[Examples] Add a CloudFormation template showing how to deploy dstack server to AWS

[peterschmidt85]

Fixes #1941

[malandrina]

Hi there! I’m working with @spott to provision and deploy resources for dstack on AWS. I’m running into an issue where creating the cluster times out due to a task failing with the error CannotPullContainerError: pull image manifest has been retried 5 time(s): failed to resolve ref docker.io/dstackai/dstack:latest: failed to do request: Head "https://registry-1.docker.io/v2/dstackai/dstack/manifests/latest": dial tcp 54.227.20.253:443: i/o timeout.

Some additional information about our setup and steps I’ve taken:

• The VPC we’re using is private, with two private subnets in two different areas
• After encountering the error the first time I attached an AWS Internet Gateway to the VPC, thinking that this would allow resources provisioned in the private VPC to access the internet, but when I recreated the CloudFormation stack the cluster creation failed with the same error

Any chance you could point us in the right direction?

[peterschmidt85]

@malandrina Thank you for sharing this! Any chance you can check that the VPC has Internet access? Perhaps can be checked by creating an instance in this private VPC and checking that Internet is accessible.

Also perhaps you could share a screenshot of the VPC resources? Just to see what subnets and Gaweway/NAT are there?

Here's ane example of a VPC that I tested the CloudFormtion stack with:

If this helps, please feel free to join Discord, I can create a private channel.

[malandrina]

Thanks for the quick response @peterschmidt85! The information you shared was exactly what I needed and I was able to create a dstack cluster via CloudFormation with the template in this PR.

It took me a while to get the private/public subnet and gateway configuration right, but (in case it's helpful to others) here's what I ended up with:

Resources:

• 1 VPC
• 2 private subnets
  • 1 in availability zone $1
  • 1 in availability zone $2
• 2 public subnets
  • 1 in availability zone $1
  • 1 in availability zone $2
• 1 Internet Gateway
• 2 NAT Gateways
  • 1 on public subnet in availability zone $1
  • 1 on public subnet in availability zone $2

Routing:

• Public subnets -> Internet Gateway
• AZ $1 private subnet -> AZ $1 NAT Gateway
• AZ $2 private subnet -> AZ $2 NAT Gateway