topic Re: NAT Gateway IP update in Administration & Architecture

NAT Gateway IP update

gillesfromparis — Mon, 02 Jun 2025 05:26:13 GMT

Hi,

My Databricks (Premium) account was deployed on AWS.

It was provisioned a few months ago from the AWS MarketPlace with the QuickStart method, based on CloudFormation.

The NAT Gateway initially created by the CloudFormation stack has been incidentally destroyed.

I created a new NAT Instance and updated the private subnets route table accordingly.

Yet the EIP (Elastic IP) associated with the previous NAT Gateway was released, and my new NAT instance gets a different EIP.

As a consequence, the Control Plane can no longer communicate with the resources deployed in the private subnets where my Databricks account is deployed, as says the message visible in my account network settings:

"Warning: Cannot find any NAT Gateways for vpc-0exxxxxxx."

Is there any way to update the NAT Gateway IP declared for my VPC workspace at the Control Plane end ?

Thanks

Gilles

Re: NAT Gateway IP update

Walter_C — Mon, 02 Jun 2025 16:48:08 GMT

Have you adjusted te network routing to point to the new nat gateway and the nat gateway to point to the internet gateway?

https://docs.databricks.com/aws/en/security/network/classic/customer-managed-vpc#additional-subnet-requirements

Re: NAT Gateway IP update

gillesfromparis — Mon, 02 Jun 2025 18:29:20 GMT

Thanks for your answer.

Yes I did all of that, as well as allowing the traffic coming from my NAT Instance as an inbound rule of the Security Group of the private instances (and the other way around), and no restriction on the outbound traffic.

I suspect that the NAT Gateway IP was "hard coded" in Databricks Control Plane (associated with my VPC reference) when it was provisioned by CloudFormation, but I don't know how it could be updated with the IP of my new NAT Instance (since this update should be made by Databricks).

Re: NAT Gateway IP update

gillesfromparis — Tue, 03 Jun 2025 05:33:38 GMT

Problem solved (the IP forward option was not declared in the right config file of the NAT Instance).

My cluster is now up and running.

Thanks anyway Walter for your support.