Despite all the planning and preparing we do to architect flawless systems, we may eventually run into issues with our design. Changing business needs will mean you need to quickly reassess your design and find reliable solutions.
For example, say you spin up several EC2 instances with unencrypted root volumes, thinking you would not need to store any sensitive data. Requirements change and you now need to encrypt those volumes. This post will walk through the steps to encrypt a root volume for an EC2 instance.
Amazon provides SDKs for many different languages. This example is written in Python and uses Amazon's Python SDK, Boto 3.
Follow the Boto 3 QuickStart Guide to get started with Boto 3.
- Install the AWS CLI
- Configure your client
These instructions will walk you through setting up your AWS credentials.
When configuring the AWS CLI it will set up your default profile.
You can also set up multiple 'Named Profiles'
with different settings. Simply set the --profile flag and it will prompt you
for input.
aws configure --profile profilename
Your config will be written to a 'config' file in the $HOME/.aws directory.
Your credentials will be written to a 'credentials' file in the same directory.
On Windows, these files are located in the $env:USERPROFILE\.aws directory.
These IAM policy actions are required in order to run this script. This is an example policy you can attach to a user to give them the proper IAM authorization.
{
"Version": "2010-10-10",
"Statement": [
{
"Effect": "Allow",
"Action": [
"ec2:DescribeInstances",
"ec2:DescribeVolumes",
"ec2:DescribeSnapshots",
"ec2:StopInstances",
"ec2:StartInstances",
"ec2:CopySnapshot",
"ec2:CreateSnapshot",
"ec2:CreateVolume",
"ec2:DeleteVolume",
"ec2:DeleteSnapshot",
"ec2:AttachVolume",
"ec2:DetachVolume"
],
"Resource": "*"
}
]
}First we will layout the overview of the script including parameters it can receive.
#!/usr/bin/env python
"""
Overview:
Take unencrypted root volume and encrypt it for EC2.
Params:
ID for EC2 instance
Customer Master Key (CMK) (optional)
Profile to use
Conditions:
Return if volume already encrypted
Use named profiles from credentials file
"""This script will use the ID parameter to find the EC2 instance. If the root volume is already encrypted the script will exit.
Also, we want to take advantage of profiles so we will receive the profile the user wants to use. The Customer Master Key will be used for the encryption, but is not required. This will be explained later.
Let's import what we need for the script and set up our argument parser.
import sys
import boto3
import botocore
import argparse
def main(argv):
parser = argparse.ArgumentParser(description='Encrypts EC2 root volume.')
parser.add_argument('-i', '--instance',
help='Instance to encrypt volume on.', required=True)
parser.add_argument('-key', '--customer_master_key',
help='Customer master key', required=False)
parser.add_argument('-p', '--profile',
help='Profile to use', required=False)
args = parser.parse_args()The next step is to get our session set up. Boto will use the default profile
unless the user passed in a --profile parameter.
""" Set up AWS Session + Client + Resources + Waiters """
if args.profile:
# Create custom session
print('Using profile {}'.format(args.profile))
session = boto3.session.Session(profile_name=args.profile)
else:
# Use default session
session = boto3.session.Session()Two key features of Boto3 are high-level object-oriented resources and low-level service connections. Low-level services map closely to the AWS service APIs whereas the high-level resource are abstracted from them.
Another feature are waiters, these are helpful for blocking until desired states are reached. Two waiters are stored, one for waiting until a snapshot is completed and the other for waiting until a volume is available. We get these waiters from the low-level client.
client = session.client('ec2')
ec2 = session.resource('ec2')
waiter_instance_exists = client.get_waiter('instance_exists')
waiter_instance_stopped = client.get_waiter('instance_stopped')
waiter_instance_running = client.get_waiter('instance_running')
waiter_snapshot_complete = client.get_waiter('snapshot_completed')
waiter_volume_available = client.get_waiter('volume_available')If the optional CMK was passed in, we will store it.
# Get CMK
customer_master_key = args.customer_master_keyBefore we proceed too far into the script we will make sure that the instance id we received can be used to retrieve an instance. If its not, we will exit the script.
""" Check instance exists """
instance_id = args.instance
print('---Checking instance ({})'.format(instance_id))
instance = ec2.Instance(instance_id)
try:
waiter_instance_exists.wait(
InstanceIds=[
instance_id,
]
)
except botocore.exceptions.WaiterError as e:
sys.exit('ERROR: {}'.format(e))We have an instance and the resources we need, let's continue the script and encrypt the volume.
You can access the volumes of the instances. These return a Boto3 'Collection' which is an iterable of resources. We can iterate through it to get access to the actual instance of the volume. Note, if the volume is encrypted, we will exit our script.
""" Get volume and exit if already encrypted """
volumes = [v for v in instance.volumes.all()]
if volumes:
original_root_volume = volumes[0]
volume_encrypted = original_root_volume.encrypted
if volume_encrypted:
sys.exit(
'**Volume ({}) is already encrypted'
.format(original_root_volume.id))The instance volume has mappings that we will want to preserve for the encrypted volume. We are able to modify these. You can do this as needed with other mappings. For this example, we are storing the 'DeleteOnTermination' information, which we will use at the end of the script to make sure it's the same.
""" Step 1: Prepare instance """
print('---Preparing instance')
# Save original mappings to persist to new volume
original_mappings = {}
original_mappings['DeleteOnTermination'] = instance.block_device_mappings[0]['Ebs']['DeleteOnTermination']We won't be able to work with this instance how we want if its running. You can check the status of instance through its state property. The different codes are:
- 0 : pending
- 16 : running
- 32 : shutting-down
- 48 : terminated
- 64 : stopping
- 80 : stopped
In this case we will exit if the state is pending, shutting-down, or terminated and stop the instance if it is running.
