tbergeron / AWS-SnapAndDelete-Lambda

collection of scripts for cloud gaming on AWS in a cost-effective way

BEWARE!

• those scripts are provided AS-IS, if you are not well experienced with AWS PLEASE DO NOT USE!
  • AWS bills can go crazy high FAST if you don't know what you're doing.
  • if something wrong happens, i am not responsible and i will provide no support.

random bits of advice

• use gp3 instead of gp2 for volume types, it is most cost efficient / cheaper by about 30%
• as the following script implies, convert unused volumes to snapshots, and if you really don't use them often don't hesitate to archive the snapshots.
  • snapshots are about half cheaper and archives are maybe 10 times even cheaper.
• as this script uses, g4dn are IMO the best instance type to use at the moment and it is good enough to run most AAA games.
• for huge games like Red Dead Redemption, a costly yet efficient way to speed up loading times and in-game texture streaming is to increase the volume Throughput.
  • BEWARE! it cost a EXPONENTIALLY more than the defaults (default is 125MB/s)
• you can use the same spawn-instance-from-image.py script for spot instances. make sure you are allowed by AWS to use them first.

snap-and-delete-on-stop.py

• create a lambda function and copy/paste the code in it.
• make sure you set the timeout to some higher value ie. 15mins
• make sure you give the lambda permission:
  • click on Permissions from the top, and click on the role that AWS automatically created for your function.
  • click Attach Policies, search for EC2 and choose the AmazonEC2FullAccess policy. click Attach Policy
• using EventBridge, create a new rule with the following event pattern:

{
  "source": ["aws.ec2"],
  "detail-type": ["EC2 Instance State-change Notification"],
  "detail": {
    "state": ["stopped"]
  }
}

• and use the lambda function you created as a target.
• that way, each time you stop an instance it will:
  • delete old AMIs and their associated snapshots
  • create a new AMI of the instance that got stopped
  • terminate the instance
  • delete all remaining volumes (be careful, it will delete ALL volumes!!!)

spawn-instance-from-image.py

• a script you can call using a bash or powershell script to redeploy the imaged instance
• follow the same lambda instructions as above (permissions, timeout, etc.)
• create a function URL (in the lambda configuration) that you can call via curl
• call that url whenever you need to spawn an instance of the last created AMI

format-root-ebs.ps1

• this script is meant to be ran ON the instance.
• CHANGE THE DRIVE LETTER based on your needs (hardcoded to G)
• script that can be executed on startup to format the ephemeral EBS root volume to NTFS.
• to execute on startup you can add a cmd script to the startup directory ie.
  • in C:\ProgramData\Microsoft\Windows\Start Menu\Programs\StartUp
  • create a startFormatRootEBS.cmd file and do something like:

PowerShell -Command "Set-ExecutionPolicy Unrestricted" >> "%TEMP%\StartupLog.txt" 2>&1
PowerShell %USERPROFILE%\Desktop\format-root-ebs.ps1 >> "%TEMP%\StartupLog.txt" 2>&1