Calculate subnet sizes for a new workspace

When you create a new workspace using the account console, Databricks creates a Google Cloud GKE cluster. By default, Databricks decides the IP ranges for the subnets. You can optionally use advanced configurations to set subnet sizes explicitly.

Important

It’s critical to configure the GKE subnets used by your Databricks workspace accurately because you cannot change them after your workspace is deployed. If the address ranges for subnets are too small, then the workspace exhausts its IP space, which in turn causes Databricks jobs to fail or clusters fail to start.

Subnet sizes are described in this article in CIDR format.

For CIDR range requirements for network resources and attributes, see Network requirements.

Quick sizing guideline

Each Databricks workspace node requires two Google Kubernetes Engine (GKE) nodes.

The maximum nodes per workspace is the total number of concurrent nodes (compute instances) that can run in this workspace at any one point in time.

Use the following table for a quick sizing estimate by subnet sizes:

Subnet Size

Maximum Databricks nodes per workspace

  • Nodes subnet size /25

  • Pods subnet size /20

  • Services subnet size /22

60

  • Nodes subnet size /24

  • Pods subnet size /19

  • Services subnet size /22

120

  • Nodes subnet size /23

  • Pods subnet size /18

  • Services subnet size /22

250

  • Nodes subnet size /22

  • Pods subnet size /17

  • Services subnet size /22

500

  • Nodes subnet size /21

  • Pods subnet size /16

  • Services subnet size /22

1000

  • Nodes subnet size /20

  • Pods subnet size /15

  • Services subnet size /21

2000

  • Nodes subnet size /19

  • Pods subnet size /14

  • Services subnet size /20

4000