# Exit if instance is pending, shutting-down, or terminated
instance_exit_states = [0, 32, 48]
if instance.state['Code'] in instance_exit_states:
sys.exit(
'ERROR: Instance is {} please make sure this instance is active.'
.format(instance.state['Name'])
)
# Validate successful shutdown if it is running or stopping
if instance.state['Code'] is 16:
instance.stop()Now that the signal to stop the instance is sent, we want to wait for the instance to be in its proper state. We can use a waiter, which polls the state of the instance at intervals, to block the code until the instance is stopped.
try:
waiter_instance_stopped.wait(
InstanceIds=[
instance_id,
]
)
except botocore.exceptions.WaiterError as e:
sys.exit('ERROR: {}'.format(e))Waiters can be configured to behave as you need, for example, the waiter will poll 40 times to check the state. If it still has not reached the desired state the waiter is looking for it will exit with a 'WaiterError'. You can change this through '.config'.
# Set the max_attempts for this waiter (default 40)
waiter_instance_stopped.config.max_attempts = 40Now that we have the id for the volume we will use that to create a snapshot of its current state. Immediately after we will use the waiter we stored earlier to wait for the snapshot to be complete. You can pass multiple ids into the waiter. In this case, we will just wait for the one we created to be complete.
""" Step 2: Take snapshot of volume """
print('---Create snapshot of volume ({})'.format(original_root_volume.id))
snapshot = ec2.create_snapshot(
VolumeId=original_root_volume.id,
Description='Snapshot of volume ({})'.format(original_root_volume.id),
)
try:
waiter_snapshot_complete.wait(
SnapshotIds=[
snapshot.id,
]
)
except botocore.exceptions.WaiterError as e:
# Clean up the snapshot to reduce clutter (optional)
snapshot.delete()
sys.exit('ERROR: {}'.format(e))To create the encrypted volumes we can simply create a copy of the snapshot of the unencrypted volume and set the 'Encrypted' flag to true. In addition to the encrypted flag, we can set other parameters for this action. If the user passed their customer master key, meaning they don't want to use the Amazon's default key management system, we will use that specific key for the encryption. Again, once the snapshot begins to be copied, we will wait for the snapshot to be complete before proceeding.
""" Step 3: Create encrypted volume """
print('---Create encrypted copy of snapshot')
if customer_master_key:
# Use custom key
snapshot_encrypted_dict = snapshot.copy(
SourceRegion=session.region_name,
Description='Encrypted copy of snapshot #{}'
.format(snapshot.id),
KmsKeyId=customer_master_key,
Encrypted=True,
)
else:
# Use default key
snapshot_encrypted_dict = snapshot.copy(
SourceRegion=session.region_name,
Description='Encrypted copy of snapshot ({})'
.format(snapshot.id),
Encrypted=True,
)
snapshot_encrypted = ec2.Snapshot(snapshot_encrypted_dict['SnapshotId'])
try:
waiter_snapshot_complete.wait(
SnapshotIds=[
snapshot_encrypted.id,
],
)
except botocore.exceptions.WaiterError as e:
snapshot.delete()
snapshot_encrypted.delete()
sys.exit('ERROR: {}'.format(e))The snapshot is complete so we can now take that and create an encrypted volume. Because the snapshot is encrypted, the volume will be too.
print('---Create encrypted volume from snapshot')
volume_encrypted = ec2.create_volume(
SnapshotId=snapshot_encrypted.id,
AvailabilityZone=instance.placement['AvailabilityZone']
)Before we can attach the new encrypted volume we need to detach the old volume.
""" Step 4: Detach current root volume """
print('---Deatch volume {}'.format(original_root_volume.id))
instance.detach_volume(
VolumeId=original_root_volume.id,
Device=instance.root_device_name,
)Once the encrypted volume is complete, we are ready to attach it. Pass in the volume id and the keep the device type constant by pulling it from the original instances property.
""" Step 5: Attach current root volume """
print('---Attach volume {}'.format(volume_encrypted.id))
try:
waiter_volume_available.wait(
VolumeIds=[
volume_encrypted.id,
],
)
except botocore.exceptions.WaiterError as e:
snapshot.delete()
snapshot_encrypted.delete()
volume_encrypted.delete()
sys.exit('ERROR: {}'.format(e))
instance.attach_volume(
VolumeId=volume_encrypted.id,
Device=instance.root_device_name
)The volume is attached so we want to bring our instance back up. We will make sure the original mappings remain. We can access those through the 'modify_attribute' and selecting the root device we just created.
""" Step 6: Restart instance """
# Modify instance attributes
instance.modify_attribute(
BlockDeviceMappings=[
{
'DeviceName': instance.root_device_name,
'Ebs': {
'DeleteOnTermination':
original_mappings['DeleteOnTermination'],
},
},
],
)We will start the instance and then wait for it to be running.
print('---Restart instance')
instance.start()
try:
waiter_instance_running.wait(
InstanceIds=[
instance_id,
]
)
except botocore.exceptions.WaiterError as e:
sys.exit('ERROR: {}'.format(e))We no longer need the snapshot because we have extracted our volume already. You can do clean up as needed. We can also delete the snapshot_encrypted resource and the original root volume, we don't want that hanging around unencrypted.
""" Step 7: Clean up """
print('---Clean up resources')
# Delete snapshots and original volume
snapshot.delete()
snapshot_encrypted.delete()
original_root_volume.delete()In this post you saw how to encrypt the root volume of an existing EC2 instance. After installing the AWS CLI and the Boto 3 Python SDK, we showed you how to create a short Python script to snapshot your existing root volume to a new encrypted root volume and restart your instance. To ensure your data is safe, the script deletes the original unencrypted volume as the last step.