In the self-terminating mode the VM will expire itself automatically after the --expiration-date, meaning the engine daemon doesn't have to be kept running in the background, which is great for automation/integration.

It works by installing a Cronjob at the end of the Ansible setup (under "root"), which runs the instance_teardown.py script that checks if we've passed the expiration_date, and if so - cleans up any volumes, elastic IPs and finally terminates itself.

Relevant CLI/Env flags:

• --self-termination / SELF_TERMINATION
• --self-termination-access-key-id / SELF_TERMINATION_ACCESS_KEY_ID
• --self-termination-secret-access-key / SELF_TERMINATION_SECRET_ACCESS_KEY

PS It is strongly recommended to generate separate AWS credentials for self-termination, only with limited EC2 privileges as the keys will be actually placed on the VMs to be able to work self-sufficiently. Also for public IP instances one should take care to configure Security Groups and pg_hba.conf rules accordingly.

Also note that not all resources are guaranteed to be auto-expired:

• Backups (if enabled) - as a security measure
• In case of non-floating IPs (by default IPs are floating) custom NICs also cannot be cleaned up as still attached to the VM at time of calling terminate on itself. Thus if using this feature with public IPs, to get rid of any left-over resources it's recommend to run the delete_all_operator_tagged_objects.sh script occasionally. Hanging NICs don't cost extra though.

Minimum AWS EC2 privileges for feature to work

"ec2:TerminateInstances",
"ec2:DisassociateAddress",
"ec2:ReleaseAddress",
"ec2:DescribeVolume*",
"ec2:DetachVolume",
"ec2:DeleteVolume",

Volume striping

To get a kind of free lunch on provisioned IOPS costs - one can stripe together a bunch of volumes on default IOPS using the below flags to get a performance boost, similar to what RDS does under the hood:

• --stripes / STRIPES 2-28 stripe volumes allowed. Default 0
• --stripe-size-kb / STRIPE_SIZE_KB 4-4096 KB range. Default 64

PS Note that for lower CPU instances you can still easily run into instance level max bandwith or IOPS limitations for heavier workloads. For example to get past 40K IOPS, one needs 16 vCPUs. AWS docs here: https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/ebs-optimized.html