Building a Custom Amazon ParallelCluster AMI
Important
We don't recommend building a custom AMI as an approach for customizing Amazon ParallelCluster.
This is because, after you build your own AMI, you no longer receive updates or bug fixes with future releases of Amazon ParallelCluster. Moreover, if you build a custom AMI, you must repeat the steps that you used to create your custom AMI with each new Amazon ParallelCluster release.
Before reading any further, we recommend that you first check out the Custom Bootstrap Actions section to determine if the modifications you want to make can be scripted and supported with future Amazon ParallelCluster releases.
Even though building a custom AMI isn't ideal (because of the reasons mentioned earlier), there are still scenarios where building a custom AMI for Amazon ParallelCluster is necessary. This tutorial guides you through the process of building a custom AMI for these scenarios.
Note
Starting with Amazon ParallelCluster version 2.6.1, most of the install recipes are skipped by default when launching
nodes. This is to improve startup times. To run all of the install recipes for better backwards compatibility at the
expense of startup times, add "skip_install_recipes" : "no"
to the cluster
key in the extra_json setting. For example:
extra_json = { "cluster" : { "skip_install_recipes" : "no" } }
Prerequisites
-
Amazon ParallelCluster is installed.
-
The Amazon CLI is installed and configured.
-
You have an EC2 key pair.
-
You have an IAM role with the permissions required to run the pcluster CLI.
How to Customize the Amazon ParallelCluster AMI
There are three ways to use a custom Amazon ParallelCluster AMI described in the next sections. Two of these three methods require you to build a new AMI that's available under your Amazon Web Services account. The third method (Use a Custom AMI at Runtime) doesn't require that you build anything in advance, but does add risk to the deployment. Choose the method that best fits your needs.
Modify an AMI
This is the safest and most recommended method. Because the base Amazon ParallelCluster AMI is often updated with new releases, this AMI has all of the components required for Amazon ParallelCluster to function when it's installed and configured. You can start with this as the base.
Build a Custom Amazon ParallelCluster AMI
If you have a customized AMI and software already in place, you can apply the changes needed by Amazon ParallelCluster on top of it.
-
Install the following in your local system, together with the Amazon ParallelCluster CLI:
-
Packer: find the latest version for your OS from the Packer website
, and install it. The version must be at least 1.4.0, but the latest version is recommended. Verify that the packer
command is available in your PATH.
Note
Before Amazon ParallelCluster version 2.8.0, Berkshelf
(which is installed by using gem install berkshelf
) was required to usepcluster createami
. -
-
Configure your Amazon Web Services account credentials so that Packer can make calls to Amazon API operations on your behalf. The minimal set of required permissions necessary for Packer to work are documented in the IAM Task or Instance Role
section of the Amazon AMI Builder topic in the Packer documentation. -
Use the command
createami
in the Amazon ParallelCluster CLI to build an Amazon ParallelCluster AMI starting from the one that you provide as base:pcluster createami --ami-id
<BASE_AMI>
--os<BASE_AMI_OS>
Important
You shouldn't use an Amazon ParallelCluster AMI from a running cluster as
<BASE_AMI>
for thecreateami
command. Otherwise, the command fails.For other parameters, see pcluster createami.
-
The command in Step 4 runs Packer, which specifically does the following:
-
Launches an instance using the base AMI provided.
-
Applies the Amazon ParallelCluster cookbook to the instance to install relevant software and perform other necessary configuration tasks.
-
Stops the instance.
-
Creates an new AMI from the instance.
-
Terminates the instance after the AMI is created.
-
Outputs the new AMI ID string to use to create your cluster.
-
-
To create your cluster, enter the AMI ID in the custom_ami field within your cluster configuration.
Note
The instance type used to build a custom Amazon ParallelCluster AMI is t2.xlarge
. This instance
type doesn't qualify for the Amazon free tier, so you're charged for any instances that are created when you build
this AMI.
Use a Custom AMI at Runtime
Warning
To avoid the risk of using an AMI that's not compatible with Amazon ParallelCluster, we recommend that you avoid using this method.
When compute nodes are launched with potentially untested AMIs at runtime, incompatibilities with the runtime installation of Amazon ParallelCluster's required software might cause Amazon ParallelCluster to stop working.
If you don't want to create anything in advance, you can use your AMI and create an Amazon ParallelCluster from that AMI.
With this method, it takes longer for the Amazon ParallelCluster to be created because all the software that's needed by Amazon ParallelCluster when the cluster is created must be installed. Moreover, scaling up also takes a longer time.
-
Enter the AMI id in the custom_ami field within your cluster configuration.