Running out of space just when pulling a docker image on an ECS spot cluster, even when requesting more tmpfs space, how should I solve this?

@batch(
        gpu=1,
        image="*.<http://dkr.ecr.eu-west-1.amazonaws.com/*|dkr.ecr.eu-west-1.amazonaws.com/*>",
        tmpfs_size=1024 * 30,
    )
    @step

it's like that the underlying node doesn't have enough disk capacity. you can bump up the disk attached (through a launch template).