aws-cloud-costs documentation

[dejanu]

Values for cloud-integration.json example are not consistent. i.e. ACCESS_KEY_ID and ACCESS_KEY_SECRET

Also, the documentation is rather unclear with regard to the minimal setup in terms of AWS resources. Is unclear if opencost needs a AWS IAM user or role, or both.

[Robbie558]

The documentation is not sufficient to setup a working AWS integration, specifically lacking information around Athena.

I am getting TABLE_NOT_FOUND errors in Athena, and there is no documentation that relates to the setup required in Athena outside:

The Athena database name is available as the value (physical id) of AWSCURDatabase in the CloudFormation stack created above

But there is no mention of the CloudFormation elsewhere in the documentation 🤷🏻‍♂️

[andreirusu99]

Hi, I am having the same problem, the docs keep saying "previous step" or "CloudFormation stack created above.", with no mention of that step or CloudFormation stack. I was able to finally track down that this refers to setting up Athena with Glue, and found this guide: https://docs.aws.amazon.com/cur/latest/userguide/use-athena-cf.html, however there is no mention of the exact AWS resources that need to be created for all this to work, if you for example want to use Terraform. Only the total number of resources to be created are mentioned.
My solution so far is to generate this template and reverse-engineer it to create Terraform files from it, otherwise this becomes configuration hell if you need it for multiple clusters.
What's more, there are now a CUR 2.0 reports, and OpenCost only talks about the 1.0 in the documentation, now called "Legacy Exports". So it looks like if you set this up, you still need to use this legacy export, so you are giving yourself tech debt. In the OpenCost code they seem to support CUR 2.0, but the documentation is nowhere for it.
Once I have a full working solution I will write it somewhere here, and maybe someone can update the website, or i could do it as well if given permission to create a PR, because this is really a tedious process, and I see many more people are trying to get it working.