Ha! Thank you!

Thank you!

I get what you are saying. Terraform would definitely suck at this. I still think k8s would be better from a DevX perspective however you still have to deploy a couple dozen tools to do this well in k8s. In terms of straight code from start to finish I do see your point that CDK is optimal.

I've gone out of my way to migrate from s3 to docker. Primary reason is being able to spin up preview environments quickly and easily with containers. Running a couple instances of a static site in docker and having CloudFront in front should not cost much more than s3.

My experience has been none of those things scale well when you have multiple developers on a team. To do an end to end test before merging to a shared environment is very challenging. It's even worse if you have a monolith of micro services.

HCL isn't hard. I've literally got an intern writing it. All one has to do is input values.... The hard part is and will always be knowing what the correct values are. Wether it's CDK, or another solution the hard part is understanding how to configure the cloud resources themselves.

I actually picked up this approach/strategy from a larger company. They did this for pretty much every SaaS including their Google Cloud. Multi-org works well for most but I rather test first when enabling org level things like AWS control tower or firewall manager.

I did recall reading something that said can template my quota requests as part of the account creation. We are not doing that yet...

haha the docs are actually for the sub accounts within an organization. We actually have 3 root orgs for extra separation:
1. Dev/nonprod (50+ sub accounts)
2. Production (20? sub accounts)
3. Backup (2 sub accounts)

💯 I've had this issue a few dozen times already. We even have a wiki for it (see attached). It might need an update because livechat seems to have been removed for activation recently:

I was trying to get 40vCPU's but I believe it's because of the account age (it's a new sub account) that's creating issues. I also ran into a verification issue and opened a ticket for verification. Usually, this takes a day, but it's also been open for almost as long as the vCPU increases tickets.

It does have it's issues. But for dev/nonprod is absolutely a no brainer for us.

I actually opened up 4 tickets for 4 regions all of them are Still pending. Not a single update/denial.

We don't have enterprise support so we tend to not be treated with any priority at AWS.

This does look cheaper for our current usage. I am reading on reddit that once you go over 50GB on a single user the costs go up a LOT. Not sure how they treat shared drives.... I'll have to contact their sales team and kick the tires on this. Thanks for the recommendation!

Never thought of CAPI for this. Do you just provision the nodes externally and let CAPI do it's thing? Or do you let CAPI provision the nodes too?

Thanks I've heard mixed things about using kubespray outside of the initial deployment. Allegedly it's difficult to maintain the kunespray playbooks with the upstream version.

Just took a quick look, I don't think that's accurate for India or with a USA bank account (e.g. chase bank).

💯 that is likely what we are going to do.