how to create table automatically in aws aurora serverless with serverless framework

Question

I'm trying to create table automatically with npm migrate whenever we deploy any changes with serverless framework. It's quite fine when I used with aurora database. But I've moved to Aurora Serverless RDS (Sydney region), it's not working at all. Because Aurora Serverless RDS itself is working inside VPC, thus when we need to access it lambda function should must be at same VPC.

PS: we're using Github Action as pipeline to deploy everything to Lambda.

Please let me know how to solve that issue, thanks.

Answer 1

There are only two basic ways that you can approach this: open a tunnel into the VPC or run your updates inside the VPC. Here are some of the approaches to each that I've used in the past:

Tunnel into the VPC:

• VPN, such as OpenVPN.

  Relatively easy to set up, but designed to connect two networks together and represents an always-on charge for the server. Would work well if you're running the migrations from, say, your corporate network, but not something that you want to try to configure for GitHub Actions (or any third-party build tool).

• Bastion host

  This is an EC2 instance that runs in a public subnet and exposes SSH to the world. You make an SSH connection to the Bastion and then tunnel whatever protocol you want underneath. Typically run as an "always on" instance, but you can start and stop programmatically.

  I think this would add a lot of complexity to your build. Assuming that you just want to run on demand, you'd need a script that would start the instance and wait for it to be ready to accept connections. You would probably also want to adjust the security group ingress rules to only allow traffic from your build machine (whose IP is likely to change for each build). Then you'd have to open the tunnel, by running ssh in the background, and close it again after the build is done.

Running the migration inside the VPC:

• Simplest approach (imo) is to just move your build inside the VPC, using CodeBuild. If you do this you'll need to have a NAT so that the build can talk to the outside world. It's also not as easy to configure CodeBuild to talk to GitHub as it should be (there's one manual step where you need to provide an access token).

• If you're doing a containerized deployment with ECS, then I recommend packaging your migrations in a container and deploying it onto the same cluster that runs the application. Then you'd trigger the run with aws ecs run-task (I assume there's something similar for EKS, but haven't used it).

• If you aren't already working with ECS/EKS, then you can implement the same idea with AWS Batch .

Answer 2

Here is an example on how you could approach database schema migration using Amazon API Gateway, AWS Lambda, Amazon Aurora Serverless (MySQL) and Python CDK.