This is built off the case study by AWS specifically creation of a custom ami where you use to be required to have a custom ami for AWS Batch.
We needed a less managed approach and the LaunchTempalate fulfills this requirement.
The parameters into this template are not required, but have been left in because this is part of a larger stack and allow us to copy and paste into another stack just replacing linting errors on non-created variables.
The launch template is the the core of allowing for managed, but modified ami's. This attribute takes two not 'required' but required parameters, an Id and version.
Version: !GetAtt SpecialComputeLaunchTemplate.LatestVersionNumber
Per the nested documentation on launch templates the BlockDeviceMappings define a remote network storage device AND UserData which will created and attach network storage to an ECS instance.
The UserData needs to be Base64 encoded and since AWS Batch adds its own user data boundry, will have to be folded into
MIME-Version: 1.0
Content-Type: multipart/mixed; boundary="==MYBOUNDARY=="
--==MYBOUNDARY==
Content-Type: text/x-shellscript; charset="us-ascii"
--==MYBOUNDARY==
Where MYBOUNDARY is your own defined boundary name, BOUNDRY seems to be reserved to AWS Batch ECS own ECS cluster management so I would avoid that.
After this the volume will be attached and available, for the use in AWS Batch we need to go one more level and make this available in docker.
For the docker container to be able to see the volume which has been created and attached on EC2 instantiation , we need to add a docker mount/volume and add a working directory.
CloudFormation
MountPoints:
- ContainerPath: "/data"
ReadOnly: false
SourceVolume: data
Volumes:
- Name: data
Host:
SourcePath: "/data"
Where SourcePath is the only external value from the LaunchTemplate which you need, the other values are up to your choosing.
Second part is in the docker file
WORKDIR /data
Defines the working directy which we need to be the attached storage that is 250G.