AWS-Reasonable-Account-Defaults

I had a friend ask me recently what his child should do to their AWS account if they wanted to stay on the free tier. I explained some basics of free tier, but it got me thinking about “rational” AWS account defaults. This project is the answer to my friend's question. Something substantial that he can give to his child. Based on 2 re:Invent chalk talks on "Ways to Avoid Cost Surprises" and the work I've done with startups around cost optimization, I've put together this CloudFormation Template. My expectation is that these resources should be a good start toward being alerted if and when costs are accumulating in an AWS account.

Where to Deploy

Most of the services used by this CloudFormation are deployed in all regions, but I have found that AWS Budgets and Amazon EventBridge Scheduler are not deployed in all regions. Therefore, you should choose a region that has both of these. You can check availability for AWS Budgets and Amazon EventBridge Scheduler.

Deploying in the older regions like us-east-1 (Virginia), us-east-2 (Ohio), us-west-2 (Oregon), or eu-west-1 (Ireland) should not be a problem.

Invocation

IAM User or SSO User

It is recommended to NOT use the root user to deploy this CloudFormation template. Either create an IAM User including MFA and the AWS Managed IAM Policy AdministratorAccess or create an Identity Center (SSO) User including MFA and the AWS Managed IAM Policy AdministratorAccess. It is best security practice to enable MFA even if using an app on your smartphone like Google Authenticator. Preventing unwanted access to your account is step 1 in preventing unexpected costs.

NOTE: Using Identity Center (SSO) will be more useful if you have multiple users and/or accounts, but requires a little more investment to setup. Creating an IAM User is easier, but less scalable across accounts.

Creating the Stack

• Download the CloudFormation Template
• Go to CloudFormation in the AWS Console

• Click Create Stack

• Select "Choose an existing template"
• Template Source, select "Upload a template file"
• click "choose file" and select the downloaded CloudFormation Template file
• Click Next

• Stack Details
  • Stack Name
    • Give a name to your CloudFormation Stack. "Reasonable-Account-Defaults" (Note: Spaces are not allowed)

• Parameters
  • Email Address - this is the email address to receive notifications to. It's important that this is an accurate address or the cost notifications will not be received
  • Spend Thresholds - this is the alert threshold that will be set in the Cost Anomaly Detector (first parameter) and in the Fixed Budget (second parameter).
    • For a Free Tier or Student account, the recommendation is the minimum of $0.01.
    • For an account with any usage, some other value should be chosen that coincides with how much spend you expect to incur
  • Log Retention - How long to retain CloudWatch Log Group data.
  • EventBridge Rule Lambda - this is the Lambda that will look for newly enabled regions and push the EventBridge rules enabling automatic S3 Bucket MPU Lifecycle Policies and CloudWatch Log Group Retentions. Default is daily (Days and 1).
• Click Next

• While Tags are optional, they are super useful for identifying resources. NOTE: the tag key and tag value are case sensitive. I suggest at least adding
  • key = project
  • value = AWS Reasonable Account Defaults

• The remaining items stay as the default

• Click Next

• Review the page

• Approve the IAM Permissions

  • "I acknowledge that AWS CloudFormation might create IAM resources with custom names."

• Click Submit

• The creation process will start
  • The process should generaly take under 5 min. During my testing, the deployment took around 2 min.
• You're done, you have taken your first step toward avoiding Cost Surprises

What Gets Deployed

• AWS Budgets
  • Auto Adjusting Budget
  • Forecasted Monthly Budget with a Target of the FixedBudgetParam you set above
• Cost Anomaly Detector
  • Monitor gets deployed with an Alert Threshold of AlterThresholdParam you set above
• CloudTrail
  • For the Event Bridge triggering rules to work, a CloudTrail Trail must be active.
  • A Lambda is deployed and run which checks all enabled regions in the account and searches for active CloudTrails. If no CloudTrail is found, it will create one.
    • The first CloudTrail Trail is available for free, therefore the Lambda checks that it will not incur an expense
    • The Lambda also creates and S3 bucket to store the CloudTrail logs in. No lifecycle policy is set to delete the logs since these are a security feature. If one is wanted, the user can go and set a lifecycle policy on the bucket created to remove all objects after X time.
• CloudWatch Log Group Retention
  • An Event Bridge Rule is added to the default Event Bridge Bus that traps all CloudWatch Log Group Creations and triggers the Lambda.
  • The Lambda that is deployed takes the Log Group from the event information and sets the Retention to what retention period was established when the CloudFormation template was run. This Lambda only finds Log Groups as creation, if the retention is changed later or if the Log Group existed already, the Lambda will not modify them.
• S3 Multipart Upload (MPU) Retention
  • An Event Bridge Rule is added to the default Event Bride Bus that traps all S3 Bucket Creations and triggers the Lambda.
  • The Lambda that is deployed takes the S3 Bucket from the event information and creates a Lifecycle Policy that deletes MPU Fragments after 7 days.
• Multi-Region / New Region Activation
  • In order for CloudWatch Log Groups and S3 Buckets created in regions other than the primary deployment to trigger the Lambdas, Event Bridge rules are deployed to all enabled regions by a Lambda.
  • The Lambda goes through all enabled regions other than the primary deployed region and if the Event Bridge forwarding rules are found, does nothing. If rules are not found, it creates them. This Lambda is scheduled to run regularly (as defined in the parameters) so that it will catch any newly enabled regions.

