Cleaning AWS Marketplace AMIs

Running (e.g.) CentOS 7 on unapproved hardware

The official CentOS 7 marketplace AMI incurs no extra fees and usually works just fine, but due to its product code and the associated marketplace metadata, AWS prohibits launching instance types not listed on that page.

At the time of posting, it appears that all AWS instance types are approved for use with the current CentOS 7 AMI, but when I originally developed these instructions, the CentOS 7 AMI could not be run on the p2.* or x1.* (etc.) instance types.

The following steps are based off jake1138’s instructions in the 2015-06-25 thread on (I originally found that page from the less-helpful but easier-to-Google blog post at However, jake1138’s method involves dd‘ing the live/root/mounted file system, which is understandably dicey, and resulted in lots of “Structure needs cleaning” errors when trying to use the resulting volume.

Here’s a better method:

Start up a new instance that will host all the dd‘ing work.

# the image id comes from ->
# "Continue" -> "Manual Launch" -> US East (N. Virginia)
aws ec2 run-instances \
  --image-id ami-46c1b650 \
  --key-name aws-ec2-20151025 \
  --instance-type m4.large \
  --block-device-mappings 'DeviceName=/dev/sda1,Ebs={VolumeSize=8,DeleteOnTermination=true}' \
  --subnet-id subnet-6df23d24 \

Store the Instance ID as HOST_INSTANCE, and Volume ID as HOST_VOLUME:

Then stop it:

aws ec2 stop-instances --instance-id <host-instance-id>

Start up a new instance:

aws ec2 run-instances \
  --image-id ami-46c1b650 \
  --instance-type m4.large \
  --block-device-mappings 'DeviceName=/dev/sda1,Ebs={VolumeSize=8,DeleteOnTermination=false}' \
  --subnet-id subnet-6df23d24

Store the Volume ID as SOURCE_VOL, and then terminate the instance (all we want is the volume):

aws ec2 terminate-instances --instance-id <its-instance-id>

Create a spare volume:

# the original CentOS image has VolumeType=standard
aws ec2 create-volume \
  --size 8 \
  --availability-zone us-east-1c \
  --volume-type standard

Store the Volume ID as TARGET_VOL.

Attach the volumes to the stopped instance:

aws ec2 attach-volume \
  --volume-id $SOURCE_VOL \
  --instance-id $HOST_INSTANCE \
  --device /dev/sdc
aws ec2 attach-volume \
  --volume-id $TARGET_VOL \
  --instance-id $HOST_INSTANCE \
  --device /dev/sdf

Start the stopped instance:

aws ec2 start-instances --instance-id $HOST_INSTANCE

SSH in and clone the source volume to the target volume:

aws ec2 describe-instances --instance-id $HOST_INSTANCE
ssh <host-instance-ip>
sudo mkfs -t xfs /dev/xvdf
sudo dd bs=65536 if=/dev/xvdc of=/dev/xvdf

Output will look something like:

131072+0 records in
131072+0 records out
8589934592 bytes (8.6 GB) copied, 334.42 s, 25.7 MB/s

Back on the local machine, stop the instance:

aws ec2 stop-instances --instance-id $HOST_INSTANCE

Detach all the volumes:

  xargs -n1 aws ec2 detach-volume --instance-id $HOST_INSTANCE --volume-id

Attach the cloned volume at the standard location:

aws ec2 attach-volume --instance-id $HOST_INSTANCE --volume-id $TARGET_VOLUME --device /dev/sda1

Trigger an AMI snapshot:

aws ec2 create-image \
  --instance-id $HOST_INSTANCE \
  --name 'centos7/base' \
  --description "CentOS 7 (x86_64) build 1602 (aw0evgkw8e5c1q413zgy5pjce) clone without product codes"

Wait a while for the image to be ready:

aws ec2 wait image-available --image-ids ami-491c095e

All done! Terminate the instance:

aws ec2 terminate-instances --instance-id $HOST_INSTANCE