I’m using boto/python to launch a new EC2 instance that boots from an EBS volume. At the time I launch the instance, I’d like to override the default size of the booting EBS volume.

I found no boto methods or parameters that might fit into my launch code:

ec2 = boto.connect_ec2( ACCESS_KEY, SECRET_KEY, region=region )

reservation = ec2.run_instances( image_id=AMI_ID, 
                                 key_name=EC2_KEY_HANDLE, 
                                 instance_type=INSTANCE_TYPE,
                                 security_groups = [ SECGROUP_HANDLE, ] )

This web page shows how to increase the size of a running EC2-instance’s EBS volume using command-line tools, but I’d like to use boto at the time the EC2 instance is specified:

You have to create a block device mapping first:

dev_sda1 = boto.ec2.blockdevicemapping.EBSBlockDeviceType()
dev_sda1.size = 50 # size in Gigabytes
bdm = boto.ec2.blockdevicemapping.BlockDeviceMapping()
bdm['/dev/sda1'] = dev_sda1 

After this you can give the block device map in your run_instances call:

reservation = ec2.run_instances( image_id=AMI_ID, 
                                 key_name=EC2_KEY_HANDLE, 
                                 instance_type=INSTANCE_TYPE,
                                 security_groups = [ SECGROUP_HANDLE, ],
                                 block_device_mappings = [bdm])

Unfortunately this is not really well documented, but the example can be found in the source code.

You can also use CloudFormation, which is used to document and automate your environment.
You can check the template for the ESB definition at: https://s3.amazonaws.com/cloudformation-templates-us-east-1/EC2WithEBSSample.template

 "Resources" : {
    "Ec2Instance" : {
      "Type" : "AWS::EC2::Instance",
      "Properties" : {
        "AvailabilityZone" : { "Fn::FindInMap" : [ "RegionMap", { "Ref" : "AWS::Region" }, "TestAz" ]},
        "SecurityGroups" : [ { "Ref" : "InstanceSecurityGroup" } ],
        "KeyName" : { "Ref" : "KeyName" },
        "ImageId" : { "Fn::FindInMap" : [ "RegionMap", { "Ref" : "AWS::Region" }, "AMI" ]},
        "Volumes" : [ 
          { "VolumeId" : { "Ref" : "NewVolume" },
            "Device" : "/dev/sdk"
          }
        ]
      }
    },

    ...

    "NewVolume" : {
      "Type" : "AWS::EC2::Volume",
      "Properties" : {
        "Size" : "100",
        "AvailabilityZone" : { "Fn::FindInMap" : [ "RegionMap", { "Ref" : "AWS::Region" }, "TestAz" ]}
      }
    }

You can then use Boto CloudFormation API to deploy your environment.

Here is a version of the code using the boto3 “everything is a resource” approach. Note also how to avoid hard-coded disk names:

import boto3

ec2 = boto3.resource('ec2')


def ec2_one_by_key_and_value(collection: str, key: str, value: str):
    handler = getattr(ec2, collection)
    return list(handler.filter(Filters=[{'Name': key, 'Values': [value]}]))[0]


image = ec2_one_by_key_and_value('images', 'name', 'ubuntu/images/hvm-ssd/...')
root_disk = None
for block_device in image.block_device_mappings:
    if block_device['DeviceName'] == image.root_device_name:
        root_disk = block_device
        assert root_disk['Ebs']
        root_disk['Ebs']['VolumeSize'] = 16 # New disk 
        break
assert root_disk
instances = ec2.create_instances(MinCount=1, ..., ImageId=image.id,
                                 BlockDeviceMappings=image.block_device_mappings)