Applying changes in a loop (one by one) - is it possible ?

[przemolb]

Hello,

I have the following use case and given cdk synth and cdk deploy I am not sure if it is possible with AWS CDK.
I want to create X (where X ~ 100) number of the same objects (i.e., ECS clusters). While bootstrapping of one object is not a problem, I wonder how can I create (or modify) all those objects in a loop, one by one.

This is one scenario:

1. create an ECS cluster "ECS-1" with a task definition
2. run some basic test (does an agent running as a container register correctly and proceed to the next step only if it succeeded)
3. create ECS cluster "ECS-2"
4. run some basic test (does an agent running as a container register correctly and proceed to the next step only if it succeeded)
5. ... carry on in a loop ...

This is another scenario:

1. modify Task definition in ECS-1 cluster (upgrade image version from X to Y)
2. run some tests to verify the image is working correctly
3. modify Task definition in ECS-2 cluster (upgrade image version from X to Y)
4. run some tests to verify the image is working correctly
5. ... carry on in a loop for the other ECS clusters ...

Given that after each change, we have to run cdk synth and cdk deploy is the above scenario possible while using AWS CDK ?

(if that matters - I was going to use Golang for that)

[przemolb]

Any hints on whether it is possible with AWS CDK ?

[przemolb]

Well, in case of EC2 instances, it (ASG) might work. But I am talking about ECS.
Each ECS cluster will be in a different, dedicated AWS account.

It looks like AWS CDK is not very well suited to meet this type of requirement?

[oshea00]

CDK helps to generate terraform and this can make the expression of an architecture perhaps easier for developers to deal with - maybe even modularizing components, but at the end of the day it's still terraform.

[przemolb]

@oshea00 you are talking about cdktf. This thread is about AWS CDK.

[oshea00]

Oh ha ha. Sorry, yeah, I use both frameworks and I have had Terraform on the brain lately. Cloudformation CDK, Terraform CDKTF, either way, same answer as they are based on the same technology - IAC.

If you did something like deploy a stack, do some testing, generate some more IaC, deploy it again, do some more testing, etc. You eventually work your way up to a final configuration that you could roll back at any point, but each deploy would just be adding to the final state. Not sure what the advantage would be.

[przemolb]

The advantage of doing the changes one by one is that I can stop and investigate once something wrong has happened. I.e. modify one ECS cluster, if it succeeded then carry on with the next one. If not then stop and let me investigate.