Minimum virtual machine sizing for Azure HDInsight cluster head nodes
Updated: 06 January, 2020
Beginning 6 April 2020, customers will be able to choose only 4-core or higher virtual machines (VMs) as head nodes for the new HDInsight clusters. Existing clusters will continue to run as expected.
Each Azure HDInsight cluster contains two head nodes to deploy and run critical Apache Hadoop & Spark services, as well as multiple worker nodes to perform the actual data analysis. Hadoop & Spark services running on the head nodes require sufficient computation and memory resources to support service deployment and job management operations. Our extensive experience with running production workloads indicates that, at a minimum, a 4-core VM is required for head nodes to ensure the high availability and reliability of your HDInsight clusters.