Building For AWS

Stacki on AWS

Stacki AWS support is included in the source tree. This procedure documents how we go from built Stacki source to a new Frontend and Backend AMI.

Frontend AMI

Launch a Frontend

Follow the Instructions for starting a new Stacki Frontend in AWS. Once the instance is up copy over a new Stacki ISO (or just Build a new one). Next disable the existing Stacki pallet and enable the new one.

Launch a Barnacle

We use the Barnacle appliance to create a new Frontend AMI, to do this following the Instructions for starting a Backend instance and supply the extra appliance information in the json user data. For example:

{
  "master"    : "172.0.0.1",
  "appliance" : "barnacle"
}

Prep the Instance

Once the Barnacle instance is finished installing login and prepare the instance.

# /opt/stack/sbin/aws-barnacle-prep-image

This command will remove any SSH keys, enable a barnacle service to run on next boot, and will shutdown the instance.

Register the AMI

Register a new AMI using the shutdown Barnacle instance as the source. For CentOS make sure the AMI is public, but for SLES keep the AMI private.

Backend AMI

The Backend AMI should rarely need to be updated. It is based on the Amazon Linux AMI with a single Stacki package added. To refresh the AMI launch the existing Backend AMI but do not supply any user data. The instance will start and you can SSH into it as the ec2-user using your AWS SSH credentials. From here do the following:

• install the stack-aws-client RPM
• rm -rf ~/ec2-user/.ssh to remove your credentials
• shutdown the instance
• register the instance as an AMI and make it public

The only reason to do this is if the stack-aws-client code has significantly changed. Also note the Backend is not operating system dependent, all it does is configure grub to trigger our registration and installation code.