Linked Accounts and AWS Organizations

If you are deploying on a linked account (a member of an AWS Organization, formerly called a child account) and your AWS Organization is using Service Control Policies (SCPs) and they have implemented the general example Deny access to AWS based on the requested AWS Region from the AWS Documentation, then you will need to add some Actions to the NotAction group. If you only have 1 account and are not part of an AWS Organization, then you can ignore this part.

• "cloudtrail:*",
• "events:ListRules*",
• "events:PutRule*",
• "events:DeleteRule*",
• "events:PutTargets*",
• "events:RemoveTargets*",
• "events:ListTargetsByRule*",

These actions are used by 2 of the Lambdas in order to check for CloudTrail Trails in all the enabled regions and to push Event Bridge rules to collect events on CloudWatch Log Group creation and S3 Bucket creation.

Costs

The resources deployed should have little to no costs. The Lambdas only run occassionally and therefore should not incur a cost under normal load and usage. The Budgets and Cost Anomaly Detection should also not incur costs. The CloudTrail that gets created should not have a direct cost since it should be the 1 Trail which supplied under the Free Tier. However, I have seen during my testing, S3 request costs for the CloudTrail logs being pushed into S3. In my test account, I saw up to $0.02 per day.

Known Issues

If the CloudFormation deployment encounters an error, the default behavior is for the process to rollback. If it rolls back, in the Events tab, you can find the probably root cause by clicking the Detect root cause button. Look on the right side of the Events table for an error.

• CloudFormation Error AnomalyMonitor: User not enabled for cost explorer access

  • If you are using the root user, you will need to enable billing access. It is NOT recommended to use the root user to deploy this CloudFormation Template.

• CloudFormation Error AnomalyMonitor: Limit exceeded on dimensional spend monitor creation

  • If you already have a Cost Anomaly Detection Monitor for All Services, then you cannot deploy a second one.
  • To enable the CloudFormation Template, you will need to, in the AWS Console, go to the Cost Anomaly Detection service --> Alert Subscriptions
    • If the subscription is for $100 or 40%, this is the default monitor.
    • If the subscription is anything else, STOP and verify outside the account where this monitor may have come from.
  • In the Cost Monitors tab, delete the monitor.

Media

• Ways to avoid cost surprises | The Keys to AWS Optimization | S10 E7
• Last week in AWS Newsletter, Tools Section
• Screaming in the Cloud Podcast
• Screaming in the Cloud on Youtube

Next Steps

• Have an option for non-student/non-free-tier accounts that splits out the fixed budget and cost anomaly detector thresholds - Done
• A Cloudtrail trail is required by the Lambdas below. So I have created a Lambda that runs once and checks if there is a Cloudtrail Trail enabled. If not, it will create it. - Done
• Implement a Lambda that pushes an S3 lifecycle policy to all new buckets so that Multipart Upload (MPU) fragments are removed after 7 days (configurable) - Done
  • push Event Bridge rules to all active regions so that a bucket created in any active region will trigger the above Lambda - Done
• Implement a Lambda that sets new CloudWatch (CW) Log Group retention to 30 days (configurable) so that any new Log Group created will not fill up forever - Done
  • push Event Bridge rules to all active regions so that a CW Log Group created in any active region will trigger the above Lambda - Done
• Implement a Lambda to run weekly (configurable) that will push the previous 2 Event Bridge rules to any newly activated region - Done
• Testing ability
  • automate the testing and cleanup of the resources deployed
• Delete Everything
  • Since there are 2 Lambdas that could possibly create resources outside of CloudFormation, there is the possability that when the CloudFormation stack is deleted, those resources will remain.
  • If the Lambda creates the CloudTrail Trail and the corresponding S3 Bucket, those will remain if the stack is deleted.
  • The Event Bridge Rule Lambda that pushes event passing rules to all enabled regions for S3 Bucket Creation and CloudWatch Log Group Creation will remain, but the Lambdas they trigger will no longer be there after the stack is deleted.
• IAM Security Hardening
• go through all the IAM permissions created and ensure they are only allowing required permissions and are scope limited to the required resources.
  • I did this, but have 1 IAM permission that I have not yet successfully limited the scope on.
• Resource Tagging
• automatically add the CloudFormation Name and StackID to all created resources and all derivative resources