What causes AWS_INSUFFICIENT_FREE_ADDRESSES_IN_SUBNET_FAILURE?

The subnet CIDR block is too small for the number of concurrent Databricks nodes being launched. Many clusters or instances are running simultaneously, exhausting the available IP range. Terminated instances are holding their IP addresses in a cooling-off period before release. Other AWS services (RDS, ECS, Lambda ENIs) in the same subnet are consuming IP addresses. A misconfigured VPC with overlapping subnets left fewer usable addresses than expected

How do I fix AWS_INSUFFICIENT_FREE_ADDRESSES_IN_SUBNET_FAILURE?

Check the subnet in the AWS VPC console to see the current available IP count. Terminate idle Databricks clusters to free up IP addresses. Use a larger subnet (e.g., /20 instead of /24) or create additional subnets for Databricks. Configure Databricks to use multiple subnets across availability zones to spread IP usage. Move other workloads (RDS, ECS) out of the Databricks subnet to free address space

High severitycluster

Power BI Refresh Error:
AWS_INSUFFICIENT_FREE_ADDRESSES_IN_SUBNET_FAILURE

What does this error mean?

The AWS VPC subnet used by Databricks has run out of available private IP addresses. No new cluster nodes can be launched until IP addresses are freed or the subnet is expanded.

Common causes

1The subnet CIDR block is too small for the number of concurrent Databricks nodes being launched
2Many clusters or instances are running simultaneously, exhausting the available IP range
3Terminated instances are holding their IP addresses in a cooling-off period before release
4Other AWS services (RDS, ECS, Lambda ENIs) in the same subnet are consuming IP addresses
5A misconfigured VPC with overlapping subnets left fewer usable addresses than expected

How to fix it

1Check the subnet in the AWS VPC console to see the current available IP count
2Terminate idle Databricks clusters to free up IP addresses
3Use a larger subnet (e.g., /20 instead of /24) or create additional subnets for Databricks
4Configure Databricks to use multiple subnets across availability zones to spread IP usage
5Move other workloads (RDS, ECS) out of the Databricks subnet to free address space

Frequently asked questions

How many IP addresses does each Databricks node use?

Each node typically uses 2 private IPs — one for the instance and one for the Databricks VPC endpoint communication. Plan subnets accordingly: a /24 (256 addresses) supports roughly 100 nodes after AWS overhead.

Can I change a subnet CIDR after creation?

No — AWS does not allow changing the CIDR of an existing subnet. You must create a new, larger subnet and migrate Databricks to use it.