📘

A worker node restart (not a node replacement) of any kind does not result in the loss of the index and cache on that node.